The Elgar Companion to Information Economics 1802203958, 9781802203950

The Elgar Companion to Information Economics dexterously navigates this interdisciplinary field of research which celebr

115 24 8MB

English Pages 578 [579] Year 2024

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Front Matter
Copyright
Contents
Contributors
Preface
PART I Information Economics Overview
1. Information economics examined through scarcity and abundance
2. Robust theory and fragile practice: Information in a world of disinformation Part 1: Indirect communication
3. Robust theory and fragile practice: Information in a world of disinformation Part 2: Direct communication
4. Information and income distribution: The perspective of information economics
PART II Information Asymmetry
5. Asymmetric information as a market failure in retrospect
6. A quadrennial review of the significance of information asymmetry in economics and finance
7. Overcoming asymmetric information: A data-driven approach
8. Asymmetric information in health economics: Can contract regulation improve equity and efficiency?
9. Consequences of information asymmetry in a syndication network: The joint investments of the Israeli venture capital funds
PART III Information Transmission and Influence
10. Disclosure of conflicts of interest: Theory and empirics
11. Information and expertise
12. Informational influence and its forecasting in e-commerce
PART IV Innovation and Intellectual Property
13. Digital innovation: An information-economic perspective
14. Innovation and information: Smooth and ongoing change, or turbulence and cognitive over-stress? On the complex deep structure of innovation
15. Intangibles, information goods, and intellectual property goods in modern economics
16. Incomplete contracts, intellectual property rights, and incentives: Investment in knowledge assets under alternative institutional configurations
PART V Payment, Value, Crowdfunding
17. Payment on information markets
18. Assessing the perceived value of information in an information immersive world
19. The role of influencer endorsements in users’ willingness to pay for knowledge products: an empirical investigation
20. Barriers to participation in cultural crowdfunding
PART VI Challenges
21. The challenge of organizational bulk email systems: Model and empirical studies
22. The effect of supervisor’s control and workload on AIS users’ perceived usefulness and approach to misuse an automated system output: The moderating role of experience of AIS practitioners
23. On the status of machine learning inferences in data privacy economics and regulation
24. The digital world has a long shadow
PART VII The future
25. The terms: The parameters of information economics
Index
Recommend Papers

The Elgar Companion to Information Economics
 1802203958, 9781802203950

  • Commentary
  • Retail Version
  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

THE ELGAR COMPANION TO INFORMATION ECONOMICS

The Elgar Companion to Information Economics Edited by

Daphne R. Raban Associate Professor of Business Administration, School of Business Administration, University of Haifa, Israel

Julia Włodarczyk Associate Professor of Economics, Department of Economics, University of Economics in Katowice, Poland

Cheltenham, UK • Northampton, MA, USA

© Daphne R. Raban and Julia Włodarczyk 2024

With the exception of any material published open access under a Creative Commons licence (see www.elgaronline.com), all rights are reserved and no part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher.

Chapter 14 is available for free as Open Access from the individual product page at www. elgaronline.com under a Creative Commons Attribution NonCommercial-NoDerivatives 4.0 International (https://creativecommons.org/licenses/by-nc-nd/4.0/) license. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA A catalogue record for this book is available from the British Library Library of Congress Control Number: 2023951002 This book is available electronically in the Economics subject collection http://dx.doi.org/10.4337/9781802203967

ISBN 978 1 80220 395 0 (cased) ISBN 978 1 80220 396 7 (eBook)

EEP BoX

Contents

viii xiv

List of contributors Preface PART I

INFORMATION ECONOMICS OVERVIEW

1

Information economics examined through scarcity and abundance Daphne R. Raban and Julia Włodarczyk

2

Robust theory and fragile practice: Information in a world of disinformation Part 1: Indirect communication Joseph E. Stiglitz and Andrew Kosenko

3

4

Robust theory and fragile practice: Information in a world of disinformation Part 2: Direct communication Joseph E. Stiglitz and Andrew Kosenko Information and income distribution: The perspective of information economics Julia Włodarczyk

PART II

2

20

53

81

INFORMATION ASYMMETRY

5

Asymmetric information as a market failure in retrospect Wojciech Giza

6

A quadrennial review of the significance of information asymmetry in economics and finance Pedro A. Martín-Cervantes and María del Carmen Valls Martínez

7

Overcoming asymmetric information: A data-driven approach Giuseppe Pernagallo

8

Asymmetric information in health economics: Can contract regulation improve equity and efficiency? Pau Olivella

154

9

Consequences of information asymmetry in a syndication network: The joint investments of the Israeli venture capital funds Ilan Talmud

170

v

106

118 135

vi 

The Elgar companion to information economics

PART III INFORMATION TRANSMISSION AND INFLUENCE 10

Disclosure of conflicts of interest: Theory and empirics Ming Li and Ting Liu

185

11

Information and expertise Filippo Pavesi, Massimo Scotti and Nicola Argelli

202

12

Informational influence and its forecasting in e-commerce Avraham Noy and Shimon Schwartz

224

PART IV INNOVATION AND INTELLECTUAL PROPERTY 13

Digital innovation: An information-economic perspective Johannes M. Bauer and Tiago S. Prado

14

Innovation and information: Smooth and ongoing change, or turbulence and cognitive over-stress? On the complex deep structure of innovation Wolfram Elsner

270

15

Intangibles, information goods, and intellectual property goods in modern economics Dominika Bochańczyk-Kupka

301

16

Incomplete contracts, intellectual property rights, and incentives: Investment in knowledge assets under alternative institutional configurations Erkan Gürpinar and Eyüp Özveren

315

PART V

246

PAYMENT, VALUE, CROWDFUNDING

17

Payment on information markets Wolfgang G. Stock

18

Assessing the perceived value of information in an information immersive world Daphne R. Raban and Niv Ahituv

364

19

The role of influencer endorsements in users’ willingness to pay for knowledge products: an empirical investigation Xiaoyu Chen and Alton Y. K. Chua

379

20

Barriers to participation in cultural crowdfunding Roei Davidson

339

395

PART VI CHALLENGES 21

The challenge of organizational bulk email systems: Model and empirical studies Ruoyan Kong and Joseph A. Konstan

407

Contents  vii 22

The effect of supervisor’s control and workload on AIS users’ perceived usefulness and approach to misuse an automated system output: The moderating role of experience of AIS practitioners Ewa Wanda Maruszewska and Maciej Andrzej Tuszkiewicz

23

On the status of machine learning inferences in data privacy economics and regulation David Bodoff

24

The digital world has a long shadow Serghei Ohrimenco, Grigori Borta and Valeriu Cernei

436

462 481

PART VII THE FUTURE 25 Index

The terms: The parameters of information economics Sandra Braman

506 535

Contributors

Niv Ahituv is a Professor Emeritus of Tel Aviv University (TAU), Israel, and a Professor in Peres Academic Center. He retired from TAU in 2011 after serving for 30 years. In addition to being a researcher and a lecturer in TAU, he served in the following positions: 1999 to 2002 – Vice President and Director General (CEO) of TAU; 1989 to 1994 – Dean of the Faculty of Management. Ahituv’s career in the IT field started in 1965 in IBM. Later, he managed the Information Systems Department at the Bank of Israel from 1969 to 1975. He has consulted to managements of large organizations in the area of information technology policy and strategic planning, and serves as a member of the board in a number of Israeli business companies. From 1997 to 2011, he represented the Government of Israel in all Information Technology related issues in UNESCO. He has been the representative of the Israeli Academy of Science to CODATA from 2006 until 2022; during that period he served four years as a member of the Executive Committee and four years as Vice President of CODATA. He has published dozens of articles and a number of books and chapters in books. He holds degrees of BSc in Mathematics, MBA, and MSc and PhD (1979) in Information Systems Management. Nicola Argelli is a PhD student in Economics and Management at the LIUC (Carlo Cattaneo University), Italy. He graduated in Economics at the Catholic University of Milan. His main research interests are in economics of education, behavioural and experimental economics. He is working in collaboration with the research centre of the LIUC Business School. Johannes M. Bauer is a Professor in the Department of Media and Information and the Director of the Quello Center for Media and Information Policy at Michigan State University, USA. His research explores institutional choice and design problems raised by emerging technologies. Among other topics, he explores how public policy and management affect the benefits and risks of advanced information and communication technologies for individuals, communities, and society. He is currently working on a book on the governance of complex adaptive socio-technical systems. Dominika Bochańczyk-Kupka is an Associate Professor of Economics at the University of Economics in Katowice, Poland. The issues connected with intellectual property protection and the role of intangibles in economics are central to her current academic work. Her recent research concerns also the re-commerce market and collectibles in the luxury market. David Bodoff is a Senior Lecturer at the University of Haifa School of Business Administration, Israel. His research regards the intersection between business and computers. One focus of work is about search behaviour and technology, especially in a consumer or financial context. A more recent focus is on perceptions of ranked lists. His research has been published in a variety of journals including MISQ, ISR, ACM TOIS, and Experimental Economics, and in technical conferences including NIPS and ACM SIGIR. Grigori Borta is a PhD in Economics of Information Technology, specializing in the research of the economics behind computer malware. The topics of information security, monetization viii

Contributors  ix of computer malware, the structure of malicious groups in cyberspace, cyberwarfare, and possible conceptual ways of combating cybercrime are especially pronounced in his research. Sandra Braman is a Professor of Communication and Journalism and John Paul Abbott Professor of Liberal Arts at Texas A&M University, USA. Her books include Change of State: Information, Policy, and Power (2006), which won the 2022 ICA Fellows Book Award for a Book of Enduring Value. She edits the Information Policy Book Series at MIT Press, is a Fellow of the International Communication Association, and is former chair of the Communication Law and Policy Division of the International Communication Association and of the Law Section of the International Association of Media and Communication Research. Her research has been funded by the Rockefeller and Ford Foundations, the First Amendment Fund, and the US Fulbright program and National Science Foundation. Valeriu Cernei is a PhD student in Information Security Risk Management, specializing in cybersecurity and security risks management. Valeriu is a practitioner with more than 20 years’ experience and University Lecturer at the Academy of Economic Studies of Moldova. He has delivered projects for financial, as well as large international organizations. Areas of research interests are cybersecurity, deep/dark web, digital risks as well as national cyber security and he has a series of publications related to these topics. Xiaoyu Chen is an Assistant Professor in Information Resources Management at the School of Cultural Heritage and Information Management of Shanghai University, China. He holds a PhD in Information Studies from Nanyang Technological University, Singapore. Dr Chen’s PhD dissertation studies a group of online influencers who create and share knowledge-intensive content on digital platforms. Alton Y. K. Chua is an Associate Professor in Information Studies at the Wee Kim Wee School of Communication and Information of Nanyang Technological University, Singapore. His research interests lie mainly in information, knowledge and technology management, with a particular focus on digital environments. An active and prolific scholar, Dr Chua has published close to 200 scholarly articles in these areas. He also serves on the editorial board of several international refereed journals such as Journal of Information Science and Journal of Knowledge Management. Roei Davidson is an Associate Professor in the Department of Communication at the University of Haifa, Israel. He studies how the information technology industry and the systems it develops – social media, artificial intelligence, digital finance – impact human autonomy and the distribution of capital. He has published in journals such as New Media and Society, Information Communication and Society, Journalism, and Public Understanding of Science. Wolfram Elsner is a Professor of Economics, University of Bremen, Germany. After his Habilitation (1986, University of Bielefeld), he worked for 10 years in regional development (city and state). He was Managing Editor, Forum for Social Economics, 2012–2018, President of the European Association for Evolutionary Political Economy (EAEPE), 2012–2014 and 2014–2016, and has taught in the USA, European countries, South Africa, Australia, Mexico and China. His Textbook is Microeconomics of Complex Economies (2015). He has edited books and book series and published numerous international book chapters and journal arti-

x 

The Elgar companion to information economics

cles. He is Editor-in-Chief of the Review of Evolutionary Political Economy (REPE) since 2018. Wojciech Giza is an Associate Professor and Head of the Department of Microeconomics at the Krakow University of Economics, Poland. His research interests include economic theory, the history of economic thought and economic methodology. In particular, he is interested in the development of microeconomics embedded in the paradigm of neoclassical economics as well as its criticism from the perspective of heterodox schools of economic thought. He is an active member of The Polish Philosophy of Economics Network. He also sits on the Editorial Board of Krakow Review of Economics and Management. Erkan Gürpinar is an Assistant Professor of Economics at the Social Sciences University of Ankara, Turkey. His main research interests are economics of knowledge, law and economics, institutional economics, evolutionary economics, and history of economic thought. He has published in a number of journals including Journal of Evolutionary Economics and Journal of Institutional Economics. Ruoyan Kong is a PhD candidate in the Department of Computer Science and Engineering at the University of Minnesota, USA. Her research interest relates to human-computer interaction, bulk communication, and recommender systems. She has published at conferences such as the ACM Conference on Computer-Supported Cooperative Work and Social Computing, the IEEE International Conference on Data Mining, and the ACM Web Conference. Joseph A. Konstan is a Distinguished McKnight Professor and Distinguished University Teaching Professor in the Department of Computer Science and Engineering at the University of Minnesota, USA, where he also serves as Associate Dean for Research of the College of Science and Engineering. His research addresses a variety of human-computer interaction issues, including personalization (particularly through recommender systems), eliciting online participation, designing computer systems to improve public health, and ethical issues in research online. He is probably best known for his work in collaborative filtering recommenders (the GroupLens project, which won the ACM Software Systems Award). Andrew Kosenko is an Assistant Professor of Economics at the School of Management at Marist College, USA. His research has two strands: information design and models of memory. The former focuses on understanding when it is possible to affect a decision maker’s action by providing them with strategically designed information. The latter focuses on models of decision-making without perfect recall. His work has been published in the B.E. Journal of Theoretical Economics and the Journal of Economic Behavior and Organization. He also teaches at Columbia University and has taught at the University of Pittsburgh. He is a member of the “Economists for Ukraine” open collective. Ming Li is a Professor of Economics at Concordia University, Canada, and a Research Fellow of CIRANO and CIREQ. His research focuses on game theoretical analysis of strategic communication and persuasion and their applications. His research articles have appeared in journals including Econometrica, Journal of Economic Theory, Journal of Public Economics, and Economic Theory. His recent research focuses on the impact of information disclosure on decision-making in economic and political contexts.

Contributors  xi  

Ting Liu is an Associate Professor of Economics at Stony Brook University, USA. Her research focuses on credence goods and markets for experts’ services. Her work appears in International Economic Review, American Economic Journal-Microeconomics, International Journal of Industrial Organization, Journal of Industrial Economics, and Journal of Economic Behavior and Organization. Her recent research analyses the welfare impact of disclosing experts’ conflicts of interest. Pedro A. Martín-Cervantes is an Assistant Professor of Economics and Business at Universidad de Valladolid, Spain. His academic background includes several master’s degrees (in finance, e-business, and secondary school teaching from the universities of Antwerp, Deusto, and Almería, respectively). His research interests include the application of econometric methods in different fields such as corporate social responsibility, health economics, political analysis, sustainability, etc. He has published work in prestigious international journals (European Research on Management and Business Economics, Sustainability, the International Journal of Environmental Research and Public Health, among others). María del Carmen Valls Martínez is an Associate Professor of Financial Economics at the University of Almería, Spain. Her current research interests include financial operations, ethical banking, gender economics studies and health economics. She has published in leading journals such as Oeconomia Copernicana, Journal of Cleaner Production, Mathematics, European Research on Management and Business Economics, Frontiers in Psychology, Corporate Social Responsibility and Environmental Management, etc. Moreover, she is Editor in the prestigious journals Plos One, Mathematics and Frontiers in Public Health. Ewa Wanda Maruszewska is an Associate Professor in the Department of Business Informatics and International Accounting at the University of Economics in Katowice, Poland. Her scientific interests relate to qualitative and quantitative methods used in behavioural and decision-making research in accounting, and their implications for financial reporting quality. She was an expert for The National Centre for Research and Development, the prosecutor’s office, the court, and for the European Financial Reporting Advisory Group. Avraham (Avi) Noy, PhD, is a researcher and was an Adjunct Lecturer in the University of Haifa, Israel and a member of the Internet Research Center. His expertise is in electronic-commerce, computer-mediated communication and simulation. He has published in journals such as European Journal of Information Systems, International Journal of Simulation and Process Modelling, Electronic Markets, Simulation & Gaming, Journal of Interactive Marketing, and Journal of Consumer Psychology. Serghei Ohrimenco, DSc, is Full Professor at the Department of Computer Science and Information Management, and Head of the Laboratory “Information Security” of the Academy of Economic Studies (Chisinau, Republic of Moldova). His main scientific interests are the following: cybersecurity, economics of information security, shadow digital economy and its segments. He conducts economic research on the impact of cybercrime on modern society and national security. Pau Olivella is a Professor of Economics at Universitat Autònoma de Barcelona and Affiliated Professor at the Barcelona School of Economics, Spain. He is mainly interested in health insurance and healthcare providers’ incentives. He has publications in The Economic Journal,

xii 

The Elgar companion to information economics

the Geneva Risk and Insurance Review, and the Journal of Health Economics, among others. He was Associate Editor of the Journal of Health Economics until 2018 and was co-Editor of the Elsevier Encyclopedia of Health Economics. With M. Vera-Hernández, he received the Royal Economic Society Prize in 2014 and the Spanish Health Economics Association Prize in 2008. Eyüp Özveren is currently an independent scholar based in Istanbul, after his early retirement as Full Professor from the Economics Department of the Middle East Technical University, Ankara, Turkey. He taught, and has researched and published widely in political economy, institutional economics, history of economic thought, and occasionally in economic history. Filippo Pavesi is an Associate Professor of Economics at LIUC (Carlo Cattaneo University), Italy and Visiting Research Associate at the School of Business of Stevens Institute of Technology. His research interests include the economics of information, behavioural economics, political economics, social networks, experimental economics, and financial economics. His research has been published in leading academic journals such as the European Economics Review, Economic Inquiry, and the Journal of Economic Behavior & Organization. Giuseppe Pernagallo is a Postdoctoral Researcher at the Department of Economics and Statistics “Cognetti de Martiis” at the University of Turin, Italy, Junior Research Fellow at the Centro di Ricerca e Documentazione Luigi Einaudi in Turin, Adjunct Professor of Mathematics and Statistics at ESCP Business School and Adjunct Professor of Econometrics and Macroeconomics at the University of Turin. His main research interests are in information economics, finance, innovation and statistics. Tiago S. Prado is a digital policy and economics consultant, with fifteen years of work experience in national governments, development banks, and international organizations, including the Inter-American Development Bank, the International Telecommunication Union, the Organization of American States, and the European Commission. He has a Master of Public Policy, an MBA focused on project management, a B.Eng. in Communications Networks, and is currently a PhD Candidate in Information and Media at Michigan State University, USA. Daphne R. Raban is an Associate Professor in the School of Business Administration, Academic Head of the Library, University of Haifa and Chair of CODATA Israel National Committee. She founded the Department of Information & Knowledge Management and was a member of LINKS, the Israeli Center of Research Excellence on Learning in a Networked Society. Her areas of research include the information society and the information economy; specifically, she studies the perceived value of information, information markets and business models, knowledge sharing, and information diffusion. Shimon Schwartz is leading research projects as a Program Advisor at the National Research Council Canada as well as an Adjunct Assistant Professor in the Faculty of Engineering at the University of Waterloo, Ontario, Canada. Shimon’s main research interest is in the field of machine learning, explainable artificial intelligence and signal processing. Shimon served as Director of Research and Development in multiple private sector technology firms. Massimo Scotti is a Lecturer in Economics at LIUC (Carlo Cattaneo University), Italy. His research interests focus on economics of information, organizational economics and political

Contributors  xiii economy. He has published in journals such as the European Economic Review and the Journal of Economic Behavior & Organization. Joseph E. Stiglitz is an American economist and a professor at Columbia University, USA. He is also the co-chair of the High-Level Expert Group on the Measurement of Economic Performance and Social Progress at the OECD, and the Chief Economist of the Roosevelt Institute. Stiglitz was awarded the Nobel Memorial Prize in Economic Sciences in 2001 and the John Bates Clark Medal in 1979. He is a former senior vice president and chief economist of the World Bank and a former chairman of the US Council of Economic Advisers. In 2000, Stiglitz founded the Initiative for Policy Dialogue, a think tank on international development based at Columbia University. In 2011 he was named by Time magazine as one of the 100 most influential people in the world. Known for his pioneering work on asymmetric information, Stiglitz’s research focuses on income distribution, climate change, corporate governance, public policy, macroeconomics and globalization. He is the author of numerous books including, most recently, People, Power, and Profits (2019), Rewriting the Rules of the European Economy (2020), and Globalization and Its Discontents Revisited (2017). Wolfgang G. Stock is an information scientist and former Professor at Heinrich Heine University Düsseldorf, Germany, and Visiting Professor at the University of Graz, Austria. His main research interests include information markets, scientometrics, smart cities, live-streaming services, and social media. He is author of about 350 articles and of some basic textbooks in information science. His articles appeared, for instance, in Journal of Information Science Theory and Practice, Journal of the Association of Information Science, Scientometrics as well as in the Proceedings of the Hawaii International Conference on System Sciences (HICSS) and of Human Computer Interaction (HCI International). Additionally, he is editor of the book series Knowledge & Information: Studies in Information Science. Ilan Talmud is an Associate Professor at the Department of Sociology, University of Haifa, Israel. His research areas are social networks, interorganizational relations, inter-industrial transactions, the global economy and war, economic sociology, knowledge management, the digital economy, Internet studies, and cryptocurrencies. His last book (with Gustavo Mesch) is Wired Youth: The Online Social World of Adolescence (2020). Maciej Andrzej Tuszkiewicz is an Assistant Professor in the Department of Business Informatics and International Accounting at the University of Economics in Katowice, Poland. He parallelly is employed in the finance manager position in business practices which further widens his research view. His research focuses on behavioural research and decision-making in accounting and its implications for financial reporting quality. He was an expert in a number of projects for the European Financial Reporting Advisory Group. Julia Włodarczyk is an Associate Professor in the Department of Economics at the University of Economics in Katowice, Poland. Her research interests include phenomena associated with agents’ heterogeneity and inequality, flows of money and flows of information. In her work, she has often adopted an interdisciplinary perspective: from thermodynamics and cybernetics to sociology and psychology. Currently, she is engaged in research at the crossroads of behavioural economics, information economics and digital economics.

Preface

Bits move markets. Bits translate into media that deliver knowledge, advice, opinions, thoughts, and preferences; they convey information. Markets are mechanisms for the exchange of goods, services, and value. The interaction between bits and markets constitutes the information economy. The information economy is huge in terms of turnover, growth, variety, and a host of other facets, yet it receives less academic research attention than public attention. Information economics is a fascinating field of research. Its theories are continually challenged by the digital revolution, and even more so by artificial intelligence, which influence the operation of markets and transform longstanding price vectors. In spite of wide coverage in the general media and over six decades of academic research, this subject has not yet found prominence relative to leading branches of economic research. In a tongue-in-cheek way we could say that information economics is in-formation. This book aims to further the development of the field by assembling the work of eminent scholars from three continents who present state-of-the-art results in the field. Major themes include information and disinformation, inequality, information asymmetry, innovation, informational influence, payment, value, challenges, and more. One thread that appears in nearly every chapter is the contrast between the traditional economic assumption of scarcity and the practical observation that information is, in fact, abundant. Although abundance is starting to receive some research attention, scarcity still dominates economic discourse. We claim that the perspective of information abundance invites innovative studies within information economics, potentially with nontrivial implications for economics in general. Information economics is a natural candidate to expand discourse on abundance because information abundance is more pervasive than other forms of economic abundance. Abundance begs theory development because, alongside scarcity, it describes information supply and demand. Abundance has substantial positive consequences accompanied by a host of evolving challenges, affecting micro and macro aspects of the economy at large. This Companion provides an intellectual journey through the information economy, including its positive and negative dynamics. For instance, abundance of information supports innovation, which creates new payment methods, new products, and new jobs. But it can also translate into information overload, mis- and disinformation, and spam, as well as aggravate certain dimensions of income inequality. Similarly, artificial intelligence can help to address the problem of information asymmetry in many areas but may be the source of information asymmetry in others, not to mention potential breaches of privacy and intellectual property. The Companion is written for a scholarly audience of academic researchers interested in information economics – economists, sociologists, information scientists, information systems researchers, communication scholars, political scientists, technology enthusiasts and others. Researchers and graduate students will find advanced knowledge and inspiration. The Companion is also a solid and up-to-date source of knowledge and policy suggestions for policy makers and decision makers in the private, public, and non-governmental sectors. We believe that the knowledge in this book will remain relevant as the information sphere continues to change. We are open and eager to further develop the discussion with our readers. xiv

Preface  xv In closing, we feel fortunate to have received contributions from an exceptional group of authors who have contributed chapters to this Companion – we are grateful for their collaborative spirit and productive discussions. Our thanks go to several key people who have helped us with the book project. We would like to thank Andrea Gurwitt, Dana Vashdi, Gadi Buskila, Barbara Probierz, Daniel Mather and the anonymous reviewers. Thanks go also to Kalanit Efrat and colleagues at the University of Agder in Norway. Lastly, we thank our families for their endless support in this journey. Daphne R. Raban Julia Włodarczyk The Companion Editors

PART I INFORMATION ECONOMICS OVERVIEW

1. Information economics examined through scarcity and abundance Daphne R. Raban and Julia Włodarczyk

1. INTRODUCTION Information economics has shown that information is a required input for and a natural output of every economic activity. A long line of Nobel laureates acknowledged information economics as a central field of study with wide ranging implications in many areas from agriculture to financial markets (Braman, 2006). Two complementary schools of thought have emerged within information economics. One school of thought investigates the fundamental assumptions of economics and shows that the availability and use of information have far-reaching effects on markets and on welfare (cf. Stigler, 1961; Akerlof, 1970; Spence, 1973; Stiglitz, 1975). The other school of thought relates to information as an economic good, an object of trade or of free sharing (cf. Benkler, 2006; Braman, 2006; Linde & Stock, 2011; Shapiro & Varian, 1999; Waldfogel, 2017). Common to both schools of thought is the recognition of inefficiency regarding information processing by markets and individuals. In this Companion we posit that this inefficiency occurs both under information scarcity and abundance. The majority of prevailing models of information economics are still based on the concept of information scarcity; however, we claim that formulation of new models and theories encompassing information abundance is an important step in the development of information economics. Information economics refers to the contribution of information to economic transactions and to the trade in information artefacts. As noted by Stiglitz (2000, p. 1471), information economics is expected to explain in detail “how and how well organizations and societies absorb new information, learn, adapt their behaviour, and even their structures; and how different economic and organizational designs affect the ability to create, transmit, absorb, and use knowledge and information”. The long tradition of economic research involves analyses conducted in the context of scarcity with information discovered as a critical component of transactions. Until the 1990s access to transaction-related information and to artefacts such as books, reports, audio-visual materials, and software was restricted and mediated. Information in its various forms was scarce and generally followed the logic of scarcity. The past 30 years have been characterized by a rapidly increasing variety of digital and networked information sources and the proliferation of both unrestricted and unmediated information artefacts. Ownership rights, which generate economic scarcity, are elusive in the context of digital information while access is a central issue. Whereas access may be restricted by techno-economic means (password, subscription), much information is freely available. In fact, information is abundantly available. We claim that abundance should be a new theoretical cornerstone of information economics to reflect reality. 2

Information economics examined through scarcity and abundance  3 To some extent this central theme of the Companion reflects Hegel’s dialectics with successive phases of thesis (information scarcity), antithesis (information abundance) and synthesis. The shift towards information abundance is not a simple, quantitative increase in the volume of information available, proportional for all economic agents, nor the disappearance of the veil of ignorance. It is a qualitative change with increases in the average volume of information per agent, coupled with substantial heterogeneity in literacy levels and unequal access to abundant information. Importantly, in the real-world, scarce information coexists with abundant information exactly in the same way as inefficiencies persist, even if efficient solutions are available. This coexistence of two contrary categories (scarcity and abundance) that are simultaneously true reflects a recognition that truth involves moments that are contradictions if separated but are reconciled and consistent in their synthesis (McTaggart, 1896, p. 10). This implies that the development of economics in general, and of information economics specifically, will be driven by the synthesis of scarcity and abundance, which is far more important than simple negation. Therefore, synthesis of scarcity and abundance seems to be a non-trivial, but fascinating task for information economics. This chapter discusses the achievements and challenges for information economics through the prism of the concepts of scarcity and abundance. First, we introduce the concept of abundance into our discourse. In the subsequent sections, we pay special attention to market operations under scarce and abundant information, as well as to prices which according to neoclassical tradition are supposed to convey all relevant market information about relative scarcity and abundance. After a short summary of conducted considerations and suggestions for future research, we present the book’s structure.

2.

FROM SCARCITY TO ABUNDANCE

2.1 Background Scarcity and abundance are inherently associated with the material and immaterial (but not necessarily digital) aspects of economic activity, with scarcity often related to material processes and abundance to immaterial. On one hand, the biological construction of human beings focuses the attention of economists and policy makers on satisfaction of basic needs of a material character, while on the other hand, human cognitive development and creativity pushes people towards higher-order needs of an immaterial character.1 Some examples of immaterial or intangible benefits include learning, self-fulfilment, satisfaction, social status, emotional well-being, and self-actualization (Swan, 2017). In economics, the focus remains on scarcity,2 which – confronted with limitless wants – becomes not one of the most important economic problems, but the economic problem of efficient allocation which theoretically can be solved via market or non-market arrangements. Despite this strong emphasis on scarcity, abundance has started to receive increasing attention by economists (Bååth & Daoud, 2021; Dugger, 2009; Peach & Dugger, 2006; Rifkin, 2014; Sheehan, 2010; Taleb, 2012), sociologists (Abbott, 2014; Hoeschele, 2010), philosophers (Swan, 2016, 2017), legal scholars (Desai & Lemley, 2022; Madison et al., 2022), and communication scholars (Boczkowski, 2021).3 Some attempts to conceptualize abundance echo early developments in economics. In particular, Robbins (1932) formulated two conclusions relevant for contemporary discussions on scarcity and abundance within

4 

The Elgar companion to information economics

information economics. First, he insisted on the shift from defining economics as a science studying the origins of material welfare towards the emphasis on human behaviour and choice. This tendency to move away from material towards immaterial underpinnings of economics paved the way for investigations on the role of information. Second, he defined economics as a science studying the relationship between the ends of human activity and scarce means that have alternative uses (Robbins, 1932, pp. 4–15). Since then, scarcity and abundance as its opposition have been interpreted in relative terms. Considering the relative perspective, abundance occurs when resources are available in larger amounts compared to wants, which may lead to negative prices (Bååth & Daoud, 2021). However, it is possible that the distribution of output and income precludes a part of society from enjoying abundance and pushes individuals into scarcity associated with poverty (Sheehan, 2010). Needs have become a moving target as marketing and advertising influence economic decisions. Moreover, marketing can be seen as an instrument of sellers to communicate a feeling of scarcity among buyers, not only in the material domain but also in the informational domain (cf. FOMO – fear of missing out). Sheehan (2010) claims that marketing is a central driver of abundance, of spending beyond the basic needs. In other words, information is a central player in the formation of general economic abundance. Spreading abundant messages (e.g., via advertising) catalyses economic transactions. Beside this Companion’s focus on the dialectics of scarcity and abundance, another state between the two is sufficiency. Every agent can be assigned to one of the three systems: scarcity, sufficiency, and abundance (Daoud, 2018; Sheehan, 2010). Individuals belonging to one system experience markedly different economic conditions from people populating other systems. Importantly, the situation of individuals is determined not only by independent factors, but also their choices and attitudes. For example, self-interest appears as only one of the available options. Supported by philosophical and religious arguments people choose from the whole spectrum of attitudes: from egoistic to altruistic, from materialistic to spiritual etc.4 Taking into account this qualitative difference between scarcity and abundance in terms of their material and immaterial aspects, Swan (2017) interprets abundance as a social good in a broad sense including intangible qualities such as autonomy, recognition, and trust. Scarcity characterizes the material economy and describes a zero-sum game leading to competition and negative social goods. Abundance, according to Swan, is an expanding-pie model. Under abundance, digital occupations replace many material jobs. Some of the new occupations may be on the borderline of work and leisure.5 What brings us closer to the issue of information abundance is the observation by Dugger that “abundance is the full participation of all in the use of the community’s joint stock of knowledge. And knowledge is not scarce in the sense that one person’s use of it precludes anyone else’s use” (Dugger, 2009, p. 21). Importantly, abundance depends on the technological advances in a society. 2.2

Defining Information Abundance

While there is no universally accepted definition, there is agreement that abundance should become more salient in economic theory. Information economics is a natural candidate to expand such discourse because regardless of socio-economic status, every user of a computer or mobile phone is exposed to abundant information and is likely to experience it with varying degrees of competence. While access and literacy may impede universal and equal usage,

Information economics examined through scarcity and abundance  5 information abundance is more pervasive than other forms of economic abundance.6 We conclude that information economics is a good starting point for understanding abundance, and in turn, abundance is a useful theoretical concept driving information economics. As a first approximation, abundance of information can be defined as a situation when the supply of information is greater than the demand for it, even when the price of information is equal to zero.7 Practically, we define abundance of information as information exceeding informational needs or personal capacities which arrives to an agent through intuitive technologies from a multiplicity of information sources presenting high substitutability and accessibility. This definition requires us to clarify five issues. First, because information nowadays is mostly digital, a technological device is required for access. Abundance depends on access.8 Due to the non-rival nature of digital goods, the quantity of information delivered by such technological devices is theoretically limitless. In practice, the limit is the throughput of a given technology and the cognitive ability of the user (Simon, 1955; Sims, 2003), which brings us to the second point. Information is not a regular market good. Information is an intellectual good. As such, it requires the user to have certain prior knowledge and competence to be able to make use of information when they are exposed to it. The feeling of scarcity or abundance of information depends on users’ personal knowledge, skills, and capacities. This dependence on a reference point makes scarcity and abundance not only relative, but also subjective due to different individual evaluations (Ahituv, 1989; Kahneman & Tversky, 1979; Raban, 2007).9 Thirdly, even though abundance may be associated with one source of information, typically it stems from a multiplicity of sources (cf. Simon, 1971, pp. 40–41), constituting a product of quantity of information and information sources. The fourth issue is that abundance of information can have positive or negative effects for individuals. Abundance is positive if users can derive meaning from information, for example, redundancy may be helpful to corroborate new incoming information. However, information may become superfluous, its quantity being too vast for cognitive absorption when literacy, attention, and time are scarce. Diminishing marginal returns to information may induce a point at which abundance becomes information overload. Finally, the fifth issue is that beyond mere quantity, the quality of information ranges from negative (mis- or disinformation) to positive. Even though these two issues are often interrelated, information overload and the spread of disinformation and misinformation are two distinctive aspects of information abundance. People having to use information confront issues of access, personal ability, quantity (of information and its sources), and quality. The myriad combinations of these issues invoke different behaviours. Rational behaviour in principle consists in maximizing effects with given resources or, alternatively, minimizing resource use to achieve a given effect. Even rational agents behave differently under scarce information when they probably try to maximize effects using all available information, while under abundant information, in line with the concept of rational inattention, they would probably strive to minimize costs associated with information activity to achieve a given effect. Typically, agents are not rational, but the perspective of abundance suggests that specific inefficiencies occur as agents try to collect all relevant information to achieve the best effect under information abundance. Moreover, the character of information as a non-excludable good implies that individually rational behaviour can result in social inefficiencies which can take the form either of masses processing partial

6 

The Elgar companion to information economics

information (with a lower or higher degree of heterogeneity in terms of information sets across agents) or trying to process as much of the available information as possible (also with some heterogeneity). The former may result in agents being underinformed, and the latter in agents bearing excessive costs of information activity. As argued above, both types of inefficiency can depend on access and computational power. Both kinds of inefficiencies coexist, giving rise to asymmetry of information and justifying coordination of information activity or public intervention.10 Summing up, the definition of information abundance relates to the relationship between the availability and variability of information itself, the users’ heterogeneity, and the affordances of technology. This special relationship translates into markets in general and information markets specifically, as discussed next.

3.

MARKETS UNDER SCARCE AND ABUNDANT INFORMATION

3.1

Examining Traditional Assumptions

The romantic idea proposed in the eighteenth century by Adam Smith that an invisible hand governs markets that achieve social good based on individuals’ self-interest has been widely accepted and just as widely criticized. Neoclassical economists explain the notion of an invisible hand by regarding people as rational agents having access to full information, efficiently allocating scarce resources in the market to maximize their utility. This neoclassical assumption of full information ignores the issues of access, prior knowledge, cognitive capacity, and quality of information. Profound criticism of the assumption of acting upon full information came from realizing that in real markets, availability of information is asymmetric (Akerlof, 1970). Asymmetry of information occurs for both sides of the market creating relative scarcity on one side of the market. Producers and sellers (supply side) usually possess more information than buyers about the quality of goods. Buyers (demand side) possess more information about their own preferences, socio-economic status, or health. Information asymmetry leads to various forms of market failure including adverse selection, moral hazard, excessive market power (monopolies) and public goods. Information economics discussed strategies, specifically signalling and screening, which are meant to mitigate asymmetry, reduce uncertainty, and enable better functioning of markets (Spence, 1973; Stiglitz, 1975).11 These approaches have special merit when scarcity is concerned. Market failure under abundance has yet to be modelled to suggest novel solutions, and, where needed, external regulation. Approximately in parallel to the realization of the importance of information for the existence and functioning of markets, development of behavioural economics has shown that the other two assumptions, rational choice and utility maximization, do not reflect the behaviour of the majority of people (Kahneman & Tversky, 1979). Through experimentation, psychologists and economists have shown that people are inefficient in their information usage which leads to suboptimal decisions and markets providing goods and services to irrational actors (Bastardi & Shafir, 1998; Shampanier et al., 2007; Thaler, 1991; Tversky & Kahneman, 1982). As mentioned above, markets operate in a continuum of scarcity – sufficiency – abundance. Theoretically, only the situation of sufficiency is informationally neutral, generating no

Information economics examined through scarcity and abundance  7 informational frictions impinging on market efficiency. However, as pointed out earlier, in practice it is unlikely that all economic agents would have similar access and competences to process this sufficient information. Moreover, the easy availability of information at no cost pushes people towards overconsumption making the state of abundance the default or the new reference point. 3.2

Information in Markets

One of the most important questions is: how does the abundance of information change the operations of markets? To answer this question, we need to realize that markets are essentially about exchanging, transforming, and monetizing information. Information is a central source of value in many material markets as manifested by the major contribution of elements such as product development, design, branding, and customer service. Markets create incentives for an exponential growth of information production, as well as for the spread of disinformation and misinformation. Abundance of information can be thus perceived as a result of market changes (concentration) and a catalyst of further changes. In other words, information is an element of a powerful feedback loop pertaining to structural changes in many markets. Exponential growth of information production requires access to advanced technologies and hardware which induces market concentration in the form of platformization in the digital world. These technologies, together with data and information constitute barriers to entry to the oligopolistic or even monopolistic market of information. Other markets become subordinate to the information markets and their operations are often limited by constraints imposed by information giants and their algorithms. When algorithms aim to monetize popularity instead of truth, the invisible hand of the market becomes the ‘wisdom’ of crowds, resulting in an ever-increasing market concentration. Platforms thrive on production of data and information; however, they tend to have little interest in the content (Ma, 2023), as long as it has no direct impact on the platform’s profits. Often the quantity rather than quality of information on platforms drives profits. Dominant platforms determine what kinds of information can be disseminated. Their concentration on profits involves actions appealing to a large audience, which effectively means an agreement on deviations from the truth. Under platformization, the informativeness of information is not as important as monetization potential. As a result, platforms have minimal interest in distinguishing between information of high or low quality, or between facts and fake news. Furthermore, even if popularity can reflect informativeness, popularity can be also shaped by algorithms (Ma, 2023). There are also subtle, personal and social, mechanisms impinging on the operations of markets under the condition of abundance. For instance, Gino and Pierce (2009) experimentally find that the context of monetary abundance induces envy and significantly changes people’s behaviour to increase cases of cheating. Even though their research focused on the abundance of wealth, not information, we observe the occurrence of behavioural effects under information abundance as well, e.g. attitudes towards vaccination. This picture becomes more complicated, because even under abundant information some agents operate in a relatively scarce environment compared to others. Agents are heterogeneous in their preferences, prior knowledge, cognitive capacity, access to digital resources, and literacy levels, all of which may generate relative scarcity, especially as they pertain to the use of information as market enabler in the exchange of various goods and services. This agents’

8 

The Elgar companion to information economics

heterogeneity may be reinforced by phenomena such as rational inattention or information self-rationing. Information abundance in this case can be instrumental to achieve potentially more efficient allocation of resources (with or without direct impact on prices of the goods involved). 3.3

Markets of Information

In the case of the markets for digital information goods, abundance and agents’ heterogeneity have an impact on market structure and prices of information goods. Such markets are especially susceptible to market failure. In fact, market failure has become de facto a business goal. Operating within the scale-free network structure, entrepreneurs understand that their strategic goal is to be the ‘winner takes all’ of the network in their selected area of activity. Monopolizing a networked market means owning the market activity data. Application of advanced artificial intelligence algorithms on owned data bolsters the monopolistic position. The proliferation of network-based monopolies means that instead of a free information market, centrally controlled information silos are the current aspiration and position. Financial, economic, and political power ensue. Recent technological developments present blockchain as a solution based on the idea of decentralization. While some perils of cryptocurrencies based on blockchain technology are evident, we assume that solutions will develop to stabilize currency markets. Blockchain evangelists envision a world where commercial, business, and managerial transactions are conducted at the users’ free will. This being an ever more virtual environment, users will need even more advanced digital skills to manage their digital identities as part of virtual life. While the early decades of internet growth have given rise to online communities in parallel to business and commerce, current developments such as blockchain technology and the vision of a metaverse focus on markets and monetization. Blockchain technology marks a major step towards a pure market economy because it is predicated on assuring scarcity and ownership rights. To what extent can consumption shift from purely material and wasteful to purely informational, partially decreasing the pressure on natural resources (with a significant exception concerning energy sources)? Is this an innovative opportunity for information economics to reduce material consumption rather than induce it? For example, buying non-fungible tokens (NFTs) in the form of pixel art can provide a modern form of conspicuous consumption. NFTs, which are based on blockchain technology, are an instrument for creating scarcity and maintaining ownership rights for digital artefacts. The opacity relating to blockchain developments and the inclination of information markets to monopolization raise an urgent need to investigate mechanism design in the presence of abundant information products.12 An example of the effect of abundance on the operation of information markets is salient in everyday life where we observe the entanglement between excessive quantity and inferior quality of information. Due to broad access, existing knowledge and beliefs, limited literacy, and zero marginal costs, we observe the ever-increasing spread of mis- and disinformation, also known as fake information, and the proliferation of spam.13 When abundant information is not necessarily reliable, metainformation (complementary information) may be needed – greater access to information thus stimulates greater demand for information. This demand for metainformation is also driven by the nature of information as an experience good. This entails economic dynamics where utility becomes known ex post, but the decision to engage in information activity has to be made ex ante.

Information economics examined through scarcity and abundance  9 3.4

Information Market Dynamics

Dugast and Foucault (2018) introduce dynamism into the analysis of abundant information. They explain that it takes time to analyse abundant data and text (tweets, company reports and more), therefore, inevitably raw information that arrives quickly is of lower quality than analysed information that takes a longer time to produce. Noise comes before the signal. The decrease in the cost of producing low precision signals due to information abundance reduces the profitability of trading on subsequent (more precise) signals and crowds out incentives for fundamental research (Dugast & Foucault, 2018). Thus, information abundance can reduce the long run price informativeness. As a result, the noise-based behaviour of agents can change the market reality, so that the behaviour of better-informed agents can potentially become irrelevant to the new market situation. More precisely, the mere presence of information delay may not be sufficient to explain potential market outcomes under information abundance, especially when it is framed within the information-delay variant of the rational expectations theory, relating behaviour to optimally formed forecasts of the future. Allowing agents to act upon inferior forecasts (e.g. due to information processing constraints) brings us closer to the rational inattention theory proposed by Sims (2003). Rational inattention theory considers systematic changes in the nature of noise due to dynamic properties of economic processes. These changes explain inter alia reallocation of information monitoring capacities towards areas characterized by greater noise associated, for example, with a greater variability of prices (Sims, 2003), because such areas offer potentially higher profits. Importantly, deviation from optimal behaviour predicted by rational expectations theory introduces heterogeneity, increasing the complexity of market phenomena pertaining to information processing. 3.5

Abundance of Data

Our focus is on information; however, an emerging literature on abundance of data, which is the raw material of information, deserves mentioning. For instance, Acemoglu et al. (2019) explain why data are given away freely by users to platforms which use data to extract high profits: when a small group of users is willing to share data as part of their participation in a network that provides insight about other users, by inference. Even if those other users regard their data as valuable, they cannot form a market because their private data have already been inferred. This is a specific form of adverse selection. Acemoglu et al. (2019) do not provide reasoning for why the first users are willing to give away their private data. The answer probably lies in the value they perceive from being active in the network they selected (Benkler, 2006; Braman, 2006). When firms own abundant data, they may overuse it and not adequately respect consumer privacy. Data are non-rival enabling infinite and simultaneous reuse. Yet, data are often difficult to copy and can be made excludable, so a market is possible, creating even more value for firms. Companies have the means to aggregate and analyse data to produce information, an ability that individual lay users currently lack. While privacy remains an important concern,14 Acemoglu et al. (2019) and Jones and Tonetti (2020) demonstrate some benefits of data abundance such as the ability to draw inferences, generate innovation, reuse resources, make better decisions, offer recommendations, create new knowledge and more. However, these benefits are asymmetric. Data held back by

10 

The Elgar companion to information economics

companies reduce social welfare. Some researchers suggest that private ownership of data will improve social welfare through the development of efficient data markets (Jones & Tonetti, 2020; Mansour et al., 2016). Yet, given what we know about information economics and agents’ heterogeneity, there is no reason to believe that individuals will exercise their market opportunities efficiently. There may be a need for the government to create or regulate incentives that would shape this new data market. In section 3 we outlined how information abundance influences markets and suggested that markets should be designed under the assumption of abundance of information and information goods. Within such a design, prices play an important role when they are present but also when they are absent. While markets rely on the informativeness of prices, information itself often appears in markets for free. Priceless information can be price-free, which we unpack next.

4.

PRICES AND BEYOND

In Oscar Wilde‘s famous play Lady Windermere’s Fan, Lord Darlington explains to Cecil Graham that a cynic “is a man who knows the price of everything and the value of nothing”. Cecil Graham replies: “And a sentimentalist, my dear Darlington, is a man who sees an absurd value in everything, and doesn’t know the market price of any single thing” (Wilde, 1893, pp. 95–96). This literary conversation is a fairly good representation of the current state of economic understanding of prices and value. When scarcity is the frame of reference, prices convey some information about wants and preferences, as predicted by Hayek (1945). However, the informativeness of prices is limited because it does not account for information asymmetry nor for the subjective perception of value. If prices had conveyed all the necessary information at no cost, markets would be frictionless and efficient, yet, more than half a century of research tells us that markets suffer from informational inefficiency (cf. Akerlof, 1970; Grossman & Stiglitz, 1980). Moreover, in a digital world where abundance of information is the frame of reference, we observe a new reality where zero prices are prevalent. In other words, information abundance is associated (but not necessarily causally related) with the price of zero. Information is known to have high costs of production and virtually zero marginal costs. These traits coupled with non-rivalry and non-excludability attributes often lead to the conclusion that a price of zero is inevitable. The result is overconsumption of free information (Dugast & Foucault, 2018; Raban et al., 2019; Raban & Koren, 2019) to the point that it becomes harmful as evidenced by the recently coined term, infodemic (Eysenbach, 2020; Global Infectious Hazard Preparedness, 2021). Is it an efficient market? Observing various phenomena such as deceitful information (e.g. fake news, urban legends, spam) and information overload casts serious doubt regarding information market efficiency. While mis- and disinformation have been practised by leaders and institutions for thousands of years, also under information scarcity (cf. Machiavelli, 1532; Tzu, 2010), nowadays they may be initiated and promoted by anyone. One of the important problems refers to information of poor quality crowding out information of good quality, because at the end of the day we can conclude that even if information in general is abundant, information of good quality becomes scarce. Research has shown that zero prices increase consumer demand regardless of need or quality (Shampanier et al., 2007). Exposing people to an economic framing (i.e. requesting

Information economics examined through scarcity and abundance  11 payment for something that was previously free) also results in a change of preference such as more focus on utility or usefulness (Heyman & Ariely, 2004; Raban et al., 2019). In other words, zero prices invite information overload while economic framing induces an evaluative mindset. Could it be that information markets are missing price informativeness when prices are absent? Whereas the absence of a price may carry implicit informativeness, we suggest that digital implementation of price informativeness such as gamified prices could be an avenue for introducing reliability while affecting users’ epistemic approach and without impeding access. The complementary aspect of prices, which received considerable attention via the prospect theory, is the notion of value indicating individual differences in the perceptions of prices and risk (Kahneman & Tversky, 1979). Subjective perceptions of value are part of the information not necessarily conveyed by prices (Raban & Ahituv, 2024; Raban & Rafaeli, 2006). For example, people may act emotionally while making economic decisions. Such emotions are triggered by the value system, not by the prices on the market. The perceived value system comes into play in participatory pricing schemes such as in bargaining or auctions. Auctions have been successfully implemented in Google’s advertising pricing; however, participatory pricing can be further developed in many other digital information markets. Similarly, voluntary pricing and payment such as pay-what-you-want or name-your-own-price have remained largely esoteric options in information markets despite their unique suitability for experience goods. It could be that companies, and especially monopolies, prefer to exercise price differentiation through big data analytics. Relying on big data analytics may be likened to driving while looking at the rear-view mirror. We suggest looking through the windshield to identify new behaviours and detect new potential via value-based participatory or voluntary pricing and payment. Through systematic discussion of payment schemes for information products, Stock (2024) exemplifies that prices are not merely supplier-provided numerical indications of market activity. Prices as carriers of value can be numeric but may also appear in terms of time, attention, reputation and more. While scarcity in information markets leads to higher prices and lower access, abundance does not guarantee universal access (cf. Plamondon, 2022). Abundant information does not even reduce price dispersion (Goldfarb & Tucker, 2017). In fact, the online environment with its abundant data and personalization options is conducive to price discrimination tactics and constitutes a significant opportunity for capturing consumer surplus. Demand for information products can be seen by vibrant search activity, news consumption, app downloads, book sales and a host of other indicators. Some of this demand activity is market based, albeit information markets tend to be monopolized (Levitan, 1982; Linde & Stock, 2011; Shapiro & Varian, 1999). Much of the demand for information addresses free sources, which is to be expected. While markets for information products exist, there is vast opportunity to innovate through pricing with money or non-pecuniary currencies such as reputation, attention, influence and others which are often complementary and non-rival and are part of the mark of abundance.

5.

SUMMARY AND IDEAS FOR FUTURE RESEARCH

The neoclassical approach, according to which information is a public good defined by what it is not – non-rival and non-excludable – has limited relevance to analyse abundant information. This chapter is an attempt to come up with new, positively framed attributes which can be used

12 

The Elgar companion to information economics

to describe information markets. Information is an intellectual good which is often abundant, has variable value, requires skills and is frequently free of direct charge. Novel digital accounting mechanisms are being continuously developed and implemented to assist in filtering and verifying abundant information. The tools of economics enabled important insights regarding asymmetry and market structure, and the resulting market mechanisms and business models. In this regard it is striking to realize that economic analysis has largely overlooked the intriguing phenomenon of information abundance. While there have been attempts by economists to quantify this phenomenon (Lyman & Varian, 2000), these attempts were soon neglected due to the rapid quantitative increase and the understanding that the significance of information is beyond quantity. The qualitative change occurring in the switch to an abundance frame of thought is multi-dimensional, with quantity being one of various aspects to analyse. Additional dimensions appeared in our definition of information abundance, and it is likely that more dimensions will arise with added research attention. What is the meaning of information abundance for any kind of market? What is its particular meaning for information markets? We have started to lay the ground for incorporating abundance into economic analyses; however, these questions await theoretical and empirical economic investigation. It may seem that if scarcity of information to consumers creates an informational problem, then abundance of information may be a solution. In fact, information abundance is not an efficient remedy for problems prevailing under information scarcity. Abundance of information does not imply the end of information imperfections. It produces new imperfections which are still not well characterized from an economic perspective. The effects of abundant information may be studied at the individual level, at the market level and from a regulatory perspective. At the individual level, access to information, prior knowledge, and literacy interact with informational variables such as price of information (zero or non-zero) and information quality in ways which require theoretical and empirical scrutiny. The heterogeneity of potential strategies associated with dealing with abundance of information (e.g. due to rational inattention) matters for market efficiency. At the market level, we observe two general types of market: markets for non-information goods and services and the market for information artefacts. Both market types suffer from a host of market failures necessitating the conceptualization of specialized innovative mechanism design that would incorporate the characteristics of abundance. We suggest studying how to include the informativeness of prices even into exchanges of free information to gain the positive effect of price information. The issue of obtaining meaning from information such as in learning or expert advice or analytics is also a good candidate for market development. Additional innovation could come from envisioning virtual marketplaces that apply reliability and verifiability as part of their structure. The notion of non-pecuniary payments such as likes or kudos give rise to specific ‘banking systems’ for such assets (sometimes in a positive sense, sometimes creating speculative and fraudulent activities including click farms). These ideas await theoretical and empirical support. The tension between information scarcity and abundance, especially the attempts of private entities to monopolize data and information to exploit and exclude worse-informed parties constitutes a rationale for public intervention. This is where policy research can develop while taking maximum care to preserve the Jeffersonian spirit of the free spread of ideas (Jefferson, 1905). Through regulation, market transparency may improve and market failure, which is always anticipated for information, could be avoided in advance, by design (Stiglitz, 2009). In

Information economics examined through scarcity and abundance  13 fact, there is also uncertainty about scarcity and abundance of information in the future. While the volume of information may increase, the proportion of information publicly available to the total volume of information may decrease over time, especially when dominant platforms will try to increase the costs of changing the standard by closing consumers and other agents in their networks (lock-in) (cf. Włodarczyk, 2020, p. 19). The science of economics has been transformed with the introduction of psychological research as indicated by the 2002 Nobel prize for behavioural and experimental economics. The current challenges in information economics, specifically the implications of abundance, invite a collaboration between economists and researchers from the fields of information science, learning, digital literacy, information systems, computer science, psychology, sociology and more.

6.

ABOUT THE CONTENT OF THE COMPANION

This Companion is an edited volume dedicated to information economics written for an audience of scholars and advanced students. The chapters are grouped into seven sections, each touching on a significant area of the field as a whole. While overviewing the chapters, some tensions and issues become visible. The longstanding tension between free markets and government regulation is most pronounced in Chapters 2, 8, 13, 23, 24, and 25 with some relation also in Chapters 3 and 16. Tension arises from the often-extreme positive network externalities when confronted with local compromises such as individual privacy or effort. When positive externalities serve the public good, then a compromise on individual concessions may be acceptable; however, due to market failure the externalities often fortify monopolistic power. This tension appears, for example, in Chapters 21 and 23. Chapters 12, 20, 21, 23, and 24 provide some details relating to issues of particular affordances of information technologies that are not always known or understood by users, which imply that technology itself creates asymmetry. Issues that relate to problems with distribution of information and resources which eventually lead to inequality are highlighted in Chapters 2, 4, and 20. The tension between scarcity and abundance is central in this volume. In Part I, entitled ‘Information Economics Overview’, Chapter 2 reviews information economics research which is premised on scarcity highlighting the critical role of information in the failure of the fundamental theorems of welfare economics, the non-existence of competitive equilibrium, and the dependence of the nature of the equilibrium, when it exists, on both what information is available, and how information can be acquired. Chapter 3 acknowledges the development of information abundance, its current manifestations, harmful effects and methods to control it. Chapter 4 goes on to show how scarcity and abundance of information play out for income distribution. Part II deals with ‘Information Asymmetry’, which occurs when one or both sides of a transaction experiences relative information scarcity. Chapter 5 offers a historical perspective on the breakthrough associated with inclusion of asymmetric information in the discourse on market failures, while Chapter 6 attempts to summarize the abundant scientific research in the area of information asymmetry (1979–2021) using bibliometric analysis. Chapter 7 claims that advancement of data collection and artificial intelligence shifts the problem of asymmetry of information and that theoretical models should be replaced or supplemented by data-driven approaches such as machine learning, provided that monopolies of knowledge do not emerge.

14 

The Elgar companion to information economics

Chapter 8 links asymmetric information to specific aspects of private information on the healthcare market (health insurance, doctors’ prescription and referral strategies), and studies the effects of contract regulation of the healthcare market. Chapter 9 discusses consequences of information scarcity and abundance epitomized by uneven social embeddedness of venture capital funds in Israel, operating under conditions of severe uncertainty, asymmetric information, and high financial risks. It emphasizes the role of network position determining inter alia accessibility to privileged information flows. Part III is about ‘Information Transmission and Influence’. In a sense, this section is about information sufficiency in the context of decision making – finding ways to convey the appropriate information in decision tasks. Chapter 10 reviews and expands the theoretical literature on disclosure of conflicts of interest in expert/decision maker relationships. Chapter 11 discusses four aspects of strategic information transmission, such as the extent to which delegation can be superior to communication, the role of experts’ competition, the effectiveness of reputation, and the role played by networks. Chapter 12 discusses social influence in e-commerce in the form of the impact of other consumers affecting the decision making of buyers and demonstrates how a machine-learning model can forecast future ratings and reviews. Part IV is about ‘Innovation and Intellectual Property’. Innovation increases information abundance. It can be motivated by information scarcity, but is usually enabled by its abundance (especially in case of combinatorial innovation). Intellectual property rights are means for implementing scarcity when innovation is concerned. Chapter 13 develops an integrative approach to innovation and focuses on digital innovation, which raises unique challenges for economic theory due to its plasticity and open-endedness. Chapter 14 investigates the relationship between information and innovation in an evolutionary-institutional perspective (demonstrating that complexity can be prohibitive or largely counterproductive to innovation), interprets the basic model of ‘evolution of cooperation’ in terms of information and innovation implications, and considers the significance of network structures and of ‘self-organization’ mechanisms. Chapter 15 makes an attempt to systematize the main concepts and approaches to intangibles, information goods, and intellectual property goods which can be described as hybrid goods made up of information and law. Chapter 16 acknowledges that the current implementation of intellectual property rights is not well-suited to knowledge innovation and suggests alternative institutional configurations to accommodate knowledge building processes. Part V on ‘Payment, Value, Crowdfunding’ highlights individual differences and social processes that underlie information monetization. Chapter 17 provides a comprehensive overview of payment methods including direct, indirect, and social forms of monetization. Chapter 18 describes the theoretical basis and empirical evidence regarding perception of value of information goods and outlines the significance of value perception for decision making and for construction of markets. Abundance of information highlights the need to polish digital literacy in addition to economic choices. Willingness-to-pay scales appear in Chapter 18 and in Chapter 19 which is devoted to the special case of knowledge influencers, who may be viewed as gatekeepers in the presence of excess information. Following theory development, the chapter examines how the attractiveness of influencers increases attachment to and willingness-to-pay for knowledge products. Chapter 20 raises social considerations of crowdfunding for cultural products by focusing on social and cultural capital and by explaining sources of inequality in raising funding.

Information economics examined through scarcity and abundance  15 Part VI is devoted to ‘Challenges’. While earlier challenges such as information asymmetry were associated mainly with information scarcity, current challenges are often about information and data abundance. Chapter 21 problematizes bulk organizational email and the associated costs. It then suggests a multi-stakeholder economic model to alleviate the problem. Describing an empirical study, Chapter 22 highlights the importance of supervision for maintaining quality of information when engaging with accounting information systems. Chapter 23 illuminates some dark corners of technological progress, namely, using advanced artificial intelligence algorithms to process large amounts of user data results in potential breaches of privacy. Technological knowledge is needed to manage these threats and devise appropriate policy. In closing the challenges section, Chapter 24 reminds our readers that alongside the technological developments (internet of things, artificial intelligence, nano-devices, virtual and augmented reality, 3-D printing and many others) and the associated information and data, the digital environment holds many opportunities for illegal activities. Finally, Chapter 25 in Part VII, ‘Future’, draws together the main threads discussed in this volume and presents cross-disciplinary approaches to information economics, starting from the historical background and moving towards the implications of artificial intelligence. The chapter discusses foundational terms that would enable information economists from different subfields to understand differences in their research agendas and methods. These terms require changing from thinking about economic agents to perceptual entities; from treating information as data points to differentiated information; and from focusing on markets to looking at probability spaces.

NOTES 1.

2.

3. 4.

5. 6.

This perspective corresponds to Maslow’s hierarchy of needs explaining differentiated motivation under scarcity (satisfaction of basic needs reduces motivation) and under abundance (motivation increases when higher-order needs are satisfied) (Maslow, 1943). This has important implications for economies, especially capitalistic ones, which are based on scarcity. A doctrine particularly explicit about scarcity was formulated by Thomas Malthus (1798) who treated poverty and famine (evident manifestations of scarcity) as natural outcomes of fast population growth relative to resources. Interestingly, Malthus acknowledged that intellectual needs are not as satiable as others and can be treated as higher-order needs: “When the mind has been awakened into activity by the passions, and the wants of the body, intellectual wants arise; and the desire of knowledge, and the impatience under ignorance, form a new and important class of excitements” (Malthus, 1798, p. 377). However, he also noted that “Could we suppose the period arrived, when there was not further hope of future discoveries, and the only employment of mind was to acquire pre-existing knowledge, without any efforts to form new and original combinations, though the mass of human knowledge were a thousand times greater than it is at present, yet it is evident that one of the noblest stimulants to mental exertion would have ceased” (Malthus, 1798, p. 383). Earlier works include e.g. Ward (1966) and Fox (1967). See for example Stoic philosophy and the recommendation to equate sufficiency with abundance: “Whatever is enough is abundant in the eyes of virtue” (Seneca, 1932). Drawing from Buddhist and Islamic readings, Rafikov and Akhmetova (2019), like Seneca, suggest simplicity and spirituality as a solution for both waste and scarcity, so that more people would feel relative abundance and contentment. Chapter 4 of this Companion alludes to the concept of playbour (Włodarczyk, 2024, pp. 81–104). Beyond information economics, abundance is typical also in 3D printing, synthetic biology (potentially revolutionizing production of food and medicines), robotics (Lemley, 2016) and more.

16  7. 8. 9.

10.

11. 12. 13. 14.

The Elgar companion to information economics From the suppliers’ perspective, this zero price of information may translate into losses as the supplier incurs costs they cannot recover directly from the consumers. Abundance of a good or service itself does not guarantee a universal access to it and may give rise to inequalities (cf. Plamondon, 2022; Włodarczyk, 2024). Interestingly, the prospect theory with its assertion that losses are felt more intensively than gains, implies that scarcity of information is experienced more intensively than abundance (which also justifies the focus on scarcity prevailing in the literature). Furthermore, when abundance transforms itself into information overload, it becomes a burning problem again. The alternative options stemming from the principle of rational behaviour can be linked to different motivation of individuals. Some of them may be motivated by aversion towards waste (e.g., want to rent their temporarily unused premises), while others remain profit-oriented (don’t mind having unused premises if they can rent them at a higher price). The former is more inclusive, the latter more exclusive, but both are stimulated by greater availability of information about capacity use etc. Chapter 2 in this Companion provides a detailed historical account and current knowledge on information asymmetry (Stiglitz & Kosenko, 2024a, pp. 20–52). Chapter 17 of this Companion describes information prices and payment for information goods (Stock, 2024, pp. 339–363), while Chapter 18 applies the approaches of behavioural economics to the study of information as a good offered on the market (Raban & Ahituv, 2024, pp. 364–378). Chapter 3 of this Companion relates to mis- and disinformation (Stiglitz & Kosenko, 2024b, pp. 53–80) and Chapter 21 relates to spam (Konstan & Kong, 2024, pp. 407–435). Chapter 23 of this Companion details how technological advances present ever-growing challenges to privacy (Bodoff, 2024, pp. 462–480).

REFERENCES Abbott, A. (2014). The Problem of Excess. Sociological Theory, 32(1), 1–26. Acemoglu, D., Makhdoumi, A., Malekian, A., & Ozdaglar, A. (2019). Too Much Data: Prices and Inefficiencies in Data Markets. National Bureau of Economic Research, No. w26296. https://​ economics​.harvard​.edu/​files/​economics/​files/​acemoglu​_spring​_2020​.pdf. Ahituv, N. (1989). Assessing the Value of Information: Problems and Approaches. In J. I. DeGross, J. C. Henderson, & B. R. Konsynski (Eds.), Proceedings of the Tenth International Conference on Information Systems (pp.  314–325). https://​aisel​.aisnet​.org/​cgi/​viewcontent​.cgi​?article​=​1007​&​ context​=​icis1989. Akerlof, G. A. (1970). The Market for “Lemons”: Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84(3), 488–500. Bååth, J. & Daoud, A. (2021). Extending Social Resource Exchange to Events of Abundance and Sufficiency. In Dictionary of Ecological Economics (pp.  2015–2017). http://​arxiv​.org/​abs/​2010​ .02658. Bastardi, A. & Shafir, E. (1998). On the Pursuit and Misuse of Useless Information. Journal of Personality and Social Psychology, 75(1), 19–32. Benkler, Y. (2006). The Wealth of Networks: How Social Production Transforms Markets and Freedom. New Haven: Yale University Press. Boczkowski, P. (2021). Abundance: On the Experience of Living in a World of Information Plenty. Oxford: Oxford University Press. Bodoff, D. (2024). On the Status of Machine Learning Inferences in Data Privacy Economics and Regulation. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 462–480). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Braman, S. (2006). The Micro- and Macroeconomics of Information. Annual Review of Information Science and Technology, 40, 3–52. Daoud, A. (2018). Unifying Studies of Scarcity, Abundance, and Sufficiency. Ecological Economics, 147, 208–217. Desai, D. R. & Lemley, M. A. (2022). Scarcity, Regulation, and the Abundance Society. Stanford Law School Working Paper Series, 572. https://​doi​.org/​10​.2139/​ssrn​.4150871.

Information economics examined through scarcity and abundance  17 Dugast, J. & Foucault, T. (2018). Data Abundance and Asset Price Informativeness. Journal of Financial Economics, 130(2), 367–391. Dugger, W. M. (2009). Economic Abundance: An Introduction. Armonk, NY: M. E. Sharpe. Eysenbach, G. (2020). How to Fight an Infodemic: The Four Pillars of Infodemic Management. Journal of Medical Internet Research, 22(6), e21820. https://​doi​.org/​10​.2196/​21820. Fox, D. M. (1967). The Discovery of Abundance: Simon N. Patten and the Transformation of Social Theory. Ithaca, NY: Cornell University Press. Gino, F. & Pierce, L. (2009). The Abundance Effect: Unethical Behavior in the Presence of Wealth. Organizational Behavior and Human Decision Processes, 109(2), 142–155. Global Infectious Hazard Preparedness (2021). An Overview of Infodemic Management during COVID-19. World Health Organization, Health Topics (May). https://​www​.who​.int/​health​-topics/​ infodemic​#tab​=​tab​_1. Goldfarb, A. & Tucker, C. E. (2017). Digital Economics. NBER Working Paper, 23684. https://​ssrn​ .com/​abstract​=​3023079. Grossman, S. J. & Stiglitz, J. E. (1980). On the Impossibility of Informationally Efficient Markets. American Economic Review, 70(3), 393–408. Hayek, F. A. (1945). The Use of Knowledge in Society. American Economic Review, 35(4), 519–530. Heyman, J. & Ariely, D. (2004). Effort for Payment: A Tale of Two Markets. Psychological Science, 15(11), 787–793. Hoeschele, W. (2010). The Economics of Abundance: A Political Economy of Freedom, Equity, and Sustainability. New York: Routledge. Jefferson, T. (1905). Thomas Jefferson to Isaac McPherson. In A. A. Lipscomb & A. E. Bergh (Eds.), The Writings of Thomas Jefferson (Document 25). Thomas Jefferson Memorial Association. Jones, C. I. & Tonetti, C. (2020). Nonrivalry and the Economics of Data. American Economic Review, 110(9), 2819–2858. Kahneman, D. & Tversky, A. (1979). Prospect Theory: An Analysis of Decision Under Risk. Econometrica, 47(2), 263–291. Konstan, J. A. & Kong, R. (2024). The Challenge of Organizational Bulk Email Systems: Models and Empirical Studies. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 407–435). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Lemley, M. A. (2016). IP in a World Without Scarcity. https://​doi​.org/​10​.31235/​osf​.io/​3vy5a. Levitan, K. B. (1982). Information Resources as “Goods” in the Life Cycle of Information Production. Journal of the American Society for Information Science, 33(1), 44–54. Linde, F. & Stock, W. G. (2011). Information Markets: A Strategic Guideline for the I-Commerce. Berlin: De Gruyter. Lyman, P. & Varian, H. R. (2000). How Much Information. http://​www​.sims​.berkeley​.edu/​how​-much​ -info. Ma, L. (2023). Information, Platformized. Journal of the Association for Information Science and Technology, 74(2), 273–282. Machiavelli, N. (1532). The Prince. https://​www​.gutenberg​.org/​ebooks/​1232. Madison, M. J., Frischmann, B. M., Sanfilippo, M. R., & Strandburg, K. J. (2022). Too Much of a Good Thing? A Governing Knowledge Commons Review of Abundance in Context. Frontiers in Research Metrics and Analytics, 7. https://​doi​.org/​10​.3389/​FRMA​.2022​.959505. Malthus, T. (1798). An Essay on the Principle of Population, as it Affects the Future Improvement of Society with Remarks on the Speculations of Mr. Godwin, M. Condorcet, and Other Writers. London: Printed for J. Johnson, in St. Paul’s Church-Yard. Mansour, E., Sambra, A. V., Hawke, S., Zereba, M., Capadisli, S., Ghanem, A., Aboulnaga, A., & Berners-Lee, T. (2016). A Demonstration of the Solid Platform for Social Web Applications. WWW 2016 Companion – Proceedings of the 25th International Conference on World Wide Web (pp. 223–226). https://​doi​.org/​10​.1145/​2872518​.2890529. Maslow, A. H. (1943). A Theory of Human Motivation. Psychological Review, 50(4), 370–396. McTaggart, J. M. E. (1896). Studies in the Hegelian Dialectic. Cambridge: Cambridge University Press. Peach, J. & Dugger, W. M. (2006). An Intellectual History of Abundance. Journal of Economic Issues, 40(3), 693–706.

18 

The Elgar companion to information economics

Plamondon, S. (2022). Inequality in Abundance. Frontiers in Research Metrics and Analytics, 7. https://​ doi​.org/​10​.3389/​FRMA​.2022​.980677. Raban, D. R. (2007). User-Centered Evaluation of Information: A Research Challenge. Internet Research, 17(3), 306–322. Raban, D. R. & Ahituv, N. (2024). Assessing the Perceived Value of Information in an Information Immersive World. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 364–378). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Raban, D. R., Barzilai, S., & Portnoy, L. (2019). The Unexpected Benefits of Paying for Information: The Effects of Payment on Information Source Choices and Epistemic Thinking. In M. Pańkowska & K. Sandkuhl (Eds.), Perspectives in Business Informatics Research (pp. 163–176). Cham: Springer. Raban, D. R. & Koren, L. (2019). Risk as a Predictor of Online Competitive Information Acquisition. Open Information Science, 3(1), 47–60. Raban, D. R. & Rafaeli, S. (2006). The Effect of Source Nature and Status on the Subjective Value of Information. Journal of the American Society for Information Science and Technology, 57(3), 321–329. Rafikov, I. & Akhmetova, E. (2019). Scarcity in the Age of Abundance: Paradox and Remedies. International Journal of Ethics and Systems, 35(1), 119–132. Rifkin, J. (2014). The Zero Marginal Cost Society: The Internet Of Things, the Collaborative Commons, and the Eclipse of Capitalism. New York: St. Martin’s Press. Robbins, L. (1932). An Essay on the Nature and Significance of Economic Science. London: Macmillan & Co. Seneca, L. A. (1932). Seneca’s Letters to Lucilius. Oxford: Clarendon Press. Shampanier, K., Mazar, N., & Ariely, D. (2007). Zero as a Special Price: The True Value of Free Products. Marketing Science, 26(6), 742–757. Shapiro, C. & Varian, H. R. (1999). Information Rules: A Strategic Guide to the Network Economy. Boston: Harvard Business School Press. Sheehan, B. (2010). The Economics of Abundance: Affluent Consumption and the Global Economy. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Simon, H. A. (1955). A Behavioral Model of Rational Choice. The Quarterly Journal of Economics, 69(1), 99–118. Simon, H. A. (1971). Designing Organizations for an Information-Rich World. In M. Greenberger (Ed.), Computers, Communication, and the Public Interest (pp.  37–72). Baltimore, MD: Johns Hopkins Press. Sims, C. A. (2003). Implications of Rational Inattention. Journal of Monetary Economics, 50(3), 665–690. Spence, M. (1973). Job Market Signaling. The Quarterly Journal of Economics, 87(3), 355–374. Stigler, G. J. (1961). The Economics of Information. Journal of Political Economy, 69(3), 213–225. Stiglitz, J. E. (1975). The Theory of ‘Screening’, Education, and the Distribution of Income. American Economic Review, 65(3), 283–300. Stiglitz, J. E. (2000). The Contributions of the Economics of Information to Twentieth Century Economics. The Quarterly Journal of Economics, 115(4), 1441–1478. Stiglitz, J. E. (2009). Government Failure vs. Market Failure: Principles of Regulation. In E. J. Balleisen & D. A. Moss (Eds.), Government and Markets: Toward a New Theory of Regulation (pp. 13–51). New York: Cambridge University Press. Stiglitz, J. E. & Kosenko, A. (2024a). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 1: Indirect Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 20–52). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. & Kosenko, A. (2024b). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 2: Direct Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 53–80). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stock, W. G. (2024). Payment on Information Markets. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 339–363). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing.

Information economics examined through scarcity and abundance  19 Swan, M. (2016). Philosophy of Social Robotics: Abundance Economics. In A. Agah, J. J. Cabibihan, A. Howard, M. Salichs, & H. He (Eds.), Social Robotics. ICSR 2016. Lecture Notes in Computer Science, vol. 9979 (pp. 900–908). Cham: Springer. Swan, M. (2017). Is Technological Unemployment Real? An Assessment and a Plea for Abundance Economics. In K. LaGrandeur & J. J. Hughes (Eds.), Surviving the Machine Age: Intelligent Technology and the Transformation of Human Work (pp. 19–33). Cham: Palgrave Macmillan. Taleb, N. N. (2012). Antifragile: Things That Gain from Disorder. New York: Random House. Thaler, R. H. (1991). The Psychology of Choice and the Assumptions of Economics. In R. H. Thaler, Quasi Rational Economics (pp. 137–166). New York: Russell Sage Foundation. Tversky, A. & Kahneman, D. (1982). Judgement Under Uncertainty: Heuristics and Biases. Science, 185(4157), 1124–1131. Tzu, S. (2010). The Art of War. Chichester: Capstone Publishing. Waldfogel, J. (2017). How Digitization Has Created a Golden Age of Music, Movies, Books, and Television. The Journal of Economic Perspectives, 31(3), 195–214. Ward, B. (1966). The Economics of Abundance. Ekistics, 21(123), 107–109. Wilde, O. (1893). Lady Windermere’s Fan: A Play about a Good Woman. London: E. Mathews and J. Lane. Włodarczyk, J. (2020). Non solum technology – korzenie i perspektywy rozwojowe gospodarki cyfrowej. In I. Ostoj & K. Bartuś (Eds.), Innowacje na poziomie mikro- i makroekonomicznym (pp. 11–23). Katowice: Uniwersytet Ekonomiczny. Włodarczyk, J. (2024). Information and Income Distribution: The Perspective of Information Economics. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 339–363). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing.

2. Robust theory and fragile practice: Information in a world of disinformation Part 1: Indirect communication Joseph E. Stiglitz and Andrew Kosenko

1.

INTRODUCTORY REMARKS AND MAIN MESSAGES

The analysis of economies where information is imperfect and asymmetric has given rise to a revolution in economics (Stiglitz, 2002, 2020). Longstanding fundamental results such as the presumption of the efficiency of competitive markets – Adam Smith’s invisible hand – and that competitive equilibrium is always characterized by demand equalling supply have been overturned. Other key results, such as those concerning the existence of competitive equilibrium and its characterization too have been overturned. Since the founding of modern economics, analyses were based on models of perfect information, with the hope that so long as information was not too imperfect, the results would be at least approximately correct. Information economics showed that this was not true: even a little bit of imperfection could drastically change the results. The insights of this literature have touched virtually every subdiscipline in both macro- and microeconomics – from labour economics to finance, from product markets to insurance markets – and provided intellectual foundations for still other subdisciplines, like accounting and corporate governance.1 With such an expansive literature, a short survey has to necessarily be selective.2 We focus particularly on adverse selection, where it is known that there are differences among individuals (among investment projects, among products, among firms), but there is imperfect information about who is who – what Stiglitz (1975a) called screening. And rather than discussing the myriad of applications, we focus on some of the general principles and insights that have emerged from the vast literature. In this chapter we look at the economics of information in instances where information is endogenous – parties form beliefs (subjective probabilities) about the unobserved characteristics of other parties as a result of actions taken. There is limited direct communication, the one exception being that individuals may disclose verifiable information about themselves (their products or their projects). The insights gleaned from the early literature in these settings – the inefficiency of markets, the sharp difference in economic outcomes that incomplete or asymmetric information environments generate relative to complete information environments, as well as the basic mechanisms for overcoming difficulties posed by incomplete or asymmetric information and the nature of welfare improving interventions – continue to be useful in applications and in advancing the theory. At the same time, recent results have modified, and in some cases overturned, key earlier results. In particular, recent literature has reinforced earlier analyses showing the fragility of the results to the precise specification of the information environment. For instance, we note the result of Kosenko, Stiglitz, and Yun (2023) that under quite general conditions, in the absence of communication, no equilibrium exists; both the 20

Robust theory and fragile practice: part 1  21 price (Akerlof, 1970) equilibrium and the quantity (Rothschild and Stiglitz, 1976) equilibrium can be broken. In the companion Chapter 3, we look at situations where information is instead communicated directly (instead of indirectly, through actions). Here we survey the advances in the fundamental models of endogenous information – signalling, screening, and adverse selection – that have been made in the decades subsequent to their formulation. The earliest literature assumed initial information asymmetries and addressed how they were addressed in competitive markets and monopolies. Here we discuss endogenizing the initial asymmetries, and consider a broader range of mechanisms by which they can be dealt with – including “mechanism design”, the most important examples of which are perhaps the design of auctions and matching algorithms. The earlier literature was written too before we moved into the age of mis- and disinformation. While the sequel chapter deals more explicitly with this issue in the broader context of communication, here we consider explicitly market incentives for obfuscation and the role of public intervention through disclosure and fraud laws. While market failures are rife in economies with costly information (that is, all economies), we explain here how many key so-called reforms of recent years – for instance, “completing markets” through the creation of derivatives and structured finance – have increased systemic informational burdens, undermining decentralization and lowering welfare. Reforms that fail to consider how the proposed changes alter the economy’s information structure may well be counterproductive. The insight that a rational, utility-maximizing decision maker may (and generally speaking, will) reveal her information through her choices and actions – a key insight in different contexts of Spence (1973), Rothschild and Stiglitz (1976), Stiglitz (1977) and Mirrlees (1971) – linked (private) information and (public) actions, and made the analyses tractable. Similarly, Robert Aumann pointed out in the context of two-agent interactions3 that: In the long run, you cannot use information without revealing it; you can use information only to the extent that you are willing to reveal it. A player with private information must choose between not making use of that information and then he doesn’t have to reveal it or making use of it, and then taking the consequences of the other side finding it out. (Hart, 2005, p. 694)

Thus, the link between information, action, and inference was made explicit, and “information” became part of the province of standard economic analysis. But critically, information could not be analysed just like an ordinary good, using standard tools, as Stigler (1961) had hoped.

2.

WHY MARKETS WITH IMPERFECT, COSTLY AND ASYMMETRIC INFORMATION ARE NOT EFFICIENT: A BRIEF SUMMARY

A central result of the economics of information is that even seemingly competitive markets are not constrained Pareto efficient – with the term “constrained” emphasizing taking into account the costs of producing and disseminating information. Importantly, there are public interventions (using the constrained set of available information) that could make everyone better off. There are several interrelated ideas that help us understand why that is so. Perhaps the most fundamental is that information is a public good – what one individual knows does

22  The Elgar companion to information economics not detract from what another does (information is non-rivalrous). The implication of this is that information should be distributed as freely as possible – but doing so would obviously not be in the interests of any individual or firm with information they believe to be valuable and not widely known. Nonetheless, it may be difficult to exclude some individual from the benefit of information provided by another.4 If informed individuals buy more of a stock because they know it will do well, the increase in the price of the stock conveys information from the informed to the uninformed. This makes it impossible for those expending resources to obtain the full social return from those expenditures. There will be underinvestment in research finding out about what are good investments, reflecting the general principle: Markets on their own are never efficient in the provision of a public good. More generally, welfare losses are associated with the incomplete appropriation of the benefits of “information externalities” (Stiglitz, 1975b,5 1975c, and Leitzinger and Stiglitz, 1984 provide a specific application in the context of oil exploration) and the pecuniary externalities that are pervasive when there is imperfect and asymmetric information (Greenwald and Stiglitz, 1986). Whenever information is costly to obtain and transmit, there is imperfect and asymmetric information, and whenever that is true, there may not be a full set of markets – certain markets may be “shut down”, in particular key risk markets.6 But while Arrow (1964) showed that the existence of a full set of markets was a sufficient condition for Pareto optimality, subsequently Stiglitz and others (Stiglitz, 1982a; Greenwald and Stiglitz, 1986; Geanakoplos and Polemarchakis, 1986) showed that generically, whenever there were not a full set of markets, the market economy is not constrained Pareto efficient, i.e. having a full set of markets was essentially necessary for Pareto efficiency.7 But “completing the market” by adding markets may be welfare-decreasing if it is not complete – just another application of the theory of the second best.8 More generally, whenever there are not a full set of markets, whenever there are incentive compatibility and self-selection constraints (which help screen individuals and ensure that they provide effort),9 and whenever there is costly search, there are pecuniary externalities that matter (Greenwald and Stiglitz, 1986, 1988; Arnott et al., 1994). Individuals assume, for instance, that health insurance premia are unaffected by how much they smoke, but when all individuals smoke, premia do increase. These results overturn a central pillar of standard economics, the first fundamental welfare theorem, the formalization of Adam Smith’s conjecture that the pursuit of self-interest would lead, as if by an invisible hand, to the well-being of society. The Greenwald-Stiglitz theorem establishing the constrained Pareto inefficiency of the market put a new light on one of the central ideas in economics, the use of the price system for decentralization. It was obvious that once it was recognized that observable actions affected information, individuals might change their actions to convey to others that they were, for instance, more able or less risky than they really were, and that others were. But given the imperfections of information, and that there would be costs, one way or another, in differentiating individuals (in the case at hand, through self-selection constraints, whether in a screening or signalling model), it was not obvious that the competitive market would not be constrained Pareto efficient nonetheless. Indeed, in some simple early examples involving only one commodity (such as the Rothschild-Stiglitz world), the economy was constrained Pareto efficient. But once one went beyond that special case, it was not in general efficient. Interactions across markets mattered, so unfettered decentralization is not efficient and designing interventions

Robust theory and fragile practice: part 1  23 that would enable efficient decentralization is very difficult; and when those cross-market externalities are particularly important, as in agrarian markets with sharecropping, decentralization simply breaks down.10 (See section 4 for a further discussion of decentralization.) While the focus of this survey is general theory/microeconomics, it should be obvious that these market failures have important implications for macroeconomic performance, including macroeconomic externalities (Korinek, 2012; Jeanne and Korinek, 2010; Davila and Korinek, 2017), financial frictions (Bernanke and Gertler, 1989; Greenwald and Stiglitz, 1993; Stiglitz and Greenwald, 2003; Stiglitz and Weiss, 1992), and labour market rigidities (Delli Gati et al., 2012).11 As we just observed, the dimensionality of the price system within existing markets is lower than the dimensionality of relevant information, so prices alone cannot in general convey all the relevant information, and existing prices may do “double duty”, conveying information not just about scarcity. Much of the formal discussion below is concerned about how, in such situations, additional information (say, about quantities) conveys further information, overturning one of the central results of standard economics that prices convey all the relevant information. And the fact that that is so not only changes the economy from what it would look like in a world where, say, quantities did not convey information, but even results in the economy not being constrained Pareto efficient. These results also overturn another pillar of standard economics, the notion that markets are informationally efficient, conveying all the relevant information from the informed to the uninformed.12 There is still one more reason that markets with imperfect and asymmetric information are not constrained Pareto efficient: a central insight of the earlier information literature was that competitive market equilibria may be characterized by markets not clearing.13 Whenever that is the case, shadow prices (say, of capital in the presence of credit rationing) will not equal market prices, and not surprisingly, market allocations will again not be Pareto efficient.14 Finally, markets with imperfect information are likely to be imperfectly competitive, for a whole variety of reasons. The fact that it is costly to search gives firms market power over their customers and workers.15 The fact that current employers have more information about their employees than others creates an impediment to labour mobility, giving even more market power to employers (Greenwald, 1979, 1986). Because of space limitations, we do not discuss this important source of market failure further in this chapter. These market failures are critical to understanding behaviour in many of the key markets in the economy (insurance markets, financial markets, labour markets), which also have large macroeconomic consequences, and obviously important policy implications.

3.

IMPERFECT AND ASYMMETRIC INFORMATION AND IMPLICATIONS FOR MARKETS

3.1

Voluntary, Truthful Disclosure and the Walras Law of Screening

There are some settings where information may be verified (perhaps by an independent third party, such as a notary, a mechanic in the car example, an evaluation by an outside healthcare professional in the insurance example, or a standardized test in the education example). What happens when information is imperfect and asymmetric, but verifiable, in the sense that there

24  The Elgar companion to information economics is something the imperfectly informed party can do to confirm (or disconfirm) the information, or find out the quality or type? The central mechanism driving outcomes in this setting was first observed by Stiglitz in 1975, when he wrote: […] assume the most able is able to provide information certifying to his abilities. The market would then, in equilibrium, pay the remaining workers their (now lower) mean marginal productivity. It would clearly pay, then, for the most able person of this group to have his ability certified. And the analysis proceeds, until information about the capabilities of all individuals except for the least capable is provided: but if we have sorted out all except for the least capable, we have also sorted out the least capable. This may be called the Walras Law of screening information. (Stiglitz, 1975a, p. 287)16

This key result, also known as the “unravelling” result, says that under certain conditions, with verifiable information, private information gets fully revealed in equilibrium (Stiglitz, 1975a). It always holds if it were costless to get a credential verifying one’s ability. This “Walras law of screening” was subsequently studied by Milgrom (1981), who confirmed that this was the unique equilibrium when information and verification is costless. Thus, the conclusion of this literature (known as the verifiable disclosure or hard information literature) is that if information is costlessly verifiable, in equilibrium it always gets revealed. Market outcomes in which different groups are not differentiated (because of a lack of information) were referred to by Rothschild and Stiglitz (1976), and Stiglitz (1975a), as “pooling” equilibria. Stiglitz (1975a) showed that if verification is costly, there may not be full information revelation; there may also be a pooling equilibrium. Only the more able individuals benefit from the validation, and so they will expend the resources to do so if the costs of screening are small enough. In the pooling equilibrium, the wages of the more able reflect the average ability. If the costs of verification are high enough, the difference between the wage they get if they show they are more able – their true productivity – and that average is smaller than the cost of verification, so a pooling equilibrium can be sustained. Thus, in this economy with costly verifiable credentials, there may be multiple equilibria – both a pooling equilibrium (where no one’s ability is revealed) and a separating equilibrium (where everyone’s ability is revealed). Interestingly, in this case with multiple equilibria, everyone in the pooling equilibrium is better off than in the “separating” equilibrium; still, the Pareto inefficient equilibrium can be sustained.17 Much of the subsequent literature has explored situations where these various forms of equilibrium may arise: pooling, separating, and hybrid (partially pooling/partially separating); whether a competitive equilibrium exists at all; and analysed the welfare economics of each pattern of equilibrium. While screening with verifiable information generated multiple equilibria, including a pooling equilibrium, in models with screening with self-selection (Rothschild and Stiglitz, 1976), there was no competitive equilibrium if the differences among individuals were small, and there never existed a pooling equilibrium. (We elaborate on this below in subsection 3.10. See also the discussion of repeated interactions, subsection 3.9 below.) The welfare analysis entails ascertaining the differences between private and social costs and benefits of differentiating. For instance, in the context of screening/signalling, social returns are related to the better resource allocations that are generated by better information – which may be nil;18 while the private returns include the increased rents associated with being identified as better (more able), in the context of labour markets, having higher returns

Robust theory and fragile practice: part 1  25 in the context of investment markets, higher probability of repaying, in credit markets, or, in insurance markets, a lower probability of the insured-against event occurring. The simple appropriation of ability rents increases inequality without increasing productivity. But these results with verifiable disclosure of information contrast markedly with those with self-selection. There, the equilibrium, if it exists, is Pareto efficient;19 but when there exists no equilibrium, Pareto efficient outcomes can be sustained with cross subsidies from the policies purchased by low-risk individuals to those purchased by the high risk (Stiglitz, 2009). 3.2

Different Structures of Information Asymmetries and the Existence of Competitive Equilibrium

In the models discussed so far, the more able individuals know they are more able. There are, however, some contexts in which that is not the natural assumption. For instance, life insurance companies may have better information about the correlates of longevity than ordinary individuals. In that case, again it is the more informed party that pays for the screening: it is advantageous for the life insurance company to screen individuals to determine which individuals are likely to have higher longevity – so long as that information does not become public. But if it does – if others can see that the insurance firm has offered the individual insurance at a low price (reflecting low risk), then other companies will do so also, and the firm doing the screening would be unable to appropriate the returns to its investment in screening. Thus, a screening equilibrium can exist only if markets are (as in Grossman and Stiglitz, 1980) at least partially informationally inefficient, i.e., only if the actions of the screener are only partially observable. Emran and Stiglitz (2009) show that this insight, the inability to appropriate returns from screening, may play an important role in inhibiting lending to young entrepreneurs. Moreover, there are problems in sustaining a competitive equilibrium when firms do the screening, even if the information they obtain doesn’t leak out. Indeed, if even two firms uncover the information about the skills of a particular individual, for example, Bertrand competition amongst them will result in all of the gains going to the individual and the firms that have expended resources to screen will lose money. With no one screening, it would pay a single firm to screen. But if two firms screen, everyone loses money. The only equilibrium is a mixed strategy equilibrium, where firms randomly choose who to screen, with some individuals then being screened only once while others have their abilities identified by multiple firms (Stiglitz, 1975b). That raises several questions, to which we now turn. 3.3

Endogeneity of Information Asymmetries: Learning About Oneself

First, if initially there are no asymmetries of information, but individuals know that there are differences in individual abilities,20 would it pay them to first identify their abilities for themselves, and then, if they are able, spend still more money verifying this for third parties? From a social perspective, such expenditures are problematic because if there were no productivity benefits from such information, ex ante, behind the veil of ignorance, all individuals are better off in the pooling equilibrium, as we have already noted. Yet Stiglitz (1984) shows that there exists an equilibrium in which nonetheless everyone pays to find out their abilities and then the more able provide verification. This is socially unproductive and simply increases inequality. On the other hand, if different individuals have different comparative advantages,

26  The Elgar companion to information economics then the information will be socially productive. Still, there is no presumption that the market equilibrium is efficient even when there are some social gains from better resource allocation, and in particular, there is no presumption that those gains outweigh the social welfare loss of increased inequality. Another strand of the endogenous information acquisition literature explores when one wouldn’t acquire a costly signal (see Sims, 2003; Matějka and McKay, 2015; Caplin and Dean, 2015; Caplin et al., 2022; Maćkowiak et al., 2019). 3.4

Excessive Investment in Information

The excessive investment which creates the asymmetries of information compounds the inefficiency noted earlier that arises in providing verifiable information about ability differences. This is not the only situation in which equilibrium may be characterized by excessive information, where there are social costs to obtaining information that are incommensurate with the benefits – indeed, the social benefits to an abundance of information may be negative (Hirshleifer, 1971). Insurance markets are particularly fragile: disclosure of genetic information may lead to the unravelling of certain insurance markets21 (see Rothschild and Stiglitz, 1997). In some cases, banning the use of certain information may be Pareto efficient (this may be the case for anti-discrimination laws which move the economy costlessly from a discriminatory equilibrium to a Pareto superior non-discriminatory equilibrium (Stiglitz, 1973, 1974b)); but in other cases, such interventions lead to the use of correlated variables, resulting in an even less desirable equilibrium (Rothschild and Stiglitz, 1982). There is another situation where too much information is “extracted”: that where firms have some market power. Such firms have an incentive to acquire information that may enhance the ability of the firm to price discriminate, to extract more of the potential consumer surplus from its customers. Extracting the consumer surplus is not simply distributive, with money often going from poorer consumers to the richer owner of firms; it is a costly “adverse distribution”. This rent extraction can be a major source of the firm’s profits as well as the distortions associated with market power (Stiglitz, 1977). 3.5

Creating Information Asymmetries

In some cases, economic agents may deliberately create information asymmetries.22 For instance, Edlin and Stiglitz study a setting where firm managers choose to invest in riskier (i.e., noisier) projects because it increases perceived uncertainty about the firm’s prospects, and therefore discourages either other firms taking over the firm or competing managers from displacing the current manager (Edlin and Stiglitz, 1995). It is not just the uncertainty that matters; it is that outsiders know less (have less precise information about the firms’ assets) than the manager. Thus, the level of informativeness of the decision maker, and the level of informativeness of others, is determined endogenously, and in ways which may be suboptimal. 3.6

Disclosure Requirements

While the analysis of sections 3.3 and 3.4 suggested that there may be situations where in equilibrium there is, in some sense, too much information, public policy has been concerned with cases where, without government intervention, there is too little disclosure. In credit markets,

Robust theory and fragile practice: part 1  27 lenders are required to disclose their true effective interest rate. The US Consumer Financial Protection Bureau (CFPB) likewise requires lenders to state their terms transparently. Here, the concern is that lenders present information in ways in which it is not fully understood by a large fraction of borrowers. There is an attempt to deceive. The US Securities and Exchange Commission (SEC) regulations typically require firms to disclose truthfully all materially relevant information about the securities they are issuing. These regulations go beyond requiring “the truth, nothing but the true, but not necessarily the whole truth”. Implicitly, they take the view that knowing that there is a serious downside risk to an investment and not disclosing it is effectively a lie. Their perspective rejects caveat emptor, which effectively puts all the burden of information on the buyer. If the seller knows something that he reasonably should know would be relevant for the buyer, he must disclose. There are several justifications for such disclosure requirements and several contexts in which there will be insufficient voluntary disclosure. A strong assumption in the above analyses yielding full disclosure is that of rational expectations – that the uninformed party understands the distribution of the qualities of the products being purchased, given the (limited) information being provided; she simply doesn’t know which item is of which quality. Moreover, she rationally makes inferences from what is observed. Behavioural economics has taught us that that is not the case. Cognitive limitations may lead individuals to make the wrong inferences, and firms have learned how to better “deceive” individuals. There are extensive literatures documenting persistent, systematic biases in consumer behaviour (including incorrect estimates of own future behaviour); the fact that these biases are exploited by firms can lead to subpar choices with regard to health, borrowing, saving, and investing behaviour.23 Providing standardized disclosure statements enables better and less costly assessments of the relative merits of different products or investment opportunities. Moreover, as we comment further below, providing only partial information imposes costs on other market participants. Some may have such high costs of screening that they simply accept the randomness of the product. But others with low costs of screening differentiate, but these are costs they would not have to bear in the presence of required disclosures.24 An analogy may be useful. In the economics of liability, Calabresi (1970) argued that the burden should be placed on the party that could avoid an accident at lowest cost. Here, not disclosing imposes significant costs on others, either in terms of making a suboptimal product selection or imposing costs of obtaining information – costs that could easily be avoided or reduced if there were disclosure.25 We noted above the importance of standardization. Standardization in the way that information is presented is important in a world with a superabundance of information: firms who do not want to make disclosures would otherwise comply with disclosure rules by burying the relevant information in a large volume of irrelevant information designed to overwhelm the attention and information processing capacities of the reader.26 It is important that information not only be disclosed, but be disclosed in a way that can be analysed and interpreted at relatively low cost. This is a form of signal jamming which is discussed more extensively in the next chapter. 3.7

Fraud Laws

If there were no costs to verifying, as we have noted, then the equilibrium that emerges is one in which everyone truthfully discloses. But if there is a cost to verification, and if everyone

28  The Elgar companion to information economics were telling the truth, it wouldn’t pay anyone to spend the upfront cost of verification. Hence, an equilibrium with full information couldn’t be sustained. There are several ways around this conundrum. One is that the individual provides a guarantee that his statement is truthful, giving the money back (plus the cost of enforcement of the guarantee plus an arbitrarily small additional cost).27 Then, no one would have an incentive to lie. (A guarantee, of course, is not just a signal; it is also an insurance policy. Even if there were no asymmetries of information, the seller might provide insurance because it is in a better position to absorb the risk of the non-performance of the product. The fact that it thereby has an incentive to make the risk of non-performance lower is also important.) An alternative is criminalizing fraud – punishing untruthful statements with a penalty at least equal to the amount received.28 Not disclosing materially relevant information, particularly information which is easily at hand, may, as we have noted, be viewed as akin to fraud. There is a remarkable paucity of literature on fraud, in spite of its importance (but see Greenwald and Stiglitz, 1992). Indeed, fraud played a critical role in the financial collapse of 2008 (Stiglitz, 2010). Since with a guarantee, there still has to be an arbiter of “truth” (is the product what the seller claims it to be?) and the guarantee may have to be enforced through a court of law, the basis of “truthfulness” in both cases is legal enforcement. This is where the assumptions of verifiability and enforceability become central. In the general case, information is only imperfectly or imprecisely verifiable, particularly to third parties. On the other hand, in the context of repeated interactions (repeated games), reputation may serve as an enforcement mechanism and third-party verifiability may not be required. See the brief discussion in section 3.9 below. The issues under discussion here become particularly relevant in a world of mis- and disinformation – a world into which we seemed to have descended. The early literature, for the most part, while recognizing information asymmetries, assumed that when information was disclosed, it was truthful. The discussion in the next chapter explores these issues further. 3.8

Costly Signals and Self-Selection

In many cases, rather than the costly acquisition of a verified credential, market participants look toward a “surrogate” for the credential, a costly signal that is correlated with the relevant, unobservable ability or quality, and a major strand of the literature has focused on this (Rothschild and Stiglitz, 1976, hereafter referred to as RS, and Spence, 1973). Sometimes the surrogate is simply a more easily observable characteristic: if race is correlated with education, and education is correlated with ability, but race is more easily observable than education, race may be used to screen individuals, resulting in a discriminatory equilibrium (Phelps, 1972; Arrow, 1973; Rothschild and Stiglitz, 1982; Stiglitz, 1973, 1974b). Most of the literature has focused though on cases where the surrogate is a decision variable, and knowing that a change in action may affect the inferences about, say, ability has fundamental effects on the nature of market equilibrium. The better-informed party may try to signal his qualities. Or the uninformed party may structure a set of choices where the choices reveal who is better: self-selection constraints work to enable the firm to “screen”. When there exists a separating equilibrium (where there is full disclosure, the “unravelling” equilibrium discussed in the previous section), a self-selection (RS) equilibrium and a signalling equilibrium are typically the same.

Robust theory and fragile practice: part 1  29 The array of actions that convey information – and therefore may well be affected – is enormous: the quantity of insurance as in RS, the level of education, as in Spence, the number of hours one is willing to work a week, as in Akerlof (1976), the willingness to engage in search, as in Salop (1977). A large literature has developed exploring the mechanisms that might be used in different contexts. A key implication was noted in the beginning of the chapter: virtually all behaviour is affected in one way or another. In the signalling model (Spence, 1973), the informed party moves first, taking a costly action (such as a level of education), which is a signal of the attribute or quality in question (in the context of education, the individual’s ability). Although expenditures on the signal are assumed not to increase the individual’s productivity – thus signalling is purely dissipative – the action verifiably differentiates her from others. In the screening models (RS), the uninformed party moves first, structuring a set of choices (actions) that differentiates among individuals.29 The two strands of literature (testing/verification, signalling/self-selection) are actually more closely related than has been recognized. Assume that anyone can pass the test for “high ability” if they put in enough effort. Thus, passing the test is a noisy signal of ability, confounded by another unobserved variable, effort. But if the test is hard enough, then it would not pay the low-ability person to exert the effort required to pass the test. Passing the test of sufficient severity thus becomes a costly signal. The cost of the signal – the cost of information – is thus both the cost of the test and the effort to pass the test. Of course, sometimes the “action” has value in its own right – getting more education increases productivity. Then the social cost of the screening/signalling information is the expenditure on education beyond that which would be desirable in a world of perfect information.30,31 Much of the theoretical work in signalling models has focused on how the different parties reason upon observing signals that should not be used, that would not be observed in equilibrium. The literature on “refinements” that operates by restricting agents’ beliefs (Banks and Sobel, 1987; Cho and Kreps, 1987), and the rest of the literature that operationalizes the “stability” notion of Kohlberg and Mertens (1986) has provided a veritable bestiary of “equilibrium refinements” that aim to obtain “reasonable” outcomes in signalling games. They proceed axiomatically, stating desiderata for what good (what they refer to as a “stable”) equilibrium (among all the possible equilibria) looks like. For instance, a “good” outcome in the Spence signalling game is the least-cost separating equilibrium32 (the “Riley outcome”, Riley, 1979)33 where the high types obtain as little education as necessary (and thus incur as low a cost as possible) to distinguish themselves, and there is information revelation by virtue of the different ability types choosing different levels of education, and thus signalling their types to the employers. In other words, there is information revelation at the lowest possible cost to the participants.34 3.9

Multiple-Period and Repeated Interactions, Reputation, and Robustness

Even when characteristics cannot be verified by a third party – and therefore guarantees and the truthfulness of statements cannot be enforced – unravelling (a separating equilibrium) may occur through reputation mechanisms. In the reputation literature, repeated interactions allow some observability (without third-party verifiability), but for such mechanisms to work, outcomes or characteristics must be observed with sufficient precision (or signals of charac-

30  The Elgar companion to information economics teristics or actions, which are sufficiently correlated with outcomes or characteristics, even if imperfectly).35 In the absence of third-party verifiability, the main enforcement is through a cut off in relationships,36 which itself can be costly to both parties. Equilibria in reputation models with, say, unobservable effort have one distinct characteristic which distinguishes them from standard competitive equilibria: there have to be “rents”, with prices, for instance, in excess of costs. What induces good behaviour is the threat of losing those rents (Shapiro, 1982; Shapiro and Stiglitz, 1984) as a result of a cut off in the relationship. Screening and signalling in multi-period contexts also change incentives in a fundamental way: if an individual reveals who he is in one period, that information can be used in all subsequent periods. It increases the incentive of the more able to have himself distinguished from the less able, but also increases the incentive of the less able not to have himself distinguished from the more able. The latter effect may dominate: in a model with a finite number of periods, there may exist a pooling equilibrium for an initial set of periods (Gale and Stiglitz, 1989; Courty and Hao, 2000; Stiglitz, 2009).37 The literature on repeated games has also investigated how robust the predictions are to a misspecification of the problem (for instance, if the beliefs about the preferences, information, or behaviour of one or more of the agents are incorrect). Hörner, Ely and others have worked on so-called “belief free” equilibria (Ely et al., 2005; Kandori and Obara, 2006; Hörner and Lovo, 2009; Hörner et al., 2018), equilibria that are completely independent of any information held by others, and rely only on individual optimization, and are accordingly extremely robust to informational assumptions.38 A profile of action sequences (one for each player) in a repeated game is belief-free if after any history of actions (some parts of which may be unobserved), a player’s plan of action from that point onward is a best response to the plans of action of the other players that are in the profile of action sequences we started with, for any beliefs the player may hold about others’ plans of action (or, equivalently, regardless of any private information a player may have). In a belief-free equilibrium of a repeated game, at every point every player’s strategy of play from that point onward is optimal, given her information, independently of the information held by the other players. They are, by definition, robust to (in fact, completely independent of) how individuals form their beliefs. This is an extremely strong assumption. Technically speaking, it is a subgame perfect equilibrium for every game of complete information that is consistent with the player’s own information. Thus, players need not use Bayesian reasoning about others and adhere to the Harsanyi doctrine (discussed below). Of course, the assumption of individual optimization in complex environments (and even of maximizing expected utility) itself is an extremely strong requirement (see the discussion in the next chapter on behavioural economics). A small diversion into the theory of repeated games may be warranted. There exists a strong result in the theory of repeated games that states the following: take the stage game (the game repetitions of which will become the repeated game). Find equilibria of this stage game, the corresponding payoffs, and now take the so-called convex combination of those payoffs (that is, if in one equilibrium a player earns a payoff of 1, and in another equilibrium she earns a payoff of 2, a convex combination is any number between 1 and 2), and suppose that for each player, they are above a lower bound (a payoff that a player can guarantee herself, for any set of actions of her opponents). The folk theorem states that any such payoff (a convex combination of equilibrium payoffs that is above a lower bound) can be a payoff in the repeated

Robust theory and fragile practice: part 1  31 game, provided players are patient enough. Thus, the repeated game has incredibly large sets of payoffs, and therefore, even larger sets of equilibria. A typical reasoning for why the folk theorem holds goes as follows – suppose a player deviates while playing a game, taking an action that gives her higher instantaneous utility, but lowering the payoffs of others. It is then in the best interest of the other players to play something that “hurts” the deviating player enough (i.e., for long enough), that the deviator will not cheat in the first place. In fact, there is a family of such results, which vary the equilibrium concept (Nash, subgame-perfect, sequential, etc.), and vary what happens after a player deviates – can there be forgiveness, what happens if the punisher(s) deviates, how to evaluate infinite streams of payoffs, and so on. One important consequence of belief-free equilibria in repeated games is that typically, the folk theorem fails to hold if belief-free equilibrium is used as a solution concept: it’s not true that nearly any feasible payoff that is above a lower bound can be the outcome in a belief-free equilibrium. In fact, a belief-free equilibrium may not exist at all. The definition of belief-free equilibrium seems to be too strong to sustain any feasible equilibrium payoff, and while it may sustain some (even a “large” set of payoffs), this is still typically smaller than the set of payoffs sustained by the folk theorem. 3.10

Equilibrium – or Non-Existence of Equilibrium – Under Different Information Structures

A key determinant of the nature of the equilibrium is what is observable and what can be communicated to whom and by what means. For instance, the adverse selection problem long investigated in the insurance market assumed that quantities of insurance were not observable; market participants only know that the average quality of what is being traded on the market (the average risk of those willing to buy insurance) is affected by the price.39 If, as in Rothschild and Stiglitz (1976), quantities of insurance purchased are observable, then that information conveys information about the individual’s type: now it is not just prices but quantities that are informative. Thus, the nature of the equilibrium is highly sensitive to whether an individual’s purchases of insurances is observable. But both the adverse selection and the RS equilibria are fragile. As Kosenko et al. (2023) (or KSY for brevity) point out, a single insurance company can obtain some information about “quantities” – even if total purchases are not observable – simply by selling a large insurance policy. KSY establish that such limited information is enough to break the standard price equilibrium. Similarly, RS had shown that with full observability of quantities purchased there exists an equilibrium in a model of two types if the two types differ by enough. But if we now modify that world to include the possibility of “secret” insurance (i.e., the possibility that some insurance contracts are not observable), then there never exists an equilibrium – neither a price equilibrium, nor a quantity (RS) equilibrium, nor a price-cum-quantity equilibrium of the kind that we just described as breaking the price equilibrium. The disturbing implication of KSY is that in this more realistic world where there is some but not full observability of insurance purchases, an equilibrium never exists. We’ll see in the next chapter that if there can be direct communication between consumers and insurance firms and insurance firms with each other, this conundrum is resolved: there always exists an equilibrium, and it entails a pooling contract.

32  The Elgar companion to information economics In practice, exclusivity cannot be enforced. In the context of insurance, there are a large number of informal implicit risk sharing mechanisms; risk is, for instance, shared within families. A natural question is: are such informal risk sharing mechanisms welfare improving? At one time, there was the hope (and it was just that, a hope) that social institutions might “fill the gap” arising from the absence of markets or other market failures, reducing the need for government intervention. For instance, concerns about moral hazard limited the amount of insurance that the market would provide (with complete insurance, individuals would take no care). Arnott and Stiglitz (1991) showed, however, that these non-market institutions, which naturally arose to fill the gap, can lower welfare, crowding out more efficient market insurance.40 3.11

Information Extraction in Monopolies and with Imperfect Competition

The discussion so far has focused on how in a competitive world with asymmetric information, those information asymmetries are (partially) overcome, sometimes at great cost, and typically affecting in fundamental ways the nature of the equilibrium. In many ways, the problem of overcoming information asymmetries is simpler for a monopolist: he alone constructs the choice set; he doesn’t have to worry about another firm making an offering. With perfect information enabling perfect price discrimination, a monopolist could and would extract all consumer surplus – and there would be no monopoly distortion. Stiglitz (1977) shows how a monopolist can maximize rent extraction (i.e., capturing as much of the consumer surplus as possible, given the limitations in his information) in a situation where it knows different consumers differ, but can’t tell which are the ones with high consumer surplus. Most importantly, as we noted earlier, contrary to standard analyses, this attempt to differentiate among consumers is the real source of monopoly distortion. Salop (1977) provides a telling example, where a monopolist charges different prices in different stores – forcing unnecessary search – simply to enable him to charge a higher price to consumers with high search costs, who in his model have higher consumer surplus, increasing thereby the rents he can extract. In the next chapter, we explain how these problems may be exacerbated with the digital platforms and artificial intelligence, in ways which fundamentally undermine the efficiency of the market economy. Markets with monopolistic competition and oligopolies lie somewhere between the two polar cases of pure monopoly and competition: firms may have to be sensitive to the offerings of others (and therefore, of the information it can extract from its own offerings alone), but have greater discretion than they do in highly competitive markets, where, for instance, an attempt to “cream skim” the best customers is met with a strong competitive response. Not surprisingly, the equilibrium in general entails distortionary screening mechanisms, including price discrimination and price dispersions. When, for instance, individuals differ in their search costs, market equilibrium will be characterized by price dispersions, inducing unnecessary search, in an attempt to extract more money out of the high search cost to individuals, not unlike that which occurs under monopoly. Indeed, the only market equilibrium may be one with price dispersion, even when all firms are initially the same (Salop, 1977; Salop and Stiglitz, 1977, 1982).

Robust theory and fragile practice: part 1  33 3.12

Information Extraction by Governments

In the previous subsection, we saw how monopolists attempt to differentiate among customers in order to extract more of their consumer surplus. Governments too want to differentiate: equalitarian governments would like to tax more those who are more able, to redistribute to those who are less able. Regulators want to ascertain whether a utility should be allowed to charge a higher price, because its costs are genuinely higher. There is a close formal similarity between the solution to the problem of maximizing social welfare by a government or regulator, in the presence of asymmetries of information, and the problem of a monopolist maximizing its profits, in the presence of asymmetries of information.41 3.13

Summary of Standard Literature

Five fundamental insights that have emerged in the voluminous literature of the past nearly half century are (a) there may exist no equilibrium;42 (b) while in certain classes of models, the only equilibrium entails “separating equilibrium”, where those with different characteristics are effectively fully identified by the actions/choices they make, more generally, the equilibrium, when it exists, may entail pooling or partial pooling, when there is costly verification or in multiple period models where the uninformed party cannot commit not to use information gleaned in one period subsequently; (c) the equilibrium is not in general constrained Pareto efficient; (d) the precise specification of the “game” describing the interaction between the two parties matters;43 and (e) there exists a variety of interventions by government that can be welfare enhancing – price interventions and disclosure and fraud laws.

4.

HOW MARKET DEVELOPMENTS ARE UNDERMINING DECENTRALIZATION AND INCREASING INFORMATION BURDENS

As we have seen, a central idea of the “information” revolution in economics is that prices do not convey all the relevant information. Changes in market structure can accordingly change the informational efficiency of the economy by increasing or decreasing the burden put on various sources of information and can affect the effectiveness of these sources in communicating the relevant information. There can be a tension here. Seeming improvements in markets, ignoring information, may actually worsen overall economic performance once the informational burdens, or the consequences of not getting relevant information, are taken into account: still another instance of the general theory of the second best. Trade liberalization is an obvious example. There are standard arguments for the benefits of trade integration.44 But as markets get integrated, the relevant set of information for market participants gets greatly increased; because volatility of prices, for example, may increase, market participants will want more information about factors affecting price. Whether welfare is increased can thus be ambiguous. An even better example is the creation of derivatives, designed to “complete markets”, or fill in for some of the missing Arrow-Debreu risk markets. But especially the way they have been constructed, they have increased information burdens enormously and undermined effective decentralization, especially the way they have been constructed. One of the great

34  The Elgar companion to information economics virtues of the market economy was supposed to be that no one had to know the preferences or technology of others. Prices were, as we have noted, sufficient statistics, conveying all the relevant information. But if, say, bank A has taken out a credit default swap (CDS) with bank B betting that C will default, A’s financial position depends on B’s ability to fulfil that contract (the counterparty risk); but if B has CDS’s with D and F, then B’s ability to fulfil its contract depends on judgements concerning D and F’s ability to fulfil their contracts, and that will depend on all of the CDS’s that they hold. In short, risk assessment of any bank requires knowledge of the balance sheets (including derivatives holdings) of all banks and firms that are intertwined in this financial network, almost surely all large banks and many other large enterprises. Decentralization has been undermined; overall information burdens increased enormously.45 Making matters even worse is that creating these new financial products opens up new betting opportunities, which at least some firms and individuals will take advantage of. This increases the volatility in their wealth positions, increasing macroeconomic volatility and increasing still further information burdens (Guzman and Stiglitz, 2020, 2021a, 2021b).46

5.

GOING BEYOND MARKETS: MECHANISM DESIGN

Markets are one way of allocating resources. Individuals reveal their preferences by their purchases and sales at particular prices; and firms, their technological capacities. With prices set according to certain rules, i.e., to clear markets, the outcomes are efficient. Once it became clear that with imperfect, asymmetric, and endogenous information price signals were no longer enough to coordinate economic activity efficiently, and that asymmetric information is a pervasive feature of economic environments, there was a push to explore other ways for market participants to communicate (e.g., information about preferences or technology) and make inferences, and other ways for allocating resources. This is especially important when there are too few market participants to make the price-taking assumption of competitive equilibrium persuasive, and in particular where there is a single agent – a monopolist or the government – that can create a “mechanism” to efficiently extract relevant information in ways that maximize its objective function. Shortly after the development of the basic information models, tools from mathematics – in particular, game theory – were beginning to be used more widely by economists, with extraordinary results. Mechanism design (sometimes known as “implementation theory”) can be thought of as the flipside of game theory – instead of starting with a game and looking for a solution, one starts with a desired outcome (the solution) and tries to find games (or, more technically, “game forms” – games where everything but the payoffs are specified) in which a specific desired outcome is an equilibrium. The participating agents send “messages” to a centralized “mechanism”, which then, depending on which messages the agents send, decides on an allocation – quantities and prices.47 The task of the mechanism designer is, given a particular social choice function, to find mechanisms in which the outcome of the game that the agents play, with each maximizing her own welfare according to the rules of the game, maximize this social choice function. Part of the reason “implementation” theory is called so is that it analyses, given a fixed planner’s preference ranking over outcomes, which outcomes can be implemented by an appropriately chosen game, and how – what does the game form look like, what are the neces-

Robust theory and fragile practice: part 1  35 sary transfers, and so on. The constraints that the planner faces may be complex – agents may have private information, may have aligned or misaligned preferences – so the implementation problem is difficult in general. Because of this, this literature is rife with impossibility theorems: Arrow, Gibbard-Satterthwaite, Muller-Satterthwaite, and others. So, the contribution is not always of the form “this is the optimal mechanism”, but rather of the form “the exists no mechanism that accomplishes all goals”, or that the only possible mechanism has very undesirable properties (such as being “dictatorial” – always ranking outcomes according to one agent’s preferences, and always disregarding those of the others). The literature on mechanism design/implementation theory, like that of game theory more generally, shows that results are highly sensitive to assumptions about information, about the players’ preferences, and about players’ beliefs, including their beliefs about the beliefs of the other players; in particular, the mechanism designer is often assumed to have a lot of information. Below, we consider some of the most salient aspects. (a) The design of economic mechanisms relies on agents’ beliefs about an uncertain, payoff-relevant random variable. The approach to studying games of incomplete information (games where players are uncertain about the payoffs that other players obtain – i.e., they do not know which game they are playing) is due to Harsanyi (1967, 1968a, 1968b); instead of analysing incomplete information games, he proposed analysing games of complete but imperfect information (where players know the payoffs, but do not observe previous moves of other players). His construction involved augmenting the game with an additional player, called Nature, who moves first, and chooses realizations of a random variable for each player according to a distribution known to all players; this realization is called the player’s type. A type summarizes all of the information about the player – whether she is of high or low ability, has high or low costs, etc. This also allows a player to form beliefs about others’ types, and their beliefs about one’s type, and beliefs about beliefs – an infinite hierarchy of beliefs; a type is assumed to include one’s entire hierarchy of beliefs. The languages of hierarchies of beliefs, and type spaces, are complementary.48 The profile of types (one for each player) is simply a vector of types. In Harsanyi’s own words: we can regard the vector ci as representing certain physical, social, and psychological attributes of player i himself in that it summarizes some crucial parameters of player i’s own payoff function Ui as well as the main parameters of his beliefs about his social and physical environment … the rules of the game as such allow any given player i to belong to any one of a number of possible types, corresponding to the alternative values of his attribute vector ci could take. … Each player is assumed to know his own actual type but to be in general ignorant about the other players’ actual types. (Harsanyi, 1967, pp. 171–172)

(b) But as early as 1972 Leo Hurwicz recognized the need for mechanisms that do not depend on assumptions about the agents’ characteristics, what he called “nonparametric” mechanisms (Hurwicz, 1972). (c) Most importantly, mechanism design depends on individuals’ beliefs about other agents’ beliefs (hence the term “hierarchies” of beliefs). Aumann’s (1976) work on “common knowledge” has been crucial for clarifying what it means for agents’ beliefs about others (and beliefs about beliefs – second-order beliefs, and beliefs about beliefs about beliefs, and so on) to be commonly known. An event E is common knowledge between two agents if agent 1 knows it, agent 2 knows, agent 1 knows that 2 knows it, and so on, ad infinitum. Aumann (1976) showed

36  The Elgar companion to information economics that if individuals have common priors, and their posteriors are common knowledge, their posteriors must be equal: they cannot agree to disagree.49 An approach based on common knowledge has been criticized, most notably by Robert Wilson. Wilson emphasized the importance of not making strong informational assumptions, such as common knowledge, in hopes of generating better, more applicable, theories. He wrote, outlining what has become known as the Wilson doctrine, “only by repeated weakening of common knowledge assumptions will the theory approximate reality” (Wilson, 1985, 1987). An example of work in this spirit is Dasgupta and Maskin (2000), who focus on “detail-free” auction rules “that are independent of the details – such as functional forms or distribution of signals – of any particular application and that work well … in a broad range of circumstances” (Dasgupta and Maskin, 2000, p. 347). 5.1

Blackwell’s Theorem: A Cornerstone of Information Measurement

We have been discussing the idea of “information”, and using terms like “information structure”; while being more or less “informed” may have a clear and intuitive meaning in many of the early “simple” applications, this is not so in general. There is, in fact, a large theory that provides measuring the amount of information,50 or at least ascertaining whether there is more information in one situation than in another, in much more complicated settings. A cornerstone of this theory is David Blackwell’s (1951, 1953) work. Observe first that generally speaking, different decision makers will value different pieces of information differently, depending on their preferences, other assets, and risk attitudes. For instance, an investment analyst who is interested in the federal funds rate may not value information about the climate, which may be very valuable to another decision maker. Thus, there can be no unanimous ranking of all information structures, by which everyone could agree that one set of information is more informative or “better” than another. But is there any context in which all expected utility-maximizing decision makers would agree that one set of information is more informative than another? As Blackwell (1951, 1953) showed, the answer is yes: information structure A is more informative than information structure B if any payoff that is attainable under B is also attainable under A. This “payoff richness” appears to be a natural criterion of evaluating information for an economist. There is another perspective one can take. Suppose one takes information structure A and adds pure noise to it, to obtain information structure B. This seems like a very plausible requirement for A to be more informative than B (we say that A is “sufficient” for B). The intuition is that one can turn A into B without knowing anything about the true state. A striking result of Blackwell is that these two ways of evaluating informativeness of information structures are, in fact, identical. Information structure A is payoff richer than B for any utility function if and only if A is sufficient for B.51 This result provides not only a completely unambiguous ranking of information – payoff richness – but also links it to a mathematically tractable and statistically appealing way of ranking information – sufficiency.52 The strength and power of this theorem – it applies to all expected utility maximizers – is also its shortcoming. The Blackwell order is not only partial (meaning most information structures are not ranked) but, loosely speaking, “very” partial. There have been many attempts to “complete” the Blackwell order (Frankel and Kamenica, 2019; Kosenko, 2022; Cabrales et al., 2013, 2017; Pęski, 2008; Mu et al., 2021); no single completion has been found to be appealing in all contexts.

Robust theory and fragile practice: part 1  37 Using the Blackwell framework, Radner and Stiglitz (1984) were able to derive a striking result about the value of information under very weak assumptions: it is always non-concave. Ignorance may not be bliss, but it is at least a local optimum. Much of standard economics (exemplified by Debreu, 1954, 1959) is based on assumptions of concavity. In a world with endogenous information, it is hard to hold that assumption.53 5.2 Auctions Auctions are perhaps the most widely developed example of a commonly used mechanism for allocating resources, as was recognized by the 2020 Nobel Memorial Prize awarded to Robert Wilson and Paul Milgrom – two of the leading figures in auction design.54 The literature on mechanism design has called attention to better ways of designing auctions under increasingly complex situations. Many of these ideas have been put into practice, with mixed results. While there have been some successes on real-world auctions (radio spectrum auctions in the US, auctions for 3G licences in the UK in 2000), there have also been many disappointments, with unexpectedly low prices (including the Australian satellite TV auction, the New Zealand radio spectrum auction, and the US Environmental Protection Agency’s SO2 auction), some of which can be attributed to collusion by the bidders, or poor auction rules design. 5.3

Matching, or Markets without Prices

Alvin Roth was awarded a Nobel Memorial Prize in economics for his work on what can be viewed as a particular class of mechanisms aimed at “matching”, e.g., kidney donors with recipients,55 scarce medical equipment (such as ventilators) with patients who need them,56 recent MD graduates with hospitals for internships, or high school applicants with public high schools.57 Matching theory is useful in situations where there is a supply and a demand of an item or service (or more generally, just two sides of the exchange), but there is no price (for legal, moral, or technical reasons). Some, like the kidney donor programme, have been huge successes; others, like the high school programmes, have been more problematic. Understanding better the nuances that lead to success or failure is a subject of ongoing research.58 Even when there is no central authority that can design a relevant mechanism for resource allocation (as in the cases just discussed), mechanism design may provide insights; for instance, the gains to trade that could be achieved under an optimal design provide an upper bound to those achievable. 5.4

Consequences of Information Assumptions for Mechanism Design

We noted earlier the sensitivity of many of the results in mechanism design to the particular informational assumptions employed. The literature has naturally asked how the optimal mechanism might be changed as the information available to the economic agents changes (for example, in agents’ beliefs about their opponents’ valuations). In particular, Bergemann and Morris (2005) have created a framework of “robust” mechanism design, investigating, for instance, how the design of auctions or of systems of price discrimination depends on the assumptions of common priors and common knowledge.59 In particular, (the main question) they consider is this: suppose that the social planner (or mechanism designer) has a social

38  The Elgar companion to information economics choice function she would like to implement. (A note on the nomenclature: this literature distinguishes between “social welfare functions”, the domain of which is individual preferences, and the range of which are rankings of alternatives, “social choice functions”, which have the same domain, output single alternatives (not a ranking, as with social welfare functions), and “social choice correspondences”, the domain of which is the same, but the output may be multi-valued – they produce a set of unranked acceptable alternatives.) What the planner should choose to maximize her social choice function may depend on variables that are only known (initially) to agents. Obviously, the planner would like the agents to convey this information, but they don’t just freely do so. They have to be induced to do so through a “game”. Bergemann and Morris (2005) ask: When does there exist a game with the property that when the agents play that game, for any information they may have, each agent has an incentive to tell the truth about her information, if she expects others to tell the truth, whatever their information turns out to be, and that the outcome of this game always implements the social choice function, i.e. entails, when all the information is revealed, the planner taking an action consistent with what maximizes her utility given the information (state of the world)? It is the requirement that the agents have an incentive to tell the truth if they expect others to tell the truth, whatever others’ information turns out to be, that makes this robust (ex post implementable) – the agent has no incentive to lie, provided others do not lie, regardless of their information. This is a very strong concept (akin to the definition of a dominant strategy – a strategy that is a best response to any strategy of one’s opponent). It is obviously desirable to find such a robust mechanism, because robustness in this sense means that other outcomes are extremely unlikely to result in practice, by construction. 5.5

Applications of Robust Mechanism Design

Two important theoretical applications of robust mechanism design, where the design reflects the lack of information of the designer, are Carroll (2015) and Guo and Shmaya (2019). Carroll (2015) studies a moral hazard problem where the principal is uncertain about the set of actions available to the agent. “She [the principal] knows some of the available actions, but other unknown actions may also exist and our principal does not even have a prior belief about these unknown actions”. He finds that the solution (under very mild assumptions concerning preferences and the ability to impose taxes, e.g., in outcome-dependent ways, under which the first best outcome could be trivially implemented) to this problem is linear contracts – i.e., that if a contract is judged by its worst possible performance, linear contracts uniquely provide the highest guaranteed return to the principal (i.e. it is the “maximin” solution). Carroll (2015) uses this to explain the pervasiveness of linear contracts in the real world – where the principal knows that there may be eventualities that matter that are not even conceivable now, and obviously can’t form beliefs about things she does not know exist – as opposed to “optimal” contracts from non-robust contract theory that often take complicated forms.60 Guo and Shmaya (2019) study from a non-Bayesian point of view the same problem of monopoly regulation (again viewed as a principal-agent problem) that Baron and Myerson (1982) studied from a Bayesian approach. They identify policies that minimize “regret” of the regulator – the difference between what he could have obtained with perfect information and what he actually obtains. The result is an interesting combination of price caps (to benefit the consumer) and piece-rate subsidies (to incentivize the producer); the precise amount and

Robust theory and fragile practice: part 1  39 combination of these tools depends on how much weight the regulator puts on the two sides of the market.

6.

CONCLUDING REMARKS

The analysis of environments with costly information has provided enormous insights into the workings of virtually all markets and of the economic system as a whole. It has provided explanations of phenomena which could not be explained by the standard theory assuming perfect information. We have seen the multiple ways in which early results concerning information revelation, screening, and signalling have been reinforced and modified, with results in some cases being shown to be more robust than was at first thought to be the case, while in others, more fragile. Most importantly, the conclusion that markets are, in general, not (constrained Pareto) efficient has, if anything, been strengthened, and indeed, we have seen how many of the reforms of recent years may have actually decreased welfare, once one takes into account the consequences of the imperfections and endogeneity of information. Recent work has highlighted the potential of information asymmetries to magnify distributional differences, entrench market power, and make efficient regulation more difficult, but has also suggested ways of at least partially overcoming these problems and improving resource allocations. In particular, we have seen how a greater understanding of how the problems posed by information asymmetries can be overcome continues to generate new insights; ingenious new mechanisms have emerged that implement desirable outcomes when intuition might suggest such outcomes may be impossible. The companion chapter that follows explores a particularly fruitful strand of work – where information is conveyed not just by making inferences based on choices and actions, but also by direct communication.

ACKNOWLEDGEMENTS We are grateful to Daphne Raban and Julia Włodarczyk, volume editors, for their work on improving the chapter, and for inviting our contribution, to Andrea Gurwitt, for editing, and to the many collaborators that worked with us over the years, Filippo Pavesi, Massimo Scotti, and all the contributors to the volume for their feedback. Joseph Stiglitz gratefully acknowledges financial support from the Institute for New Economic Thinking.

NOTES 1.

2.

See, e.g., Stiglitz (1985b) and Stiglitz and Wolfson (1988). There is a wealth of other applications, most which we cannot explore in this short chapter. These include regulation (itself a huge literature, including Sappington and Stiglitz, 1987a), the exercise of and regulation of market power (e.g., Baron and Myerson, 1982; Rey and Tirole, 1986; Laffont and Tirole, 1986, 1988, 1990), and privatization (Sappington and Stiglitz, 1987b). Indeed, there have been multiple surveys, both of the field in general (e.g., Stiglitz, 1975a, 2000, 2002, 2013b, 2020; the introductory essays in Stiglitz, 2009, 2013a; and Kamenica, 2017) and its application to particular subdisciplines, including labour (see Ashenfelter and Card, 2011a, 2011b), product (see Stiglitz, 1989), and insurance (see Dionne, 2013) markets. Veldkamp (2012) discusses

40  The Elgar companion to information economics

3. 4.

5.

6.

7.

8.

9. 10.

11. 12.

13.

applications of imperfect information in macroeconomics and financial markets. By the same token, we necessarily must be selective in the references we choose to cite. We make no attempt either at completeness or comprehensiveness. Although, notably, this insight fails to hold with more than two decision makers, as the following discussion will make clear. The two key properties of a pure Samuelsonian public good are non-rivalrous consumption (what one person consumes doesn’t subtract from what another can) and non-excludability. Information has both properties, but we have noted even if it were possible to exclude partially (as patents do for certain kinds of information), the resulting market equilibrium is not efficient. This chapter notes the inefficiency associated with obtaining a particular kind of information, knowledge associated with increased efficiency in production, namely, research and development. Thus, the market failures noted here in the context of screening models are also relevant for investments in innovation. These inefficiencies are more extensively discussed in Stiglitz and Greenwald (2015). Akerlof (1970) showed that with information asymmetries, there might not exist a market for used cars, a specific application of a more generalized “no trade theorem” (Milgrom and Stokey, 1982; Stiglitz, 1982b). For applications of these ideas to labour markets, see Greenwald (1979, 1986). For applications to equity markets, see Greenwald et al. (1984). There are also fixed costs associated with establishing and running markets, providing an alternative explanation for the absence of markets. Diamond (1967) provided a weaker set of sufficient conditions for a weaker notion of “constrained Pareto efficiency”, but Stiglitz (1982a) showed that whenever there were more than one good or bankruptcy costs, markets were not constrained Pareto efficient even in the weaker sense that Diamond had proposed. Earlier, Stiglitz (1972) had shown even in the more restrictive world of mean-variance (the basis of the capital asset pricing model) the economy was not efficient. More generally, Greenwald and Stiglitz showed that given the set of markets in existence, there exist interventions in the market allocation of goods that could improve the welfare of some without decreasing that of others. Their result can be seen as a generalization not only of Stiglitz (1982a), but also of Newbery and Stiglitz (1982), who show the (constrained) inefficiency of markets even with rational expectations. Newbery and Stiglitz (1984) illustrate showing that opening up new opportunities to trade between two economies may lower the welfare of all individuals in both economies if there are not risk markets. Later, we expand on the specific reasons that adding additional securities – short of a complete set of risk markets – may be welfare decreasing. Or when there are collateral constraints, as in many recent macroeconomic models. The economics of information help explain the widespread institutional arrangement in agriculture of sharecropping, which from the perspective of standard economics seemed a peculiar economic arrangement, significantly reducing workers’ incentives. Indeed, sharecropping provided the paradigm of the “agency” problem, where effort was unobservable, but workers were risk averse (Stiglitz, 1974a). Braverman and Stiglitz (1982) showed in that context that there would be interlinking of markets – decentralization failed. For an application of these externalities to the Covid-19 pandemic, see Guzman and Stiglitz (2021c). There is a vast literature claiming that markets are informationally efficient (see, e.g., Fama, 1970, 1991), but an even more important literature establishing that they are not (Shiller, 1981, 2000). Grossman (1976, 1977) and Grossman and Stiglitz (1980) not only established that markets could not be informationally efficient in transmitting information from the informed to the uninformed, but that they even failed to aggregate disperse information well. Similarly, Gale and Stiglitz (2013) showed that futures markets are almost always informationally inefficient. One of the most curious Nobel Prize awards was that of 2013, awarded both to Fama, for his work on informationally efficient markets, and to Shiller, for showing that markets were not informationally efficient. First noted in Stiglitz (1969, 1974c, 1982d). Since then there has been an enormous literature on the efficiency wage model (see, e.g., Weiss, 1980) and on credit rationing (Stiglitz and Weiss, 1981). For an overview, see Stiglitz (1987).

Robust theory and fragile practice: part 1  41 14. For instance, many of the seeming anomalies in macroeconomics arise because market prices (say, interest rates) and shadow prices may move in different directions in the presence of credit rationing. 15. See, for instance Diamond (1971) who shows that even with small search costs, the market price will be the monopoly price. Stiglitz (1985a) shows analogous results in labour markets, and established that there may be multiple equilibria wage distributions. Stiglitz (2013a) shows, moreover, that in the Diamond model, in the absence of heterogeneity, there is in general no equilibrium. 16. Walras’ Law held that if there are N markets, and N-1 clear, the Nth must clear. 17. While Rothschild and Stiglitz (1976) showed that in their one-period model based on self-selection and contract exclusivity, there could not be a pooling equilibrium, in other contexts, there can be a pooling equilibrium. See the discussion below. Note that in Stiglitz (1975a) the market equilibrium may be inefficient even with a single commodity; in the RS (1976) analysis, the market equilibrium, when it exists, will be Pareto efficient, but only with a single commodity. With multiple goods, self-selection equilibria are generally inefficient (see Greenwald and Stiglitz, 1986, and Arnott et al., 1994). 18. In the simple early models, such as Spence (1973), Stiglitz (1975a), and Mirrlees (1971), more able individuals were proportionately better at every task; there was no comparative advantage. But when there exists comparative advantage, with one type of individual’s having better relative performance at some task than other types, but poorer relative performance in others, knowing individuals’ abilities allows a better allocation of individuals to different tasks. There is a social return to having information about individuals’ relative abilities. Of course, there will always be some comparative advantage, except in the trivial case – which is what the literature focused upon. The results of that literature appear robust, so long as the magnitudes of differences in comparative advantages are not large. Surprisingly, little of the literature has explored such situations. 19. Assuming there is only one good. As we have noted, if there is more than one commodity, the Greenwald-Stiglitz theorem applies, so that the economy is not Pareto efficient (see Greenwald and Stiglitz, 1986). 20. Throughout the chapter, the exposition focuses on identifying differences in individual abilities; the discussion could have been exposited as well in terms of differences in project returns, risks of individuals seeking insurance, or qualities of products. 21. This might not be a problem if one could buy insurance before the information were available to any party, e.g., before the individual is born. But that is not possible – partly because genetic information about parents (or asymmetries about such information) may lead to the unravelling/non-existence of that market. 22. Similarly, Stiglitz (1985a) and Salop and Stiglitz (1977, 1982) show that the market may create price dispersions, even when there is no intrinsic difference among firms. See also Stiglitz (2013a). When it does this, it imposes unnecessary search costs on consumers. See the discussion below in sections 3.11 and 3.12. 23. See, inter alia, Thaler and Benartzi (2004), Benartzi and Thaler (2007), and DellaVigna and Malmendier (2006). 24. There is an analogy here to the costs imposed on consumers in markets with endogenous price dispersion noted in note 22 and discussed further below. The high price firm exploits the high search cost individuals, but simultaneously imposes a search cost externality on the low search cost individuals. There is a further rationale for compulsory disclosure when verification is costly. We showed that the pooling equilibrium is Pareto superior to the screening equilibrium if the supply of the two types is fixed. But the pooling equilibrium will lead to an oversupply of “bad” products and an undersupply of good products, since the price they receive is the same, i.e., incentives are distorted. Taking these into account, the pooling equilibrium may well be welfare-inferior; and disclosure mandates may be necessary and desirable. 25. Obviously, there are some circumstances in which the seller may not be fully informed about the characteristics of his or her product; but typically, it is more efficient/less costly for the seller to gather relevant information than for a multitude of buyers. On the other hand, the information may be less trustworthy, especially in the presence of imperfect and costly verifiability (so mild deviations from the truth are hard to prosecute). In such cases, third-party provision of information (like Consumers’ Reports) may be desirable, either as a substitute or complement. Thus, the testing of

42  The Elgar companion to information economics

26. 27.

28. 29.

30.

31. 32.

33. 34.

35.

36.

drugs for safety and efficacy by the pharmaceutical companies selling those drugs has been questioned, with some advocating that third-party testing should play a more important role (Jayadev and Stiglitz, 2008, 2010). Thus, securities laws requiring disclosure of risks have proven to be less effective than hoped, because firms bury the real and important risks to which attention should be drawn within a long list of ordinary risks to which investors are exposed. On the role of guarantees, see Grossman (1981), and Heal (1977). A money back guarantee by itself does not suffice in the presence of costly enforcement. Note that increasing the costs of enforcement (e.g., by not allowing class action suits) increases opportunities for low-information (“bad”) equilibrium to occur (e.g., where firms exploit poorly informed consumers) – and increases the likelihood of a no-trade equilibrium (with critical markets being absent). Or in the case of confinement, where the length of confinement is sufficient to deter fraudulent behaviour. Stiglitz and Weiss (1994) further clarify the distinction between signalling and screening models. In Stiglitz (1977), a single firm (a monopolist) constructs the choice set to enable the differentiation among individuals of different types. RS considers the more complicated situation where the choice set emerges out of a competitive equilibrium. Similarly, earlier we noted that a guarantee is a costly signal. The cost to providing a guarantee for the good product, one where the probability of a critical defect is lower, for example, is less than for the bad; but, as we saw in the previous section, the guarantee can be thought of as part of the “verification of truth”. To put it rather inelegantly, the party providing the guarantee is putting his money where his mouth is. As we noted earlier, a key insight in these screening and signalling models is that the social costs and benefits of signalling and screening may differ markedly from the private costs and benefits. In the Spence model, equilibrium is defined simply as a set of self-confirming beliefs: given wage differences between the educated and the uneducated, those choosing to get educated (not educated) had productivities precisely corresponding to those beliefs. There might, of course be multiple such sets of equilibrium beliefs. These refinements eliminate the multiplicity of equilibrium, doing so typically by asking what inferences the uninformed would make were they to see an action that was deviant. See also Stiglitz (2002), putting forward the almost obvious point: in any putative signalling equilibrium, the least able individual has nothing to lose by simply investing nothing in education. The worst that could happen is that he would be treated as if he were the least able – and then he would still be better off than he is in the signalling equilibrium where he invests in education. So too, the more able could invest far more in education than needed to “signal” that he is of better quality. That, in Spence’s sense, would be an equilibrium. But it is obvious that any high ability individual who invested at least enough such that no low individual would invest that much (given their cost differences and putative returns) would do so, would still convey (signal) that he is high ability. The Riley “equilibrium” is a reactive equilibrium, inconsistent with the spirit of competitive analysis that motivated, for instance, Rothschild and Stiglitz. Acevedo and Gottlieb (2019) explore the non-existence of the Riley equilibrium. But this does not address the issue of whether full revelation is efficient. In general it is not. See Stiglitz (2009). This helps explain why there does not exist a screening equilibrium with a continuum of individuals. Recall the discussion above where, with two types, an equilibrium only exists if the types are different enough. Obviously, with a continuum, individuals are arbitrarily similar to others. It can be shown that it does not pay those near the least able individuals to distinguish themselves from that type. For an overview of the repeated games literature, see Mailath and Samuelson (2006). Casella (2005) and Jackson and Sonnenschein (2007) discuss other mechanisms. The titles of the latter two works are rather evocative: “Storable Votes” and “Overcoming Incentive Constraints by Linking Decisions”. As in Shapiro and Stiglitz (1984) or Stiglitz and Weiss (1983). These results are related to the earlier discussion of decentralization: here, intertemporal linkages are critical. See also Shapiro (1982) and Klein and Leffler (1981). Under certain conditions in labour markets, firms can punish workers for

Robust theory and fragile practice: part 1  43

37. 38. 39. 40. 41.

42.

43. 44.

45. 46.

47.

48.

49.

50. 51.

whom there has been a (noisy) signal of shirking by lowering wages (rather than terminating the contract). See Rey and Stiglitz (2023) and Arnott et al. (1988). In Chapter 3, we consider models with direct communication, giving rise to the problem of “information design”. An information design approach to the sequential screening problem is studied by Krämer and Strausz (2015) and Heumann (2020). This is analogous to “robust” mechanism design discussed below. In Akerlof’s model, where individuals just purchase or sell one car, quantities are not relevant. In the insurance market (and many other markets), there may be other aspects of the economic transaction that are observable and convey information. The literature on equilibrium with non-exclusivity is large and complex. See Arnott and Stiglitz (2013a, 2013b) and the discussion in KSY (2023). In both cases, for instance, the maximization problem entails self-selection constraints. There are many institutional details that need to be incorporated, and when this is done, there are salient differences, accounting for the huge literatures which have developed in each of these separate fields. Early contributions noting the formal similarity include Sappington and Stiglitz (1987a) and Stiglitz (1982c). Rothschild and Stiglitz (1976). In fact, when the types are not too different – as when there is a continuum of types – there may never exist an equilibrium. See Riley (1975, 1979, 2001) and Stiglitz (2009). Dasgupta and Maskin (1986a, 1986b) provide a game theoretic formulation in which there is a mixed strategy equilibrium. For instance, it matters whether the informed party “moves” before the uninformed (as in the typical signalling game) or vice versa (Stiglitz and Weiss, 1994). As we noted earlier, those do not hold in the absence of perfect risk markets. Indeed, trade liberalization can be Pareto inferior (Newbery and Stiglitz, 1984). The discussion here goes beyond Newbery and Stiglitz in noting that trade integration leads to (greater) price variability, and therefore the value of information about what prices will be increases. Indeed, derivatives may greatly increase the complexity of markets and give rise to a fundamental indeterminacy of equilibria. See Roukny et al. (2018) and Battiston et al. (2016). These are not the only examples of misguided market “reforms”. In some quarters, there is support for moving from bank lending to capital markets, in the belief that the latter do a better job in spreading risks. But as Grossman and Stiglitz (1980) and Stiglitz and Greenwald (2003) emphasize, because of informational spillovers, capital markets won’t have sufficient incentives for gathering information. See also Stiglitz (1992). There are various strands within this literature where the mechanism designer may not be able to completely control the behaviour of agents (the allocations) (as in our earlier discussion of the insurance market where customers may purchase secret insurance); and where the mechanism designer may not be able to commit to a particular rule before the messages are sent. See, inter alia, Rahman (2012), whose work is evocatively titled “But Who Will Monitor the Monitor?” Mertens and Zamir (1985) and Brandenburger and Dekel (1993) provided a critical mathematical foundation for these ideas, by constructing a universal type space, such that any “reasonably consistent” infinite hierarchy of beliefs can be represented by a type in this universal type space, and along the way explicitly constructing a mathematical mapping that maps types to infinite hierarchies of beliefs, and vice versa. This mapping is one-to-one (injective), onto (surjective), continuous, and has a continuous inverse (a “homeomorphism”); by virtue of satisfying all of these conditions, this map “preserves” all the features of one space when mapping it to another, and vice versa. Of course, without common priors, even sharing information will not lead to common posteriors. One can then agree to disagree, as was recognized and discussed by Aumann (1976). Chapter 3 discusses some of the problems of societal polarization which arise in the presence of different priors. The assumption of common knowledge is standard in the rational expectations macroeconomic literature and has provided a central point of critique of dynamic stochastic general equilibrium (DSGE) models (see, e.g., Stiglitz, 2018 and Guzman and Stiglitz, 2021a, 2021b). These ideas are related to, but different from “information theory” where the focus is on the information transmitted through a “channel” – an idealized communication device. In more mathematical terms, there exists a “garbling” matrix (in the finite case), or a Markov kernel (in the infinite case) that when applied to A, yields B.

44  The Elgar companion to information economics 52. The link between Rothschild and Stiglitz’s work (1970, 1971) ranking probability distributions for any risk averse individual and that of Blackwell should be clear; they provide additional equivalent characterizations, as does Atkinson (1970), in the context of rankings of distributions of income. For extensions in that context, beyond separable utility functions, see Rothschild and Stiglitz (1973). For analyses of behavioural implications, see Rothschild and Stiglitz (1971) and Diamond and Stiglitz (1974). We make one clarifying remark – while Rothschild and Stiglitz (1970, 1971) characterize mean-preserving spreads of distributions of monetary lotteries, Blackwell informativeness is equivalent to mean-preserving spreads of distributions of beliefs; in the context considered by Rothschild and Stiglitz, the two are equivalent. 53. Similar results were shown to hold in models with moral hazard (Arnott and Stiglitz, 1988, 2013a, 2013b). 54. Even before that, the Nobel Memorial Prize was awarded to William Vickrey (along with James Mirrlees) in 1996, partly for his work on second price auctions (Vickrey, 1961). 55. See Roth (2003) and Roth et al. (2004, 2005a, 2005b, 2006, 2007, 2008). 56. See Pathak et al. (2020) for an example motivated by the coronavirus pandemic; different US states had different priority rules for assigning scarce medical resources, which aimed to balance various ethical and practical considerations. Pathak and co-authors proposed a better mechanism, which was subsequently adopted by the University of Pittsburgh Medical Center. 57. Roth and Sotomayor (1990) summarize much of the theoretical work in this area; Teytelboym et al. (2021) provide a more recent summary. 58. See also the references to work by Atila Abdulkadiroglu, Parag A. Pathak, Alvin E. Roth, Tayfun Sonmez and M. Utku Ünver. 59. The “robustness” of mechanism design has a similar flavour to the belief-free equilibria of the context of repeated games, where the term “robust” was also used, but the precise meaning differs because the contexts are different. Belief-free equilibria are robust to individuals’ beliefs about others in repeated games; robust implementation occurs when a mechanism designer implements a social choice function as an equilibrium where individuals tell the truth, provided they expect others do as well, for any information they may have. 60. See also Diamond (1998), who also finds linear contracts to be optimal in a different, simpler, though more restrictive setting. See also Allen (1985), who focuses on the implications of non-observable trades, and which may be a better explanation of linear contracts than that provide by Carroll’s theory, since the critical unknown, the weather, has (particularly before the onset of climate change) a relatively well-defined probability distribution.

REFERENCES Acevedo, E. & Gottlieb, D. (2019). An Example of Non-existence of the Riley Equilibrium in Markets with Adverse Selection. Games and Economic Behavior, 116(C), 152–157. Akerlof, G. A. (1970). The Market for “Lemons”: Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84(3), 488–500. Akerlof, G. A. (1976). The Economics of Caste and of the Rat Race and Other Woeful Tales. The Quarterly Journal of Economics, 90(4), 599–617. Allen, F. (1985). On the Fixed Nature of Sharecropping Contracts. Economic Journal, 95(377), 30–48. Arnott, R., Greenwald, B., & Stiglitz, J. E. (1994). Information and Economic Efficiency. Information Economics and Policy, 6(1), 77–82. Arnott, R., Hosios, A. J., & Stiglitz, J. E. (1988). Implicit Contracts, Labor Mobility and Unemployment. American Economic Review, 78(5), 1046–1066. Arnott, R. & Stiglitz, J. E. (1988). The Basic Analytics of Moral Hazard. Scandinavian Journal of Economics, 90(3), 383–413. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 417–443. Arnott, R. & Stiglitz, J. E. (1991). Moral Hazard and Non-Market Institutions: Dysfunctional Crowding Out or Peer Monitoring. American Economic Review, 81(1), 179–190. Reprinted in Selected Works of

Robust theory and fragile practice: part 1  45 Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 589–603. Arnott, R. & Stiglitz, J. E. (2013a). Equilibrium in Competitive Insurance Markets with Moral Hazard. In Selected Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis: Applications to Capital, Labor, and Product Markets. Oxford: Oxford University Press, 2013, pp.  660–689. Originally Princeton University Discussion Paper 4, 1987. Also NBER Working Paper 3588, 1991. Arnott, R. & Stiglitz, J. E. (2013b). Price Equilibrium, Efficiency, and Decentralizability in Insurance Markets. In Selected Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis: Applications to Capital, Labor, and Product Markets. Oxford: Oxford University Press, 2013, pp. 632–659. Originally NBER Working Paper 3642, 1991. Arrow, K. J. (1964). The Role of Securities in the Optimal Allocation of Risk-Bearing. The Review of Economic Studies, 31(2), 91–96. Arrow, K. J. (1973). The Theory of Discrimination. In O. Ashenfelter and A. Rees (Eds.), Discrimination in Labor Markets. Princeton, NJ: Princeton University Press, pp. 3–33. Ashenfelter, O. & Card, D. (Eds.) (2011a). Handbook of Labor Economics Volume 4 Part A. North Holland: Elsevier. Ashenfelter, O. & Card, D. (Eds.) (2011b). Handbook of Labor Economics Volume 4 Part B. North Holland: Elsevier. Atkinson, A. B. (1970). On the Measurement of Inequality. Journal of Economic Theory, 2(3), 244–263. Aumann, R. (1976). Agreeing to Disagree. The Annals of Statistics, 4(6), 1236–1239. Banks, J. S. & Sobel, J. (1987). Equilibrium Selection in Signalling Games. Econometrica, 55(3), 647–661. Baron, D. P. & Myerson, R. B. (1982). Regulating a Monopolist with Unknown Costs. Econometrica, 50(4), 911–930. Battiston, S., Caldarelli, G., May, R. M., Roukny, T., & Stiglitz, J. E. (2016). The Price of Complexity in Financial Networks. Proceedings of the National Academy of Sciences of the United States of America, 113(36), 10031–10036. Benartzi, S. & Thaler, R. (2007). Heuristics and Biases in Retirement Savings Behavior. Journal of Economic Perspectives, 21(3), 81–104. Bergemann, D. & Morris, S. (2005). Robust Mechanism Design. Econometrica, 73, 1771–1813. Bernanke, B. & Gertler, M. (1989). Agency Costs, Net Worth, and Business Fluctuations. American Economic Review, 79(1), 14–31. Blackwell, D. (1951). Comparison of Experiments. Proceedings of the Second Berkeley Symposium on Mathematical Statistics and Probability, 2, 93–103. Blackwell, D. (1953). Equivalent Comparisons of Experiments. The Annals of Mathematical Statistics, 24(2), 265–272. Brandenburger, A. & Dekel, E. (1993). Hierarchies of Beliefs and Common Knowledge. Journal of Economic Theory, 59(1), 189–198. Braverman, A. & Stiglitz, J. E. (1982). Sharecropping and the Interlinking of Agrarian Markets. American Economic Review, 72(4), 695–715. Cabrales, A., Gossner, O., & Serrano, R. (2013). Entropy and the Value of Information for Investors. American Economic Review, 103(1), 360–377. Cabrales, A., Gossner, O., & Serrano, R. (2017). A Normalized Value for Information Purchases. Journal of Economic Theory, 170, 266–288. Calabresi, G. (1970). The Cost of Accidents: A Legal and Economic Analysis. New Haven: Yale University Press. Caplin, A. & Dean, M. (2015). Revealed Preference, Rational Inattention, and Costly Information Acquisition. American Economic Review, 105(7), 2183–2203. Caplin, A., Dean, M., & Leahy, J. (2022). Rationally Inattentive Behavior: Characterizing and Generalizing Shannon Entropy. Journal of Political Economy, 130(6), 1676–1715. Carroll, G. (2015). Robustness and Linear Contracts. American Economic Review, 105(2), 536–563. Casella, A. (2005). Storable Votes. Games and Economic Behavior, 51(2), 391–419. Cho, I.-K. & Kreps, D. M. (1987). Signalling Games and Stable Equilibria. The Quarterly Journal of Economics, 102(2), 179–221. Courty, P. & Hao, L. (2000). Sequential Screening. Review of Economic Studies, 67(4), 697–717.

46  The Elgar companion to information economics Dasgupta, P. & Maskin, E. (1986a). The Existence of Equilibrium in Discontinuous Economic Games, Part I (Theory). Review of Economic Studies, 53(1), 1–26. Dasgupta, P. & Maskin, E. (1986b). The Existence of Equilibrium in Discontinuous Economic Games, Part II (Applications). Review of Economic Studies, 53(1), 27–44. Dasgupta, P. & Maskin, E. (2000). Efficient Auctions. The Quarterly Journal of Economics, 115(2), 341–388. Davila, E. & Korinek, A. (2017). Pecuniary Externalities in Economies with Financial Frictions. The Review of Economic Studies, 85(1), 1–44. Debreu, G. (1954). Valuation Equilibrium and Pareto Optimum. Proceedings of the National Academy of Sciences of the United States of America, 40(7), 588–592. Debreu, G. (1959). The Theory of Value: An Axiomatic Analysis of Economic Equilibrium. New Haven: Yale University Press. DellaVigna, S. & Malmendier, U. (2006). Paying Not to Go to the Gym. American Economic Review, 96(3), 694–719. Delli Gatti, D., Gallegati, M., Greenwald, B., Russo, A., & Stiglitz, J. E. (2012). Mobility Constraints, Productivity Trends, and Extended Crises. Journal of Economic Behavior & Organization, 83(3), 375–393. Diamond, P. (1967). Cardinal Welfare, Individualistic Ethics, and Interpersonal Comparison of Utility: Comment. Journal of Political Economy, 75(5), 765–766. Diamond, P. (1971). A Model of Price Adjustment. Journal of Economic Theory, 3(2), 156–168. Diamond, P. (1998). Managerial Incentives: On the Near Linearity of Optimal Compensation. Journal of Political Economy, 106(5), 931–957. Diamond, P. & Stiglitz, J. E. (1974). Increases in Risk and in Risk Aversion. Journal of Economic Theory, 8(3), 337–360. Dionne, G. (Ed.) (2013). Handbook of Insurance. Cham: Springer. Edlin, A. S. & Stiglitz, J. E. (1995). Discouraging Rivals: Managerial Rent-Seeking and Economic Inefficiencies. American Economic Review, 85(5), 1301–1312. Ely, J. C., Hörner, J. & Olszewski, W. (2005). Belief-Free Equilibria in Repeated Games. Econometrica, 73(2), 377–415. Emran, M. S. & Stiglitz, J. E. (2009). Financial Liberalization, Financial Restraint, and Entrepreneurial Development (SSRN Scholarly Paper No. 1332399). Social Science Research Network. https://​doi​ .org/​10​.2139/​ssrn​.1332399. Fama, E. F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. The Journal of Finance, 25(2), 383–417. Fama, E. F. (1991), Efficient Capital Markets: II. The Journal of Finance, 46, 1575–1617. Frankel, A. & Kamenica, E. (2019). Quantifying Information and Uncertainty. American Economic Review, 109(10), 3650–3680. Gale, I. & Stiglitz, J. E. (1989). The Informational Content of Initial Public Offerings. The Journal of Finance, 44(2), 469–478. Gale, I. & Stiglitz, J. E. (2013). A Simple Proof that Futures Markets are Almost Always Informationally Imperfect. In Selected Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis: Applications to Capital, Labor, and Product Markets. Oxford: Oxford University Press, pp. 122–132. (Revised version of “Futures Markets Are Almost Always Informationally Inefficient”. Princeton University Financial Research Center Memorandum No. 57, February 1985.) Geanakoplos, J. D. & Polemarchakis, H. M. (1986). Existence, Regularity, and Constrained Suboptimality of Competitive Allocations When the Asset Market Is Incomplete. In W. P. Heller, R. M. Starr, & D. A. Starrett (Eds.), Essays in Honor of Kenneth J. Arrow, Volume 3: Uncertainty, Information, and Communication. New York: Cambridge University Press, pp. 65–96. Greenwald, B. (1979). The Labor Market as a Market for Lemons. New York: Garland Publishing. Greenwald, B. (1986). Adverse Selection in the Labour Market. Review of Economic Studies, 53(3), 325–347. Greenwald, B. & Stiglitz, J. E. (1986). Externalities in Economies with Imperfect Information and Incomplete Markets. The Quarterly Journal of Economics, 101(2), 229–264. Greenwald, B. & Stiglitz, J. E. (1988). Pareto Inefficiency of Market Economies: Search and Efficiency Wage Models. American Economic Review, 78(2), 351–355.

Robust theory and fragile practice: part 1  47 Greenwald, B. & Stiglitz, J. E. (1992). Information, Finance and Markets: The Architecture of Allocative Mechanisms. Industrial and Corporate Change, 1(1), 37–63. Greenwald, B. & Stiglitz, J. E. (1993). Financial Market Imperfections and Business Cycles. The Quarterly Journal of Economics, 108(1), 77–114. Greenwald, B., Stiglitz, J. E., & Weiss, A (1984). Informational Imperfections in the Capital Market and Macroeconomic Fluctuations. American Economic Review, 74(2), 194–199. Grossman, S. (1976). On the Efficiency of Competitive Stock Markets Where Traders Have Diverse Information. Journal of Finance, 31(2), 573–585. Grossman, S. (1977). The Existence of Future Markets, Noisy Rational Expectations and Informational Externalities. Review of Economic Studies, 44(3), 431–449. Grossman, S. (1981). The Informational Role of Warranties and Private Disclosure about Product Quality. Journal of Law and Economics, 24(3), 461–483. Grossman, S. & Stiglitz, J. E. (1980). On the Impossibility of Informationally Efficient Markets. American Economic Review, 70(3), 393–408. Guo, Y. & Shmaya, E. (2019). Robust Monopoly Regulation. Working Paper. https://​arxiv​.org/​pdf/​1910​ .04260​.pdf. Guzman, M. & Stiglitz, J. E. (2020). Towards a Dynamic Disequilibrium Theory with Randomness. Oxford Review of Economic Policy, 36(3), 621–674. Guzman, M. & Stiglitz, J. E. (2021a). Pseudo-Wealth and Consumption Fluctuations. The Economic Journal, 131(633), 372–391. Guzman, M. & Stiglitz, J. E. (2021b). Economic Fluctuations and Pseudo-Wealth. Industrial and Corporate Change, 30(2), 297–315. Guzman, M. & Stiglitz, J. E. (2021c). The Pandemic Economic Crisis, Precautionary Behavior, and Mobility Constraints: An Application of the Dynamic Disequilibrium Model with Randomness. Industrial and Corporate Change, 30(2), 467–497. Harsanyi, J. C. (1967). Games with Incomplete Information Played by “Bayesian” Players, I–III Part I. The Basic Model. Management Science, 14(3), 159–182. Harsanyi, J. C. (1968a). Games with Incomplete Information Played by “Bayesian” Players, I–III. Part II. Bayesian Equilibrium Points. Management Science, 14(5), 320–334. Harsanyi, J. C. (1968b). Games with Incomplete Information Played by “Bayesian” Players, I–III. Part III. The Basic Probability Distribution of the Game. Management Science, 14(7), 486–502. Hart, S. (2005). An Interview with Robert Aumann. Macroeconomic Dynamics, 9(5), 683–740. Heal, G. (1977). Guarantees and Risk-Sharing. The Review of Economic Studies, 44(3), 549–560. Heumann, T. (2020). Information Design and Sequential Screening with Ex Post Participation Constraint. Theoretical Economics, 15, 319–359. Hirshleifer, J. (1971). The Private and Social Value of Information and the Reward to Inventive Activity. American Economic Review, 61(4), 561–574. Hörner, J. & Lovo, S. (2009). Belief-Free Equilibria in Games with Incomplete Information. Econometrica, 77(2), 453–487. Hörner, J., Lovo, S., & Tomala, T. (2018). Belief-Free Price Formation. Journal of Financial Economics, 127(2), 342–365. Hurwicz, L. (1972). On Informationally Decentralized Systems. In C. McGuire & R. Radner (Eds.), Decision and Organization: A Volume in Honor of Jacob Marschak. Amsterdam: North-Holland, pp. 297–336. Jackson, M. O. & Sonnenschein, H. F. (2007). Overcoming Incentive Constraints by Linking Decisions. Econometrica, 75(1), 241–257. Jayadev, A. & Stiglitz, J. E. (2008). Two Ideas to Increase Innovation and Reduce Pharmaceutical Costs and Prices. Health Affairs, 27(Supplement 1), w165–w168. Jayadev, A. & Stiglitz, J. E. (2010). Medicine for Tomorrow: Some Alternative Proposals to Promote Socially Beneficial Research and Development in Pharmaceuticals. Journal of Generic Medicines, 7(3), 217–226. Jeanne, O. & Korinek, A. (2010). Excessive Volatility in Capital Flows: A Pigouvian Taxation Approach. American Economic Review, 100(2), 403–407. Kamenica, E. (2017). Information Economics. Journal of Political Economy, 125, 1885–1890.

48  The Elgar companion to information economics Kandori, M. & Obara, I. (2006). Efficiency in Repeated Games Revisited: The Role of Private Strategies. Econometrica, 74(2), 499–519. Klein, B. & Leffler, K. B. (1981). The Role of Market Forces in Assuring Contractual Performance. Journal of Political Economy, 89(4), 615–641. Kohlberg, E. & Mertens, J.-F. (1986). On the Strategic Stability of Equilibria. Econometrica, 54(5), 1003–1037. Korinek, A. (2012). Capital Flows, Crises, and Externalities. In F. Allen et al. (Eds.), The Global Macro Economy and Finance: Proceedings of the 16th IEA World Congress, vol. 3. Basingstoke: Palgrave Macmillan, Chapter 5. Kosenko, A. (2022). Algebraic Properties of Blackwell’s Order and a Cardinal Measure of Informativeness. Working Paper. Kosenko, A., Stiglitz, J. E., & Yun, J. (2023). Bilateral Information Disclosure in Adverse Selection Markets with Nonexclusive Competition. Journal of Economic Behavior & Organization, 205, 144–168. Krähmer, D. & Strausz, R. (2015). Optimal Sales Contracts with Withdrawal Rights. Review of Economic Studies, 82, 762–790. Laffont, J.-J. & Tirole, J. (1986). Using Cost Observation to Regulate Firms. Journal of Political Economy, 94(3), 614–641. Laffont, J.-J. & Tirole, J. (1988). The Dynamics of Incentive Contracts. Econometrica, 56(5), 1153–1175. Laffont, J.-J. & Tirole, J. (1990). Adverse Selection and Renegotiation in Procurement. The Review of Economic Studies, 57(4), 597–625. Leitzinger, J. & Stiglitz, J. E. (1984). Information Externalities in Oil and Gas Leasing. Contemporary Economic Policy Issues, 5, 44–57. Maćkowiak, B., Matejka, F., & Wiederholt, M. (2019). Rational Inattention: A Review. Working Paper. Mailath, G. & Samuelson, L. (2006). Repeated Games and Reputations: Long-Run Relationships. New York: Oxford University Press. Matějka, F. & McKay, A. (2015). Rational Inattention to Discrete Choices: A New Foundation for the Multinomial Logit Model. American Economic Review, 105(1), 272–298. Mertens, J.-F. & Zamir, S. (1985). Formulation of Bayesian Analysis for Games with Incomplete Information. International Journal of Game Theory, 14(1), 1–29. Milgrom, P. R. (1981). Good News and Bad News: Representation Theorems and Applications. The Bell Journal of Economics, 12(2), 380–391. Milgrom, P. R. & Stokey, N. (1982). Information, Trade, and Common Knowledge. Journal of Economic Theory, 26, 17–27. Mirrlees, J. A. (1971). An Exploration in the Theory of Optimum Income Taxation. Review of Economic Studies, 38(2), 175–208. Mu, X., Pomatto, L., Strack, P., & Tamuz, O. (2021). From Blackwell Dominance in Large Samples to Rényi Divergences and Back Again. Econometrica, 89(1), 475–506. Newbery, D. M. G. & Stiglitz, J. E. (1982). The Choice of Techniques and the Optimality of Market Equilibrium with Rational Expectations. Journal of Political Economy, 90(2), 223–246. Newbery, D. M. G. & Stiglitz, J. E. (1984). Pareto Inferior Trade. Review of Economic Studies, 51(1), 1–12. Pathak, P., Sönmez, T., Ünver, M., & Bumin, Y. M. (2020). Leaving No Ethical Value Behind: Triage Protocol Design for Pandemic Rationing. Working Paper No. 2695. National Bureau of Economic Research. https://​www​.nber​.org/​papers/​w26951. Pęski, M. (2008). Comparison of Information Structures in Zero-Sum Games. Games and Economic Behavior, 62(2), 732–735. Phelps, E. S. (1972). The Statistical Theory of Racism and Sexism. American Economic Review, 62(4), 659–661. Radner, R. & Stiglitz, J. E. (1984). A Nonconcavity in the Value of Information. In M. Boyer & R. Khilstrom (Eds.), Bayesian Models in Economic Theory. Amsterdam: Elsevier Science Publications, pp. 33–52. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 537–554. Rahman, D. (2012). But Who Will Monitor the Monitor? American Economic Review, 102(6), 2767–2797.

Robust theory and fragile practice: part 1  49 Rey, P. & Stiglitz, J. E. (2023). Moral Hazard and Unemployment in Competitive Equilibrium. In Selected Works of Joseph E. Stiglitz, Volume IV: Rethinking Macroeconomics. Oxford: Oxford University Press, Chapter 12. Rey, P. & Tirole, J. (1986). The Logic of Vertical Restraints. American Economic Review, 76(5), 921–939. Riley, J. G. (1975). Competitive Signalling. Journal of Economic Theory, 10(2), 174–186. Riley, J. G. (1979). Informational Equilibrium. Econometrica, 47(2), 331–359. Riley, J. G. (2001). Silver Signals: Twenty-Five Years of Screening and Signalling. Journal of Economic Literature, 39(2), 432–478. Roth, A. E. (2003). The Origins, History, and Design of the Resident Match. JAMA, 289(7), 909–912. Roth, A. E., Sönmez, T., & Ünver, M. U. (2004). Kidney Exchange. The Quarterly Journal of Economics, 119(2), 457–488. Roth, A. E., Sönmez, T., & Ünver, M. U. (2005a). A Kidney Exchange Clearinghouse in New England. American Economic Review, 95(2), 376–380. Roth, A. E., Sönmez, T., & Ünver, M. U. (2005b). Pairwise Kidney Exchange. Journal of Economic Theory, 125(2), 151–188. Roth, A. E., Sönmez, T., & Ünver, M. U. (2007). Efficient Kidney Exchange: Coincidence of Wants in Markets with Compatibility-Based Preferences. American Economic Review, 97(3), 828–851. Roth, A. E., Sönmez, T., & Ünver, M. U. (2008). Kidney Paired Donation with Compatible Pairs. American Journal of Transplantation, 8(2), 463–463. Roth, A. E., Sönmez, T., Ünver, M. U., Delmonico, F. L., & Saidman, S. L. (2006). Utilizing List Exchange and Nondirected Donation through ‘Chain’ Paired Kidney Donations. American Journal of Transplantation, 6(11), 2694–2705. Roth, A. E. & Sotomayor, M. (1990). Two-Sided Matching: A Study in Game-Theoretic Modeling and Analysis. New York: Cambridge University Press. Rothschild, M. & Stiglitz, J. E. (1970). Increasing Risk: I. A Definition. Journal of Economic Theory, 2(3), 225–243. Rothschild, M. & Stiglitz, J. E. (1971). Increasing Risk: II. Its Economic Consequences. Journal of Economic Theory, 5(1), 66–84. Rothschild, M. & Stiglitz, J. E. (1973). Some Further Results on the Measurement of Inequality. Journal of Economic Theory, 6(2), 188–204. Rothschild, M. & Stiglitz, J. E. (1976). Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information. The Quarterly Journal of Economics, 90(4), 629–649. Rothschild, M. & Stiglitz, J. E. (1982). A Model of Employment Outcomes Illustrating the Effect of the Structure of Information on the Level and Distribution of Income. Economic Letters, 10, 231–236. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 122–126. Rothschild, M. & Stiglitz, J. E. (1997). Competition and Insurance Twenty Years Later. Geneva Papers on Risk and Insurance Theory, 22(2), 73–79. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 160–167. Roukny, T., Battiston, S., & Stiglitz, J. E. (2018). Interconnectedness as a Source of Uncertainty in Systemic Risk. Journal of Financial Stability, 35, 93–106. Salop, S. (1977). The Noisy Monopolist: Imperfect Information, Price Dispersion and Price Discrimination. The Review of Economic Studies, 44(3), 393–406. Salop, S. & Stiglitz, J. E. (1977). Bargains and Ripoffs: A Model of Monopolistically Competitive Price Dispersion. The Review of Economic Studies, 44(3), 493–510. Salop, S. & Stiglitz, J. E. (1982). The Theory of Sales: A Simple Model of Equilibrium Price Dispersion with Identical Agents. American Economic Review, 72(5), 1121–1130. Sappington, D. & Stiglitz, J. E. (1987a). Information and Regulation. In E. Bailey (Ed.), Public Regulation. London: MIT Press, pp. 3–43. Sappington, D. & Stiglitz, J. E. (1987b). Privatization, Information and Incentives. Journal of Policy Analysis and Management, 6(4), 567–582. Reprinted in E. Baily & J. Hower (Eds.), The Political Economy of Privatization and Deregulation. Aldershot, UK and Brookfield, VT, USA: Edward Elgar Publishing, 1993.

50  The Elgar companion to information economics Shapiro, C. (1982). Consumer Information, Product Quality, and Seller Reputation. Bell Journal of Economics, 13, 20–35. Shapiro, C. & Stiglitz, J. E. (1984). Equilibrium Unemployment as a Worker Discipline Device. American Economic Review, 74(3), 433–444. Shiller, R. J. (1981). Do Stock Prices Move Too Much to Be Justified by Subsequent Changes in Dividends? American Economic Review, 71, 421–436. Shiller, R. J. (2000). Irrational Exuberance. Princeton, NJ: Princeton University Press. Sims, C. A. (2003). Implications of Rational Inattention. Journal of Monetary Economics, 50(3), 665–690. Spence, M. (1973). Job Market Signalling. The Quarterly Journal of Economics, 87(3), 355–374. Stigler, G. (1961). The Economics of Information. Journal of Political Economy, 69(3), 213–225. Stiglitz, J. E. (1969). Alternative Theories of Wage Determination and Unemployment in L.D.C.’s. Nairobi: Institute for Development Studies, mimeo. (Parts of which subsequently published as Stiglitz (1974c) and (1982d).) Stiglitz, J. E. (1972). On the Optimality of the Stock Market Allocation of Investment. The Quarterly Journal of Economics, 86(1), 25–60. Stiglitz, J. E. (1973). Approaches to the Economics of Discrimination. American Economic Review, 62(2), 287–295. Stiglitz, J. E. (1974a). Incentives and Risk Sharing in Sharecropping. Review of Economic Studies, 41(2), 219–255. Stiglitz, J. E. (1974b). Theories of Discrimination and Economic Policy. In G. von Furstenberg et al. (Eds.), Patterns of Racial Discrimination. Lexington, MA: D. C. Heath and Company (Lexington Books), pp. 5–26. Stiglitz, J. E. (1974c). Alternative Theories of Wage Determination and Unemployment in L.D.C.’s: The Labor Turnover Model. The Quarterly Journal of Economics, 88(2), 194–227. Stiglitz, J. E. (1975a). The Theory of “Screening”, Education, and the Distribution of Income. American Economic Review, 65(3), 283–300. Stiglitz, J. E. (1975b). Information and Economic Analysis. In J. M. Parkin & A. R. Nobay (Eds.), Current Economic Problems. Cambridge: Cambridge University Press, pp.  27–52. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 29–52. Stiglitz, J. E. (1975c). The Efficiency of Market Prices in Long Run Allocations in the Oil Industry. In G. Brannon (Ed.), Studies in Energy Tax Policy. Cambridge: Ballinger Publishing, pp. 55–99. (Report written for the Ford Foundation Energy Policy Project, August 1973.) Stiglitz, J. E. (1977). Monopoly, Non-Linear Pricing and Imperfect Information: The Insurance Market. Review of Economic Studies, 44(3), 407–430. Stiglitz, J. E. (1982a). The Inefficiency of the Stock Market Equilibrium. Review of Economic Studies, 49(2), 241–261. Stiglitz, J. E. (1982b). Information and Capital Markets. In W. F. Sharpe & C. M. Cootner (Eds.), Financial Markets: Essays in Honor of Paul Cootner. Englewood Cliffs, NJ: Prentice-Hall. Stiglitz, J. E. (1982c). Self-Selection and Pareto Efficient Taxation. Journal of Public Economics, 17, 213–240. Stiglitz, J. E. (1982d). Alternative Theories of Wage Determination and Unemployment: The Efficiency Wage Model. In M Gersovitz et al. (Eds.), The Theory and Experience of Economic Development: Essays in Honor of Sir W Arthur Lewis. London: Allen & Unwin, pp. 78–106. Stiglitz, J. E. (1984). Information, Screening and Welfare. In M. Boyer & R. Khilstrom (Eds.), Bayesian Models in Economic Theory. Amsterdam: Elsevier Science Publications, pp. 209–239. Stiglitz, J. E. (1985a). Equilibrium Wage Distributions. Economic Journal, 95, 595–618. Stiglitz, J. E. (1985b). Credit Markets and the Control of Capital. Journal of Money, Banking, and Credit, 17(2), 133–152. Stiglitz, J. E. (1987). The Causes and Consequences of the Dependence of Quality on Price. Journal of Economic Literature, 25(1), 1–48. Stiglitz, J. E. (1989). Imperfect Information in the Product Market. In R. Schmalensee & R. Willing (Eds.), Handbook of Industrial Organization, Vol. 1. Amsterdam: Elsevier, pp. 769–847.

Robust theory and fragile practice: part 1  51 Stiglitz, J. E. (1992). Banks versus Markets as Mechanisms for Allocating and Coordinating Investment. In J. A. Roumasset & S. Barr (Eds.), The Economics of Cooperation: East Asian Development and the Case for Pro-Market Intervention. Boulder, CO: Westview Press, pp. 15–38. Reprinted in Selected Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis: Applications to Capital, Labor, and Product Markets, Oxford: Oxford University Press, 2013. Stiglitz, J. E. (2000). The Contributions of the Economics of Information to Twentieth Century Economics. The Quarterly Journal of Economics, 115(4), 1441–1478. Stiglitz, J. E. (2002). Information and the Change in the Paradigm in Economics (Abbreviated version of Nobel lecture). American Economic Review, 92(3), 460–501. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 53–98. Stiglitz, J. E. (2009). Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press. Stiglitz, J. E. (2010). Freefall: America, Free Markets, and the Sinking of the World Economy. New York: W. W. Norton. Stiglitz, J. E. (2013a). Introduction to Part V. In The Selected Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis: Applications to Capital, Labor, and Product Markets. Oxford: Oxford University Press, pp. 695–704. Stiglitz, J. E. (2013b). Information and Competition. In Selected Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis: Applications to Capital, Labor, and Product Markets. Oxford: Oxford University Press, pp. 20–35. (Originally the Inaugural Lecture presented at All Souls College, Oxford, June 1978.) Stiglitz, J. E. (2018). Where Modern Macroeconomics Went Wrong. Oxford Review of Economic Policy, 34(1–2), 70–106. Stiglitz, J. E. (2020). The Revolution in Information Economics: The Past and the Future. In K. Basu, D. Rosenblatt & C. Sepulveda (Eds.), The State of Economics, the State of the World. Cambridge, MA: MIT Press, pp. 101–150. Stiglitz, J. E. (2023). The Selected Works of Joseph E. Stiglitz, Volume IV: Rethinking Macroeconomics. Oxford: Oxford University Press. Stiglitz, J. E. & Greenwald, B. (2003). Towards a New Paradigm in Monetary Economics. Cambridge: Cambridge University Press. Stiglitz, J. E. & Greenwald, B. (2015). Creating a Learning Society: A New Approach to Growth, Development, and Social Progress. New York: Columbia University Press. Stiglitz, J. E. & Guzman, M. (2021a). Where Modern Macroeconomics Went Wrong. Oxford Review of Economic Policy, 34(1–2), 70–106. Stiglitz, J. E. & Guzman, M. (2021b). Economic Fluctuations and Pseudo-Wealth. Industrial and Corporate Change, 30(2), 297–315. Stiglitz, J. E. & Weiss, A. (1981). Credit Rationing in Markets with Imperfect Information. American Economic Review, 71(3), 393–410. Stiglitz, J. E. & Weiss, A. (1983). Incentive Effects of Terminations: Applications to the Credit and Labor Markets. American Economic Review, 73(5), 912–927. Stiglitz, J. E. & Weiss, A. (1992). Asymmetric Information in Credit Markets and Its Implications for Macroeconomics. Oxford Economic Papers, 44(4), 694–724. Stiglitz, J. E. & Weiss, A. (1994). Sorting Out the Differences Between Screening and Signalling Models. In M. O. L. Bacharach, M. A. H. Dempster, & J. L. Enos (Eds.), Mathematical Models in Economics. Oxford: Oxford University Press. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic Analysis. Oxford: Oxford University Press, 2009, pp. 223–231. Stiglitz, J. E. & Wolfson, M. (1988). Taxation, Information, and Economic Organization. Journal of the American Taxation Association, 9(2), 7–18. Teytelboym, A., Li, S., Kominers, S. D., Akbarpour, M., & Dworczak, P. (2021). Discovering Auctions: Contributions of Paul Milgrom and Robert Wilson. Scandinavian Journal of Economics, 123, 709–750. Thaler, R. H. & Benartzi, S. (2004). Save More TomorrowTM: Using Behavioral Economics to Increase Employee Saving. Journal of Political Economy, 112(S1), S164–S187.

52  The Elgar companion to information economics Veldkamp, L. L. (2012). Information Choice in Macroeconomics and Finance. Princeton, NJ: Princeton University Press. Vickrey, W. (1961). Counterspeculation, Auctions, and Competitive Sealed Tenders. The Journal of Finance, 16(1), 8–37. Weiss, A. (1980). Job Queues and Layoffs in Labor Markets with Flexible Wages. Journal of Political Economy, 88(3), 526–538. Wilson, R. (1985). Incentive Efficiency of Double Auctions. Econometrica, 53(5), 1101–1115. Wilson, R. (1987). Game-Theoretic Analyses of Trading Processes. In T. F. Bewley (Ed.), Advances in Economic Theory: Fifth World Congress. Cambridge: Cambridge University Press, pp. 33–70.

3. Robust theory and fragile practice: Information in a world of disinformation Part 2: Direct communication Joseph E. Stiglitz and Andrew Kosenko

1. INTRODUCTION In this chapter we provide an interpretive survey of recent work on endogenous information structures, where the information that is obtained by agents – who knows what – is determined at least in part by the agents communicating among themselves. This chapter follows our earlier companion chapter discussing the economics of information where direct communication is limited, and where much of the relevant information is gleaned from making inferences based on observable actions, but is self-contained. The problem of information control where a party (perhaps more than one) can determine what information others receive is a natural one to study when information is either abundant or can be made so (through investigations or experiments). Indeed, in a world where there is more information than can be reasonably processed, the “information problem” is no longer about lack of information, but about deciding what information to gather and to attend to. Thus, in this chapter we focus mostly on the perils of abundance of information, especially if this information is produced strategically or is subject to manipulation. As we discuss, such perilous abundance, often marked by strategic mis- and disinformation, can undermine both markets and democratic mechanisms. The chapter is divided into six sections beyond this introduction and brief concluding remarks. The first three address communication in a world of rationality; at the centre are rational individuals updating priors according to Bayes’ rule. We begin in section 2 by explaining the constraints imposed by the rationality hypothesis. Section 3 discusses the burgeoning literatures on communication and information design, where an agent designs what information to communicate to others, in contexts where such information is not verifiable (“cheap talk”) and there is no punishment for lying – other than the loss of credibility; or where communication is verifiable, but probabilistic, possibly selective and incomplete). Section 4 shows that combining direct communication with the indirect mechanisms that were at the centre of discussion in the previous chapter generates markedly different outcomes – most strikingly, in the standard insurance model, an equilibrium always exists and entails a pooling policy bought by both high-risk and low-risk individuals. The last two sections focus on the more sinister side of the superabundance of information, including how mis- and disinformation is leading to the polarization of society, a polarization which is best understood by going beyond the standard model of rationality and how social media and virality are worsening polarization and giving rise to a variety of social harms (section 5); and what can be done about these harms and the increase in market power that is

53

54  The Elgar companion to information economics associated with social media – with their power, based on the abuse of information, undermining the foundations of competition through the economy (section 6).

2.

CONSTRAINTS ON CREDIBILITY IMPOSED BY RATIONALITY

It may seem that if the information designer can control what information others obtain, she has in effect complete freedom to choose the actions of others. She simply conveys the information that induces those she wants to manipulate to act in the way she wished. But if those she is attempting to manipulate are rational, this is, in fact, not so – it’s not necessarily true that “anything goes”. If the agents interpret the information they receive in a Bayesian fashion, there exist significant constraints on what the information designer can “persuade” her audience of. Consider a simple example – suppose our information designer creates an information structure that always provides the same information – regardless of the truth. For instance, an investment adviser who always recommends purchase, or an attorney who always recommends conviction. If the agent on the receiving end knows this, she will realize that this signal is effectively uninformative and disregard it. Therefore, the information designer would not provide such a signal in the first place; in fact, the constraints (known as Bayesian plausibility,1 Kamenica and Gentzkow, 2011) placed on the designer by a rational, Bayesian receiver are very significant, and the work in this area nearly always assumes that these constraints are respected. In settings where instead of providing information and then relying on agents to take the optimal action given that information, the information designer directly recommends an action, a similar mechanism is in play; the recommendation cannot be arbitrary because it has to be incentive compatible for a rational agent (who knows that the information designer is in charge of the information flow, and may have her own, ulterior, motives) to obey the recommendation; Bergemann and Morris (2016) discuss the obedience constraint. In much of the theoretical work on endogenous information structures, the agents know the distributions from which signals are drawn. Yet, the Wilson critique2 – that knowledge of the distributions is too strong of an assumption – applies here as well; robust information design, which we discuss at the end of the chapter, has explored the possibilities of what is and is not achievable when assumptions of common knowledge of signal distributions (and related assumptions) are relaxed. Importantly, and much more relevantly, when decision makers are not Bayesian (a topic to which we turn in the second half of this chapter), these constraints are not present; the information designer then has much more freedom to affect action via information control. 2.1

Cheap Talk

Implicit in much of the earlier literature on asymmetric information is the assumption that simply sending unverifiable messages couldn’t convey meaningful information, and in the context in which that literature developed, that was true: less able individuals would simply say they are more able. Actions speak louder than words: because some actions are more costly (or some decisions are preferable) for low ability individuals than high ability, one can make

Robust theory and fragile practice: part 2  55 inferences about ability by observing actions or decisions. But in other contexts, messages can convey meaningful information. In a seminal model of costless (and therefore known as “cheap talk”) communication, Crawford and Sobel (1982) establish conditions under which there is some information revelation, even if the communication is unverifiable – and therefore “fraud” is not punishable (as in the analysis of the previous chapter). The intuition is that, provided there is not “too much” misalignment in preferences, it is in the “sender’s” interest to reveal some information correctly, and in fact, in equilibrium, this information will be believed by the “receiver”. In other words – and this is the striking interpretation of the mathematical result – there can be some degree of credible information transmission even if there is no way to verify that information. Crucially, the sender is unable to commit to an information disclosure strategy;3 this is one way in which cheap talk differs from information design, where the sender can commit to a disclosure strategy. Subsequently, this model has been used to incorporate many other elements: costs of misrepresentation (“lying”, Kartik, 2009),4 multiple dimensions of information (such as would arise if there are for instance, incompatible considerations, say, climate damage and private costs (Battaglini, 2002), multiple senders (Ambrus and Takahashi, 2008), partial rankings or comparative (as opposed to absolute) statements (Chakraborty and Harbaugh, 2007), and repeated interactions (Ambrus et al., 2013). Sobel (2013) provides an insightful and exhaustive overview of the literature up to that point. Typical results state conditions under which some information revelation is an equilibrium outcome, and study the implications of these conditions (such as what will be communicated, what kinds of assumptions and reasoning are necessary and sufficient for communication, and what the preferences of the agents must be). The results are sometimes of the flavour “one needs at least this much alignment in preferences” for information revelation. Crawford and Sobel (1982) also provide a simple example, and one that is widely used in applications (in particular, in experimental economics); the sender is privately informed about the unknown payoff-relevant state – simply a real number between 0 and 1, while the receiver knows only the distribution of the number, which is uniform. The sender sends a costless, unverifiable message (that is intended to convey some information about the state, but may be used to deceive the receiver). The receiver wishes to take an action (choose a number) that is as close to the true number as possible, while the sender has an upward bias – she wants the receiver to “inflate” the number the receiver chooses. The bias is capturing the misalignment in preferences (such as would arise say, in financial advice or sales applications). There is always an uninformative (“babbling”) equilibrium, where the receiver disregards any message. An imperfectly informative equilibrium of this model (what Crawford and Sobel refer to as a “partition” equilibrium) takes the following form: all types of senders below a certain cut-off choose a low message, while higher types choose a higher message. Thus, there is some information conveyed: the senders inform the receiver that the true state is within some interval (the element of the partition), but do not state the exact number. The receiver is thus better informed than in the babbling equilibrium, but is not perfectly informed. Of course, the receiver is also never systematically (meaning, in expectation, ex ante) misled; as we have already noted, systematic deception is impossible to obtain in standard models. In an important example of cheap talk with multiple senders, as opposed to one sender, as in Crawford and Sobel (1982), Battaglini (2002) leverages conflict of interest across dimensions of information to – strikingly – obtain full revelation (although Ambrus and Takahashi,

56  The Elgar companion to information economics 2008, qualify this result by pointing out the necessity of the richness of the space across which dimensions of agreement can be exploited to obtain full revelation). In Battaglini’s (2002) leading example, there are two senders (“experts”) – one with expertise in carbon emissions and another with expertise in economic policy. The receiver has to take an action that is two-dimensional: a recommended level of carbon emissions, and a tax/ subsidy policy. Both senders have preferences over both dimensions; their ideal points may be arbitrarily far apart (and thus the conflict of interest with the receiver may be arbitrarily large). Battaglini (2002) constructs a fully revealing equilibrium in the following fashion: Both senders send two-dimensional signals. Suppose that they both tell the truth, and the receiver implements the first dimension of the first sender’s (two-dimensional) message, and the second dimension of the second expert’s message. It turns out that (unless the experts’ ideal points lie on a line in two-dimensional space – a very restrictive condition that is unlikely to be satisfied in applications) neither expert has an incentive to lie, and the receiver obtains full revelation, even in this setting of unverifiable information and preference conflict. 2.2

Endogenizing Information and Conflict of Interest in Cheap Talk Models

It is intuitive that if the sender doesn’t use all the information he has and if it is costly to obtain information, he would not obtain information he did not use, in which case, the Crawford-Sobel results would have to be modified. Argenziano et al. (2016) and Pei (2015) study cheap talk with endogenous information, where information can be obtained at a cost. The difference between their settings lies in the modelling of the sender’s information: Argenziano and co-authors use repeated Bernoulli experiments, while Pei allows the sender to choose partitions of the state space. Argenziano and co-authors study two variations: a game that proceeds à la Crawford and Sobel where the receiver observes the fact that the sender has become informed, and a game where the receiver does not know this. They find inefficient overinvestment in information – the sender’s precision is too high in both variations (provided the expert’s bias is not too large); i.e., the sender acquires more costly information than the receiver (who is also the decision maker) would, if she could do so, in all Pareto-efficient equilibria. Overinvestment occurs even in a setting where the sender is unbiased – for any number of Bernoulli experiments he chooses to acquire, the optimal action of the sender and receiver coincide. If the receiver observes that the sender is informed, the result is driven by “equilibrium pessimism” of the receiver: she adopts a sceptical posture (see also Milgrom and Roberts, 1986), completely ignoring the transmitted information (which is bad for the sender), unless the sender acquires the equilibrium amount of information (which is too high). If the receiver does not observe whether the sender is informed, in equilibrium, the receiver believes that the sender has acquired the equilibrium amount of information even if she has not; this belief turns out to greatly constrain the kinds of deviations (to acquiring less costly information) that the sender prefers, and she has no choice but to acquire too much information. In closely related work, Pei studies a cheap talk setting where the sender may become more informed (in the Blackwell sense)5 at a cost. Consistent with the intuition above, Pei finds that the sender always communicates all of the obtained information. However, equilibrium may be more or less informative than the Crawford-Sobel one. Thus communicating “everything she knows” may entail communicating more or less information than what is communicated

Robust theory and fragile practice: part 2  57 in Crawford-Sobel’s model with an exogenously informed receiver in the most informative equilibrium of their game.6

3.

INFORMATION DESIGN

When one party has more information about a variable of general relevance to others (as in the analysis of cheap talk of the previous section), it can choose what information to release. The informed party wants to do so to induce others to act in ways that maximize its utility. This general problem is referred to as the information design problem, defined by Taneva (2019, p. 151) as follows (italics in original):7 Mechanism design takes the informational environment as given and focuses on providing incentives for desired equilibrium behavior by committing to an extensive form of the strategic interaction, i.e., a mechanism. In contrast to this, information design studies the way a designer can manipulate the equilibrium behavior of agents by selecting the informational environment under which they operate while holding the mechanism fixed. Information design thus applies to situations where a designer is able to influence the optimal behavior of agents only through the information she provides about the state, without being able to change any aspects of the mechanism.

A typical question of mechanism design is: For some specification of payoffs, what are the outcomes as a function of information structure (a set of signals for agents that they use to update their beliefs and play the game)? What payoffs (and beliefs) are feasible? And among these, which mechanism optimizes the well-being of the mechanism designer? By contrast, information design focuses on the selection of information structures: Of all the possible information structures, which is optimal, i.e., maximizes a given objective function, given that agents will obtain information from this information structure and then act “selfishly”? Generally, the information designer cannot provide information that always incentivizes the agent to act in a particular way; however, it is possible in many cases to significantly influence behaviour. In the typical view, the information designer commits to an information disclosure rule that is state dependent (before the designer herself observes the state), and possibly probabilistic, the state is realized according to a prespecified commonly known distribution, the signals are realized according to the distribution chosen by the information designer (this is also known by the agents), and agents proceed to play a game. Because the information designer committed to the information structure before she observed the state, the signal realizations are now credible (in the sense of being drawn from a distribution, and not subject to manipulation by the designer). But of course, if the signal distributions are non-degenerate (as is typically the case), these signals are also possibly wrong (because of the stochasticity); the agents take this into account, as does the designer when choosing the distributions. This literature elucidates the limits of what exactly is possible, and how this depends on the preferences of those that the information designer wishes to influence and on the environment. 3.1

Information Design, Commitment, and Persuasion

Kamenica and Gentzkow (2011) illustrate the possibilities and limits of information design, taking into account the fact that if agents are rational (in particular, use Bayes’ rule), one cannot mislead them systematically. They observe that one can provide information so that

58  The Elgar companion to information economics they act in a way that is at least probabilistically much more favourable to the information designer. As noted in the earlier discussion of cheap talk (see also Bergemann and Morris, 2019), it may be optimal to release some but not all the information, i.e., to partially obfuscate. By doing so, one can “persuade” agents to take actions that are more favourable to the informed party – actions that she might not otherwise have undertaken. “Persuasion” within the information design framework is understood to mean provision of strategically designed information to influence action. As discussed earlier, this literature typically assumes commitment power on the part of the party controlling the information (“information control” typically refers to a setting where the sender can provide more information than she has initially, by designing appropriate experiments, and “commitment power” refers to the fact that the sender cannot hide the signal realizations once they are realized – but the probability with which they are realized in each state is under the control of the sender. In other words, the sender commits to running an experiment the results of which will be observed whether they are favourable to the sender or not.), and in this way differs from the cheap talk literature; commitment refers not just to the provision of evidence but also, for instance, the design of an investigation or a clinical trial. Typically this literature assumes common prior beliefs about the underlying payoff relevant state. Hedlund (2017) and Kosenko (2022) extend the persuasion problem to include private information on the part of the sender: the sender may have more precise (but perhaps still imperfect) information than the receiver when designing the experiments (and, crucially, the sender is the only party able to do experiments to obtain more information). In a word, the sender gets an additional informative signal about the state, that the receiver is not privy to. Hedlund (2017) shows that with such private information, if the sender can choose among all possible information structures, and they all have the same cost – zero, strikingly, private information in a model of persuasion is irrelevant – in all equilibria, either the private (imperfect) information is revealed, or even more strikingly, the true state of the world is revealed. Kosenko (2022) qualifies this by pointing out that this argument relies on the availability (even if it is not used in equilibrium) of a very special information structure – one that reveals the true state of the world in every state. If this structure is unavailable (or, to put it differently, there is an arbitrarily small amount of noise, as would be the case in most applications), Kosenko (2022) shows that there are many equilibria, many of which are uninformative, and the Hedlund (2017) result does not apply. Thus, in any realistic application, private information in a model of persuasion matters a great deal. Faced with this multiplicity of equilibria, Kosenko (2022) then provides an equilibrium selection procedure (a “refinement”) that selects the most informative equilibria in this setting.8 Kolotilin et al. (2017) study persuasion with private information on the part of the receiver (as opposed to a privately informed sender, as in Hedlund, 2017, and Kosenko, 2022). In general, without full information, the sender can’t be sure about how the different types of receivers will interpret different signals, and respond to them. Secondly, we wish to understand how the sender should design different communication strategies for different types. Thus, in principle, there can be many ways of communication in this setting, in particular, the sender may ask the receiver about her type, or just choose an experiment for all types of receivers. Within their framework (linear utilities, and binary actions for the receiver), they show that these two channels – private communication (where the sender asks the receiver for her type, and then a type-dependent experiment produces a signal), and public communication (where the sender chooses an experiment, aware of the possible private information of the

Robust theory and fragile practice: part 2  59 receivers, and how it might affect interpretation of the signal realizations) – are equivalent. In a word, under the stated restrictions, public communication is equivalent to private communication, although the strong restrictions are necessary for this result (in general the equivalence fails). Moreover, Kolotilin et al. (2017) show that private information on the part of the receivers makes a great deal of difference; in particular, they establish an “anything goes” result – any interim (i.e., after observing her own type) utility for the receiver that is achieved (between complete information about the state, and no information about the state), can be achieved by some experiment (“public” communication) or a persuasion mechanism (“private” communication). In other words, there are many equilibria, which differ in the amount of information transmitted. Of course, which receiver utility will be implemented depends on the preferences of the sender. 3.2

Robust Information Design

Most of the literature in this area relies on very strong, and implausible, assumptions concerning what information the parties have (namely, the distribution of the underlying state, the preferences of the agents, and how they form and update their beliefs upon observing signal realizations). There is an extensive literature exploring ways of weakening the informational assumptions that go into information design, just as we noted in the previous chapter for analogous work on robust mechanism design.9 For instance, Bergemann and Morris (2016) define “Bayes-correlated equilibria”. Outcomes of Bayes-correlated equilibria (BCE) are Bayes-Nash equilibrium outcomes that could arise across all information structures such that there is a strict lower bound on the information that each agent has. In other words, the mechanism designer can be mistaken about some aspect of the environment, or the agents may even be behaving in an adversarial fashion10 (as they are in Mathevet et al., 2020, where the agents are coordinating on an outcome that is worst for the mechanism designer). Yet Bergemann and Morris show that the predictions of the solution concept are reliable, in the sense that they are invariant to the information. However, as Bergemann and Morris (2016) illustrate, this power comes at a cost – BCE typically make weak predictions. Many allocations (in particular, many more than under Bayes-Nash equilibrium) can be supported as a BCE.11 Nonetheless, Bergemann and Morris (2016) show that the set of BCE shrinks if and only if the informativeness of the information structure increases. An increase in the abundance of information restricts the set of equilibria. Many of the results in this literature are technical, and, like those in mechanism design (apart from the work on auctions and matching), so far have found limited application. Sometimes, the central result is only to provide bounds on what can be achieved.12 Mathevet et al. (2020), as well as Carroll (2015) and Guo and Shmaya (2019) analysing robust-mechanism design, use a worst-case scenario approach to evaluate solutions. In the latter two papers, common prior beliefs and Bayesian updating are assumed to play no role at all. Perhaps the best way to see this new strand of literature is that it explores the opposite polar case to that which previously dominated the literature, where there is always perfect Bayesian updating. In terms of applications, Guo and Shmaya (2019) illustrate robust information design by showing when a regulator should use a price cap versus using a subsidy, so that the policy works well (in a particularly defined way) in all circumstances.

60  The Elgar companion to information economics While the applications of this literature have, so far, been relatively limited, one potential area where information design may be relevant in the future relates to the platforms over which so many transactions occur and which can observe so much behaviour. These platforms clearly have much more information than others – an abundance of information. Because what information they choose to disseminate can have significant consequences, there is considerable value in understanding better what information they might choose to disseminate under various conditions. For instance, Kanoria and Saban (2017) study an example where platforms (such as dating or ride-sharing apps) can improve welfare by restricting what information agents have access to. But as we note below, the objectives of the platforms and those of society may differ markedly, giving rise to the necessity of regulating platforms, a subject which we discuss further below. 3.3

Is Information Revelation Necessarily Welfare Enhancing?

Another arena in which information design is relevant concerns the disclosure of information by central banks, who typically have more information than other agents in the economy, both about what they may be thinking of doing and about the state of the economy. The policy discourse has centred around the question: How much information should central banks disclose? How transparent should they be? Traditionally, they were very non-transparent, but there have been marked moves to increase transparency. This has given rise to some controversy. Morris and Shin provide a model in which public information may have a detrimental effect on welfare (Morris and Shin, 2002), in a setting reminiscent of Keynesian beauty contests (Keynes, 1936). In this context, there may be (private) benefits to coordination. Agents have private information to which they pay insufficient attention as they overreact to public information which can serve to coordinate, so that public disclosures may lower welfare. (Their result parallels some noted in the previous chapter, where more information may be welfare decreasing.) However, Svensson (2006) shows that even in their restrictive model, that is not likely to be the case: under plausible assumptions about the precision of the information available to private actors versus that of the monetary authorities and the ambient noise in the economy, greater transparency is welfare improving.13 More generally, whether the conclusion of Morris and Shin’s (2002) work supports or opposes public release of information (i.e., transparency) depends on the model specification. Angeletos and Pavan (2004) find that public information may decrease or increase welfare, depending on the strength of complementarities, the link between individual returns on investment and aggregate return; if the complementarities are weak, more transparency in public information increases welfare. If, on the other hand, the complementarities are strong, there may be multiple equilibria, in some of which greater transparency is welfare-decreasing. Angeletos and Pavan (2007, p. 1104) conclude: Because we allow for various strategic and external effects, there is no simple answer [to the question of whether more transparency is welfare increasing]. For example, there are economies where welfare would be higher if agents were to raise their reliance on public information and economies where the converse is true. Similarly, there are economies where any information is socially valuable and economies where welfare decreases with both private and public information. This is consistent with the folk theorem that “anything goes” in a second-best world.

Robust theory and fragile practice: part 2  61 This is a large literature. In macroeconomics beauty contests appear in the literature on monetary policy (Woodford, 2002), and business cycles (Angeletos and La’O, 2013; Benhabib et al., 2015); Huo and Pedroni (2020) provide a unifying framework for this work. Jackson and Pernoud (2021) survey some of the related literature on financial fragility, while Blanchard (2009), Angeletos et al. (2021), and Goldstein and Yang (2017) offer additional reviews on the different facets of the question of the value of information disclosure in macroeconomics and finance. 3.4

Beyond Standard Information Design

In practice, persuasion and selective information disclosure may be greatly affected by (rational) understandings of individual irrationalities (in the words of Daniel Ariely, their predictable irrationalities (Arieli, 2008)).14 There is by now a large literature in behavioural economics showing that at least in a wide variety of circumstances, individuals do not behave in the way assumed by the literature on information design (in particular, in forming expectations using Bayes’ theorem). It is this reality that has given rise to the huge advertising industry (with much, if not most, of the “information” provided being not informative at all) and to the problems of mis- and disinformation to which we turn below. The information design literature has focused on one aspect of communication: what information is at the disposal of an informed party to disclose to others. There are several related issues, some of which have been touched on in the earlier literature already referred to in this and the previous chapter: (a) What information should the parties themselves gather (as in Stiglitz, 1984), where an initially uninformed individual in equilibrium acquires information to make himself more informed than his trading partners), and of the information they have, what should and can they credibly convey, and how (the issue of information design associated with direct communication upon which this section has focused)?; (b) What actions should the uninformed take to extract information from the informed (as in the screening literature)?; (c) What actions should the informed take to convey whatever information they decide to convey (i.e., signalling through actions – going beyond direct communication); (d) What actions should the informed take to make it more difficult to extract such information (as in the Edlin and Stiglitz, 1995, analysis where managers make portfolio decisions that make it more difficult for outsiders to assess the net worth of the firm)?; (e) What actions should the government undertake that affect communication – both explicit disclosure rules (with penalties for incomplete disclosures) and policies that affect payoffs (e.g. “mechanism design”) and therefore incentives for disclosure?; and (f) How transparent should the government itself be, i.e., how much of the information at its disposal should it disclose? In short, asymmetries of information (both ex ante, before communication, and ex interim, after communication, but before all uncertainty has been resolved) are endogenous and need to be modelled. Moreover, of increasing importance are the consequences (including potential penalties) of providing mis- and disinformation going beyond partial information disclosure. And the behavioural responses have to be assessed in models which do not assume full rationality. We turn to this subject shortly.

62  The Elgar companion to information economics

4.

COMBINING COMMUNICATION AND INDIRECT INFERENCE

In the previous chapter, we noted a conundrum: if there exist secret contracts, there appeared to be no competitive equilibrium in the standard model with asymmetric information. The standard adverse selection price equilibrium could be broken by a firm offering a large quantity contract (knowing the individual who is purchasing the contract) at a price slightly below the pooling price and make a profit. Moreover, the standard Rothschild-Stiglitz (1976) exclusive contract quantity equilibrium can’t be enforced, because of the secret contract, and it turns out that the separating equilibrium that they identify doesn’t work: high risk individuals buy the contract intended for the low risk individual and supplement it with secret insurance. Kosenko, Stiglitz, and Yun (2023, KSY for brevity) show more generally that there never exists an equilibrium: every proposed set of separating contracts can be broken, every proposed pooling equilibrium can be broken, and there is no hybrid equilibrium. KSY go on to investigate an equilibrium in which individuals and insurance firms can communicate directly anything (that they know) with each other. This stands in marked contrast to the earlier work of both Akerlof (1970) and Rothschild-Stiglitz. There, individuals’ only information derives from what they themselves observe. In the case of the adverse selection equilibrium, they only observe whether the individual has bought a policy at a particular price – this conveys a limited amount of information. In Rothschild-Stiglitz, they observe that they chose a particular policy from a known set of policies; this can convey more information, though that information too is limited. In KSY, individuals can’t lie, but they don’t have to be fully forthcoming. The remarkable result is that with the seemingly small change in assumptions, allowing secret contracts and direct communication between consumers and firms, all the standard results are reversed: under very weak conditions,15 there always exists an equilibrium and it entails partial pooling, with the low- and high-risk individuals purchasing a common policy (the pooling policy that maximizes the utility of the low-risk individual) and the high-risk individual purchasing supplemental (undisclosed) insurance to become fully insured.16 More generally, communication, learning and inference are all important. Crawford et al. (2013) clarify this by distinguishing between learning and thinking. They argue that learning is an appropriate prism when a situation is repeated, and agents can adjust behaviour over time, which they do, depending on their payoffs; often behaviour with learning in this sense converges to game-theoretic equilibrium predictions, where players make fully rational inferences, including taking into account fully the inferences that will be made by others. Learning players are strategic (they reason in terms of best-response and iterated best-response – best-response by one player to a best-response by another), based on their beliefs, and are probabilistically sophisticated (they compute conditional probabilities according to Bayes’ rule). If on the other hand, a situation is truly novel, and players have no prior experience, models of thinking (or inference) are necessary to understand players’ initial responses. “Thinking” in this instance is an explicit process for how players decide which actions to take, as they think about how others will respond to their action.17 As they think about how others will respond, they have to ask how sophisticated other players are. If they are very sophisticated, the other players will be thinking about how others will interpret their actions, who will be thinking about how others will be interpreting their actions, ad infinitum. This is, in most situations,

Robust theory and fragile practice: part 2  63 an extraordinarily difficult task; major strands of analysis attempt to analyse more “bounded” rationality. One prominent model of thinking is the level-k model and the closely related cognitive hierarchies model. In level-k models (Stahl and Wilson, 1994; Nagel, 1995; Costa-Gomes et al., 2001) all agents in a game have a type (level-0, often called L0; level-1, called L1; level-2, called L2, and so on), each type occurring with a pre-specified probability. Higher types are more sophisticated in their reasoning. L0 is the least sophisticated type; they are non-strategic, or as Crawford et al. (2013, p. 14), put it, they have “a strategically naïve initial assessment of others’ likely responses to the game”. They take an action specified by the modeller (often their choice is modelled by assuming that they take all actions with equal probability – a uniform random choice). L1s best respond to the L0 types. L2s best respond to L1s, L3s best respond to L2s, and so on. Thus, in level-k thinking models agents best respond to the level of reasoning sophistication below them.18 (Note that this is an explicit process for how players reason when deciding – a model of thinking. This model has the potential to out-predict what is called equilibrium behaviour (i.e., behaviour in which there is full rationality, taking in the infinite set of inferences described above) in novel situations. Empirically, most players’ behaviour is well-described by a relatively low level of k – from one to four (Crawford and Iriberri, 2007; Arad and Rubinstein, 2012).) Level-k models have some attractive features – they respect restrictions placed by rationality and common knowledge of rationality (more precisely, has k-level-rationalizability, for finite k. A strategy is 1-rationalizable if it is a best response to some strategy profile of the opponents, and a k-rationalizable strategy is a best response to a profile that is (k-1)-rationalizable). Plainly, this means that players will never choose actions they know are bad for them, and can infer that others will not do so either, so their choices are best responses to some plausible conjecture about how other players are playing (namely, that opponents will never play strategies that are never a best response) (Bernheim, 1984; Pearce, 1984). These restrictions are less restrictive than those imposed by Nash equilibrium. Furthermore, as play continues (i.e., as the number of times the game is played grows), predictions of level-k models often converge to the Nash equilibrium behaviour (for instance, in “beauty contests”). There are also some drawbacks of level-k models – the assumption of non-strategic behaviour for the L0 types and the distribution of types give the modeller two degrees of freedom, and often, additional ad hoc assumptions are necessary. L0’s actions are sometimes (Crawford and Iriberri, 2007; Penczynski, 2013) assumed to be specific to the game, and while their empirical frequency is low (in fact, in applications their frequency is often assumed to be zero – meaning these types only exist in the minds of higher types), it does act as an “anchor” for the actions of others. By changing this anchor, it may be possible to change predictions. Furthermore, the sophistication levels may not be stable and fixed – there is some evidence that players endogenously change their level in response to incentives (Alaoui and Penta, 2016).

5.

MIS- AND DISINFORMATION, SOCIAL MEDIA, AND THE POLARIZATION OF SOCIETY

The revolution in economics brought about some fifty years ago by information economics focused on asymmetries in information – how informed parties can convey favourable infor-

64  The Elgar companion to information economics mation to the uninformed and how the uninformed parties can elicit information from the informed. In the previous chapter, we discussed the role of “statements” that might convey information, noting the critical role played by verifiability. Similarly, in this chapter, we have explored how under some circumstances, statements may convey information, even without verifiability. In the earlier literature discussed, individuals might not disclose all the relevant information; they might take actions to obfuscate information (as in Edlin and Stiglitz, 1995); they strategically decide on how much information and what information to release; but they could be prosecuted for lying, e.g., under fraud laws, “truth in advertising”, libel laws, etc. In the absence of such laws and with costly verification, economic agents might have an incentive to lie. They may be able to “get away with it”. The verisimilitude of this literature to what actually occurs in markets and the policy questions countries face may seem weak. Most advertising is not about the provision of information – it is about preying on individuals’ aspirations and vulnerabilities: the Marlboro (cigarette) man is emblematic – the hugely successful ad campaign did not provide information that smoking the cigarette would make one a rugged cowboy, information which in any case would be irrelevant for the majority of smokers living in urban areas. Philip Morris might have provided relevant information, such as that its product was deliberately designed to be addictive or that a succession of Marlboro men had died of lung cancer, and that the smoker too might die of this or a number of other health risks arising from smoking, but chose not to do so. Advertisers attempt to induce people to buy their product, but typically not by the kind of “persuasion” that has been discussed earlier in this chapter. Part of the real-world information revolution (as opposed the information revolution within the academic economic discipline) is the growth of social media, which has enhanced not just the ability to target better such advertising on those most susceptible, but also the ability to rapidly spread mis- and disinformation. 5.1

The Rationality Conundrum

But, apart from a very limited literature within economics on fraud, little attention has been paid to concerns about mis- and disinformation. This is perhaps not surprising, given economists’ predilection for rationality and rational expectations. Indeed, the success of mis- and disinformation represents a puzzle for standard economics, which begins by assuming individual rationality, including an individual’s ability to rationally evaluate the accuracy of information and update priors, using Bayes’ theorem.19 In this perspective, individuals should put little weight on unverified “information”, putting greater weight on sources of information that have established a reputation for accuracy. So too, presumably, information from a source that repeatedly provided mis- and disinformation would lose credibility and therefore would play no role in decision making – and so would not be a problem. These are issues, of course, that are at the centre of the information design problem discussed in sections 2 and 3. In particular, section 2 discussed the strong constraints imposed by the hypothesis of rationality and Bayesian reasoning. But in fact, a central problem confronting society today is the provision and spread of mis- and disinformation, and its rapid dissemination over social media. There is a small recent literature trying to come to grips with these issues, including deriving policies that might militate against the social harms to which dis- and misinformation give rise. Not surprisingly, much of this goes beyond the Bayesian framework that has been central to the analysis so far.

Robust theory and fragile practice: part 2  65 5.2

Why It Matters

This mis- and disinformation is associated with high levels of social harms, including inducing people not to get vaccinated, inciting violence, and stimulating racial bigotry. The magnitude of these problems seems hard to reconcile with any model of individual rationality. Mis- and disinformation have also contributed to the polarization of society. In sections 6.5, 6.6, and 6.7, we’ll describe some of the mechanisms by which this occurs, both with and without the assumption of rationality. This polarization is rightly viewed as one of the fundamental problems facing society today. If individuals differed only in the judgements about whether red or green lettuce were healthier, such differences would be of limited significance: those believing the red lettuce was healthier could consume more red lettuce. But there are a host of important decisions that are made collectively (including the rules that underlie any economy), and differences in worldviews are associated with major differences in views about these decisions. The pandemic brought the issues to the fore: requirements over vaccines and masks in the pandemic were framed by the Right as an infringement on individual liberty while the Left correctly emphasized the importance of public health externalities, and saw mandates such as those associated with masking as appropriate regulatory responses. In the presence of externalities, there is a need for collective action. But it is hard to take the appropriate collective actions when there is the level of polarization in worldviews that is evident in many societies. One has to ask, how can there be such differences in worldviews, when the evidence is there for all to see? 5.3

“In a Free Marketplace of Ideas, Only the Best Win Out”: A Misguided Metaphor20

There are some who say not to worry. Just as in competitive markets the best producers – the most efficient, those who produce the goods that consumers want – survive, so too, in the competitive marketplace of ideas. The Greenwald-Stiglitz Theorem (1986) provides the obvious caveat: in the presence of imperfect information, markets are not in general efficient; and of necessity, the marketplace of ideas is one in which a priori there cannot be perfect information. The discussion in the previous chapter highlighted the importance of regulations such as “truth in advertising” and fraud laws. There is a consensus that cigarette advertising, the objective of which is to induce individuals to engage in a harmful activity, should be highly regulated, with most countries requiring some disclosure of some of the harmful effects. (One reason is that there is no adequate remedy through tort law of the harm that may follow from such advertising; even if truth eventually “wins out”, i.e., even if eventually the harmful effects of cigarettes became known, those who died as a result have no adequate recourse.) The analogy to the competitive marketplace of goods is flawed in several other ways. Typically, as we have noted, competition is limited, and that is especially so in the social media platforms, marked by high levels of network externalities. Market power, the importance of which we have already noted in this context, means lack of equal access. The intermediaries control access and money matters both in the control of the intermediaries and in getting access. Those with enough money can flood the intermediaries, including by using bots. That’s not a free market.

66  The Elgar companion to information economics Moreover, the first principle of a competitive “free” market is transparency. But a market in which no one knows what messages have been sent to whom is intrinsically non-transparent. To put it another way, good information is necessary to make the marketplace for goods work. But as we explained in the previous chapter, markets simply won’t ensure this on their own. For instance, we regulate securities markets to ensure equal access to information in the form of the SEC’s fair disclosure requirement and to ensure greater access to information through a variety of disclosure requirements. There is at least one more ingredient necessary to make markets work well: the absence of the use of force and intimidation. Regrettably, unregulated trolling on social media has become a fact of life. Thus, Schiffrin and Stiglitz (2020) concluded their discussion of the idea of a free marketplace of ideas as follows: In short, without full transparency, without a mechanism for holding participants to account, without equal ability to transmit and receive information, and with unrelenting intimidation, there is no free marketplace of ideas. One of the major insights of modern economics is that private and social incentives are often not well-aligned. If those who want to spread misinformation are willing to pay more than those who want to counter it, and if lack of transparency is more profitable than transparency, then [if we simply say] “so be it” we won’t get a well-functioning marketplace of ideas.

5.4

Evolution, Selection, and Divergence

There is another strand of thought that says, not to worry, but for a different reason: those who are more rational will win out, they will prosper and dominate, so that eventually, through an evolutionary process, the economy converges to one well-described by full rationality, with decisions being made that incorporate all the relevant information. As we have already noted, however, even if that were true there is ample evidence that today we’re far from such a world. Moreover, there are no strong results suggesting that the economy will converge to such a world. Indeed, even in the simpler context of competitive models, there is no assurance of convergence (Bray, 1978, 1981). More recently, Dosi et al. (2020) have shown in the context of a simple macroeconomic model with endogenous technological change that if market participants switch to and from simple rules for expectation formation (say simple extrapolative rules) to more sophisticated rules (least square regressions based on past data) based on their relative performance, there is not convergence to the more sophisticated rules; and that overall economic performance (both growth and volatility) is poorer with more sophisticated rules. There is no natural selection towards more rationality; and it may not even be rational to be (seemingly) more rational. 5.5

Polarization and Mis- and Disinformation in a Rational Framework: Bayes’ Theorem and its Critics

In this and the next subsection, we discuss briefly research attempting to help us understand this polarization and its persistence. We do so through the lenses both of the standard model of rationality with disparate priors and through that of modern behavioural economics. In this subsection, we provide a selective discussion of the economic literature on how it is possible that rational individuals would differ so markedly in their beliefs.

Robust theory and fragile practice: part 2  67 There are many taxonomies of this literature. One useful for our purpose is to divide the work according to the kind of probabilistic reasoning it uses. Many (perhaps most) models use Bayesian updating, dating back to the work of Jerzy Neyman, L. J. Savage, and Harold Jeffreys in the 1950s.21 In the previous chapter, we discussed Aumann’s (1976) clarification that agents can only “disagree” – have different posteriors, and know each other’s posteriors – to the extent that they have different priors. A recent literature on polarization and disagreement explores the effect of assuming different priors. Sethi and Yildiz (2012) for example, assume different priors and incomplete information, and endow their agents with the ability to communicate their beliefs; they show that in this setting communication need not result in information revelation, and identify the cases in which this communication breakdown can occur. Intriguingly, they show that even if priors are heterogeneous and unobserved but correlated (as would be in a society that is in some sense relatively homogeneous), communication results in an outcome that is the same as the one where priors are observed. More generally, however, outcomes are less salutary. Kartik et al. (2021), for instance, study agents with different priors who are otherwise fully Bayesian; they show that if agents’ priors are different, observing the same event leads them both to update their belief about the other agent’s belief to be closer to their own prior – a result they dub “information validates the prior”.22 An implication of this statistical result is that that if we start polarized, we expect more information to confirm our priors, and remain more entrenched in our own worldview. There is still another reason for the perpetuation and amplification of differences in priors. If individuals differ in their priors, they will differ in their judgements of the accuracy of information provided by different suppliers of information. Given the scarcity of time, even if information were free they would turn to suppliers of information that are, from their perspective, “better”. Indeed, Sethi and Yildiz (2016) study a setting with heterogeneous priors and consider the trade-off between attending to information sources that are well understood (i.e., perhaps biased, but whose bias is known) and well informed (in the sense of precision of their information). Broadly speaking, their main result is that nearly anything can happen in this setting. Many kinds of behaviour, including opinion leadership (where weight is given to the views of particular individuals or sources) and information segregation “groups” (where different individuals live within different information bubbles: “individuals observe only those within their own subgroup”) can arise. Changes in technology and policy affect the extent of such fragmentation. In the era after the Second World War, when TV was a major media for providing new information, there were only three major national networks in the US and all aimed to provide broad and unbiased information. News programmes were treated as a public service by networks (a practice that was changed, in part, by the programme 60 Minutes on CBS, which showed that news programmes can also bring in revenue). Fairness doctrines ensured that major different views were given airtime. Those across the political spectrum were at least exposed to similar information. But the elimination of fairness obligations in 1987 by the US Federal Communications Commission, combined with proliferation of cable TV, and then the internet, meant that the information to which those of different beliefs were exposed became markedly different. The consequences of these changes have been discussed and modelled, among others, by Glaeser (2005), Pickard (2015), Guriev and Treisman (2020), Ash et al. (2021), and Szeidl and Szucs (2022).

68  The Elgar companion to information economics The Bayesian approach has many advantages: it is mathematically tractable, and is consistent with vast amounts of research in other fields. But it suffers from a critical defect. It’s long been known to be inconsistent with a wealth of evidence of behaviour in a wide variety of circumstances – humans don’t typically reason about probabilities in a way that is predicted by the Bayes rule (Keynes, 1921; Allais, 1953 (presenting the famous Allais Paradox23); Ellsberg, 1961 (presenting the famous Ellsberg paradox); Kahneman and Tversky, 1979; for an influential early review, see Machina, 1987).24 5.6

Behavioural Economics: Beyond Bayes

None of the approaches based on Bayesian rationality, as sophisticated as they may be, can really account for the observed divergences in views. In the standard economists’ model, non-scientific (for example, anti-vaccine) information would simply have no impact. The evidence is that it does, and that this is so, is consistent with a large literature in behavioural economics stressing an individual’s cognitive limitations, particularly in processing statistical information and especially when such information (data) is contrary to prior beliefs. One then has to model human probabilistic information processing in some other way; other updating procedures (for instance, probability weighting (Tversky and Kahneman, 1992; Prelec, 1998), over/underreaction to new information (Epstein et al., 2008, 2010), the peak-end rule according to which intense experiences (“peak”) and experiences which come last (“end”) are remembered (Fredrickson and Kahneman, 1993; Kahneman et al., 1993), and attaching disproportionate weight to initial observations (Rabin and Schrag, 1999)) may be perhaps more plausible in some contexts, but are typically imposed in a somewhat ad hoc fashion. Ortoleva (2012) provides a prominent example of this literature in which he posits a threshold probability above which a decision maker acts in a Bayesian fashion, and below which (i.e., for low probabilities – “unexpected news”) the decision maker acts differently.25 Twenty-first century behavioural economics has, in addition, emphasized the importance of the social formation of beliefs. Beliefs are interdependent, with information about a particular subject (such as the safety and efficacy of vaccines or the role of masks during a pandemic) interpreted through a cultural lens, which “prejudice” the assessments, and affected by the beliefs of those with whom one interacts. This is especially so if those providing the information succeed in framing the information in ways that embed it into a cultural context.26 Given the polarization of views around central themes of individual liberties and collective action and the centrality of these issues to today’s critical policy debates, it is not surprising that there is deep polarization around what would seem to be scientific issues like climate change or the efficacy and safety of vaccines. While there is, at this juncture, no consensus on precisely how individuals actually process information to form beliefs (in contrast to the consensus over how “rational” individuals should form beliefs through Bayesian statistics), there is widespread understanding of some of the constitutive elements and key properties, e.g., on the importance of framing and of confirmatory bias. As Hoff and Stiglitz (2010) have shown, confirmatory bias can easily give rise to “equilibrium fictions”, beliefs that are self-sustaining, even in the presence of evidence against them. And this may be even more so if individuals not only start with different priors, as in Kartik et al. (2021) and Sethi and Yildiz (2012), and are influenced by the beliefs of others with whom they interact, but interact only or mostly with people whose priors are close to theirs but different from others.27

Robust theory and fragile practice: part 2  69 Of course, those in marketing have long sought to understand how to influence individuals’ beliefs, even with “non-informative” advertising (the Marlboro Man being the quintessential example), with a modicum of success, enough evidently to justify the billions of dollars spent every year on such advertising. 5.7

Competition in Worldviews and Signal Jamming

Building on these insights, there are two important directions to be addressed in future research. One is “competition in worldviews”: there is a need for a theory of metanarratives, the “lens” through which we see the world, which result in competing interpretations of the same pieces of information. How do we explain the persistence of differences in priors, even when so much of the evidence on which judgements are made is widely available? As we noted before, the question of how competition in worldviews plays out is of crucial importance. The fact of the matter is that many critical events happen with such rarity that there is ample opportunity for differing interpretation of their origins and consequences (Guzman and Stiglitz, 2020, 2021). The analysis of the preceding subsection helps our understanding of how differences in worldviews could persist. One way in which deception and (incorrect, easily disconfirmed, yet persistently held) different worldviews may persist and succeed, arises from the presence (potential or actual) of boundedly rational types. A decision-maker acting according to an explicit procedure (such as satisficing – a portmanteau of “satisfy” and “suffice” (Simon, 1955), where agents sample items in order until they find one that is acceptable), based on “simple” decision rules, or in a way that it different from the full strategic and inferential paradigm discussed above, is said to be boundedly rational. Crawford (2003) discusses a compelling model that combines bounded rationality and strategic communication, in which deception can exist in equilibrium. In this game with pre-play communication senders send costless, noiseless messages and then both senders and receivers take payoff-relevant actions; players can also be rational (“sophisticated”) or boundedly rational (“mortal”). The mortal players act like level-k players, best-responding to lower levels of other “mortal” players and forming beliefs in a mechanistic way. A “lie” in this setting is sending one message, and later playing an action that corresponds to a different message. Crawford (2003) focused on the case where the presence of the boundedly rational types enables a false message – a lie – to persist. The result holds if the probabilities of the sophisticated types are relatively low. Intuitively (although the result is somewhat more subtle), if a sophisticated sender type is unlikely, a sophisticated receiver will best respond under the premise that the message is coming from a mortal type. This, in turn, creates a possibility that the receiver plays an action that is a best response to the mortal types but not to the sophisticated types (because they are unlikely), and therefore the sophisticated senders get away with lying. As Crawford (2003, p. 133) puts it, “rational players exploit boundedly rational players, but are not themselves fooled”. (As we noted above, systematic deception is impossible to obtain in standard equilibrium models. Even here a sophisticated sender cannot truly fool the sophisticated receiver in the sense of using a strategy that is unknown to the receiver. Rather, the sophisticated receiver knows the sophisticated sender’s strategy, but the payoffs and the low probability of a sophisticated sender nevertheless make it optimal for the receiver to act in a way that disregards this, knowing that some of the time that action will not be optimal, and thus, the sophisticated receiver ends up acting as if he were “fooled”.)

70  The Elgar companion to information economics In the same vein, Szeidl and Szucs (2022) present a political economy model where, although information is verifiable (there is an “objective reality” in their world), some politicians are nevertheless able to persuade some voters of something false. The key assumption of their model is that some voters may believe a verifiably false message, because they believe that with some small probability that message could be true.28 In other words, “propaganda makes the voter assign positive probability to a non-existent alternative reality” (Szeidl and Szucz, 2022, p. 13). One of the key theoretical findings of their framework is that once such an alternative reality is created, it can persist, even in presence of clear evidence of its falsity, and will cause the persuaded voter to act against her best interest. A second promising line of thought is “signal jamming” as information manipulation. A malicious actor may exploit the fact that audiences have a limited capacity for information processing. The possibility of such obfuscation has been critical in the design of disclosure requirements. (See the discussion of the companion chapter.) Burying the required disclosure (nutritional information for food, risk for investments) in a barrage of other information attenuates the value of the information being provided. The audience may then either tune out informative signals, or, once information processing capacity has been reached, tune out all signals. In other words, one may “crowd out” informative signals by providing a surfeit of uninformative ones, with the aim of limiting the ability to process the informative ones. In some cases, e.g., in the health hazards of cigarettes, governments have specified how the relevant information is to be disclosed to prevent such obfuscation.

6.

SOCIAL MEDIA, SOCIAL DIVISIONS, AND PUBLIC POLICY

Social media platforms have been able to take advantage of not only advances in artificial intelligence (AI), but of understandings of human behaviour and information processing in ways that have increased their profits while imposing large costs on society, exacerbating societal polarization. Earlier we noted a business model that profits from engagement (sometimes through enragement) and AI algorithms have targeted different individuals with different information designed to enhance engagement, fragmenting the information structure beyond anything that had previously been possible and in ways that have enhanced polarization. The ability to create separate communities, reinforcing the disparate beliefs, has made matters still worse. Virality meant information could spread quickly, more quickly than “antidotes” to the misinformation could be designed. The lack of transparency in who gets what messages has meant that the antidotes could not be effectively delivered in the relevant time span, if at all. In the previous chapter we emphasized the interaction between information and market power. Information economics helps explain the limited competition in media, including social media. Later in this section, we will explore in greater detail how this plays out with vengeance in the context of social media. Market power, in turn, enhances the power to exploit limitations in information. Another central theme of modern economics is the link between economic power and political power – a vicious circle where the concentration of economic power leads to a concentration of political power, which results in rules of the game enhancing the concentration of economic power.29

Robust theory and fragile practice: part 2  71 It not only pays the very rich to create a “narrative” that supports an economic environment that enriches them (say with low taxes and the ability to engage in exploitation of others), but many of them also come to believe this narrative. That is why market power in the media may be particularly invidious. It gives those with the wealth and the desire to control dominant media the power to shape the societal narrative (discussed in 5.7) in ways which affect the interpretation of data.30 This effect may be even stronger than the more narrowly defined distortions in the information disseminated discussed above; it is this which enables the success of mis- and disinformation campaigns. And when there is not sufficient media diversity, the ability to counter the narrative is limited. But even with some media diversity, the polarization effects described earlier mean that even if there are outlets providing counter-narratives and “true” facts, their impacts may be limited. The invidious effects of social media are even greater, not just because competition may be more limited and they have a greater ability to target information, but also because of policy. In the US, social media platforms are shielded by section 230 of the Communications Decency Act (1996) from liability for what they transmit across their platforms, in a way that standard media are not. Conventional media are subject, for instance, to being sued under libel and fraud laws; not so for the platforms. A provision originally designed to encourage a nascent industry has led to the viral dissemination of mis- and disinformation, which has resulted in enormous social harms with no accountability. 6.1

Regulating the Societal Harms of Social Media and Mis- and Disinformation in a Democratic Society

Today, many countries, recognizing the variety of societal harms arising from mis- and disinformation, especially over social media, are debating how to regulate them.31 The EU, for instance, has adopted the Digital Services Act. A central question in the design of such regulations (a full discussion of which is beyond the scope of this chapter) is how to prevent such harms within democratic frameworks that emphasize, for instance, freedom of speech (First Amendment rights). Societies including the US have not, however, taken absolutist positions: there are prohibitions against fraud (“lying” in commercial contexts where the lie results in harm), false advertising, child pornography, and crying fire in a crowded theatre. Some countries ban hate speech. Clearly, the greater harms emanating from mis- and disinformation on, say, social media change the “balancing” entailed in the design of regulations towards greater intervention, especially when such intervention is directed at the extent of virality.32 Indeed, in some way, there is already a form of regulation of virality, but it is effectively regulation delegated to the platforms themselves – virality is determined by the social media platforms in ways which are not transparent but which maximize their profits, almost regardless of the social harms generated, with a limited role given to content moderation. There is a growing consensus that such self-regulation is no more effective in this context than it was in financial markets. 6.2

The Market Power of Social Media and the Need to Control It

The enormous profits of the social media companies are a strong sign of the lack of competition. It is not that the successful companies are that much more productive or innovative than their rivals. (Indeed, in many cases, they made only small (but still important) innovations

72  The Elgar companion to information economics that gave them some advantage over rivals.) Normally, such large profits would attract entry, which in turn would lead to the dissipation of the profits. This, however, has not happened. The reason is simple: network externalities. The value of being on a platform like Facebook depends on the presence of others being on the platform. It is hard, in such a situation, to displace even a relatively inefficient incumbent, one who does not serve the interests of those on the platform as well as others might. But there is another element to the vicious circle which has given rise to their enormous market power and profits. Their business model, based on the use of information garnered from interactions that occur over their platform, has been a double-edged sword. (A key element in this process is that this data creation (data that is then monetized by the platforms) by virtue of mere interaction with the platform is essentially independent of the content of the interaction.) The more efficient use of the greater information that they have has allowed them to better target messages in ways that engender more engagement (and thus generate still more information). With attention (and time) a scarce commodity, “better” targeting could mean individuals receive messages that are more relevant, thereby leading to more efficient resource allocations (purchases that result in individuals enjoying a higher level of well-being). Unfortunately, that is not the objective of the better targeting. The objective is more profits, which are derived from advertising revenues, which in turn are derived from inducing more profitable purchases. Increased profits from sales, in turn, can result from more effective price discrimination – capturing more of the individual’s consumer surplus. They can also arise from increased sales, including to people whose weaknesses they are exploiting, e.g., to gambling addicts. In addition, a platform can increase its profits by increasing its competitive advantage over rivals by “hoarding” information, enabling it to engage in this targeting better than others. Of course, if the information were used in a socially productive way, economic efficiency would require its sharing, since information is a public good. (See previous chapter.33) Hoarding such information, while privately profitable, is doubly inefficient because it not only prevents its full use, but endows the platform with market power. With data being a key (largely unpriced) resource, especially important in AI, there is a vicious circle. Larger platforms get more data, which give them a competitive advantage over rivals, enhancing still further their market power, with profits often generated from their better ability to exploit consumers, not their better ability to serve consumers. There are tensions, of course, between the efficient use of information, the anti-competitive hoarding of information, and privacy concerns. One of the reasons that individuals are concerned with privacy is that the disclosure of information can, as we have already noted, allow a variety of forms of exploitation. In the standard competitive market, there is no value to information about consumer preferences. As we discussed in the previous chapter, price is a “sufficient statistic” for conveying all relevant information. But while such information has no incremental private or social value in a competitive market, in the real world, with imperfect competition and incomplete markets, it can be enormously valuable to a firm, increasing significantly its profits. While it is difficult to ascertain the extent, if any, of improvement in resource allocation resulting from the information being exploited by the platforms, one analytic result is clear: the use of this information to engage in price discrimination undermines the standard argument for the efficiency of competitive markets (the first welfare theorem), which is premised on every household and firm facing the same prices.34

Robust theory and fragile practice: part 2  73 There is another adverse by-product of the business model of social media which entails maximizing engagement: engagement is enhanced by enragement, and especially of a kind that has been associated with polarization (discussed above). Thus, one of the societal harms of social media, as it operates today, is that it has created a more divided society, making cooperative actions to address society’s common problems far more difficult. There is an obvious social externality – but one which the social media companies, in their search for ever-increasing profits, pay no attention to. Importantly, the informational advantages that are obtained by the large technological firms of today are qualitatively different from the monopoly or collusive advantages that the main antitrust laws enforced by the Federal Trade Commission and the Department of Justice (the Sherman Act and the Clayton Act) were designed to regulate. This observation suggests that there is a need for a new generation of antitrust legislation that is designed specifically for settings in which differential access to information confers an advantage. Because information itself is so special (one cannot “unknow” information, but one can resell a good), and because of the complex interactions between consumer and user-provided information and industry-obtained or metadata information, this presents a formidable task, one the European Union is beginning to undertake systematically (though incompletely) in its Digital Marketing Act (DMA), the Digital Services Act (DSA), and the General Data Protection Regulation (GDPR). Such legislation, designed to promote the functioning of competitive markets with endogenous asymmetric information while ensuring appropriate levels of consumer privacy and control of user data and other basic rights (as described at the end of the previous sub-section), is absolutely necessary if the extremes of either market breakdown à la Akerlof or exploitative informational monopolies are to be avoided.

7.

CONCLUDING COMMENTS

The development of information economics a half century ago unleashed a revolution in economics that touched upon every aspect of economics. Key presumptions – that markets were efficient and that demand equalled supply in equilibrium – were undermined. The new theories provided new insights into why markets might be absent, into why governance issues were so important, and into why prices did not convey all the information that was relevant for decisions of firms and households. This, in turn, opened up new avenues of enquiry, including into the design of other mechanisms besides competitive markets for resource allocations and for the transmission of relevant information. While there has been considerable excitement about such mechanisms, their domain of applications has remained limited. These limitations, combined with the complexity of the information and mechanism design literature based on rational behaviour and the limitations of even the theoretical results obtained so far, may have reinforced the conviction that what is needed is more analyses based on behavioural economics, recognizing cognitive limitations and the importance of the social determination of beliefs. Firms are always trying to design behaviour and information policies that maximize their profits. As we have seen, that may entail not only costly efforts to overcome information asymmetries but the selective disclosure of information, the creation of information asymmetries, and efforts to undermine the ability to overcome such asymmetries. Because what is privately profitable may not be socially desirable, governments are engaged in designing rules and regulations, including policies that affect the collection, use, and dissemination of

74  The Elgar companion to information economics information, to enhance societal welfare, taking into account how private actors will respond, including with respect to actions related to information. Governments (like firms) know too that consumers may have cognitive limitations and may not be fully rational, and they know that firms know that, and are willing and able to take advantage of these limitations. And because almost all actions (or “non-actions”) can potentially convey information, public policies need to be all-encompassing. All of the problems posed by imperfect and asymmetric information have become worse with digitalization and AI. For instance, there are also pervasive novel externalities associated with algorithms – they may accurately predict behaviour (using information on behaviour of other similar users) even if this particular user has not interacted with this algorithm before. Thus, users interacting with an algorithm exert an externality on future users. This “forward” externality has not been adequately studied, or regulated. An essential part of the distributive battle is the battle over information; and while with perfect information outcomes in (perfectly) competitive markets are Pareto efficient, there is no such presumption that that is the case for the outcomes in these information battles. To the contrary, there is a presumption that this is not so and that appropriate strong government intervention can be welfare-increasing; whether it will be depends on political processes. What was supposed to be the information age has become the “dis- and misinformation” age, with sustained dis- and misinformation – often seemingly inconsistent with the economists’ standard model of rationality. This surfeit of mis- and disinformation has had marked effects on economics, politics, and society. Abundance of information has its perils. Understanding better the social interactions and cognitive functions that make such dis- and misinformation so salient, and devising better policies to combat it, should be one of the main objectives of information economics going forward.

NOTES 1.

2. 3.

4. 5. 6.

In fact, this constraint turns out to be equivalent to a fundamental property of mathematical expectation – the law of iterated expectation, which states that the expectation of the conditional (with respect to some information) expectation of a random variable equals the unconditional expectation. Rephrasing this in the language of beliefs (estimates determined by decision makers about event probabilities), the expected posterior belief has to equal the prior belief (sometimes expressed by saying that “beliefs are a martingale”), and thus, a Bayesian agent cannot be systematically (meaning, in expectation, or on average) misled. Discussed more fully in the previous chapter. If the sender could, instead, commit to an information disclosure strategy before observing her information, the interpretation of “cheap talk” would no longer apply. Because the message is now credible (it’s coming from a communication device the parameters of which were fixed before any private information was observed), the communication now has a “hard information” flavour. The costs of misrepresentation are “internal”, i.e., not imposed by a third party, as in the case of fraud laws. See the previous chapter for a discussion Blackwell informativeness. Antić and Persico (2020) endogenize the difference in preferences in the cheap talk model – the magnitude of the “conflict of interest”. Recall that in the workhorse Crawford-Sobel model the preferences of the sender and receiver differ by a “bias parameter”, interpreted as a difference between the sender-optimal and receiver-optimal actions given any information. The work of Antić and Persico can be thought of as endogenizing this bias parameter (although their setting in much more general); agents can “purchase” elements that change their utility functions. The authors then investigate informativeness of equilibria in two important settings – a competitive market (for the

Robust theory and fragile practice: part 2  75

7. 8.

9.

10. 11. 12.

13. 14. 15. 16. 17. 18. 19.

preference parameters), where they exhibit a condition under which equilibria are as informative as possible, and a principal-agent problem where information transmission will be partial. For a literature review, see Bergemann and Morris (2019). Mechanism design is discussed in the previous chapter. There are many different situations with a multiplicity of equilibria, and therefore, many such refinements. They differ in the strength of their assumptions (stronger assumptions on, say, the inferences that agents make upon observing messages that they should not normally observe typically yield stronger predictions – more equilibria fail such refinements) and the games to which they apply. In this setting the existing refinements turn out to fail to “refine away” the nuisance equilibria, i.e., the equilibria that don’t seem to make much sense. The refinement proposed by Kosenko (2022) operates by asking agents to ascribe actions to types who benefit relatively more than other types; hence the name “belief-payoff monotonicity”. Beliefs (upon observing an action) are monotonically increasing as the payoff of the type that is taking that action. This is a weak, but highly plausible, restriction on beliefs. Two important examples of other models that are robust to misspecification are Bohren (2016) and Bohren and Hauser (2021) which we do not discuss because they deal with sequential social (“observational”) learning, which we do not cover in this review. See Banerjee (1992), Bikhchandani et al. (1992) and the subsequent literature. It is important to note the multiplicity of meanings to the term “robust”. It simply refers to results that hold beyond the specifications that have previously been explored, e.g., concerning the information available to various parties. As bidders might be supposed to be, vis-à-vis the auction designer. Du (2018), Bergemann and Morris (2016), and Brooks and Du (2021a, 2021b, 2021c) extend this literature. Taneva (2019) provides a characterization of the optimal information environment in the static case, using BCE as the solution concept. Mathevet et al. (2020) study a similar problem, but focus on hierarchies of beliefs, which they view as useful for studying “robustness, bounded rationality, collusion, or communication”. One of the main features distinguishing their work is their attention to equilibrium selection; whereas most of the Bayesian persuasion and information design literature has assumed designer- or sender-optimal equilibria (reasoning that the party that moves first can “steer” the game into a particular equilibrium), they explicitly include the possibility of adversarial equilibria: those in which agents may collude on the equilibrium that is worst (not best) for the designer. This in itself is a form of robust information design. Morris et al. (2006) offer a response to Svensson (2006) in which they emphasize the importance of different criteria of efficiency for evaluating outcomes. Akerlof and Shiller (2015) explore related ideas. Convexity of preferences is a sufficient condition. Not even the single crossing property (entailing the indifference of the high-risk individual and the low-risk individual only cross once) has to be satisfied. The disclosure rule is also simple: the insurance firm discloses all of its sales to all firms that have not been disclosed as sellers of insurance to the individual. Models of thinking in this sense fall under the umbrella of epistemic game theory (see, inter alia, Brandenburger, 1992, and Dekel and Siniscalchi, 2014). In the closely related cognitive hierarchies model (Camerer et al., 2004), types best respond to a distribution of types below them, whereas in level-k models types assume that everyone is the next lowest type. In this approach, decision makers are thought to start with an a priori probability distribution (the “prior”) over the unobservable state of the world, and upon observing any information, use Bayes’ theorem to compute “posterior” (or a posteriori) probability distributions. While this approach has the advantage of being well-formulated and mathematically tractable, as we note below, it has the marked disadvantage that it seems counter to how individuals actually behave. Moreover, the questions of how priors are formed, where they come from, whether they are shared by the agents, and how important they are, are all crucial. This approach typically abstracts from investigating these questions, and starts by just assuming that there is a given prior. The following discussion suggests that priors themselves may not be as “rational”, or shared (“common”), as much of this literature assumes.

76  The Elgar companion to information economics 20. This section contains ideas from and is partially borrowed from Schiffrin and Stiglitz (2020). 21. Note 19 raised several critical issues concerning the standard usage of priors. 22. Thus, Kartik et al.’s results, set in the context of rational Bayesian agents, are in line with those obtained in the behavioural economics/psychology literature based on confirmatory bias, where individuals discount information that is not aligned with their priors. Hoff and Stiglitz (2010) show that with confirmatory biases, there can be equilibrium fictions, and that in equilibrium, those with different priors can, in equilibrium, sustain their differences in beliefs. We discuss non-Bayesian belief formation more extensively below. Lipnowski and Mathevet (2018) study information disclosure to an agent with psychological biases and belief-based utility, and show how to take these into account in particular cases. De Clippel and Zhang (2020) study more general persuasion of an agent with non-Bayesian updating rules. They show (among other results) the optimal policy if agents distort their beliefs in a subset of non-Bayesian ways. 23. Maurice Allais was the 1988 recipient of the Nobel Memorial Prize in economics. 24. More recent literature includes De Bondt and Thaler (1985, 1987, 1990), Camerer (1998), Angrisani et al. (2021), Levy and Razin (2015), and Bohren and Hauser (2021). 25. Ortoleva (2012) also cites much of the evidence for non-Bayesian behaviour. 26. See, e.g., Hoff and Stiglitz (2016), and Demeritt et al. (2023). 27. A key aspect of the work cited in note 22 is the social formation of beliefs. 28. This is a very different use of the word “persuasion” than that employed earlier in section 3. If the cost of verification were really zero, they would engage in verification (even if they thought there was some probability that the information was true), in which case the false message would have no impact. But, in practice, there are costs, e.g., of attention diverted from elsewhere. 29. See Stiglitz (2012, 2015, 2019). 30. See Prat (2018) for a definition of “media power” and an application to the case of the US. 31. There is a nascent literature on the subject in media studies. 32. Note that historically, censoring has sometimes taken the form of “regulating” virality, by insisting that adverse news be placed in a position in the newspaper where it is less likely to be read. 33. Because information is a public good, disinformation a public bad, and it may not pay anyone individually to stop its production and dissemination or to engage in activities (like showing its untruthfulness) that might undermine it. Stopping mis- and disinformation is a public good. Without public action, there will be an undersupply of efforts at countering mis- and disinformation. There is thus a strong argument for fraud laws, even if such laws might be viewed as an infringement on “free speech”, interpreted in an absolutist way. Even more so, there is a strong argument for laws restricting virality. 34. Although efficiency can also be sustained by perfect price discrimination, the information generated by platforms and employed by firms engaged in price discrimination is far from sufficient to enable perfect price discrimination (although modern algorithms can come much closer than before). Stiglitz (1977) showed that, in the presence of imperfect information, the welfare losses associated with monopoly arise from the attempt to engage in imperfect price discrimination.

REFERENCES Akerlof, G. A. (1970). The Market for “Lemons”: Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84(3), 488–500. Akerlof, G. A. & Shiller, R. J. (2015). Phishing for Phools: The Economics of Manipulation and Deception. Princeton, NJ: Princeton University Press. Alaoui, L. & Penta, A. (2016). Endogenous Depth of Reasoning. The Review of Economic Studies, 83(4), 1297–1333. Allais, M. (1953). Le comportement de l’homme rationnel devant le risque: critique des postulats et axiomes de l’Ecole Americaine. Econometrica, 21(4), 503–546. Ambrus, A., Azevedo, E. M., & Kamada, Y. (2013). Hierarchical Cheap Talk. Theoretical Economics, 8(1), 233–261.

Robust theory and fragile practice: part 2  77 Ambrus, A. & Takahashi, S. (2008). Multi-Sender Cheap Talk with Restricted State Spaces. Theoretical Economics, 3(1), 1–27. Angeletos, G.-M., Huo, Z., & Sastry, K. A. (2021). Imperfect Macroeconomic Expectations: Evidence and Theory. NBER Macroeconomics Annual, 35, 1–86. Angeletos, G.-M. & La’O, J. (2013). Sentiments. Econometrica, 81(2), 739–779. Angeletos, G.-M. & Pavan, A. (2004). Transparency of Information and Coordination in Economies with Investment Complementarities. American Economic Review, 94(2), 91–98. Angeletos, G.-M. & Pavan, A. (2007). Efficient Use of Information and Social Value of Information. Econometrica, 75(4), 1103–1142. Angrisani, M., Guarino, A., Jehiel, P., & Kitagawa, T. (2021). Information Redundancy Neglect versus Overconfidence: A Social Learning Experiment. American Economic Journal: Microeconomics, 13(3), 163–197. Antić, N. & Persico, N. (2020). Cheap Talk with Endogenous Conflict of Interest. Econometrica, 88, 2663–2695. Arad, A. & Rubinstein, A. (2012). The 11–20 Money Request Game: A Level-k Reasoning Study. American Economic Review, 102(7), 3561–3573. Argenziano, R., Severinov, S., & Squintani, F. (2016). Strategic Information Acquisition and Transmission. American Economic Journal: Microeconomics, 8(3), 119–155. Ariely, D. (2008). Predictably Irrational: The Hidden Forces That Shape Our Decisions. New York: HarperCollins. Ash, E., Mukand, S., & Rodrik, D. (2021). Economic Interests, Worldviews, and Identities: Theory and Evidence on Ideational Politics. Working Paper 29474, National Bureau of Economic Research. Aumann, R. (1976). Agreeing to Disagree. The Annals of Statistics, 4(6), 1236–1239. Banerjee, A. V. (1992). A Simple Model of Herd Behavior. The Quarterly Journal of Economics, 107(3), 797–817. Battaglini, M. (2002). Multiple Referrals and Multidimensional Cheap Talk. Econometrica, 70(4), 1379–1401. Benhabib, J., Wang, P., & Wen, Y. (2015). Sentiments and Aggregate Demand Fluctuations. Econometrica, 83(2), 549–585. Bergemann, D. & Morris, S. (2016). Bayes Correlated Equilibrium and the Comparison of Information Structures in Games. Theoretical Economics, 11(2), 487–522. Bergemann, D. & Morris, S. (2019). Information Design: A Unified Perspective. Journal of Economic Literature, 57(1), 44–95. Bernheim, B. D. (1984). Rationalizable Strategic Behavior. Econometrica, 52(4), 1007–1028. Bikhchandani, S., Hirshleifer, D., & Welch, I. (1992). A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades. Journal of Political Economy, 100(5), 992–1026. Blanchard, O. (2009). The State of Macro. Annual Review of Economics, 1, 209–228. Bohren, J. A. (2016). Informational Herding with Model Misspecification. Journal of Economic Theory, 163(C), 222–247. Bohren, J. A. & Hauser, D. N. (2021). Learning with Heterogeneous Misspecified Models: Characterization and Robustness. Econometrica, 89, 3025–3077. Brandenburger, A. (1992). Knowledge and Equilibrium in Games. Journal of Economic Perspectives, 6(4), 83–101. Bray, M. (1978). Information in Futures Markets. BPhil thesis, University of Oxford. Bray, M. (1981). Futures Trading, Rational Expectations, and the Efficient Markets Hypothesis. Econometrica, 49(3), 575–596. Brooks, B. & Du, S. (2021a). Maxmin Auction Design with Known Expected Values. Working Paper. Brooks, B. & Du, S. (2021b). Optimal Auction Design with Common Values: An Informationally Robust Approach. Econometrica, 89(3), 1313–1360. Brooks, B. & Du, S. (2021c). On the Structure of Informationally Robust Optimal Auctions. Working Paper. Camerer, C. (1998). Bounded Rationality in Individual Decision Making. Experimental Economics, 1, 163–183. Camerer, C., Ho, T.-H., & Chong, J.-K. (2004). A Cognitive Hierarchy Model of Games. The Quarterly Journal of Economics, 119(3), 861–898.

78  The Elgar companion to information economics Carroll, G. (2015). Robustness and Linear Contracts. American Economic Review, 105(2), 536–563. Chakraborty, A. & Harbaugh, R. (2007). Comparative Cheap Talk. Journal of Economic Theory, 132(1), 70–94. Costa-Gomes, M. A., Crawford, V. P., & Broseta, B. (2001). Cognition and Behavior in Normal-Form Games: An Experimental Study. Econometrica, 69, 1193–1235. Crawford, V. P. (2003). Lying for Strategic Advantage: Rational and Boundedly Rational Misrepresentation of Intentions. American Economic Review, 93(1), 133–149. Crawford, V. P., Costa-Gomes, M. A., & Iriberri, N. (2013). Structural Models of Nonequilibrium Strategic Thinking: Theory, Evidence, and Applications. Journal of Economic Literature, 51(1), 5–62. Crawford, V. P. & Iriberri, N. (2007). Fatal Attraction: Salience, Naïveté, and Sophistication in Experimental “Hide-and-Seek” Games. American Economic Review, 97(5), 1731–1750. Crawford, V. P. & Sobel, J. (1982). Strategic Information Transmission. Econometrica, 50(6), 1431–1451. De Bondt, W. F. M. & Thaler, R. H. (1985). Does the Stock Market Overreact? The Journal of Finance, 40(3), 793–805. De Bondt, W. F. M. & Thaler, R. H. (1987). Further Evidence on Investor Overreaction and Stock Market Seasonality. The Journal of Finance, 42(3), 557–581. De Bondt, W. F. M. & Thaler, R. H. (1990). Do Security Analysts Overreact? American Economic Review, 80(2), 52–57. de Clippel, G. & Zhang, X. (2020). Non-Bayesian Persuasion. Working Paper. Dekel, E. & Siniscalchi, M. (2014). Epistemic Game Theory. Mimeo, https://​faculty​.wcas​.northwestern​ .edu/​msi661/​EpistemicGameTheory​-131120​.pdf. Demeritt, A., Hoff, K., & Stiglitz, J. E. (2023). The Other Invisible Hand: The Power of Culture to Promote or Stymie Progress. New York: Columbia University Press. Dosi, G., Napoletano, M., Roventini, A., Stiglitz, J. E., & Treibich, T. (2020). Rational Heuristics? Expectations and Behaviors in Evolving Economies with Heterogeneous Interacting Agents. Economic Inquiry, 58(3), 1487–1516. Du, S. (2018). Robust Mechanisms Under Common Valuation. Econometrica, 86(5), 1569–1588. Edlin, A. S. & Stiglitz, J. E. (1995). Discouraging Rivals: Managerial Rent-Seeking and Economic Inefficiencies. American Economic Review, 85(5), 1301–1312. Ellsberg, D. (1961). Risk, Ambiguity, and the Savage Axioms. The Quarterly Journal of Economics, 75(4), 643–669. Epstein, L. G., Noor, J., & Sandroni, A. (2008). Non-Bayesian Updating: A Theoretical Framework. Theoretical Economics, 3(2), 193–229. Epstein, L. G., Noor, J., & Sandroni, A. (2010). Non-Bayesian Learning. B. E. Journal of Theoretical Economics: Advances in Theoretical Economics, 10(1), 3. Fredrickson, B. L. & Kahneman, D. (1993). Duration Neglect in Retrospective Evaluations of Affective Episodes. Journal of Personality and Social Psychology, 65(1), 45–55. Glaeser, E. L. (2005). The Political Economy of Hatred. The Quarterly Journal of Economics, 120(1), 45–86. Goldstein, I. & and Yang, L. (2017). Information Disclosure in Financial Markets. Annual Review of Financial Economics, 9, 101–125. Greenwald, B. & Stiglitz, J. E. (1986). Externalities in Economies with Imperfect Information and Incomplete Markets. The Quarterly Journal of Economics, 101(2), 229–264. Guo, Y. & Shmaya, E. (2019). Robust Monopoly Regulation. Working Paper. https://​arxiv​.org/​pdf/​1910​ .04260​.pdf. Guriev, S. & Treisman, D. (2020). A theory of informational autocracy. Journal of Public Economics, 186, 104–158. Guzman, M. & Stiglitz, J. E. (2020). Towards a Dynamic Disequilibrium Theory with Randomness. Oxford Review of Economic Policy, 36(3), 621–674. Guzman, M. & Stiglitz, J. E. (2021). Pseudo-Wealth and Consumption Fluctuations. The Economic Journal, 131(633), 372–391. Hedlund, J. (2017). Bayesian Persuasion by a Privately Informed Sender. Journal of Economic Theory, 167, 229–268.

Robust theory and fragile practice: part 2  79 Hoff, K. & Stiglitz, J. E. (2010). Equilibrium Fictions: A Cognitive Approach to Societal Rigidity. American Economic Review, 100(2), 141–146. Hoff, K. & Stiglitz, J. E. (2016). Striving for Balance in Economics: Towards a Theory of the Social Determination of Behavior. Journal of Economic Behavior & Organization, 126, 25–57. Huo, Z. & Pedroni, M. (2020). A Single-Judge Solution to Beauty Contests. American Economic Review, 110(2), 526–568. Jackson, M. O. & Pernoud, A. (2021). Systemic Risk in Financial Networks: A Survey. Annual Review of Economics, 13, 171–202. Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A. (1993). When More Pain Is Preferred to Less: Adding a Better End. Psychological Science, 4(6), 401–405. Kahneman, D. & Tversky, A. (1979). Prospect Theory: An Analysis of Decision Under Risk. Econometrica, 47(2), 263–291. Kamenica, E. & Gentzkow, M. (2011). Bayesian Persuasion. American Economic Review, 101(6), 2590–2615. Kanoria, Y. & Saban, D. (2017). Facilitating the Search for Partners on Matching Platforms: Restricting the Agents’ Actions. Working Paper. Kartik, N. (2009). Strategic Communication with Lying Costs. The Review of Economic Studies, 76(4), 1359–1395. Kartik, N., Lee, F. X., & Suen, W. (2021). Information Validates the Prior: A Theorem on Bayesian Updating and Applications. American Economic Review: Insights, 3(2), 165–182. Keynes, J. M. (1921). A Treatise on Probability. London: Macmillan & Co. Keynes, J. M. (1936). The General Theory of Employment, Interest and Money. London: Macmillan. Kolotilin, A., Mylovanov, T., Zapechelnyuk, A., & Li, M. (2017). Persuasion of a Privately Informed Receiver. Econometrica, 85, 1949–1964. Kosenko, A. (2022). Constrained Persuasion with Private Information. The B.E. Journal of Theoretical Economics. https://​doi​.org/​10​.1515/​bejte​-2021​-0017. Kosenko, A., Stiglitz, J. E., & Yun, J. (2023). Bilateral Information Disclosure in Adverse Selection Markets with Nonexclusive Competition. Journal of Economic Behavior & Organization, 205, 144–168. Levy, G. & Razin, R. (2015). Correlation Neglect, Voting Behavior, and Information Aggregation. American Economic Review, 105(4), 1634–1645. Lipnowski, E. & Mathevet, L. (2018). Disclosure to a Psychological Audience. American Economic Journal: Microeconomics, 10(4), 67–93. Machina, M. J. (1987). Choice Under Uncertainty: Problems Solved and Unsolved. Journal of Economic Perspectives, 1(1), 121–154. Mathevet, L., Perego, J., & Taneva, I. (2020). On Information Design in Games. Journal of Political Economy, 128(4), 1370–1404. Milgrom, P. & Roberts, J. (1986). Relying on the Information of Interested Parties. The RAND Journal of Economics, 17(1), 18–32. Morris, S. & Shin, H. S. (2002). Social Value of Public Information. American Economic Review, 92(5), 1521–1534. Morris, S., Shin, H. S., & Tong, H. (2006). Social Value of Public Information: Morris and Shin (2002) Is Actually Pro-Transparency, Not Con: Reply. American Economic Review, 96(1), 453–455. Nagel, R. (1995). Unraveling in Guessing Games: An Experimental Study. American Economic Review, 85(5), 1313–1326. Ortoleva, P. (2012). Modeling the Change of Paradigm: Non-Bayesian Reactions to Unexpected News. American Economic Review, 102(6), 2410–2436. Pearce, D. G. (1984). Rationalizable Strategic Behavior and the Problem of Perfection. Econometrica, 52(4), 1029–1050. Pei, H. D. (2015). Communication with Endogenous Information Acquisition. Journal of Economic Theory, 160, 132–149. Penczynski, S. P. (2013). Herding Differently: A Level-k Model of Social Learning. Working Papers 13-01, University of Mannheim, Department of Economics. Pickard, V. (2015). America’s Battle for Media Democracy: The Triumph of Corporate Libertarianism and the Future of Media Reform. New York: Cambridge University Press.

80  The Elgar companion to information economics Prat, A. (2018). Media Power. Journal of Political Economy, 126(4), 1747–1783. Prelec, D. (1998). The Probability Weighting Function. Econometrica, 66(3), 497–527. Rabin, M. & Schrag, J. L. (1999). First Impressions Matter: A Model of Confirmatory Bias. The Quarterly Journal of Economics, 114(1), 37–82. Rothschild, M. & Stiglitz, J. E. (1976). Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information. The Quarterly Journal of Economics, 90(4), 629–649. Schiffrin, A. & Stiglitz, J. E. (2020). Facebook Does Not Understand the Marketplace of Ideas. The Financial Times, 17 January. Sethi, R. & Yildiz, M. (2012). Public Disagreement. American Economic Journal: Microeconomics, 4(3), 57–95. Sethi, R. & Yildiz, M. (2016). Communication with Unknown Perspectives. Econometrica, 84(6), 2029–2069. Simon, H. A. (1955). A Behavioral Model of Rational Choice. The Quarterly Journal of Economics, 69(1), 99–118. https://​doi​.org/​10​.2307/​1884852 Sobel, J. (2013). Giving and Receiving Advice. In D. Acemoglu, M. Arellano, & E. Dekel (Eds.), Advances in Economics and Econometrics. Cambridge: Cambridge University Press, pp. 305–341. Stahl, D. O. & Wilson, P. W. (1994). Experimental Evidence on Players’ Models of Other Players. Journal of Economic Behavior & Organization, 25(3), 309–327. Stiglitz, J. E. (1977). Monopoly, Non-Linear Pricing and Imperfect Information: The Insurance Market. Review of Economic Studies, 44(3), 407–430. Stiglitz, J. E. (1984). Information, Screening and Welfare. In M. Boyer & R. Khilstrom (Eds.), Bayesian Models in Economic Theory. Amsterdam: Elsevier Science Publications, pp. 209–239. Stiglitz, J. E. (2012). The Price of Inequality: How Today’s Divided Society Endangers Our Future. New York: W. W. Norton. Stiglitz, J. E. (2015). New Theoretical Perspectives on the Distribution of Income and Wealth among Individuals: Part IV: Land and Credit. NBER Working Paper No. 21192. Stiglitz, J. E. (2019). People, Power, and Profits: A Progressive Capitalism for an Age of Discontent. New York: Penguin Books. Svensson, L. E. O. (2006). Social Value of Public Information: Comment: Morris and Shin (2002) Is Actually Pro Transparency, Not Con. American Economic Review, 96(1), 448–452. Szeidl, A. & Szucs, F. (2022). The Political Economy of Alternative Realities. Mimeo. Taneva, I. (2019). Information Design. American Economic Journal: Microeconomics, 11(4), 151–185. Tversky, A. & Kahneman, D. (1992). Advances in Prospect Theory: Cumulative Representation of Uncertainty. Journal of Risk and Uncertainty, 5(4), 297–323. Woodford, M. (2002). Imperfect Common Knowledge and the Effects of Monetary Policy. In P. Aghion (Ed.), Knowledge, Information, and Expectations in Modern Macroeconomics. Princeton, NJ: Princeton University Press, pp. 25–58.

4. Information and income distribution: The perspective of information economics Julia Włodarczyk

1. INTRODUCTION Information and income distribution are inherently interdependent. Information flows shape the income distribution and income distribution determines various flows of information. To a large extent, income distribution reflects the informational role of wages and other sources of income and mirrors information imperfections and asymmetries on the labour market. It bears information about objective differences between individuals, subjective valuations of their potential productivity and exerted effort, as well as evolution of institutions and distributive conflicts on the labour and capital market. The character of the relationship between information and income distribution has evolved over time, mainly due to technological progress, with the special role of communication technologies, both traditional and digital. From the very beginning communication technologies (such as speech or writing) enabled cooperation and negotiations over the division of labour and income. In the analogue world, the logic of technological progress regarding communication comes down to the tendency of substituting media offering rich content, but a relatively short information chain (e.g., speech) with media transmitting more deficient information, but at a longer distance (cf. Blain, 2002, p. 40). Recently, the digital revolution has allowed not only to transmit information rich in content at any distance, but also to reverse the traditional emphasis on information scarcity into abundance. Without communication, conflicts over scarce resources are usually solved by force. On one hand, communication brings about the possibility of negotiating peaceful (market) solutions. On the other hand, it motivates individuals to pursue higher incomes – not necessarily in absolute terms, but often in relative terms: individuals desire higher incomes than others. In this pursuit individuals resort to and are subject to omnipresent practices reflecting information imperfections, such as signalling or screening. The goals of the chapter are to: (1) increase the understanding of mechanisms that shape the income distribution adopting the perspective of mainstream economics confronted with information economics; (2) demonstrate differences between the traditional perspective of information economics accentuating scarcity of information and the new digital economics perspective highlighting abundance of information, analysing mechanisms potentially responsible for shaping prevailing inequalities and emergence of new forms of inequality; (3) discuss some implications of information flows for individual perception of income inequalities and life satisfaction. The structure of the chapter reflects formulated goals. Section 2 has an introductory character and starts with an exposition of selected observations concerning income distribution before the advent of information economics. Even though early literature did not make any substantial references to information flows, transmission and accumulation of information 81

82  The Elgar companion to information economics stands behind many processes shaping income distribution. Then, the emphasis shifts to rents and rent-seeking activities which require information imperfections and determine the distribution of income as well. The core of the chapter consists of sections 3 and 4 which contrast the relationship between information and income distribution under relative scarcity and abundance of information, mirroring the traditional and more recent approaches to information economics. Both sections begin with an overview of market operations under imperfect information and subsequently discuss issues directly linked with income distribution. Section 5 extends the analysis to present selected behavioural and social implications of income differentials in an increasingly, but never fully, transparent world of the digital economy and refers to concepts such as relative deprivations, job satisfaction and playbour. The chapter ends with concluding remarks.

2.

INCOME DISTRIBUTION AND ITS ORIGIN

Personal income distribution is a statistical representation of the division of income among individuals from a given population. The origins of investigations in income distribution are usually traced back to the nineteenth century, when statistical information on income (stemming from tax data) became available. Although the early works did not refer to information explicitly, information can be treated as an important component of education, training, abilities (e.g., more able individuals tend to possess more information about how to perform their tasks), and institutions (cf. ignorantia iuris nocet). Even individual motivations require being informed about what is valued as good or bad for an agent and the whole society. There is a certain dialectics behind the scene. As argued below, individual differences in innate abilities, gender or life expectancy coupled with unanticipated idiosyncratic shocks that differentiate incomes can be attributed to chance or risk. To some extent, these factors can be associated with exogenous sources of imperfect (scarce) information. On the other hand, a portion of income differences is created by agents attempting to overcome scarcity of information. Intentional exploitation of emerging opportunities often implies blocking access to information or its economic use to others. This procedure is closely related to rent seeking. Therefore, after an exposition of early scientific findings concerning income distribution which did not clearly distinguish between chance and design, special attention is paid to rents and rent seeking. This section is completed with a stylized comparison of two different types of income distribution which roughly correspond to the exogenous and endogenous forces shaping the income distribution and information flows in the economy. 2.1

Positive Skewness of Income Distribution: Early Empirical and Theoretical Observations

In one of the first widely recognized works, Ammon (1895) made an attempt to demonstrate that the income distribution (with the exception of lowest incomes) almost coincides with the Galton’s ability curve following the Gaussian (normal) distribution.1 Ammon observed a lack of symmetry of the income distribution with regard to its tails and he also acknowledged certain differences between the Galton’s ability curve and the income curve, but attributed them mostly to the method of collecting income statistics (Ammon, 1895, pp. 127–133; cf.

Information and income distribution  83 Ammon, 1899). Ammon (1899, p. 229) further discussed differences in incomes of individuals driven by egoistic and altruistic motives. Manufacturers, land owners or bankers earned more than the highest public officials, statesmen or scholars, not because of a lower importance of the services provided by the latter, but due to different goals, tastes and preference pursued by both groups. In turn, low incomes of representants of the working class are not the result of altruism, but potential deficiencies in intellectual, moral and economic traits. Still, despite these differences Ammon believed in normality of income distribution. Ammon’s belief in theoretical concepts yielding normal distribution of income was criticized by Pareto (1896), whose work demonstrating a significant skewness of the income distribution due to a power-law relationship (a linear inverse relationship between logarithm of income y and logarithm of the number of agents receiving income equal to or greater than y) inspired generations of researchers. Importantly, Pareto (1896) noted that the observed tendency for incomes to cluster in a certain manner is characteristic for societies and periods under analysis, but is not necessarily universal. In particular, the law of distribution of different categories of income is not the same. Extreme skewness of the income distribution (with small average differences in the incomes at the bottom tail and greater differences with successively higher incomes) has been emphasized by many writers, including Young (1917), who perceived it as a result of a significant inequality of opportunity. Young (1917) also predicted an increase in income inequality resulting from an increasing life expectancy coupled with the process of accumulation of wealth over the life cycle. Although some individuals grow poorer over time, usually individuals become richer as they age, but there is a significant heterogeneity in the individual rate of wealth accumulation which determines the value of bequests. Young (1917) concluded that analysis of inequality stemming from inheritance of property requires taking into account the age profile of the decedents. Apart from inequality of opportunity (in some cases implicitly related to information asymmetry) and inheritance of property, other explanations of positive skewness of income distribution have been discussed in the literature over the decades. For instance, according to Pigou (1932, p. 651) the skewness of income distribution might be the result of merging different subpopulations, each characterized by a normal distribution of income. Nevertheless, he pointed also at a more important explanation, namely that income depends on a combination of individual capacities (manual and mental) and inherited property that allows for increasing individual capacities, but is highly concentrated. Pigou (1932, p. 652) also accentuated the difference between income earned as a result of manual or mental work and income from investment or property, with the latter being spread less uniformly than the former. Besides, with increasing total income, the share of the investment or property income will continually grow, while income earned from personal activity is likely to increase at first (as property offers new opportunities), but then to decrease (both in relative and absolute terms) when individuals are no longer motivated to engage in any profitable activity personally (Pigou, 1932, p. 654). The idea that the shape of the income distribution depends not only on external circumstances, but is the matter of individual choice and preferences (expressed in other ways than collective action) was further developed by Friedman (1953). In particular, Friedman (1953) emphasized the role of individual preferences towards risk. If two populations differ only in the degree of risk aversion, they still may have normal distributions of income: one with a lower mean and a lower variance and the other with a higher mean and a higher variance.

84  The Elgar companion to information economics However, the sum of such distributions will be positively skewed. Friedman (1953) concluded that tastes and preferences of the members of the society are important in explaining the shape of income distribution and that institutional arrangements are in fact instruments of achieving the distribution of income that conforms to these tastes and preferences (see e.g., Kanbur, 1979 for critique and extension of analysis). Miller (1955), similarly to Friedman, attributed the skewness of income distribution to mergers of several distributions that can be symmetrical. In particular, he found that the skewness of the income curve was due to merging income distributions for men and women. These differences were likely to be related to customs and social preferences. Mincer (1958) criticized Pigou, Friedman and Miller as they did not explain why merging symmetric distributions results in an aggregate distribution that is positively skewed. Mincer (1958) demonstrated that the existence of positive skewness of income distribution was due to differences in time spent on training. More precisely, positive skewness is a consequence of inter-occupational and intra-occupational income disparities, with the former linked to differences in training, and the latter adding experience collected over time. Trained individuals receive higher wages. The difference between earnings of individuals with identical abilities, but differing by the number of years of training depends positively on the discount rate (depicting individual devotion due to income postponement) and negatively on the span of professional life during which the costs of training must be recovered. As a consequence, income earned at various levels of training (differing additively by the same number of years) differs by a multiplicative factor.2 Therefore, even if the distribution of training is symmetrical, the distribution of earnings will be asymmetrical and positively skewed (log-normal income distribution is one of the possible outcomes). Importantly, if individuals are heterogeneous in terms of abilities, one can expect a positive correlation between the level of individual abilities and the amount of training, increasing skewness of the distribution even further. Thus Mincer (1958) showed how the element of choice concerning training also yields positive skewness of the distribution. (Naturally, the structure of the educational system can potentially reduce this impact, depending e.g., on the years of compulsory education regardless of individual capacities and tastes.) To recapitulate, early literature offered explanations for the skewness of income distribution, including combination of factors independent from individuals (such as innate abilities) and those closely related to their choices and preferences. Even though the issue of inequality of opportunities associated e.g., with individual choices concerning education and training (accumulation of knowledge) or saving decisions and bequests (accumulation of wealth) was frequently discussed, researchers focused mostly on earnings and did not analyse rents and rent-seeking activity in detail. 2.2

Rents and Rent-Seeking Activity as Drivers of Excessive Skewness of Income Distribution

Rent is an economic surplus and can be treated as an unearned increment (Bronfenbrenner, 2007, p. 355). This definition puts rents in sharp contrast to profits which are earned over the course of business operations associated with bearing risk. Thus, to simplify considerations, in this chapter profits are assumed to be the result of productive, while rents of unproductive activities. Similarly, profit seeking means the pursuit of revenues stemming from risky operations, while rent seeking is oriented towards eliminating risk.

Information and income distribution  85 The literature discusses various types of rents: from Ricardian rents derived from fixed stocks such as land, the Marshallian concept of quasi-rents referring to stocks that are constant in the short run only (extended to the rent of ability, associated with naturally or artificially limited skills of labourers), monopoly rents obtained through the price mechanism and other transfers created within the political mechanism (e.g., taxes, subsidies) to Schumpeterian rents that emerge due to innovation and are closely related to information costs or information imperfections (cf. e.g., Bronfenbrenner, 2007, p. 349; Birch, 2020). The logic of Schumpeterian rents can be extended to capture incentives necessary to generate all types of information. As the use of new information is rewarded with higher returns, better informed agents receive rents, but only until this information is diffused and publicly available. With instant access to information rents as a source of income disappear, leaving no incentives to generate new information. Contrarily, protection preventing the spread of information stimulates innovation, but creates social costs. Therefore, regulations ruling the period during which rents are secured have important distributive effects. Rent seeking means a specific use of resources to obtain an uncompensated transfer from others, often as a result of securing or blocking changes of law (Hartle, 1983). Rent seeking in the aggregate means substantial losses to a society, because it diverts resources towards unproductive activities (Tollison, 1997, p. 506). Notably, this reallocation may disproportionately affect the situation of agents from various segments of the income distribution, reconfiguring the institutional environment and opportunities of generating incomes. Generally, the incentive to engage in rent seeking is relatively stronger for the rich, who are motivated to protect their wealth from expropriation by others, especially when property rights are insecure (Chakraborty & Dabla-Norris, 2006, p. 29). This is in line with findings of Pigou mentioned earlier according to which individuals achieving higher incomes are less motivated to engage in any profitable activity personally (Pigou, 1932, p. 654). In fact, there is strong feedback between rent seeking and income inequalities: income inequalities promote rent seeking (especially as an instrument of protection from expropriation and of augmentation of wealth and incomes) and – inversely – rent seeking increases income inequalities and leads to wealth condensation and polarization of the society (Włodarczyk, 2013b). In the extreme case, if the richest were able to exploit other strata of the society, depriving other individuals one by one from sources of income that can be reappropriated, the poorest could also engage in rent seeking (having nothing to lose), while the middle (productive) class would vanish in accordance with Marxian arguments. Naturally, due to innovation and diffusion of information and rents, such an extreme polarization is rather unlikely. Interestingly, the lack of deeper reflections of distributive consequences of rents and rent seeking within mainstream economics is to some extent paradoxical. The origin of the neoclassical theory lies in the marginal revolution. Marginal revolution effectively meant the generalization of the marginal principle underlying the differential theory of rent to the whole theory of production and distribution, and thus required assumptions that all the factors of production can be treated like land (Pasinetti, 2000). In general, the division of income in the economy depends on who performs the process of production and is able to capture the residual. However, equal treatment of factors of production included in a single neoclassical production function (motivated to some extent by the beauty of simple representations) resulted in the assumption that the marginal product of labour is wage, the marginal product of land is land rent, the marginal product of capital is profit, and that no residual is left (Pasinetti, 2000).

86  The Elgar companion to information economics As a consequence, rents became a sign of inefficiency in neoclassical models, while their role associated with creating incentives for information generation was overlooked despite their significance for market operations and distributive consequences. In particular, the presence of rents among other sources of income contributes to a greater skewness of income distribution, because rents tend to be concentrated among high-income individuals and often reflect more or less formalized hierarchies designed to exploit market opportunities. 2.3

Chance and Design: Dialectics of Income Distribution Skewness

Preceding subsections discussed two major sources of income differentials – an exogenous one (attributed mostly to chance) and an endogenous one (designed by agents). For illustrative purposes it is useful to contrast stylized differences between distributions exemplifying both regimes. These two regimes correspond inter alia to the opposition between spontaneous market interactions and hierarchies, profits and rents as sources of income, but also adaptive and exploitative behaviour with regard to information. They can be illustrated by the log-normal and the power-law distribution, respectively (see Table 4.1).3 As already mentioned, differences in innate abilities, preferences or training can potentially lead to a log-normal income distribution. Besides, if individual incomes are subject to a series of random proportional changes (e.g., capital gains or losses, potentially associated with bearing risk), then under the Central Limit Theorem the distribution of logarithms of these incomes will tend to normality, producing a log-normal income distribution (Bronfenbrenner, 2007, p. 53). Within this simplified framework, log-normal income distribution appears as a stable, stochastic outcome of spontaneous market transactions conducted by agents without any significant monopoly power. The additive character of accumulation prevents them from obtaining substantial income from savings and thus does not disturb consumption-saving decisions, stabilizing income distribution. From the point of view of information, all agents are exposed to imperfect, but more or less symmetrical information. Their main motivation to engage in informational activity is to adapt to existing imperfections. In turn, the power-law distribution reflects a hierarchical pyramid of incomes and can be observed in a situation when every worker (except the ones who are the lowest in the hierarchy) receives an income proportional to the total income received by his or her subordinates (Bronfenbrenner, 2007, p. 54). This allows for the multiplicative character of accumulation and stronger linkages with the capital market. Hierarchies, especially in the form of complex firm structures, arise as a preferable form of organization in a situation when transaction costs on the market are high and information is scarce (cf. Williamson, 1981). In fact, hierarchies, ranks and orders issued by superiors to subordinates can actually be treated as substitutes of perfect information (Kelly, 1998, p. 119). Therefore, hierarchies determine not only differences in incomes earned at particular levels, but also the flow of information between these levels. Under the power-law regime the flow of information is controlled to a larger extent than under the log-normal one. Intentionally, agents (especially on the higher levels of hierarchies) are not only looking for information, but also they select what information to convey or conceal. As a result, information asymmetry between various levels of organizational or market hierarchies gives rise to phenomena like adverse selection or moral hazard.

Information and income distribution  87 Table 4.1

Stylized characteristics of a log-normal and power-law income distribution

Criterion

Log-normal income distribution

Mechanisms generating income

Inequality by chance: spontaneous interactions Inequality by design: hierarchy in which

Power-law income distribution

distributions

between agents (with more or less equal

each agent receives income proportionately

position) subject to random and proportional

higher than agents located lower in this

changes (gains and losses)

hierarchy

Market

Competitive (imperfect competition)

Monopolized (imperfect competition)

Main sources of income

Salaries, remuneration for productive work,

Capital gains, remuneration for

profits

non-productive work, rents (as well as sources mentioned in the previous column)

Sources of income differences

Natural, exogeneous

Institutional, endogenous

Model of interactions (exchange/

One-to-one – spontaneous

One-to-many – designed, institutionalized

Additive (agents save according to their life

Multiplicative (based on compound

cycle, saving a small proportion of their

interest)

communication) Character of accumulation

income) Connectivity

Low (low connectivity disables agents from

High (with high variability)

reaching high profits, but it is also a barrier preventing them from incurring high losses) Character of income distribution

Stochastic, stable, dependent to a large extent

Deterministic, more variable, dependent to

on the situation on the labour market

a large extent on the situation on the capital market

Information

Information activity

Imperfect, scarce or abundant, but more or less Imperfect, scarce or abundant, and symmetrical (all agents operate under scarcity

asymmetrical (scarcity and abundance of

or abundance of information)

information may coexist)

Adapting to information imperfections, dealing Overcoming and exploiting information with risk and uncertainty, learning

imperfections (often in the form of opportunistic behaviour), hoarding information

Source: Own elaboration based on Milaković (2005), Risau Gusman et al. (2005), Sinha (2005), Yarlagadda & Das (2005), Włodarczyk (2013a).

In line with the stylized characteristics presented in Table 4.1, one may conclude that in a risky and uncertain world, skewness of income distribution arises naturally and could be potentially proxied by a log-normal distribution,4 while a power-law distribution reflects artificially created inequality.5 Both types of inequality coexist in the real world, because many agents receive income from various sources.6 However, power-law distribution is characteristic for the top of the distribution and often reflects not only high income, but also a greater number of income sources. This section leads to two important conclusions. First, deviations from log-normal income distribution can be attributed to phenomena such as information asymmetry or an institutional equivalent of asymmetric information e.g., property rights that prevent not necessarily the spread of information per se, but the spread of its economic use. Second, because property rights are often used to secure rents, rents can be treated as an important channel translating information imperfections into income inequality depicted e.g., by the excessive skewness of the income distribution. To recapitulate, the exposition of potential mechanisms shaping income distribution presented above, even though simplified, clearly illustrates not only the relevance, but also

88  The Elgar companion to information economics potential channels through which information economics can be linked with investigations in income distribution. The opposition between profit and rent, productive and unproductive activities, chance and design, as well as market and hierarchy reflects existing information imperfections and the way economic agents deal with them.

3.

INFORMATION IMPERFECTIONS AND INCOME DISTRIBUTION FROM THE TRADITIONAL PERSPECTIVE OF INFORMATION ECONOMICS

As sketched above, neither the problems of information nor income distribution were central to neoclassical economics. Representative agents from the neoclassical tradition were supposed to possess perfect information about all the technicalities associated with the process of production (both of goods and know-how) and about their intertemporal utility function, so that they were able to maximize the present value of their satisfaction in an infinite time horizon (Pasinetti, 2000). Besides, the standard paradigm argued that institutions and history did not matter, nor did incentives and motivation (Stiglitz, 2002). The perspective of information economics stands in stark contrast to the neoclassical one. As argued below, information imperfections exert significant impact on market operations, behaviour of individual agents and income distribution. Economic agents spend a fraction of their time on information activity. Agents looking for attractive job offers or productive job applicants engage in screening, agents trying to sell high-quality goods and competences bear costs of signalling the characteristics of their goods or their job offer, while other agents try to exploit existing information asymmetry to sell products of poor quality. All these activities (successful or not) shape the income distribution. 3.1

Information and Market Imperfections: Critique of Mainstream Economics

There were some attempts to emphasize the role of knowledge and information in economics before the advent of information economics. For instance, Hayek (1945), similarly to Knight (1921)7 stressed the importance of knowledge closely connected with change. As the imperfection of human knowledge is unavoidable, Hayek (1945) saw a need for a mechanism by which knowledge can be constantly communicated and acquired. For him the real function of the price system is communicating information. This focus on prices (associated later with the privileged position of the Arrow-Debreu Walrasian general equilibrium theory) and on representative agents effectively crowded out the analysis of income distribution from the discourse, as it made no sense to discuss division of income among representative individuals (Pasinetti, 2000). Importantly, information communicated by prices does not explain the whole complexity of market processes. Information economics has challenged the fundamentals of mainstream economics, including competitive equilibrium analysis, the law of one price or efficient market hypothesis. Accordingly, information economics has explained why in equilibrium demand may not be equal to supply, why firms may charge prices exceeding their marginal costs and pay wages exceeding workers’ reservation wage (thus generating and sustaining price and wage distributions), and why prices (including stock prices) may aggregate information imperfectly

Information and income distribution  89 (Stiglitz, 2002, cf. Grossman & Stiglitz, 1980). One of the reasons for this state of affairs is rent seeking and attempts to prevent immediate access to information to a wider public. Information economics along with institutional economics contributed to deepening of the analysis of rent-seeking activities shaping the income distribution. Under perfect competition there are no rents. When a monopolist captures all the rents, there are no battles over rents. However, in a typical situation of imperfect competition and imperfect information, conflicts over rents and attempts to capture or restrict them arise somehow automatically and their result has important implications for the market structure as well as the institutional order (Stiglitz, 2017). Information economics questioned the fundamental theorems of welfare economics. In particular, it showed inseparability of efficiency and distributive effects: efficiency gains can be lost due to their distributive consequences, while changes in the distribution may have efficiency consequences (Stiglitz, 2017). One of the fundamental differences between markets operating under complete and imperfect information is that, with imperfect information, market behaviour of individual agents conveys information. Sometimes agents attempt to conceal information (this activity may reflect rent seeking) or even behave in a dishonest way (Stiglitz, 2002), including behaviour associated with moral hazard (cf. Arrow, 1971). Besides, deliberate misinformation is often spread not only by individual agents, but also by whole groups (Tullock, 1988, p. 23).8 Interestingly, in sectors relying significantly on information flows the prices and wages cannot be defined with a similar precision as in the industrial sector. Markets tend to overestimate unknown or highly abstract values (e.g., high skills). Setting high prices and offering high wages results from the margin of error that is much higher for a highly complex organizational order than pricing of physical goods (Chalidze, 2000, pp. 175–179). The omnipresent and increasing uncertainty (e.g., about the quality of a product or a worker) is also an inevitable result of differentiation of products, progressive division of labour, as well as heterogeneity and increasing complexity of economic reality.9 This makes information flows increasingly relevant to operations of many markets, but potentially also more disruptive. 3.2

Income Distribution: Search Costs and Beyond

Imperfect information is closely related not only to market imperfections in general, but also to specific circumstances shaping the income distribution. A non-uniform income distribution is not the market failure to arbitrage differences in wages and other sources of income or prices, but is created by the market (Stiglitz, 2002). One of the channels through which information exerts a direct impact on income distribution is the continuous growth of the information sector over many decades. Acceleration of this growth rate implies greater demand for workers performing cognitive tasks at the cost of manual workers which leads to a problem of employability of the latter (cf. Machlup, 1962; Porat, 1977) and differentiates the evolution of incomes in both groups of workers.10 Besides, along with the process of economic growth the structure of the demand changes. Growing demand for information goods and services results in pronounced information imperfections in many areas, again including changes in the income distribution (Stiglitz, 2017). Wage (or price) distribution is often treated as given in the literature. Information economics showed that search costs lead to an equilibrium characterized by a wage (or price) distribution

90  The Elgar companion to information economics even if search costs are small and uniform for all agents (Stiglitz, 2002). Actually, information economics provides additional explanations of income distribution other than search costs. Generally, knowledgeable and productive workers have a good bargaining position on the labour market and receive higher wages. Therefore, under perfect information, identical untrained workers are ready to accept lower wages, if they know that they will receive higher incomes after training. However, in reality job applicants are not perfectly informed about the quality of a job they are applying for and the amount of training provided by the employer. In this situation, firms are likely to use their wage offers to untrained workers to signal information about the job and training. Highly productive companies will offer higher wages to distinguish themselves from low-productivity companies. As a result, demand and supply may not be equal for all jobs and job rationing may occur.11 Candidates rejected at high quality jobs have to accept employment in companies offering a lower amount of training and lower incomes. Therefore, under imperfect information, equilibrium on the labour market results in unequal treatment of identical job applicants which differentiates incomes (Bester, 2004). When individuals are not identical (e.g., later in their professional career), more productive workers will have an incentive to convince the potential employer that they are. If there are no costs of such an activity, then there will be a fully revealing equilibrium, even though less productive individuals would prefer to conceal information about their characteristics (Stiglitz, 2002).12 If such a communication is costly, job applicants will engage in signalling resulting in multiple equilibria (Spence, 1973). Employers have incentives to gather information about workers and engage in screening. Finding workers that are more productive than is recognized by other employers, they can pay the wage determined by public information and appropriate the surplus stemming from their private information about worker productivity. Such rents are likely to dissipate as soon as the quality of the worker becomes known to other employers (Stiglitz, 2002). If this is the case, it will be the worker that might be able to capture the rent of ability at the cost of the employer. Employers invest in sophisticated instruments to find appropriate workers and appropriate rents from hiring them. Human resources management thus has an important dynamic component associated with creating new rents for the employer. This dynamic aspect of human resources management refers also to monitoring the risk of moral hazard as many employees change behaviour after signing the contract or – as in the example discussed above – after learning how the market values their productivity. There are significant information-related externalities with regard to income distribution. In equilibrium, gains of the better-informed agent (e.g., employer) come at the cost of less-informed agents (e.g., other employers and highly productive workers). Similarly, gains of the more productive agents come at the cost of the less productive ones as higher wages of more able individuals drive down wages of those that are less able (Stiglitz, 2002). To recapitulate, information imperfections in general and returns to information in particular create the possibility to capture information rents and thus have significant distributive effects. Distributive consequences of information imperfections are related to endogenous dynamics of income and unemployment. Information imperfections amplify the trade-off between high incomes and unemployment as well as between chance and design. Sometimes identical workers due to job rationing may end up on significantly different income trajectories which can lead to excessive inequalities and greater skewness of income distribution.

Information and income distribution  91

4.

INCOME DISTRIBUTION UNDER IMPERFECT, BUT ABUNDANT INFORMATION: IMPACT OF DIGITALIZATION

Scarcity of information was the central concept in the early information economics literature. The digital revolution has made it possible to reduce adverse effects of scarcity of information and offer information in abundance, but it has also propelled new forces that exert pressures on the income distribution. For example, digitalization significantly reduces many costs, including search costs, replication costs, tracking costs and verification costs (Goldfarb & Tucker, 2019). By lowering search costs digital technologies have increased competition (especially in case of homogeneous or specified goods), however, they have also increased the ability to exploit less-informed agents by the better-informed (Stiglitz, 2017). Reduction of these costs has important implications for prices (in the extreme case goods are offered for free) and wages (also subject to a downward pressure). On the other hand, digitalization creates new opportunities to generate and capture rents. This section points at several aspects of digitalization that change the market structure, behaviour of individual agents and income distribution compared to the situation of scarce information. Special attention is paid to network effects, digital platforms, status of platform workers and their remuneration. 4.1 Markets, Networks, Platforms Digital networks in general and digital platforms (multi-sided markets) in particular, have become the dominant form of organizing economic activity resulting from network effects, economies of scale and economies of scope.13 There are important qualitative differences between economic entities operating in the traditional and digital sector of the economy. Industrial monopolies are hierarchical and take advantage from economies of scale achieved within their own structures, while network effects go beyond organizational structures and refer to the value created and captured outside organizations (Kelly, 1998, p. 28). As mentioned earlier, hierarchies operate under scarce information (rank can be perceived as a substitute for information), while networks are associated with abundant information which gives agents similar power and opportunities (Kelly, 1998, p. 119). This opposition between hierarchies and networks may imply that networks promote greater equality of opportunities and incomes. In fact, networks have highly skewed distributions in which some nodes become privileged hubs, potentially monopolizing the network. In other words, networks can be a way to achieve a new, but not necessarily a more equal income distribution. Indeed, the digital economy favours dominant agents in terms of collecting data and information. This creates distortions on many markets, because of increased opportunities of price discrimination, stable market position of incumbents, and biased incentives to innovate due to informational rents (Stiglitz, 2017).14 Already Galbraith (1998) argued that increasing income inequality is not because of the skill-biased technical change, but due to innovation aimed at increasing market power and securing rents.15 As far as the analysis of the role of platforms in shaping the income distribution is concerned, information economics finds support in many strands of economics, such as insti-

92  The Elgar companion to information economics tutional economics or critical political economy.16 Institutional analysis acknowledges that platforms organize content and influence the information environment of individual agents, shaping reality through non-transparent processes of data collection and processing. Critical political economy in turn emphasizes the exploitative character of digital platforms based on the asymmetric distribution of power exerted through advertiser-based business models, abusive practices on the labour market, callous algorithms and subordination tactics (Mansell & Steinmueller, 2020, pp. 49–51). In terms of incomes, digital platforms generate (literally!) power laws. These power-law relationships are also associated with the tendency of globalization and digitalization to promote the most productive firms in various industries which induces increasing market concentration until industries become dominated by “superstar companies” with high markups and a low labour share (Autor et al., 2020). Digital platforms are often perceived as the fulfilment of unprecedented freedom of access to information, but they actually try to achieve a balance between scarce and abundant information. Platforms are the markets that artificially create scarcity of information using instruments allowing to control data and information (Mansell & Steinmueller, 2020, p. 24). Thus, the economic value of a platform can be typically attributed to rents. Traditionally, monopoly rents were perceived as unearned transfers from consumers and were obtained through the price mechanism thanks to barriers of entry. New digital monopolies are different in the sense that they receive rents transferred both from consumers (e.g., via subscriptions) and other companies e.g., via revenues from advertising. Advertising provides information about potential market transactions and influences the preferences of consumers via both analogue and digital channels. Digital channels offer significantly lower costs of advertising and allow for personalization or targeting that requires compromising traditional public/private perspectives. Interpretation of individual preferences by machine learning systems (algorithms) is based on increasingly precise computations of users’ characteristics such as demographics, interests or mobility which enables discrimination on a large scale (Mansell & Steinmueller, 2020, p. 10, cf. Bodoff, 2024).17 Supported by new technologies (including artificial intelligence), digital platforms in fact conduct endless experiments (resembling thorough medical examinations) to discover and shape user preferences, reorientate user attention and capture consumers’ surplus. Essentially, algorithms make it easier to discriminate against low-income households which may not only increase the costs of living for low-income groups (Bauer, 2018) due to their low elasticity of demand, but also impinge on income distribution as it distorts their decisions concerning consumption and saving. 4.2

Income Distribution: New Opportunities and Threats

Under digitalization changes in income distribution reflect deep structural changes on the labour market. For instance, digitalization stimulated the expansion of the service sector at the cost of manufacturing in many countries. One of the most frequently mentioned suggestions for displaced workers was to acquire digital skills and engage in informational work (e.g., data processing, advertising and other business services). However, this recommendation was based on the assumption of global labour immobility, while due to digitalization informational tasks have become increasingly mobile globally. It turned out that informational labour is even more susceptible to task migration than manufacturing work, because it does not require

Information and income distribution  93 large investments in physical capital or access to cheap transportation (May, 2000).18 This task migration of informational work, supported by the rise of digital platforms, has changed the mechanics of income distribution on the global scale. On one hand, it has stimulated employment in developing countries, while on the other hand it has stretched the income distribution in most developed countries (May, 2000). The COVID-19 pandemic accelerated these developments. Indeed, the impact of digital technologies on income distribution can be inequality-increasing or decreasing, moderated by institutional, as well as economic, social and political forces which also shape the flows of information. According to Bauer (2018), the digital revolution affected income distribution in several (often contradictory) ways. In particular, it has operated through: ● changes in the relative demand for capital, labour and their remuneration; ● demonetization of activities earlier generating income; ● inclusion of individuals previously excluded from the labour market (generating new sources of income or just reducing its volatility); ● rents; ● changes in wealth distribution (often in a winner-takes-all dynamics). Importantly, under heterogeneity of education and training, increasing digital connectivity is likely to be positively correlated with income inequality (Bauer, 2018, cf. Mincer, 1958). Digital platforms reinforce social inequalities. They provide new employment opportunities, but also reduce employment security and lead to precarious employment (Mansell & Steinmueller, 2020, p. 11). Their disruptive effects on labour relations are particularly visible in case of workers in the service sector in general and platform workers in particular who are not as protected as industrial workers. Lower level of unionization puts them in an inferior position compared to more traditional sectors. Besides, there is a greater risk of alienation of online workers due to their spatial dispersion. This dispersion can lead to further erosion of their bargaining power and remuneration, even though digitalization significantly decreases the costs of organization. Platformization of work often leads to its degradation. Remuneration of platform workers (treated not as employees but independent contractors) in many countries is lower than the minimum wage (Ostoj, 2020, p. 67). Frequently, the benefits from technological progress do not accrue to workers. Platform work resembles the early industrial model with its pressure on productivity, but without its employment stability and security. This induces questions about the future of work as such (Ostoj, 2020, p. 75), echoing thoughts about transforming the service sector into the “servant class” (Galbraith & Hale, 2004). This “servant class” can be potentially extended to the majority of users who produce content without any remuneration.19 User-produced content raises many controversies around its commodification (transforming content into an object of exchange) and monetization (converting content into a source of income for the owner of rights to this content). Exploitation of users occurs through appropriation both of user-produced content and data resulting from monitoring user behaviour. Often “voluntary” contributions of platform users are in fact unpaid labour, with all its distributive consequences (Mansell & Steinmueller, 2020, p. 32). Commodification of user-produced content reminds of the commodity fiction discussed by Polanyi (2001 [1944]) with regards to labour, land, and money that can be extended to data. According to Polanyi, labour, land, and money are organized in markets, even though they are

94  The Elgar companion to information economics not commodities (they are not produced for sale). However, the commodity fiction is a fundamental principle organizing the life of the whole society. Labour, land, and money are being bought and sold on the market, while their demand and supply are expressed as real magnitudes. Polanyi also admitted that allowing the market to be the only mechanism responsible for the fate of human beings, their natural environment, and even their use of purchasing power, would have a devastating impact on the society (Polanyi, 2001 [1944], pp. 75–76). Similarly, data are not produced for sale,20 but once they are generated, they can be monetized and sold with all the risks associated not only with adverse effects for income distribution, but for the society as a whole. Another tendency amplified by digitalization is associated with companies communicating the availability of job offers not publicly, but through the networks of professional connections of their employees (cf. Ostoj, 2015). Networks of social connections constitute a part of the endowment of agents in the economy. Both job applicants and employers achieve informational returns from referrals within their network of connections and in general such practices often yield better matches than other recruiting methods. Therefore, persistent income differences across groups of individuals characterized by identical abilities and training can be attributed to differences in their network composition (Calvó-Armengol, 2006). Employment status of individuals is determined by their past employment and the network of connections (it follows a finite state Markov process with transition probabilities dependent on the network of relationships). In the short run, the correlation between the employment status of interconnected individuals is negative due to competition for jobs and information between them; however, in the long run this correlation is positive. This induces clustering of agents according to their employment status (Calvó-Armengol & Jackson, 2004). Therefore, education does not only increase productivity and income due to years of training (human capital), but also enables establishing network capital. Network capital is yet another factor responsible for increasing skewness of income distribution. To recapitulate, growing abundance of information in the digital economy could have brought the world closer to the transparency of the neoclassical theory. However, competition did not become more intense. A new type of monopoly emerged giving rise to new sources of rents and inequalities via artificially created scarcity. Both scarcity and abundance of information have distributive effects. Scarcity and income inequality often coincide, because scarcity precludes an equal and universal access to resources and benefits from their use. However, inequality may persist under conditions of abundance due to political, structural and even social and psychological factors (cf. Plamondon, 2022). In other words, abundance creates the potential for, but does not guarantee an equal and universal access. In particular, demonetization (observed under abundance) decreases the income of some agents, while it is a new source of rents for others. Networks create new opportunities, but their structure is important for income distribution: distributed networks offer more equal opportunities, while centralized and decentralized ones promote polarization of incomes and a greater positive skewness of income distribution. Thanks to their position in a hierarchical network, “superstar” individuals are likely to receive higher income stemming from a greater variety of sources, while the majority of agents depend usually on one main source of income. Finally, the commodity fiction with regard to data is a source of pressures on income distribution and social cohesion. In fact, users are treated like land or natural resources and

Information and income distribution  95 constitute a source of rents (understood as unearned income). They are fertilized (sometimes literally – with ideas) and harvested. Thus, digitalization allowed the biggest digital platforms to shift from Schumpeterian rents to Ricardian rents. This means that they are capable of exploiting informational advantage without the risk of losing sources of income due to diffusion of information.

5.

SELECTED BEHAVIOURAL IMPLICATIONS OF INFORMATION FLOWS FOR INCOME DISTRIBUTION

Information influences behaviour and behaviour conveys information. Mainstream economics assumes that individual tastes are independent. Information economics allows to ascribe greater importance to interdependence of preferences and strategic complementarity of individual behaviour. When making decisions, individuals follow their own preferences and information about others and their behaviour. Besides, they also take into account how their behaviour influences others’ beliefs about them. This is one of the reasons why individuals have an incentive to manipulate information, e.g., to pretend that they are more productive or wealthier than they are. Agents typically like to know more about others, but to hide information about themselves. These processes have important distributive consequences. This section links information and distribution of individual incomes with individual reactions, feelings and behaviour. First, relative deprivation describes dissatisfaction of an individual receiving information that other individuals earn more. Then, the focus is on work as a source of income and satisfaction that an individual would like to signal to other individuals. Finally, the concept of playbour exemplifies a particular case of the recent tendency associated with blurring of work and leisure which complicates considerations on individual income and satisfaction in the presence of imperfect information. 5.1

Information and Relative Deprivation

As already mentioned, the gains of more productive and more convincing (or more able in transmitting information and exerting influence) individuals are translated into losses of others. It follows that achieving greater economic efficiency associated with reallocation of resources towards more productive agents is associated with increasing social dissatisfaction due to lower incomes in some segments of the labour market and potentially increasing awareness about existing income inequalities. In line with the prospect theory, losses are felt more intensively than gains (Kahneman & Tversky, 1979). Thus, for society as a whole, the net welfare effect is likely be negative. The efficiency-social satisfaction trade-off can lead to greater polarization if the more productive individuals (or more influential, able to capture rents) become more motivated to continue engaging in their activities, while the less productive (or less convincing) lose their motivation and decrease their efforts. Following this thread, the difference between private and social returns to information can be linked with a concept that is not discussed frequently in the information economics literature, but captures this dark side of information flows, namely relative deprivation. According to Runciman (1966, p. 10), individuals are relatively deprived if they: don’t possess X, see others possessing X (individual perceptions do not have to conform to reality), want to possess X and think that it is feasible to possess X. The concept of relative deprivation

96  The Elgar companion to information economics is quite capacious, as the object X can refer to material, immaterial and even social spheres. In the context of income distribution, relative deprivation can be interpreted as a measure of stress of an individual resulting from interpersonal income comparisons and calculated as the sum of the extra income units that others in the population have, normalized by the size of the population in which comparisons take place (Stark & Włodarczyk, 2015).21 Two interesting phenomena are associated with imperfect information and relative deprivation. On one hand, information imperfections reduce the scale of conducted comparisons (screening of the surroundings). This means that if individuals compare their incomes with others from their vicinity that are often similar to them (within the network of colleagues not only employment status, but also incomes are likely to be positively correlated), their relative deprivation can be potentially lower than under large-scale comparisons. On the other hand, due to the tendency to overestimate unknown values people are likely to overestimate incomes of others.22 This may induce a more intense feeling of relative deprivation compared to the transparent situation and exert a greater behavioural impact than in a situation with perfect or no informational transparency. In other words, under imperfect information due to potentially biased reference levels, individuals are likely to overreact when receiving information about others. This has potentially destabilizing impact on the society. 5.2 Work, Income and Satisfaction Interdependence of preferences and relative deprivation shed new light on various functions that work plays in a society in the context of information flows and their distributive effects. In an unequal society, the realization of the function of work as a means of achieving fulfilment is more complicated than in a relatively equal one. Inequalities exert pressure on households towards greater consumption to maintain their social status, and thus work becomes a means to achieve a higher level of consumption. As a result, when there is a trade-off between work quality and income, inequality and relative deprivation encourage workers to pursue the latter (Wisman, 2022, p. 408). In line with the law improperly called Gresham’s law (cf. e.g., Balch, 1908), income-generating work crowds out satisfaction-generating work. Thus, communication (verbal or non-verbal) concerning the status can increase frustration despite achieving a higher standard of living. The problem can be also presented from a different angle. An individual deciding on a job considers not only his potential income and satisfaction, but also the opinions of others to whom he would like to signal his material status and satisfaction. With small reference groups (e.g., family members) it is relatively easy to communicate both. However, under imperfect information and larger reference groups it might be much easier to signal material status than satisfaction. As a result, income-oriented individuals put pressure on satisfaction-oriented ones making them potentially deprived twice: first, in terms of income (income-oriented individuals enjoy first-mover advantage – they were earlier motivated to look for promising jobs that may no longer be available on the market), and second, in terms of satisfaction (due to lack of recognition). Digitalization potentially amplifies these phenomena.23 As it enables contacts with a greater number of individuals, but shallower compared to traditional social relations, it also promotes income-oriented individuals at the cost of satisfaction-oriented ones. Importantly, as income can consist of not only wages, but also rents, digitalization is likely to promote rent seeking as a way of securing higher total income from various sources.

Information and income distribution  97 Moreover, information about income distribution has an impact on the behaviour of people, shaping the income distribution in the future and future information flows. Information about income distribution can spread through mass media which for decades have provided information and influential comments on inequality levels shaping redistribution preferences (cf. Grisold & Preston, 2020), as well as rent-seeking orientation of individuals. Under the domination of mass media, the whole population was exposed roughly to the same set of information. Social media in turn offer personalized newsfeed with algorithms supporting the prevailing opinions of users. Emerging social bubbles make it difficult for the whole community to communicate, creating cultural divisions that resemble biblical babelization (cf. Włodarczyk, 2020, p. 20)24 or balkanization. In the long run, social divisions may impact employment decisions and wage negotiations, especially if the role of the network capital is weighed in. Obviously, there is also a positive aspect of such divisions. As already mentioned, within homogeneous groups relative deprivation experienced by its members can be lower. 5.3

Dialectics of Playbour

Currently, the discourse becomes more complicated due to an ever-increasing blurring of work and leisure, as well as processes of monetization and demonetization,25 associated with creating rents (as a result of paid unproductive activities) and unpaid labour.26 One of the manifestations of these tendencies is playbour. Playbour can be interpreted as the activity in which consumption (leisure, play) is transformed into production (work, labour). In other words, it consists in monetization of leisure (e.g., due to exclusion from other segments of the labour market, job rationing, etc.).27 Manifestations of playbour evolved over the last decades from the monetization of symbols in an analogue form (such as hip-hop songs, cf. Kelley, 1997) to the explosion of digital forms, including computer games modifications (Kücklich, 2005) or “play-to-earn” games (e.g., Vidal-Tomás, 2022). Interestingly, Kelley (1997) discusses playbour (originally “playlabour”) in the context of self-commodification which in the digital world leads to double commodification – of the individuals and the data they generate. Playbour raises several controversies associated with measuring the duration of work and productivity, worker’s freedom, exploitation, job and life satisfaction, work-life balance, etc. The picture is further complicated when interdependence of preferences is considered. For instance, collating the concept of playbour, the theory of signalling and Veblenian conspicuous consumption28 provokes questions whether, on some occasions, agents engaged in playbour pretend to work when having fun or pretend to have fun at work. Agents deciding to engage in playbour may be those caught in the trap between expectations of the employers and peers, balancing on the utility curve to choose an optimal combination of income and satisfaction in front of others.29 To recapitulate, flows of information have direct and indirect effects on the income distribution. Information flows arise because of inequalities and create them. In general, information flows are propelled by differences between agents including differences in income which can have various consequences depending not only on individual characteristics, but also on informational transparency. Constraining informational transparency can induce more pronounced behavioural reactions of individuals compared to the situation of perfect or no transparency at all. If different segments of the income distribution have their own dynamics, imperfect information (even relatively abundant) can accelerate the polarization of incomes.

98  The Elgar companion to information economics Greater abundance of information associated with digitalization significantly modifies the mechanisms of signalling and screening. For instance, existing inequalities give rise to information flows in the form of conspicuous consumption, transmitted in person or via social media.30 Simultaneously, these information flows increase inequalities inasmuch as they create disruptions in the consumption-saving decisions, leading to increases in aggregate demand and prices. Poorer households are more affected by inflationary processes (even in the presence of goods offered for free). Proliferation of new consumptions norms can decrease savings and contribute to lower wealth accumulation with long run repercussions for the income distribution. Naturally, there are many flows of information decreasing inequalities (e.g., about job opportunities or training), but these are potentially constrained by the scale of the network. Finally, digitalization is likely to promote income-oriented individuals at the cost of satisfaction-oriented ones.

6.

CONCLUDING REMARKS

This chapter has shown that information economics is very relevant for investigations of the origin, evolution and persistence of income inequality, although such references are not common in the literature. In particular, information flows (realized via education, training, professional contacts, mass media and social media, advertising, etc.) can increase or decrease income inequalities in various segments of income distribution, producing an excessive positive skewness of income distribution on the global, macro and micro levels (even within a household). Early literature has shown that differences in income are a product of individual abilities and training, as well as bequests. Information economics adds new multipliers to this equation. Access to information increases average income and its variance. Exact effects on parameters of income distribution are yet to be discussed. Greater skewness (but not necessarily a different type of distribution) arises when information imperfections are combined with other spheres. When information becomes abundant it creates new opportunities and multiplies existing ones; however, exploitation of these new opportunities will not be the same for all the agents from the income distribution (depending e.g., on access to internet, digital skills of individuals, size and composition of their private and professional networks).31 Therefore, modern changes in the income distribution are de facto a product of prevailing inequality coupled with digital inequality, while new technological advances such as artificial intelligence are likely to perpetuate or even augment existing inequalities. Altogether, they impinge on individual satisfaction and behaviour. The chapter emphasized the role of property rights and rents. In the world of perfect information there is no informational advantage to be monetized. Any innovation is imitated, so there are no important incentives for innovation. Securing property rights is a step towards growth, but also inequalities. Interestingly, information economics provides yet another channel linking economic growth and income inequality. Economic growth and development offer a greater freedom of choice (cf. Sen, 1999), while greater freedom of choice is associated with greater uncertainty (cf. Machlup, 1962, p. 8). Greater uncertainty in turn translates into greater information imperfections and potentially greater income inequality. There is a risk that increased economic inequality (translated into political inequality) will promote institutional changes enhancing market power and increasing inequality further,

Information and income distribution  99 simultaneously weakening economic performance (Stiglitz, 2017). Market imperfections associated with problems of transparency and privacy, as well as ownership rights of information and data are among the issues justifying the intervention of the government, even though this chapter did not discuss these issues. In the future, greater availability of microdata will make it feasible to analyse information flows, network connections and income generation in detail in a dynamic framework. For instance, there is a need for in-depth research demonstrating to what extent greater information transparency achieved due to digitalization increases the size of reference groups, intensity and scale of information flows within social networks and their impact on individual motivation. In general, information economics should put more weight on the transmission of information within and between networks (and their distributive effects), as well as the analysis of the quality of information within and beyond the context of its scarcity and abundance. Such analyses could for instance assess the distributive effects of fake news. Even circulation of “useless” information has potentially distributive effects as its consumption requires the attention of users and reallocates their time towards unproductive purposes. Thus, distributional issues associated with circulation of scarce and abundant information are likely to inspire future theoretical discussions and empirical investigations.

NOTES 1. 2. 3.

4.

5.

6. 7.

Galton (1892), following Quetelet, applied the law of “deviation from an average” in many areas, including results of examinations of college candidates, which let him conclude that natural gifts are characterized by a normal distribution. In line with the example provided by Mincer (1958), relative income differences between individuals with 10 and 8 years of training are significantly larger than the relative differences between individuals that completed 4 and 2 years of training, respectively. This chapter does not discuss mathematical forms of distributions with the best fit to the real-world income data. Distributions discussed were chosen both to reflect theoretical considerations and to fit real-world data at least in a satisfactory way. In fact, both the log-normal and the power-law distribution can be treated as a result of special conditions imposed on transition possibilities within a more general Markov chain approach. Importantly, the stability of income distributions which results from certain income transition probabilities raises the question about the effectiveness of income redistribution, which may require not simple transfers, but changes in transition probabilities (Bronfenbrenner, 2007, pp. 56–57). Such changes would reflect institutional change. Interestingly, maximum entropy methods from statistical mechanics and information theory lead to log-normal distribution in equilibrium (e.g., Venkatasubramanian et al., 2015), which means that an egalitarian income distribution, even though popular in the literature, ceases to be a natural benchmark. The opposition between the natural and artificial character of distributions under discussion to some extent can be interpreted in terms of natural processes: log-normal distribution corresponds to diffusion of information and increase in entropy, while power-law distribution corresponds to emergence of innovation and anti-entropic processes characteristic for open systems. Even wages are in fact composed from earnings corresponding to the market-clearing wage and rents (cf. note 11). Knight (1921) linked profits and losses with risk and uncertainty resulting from imperfect knowledge that occurs due to changes in market conditions. His argument is elicited from the observation that competition is based on anticipations. Under perfect information all market developments can be foreseen in an indefinite time horizon and thus no profit (or loss) arises. In turn, not anticipated changes in market conditions produce a divergence between costs and prices, which cannot be equalized by competition (Knight, 1921, pp. 197–198), leaving some agents better off or worse off.

100  The Elgar companion to information economics 8. 9.

10.

11.

12.

13.

14.

15. 16.

17.

18.

See also the discussion on dis- and misinformation in Chapter 3 of the Companion (Stiglitz & Kosenko, 2024b, pp. 53–80). Uncertainty has important implications for the labour market. For instance, imperfect information practices bearing the hallmark of racial or sexual discrimination can be actually related to statistical inference based on the knowledge of average characteristics differentiating various groups (e.g., quality of schooling) without information about the true distribution of individual characteristics (cf. Akerlof, 1970). Importantly, according to Spence (1973) race and sex are generally not alterable attributes, so they should not be treated as signals, but as indices. Nevertheless, potential employers take into account both: indices and signals. As a result, the Phillips curve seems to be stretched at both ends simultaneously – one segment of the labour market exerts an upward pressure on wages, while the other experiences a greater risk of unemployment. Importantly, both segments may engage in rent seeking: the first segment will try to capture rent of abilities, while the unemployed can also engage in rent-seeking activities associated with political mechanisms (trying to secure transfers). Job-rationing theory is one of the theories opposing the neoclassical tradition assuming market clearing. Importantly, under job rationing even the low-quality jobs are remunerated with wages in excess of market clearing (this is consistent e.g., with the efficiency wage theory). Thus, workers receive rents. This explains persistent wage dispersion in case of workers of identical characteristics (cf. Akerlof et al., 1988). In this situation the most productive individuals would prove their abilities (because of the strongest incentive to do so), leaving the group of less able with lower wages. Then the most productive individuals from this group would like to differentiate themselves as being characterized by productivity above average, again leaving the group of less able with lower wages. This process would continue until full revelation (Stiglitz, 2002). Network effects are associated with a nonlinear, often exponential increase in value of the network relative to the number of users, economies of scale with a decrease in unit costs due to a larger scale of output, while economies of scope with cost efficiencies stemming from offering additional products or services (e.g., bundling) compared to offering them separately (cf. Mansell & Steinmueller, 2020, p. 39). In Chapter 2 of this Companion, Stiglitz & Kosenko (2024a, pp. 20–52) discuss these negative aspects of abundance of information (linked to the price discrimination and extraction of consumer surplus), using the term “adverse distribution” to depict money flows from poorer consumers to the richer owners of firms. See also the discussion concerning the link between inequalities, redistribution to the top, rent seeking and deregulation of markets by Elsner (2024). Neoclassical analysis does not focus on power asymmetries in the digital world, because of the assumption of competitive markets and treats the power of the platforms as an opportunity to serve consumers better (Mansell & Steinmueller, 2020, p. 51) and therefore is not a natural point of reference in this regard. In particular, digital tracking and automated decision-making in social welfare provision often reflect prevailing gender, racial and economic discrimination giving birth to automated inequality. When algorithms (sometimes significantly flawed) make decisions about whom to support, the public is likely to lose from sight the real problem of poverty. Thus, repercussions of automated decision-making with reference to the poorest are likely to be further reaching than previous non-digital mechanisms (cf. Eubanks, 2018). An important step to make the transfer of informational work and data processing less risky was the adoption of the Trade Related Intellectual Property Rights agreement (TRIPs) in 1995 that harmonized the protection of intellectual property globally. Intellectual property rights are intangible assets that do not depreciate over time like tangible assets and offer the opportunity of continuous income (rents), which stimulates rent seeking. Rent seeking consists in the process of assetization, i.e., purposeful transformation of knowledge into an asset allowing for extraction of value from this asset through rents (Birch, 2020). Thus, intellectual property rights institutionalized rent seeking, enabling transformation on informational advantage into pressures on income distribution. A broader discussion on intellectual property rights is included in Chapters 15 and 16 of this Companion (Bochańczyk-Kupka, 2024, pp. 301–314; Gürpınar & Özveren, 2024, pp. 315–337).

Information and income distribution  101 19. Some users (celebrities or influencers) receive a share of the advertising revenue they generate (Mansell & Steinmueller, 2020, p. 38) or even are able to collect income from various sources (Stock, 2024). However, new algorithms and models of content distribution may challenge their position in the future. 20. Datafication is not associated with production of data, but transforming data in a socially useful format (it is a purposeful use of technology to collect data in order to sell them). 21. For example, if the population consists of three individuals receiving income of 100, 100 and 50 dollars, then the first two individuals do not experience relative deprivation (there is no one earning more), while the relative deprivation of the third individual is equal to (50+50)/3 = 33.33 dollars. Similarly, if a society is composed of three strata, receiving on average 20,000, 2,000 and 200 dollars and being composed of 10, 100 and 1,000 individuals, respectively, then the relative deprivation of each individual from the first stratum is nil, of the second 17.80 dollars, while from the third 35.61 dollars. In this case the aggregate relative deprivation (measure of social dissatisfaction) is equal to 357,863.50 dollars which corresponds to 16.2 per cent of total income of this society. 22. They may not see the effort of others and attribute this state of affairs to chance and treat it as unfair. In fact, both under scarcity and abundance individuals are often uncertain whether scarce or abundant resources are fairly distributed. However, abundance produces not only more frequent cases of dishonest behaviour, but also increases psychological stress due to concerns about fairness (Gino & Pierce, 2009). 23. There is a bitter addendum to Ammon’s (1899, p. 229) observations concerning differences in incomes of individuals driven by different motives. Public officials, statesmen or scholars driven by altruistic motives not only earn less than individuals driven by egoistic motives, but in a more transparent world of the digital economy they are more frequently exposed to criticism, hate, loss of privacy etc. In terms of overall life satisfaction, they may be well behind the third group mentioned by Ammon, namely low-income individuals characterized by deficiencies in intellectual, moral, and economic traits. 24. This secondary babelization can be called bubble-ization. 25. Offering goods for free is affecting the income distribution in a longer perspective, especially if it resembles giving fish (consumption good) instead of a fishing rod (production good) and can be a way not only to consumer lock-in, but also to social enslavement due to created market distortions. 26. Naturally, technological progress in general and the digital revolution in particular also create new sources of income (e.g., offer the possibility to trade with cryptocurrencies). However, it is unlikely that losses in one domain would be proportionately compensated in another one. 27. The opposite phenomenon, namely the transformation of production into consumption or demonetization of labour (e.g., due to decreasing costs in the digital economy) is often associated with gamification (cf. Werczyńska & Włodarczyk, 2023). 28. Conspicuous (vicarious) consumption refers to intentional and unproductive (or even wasteful) consumption of goods not associated with satisfaction of basic needs of an individual, but serving the purpose to publicly display individual wealth or status (cf. Veblen, 1899, pp. 68–69). 29. This balancing can be costly. Törhönen et al. (2019) show differences in income between individuals interpreting their activity (content creation in this case) as work, play or work combined with play. Playbourers in their sample were characterized by the lowest income levels. 30. Digitalization enables new manifestations of conspicuous consumption, e.g., via non-fungible tokens. 31. See also the discussion in Chapter 1 of this Companion (Raban & Włodarczyk, 2024, pp. 2–19).

REFERENCES Akerlof, G. A. (1970). The Market for “Lemons”: Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84(3), 488–500. Akerlof, G. A., Rose, A. K., & Yellen, J. L. (1988). Job Switching and Job Satisfaction in the U.S. Labor Market. Brookings Papers on Economic Activity, 1988(2), 495–594. https://​doi​.org/​10​.2307/​2534536.

102  The Elgar companion to information economics Ammon, O. (1895). Die Gesellschaftsordnung und ihre natürlichen Grundlagen. Entwurf einer Sozial-Anthropologie zum Gebrauch für alle Gebildeten, die sich mit sozialen Fragen befassen. Jena: Verlag von Gustav Fischer. Ammon, O. (1899). Some Social Applications of the Doctrine of Probability. Journal of Political Economy, 7(2), 204–237. Arrow, J. K. (1971). Essays in the Theory of Risk-Bearing. Chicago: Markham Pub. Co. Autor, D., Dorn, D., Katz, L. F., Patterson, C., & Reenen, J. V. (2020). The Fall of the Labor Share and the Rise of Superstar Firms. The Quarterly Journal of Economics, 135(2), 645–709. Balch, T. W. (1908). The Law of Oresme, Copernicus and Gresham. Proceedings of the American Philosophical Society, 47(188), 18–29. Bauer, J. M. (2018). The Internet and Income Inequality: Socio-Economic Challenges in a Hyperconnected Society. Telecommunications Policy, 42(4), 333–343. Bester, H. (2004). Specific Skills, Imperfect Information and Job Rationing. In S. Huck (Ed.), Advances in Understanding Strategic Behaviour: Game Theory, Experiments and Bounded Rationality (pp. 73–84). Basingstoke: Palgrave Macmillan. Birch, K. (2020). Technoscience Rent: Toward a Theory of Rentiership for Technoscientific Capitalism. Science, Technology, & Human Values, 45(1), 3–33. Blain, R. R. (2002). The Hour is de facto World Money Unit. In M. Dobija (Ed.), Monetary Unit Stability in Holistic Approach. Warsaw: Leon Koźmiński Academy of Entrepreneurship and Management. Bochańczyk-Kupka, D. (2024). Intangibles, Information Goods, and Intellectual Property Goods in Modern Economics. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 301–314). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Bodoff, D. (2024). On the Status of Machine Learning Inferences in Data Privacy Economics and Regulation. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 462–480). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Bronfenbrenner, M. (2007). Income Distribution Theory. Chicago: Aldine Transactions. Calvó-Armengol, A. (2006). Social Networks and Labour Market Outcomes. Barcelona: Centre de Recerca en Economia Internacional (CREI). Calvó-Armengol, A. & Jackson, M. O. (2004). The Effects of Social Networks on Employment and Inequality. American Economic Review, 94(3), 426–454. Chakraborty, S. & Dabla-Norris E. (2006). Rent seeking. IMF Staff Papers, 53(1). Chalidze, V. (2000). Entropy Demystified: Potential Order, Life and Money. Irvine, CA: Universal Publishers. Elsner, W. (2024). Innovation and Information: Smooth and Ongoing Change, or Turbulence and Cognitive Over-Stress? On the Complex Deep Structure of Innovation. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 270–300). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Eubanks, V. (2018). Automating Inequality: How High-Tech Tools Profile, Police and Punish the Poor. New York: St. Martin’s Press. Friedman, M. (1953). Choice, Chance, and the Personal Distribution of Income. Journal of Political Economy, 61, 277–290. Galbraith, J. K. (1998). The Distribution of Income. UTIP Working Paper No. 2. https://​utip​.gov​.utexas​ .edu/​papers/​utip​_02​.pdf. Galbraith, J. K. & Hale, T. (2004). Income Distribution and the Information Technology Bubble. UTIP Working Paper No. 27. https://​utip​.lbj​.utexas​.edu/​papers/​utip​_27​.pdf. Galton, F. (1892). Hereditary Genius: An Inquiry into its Laws and Consequences. London: Macmillan. Gino, F. & Pierce, L. (2009). The Abundance Effect: Unethical Behavior in the Presence of Wealth. Organizational Behavior and Human Decision Processes, 109(2), 142–155. Goldfarb, A. & Tucker, C. (2019). Digital Economics. Journal of Economic Literature, 57(1), 3–43. Grisold, A. & Preston, P. (Eds.) (2020). Economic Inequality and News Media: Discourse, Power, and Redistribution. New York: Oxford University Press. Grossman, S. J. & Stiglitz, J. E. (1980). On the Impossibility of Informationally Efficient Markets. American Economic Review, 70(3), 393–408. Gürpınar, E. & Özveren, E. (2024). Incomplete Contracts, Intellectual Property Rights, and Incentives: Investment in Knowledge Assets under Alternative Institutional Configurations. In D. R. Raban & J.

Information and income distribution  103 Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 315–337). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Hartle, D. G. (1983). The Theory of “Rent-Seeking”: Some Reflections. Canadian Journal of Economics, 16(4), 539–554. Hayek, F. A. (1945). The Use of Knowledge in Society. American Economic Review, 35(4), 519–530. Kahneman, D. & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), 263–291. Kanbur, S. M. (1979). Of Risk Taking and the Personal Distribution of Income. Journal of Political Economy, 87(4), 769–797. Kelley, R. D. G. (1997). Yo’ Mama’s Disfunktional! Fighting the Culture Wars in Urban America. Boston: Beacon Press. Kelly, K. (1998). New Rules for the New Economy: 10 Radical Strategies for a Connected World. New York: Viking Penguin. Knight, F. H. (1921). Risk, Uncertainty and Profit. Boston: Houghton Mifflin. Kücklich, J. (2005). Precarious Playbour: Modders and the Digital Games Industry. The Fibreculture Journal, No. 5. https://​five​.fibreculturejournal​.org/​fcj​-025​-precarious​-playbour​-modders​-and​-the​ -digital​-games​-industry/​. Machlup, F. (1962). The Production and Distribution of Knowledge in the United States. Princeton, NJ: Princeton University Press. Mansell, R. & Steinmueller, W. E. (2020). Advanced Introduction to Platform Economics. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. May, C. (2000). Information Society, Task Mobility and the End of Work. Futures, 32, 399–416. Milaković, M. (2005). Do We All Face the Same Constraints? In A. Chatterjee, S. Yarlagadda, & B. K. Chakrabarti (Eds.), Econophysics of Wealth Distributions (pp. 184–191). Berlin: Springer Verlag. Miller, H. P. (1955). Elements of Symmetry in the Skewed Income Curve. Journal of the American Statistical Association, 50(269), 55–71. Mincer, J. (1958). Investment in Human Capital and Personal Income Distribution. Journal of Political Economy, 66(4), 281–302. Ostoj, I. (2015). Pracodawcy w obliczu niepełnej informacji na rynku pracy – studium teoretyczne i empiryczne. Zeszyty Naukowe Uniwersytetu Szczecińskiego. Studia i Prace Wydziału Nauk Ekonomicznych i Zarządzania, 40(1), 47–56. Ostoj, I. (2020). Praca w czasach cyfrowych platform technologicznych: w sieci gig economy. Katowice: Uniwersytet Ekonomiczny w Katowicach. Pareto, V. (1896). La courbe de la répartition de la richesse. Lausanne: Université de Lausanne. Pasinetti, L. L. (2000). Critique of the Neoclassical Theory of Growth and Distribution. PSL Quarterly Review, 53(215), 383–431. Pigou, A. C. (1932). Economics of Welfare. London: Macmillan and Co. Plamondon, S. (2022). Inequality in Abundance. Frontiers in Research Metrics and Analytics, 7, 980677. https://​doi​.org/​10​.3389/​frma​.2022​.980677. Polanyi, K. (2001 [1944]). The Great Transformation: The Political and Economic Origins of Our Time. Boston, MA: Beacon Press. Porat, M. U. (1977). The Information Economy: Definition and Measurement. Office of Telecommunications (DOC), Washington, DC, OT-SP-77–12(1). Raban, D. R. & Włodarczyk, J. (2024). Information Economics Examined through Scarcity and Abundance. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 2–19). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Risau Gusman, S., Laguna, M. F., & Iglesias, J. R. (2005). Wealth Distribution in a Network with Correlations Between Links and Success. In A. Chatterjee, S. Yarlagadda, & B. K. Chakrabarti (Eds.), Econophysics of Wealth Distributions (pp. 149–158). Berlin: Springer Verlag. Runciman, W. G. (1966). Relative Deprivation and Social Justice: A Study of Attitudes to Social Inequality in Twentieth-Century England. Berkeley: University of California Press. Sen, A. (1999). Development as Freedom. New York: Alfred Knopf. Sinha, S. (2005). The Rich Are Different! Pareto Law from Asymmetric Interactions in Asset Exchange Models. In A. Chatterjee, S. Yarlagadda, & B. K. Chakrabarti (Eds.), Econophysics of Wealth Distributions (pp. 177–183). Berlin: Springer Verlag.

104  The Elgar companion to information economics Spence, M. (1973). Job Market Signaling. The Quarterly Journal of Economics, 87(3), 355–374. Stark, O. & Włodarczyk, J. (2015). European Monetary Integration and Aggregate Relative Deprivation: The Dull Side of the Shiny Euro. Economics and Politics, 27(2), 185–203. Stiglitz, J. E. (2002). Information and the Change in the Paradigm in Economics. American Economic Review, 92(3), 460–501. Stiglitz, J. E. (2017). The Revolution of Information Economics: The Past and the Future. NBER Working Paper Series, no. 23780. https://​doi​.org/​10​.3386/​w23780. Stiglitz, J. E. & Kosenko, A. (2024a). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 1: Indirect Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 20–52). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. & Kosenko, A. (2024b). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 2: Direct Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 53–80). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stock, W. G. (2024). Payment on Information Markets. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 339–363). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Tollison, R. D. (1997). Rent Seeking. In D. Mueller (Ed.), Perspectives on Public Choice (pp. 506–525). Cambridge: Cambridge University Press. Törhönen, M., Hassan, L., Sjöblom, M., & Hamari, J. (2019). Play, Playbour or Labour? The Relationships between Perception of Occupational Activity and Outcomes among Streamers and YouTubers. Proceedings of the 52nd Hawaii International Conference on System Sciences. http://​hdl​ .handle​.net/​10125/​59694. Tullock, G. (1988). The Costs of Rent Seeking: A Metaphysical Problem. Public Choice, 57(1), 15–24. Veblen, T. (1899). The Theory of the Leisure Class: An Economic Study in the Evolution of Institutions. New York and London: Macmillan. Venkatasubramanian, V., Luo, Y., & Sethuraman, J. (2015). How Much Inequality in Income Is Fair? A Microeconomic Game Theoretic Perspective. Physica A: Statistical Mechanics and its Applications, 435, 120–138. Vidal-Tomás, D. (2022). The New Crypto Niche: NFTs, Play-to-Earn, and Metaverse Tokens. MPRA Paper No. 111351. https://​mpra​.ub​.uni​-muenchen​.de/​111351/​. Werczyńska, D. & Włodarczyk, J. (2023). Between Employment and Nonemployment: The Ambiguity of Work and Leisure in the Contemporary Labor Market, Forum for Social Economics, 52(4), 416–432. Williamson, O. E. (1981). The Modern Corporation: Origins, Evolution, Attributes. Journal of Economic Literature, 19(4), 1537–1568. Wisman, J. (2022). The Origins and Dynamics of Inequality: Sex, Politics, and Ideology. New York: Oxford University Press. Włodarczyk, J. (2013a). Nierówności dochodowe w Polsce według rozkładów Pareto i Boltzmanna-Gibbsa. Studia Ekonomiczne, 130, 76–87. Włodarczyk, J. (2013b). Nierówności dochodowe, pogoń za rentą i dyscyplina fiskalna – konceptualizacja powiązań. Zeszyty Naukowe no. 27, Wyższa Szkoła Handlu i Usług w Poznaniu 2013, 179–189. Włodarczyk, J. (2020). Non solum technology – korzenie i perspektywy rozwojowe gospodarki cyfrowej. In I. Ostoj & K. Bartuś (Eds.), Innowacje na poziomie mikro- i makroekonomicznym (pp. 11–23). Katowice: Uniwersytet Ekonomiczny w Katowicach. Yarlagadda, S. & Das A. (2005). A Stochastic Trading Model of Wealth Distribution. In A. Chatterjee, S. Yarlagadda, & B. K. Chakrabarti (Eds.), Econophysics of Wealth Distributions (pp. 137–148). Berlin: Springer Verlag. Young, A. A. (1917). Do the Statistics of the Concentration of Wealth in the United States Mean What They Are Commonly Assumed to Mean? American Economic Review, 7(1), 144–156.

PART II INFORMATION ASYMMETRY

5. Asymmetric information as a market failure in retrospect Wojciech Giza

1. INTRODUCTION When in the 1920s the Austrian economists discerned the problems related to knowledge and information processes, hardly anyone realized that their findings would contribute to the development of the so-called information society. Technological changes connected with information technology in the post-war period made by knowledge and information drew the attention of economists. It was observed that the market is not only a mechanism driven by demand and supply. The market, as F. A. Hayek emphasized, is also a telecommunication mechanism. It should be noted, though, that in the approach proposed by Hayek, the information issue was not treated as the cause of market failure, but rather as a condition for market opportunities and innovativeness. The aim of the chapter is to reconstruct the dispute concerning market failure on the basis of the subject literature and to demonstrate the relationship between market failure and asymmetry of information. When identifying market failure I apply the criterion of market economy efficiency developed on the basis of neoclassical economics. The differentiation between the old and the new approach to market failure was based on asymmetry of information, as proposed by Jospeh Stiglitz. The chapter is divided into three sections. The first one offers an interpretation of such issues as rationality, knowledge and information. It shows the way in which these issues have been incorporated into economic discourse. In the second part, I distinguish market failure on the basis of efficiency criteria formulated in general equilibrium. The third part focuses on the implementation of asymmetric information understood as a new kind of market failure.1

2.

RATIONALITY, KNOWLEDGE AND INFORMATION AS THE SUBJECT OF ECONOMIC ANALYSIS

Asymmetric information is one type of market failure. Therefore we must ask about the way in which knowledge and information are perceived in a broader context. The development of economics as an independent branch of science was started by Adam Smith in the second half of the eighteenth century. Smith was a moral philosopher and a prominent representative of the Scottish Enlightenment. At that time the British Isles experienced profound transformations which led to the appearance of a new form of social relations. In the Middle Ages as well as in the era of economic mercantilism societies were bonded with custom and tradition ties. They were guarded by the state supported by the authority of the church. Novel ideas presented by Bernard Mandeville in The Fable of the Bees (1970 [1714]) and developed by Adam Smith in An Inquiry into the Nature and Causes of the Wealth of Nations (1976 [1776]) put the market, 106

Asymmetric information as a market failure in retrospect  107 understood as a mechanism determining social relations, at the center of research. The state authority was replaced with the faith in Enlightenment reason. The dominant mercantilism conviction that an individual must subordinate to the goals defined by the state was replaced with postulates of social and economic freedom. The founders of classical political economics came to the conclusion that laissez faire is the best kind of economic policy. The social and economic system based on individual freedom and respect for private property ensures the welfare of the whole society. This was possible thanks to giving individuals a broadly understood area in which they could act and believing that they are driven by rational reasons and knowledge. In this way such categories as rationality, knowledge, and then information, became important to economists. The rationality of individuals is one of the key assumptions of neoclassical economics (Colander, 2000, p. 134). However, its interpretation was different for representatives of classical political economics – David Hume, Adam Smith, David Ricardo – and different for neoclassical economists. The former interpreted rationality in the spirit of Enlightenment philosophy. The essence of this interpretation consists in a conviction that “development of the history of civilization has always been linked to the idea of the overcoming of dogmatic traditions by means of rational insight” (Honneth, 1987, p. 692). In the economic dimension, overcoming the dogmatically understood tradition consisted in limiting the role of the state and negating state interventionism by widening the sphere of people’s individual liberties. In neoclassical economics the concept of rationality was significantly narrowed. This was due to the development of quantitative methods, allowing researchers to precisely examine the decision process according to unambiguously defined criteria of efficiency. The direction of research adopted by neoclassicists resulted from the methodological approach modeled on the development of science. In 1900, at the 2nd Congress of Mathematicians in Paris, David Hilbert formulated 23 problems. The sixth one referred to the axiomatic treatment of physics (Grattan-Guinness, 2000). Thirteen years later, a logician from Cambridge, William Ernest Johnson used formal analysis to derive a negatively sloped indifference curve (Johnson, 1913).2 Currently, indifference curves constitute a mathematical reflection of consumer preferences, which combined with budget constraint and the rule of optimization, constitute a formal choice scheme. In this way, neoclassical economics gave new meaning to Enlightenment rationalism. The essence of economic rationality is expressed by rational behavior axioms which justify the shape of the indifference curve. These are: completeness (individuals can compare all the available alternatives), reflexivity (any bundle is certainly at least as good as an identical bundle), transitivity (if X is preferred to Y and Y is preferred to Z, X is preferred to Z), and monotonicity (a rational agent prefers more than less). The adoption of these axioms constitutes justification for the indifference curve reflecting so-called well-behaved preferences (Varian, 2010, pp. 44–48). The first of the above-listed axioms does not seem controversial. However, it may be interpreted as a statement referring to the scope of knowledge possessed by market game players. This axiom implies full and far-reaching rationality of homo oeconomicus. This view is rightly criticized by supporters of heterodoxy, who draw extensively on the achievements of behavioral economics, anthropology and various trends of social philosophy. In the field of mainstream economics, one of the most serious criticisms of rationality understood in this way was proposed by Herbert Simon. He presented the theory of bounded rationality, stemming from the analysis of decision processes in public administration. In his work Administrative

108  The Elgar companion to information economics Behavior: A Study of Decision-Making Processes in Administrative Organization (Simon, 1947) he presented an interpretation of rationality in the context of management effectiveness. Simon rejected the assumption of full rationality (Simon, 1955, 1956, 1978). His skepticism resulted both from the limitation of the set of information an individual possesses as well as from the possibility of processing such information. Simon also emphasized that people do not aim at achieving an optimal solution, and are happy with a satisfactory one. Pointing at differences between rationality and knowledge in economics we should emphasize that rationality is a broadly understood meta-category dating back to the Enlightenment belief in reason. Through axioms of rational behavior it organizes the structure of our preferences. It constitutes a normative behavior model. Knowledge is something that allows entities operating in the economic sphere to make optimal decisions. Information and data constitute a narrower category than knowledge. Rational knowledge of the world is based on true information, which should be based on data.3 How should we then understand data and information? Let us use the following example to explain this. Let us imagine some numerical data, for example: 20, 40, 80. What can we derive from them? Nothing, unless we learn what they mean. Data become information when we specify their meaning. For example, if the above-quoted data represent the number of points obtained by students in a mathematics test, then knowing the point value needed to pass the test, we know how well students mastered the material covered by the mathematics course. We know whether the education process is effective. Therefore we can present the following hierarchy of analyzed categories. Data constitute the foundation. Information relies on data, as we add context to data. Based on information we gain knowledge. What differentiates knowledge from information is the developed structure and presence of interpretation rules. (Stępniak, 2014, p. 40). Explaining the difference between knowledge and information, J. M. Dunn stated that information is something that remains when knowledge is deprived of such elements as: truthfulness, credibility, justification, conviction and reference to the learning entity (Dunn, 2008, p. 581). In order to understand the essence of knowledge, information and data in the economic process we need to adopt a wider approach than that offered by economists. The development of IT after the Second World War increased the calculation capabilities allowing us to process a huge amount of information. In this context, A Mathematical Theory of Communication (Shannon, 1948) was of key importance. The author used quantitative analysis to draw our attention to the issues related to communication and passing information. Before Shannon presented his theory, economists from two opposing teams – supporters of capitalistic market economy and proponents of socialism based on central planning – had conducted a debate on the possibility of rational economic calculation in centrally-planned economies. Its essence concerned the possibility of using knowledge and processing information in the process of economic calculation that allows optimal allocation of resources. The debate was started by Ludwig von Mises with his article “Die Wirtschaftsrechnung im sozialistischen Gemeinwesen” (Mises, 1920). He negated the possibility of rational economic accounting in socialist economy, in which there is no market, and thus there are no prices interpreted as parameters (data) that allow making rational decisions. In a reply to Mises, Oskar Lange, a Polish economist, proposed a model of quasi-market socialism. He showed how a Central Pricing Board, using statistical methods, may use data concerning the volume of demand and supply to determine market equilibrium prices (Lange, 1936, 1937). Lange presented a logically coherent model corresponding to the general equilibrium proposed by Léon Walras.

Asymmetric information as a market failure in retrospect  109 He demonstrated, albeit only theoretically, how a Central Planning Board could process data in order to balance demand-supply transactions in the whole economy. He did not prove, however, that this theory might be used in practice to replace the market with such a Board. Mises’ disciple, F. A. Hayek, criticized Lange’s concept pointing at limitations of knowledge possessed by an individual. In his article “The Use of Knowledge in Society” Hayek claims: “the knowledge of the circumstances of which we must make use never exists in concentrated or integrated form, but solely as the dispersed bits of incomplete and frequently contradictory knowledge which all the individuals possess” (Hayek, 1945, p. 519). In Hayek’s interpretation, the market is a telecommunication mechanism through which information is passed and knowledge is transferred. The socialism versus capitalism debate was to determine which of these social and economic mega-systems was characterized by higher efficiency. Would it be market-based capitalism? Or centrally-planning socialism? The events that started in 1989 clearly proved that social and economic systems based on the market have greater economic efficiency. At the beginning of the 1990s, the Soviet bloc collapsed, strengthening our belief in the driving force of market solutions. In the historical perspective, the market turned out to be the most effective mechanism of resource allocation. This, however, does not mean that it is perfect. Long before the dispute concerning the possibility of rational economic accounting in a centrally-planned economy, economists discerned manifestations of market failure. Representatives of the Austrian school, Mises and Hayek, enhanced the discourse with a new element concerning knowledge and the way in which individuals process information.

3.

FROM CELESTIAL MECHANICS OF NON-EXISTENT SPHERES TO MARKET FAILURE

Economic literature is full of reservations concerning the way in which the market operates. On the macroeconomic plain, such criticism usually refers to problems connected with unemployment, instability of economy demonstrated in fluctuations of business cycles, or relatively low effectiveness of market solutions in eliminating the difference in economic development between developed and underdeveloped countries. The criticism of market solutions is often expressed in normative judgments. Their authors point at growing inequalities and unfair levels of income, which lead to social tensions. Finally, we have the opinions of such social philosophers as Michael Sandel, who showed the moral limits of the market and posed the question of what money can’t buy (Sandel, 2012). Like every mechanism, the market has numerous limitations and may fail in some circumstances. To speak reasonably of market failure, one must provide clearly defined criteria of efficiency. They are necessary, since otherwise it would be difficult to determine whether something is a market failure or not. Of many existing theories, neoclassical economics is the one that provided in the most unambiguous way the criteria of efficiency allowing us to maximize social prosperity. They cover both the sphere of consumption and the sphere of production, as well as relations between the production sphere and the consumption sphere. The most advanced presentation of Walras’ idea of general equilibrium was presented by K. J. Arrow and G. Debreu (1954) as well as L. W. McKenzie (1954), who used advance mathematics to prove the existence of general equilibrium in economy. This was undoubtedly a significant theoretical achievement which dominated the thinking on economy of future

110  The Elgar companion to information economics generations of economists. However, it did not provide the proof of the existence of such state of equilibrium in the real world. Therefore the achievements of Arrow and Debreu, as well as McKenzie, can be defined as celestial mechanics of non-existent spheres. We should also agree with the opinion that Arrow and Debreu’s genius consisted in the fact that they were able to discern a case of an efficiently operating market (assuming full information of entities and completeness of markets) in a world in which inefficiency is a common phenomenon (Arnott et al., 1993, p. 14). Market failure is a certain state of economic reality diverging from the adopted model assumption. Such an approach was proposed by Francis M. Bator. He was not the first economist to observe market failures, but it was he who distinguished market failure on the basis of clearly defined criteria. In his article titled “The Simple Analytics of Welfare Maximization” (Bator, 1957), taking advantage of the Edgeworth-Bowley box he used indifference curves and isoquants (being production functions) to show the essence of the equilibrium of the whole economy. He also presented relevant equilibrium conditions. In another article, “The Anatomy of Market Failure” (Bator, 1958, p. 351) he claimed that “typically, at least in allocation theory, we mean the failure of a more or less idealized system of price-market institutions to sustain ‘desirable’ activities or to stop ‘undesirable’ activities. The desirability of an activity, in turn, is evaluated relative to the solution values of some explicit or implied maximum-welfare problem”. Expanding the presented definition, he described the New Welfare Economics as an analytical plain with reference to which market failure is considered.4 He sought an answer to the question of the extent to which Pareto optimal allocation described with a mathematical model reflects the reality. In “The Simple Analytics” paper, Bator showed an ideal model of the world in which the market ensures maximization of welfare for particular individuals and the whole society. On the other hand, in “The Anatomy” paper he rejected this idealized picture of the world, explaining why maximization of the individual utility function is not identical with maximization of the social welfare function, and he demonstrated what market failures are (Giza, 2013, pp. 48–49). The most common types of market failure include: failure of competition, public goods, externalities, incomplete markets, imperfect information (asymmetric information). The first of these market failures stems from the fact that a perfect competition market is an extremely rare phenomenon. Even authors of economics textbooks find it extremely difficult to provide examples that could be defined as perfect competition. The overwhelming majority of existing markets do not meet the following criteria: existence of a very large number of entities, both on the supply and demand side; companies are price takers; a product is homogeneous; there are no market entry barriers. The most frequently given example is the market for some agricultural produce. In reality, however, we have imperfect competition markets. They include: monopolistic competition, monopsonistic competition, oligopoly, oligopsony, monopoly, and monopsony. In the 1930s J. V. Robinson (1933) and E. H. Chamberlin (1933) argued that the ideal of perfect competition should be abandoned and a more realistic approach based on the analysis of imperfect competition should be adopted. Their proposal, however, did not bring any breakthrough changes in the theory of economics. A perfect competition market is able to effectively allocate private good. However, as far as public good is concerned,5 the market fails. The essence of public goods was defined by such economists as E. Lindahl (1919), H. R. Bowen (1943), P. A. Samuelson (1954) and (1955), and R. A. Musgrave (1939). Both Lindahl and Bowen considered public goods in the context of seeking optimal ways of financing them. They also tried to capture the characteristics

Asymmetric information as a market failure in retrospect  111 of such goods. Bowen distinguished individual goods and social goods on the basis of the divisibility criteria. Individual goods are divisible. Therefore, it is possible to provide them in quantities strictly corresponding to the volume of reported demand. In the case of social goods, their indivisibility makes it impossible to provide an individual with the exact amount of a particular good which they report demand for. This led Bowen to conclude that demand for social goods (which we currently define as public goods) is not determined in terms of an individual dimension, but in a social and political dimension (Bowen, 1943, p. 27). Referring to Bowen’s classification, Samuelson defined public goods not through non-divisibility, but through non-rivalry. He stated that “A public consumption good, like an outdoor circus or national defense … is provided for each person to enjoy or not, according to his tastes” (Samuelson, 1955, p. 350). Non-rivalry accounts for the fact that consumption of a public good in a particular quantity by one consumer does not decrease the possibility of consuming the same good by another consumer. Another feature of public goods is their non-excludability, which stands for the lack of possibility of excluding other individuals from consumption. Based on this criterion, J. M. Buchanan developed the theory of club goods (Buchanan, 1965). Club goods lie between purely public goods, such as Samuelson’s example of national defense, and private goods (for instance a cup of coffee). The application of the above-indicated criteria allows us also to distinguish common resources. Fish swimming in the river are an example of such a good. It is characterized by rivalry and non-excludability. We can also take a broader look at public goods. We owe this perspective to Musgrave, who introduced the category of merit wants to the economic discourse. This denotes wants, which, even though desirable from the perspective of the whole society, are not perceived in this way by an individual. The goods that satisfy such needs are merit goods. Musgrave negates the extremely individualistic assumption concerning rationality of consumers and adds elements of paternalism to the discussion, as well as the thread concerning the role of a state – the arbiter correcting the wrong decisions of individuals. Referring to Musgrave, J. G. Head defined merit goods as those which, thanks to imperfect knowledge possessed by individuals, are consumed to an insufficient extent. On the other hand, demerit goods are goods that due to imperfect knowledge might be consumed by entities in an excessive way (Head, 1966, p. 3). Another type of market failure are externalities. They were observed by Alfred Marshall (1920, p. 266) in his analysis of the relationship between externalities and a company’s scale of production. He also noticed that entrepreneurs who have access to information are able to use their capital and labor resources more efficiently (Boudreaux & Meiners, 2019, p. 3). However, he did not notice the relationship between externalities and the level of social welfare. This issue was analyzed by A. C. Pigou in the 1920s. He proposed to levy a special tax to prevent externalities (Pigou, 1920, pp. 131–135). This concept was developed by J. E. Meade (1952) and T. Scitovsky (1954). Currently, external economies are defined as benefits or costs that go beyond the activities of one entity and are borne by other entities, without any compensation. The essence of external effects was presented by Meade in a simple model that takes into consideration two producers: an owner of an orchard and a beekeeper (Meade, 1952, pp. 54–61). For the owner of an orchard, the volume of obtained produce depends on the expenditure of land, work and capital made in the process of producing apples. In order to produce honey it is not only necessary to make expenditure connected with keeping beehives, but also to have trees from which bees will collect nectar. Therefore, the growing area of orchards provides the beekeeper with additional product without him having to make any additional payments. In the production process we can observe a production factor which Meade

112  The Elgar companion to information economics defines as unpaid. In this case, the beekeeper is a beneficiary of a positive external effect generated by the owner of the orchard. Increased production of apples leads to a situation in which the social cost of production differs from individual cost, thus making maximization of welfare impossible. Scitovsky made a generalization of Meade’s concept. He claimed that external effects are the reason why individual benefit does not coincide with social benefit. This leads to a solution which does not meet Pareto’s optimum requirements. Analyzing types of interaction between producers and consumers, Scitovsky distinguished external economies, which refer both to the production sphere and the consumption sphere (Scitovsky, 1954, p. 144). The formal presentation of external effects was proposed by J. M. Buchanan and W. C. Stubblebine. Total utility achieved by entity A will take the following shape (Buchanan & Stubblebine, 1962, p. 372): UA = UA(X1, X2 ,…, Xm , Y1). According to this formula, the utility of entity A depends on actions resulting from their own decisions (X1, X2,…,Xm), as well as from actions taken by other entities (Y1). Decisions taken by other members of the society affect positively or negatively the satisfaction level of a given individual. In both cases they will disturb the process of achieving Pareto optimal allocation.

4.

ASYMMETRIC INFORMATION AND ITS SIGNIFICANCE

Taking into account asymmetric information constituted a breakthrough point in the analysis of market failure. Stiglitz claimed that the discovery of asymmetric information marked the boundary between the old and the new presentation of market failure (Stiglitz, 2002, pp. 56–57). In neoclassical economics market analysis is performed assuming that there is full and free information. In the 1970s, George A. Akerlof presented a model of the market functioning in a situation when one side of the transaction possesses more information than the other (Akerlof, 1970). In this way asymmetric information has become a central element in the analysis of market failure. Its existence accounts for the appearance of adverse selection and moral hazard. These phenomena make it difficult to find a Pareto-optimal solution (Prescott & Townsend, 1984). In his article Akerlof used an example of the market for used cars (Akerlof, 1970). He wondered why cars that entered this market right after leaving the dealer’s showroom, immediately lose value. This cannot be attributed to their loss of utility value. He rejected the explanation that a consumer pays more for the possibility of owning a brand new good. In the model presented by Akerlof, asymmetric information makes buyers determine the price they are willing to pay for a used car not on the basis of its actual technical condition, but on the basis of averaged expectations concerning the technical condition of similar cars. Empirical research proves that cars characterized by higher failure rate more frequently enter the used cars market. Therefore potential buyers are willing to offer lower prices for such products. This situation is disadvantageous for owners of cars in perfect technical condition. They find it difficult to get the price adequate to the quality of cars they offer. Adverse selection is a consequence of this process. Akerlof’s argumentation allows us to explain logically different prices of cars bought from car dealers and nearly new cars bought in the used cars market on the basis of rational calculation of benefits and costs. Akerlof did not content himself with the explanation of asymmetric information. He also tried to find markets in which it plays a vital role. In his 1970 article he listed the insurance

Asymmetric information as a market failure in retrospect  113 market, medical services market,6 labor market and bank loans market as each characterized by asymmetric information that significantly disturbs its operations. Determining market failure resulting from asymmetric information is important, but an equally important issue is to seek means of counteracting these imperfections. Such means include: institutional solutions concerning warranty granted by manufacturers and brands that stand for high quality products. The research initiated by Akerlof was continued by Michael Spence in his article “Job Market Signaling”. He examined how employers read signals in a situation when asymmetric information does not allow an employer clearly to distinguish the qualifications of an employed worker. An important signal allowing the employer to determine the competencies of a new worker is the candidate’s education. Gaining sound education usually requires predisposition and expenditure. If the employed person has a diploma from a prestigious university which sets high standards of education, it means that a graduate has considerable potential. Therefore, education is an important signal. It allows the employer to make the right selection and to optimize the offered pay (Spence, 1973). Another important article devoted to the way in markets function was “Adverse Selection in the Labor Market” by Bruce Greenwald who analyzed the consequences of adverse selection for a company’s employment policy (Greenwald, 1986). The insurance market was analyzed, inter alia, by Shavell (1979)7 and Stiglitz and Weiss (1981). Research on asymmetric information and incompleteness of markets led Greenwald and Stiglitz to formulate some general statements in their article “Externalities in Economies with Imperfect Information and Incomplete Markets”. They claimed that contrary to the approach represented by Bator, market failure is not only a type of deviation from the optimal equilibrium state. It is an integral feature of market mechanism. The Greenwald-Stiglitz theorem states that “economies in which there are incomplete markets and imperfect information are not, in general, constrained Pareto efficient. There exist government interventions (e.g., taxes and subsidies) that can make everyone better off” (Greenwald & Stiglitz, 1986, p. 230). This theorem can be interpreted as criticism of the paradigm of neoclassical economics – especially the Arrow and Debreu model. Their genius consisted in ability to discern effectively operating markets in the world in which inefficiency is a common phenomenon. Nevertheless, they adopted an assumption of full information and completeness of markets (Arnott et al., 1993, p. 14). Considering market failures as a rule rather than an exception brings significant consequences for welfare economics. In his work The Invisible Hand and Modern Welfare Economics Stiglitz referred to the first and second fundamental theorems of welfare economics. He showed the relation between the formal presentation given by Arrow and Debreu and Smith’s invisible hand. He objected to idealizing the market mechanism on the basis of the formal presentation of Arrow and Debreu and using such idealization in ideological disputes concerning the relation between the state and the market. The concealment of vital assumptions leads to unjustified extrapolation of conclusions drawn from the mathematical formula into specific situations in economic life (Stiglitz, 1991, pp. 2–6). Apart from asymmetric information, Greenwald and Stiglitz also paid attention to the incompleteness of markets. The fundamental question is where one can observe incomplete markets and what are the reasons preventing the emergence of markets for all goods. The above questions were answered by Adam Przeworski, who emphasized that the incompleteness of markets is mostly noticeable in the following cases: futures markets, risk markets, and markets for future labor (Przeworski, 2003, pp. 43–46).

114  The Elgar companion to information economics Seeking the reasons behind incompleteness of markets, Przeworski concluded that each good that is an object of a transaction must be defined on the basis of its physical features and place and time of its delivery. A consumer optimizes the structure of consumption not only in the present period, but also in the inter-period dimension. How can we then determine the optimal structure of consumption if a specific good does not exist in the current period? Incompleteness of markets in which futures contracts are concluded results from the specificity of this good. Futures contracts are usually concluded for homogeneous goods (for example some raw materials). However, for heterogeneous goods, such as a new model of car, such transactions generally do not happen. It is difficult to determine the value of particular goods in a situation when they have not been produced yet and we do not know their utility features. Another example of market incompleteness is the lack of markets for future labor. The specificity of this production factor inclines us to treat it differently than other production factors (for instance capital and land). This is connected with ethical beliefs concerning individual freedom. Therefore Przeworski argues that markets for future labor practically do not exist. Risk can also be the reason why there are no markets for certain goods. This is because entities, having taken out an insurance policy, engage in riskier activities, convinced that the insurance company will cover the costs of potential damage. This phenomenon is known as moral hazard. It accounts for the fact that insurance companies are not prone to insure against certain kinds of risk (Arrow, 1963; Pauly, 1968). The incompleteness of markets may be caused by external effects and public goods. In the latter case we observe the so-called fare dodger effect. If there is no possibility of excluding a particular public good from consumption, some consumers will take actions aimed at avoiding participation in costs of financing this good. This, in turn, may make the means that could be allocated to such goods insufficient. Analogically, we can formulate an argumentation concerning the operation of external effects.

5. CONCLUSIONS Asymmetric information is a phenomenon which seems apparently obvious. However, it brings about a series of important consequences in economic life. When economists discerned asymmetric information, it changed the perception of market economy efficiency. The traditional approach to market failure, initiated by Marshall at the end of the nineteenth century in Cambridge and developed by Bator, determined market failure as an inherent part of economic reality. Markets fail because they either diverge from the adopted model of perfect competition, or because goods which are exchanged through them have some specific features that do not allow to reach the maximum level of social welfare. Hayek’s interpretation of the market as a telecommunication mechanism encouraged economists to pay attention to the issues of knowledge and information. Hayek’s approach, in spite of its novelty, was treated as a manifestation of economic heterodoxy. Mainstream economists are inclined to treat market failure as something related to reality perception and the cognitive limitations of homo oeconomicus. This discovery led such economists as Akerlof, Stiglitz, Greenwald, Spence and others to show the complexity of economic relations in the perspective of an information society in which we are living. The economists analyzing market failure made the Holy Grail

Asymmetric information as a market failure in retrospect  115 of neoclassical economics – general equilibrium – merely a desirable state of the world rather than a description of reality.

NOTES 1. 2. 3. 4.

5. 6. 7.

Chapters 2 and 3 of this Companion allude to the contemporary analysis of economics of information from the 1970s to today (Stiglitz & Kosenko, 2024a, pp. 20–52, 2024b, pp. 53–80). The negatively sloped indifference curve proposed by Johnson can be found in contemporary microeconomics textbooks. A little earlier, F. Y. Edgeworth presented a positively sloped indifference curve (Edgeworth, 1881, p. 28). Chapter 1 of this Companion presents the significance of the excess and scarcity of information in the perspective of economic theory (Raban & Włodarczyk, 2024, pp. 2–19). The difference between the old and the new economics of welfare consists in the method of calibrating utility. In old economics of welfare, such economists as Marshall relied on the cardinal theory of utility, trying to apply it to measuring welfare with the concept of consumer’s rent (surplus). On the other hand, new economics of welfare is based on ordinal utility and the Pareto criterion. Taking into account the neoclassical criterion of efficiency one might doubt whether information is the same public good as, for example, national defense. However, one can follow the example of Stiglitz (2021) and take a broader look at this issue. Chapter 8 of this Companion presents the functioning of the medical services market in conditions of information asymmetry (Olivella, 2024, pp. 154–169). Steven Shavell cites in his articles the names of Kenneth Arrow and Mark Pauly as the first scientists who, in the 1960s, pointed at the phenomenon of moral hazard.

REFERENCES Akerlof, G. A. (1970). The Market for “Lemons”: Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84(3), 488–500. Arnott, R., Greenwald, B., & Stiglitz, J. E. (1993). Information and Economic Efficiency. Working Paper No. 4533. Cambridge: National Bureau of Economic Research. Arrow, K. J. (1963). Uncertainty and the Welfare Economics of Medical Care. American Economic Review, 53(5), 941–973. Arrow, K. J. & Debreu, G. (1954). Existence of an Equilibrium for a Competitive Economy. Econometrica, 22(3), 265–290. Bator, F. M. (1957). The Simple Analytics of Welfare Maximization. American Economic Review, 47(1), 22–59. Bator, F. M. (1958). The Anatomy of Market Failure. The Quarterly Journal of Economics, 72(3), 351–379. Boudreaux, D. J. & Meiners, R. (2019). Externality: Origins and Classifications. Natural Resources Journal, 59(1), 1–34. Bowen, H. R. (1943). The Interpretation of Voting in the Allocation of Economic Resources. The Quarterly Journal of Economics, 58(1), 27–48. Buchanan, J. M. (1965). An Economic Theory of Clubs. Economica, New Series, 32(125), 1–14. Buchanan, J. M. & Stubblebine, W. C. (1962). Externality. Economica, New Series, 29(116), 371–384. Chamberlin, E. H. (1933). Theory of Monopolistic Competition. Cambridge, MA: Harvard University Press. Colander, D. (2000). The Death of Neoclassical Economics. Journal of the History of Economic Thought, 22(2), 127–143. Dunn, J. M. (2008). Information in Computer Science. In P. Adriaans & J. van Benthem (Eds.), Philosophy of Information (pp. 581–608). Amsterdam: Elsevier.

116  The Elgar companion to information economics Edgeworth, F. Y. (1881). Mathematical Psychics: An Essay on the Application of Mathematics to the Moral Sciences. London: Kegan Paul & Co. Giza, W. (2013). Zawodność rynku. Powstanie i rozwój idei [Market Failure. The Formation and Development of the Idea]. Kraków: Wydawnictwo Uniwersytetu Ekonomicznego w Krakowie. Grattan-Guinness, I. (2000). A Sideways Look at Hilbert’s Twenty-three Problems of 1900. Notices of the American Mathematical Society, 47(7), 752–757. Greenwald, B. C. (1986). Adverse Selection in the Labor Market. Review of Economic Studies, 53(3), 325–347. Greenwald, B. C. & Stiglitz, J. E. (1986). Externalities in Economies with Imperfect Information and Incomplete Markets. The Quarterly Journal of Economics, 101(2), 229–264. Hayek, F. A. (1945). The Use of Knowledge in Society. American Economic Review, 35(4), 519–530. Head, J. G. (1966). On Merit Goods. FinanzArchiv, 25(1), 1–29. Honneth, A. (1987). Enlightenment and Rationality. The Journal of Philosophy, 84(11), 692–699. Johnson, W. E. (1913). The Pure Theory of Utility Curves. The Economic Journal, 23(92), 483–513. Lange, O. (1936). On the Economic Theory of Socialism. Part One. The Review of Economic Studies, 1, 53–71. Lange, O. (1937). On the Economic Theory of Socialism. Part Two. The Review of Economic Studies, 2, 123–142. Lindahl, E. (1919). Die Gerechtigkeit in der Besteuerung. Eine Analyse der Steuerprinzipien auf Grundlage der Grenznutzentheorie. Lund: Gleerupska Universitets-Bokhandeln. Mandeville, B. (1970 [1714]). The Fable of the Bees. London: Penguin Books. Marshall, A. (1920). Principles of Economics, 8th edition. London: Macmillan and Co. McKenzie, L. W. (1954). On Equilibrium in Graham’s Model of World Trade and Other Competitive Systems. Econometrica, 22(2), 147–161. Meade, J. E. (1952). External Economies and Diseconomies in a Competitive Situation. The Economic Journal, 62(245), 54–67. Mises, L. (1920). Die Wirtschaftsrechnung im sozialistischen Gemeinwesen. Archiv für Sozialwissenschaft und Sozialpolitik, 47, 86–121. Musgrave, R. A. (1939). The Voluntary Exchange Theory of Public Economy. The Quarterly Journal of Economics, 53(2), 213–237. Olivella, P. (2024). Asymmetric Information in Health Economics: Can Contract Regulation Improve Equity and Efficiency? In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 154–169). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Pauly, M. V. (1968). The Economics of Moral Hazard: Comment. American Economic Review, 58(3), 531–537. Pigou, A. C. (1920). The Economics of Welfare. London: Macmillan. Prescott, E. C. & Townsend, R. M. (1984). Pareto Optima and Competitive Equilibria with Adverse Selection and Moral Hazard. Econometrica, 52(1), 21–46. Przeworski, A. (2003). States and Markets: A Primer in Political Economy. Cambridge: Cambridge University Press. Raban, D. R. & Włodarczyk, J. (2024). Information Economics Examined through Scarcity and Abundance. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 2–19). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Robinson, J. V. (1933). The Economics of Imperfect Competition. London: Macmillan. Samuelson, P. A. (1954). The Pure Theory of Public Expenditure. The Review of Economics and Statistics, 36(4), 387–389. Samuelson, P. A. (1955). Diagrammatic Exposition of a Theory of Public Expenditure. The Review of Economics and Statistics, 37(4), 350–356. Sandel, M. J. (2012). What Money Can’t Buy: The Moral Limits of Markets. New York: Farrar, Straus & Giroux. Scitovsky, T. (1954). Two Concepts of External Economies. The Journal of Political Economy, 62(2), 143–151. Shannon, C. (1948). A Mathematical Theory of Communication. The Bell System Technical Journal, 27, 379–423.

Asymmetric information as a market failure in retrospect  117 Shavell, S. (1979). On Moral Hazard and Insurance. The Quarterly Journal of Economics, 94(4), 541–562. Simon, H. (1947). Administrative Behavior: A Study of Decision‑Making Processes in Administrative Organization. New York: Macmillan. Simon, H. (1955). A Behavioral Model of Rational Choice. The Quarterly Journal of Economics, 69(1), 99–118. Simon, H. (1956). Dynamic Programming Under Uncertainty with a Quadratic Criterion Function. Econometrica, 24(1), 74–81. Simon, H. (1978). Rationality as Process and as Product of Thought. American Economic Review, 68(2), 1–16. Smith, A. (1976 [1776]). An Inquiry into the Nature and Causes of the Wealth of Nations. Oxford: Oxford University Press. Spence, M. (1973). Job Market Signaling. The Quarterly Journal of Economics, 87(3), 355–374. Stępniak, A. (2014). Różnice między informacją a wiedzą w kontekście zarządzania [Differences Between Knowledge and Information with Reference to Management]. Studia Metodologiczne, 32, 29–47. Stiglitz, J. E. (1991). The Invisible Hand and Modern Welfare Economics. Working Paper No. 3641. Cambridge: National Bureau of Economic Research. Stiglitz, J. E. (2002). Keynesian Economics and Critique of First Fundamental Theorem of Welfare Economics. In T. Cowen & E. Crampton (Eds.), Market Failure or Success: The New Debate (Chapter 3). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. (2021). World Press Freedom Day 2021. Information as a Public Good: 30 Years of the Windhoek Declaration. https://​en​.unesco​.org/​sites/​default/​files/​wpfd​_2021​_concept​_note​_en​.pdf. Stiglitz, J. E. & Kosenko, A. (2024a). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 1: Indirect Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 20–52). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. & Kosenko, A. (2024b). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 2: Direct Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 53–80). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. & Weiss, A. (1981). Credit Rationing in Markets with Imperfect Information. American Economic Review, 71(3), 393–410. Varian, H. R. (2010). Intermediate Microeconomics: A Modern Approach. New York: W. W. Norton & Company.

6. A quadrennial review of the significance of information asymmetry in economics and finance Pedro A. Martín-Cervantes and María del Carmen Valls Martínez

1. INTRODUCTION In the field of social sciences, sometimes researchers are not completely conscious of what lies behind the abstraction of reality from a series of explanatory variables or the enunciation of a starting hypothesis. Such abstraction of the economic-financial reality is not entirely aseptic, since scholars are subsumed within a socio-cultural environment which has a decisive influence when explaining economic transactions by reducing them to the minimum expression through variables or drawing generalizations through theories linked to hypotheses formulated with a greater or lesser degree of coherence. These factors are clearly perceptible in the first economic-financial models that were defined at the beginning of the twentieth century and, even more so, if we observe in detail how the “information” that transcends the economies or financial markets was conceptualized. Indeed, this is reflected in the pioneering works of Bachelier, such as “Theory of Speculation” (Bachelier, 1900) or “The Periodicity of Hazard” (Bachelier, 1915), in which he explores themes as close to the present-day researcher as the economics of information or establishes the bases for the analysis of moral hazard. In fact, Bachelier’s oeuvre epitomizes how behind every economic-financial theory there are often sociological and axiological conditioning factors which, just as they can arbitrarily reject a scientifically proven theory, can also validate a theory or hypothesis that is weak from an empirical point of view. Bachelier’s findings are inextricably linked with M. Smoluchowski and B. B. Mandelbrot. In the first case, Bachelier applies the theory of Brownian motion recently discovered to the analysis of the evolution of the prices of financial assets (Smoluchowski, 1906) while in the second, it is verified how the theoretical schemes provided by Bachelier could be valid under certain conditions (Mandelbrot, 1963). Nevertheless, Bachelier’s preliminary work needed to be supported by a certain specific conceptual framework, and this would eventually be none other than the Hypothesis of Efficient Markets that he would anticipate in his work Le Jeu, la Chance et le Hasard (Bachelier, 1914). In this work it was assumed that the participants in a financial market have the same level of information and, in addition, they take completely rational decisions from a mathematical point of view. Today, this assumption, which represents the antithesis of the concept of Information Asymmetry, could be automatically discarded but it played an essential role in the development of neoclassical economic thought, which was the predominant approach until well into the 1960s. Note that Bachelier’s papers were literally despised for two reasons: on the one hand, they promulgated speculative practices that were highly reviled at that time (Bachelier, 1900). On the other hand, his works encouraged the bourgeoisie (Foster, 2018) and the middle classes to 118

Information asymmetry in economics and finance  119 make investments, a possibility that formerly was only afforded to large mercantile corporations. Jovanovic (2001) locates the entry of the popular classes into financial markets during the epoch of the construction of the Suez Canal, a period contemporary to the life and work of Bachelier. Actually, the popular classes massively participated in the issuance of bonds of the “Compagnie Universelle du Canal Maritime de Suez” instead of the large European or American investment banks who rejected the viability of this venture because of the enormous risks it might entail. Obviously, those small investors were led to believe that they were fully informed and that the risks of their investments would be perfectly safe. As we can see, this was the social germ of the Efficient Markets Hypothesis, based on the assumption that the information available to buyers and sellers is perfect by definition. However, although the mathematical bases of the “Theory of Speculation” were absolutely congruent (Bachelier, 1900), this work remained for more than 50 years completely anonymous since a monograph that was entirely dedicated to the practice of speculation contravened the moral-ethical norms of the French haute bourgeoisie of that time (Jovanovic, 2001). Consequently, the theory of Brownian motion developed by Einstein and Smoluchowski was anticipated by the work of Bachelier, a fact that continues to be largely unknown by the Academy and that should be recognized in justice (Bernstein, 2005). It seems that the oeuvre of the French mathematician was forgotten due to the conservatism of that society, which refused to accept any form of speculation as an investing practice. Nevertheless, this manuscript would have enormous repercussions on the future of economic and financial modeling in the long run, contributing to the formulation of the “Paradigmatic Economy” (Stiglitz, 2002). Fundamentally, two facts led to the rediscovery of Bachelier’s work, once rejected for simple axiological constraints. First, it served as a foundation on which P. Samuelson established that financial asset prices describe a geometric Brownian motion (Samuelson, 1965a, 1965b), that is, they follow a completely random behavior or movement. Second, such an assumption justified one of the most relevant hypotheses in the scientific literature, the Efficient Capital Markets Hypothesis (Fama, 1970), which was accepted by the Academy almost unanimously. At the same time, the Efficient Capital Markets Hypothesis transcended to public opinion, to financial-economic institutions and even to legal institutions since as Jovanovic et al. (2016) emphasize, it has been part of the legal corpus of US courts since 1973 (“Fraud on the Market Doctrine”). The underlying basic approach is quite simple: in the same way that an investor holds an aliquot share of the capital of a company, he or she is also entitled to hold a share of the information that comes out of the markets, which, moreover, is also assumed to be truthful and meaningful (Martín-Cervantes, 2020). However, its empirical fragility is evidenced by the fact that financial markets1 occasionally experience sudden shocks (Stiglitz, 1990). Similarly, in the context of contemporary economies, money constitutes a non-neutral good from which networks can be established between different economic partners whose knowledge of the surrounding information is conditioned by a sociological order (Włodarczyk, 2014, 2015). The Efficient Markets Hypothesis is not the only erroneous approach, but there are a plethora of theories of a similar nature, such as the Rational Expectations Theory (Muth, 1961), which were also based on somewhat fallacious approaches. These formulations assumed that all stakeholders involved in a given transaction or market count on transparent information that, a priori, is shared by all parties. In response to this conceptual scheme, further investigations began to emerge taking into account the reality of markets and human behavior which, on the other hand, does not always move under strictly rational parameters. Under this umbrella, the term “information asymmetry” was born, generating a disruptive point of view that this

120  The Elgar companion to information economics chapter aims to summarize. A procedure similar to Cuccurullo et al. (2013) or Raban and Gordon (2020) has been employed which, to the best of our knowledge, has allowed us to elaborate the first work that, in a novel way, specifically performs a bibliometric analysis centered on the analysis of information asymmetry. The rest of the chapter is structured as follows: the essential conceptual bases of information asymmetry are outlined, followed by a brief outline of bibliometric analysis, determining the database and methodology used, as well as the main features related to citations and scientific production in this field of knowledge. Likewise, the main characteristics, trends, and perspectives detected through the application of bibliometrics are also indicated. Finally, it ends with a summary highlighting the main findings and conclusions obtained in this research.

2.

RATIONALE OF INFORMATION ASYMMETRY

The Efficient Markets Hypothesis must be understood within the sociological framework in which it was enunciated during the 1970s. At that time, the implementation of nuclear energy was in full swing and society conferred it a certain mystical halo, becoming a factotum around which any order of human development would revolve. In this regard, the implications and similarities between the works of Bachelier and Einstein (Read, 2012) gave the ordinary citizen the possibility of investing in a scientifically contrasted way on the basis of a series of theories whose mathematical development was impeccable. However, these theories approached the phenomenon of information on the basis of general hypotheses whose empirical verification in practice proved to be a chimera. The iconic Black-Scholes model (Black and Scholes, 1973) was based on this same conditioning factor, assuming, among other points, the prevalence of the Efficient Markets Hypothesis in an analytical scheme that dismissed the occurrence of extreme movements in financial markets, given that the prices of financial assets would tend to adjust in the long run around their average values according to the information obtained (Clarkson, 1995). This formula, considered by Stewart (2012) as one of the 17 most important equations in the history of mankind, faithfully embodies the weaknesses of rational equilibrium models: they were intended to evaluate human behaviors, which are characterized by usually exhibiting certain doses of irrationality (Akerlof and Yellen, 1987). Sometime later, just when paradigmatic economics was at its peak, some papers began to appear from different theoretical-empirical points of view proposing revolutionary lines of research that a priori considered the information as an “asymmetric phenomenon” which could not be unambiguously modeled, nor could it be contrasted according to excessively generic hypotheses. This resulted in the appearance of novel models such as the Glosten-Milgrom Model (Glosten and Milgrom, 1985) aimed at studying the microstructure of markets from a heterogeneous point of view that was much closer to reality, in which both types of informed and uninformed individuals coexisted. To achieve the change in the orientation of the aims and objectives of economic science as outlined by Stiglitz (2000), Akerlof’s “market for lemons” (Akerlof, 1970) laid the foundation for a new and seminal line of research: work on the concept of Information Asymmetry, whether from its public or private side, or focused on general economics or financial economics, was the primary line of research that would lead to the awarding of the Nobel Prize in

Information asymmetry in economics and finance  121 Economics to scholars such as G. A. Akerlof, M. Spence, J. A. Mirrlees, W. S. Vickrey, and J. E. Stiglitz (Sandmo, 1999; Löfgren et al., 2002; Rosser, 2003). Given the multifaceted nature of asymmetries of information from theoretical and empirical perspectives, the growth of information economics has been exponential from the 1970s to the present, diversifying towards numerous and varied fields. This fact highlights another characteristic of information economics, its ability to adapt to changing times or to continually redefine its objectives. Therefore, it is essential to know in a rigorous manner how this discipline has been evolving since its initial stages, specifying its main drivers, trends, sources, or geographical areas which have contributed to its development. To achieve this purpose, bibliometrics is a useful tool when considering its ability to dissect any topic based on multiple scientific parameters that allow performing long-term empirical inferences and contrasts on subjects as diverse as the effect of technology on learning research trends or the conceptual evolution of the term “Corporate Social Responsibility” (Raban and Gordon, 2015; Martín-Cervantes et al., 2021). In summary, information asymmetry analyzes decisions taken in an environment in which one of the parties has a greater quantity and/or quality of information than the other (Postlewaite, 2018), usually from the perspective of the agent who is acting as a seller. Therefore, information asymmetry can be characterized as unbalanced information, as specified in Figure 6.1. The results of this phenomenon, among many others, can be the following: the emergence of various types of market failures2 (Ledyard, 2018), the tendency to the create monopolistic or oligopolistic practices (West, 2018) given that markets lose the efficiency of their capacity of assignation, contexts of total or partial uncertainty (Wakker, 2018), disincentives to produce under quality standards, in addition to entailing an obvious circumstance for the emergence of moral hazard (Kotowitz, 2018).

Source: Own elaboration.

Figure 6.1

Basis of information asymmetry

122  The Elgar companion to information economics Note that accepting any of these effects by the mere fact of assuming that some economic operators may have more information than others, would be completely inadmissible for neoclassical theory. However, from a behavioral point of view, behavioral theory finds an explanation in the term “bounded rationality” (Simon, 2018). According to this behavioral precept, human knowledge is not unlimited, it has its own constraints and barriers. Humans are not computing machines capable of establishing a utility function in which to order all possible alternatives based on available information. Therefore, according to Simon (2018), they are unable to assign consistently all viable alternatives in terms of probability, particularly in those situations where there is uncertainty, as is the case when information asymmetry arises.

3.

BIBLIOMETRIC ANALYSIS

Notwithstanding the emergence in the literature of bibliometric methods and procedures over the last decade, the origin of this field of knowledge remains somewhat unknown to the general public and, to a certain extent, it maintains a common link with the economics of information: its pragmatic character, or the fact of being able to provide conclusive answers to a priori non-stereotyped research. Grosso modo, the following stages can be distinguished in the timeline that led to the current configuration of bibliometrics that was initially implemented mainly in psychology-related fields (Godin, 2006): (a) Primitive works that Shapiro (1992) locates in the “publication counts” of the nineteenth century, coming from the legal sphere. Examples of these are the papers by Lotka (1926) or Hulme (1923) that advocated the use of quantitative techniques in the analysis of bibliographic production, although the perspective adopted was still somewhat rudimentary and was defined as “statistical bibliography”. (b) Coining of the term “bibliometrics” by Pritchard (1969) who defines it as: “the application of mathematics and statistical methods to books and other media of communication”, although there was still some ambiguity and words such as “biometrics” or “scientometrics” continued to be practically analogous (Lawani, 1981). (c) The further theorization of the discipline through the contributions of E. Garfield and D. J. Price (see, e.g., Price, 1976; Garfield, 2007). 3.1

Data and Methodology

Data for the present bibliometric analysis were retrieved from the Scopus database with a time span of 01/01/1979–07/31/2021. Subsequently, an extensive search was performed for bibliographic works of any kind, in English language, limited to the formal SCOPUS categories “Economics, Econometrics, and Finance” and “Business, Management, and Accounting”. The selector used was the TITLE-ABS-KEY (“Asymmetric Information”) order, i.e., by selecting from the database the term “Asymmetric Information”, i.e., limiting its inclusion in the title or abstract or keyword of each paper. It should be noted that after the initial collection of the bibliographic data, each record was checked and scrubbed since the metadata provided are not always correct and therefore it is necessary to “purify” them. In this sense, we have eliminated duplicate metadata or those that lacked a title or keywords, all of which would have distorted our analysis. Thus, as summarized in Table 6.1, 4,799 papers coming from 14 different types of documents (articles, books, conferences, letters, etc.) have been compiled. The methodological approach of Aria and Cuccurullo (2017) has been employed, which has been previously

Information asymmetry in economics and finance  123 Table 6.1

Stylized facts and figures of the research

Main information about data

Document type

Timespan 1979–2021

Article

Sources

Article review

975

Book

Documents 4,799

4,291 3

16

Average years from publication

11.7

Book chapter

114

Average citations per documents

22.42

Book chapter review

1

Average citations per year per doc

1.555

Book review

1

Conference paper

217

References 145,804

Editorial

6

  Erratum

2

Letter

2

Authors of single-authored documents 1,340

Note

2

Authors of multi-authored documents 5,777

Review

140

Single-authored documents

Review review

Authors Authors

7,117

Author Appearances

9,874

1,550

1

Short survey 3

Documents per Author 0.678 Authors per Document 1.48  

Document contents

Co-Authors per Documents Collaboration Index

2.05

1.77

Keywords Plus (ID)

3,916

Author’s Keywords (DE) 8,106

utilized with efficiency in research coming from very different fields of knowledge such as Artificial Intelligence (Alonso et al., 2018), Logistics (Agostino et al., 2020), Ecological Economics (Ballandonne, 2018), Heterodox Economics (Almeida and de Paula, 2019), or Tourism (Damayanti et al., 2017). Likewise, the methodological part has been structured as far as possible according to Raban and Gordon (2020), since the key aspects of the bibliographic research are specified jointly in this work. 3.2

Bibliometric Results

As shown in Figure 6.2, from a bibliometric point of view, work on asymmetric information reached its production peak twice: once in the early 1980s and again between 2000 and 2010. Today it could be said that this production remains constant, given that its volume is very similar to that of the 1980s, just when this research began to become popular. At the level of citations, the patterns are somewhat different; two consecutive peaks are reached between 1980 and 1990 and the highest in 2020. The five countries that have contributed most to the analysis of information asymmetry, taking into account the origin of the corresponding authors are: United States, China, United Kingdom, Germany and France, as detailed in Figure 6.3, which also shows that this is research in which work carried out in a single country predominates (Intra-country, SCP or Single Country Production) as opposed to those that were carried out in more than one (Inter-country, MCP or Multiple Country Production). On the other hand, SCP refers to bibliographic works belonging to authors from the same country, while MCP denotes those from more than one country. Note how the US is three times more productive than China, the next country in the ranking established in Figure 6.3, and how this fact can be explained by the fact that the papers coming from the US are those that have had the greatest impact on the study of information asymmetry.

124  The Elgar companion to information economics

Note: a. Different bibliographic production metrics (I); b. Different bibliographic production metrics (II).

Figure 6.2

Evolution of the scientific production and citations (1979–2021)

 

The collaborative networks that have been created between different countries are shown in the world map outlined in Figure 6.4, in which it can be verified that most of the research has as its original node the five countries mentioned above, in addition to others such as Brazil, Canada, Australia, Spain or South Africa. For their part, the most relevant key topics of the research are listed in Table 6.2: in addition to the term “asymmetric information” per se, there are others with which it is closely related, mainly those related to its more direct consequences

Information asymmetry in economics and finance  125

Figure 6.3

Most productive countries

such as “adverse selection”, “moral hazard”, “uncertainty”, “incentives” or “liquidity”. There are also others related to its modeling, such as “signaling”, “mechanism design”, “market microstructure”, or “game theory”. These terms are graphically represented in the word cloud shown in Figure 6.5.

Figure 6.4

Country collaboration map (1979–2021)

126  The Elgar companion to information economics

Figure 6.5

Keywords word cloud

Table 6.2

Most relevant keywords

Rk.

Author Keywords (DE)

1

ASYMMETRIC INFORMATION 1,741

Articles

Rk.

Author Keywords (DE)

Articles

11

CORPORATE GOVERNANCE

53

2

ADVERSE SELECTION

170

12

MARKET MICROSTRUCTURE

51

3

INFORMATION ASYMMETRY

153

13

D82

50

4

MORAL HAZARD

112

14

REPUTATION

40

5

SIGNALING

90

15

BARGAINING

39

6

REGULATION

81

16

SCREENING

39

7

MECHANISM DESIGN

72

17

INCENTIVES

37

8

GAME THEORY

61

18

UNCERTAINTY

37

9

CAPITAL STRUCTURE

56

19

INFORMATION

36

10

LIQUIDITY

54

20

COLLATERAL

35

 

The analysis of the original sources of the research highlights an aspect to be taken into account: it cannot be established that there is one publication that stands out from the others in the analysis of information asymmetry. In fact, Table 6.3 shows that the three most relevant publications in this field of knowledge (“Economic letters”, “Journal of Economy Theory”, and “Economic Theory”) oscillate in very close numbers. For greater detail, the bibliometric analysis includes a “who’s who” in the investigation of information asymmetry: Table 6.4 shows the top-10 papers with the most influential papers in terms of citations (TC = Total Citations, TC per Year = Total Citations per Year, and NTC = Normalized Total Citations) while Figure 6.6 details the contributions of the most productive authors throughout the study period.

 

Information asymmetry in economics and finance  127 Table 6.3

Most relevant sources

Rk.

Sources

Articles

Rk.

Sources

Articles

1

ECONOMICS LETTERS

102

11

INTERNATIONAL JOURNAL OF

52

2

JOURNAL OF ECONOMIC

97

12

JOURNAL OF MATHEMATICAL

INDUSTRIAL ORGANIZATION THEORY

52

ECONOMICS

3

ECONOMIC THEORY

90

13

MANAGEMENT SCIENCE

52

4

JOURNAL OF ECONOMIC

78

14

RAND JOURNAL OF ECONOMICS

46

73

15

APPLIED ECONOMICS

42

71

16

JOURNAL OF FINANCIAL

40

BEHAVIOR AND ORGANIZATION 5

EUROPEAN ECONOMIC REVIEW

6

JOURNAL OF BANKING AND FINANCE

7

JOURNAL OF PUBLIC

INTERMEDIATION 68

17

ECONOMICS 8

GAMES AND ECONOMIC

JOURNAL OF CORPORATE

37

FINANCE 65

18

REVIEW OF FINANCIAL STUDIES

37

59

19

AMERICAN JOURNAL OF

33

BEHAVIOR 9

JOURNAL OF FINANCIAL ECONOMICS

10

REVIEW OF ECONOMIC

AGRICULTURAL ECONOMICS 53

20

ECONOMETRICA

32

STUDIES

Table 6.4

Top manuscripts per citations

Rk.

Article

TC

 

NTC

1

Graham JR, Harvey CR. The theory and practice of corporate

2106

100.3

34.03

1535

49.5

16.85

1412

38.2

7.6

1060

46.1

23.4

1047

47.6

20.02

1044

52.2

22.91

873

27.3

12.23

823

20.6

4.83

810

45

21.85

685

20.1

9.96

finance: Evidence from the field. Journal of Financial Economics. 2001;60(2–3):187–243. 2

Harris M, Raviv A. The Theory of Capital Structure. Journal of Finance. 1991;46(1):297–355.

3

Miller MH, Rock K. Dividend Policy under Asymmetric Information. Journal of Finance. 1985;40(4):1031–51.

4

Coval JD, Moskowitz TJ. Home bias at home: Local equity preference in domestic portfolios. Journal of Finance. 1999;54(6):2045–73.

5

Kirmani A, Rao AR. No pain, no gain: A critical review of the literature on signaling unobservable product quality. Journal of Marketing. 2000;64(2):66–79.

6

Ritter JR, Welch I. A review of IPO activity, pricing, and allocations. Journal of Finance. 2002;57(4):1795–828.

7

Sharpse SA. Asymmetric Information, Bank Lending, and Implicit Contracts: A Stylized Model of Customer Relationships. Journal of Finance. 1990;45(4):1069–87.

8

Milgrom P, Roberts J. Predation, reputation, and entry deterrence. Journal of Economic Theory. 1982;27(2):280–312.

9

Klapper LF, Love I. Corporate governance, investor protection, and performance in emerging markets. Journal of Corporate Finance. 2004;10(5):703–28.

10

Glosten LR, Harris LE. Estimating the components of the bid/ask spread. Journal of Financial Economics. 1988;21(1):123–42.

128  The Elgar companion to information economics

Figure 6.6

Productivity of the most prolific authors over time

A key point of this study has been to carry out a taxonomy of the different lines of research. In this sense, a thematic map has been used, as illustrated in Figure 6.7, which employs the procedure of Cobo et al. (2011), based in turn on two differentiating parameters: density and centrality (Callon et al., 1983, 2005). According to this methodology, the topics studied and/or related to the information asymmetry can be classified as follows: ● Topics in the lower left quadrant. They are considered to be “underdeveloped and marginal”. Themes in this quadrant have low density and low centrality, presenting reduced centrality and density. They are referred to as “emerging or disappearing topics”. In this case, we would have “Asymmetric information”, “Costs”, and “Commerce”. ● Topics in the upper left quadrant. They are considered to “have well-developed internal linkages, but unimportant external linkages”. In this sense, they have been considered items with marginal importance given their high specialization. They are denoted as “Niche themes”. In this case, we would have the topic “Economics”. ● Topics in the upper right quadrant. They are considered to be “well developed and important for the structuring of a research field”. They are called “motor themes” of the research. In this case, none would exist as such. ● Topics in the lower right quadrant. These are considered “important to a field of research but are not developed”. Therefore, this quadrant is considered transversal, is called “basic topics”, and is related to future lines of research. In this case, we would have “Theoretical study”, “Modeling” and “Price dynamics”.

Information asymmetry in economics and finance  129

Source: Own elaboration.

Figure 6.7

4.

Thematic map based on co-word network cluster analysis

SUMMARY

Regarding information asymmetry, a well-known quote by N. Lee is perfectly applicable for the conclusion of this chapter: “Information is power. Disinformation is an abuse of power”. In fact, our work highlights the fact that all the information available to the participants in an economic-financial transaction does not always have to be accurate or perfectly shared, as pointed out by the paradigm of neoclassical economics which, in turn, is strengthened by economic theories of relative empirical evidence such as the Efficient Market Hypothesis. Interestingly, this perspective is quite close to other domains of knowledge such as International Relations, since the asymmetries currently existing between countries and geographical areas lead to an erroneous perception of risk, eroding the relations between nations (Womack, 2016). Similar effects are the information asymmetry in a market economy on economic exchanges, insofar as market failures can systematically occur. To some extent, the functioning of information asymmetry can be explained by “Black’s noise” (Black, 1986), that is, by an element of an intangible nature that transcends all levels of economic transactions and that, in one way or another, has the result that not all involved parties have the same level of information, distorting reality to the fullest extent. At a formal level, from a bibliometric point of view, our work shows that bibliographic production has been decreasing in recent times, but not the number of citations, which has been progressively increasing. Five countries have the highest bibliographic production (United States, China, United Kingdom, Germany, and France) and, in turn, these nations are the ones that have created the greatest number of collaborations at a global level. It is also noteworthy that this research is relatively evenly distributed among the top five positions in the ranking of publications that have chosen to include information asymmetry as a subject matter. In the same way, its transverse character can also be confirmed, as can be deduced from the predominant keywords. Something similar is also indicated by the articles with the highest number of citations, which cover multiple fields of the economic-financial world, such as: Bank lending

130  The Elgar companion to information economics (Sharpse, 1990), Market competition (Klapper and Love, 2004), Corporate finance (Graham and Harvey, 2001; Harris and Raviv, 1991; Ritter and Welch, 2002; Milgrom and Roberts, 1982), Dividends policy (Miller and Rock, 1985), Micromarkets/Supply and demand analysis (Glosten and Harris, 1988), Portfolio allocation (Coval and Moskowitz, 1999) and Product Quality (Kirmani and Rao, 2000). It is interesting to note that the main identifiable lines of research suggest that in a certain sense this study has to return to its origins, i.e. to focus on theoretical studies, the development of models, or the analysis of the dynamics of prices. The main limitation of this chapter results from the very nature of bibliometric models, i.e., the aim is to obtain the clearest possible picture of a given bibliographic base constructed ad hoc, however, the results can hardly be considered exhaustive, a factor that is one of the four challenges related to the elaboration of bibliometric reviews (Romanelli et al., 2021). This lack of exhaustiveness, typical of any bibliometric analysis, may lie in the choice of the scientific database used as a starting point, in the time horizon employed, or in the key terms used to articulate the database finally created. In this sense, it would have been interesting to distinguish between two fundamental areas of knowledge, economics, and finance, but this discrimination could be spurious since not all scientific databases apply the same specific criteria when it comes to including their records in each field of knowledge. For this reason, the Scopus database was used, since it is the scientific database with the largest number of registries. However, this constitutes one of the main handicaps of bibliometrics because each scientific database uses a different metadata structure, which hinders such joint analysis. Again arguing for completeness, the key term used (“information asymmetry”) is sufficiently representative to create a complete and robust database, regardless of the fact that it was probably used more frequently in the primary stages of the research. In addition, the elaboration of the database could have included Boolean searches of many other terms that, although not expressly synonymous with “information asymmetry”, are intimately related to it or belong to its scope of study such as “sender-receiver game”, “private information”,3 “cheap talk”4 (Crawford and Sobel, 1982; Kellner and Le Quement, 2018), “disclosure”5 (Grossman, 1981; Guttman et al., 2014; Milgrom, 2008; Verrecchia, 1983), “persuasion” (Beauchêne et al., 2019; Kamenica and Gentzkow, 2011; Milgrom, 2008) or “lemon markets” (Akerlof, 1970), just to name a few. However, since one of the goals of bibliometrics is to summarize a phenomenon to its minimum expression, we have chosen not to additionally include these terms but to rely on a concise database that succinctly manifests the most elementary characterizing facts of “information asymmetry” from a bibliometric perspective. Clearly, in the light of this chapter, new lines of bibliometric research may emerge. For instance, to delineate the behavioral economics of information asymmetries from models that contemplate the rational interaction between economic-financial agents (Kamenica and Gentzkow, 2011) to those that start from postulates that are close to bounded rationality (Beauchêne et al., 2019; Crawford and Sobel, 1982; Kellner and Le Quement, 2018).

NOTES 1. 2. 3.

Chapter 7 of this Companion alludes the repercussions of information asymmetry on financial markets (Pernagallo, 2024, pp. 135–153). Chapter 5 of this Companion alludes the concept of market failure (Giza, 2024, pp. 106–117). Chapter 4 of this Companion alludes the concept of private information (Włodarczyk, 2024, pp. 81–104).

Information asymmetry in economics and finance  131 4. 5.

Chapter 11 of this Companion alludes the concept of cheap talk (Pavesi et al., 2024, pp. 202–223). Chapter 10 of this Companion alludes the concept of disclosure (Li & Liu, 2024, pp. 185–201).

REFERENCES Agostino, R. S., Ristow, C., Morosini Frazzon, E., & Taboada Rodriguez, C. M. (2020). Perspectives on the Application of Internet of Things in Logistics. In M. Freitag, H. D. Haasis, H. Kotzab, & J. Pannek (Eds.), Dynamics in Logistics. LDIC 2020. Lecture Notes in Logistics (pp. 387–397). Cham: Springer. Akerlof, G. A. (1970). The Market for “Lemons”: Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics 84(3), 488–500. Akerlof, G. A. & Yellen, J. L. (1987). Rational Models of Irrational Behavior. American Economic Review 77(2), 137–142. Almeida, F. & de Paula, L. G. (2019). The Place of Uncertainty in Heterodox Economics Journals: A Bibliometric Study. Journal of Economic Issues 53(2), 553–562. Alonso, J. M., Castiello, C., & Mencar, C. (2018). A Bibliometric Analysis of the Explainable Artificial Intelligence Research Field. In J. Medina, M. Ojeda-Aciego, J. L. Verdegay, D. A. Pelta, I. P. Cabrera, B. Bouchon-Meunier, & R. R. Yager (Eds.), Information Processing and Management of Uncertainty in Knowledge-Based Systems: Theory and Foundations. IPMU 2018. Communications in Computer and Information Science, Vol. 853 (pp. 3–15). Cham: Springer International. Aria, M. & Cuccurullo, C. (2017). Bibliometrix: An R-tool for Comprehensive Science Mapping Analysis. Journal of Informetrics 11(4), 959–975. Bachelier, L. J.-B. A. (1900). Théorie de la spéculation. Annales Scientifiques de l’École Normale Supérieure 17, 21–86. Bachelier, L. J.-B. A. (1914). Le Jeu, la Chance et le Hasard. Paris: Ernest Flammarion. Bachelier, L. J.-B. A. (1915). La périodicité du hasard. L’Enseignement Mathématique 17, 5–11. Ballandonne, M. (2018). The Historical Roots (1880–1950) of Recent Contributions (2000–2017) to Ecological Economics: Insights from Reference Publication Year Spectroscopy. Journal of Economic Methodology 26(4), 307–326. Beauchêne, D., Li, J., & Li, M. (2019). Ambiguous Persuasion. Journal of Economic Theory 179, 312–365. Bernstein, J. (2005). Bachelier. American Journal of Physics 73(5), 395–398. Black, F. (1986). Noise. The Journal of Finance 41(3), 528–543. Black, F. & Scholes, M. (1973). The Pricing of Options and Corporate Liabilities. Journal of Political Economy 81(3), 637–654. Callon, M., Courtial, J.-P., & Laville, F. (2005). Co-Word Analysis as a Tool for Describing the Network of Interactions between Basic and Technological Research: The Case of Polymer Chemistry. Scientometrics 22, 155–205. Callon, M., Courtial, J.-P., Turner, W. A., & Bauin, S. (1983). From Translations to Problematic Networks: An Introduction to Co-Word Analysis. Social Science Information 22(2), 191–235. Clarkson, R. S. (1995). Some Observations on the Black-Scholes Option Pricing Formula. Transactions of the 5th AFIR International Colloquium. Vol. 3 (pp. 1125–1136). Brussels. Cobo, M., López-Herrera, A., Herrera-Viedma, E., & Herrera, F. (2011). An Approach for Detecting, Quantifying, and Visualizing the Evolution of a Research Field: A Practical Application to the Fuzzy Sets Theory Field. Journal of Informetrics 5(1), 146–166. Coval, J. & Moskowitz, T. (1999). Home Bias at Home: Local Equity Preference in Domestic Portfolios. Journal of Finance 54(6), 2045–2073. Crawford, V. P. & Sobel, J. (1982). Strategic Information Transmission. Econometrica 50(6), 1431–1451. Cuccurullo, C., Aria, M. & Sarto, F. (2013). Twenty Years of Research on Performance Management in Business and Public Administration Domains. Academy of Management Proceedings 2013, 14270. Damayanti, M., Scott, N. & Ruhanen, L. (2017). Coopetitive Behaviours in an Informal Tourism Economy. Annals of Tourism Research 65, 25–35. Fama, E. F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. Journal of Finance 25(2), 383–417.

132  The Elgar companion to information economics Foster, J. (2018). Bourgeoisie. In M. Vernengo, E. Perez Caldentey, & B. J. Rosser Jr. (Eds.), The New Palgrave Dictionary of Economics, 3rd edition (pp. 1038–1041). London: Palgrave Macmillan. Garfield, E. (2007). From the Science of Science to Scientometrics: Visualizing the History of Science with HistCite Software. Proceedings of the 11th ISSI International Conference (pp. 1–11). Madrid. Giza, W. (2024). Asymmetric Information as a Market Failure in Retrospect. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 106–117). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Glosten, L. R. & Harris, L. E. (1988). Estimating the Components of the Bid/Ask Spread. Journal of Financial Economics 21(1), 123–142. Glosten, L. R. & Milgrom, P. R. (1985). Bid, Ask and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders. Journal of Financial Economics 14(1), 71–100. Godin, B. (2006). On the Origins of Bibliometrics. Scientometrics 68, 109–133. Graham, J. & Harvey, C. (2001). The Theory and Practice of Corporate Finance: Evidence from the Field. Journal of Financial Economics 60(2–3), 187–243. Grossman, S. J. (1981). The Informational Role of Warranties and Private Disclosure about Product Quality. Journal of Law and Economics 24(3), 461–483. Guttman, I., Kremer, I., & Skrzypacz, A. (2014). Not Only What But Also When: A Theory of Dynamic Voluntary Disclosure. American Economic Review 104(8), 2400–2420. Harris, M. & Raviv, A. (1991). The Theory of Capital Structure. The Journal of Finance 46(1), 297–355. Hulme, E. W. (1923). Statistical Bibliography in Relation to the Growth of Modern Civilization: Two Lectures Delivered in the University of Cambridge in May 1922. Nature 112, 585–586. Jovanovic, F. (2001). Pourquoi l’hypothèse de marche aléatoire en théorie financière? Les raisons historiques d’un choix éthique. Revue d’économie financière 61(1), 203–211. Jovanovic, F., Andreadakis, S., & Schinckus, C. (2016). Efficient Market Hypothesis and Fraud on the Market Theory: A New Perspective for Class Actions. Research in International Business and Finance 38, 177–190. Kamenica, E. & Gentzkow, M. (2011). Bayesian Persuasion. American Economic Review 101(6), 2590–2615. Kellner, C. & Le Quement, M. T. (2018). Endogenous Ambiguity in Cheap Talk. Journal of Economic Theory 173, 1–17. Kirmani, A. & Rao, A. R. (2000). No Pain, No Gain: A Critical Review of the Literature on Signaling Unobservable Product Quality. Journal of Marketing 64(2), 66–79. Klapper, L. F. & Love, I. (2004). Corporate Governance, Investor Protection, and Performance in Emerging Markets. Journal of Corporate Finance 10(5), 703–728. Kotowitz, Y. (2018). Moral Hazard. In M. Vernengo, E. Perez Caldentey, & B. J. Rosser Jr. (Eds.), The New Palgrave Dictionary of Economics, 3rd edition (pp. 9131–9132). London: Palgrave Macmillan. Lawani, S. M. (1981). Bibliometrics: Its Theoretical Foundations, Methods and Applications. Libri 31, 294–315. Ledyard, J. O. (2018). Market Failure. In M. Vernengo, E. Perez Caldentey, & B. J. Rosser Jr. (Eds.), The New Palgrave Dictionary of Economics, 3rd edition (pp. 8246–8250). London: Palgrave Macmillan. Li, M. & Liu, T. (2024). Disclosure of Conflicts of Interest: Theory and Empirics. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 185–201). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Löfgren, K.-G., Persson, T., & Weibull, J. W. (2002). Markets with Asymmetric Information: The Contributions of George Akerlof, Michael Spence and Joseph Stiglitz. The Scandinavian Journal of Economics 104(2), 195–211. Lotka, A. J. (1926). The Frequency Distribution of Scientific Productivity. Journal of the Washington Academy of Sciences 16(12), 317–323. Mandelbrot, B. (1963). The Variation of Certain Speculative Prices. The Journal of Business 36(4), 394–419. Martín-Cervantes, P. A. (2020). Hacia un modelo estocástico eficiente para la valoración de activos financieros basado en el volumen de negociación: fundamentos teóricos e implementación práctica, Vol. 370. EDUAL, Almería.

Information asymmetry in economics and finance  133 Martín-Cervantes, P. A., Valls Martínez, M. d. C., & Cruz Rambaud, S. (2021). Corporate Social Responsibility: A Bibliometric Research. In D. C. Poff & A. C. Michalos (Eds.), Encyclopedia of Business and Professional Ethics. Cham: Springer. Milgrom, P. (2008). What the Seller Won’t Tell You: Persuasion and Disclosure in Markets. The Journal of Economic Perspectives 22(2), 115–132. Milgrom, P. & Roberts, J. (1982). Predation, Reputation, and Entry Deterrence. Journal of Economic Theory 27(2), 280–312. Miller, M. H. & Rock, K. (1985). Dividend Policy under Asymmetric Information. The Journal of Finance 40(4), 1031–1051. Muth, J. F. (1961). Rational Expectations and the Theory of Price Movements. Econometrica 29(3), 315–335. Pavesi, F., Scotti, M., & Argelli, N. (2024). Information and Expertise. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp.  202–223). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Pernagallo, G. (2024). Overcoming Asymmetric Information: A Data-Driven Approach. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 135–153). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Postlewaite, A. (2018). Asymmetric Information. In M. Vernengo, E. Perez Caldentey, & B. J. Rosser Jr. (Eds.), The New Palgrave Dictionary of Economics, 3rd edition (pp. 510–513). London: Palgrave Macmillan. Price, D. J. d. S. (1976). A General Theory of Bibliometric and Other Cumulative Advantage Processes. Journal of the American Society for Information Science 27(5), 292–306. Pritchard, A. (1969). Statistical Bibliography or Bibliometrics? Journal of Documentation 25(4), 348–349. Raban, D. R. & Gordon, A. (2015). The Effect of Technology on Learning Research Trends: A Bibliometric Analysis Over Five Decades. Scientometrics 105, 665–681. Raban, D. R. & Gordon, A. (2020). The Evolution of Data Science and Big Data Research: A Bibliometric Analysis. Scientometrics 122, 1563–1581. Read, C. (2012). Discussion and Applications: Einstein and Bachelier. In C. Read (Ed.), The Efficient Market Hypothesists: Bachelier, Samuelson, Fama, Ross, Tobin and Shiller (pp. 33–43). Basingstoke: Palgrave Macmillan. Ritter, J. R. & Welch, I. (2002). A Review of IPO Activity, Pricing, and Allocations. The Journal of Finance 57(4), 1795–1828. Romanelli, J. P., Pereira Gonçalves, M. C., de Abreu Pestana, L. F., Hitaka Soares, J. A., Stucchi Boschi, R. & Fernandes Andrade, D. (2021). Four Challenges When Conducting Bibliometric Reviews and How to Deal with Them. Environmental Science and Pollution Research 28(43), 60448–60458. Rosser, J. B. (2003). A Nobel Prize for Asymmetric Information: The Economic Contributions of George Akerlof, Michael Spence and Joseph Stiglitz. Review of Political Economy 15(1), 3–21. Samuelson, P. A. (1965a). Proof that Properly Anticipated Prices Fluctuate Randomly. Industrial Management Review 6(2), 41–49. Samuelson, P. A. (1965b). Rational Theory of Warrant Pricing. Industrial Management Review 6(2), 13–39. Sandmo, A. (1999). Asymmetric Information and Public Economics: The Mirrlees-Vickrey Nobel Prize. The Journal of Economic Perspectives 13(1), 165–180. Shapiro, F. R. (1992). Origins of Bibliometrics, Citation Indexing, and Citation Analysis: The Neglected Legal Literature. Journal of the American Society for Information Science 43(5), 337–339. Sharpse, S. A. (1990). Asymmetric Information, Bank Lending, and Implicit Contracts: A Stylized Model of Customer Relationships. The Journal of Finance 45(4), 1069–1087. Simon, H. A. (2018). Behavioural Economics. In M. Vernengo, E. Perez Caldentey, & B. J. Rosser Jr. (Eds.), The New Palgrave Dictionary of Economics, 3rd edition (pp.  846–853). London: Palgrave Macmillan. Stewart, I. (2012). Seventeen Equations that Changed the World. London: Profile Books. Stiglitz, J. E. (1990). Symposium on Bubbles. The Journal of Economic Perspectives 4(2), 13–18. Stiglitz, J. E. (2000). The Contributions of the Economics of Information to Twentieth Century Economics. The Quarterly Journal of Economics 115(4), 1441–1478.

134  The Elgar companion to information economics Stiglitz, J. E. (2002). Information and the Change in the Paradigm in Economics. American Economic Review 92(3), 460–501. Verrecchia, R. E. (1983). Discretionary Disclosure. Journal of Accounting and Economics 5, 179–194. Smoluchowski, M. (1906). O pewnym zagadnieniu z teorii sprężystości i jego związku z wytwarzaniem się gór fałdowych. Rozprawy Wydziału matematyczno-przyrodniczego Polskiej Akademii Umiejętności w Krakowie 49, Seria A, 223–236. Wakker, P. P. (2018). Uncertainty. In M. Vernengo, E. Perez Caldentey, & B. J. Rosser Jr. (Eds.), The New Palgrave Dictionary of Economics, 3rd edition (pp. 13969–13981). London: Palgrave Macmillan. West, E. G. (2018). Monopoly. In M. Vernengo, E. Perez Caldentey, & B. J. Rosser Jr. (Eds.), The New Palgrave Dictionary of Economics, 3rd edition (pp. 9099–9104). London: Palgrave Macmillan. Włodarczyk, J. (2014). Nonneutrality of Money in a Social Perspective. Economics & Sociology 7(2), 199–208. Włodarczyk, J. (2015). Money as a Network Good. Journal of Economics and Management 20(2), 199–208. Włodarczyk, J. (2024). Information and Income Distribution: The Perspective of Information Economics. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 81–104). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Womack, B. (2016). Asymmetry and International Relationships. Cambridge: Cambridge University Press.

7. Overcoming asymmetric information: A data-driven approach Giuseppe Pernagallo1

1. INTRODUCTION Asymmetric information is a long-standing theme in the economic literature, with many of the most influential works in economics published in this field (e.g., Akerlof, 1970; Spence, 1973; Rothschild & Stiglitz, 1976; Stiglitz & Weiss, 1981). Before the massive spread of information technology, obtaining information was difficult and asymmetries could undermine market functioning. In this scenario, information was a valuable commodity, and its cost of acquisition and processing was not negligible. With the advent of the “data revolution”, the amount of information available has grown exponentially, to the point that today we are no longer faced with a problem of scarce and costly information, but of managing huge amounts of data. Innovation, especially in information technology, is radically changing the way we approach this problem. In particular, the rise of artificial intelligence is changing the branches of the economics of information and knowledge, making it easier to access, screen and process massive amounts of data. It is inevitable that this will change the way information is accessed and exploited, so it is very likely that soon in many markets where asymmetric information was considered an inherent feature, the problem will be solved through technology. However, there are many other situations where technological innovation itself generates new asymmetries as, for example, in personal data markets. Artificial intelligence is the key to better understanding our current limitations and possibilities for making relevant information accessible to all. The contribution of this chapter is twofold. First, I show that, in many cases, the problem of asymmetries can become marginal, thanks to the advancement of data collection systems and artificial intelligence tools that can handle huge streams of data. In these situations, traditional approaches based on mere theoretical models should be replaced or supplemented by data-driven approaches such as machine learning. Second, I provide a demarcation criterion that allows us to distinguish in which cases artificial intelligence and machine learning can really solve the problem and in which they cannot. This demarcation criterion is based on the idea of monopolies of knowledge, which is the content of section 4. The rest of the chapter is structured as follows. Section 2 introduces concepts useful for understanding the argumentation of this chapter, while section 3 presents old and new asymmetric information problems and applications of machine learning and artificial intelligence to solve them. The last section concludes the chapter.

135

136  The Elgar companion to information economics

2.

BASIC CONCEPTS

2.1

Information Asymmetry

Information asymmetry (also referred to as “imperfect information”) occurs when some agents are endowed with “more” or “better” information than the others, which is associated with undesirable situations such as moral hazard or adverse selection. The issue is so relevant because it affects many important markets, such as the credit market (e.g., Stiglitz & Weiss, 1981; Bester, 1994), the labour market (e.g., Spence, 1973), or the insurance market (Rothschild & Stiglitz, 1976). In undergraduate courses, the recurring example is the market for used cars (“lemons”), which was target of Akerlof’s famous paper (Akerlof, 1970) and a prime example of adverse selection. In this market, the fact that the potential buyer cannot observe the quality of the car puts the seller in a privileged position. Consequently, bad cars and good cars are sold at the same price, pushing good cars out of the market since they are valued by their sellers more than the market price. A typical example of moral hazard involves agency relationships. An agency relationship is “a contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision making authority to the agent” (Jensen & Meckling, 1976, p. 308). The agency problem (or principal-agent problem) revolves around the fact that the principal wants to induce the agent to behave as if she is maximizing the principal’s welfare. In this case, we may have hidden information or hidden actions problems (Mas-Colell et al., 1995), the latter known as moral hazard since the principal cannot observe the agent’s effort after signing the contract. Indeed, the timing of when the asymmetry occurs is another useful demarcation criterion between adverse selection and moral hazard, since in adverse selection the asymmetry exists at the time the contract is made, whereas in moral hazard it is after the contract is signed. 2.2

Artificial Intelligence and Machine Learning

The term machine learning generally indicates the automated detection of relevant patterns in data (Shalev-Shwartz & Ben-David, 2014). In simple terms, machine learning aims to allow the machine itself to perform tasks decided by the programmer. The meaning of the word “learning” can be illustrated with an example from the animal world. When some animal species are baited with unfamiliar types of food, they first taste a little bit of the new food and then, based on their physiological response, decide whether it is worth eating it all. If that small amount of food causes disease, the animal learns to avoid that new food. By associating the characteristics of the new food with a dangerous or repulsive food, the animal can avoid, for example, food poisoning. With machine learning, attempts are made to reproduce this behaviour using computers, thereby automating procedures and tasks. The main tasks of machine learning are prediction (regression) and classification, which are examples of supervised learning, and clustering (or grouping tasks), which falls under unsupervised learning (Athey, 2019). Following an established definition (James et al., 2021), in supervised learning we have for each observation of the independent variable(s), xi, i = 1, 2,…, n, an observation of the associated outcome variable, yi, with n being the sample size. The purpose is to fit a model that relates the outcome variable to the covariates to predict values of the response variable for future observations (prediction) or simply to study the relationship

Overcoming asymmetric information  137 between the outcome variable and the predictors (inference). If we let Y be an n × 1 column vector containing all the observations of the dependent variable, and X be the n × p matrix of covariates, with p the number of covariates, then we want to construct an estimator of the conditional expectation, ​μ​(​x)​ ​  =  𝔼​[​Y|​​X   =   x​]​​, to make predictions as close as possible to the true values of Y in an independent dataset. Observations are generally assumed to be independent, and data are split in two sets: a training and a test set. The training dataset or training set is the set of data used to train the algorithm. Only part of the data is used to allow the algorithm to learn the relevant features of the data. Once we have obtained a learner (i.e., a machine learning algorithm), we test its performance on the other unused portion of the data, called the test dataset or test set. The test set contains elements that are not known by the model, so predictions about the test set allow the accuracy of the model to be properly assessed. Others prefer to articulate the learning in three phases by adding a phase between the training and the test stages. The validation set can be used instead of the test set to evaluate the performance of the algorithm, while the test set in this subdivision has the role of applying the model to real-world data. This division is necessary to prevent the algorithm from learning and evaluating on the same data, which in turn would cause several biases. The main problem with not splitting data into sets is overfitting, which means that the model follows errors, or noise, too closely (James et al., 2021). A simple rule for dividing the data into training and test sets is the Pareto principle or 80/20 rule, which is to assign 80 per cent of the observations to the training set, and the remaining 20 per cent of the observations to the test set. This rule is quite common in many applied works (Koch, 1999; Pernagallo & Torrisi, 2020; Pernagallo et al., 2021; Warnat-Herresthal et al., 2021). In the case of classification, algorithms are used to classify observations. For example, given a dataset with spam emails and ham emails (emails wanted by the recipient), we can construct an algorithm that, based on a set of features, classifies new emails into spam or ham. In this case the estimation problem is to estimate the probability ​P​(​Y  =  k​|​ X  =  x​)​​ for each k = 1,…, K possible realizations of Y. The best model should ideally minimize the deviations between actual and predicted outcomes, so we can use different selection criteria, such as information criteria (e.g., Akaike information criterion or Bayes information criterion) or the percentage of correct predictions. When K = 2, we have a classification problem with a dummy or binary dependent variable. In section 3.3 we present many applications of this classification problem applied to information asymmetry. Famous supervised algorithms include regression, support-vector machines, or decision trees. In unsupervised learning, a learning algorithm provides unlabelled data that the machine tries to define by extracting features and patterns on its own, unlike supervised learning algorithms that learn on a labelled dataset (in the sense that the data are ingested in the form of input-output pairs). Clustering, a primary example of unsupervised learning, involves dividing the data into a number of clusters within which the data are more similar to each other than those in the other clusters. Other examples of unsupervised learning are tools for dimension reduction or matrix completion (e.g., Principal Component Analysis). Thus, the word “unsupervised” is used because hidden patterns in the data are discovered without human intervention. The demarcation between supervised and unsupervised learning is not always so clear as in the case of the so-called semi-supervised learning problems (see Engelen & Hoos, 2020). Machine learning is a branch of Artificial Intelligence (AI), the latter defined by IBM2 as “the broadest term used to classify machines that mimic human intelligence”. There are several definitions of AI, each covering different aspects (thought and reasoning processes, or

138  The Elgar companion to information economics behaviour), for which the interested reader can consult Russell and Norvig’s book (Russell & Norvig, 2010). The goal of AI is to predict, automate, and optimize human tasks faster and, ideally, achieve higher levels of accuracy. In simple terms, AI aims to reproduce human intelligence by means of machines. The term “data-driven approaches” used in this chapter refers primarily to AI tools and techniques. Scientific production on AI and machine learning has increased exponentially over the past 50 years (Figure 7.1), and economics is no exception. Indeed, both AI and machine learning are increasingly being applied to various economic problems (Varian, 2014; Mullainathan & Spiess, 2017; Athey, 2019),3 and in this chapter we focus on its application to asymmetric information problems.

Source: Lens.org.

Figure 7.1

2.3

Count of scholarly works (e.g., papers, book chapters, etc.) on machine learning (left panel) and artificial intelligence (right panel) from 1950 to 2020

Data, Information, and Knowledge

We need to clarify at this stage the difference between data, information, and knowledge. According to basic information theory, data are the original representation of a phenomenon, while information is the result of interpreting, filtering, and organizing data. Finally, abstraction and information processing provide knowledge. Knowledge is thus the result of sorting, processing, evaluating, and making sense of data and signals, screening between reliable and false signals, and on aggregating them to the existing stock of information (Antonelli, 2018). According to the Bayesian approach of Arrow (Arrow, 1969, 1996), signals constitute information, they are not given, and their knowledge changes the (conditional) probability distribution of an unknown state of the world, driving the value realized by the economic agent’s action. To help the reader better understand the distinction, we can provide a simple example. Take the vector Age = (35, 21, 23, 44, 37, 50), this is a clear example of data, i.e., a collection of text, numbers, or symbols in raw or unorganized form. We can then interpret these data by

Overcoming asymmetric information  139 assigning a meaning or a context, for example, by saying that these numbers represent the age of six individuals: we now have information. Finally, we can use this information to gain knowledge, for example by saying that the oldest individual in our dataset is 50 years old. In economics, the distinction between the three terms is sometimes very blurry. The three words are usually used without a clear demarcation in some econometric work. The difference does not seem to be of much interest to economists, yet it is crucial. On this issue, the interested reader is referred to Boisot and Canals (2004). For a more in-depth analysis of the three terms, the interested reader is referred to other works (e.g., Stock & Stock, 2013). Another important distinction to make is that between knowledge, technological knowledge, and scientific knowledge. The boundary between these concepts is not easy to highlight. In simple terms, “technological knowledge” and “scientific knowledge” can be seen as subsets of the larger set of “knowledge”. The two subsets sometimes intersect, but they are distinct. The key difference is related to the adjective “technological”, which is defined by the online Cambridge Dictionary as “relating to, or involving, technology”. Instead, “technology” is defined by the same dictionary as “(the study and knowledge of) the practical, especially industrial, use of scientific discoveries”. On the other hand, “scientific” is about science, in the sense of knowledge provided, for example, by scholars or researchers. The adoption of information and communication technologies (ICTs) has reduced the separation between scientific and technological knowledge, reducing barriers to access to knowledge and fostering the interaction between science and technology (Antonelli, 2017). The gap between scientific and technological knowledge has narrowed even further with the adoption of artificial intelligence. Scientific and technological knowledge producers can speak the same algorithmic language and can access and process unprecedented amounts of data through big data analytics, changing the way new knowledge is generated. 2.4

AI and Knowledge Generation

AI is hardly a new idea, as it dates back to the 1950s. The real radical innovation is the development of computational systems and skills that have made the application of AI effective and pervasive. AI is influencing the knowledge generation process by reducing its cost and increasing its spillovers. As with information and communication technologies (Antonelli et al., 2000; Antonelli, 2017), AI allows the user to acquire massive amounts of external knowledge while reducing the cost of collecting and processing it. Unlike information and communication technologies, AI has made it possible to access new knowledge that was precluded prior to its introduction and enhancement. For example, in 2019, for the first time, astronomers were able to produce an image of a black hole.4 The researchers developed a series of algorithms that converted the telescopic data into this historical image, a process that required the collection of about five petabytes of data.5 Without modern computing power and advances in data analysis techniques, obtaining this knowledge would have been impossible. The fact that AI can collect and process huge amounts of data is one of its salient features as it allows agents to make intensive use of external knowledge surpassing any other available alternative. If new knowledge is generated by recombining existing knowledge (Weitzman, 1996), then AI offers unprecedented possibilities in the knowledge generation process as the same algorithm can be implemented in countless applications. The polymorphic nature of AI algorithms can generate knowledge in a variety of fields, from finance to medicine, from

140  The Elgar companion to information economics biology to the arts, at no cost other than the initial creation of the mathematical and computer framework. This has tremendous repercussions on society as production, distribution and use of knowledge is the foundation for modern economic growth (Geuna, 1999; Antonelli, 2019). Section 3 shows how AI can turn data into valuable knowledge in many situations. Thus, AI can generate knowledge in financial markets from microeconomic data on borrowers and stocks to assess the riskiness of future borrowers and companies to be financed. The same mechanism can be used to evaluate candidates for a job position in the labour market, or to assess the quality and reliability of news or news creators. Electronic markets are a source of huge datasets, many times in unstructured form, which can be processed using, for instance, text mining techniques. Finally, AI also brings value to Public Administration, for instance, by making government action more transparent and citizens more informed.

3.

A NEW PARADIGM: AI AND INFORMATION ASYMMETRY

3.1

The Data Revolution

Big data processing undermines the generation of new knowledge, as humans can only process a limited amount of data (Pernagallo & Torrisi, 2022). Instead, big data makes AI applications effective, because the more data there is, the more accurate the results produced by the machines (Bag et al., 2021). Two relevant problems in applying AI algorithms in knowledge extraction are the black box syndrome and the interpretation of results. The first problem occurs because AI allows for standardization of procedures, in the sense that once the best algorithm has been identified, it is sufficient to feed the data to the machine to produce a result. This obviously severely penalizes the robustness of the knowledge production process, as researchers must rely on some sort of invisible action. The second problem is more extensive as it concerns the interaction between AI and the individuals producing knowledge, such as researchers or practitioners. Turning the results of AI algorithms, the raw computer output, into valuable knowledge is a difficult process. There are two pivotal figures who play this important role. Researchers and practitioners act as interpreters of the language of AI. Researchers use and develop AI techniques to produce scientific results, which can then be converted into something practical for society (on this issue, see Chubb et al., 2021). Practitioners use and develop AI techniques to perform their tasks within a company or to earn a profit by selling algorithms and their knowledge to other individuals. The two figures many times are closely related or may even coincide. The fact that the field is very active nowadays for both figures is demonstrated by the large number of academic papers published in recent years (see Figure 7.1) and the high number of patents (e.g., Balland & Boschma, 2021). The potential of AI is not limited to the extraction of data, information and knowledge from common sources or mainstream literature. Potentially, everything that can be turned into input for the machine can be sifted and processed by AI algorithms and turned into knowledge. Some examples are the so-called grey literature (which is less used in academic research), images, or even audio. The real problem today is not finding data mines but finding capable data miners. Clearly, AI can create “public” knowledge when agents have free and unrestricted access to data, information, and existing knowledge. When this is not possible, sources of asymmetry may arise.

Overcoming asymmetric information  141 3.2

The Relationship between Innovation and Imperfect Information: Old and New Problems

Asymmetric information can influence innovation in many ways. The credit market is a common example of a market with asymmetric information (e.g., Stiglitz & Weiss, 1981; De Meza & Webb, 1987; Bester, 1994) where information plays a crucial role as access to credit promotes entrepreneurship and innovation (Antonelli, 2019). This market is asymmetric because a borrower usually has better information about the potential returns and risks associated with the investment projects than the lender (Mishkin & Eakins, 2018). Formally, there is a distribution of borrowers, e.g., firms or consumers, each with a given probability of success, p, representing the probability of being able to repay the debt. The interest rate should act as a screening mechanism to separate risky borrowers from safe ones. With perfect information, riskier borrowers with low p should expect higher interest rates; unfortunately, lenders cannot observe the riskiness of borrowers, and this introduces a source of asymmetry. This lack of information can lead to credit rationing (Stiglitz & Weiss, 1981), which causes society to miss the opportunity to finance many worthy entrepreneurial projects. On the other hand, the academic job market is a good example of how moral hazard can affect innovation. Every researcher’s dream is to obtain tenure; however, the system with tenure is easily exposed to moral hazard (e.g., McDonald, 1974; Byford, 2017), pre- and post-tenure. Pre-tenure, a newly appointed researcher may decide to pursue a research agenda without significant contributions that involve high risk-taking (Byford, 2017). Post-tenure, the researcher may lose the incentive to do meaningful research since they have acquired lifetime job security. In either case, agents may produce less innovative research (for a broader model on the relationship between imperfect information and innovation in the research market, see Pernagallo, 2023). The relationship between innovation and asymmetric information is not unidirectional, as innovation also affects asymmetric information. Technological innovation is slowly solving the asymmetry problem in many situations due to the large use of data collection systems.6 We can provide several examples. One of the most common solutions to the problem of adverse selection in the market for used cars is the adoption of warranties, which protect the buyer in the event of fraudulent behaviour by the seller or simply if the product does not conform to the description. Nevertheless, a simple data-driven approach is enough to solve (or mitigate) the problem. The odometer reading is sufficient to provide the buyer with important information about the state of the car. The future seems to be even more transparent, as many vehicles are now equipped with an event data recorder (or black box) that collects various useful information such as speed, braking, or even location, which could be used to determine as precisely as possible the condition of the vehicle at the time of sale. Economic theory predicted that information asymmetry could lead to a collapse of the market for used cars; today, this market is more than double that of new cars, with about 39 million used cars sold in 2020 in the United States versus 14 million new cars.7 A similar approach could be used in the insurance market, where moral hazard problems are recurrent, to ensure that the customers behave properly after signing an insurance contract. Another interesting example comes from the monitoring systems employed by large companies. Monitoring is a possible solution to the principal-agent problem in the case of hidden actions because it encourages the agent to increase effort (e.g., Jost, 1991). However, monitoring has generally been considered costly, especially in large companies where the high number

142  The Elgar companion to information economics of employees makes it difficult to implement monitoring of their actions. The advent of new technological systems, such as video cameras, has greatly reduced this problem. As documented by the CNBC,8 according to a 2018 survey by Gartner, 22 per cent of organizations worldwide in various industries are using employee-motion data, 17 per cent are monitoring work-computer-usage data, and 16 per cent are using Microsoft Outlook- or calendar-usage data. This is the case with companies such as Walmart or UPS, which, thanks to sensors, can collect data such as scanner beeps at the checkout, chats between employers and customers, or the status of the vehicle being driven. Processing such a large amount of data requires the application of appropriate statistical techniques and has, de facto, reduced the prominence of traditional approaches (e.g., game theoretic models). Unfortunately, although technological innovation is solving many old asymmetric information problems, it has simultaneously created new ones. These data gathering systems are considered an invasion of privacy, which raises serious concerns about their legitimacy. In addition, there are many other situations where our data are being collected, think of data collected by smartphones and other technological devices, or by social media platforms. This has created a marketplace for our personal data where companies sell this data to third parties without the consumer’s knowledge, creating a new source of asymmetry because “data-driven companies collect much more personal data than the consumer knows or can reasonably oversee, and data-driven companies have much more (technical) information about how this data is processed than consumers would be able to understand” (Waerdt, 2020). Another new source of information asymmetry is provided by the news market, which has been profoundly changed by the daily use of the Internet and social networks. In this case, the relevant information is the veracity of a news story shared by a news website, which is not revealed to consumers. Some authors have shown that in this market, competition is not very effective in promoting news accuracy, and that this asymmetry can lead to the unintended effect of “bad” news (e.g., fake news) replacing “good” news (Mullainathan & Shleifer, 2005; Pernagallo et al., 2021). Increasing returns triggered by economies of density and network externalities characterize both the generation and exploitation of technological knowledge and the use of the array of digital technologies. Large firms enjoy the benefits triggered by the use of the very same software to large stocks of data and information. The larger the number of data-rich interactions and transactions, the larger the amount of information that can be extracted and the larger the amount of knowledge that can be generated and exploited. Increasing returns associated with digital technologies and the generation of technological, commercial, financial, and economic data are a new source of information asymmetries: evidence from platform and grid corporations provides substantial support for this aspect (Rikap, 2022). 3.3

Applications of Machine Learning and AI with Imperfect Information

The creation of data-driven markets poses the need to address asymmetric information using a different paradigm. Artificial Intelligence offers a framework that can be applied to both traditional asymmetric information problems and new ones. The following is a review of some of the potential applications.

Overcoming asymmetric information  143 3.3.1 Financial markets As discussed above, asymmetric information could cause credit rationing, limiting the innovation potential of the economic system. One device commonly used in response to asymmetric information in the credit market is the collateral, or the pledge of the borrower’s properties to the lender, to ensure repayment of a loan (e.g., Bester, 1987, 1994). However, the introduction of a collateral would not eliminate the possibility of credit rationing (Stiglitz & Weiss, 1992). Traditionally, research on this topic has dealt with microeconomic models, which has two major shortcomings. First, they are (for the most part) purely theoretical, which has certainly been the dominant paradigm in the past, but which reveals little appeal in the era of the data revolution. Second, the models tacitly assume that borrowers have the relevant information or, in other words, that they know, more or less, their riskiness. There are many reasons why this might not be the case. In fact, several biases may influence the borrower’s judgement, such as overconfidence (e.g., Invernizzi et al., 2016), and borrowers are not always able to properly assess their projects due to a lack of adequate economic expertise, which can be a cause of over-indebtedness (e.g., Gutiérrez-Nieto et al., 2016). Nowadays we have enough data and statistical tools to change approach and try to solve the problem of information asymmetry using data analysis. Machine learning offers a quantitative framework that allows us to compute the probability of success (or default) of borrowers without relying on theoretically rigid assumptions. The problem of estimating the probability of success is a problem of classification. These problems are extensively studied in the machine learning literature and can be easily adapted to economic contexts. These approaches can be used, among the others, by banks to estimate credit risk (Khandani et al., 2010), or by insurance companies for risk prediction (Boodhun & Jayabalan, 2018). The usual approach is to obtain microdata on borrower default, consisting of a dummy dependent variable, Default, and independent variables, such as personal information about the borrower (age, income, marital status, etc.). Then the researchers use machine learning algorithms to perform the task of correctly classifying borrowers who are in default and those who are not, so that the algorithm can be used in the future to evaluate the default risk of new applicants. Famous algorithms in literature are the simple logit model (e.g., Ge et al., 2017) or more sophisticated approaches like random forests, extreme gradient boosting, and neural networks (e.g., Xu et al., 2021). Data-driven credit scoring systems are already in use. For example, Dun & Bradstreet, a well-known company that provides business data, analysis, and insights for companies, already offers automated solutions for credit decisions based on several variables.9 The benefits of this approach are considerable, and the following is a partial list of them. 1. 2. 3. 4.

We have an accurate measure of the borrower’s probability of success (or default). This probability is computed “objectively” by algorithms based on available data. This method can deal with big data. We do not have to make rigid theoretical assumptions (for example, agents are all risk-neutral and described by the same utility function). 5. Greater degree of flexibility. It is very difficult to move from one theoretical model (made of many assumptions) to another, because even changing one assumption can lead to different conclusions. Consider Stiglitz and Weiss (1981) who use second-order stochastic dominance in their model and conclude that the credit market can experience credit rationing, while De Meza and Webb (1987) use first-order stochastic dominance and arrive at

144  The Elgar companion to information economics the opposite conclusion of excessive lending. On the other hand, changing one algorithm to another requires less effort. 6. We can determine “objectively” the best model, for example, by using information criteria or the percentage of correctly predicted cases. One of the main flaws is that this approach can easily become a black box approach, which can lead researchers to neglect investigating the economic logic behind the observed phenomena. In addition, AI also raises some ethical issues related to problems of unjustified actions, opacity, bias, discrimination, information privacy, and moral responsibility (Mittelstadt et al., 2016). Unjustified actions can arise because highlighting causal relationships is very difficult and relying on inductive correlations can have a negative impact on humans in terms of actions taken based on algorithms. Algorithms can also be opaque because of their lack of accessibility and comprehensibility, a condition exacerbated by the need to keep the mechanism behind a method secret to preserve its profitability. The fact that AI can be biased follows from the fact that the developer can be biased; in fact, many choices are made during the process of algorithm development that can lead to certain outcomes and that can also naturally result in discrimination, for example, based on gender or ethnicity. We also pointed out how accessibility to our personal data can generate asymmetries and unregulated markets, and the use of AI algorithms contributes to this problem as it allows easier use and extraction of this data. Finally, AI poses a liability problem, since in the event of an error it is unclear who should be blamed. This is certainly not a comprehensive review of all AI-related problems, so we refer the interested reader to other works (e.g., Allen et al., 2006; Mittelstadt et al., 2016; Hagendorff & Wezel, 2020). Asymmetric information in equity markets has also been vastly investigated in the literature (Narayanan, 1988; Dierkens, 1991; Welker, 1995; Park et al., 2021). The asymmetry arises because managers should normally have an information advantage over the market, so lenders do not know whether they are financing a worthy firm or a “lemon”. As noted by Hubbard (1990, p. 2), “For equity finance, new shareholders demand a premium to purchase the shares of relatively good firms to offset the losses arising from funding lemons […] This premium raises the cost of new equity finance faced by managers of relatively high-quality firms above the opportunity cost of internal finance faced by existing shareholders”. If there are good and bad companies issuing stocks in the equity market, machine learning can be used to choose the best stocks to invest in. The field is receiving increasing interest in recent years. For example, Gu et al. (2020) provide a comparative analysis of various machine learning tools to assess the problem of measuring asset risk premiums, which is the canonical problem in empirical asset pricing. The authors show that the use of machine learning-based predictions, using decision trees and neural networks, yields great advantages over canonical methods, allowing investors to earn large economic gains. Although data-driven approaches have the potential to mitigate asymmetry in financial markets, they are not enough to completely solve the problem. In fact, there are many other factors that machines cannot evaluate. When a bank finances an entrepreneur, elements such as talent or even luck cannot be evaluated by algorithms, so human evaluation is still a relevant component in the process. The same is true in the stock market, where private information cannot be fully disclosed, otherwise informed traders could gain huge advantages over other market participants. For this reason, there are strict rules on insider trading that highlight the key role of the policy maker in this market.

Overcoming asymmetric information  145 3.3.2 The labour market AI is also a formidable tool for reducing information asymmetry in the labour market. As discussed in section 3.2., companies are gathering huge streams of data through sensors and other data collection systems. It is obvious that these data are humanly impossible to manage, so AI offers appropriate tools to manage and process the data in these contexts. This kind of continuous, automated supervision would provide an appropriate incentive for the worker to behave properly. In addition to this, machine learning could also be used to predict the skill level of job candidates when hiring (Fumagalli et al., 2022). Hiring is notoriously considered an investment under uncertainty since, in most labour markets, the employer does not know an applicant’s productive skills at the time of the interview, and this information may become available to the employer after an unknown period if the applicant is hired (Spence, 1973). This is clearly an asymmetric information problem, as the candidate has more information about his or her capabilities than the employer. Following a similar approach to that described in section 3.3.1, we assume for simplicity that there are only two types of applicants, low-skilled (L) and highly skilled (H) applicants. If we collect data on past applicants and their job performance after being hired, we have again a classification problem in which we want to predict p, in this case the probability of hiring a highly skilled worker, using a set of covariates, such as education, age, previous work experience, and so on. This approach is certainly feasible these days given that large (as well as mid-sized) companies are creating databases of candidates and employees. AI has the potential to profoundly change human resource management and offer companies the opportunity to gain a competitive advantage (Pan et al., 2022). The main disadvantage of this approach is that a purely algorithmic selection process would penalize assessment elements that can only be gathered during personal exchanges with the recruiter or employer; therefore, this innovative process should be used to complement the normal assessment process. As noted in a Forbes10 article, recruiting that leverages artificial intelligence with experienced human talent acquisition professionals will be a successful strategy to recruit the best candidate. 3.3.3 Digital marketplaces Digital marketplaces have grown to the point where they are now perceived as a common way of doing business. In principle, in these platforms anything can be offered in exchange for money, from items that users no longer use (e.g., eBay) to hotel reservations (e.g., Booking. com), from freelance services (e.g., Fiverr) to used and vintage clothing pieces (e.g., Depop). The initial mistrust of many users for this way of trading has now been overcome, and today online platforms such as eBay have billions of visits every month.11 Digital markets present an asymmetric information problem; in fact, strangers engaged in a transaction are reluctant to trust the counterparty. Fraud can occur on both the seller’s and the buyer’s side. On platforms like eBay, the buyer may not receive the product or may receive a different one, while on platforms like Booking.com, fraudulent reservations made by the “buyer” can cause the hotelier lost revenue. The introduction of feedback and reputation mechanisms changed the game. However, many studies have shown that user-generated feedback is not entirely reliable because it is often biased (“grade inflation”) and can be manipulated by sellers (Milgrom & Tadelis, 2019). Fortunately, the amount of data generated by these marketplaces offers alternative solutions to address the asymmetry problem.

146  The Elgar companion to information economics A first application of AI to digital marketplaces comes from the huge amount of data generated by users during private chat conversations before and after the transaction. In many platforms, such as eBay or Booking.com, it is customary to contact the counterparty before finalizing the transaction to ask specific questions that are not provided in the insertion, or afterwards to provide feedback. This helps the buyer gain all the necessary information before making a choice, and the seller get feedback and design a better customer experience. Techniques such as Natural Language Processing (NLP) or other text mining tools can be used to analyse the elements that characterize successful transactions in terms of customer and seller experience. This would help generate more comprehensive feedback from users. As a matter of fact, AI is already changing the user experience on these websites, as search engines rely on algorithms that try to come up with the most relevant results for the user as the first results. For example, eBay uses the Best Match method, which is “designed to show the most relevant listings, taking into account the things our users find most important when they’re deciding what to buy”.12 This algorithm weighs several factors, such as how closely the listing matches the buyer’s search terms, the price of the item, or the seller’s track record. This saves the buyer a lot of time and increases the likelihood of finding a good seller. Indeed, the disclosure provided in the listing in these markets is an effective remedy against information asymmetry (on the issue, see Lewis, 2011). Unfortunately, a major problem of asymmetric information against new users persists. New buyers or sellers without a record of their activity in the platform could easily be marginalized as users would not trust them. This is known as the “cold-start” problem (Milgrom & Tadelis, 2019) and can be addressed in several ways. For example, NLP AI algorithms can be used to provide feedback in such situations (Milgrom & Tadelis, 2019) or deep learning can be used to overcome the item cold-start problem (Yuan et al., 2016), e.g., when items added to a catalogue have no or very few interactions. 3.3.4 The market for news With the advent of social networking, news sharing and production have been revolutionized. Nowadays, we are constantly and immediately informed about what is happening in the world. The accessibility and shareability of news are immediate in the internet age; however, the fact that almost anyone can produce news poses serious problems in the process of information dissemination. For example, one of the main problems is the spread of fake news, which can be defined as unfounded information that mimics the form of reliable content in news media (Lazer et al., 2018). The presence of good and bad news, and asymmetric information, exposes this market to the same problems as Akerlof’s market for used cars. To solve the problem of asymmetry in the news market, researchers have designed online tools that provide guidelines for unaware users to recognize bad news. The purpose is to reveal to users the relevant information, i.e., the quality of the news or the reliability of the author. For example, Botometer13 monitors the activity of Twitter accounts and rates them on how likely they are to be bots, while FactCheck.org is a website that checks the factual accuracy of what is stated by leading US politicians to the media. Pernagallo et al. (2021) use machine learning algorithms to assign a trustworthiness score to news websites. Logit and probit models are used to rank news websites based on a set of features, such as the presence of a “contact us” section or a secure connection. These features are manifestations of a legitimate and editorially compliant organization. The use of algorithms to combat the spread of bad news proves to be a formidable tool for disseminating information to users quickly and easily.

Overcoming asymmetric information  147 3.3.5 Public administration The relationship between Public Administration (PA) and citizens is also characterized by the presence of information asymmetry (Mayston, 1993; Pernagallo & Torrisi, 2020), because the “principal” (citizens) cannot directly observe the actions of the “agent” (PA). Then, transparency becomes crucial to make the “action” of the agent observable. To solve this problem, it is now customary to disseminate relevant information through official PA websites, a phenomenon generally referred to as e-government. In this way, citizens can monitor PA actions, for example, by reading spending reports and other documents. One of the main problems associated with such a solution is that, especially at the municipal level, websites can be poorly constructed, so their actual use by citizens is very limited (Pernagallo & Torrisi, 2020). Machine learning algorithms can be used to verify the quality of websites and their compliance with transparency regulations. Learning algorithms can assess the transparency of institutional websites based on various features available on the website. In general, the implementation of AI can be used to improve public decision-making and increase trust in citizens, thereby decreasing the information discrepancy between citizens and administrators (Zuiderwijk et al., 2021). However, the implementation of AI in the management of res publica carries several risks as many citizens may lose trust in the government due to low protection of their privacy and the creation of black box systems. In addition, the use of AI creates a serious liability issue: since AI cannot be legally prosecuted, who should be blamed for mistakes? These issues will be a topic of debate in the coming years, and the literature on the subject is growing at an increasing rate (on this issue, see Pernagallo & Torrisi, 2020; de Fine Licht & de Fine Licht, 2020; Zuiderwijk et al., 2021).

4.

AI AND INFORMATION ASYMMETRIES: PRESENT AND FUTURE PERSPECTIVES

As discussed in section 3, there are many situations where data-driven approaches have reduced the problem of asymmetric information. However, there are many other situations where asymmetric information is still relevant. Two examples are personal data markets and the market for news. In both cases, the problem is so severe that it requires policy maker intervention to prevent misuse of private data and dissemination of misleading news. In these cases, the use of algorithms is certainly useful to inform users (Waerdt, 2020; Pernagallo et al., 2021), but insufficient without the support of laws that disincentivize misbehaviour. Thus, the question is: how can we distinguish between cases where the asymmetric problem can be addressed only with data and cases where data-driven approaches are insufficient? My answer to this question is based on the concept of monopolies of knowledge. This idea was developed by historian and economist Harold Innis (Innis, 1946, 1986; Heyer, 2003) to denote situations in which a group monopolizes knowledge by controlling both the information available and how it is interpreted (Comor, 2018). Innis gives the example of ancient societies, where mastery of writing systems gave priests a monopoly on knowledge (Innis, 1986). We can extend this idea to modern societies by using the Internet as our main example. In fact, the Internet resembles a complex system that allows capable users to control information and, consequently, gain a better position than naive users. Large users take advantage of the new forms of increasing returns that characterize both the use of digital technologies and the generation of an array of knowledge(s) ranging from technological, to financial, commercial, and

148  The Elgar companion to information economics economic knowledge. In economics, similar research into knowledge monopolies has found breeding ground in the more recent literature on intellectual capital and intellectual monopoly capitalism (Pagano & Rossi, 2009; Pagano, 2014; Rikap & Lundvall, 2022). We can see how monopolies of knowledge play a role in data-driven markets or the market for news. In data-driven markets such as the market for personal data, the asymmetry arises because companies through the Internet can collect much more personal data than consumers think to give. This is the result of better knowledge about how private data information is processed than consumers. The problem is also broader, as it affects different types of users besides consumers, such as citizens or workers. Consider how citizens’ personal information is stored online by the public administration or how private companies process the personal information of applicants and employees. This information is easily accessible via the Web and can be exploited by third parties. In the market for news, news creators know the quality of news because they know the quality of their sources or how to look for news veracity, while users generally lack media literacy skills, attention, or more generally, relevant knowledge (Guess et al., 2020; De Paor & Heravi, 2020; Pennycook & Rand, 2021; Pennycook et al., 2021). In both examples, one party profits from its better knowledge of the Internet at the expense of the other party. The technology itself creates information asymmetries that are bound to persist if the policy maker does not intervene (e.g., through legislation) or if the monopoly on knowledge is not extinguished (e.g., by increasing user knowledge). The use of data-driven approaches such as AI or machine learning in contexts with monopolies of knowledge may or may not extinguish the asymmetry. To better understand this point, Figure 7.2 shows a matrix, where the horizontal axis represents the concentration of knowledge, and the vertical axis represents the level of pervasiveness of AI. The term “pervasiveness” is used to indicate the effectiveness with which AI can be applied to solve a problem. Obviously, the more concentrated the knowledge, the closer we get to a monopoly of knowledge. We obtain four regions of the plane based on the level of AI pervasiveness and knowledge concentration, into which, for example, we can classify some of the markets discussed in section 3. Based on the discussion made in section 3, we can say that the personal data market is an example of a market with a significant concentration of knowledge and a low level of AI pervasiveness.14 Legislation (such as laws protecting user privacy) and user awareness are essential, so this market has a significant level of asymmetry. On the other hand, the used car market is a good example of a market with low knowledge concentration and high AI pervasiveness, as data-driven approaches are easy to implement and can drastically reduce, or even solve, asymmetry. The labour market has a low level of knowledge concentration, but the asymmetry persists even with the application of AI tools. Indeed, machines cannot assess the talent or many skills of employees and candidates, so human intervention is still needed in the process. Finally, the news market or PA are situations with relevant knowledge concentration but high AI pervasiveness because algorithms can be used to disseminate relevant information to users and reduce information discrepancy. This demarcation criterion seems to work well, although it is only an exemplification of reality and serves only to illustrate how the idea of monopolies of knowledge can help us distinguish precise areas in which to apply data-driven approaches. The argument also stimulates another consideration. As modern society is increasingly dependent on digital technologies, the problem of asymmetric information is likely to change and evolve over time. For example,

Overcoming asymmetric information  149

Figure 7.2

Markets by knowledge concentration and AI pervasiveness

the use of sensors in the workplace on the one hand reduces the problem of moral hazard, but on the other hand can introduce a knowledge differential between employers and employees as occurs in personal data markets. If this issue is not managed by the policy maker, this can create a monopoly of knowledge and a new source of asymmetry.

5. CONCLUSIONS Innovation is changing our daily lives in many ways. Digital technologies have become an integral part of our routines and it is inevitable that this changes the way we deal with many problems. The problem of information asymmetries in many cases seems to be obsolete in the era of big data, in which decision makers struggle to process huge amounts of information (Pernagallo & Torrisi, 2022). In fact, the use of data represents a formidable tool against asymmetric information. A first purpose of this chapter is to show how data-driven approaches are effective in dealing with information asymmetries. This is certainly true in “traditional” markets such as credit, labour, or insurance; however, there are many instances where technology itself is creating new sources of asymmetry. In some cases, AI can mitigate the asymmetry, in others human intervention is still needed. Thus, a second purpose is to propose the idea of monopolies of knowledge to distinguish between these two situations. Machine learning and AI can be used, for example, to estimate the default probability of borrowers (Ge et al., 2017; Xu et al., 2021), to predict risk in actuarial problems (Boodhun & Jayabalan, 2018), to monitor workers and in the candidate selection process (Fumagalli et al., 2022), to address the “cold-start” problem in digital marketplaces (Yuan et al., 2016;

150  The Elgar companion to information economics Milgrom & Tadelis, 2019), to increase the transparency of PA action (Pernagallo & Torrisi, 2020; Zuiderwijk et al., 2021), or to reveal the quality of news or news sources to internet users (Pernagallo et al., 2021). In all these cases, data-driven approaches mitigate the asymmetry problem. However, digital technologies inevitably create monopolies of knowledge, as only a fraction of users can master these new tools and new forms of increasing returns are at work in their use. This introduces new scenarios with asymmetries, as is the case of data-driven markets. Exactly as in ancient cultures the class that dominated the writing systems exerted a form of power over other classes, the same is happening with the Internet. In these cases, algorithmic solutions, although useful, must be supported by policy maker intervention and by fostering people’s information literacy to avoid digital monopolies of knowledge.

NOTES 1.

This work is part of the author’s PhD thesis. The author is grateful for the valuable comments of Prof. Cristiano Antonelli, whose support was fundamental, and for the remarks of the participants at the 9th International PhD Workshop in Economics of Innovation, Complexity and Knowledge (WICK#9) held in Turin. Special thanks go to the editors for their work and to the two anonymous reviewers for their insightful comments. 2. See https://​www​.ibm​.com/​cloud/​blog/​ai​-vs​-machine​-learning​-vs​-deep​-learning​-vs​-neural​ -networks. 3. See also the special issue “The Economics of Artificial Intelligence and Machine Learning” published in Information Economics and Policy: https://​www​.sciencedirect​.com/​journal/​information​ -economics​-and​-policy/​vol/​47. 4. See https://​www​.nature​.com/​articles/​d41586–019–01155–0. 5. See https://​www​.nationalgeographic​.com/​science/​article/​first​-picture​-black​-hole​-revealed​-m87​ -event​-horizon​-telescope​-astrophysics. 6. For a discussion on digital innovation, see Chapter 13 of this Companion (Bauer & Prado, 2024, pp. 246–269). 7. See https://​www​.statista​.com/​statistics/​183713/​value​-of​-us​-passenger​-cas​-sales​-and​-leases​-since​ -1990/​. 8. See https://​www​.cnbc​.com/​2019/​04/​15/​employee​-privacy​-is​-at​-stake​-as​-surveillance​-tech​ -monitors​-workers​.html. 9. See https://​www​.dnb​.com/​resources/​business​-credit​-scorecard​.html. 10. See https://​www​.forbes​.com/​sites/​forbeshu​manresourc​escouncil/​2021/​06/​16/​is​-ai​-the​-answer​-to​ -recruiting​-effectiveness/​?sh​=​10ecdb352d7c. 11. See https://​www​.statista​.com/​topics/​2181/​ebay/​#dossierKeyfigures. 12. See https://​www​.ebay​.com/​help/​selling/​listings/​listing​-tips/​optimising​-listings​-best​-match​?id​=​ 4166. 13. See https://​botometer​.osome​.iu​.edu/​. 14. On this issue, Chapter 23 of this Companion offers a discussion on the role of machine learning in data privacy (Bodoff, 2024, pp. 462–480).

REFERENCES Akerlof, G. A. (1970). The market for “lemons”: Quality uncertainty and the market mechanism. The Quarterly Journal of Economics, 84(3), 488–500. Allen, C., Wallach, W., & Smit, I. (2006). Why machine ethics? IEEE Intelligent Systems, 21(4), 12–17. Antonelli, C. (2017). Digital knowledge generation and the appropriability trade-off. Telecommunications Policy, 41, 991–1002.

Overcoming asymmetric information  151 Antonelli, C. (2018). From the economics of information to the economics of knowledge. In C. Antonelli (Ed.), Recent Developments in the Economics of Information. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Antonelli, C. (2019). The Knowledge Growth Regime: A Schumpeterian Approach. Cham: Palgrave Macmillan. Antonelli, C., Geuna, A., & Steinmueller, E. (2000). New information and communication technologies and the production, distribution and use of knowledge. International Journal of Technology Management, 20(1/2), 72–94. Arrow, K. J. (1969). Classificatory notes on the production and transmission of technical knowledge. American Economic Review, 59(2), 29–35. Arrow, K. J. (1996). The economics of information: An exposition. Empirica, 23, 119–128. Athey, S. (2019). The impact of machine learning on economics. In A. Agrawal, J. Gans, & A. Avi Goldfarb (Eds.), The Economics of Artificial Intelligence: An Agenda (pp.  507–552). Chicago: University of Chicago Press. Bag, S., Gupta, S., Kumar, A., & Sivarajah, U. (2021). An integrated artificial intelligence framework for knowledge creation and B2B marketing rational decision making for improving firm performance. Industrial Marketing Management, 92, 178–189. Balland, P. & Boschma, R. (2021). Mapping the potentials of regions in Europe to contribute to new knowledge production in Industry 4.0 technologies. Regional Studies, 55(10–11), 1652–1666. Bauer, J. M. & Prado, T. S. (2024). Digital innovation: An information-economic perspective. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 246–269). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Bester, H. (1987). The role of collateral in credit markets with imperfect information. European Economic Review, 31(4), 887–899. Bester, H. (1994). The role of collateral in a model of debt renegotiation. Journal of Money, Credit and Banking, 26(1), 72–86. Bodoff, D. (2024). On the status of machine learning inferences in data privacy economics and regulation. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 462–480). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Boisot, M. & Canals, A. (2004). Data, information and knowledge: Have we got it right? Journal of Evolutionary Economics, 14, 43–67. Boodhun, N. & Jayabalan, M. (2018). Risk prediction in life insurance industry using supervised learning algorithms. Complex & Intelligent Systems, 4, 145–154. Byford, M. C. (2017). Moral hazard in strategic decision making. International Journal of Industrial Organization, 55, 114–136. Chubb, J., Cowling, P., & Reed, D. (2021). Speeding up to keep up: Exploring the use of AI in the research process. AI & Society, 37(1). Comor, E. (2018). Ubiquitous media and monopolies of knowledge: The approach of Harold Innis. In M. Daubs & V. Manzerolle, (Eds.), From Here to Ubiquity: Critical and International Perspectives on Mobile and Ubiquitous Media (pp. 183–200). New York: Peter Lang. de Fine Licht, K. & de Fine Licht, J. (2020). Artificial intelligence, transparency, and public decision-making. AI & Society, 35, 917–926. De Meza, D. & Webb, D. C. (1987). Too much investment: A problem of asymmetric information. The Quarterly Journal of Economics, 102(2), 281–292. De Paor, S. & Heravi, B. (2020). Information literacy and fake news: How the field of librarianship can help combat the epidemic of fake news. Journal of Academic Librarianship, 46(5), 102218. Dierkens, N. (1991). Information asymmetry and equity issues. Journal of Financial and Quantitative Analysis, 26(2), 181–199. Engelen, J. E. van & Hoos, H. H. (2020). A survey on semi-supervised learning. Machine Learning, 109, 373–440. Fumagalli, A., Rezaei, S., & Salomons, A. (2022). OK computer: Worker perceptions of algorithmic recruitment. Research Policy, 51(2), 104420. Ge, R., Feng, J., & Zhang, P. (2017). Predicting and deterring default with social media information in peer-to-peer lending. Journal of Management Information Systems, 34(2), 401–424.

152  The Elgar companion to information economics Geuna, A. (1999). The Economics of Knowledge Production: Funding and the Structure of University Research. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Gu, S., Kelly, B., & Xiu, D. (2020). Empirical asset pricing via machine learning. Review of Financial Studies, 33, 2223–2273. Guess, A. M., Lerner, M., Lyons, B., Montgomery, J. M., Nyhan, B., Reifler, J., & Sircar, N. (2020). A digital media literacy intervention increases discernment between mainstream and false news in the United States and India. Proceedings of the National Academy of Sciences, 117(27), 15536–15545. Gutiérrez-Nieto, B., Serrano-Cinca, C., & De la Cuesta, M. (2016). A multivariate study of over-indebtedness’ causes and consequences. International Journal of Consumer Studies, 41(2), 188–198. Hagendorff, T. & Wezel, K. (2020). 15 challenges for AI: Or what AI (currently) can’t do. AI & Society, 35, 355–365. Heyer, P. (2003). Harold Innis. Toronto: Rowman & Littlefield Publishers. Hubbard, R. G. (1990). Introduction to “Asymmetric information, corporate finance, and investment”. In R. Glenn Hubbard (Ed.), Asymmetric Information, Corporate Finance, and Investment (pp. 1–14). Chicago: University of Chicago Press. Innis, H. A. (1946). Political Economy in the Modern State. Toronto: The Ryerson Press. Innis, H. A. (1986). Empire & Communications. Ontario: Press Porcépic Limited. Invernizzi, A. C., Menozzi, A., Passarani, D. A., Patton, D., & Viglia, G. (2016). Entrepreneurial overconfidence and its impact upon performance. International Small Business Journal: Researching Entrepreneurship, 35(6), 709–728. James, G., Witten, D., Hastie, T., & Tibshirani, R. (2021). An Introduction to Statistical Learning with Applications in R. New York: Springer. Jensen, M. C. & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305–360. Jost, P.-J. (1991). Monitoring in principal-agent relationships. Journal of Institutional and Theoretical Economics, 147(3), 517–538. Khandani A. E., Adlar, J. K., & Lo, A. W. (2010). Consumer credit-risk models via machine-learning algorithms. Journal of Banking & Finance, 34(11), 2767–2787. Koch, R. (1999). The 80/20 Principle: The Secret to Achieving More with Less. London: Currency Illustrated Edition. Lazer, D. M. J., Baum, M. A., Benkler, Y., Berninsky, A. J., Greenhill, K. M., Menczer, F., et al. (2018). The science of fake news. Science, 359(6380), 1094–1096. Lewis, G. (2011). Asymmetric information, adverse selection and online disclosure: The case of eBay motors. American Economic Review, 101, 1535–1546. Mas-Colell, A., Whinston, M. D., & Green, J. R. (1995). Microeconomic Theory. Oxford: Oxford University Press. Mayston, D. (1993). Principals, agents and the economics of accountability in the new public sector. Accounting, Auditing and Accountability Journal, 6(3), 68–96. McDonald, J. G. (1974). Faculty tenure as a put option: An economic interpretation. Social Science Quarterly, 55(2), 362–371. Milgrom, P. R. & Tadelis, S. (2019). How artificial intelligence and machine learning can impact market design. In A. Agrawal, J. Joshua Gans, & A. Goldfarb (Eds.), The Economics of Artificial Intelligence: An Agenda (pp. 567–585). Chicago: University of Chicago Press. Mishkin, F. S. & Eakins, S. G. (2018). Financial Markets and Institutions. Harlow: Pearson. Mittelstadt, B. T., Allo, P., Taddeo, M., Wachter, S. & Floridi, L. (2016). The ethics of algorithms: Mapping the debate. Big Data & Society, 3(2). Mullainathan, S. & Shleifer, A. (2005). The market for news. American Economic Review, 95(4), 1031–1053. Mullainathan, S. & Spiess, J. (2017). Machine learning: An applied econometric approach. Journal of Economic Perspectives, 31(2), 87–106. Narayanan, M. P. (1988). Debt versus equity under asymmetric information. Journal of Financial and Quantitative Analysis, 23(1), 39–51. Pagano, U. (2014). The crisis of intellectual monopoly capitalism. Cambridge Journal of Economics, 38, 1409–1429.

Overcoming asymmetric information  153 Pagano, U. & Rossi, M. A. (2009). The crash of the knowledge economy. Cambridge Journal of Economics, 33, 665–683. Pan, Y., Froese, F., Liu, N., Hu, Y., & Ye, M. (2022). The adoption of artificial intelligence in employee recruitment: The influence of contextual factors. International Journal of Human Resource Management, 33(6), 1125–1147. Park, H., Kim T. S., & Park, Y. J. (2021). Asymmetric information in the equity market and information flow from the equity market to the CDS market. Journal of Financial Markets, 55, 100607. Pennycook, G., Epstein, Z., Mosleh, M., et al. (2021). Shifting attention to accuracy can reduce misinformation online. Nature, 592, 590–595. Pennycook, G. & Rand, D. G. (2021). The psychology of fake news. Trends in Cognitive Sciences, 25(5), 388–402. Pernagallo, G. (2023). Science in the mist: A model of asymmetric information for the research market. Metroeconomica, 74(2), 390–415. Pernagallo, G. & Torrisi, B. (2020). A logit model to assess the transparency of Italian public administration websites. Government Information Quarterly, 37(4), 101519. Pernagallo, G. & Torrisi, B. (2022). A theory of information overload applied to perfectly efficient financial markets. Review of Behavioral Finance, 14(2), 223–236. Pernagallo, G., Torrisi, B., & Bennato, D. (2021). A classification algorithm to recognize fake news websites. In P. Mariani & M. Zenga (Eds.), Data Science and Social Research II (pp.  313–329). Cham: Springer. Rikap, C. (2022). Becoming an intellectual monopoly by relying on the national innovation system: The State Grid Corporation of China’s experience. Research Policy, 51(4), 104472. Rikap, C. & Lundvall, B. (2022). Big tech, knowledge predation and the implications for development. Innovation and Development, 12(3), 389–416. Rothschild, M. & Stiglitz, J. (1976). Equilibrium in competitive insurance markets: An essay on the economics of imperfect information. The Quarterly Journal of Economics, 90(4), 629–649. Russell, S. J. & Norvig, P. (2010). Artificial Intelligence: A Modern Approach. Upper Saddle River, NJ: Prentice Hall. Shalev-Shwartz, S. & Ben-David, S. (2014). Understanding Machine Learning: From Theory to Algorithms. Cambridge: Cambridge University Press. Spence, M. (1973). Market Signalling: Information Transfer in Hiring and Related Processes. Cambridge, MA: Harvard University Press. Stiglitz, J. E. & Weiss, A. (1981). Credit rationing in markets with imperfect information. American Economic Review, 71(3), 393–410. Stiglitz, J. E. & Weiss, A. (1992). Asymmetric information in credit markets and its implications for macro-economics. Oxford Economic Papers, New Series, 44(4), Special Issue on Financial Markets, Institutions and Policy, 694–724. Stock, W. G. & Stock, M. (2013). Handbook of Information Science. Berlin: De Gruyter Saur. Varian, H. R. (2014). Big data: New tricks for econometrics. Journal of Economic Perspectives, 28(2), 3–27. Waerdt, P. J. van de (2020). Information asymmetries: Recognizing the limits of the GDPR on the data-driven market. Computer Law & Security Review, 38, 105436. Warnat-Herresthal, S., Schultze, H., Shastry, K. L., et al. (2021). Swarm learning for decentralized and confidential clinical machine learning. Nature, 594, 265–270. Weitzman, M. L. (1996). Hybridizing growth theory. American Economic Review, 86, 207–212. Welker, M. (1995). Disclosure policy, information asymmetry, and liquidity in equity markets. Contemporary Accounting Research, 11(2), 801–827. Xu, J., Lu, Z., & Xie, Y. (2021). Loan default prediction of Chinese P2P market: A machine learning methodology. Scientific Reports, 11, Article 18759. Yuan, J., Shalaby, W., Korayem, M., Lin, D., Al Jadda, K., & Luo, J. (2016). Solving cold-start problem in large-scale recommendation engines: A deep learning approach. 2016 IEEE International Conference on Big Data (Big Data), 1901–1910. Zuiderwijk, A., Chen, Y., & Salem, F. (2021). Implications of the use of artificial intelligence in public governance: A systematic literature review and a research agenda. Government Information Quarterly, 38(3), 101577.

8. Asymmetric information in health economics: Can contract regulation improve equity and efficiency? Pau Olivella

1. INTRODUCTION This chapter covers the application of concepts and tools of the theory of asymmetric information (AI) to the realm of health economics. As Arrow already pointed out in 1963, problems of AI are pervasive in this industry. We concentrate here on pure asymmetric information issues (PAI), where one of the parties has private information either on the environment or on his or her own preferences or abilities.1 Hence, we leave aside the – also very relevant – problems of moral hazard (hidden action).2 Hence this chapter deals with the scarcity of information in one side of the relationships studied. In the examples of PAI that follow, we indicate in parenthesis which is the party that has privileged information. PAI arises in the prescription of drugs (the physician), the purchase of health insurance (the individual), the procurement of health services by a health authority (providers), and in the negotiation of prices between payers and big pharma (the pharmaceutical firm), to mention a few examples. These issues are as of now well understood in a general setting, but the healthcare market is affected by additional forces that require revisiting the standard models and predictions. First, the healthcare market is heavily regulated. Take for instance the menus of insurance plans that an insurance firm may put on the table. As is well known, these menus aim at screening low risks (who will choose cheap but not very generous coverage) from high risks (who will purchase expensive and generous coverage). However, the Affordable Care Act in the US (also known as Obamacare) imposes two sorts of restrictions in the market. The first restriction is mandatory enrolment, which is basically aimed at avoiding the so-called “spiral of death”, a phenomenon that we will illustrate with a simple numerical example later on. The second one is the establishment of some minimum coverage (aimed at limiting the distortions imposed by the existence of PAI). All these issues are treated in section 2. Similarly, in the relationship between physicians and patients (in section 3), it is usually the case that a third party is involved, namely the “third party payer” (3PP), be it a private insurer or a national health system (NHS). This implies that physicians’ remuneration and their actions are governed by contracts and protocols and subject to malpractice litigation.3 For reasons of space, I will restrict attention to two issues in this relationship: inappropriate prescription and inappropriate referrals by primary care physicians. Again, moral hazard (lack of diagnostic or treatment effort) will not be discussed. In other words, I will assume that effort is observable and contractable.4 The existence of a conflict of objectives between the players involved in these relationships gives rise to contracts (e.g., employment or insurance contracts) that aim at harmonizing these 154

Asymmetric information in health economics  155 objectives. However, these contracts often introduce severe distortions both in the allocation of risk and in the actions of players. Following the examples above, screening individuals in the insurance realm requires that some individuals (the low-risk individuals) bear more risk than would be otherwise efficient. Hence the interventions observed in reality, like minimum coverage legislation. As for the doctor–patient relationship, contracts often induce doctors to exert suboptimal effort in order to induce revelation of their private information. Contracts also impinge on the equity of the final allocation. For instance, low-risk individuals may cross-subsidize those who bear a higher risk. Similarly, two patients with the same medical needs may happen to visit doctors with different abilities or levels of altruism. It can also be the case that altruism is exploited in the market so that altruistic doctors end up receiving lower salaries. The main question that arises is whether contract regulation can bring improvements in these two matters, i.e., can efficiency be restored? and, can equity be further promoted? In all sections, we discuss both the theoretical literature and the empirical evidence (both in-the-field and experimental). The latter provides support to either the existence of the problem in the first place (do some agents really have privileged information? and, if so, do they act on it?) or to the predicted market reaction (does the less informed party respond to PAI as predicted by the theory?).

2.

ASYMMETRIC INFORMATION IN HEALTH INSURANCE MARKETS

2.1

The Framework

Individuals have privileged information on several dimensions that determine their final demand for health insurance. These dimensions can be divided into two main groups: behaviour and environment. Some of these dimensions also determine their expected future healthcare expenditures, which in turn affect the coverage costs borne by the insurer. As explained in the introduction, we will not address behaviour factors, which imply that either health risk itself (ex-ante moral hazard) or the demand of health services (ex-post moral hazard) become endogenous. Smoking or driving a motorbike are examples of the former, requesting cataract surgery earlier when the insurance pays part of the cost is an example of the latter. We instead concentrate on exogenous determinants of the demand for health insurance. These include (a) the family health history (for genetically transmitted diseases) or the results of medical tests, which provide informative signals on the likelihood of becoming ill, (b) an individual’s preferences for health as a good, (c) his or her cognitive ability, or (d) his or her risk tolerance. Although it is true that some of these variables can be partially inferred from readily available data on the individual (age, gender, past health expenditures), some information remains private for the individual (think of risk tolerance or preference for health). In fact, some countries put limits on the extent to which observable variables can be used in conditioning health contracts, a regulation that de facto exacerbates the informational asymmetry between the insurer and the individual. Such regulation (sometimes referred to as “community rating”) is quite specific to the health insurance industry. The main problem that arises due to such informational asymmetries is best understood if one assumes that the insurer offers a single contract to all individuals in a given class (say

156  The Elgar companion to information economics 35-year-old males, or class A). This gives rise to the so called “adverse selection problem” strictu sensu.5 A numerical example will prove useful. Suppose a health insurer offers a contract covering 20,000 euros of hospitalization bills to all individuals in class A. Suppose also that within class A individuals, there are three types of individuals, each with a different probability of needing hospitalization. Say these probabilities are 0.05, 0.25, and 0.3, respectively. Hospitalization costs 30,000 euros. Suppose that the shares of types among the population are the same, 1/3 each type. Notice that the expected indemnity per capita within class A is 1/3*0.05*20,000+1/3*0.25*20,000+1/3*0.3*20,000 = 4,000. Now suppose that an insurer sets the premium at 4,500 euros (in order to obtain a margin of 12.5%). Let us assume for the moment that individuals only care about the expected gain they obtain from accepting the contract. In other words, individuals are risk neutral. The first type of individual would expect to gain 0.05*20,000–4,500=-3,500. The same calculation for the second type would yield 0.25*20,000–4,500=500, and 0.3*20,000–4,500=1,500 for the third type. As a consequence, the actual selection of individuals demanding the insurance contracts would only be made up by the second and third types (50 per cent of each type), which implies that the selection of individuals is adverse for the insurer, hence the term “adverse selection”. The true expected gain for the insurer becomes negative: 4,500-(1/2*0.25*20,000 +1/2*0.3*20,000)=-1,000. The insurer may now be tempted to increase the premium to avoid the loss, but this will make things worse. Indeed, suppose that he raises the premium to 4,500+1,000=5,500. The second type now gains 0.25*20,000–5,500=-500, and hence he also drops from the market. Now only individuals with the highest risk are willing to purchase the insurance contract, and the insurer’s profit per enrolee becomes 5,500-(0.3*20,000)=-500. The final outcome now depends on whether the insurer can readjust the premium to still attract the high risks and break even. But the important lesson is that only individuals with the highest risks may be insured. This is the phenomenon known as the spiral of death.6 The reader may think that the example is concocted and that it is easy to change the data slightly to arrive to a different conclusion, and he or she would be right. However, the example points out several important elements in the discussion. The first one is how asymmetric the information is. Can’t the insurer try to make a better guess at who is which type of individual? Suppose the insurer is able to condition the contract on whether the individual had any health expenditures in the previous two years. He could then refine the class to which the contract given above is offered. Suppose class A-1 is composed of 35-year-old males with no health expenditures in the last 2 years. Assume that the proportions of each type within class A-1 is 3/6 (the low risk), 2/6 (the medium risk), 1/6 (the high risk), respectively. That is, given the new information, the proportion of high (low) risks has decreased (increased) a lot. As computed above, it is still true that only the 2nd and 3rd types would accept the given contract, but now the insurer makes positive profits. Indeed, 4,500-(2/6*0.25*20,000+1/6*0.3*20,000)=1833.3 euros. Hence the adverse selection problem persists but is less severe. The strategy of using observables to underwrite contracts is referred to as “risk categorization” and does indeed make the information between insurers and individuals more symmetric.7 However, many countries put limits on the extent to which categorization is allowed. For instance, categorization by gender has been banned in the EU since 2012 (European Union, 2012). A second feature of the numerical example is that coverage is relatively low: 20,000/30,000=66.67 per cent. How is the level of coverage decided? Shouldn’t the insurer use coverage as well as premium in order to maximize profits? In the same vein, is the margin over costs sustainable? Wouldn’t competition among insurers force our insurer to reduce it?

Asymmetric information in health economics  157 Similarly, if another insurer is offering a more generous coverage, say 75 per cent, at the same premium, wouldn’t our insurer be forced to do the same? Finally, if the second type of individual had a slightly lower probability of requiring hospitalization and hence his expected gain from accepting the contract was slightly negative, he might still take it because he is risk averse. In fact, it is precisely because individuals are risk averse that an insurance market exists in the first place. As we have seen, some of the elements just discussed hinge on the presence of other health insurers. We will address this case in the next subsection, and we do it for the case where insurers are indistinguishable except if they offer different contracts. This means that, for a given coverage, a contract with a smaller premium than the rivals’ contract, and no matter how small the difference is, will attract all individuals. The same goes for a fixed premium and a higher coverage. In other words, an individual’s valuation of each dollar spent in providing this coverage is the same across insurers. We refer to this case as “the competitive scenario”. A well-studied departure of these assumptions is the case where each insurer subcontracts with providers that are differentiated, say geographically. For someone living in Madrid, a dollar spent in hip replacement in a hospital in Madrid is valued differently from a dollar spent in hip replacement in a hospital in Barcelona. We will get back to this imperfectly competitive scenario below. 2.2

The Competitive Scenario

The prediction for the case of competition among insurers and whenever a single source of asymmetric information (true risk) exists is clear.8 Insurers will offer a menu of contracts aimed at screening individuals, and each element in the menu will bring zero profits to the insurer. By screening, we mean that each type of individual will choose a different contract in the menu.9 More specifically, and following the previous numerical example, the menu will be composed of three contracts: a very cheap contract with very limited coverage (let me call it a “bronze” contract); a more expensive contract with a more generous coverage (the “gold” contract), and a very expensive and full coverage contract (the “platinum” contract). The idea is that the first type in the example chooses the bronze contract, the second type the gold contract, and the third type the platinum contract. The premium and coverage pairs must, as mentioned before, yield zero profits per insure. However, they also need to satisfy the so-called incentive compatibility constraints, which ensure that no individual is willing to accept the contract aimed to another type. Before addressing this scenario, we need to tackle a rather technical but important issue: does an equilibrium always exist? Notice that in the predicted market equilibrium, some individuals will be purchasing a contract that is cheap but with meagre coverage. Let us keep the term “bronze contract“ to refer to this contract. This is an inefficient situation in regard to the optimal sharing of risk. Indeed, the insurer can hedge her risks among a large number of clients whereas the individual is risk averse, so efficiency in risk sharing mandates that individuals enjoy full insurance. Why doesn’t the market by itself fix this? Suppose that an insurer improves the coverage aimed at the low-risk individual (at a slightly higher premium). This will induce individuals with a higher risk to abandon their contract and switch to the improved contract. This can be shown to bring losses to the insurer from each high-risk individual. However, while this is true, it is also true that the gain in efficiency at the bronze contract allows the firm to make positive profits from the low-risk individuals. If the proportion of low

158  The Elgar companion to information economics risks in the market is sufficiently large, these profits compensate the losses brought about by the high-risk individuals. Technically, there does not exist an equilibrium in this market if the proportion of low risks is sufficiently high. There have been several attempts in the literature to fix this inexistence problem. The one that is most used nowadays is the so-called Wilson-Miyazaki-Spence (WMS) solution, which establishes that the equilibrium menu of contracts is the one such that it is “constrained Pareto-efficient” (Wilson, 1977; Miyazaki, 1977; Spence, 1978). In other words, the equilibrium menu keeps profits to be zero for all insurers and there does not exist any other menu of contracts (making non-negative profits) that makes all individuals better-of and all individuals choose the contract aimed at their type. An important consequence of applying this equilibrium concept is that when the proportion of low risks is quite high, the equilibrium menu of contracts involves some degree of cross-subsidization. More precisely, the low risks bring positive profits by accepting the (improved) bronze contract whereas the high risks bring losses by accepting the platinum contract. An insurer offering the equilibrium menu of contracts does however make zero profits on average. However, how can this be an equilibrium? If I am an insurer making losses from my platinum contract, why should I offer it? Shouldn’t I drop my platinum contract and specialize in attracting low-risks individuals? A negative answer was first proposed by Wilson (1977). Suppose that I drop my platinum contract. Which contract do high-risk individuals purchase? They go to my rivals, who are still offering the platinum contract. My rivals are now making losses since the proportion of high risks has increased for them. The key element in Wilson’s solution is that my rivals respond to this by dropping their platinum contracts themselves. Some high risks now accept my bronze contract, and this (as mentioned above) brings losses to me. My original deviation (dropping my platinum contract) is no longer profitable. A second obvious and direct consequence of using the WMS equilibrium notion is that, no matter the proportion of low risks in the population, the WMS equilibrium cannot be Pareto-improved upon unless the regulator has more information than insurers on who is of which type. Notice that this would be quite an unrealistic premise. However, there is another thing that the regulator can do. Suppose that the regulator forbids insurers from using some observable variable, say age, to condition insurance contracts. Then it is as if the regulator does have superior information on individuals’ types. An insurer knows your age but must treat you and an older individual in the same manner. Suppose now that the regulator collects the premia from the individuals, deposits them in a common fund F, and F pays each insurer according to its clients’ age. Glazer and McGuire (2000) show that, by over-compensating the costs of old individuals and under-compensating the costs of young individuals, all parties in the industry are at least as well-off, and individuals are strictly better-off. This may seem a far-fetched mechanism but is in fact at the heart of the so called “National Health Insurance” or “Statutory Health Insurance” that one finds in the Netherlands, Germany, or in part of the Medicare system for the old in the US. Insurers compete for enrolees, but enrolees do not pay their premium (at least not in full) to the insurer of their choice. It is the common fund that pays the insurer a “capitation rate”. Finding the right adjustment of these capitation rates to actual expected costs has given rise to a large body of literature (see for instance the survey by Ellis and Layton, 2014). Obviously, this adjustment is continuously being improved by means of more powerful diagnostic procedures and statistical techniques.

Asymmetric information in health economics  159 Therefore, interventions can indeed improve efficiency and vertical equity. Distortions are palliated and further cross-subsidization is implemented trough a risk-adjustment formula that overcompensates the healthcare costs of the old and undercompensates those of the young. 2.3

Regulating a Competitive Health Insurance Market

Even at the second-best menu of contracts, some individuals end up with less than full insurance. Several real-world regulations try to directly increase the coverage of these contracts. One of them is the so-called “minimum coverage legislation” (MCL). This regulation is already present in many countries, including the US even before the Affordable Care Act. The idea is simple, an insurer cannot offer a coverage below some threshold, say the coverage that is present in the “gold” contract. Notice that this regulation still allows insurers to use the gold and platinum contracts to screen individuals. A problem with such an arrangement is that (if the minimum coverage is set too high), low risks will not be willing to pay the premium that such a generous contract requires to be financially feasible. This is the reason MCLs are usually coupled with mandatory enrolment, a policy that in the US has been implemented by means of the “right to tax”. It is quite intuitive that this may make the gold contract (which now acts as the bronze contract vis-à-vis the platinum contract) more profitable. After all, this is similar to imposing a minimum quality standard in a market and then making purchase of the product compulsory. Hence the regulator must be vigilant that the minimum coverage legislation does not imply a transfer from the low risks to the insurers (McFadden et al., 2015). As mentioned above, another common regulation is to limit the extent of categorization. However, such a policy only affects the equilibrium set of contracts if these menus entail some degree of cross-subsidization, and this only occurs if the proportion of low risks is high. Suppose that the original proportion of low risks in (sub) class A-1 in the example above is low so that the competitive equilibrium entails zero profits per contract, no matter if it is the bronze or the platinum contract. If categorization based on past expenditures is banned, then class A will have an even lower proportion of low risks than subclass A-1, since individuals with and without past expenditures are pooled in the class A. The new competitive equilibrium will still have zero profits per contract. This implies that the equilibrium menu of contracts remains unchanged after the categorization ban. This constitutes an example where regulation becomes completely ineffective. 2.4

Market Power

We say that insurers have some degree of market power if individuals either have some exogenous preference for some particular insurer or if individuals have difficulties in switching from one insurer to another. The first case arises if, for instance, insurers subcontract healthcare providers that are located in a particular area. The second case arises if insurers offer some advantages to their enrolees that are lost if the individual switches to another insurer.10 Interestingly, the fact that insurers respond to PAI by offering menus of contracts still holds in the presence of market power. This is important because one cannot infer that there is no market power from the presence of such menus. However, it is no longer the case that insurers make zero profits no matter the type they attract. The reason is that each insurer can set its premium slightly above the expected cost because a rival cannot steal its customers by lowering the premium by just one cent. Nevertheless, insurers must still screen their insurees by

160  The Elgar companion to information economics means of a menu of contracts (Jack, 2006; Olivella and Vera-Hernández, 2007). In doing so, it turns out that the rents captured by the insurer from the high risks are lower than the rents captured from low risks, the reason being that it is the high risks that really have an informational advantage. Indeed, high risks have an incentive to pretend to be low risks while the opposite is not true. Even more interestingly, if market power is not too strong, it may be the case that the insurer makes positive profits from the low risks at the bronze contract and losses from the high risks at the platinum contract while insurers obtain positive (but small) profits on average. Notice that this means that market power can lead to cross-subsidization amongst types, which fosters vertical equity in the health-status dimension. Notice that this cross-subsidization is not sustained via Wilson’s argument, but directly by market power. Intuitively, just as before one could ask: why doesn’t an insurer drop the platinum contract? Isn’t this contract yielding losses? The idea is that, as soon as this contract is dropped, high-risk individuals who have the exogenous preference for this insurer will be choosing the bronze contract that remains in the menu. This will bring heavy losses to the insurer in question. This argument also suggests that the existence problem that is present in the competitive case becomes much weaker. Another important consequence of the existence of cross-subsidization is that limiting the extent of categorization (for instance by forbidding gender discrimination) will indeed change the equilibrium contracts. This means these limitations may induce winners and losers. If we take market power to an extreme, we face the typical monopolistic screening problem, where predictions are forthcoming: Low risks will be “pushed down to their voluntary participation constraint”, which means that, in equilibrium, low risks are indifferent between accepting their contract and remaining uninsured. High risks will instead enjoy some informational rents (they will strictly prefer to accept their contract). Still, contracts aimed at attracting the high risk will be expensive and coverage-generous and the opposite will be true of the contracts aimed at the low risks. 2.5

Empirical Evidence

The empirical evidence of the presence of asymmetric information in health insurance is well-established. Less so is whether the presence of asymmetric information leads to adverse selection in this industry. In fact, there are areas, like life insurance, where no evidence of adverse selection is found (McCarthy and Mitchell, 2010). To make things worse, the true probability of falling ill of a given individual is not directly observable and therefore actual healthcare usage is used to proxy this probability. This brings the so-called moral hazard bias, by which an individual who enjoys generous coverage faces a lower spot price of using healthcare services, which induces a higher use and therefore an overestimation of the risk-based adverse selection. In order to clean this bias, some authors (Ettner, 1997; Cardon and Hendel, 2001; Olivella and Vera-Hernández, 2013) have compared the usage of individuals who chose to be insured with those whose insurance was part of their job benefits. Hence, both groups face the same access conditions (which turns off the moral hazard bias) while the first group will in principle act on their private information. In any case, it is incorrect to attribute the lack of adverse selection to the absence of asymmetric information. In fact, several theoretical and empirical works suggest that the true reason for the mixed evidence on adverse selection is in fact the presence of additional sources of asymmetric information. The first theoretical proposals were based on heterogeneous risk aversion. The more risk averse an individual is, the more is he willing to pay for coverage.

Asymmetric information in health economics  161 At the same time, the more risk averse an individual is, the more cautious he will be in his lifestyle, leading to a lower actual risk of requiring healthcare. This may lead to “propitious” (or “advantageous”) rather than “adverse” selection (De Meza and Webb, 2001). However, empirical evidence suggests that heterogeneity in risk aversion does not seem to explain such propitious selection. Rather, it seems that heterogeneity in cognitive ability constitutes a better explanation (Fang et al., 2008; Keane and Stavrunova, 2016). Individuals with higher cognitive ability are both able to predict the risk involved in some activities and, at the same time, understand the benefits of insurance. Finally, as initially suggested by Chiappori et al. (2006), the presence of market power could induce propitious selection even in the presence of asymmetric information on risk. In an extreme, under a monopolistic insurer, the contracts aimed to separate relatively risk-tolerant high risks from relatively risk-averse low risks could provide more coverage to the latter (Olivella and Schroyen, 2014). Notice that this has nothing to do with the endogenous probability of falling ill (through lifestyle) that was used by the early theoretical explanations for propitious selection.

3.

THE PHYSICIAN AS AN EXPERT

3.1 Introduction There are many instances where the physician (she) has privileged information on the patient’s true health status and the appropriate course of action. This is true both vis-à-vis the patient (he) and vis-à-vis the 3PP when the patient is insured. For instance, the patient may at most recognize a few symptoms and, even after some diagnostic procedure, he may be unable to understand the implications of the diagnosis. Hence, the physician acts an expert and may use her informational advantage to her own benefit. This opportunistic behaviour may be limited if the physician is altruistic towards the patient and/or is sensitive to the costs of her treatment decisions. Even if strict protocols governing the appropriate treatment are in place, a large degree of freedom rests in the physician’s hands.11 I will concentrate on two issues in this respect. The first one is related to the correct prescription of drugs, where overtreatment, undertreatment, or incorrect prescription may arise. Think for instance of the opioid crisis or the abuse of antibiotics. The other issue is that of a general practitioner (GP) referring her patient to a specialist, where the financial responsibility that accrues to the GP when referring the patient becomes important. We address these two issues in turn next. 3.2

Prescription Behaviour

Suppose a patient suffering from a certain illness, say hypertension, may be in two alternative conditions. A severe condition, which requires an intensive and expensive drug treatment, say E (for “expensive”); or a mild condition, which requires a less expensive treatment, say C (for “cheap”). We speak of overtreatment if the patient’s condition is mild and the prescription is E, and undertreatment if the opposite holds. In evaluating the consequences of each of these incorrect prescriptions, a large number of cases arise depending on the effectiveness of each of the two treatments as a function of the actual condition of the patient. One could be tempted to

162  The Elgar companion to information economics assume that treatment E is at least as effective as treatment C irrespective of the actual condition of the patient, but this is incorrect if treatment E has severe adverse effects. However, in terms of the financial costs, overtreatment is obviously worse. As a first exercise, suppose that (i) the doctor is salaried (her wage does not depend on either the health or the financial outcome), that (ii) she does not receive any bonus from the producer(s) of either treatment, and that (iii) she is completely selfish. In this case, the physician is absolutely indifferent between the two treatments, and she will prescribe, out of indifference, the treatment that is most efficient in terms of costs and benefits. Suppose that the patient is in mild condition, that treatment E is just a bit more beneficial than treatment C, and that E is much more costly. Then the correct prescription is C. However, if the physician cares intensely enough about the patient, she will prescribe treatment E. This may change if the doctor is either sensitive to the financial costs of the system or if she is held partially responsible for the treatment costs. However, if too much of the treatment cost is borne by the physician, it may be the case that the doctor prescribes C even if the patient is in severe condition. This implies that the optimal remuneration scheme must depend on her other-regarding preferences (altruism towards the patient and cost sensibility). In addition, if E and C are produced by two different laboratories, or if the profit margin is larger for E than for C, incentives coming from the laboratories (“detailing”) may come into play. The doctor is now serving three conflicting parties: the patient, the 3PP, and the pharmaceutical firm. Her behaviour will be shaped by the power of altruism towards the patient, her internalization of financial costs, the pecuniary incentives put in place by the 3PP, and the promotion activities performed by the labs. Unsurprisingly, no theoretical model containing such a large set of ingredients exists. There are, however, some works that are able to tackle a few of these ingredients simultaneously. The literature can be split into two large sets of works. In the first set, some real-world remuneration schemes are taken as given and compared (the positive approach). For instance, one may compare a system where physicians are salaried with one where physicians are paid on a fee-for-service (FFS) basis. In the second set of works, authors aim at characterizing the optimal set of contracts from the point of view of the 3PP, who is assumed to maximize a social welfare function that takes into account both patient’s welfare and healthcare costs (the normative approach). In this latter approach, the Revelation Principle (Myerson, 1979) is usually invoked, by which only very special remuneration arrangements need to be considered when searching for the optimal one. Namely, candidate arrangements must satisfy two properties: 1. The arrangement is a “direct revelation mechanism”, meaning that the physician is asked to report her private information (say, her patient’s health status or her own level of altruism), and then her actions (which are observable and contractable) are functions of the report. This leads to a menu of contracts where different contracts are in principle assigned to different reports. 2. The menu of contracts induces the physician to reveal her private information truthfully. The next step is to find the 3PP’s optimal contract within the family mechanisms satisfying 1 and 2. Once the optimal contract is found, one can compare it with the actual functioning of several real-world systems and provide, if need be, some reform recommendations. Starting with the positive approach, Blomqvist (1991), is the first author recognizing the fact that physicians take decisions that simultaneously affect patients and 3PPs. This author makes two simplifying assumptions. First, the 3PPs are private insurers that perfectly compete

Asymmetric information in health economics  163 for enrolees. Second, the physicians are assumed to be selfish. Physicians’ private information lies in their knowledge of the severity (and therefore need) of their patients. Larger severities deserve larger quantities of treatment. This author compares two mechanisms that can be found in the real world. In the first mechanism the doctor is paid on a FFS basis. In the second one, the doctor is salaried. It is quite obvious that doctors will have an incentive to oversupply healthcare in the first system, whereas they may undersupply it in the second. Neither of the two mechanisms is able to implement the first best levels of care (those that would arise if patient’s severity was observable and contractable). Therefore, an additional source of incentives must be in place. The author proposes malpractice liability as the additional ingredient that balances incentives. Since malpractice liability is assumed to be the more likely the larger the difference between the correct quantity of treatment and the quantity of treatment actually provided, this mechanism, taken in isolation, would induce an oversupply of treatment. Hence, if malpractice is combined with salaried doctors, the first best treatments could be implemented. As for the normative approach, Choné and Ma (2011) propose a model where asymmetric information appears in more than one dimension, namely, the doctor’s degree of altruism and the patient’s healthcare needs are both unknown to the 3PP. Interestingly, these authors show that both undertreatment and overtreatment (as compared to the first best) are possible. More precisely, the optimal mechanism will induce overtreatment by physicians that are (or rather, that report to be) not very altruistic and undertreatment for physicians that report to be highly altruistic.12 Assigning the amount of treatment in this fashion is the only way to induce physicians to reveal their private information. This is an important message: asymmetric information and the possibility to implement the efficient levels of treatment may be incompatible. Some distortions are unavoidable if one wants to maximize social welfare under asymmetric information. In a quite related work, Liu and Ma (2013) also analyse a situation where both doctor’s level of altruism and the appropriate treatment are the physicians’ private information. The fundamental contribution of their analysis is to show that one can combine altruism and the existence of a limited liability constraint to induce doctors, no matter their level of altruism, to take the right treatment decisions. This requires, however, that the physician be able to commit to a treatment plan before he learns the true health status of the patient. Surprisingly, this is accomplished with a single contract rather than with a menu of contracts as in Choné and Ma (2011). However, if the doctor is not able to commit in this way, then a screening menu of contracts must be in place to approach the efficient allocation as close as possible. This entails doctors with different levels of altruism to vary in their treatment practices. In any case, the efficient allocation can never be implemented without commitment. In both cases (commitment and no commitment) the authors provide a characterization of the contracts, which will consist of a capitation fee and a proportional co-payment of treatment costs.13 Many other sources (or combinations thereof) of asymmetric information have been studied. As one adds more dimensions of physicians’ private information, additional tools for screening need to be used in order to improve efficiency. As we have seen, Blomqvist proposes the possibility of malpractice liability. Barigozzi and Burani (2016) assume, realistically, that not only a physician’s level of altruism is unobservable, but also her ability (to be interpreted as a lower cost of providing any level of quality of care). This allows these authors to tackle the issue of whether providers of certain levels of ability and certain levels of altruism do self-select into different types of organizations, namely for profit vs. non-profit hospitals. Several authors have proposed that altruism brings some “wage penalty”: individuals feel rewarded not only

164  The Elgar companion to information economics by the extrinsic (i.e., pecuniary) incentives, but also by the “warm glow” of providing pro-bono treatment. This implies that higher levels of effort do not require so powerful (and therefore expensive) incentives as would be needed to motivate a selfish provider. The employer (say a hospital) will be able to screen physicians according to their ability by offering them a menu of wage and effort contracts, and the type of hospital induces self-selection according to the physician’s pro-social preferences. Hence the existence of different institutions serves as the additional screening device. As one would expect, the relatively more selfish providers self-select into for-profit hospitals. Within such providers, the more able choose demanding but highly paying tasks while the less able self-select into less-demanding and low-pay tasks. In contrast, the altruistic physicians self-select into non-profit hospitals (and choose tasks in the same fashion as the less altruistic). An important but credible assumption here is that ability and altruism are distributed independently across the population. This brings light into the true existence of a wage penalty. In this regard, it turns out that the selection into profit vs. non-profit hospitals is ability-neutral, so wage disparities can entirely be attributed to the level of altruism. Still, the non-profit hospital manages to implement higher levels of quality than the for-profit hospital. 3.3 Referrals The same simple model offered in the previous subsection can of use here. Suppose that C (for cheap) represents the outcome where the GP treats the patient herself (no referral), and E (for expensive) represents a referral to the specialist. To make things interesting, let us assume that it is appropriate (from the cost/benefit analysis) to refer the patient if he is in severe condition whereas the GP should treat the patient herself if his condition is mild. It quite easy to see that a FFS payment system will tend to induce “undertreatment”, by which I mean that the patent is not referred despite being in severe condition. Conversely, if a patient in mild condition is referred, we can speak of “overtreatment”. The usual assumption giving rise to both phenomena is that diagnostic results are the GP’s private information. The literature has focused on the positive approach. In other words, the authors take the remuneration mechanism as given (based on the real-world cases) and compare the incentive effects and the performance of each mechanism. Allard et al. (2011) study the game that starts when the patient is already in the GP’s office, whereas Brekke et al. (2007) and González (2010) compare the GP gatekeeping system where visiting the GP for a possible referral to a specialist is mandatory with a system where the patient is allowed to visit the specialist directly. We will concentrate on the first work, as the other works do not focus on the asymmetric information problem.14 Allard et al. (2011) also introduce diagnostic ability as an additional source of private information. They compare three remuneration schemes that are (or have been) in place in different countries: from the most retrospective to the most prospective; FFS, capitation, and Fundholding. The first two have already been defined. In the Fundholding system, the GP is not only held responsible for the treatment costs at her office, but also for the specialist’s treatment costs. An important ingredient in Allard et al.’s (2016) model is whether the GP “follows” the result of the diagnostic procedure, that is, whether the GP refers the patient if and only if the test results suggest that the patient is in severe condition. This is determined by two forces: first, how trustworthy is the diagnostic signal, which is in turn determined by the GP’s diagnostic ability; and second, the remuneration system that she is subject to. It is quite obvious that under FFS the GP will tend to ignore the diagnostic results and treat

Asymmetric information in health economics  165 the patient herself. Under capitation, the GP is bearing the full cost of the treatment so now the GP will tend to again ignore the diagnostic results but now always refer the patient to the specialist. This last effect is corrected under the Fundholding system, since the GP bears the specialist’s costs as well. The authors introduce an additional source of incentives. Namely, if a patient in severe condition is not referred, his condition worsens, which is internalized by the doctor due her altruistic preferences. One of the main insights from their analysis is that Fundholding and FFS lead to similar GP behaviour. Less altruistic GPs will tend to treat patients themselves, either to increase revenues in the FFS system, or to save specialist costs in the Fundholding system. Moderately altruistic GPs will behave according to their ability. More able doctors will follow the diagnostic results. Less able doctors will tend to always refer the patients in order to avoid the worsening of the patient’s health. Finally, under capitation, as mentioned above the GP tends to always refer no matter the diagnostic result, and this is so no matter what their level of ability or altruism is. Now, which incentive mechanism is best for the regulator? This depends on the weights that this regulator puts on saving healthcare costs and on patients’ well-being. If the first objective dominates then avoiding unneeded referrals is optimal, so the regulator should opt for either FFS or Fundholding. In the opposite case, capitation is the optimal arrangement. This last result stands in sharp contrast with the common wisdom that prospective payments systems tend to save costs. This is only true if the doctor is forced to treat the patient. Here, in contrast, the GP is allowed to transfer the patient to a more expensive treatment, that is, the one carried out by the specialist. 3.4

Empirical Evidence

Although empirical evidence of the impact of other-regarding preferences on actual behaviour is scarce, the existence of heterogeneity in these preferences has been pointed out as one of the reasons for the mixed evidence on the effect of pay for performance (P4P).15 Unfortunately, it is quite hard to elicit such preferences from observational data, so most of the current evidence is based on laboratory experiments, which we will discuss in the next subsection. Basically, such experiments allow us to clear up several confounding factors that come in the way when using observational data: “financial incentives, peer pressure, uncertainty about costs and benefits, and fear of being sued for malpractice” (Kesternich et al., 2015, p. 2). 3.4.1 Experimental evidence Controlled experiments are key in eliciting the effects of non-pecuniary incentives. They allow to design treatments that screen out the confounding effects mentioned above. They can even be designed in order to identify and separate the two types of other-regarding preferences that were mentioned in subsection 3.1, namely, caring for the patient’s well-being and caring for society as a whole (the latter in reference to financial costs). Kesternich et al. (2015) conduct an experiment with medical students where an increase in treatment costs brings a reduction in the (factual) contribution to a charity fund. They find that altruism towards the patients may induce an inefficient increase in costs. Experiments can also be used to study the interaction between remuneration systems. Brosig-Koch et al. (2021) compare the combination of FFS and P4P with the combination of capitation and P4P. They find that, even after controlling for all the treatment variables, some remaining heterogeneity remains in subjects’ choices, and propose that heterogeneity in social preferences may explain this phenomenon.

166  The Elgar companion to information economics

4.

CONCLUDING DISCUSSION

Contract theory and, more generally, the economics of information, have been proven to be useful tools in understanding the phenomena that arise when agents possess private information. They are also useful in the design of optimal contracts. These tools, combined with the now well-understood role of other-regarding preferences, like altruism and cost sensitivity, are useful in explaining several apparently paradoxical phenomena, like some unexpected effects of pecuniary remuneration (inappropriate referrals or induced demand, to mention two examples) or the choice of excessive insurance coverage. I have focused on issues related to pure asymmetric information, or hidden type and in doing so I have neglected important issues related to moral hazard, or hidden action. Examples of the latter are precautionary effort on the part of the insuree, or cost containment and diagnostic and treatment effort on the part of doctors, for example. It is often the case that both asymmetric information and moral hazard coexist, and this requires even more sophisticated incentive mechanisms. Yet, in reality, contracts are relatively simple, like FFS or capitation. Theoretical models are useful in predicting the adverse effects of these simple mechanisms and to propose contract regulation policies. Further experimental evidence is needed to confirm whether these adverse effects are in fact present. This evidence can come both from randomized control trials like the Rand Health Insurance Experiment conducted between 1974 and 1977, or from laboratory experiments, like those discussed in subsection 3.3.1. Finally, PAI theory and empirical enquiry have allowed us to provide a positive answer to the formulated in the title. Let us start with the health insurance market. On the one hand, in the absence of market power, we have seen that the insurance contracts arising in the laissez faire situation are constrained-efficient, which means that a regulatory authority that is able to use the same variables as insurers cannot bring changes that benefit all parties simultaneously. However, we have also seen that efficiency and cross-subsidization (vertical equity) can be fostered by limiting categorization (whereby insurers are not allowed to use certain variables to underwrite contracts). In contrast, introducing minimum coverage legislation requires imposing mandatory enrolment, which is binding for the low risks. This points to the fact that low risks may be harmed by such legislation. As for the doctor–patient sphere, an important lesson derived from the normative approach is that some degree of over- and under-treatment is unavoidable (even under optimal contracts) in the presence of PAI. This means that observing some distortions, e.g., excessive referrals, is not a reason per se to introduce contract regulations. However, the positive approach has shown that some commonly-used contracts (like FFS or capitation) can be improved upon in order to achieve both efficiency and equity gains.

NOTES 1.

For a thorough review of the models and tools used to deal with the presence of both asymmetric information and imperfect information, see Chapter 2 (Stiglitz & Kosenko, 2024a, pp. 20–52). Another constraint of our analysis is that we solve the game once all agents have already decided how much information to acquire (exogenous information structure). This additional restriction impedes the analysis, for example, of the decision to take a genetic test, an issue that has received a lot of attention in the last three decades. For a review of the models of information acquisition (endogenous information structures) see Chapter 3 (Stiglitz & Kosenko, 2024b, pp. 53–80).

Asymmetric information in health economics  167 2. 3. 4.

5. 6. 7.

8. 9.

10. 11.

12. 13. 14.

15.

For works that characterize the optimal contracts when both the physician’s level of altruism is her private information (leading to PAI) and her treatment effort is unobservable (leading to moral hazard), see, for instance, Jack (2005) and Jelovac and Kembou Nzale (2020). In relation to this, an interesting PAI problem arises when the true extent of damages is only known by one of the parties. See, for instance, Daughety and Reinganum (2005). Other issues in the patient–physician relation bring in pharmaceutical firms as a third player. Policy discussion focus on the possible prohibition of direct-to-consumer advertising and the limitation of detailing strategies (where pharma representatives visit physicians’ offices to promote the prescription of specific drugs). For the first issue see, for instance, Brekke and Kuhn (2006). On the second issue, see, for instance, Beilfuss and Linde (2021). The term “adverse selection” is often used as a synonym for “asymmetric information”. I will stick to the stricter sense of the term, that is, the fact that riskier individuals purchase more coverage. See Buchmueller and Dinardo (2002) for some evidence on this phenomenon. The fact that massive amounts of data together with artificial intelligence may drastically reduce the extent to which two players have asymmetric information is largely discussed in Chapter 7 (Pernagallo, 2024, pp. 135–153). One question here is whether regulation could impede the use of some of this information in underwriting contracts. Here I draw heavily from the seminal article by Rothschild and Stiglitz (1976). Notice that, in general, it is impossible to screen agents if contracts are unidimensional. In our case, imagine two contracts with the same coverage and different premiums are offered (everything else equal). It is obvious that agents will choose the cheaper contract. The same goes if two contracts require the same premium but one is more generous than the other in terms of coverage. It turns out that such unidimensionality (in the variation among contracts) is what leads to market unravelling, which, in the context of insurance, is also referred to the “spiral of death” we saw in section 2.1. See also the discussion in section 4 of Chapter 5 (Giza, 2024, pp. 106–117). Lamiraud and Stadelmann (2020) show that this may explain the lack of competition in the Swiss health system. We focus on situations where physicians’ (our experts) private information and conflict of objectives vis-à-vis patients can be dealt with by contracts. Moreover, the decision (prescription, referral) is taken by the more informed party, so we are in the framework termed “delegation” in the experts literature. See Chapter 11 for an overview of the literature on experts where contracts are not available but the decision may rest on the less-informed party (Pavesi et al., 2024, pp. 202–223). This result is valid for a large set of distributions of patients’ needs and physicians’ altruism. Another contribution of this work is to analyse the possibility that a sequence of treatments (rather than a single treatment) may be prescribed. González (2010) introduces the possibility that the patient has some information on her true need of referral and the patient can put pressure on the GP to refer him or her to the hospital. She shows GP gatekeeping is only welfare improving if the patient’s information is either very inaccurate or, surprisingly, too accurate. Brekke et al. (2007) introduce competition among hospitals that may differentiate their healthcare through specialization. GP gatekeeping would make the patient more aware of the differences among hospitals and this would soften competition, as hospitals incur excessive specialization. See, for instance, the discussion in Brosig-Koch et al. (2021).

REFERENCES Allard, M., Jelovac, I., & Leger, T. (2011). Treatment and referral decisions under different physician payment mechanisms. Journal of Health Economics, 30(5), 880–893. Arrow, K. J. (1963). Uncertainty and the welfare economics of medical care. American Economic Review, 53(5), 941–973. Barigozzi, F. & Burani, N. (2016). Competition and screening with motivated health professionals. Journal of Health Economics, 50, 358–371.

168  The Elgar companion to information economics Beilfuss, S. & Linde, S. (2021). Pharmaceutical opioid marketing and physician prescribing behavior. Health Economics, 30(12), 3159–3185. Blomqvist, A. (1991). The doctor as a double agent: Information asymmetry, health insurance, and medical care. Journal of Health Economics, 19(4), 411–432. Brekke, K. R. & Kuhn, M. (2006). Direct to consumer advertising in pharmaceutical markets. Journal of Health Economics, 25(4), 102–130. Brekke, K., Nuscheler, R., & Straume, O. (2007). Gatekeeping in health care. Journal of Health Economics, 26(1), 149–170. Brosig-Koch, J., Gross, M., Hennig-Schmidt, H., Kairies-Schwarz, N., & Wiesen, D. (2021). Physicians’ Incentives, Patients’ Characteristics, and Quality of Care: A Systematic Experimental Comparison of Fee-for-Service, Capitation, and Pay for Performance. Ruhr Economic Papers, No. 923. https://​ideas​ .repec​.org/​p/​zbw/​rwirep/​923​.html. Buchmueller, T. & Dinardo, J. (2002). Did community rating induce an adverse selection death spiral? Evidence from New York, Pennsylvania, and Connecticut. American Economic Review, 92(1), 280–294. Cardon, J. H. & Hendel, I. (2001). Asymmetric information in insurance: Evidence from the National Medical Expenditure Survey. The RAND Journal of Economics, 32(3), 408–427. Chiappori, P. A., Jullien, B., Salanié, B., & Salanié, F. (2006). Asymmetric information in insurance: General testable implications. The RAND Journal of Economics, 37(4), 783–798. Choné, P. & Ma, C. A. (2011). Optimal health care contract under physician agency. Annals of Economics and Statistics, 101/102, 229–256. Daughety A. F. & Reinganum, J. F. (2005). Economic theories of settlement bargaining. Annual Review of Law and Social Science, 1, 35–59. De Meza, D. & Webb, D. C. (2001). Advantageous selection in insurance markets. The RAND Journal of Economics, 32(2), 249–262. Ellis, R. P. & Layton, T. J. (2014). Risk selection and risk adjustment. In T. Culyer (Ed.), Encyclopedia of Health Economics (pp. 289–297). Amsterdam: Elsevier. Ettner, S. L. (1997). Adverse selection and the purchase of Medigap insurance by the elderly. Journal of Health Economics, 16(5), 543–562. European Union (2012). Guidelines on the application of council directive 2004/113/EC to insurance, in the light of the judgment of the court of justice of the European Union in case C-236/09 (Test-Achats). Document 52012XC0113(01). Official Journal of the European Union, 55, 1–11. Fang, H., Keane, M., & Silverman, D. (2008). Sources of advantageous selection: Evidence from the Medigap insurance market. Journal of Political Economy, 116(2), 303–350. Giza, W. (2024). Asymmetric information as a market failure in retrospect. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 106–117). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Glazer, J. & McGuire, T. G. (2000). Optimal risk adjustment in markets with adverse selection: An application to managed care. American Economic Review, 90(4), 1055–1071. González, P. (2010). Gatekeeping versus direct-access when patient information matters. Health Economics, 19(6), 730–754. Jack, W. (2005). Purchasing health care services from providers with unknown altruism. Journal of Health Economics, 24(1), 73–93. Jack, W. (2006). Optimal risk adjustment with adverse selection and spatial competition. Journal of Health Economics, 25(5), 908–926. Jelovac, I. & Kembou Nzale, S. (2020). Regulation and altruism. Journal of Public Economic Theory, 22(1), 49–68. Keane, M. & Stavrunova, O. (2016). Adverse selection, moral hazard and the demand for Medigap insurance. Journal of Econometrics, 190(1), 62–78. Kesternich, I., Schumacher, H., & Winter, J. (2015). Professional norms and physician behavior: Homo oeconomicus or homo hippocraticus? Journal of Public Economics, 131, 1–11. Lamiraud, K. & Stadelmann, P. (2020). Switching costs in competitive health insurance markets: The role of insurers’ pricing strategies. Health Economics, 29(9), 992–1012. Liu, T. & Ma, A. (2013). Health insurance, treatment plan, and delegation to altruistic physician. Journal of Economic Behavior & Organization, 85, 79–96.

Asymmetric information in health economics  169 McCarthy, D. & Mitchell, O. S. (2010). International adverse selection in life insurance and annuities. In S. Tuljapurkar, N. Ogawa, & A. H. Gauthier (Eds.), Ageing in Advanced Industrial States: Riding the Age Waves (pp. 119–138). Dordrecht: Springer Verlag. McFadden, D., Noton, C., & Olivella, P. (2015). Minimum coverage regulation in insurance markets. SERIEs, 6(3), 247–278. Miyazaki, H. (1977). The rat race and internal labor markets. Bell Journal of Economics, 8(2), 394–418. Myerson, R. B. (1979). Incentive compatibility and the bargaining problem. Econometrica, 4(1), 61–73. Olivella, P. & Schroyen, F. (2014). Multidimensional screening in a monopolistic insurance market. The Geneva Risk and Insurance Review, 39(1), 90–130. Olivella, P. & Vera-Hernández, M. (2007). Competition among differentiated health plans under adverse selection. Journal of Health Economics, 26(2), 233–250. Olivella, P. & Vera-Hernández, M. (2013). Testing for asymmetric information in private health insurance. The Economic Journal, 123(567), 96–130. Pavesi, F., Argelli, N., & Scotti, M. (2024). Information and expertise. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp.  202–223). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Pernagallo, G. (2024). Overcoming information asymmetry: A data-driven approach. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 135–153). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Rothschild, M. & Stiglitz, J. E. (1976). Equilibrium in competitive insurance markets: An essay on the economics of imperfect information. The Quarterly Journal of Economics, 90(4), 630–649. Spence, M. (1978). Product differentiation and performance in insurance markets. Journal of Public Economics, 10(3), 427–447. Stiglitz, J. E. & Kosenko, A. (2024a). Robust theory and fragile practice: Information in a world of disinformation. Part 1: Indirect communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 20–52). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. & Kosenko, A. (2024b). Robust theory and fragile practice: Information in a world of disinformation. Part 2: Direct communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 53–80). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Wilson, C. (1977). A model of insurance markets with incomplete information. Journal of Economic Theory, 16(2), 167–207.

9. Consequences of information asymmetry in a syndication network: The joint investments of the Israeli venture capital funds Ilan Talmud1

1. INTRODUCTION Economic sociologists have demonstrated how relational patterns create power inequality between economic actors, resulting in differential returns (or rents). An actor’s social position in a network of economic exchange impacts her relative position vis-à-vis others (Burt, 1983, 1992, 2007; Talmud, 1992, 1994; Talmud & Mesch, 1997). Venture capital funds (VCFs) have emerged as an important intermediary in financial markets, providing capital to young high-technology firms that might have otherwise gone unfunded. Venture capital (VC) serves as a professional asset management activity that by raising money from wealthy individuals and institutional investors invests into new ventures with risky ideas, but also with a high potential to grow (Sahlman, 1990). VC faces dramatic information asymmetry and market uncertainty while investing in entrepreneurial technological firms, especially at the start-up stage of emerging technology (Darr & Talmud, 2003; Yitshaki-Hagai, 2003; Du et al., 2020). These information asymmetries stem mainly from the contextual and latent nature of the knowledge economy of technological firms (Darr & Talmud, 2003). To overcome these impediments, VCFs tend to specialize in a technological cluster or a market niche, and to co-invest with other VCFs. The combination of these two contrasting principles, specialization and co-investment, creates a network of asymmetric exchange among VC funds. More specifically, the attempt to cope with severe degrees of uncertainty, relatively turbulent environment, high rate of start-ups turnover, and information asymmetry, VC funds co-create a network of co-investments, leading to the creation of a network of co-determination (see Talmud, 1994). However, the co-evolution of the network does not at all imply an equality in the outcome of VCFs’ investments. On the contrary, a VCF’s position in a network may determine its ability to cope with information asymmetry and thus predict its overall performance. The network of shared investments thus benefits certain VCFs at the expense of others. Joint investments, accordingly, depict the structural inequality between VCFs and their leverages to grapple with information asymmetry and to accrue resources from the VCFs’ social space.

2.

STATE OF THE ART

According to a literature review, the main topics in VC research are: (1) heterogeneity (e.g. in affiliation, experience, reputation) within the VC and its effects; (2) the causal link between VC financing and various aspects of company performance, such as growth or innovation; 170

Consequences of information asymmetry in a syndication network  171 (3) the performance of VCF and how to measure it properly; (4) internationalization; and (5) the processes through which VCFs and entrepreneurs select and match with one another. Additionally, most studies use data on the USA (Tykvová, 2018). This chapter examines the links between VC’s affiliation network and performance. Information is a critical resource for any organization. For venture capitalists, information is a cardinal resource. Asymmetric information is a common feature of market interaction, but in the high technology arena it is more pronounced. Due to the high returns and high-risk nature of venture capital, the consequences of the information asymmetry problem will be far greater than for those in other industries. Notably, asymmetric information may arise following first contracting (Trester, 1998). It is not merely that the seller of goods often knows more about their quality than the prospective buyer. In this sector, the “buyer”, the entrepreneur, seeking an investment from a VCF is often not very knowledgeable about the VCF’s preferences and overall interest. On the other hand, investors typically do not possess complete information regarding the entrepreneur’s future conduct, and similarly they are lacking complete information on their potential and actual partners in their joint investments network. Other sources of information asymmetry are: limited rationality, investment and agency costs, and monopoly over information resources (Du et al., 2020). Often, VCFs operate in a risky area, typified by technological discontinuity (Tushman & Anderson, 1986), high turnover rate, and the prominence of tacit knowledge (Aspers, 2009). Over and above the governing condition of information asymmetry, start-ups and VCFs face preconditions of “inscrutable markets”. Unlike markets that are subject to “simple” asymmetric information, in inscrutable markets not just buyers but sellers too do not know much about the quality of their products (Gambetta, 1994; Darr & Talmud, 2003). Because buyers cannot discern which products are of good quality and sellers cannot construct “honest” signals to guide them, rational choice theory predicts their failure. Yet these markets do emerge. Gambetta (1994) showed that the market is constructed and maintained by agents provoking associations between the inscrutable commodity and a string of interrelated signals that are costly or impossible to imitate. It consists of gaining a distinctive identity. Inscrutable markets are fundamentally uncertain. Empirically examining the VC market, Podolny and Castellucci (1999) distinguish between VCF’s altercentric and VC’s egocentric uncertainty. Altercentric uncertainty refers to the uncertainty that buyers face about the product quality of a focal producer (ego). By contrast, egocentric uncertainty refers to the uncertainty that the producer herself faces about the resource allocation decisions that will result in a product regarded as high quality by buyers. Podolny and Castellucci show that high status producers seek out markets or market segments where egocentric uncertainty is low. Comparably, Hughes-Morgan and Yao (2016) show that firms with a favorable network position are better at appropriating rents, spotting opportunistic behavior of partners and discouraging involuntary knowledge leakage to partners (see also Niesten & Stefan, 2019; Lebedev et al., 2021). As the capability of a VCF is often limited in scope, it needs to form a partnership with other VCFs. Nevertheless, maintaining a partnership in a turbulent, relatively erratic domain involves a greater uncertainty about the prospective behavior of the partner(s).

172  The Elgar companion to information economics

3.

A NETWORK APPROACH TO ECONOMIC PERFORMANCE

Network analysis inquires into the impact of relationship structure on interdependencies and inequalities among actors (Talmud, 1994). As VCFs use joint investments, attempting to diminish their degree of information asymmetry, network analysis is a promising, efficacious tool to explain unequal outcomes between VCFs. This chapter, then, adds to a long legacy of network analytic studies of firms. According to Burt (1992, 2007), the bridging function between “structural holes” enables entrepreneurs to access information about new opportunities and referrals, and can assist them in using this information at the right time. Thus, it is not just the strength of relations, but the structure of those relations that define market power. Bridging ties provide actors with higher degrees of control, holding “non-redundant information” (Burt, 1992), a better market reach, and a greater ability for calculating timing (Stuart & Sorenson, 2007). VC supports entrepreneurial firms in developing new connections with distant sectors (Yitshaki-Hagai, 2003), enabling the entrepreneurial firms to access new information and opportunities (Powell, 1990; Nohria, 1992). It seems that the ability of VCFs to develop connections for their portfolio companies depends on their centrality in the network, and the degree of expertise of the VCF in the industry in which the entrepreneurial firm operates (Tenenbaum & Norton, 1993; Podolny & Feldman, 1997; Stuart et al., 1999). Studies consistently show that though strategic networks and syndicated ties are ubiquitous, they are far from being isomorphic (Sorenson & Stuart, 2001). Network structures dramatically vary across industrial and business contexts in ways that hamper oversimplification, often making network strategy context-specific. More to the point, while it is self-evident that inter-organizational networks formed by joint ventures and alliances are inherently strategic in nature, it is not at all clear what types of network structure yield strategic success and entrepreneurial efficacy. Numerous network studies have demonstrated how uncertainty is managed via concrete social ties (Podolny, 1993, 1994, 2001, 2005; Podolny & Feldman, 1997; Podolny & Castellucci, 1999; Uzzi & Gillespie, 1999). For example, investment banks create tight networks to overcome financial market uncertainty (Podolny, 1993, 1994). Strong ties assist in overcoming the uncertainty associated with the introduction of new products (e.g. Krackhardt, 1992; Rogers, 1995; Freeman, 1999; Darr & Talmud, 2003). Female middle managers who are successful in breaking through the corporate “glass ceiling“ are typically promoted through strong ties, while male managers are promoted via weak ties (Granovetter, 1982; Burt, 1992, chapter 4). Uzzi (1997) claims that sellers and buyers perceive the construction of embedded, rather than arm’s-length ties, as an efficient exchange behavior, which are devised to counter the uncertainties that plague the New York garment industry. In a similar way, in the hyper-competitive and uncertain bio-technology industry, inter-organizational exchange and learning is facilitated mainly through trustful cooperation between firms (Powell et al., 1996; Lundvall, 1995; Liebeskind et al., 1996; Lütz, 1997; Lipparini & Sobrero, 1999). Thus, embedded ties are a form of economic exchange designed to cope with market uncertainty and to facilitate the transfer of contextual or expert knowledge (Gambetta, 1994; Knorr-Cetina, 1999; Oliver & Ebers, 1998; Darr & Talmud, 2003). VCFs particularly confront information asymmetry as well as technological and market uncertainty. To succeed, VCFs and entrepreneurial firms seek to lower the level of uncertainty and to balance against information asymmetry (Larson, 1992; Cooper & Kleinschmidt, 1995)

Consequences of information asymmetry in a syndication network  173 by working with a few VCF funds that specialize in different technological niches. On the other side, VC faces a two-fold risk: (a) risks associated with informational asymmetry about the entrepreneur’s ability (Amit et al., 1990; Brander et al., 2002), and (b) uncertainties stemming from the contextual, non-standard nature of the firm’s technological knowledge (Darr & Talmud, 2003). Often, information asymmetry stems from the nature of uneven ownership of contextual or tacit knowledge that is transferred only by social interaction among agents (Collins, 1993, 1995; Knorr-Cetina, 1981, 1999; Nahapiet & Ghoshal, 1998; Powell, 1990; Aspers, 2009). Darr and Talmud (2003) compared an inter-organizational network in markets for emergent technologies, in which products are fully customized, to a network observed within a more standard mass market. They found that the inherent problem associated with the transfer of contextual and non-standard knowledge tightened the structure of micro-level ties, placing technological experts at the center of communication (Darr & Talmud, 2003). Information between entrepreneurs and venture capitalists is often shared unequally. VCF managers are experienced and professional deal makers, while entrepreneurs have great knowledge about their venture but usually limited knowledge about venture capitalists’ financing process and requirements. For entrepreneurs, venture capitalists’ asymmetric information advantage can lead to difficulties in receiving funding, unfavorable terms, or negative start-up experiences (Glücksman, 2020). In order to reduce uncertainty and risk, VCF managers tend to cooperate and co-invest (Bygrave, 1988; Podolny & Castellucci, 1999; Hellmann, 2002; Yitshaki-Hagai, 2003). Cooperation between them expands their social connections, hence providing a greater accessibility to networks for their portfolio companies (Stuart & Sorenson, 2007; Hughes-Morgan & Yao, 2016; Lebedev et al., 2021). It seems that information asymmetry has even increased, as works published in the last two decades show the evolution in the preference of factors with the focus shifting from the venture team and product to factors such as intellectual property rights, economic crisis and social capital (Vazirani & Bhattacharjee, 2021). Most of the strategic advice offered for mitigating information asymmetry’s hurdles is based on a dyadic-level analysis (Cumming & Johan, 2008; Du et al., 2020; Glücksman, 2020). In reality, however, VCF managers exchange information with a web of partners (Sorenson & Stuart, 2001). Consequently, they can have additional information sources, such as through third parties, about the activities their partners are undertaking. Nevertheless, while information is spreading via social and economic networks, its diffusion is unevenly distributed (Jackson, 2008; Knoke, 2012). Network structure also affects the information benefits a firm can obtain, thus influencing the ease of monitoring and deterrence for certain firms. Different structural locations in inter-organizational networks may variably affect a venture capitalist’s ability to benefit from the opportunity structure implied in the network. More precisely, it has been recurrently found that centrally located firms in a network will have a higher rate of success (Burt, 1992, 2000; Talmud, 1992, 1994; Talmud & Mesch, 1997; Hughes-Morgan & Yao, 2016). More specifically, centrally located firms are exposed to richer external resources, having higher control and flexibility of resource allocation to achieve their organizational goals. Because these central firms have a wider range of partners from which to acquire such knowledge and access resources, they have more privileged channels and own more stockpiles of asymmetric information. Yet, asymmetric information increases also because of a better access to alternative alliances, which, in turn, improve a firm’s bargaining power (Lavie, 2006, 2007). Central network locations can also help firms to better select partners and even monitor

174  The Elgar companion to information economics partners’ behavior, thus deterring them from taking opportunistic action (Hughes-Morgan & Yao, 2016). A brokerage position in a network provides a VC firm with structural leverage to form more non-redundant ties which facilitate information channels, while a one that is located at a more peripheral position will have fewer connections in the network, a narrower market reach, and hence is more likely to experience delayed or missed information relevant to investment opportunities. Additionally, this difference implies asymmetric ability to monitor partners’ conduct. VC firms located in a closed dense network cluster are exposed to more transparent information and peer control, and therefore are deterred from opportunistic behavior, while VC firms with high betweenness centrality have fewer external network constraints, thus having a better chance to take unobserved actions for their benefit (Hughes-Morgan & Yao, 2016; Lebedev et al., 2021). Brokerage and closure are two mechanisms of handling information asymmetry and market uncertainty (Burt, 2007). While brokerage facilitates the creation of exploitative rent appropriation, it is not the only strategic device used by business organizations to cope with information asymmetry. In many cases, firms utilize – paradoxically – the opposite mechanism: embedding transactions in a cluster of strong, transitive ties. Transitivity serves as a mutual control mechanism and reputation device. It prevents opportunistic behavior by providing shared information flows regarding actors’ conduct across the dense cluster. Social transitivity thus assures the manufacturing and maintenance of trust over time, thus raising the odds of spreading uniform information across the networks. More to the point, research found that strong ties and dense networks are key structural foundations in the creation of social capital and the production of composite rent by means of joint value creation and risk sharing. Moreover, this effect was found in many empirical instances, from education to the garment industry (Coleman, 1988, 1994; Uzzi, 1997; Talmud & Mesch, 1997). In another context of market information asymmetry, it was found that co-affiliation in a bank by actors from a credit-seeking firm’s network increases the amount and quality of the information available pertinent to that firm, hence providing the firm with the opportunity to access capital more closely aligned with its true credit-worthiness. In line with network theory, it was shown that even very low levels of co-affiliation increase a firm’s credit availability (Sasson & Fjeldstad, 2009). In other words, relational management reduces information asymmetry (Jolink & Niesten, 2021). Moreover, these relational devices take many forms (Jackson, 2008; Knoke, 2012). Talmud and Mesch (1997) and Uzzi (1997) show how two mechanisms – the creating of non-redundant ties to otherwise unrelated market segments and highly dense local network clusters – balance off one another, thus increasing the odds of organizational survival. It was found that organizational survival was positively associated with a “mixed model” network, comprising both brokerage and cohesion, providing a combined advantage. Put differently, while betweenness centrality provides the firm with effective capacity for information exploration, local density furnishes the valuable capacity for information exploitation (Talmud & Mesch, 1997; Uzzi, 1997; Hansen et al., 2001).

4.

DATA AND METHODOLOGY

The analytical strategy of this study is a mixed method approach (Levasseur et al., 2022).

Consequences of information asymmetry in a syndication network  175 Quantitative data, which is the main set of the analysis, is available through the Israel Venture Capital Research Center’s annual sourcebooks of the Giza Group, 1995–2002 (Giza Group, 2000–2002). These volumes annually report key attributes of VC funds (such as the annual amount of capital raised), specifying each investment of a VC fund i in an entrepreneurial technological firm k for each year. To provide qualitative supplementary data, 53 semi-structured interviews were conducted with key personnel and senior industry analysts, reflecting various industry segments. The aims of the interviews were twofold: (a) to gain additional sensitizing concepts of relational management (such as informal networks of VC partners), and (b) to verify the external validity of the existing research framework (Figure 9.1). The dependent variables are as follows: 1. 2. 3. 4.

Number of exits Rate of exits Number of failures2 Rate of failures.

The independent variables are network betweenness centrality and VC’s local density. Network betweenness centrality (Freeman, 1977, 1979; Wasserman and Faust, 1995), is the degree to which an actor controls access to non-linked clusters. It is expressed as: ​CB​  ​ (​ni​ ​) ​  =​   ∑ ​ ​ ​g  jk​  ​(​ ​ni​ ​)​ / ​gjk​  ​ j  0​ and ​p  =  1/2​) that demonstrates that nondisclosure of the expert’s COI may be beneficial to informative communication between the expert and the decision maker. Morgan and Stocken (2003) consider the context where a stock analyst may or may not have a COI that compels him to pump up the price of a stock. Their theoretical setting (​​bl​ ​  =  0  d and a >! (b  +   d)/2.

For a “well-mixed” population (with initial random encounters of each two agents) in an evolutionary sequential interaction and replication process (the EGT frame), a differential replication of population shares of many possible strategies applies. Then, a condition for the non-invadability of a population applying the strategy of institutionalized (conditional) cooperation (the well-known “tit-for-tat” strategy – TFT) by defectors (the always defecting strategy – ALL-D) is that incumbent cooperators should at least play better against each other in indefinitely repeated games (SG) than defectors against incumbent cooperators: PTFT/TFT >! PALL-D/TFT.

280  The Elgar companion to information economics The superiority (in the sense of non-invadability) of such contingent cooperation (a cooperative but also responsive strategy, a so-called “trigger strategy”) then is given by the capitalized current values of the different relative future SG-payoffs (the capitalized limit values of the sums of two infinite geometric series of future payoffs), δ being the discount factor: ​a / ​ (1 − δ) ​   >  ! b − c  +   c / (​ ​1 − δ​)​

or ​δ  >  ! ​(​ ​b − a​)​ / ​ (b − c) ​.​

Note, cooperators interacting with their kind always get a, while ALL-D-defectors against TFT-cooperators receive b upon their first exploitation of conditional TFT-cooperators (as cooperators always start cooperative), but thereafter only c (as conditional cooperators will then retaliate). This simple (but not simplistic) solution established a class of models determining conditions for behavioural institutionalization, i.e., an emergent institutionalized cooperation, for the superiority of instrumental (i.e., problem-solving) cooperation to solve ubiquitous SDs. Behavioural cooperation would then be the superior, more successful, and thus an evolutionarily stable strategy (ESS). Emergent cooperative behaviour, solving the SD-problem, then turns out to be an or (THE) emergent innovative behaviour as compared to the (presumedly incumbent) short-run individualist hyper-rational benchmark behaviour ALL-D. Note that obviously, this solution (an interactively learned and habituated behaviour in the process of innumerable interactions and many evolutionary differential replications among the competing strategies) can only be realized non-hyper-rationally, as a learned and habituated behavioural institution. A social institution thus must be a social rule with an endogenous sanction, i.e., a threat of retaliation by a cooperative trigger strategy, in order to prevent defective exploitative short-run maximization, which would result in general defection. With hyper-rational behaviour, a solution could never come about, as agents would myopically play a series of one-shot games, always defecting, and never learning, never developing and culturally acquiring a longer-run perspective and related longer-run rationality and calculation. While only the latter can render cooperation superior, as formally represented by the infinite geometric-series calculus (substantially an indefinite future between each two agents), then benefiting all in the long-run. Hyper-rationality thus must be replaced by a rule-based behaviour, justified through the longer-run perspective, rationality, and calculation as indicated. Longer-run oriented agents, then, will be “rational fools” (Sen, 1977), i.e., fools in the strictly neoclassical sense, but rational with their collective long-run improvement. In all, an innovative behaviour. Also note, the longer time horizon is reflected in a larger δ, equivalent with the probability in each interaction that the next interaction will take place, with some positive probability of meeting the same interaction partner again and potentially sanction him (positively or negatively) for his earlier behaviour (or being sanctioned oneself, resp.). Such expectation of futurity will typically feed back into agents’ current considerations, calculus (if applicable), and behaviour. Note, finally, that the reverse question of invadability of an incumbent defective population by invading TFT-cooperators, a more demanding question than the preceding one, is formally

Innovation and information  281 not symmetric to that approach, since TFT, according to the preceding EGT-criterion would not be an ESS, as TFT-cooperators against ALL-Ds always lose in the first interaction and then perform as bad as the ALL-D players, so will lose in total. They can only play better than incumbent defectors through interacting in relevant minimal critical masses of their own kind, which is equivalent with making themselves meet their kind often enough (through some clustering in space and/or through somehow extending the more favourable interactive relations that exist among among their own kind), in order to fare better overall. This more general condition for the overall, population-wide superiority of institutionalized contingent cooperation (with k the number of cooperators in a population of size n), then, is the general population approach where both cooperators and defectors meet cooperators with probability k/n and defectors with (n-k)/n: ​(​k​ / ​n)​ ​[​a / (​ ​1 − δ​)​]​  +   ​[​ (n − k) ​ / ​n]​ ​[​c / ​ (1 − δ) ​  +   d − c​]​ ​>  ! ​(​ ​k​ / n​ )​ ​ [​ ​c / ​ (1 − δ) ​​  +   b − c​]​  +   [​ ​(​n − k​)​ / ​n]​ ​[​c / (​ ​1 − δ​)​]​

Thus, conditional cooperators will be more successful overall, and capable of invading and taking over a population of defectors if they 1. are sufficiently many (a sufficiently large k, or k/n, resp.), the critical minimum mass of cooperators or cooperative strategies or actions, and/or 2. playing relatively successful with each other (a relatively large a, and a relatively low c and b, in the given basic real interdependence structure), while defectors against each other are relatively unsuccessful anyway (through the PD structure: a > c), and/or 3. are able to play relatively more frequently and/or longer with their own kind, i.e., with relatively higher probability in each interaction to meet each other again rather than meet with defectors. In this way they generate a higher δ for themselves, as far as possible, while defectors play with a very low δ anyway: They just don’t care and play in a one-shot perspective anyway. For instance, cooperators may group in clusters, if they can change (spatial, social) locations on the network landscape, or try to reduce the change rate of partners by extending successful cooperative relations (some form of partner selection), reduce disembedding and uprooting forms of mobility etc., as far as possible. Certain social agency capacities are required for such a risky behavioural innovation in favour of cooperation (the risk of losing overall against ALL-Ds), when agents strive for Pareto-superior cooperation, improving themselves and the entire socio-economy above the individualistic hyper-rational “benchmark”, as modelled in full-fledged formalizations of an evolutionary process, with differential replication. These usually include, across many different model settings in the literature: ● Innovative agents need to be non-risk-averse, as they easily may be exploited at least once. ● Similarly, they need to be non-envious, as the ALL-D-exploiter (even if the TFT is exploited only once and both switched over to joint cooperation thereafter) would have fared better at the end of the day. What counts for the non-envious innovator, then, is own and general improvement over the individualistic status quo, with a sufficiently large δ so that her payoff from some learned cooperation [d-a+a/(1-δ)] will be larger than her

282  The Elgar companion to information economics status-quo payoff c/(1-δ) (if she previously were a defector) or [d-c+(c/(1-δ)] (if she were a TFT-cooperator in a population of defectors, i.e., short-run maximizers). ● Furthermore, they have to have some memory of their earlier interactions (a logical requirement of cooperative SG-strategies; e.g., one period memory in the simplest case of TFT), while defectors logically have no memory at all, they just do not care. ● Innovators should further be able to increase their knowledge about other agents’ behaviours beyond just their own interaction experience with them, through some monitoring of others’ interactions, through building own reputation (signalling) and using reputation chains and communication of related information with others, in order to improve their information about their social environment and the entire population to increase the probability of welfare-enhancing cooperation from the very beginning of any interaction. ● Innovators should be able to develop some capacity of partner selection in favour of other cooperators (see on self-organization mechanisms below) and of extending the length of successful relations and reducing the length of unsuccessful ones. To develop such capacities, they would be motivated by ● evading the repeated frustration from the negative unintended consequences of hyper-rational one-shot behaviour (always aspiring the maximum, but collectively ending up in the second worst situation, the common defective one-shot original Nash equilibrium, the very collective dilemma); and/or ● aspiring economic improvement (a Pareto superiority of the cooperative solution over the defective original Nash-equilibrium, as recognizable from the PD normal form payoffs) by, e.g., “idle curiosity” (Veblen) to find out what could be gained by innovating one’s behaviour from benchmark status-quo defection to innovative cooperation; and/or ● expressing some “instinct of workmanship” (Veblen), finding some benefit or pleasure in creating something (new) in institutionalized coordination, i.e., cooperatively (e.g., Elsner, 2017). Most crucially, such solution depends on the fierceness of the dilemma (payoff structure) and on the innovators’ culturally acquired longer-run perspective, or “futurity” (Commons, 1990 [1934]; Jennings, 2005), represented formally by a high (subjective, perceived) discount factor, i.e., a (experienced and expected) high probability of “meeting (again)” next interaction. In an evolutionary population, this applies to: ● either the same interaction partner as before, ● or a “knowing” partner, informed about my earlier behaviour through memorizing, monitoring or the reputation chain, ● or in the general population encounter process, to a cooperator occurring with an average probability as experienced in many interactions throughout the population, implying a high societal probability of being punished for past or current defective behaviour next time. Such futurity (δ) must be in a certain inequality relation to the strength (fierceness) of the incentive structure, as the formal inequality solution above illustrates. The fiercer the dilemma interdependence (in terms of the relations of a, b, and c), the larger the future must loom (a high delta) in order to overcome the SD and for cooperation to emerge from an initial population of general defectors.

Innovation and information  283 In a more substantial consideration, the typical (behavioural) innovator must not only be risk-friendly and non-envious, as mentioned, but also must develop and follow her longer-run perspective, largely independent of what others do. So she will start friendly and cooperative, but will not be a masochist and will therefore properly retaliate should she experience attempts of others to exploit her. But she will also look for an appropriate partner selection and social clustering, will memorize, monitor others’ interactions, communicate in the reputation chains, and signal own trustworthiness. But she will be evolutionarily successful also by forgiving, no longer punishing the other one should he switch himself to longer-run rationality as well. This is what Axelrod and many others had found in their countless modelling and simulations of evolutionary selection and replication processes among many potentials more or less cooperative vs. more or less defective and exploitative strategies. And this seems to be exactly what Adam Smith had developed as “prudent” behaviour or what Axelrod derived from his computer tournaments as evolutionarily stable forms of trigger strategies. The evolutionary idea of “fitness” then notably includes, not always to win and maximize in the short-run, but to be successful overall, at the end of the day, fare best across all interactions with all kinds of heterogenous agents and strategies in a population. But also, in all, another fundamental “message” is that “good” innovators can not be just technological inventors/innovators but must be prudent and mentally and behaviourally smart (e.g., Leliveld & Bhaduri (Eds.), 2023). Note that we also are talking here about the (in case) stronger fundamentals of emerged informal structures, namely informal social institutions (also, e.g., Gault et al. (Eds.), 2023). With increasing futurity, an increasing average expectation “to meet (again)”, as, e.g., in a decreasing population or “arena” size (a smaller interaction arena, neighbourhood, peer group, locality, firm cluster, firm network etc.), or with less enforced disembedding/uprooting mobility, thus with less perceived turbulence, or, put differently, with a stronger expectation of meeting future cooperative behaviour, a higher general trust, the disposition towards cooperation will increase (e.g., Elsner, 2012; García-Vega & Huergo, 2017; Lueders et al., 2017; van der Wouden, 2020). So, with more favourable cognitive conditions, SDs will be easier solved in an evolutionary population process, and learned conditional coordination and cooperation as the innovative behaviour will thrive and expand in the population and take it over as the dominant culture. Behavioural innovation and corresponding institutionalization of contingent instrumental cooperation, thus, will have to co-evolve with technological invention and innovation. It will often even (have to) come prior to it, as it lays the foundations for agents to be cognitively able (and then probably emotionally willing and apt, too) to generate or adopt technological innovations (e.g., Frey, 2019). Action, governance, deliberate agents’ self-governance and policy orientations then include: ● The recognition of interdependence, i.e., of the incentive structure (the kind of “game”) and its quantitative fierceness (e.g., a strong or weak SD), therefore, need to be trained among agents. “Recognized interdependence” and feedback-based learning has in fact been an old evolutionary-institutional issue (e.g., Bush, 1999) as well as an elementary EGT condition (looming large in, e.g., E. Ostrom’s work as well). The more directly the relation between own action and others’ reactions can be (cognitively) perceived, the better.

284  The Elgar companion to information economics ● Which, again, is also a question mirrored by the size of the relevant arena, group, cluster, or population. ● Also, the awareness and the weight of futurity are to be increased and developed as a learned and habituated culture, e.g., by supporting a culture of longer-run calculation and rationality, namely in face of often still prevailing adverse incentive structures (favouring short-termism). ● Further, memory, monitoring, and reputation (chain) capacities, a transparency of interaction histories, have to be supported, in this way strengthening expectations “to meet (again)”, and with this, futurity generating cooperative dispositions.10 ● Partly based on those factors, agents need to expect reciprocation, positive as negative, particularly general retaliation (punishment) in case of own exploitative defection. However, as both many model simulations and anthropological research have demonstrated, an overly punishing population may bear excessive punishment costs and usually will not be evolutionarily successful (e.g., Elsner 2021). 3.3

Network Structures, Self-Organization Mechanisms, Optimal Forgetting, and Policy Implications

3.3.1 Network size and structures, cognitive capacity, and innovation In a full-fledged complex modelling and simulation, the basic network structures, initial and evolving, of course need to be specified, as there are many and their interrelations are critical for information, institutionalization, and innovation. “Markets”, in reality, are networks. Ideal, perfect “markets”, the largely deregulated ones in reality (i.e., designed, created, shaped, regulated, “leaned” and institutionally “attenuated” after the simplistic ideal), then would have to be considered, if not the textbook-like dimensionless “point markets”, nor Walrasian-auctioneer systems (i.e., star networks), as would-be complete networks, where ideally each agent is connected to, and might have to interact anytime with, each other, while as said all are considered the “same”. However, a truly complete network, in reality, would quickly generate over-complexity, at least in terms of the number of relations (r), as with a growing number of agents (n), r(n) = n(n-1)/2, an exponential increase of potential interactions / relations. In addition, if agents have to deal with real-world problems, such as coordination or (as seen) the more intricate social dilemma problems, they will always have several options to behave in an ongoing process of many 2×2 interactions in a population. Then, a minimum of four potential relations between each two players exist, and complexity tends to further increase, in terms of the number (and then also quality) of relations: r = 2n(n-1). With, say, eight agents, the number of potential relations would increase accordingly to r = 112, a considerable complexity emerging already. And the mere increase of the number of relations may, in a real-world evolutionary process in a population, be complemented, and in fact worsened for agents’ cognition capacity, by a realization of a heterogeneity of agents (and their many different potential strategies): Heterogeneous agents might emerge according to their different life-experience paths, further causing higher complexity. Namely, with different real-world interdependence structures such as coordination, anti-coordination, or dilemma problems of different variants (stag hunt, chicken, and many more game types) and with different levels of (numerical) fierceness, many different kinds of agents and strategies may emerge in an evolutionary process with strategy

Innovation and information  285 search and ongoing differential replication. So, we stepwise would approach the complexity of reality: many and heterogeneous agents are at the core of systemic complexity. Anthropological, cognition/brain, and behavioural research tell us that human agents have definite limits of cognitive and computational capacity, and, compared to the potential information always generated in a CAS, act under radical uncertainty (as distinct from calculable risk), and, particularly when meeting strangers, ideally act under some initial radical strategic uncertainty. In fact, anthropological research informs us that stable migrating and settling bands of hunters and gatherers had the size order of 35–40 agents, and larger clans (with, on average, less intensive relations and less frequent interactions between any two members) displayed the size order of 150 members, often only transitorily, e.g. for winter settlements. This was cognitively reflected in the phase of early human development, when social-emotional cognitive capacities such as face recognition etc. developed together with the (biologically relatively young) development of the neocortex of the human brain (e.g., Dunbar, 1993, 2008). Genetic, cognitive, neuro, and brain sciences have confirmed this social-brain thesis, and the relevant maximum group sizes that humans can cognitively accommodate even today, i.e., around 40 at relatively high interaction-intensity levels (e.g., Marlowe, 2005). At this size order, if the social structure is sufficiently stable, they may be able to mutually generate sufficiently high expectations of mutual cooperation, or general trust. General trust then will eventually be embodied, and potentially generalized and extended, as a cooperation expectation even

Figure 14.1

Illustration of a network structure with a resulting right-skewed (particularly power-law) centrality-degree distribution of agents (“scale-free network”, (b)), as compared to a “random network” (a)

286  The Elgar companion to information economics vis-à-vis a stranger and even in an expected one-shot interaction with a stranger (e.g., Elsner & Schwardt, 2015, 2015). Considering real-world deregulated markets with a high power concentration, oligopolistic interaction and a resulting structure of hierarchies of hubs and sub-hubs, we will usually end up with so called scale-free networks, displaying same structures at all size scales and with a resulting right-skewed, often even power-law centrality-degree distribution, in turn displaying the same slope at all size orders (scales) of agents. Figure 14.1 shows a quite simple synthetic example of a scale-free network (b) with its strong hub-centrality, as compared to a random network (a) with a resulting normal distribution of centrality degrees (distributions illustrated in Figure 14.2).

Source: Barabási (2002, p. 71).

Figure 14.2

Illustration of a “normal” distribution (based on a “random network”) as compared to a right-skewed (particularly power-law) distribution (based on a “scale-free network”)

Such “long-tail distributions” were empirically found for income and wealth (Pareto, 1896–7), firm sizes (Gibrat’s law), city sizes (Zipf’s Law), asset prices and returns (e.g., Mandelbrot & Hudson, 2006), price changes, or financial-speculation sector agents’ centralities (e.g., Lux & Marchesi, 1999), with further appearances of words in languages, citations in citation networks, and many others in both natural and social systems (earthquakes, avalanches etc.). However, exactly these types of networks and size and centrality distributions that dominate the world through the dominating kind of hierarchical and power-based “market” economies are not particularly innovation-friendly. The great powers may be innovative for themselves in some ways and even extract resources of innovation from dependent agents, but the system as such has systemic disadvantages in terms of innovation: first, relatively little independent clustering, where institutions of cooperation and cultures of innovation would tend to easily emerge, and, second, often a relatively long “diameter”, so that the diffusion of information (and thus also of innovation) would take a relatively large number of steps from any one agent to any other in the system, namely when oligopolistic powers separate their supply chains in fierce rivalry.

Innovation and information  287 In contrast, we should preferably consider so-called “small-world” networks that have some combination of a high cluster coefficient and a short diameter. These would be more affine to both relatively easy institutions/innovation generation and relatively fast information/ innovation diffusion, if not dominated, walled off, and degenerated by oligopolistic power rivalry. For avoiding any “sclerotization” through (over-)opportunism and (over-)conformism, proximity and interaction frequency in neighbourhood clusters with fast institutionalization and standardization need to be balanced with a fast refreshment of ideas through the entire population network by means of long-distance relations. Well-shaped small-world networks display such properties (Watts, 1999).

Source: https://en.wikipedia.org/wiki/Small-world_network.

Figure 14.3

Illustration of a small-world network (resulting in some right-skewed (but not power-law) centrality distribution)

Figure 14.3 illustrates a small-world network, which displays considerably less and “smaller” hubs than scale-free networks, with some clustering around them and more critical long-distance relations. Its centrality-degree distribution will be less right-skewed than a power-law distribution, with less (oligopolistic, power-based) tail risks, somewhat more similar to a normal distribution. Thus, it will also be less prone to major crises. But those gatekeeper positions that connect local clusters might nevertheless become “tail risks” in potentially evolving right-skewed centrality and power distributions, which therefore should be exchanged both in network positioning and among several real human agents in due time periods through a proper complexity policy (more below). So, whether small-world networks will evolve at all or will degenerate into more hierarchical scale-free networks, is in the first instance a question of the basic self-organization mechanisms underlying the formation and evolution of network structures. Also in terms of policies, the tail risks have gained considerable attention. For instance, complexity economists have derived a stance against (informational) protection of large oligopolistic corporations and in favour of fostering informational openness so that strategic

288  The Elgar companion to information economics information will not be monopolized or locked away by some few, but will broadly flow through the economic system (e.g., Gallegati et al., 2007; Allen et al., 2020). In the same vein, others have derived recommendations in particular on patent protection policy (the “intellectual property rights” protection as fundamentally increased in the neoliberal era from the late 1970s onwards) and found that the “battle over patents” still is unsettled and needs to be shifted back by policy intervention towards a much more limited and more temporary right (of corporations) in order to favour information and innovation flows to independent founders and SMEs (e.g., Haber & Lamoreaux, 2021; Krasteva et al., 2020; Coad et al., 2021). Yet other complexity economists found, for instance in the case of China’s developmental catch-up, productivity gains through industrial-policy regulations that have indeed shifted the centrality distributions of entire industries to the “left”, by fostering millions of spin-offs and SMEs, and by controlling and neutralizing the power of the big public and private enterprises through regulations (e.g., Heinrich et al., 2020). Considering further the quality of information flows and of innovation in networks, we may go beyond the general kind of network, into more specific network properties. So, we will first take a brief and exemplary look at some specific aggregate, systemic network property. Thereafter we will briefly consider some, micro-level condition of information exchange and innovation generation among individual agents. Note that the two levels of conditions, systemic and individual, do not have a unique and clearcut relation per se, unless explicitly modelled. Some network properties that facilitate the emergence of innovation cooperation have been identified in a number of models, such as heterogeneous degree distributions and high degree correlations. Both refer to the centrality-degree distributions of sub-networks, i.e., compartments of larger networks, e.g., clusters around hubs. They may have different structures and, related to that, some heterogeneity in internal relations and decision-making, somewhere between full defection and full cooperation: heterogeneous local cultures or different “self-organization mechanisms” (below). High-degree correlation then is a property of a network according to which high-degree nodes (hubs) tend to have high-degree neighbours. This will facilitate the spreading of innovation cooperation among high-degree nodes (e.g., Rong & Wu, 2009). With a lower degree correlation, a network would have a substantial number of more isolated hub structures, in which defective behaviour may more easily persist (e.g., Cordes et al., 2021). Heterogeneous degree distributions will facilitate innovation cooperation by generating a negative feedback for defecting hubs, and a positive feedback for cooperative hubs (e.g., Santos et al., 2012; Pacheco et al., 2009). Defecting hubs, i.e., defectors with many connections, will usually make their less well-connected neighbours copy their defective strategy, while well-connected neighbours will be less easily compelled to become defectors. But the spread of defection to the neighbourhood of a defective hub reduces the hub’s success and forces it to copy the cooperative strategy from other hubs that it remains connected to. Barabási and Bonabeau (2003) showed that heterogeneous degree distributions are relatively common, a possibility in social networks that are scale-free. On the micro-level of individual agents and their relative positions in the network, we have already mentioned some particular positions such as hubs and others with long-distance relations, “gatekeepers” or “brokers” that connect different clusters and bridge “structural holes”. Both the conception of structural holes among sub-networks, clusters or groups and of between-group brokerage of central agents were developed by R. S. Burt and combined in

Innovation and information  289 Burt (2004). Burt argued that innovative ideas are disproportionately developed by central agents connected across groups, spanning such structural holes, and knowing different cultures. They seem to be able to develop and apply the important cultural technique of “switching across paradigms”, in this way generating innovative impulses. On top of such distance-bridging functions, Vedres and Stark (2010) developed the conception of a structural fold (rather than hole), where a joint member of two sub-networks provides intercohesion (rather than just brokerage and closure of the hole). In this function, the agent at the fold directly combines and recombines information from two areas and may generate new knowledge, equivalent with the recombinant effect of innovation proper. Both relative positions and functions are examples of micro-functions of individual agents in networks, emerging from certain larger network structures in real-world (i.e., non-complete and non-random) networks. Networked systemics tend towards structuration, clustering, and hierarchization (in addition to institutionalizations) in order to reduce their complexity down to (cognitively) manageable levels (away from perceived over-complexity) and towards higher stability (away from repeated crisis (below)). Under these conditions, bridging, brokering etc. have the function of maintaining the entire system, its information flows, and innovation at relatively high levels of complexity and, not pushing them into deterministic chaos, with this, keeping them relatively successful (e.g., Watanabe & Takagi, 2022). 3.3.2 Self-organization mechanisms and innovation What will agents do in face of ubiquitous (latent) issues of perceived over-complexity and over-turbulence, in order to solve their dilemma-prone decision problems, and weaken the impact of their bounded rationality? They will strive to reduce complexity and adapt it to their cognitive capacity. Besides striving to restrict arena and group sizes, attenuate other factors of turbulence, such as high uprooting in-and out-mobility, and develop other capacities (see above), they will, in order to develop expectations-stabilizing institutions, strive to select partners by some so-called preferential attachment or selective mixing: They will try to extend interactions with those partners, with whom they perform successfully and whom they have learned to trust. And they will establish stable patterns in their relations, not only according to family and kin, but also to spatial/social neighbourhood, etc. For example, in the pursuit of reducing complexity, people usually invent, develop, and carry out Polanyian countermovements against deregulated, globalized, and highly anonymous “markets”. Even the largest global (firm) players strive to reduce perceived over-turbulence, will concentrate and centralize, chasing after growth and power, establish long-run relations, and partially exclude the “market”. Also, their most sensitive research, technology, and production divisions will be locally clustered and networked in a more cooperative institutional arrangement and local culture, together with manifold forms of frequent and stable cooperative interactions, as known since Marshall’s industrial agglomeration studies (Marshall, 1920 [1890]). A typical self-organization mechanism in physical, linguistic, financial, social, and economic network analyses has been coined (by physicists) the rich-get-richer mechanism or the preferential-attachment mechanism (Barabási & Albert, 1999): the more relations, power and centrality an entity has, the more it tends to get on top, and such positive cumulative feedback mechanism will be reflected then in right-skewed centrality distributions, with its crisis-prone implications mentioned (e.g., Schlossberger, 2017).

290  The Elgar companion to information economics With the wide variety of dynamics of CAS, i.e., high idiosyncrasies with potential tipping points, bifurcations and sometimes deterministic chaos (where short-term prediction will no longer be possible), often high rigidities (fixed points/attractors), usually transitory, and a number of other system properties will emerge, such as these emergent networks and distributions mentioned, all indicating self-organization mechanisms, i.e., mechanisms of some systematic social action-reaction sequences. These are by no means necessarily “good” or problem-solving, as scale-free networks, power-law distributions, and the rich-get-richer self-organization illustrate. Thus, all dynamic properties and emergent structures of CAS stem from self-organization mechanisms, which are at the core of the theoretical-analytical research programme of CE. And, in fact, the “central-get-more-central”, “powerful-get-more-powerful” or “rich-get-richer” mechanism is a most relevant real-world socio-economic interaction mechanism in existing capitalist economies. Real-world economies, thus, are all but representing random process, which then would entail “normal” centrality distributions, as were assumed in stochastic versions of neoclassical (financial “market”) models, which by putting the financial sector fundamentally wrong, contributed to the big financial crisis. Given adverse self-organization mechanisms (power structures and differential attachment to the largest powers), namely fierce versions of them dominating in deregulated neoliberal capitalist “market” economies, we have to expect badly adapted socio-economic systems, which may well run into repeated crises (self-organized criticality, Bak et al. 1897). And exactly these will have difficulties in generating a continuous high system-wide innovation stream. We will have to expect, rather, CAS that generate, last not least (cycles of), over- and under-innovation. 3.3.3 Insufficient vs. excessive (scarce vs. abundant) information We may reflect, as indicated earlier, CAS with adverse network and distributional structures, and self-organization mechanisms under the informational dimension of forgetting, or information loss, vs. information generation and retention. Under combinations of insufficient vs. excessive diversity generation (i.e., too little or too much innovation production) and insufficient vs. excessive selection pressure (too little or too much information loss) in a system both reflecting different critical phases of open oligopolistic rivalry, only certain areas of combinations of levels of diversity generation and selection pressure will entail proper rates of information loss and information retention, and thus system resilience (Heinrich, 2016, 2018a, 2018b). Outside these areas, systems will be subject to insufficient adaptability, or even collapse. In a related vein, socio-economic CAS are developing today under ubiquitous (1) microelectronic digitization and (2) positive network effects (with ever more users entering just the largest technical networks). They may, under the first condition, generate a volatile innovation with close-to-zero marginal costs, and the system thus easily being subject to too high rates of change (too high diversity generation or over-turbulent innovation) (Heinrich, 2018b), a subsequent lack of stability and a potential “error catastrophe” (carrying too much information along). Under the second condition, the system may generate few and increasingly dominant standards, homogenization, monopolization/oligopolization, and related (technological) lock-in. It may become, through insufficient diversity/innovation, subject to excessive loss of information, some petrifaction, and a subsequent potential crisis, and non-resilience. In both cases they may eventually collapse, or face repeated major crises. The two cases together even may render the system insufficiently innovative across repeated crises, when change rates

Innovation and information  291 overall become too volatile between high turbulence and lock-in (also, e.g., Bloom et al., 2005; Allen et al., 2020). 3.3.4 Policy orientations In the evolution of a CAS, the continuously changing social ecology (strategy composition of the population) will render it impossible that a certain behavioural strategy (institution) can be persistently “optimal” or “fit”. In such process, there will be no sufficient structural stability to provide sufficient time for some selection mechanism to operate in any “meliorating” way. Rather, there will result a “rugged and moving fitness surface” with usually no clear-cut direction for such systems other than attractors or tending towards the “edge of chaos” (as they linger between under- and over-complexity, e.g., Kauffman et al., 2013), and particularly so when they are insufficiently institutionally embedded and insufficiently politically regulated. This all undermines any simple comprehension of innovation, and innovation of policy as well. As Axelrod had already suggested, problem-solving institutionalization can be supported by drawing agents into sequential cooperative projects (at all levels), always overlapping in time, in this way promoting opportunities of “meeting again”, reciprocation in the future, and extending time horizons. In the same vein, policies need to address and shape network structures (above), “shaping the connective geometry” (Room, 2011; also, e.g., Ormerod, 2012; Erixon & Weigel, 2016; Grant & Moses, 2017; Perilla Jimenez, 2019). Specifically, the system’s modularity (Simon, 1962) needs to be scrutinized: functionally overlapping and hierarchically layered and staged arenas and networks (“multiplex” networks), as related to the spatial and social reaches of the “goods” and services at stake, public and private, need to come into the focus of a future complexity policy (e.g., Ormerod, 2012; Elsner & Schwardt, 2015; Gilles et al., 2015). Also, the incentive structures need to be improved, and perhaps even the game type as such (from more to less intricate problems, e.g., from intricate anti-coordination and dilemma problems to less intricate coordination problems). In the case of SDs, this may imply extra rewarding cooperation payoffs, punishing defection, thus reducing “winner-takes-all” one-shot incentives. This must not necessarily imply massive subsidies, let alone pecuniary ones, and by no means must dissolve the PD/SD structure as such. This structure will then transform into a less intricate coordination problem through the general solution of the PD-SG. Namely, the SG-payoff matrix (row player’s payoffs) ​  / (​ ​1 − δ​)​d – c +​ c / (​ ​1 − δ​)​ a b – c - ​c / (​ ​1 − δ​)​​ ​c / (​ ​1 − δ​)​ will, under the basic solution condition above [δ > (b-a)/(b-c)], be transformed into a less “wicked” coordination problem, where coordination is in everyone’s immediate interest, thus easier solvable in favour of a convergent process towards the Pareto-superior coordination, i.e., cooperation. All this is about shaping relevant framework conditions in order to trigger problem-solving adaptive interaction processes of the private. We have called such complexity policy a framework policy approach or an institutional policy (Elsner, 2012), as it aims at shaping those critical factors that tend to structure the interaction processes of the private towards instrumental institutionalizations (building “social capital”), thus generating some socially “responsible innovation” (Schomburg & Hankins (Eds.), 2019). We also have called this new policy

292  The Elgar companion to information economics perspective double-interactive, as the policy system interacts adaptively with the interactive target CAS (while having to be the more complex system of the two), i.e., with the interaction system of the privates, which then is forced to adapt (also, e.g., Elsner, 2017). Such policy orientation improves innovation through coordination, cooperation, and “improving collective capabilities” (Perry, 2020). Others have called this a collaborative innovation philosophy (Ziegler, 2020).

Source: Glaeser et al. (2002, p. F451).

Figure 14.4

Mobility and social capital: the decline of social capital (indicated by organizational membership) upon increasing (uprooting) spatial mobility – an empirical illustration

As an example, enforced (“by the market”) uprooting and disembedding spatial mobility is to be reduced to improve the cognitive and “expectational” conditions, as has been increasingly addressed in a world of growing volatility, turbulence, and enforced mobility, including naïve globalization (gone astray) or increasing refugee and poverty migrations of all kinds nowadays (e.g., Solari & Gambarotto, 2014; Glaeser et al., 2002; Battiston et al., 2015). This obviously is in stark contrast to the conventional wisdom of the superiority of an always productivity-increasing and welfare-enhancing mobility. As an empirical illustration, see Figure 14.4 for the example of (enforced “by the market”, in this case professionally conditioned, domestic) mobility and social-capital deterioration.

Innovation and information  293

4.

CONCLUSIONS

Policy will have to sometimes intervene heavily and lastingly, sometimes only briefly “nudging”, depending on the status and direction of the system. Such innovation policy must be based on a superior complexity of the policy system itself, which thus must be able to assume more states than the target system, while in fact caring for reducing the target system’s state space. It must further be based on the state’s deeper knowledge and permanent anticipatory analysis of the socio-economic target system. Qualified government and state action, thus, has to use itself big data, big computation, modelling, simulations, and algorithms, machine learning and AI (e.g., Elsner, 2012, 2017; Page, 2012; Gilles et al., 2015; Mulgan, 2017; Pyke, 2018; Allen et al., 2020) to shape interaction conditions towards system attractors that are expected to be superior. It is obvious that this requires a “strong” and qualified state, the action capacity of which does not simply depend on its legal capacity or its formally exclusive power. And the state will have to head for permanent learning and adaptation itself, always being alert to intervene massively and lastingly, as required, namely in face of adverse system attractors. Such policy will have to generate no less than the broadest and most inclusive innovative capacities of the socio-economic agents (e.g., Pyke, 2018). Such policy then will foster broad behavioural capacities and willingness to innovate. As a research outlook, we just mention further cutting-edge innovation research in the broadest sense, which deals with radical surprise and novelty, even beyond still calculable white-swan probabilities of innovations. This is, in fact, undecidability and non-computability of novel constellations, which are not at all in the knowledge and experience fund of the agent yet. It fuses developments in mathematical undecidability theory, paleobiology, genome biology, virology / immunology / epidemiology (in medicine and in computer science) and refers to capacities that early eukaryotes had developed billions of years ago: developing novel paths towards “self-other” comparisons (be it “good” innovations or “bad” infections/ malware/virus invasions, both types completely unknown to the agent) and radical agent proteanism (e.g., Casti, 1994; Markose, 2004, 2021). And again, this suggests that the issues of information, cognition and innovation are, first, closely interconnected, and, second, all but settled. Rather, they display even further dimensions in the sciences, still largely unknown to economics. Information and innovation will then have to be scrutinized in yet another dimension.

NOTES 1.

For a modern, neo-Schumpeterian analysis of innovation, scrutinizing both creation and destruction, see e.g., Buenstorf et al. (2013). 2. See also Chapter 1 in this Companion (Raban & Włodarczyk, 2024, pp. 2–19) on informational scarcity and abundance, of course a major cross-sectional theme in this volume. 3. See also Chapters 2 and 3 in this Companion (Stiglitz & Kosenko, 2024a, pp. 20–52, 2024b, pp. 53–80) on disinformation. 4. Note that we use the following terminology: a coordination problem is modelled as a coordination game, where the solution is realized coordination, which in turn is just a common problem (e.g., right- or left-driving on the street) as it requires just parallel behaviour, and coordination is in everyone’s immediate short-run (“hyper-rational”) interest. So the tool for solution (i.e., convergent coordination) is a learned social rule, obviously already difficult enough. A collective-good

294  The Elgar companion to information economics

5.

6.

7.

8.

9. 10.

or social-dilemma problem (SD) is more intricate, as defection is in everybody’s hyper-rational, short-run interest. The solution thus requires a learned and habituated sacrifice of the short-run maximum in favour of the longer-run Pareto-superior solution (and potential new Nash equilibrium of the recurrent game). This is considered a collective problem, and the solution is called cooperation. Cooperation thus is coordination plus a sacrifice. The latter can be realized among “rational” agents only through a credible threat of a sanctioning through some cooperative but responsive (a “trigger”) strategy, a threat of retaliation with a joint setback to the short-run defective Nash equilibrium. The tool for this we call a social institution, i.e., a social rule plus (a credible threat of an) endogenous sanction to generate the sacrifice of the short-run maximum. This terminology has long been established in the theory of institutional emergence and evolution (e.g., Schotter, 1981; Axelrod, 2006 [1984]; Elsner, 2021). Note that our consideration here is methodological: In a formal complex modelling and computational simulation, we would have to define initial sizes and structures, but usually also would reach conclusions about which sizes and kinds of structures, be it incentive structures (“games”), group and arena sizes, network topologies or the reach of learned and habituated behaviour (social institutions), may emerge in the evolutionary process (e.g., Elsner, 2017). More below. We should stress already at this point that systemic “lock-in” usually has its microeconomic correspondence in some coordination or cooperation attained, also to be considered a (technical and behavioural) standardization. Note that in one of the original, and still simple, models (namely Arthur, 1989), lock-in just had the neutral (rather than negative), and just “technical” connotation of a cumulative process leading the system into a (stable) fixed point (attractor), which it is unable to break off from by itself (given its working mechanisms, namely with positive feedback loops). So, while microeconomic coordination / cooperation is an achievement with some systemic stabilization, this by no means determines that such stabilization takes place at some superior attractor. And if it were cooperative institutionalization indeed in a “positive”, superior attractor, the negative connotation of “lock-in” might apply soon, when the institutionalized standardization would remain for too long, relative to some ongoing change of objective conditions (lock-in as institutional hysteresis). See also Chapter 16 in this Companion (Gürpinar & Özveren, 2024, pp. 315–337) on incomplete contracts, which are a particular case, if not fully equivalent with incomplete information. Any incomplete information in any interaction may of course be considered an informal incomplete contract. See Chapter 5 in this Companion (Giza, 2024, pp. 106–117) on asymmetric information as a (structural) market failure, and the entire second part of this volume, of course another major cross-sectional field in this volume, given the importance of asymmetric information in economics in the last three to four decades. “Market failure”, or market structure and market regulation, is the critical aspect of the present chapter as well, and one could argue, indeed, that even if information asymmetry might be considered a technical property of information, a well-regulated market should be structured such that any asymmetry of information, power, elasticities etc. should be mitigated, neutralized, compensated and removed, as introductory textbooks on markets in fact imply. Agent-based modelling, of course, does not need to be game-based but can be modelled in manifold different ways. As an example of another direct modelling of innovative micro-behaviour and its emergent sectoral effects, see, e.g., Beckenbach et al. (2013). Note that this does not simply mean a maximum memory length is suggested. There are evolutionarily advantageous combinations of both proper memory and forgetfulness under different levels of stability or turbulence (e.g., Dosi et al., 2017).

REFERENCES Acemoglu, D., Ozdaglar, A., & Tahbaz-Salehi, A. (2017). Microeconomic origins of macroeconomic tail risks. American Economic Review, 107, 54–108. Albin, P. S. & Foley, D. K. (1992). Decentralized, dispersed exchange without an auctioneer: A simulation study. Journal of Economic Behavior & Organization, 18(1), 27–52.

Innovation and information  295 Allen, D. W. E., Berg, C., Markey-Towler, B., Novak,. M., Potts, J. (2020). Blockchain and the evolution of institutional technologies: Implications for innovation policy. Research Policy, 49(1), DOI: 10.1016/j.respol.2019.103865. Antonelli, C. (2017a). The Evolutionary Complexity of Endogenous Innovation: The Engines of the Creative Response. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Antonelli, C. (2017b). Endogenous Innovation: The Economics of an Emergent System Property. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Arrow, K. J. & Debreu, G. (1954). Existence of an equilibrium for a competitive economy. Econometrica, 22, 265–290. Arthur, W. B. (1989). Competing technologies, increasing returns, and lock-in by historical events. Economic Journal, 99, 116–131. Aspara, J., Pajunen, K., Tikkanen, H., & Tainio, R. (2014). Explaining corporate short-termism: Self-reinforcing processes and biases among investors, the media and corporate managers. Socio-Economic Review, 12(4), 667–693. Axelrod, R. (2006 [1984]). The Evolution of Cooperation (2nd ed.). New York: Basic Books. Bak, P., Tang, C., Wiesenfeld, K. (1987). Self-organized criticality: An explanation of the 1/f noise. Physical Review Letters, 59(4), 381–384. Barabási, A.-L. (2002). Linked: The New Science of Networks. New York: Perseus Books. Barabási, A.-L. & Albert, R. (1999). Emergence of scaling in random networks. Science, 284(5439), 509–512. Barabási, A.-L. & Bonabeau, E. (2003). Scale-free networks. Scientific American, 288(5), 60–69. Battiston, S., Caldarelli, G. R., Roukny, T., & Stiglitz, J. E. (2015). The price of complexity in financial networks. http://​papers​.ssrn​.com/​sol3/​papers​.cfm​?abstract​_id​=​2594028. Battiston, S., Vitali, S., & Glattfelder, J. B. (2011). The network of global corporate control. PLoS One, 6(10), e25995. https://​doi​.org/​10​.1371/​journal​.pone​.0025995. Beckenbach, F., Daskalakis, M., & Hofmann, D. (2013). Agent-based modelling of novelty creating behavior and sectoral growth effects: Linking the creative and destructive side of innovation. In G. Buenstorf, U. Cantner, H. Hanusch, M. Hutter, H.-W. Lorenz, & F. Rahmeyer (Eds.), The Two Sides of Innovation: Creation and Destruction in the Evolution of Capitalist Economies (pp.  129–158). Cham: Springer International. Bertani, F., Ponta, L., Raberto, M., Teglio, A., & Cincotti, S. (2019). The complexity of the intangible digital economy: An agent-based model. MPRA Paper No. 97071. Munich. https://​mpra​.ub​.uni​ -muenchen​.de/​97071/​. Bettignies, J.-E. de & Ries, J. (2023). When less is more: Information and the financing of innovation. Journal of Economic Behavior & Organization, 214, 346–369. Bloom, N., Aghion, P., Blundell, R., Griffith, R., & Howitt, P. (2005). Competition and innovation: An inverted U relationship. The Quarterly Journal of Economics, 120(2), 701–728. Boudreau, K. J., Lacetera, N., & Lakhani, K. R. (2008). Parallel search, incentives and problem type: Revisiting the competition and innovation link. Working Paper No. 09–041. Harvard Business School. https://​pdfs​.semanticscholar​.org/​1687/​4121be​a49c69b7fd​8252f83546​271df02f08​.pdf. Buenstorf, G., Cantner, U., Hanusch, H., Hutter, M., Lorenz, H.-W., & Rahmeyer, F. (Eds.) (2013). The Two Sides of Innovation: Creation and Destruction in the Evolution of Capitalist Economies. Cham: Springer International. Burt, R. S. (2004). Structural holes and good ideas. American Journal of Sociology, 110(2), 349–399.‫‏‬ Bush, P. D. (1999). Recognized interdependence. In P. A. O’Hara (Ed.), Encyclopaedia of Political Economy (Vol. 2, pp. 963–964). New York: Routledge. Casti, J. (1994). Complexification: Explaining a Paradoxical World Through the Science of Surprises. New York: HarperCollins. Cetrulo, A., Cirillo, V., & Guarascio, D. (2018). Weaker jobs, weaker innovation: Exploring the temporary employment-product innovation nexus. ISI Growth Working Paper 6/2018. http://​www​ .isigrowth​.eu/​2018/​02/​17/​weaker​-jobs​-weaker​-innovation​-exploring​-the​-temporary​-employment​ -product​-innovation​-nexus/​. Chen, P. (2010). Economic Complexity and Equilibrium Illusion: Essays on Market Instability and Macro Vitality. New York: Routledge.

296  The Elgar companion to information economics Christensen, C. (1997). The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail. Cambridge, MA: Harvard University Press. Coad, A., Nightingale, P., Stilgoe, J., & Vezzani, A. (2021). The dark side of innovation. Editorial, Special Issue Industry and Innovation, 28(1), 102–112. Commons, J. R. (1990 [1934]). Institutional Economics: Its Place in Political Economy. Basingstoke: Macmillan. Cordes, C., Elsner, W., Graebner, C., Heinrich, T., Henkel, J., Schwardt, H., Schwesinger, G., Su, T.-Y. (2021). The collapse of cooperation: The endogeneity of institutional break-up and its asymmetry with emergence. Journal of Evolutionary Economics, 31(4), 1291–1315. Debreu, G. (1974). Excess demand functions. Journal of Mathematical Economics, 1(1), 15–21. Dosi, G., Marengo, L., Paraskevopoulou, E., & Valente, M. (2017). A model of cognitive and operational memory of organizations in changing worlds. Cambridge Journal of Economics, 41(3), 775–806. Dosi, G., Pereira, M. C., Roventini, A., & Virgillito, M. E. (2018). What if supply-side policies are not enough? The perverse interaction of flexibility and austerity. ISI Growth Working Paper 1/2018. http://​www​.isigrowth​.eu/​2018/​01/​09/​what​-if​-supply​-side​-policies​-are​-not​-enough​-the​-perverse​ -interaction​-of​-flexibility​-and​-austerity/​. Dunbar, R. I. M. (1993). Co-evolution of neocortex size, group size and language in humans. Behavioral and Brain Sciences, 16(4), 681–735. Dunbar, R. I. M. (2008). Cognitive constraints on the structure and dynamics of social networks. Group Dynamics: Theory, Research, and Practice, 12(1), 7–16. Elsner, W. (1989). Adam Smith’s model of the origins and emergence of institutions: The modern findings of the classical approach. Journal of Economic Issues, 23(1), 189–213. Elsner, W. (2012). The theory of institutional change revisited: The institutional dichotomy, its dynamic, and its policy implications in a more formal analysis. Journal of Economic Issues, 46(1), 1–43. Elsner, W. (2017). Complexity economics as heterodoxy: Theory and policy. Journal of Economic Issues, 51(4), 939–978. Elsner, W. (2021). Collapse: Institutional decline and breakdown, its endogeneity and its asymmetry vis-à-vis emergence. Journal of Economic Issues, 55(1), 79–102. Elsner, W., Heinrich, T., & Schwardt, H. (2015). The Microeconomics of Complex Economies: Evolutionary, Institutional, Neoclassical and Complexity Perspectives. Amsterdam: Elsevier / Academic Press. Elsner, W. & Schwardt, H. (2014). Trust and arena size: Expectations, institutions, general trust, and critical population and group sizes. Journal of Institutional Economics, 10(1), 107–134. Elsner, W. & Schwardt, H. (2015). From emergent cooperation to contextual trust, and to general trust: Overlapping meso-sized interaction arenas and cooperation platforms as a foundation of pro-social behavior. Forum for Social Economics, 44(1), 69–86. Erixon, F. & Weigel, B. (2016). The Innovation Illusion: How So Little is Created by So Many Working So Hard. New Haven, CT: Yale University Press. Fagerberg, J. (2018 [2004]). Innovation: A guide to the literature. In J. Fagerberg, Innovation, Economic Development and Policy: Selected Essays (pp. 3–28). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Frey, C. B. (2019). The Technology Trap. Capital, Labor, and Power in the Age of Automation. Princeton, NJ: Princeton University Press. Gallegati, M., Kirman, A., & Palestrini, A. (2007). Implications of scaling laws for policy-makers. In M. Salzano & D. Colander (Eds.), Complexity Hints for Economic Policy (pp. 291–302). Cham: Springer. García-Vega, M. & Huergo, E. (2017). Trust and technology transfers. Journal of Economic Behavior & Organization, 142(C), 92–104. Gault, F., Arundel, A., & Kraemer-Mbula, E. (Eds.) (2023). Handbook of Innovation Indicators and Measurement. 2nd ed. Cheltenham, UK, and Northampton, MA, USA: Edward Elgar Publishing. Georgescu-Roegen, N. (1975). Energy and economic myths. Southern Economic Journal, 41(3), 347–381. Gilles, R., Lazarova, E., & Ruys, P. (2015). Stability in a network economy: The role of institutions. Journal of Economic Behavior & Organization, 119, 375–399.

Innovation and information  297 Giza, W. (2024). Asymmetric information as a market failure in retrospect. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 106–117). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Glaeser, E. L., Laibson, D., & Sacerdote, B. (2002). An economic approach to social capital. Economic Journal, 112(483), F437–F458. Godin, B. & Vinck, D. (Eds.) (2017). Critical Studies of Innovation: Alternative Approaches to the Pro-Innovation Bias. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Goerner, S. (2017). The science of flow says extreme inequality causes economic collapse. Evonomics. http://​evonomics​.com/​science​-flow​-says​-extreme​-inequality​-causes​-economic​-collapse/​. Grant, D. & Moses, L. B. (2017). Technology and the Trajectory of Myth. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Gürpinar, E. & Özveren, E. (2024). Incomplete contracts, incentives, and investment in knowledge assets under alternative institutional configurations. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 315–337). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Haber, S. H. & Lamoreaux, N. R. (2021). The Battle Over Patents: History and the Politics of Innovation. NBER Working Paper No. 28774. Heinrich, T. (2016). Evolution-based approaches in economics and evolutionary loss of information. Journal of Economic Issues, 50(2), 390–397. Heinrich, T. (2018a). A discontinuity model of technological change: Catastrophe theory and network structure. Computational Economics, 51(3), 407–425. Heinrich, T. (2018b). The rate of change in evolutionary systems and evolutionary economic modeling. Journal of Economic Issues, 52(2), 570–579. Heinrich, T., Yang, J., & Dai, S. (2020). Growth, development and structural change at the firm-level: The example of the PR China. University of Oxford, Oxford Martin School, Working Paper No. 2020–8. https://​www​.oxfordmartin​.ox​.ac​.uk/​publications/​growth​-development​-and​-structural​-change​ -at​-the​-firm​-level​-the​-example​-of​-the​-pr​-china/​. Herrmann-Pillath, C. (2018). The case for a new discipline: Technosphere science. Ecological Economics, 149, 212–225. Jayasinghe, P. (2017). On microfoundations of macroeconomics. Real-World Economics Review, 82, 60–75. Jennings, F. B. (2005). How efficiency/equity trade-offs resolve through horizon effects. Journal of Economic Issues, 39(2), 365–373. Kauffman, S. A., Farmer, D., Solée, R. V., Valverde, S., Casals, M. R., & Eldredge, N. (2013). The evolutionary ecology of technological innovations. Complexity, 18(4), 15–27. Kendall, G., Yao, X., & Chong, S. Y. (2007). The Iterated Prisoner’s Dilemma: 20 Years On. Singapore: World Scientific. Keynes, J. M. (1921). A Treatise on Probability. London: Macmillan. Keynes, J. M. (1936). The General Theory of Employment, Interest and Money. London: Macmillan. Kingston, W. (2017). How Capitalism Destroyed Itself: Technology Displaced by Financial Innovation. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Kirman, A. P. (1992). Whom or what does the representative individual represent? Journal of Economic Perspectives, 6(2), 117–136. Kleinknecht, A., Kwee, Z., & Budyanto, L. (2016). Rigidities through flexibility: Flexible labour and the rise of management bureaucracies. Cambridge Journal of Economics, 40(4), 1137–1147. Komlos, J. (2016). Has creative destruction become more destructive? B.E. Journal of Economic Analysis & Policy, 16(4), 1–12. Krasteva, S., Sharma, P., & Wang, C. (2020). Patent policy, imitation incentives, and the rate of cumulative innovation. Journal of Economic Behavior & Organization, 178, 509–533. Langley, G. & Tulloch, J. (2017). Capitalism’s creation myth revisited. https://​projectm​-online​.com/​risk/​ capitalisms​-creative​-destruction/​. Leliveld, A., Bhaduri, S., Knorringa, P., & van Beers, C. (Eds.) (2023). Handbook on Frugal Innovation. Cheltenham, UK, and Northampton, MA, USA: Edward Elgar Publishing.

298  The Elgar companion to information economics Lindgren, K. (1989). Evolutionary dynamics in game-theoretic models. In W. B. Arthur, S. Durlauf, & D. Lane (Eds.), The Economy as an Evolving Complex System II (pp.  337–367). Reading, MA: Addison-Wesley. Lipsey, R. G. & Lancaster, K. J. (1956). The general theory of second best. Review of Economic Studies, 24(1), 11–32. Lueders, M., Andreassen, T. W., Clatworthy, S., & Hillestad, T. (2017). Innovating for Trust. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Lux, T. & Marchesi, M. (1999). Scaling and criticality in a stochastic multi-agent model of a financial market. Nature, 397, 498–500. Mandelbrot, B. & Hudson, R. L. (2006). The Misbehavior of Markets: A Fractal View of Financial Turbulence. New York: Basic Books. Mantel, R. R. (1976). Homothetic preferences and community excess demand functions. Journal of Economic Theory, 12(2), 197–201. Markose, S. (2004). Novelty in complex adaptive systems (CAS): A computational theory of actor innovation. Physica A, 344, 41–49. Markose, S. (2021). Genomic intelligence as Über bio-cybersecurity: The Gödel sentence in immuno-cognitive systems. Entropy, 23(4), 405. Marlowe, F. W. (2005). Hunter-gatherers and human evolution. Evolutionary Anthropology, 14, 54–67. Marshall, A. (1920 [1890]), Principles of Economics (8th ed.). London: Macmillan. Mazzucato, M. (2016). Innovation policy as creating markets, not only fixing them: Implications for complexity theory. In D. S. Wilson & A. Kirman (Eds.), Complexity and Evolution: Towards a New Synthesis for Economics (pp. 271–284). Cambridge, MA: MIT Press. Mirowski, P. (1989). More Heat than Light: Economics as Social Physics, Physics as Nature’s Economics. Cambridge: Cambridge University Press. Mulgan, G. (2017). Cognitive economics: How self-organization and collective intelligence works. http://​evonomics​.com/​cognitive​-economics​-intelligence​-mulgan/​. Ormerod, P. (2012). Networks and the need for a new approach to policymaking. In T. Dolphin & D. Nash (Eds.), A Complex New World: Translating New Economic Thinking into Public Policy (pp. 28–38). London: Institute for Public Policy Research. Pacheco, J. M., Pinheiro, F. L., & Santos, F. C. (2009). Population structure induces a symmetry breaking favoring the emergence of cooperation. PLoS Computational Biology, 5(12), e1000596. Page, S. (2012). A complexity perspective on institutional design. Politics, Philosophy & Economics, 11(1), 5–25. Pareto, V. (1896–7). Cours d’Économie Politique. Berlin: Duncker & Humblot. Perilla Jimenez, J. R. (2019). Mainstream and evolutionary views of technology, economic growth and catching up. Journal of Evolutionary Economics, 29(2), 823–852. Perry, K. K. (2020). Innovation, institutions and development: A critical review and grounded heterodox economic analysis of late-industrializing contexts. Cambridge Journal of Economics, 44(2), 391–415. Pogach, J. (2018). Short-termism of executive compensation. Journal of Economic Behavior & Organization, 148, 150–170. Polanyi, K. (2001 [1944]). The Great Transformation: The Political and Economic Origins of Our Time. Boston: beacon Press. Prodi, G., Nicolli, F., & Frattini, F. (2020). Embeddedness and local patterns of innovation: Evidence from Chinese prefectural cities. Journal of Evolutionary Economics, 30, 1219–1242. Pyke, F. (2018). Managing technological change for inclusive growth. Cambridge Journal of Economics, 42(6), 1687–1695. Raban, D. R. & Włodarczyk, J. (2024). Information economics examined through scarcity and abundance. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 2–19). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Rasmus, J. (2016). Systemic Fragility in the Global Economy. Atlanta, GA: Clarity Press. Rong, Z. & Wu, Z. X. (2009). Effect of the degree correlation in public goods game on scale-free networks. Europhysics Letters, 87(3), 30001. Room, G. (2011). Complexity, Institutions and Public Policy: Agile Decision-Making in a Turbulent World. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing.

Innovation and information  299 Ruth, M. & Goessling-Reisemann, S. (Eds.) (2018). Handbook on Resilience of Socio-Technical Systems. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Samuelson, P. A. (1948). Foundations of Economic Analysis. Cambridge, MA: Harvard University Press. Santos, F. C., Pinheiro, F. L., Lenaerts, T., & Pacheco, J. M. (2012). The role of diversity in the evolution of cooperation. Journal of Theoretical Biology, 299, 88–96. Sautua, S. I. (2017). Does uncertainty cause inertia in decision making? An experimental study on the role of regret aversion and indecisiveness. Journal of Economic Behavior & Organization, 136, 1–14. Schlossberger, J. (2017). The network topology of locally interacting agents and system-level distributions of agent actions. https://​papers​.ssrn​.com/​sol3/​papers​.cfm​?abstract​_id​=​3085658. Schotter, A. (1981). The Economic Theory of Social Institutions. Cambridge: Cambridge University Press. Schrepel, T. (2017). Predatory innovation: The definite need for legal recognition. https://​papers​.ssrn​ .com/​sol3/​papers​.cfm​?abstract​_id​=​2986246. Schubert, C. (2013). Is novelty always a good thing? Towards an evolutionary welfare economics. In G. Buenstorf, U. Cantner, H. Hanusch, M. Hutter, H.-W. Lorenz, & F. Rahmeyer (Eds.), The Two Sides of Innovation: Creation and Destruction in the Evolution of Capitalist Economies (pp.  209–242). Cham: Springer International. Sen, A. K. (1977). Rational fools: A critique of the behavioral foundations of economic theory. Philosophy & Public Affairs, 6(4), 317–344. Seo, H. J., Kang, S. J., & Baek, Y. J. (2020). Managerial myopia and short-termism of innovation strategy: Financialization of Korean firms. Cambridge Journal of Economics, 44, 1197–1220. Shiozawa, Y. (2020). A new framework for analysing technological change. Journal of Evolutionary Economics, 30, 989–1034. Simon, H. A. (1955). A behavioral model of rational choice. The Quarterly Journal of Economics, 69(1), 99–118. Simon, H. A. (1962). The architecture of complexity. Proceedings of the American Philosophical Society, 106, 467–482. Solari, S. & Gambarotto, F. (2014). Territorial rooting as an element of wellbeing. Review of Social Economy, 72(4), 504–522. Sonnenschein, H. (1973). Do Walras’ identity and continuity characterize the class of community excess demand functions? Journal of Economic Theory, 6(4), 345–354. Sraffa, P. (1926). The laws of returns under competitive conditions. Economic Journal, 36(144), 535–550. Srinivas, S. (2021). Heuristics and the microeconomics of innovation and development. Innovation and Development, 11(2–3), 281–302. Stiglitz, J. E. & Kosenko, A. (2024a). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 1: Indirect Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 20–52). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. & Kosenko, A. (2024b). Robust Theory and Fragile Practice: Information in a World of Disinformation. Part 2: Direct Communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 53–80). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Swann, G. M. P. (2015). Common Innovation: How We Create the Wealth of Nations. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Tainter, J. A., Strumsky, D., Taylor, T. G., Arnold, M., & Lobo, J. (2018). Depletion vs. innovation: The fundamental question of sustainability. In R. Burlando & A. Tartaglia (Eds.), Physical Limits to Economic Growth (pp. 64–93). New York: Routledge. Van der Wouden, F. (2020). A history of collaboration in US invention: Changing patterns of co-invention, complexity and geography. Industrial and Corporate Change, 29(3), 599–619. Veblen, T. B. (1899). The Theory of the Leisure Class: An Economic Study of Institutions. London: Macmillan. Vedres, B. & Stark, D. (2010). Structural folds: Generative disruption in overlapping groups. American Journal of Sociology, 115(4), 1150–1190.

300  The Elgar companion to information economics Von Schomberg, R. & Hankins, J. (Eds.) (2018). Handbook on Responsible Innovation. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Watanabe, I. & Takagi, S. (2022). NK model-based analysis of technological trajectories: A study on the technological field of computer graphic processing systems. Evolutionary and Institutional Economics Review, 19, 119–140. Watts, D. J. (1999). Small Worlds: The Dynamics of Networks between Order and Randomness. Princeton, NJ: Princeton University Press. Ziegler, R. (2020). Innovation, Ethics and our Common Futures: A Collaborative Philosophy. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing.

15. Intangibles, information goods, and intellectual property goods in modern economics Dominika Bochańczyk-Kupka

INTRODUCTION Greek Philosopher Heraclitus’s phrase Panta rhei kai ouden menei translated as “everything flows” is one of the most famous philosophical quotes of all time. This statement perfectly captures the changing economic reality and understanding of economic categories, especially nowadays in the globalized digital era. Despite the growing relevance of intangible goods in economies, the difficulty in defining and measuring such assets has often led them to be omitted by economic theory, resulting in flawed theories and poor decision-making. Paul Samuelson argued that two major sources of confusion in economics are: giving the same thing different names and calling different things by the same name (Boettke, 2002, p. 272). His argument advocated the use of mathematics in economics. However, this usage has not solved problems and even has exacerbated difficulties, as the intangibles are usually unmeasurable, therefore the discussion over the standardization of concepts and theories is from a mathematical point of view pointless. Additionally, many fields of interest, the wide range of opinions and disputes, and problems of translation processes have caused a lack of transparency and sharp boundaries between concepts. In economic literature the concepts of intangibles, knowledge, information goods, and intellectual property have been developed in parallel, often without mutual recognition and cooperation on a conceptual level. This lack of cohesion causes uncertainty and misunderstanding, so comparing existing categories and theories can eliminate ambiguity and simplify future considerations.

1.

INTANGIBLES AND TANGIBLES

Probably the broadest category among the concepts under consideration is intangibles. In economics, intangible assets are assets that do not have a physical or financial embodiment. A clear, brief definition of intangible goods is difficult to give. Koppius (1999) defines this kind of good literally. The literal meaning of the term “intangible goods” allows the creation of the colloquial definition of “being a product that you can drop on your foot without feeling it”. The study of intangibles faces similar challenges, because of their inherent characteristics, that is, they are intangible and difficult to conceptualize and measure. Another reason is that what is meant by intangible assets and what is or is not covered by this term differs from author to author, study to study, and project to project. A detailed study of the diverse understanding of intangibles in the economics literature was given by YoungGak Kim (2007). Nowadays researchers use mainly the World Intellectual Property Organization (WIPO) or OECD classifications (which also have been evaluated and developed in the last years). WIPO, the 301

302  The Elgar companion to information economics global forum for intellectual property services, policy, information, and cooperation which is a self-funding agency of the United Nations, with 193 member states, divides intangible assets into five broad groups: marketing-related intangible assets, consumer-related intangible assets, artistic-related intangible assets, contract-based intangible assets, and technology-based intangible assets.1 In the past, in previous OECD works, the term “intellectual assets” has been used as referring to knowledge assets or intellectual capital. Much of the focus on intangibles has been on R&D, key personnel, and software. But the range of intangible assets is considerably broader. Nowadays, the OECD classifies intangibles into three types (OECD, 2012): ● computerized information (such as software and databases), ● innovative property (such as scientific and nonscientific R&D, copyrights, designs, trademarks), ● economic competencies (including brand equity, firm-specific human capital, networks joining people and institutions, organizational know-how that increases enterprise efficiency, and aspects of advertising and marketing). A significant distinction can be drawn between “hard” intangibles, which are tradable in the marketplace, and “soft” intangibles, which cannot be sold or negotiated (OECD, 2014; Zambon et al., 2019). Intangible assets are of growing importance to the economy of the twenty-first century and are attracting interest from diverse fields including business, finance, law, economics, statistics, and accounting. Numerous academic studies show the importance of intangible assets in national growth accounting (Corrado et al., 2005; McGrattan & Prescott, 2005a, 2005b; Fukao et al., 2007). There is also a vast literature pointing out the importance of various types of intangible assets in both the microeconomic and macroeconomic sense. Many studies have examined the role of particular types of intangibles such as research and development patents (Griliches, 1995), innovation, and human capital (Rajan & Zingales, 1998; Garicano, 2000). In a macroeconomic sense, intangibles are disputed in the context of economic growth, economic crises, or monetary and fiscal policy (Ahn et al., 2020; Aghion et al., 2010, 2014; Correa-Caro et al., 2018) and role of investment in the economy (Corrado et al., 2005, 2009). The process of distinguishing the intangibles from tangibles and understanding the uniqueness of each requires the analysis of the existence of a physical form, the length of the protection period, the available method of measurement, position in conventional accounting systems, availability, resistance to copy, legal protection, rate of depreciation, the transfer possibility and its cost, the possibility of simultaneous multiple-use, and realization through people. The tangible property right is perpetual. Property rights are assigned to economic goods to create maximum utility in their use. In terms of physical goods, their scarcity is a presupposition of property existence. Intangibles are not perpetual and with time-limited protection (except for distinctive signs such as trademarks, slogans, and brand names, as long as they are used and their registration is kept up-to-date, and geographical indications).2 Moral rights, the rights of the creator to be recognized as an author, also has a timeless character (copyright), they cannot be resold or inherited. As the time of exclusive rights of other types of tangibles is limited when the period of protection expires, they become common goods or “public domain” goods commonly available without any permission and additional costs. The physical nature of tangible assets implies that they can be seen, felt, or touched, allowing measurement of the

Intangibles, information goods, and intellectual property goods in modern economics  303 value loss. Tangible assets depreciate over time. There are well-known standards of tangible resource valuation that are transparent, efficient, and comparable. Additionally, tangible assets possess a scrap or residual value, and they can be used as collateral to obtain loans. Intangible assets cannot be depreciated. They are amortized (except for goodwill) over the useful life of the asset. Generally, intangible assets are amortized using the straight-line expense method. Additionally, the valuation of intangibles is perceived as very subjective and modifiable. Intangible assets represent a major share of the value of modern firms and play an important role in their strategies and they are also usually long-term assets. “Intangible assets have long been the engine for value creation in the world’s developed economies”, says the International Valuation Standards Council (IVSC, 2021). Companies’ investment in intangible assets, and investors’ ability to identify companies able to make the best return on these assets, are critical. Yet, “despite the importance of intangible assets to the capital markets, only a small percentage are recognized on balance sheets, typically via acquisition from a third-party transaction” (IVSC, 2021). Intangible assets can be classified as identifiable or non-identifiable. Identifiable intangible assets are those that can be separated from other assets (such as patents, copyrights, trademarks, and trade names). Unidentifiable intangible assets are those that cannot be physically separated from the company, i.e., goodwill, branding, and reputation. The accurate valuation of intangibles remains a challenge, although there are several accepted ways to measure the value of the intellectual property such as the cost approach, income approach, market approach, and debatable relief from the royalty approach (Bochańczyk-Kupka, 2017). Their material character makes the process of tangible goods’ duplication more expensive. Costs and availability of intermediates and time limits influence the higher cost of copy creation. Intangible goods can be duplicated immediately in massive numbers at zero marginal cost. Tangible goods are usually limited in number and their consumption shortens their availability, and the shortage determines the price level. The simultaneous consumption by more than one person at the same time is impossible. Intangible goods can be consumed concurrently by an infinite number of individuals and making a copy does not deprive anyone of their possessions. The consumption of intangibles does not influence their quality, availability, or volume, and this production factor is inexhaustible. All the above-mentioned features of both types of assets indicate that the tangible and intangible are theoretically and practically distinct (Haskel & Westlake, 2017). Haskel and Westlake (2017) describe the “unusual economic characteristics of intangibles”, drawing attention to their scalability, sunkenness, spillovers, and synergies. Scalability means intangible assets can be used repeatedly and simultaneously in multiple places at the same time (in contrast to tangible assets). Sunkenness concerns irrecoverable costs; additionally, information is almost impossible to liquidate. Intangibles share with tangible sunk costs the characteristic that they are highly specialized and nearly irrecoverably lost upon exit. However, intangibles investment often has little or no salvage value. Tangibles are marketable goods, and in the event of failure, they can be sold to provide investors with some compensation. In contrast, if a business invests in intangibles without success, there is little or nothing that can be sold in the market to recover any of the cost of the bad investment. Investments with high sunk costs can be difficult to finance. Spillovers can be created by intangible investments, especially because of their non-excludable character. Synergies are important in intangible-based economies, as information often works well together. The concept of “open innovation“, defined as when a firm directly engages with and benefits from new knowledge gathered from outside the company, is a major driver of intangible synergy at both the microeconomic and macroeconomic level.

304  The Elgar companion to information economics

2.

INTELLECTUAL PROPERTY AND ITS PROTECTION IN ECONOMICS (LAW AND ECONOMICS PERSPECTIVE)

As Machlup and Mansfield (1983, p. 642) have noted, the original meaning of the word information derives from the Latin informare, which means “to put into form”. In each social discipline, the term information is applied to processes that involve a flow, impulse, etc. Historically information holds little significance in history as a term of broad definitional power. Western philosophy used the term knowledge as its keyword within the broader project of epistemological grounding (tied to the material projects of the domination of nature and the conceptualization of the human individual as a juridical, political, and economic subject).3 Information emerges as an important concept in the mid-twentieth century as industrial capitalism started to use intelligent machine tools (computers and then artificial intelligence), and intelligent marketing which is called direct marketing. Information is not knowledge but becomes knowledge representation (Machlup & Mansfield, 1983, p. 34).4 In economics, the concept of intellectual property is relatively new. It became recognizably important and globally discussed due to the implementation of the Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS), which came into effect on January 1, 1995. The idea of intellectual property was not considered and discussed by representatives of classical economics, neoclassical economics, Keynesian economics, and even the new institutional economics so its presence in the economics literature is recent; however, much important research has been conducted so far. A few decades ago intellectual property was regarded as a rather obscure and irrelevant field of legal regulations, but nowadays intellectual property and its protection are recognized as a driving force for economic growth and development and a crucial factor in firms’ development. The mainstream concept of intellectual property assumes that certain products of human intellect should be afforded the same protective rights that apply to tangibles. As mentioned above, these two types of assets are significantly different. Intangibles’ novelty and dissimilarity cause many theoretical and practical problems, and their increasing popularity was not accompanied by the solution to the underlying conceptual problems (WIPO, 2016; Yu, 2016). The concept of intellectual property derives from property rights considerations but has to consider the unique and extraordinary character of intangibles. The term “property” is the legal concept, which in a market economy is secured by law and fully enforced and protected by the state. Although economics defines property “rights” in ways that diverge significantly from standard legal conceptions, those divergent definitions can bias economic analyses and create the potential for misunderstanding. In other words, the definitions offered by economists sometimes are distinctly at odds with the conventional understandings of legal scholars and the legal profession. Cole and Grossman (2002) wrote: “It is careless (to say the least) for economists writing about property rights simply to presume that such ‘rights’ arise from mere use, without acknowledging that such a presumption (1) runs contrary to the substantial jurisprudence on the definition and allocation of property rights and (2) may preordain suboptimal economic outcomes”. This problem concerns intellectual property. It is questionable whether intellectual property should be considered property because its owner usually is not able to control its consumption and dissemination and obtain the right profits. Intellectual property is defined as any creation of the human mind, which means that it is always intangible. Therefore, it is an intangible asset with legal rights, a non-physical property that is the product of original thought (Zalta, 2011). Lindberg (2009) characterizes intellectual

Intangibles, information goods, and intellectual property goods in modern economics  305 property as a hybrid good made up of equal parts information and law. This definition includes innovation-related intangibles (R&D, patents), but also market-related (brands), human resources (competencies & skills, training), and organizational intangibles (internal structures, systems, procedures, routines, and processes). Intangible assets may be owned, possessed, or accessed. Intellectual property may be used in a variety of ways: waived (defer ownership responsibilities to others, or dedicate to the public domain), asserted (use ownership positions to require payment from others for the use of, or right to use), shared, transferred (convey by assignment or license the responsibility to manage intellectual property and thus exercise control over related assets), excluded (exclude others from practicing or exploiting intangible assets), and squandered (fail to use claimed and established ownership positions to secure value, promote use, or enhance public benefit) (American Association of University Professors, 2021). The WTO (2018) describes intellectual property rights as the rights given to persons over the creations of their minds which usually give the creator an exclusive right over the use of his/her creation for a certain period. This originality is the main determinant of intellectual property but originality is not an innovation. According to the OECD (2008, p. 276), intellectual property rights is the general term for the assignment of property rights through patents, copyrights, and trademarks. These rights allow the holder to exercise a monopoly on the use of the item for a specified period. Among types of intellectual property, there is ownership of ideas, including literary and artistic works (protected by copyright), inventions (protected by patents), signs for distinguishing goods of an enterprise (protected by trademarks), and other elements of industrial property (Khemani & Shapiro, 1993). The Convention establishing the World Intellectual Property Organization (WIPO, 1967), was signed in Stockholm on July 14, 1967. Article 2 (viii) provides that intellectual property shall include rights relating to literature, artistic and scientific works, performances of performing artists, phonograms and broadcasts, inventions in all fields of human endeavor, scientific discoveries, industrial designs, trademarks, service marks, and commercial names and designations, protection against unfair competition, and all other rights resulting from intellectual activity in the industrial, scientific, literary or artistic fields. During that time the list of intellectual property types was open to new types which means that any new rights resulting from any intellectual process of the human mind should be protected as intellectual property. The TRIPS Agreement, which came into effect on January 1, 1995, is nowadays the most comprehensive multilateral agreement on intellectual property. The areas of intellectual property that it covers are as follows: copyright and related rights (i.e. the rights of performers, producers of sound recordings, and broadcasting organizations), trademarks including service marks, geographical indications including appellations of origin, industrial designs, patents including the protection of new varieties of plants, the layout-designs of integrated circuits, undisclosed information including trade secrets and test data. The TRIPS approach limited the number of types of intellectual property that could be protected (Aplin & Davis, 2017, p. 28). It means that any other types of intellectual property which are not mentioned in this Agreement cannot be protected by it without changes in the Agreement. As the impact of the TRIPS Agreement on worldwide economic and political relations was undoubtedly huge, therefore, the perception is that intellectual property has become naturally accepted worldwide. However, it is necessary to notice that all these definitions do not refer to the nature of the intellectual property and mainly exemplify some of its types.

306  The Elgar companion to information economics The origins of intellectual property economics are based on the “theory of public goods”.5 It states that the information embodied in the majority of intellectual property types has two characteristics that distinguish it from tangible property. Like other public goods, copyrighted expressions and patented inventions are non-rivalrous (non-rival) and non-excludable. If a good is non-rivalrous it is costless to allow additional consumers simultaneously to enjoy the benefits of it once it has been produced. If a good is non-excludable, it is difficult for producers to force consumers to pay for the privilege of using it (Barnes, 2011). These features are common to most types of intellectual property goods, but there are two exceptions: geographical indications and trademarks. Geographical indications are non-rival but excludable. It means then they can be treated as the “club asset” belonging to club members consisting of firms located in a specific territory to which the particular geographical indication is attached, and that share the reputation of the geographical indication. The trademarks are rival-in-consumption and excludable so they should be treated as pure private goods. To sum up, it is worth mentioning that different types of intellectual property can be classified as private goods, public goods, and club goods. Individual types of intellectual property differ significantly. The analysis of the specific features of intellectual property goods and their protection reveals the diversity of approaches and the lack of a single concept describing them all. The increasing role of intellectual property in the economy and the necessity of its protection along with a lack of theoretical background have increased the confusion and misunderstanding of existing concepts. This field of research exploded and theoretical consideration was joined by the new discipline of economics and law.6 The discipline is now well established, with eight associations, including the American, Canadian, and European law and economics associations, and several journals (Rubin, 2020). Papers dedicated to law and economics appear regularly in major economics journals. Unfortunately, the growing interest in this subject has not solved the key theoretical problems and has not led to the development of universally widely accepted, and recognizable concepts concerning the economic aspects of intellectual property and its protection. The main feature of the social sciences is their continuous development. Also within economics, many different trends arise that try to explain the complicated reality of the contemporary economy. One of them is the economics of information, which seems to be an appropriate research field for considering the specifics and meaning of intangibles.

3.

INFORMATION AND INFORMATION ECONOMICS

Information (as well as intellectual property) as an economic concept poses a lot of difficulties for neoclassical economics (Raban & Włodarczyk, 2024). Classical economics assumed perfect information, which is only a theoretical assumption. In reality, information is never perfect, and in many instances, economic agents are faced with information asymmetry. The problem of information being treated as a good was not considered in classical economics, and, oddly enough, information was not even classified as a factor of production. Prices revealed all the relevant information; no more consideration was needed. The fundamental breakthrough of modern economic theory was the recognition that information matters and is fundamentally different from traditional goods, mainly understood as commodities (tangibles). Joseph Stiglitz, the founder of the economics of information, noticed that each piece of information is different and unique, and markets of information are virtually characterized

Intangibles, information goods, and intellectual property goods in modern economics  307 by imperfect information (Stiglitz, 2000, p. 1499). In 1999 Shapiro and Varian recognized the role of communication in the transformation of information goods from a material to an immaterial character. They defined an information good as anything that can be encoded, or digitized, as a stream of bits. They noted that in economics and law, information good is generally defined as a commodity that derives its main market value from the information it contains (e.g., books). First, telecommunication technologies (e.g., radio) and later digitization enabled the detachment of information goods from the medium of transfer. This change had tremendous effects on the production, exchange, and consumption of information that could not be fully captured by the traditional conceptualizations (Shapiro & Varian, 1999). Modern information goods do not need a material medium anymore. Stiglitz agreed with Hayek that the central problem of economics was a problem of knowledge and information (Stiglitz, 2000, p. 1468) but pointed out that Hayek’s point of view was too narrow and focused on the once-and-for-all allocation problem and information is multidimensional and has many other dimensions beyond the problem of scarcity. Additionally, information is not just related to scarcity. The contemporary approach redefined information goods as anything that can be digitized but the traditional medium is not necessary. Moreover, these digital information goods may be copied, shared, resold, or rented to provide revenue (Vafopoulos, 2012) and do not require a material medium. As large amounts of information become available and can be communicated more easily and processed more effectively, information has come to play a central role in economic activity and welfare in our age. Information changes and redefines the principles of contemporary market operation. The crucial element of proper recognition of contemporary information is understanding its unique features. Information goods have a few main properties that would seem to cause difficulties for market transactions (i.e., problems in measuring value or uncertainty associated with a sharp loss in value upon disclosure). As recognized at least since Arrow (1962), information goods are characterized by several peculiar features, such as uncertainty, indivisibility, and inappropriability. Information goods are typically hard to evaluate and their real value estimation is subjective and time-varying. It is often difficult for a user to ascertain the quality of innovation before applying it and sometimes even long after that. Information goods are easy to duplicate. On the supply side, the production of information typically entails a high initial fixed cost (typically it is nonreversible and thus it is a sunk cost). But it is cheap to duplicate the next pieces of information at near-zero marginal cost and with virtually infinite capacity. It means that information goods are non-rival: their use by a consumer does not prevent use by other consumers. Information goods are difficult to exclude. Once information becomes available to one user, it is costly to prevent other users, or competing sellers, from having access to it. Varian (1998) specified three main features of information goods: the necessity of experience, the low marginal cost of reproduction, and their non-rival and sometimes non-excludable character. These features are similar to Arrow’s description but Varian’s explanation was extended. The necessity of experience makes the revealed information priceless and unrevealed unknown. It causes two main concerns: problems of asymmetric information and the necessity to prove the value (by previewing and browsing, creating reviews, or creating a reputation) (Giza, 2024). Before providing the information, its value is contractual, and it is not possible to get to know it exactly without disclosing its content. Disclosure causes a significant loss in value. Additionally, the information has a subjective value, and this is described in economic literature as a credence-quality problem7 (Dulleck et al., 2011). The buyer has to

308  The Elgar companion to information economics trust the seller, and the quality of the information provided may be built over a long period (it cannot be checked immediately after its delivery; the results can be seen in the long run). This leads to the increase of market uncertainty and causes the possibility of moral hazard or holdup8 (hidden intention). Information duplication is possible without loss of utility, at a very low cost (marginal cost equals zero). When selling, the market deals with diminishing average costs, so the seller achieves economies of scale (supply scale effects). This is a very large entry barrier and enhances the competition. Zero marginal cost contributes to an increased risk of any kind of theft. Information is a non-rival good. Some information is of a public good nature due to its non-exclusion and non-competition nature. Currently, it is believed that the high cost of obtaining information and the loss of its value at the time of disclosure results in the desire to exclude entities that did not participate in the cost of obtaining the information. Other important and unique features of information goods are also debated in the economic literature. Information goods do not wear during consumption, transfer, and duplication. Information can be reused indefinitely without loss of quality or value. It has a defined life cycle and can be both destroyed and recovered. Information may be subject to deformation, and falsification as a result of conscious human actions as well as accidental events, and information may increase or decrease the uncertainty depending on the content it contains. Therefore it is a durable good (that cannot be destroyed during consumption; Raine, 2002). The inexhaustibility of this resource means that there is an infinite set of information. Information flows may be subject to retardation, i.e. deliberate delay in its disclosure – the information is cumulative, it grows over time, but some of it becomes unnecessary and out of date. Information can exist objectively (without people’s awareness). The information has an intrinsic value of its own, which varies over time and is subject to the principles of market behavior. Disclosing information causes impairment (or significantly changes it). It means that revealed information loses its market value in transactions as its cost is zero. Arrow (1962) noticed that the value of information is a growing function of wealth; having information makes the rich richer and the poor poorer. Additionally, the value of the information may be cumulative but also the information is not fully divisible. The customer is not sure whether the information received will give him the complete information necessary to make the right decision. Additionally, information is subject to aging, but it is a very individual feature. Information is priceless when needed, then it can be immediately worthless. It is characterized by full asymmetry referred to as the “information paradox”9 or “the pig in a poke syndrome”.10 Machlup describes information goods as unappropriated. After the transaction, both the seller and the buyer are in possession, but the possibility of further use and sale depends on the arrangements set out in the contract. Information can be perceived as a positive externality. Information often benefits many people, not just the owner (person or company). Consumption effects caused by having and revealing information increase welfare and well-being. The analysis of the total benefits of consumption of information goods shows that it would be profitable to produce more of it at a lower price. If this is true, this leads to inefficiency in the Pareto sense: too few resources are allocated to the production and distribution of such information. Intellectual property protection restricts accessibility; even if the information is disclosed it cannot be used. Therefore, the problems of monopolization of information and information and digital exclusion arise as information is known but cannot be used. Intellectual property protection makes information easily accessible; however, it limits free usage. Additional costs can increase exclusion.

Intangibles, information goods, and intellectual property goods in modern economics  309 Drucker (1993) claimed that information should be accepted as the fourth factor of production. He wrote: “there is less and less return on traditional resources: labor, land and (money) capital. The main producers of wealth have become information and knowledge” (Drucker, 1993, p. 183). Traditional factors of production have become secondary to information because having information allows for obtaining each of these factors, therefore it is indispensable in creating and acquiring all other factors of production.11

CONCLUSIONS All the categories described above – intangibles, intellectual property goods, and information goods – have many features in common. All of them are not well-described in economics and existing theories are not universal and widely accepted by researchers and scholars. The creation of proper definitions is doubtful and even the formation of a comprehensive classification of their types causes unsolvable problems. Not to mention the difficulties associated with creating the principles of their market valuation. Nevertheless, the relevance of information has been appreciated by the Sveriges Riksbank (the central bank of Sweden), which in recent years has awarded almost half of all of the Nobel Memorial Prizes to economists dealing with widely understood intangibles. Intangibles always matter and nowadays their volume, availability, and ease of spread cannot be disregarded but their conceptualization still causes problems. The detailed analysis of the special features of intangibles, information goods, and intellectual property goods reveal many similarities, as they all involve very similar, or a large part of the same, issues. But each category is different and has its peculiarities, which distinguish it from the others. The broadest meaning characterizes intangibles. All intellectual goods and information goods can be classified as intangibles, without any exception. However, among intangible goods, some goods share the characteristics not found in information goods or intellectual property goods, such as colors, books containing outdated information, personal feelings, and moods. Intellectual property goods are intangible assets with legal rights, they are non-physical properties that are the products of original thought and can be described as hybrid goods made up of both information and law. But not all intellectual property can be protected by intellectual property rights. Works that are in the public domain are not protected, and copyright does not protect ideas or useful items. Works that have not been fixed in a tangible form of expression are not protected: titles, names, short phrases, slogans, simple product lettering or coloring, or the mere listing of product ingredients or contents, ideas, procedures, principles, discoveries, ideas, procedures, principles, discoveries, and devices are all excluded from intellectual protection. An increasing quantity of various kinds of information is traded daily, but its property rights are unavailable or unenforceable, e.g., financial and market information, news, weather forecasts, databases, and encyclopedic, professional, and leisure-related knowledge (Boldrin & Levine, 2008). Therefore, information is a much wider conceptual category than intellectual property (protected intellectual property). However, information can be defined as intellectual property which can be either protected or unprotected, and protected information is a conceptual synonym only for an intellectual property good. Analyzing and defining the boundaries between intangible goods, information goods, and intellectual property goods is still a current and important research question. The specificity

310  The Elgar companion to information economics of the categories in question makes it impossible to develop a complete theory. Nevertheless, due to the role and importance of these goods in the economy, theoretical considerations have to be undertaken. In conclusion, it is worth quoting the words of Varian, who wrote “if information poses problems for economic theory, so much the worse for economic theory: real markets seem to deal with information rather well” (Varian, 1998, p. 1). Economics describes the economy – it would be good if it kept up with its changes. The modern world is experiencing profound changes in practical dimensions which need similar change and a “new opening” in the theory of social sciences in many fields of interest. The new phenomena – especially the growing role and importance of the new type of goods called intangibles, information goods, and intellectual property goods – are reshaping reality and creating new foundations and rules for market behavior. Classical economics was based on the perfectly competitive general equilibrium model which assumes equal access to the same information and rapid dissemination of information to all market participants. This model was based on the concept of rational decision. This paradigm was changed first by neoclassical economists, who attempted to integrate psychology into economics (however, still assuming utility maximization). In the 1960s, economists changed this paradigm by drawing upon how people irrationally seek satisfaction, rather than maximizing utility. Law and economics has focused on the legal aspect of intangibles’ protection. Economic concepts have been used to explain the effects of laws, assess which legal rules are economically efficient, and predict which legal rules will be promulgated (mainly on the microeconomic level). One of the two main branches of that school has focused on an institutional analysis of law and legal institutions, with a broader focus on economic, political, and social outcomes. Information economics has introduced the concepts of information asymmetry, moral hazard, and adverse selection. These concepts helped to explain market inefficiencies in a wide variety of industries and markets. But information economics, welfare economics, and economics and law have not been able to build a concise theory concerning intangibles. Contemporary economic theory can barely keep up with changes in the economy. New phenomena, such as digitalization, market interdependencies, and the growing importance of intangibles and their varieties are transforming the economy. While intangibles, information goods, and intellectual property are not yet the main focus of contemporary economics, the growing interest in the market, consumer, and firms at the microeconomic level should allow for the reshaping and redefinition of basic market categories, which include, inter alia, all goods, including intangible goods.

NOTES 1.

Marketing-related intangible assets: trademarks, trade names, collective marks, certification marks, trade dress (unique color, shape, or package design), newspaper mastheads, internet domain names, noncompetition agreements. Customer-related intangible assets: customer lists order or production backlog, customer contracts and related customer relationships, non-contractual customer relationships. Artistic-related intangible assets: plays, operas, ballets, books, magazines, newspapers, other literary works, musical works, compositions, song lyrics, jingles, pictures, photographs, video and audiovisual material, including motion pictures, music videos, television programs. Contract-based intangible assets: licensing, royalty, standstill agreements, advertising, construction, management, service or supply contracts, lease agreements, construction permits, franchise agreements, operating and broadcast rights use rights such as drilling, water, air, mineral, timber cutting, and route authorities servicing contracts such as mortgage servicing contracts employment contracts.

Intangibles, information goods, and intellectual property goods in modern economics  311

2. 3.

4. 5.

6.

7.

8.

9.

Technology-based intangible assets: patented technology, computer software and mask works, unpatented technology, databases, including title plants, trade secrets, such as secret formulas, processes, recipes (Nanayakkara, 2012). The exigency of timeless protection of geographical indications and trademarks comes from their rival-in-consumption character as their main economic function is to distinguish some products from others. The two economic categories – information and knowledge – can be distinguished based on three main axes: multiplicity, temporal and spatial. Multiplicity: information is piecemeal, fragmented, particular and knowledge is structured, coherent and universal. Temporal: information is time limited, transitory, even ephemeral, and knowledge is enduring and temporally expansive. Spatial: information is a flow and knowledge is a stock. In conclusion, information is perceived as a process, whereas knowledge is a state (Machlup & Mansfield, 1983, p. 642). What is Information? The Flow of Bits and the Control of Chaos. http://​web​.mit​.edu/​comm​-forum/​ legacy/​papers/​sholle​.html. Public goods theory has been a cornerstone of the economic theory of the public sector since the 1950s. It was inspired by two Paul Samuelson papers, published in 1954 and 1955. On this basis economists have accepted a rigorous definition of the term “public good” (Holcombe, 2000). Property rights define who gets what when a given stream of income is divided. But different kinds of property and thus different kinds of property rights exist. The Ostroms (1977) defined goods using the two binary characteristics of excludability and rivalry/subtractability in consumption. These characteristics define four ideal types of goods with potentially different kinds of property rights: private goods, public goods, common pool goods, and club/franchise goods. Modern law and economics dates from about 1960, when Ronald Coase published “The Problem of Social Cost”. Gordon Tullock and Friedrich Hayek also wrote in the area, but the expansion of the field began with Gary Becker’s 1968 paper on crime. In 1972, Richard Posner, a law and economics scholar and the major advocate of the positive theory of efficiency, published the first edition of Economic Analysis of Law and founded the Journal of Legal Studies, both important events in the creation of the field as a thriving scholarly discipline. The credence goods literature is based on the pioneering paper by Darby and Karni (1973), who introduced this term and added this type of good to Nelson’s (1970) classification which divides goods into ordinary, search and experience goods. Ordinary goods (such as petrol) have well-known characteristics, and subjects know where to get them. Search goods (like clothes) need to be inspected before buying in order to observe their characteristics. Experience goods (like wine) have unknown characteristics, but they are revealed after buying or consuming them. Taylor (1995) provides a theoretical micro-foundation for some features typical for credence goods markets, for instance, the heavy reliance on ex post contracts and the prevalence of free diagnostic checks. Dulleck and Kerschbamer (2006) described liability and verifiability as the most important institutional factors for experts’ behavior and reputation and competition as the most important market factors. Credence goods markets are also characterized by asymmetric information between sellers and consumers that may give rise to inefficiencies, such as under- and overtreatment or market breakdown (Dulleck et al., 2011). The hold-up problem is an economic situation connected with underinvestment. It may happen in the situation when two parties could work most efficiently by cooperating but forbear from doing so because they are afraid that may give the other party increased bargaining power and thus decrease their own profits. A literature review on this issue can be found in Yang (2021). Arrow’s information paradox asserts that demand for undisclosed information is undefined. The information paradox, known also as Arrow’s disclosure paradox (Arrow, 1962) states that the potential consumer of the information or technology wants to have detailed knowledge to understand its capabilities or detailed information about the facts or products to decide whether or not to buy it. Revealing it means that the seller has in effect transferred the technology/information to the customer without any compensation. Arrow wrote: “there is a fundamental paradox in the determination of demand for information: its value for the purchaser is not known until he knows the information, but then he has in effect acquired it without cost” (Arrow, 1962, p. 615). It means that ex ante the buyer cannot assess the value of some particular information, and it can be known only after it has been revealed. But in that situation the buyer has no reason to compensate the seller

312  The Elgar companion to information economics ex post. Hence, there is no demand for information as such. The paradox can be solved through protection for intellectual property, such as patents, not by making the pre-disclosure valuation easier but by removing the disincentive to disclose the information (Gans & Stern, 2003). 10. The “pig in a poke syndrome” is connected with the increasing risk of trust abuse (moral hazard) by the supplier who is conscious of information asymmetry and may try to ruthlessly maximize the profit by limiting the cost. Zielińska (2016, p. 32) lists the main features of the “pig in a poke syndrome”: 1. The buyer does not know what he is buying. 2. The object of the market transaction is the information whose property rights mean transferring a material carrier of the information. 3. The transactions made are of high risk. The buyer is unable to evaluate the information utility ex ante, especially its relevance and pertinence. Also seller’s knowledge of the information’s future usage is limited, therefore, to avoid illegal copying in that the information is offered not as a product, but as a service. 4. The price of the information is often strictly correlated with the costs. 5. The buyer’s needs are not always defined explicitly, therefore, the product is evaluated only by the seller. 11. Key concepts and contributions of Peter Drucker in research upon information and knowledge are described by Bang et al. (2010).

REFERENCES Aghion, P., G. Angeletos, A. Banerjee, & K. Manova (2010). Volatility and growth: Credit constraints and the composition of investment. Journal of Monetary Economics, 57(3), 246–265. Aghion, P., D. Hemous, & E. Kharroubi (2014). Cyclical fiscal policy, credit constraints, and industry growth. Journal of Monetary Economics, 62(C), 41–58. Ahn J., R. Duval, & C. Sever (2020). Macroeconomic Policy, Product Market Competition, and Growth: The Intangible Investment Channel. IMF Working Paper WP/20/25. https://​www​.imf​.org/​-/​media/​ Files/​Publications/​WP/​2020/​English/​wpiea2020025​-print​-pdf​.ashx. American Association of University Professors (2021). https://​www​.aaup​.org. Aplin, T. & J. Davis (2017). Intellectual Property Law: Text, Cases, and Material. Oxford: Oxford University Press. Arrow, K. J. (1962). Economic welfare and the allocation of resources for invention. In National Bureau of Economic Research, The Rate and Direction of Inventive Activity: Economic and Social Factors (pp. 609–626). Princeton, NJ: Princeton University Press. Bang, A., C. M. Cleemann, & P. Bramming (2010). How to create business value in the knowledge economy: Accelerating thoughts of Peter F. Drucker. Management Decision, 48(4), 616–627. Barnes, D. W. (2011). Congestible intellectual property and impure public goods. Northwestern Journal of Technology and Intellectual Property, 9(8), 533–563. Bochańczyk-Kupka, D. (2017). The measurement of the intellectual property value at the microeconomic level. Organization and Management, 114, 51–64. Boettke, P. J. (2002). Information and knowledge: Austrian economics in search of its uniqueness. The Review of Austrian Economics, 15(4), 263–274. Boldrin, M. & D. K. Levine (2008). Against Intellectual Monopoly. Cambridge: Cambridge University Press. Cole, D. H. & P. Z. Grossman (2002). The meaning of property “rights”: Law versus economics? Land Economics, 78(3), 317–330. Corrado, C., C. Hulten, & D. Sichel (2005). Measuring capital and technology: An expanded framework. In C. Corrado, J. Haltiwanger, & D. Sichel (Eds.), Measuring Capital in the New Economy (pp. 11–46). Chicago, IL: University of Chicago Press. Corrado, C., C. Hulten, & D. Sichel (2009). Intangible capital and U.S. economic wealth. Review of Income and Wealth, 55(3), 661–685. Correa-Caro, C., L. Medina, M. Poplawski-Ribeiro, & B. Sutton (2018). Fiscal Stimulus Impact on Firms’ Profitability During the Global Financial Crisis. IMF Working Paper No.18/251. Darby, M. R. & E. Karni (1973). Free competition and the optimal amount of fraud. Journal of Law and Economics, 16, 67–88. Drucker, P. F. (1993). Post-Capitalist Society. New York: Harper Business.

Intangibles, information goods, and intellectual property goods in modern economics  313 Dulleck, U. & R. Kerschbamer (2006). On doctors, mechanics, and computer specialists: The economics of credence goods. Journal of Economic Literature, 44, 5–42. Dulleck, U., R. Kerschbamer, & M. Sutter (2011). The economics of credence goods: An experiment on the role of liability, verifiability, reputation, and competition. American Economic Review, 101(2), 526–555. Fukao, K., S. Hamagata, T. Miyagawa, & K. Tonogi (2007). Intangible Investment in Japan: Measurement and Contribution to Economic Growth. RIETI Discussion Paper Series 07-E-034. Gans, J. S. & S. Stern (2003). The product market and the market for ‘ideas’: Commercialization strategies for technology entrepreneurs. Research Policy, 32(2), 333–350. Garicano, L. (2000). Hierarchies and the organization of knowledge in production. Journal of Political Economy, 108, 874–904. Giza, W. (2024). Asymmetric information as a market failure in retrospect. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 106–117). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Griliches, Z. (1995). R&D and productivity: Econometric results and measures. In P. Stoneman (Ed.), Handbook of the Economics of Innovation and Technological Change (Chapter 3). Oxford: Blackwell. Haskel, J. & S. Westlake (2017). Capitalism without Capital. Princeton, NJ: Princeton University Press. Holcombe, R. G. (2000). Public goods theory and public policy. In J. Narveson & S. Dimock (Eds.), Liberalism: New Essays on Liberal Themes (pp. 125–138). Dordrecht: Kluwer. International Valuation Standards Council (IVSC) (2021). The case for realigning reporting standards with modern value creation. https://​www​.ivsc​.org. Khemani, R. S. & D. M. Shapiro (1993). Glossary of Industrial Organization Economics and Competition Law. Paris: OECD. http://​www​.oecd​.org. Koppius, O. (1999). Dimensions of intangible good. Proceedings of the 32nd Hawaii International Conference on System Sciences, vol. 5. Lindberg, V. (2009). Intellectual Property and Open Source: A Practical Guide to Protecting Code. Sebastopol, CA: O’Reilly. Machlup, F. & U. Mansfield (1983). The Study of Information: Interdisciplinary Messages. New York: John Wiley & Sons. McGrattan, E. & E. Prescott (2005a). Expensed and Sweat Equity. Federal Reserve Bank of Minneapolis, Working Paper no. 636. McGrattan, E. & E. Prescott (2005b). Taxes, regulations, and the value of U.S. and U.K. corporations. Review of Economic Studies, 72, 767–796. Nanayakkara, T. (2012). An Introduction to the Intellectual Property System and the Work of the World Intellectual Property Organization. https://​www​.wipo​.int/​edocs/​mdocs/​sme/​en/​wipo​_smes​_co​_11/​ wipo​_smes​_co​_11​_ref​_theme​_01​_01​.pdf. Nelson, P. (1970). Information and consumer behavior. Journal of Political Economy, 78, 311–329. OECD (2008). Glossary of Statistical Terms. https://​www​.oecd​-ilibrary​.org/​economics/​oecd​-glossary​-of​ -statistical​-terms​_9789264055087​-en. OECD (2012). New Sources of Growth: Intangible Assets. https://​www​.oecd​.org/​sti/​inno/​46349020​.pdf. OECD (2014). Guidance on Transfer Pricing Aspects of Intangibles: OECD/G20 Base Erosion and Profit Shifting Project. Paris: OECD Publishing. http://​dx​.doi​.org/​10​.1787/​9789264219212​-en. Ostrom, V. & E. Ostrom (1977). Public goods and public choices. In E. S. Savas (Ed.), Alternatives for Delivering Public Services: Toward Improved Performance (pp. 7–49). Boulder, CO: Westview Press. Raban, D. R. & J. Włodarczyk (2024). Information economics examined through scarcity and abundance. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 2–19). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Raine, A. (2002). A model for buying and renting information goods incorporating incomplete information. Honours thesis, University of Queensland, Brisbane. Rajan, R. & L. Zingales (1998). The Firm as a Dedicated Hierarchy: A Theory of the Origin and Growth of Firms. NBER Working Paper no. 7546. Rubin, P. H. (2020). Law and Economics. https://​www​.econlib​.org/​library/​Enc/​LawandEconomics​.html. Shapiro, C. & H. R. Varian (1999). Information Rules: A Strategic Guide to the Network Economy. Boston: Harvard Business School Press.

314  The Elgar companion to information economics Stiglitz, J. E. (2000). The contributions of the economics of information to twentieth-century economics. The Quarterly Journal of Economics, 115(4), 1441–1478. Taylor, C. (1995). The economics of breakdowns, checkups, and cures. Journal of Political Economy, 103, 53–74. Vafopoulos, M. (2012). The Web economy: Goods, users, models, and policies. Foundations and Trends in Web Science, 3(1–2), 1–136. Varian, H. R. (1998). Markets for information goods. University of California, Berkeley. https://​archive​ .ph/​wC8T9​#selection​-27​.0–41​.38. WIPO (1967). Convention establishing the World Intellectual Property Organization. https://​www​.wipo​ .int. WIPO (2016). Understanding Industrial Property. https://​www​.wipo​.int. WTO (2018). What Are Intellectual Property Rights? https://​www​.wto​.org. Yang, Y. (2021). A survey of the hold-up problem in the experimental economics literature. Journal of Economic Surveys, 35(1), 227–249. YoungGak, K. (2007). A Survey on Intangible Capital. Center for Economic Institutions. Working Paper Series, no. 2007–10. Yu, P. K. (2016). Five decades of intellectual property and global development. The WIPO Journal, 8(1), 11–22. Zalta, E. N. (Ed.) (2011). The Stanford Encyclopedia of Philosophy. https://​plato​.stanford​.edu/​. Zambon, S., G. Marzo, L. Girella, M. Abela, & N. D’Albore (2019). An Academic Literature Review on Reporting on Intangible Assets for the European Financial Reporting Advisory Group (EFRAG). University of Ferrara. https://​www​.efrag​.org/​. Zielińska, A. (2016). Information as a market product and information markets. Czech Journal of Social Sciences, Business and Economics, 5(4), 31–38.

16. Incomplete contracts, intellectual property rights, and incentives: Investment in knowledge assets under alternative institutional configurations Erkan Gürpinar and Eyüp Özveren

INTRODUCTION Information economy offers new prospects for production organization. Organizational ecology is no longer dominated overwhelmingly by business corporations and producer oriented production of information and knowledge, where the prime motive has always been profit-making. Far from streamlining in this respect, we have witnessed instead the proliferation of alternative institutional arrangements over the last few decades. For example, there are novel forms of information sharing and production that rely on direct individual collaboration, sometimes dubbed networked information economy or commons-based peer production (Benkler, 2002a, 2006; Baldwin and von Hippel, 2011). One factor enabling this transformation is the cheap and efficient interaction of individuals. It is the Internet and the so-called digital economy that create opportunities, in which tangible assets play only a negligible role in the production process, e.g., software development projects. Direct collaboration among individuals, in general, enables the production of user-generated content in information goods. Importantly, profit-making is not the sole motive in many of these novel institutional arrangements. Individual contributors freely share information in multiple collaborative projects (Baldwin and von Hippel, 2011; von Hippel, 2005). There has been, therefore, a renewed interest among researchers to make sense of this shared resource, i.e., information, which was confined to libraries, archives, and paper based distribution of information in the past (Hess and Ostrom, 2011). Information economy also challenges intrafirm relations, and traditional models of doing business. Large corporations started to implement various strategies to take the control of production knowledge at the shop floor around the turn of the twentieth century. Friedrich W. Taylor’s (1911) scientific management prescribed to separate thinking (intellectual skills) from doing (manual labour), and relocating the thinking function exclusively to the managerial authority. It was believed that corporations were not in need of the creativity and intellectual skills their workers possess (Stone, 2013). Today, on the contrary, the tacit and dispersed knowledge workers possess are regarded as scarce and valuable assets of business firms (Benkler, 2002b; Stone, 2013; Teece, 1986, 1998; Teece et al., 1997). Scientific management and managerial authority are no longer regarded as efficient methods of management, since these strategies may be impediments to learning and intellectual skill development (of knowledge workers) (Fisk, 2014; Zuboff, 1989). Any narrative of technological change has as its backdrop a broader social and institutional transformation. This is evident in contemporary discussions regarding the information 315

316  The Elgar companion to information economics economy. We now talk about worker motivation and incentives, since the skills and talent these workers have are believed to be one of the most important knowledge assets of firms. In the information economy, therefore, the range of intangible (knowledge) assets has greatly expanded. In addition to (intangible) information (goods) such as software, blueprints and drawings, human creativity and intellectual skills that are embodied in individuals have started to be treated as valuable resources. In general, knowledge assets are even more difficult to identify and specify (in a contract) in comparison with tangible assets. These assets vary in quality, that is, they are difficult to standardize (Benkler, 2002b). In theory, complete (or efficient) contracting in knowledge assets is possible when there are zero transaction costs, i.e., when specification and exchange of a good in contractual terms bear no significant costs. Under these assumptions, new interdependencies which the information economy has ushered in among agents (firms, knowledge workers and individual contributors), and any (negative or positive) external effect they have on one another could be eliminated, since costless exchange of (property) rights is possible. Thereby, the final allocation of assets could be efficient (Coase, 1960; Williamson, 1985). In the information economy, this outcome depends on, foremost, a well-defined intellectual property rights (IPR) regime, which could enable efficient exchange and bargaining of knowledge assets without imposing significant transaction costs on market participants. Notwithstanding that, even in tangible assets, a well-defined property rights regime and efficient exchange and bargaining of assets are challenging tasks (Alchian and Demsetz, 1972; Coase, 1960; Demsetz, 1967; Williamson, 1985). The design of such a system for knowledge assets is even more difficult due to above mentioned features of knowledge assets. In the information economy, knowledge is the most valuable asset for business firms. When dividing the gains from knowledge production, the management of these assets via legal tools such as patents, trade secrets or non-compete agreements has important consequences on both interfirm and intrafirm relations. Since knowledge assets are largely produced by knowledge workers, intrafirm relations are an indispensable part of an analysis of interfirm relations of knowledge production as well. Firms devise various incentive mechanisms in order to motivate knowledge workers to contribute to knowledge production, e.g., flattened hierarchies, and company shares. Incentives to knowledge workers also matter, because collaboration and cooperation (e.g., licensing agreements) rest not only on legal arrangements at the managerial (or firm) level. This is so, because technology specified in a patent is significantly embodied in knowledge workers’ skills and expertise. These types of knowledge assets are left untouched by patents, and are subject to other legal tools such as trade secrets and non-compete agreements. Overall, worker incentives are key to understanding not only how firms produce knowledge assets but also how they collaborate (i.e., exchange knowledge assets) in the market (Haber and Lamoreaux, 2021; Marx and Fleming, 2012). Indeed, collaborative projects among individual participants address the same incentive problem in a novel way: self-identification with the tasks in projects. It is assumed that individuals possess better information than managerial authority on how to match their skills with various tasks in projects (Benkler, 2002b, 2006). Finding the right incentives for knowledge workers is especially challenging for firms. This is because the rising importance of knowledge workers takes place in an environment where start-ups, direct collaboration projects and small entrepreneurial firms all contest the standard (employment) contract that has been dominant in the past. Independent contracting, and temporary working are now more widespread than they were in the twentieth century (Simon, 1979; Stone, 2013). Therefore, firms should

Incomplete contracts, intellectual property rights, and incentives  317 not only find out the right incentives for workers, but at the same time, figure out ways (usually legal arrangements) to retain these assets developed by workers within the firm, in case of job termination (Fisk, 2014; Hyde, 2003; Marx and Fleming, 2012; Pagano and Rossi, 2004). Today, various strategies leading to alternative institutional configurations in production are used to address the need to balance the firms’ interest with those of the individuals. In this regard, IPR stand out as one of the most important factors that explain institutional diversity, at both intra-organizational and inter-organizational level. We will nevertheless insist that a more elaborate alternative (yet complementary) perspective on the original problem, i.e., incentives to investment in knowledge assets, is badly needed. To this effect, we proceed first by taking up the intricate connections between the free-rider problem, the IPR regime and the prospects of investment in knowledge assets. We then address how the alternative perspective would in fact suit better to the needs of cooperation, coordination and knowledge creation in our times. We use simple game theoretic illustrations in order to compare and contrast the contents of the problem addressed in alternative approaches. In the concluding section, we summarize the fundamental differences between physical assets and knowledge assets, and derive some policy implications.

1.

FREE-RIDERS, IPR AND INVESTMENT IN KNOWLEDGE ASSETS

It is known that information is a unique good (Arrow, 1962). Of course, this does not mean that information is not amenable to economic analysis. Its uniqueness partially comes from the fact that it is a (pure) public good, having non-rivalry and non-excludability features. While non-rivalry enables easy dissemination of existing information, hence creating positive externalities, non-excludability discourages producing new information by preventing full appropriation of its benefits by the producers themselves. That is, involuntary dissemination results in underinvestment in information production. According to this formulation, then, a social dilemma is potentially abundance (or not scarcity) ridden. The main obstacle to efficient allocation of the resource, i.e., inadequate appropriation, is nothing but an example of externalities and contractual incompleteness, which are widespread in many market and non-market interactions.1 Alternative institutional configurations, then, are means of alleviating the severity of contractual incompleteness. In theory, institutional configurations having lower transaction costs are chosen by market participants. In practice, however, almost every institutional solution includes a mixture of public and private mechanisms (Coase, 1960; Williamson, 1985). The former method (sometimes dubbed public ordering) is usually not enough, since firms (private orderings) are able to better use (that is incurring lower transaction costs) managerial discretion for their specific (e.g., sectoral) needs (Pagano, 2007; Williamson, 1985). In information production, historically, various mechanisms of public support have been central to alleviating the underinvestment problem. The university system, public R&D, government sponsored prizes and grants are all examples of public involvement in information production. The alternative, and complementary mechanism is to create a well-functioning market system for information and knowledge. In this regard, IPR is the main tool, since without well-defined property rights, an efficient market mechanism cannot exist (Demsetz, 1967). Before digging into the merits and shortcomings of the IPR system, we need to disclose another important ingredient of the public good approach.

318  The Elgar companion to information economics Overlooking the diversity in human motivations, this approach puts utmost importance on self-interest and profit-making. Indeed, interdependence between behaviours of agents, and the (positive and negative) external effect they have on one another becomes a problem worthy of consideration only under self-interest (Williamson, 1985). Self-interest, and opportunistic behaviour are crucial under contractual incompleteness, since they may lead to the well-known free-rider problem in strategic interactions. Overall, contractual incompleteness combined with self-interest constitute the essence of the underinvestment problem in information production as we will see below. In collaborative projects, free-riding refers to any action that an individual takes in order to increase one’s own benefit at the expense of (that is, with the effect of decreasing) common benefit. Free-riders, in general, undermine the success of collective projects (Olson, 1965). In this regard, involuntary dissemination of information (the characteristic of the asset), and imitation by free-riders (the behaviour of self-interested agents) constitute the public good problem in information production. 1.1

Knowledge Appropriation Problem as a Prisoner’s Dilemma Game

The inner structure of the problem can be better illustrated by using the Prisoner’s Dilemma (PD) game, which is used to represent a set of undesirable outcomes in social dilemmas. Consider two individuals who strategically interact in order to produce information. Each of them decides whether to invest in information production (choose invest strategy) or not (choose don’t invest strategy), i.e., each player has two (pure) strategies: invest or don’t invest (see Table 16.1). Assume that the agent who chooses invest incurs a cost. Assume that cost of investment is 3. Importantly, since information is a public good, that is both non-excludable and non-rival the agent who chooses don’t invest could still reap the (full) benefits of information produced (without incurring any cost) by the other. The agent who decides not to invest is the so-called free-rider, who simply benefits from the investment of the other agent. Assume that when both choose invest, they each receive 5. On the other hand, when only one agent chooses invest, the benefit is 3. Taking into account the cost of investment, benefits to alternative strategies are as follows: When both invest, each receive 2; when only one agent chooses invest her benefit is 0, and the benefit to the free-rider is 3. Lastly, assume that when both agents choose don’t invest, benefit is negligible, e.g., 1. The (Nash) equilibrium of this game is {don’t invest, don’t invest}, because don’t invest is a best response to both invest (since 3>2), and don’t invest (since 1>0). On the other hand, {invest, invest} outcome is actually better for both players (since 2>1). This interaction describes the well-known PD game, where both agents choose don’t invest in information production, even though there is a better alternative, in which both would choose invest. Yet, this outcome is not realized, since each agent is tempted to free-ride on the contribution of the other agent.2 Table 16.1

Information production as a PD game

   

invest

don’t invest

invest

 ​2, 2​

 ​0, 3​

don’t invest

 ​3, 0​

 ​1, 1​

As argued above, the underinvestment problem is overcome by relying on an alternative incentive mechanism (such as public subsidies, grants, and IPR). Importantly, depending on the context, institutional configurations having lower transaction costs are chosen, that is, efficient

Incomplete contracts, intellectual property rights, and incentives  319 allocation of the resource is still possible (Aghion and Tirole, 1994; Alchian and Demsetz, 1972; Hart, 1995; Grossman and Hart, 1986; Hart and Moore, 1990; Williamson, 1985). In information production there are alternative institutional arrangements as well. Public grants and prizes are examples of centralized mechanisms, whereas IPR is the most well-known and widespread decentralized mechanism (Scotchmer, 2004). Public grants and prizes are usually provided by the wealthy (patrons, firms, etc.) and public authorities. IPR, on the other hand, aims to turn information into a (temporarily) excludable good. Therefore, any IPR regime determines to what extent society wants to convert information into an excludable good. 1.2

Information, Knowledge Assets and IPR as an Incentive Mechanism

IPR refers to several legal arrangements including patents, copyrights and trade secrets (Besen and Raskind, 1991). The history of patents, for example, dates back to the Statue of Monopolies in Britain (1623), or even to monopolies granted in Venice (1474) (Moser, 2012; Scotchmer, 2004).3 IPR protects the knowledge of (or template for) making something (Scotchmer, 2004). The IPR system is unique (among other incentive mechanisms) insofar as it is the only decentralized mechanism by which the underinvestment problem is addressed.4 The decentralization of investment decisions in information production is the main means of creating an efficient market for information goods (Gallini and Scotchmer, 2007; Scotchmer, 2004). This is where the comparative advantage of the IPR system lies: it facilitates the creation of a market for various kinds of information goods (Gans and Stern, 2010). To what extent an IPR regime achieves its promises is related to its ability to alleviate contractual incompleteness. Before delving into this discussion, note that the market solution favoured by IPR is not costless, since it creates a temporary monopoly, itself an anticompetitive solution (Ramello, 2005). Therefore, consumer welfare decreases due to deadweight loss, i.e., the monopoly price is always above the marginal cost of production. Yet, when there are (positive) transaction costs, there is no free-lunch. The consumer welfare loss is (partially) compensated by another merit of the IPR system. The costs of creating new information by (private) agents are only covered by users, and not by all taxpayers as in the case of prizes or grants (Scotchmer, 2004, p. 39). Notwithstanding that, the main difficulty of running an IPR system lies elsewhere. Uncertain and overlapping IPRs can make (efficient) negotiation and exchange of information goods difficult, when different pieces of (useful) information are owned by multiple agents.5 There may be high transaction costs for efficient market exchange to take place in this case (Gans and Stern, 2010). This issue arises when IPR is poorly defined. In patents, for example, the fundamental difficulty is determining the boundaries of a technique covered by a (single) patent. That is, IPR is usually unable to (perfectly) capture the truth about the boundaries between knowledge assets. Therefore, for example, different patents may cover overlapping areas of closely related (overlapping) technologies (Heller, 1998; Moser, 2012).6 At the extreme case, multiplicity of overlapping patents may block one another and lead to the, now famous, tragedy of the anti-commons (Heller, 1998; Heller and Eisenberg, 1998) (see next section). Licensing is a solution to overcome blocking patents (Gallini and Scotchmer, 2007; Scotchmer, 2004). It is an important tool especially for small firms and start-ups, which specialize in trading a certain type of information goods: knowledge assets which are detached from tangible assets as well as knowledge workers, i.e., information goods. Technology firms in biotechnology and software industries, for example, do not have downstream marketing and production capabilities. Therefore, they could not profit from embedding technological

320  The Elgar companion to information economics knowledge in final products. Instead, they rely on the creation of a market, and tradable rights via IPR. When these are combined with licensing, the resulting institutional configuration facilitates the proliferation of science driven small firms (Arora and Gambardella, 2010; Haber and Lamoreaux, 2021). Of course, the solution is not automatic. The need to license various technologies from several IPR owners may still be prohibitively costly. Transaction costs of such multilateral bargaining include not only expensive licence demands, but also indirect costs of longevity of bargaining, and the more fundamental uncertainty of dividing the gains from such agreements (Bessen and Maskin, 2009; Boldrin and Levine, 2008, 2012). The latter, dividing the gains from cooperation, concerns the more fundamental problem of information asymmetries over bargaining, when each party’s bargaining power helps determine the final outcome (Scotchmer, 2004). Put simply, licensing may also fail.7 In addition to being complementary, knowledge assets have a high degree of cumulativeness, i.e., ideas almost always build on the ideas of others (Scotchmer, 1991). In other words, existing information is an input to new information production. Efficient bargaining over dividing the gains from information production across time is even more difficult, since one has to make sure that earlier contributors to information production are compensated for their efforts, while at the same time, keeping prospects and incentives open for subsequent contributors. Asymmetric information is still the most important impediment to such bargaining. The one who comes first may hold up subsequent contributors by means of demanding high licensing costs. Or, if the costs are already incurred by the first contributor, her bargaining position could be greatly reduced in the process (Bessen and Maskin, 2009; Scotchmer, 2004). The difficulties of efficient exchange and bargaining, when information production is cumulative, are found in dividing the gains from basic research and commercial applications.8 Investors in basic research could make profits only if they are able to receive licensing fees from subsequent developers at the commercial application stage. Yet, recall that licensing is possible when subsequent developments infringe upon previous developers’ patents. However, basic research is usually not patentable! (Scotchmer 2004, p. 129). The distinction between science (basic research) and technology (applied research and commercial development) depends on whether knowledge created is appropriable by IPR or not. The fuzzy line between science and technology, and concomitant legal amendments since the 1980s are partly responsible for this novel problem (Scotchmer, 2004).9 Today, there are university researchers who found start-ups, and at the same time, there are science oriented companies doing more basic research than traditional firms, e.g., biotechnology firms. The difficulty of drawing borders between science and technology reveals how the sequential and cumulative nature of information complicates solutions based on IPR. Overall, the efficiency of IPR may be less disputable in static settings, whereas it raises doubts in dynamic interactions. In the latter case, a strong IPR regime may actually discourage investment in information production (Bessen and Maskin, 2009; Ramello, 2005; Scotchmer, 1991). The efficient market hypothesis for knowledge assets may fail for another specific reason. Efficient exchange usually entails more knowledge than covered by a patent or IPR. Licensing agreements do not permit easy access to uncodified knowledge embodied in skilled workers, who produced the (patented) knowledge asset (Moser, 2012). As argued above, knowledge assets include not only information, but also skills and capabilities at the firm and individual level. In this regard, when IPR is exchanged in licensing agreements, the transfer of know-how is extremely difficult to specify in a contract (Arora, 1995; Teece, 1977). In essence, efficient exchange of knowledge assets entails more than information exchange. Information (blueprint,

Incomplete contracts, intellectual property rights, and incentives  321 drawings, etc.) exchange explains only a small part of how the market for knowledge assets operates; therefore, know-how embodied in human intellectual skills is needed to efficiently utilize information covered by a license (Boldrin and Levine, 2021). Overall, in order to understand how knowledge is produced within and shared between firms, we have to analyse the implications of the distinction between information goods and the broader set of knowledge assets firms possess.10 Because, when firms collaborate, what they exchange goes beyond easily contractible information goods, and includes a broader set of skills and intellectual assets embodied in knowledge workers. At the turn of the twentieth century, business corporations started to conduct R&D internally. This was partly because of technology. Several new technologies were beyond the reach of individual producers (Fisk, 2014; Stone, 2013). The legal counterpart of this development, the so-called corporatization of IPR, is visible in statistics. Whereas at the end of nineteenth century roughly 10 per cent of patents were issued to firms and the rest to independent inventors, at the end of the twentieth century the figures were reversed (Merges, 2000). In a certain sense, the rise of corporate R&D complemented scientific management, in which legal default rules were devised (with employment contracts) to assign ownership of knowledge assets to firms (Fisk, 2014; Merges, 2000). Today, this practice comes under close scrutiny due to (1) the decline of the classical authority relation and the rising importance of workers’ knowledge assets in the production process, and (2) the renewed growth in technology markets. Firms extensively rely on external sources of acquiring knowledge assets (Arora and Gambardella, 2010). These external sources include information as well as knowledge embodied in workers’ (intellectual) skills. Therefore, firms should protect their knowledge assets against worker turnover (via strong IPR implemented against workers), and at the same time, motivate workers to invest in their (usually firm related) intellectual skills. To this end, firms deploy strategies that restrain workers from owning their own knowledge assets developed during the employment period. These strategies explain the renewed interest of firms in trade secrets and restrictive non-compete covenants (see next section). In essence, exchange in knowledge assets goes beyond what is covered by information (hence patents), and includes workplace knowledge embodied in workers (Fisk, 2014; Merges, 2000; Stone, 2013). Although important, IPR is not the only tool that facilitates information and knowledge exchange. For example, public funding of (especially) basic research has been the central element of science and technology policy in the twentieth century (Bush, 1945; Stokes, 1997). Basic research still relies on (publicly funded) universities, and government research laboratories. Moreover, firms receive significant R&D funding from governments. For example, the decline of basic research in corporate R&D has been quite visible over the last few decades (Arora et al., 2018). This is not because basic research is deemed unimportant by firms. Firms are reluctant to invest in basic research themselves, and instead, specialize in applied research and commercial applications. They leave science to small start-ups and (publicly funded) universities (Arora et al., 2018; Fleming et al., 2019). Importantly, there is also ample evidence that agents exchange information and knowledge outside the market system, and even without any government intervention to this effect (Arora and Gambardella, 2010; Hess and Ostrom, 2011; Potts, 2019). Knowledge sharing and collaborative production, of course, are not new phenomena in the history of science and technology (Allen, 1983; Nuvolari, 2004). However, these developments are not easy to reconcile with the conventional public goods approach. We therefore need an alternative (yet complementary) perspective on the original problem, i.e., incentives to investment in knowledge assets.

322  The Elgar companion to information economics

2.

COOPERATION, COORDINATION AND THE KNOWLEDGE CREATION PROBLEM

Rightly assuming that the production and exchange of knowledge assets take place under contractual incompleteness (with positive transactions costs), we can now take the issue one step further. The fundamental problem may not be appropriability, hence the opportunistic behaviour of free-riders, but instead, the uncertainty and difficulty in creating a non-existing resource, i.e., knowledge (Potts, 2019). Hence, free-riders may not be the fundamental threat, since the resource is not available at first hand. In other words, unless certain conditions are met, there would be scarcity in knowledge production. It is only because agents voluntarily collaborate and cooperate in knowledge production and dissemination in both scientific and technological communities that knowledge is continuously produced and reproduced. On the contrary, for tangible assets such as natural resources, free-riders pose the fundamental threat, since for those resources, benefits available to individuals decrease with use and over-consumption, i.e., the resource is rival. However, open access to knowledge assets is not like open access to some other common pool resources such as land or water, since subtractability (or rivalry) is not a fundamental threat. Recall that information is a non-rival good, simply because there is no depletion in use.11 This feature of knowledge assets partially explains why, historically, open access regimes have been so common in information and knowledge. Researchers point out that universities and academia have flourished with the ethics of full disclosure (openness) and sharing of individual contributions (Bollier, 2011; David, 1993). For example, disclosure of findings (combined with the scientific method) greatly contributed to the reliability of science as a form of knowledge (Ziman, 1976; Schweik, 2011). The survival of scientific communities, for example, could be explained by the “invisible college” argument, in which agents collaborate in order to create a non-existing resource (Kealey and Ricketts, 2014, 2022). The argument is similar to what Allen (1983) calls “collective invention”. His historical analysis shows how technology sharing among competing firms in iron and steel industries during the nineteenth and twentieth centuries, contributed to cumulative knowledge creation and accumulation. He even notes that this phenomenon is observed in many other industries, hence has a broader application (Allen, 1983, p. 1). The analysis of Nuvolari (2004) goes back to the industrial revolution. He shows that knowledge sharing among agents was central to the creation of technologies such as the pumping engine. Let us take a more recent example. In free and open source software development, individual contributors are usually unpaid. Management and direction, hence hierarchy, are very limited. Moreover, legal restrictions are almost non-existent (Maurer and Scotchmer, 2006; Lerner and Tirole, 2004). Yet, individuals usually share their contributions with the sole expectation that the others will do the same, i.e., there is reciprocity (Maurer and Scotchmer, 2006; Benkler, 2002b). A more recent and well-known firm level evidence is cooperation and collaboration practices among firms in Silicon Valley (Saxenian, 1994; Gilson, 1999; Hyde, 2003). Gilson (1999) argues that such knowledge sharing practices resemble the industrial district arguments of Alfred Marshall (1890), with the important difference that, today, economies of scale and scope are achieved through an intangible asset, i.e., knowledge. All these examples about sharing and collaboration in scientific endeavour, industry or direct individual collaboration are not easy to reconcile with the free-riding narrative. Importantly, almost all these studies share the belief that creation of knowledge is not an individual level phenomenon, that is, collaboration and cooperation of agents is necessary.12

Incomplete contracts, intellectual property rights, and incentives  323 2.1

Knowledge Creation as a Coordination Game

Knowledge is a non-existing asset (Hess and Ostrom, 2011; Frischman et al., 2014). The fundamental challenge is the creation of knowledge assets, which could better be illustrated as a coordination problem (Kealey and Ricketts, 2014, 2022; Potts, 2019). This problem could not be reduced to the free-rider problem, since (a) it is not easy to imitate the contributions of others, but also, (b) almost impossible to create knowledge in isolation, i.e., knowledge production is an activity strictly based on collaboration and coordination among agents.13 The fundamental problem is dispersed and tacit knowledge. In this framework, transaction costs arise due to the need of pooling and sharing fragmented knowledge possessed by individuals (Hayek, 1937, 1945; Polanyi, 1958, 1967; Potts, 2019). The alternative framework therefore emphasizes the need for coordination and cooperation in knowledge creation, in which there are differential benefits to agents who invest in knowledge production vis-à-vis agents who don’t invest. Hence, differential benefits to contributors over non-contributors alleviate the free-rider problem (Aydogmus and Gürpinar, 2022; Kealey and Ricketts, 2014, 2022; Potts, 2019). The structure of any coordination problem, in its simplest form, could be illustrated as a stag-hunt game, attributed to Jean-Jacques Rousseau (1992 [1755]). It is also known as the common interest, or assurance game (Bowles, 2004). Consider the strategic interaction structure put forward in the previous section. In a similar vein, each agent decides whether to invest in knowledge production or not (see Table 16.2). Recall that the agent who chooses to invest incurs a cost, i.e., the cost of investment is 3. Yet, different than the previous case, in this setting, knowledge is not assumed to be a (pure) public good. Since knowledge is tacit, and partially embodied in individual talent and skills, it does not disseminate automatically to the agent who does not contribute to knowledge production, i.e., it confers differential benefits to agents. In other words, the agent who chooses don’t invest is not able to reap the full benefits of the contributor’s investment. Moreover, pooling of investment efforts is crucial, since, alone, agents are not able to reap any benefits from knowledge assets at their possession: ideas in isolation barely have any value. Assume that when both agents choose invest, they each receive 6. If only one agent chooses invest her return is 3, while the agent who chooses don’t invest receives only 2, i.e., different than the PD game, there are less benefits to the free-riders. Similarly, assume that when both agents choose don’t invest, benefits are negligible, e.g., 1. Table 16.2

Knowledge creation as a coordination game

   

invest

don’t invest

invest

 ​3, 3​

 ​0, 2​

don’t invest

 ​2, 0​

 ​1, 1​

The (Nash) equilibria of the game are {invest, invest}, and {don’t invest, don’t invest}. Because, invest is the best response to invest (since 3>2), and don’t invest is the best response to don’t invest (since 1>0). On the other hand, one of the outcomes, {invest, invest} is better for both agents (since 3>1). This interaction illustrates a coordination game with two possible outcomes: Either both agents choose invest, or both restrain from investment and choose don’t invest, i.e., they either overcome the coordination problem by pooling their knowledge assets, or fail in achieving cooperation.14 Importantly, instead of always ending up in abstaining from contribution, in this alternative setting, there is a (better) possibility that they both contribute

324  The Elgar companion to information economics to knowledge production. What makes the difference? In this setting, agents do not unconditionally free-ride and avoid cooperation, i.e., an agent may decide to invest in knowledge production if s/he believes that the other agent will do the same. Therefore, mutual coordination, or cooperation is possible. 2.2

Moving Beyond the Public Good Approach: Some Implications

The knowledge creation problem points out the relevance of two features of knowledge assets: First, open access does not mean that knowledge sharing has no boundaries at all. The difficulty of knowledge dissemination creates natural boundaries (Potts, 2019). In other words, as we have pointed out above, there are differential returns to contributors over non-contributors. This concern is relevant both at the firm and individual level, and partly explains knowledge sharing in basic science as well as industry level commercial applications. Knowledge is not a (pure) public good, and its dissemination is not automatic. This is why free-riders are a lesser threat than originally formulated by the public good approach. This fact is well known to the economists of technology. Free-riders (or imitators) cannot easily enter the market (by copying the idea of an inventor), and drive the revenues of the inventor to zero. Rather, imitation and entry are costly, since they require access not only to information (blueprints, drawing, software, etc.), but also a broad range of assets including human, and specialized non-human capital (Bessen and Maskin, 2009; Boldrin and Levine, 2021). As we repeat time and again, knowledge assets consist of multiple types of goods, practices and capabilities. This is why they are subject to varieties of legal regimes and governance mechanisms. Indeed, the now-famous tripartite division as data, information and knowledge, introduced by Machlup (1984), captures these variations, and shows how production and dissemination of each type of knowledge assets is subject to alternative institutional arrangements. To sum up, whereas some features of knowledge assets are amenable to the public goods approach, others require that we broaden our perspective by bringing into consideration the specific nature of the asset concerned. Second, knowledge as a resource is not naturally given. Its production requires permanent interaction of individuals. This is why, simply, lack of communication and interaction among agents may explain coordination failures. A well-known remedy is sustained interaction between agents within (usually) small communities. Foremost, this allows alleviation of communication failures that may lead to the failure of cooperation among agents (Hess and Ostrom, 2011; Frischmann et al., 2014; Ostrom, 1990). Even so, knowledge production is not only social but also a highly personal process (Polanyi, 1958, 1967; Nelson and Winter, 1982). This is why individual motivations and incentives matter. These motivations are psychologically richer in content as well as in consequence, and some of them go well beyond the profit-making motive and its self-interest assumption. A good example is belonging to a community along with assuming the responsibilities this brings (Maurer and Scotchmer, 2006). At a broader level, the spectrum includes intrinsic and extrinsic motivations. For example, in software projects, social psychology incentives (such as reciprocity, sharing, etc.) matter, when contributors are asked to contribute relatively small levels of extra effort (Benkler, 2002b; Maurer and Scotchmer, 2006). These communities also deploy legal measures to avoid socially undesirable outcomes, e.g., a General Public Licence is used to restrict an opportunistic developer who can free-ride on earlier contributions (Maurer and Scotchmer, 2006). Therefore, usually a mixture of motivations including self-interest, profit-making as well as

Incomplete contracts, intellectual property rights, and incentives  325 joy, reputation, satisfaction of sharing, and reciprocity enable these projects to succeed. That is, communities of individuals themselves are able to design informal and formal methods to overcome various social dilemmas in knowledge production. These collaborative projects, sometimes dubbed “commons based peer production”, are examples of a system of rules of cooperation that facilitate pooling of knowledge assets under high uncertainty (Benkler, 2002b; Hess and Ostrom, 2011; Potts, 2019).15 Incentives at the individual level matter not only for collaborative projects among individuals themselves, but also for intrafirm production of knowledge. Here lies the fundamental difference between non-human and human assets: Allocation of knowledge assets usually entails the allocation of humans themselves (Pagano, 2007). Therefore, production organization, and legal arrangements could boost as well as undermine investment in human capital and intellectual assets in firms. When it comes to incentives to investment in knowledge assets, then, the existing institutional structure, which mainly consists of the IPR regime, matters at the interfirm and intrafirm relations. This is why the implications of the incomplete contract framework are more nuanced. When contracts are incomplete, agents may not and even cannot coordinate their interactions across several interdependent domains (Aoki, 2001; Pagano and Rowthorn, 1994). In addition, one further implication of such complexity in the information economy is that firms, knowledge workers as well as individuals who collaborate outside the market system do not interact in a neutral institutional environment. That is, incentives to invest in knowledge assets are affected by the prevailing institutions, i.e., the IPR regime. The challenge is related to the fact that dividing the gains from knowledge creation between firms, as well as firms and workers may not be as easy as illustrated by the coordinating game structure introduced above. What is more, incentives of agents may be positively or negatively affected by this conflictual nature of the division process. Hence, IPR and incentives should be studied under a unified framework, in which they affect each other.

3.

CONFLICT OVER DIVIDING THE GAINS FROM KNOWLEDGE CREATION

When the premises of the coordination problem framework are made explicit, an important result follows. The incentive effect of the allocation of IPR among agents is vital for another reason apart from the free-rider problem. The existing distribution of knowledge assets and the control rights over them (via IPR) could actually prevent the efficient re-allocation of assets to agents who value them more (Pagano and Rossi, 2004). This is because the existing IPR allocation on knowledge assets may discourage firms as well as individuals from further investment in knowledge assets. Returns to exclusive control rights over knowledge assets may be enormous, which may then push firms to use these rights as part of their business strategy to deter other firms from investing in knowledge assets (e.g., via threat of infringement), instead of encouraging them to collaborate in knowledge production (Boldrin and Levine, 2008, Jaffe and Lerner, 2004). Similarly, at the intrafirm level, the exclusion of workers from having access and usage rights to knowledge assets (via trade secrets and non-compete agreements) that they have already invested in may lead to a similar disincentive effect. Hence, under IPR, there is no direct mechanism that guarantees cooperation and collaboration. That is, potential abundance produces new dilemmas (or inefficiencies) in knowledge production and dissemination.16 In this respect, the IPR regime may lead to a vicious circle, a self-reinforcing obstacle

326  The Elgar companion to information economics to efficient allocation of resources between agents who already possess knowledge assets, and who are initially deprived of owning any knowledge assets.17 3.1

Knowledge Creation as a Disagreement Game

The nature of this problem could better be illustrated by adding a further twist to the coordination game introduced in the previous section. As we have already pointed out, collaboration and cooperation are necessary in order to create new knowledge, which could simply be illustrated as a coordination problem. However, there may be conflict over how to divide gains from such collaboration and cooperation when the effect of IPR is taken into account. The basic structure and implications of such a conflict, and the unintended consequences (regarding the premises of IPR regime) it brings about could simply be illustrated as a Disagreement Game (Bowles and Halliday, 2022), also known as Battle of Sexes (Luce and Raiffa, 1957, pp. 90–94). Consider an interaction between two agents who collaborate in order to produce knowledge. Yet, the agents have different roles (see Table 16.3).18 One agent (row player) has control rights over knowledge assets, and thereby is able to decide on openness (share strategy) and secrecy (withhold strategy) over knowledge assets. The other agent (column player) does not own knowledge assets, and decides to invest in knowledge assets. In this framework while the row player represents actors who already own knowledge assets, the column player represents agents who want to invest in knowledge creation, yet are deprived of owning any knowledge assets. Hence, this simple formulation allows us to analyse the cumulative and sequential nature of knowledge production in a unified framework, in which IPR and incentives to invest are co-determined. When the row player chooses withhold, the ownership of the asset is not shared, and is kept under exclusive control. This strategy corresponds to strict control over knowledge assets, i.e., implementing strong IPR. The agent who chooses withhold, then, receives extra benefit (that is profit) from the ownership of the asset. This is the essence of the argument for property rights on knowledge assets, i.e., reap the benefits of the asset by holding control rights on them. On the other hand, share corresponds to weak IPR protection, i.e., relinquish some ownership rights on knowledge assets. Forgone benefits due to sharing, at the same time may function as a signal for collaboration to the other agent. The choice of such weak IPR protection, in other words, makes investment in knowledge assets less costly (and more encouraging) for the other agent. On the contrary, withhold leads to the risk of non-compensation (or under-compensation) of the column player, if she is not able to (fully) reap the benefits of his investment specific to the knowledge asset under the control of the row player. In other words, weak IPR regime (hence, sharing benefits) with the column player may only be preferrable when she is expected to collaborate in knowledge production, i.e., share is best response invest (2>1). On the contrary, having a strong IPR regime is more preferrable when the other agent is expected to free-ride in knowledge production, i.e., withhold is the best response to don’t invest (3>0). In essence, the row player is willing to implement weak IPR only when she expects the column player to reciprocate, i.e., does not free-ride in knowledge creation. The column player decides to contribute to knowledge production or not. She chooses between two strategies: invest or don’t invest. In a similar vein, the choice of the column player is conditional on the choice of the row player. That is, collaboration and investment in knowledge creation is preferrable when the row player is expected to choose weak IPR protection, i.e., invest is the best response to share (3>0). On the contrary, defection or no contribution

Incomplete contracts, intellectual property rights, and incentives  327 may be preferrable when the row player is expected to choose strong IPR, i.e., don’t invest is the best response to withhold (2>1). In the latter case, the possibility of non-compensation of her investment creates a disincentive effect for the column player.19 This interaction has two outcomes: Either the row player retains exclusive control rights over knowledge, and the column player has almost no incentives to invest in knowledge creation; or the row player chooses weak IPR, and the column player reciprocates by investing in knowledge creation. That is the Nash equilibria of the game are {share, invest}, and {withhold, don’t invest}. Importantly, the gist of the argument lies in the fact that none of the outcomes is mutually preferable by both players, i.e., outcomes are not (Pareto) comparable. The row player prefers {withhold, don’t invest}, whereas the column player prefers {share, invest} outcome. That is, knowledge sharing (secrecy), and investment (no investment) feedback (or reinforce) each other. This type of interaction describes a situation which is also known as institutional complementarities.20 In the information economy, such complementarities between the strategies of agents are found both at the interfirm and intrafirm level (Gürpinar, 2016a, 2016b; Landini, 2012, 2013). Table 16.3

Knowledge creation as a disagreement game

   

invest

don’t invest

share

 ​2, 3​

 ​0, 0​

withhold

 ​1, 1​

 ​3, 2​

Overall, in the information economy, the decision to invest in knowledge assets cannot be studied in isolation. It is affected by the existing IPR regime (Elkin-Koren and Salzberger, 2004, 2012; Pagano and Rossi, 2004). The opposite direction of causality is also relevant. The decision to share ownership rights on knowledge assets is affected by the investment decisions of agents who are willing to contribute to subsequent knowledge creation.21 3.2

Cooperation and Conflict in Knowledge Creation: Some Implications

This illustration also explains how the characteristics, i.e., legal status of knowledge assets is determined within the institutional system (largely by firms and individuals who are in a position to decide whether or not to make knowledge a shared resource). This decision, in turn, affects the incentives for agents who want to invest in knowledge production. Recall that, for scarce resources (i.e., natural resources such as oceans, lakes, land) it is their biophysical features that are crucial in determining the nature of the social dilemma (Frischmann et al., 2014; Hardin, 1968; Hess and Ostrom, 2011; Ostrom, 1990). These features (of land, water, forests, etc.) are rather exogenous to the interaction, e.g., rivalry (for a depletable resource) leads to over-usage when exclusion is not possible, for example, in the case of common pool resources. This is known as the tragedy of the commons since Hardin (1968). On the other hand, knowledge assets are not given by nature, therefore, their features are partially determined when individuals and firms create the resource themselves (Frischmann et al., 2014). That is, the characteristics of the resource are endogenous to the interaction. Whether or not knowledge is available to a larger community of producers, or if there are barriers to entry of the community of producers are determined in part by the decisions of firms and individuals who are in a position to decide on the legal status of knowledge owned by them. By implication, then, existing allocation of ownership rights on knowledge assets may have enduring

328  The Elgar companion to information economics effects that manifest themselves on the investment decisions of some other individuals, i.e., firms and individuals who are eager to invest in creating new knowledge in the future. What is more, beware that the investment decisions of individuals (who are sensitive to such legal arrangements) could also affect the choice made for the openness or secrecy (hence, the design of the IPR regime). That is, in the information economy, the institutional configuration is determined by the choices of agents with different roles, i.e., the choice of strong or weak IPR may also respond to the incentives and motivations of individuals as well as firms. This is why we have to take into account the incentive effects of IPR, when analysing knowledge creation as a coordination problem. Because, to repeat, the prospects and challenges brought by information economy can only be analysed as a joint determination by the IPR and incentives for investment in knowledge assets. An important implication of this framework is that, when contracts are incomplete, underinvestment in knowledge assets, and thereby, the inefficient allocation of knowledge assets are, in fact, possible. The focal point of contractual incompleteness in the information economy is the ownership and control of knowledge assets. Let us recall that knowledge production is sequential and cumulative. If some actors cannot own the knowledge assets which they invest in, they will obviously be reluctant to invest in these assets. However, if knowledge assets (including the skills and expertise of individuals and knowledge workers) are the most important input in the information economy, efficiency requires that these rights should accrue to the agents who value them the most (Demsetz, 1967; Pagano and Rossi, 2004; Pagano, 2007; Williamson, 1985). The task is challenging, since there are interdependencies as well as conflict of interest among agents over dividing the gains from knowledge creation in the information economy. This is why initial distribution of rights regarding knowledge assets could encourage or inhibit the accumulation of knowledge for firms, and even intellectual skills for knowledge workers who are initially deprived of holding such assets. However, there is more to it. When incentives solely target pecuniary gains and profit-making, e.g., with a strong IPR regime, there is a possibility of crowding out of intrinsic motivations such as sharing and reciprocity, which are believed to be the cornerstone of knowledge sharing communities (Bowles, 2016; Cohen and Sauerman, 2007; Ryan and Deci, 2000). Such adverse behavioural adaptation to pecuniary incentives may be an unintended consequence of too much emphasis on secrecy, or IPR in general (Benkler, 2002a; Ramello, 2005). Consequently, pecuniary rewards may undermine cooperative behaviour and investment in knowledge assets. This possibility is not a mere theoretical speculation. It is well known that non-pecuniary incentives substantially affect the innovation performance of workers in firms (Cohen and Sauerman, 2007; Stern, 2004). For example, sharing (and weak IPR) could motivate science oriented workers. These workers usually accept lower wages in return for permission to publish at least partially their findings. In a famous article, Stern (2004) finds a negative relation between wages and doing science. In essence, scientists are willing to give up their claim to a higher level of income for the sake of this opportunity. Policy makers should take into account the diversity of motivations individuals have. Otherwise, for example, institutions of open science could be marginalized by insisting solely on pecuniary motivations. Some degree of insulation for science and the isolation of science oriented individuals may even be necessary to prevent such an adverse outcome (Dasgupta and David, 1994; Stern, 2004). Another example is free and open source software projects. Computer programmers usually forgo income to participate in open software projects. This is a choice that also reflects the role of intrinsic motivations (such as sharing, cooperation) over career concerns and pecuniary gains

Incomplete contracts, intellectual property rights, and incentives  329 (Lerner and Tirole, 2004; Stern, 2004). For software engineers, for example, firms devise contracts that include the option to devote part of their time to such projects (Gambardella et al., 2009). It should be noted that one of the motivations behind the GNU/Linux operating system project was to have a free alternative to software produced by large corporations (Stallman, 2002). The message is clear; diversity of motivations at the individual level should not be overlooked (Cohen and Sauerman, 2007).22 Firms need to find the right balance between openness and secrecy. Too much IPR may lead to a vicious circle in which knowledge workers do not invest in knowledge assets. The opposite extreme of having no IPR and exclusive control rights on knowledge, is not desirable by firms either. The increase in worker mobility, and the rise of markets for knowledge are among the factors that significantly influence the firms’ choices. A recent example of strong IPR is how firms strategically deploy non-compete agreements to this end. The employees who sign non-compete covenants agree not to work for a competitive firm (or in the same industry) for a fixed (usually one or two years) period of time after job termination (Fisk, 2014; Marx, 2015). Firms argue that non-compete agreements prevent the leakage of trade secrets23 by enabling the firms to restrain their former employees from using elsewhere the knowledge they acquired in their firm. Even though knowledge assets that include intellectual skills and capabilities are impossible to separate from workers (Becker, 1964), workers can be constrained in using their knowledge assets by such legal arrangements (Marx and Fleming, 2012). Indeed, the same outcome could be achieved by non-disclosure agreements, which are less restrictive (for workers) than non-compete agreements. However, non-disclosure agreements are difficult to enforce, since the transaction costs of knowing whether an ex-employee is abiding by the agreement is very high. Thus, non-competes are a (transaction) cost effective tool to protect the involuntary dissemination of trade secrets for firms (Marx and Fleming, 2012). Notwithstanding that, contrary to several other IPR regulations that protect the final output, non-competes restrict the usage of human input: workers are deprived of using their skills and expertise, i.e., knowledge assets (Marx and Fleming, 2012). Ironically, it is these skills and expertise of workers that are usually responsible for creating knowledge intensive output protected by the trade secret. Under such an institutional configuration, workers will not invest in skills they could not use in subsequent employment in a different firm within the same industry (Marx and Fleming, 2012). This is why some argue that in Silicon Valley, for example, non-compete agreements are not enforced by the state (Saxenian, 1994; Hyde, 2003). In this regard, it has been pointed out that the Silicon Valley example shows how knowledge creation is a collective phenomenon. What is more, research on Silicon Valley is a clear example of how interfirm and intrafirm concerns over knowledge creation are strictly intertwined. Collaboration among firms has traditionally been achieved through labour mobility, which is one of the most effective methods of knowledge dissemination. Various forms of weak IPR (e.g., not enforcing non-compete agreements) have been the key to investment in intellectual skills for workers. These very same legal arrangements, by allowing higher labour mobility contributed to interfirm knowledge sharing and collaboration. Recall that firms’ knowledge assets are also embodied in worker skills. Gilson (1999) provides an excellent account of this phenomenon, i.e., the dynamics of collective knowledge creation at the firm and worker level (see also Hyde, 2003).

330  The Elgar companion to information economics

CONCLUSION In the information economy, we have witnessed the proliferation of various institutional configurations ranging from science driven start-ups to direct individual collaboration. Increasing diversity necessitates to reconsider the variety found in incentives for firms, knowledge workers, scientists and individual enthusiasts. Indeed, user-driven innovation and direct individual collaboration are so significant phenomena that researchers invent concepts such as commons based production, or innovation commons to capture transformation of production organization (Baldwin and von Hippel, 2011; Benkler, 2006; Potts, 2019). In this regard, researchers who conceptualize knowledge as a commons (Hess and Ostrom, 2011; Frischman et al., 2014) call for a broader perspective when analysing information and knowledge as an economic resource. There is one common feature of all the diversity found in the information economy. Knowledge assets of individuals are regarded as the most valuable (and scarce) resource. However, the benefits of this human asset (which is embodied in intellectual skills) could only be utilized when individuals have the rights and incentives to invest in these skills. Hence, the institutional environment has a lasting effect on the strength and scope of individual incentives. In the past, pecuniary incentives (e.g., high wages, permanent employment) were almost the only strategy in the toolbox of large corporations, partly because of the relative unimportance of such skills in the production process (Fisk, 2014; Stone, 2013). Today, the reverse is the case; human assets matter. This is why there is a renewed interest in grasping the content of individual motivations and incentives. The outcome is not surprising: individual motivations are psychologically rich, and certainly include non-pecuniary incentives (Bowles, 2016; Cohen and Sauermann, 2007). These issues could easily be overlooked when two interrelated but distinct problems (creation and appropriation) regarding knowledge are conflated. In the latter, free-riders lead to underinvestment in knowledge assets. Policy implications immediately follow: either create a market (and IPR regime) or provide public subsidies. Notwithstanding the merits of this approach, it is also true that open innovation, collaboration and sharing in both science and technology are widespread, i.e., free-riders do not divert individuals from contributing to knowledge commons (Allen, 1983; Nuvolari, 2004; David, 1993; Potts, 2019). Indeed, cooperation and social norms of sharing are not only relevant in the context of knowledge. These behavioural adaptations are as old as the history of mankind (Bowles and Gintis, 2011; Richerson and Boyd, 2005; Fehr and Gachter, 2000). In essence, the self-organizing capacity of individuals should not be overlooked. A more comprehensive perspective requires two things: first, the acknowledgement of the fact that information (as goods) is only a subset of knowledge assets. Human intellectual skills, talent and capabilities are all examples of a broader set. They are extremely difficult to specify in a contract, and as a consequence, we are faced with contractual incompleteness. The IPR regime and market exchange do not automatically lead to an efficient outcome. However, firms as well as other institutional configurations could alleviate the dilemma (in knowledge creation) by a complementary mechanism: distribute the ownership rights in such a way that individuals have enough incentives to invest in knowledge assets. The proliferation of institutional configurations signals how alternative legal and managerial tools are deployed to this end. Whereas some prioritize exclusive property rights, sometimes, at the expense of incentives for workers and firms, other institutional configurations do not overlook the importance

Incomplete contracts, intellectual property rights, and incentives  331 of employee motivations. This takes us to the second ingredient of the broader perspective: there is no homogeneity in human motivations, and this heterogeneity should seriously be taken into account. A policy design based on a simplified vision of human motivations and incentives may crowd-out cooperative motivations and their corresponding pivotal institutions. Individuals do not only respond to pecuniary motivations. Sharing, cooperation, and reciprocity have been central ingredients of regulation interactions among individuals in user communities, as well as science, business firms, and even market exchange (Bowles, 2016; Ostrom, 1990). Overall, there are several features of knowledge that make it intrinsically different from natural resources. In natural resources, scarcity (and depletion due to overuse) is the essence of social dilemmas, sometimes dubbed the tragedy of the commons (Ostrom, 1990). For knowledge assets, overuse is not the fundamental problem. The knowledge stock of any society increases with use. Yet, the social dilemma in knowledge assets is twofold: first, knowledge is a non-existing resource; individuals have to overcome this scarcity problem via institutional arrangements such as collaboration, cooperation and sharing. Yet, there is a second dilemma to deal with. This dilemma concerns the efficient production and exchange of a potentially abundant resource, and requires addressing head on the unintended effects of fragmented (intellectual) property on information and knowledge. The essence of the second problem is to appreciate the fact that knowledge production is sequential and cumulative. Ironically, then, underuse (of a potentially abundant resource) is the real dilemma under a fragmented property rights regime imposed upon the many owners (Heller, 1998). This also explains why scholars have always been more sceptical towards the IPR (compared to property rights on tangible goods) since, if not before, the nineteenth century (Boyle, 2003; Machlup and Penrose, 1950; Fisk, 2014). The lesson to be taken from such scepticism is not, however, to abolish any kind of IPR from the toolbox of policy makers. It would also be a grave error to believe that the market for ideas could better be regulated without IPR. Instead, it shows the limits of one-size-fits-all measures. The once generalization of the celebrated solution of property rights and market exchange (Alchian and Demsetz, 1972; Demsetz, 1967), which is believed to work for tangible assets, does not necessarily function equally efficiently for knowledge assets. Not only public authorities but also firms are aware of the fact that voluntary and involuntary dissemination of knowledge are at the heart of knowledge production. If we all stand on the shoulders of giants, as Newton rightly pointed out long ago (Scotchmer, 1991), then the multiple pathways for investment in knowledge assets need to be kept wide open for the benefit of all parties concerned.

ACKNOWLEDGEMENTS We would like to thank the editors and two anonymous reviewers for their valuable comments that helped us significantly improve the chapter. The usual caveats apply.

332  The Elgar companion to information economics

NOTES 1.

2.

Chapters 2 and 3 in this Companion neatly summarize the theoretical underpinnings of contractual incompleteness in market and non-market interactions that stem (mainly) from asymmetric information (Stiglitz and Kosenko, 2024a, pp. 20–52, 2024b, pp. 53–80). Chapter 5 in this Companion discusses the relationship between market failures and asymmetric information (Giza, 2024, pp. 106–117). A more general illustration of the game is as follows (a, b, d, and e represent payoffs, and c is the cost of investment):

     

invest

don’t invest

invest

​a-c, a-c​

​b-c, b​

don’t invest

​d, d-c​

​e, e​

In this formulation, PD game obtains when d​ > a-c​, ​e> b-c​, and ​a-c> e​. The last condition guarantees that {invest, invest} is the best alternative, i.e., Pareto efficient outcome. For further analytical details of this model and relevant assumptions, see Aydogmus and Gürpinar (2022), Kealey and Ricketts (2014). Dasgupta and David (1994, pp. 502–503) is one of the first who discuss the intuition behind this kind of formulation regarding scientific information. For a coordination game analysis of the interaction see next section, and note 14. 3. International treaties such as the Paris Convention (1883) on patents, and Berne Convention (1886) for copyright related works are more recent phenomena concerning an IPR regime of transnational scope (Scotchmer, 2004). 4. Another way to classify incentive mechanisms is to look at whether they are ex-ante or ex-post solutions. Whereas grants are an example of the former, prizes and IPR exemplify the latter (Scotchmer, 2004). 5. Chapter 1 of this Companion elaborates on how inefficiencies in information processing arise in markets. It shows that regardless of conceptualizing information as a scarce or abundant resource, both approaches have to deal with inefficiencies in information markets (Raban and Włodarczyk, 2024, pp. 2–19). 6. The breadth of the patent determines how different a technique should be in order to avoid infringement (Scotchmer, 2004). 7. Patent pools is yet another solution to overcome high transaction costs when ideas are complementary (Lerner and Tirole, 2004). The danger in patent pools is that they may also be formed when ideas are substitutes. Such a collusion among agents in the market usually raises antitrust concerns (Aydogmus, 2022; Ramello, 2005; Scotchmer, 2004). This is part of a more general problem, in which patents (and IPR) are used as legal weapons to avoid competition in the market (Moser, 2012; Scotchmer, 2004). Put simply, large monopolies of patent holders could reduce the incentives for information production, since newcomers are under constant threat of legal action (i.e., infringement) and licensing demands (Boldrin and Levine, 2012). 8. Note that efficiency is central to the public goods approach. It is part of a broader framework regarding (private) property and efficient allocation of resources (Alchian and Demsetz, 1972; Demsetz, 1967). Notwithstanding that, concerns over efficiency could be separated from the right to own ideas (see, for example, Merges, 2012). Hence, efficiency considerations and welfare implications cannot be the founding principle of an IPR regime. For a detailed discussion, see Elkin-Koren and Salzberger (2012), and Merges (2012). 9. These legal changes include the Bayh-Dole Act (enabling patenting basic research carried out in universities), and Stevenson-Wydler Act (authorizing licensing for universities) in 1980 (Scotchmer, 2004).

Incomplete contracts, intellectual property rights, and incentives  333 10. Chapter 15 of this Companion discusses in depth the meaning of and relationship between the concepts of intangible assets and information goods (Bochańczyk-Kupka, 2024, pp. 301–314). 11. A caveat is needed: Some researchers argue that information about entrepreneurial opportunity could still be rivalrous. When an agent exploits the entrepreneurial opportunity, there will be less (pecuniary) benefits for the others to reap (Potts, 2019). Gans and Stern (2010) share a similar viewpoint: users’ willingness to pay for an idea will decline as it gets more widespread. In essence, ideas may be non-rival, but there may be rivalry regarding the value obtained from the application of an idea. 12. Chapter 14 of this Companion reviews alternative approaches to innovation, and discusses how they differ in conceptualizing individuals (and their preferences) and institutions and institutional change (Elsner, 2024, pp. 270–300). 13. As more recent literature puts it, any particular piece of knowledge does not have much value in isolation. Individuals are highly dependent on one another, i.e., techniques or ideas are usually complements. The real value of any knowledge asset is realized only when it is matched with complementary assets (Teece, 1986, 1998; Gans and Stern, 2010). 14. Following the more general representation in note 2, a coordination game obtains when ​a-c>d​ ,​ e> b-c​, and ​a-c> e​. An evolutionary game theoretic extension of both the PD game and coordination game are relevant to study population level outcomes and stability of equilibrium/equilibria (Aydogmus and Gürpinar, 2022). 15. The need to pool dispersed knowledge may even precede entrepreneurial activity, which is believed to be at the heart of innovation (Schumpeter, 1934 [1912]; Potts, 2019). In other words, the starting point of any innovative activity is not entrepreneurial creativity, but rather the creation of a pool; and only in later stages, do the entrepreneurs use this pool to discover opportunities (Potts, 2019). 16. Chapter 1 of this Companion elaborates further on how both scarcity and abundance conceptualization of information (and knowledge) produce inefficiencies (Raban and Włodarczyk, 2024, pp. 2–19). 17. Indeed, similar concerns apply to the country level analysis as seen in the TRIPS negotiations. While developed countries push for strong IPR regime, developing and under-developed countries accuse them of kicking away the ladder, and point out the adverse effects of such policies on investment in knowledge assets for less-developed countries (Chang, 2001, 2002). Thereby, the game theoretic model developed in this section can well be used for a country level analysis of the implications of strong IPR regime. 18. A more comprehensive analysis of the game theoretic model introduced in this section could be found in Gürpinar (2016a, 2016b) and Landini (2012, 2013). Moreover, a general framework regarding the interaction between contracts/rights and preferences/behaviour of agents can be found in Bowles (2004, pp. 261–264). 19. In this formalization, {withhold, invest} is more preferable to {share, don’t invest}, since 1>0 for both players. This could be justified by the fact that withhold, and invest are more preferrable strategies by the row and column player, respectively. Yet, a disagreement game could still be obtained under an alternative payoff structure, e.g., both off-diagonal entries are (0, 0). 20. The existence of institutional complementarities in production organization is studied in Milgrom and Roberts (1990), Aoki (2001, pp. 225–229), and Pagano and Rowthorn (1994). The same phenomenon is also known as strategic complementarity (Bowles, 2004, pp. 158–160). 21. The model (and framework) introduced in this section could be extended in several directions. One such direction is to have a sequential game, in which the agent who decides on IPR chooses first. Such a first mover advantage could lead to the selection of the Nash equilibrium that favours her. Another possible extension is to study the population level dynamics of the interaction, i.e., an evolutionary game theoretic extension is possible (Aydogmus and Gürpinar, 2022). 22. Chapter 13 of this Companion takes up the dynamics of digital innovation (Bauer and Prado, 2024, pp. 246–269). 23. Unlike patents, trade secrets are less restrictive (for firms), since in order to have legal protection, novelty is not required. There is even no need for relating trade secrets to any particular technology (Scotchmer, 2004). The advances in scientific method and codification lower the effectiveness of secrecy (Moser, 2012).

334  The Elgar companion to information economics

REFERENCES Aghion, P. & Tirole, J. (1994). The management of innovation. The Quarterly Journal of Economics, 109(4), 1185–1209. Alchian, A. & Demsetz, H. (1972). Production, information costs, and economic organization. American Economic Review, 62(5), 777–795. Allen, R. C. (1983). Collective invention. Journal of Economic Behavior & Organization, 1, 1–10. Aoki, M. (2001). Toward a Comparative Institutional Analysis. Cambridge, MA: MIT Press. Arora, A. (1995). Licensing tacit knowledge: Intellectual property rights and the market for know-how. Economics of Innovation and New Technology, 4(1), 41–60. Arora, A., Belenzon, S., & Patacconi, A. (2018). The decline of science in corporate R&D. Strategic Management Journal, 39, 3–32. Arora, A. & Gambardella, A. (2010). Ideas for rent: An overview of markets for technology. Industrial and Corporate Change, 19(3), 775–803. Arrow, K. J. (1962). Economic welfare and the allocation of resources for invention. In National Bureau of Economic Research, The Rate and Direction of Inventive Activity: Economic and Social Factors (pp. 609–626). Princeton, NJ: Princeton University Press. Aydogmus, O. (2022). Increasing returns and path dependence in knowledge creation and their effects on patent pools. Structural Change and Economic Dynamics, 62, 462–477. Aydogmus, O. & Gürpinar, E. (2022). Science, technology and institutional change in knowledge production. Dynamic Games and Applications, 12(4), 1163–1188. Baldwin, C. & von Hippel, E. (2011). Modeling a paradigm shift: From producer innovation to user and open collaborative innovation. Organization Science, 22(6), 1399–1417. Bauer, J. M. & Prado, T. S. (2024). Digital innovation: An information-economic perspective. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 246–269). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Becker, G. (1964). Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education. Chicago: University of Chicago Press. Benkler, Y. (2002a). Coase’s penguin, or, Linux and the nature of the firm. Yale Law Journal, 112(3), 339–446. Benkler, Y. (2002b). Intellectual property and the organization of information production. International Review of Law and Economics, 22, 81–107. Benkler, Y. (2006). The Wealth of Networks. New Haven: Yale University Press. Besen, S. M. & Raskind, L. J. (1991). An introduction to the law and economics of intellectual property. Journal of Economic Perspectives, 5(1), 3–27. Bessen, J. & Maskin, E. (2009). Sequential innovation, patents, and imitation. The RAND Journal of Economics, 40(4), 611–635. Bochańczyk-Kupka, D. (2024). Intangibles, information goods, and intellectual property goods in modern economics. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 301–314). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Boldrin, M. & Levine, D. (2008). Against Intellectual Monopoly. Cambridge: Cambridge University Press Boldrin, M. & Levine, D. (2012). The Case against Patents. Federal Reserve Bank of St. Louis WP Series, 1–24. Boldrin, M. & Levine, D. (2021). Reforming patent law: The case of Covid-19. Cato Journal, 41(3), 773–784. Bollier, D. (2011). The growth of the commons paradigm. In C. Hess & E. Ostrom (Eds.), Understanding Knowledge as a Commons (pp. 27–40). Cambridge, MA: MIT Press. Bowles, S. (2004). Microeconomics. New York: Russell Sage Foundation. Bowles, S. (2016). The Moral Economy. New Haven: Yale University Press. Bowles, S. & Gintis, H. (2011). A Cooperative Species. Princeton, NJ: Princeton University Press. Bowles, S. & Halliday, S. (2022). Microeconomics. Oxford: Oxford University Press. Boyle, J. (2003). The second enclosure movement and the construction of the public domain. Law and Contemporary Problems, 66(1–2), 33–75. Bush, V. (1945). Science: The Endless Frontier. Washington, DC: US Government Printing Office.

Incomplete contracts, intellectual property rights, and incentives  335 Chang, H.-J. (2001). Intellectual property rights and economic development: Historical lessons and emerging issues. Journal of Human Development and Capabilities, 2(2), 287–309. Chang, H.-J. (2002). Kicking Away the Ladder. London: Anthem. Coase, R. (1960). The problem of social cost. Journal of Law and Economics, 3, 1–44. Cohen, W. M. & Sauermann, H. (2007). Schumpeter’s prophecy and individual incentives as a driver of innovation. In F. Malerba & S. Brusoni (Eds.), Perspectives on Innovation (pp. 73–104). Cambridge: Cambridge University Press Dasgupta, P. & David, P. (1994). Toward a new economics of science. Research Policy, 23(5), 487–521. David, P. A. (1993). Intellectual property institutions and the panda’s thumb: Patents, copyrights, and trade secrets in economic theory and history. In M. Wallerstein, M. E. Mogee, & R. Schoen (Eds.), Global Dimensions of Intellectual Property Rights in Science and Technology (pp. 19–61). Washington, DC: National Academy Press. Demsetz, H. (1967). Toward a theory of property rights. American Economic Review, 57(2), 347–359. Elkin-Koren, N. & Salzberger, E. (2004). Law, Economics and Cyberspace: The Effects of Cyberspace on the Economic Analysis of Law. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Elkin-Koren, N. & Salzberger, E. (2012). The Law and Economics of Intellectual Property in the Digital Age. London and New York: Routledge. Elsner, W. (2024). Innovation and information: Smooth and ongoing change, or turbulence and cognitive over-stress? On the complex deep structure of innovation. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 270–300). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Fehr, E. & Gachter, S. (2000). Fairness and retaliation: The economics of reciprocity. Journal of Economic Perspectives, 14(3), 159–182. Fisk, C. L. (2014). Working Knowledge. Chapel Hill: University of North Carolina Press. Fleming, L., Greene, H., Li, G., Marx, M., & Yao, D. (2019). Government-funded research increasingly fuels innovation. Science, 364(6446), 1139–1141. Frischmann, B., Madison, M., & Strandburg, K. (Eds.) (2014). Governing Knowledge Commons. New York: Oxford University Press. Gallini, N. & Scotchmer, S. (2007). Private orderings and intellectual property: What is the best incentive system? In F. Cafaggi, A. Nicita, & U. Pagano (Eds.), Legal Orderings and Economic Institutions (pp. 133–153). London and New York: Routledge. Gambardella, A., Giarratana, M. S., & Panico, C. (2009). How and when should companies retain their human capital? Contracts, incentives and human resources implications. Industrial and Corporate Change, 19(1), 1–24. Gans, J. S. & Stern, S. (2010). Is there a market for ideas? Industrial and Corporate Change, 19(3), 805–837. Gilson, R. J. (1999). The legal infrastructure of high technology industrial districts: Silicon Valley, Route 128, and covenants not to compete. New York University Law Review, 74, 575–629. Giza, W. (2024). Asymmetric information as a market failure in retrospect. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 106–117). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Grossman, S. & Hart, O. (1986). The costs and the benefits of ownership: A theory of vertical and lateral integration. Journal of Political Economy, 94(4), 691–719. Gürpinar, E., (2016a). Organizational forms in the knowledge economy: A comparative institutional analysis. Journal of Evolutionary Economics, 26(3), 501–518. Gürpinar, E. (2016b). Institutional complementarities, intellectual property rights and technology in the knowledge economy. Journal of Institutional Economics, 12(3), 565–578. Haber, S. & Lamoreaux, N. (2021). The Battle over Patents: History and Politics of Innovation. NBER Working Paper 28774, 1–38. Hardin, G. (1968). The tragedy of the commons. Science, 162, 1243–1248. Hart, O. (1995). Firms, Contracts and Financial Structure. Oxford: Oxford University Press. Hart, O. & Moore, J. (1990). Property rights and the nature of the firm. The Journal of Political Economy, 98(6), 1119–1158. Hayek, F. (1937). Economics and knowledge. Economica, 4(13), 33–54.

336  The Elgar companion to information economics Hayek, F. (1945). The use of knowledge in society. American Economic Review, 35(4), 519–530. Heller, M. A. (1998). The tragedy of the anti-commons: Property in the transition from Marx to markets. Harvard Law Review, 111(3), 622–688. Heller, M. A. & Eisenberg, R. S. (1998). Can patents deter innovation? The anticommons in biomedical research. Science, 280(5364), 698–701. Hess, C. & Ostrom, E. (Eds.) (2011). Understanding Knowledge as a Commons. Cambridge, MA: MIT Press. Hyde, A. (2003). Working in Silicon Valley. Armonk, NY: M. E. Sharpe. Jaffe, A. & Lerner, J. (2004). Innovation and Its Discontents. Princeton, NJ: Princeton University Press. Kealey, T. & Ricketts, M. (2014). Modelling science as a contribution good. Research Policy, 43, 1014–1024. Kealey, T. & Ricketts, M. (2022). The contribution good as the foundation of the industrial revolution. In E. Dekker & P. Kuchar (Eds.), Governing Markets as Knowledge Commons (pp. 19–57). Cambridge: Cambridge University Press. Landini, F. (2012). Technology, property rights and organizational diversity in the software industry. Structural Change and Economic Dynamics, 23(2), 137–150. Landini, F. (2013). Institutional change and information production. Journal of Institutional Economics, 9(3), 257–284. Lerner, J. & Tirole, J. (2004). Efficient patent pools. American Economic Review, 94(3), 691–711. Luce, R. D. & Raiffa, H. (1957). Games and Decisions: Introduction and Critical Survey. New York: Wiley. Machlup, F. (1984). The Economics of Information and Human Capital. Princeton, NJ: Princeton University Press. Machlup, F. & Penrose, E. (1950). The patent controversy in the nineteenth century. Journal of Economic History, 10(1), 1–29. Marshall, A. (1890). Principles of Economics. London: Macmillan. Marx, M. (2015). The firm strikes back: Non-compete agreements and the mobility of technical professionals. American Sociological Review, 76(5), 695–712. Marx, M. & Fleming, L. (2012). Non-compete agreements: Barriers to entry … and exit? Innovation Policy and the Economy, 12, 39–64. Maurer, S. & Scotchmer, S. (2006). Open source software: The new intellectual property paradigm. NBER Working Paper 12148, 1–36. Merges, R. P. (2000). One hundred years of solicitude: Intellectual property law, 1900–2000. California Law Review, 88, 2187–2240. Merges, R. P. (2012). The relationship between foundations and principles in IP Law. San Diego Law Review, 49, 957–978. Milgrom, P. & Roberts, J. (1990). The economics of modern manufacturing: Technology, strategy and organization. American Economic Review, 80(3), 511–528. Moser, P. (2012). Patent laws and innovation: Evidence from economic history. NBER Working Paper 18631, 1–31. Nelson, R. & Winter, S. (1982). An Evolutionary Theory of Economic Change. Cambridge MA: Belknap Press. Nuvolari, A. (2004). Collective invention during the British industrial revolution: The case of the Cornish pumping engine. Cambridge Journal of Economics, 28(3), 347–363. Olson, M. (1965). The Logic of Collective Action. Cambridge, MA: Harvard University Press. Ostrom, E. (1990). Governing the Commons. New York: Cambridge University Press. Pagano, U. (2007). Legal positions and institutional complementarities. In F. Cafaggi, A. Nicita, & U. Pagano (Eds.), Legal Orderings and Economic Institutions (pp. 54–83). London and New York: Routledge. Pagano, U. & Rossi, M. A. (2004). Incomplete contracts, intellectual property and institutional complementarities. European Journal of Law and Economics, 18(1), 55–76. Pagano, U. & Rowthorn, R. (1994). Ownership, technology and institutional stability. Structural Change and Economic Dynamics, 5(2), 221–242. Polanyi, M. (1958). Personal Knowledge. Chicago: University of Chicago Press. Polanyi, M. (1967). The Tacit Dimension. New York: Doubleday Anchor.

Incomplete contracts, intellectual property rights, and incentives  337 Potts, J. (2019). Innovation Commons: The Origin of Economic Growth. Oxford: Oxford University Press. Raban, D. R. & Włodarczyk, J. (2024). Information economics examined through scarcity and abundance. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 2–19). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Ramello, G. (2005). Intellectual property and the markets of ideas. In J. G. Backhaus (Ed.), The Elgar Companion to Law and Economics (pp. 127–144). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Richerson, P. & Boyd, R. (2005). Not by Genes Alone. Chicago: University of Chicago Press. Rousseau, J. J. (1992 [1755]). Discourse on the Origin of Inequality. Indianapolis: Hackett Publishing. Ryan, R. M. & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55(1), 68–78. Saxenian, A. (1994). Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge, MA: Harvard University Press. Schumpeter, J. A. (1934 [1912]). The Theory of Economic Development. Trans. R. Opie. New York: Oxford University Press. Schweik, C. M. (2011). Free/open source software as a framework for establishing commons in science. In C. Hess & E. Ostrom (Eds.), Understanding Knowledge as a Commons (pp. 277–309). Cambridge, MA: MIT Press. Scotchmer, S. (1991). Standing on the shoulders of giants: Cumulative research and the patent law. Journal of Economic Perspectives, 5(1), 29–41. Scotchmer, S. (2004). Innovation and Incentives. Cambridge, MA: MIT Press. Simon, H. (1979). Rational decision making in business organization. American Economic Review, 69(4), 493–513. Stallman, R. M. (2002). Free Software, Free Society. Boston: GNU Press. Stern, S. (2004). Do scientists pay to be scientists? Management Science, 50(6), 835–853. Stiglitz, J. E. & Kosenko, A. (2024a). Robust theory and fragile practice: Information in a world of disinformation. Part 1: Indirect communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 20–52). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stiglitz, J. E. & Kosenko, A. (2024b). Robust theory and fragile practice: Information in a world of disinformation. Part 2: Direct communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 53–80). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stokes, D. E. (1997). Pasteur’s Quadrant: Basic Science and Technological Innovation. Washington, DC: Brookings Institution Press. Stone, K. (2013). The decline in the standard employment contract: A review of the evidence. In K. Stone & H. Arthurs (Eds.), Rethinking Workplace Regulation: Beyond the Standard Labor Contract of Employment. New York: Russell Sage Foundation. Taylor, F. W. (1911). The Principles of Scientific Management. New York and London: Harper & Brothers. Teece, D. (1977). Technology transfer by multinational firms: The resource cost of transferring technological know-how. Economic Journal, 87(346), 242–261. Teece, D. (1986). Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy. Research Policy, 15(6), 285–305. Teece, D. (1998). Capturing value from knowledge assets: The new economy, markets for know-how, and intangible assets. California Management Review, 40(3), 55–79. Teece, D., Pisano G. & Shueni A. (1997). Dynamic capabilities and strategic management. Strategic Management Journal, 18(7), 509–533. von Hippel, E. (2005). Democratizing Innovation. Cambridge, MA: MIT Press. Williamson, O. E. (1985). The Economic Institutions of Capitalism. New York: Free Press. Ziman, J. (1976). Reliable Knowledge: An Exploration of the Grounds for Belief in Science. Cambridge: Cambridge University Press. Zuboff, S. (1989). In the Age of the Smart Machine. New York: Basic Books.

PART V PAYMENT, VALUE, CROWDFUNDING

17. Payment on information markets Wolfgang G. Stock

INFORMATION GOODS AND THE ARITY OF INFORMATION MARKETS E-commerce offers goods via digital channels and the delivery of those goods mainly by parcel post. In contrast, i-commerce offers information goods – content and software – also via digital channels, but the delivery happens exclusively via digital channels, too. Our topic is i-commerce, which is embedded in the market of digital information. Such information markets are very special leading to peculiar business and payment models on the World Wide Web (WWW) and partly also in the offline world (e.g. the offer of CDs, DVDs, etc.). Once digital information is created, it can easily be redistributed by simply copying it. That is why digital information can never actually be scarce. However, due to high prices and legal protective measures as, for example, copyright, digital information can become artificially scarce. This can best be done in markets where money is used to pay. As the information market moved from largely scarce goods to goods in abundance (Raban & Włodarczyk, 2024), the payment options also changed. In the last decades of the twentieth century, highly priced specialized information offered by content aggregators as, for instance, DIALOG or LexisNexis dominated the market; in the twenty-first century, we find large amounts of free information on the WWW. Some years ago, pricing information goods and services were topics of special interest for librarians and information brokers (Akeroyd, 1991; Braitacher et al., 1994; Broadbent, 1981; Rowley, 1998; Snyder & Davenport, 1997); nowadays they are important for nearly everyone living in the information, knowledge, or “smart” societies of the twenty-first century. Information markets contain a growing number of payment models for digital content (Bichler & Löbbecke, 2000) and for software (Gallaugher & Wang, 2002) – from paying with money, i.e., fixed prices as well as differential pricing models, via paying with attention and personal data, paying with fan loyalty to simply paying nothing, all latter cases being forms of noncash payments. In information markets, digital content and software products and services are offered and demanded. If one sells a piece of information, the customer will get the information, of course; however, the seller still owns it, too, as the buyer only got a copy. One may observe special market mechanisms on information markets in contrast to traditional non-digital goods, namely there are dominant fixed costs for the first data set and low costs for all copies, we see distinct information asymmetries between informed information providers and much less informed customers, there are pronounced direct and indirect network effects leading to the winner-takes-it-all phenomenon, and, finally, there is a tendency towards mutating every digital information into a public good, due to simply made copies. All these regularities are well known in information economics. What characterizes trading an information good? “An information good is everything that is or can be available in digital form, and which is regarded as useful by economic agents” (Linde 339

340  The Elgar companion to information economics & Stock, 2011, p. 24). Information goods are both products and services; main product groups are software and content including goods of software companies, publishing houses (articles, books, and proceedings), search services, digital libraries, social media, and live-streaming services. Information markets may exhibit network effects, namely direct network effects (the more users a network attracts, the more its value – the more a network is worth, the more it attracts new users) and indirect network effects (the greater the network, the more complementary offers – the more complementary products and services, the greater the network’s value). If there is no primary and complementary good, but both goods are of equal importance, we speak of “two-sided indirect network effects” (Linde & Stock, 2011, p. 60). A classic example is the interplay of an operating system (e.g. Android) and an application software (e.g. Office on Android). An operating system without applications is useless, and an application without an operating system cannot run. Markets with two-sided indirect network effects are called “platforms”, they are “two-sided markets” (Rochet & Tirole, 2003, 2006) or – in the general case – n-sided markets and multi-sided platforms (Evans & Schmalensee, 2016; McAfee & Brynjolfsson, 2017). Multi-sided platforms “connect two or more independent user groups, by playing an intermediation or a matchmaking role” (Abdelkafi et al., 2019, p. 553). All search engines, all social media, and all live-streaming services are platforms. Especially on platform markets, information providers do not or not always ask for money. Widely known examples are Google and Meta (formerly Facebook). But are their goods really free? Following Larnier (2014), such platforms are “siren servers”, they delude their customers into thinking that their goods are free; however, in fact they are not free, but take some customers’ data as payment. So they follow the tradition of the ancient sirens luring sailors with their singing (allegedly for free), but the sirens hoped for shipwrecking and subsequently killing the sailors (as payment). If we believe in Homer’s epic, Ulysses was a prominent customer of the sirens, but he refused to pay. In this chapter, we describe theoretically and terminologically the different kinds of payments on information markets including the economically very interesting platform markets. The sellers of digital information make their claims, partly open and ask for money, partly hidden and without even asking and take customers’ data, their attention, or their loyalty as payment. Obviously, there are different market sides concerning their kind of payment. As this is also a broadly known fact, what is new in this chapter? We will comprehensively give an overview on the payment structures on information markets. Will payment with attention, personal data, or fan loyalty be new currencies besides money? On one market side, there are the customers, in computer science or information science often called “users”. On the supply side of the information market, we find professional content and software suppliers, streamers on live-streaming services, micro-celebrities as influencers and social media entrepreneurs, which are called wanghongs especially in China (Craig et al., 2021). While their goods are called “free”, Google had revenues of $257,637 million and a net income of $76,033 million (in 2021), Meta (Facebook) reported on revenues of $117,929 million and a net income of $39,370 million (also in 2021), and a prominent Chinese wanghong named Zhang Dayi made $46 million in one year (in 2016). So, what is the value of information (Raban & Ahituv, 2024)? As the designation “two-” or “n-sided” market is occupied by the players on platforms, we introduce “arity” for the denotation of information markets with different market relations concerning the kind of payment (Stock, 2020). In short, arity is the number of different currencies occurring in an individual purchase process. A nullary market has no payment

Payment on information markets  341 relation, meaning that all products and services are free of charge. A unary information market is typical for direct payment with money. One market side (the seller) provides a product or a service and the other market side (the buyer), if she or he is willing to buy, pays for it with money. Here we find fixed prices as well as price discriminations. On some live-streaming services (e.g. Twitch) audience members pay with bought tokens as gifts or tips for the streamers and indirectly also for the platform. Many information products on unary markets are offered for subscription. A binary market has two exchange relationships. One market side (e.g. Google or Facebook) provides a product or service and the other side pays with personal data (on Facebook) or attention (on Google), but not with money. In turn, the seller provides the users’ data or their attention to advertisers, which are now paying with money. Finally, a ternary market consists of three different relationships. An example is the influencer market on some social media (for instance, on YouTube, TikTok, and Instagram), which is similar to a binary market with the additional payment of the users clearing with their loyalty as fans.

INTERPERSONAL INTERACTIONS AND PAYMENT ON INFORMATION MARKETS Especially in the wanghong economy, including influencer and live-streaming markets, interpersonal interactions play important roles, also when it comes to pay – in this case, to tip or to subscribe, or, additionally, negotiate with advertisers. The audience members’ willingness to pay for the wanghongs depends on the viewers’ relation towards the broadcaster and on their relations to other viewers (Lin, 2021). The more viewers are engaged in the stream or in the service, the more likely they are to donate gifts (Yu et al., 2018). In general, it seems that positive social influence “makes a platform better” (Wang & Guo, 2023). Many human actions are social actions. For Max Weber (1978, p. 4), “action is ‘social’ insofar as its subjective meaning takes account of the behavior of others and is thereby oriented in its course”. Information behavior on wanghong markets is mostly oriented on the behavior of others, be it influencers or streamers or other users. So it is social action. If there are concrete interactions between two or more persons, i.e. the persons have contact and acknowledge that they are connected, we speak of “social interaction”. Basic elements of social interactions include bodily contact, proximity, orientation, gesture, facial expression, eye-movement as well as verbal and non-verbal aspects of speech (Argyle, 1969). In mediated contexts – for instance, a TV show, a movie, or on social media – an audience member does sometimes not only passively consume the content, but he or she builds up a kind of relationship to an actor, streamer, presenter, or celebrity. The “media figure” is not (or only very rarely) aware of the relationship, but of course the spectator is. Horton and Wohl (1956) named such mediated social interactions “parasocial interactions”. The crucial difference between social interactions and parasocial interactions “lies in the lack of effective reciprocity”, establishing an “intimacy at a distance” (Horton & Wohl, 1956, p. 215) as bodily contact is not given as well. In media and communication science, “parasocial relationship” is an established concept to name the relations between media users and media figures (Giles, 2003). Interpersonal relations on live-streaming services are neither social relations (there are no spatial proximity and no bodily contact) nor parasocial relations (as there is reciprocity and temporal proximity), but cyber-social relations (Scheibe et al., 2022; Stock et al., 2022). Cyber-social relations occupy

342  The Elgar companion to information economics a position in between social and parasocial relations, giving live-streaming an exceptional position in the entire landscape of social media (Figure 17.1). Following Shao (2009), there are three user types on social media and also on wanghong markets, namely actors (influencers or streamers with active cyber-social behavior), consumers (the purely passive viewers), and, finally, participants (for instance, on live streaming services, the consumers with active cyber-social behavior). Special groups of actors are micro-celebrities and influencers, both being wanghongs; however, these groups partly overlap. A micro-celebrity is a star on social media or on a specific service (Khamis et al., 2017); influencers are endorser shaping audience attitudes through the use of social media (Freberg et al., 2011). In China, wanghongs are influencers or micro-celebrities acting as social media entrepreneurs (Craig et al., 2021). In this chapter, we use the term wanghong for all micro-celebrities on the Web (i.e., influencers, streamers, bloggers, etc.) who try to make money with their online actions. Wanghong is a short form for wǎng luò hóng rén (网络红人; Chinese for “people who have gone viral on the internet”). Following Han (2021, p. 317) it is used “as the vernacular term for influencers and microcelebrities in China, which are increasingly defined by their acute ability to convert internet viewer traffic to money with diverse economic models in its contemporary context of wanghong economy”. Influencers seek money with the exploitation of their parasocial relations, and broadcasters on live-streaming services do the same with their cyber-social relations.

Note: There are social interactions (prototype: personal conversations), cyber-social interactions (actions using the internet; prototype: social live-streaming), and parasocial interactions (between celebrities and audience members; prototype: fandom of a star).

Figure 17.1

Interpersonal interactions

Payment on information markets  343

FREE INFORMATION GOODS ON NULLARY INFORMATION MARKETS Next, we describe the different payment models (for an overview, see Table 17.1) and provide each time a few paradigmatic examples found in contemporary information economy. Additionally, we present some bibliographic references for further reading and better understanding. All price quotations are as of January, 2022. Some information products and services are free from any payment: no money, no attention, no data, and no fan loyalties are involved. We will describe five versions of such nullary information markets (Table 17.1, column 0), namely trial offers, free basic goods, cross-subsidized goods, universal services, and open access. Of course, producers of free information products have expenses to deal with, but they refinance their costs without payment of the end users and apply internal ways (basic goods, trial offers, cross-subsidized goods) or external sources (universal services, open access). Trial offers and free basic goods are also called freemium payment approaches (Anderson, 2009). Here, a product version (maybe with limited functionality or with a short time of usage) is free for customers; however, “monetization is the nucleus of the freemium experience around which all product features are organized” (Seufert, 2014, p. 7). We found trial offers on content as well as on software markets. There are many information products and services, which are offered for free, but their developers hope to monetize their endeavors with side-products. This may be a free version of the online issue of a newspaper or magazine and the options to subscribe (and paying with money) to the complete digital version of the newspaper or to register (and now paying with personal data) to continuously read the online news. For instance, the Austrian newspaper Standard has a free online version, but users have only limited access concerning the number of free articles. If one wants to see all articles, the user has to register with his or her email address; if no data should be stored the user has to pay €8 every month and then no advertisements will appear and the user’s data are not tracked. If another customer is impressed by Standard’s content, they may subscribe to the complete newspaper (called “e-paper”) for monthly €25.99. If users only read the free versions of content-oriented products, the supplying company has to refinance their expenses with advertising revenues (Linde & Stock, 2011, pp. 413ff.), and now we enter a binary market with advertisers as new players. Many apps are free, but there are options to pay for advanced functionality. The German Food Database (FDDB) is free in its standard version, but there is a premium version with more lists and recipes, no advertising, and the option to connect FDDB with fitness tracking devices (e.g. Fitbit) for yearly €19.90. Often, those paid options are offered as in-app purchases. If an information company offers a variety of products, whereby one product is a basic good and all other products or complements need this basic product, the basic product can be distributed for free, but all others for money (Gallaugher & Wang, 1999). The vendor tries to create a lock-in effect of the users and hopes to generate revenues with complementary products or premium versions (Lehmann & Buxmann, 2009). Oracle’s Java is such a free version under an Open Source license, but is only free for personal or development use; for production use, organizations have to pay. Another example is the product and pricing policy of Adobe. They distributed their Acrobat PDF reader free for everyone, but sell Acrobat Pro to create PDF documents only with subscription fees (Eisenmann et al., 2006). As a result of this strategy,

344  The Elgar companion to information economics Table 17.1

Price models and currencies on information markets

n-ary information

(0)

(1)

(2)

markets

Nullary markets

Unary markets

Binary markets

(3) Ternary markets

Currencies

No

Money

Money, attention / data

Currencies: Money, attention / data, fan loyalty

Prices

Free

Fixed prices

   

– freemium (trial offers,

– fixed single item prices

free basic goods)

– fixed subscription prices attention

 

– cross-subsidization

1st degree price

– personal data

– personal data

– universal service

discrimination

 

– fan loyalty (following /

– open access  

– price negotiations

Prices for advertisers

viewing)

– pay-what-you-want

(money):

– tips, subscriptions

(tipping, social paying,

– bidding

Prices for platforms

   

End user prices: – search arguments /

End user prices: – search arguments / attention

 

 

donations)

– pay per click, pay per

(money):

 

 

– hidden auctions

impression, pay per

– sharing revenues

2nd degree price

action

Prices for advertisers

discrimination

– quality of ad and

(money)

– versioning

landing page (user

– price negotiations

– windowing

experience)

– bundling / unbundling

– hidden auctions  

 

 

 

 

 

3rd degree price

 

 

 

 

Main players

Information companies,

discrimination Information companies,

Platforms, customers,

Wanghongs, platforms,

customers

customers

advertisers

customers, advertisers

PDF became a world-wide standard for document exchange; indeed there exists even an ISO standard for PDF. A company may distribute an information good for free if it sees benefits in this offer and the company is able to cross-subsidize the free product with other paid products. We found a typical example with the free reference management software Mendeley, which was acquired by the publishing house Elsevier. Elsevier possesses one of the most comprehensive collections of scholarly journals on a global scale; Mendeley is a social networking service specially for researchers and brings user data to Elsevier. Additionally, Elsevier is now able to store all metadata of their journal articles on Mendeley leading to higher visibility for these papers. Similarly, but in this case externally financed are information goods which are produced and distributed as universal services (Stock, 1997). In the United States, universal services have been known as free or very cheap telecommunication services for decades. However, especially in the US, universal services include content if the service is essential to education, public health, or public safety (Cremer et al., 2001). A prominent example is PubMed which is the most comprehensive information service for biomedical and life-science literature. PubMed is located at the US National Library of Medicine, is a public good and free for everyone in the world, and is approved and funded by the US government. As research results may have importance for both other researchers and the public, open access to research articles seems to be essential. Open access means that articles from the sciences, social sciences, engineering, and humanities are freely accessible for everyone in one of three ways: gold, green, and black (Björk, 2017). Concerning the gold way, the authors

Payment on information markets  345 or their institutions pay for the publication; a typical golden-way journal, PLOS ONE (Fein, 2013), works with an article processing charge of $1,749. Gold open access demonstrates a shift from user-oriented payment to author-oriented payment of research literature. When authors publish their own manuscripts or preprints on their individual homepages or on university repositories, we speak of the green way. The black way is the illegal one. In violation of copyright rules (however, accepted by many users but of course not by publishing houses), Sci-Hub bypasses all paywalls and publishes millions of scholarly articles in their original layout for free (Greshake, 2017). Users on nullary markets may be confronted with an overabundance of information. Is “the more information the better” really true? Raban et al. (2019) showed in an experimental setting that paying participants in contrast to participants working on nullary markets accessed fewer information sources but that these sources were more diverse and balanced in their positions.

PAYMENT ON UNARY INFORMATION MARKETS: MONEY On unary information markets, we only find one payment relation between the seller and the customer – the classic currency of money (Table 17.1, column 1). There are both fixed prices and price discriminations. Due to the dominant costs for the production of the first information product and low costs for all copies the marginal costs of the copies are low or even close to zero. However, there is a big problem on unary information markets: As the value and utility of information varies by person and circumstance (Rafaeli & Raban, 2003) and as there is a difference between the price of information and its value perception (Rusho & Raban, 2021) all price calculations in terms of money are difficult. After the income for the first-copy-costs the seller has lots of options for calculating the prices. Following Pigou (1932, pp. 279ff.), Varian (1996) distinguishes between three degrees of price discrimination on information markets. A 1st degree price discrimination means that prices differ from person to person depending on the amount of money a customer is willing to pay. A 2nd degree price discrimination depends on how and how much a customer purchases. Finally, in a 3rd degree price discrimination the seller is able to define different customer groups with different willingness to pay. Figure 17.2 illustrates the payment on a unary information market.

346  The Elgar companion to information economics

Note: The seller (market side 1) provides digital content or software, and the buyer (market side 2) pays with money. Due to the seller’s price policy there are fixed prices and price discriminations (on all three degrees).

Figure 17.2

Payment on a unary information market

 

Fixed Prices There are fixed prices for single purchases and also fixed prices for subscriptions. Offered products and services include single content items, software, mobile applications (apps), digital games, video-on-demand services, in-app purchases, and in-game purchases. We find fixed prices for content from low end (some cents) via middle range (some dollars) to high end (some thousand dollars) for one piece of information. A classic example for pricing single content items is Apple’s iTunes Store offering downloads of music pieces and other digital content (e.g. TV shows or videos) for $0.99 for most of the songs. Especially for small amounts of money companies created micropayment systems (Baddeley, 2004); however, such systems seem to be rather impractical when transacting very small sums as there are always some transaction fees. As an example of a middle-range fixed price, the New York Times sells reprints of front pages of their archived newspapers for $60. There are also high-end prices for content. On MarketResearch.com, a customer has to pay some hundreds or even thousands of dollars for one single market report (“Sub-Saharan Africa ICT Regulations” costs $4,500). This high price protects buyers from competitors who are not willing or not able to pay the price and provides the buyers with competitive advantages due to valuable information. A special problem arises with the prices of some scholarly journals. For instance, Elsevier’s Journal of Chromatography (series A and B) has an annual subscription price of €25,275 for libraries. As academic libraries often have restricted budgets it is not possible to subscribe to all journals which the researchers of their institution actually need. For decades, the problem has been known as the “serials crisis” (McGuigan, 2004) and led researchers to call for a boycott of journals especially from the publishing house Elsevier. The main reason for overpriced journal offers is due to the oligopolistic structure of academic publishers (Larivière et al., 2015). As researchers are both authors and readers of scholarly articles, the science system has to buy back its own results (Linde & Stock, 2011, p. 221) and this sometimes at high prices. As a solution, a nullary market for end users via “open access” to all research articles was proposed.

Payment on information markets  347 Software pricing leads to a labyrinth as we find very different pricing strategies on this market segment (Cusumano, 2007). So we will come back to this topic in later paragraphs. However, prices for software development may be sometimes fixed-price contracts (Gäbert, 2015). A special segment of software is the market for digital games, games for consoles and for PCs. Some games are free (e.g. Dota 2), some are subscription-based with price discrimination (e.g. World of Warcraft), and others work with a fixed price including, for instance, games for the PlayStation (such as Star Wars Jedi: Fallen Order for about €20 or FIFA 22 for about €100). Also and especially in free games, game providers offer virtual items as in-game purchases with a fixed price (Jiao et al., 2021). We can distinguish between virtual items (e.g. “Cluckles the Brave” – a chicken courier – for Dota 2 for about $10, up to “Dark Moon Baby Roshan” for $2,100 on Loot Market) and pay-to-win elements, i.e. paying to advance in the game (Lelonek-Kuleta et al., 2021). Additionally, there are loot boxes, which are random bundles of virtual items for a specific game (Chen et al., 2020); however, in several countries, loot boxes are considered illegal gambling. On stores for mobile applications for smartphones or tablets such as Google Play or Apple App Store one can find more than two million different products (as of January, 2022). While many apps are distributed without asking the end user for money, some apps are to be paid with a fixed price (Gans, 2012). We find payment models with a single payment and those with subscriptions. A one-time payment is, for instance, realized on the market for fitness trackers such as Fitbit and Garmin for the hardware and the basic app (Heidel et al., 2021); a subscription-based model is, for instance, applied by Lingvist, a language learning app with annual fees of about €80. Especially some free apps offer in-app purchases, sometimes presenting directly the price and sometimes later in the purchase process (Shulman & Geng, 2019). In-app sold goods include fee-based upgrades of the app, paid feature unlocks, items for sale, or other paid apps and services. All in all, there are not that many information goods and services with only concrete fixed prices. Many companies in the information economy bank on price discrimination or abstain from money as a currency for end users. 1st Degree Price Discrimination: Customer-Specific Paying 1st degree price discrimination means that there are user-specific prices for the same information good. 1st degree price models include price negotiations, pay-what-you-want (tips, social paying, donations), and auctions. Concerning price negotiations both market sides deal with the price which should be paid by the customer (Blount et al., 1996). Our example comes from the academic world. Clarivate Analytics’ Web of Science (WoS) is one of the most prestigious scientific information services worldwide. Their customers are academic or research libraries. The price of WoS considers the number of students of a university and the institution’s country, but there are negotiations between Clarivate Analytics and the library in question. To establish a certain market power for the negotiations most libraries form consortia for all their acquisitions of digital resources, i.e. information services as WoS and journals as well as book series subscriptions (Turner, 2014). In some countries the consortia work even country-wide and enable access to academic information goods for all their citizens with by means of national licenses (Filipek, 2010).

348  The Elgar companion to information economics On voluntary payment markets (Natter & Kaufmann, 2015) for content and software, price models of pay-what-you-want are realized. In a closer look, we can differentiate between tipping including voluntary subscribing as a form of tipping, social paying, and donations. We now make a big leap from the academic world into the world of social live-streaming services (Scheibe et al., 2016). On some of these services (as, e.g. YouNow or Twitch) audience members pay – if at all – with an amount of money at their own choice or they sometimes subscribe to a broadcaster leading this form of price discrimination to a kind of fan-based tipping economy. Some broadcasters make their live-videos just for fun and play, for instance, on Twitch; for others, it is a position between play and labor (called “playbour” by Törhönen et al., 2019; see also Włodarczyk, 2024, section 5.3); and for the last group it is an occupation. On YouNow, viewers can reward streamers with gifts and bars (which is the currency of YouNow), the gifts being free for the viewer, while the bars have to be paid for (Scheibe et al., 2018). Gamification elements (e.g. levels, coins, or badges) boost customer loyalty to the streamer and the service and encourage users’ tipping behavior (Scheibe, 2018). The better a streamer’s cyber-social relations are (i.e., the more bars fans will present), the better will be their income. We found a very special form of paying on an American website, which is an adult sex-oriented live-streaming service. Here, the streamer may define a goal and set the price in tokens, which is the website’s currency (Hernandez, 2020). Say, a broadcaster wants to see 200 tokens. It is possible (but not very probable) that one viewer pays the 200 tokens alone, but it is also possible that different viewers pay different amount of tokens till the goal is finally reached. We will call the kind of pooling money for one information good by several customers “social paying”. This is similar to crowdfunding but on a smaller scale, not by a crowd but by a small group. On the adult website, 60 percent of the revenues go to the broadcaster and 40 percent to the service provider. Tipping, subscribing, or social paying on live-streaming services bring income for individual streamers as well as for platforms. Donations are income sources for other content and software products; however, the receiving entity has to be a qualified nonprofit organization. Donation-based support is essential for open and peer-dependent projects such as, for instance, open-source software products or the online encyclopedia Wikipedia. Wikipedia is free for many users, but it does not only need the free contributions of the authors, but also some money in order to run this service and to invest in future activities. Donations are the most important sources of Wikipedia’s income. It ensures, Wikipedia writes on its Web page, that “Wikipedia can remain independent of advertising, commercial interests or third-party funding”. Donations seem to be an aspect of a “reciprocity mechanism” in fundraising (Kocielnik et al., 2018), which means that Wikipedia entries with a high utility value for a user attract high rates of donations. Additionally, via Wikimedia Enterprise, Wikipedia runs a paid service for high volume commercial use of its content for Big Tech companies such as Google, Meta, or Apple. At auctions, one market side pays in dependence of their bids. There are public auctions (English auctions: increasing the offers until there is only one bidder left, or Dutch auctions: a starting price will be lowered until one bidder accepts it) and hidden auctions, i.e. single-shot first price auctions: the highest bidder wins, or Vickrey auctions (Vickrey, 1961, pp. 22–23): the highest bidder wins but pays the price of the second highest bid (Linde & Stock, 2011, p. 386). On binary information markets (see below!), we only found hidden auctions. The advertiser’s payment on Facebook will be calculated in a single-shot first price auction (Boyd

Payment on information markets  349 & Sanchez, 2018); similarly there is real-time advertising, where the highest bid wins the ad impression on a Web page, which an identified individual customer has opened (Stange & Funk, 2014). Google, on its search engine, applies a Vickrey auction (Lucking-Reiley, 2000) as an aspect of the price calculation an advertiser has to pay (Jahan et al., 2015). We will discuss those auction-based price models in more detail in the paragraphs on binary information markets. 2nd Degree Price Discrimination: Versioning, Windowing, (De-)bundling Following Linde (2009), 2nd degree price discrimination covers the price models of versioning, windowing, and bundling as well as unbundling. In versioning, the selling company offers an information product in different versions and leaves it to the customers to select his or her personally best variant leading to performance-oriented price discrimination (Bhargava & Choudhary, 2008). Forms of versioning are, for instance, the versions’ up-to-dateness, the access options, the range of functionalities, the resolution of images, or the speed of processing (Linde & Stock, 2011, p. 394). One will often find three versions of a good, as customers tend to buy the middle-range product and avoid the extremes. On the video-on-demand service Netflix, indeed, three subscription-based versions are offered: basic, standard, and premium leading to different numbers of screens or the functionality of HD or Ultra-HD (Kweon & Kweon, 2021). Adobe’s Acrobat offers only two different versions with different functionality. Besides the free version of the Acrobat reader (with very limited functionality) there are subscription-based Acrobat Standard and Acrobat Pro versions. In the last years, Adobe radically changed their price policy from product sales to subscriptions (Hinterhuber & Liozu, 2020). On Twitch one can find an interesting option of versioning (Figure 17.3). Twitch is a free social live-streaming service which partly specializes in e-sports. Audience members can interact with the streamers and with other users. User participation touches not only the cyber-social interactions but also the payment. Twitch works with subscription-based versions which only differentiate in the price (Wolff & Shen, 2022). With the selected version a user exhibits his or her support for the streamer and the broadcasted content. Wohn et al. (2019) call such a user behavior “digital patronage”. Here, versioning means different versions of support for the streamer. Windowing refers to time-oriented price discrimination. There are different modes of transmission of an information good at different times with also different profit windows for the seller (Owen & Wildman, 1992). Linde (2009) presents formats and times of a movie distribution as an example. All formats are based on one and the same first copy; however, the distribution channels vary over time. Typically, a movie will be distributed primarily in cinemas (time window no. 1), subsequently on pay TV or on video on demand services (time window no. 2), later as DVD or Blu-Ray (time window no. 3), and finally, if at all, as movie on free TV, whereby the time windows may partly overlap. With the specific time-dependent distribution channels and different media, the prices are different in the respective time periods and allow for different income sources for the sellers for the initially same good. Commodity bundling is a product packages-oriented price discrimination (Adams & Yellen, 1976). Pure bundling is the offer of a product package with several components only as a bundle, mixed bundling includes the offer of the complete package and also of its parts, and unbundling means that information goods of a bundle are sold individually.

350  The Elgar companion to information economics

Source: Twitch. Note: There is always the same service – which is basically free for all users – but the versions correlate with different levels of support for the streamer and depend on the extent of the cyber-social relationship between the streamer and the respective user.

Figure 17.3

Three subscription-based versions for tiers 1, 2, and 3 on Twitch

Pure bundling can be found on markets for newspapers and magazines insofar as the articles are not sold individually. If one subscribes to National Geographic in the US for $24 annually for the print and digital version, one has access to the complete bundle of all articles in the magazine. Another example for pure bundling is Amazon Prime. For a monthly €7.99, customers subscribe to Prime video, music, games, premium shipping for e-commerce goods, and one free subscription on Twitch, which is owned by Amazon. Amazon’s subscription price for the bundle is combined with target group-specific pricing, as the customer group of students pays only 50 percent of the standard price. A typical example of mixed bundling is the 365 (formerly Office) package of Microsoft. An annual subscription for Word, Excel, and PowerPoints costs $69.99 for one person. Single software programs are additionally sold by Microsoft with a unique fixed price, e.g. $159.99 for Microsoft Word. Following an unbundling price strategy, some information suppliers untie their bundles to individual products to be sold by pay-per-use (Balasubramanian et al., 2015). The German information aggregator Genios, similar to many scholarly publishing houses, offers subscriptions to complete databases, but also purchases of single articles asking for (relatively high) fixed prices (Figure 17.4).

Payment on information markets  351

Source: Genios Note: Many newspapers and magazines are sold as subscription-based bundles including our two examples Industry & Higher Education and Der Betrieb. However, the content aggregator Genios offers every single paper of such bundles article-by-article.

Figure 17.4

Unbundling information products

3rd Degree Price Discrimination: Target Group-Specific Pricing 3rd degree price discrimination includes personal, spatial, or temporal price discriminations. Personal price discrimination exhibits different prices for different user groups, as, for example, the students’ prices on Amazon Prime. Spatial price discrimination (Adhikari et al., 2022) means different prices for the same information good in different regions, i.e. countries. For instance, Netflix Premium costs €17.99 in Austria, but in Yemen only €11.99 a month. Applying temporal price discrimination, the price for an information good changes over time. There are two typical strategies, namely penetration and skimming (Linde & Stock, 2011, pp. 407ff.). Following a skimming strategy, there are high prices at first and later on lower prices. This strategy is rarely used on information markets. The penetration strategy is much more common; here, the products are sold initially for a low price or even for free in the seller’s hope of network effects which establish the product as a standard in its market. Some years ago, Microsoft’s Bill Gates said, “although 3 million computers get sold each year in China, people don’t pay for our software. Someday they will, though, and as long as they are going to steal it, we want them to steal ours” (Economic Times, 2011). Also piracy may lead to a positive development of a software product towards establishing a standard. And, indeed, Windows became standard in China and nowadays people also in this country pay for Microsoft products. We find a special target group identification on payment-based social question and answer (Q&A) platforms (Zhao et al., 2020). There are different prices for the answers to be paid by the initial asker and later by other users. The asker has to pay the answerer, subsequent users – called “listeners” by Zhao et al. (2020) – pay a lower price to the company which operates the Q&A platform.

PAYMENT ON BINARY MARKETS: ATTENTION OR PERSONAL DATA On binary information markets, we find two different payment methods (Figure 17.5). We have to presuppose that one market side operates a platform. The buyer gets the desired content and pays either with their attention or with their personal data. In turn the platform

352  The Elgar companion to information economics owner works on the attention-based data or the personal data and resells them to advertisers who pay with money. Finally, ads will be presented to the buyer. If the currency is attention, we speak of attention economy, if the currency is personal data, it is personal data economy (Table 17.1, column 2). If a pay-for-privacy model (Elvy, 2017) is realized, a binary information market collapses to a unary market. Here, the buyer does not pay with attention or with personal data, but – as is normal on unary markets – with money. Examples are the already discussed software FDDB or the music streaming service Spotify; one can use it for free (with advertising) on a binary information market and one can subscribe to the service (without advertising and with a greater functionality) on a unary market. A similar model can be identified for many newspapers: the content with ads on a binary market and the pure content with subscription prices on a unary information market.

Note: The seller, running a platform (market side 1) provides digital content, and the buyer (market side 2) pays with attention (e.g. on Google) or with personal data (e.g. on Facebook or Instagram) (payment form 1). On the platform, the customer-specific data become processed and sold to advertisers (market side 3), who pay with money (payment form 2) and present their ads to the buyer.

Figure 17.5

Payment on a binary information market

Attention Economy Since the time of commercial TV, there is an “attention economy”, as the TV companies sell their viewers’ attention to advertising firms in commercial breaks (Davenport & Beck, 2001). On digital information markets, for instance, the search engine company Google sells the users’ search arguments to advertisers leading to context-specific online advertising in hopes of the users’ attention for the displayed ads (Ruhrberg et al., 2017). The money-making company part of Google is their Google Ads (formerly AdSense) service. Typically, a customer places a search argument on Google, and Google searches twofold, on the one hand in its repository of stored Web pages presenting the “objective” results to the user, on the other hand in its repository of keywords (in the Google Ads service) presenting the found ads to the user in a special section of the search engine results pages. For every search, Google creates a customer search profile, which includes the concrete search argument, the

Payment on information markets  353 country and additional location data, data on the customer (if the customer runs an Android system: e.g. data from the browser history), and (if the user is logged-in to Google or uses a cell phone: the user’s search history on Google). The kind of data which the users have to pay in this example of Web search engines is their search argument and their attention when seeing the ads and clicking on them. On the platform of Google Ads (Linde & Stock, 2011, pp. 335–339), the advertiser books a keyword, which may be without any format (e.g. fair), a phrase (e.g. “motor show”), or an exact keyword or phrase (e.g. [motor show]). If the advertiser chooses the phrase, their ad will be presented if the phrase occurs somewhere in the search argument (e.g. looking for “Frankfurt motor show”), while the exact keyword or phrase means the identity between search argument and keyword (in this case, the search argument “Frankfurt motor show“ will not match with the keyword [motor show]). Additionally, the advertiser can select excluding keywords (i.e., keywords, which may not occur in a search argument, e.g. Frankfurt in connection with “motor show” to exclude the Frankfurt motor show). Further information from the advertisers are the ad text (including a title and a link to the landing page), the scope (from local to global), the budget (i.e., the monthly maximum sum of expenses), start and end dates for the ad campaign, the preferred networks (Google search engine homepage, Google affiliate program, YouTube, Gmail, etc.), a targeted language, exclusions (e.g. URLs from competitors), and – most important – the bid, which is the highest price per click. Additionally, the advertiser may adjust the bid (i.e., increase or decrease the bid in percent) depending on the user’s context (e.g. using a smartphone, at a certain time of day, or from a specific location). All those data are stored in an advertiser keyword profile. The budget can be problematic, when competitors click frequently on the ad so that it will be hidden when the budget limit is reached – and no consumer of the target group has seen it. However, Google tries to counteract such click fraud (Soubusta, 2008) by elaborate algorithms. If a customer search profile matches an advertiser keyword profile, which is a common task in information retrieval (Stock & Stock, 2013, pp. 89ff.), the ad will be presented (currently marked with “Ad”) on the search engine results page. All ads for a search argument are ordered by a kind of relevance ranking in a separate list besides the “objective” search results (Stock & Stock, 2013, pp. 356f.). Only if the user clicks on the ad, the advertiser’s payment becomes due. On the Google search engine, the payment scheme of pay per click is realized. Google files for patents to protect inventions concerning the search engine and advertising. Yet, not all details from the patents are actually implemented in practice. There are hundreds of patent applications and granted patents dealing with their ad technology (e.g. Chitilian et al., 2021, p. 9, which describes advertiser campaigns on Google including payment schemes). Google applies not only one form of auction. For Google search (which we describe here) they work with a Vickrey auction (Jahan et al., 2015), for ads concerning videos and games it is a first-price auction. The patent by Monkman et al. (2020) describes in detail the payment process of Google advertisers. The baseline cost is determined in part by a next highest auction score. This is a variant of the Vickrey auction, orienting the price component of every bidder on the price of the lower bidding neighbor. The price calculation algorithm for an ad i follows the formula: NextHighestAS  +   OverallNegative Effect​ (i) ​ ​∑ 1 ​m​ (s) ​* pAction​ (x) ​

_______________________________ ​ BaselineCost​ (i) ​  ​=   ​           ,​​ s

354  The Elgar companion to information economics where NextHighestAS is the bid of the direct competing neighbor, m(s) is a bid modifier (e.g. raising or lowering the bids in the bid adjustments), pAction is the probability p that a user will invoke the action in question (e.g. clicking on an ad of a mobile impression), and, finally, OverallNegative Effect is calculated by the formula x

​∑   ​BadVisitCost​ (x) ​* pBadVisit​ (x) ​* pAction​ (x) ​,​ 1

where BadVisitCost is a monetary value of a bad user experience when they click on an ad, pBadVisit is the probability p that a user will have a bad experience, and pAction is the probability p that the user clicks on the ad i if i is presented (Monkman et al., 2020, p. 16). The BadVisitCost can be specified on a per-query basis or on a global basis (for all queries), for each particular category of queries (e.g. automobiles, art, or sports), or for each particular query. The probability specified by pBadVisit can be determined based on user feedback (e.g. survey information) or inferred based on previous actions of users. A bad user experience includes a failed page load or the user’s dissatisfaction with the landing page content. In short, payment on Google Ads depends (1) on an auction, which is a second price auction following Vickrey and (2) on bad user experience when the user clicks on the ad. What do the players know about their payments? As a typical siren server, Google does not reveal anything on paying to their search engine end users. But also advertisers cannot know

Note: For end users, all price components are unknown. For advertisers, there are known as well as unknown price components. So also advertisers cannot know the exact price per click before the purchase

Figure 17.6

Payment on Google search engine as a part of attention economy

Payment on information markets  355 the final price they have to pay as both the NextHighestAS and the OverallNegative Effect are unknown before the ad will be displayed. There are tools for the estimation of a price an ad will have as the keyword plan on Google Ads, but it is by no means a concrete price calculation. Geradin and Katsifis (2019, 2020) conclude that online display advertising à la Google lacks transparency and is characterized by a high degree of opacity. Figure 17.6 gives an overview of important price components on Google Ads. Personal Data Economy In the “personal data economy” (Elvy, 2017), data on and of the users are sold to advertisers, which is the business model of many social media companies including Meta (formerly Facebook Inc.) called personalized online advertising (Ruhrberg et al., 2017). Analogous to Google, Facebook protects its inventions on online advertising and pricing with patents (Kendall et al., 2018). Social networking services such as Facebook collect detailed person-related data in an amount which is dependent on users’ willingness to provide information in their personal profiles (e.g. user’s gender, actual location, places born and lived, partner, family members, education, hobbies, birthday/age; Tang & Lian, 2018, p. 2) and, additionally, on users’ digital traces, which comprise their posts (texts, images, or videos), their “friends”, viewed posts, the frequency of viewing these posts, their reactions on posts (likes, shares, and comments), and visits to websites of companies which want to track those visiting users. Using these data from personal profiles and from digital traces, Facebook creates a comprehensive description for every user. On the other market side there are Facebook’s advertising customers. For an ad, they have to define their goals (e.g. increasing traffic on their website), their objectives (awareness, consideration, and conversion, including several subcategories), and their audience, which is defined by location, age, gender, languages, interests (e.g. liked Facebook pages), behavior (as, for instance, purchase behavior), and connections (between the concrete user and the advertiser on Facebook) (Curran et al., 2011). As Facebook applies an auction, advertisers have to make their bids. The bidding includes the distribution of the budget (spending lowest or highest costs or even the full budget for one ad), the goal (cost cap: maximum cost per purchase that will keep the purchase profitable for the advertising company) or a manually defined bid cap (maximum bid across auctions). As advertisers define target groups, those groups may overlap. To cite Facebook’s example, one advertiser may target all women who like skiing, but another advertiser targets all skiers in California. Now it is conceivable that one and the same person belongs to both target groups: this would be the case with a female skier living in California. Here the auction starts. If a user description and an ad profile match, the winning ad is displayed to the user. At first sight, the more information is collected from both market sides, i.e., from customers and from advertisers, the better and the more purposeful is the ad. However, there are problematic issues for the advertisers, as end users are often very aware of their privacy so that the over-exploitation of users’ personal data would invade their privacy (Liao et al., 2022). Users also tend to ignore banner ads and reveal banner blindness (Benway & Lane, 1998), and users tend to ignore settings of preferred or declined ads, which Facebook actually offers (Haji & Stock, 2021). The payment method corresponds with the declared objective. Facebook differentiates between pay per click (e.g. traffic objective), pay per impression (e.g. reach objective), and

356  The Elgar companion to information economics pay per action (e.g. conversion objective). The final price depends on the advertiser’s and the competitor’s bids as the highest bid for the same user wins (Hegeman & Yan, 2013, p. 2), the probability of interaction (i.e., the number of users the ad was presented to), and the quality of the banner ad. As the circumstances vary, the price for the same ad is not constant, but changes over time (Ramnik & Runke, 2018, p. 6). The payments of the end users include the personal data the users are willing to share with Facebook and the users’ digital traces on social media and on the WWW in general. As all data are stored within Facebook’s repositories, services used on Facebook are prepaid. Facebook is a siren server, so nearly all price components are unknown to the end user. For advertisers it is complicated: there are lots of options and an advertiser cannot know all its competitors and their bids and therefore cannot know the price to pay before the purchase. The complete price calculation on Facebook is performed by algorithms, so also the company does not exactly know the respective final price before the purchase. The price calculation lacks transparency for both market sides (Alleman, 2018) and, additionally, information asymmetry (Akerlof, 1970; Stiglitz & Kosenko, 2024; Giza, 2024) to the disadvantage of the advertiser, as Facebook has at least knowledge about its concrete algorithms.

PAYMENT ON INFLUENCER MARKETS: FAN LOYALTY On social media, internet entrepreneurs, called wanghongs in China, act as micro-celebrities and monetize the loyalty of their fans (Abidin, 2018). On social media in general, users create and find user-generated content; on influencer markets, users find influencer-generated content (Chen & Chua, 2020, 2024). Marwick (2015, p. 142) calls the user-loyalty on fan-based markets a “social currency”. Celebrities and micro-celebrities have “attention income”, i.e., the attention of their fans to their performances, which is – for Frank (2019, p. 10) – a “kind of capital”. Micro-celebrities are influencers, e.g. on social media or blogs with strong parasocial relationships, streamers on live-streaming services with likewise strong cyber-social relationships, or are both (Fietkiewicz et al., 2018). Streamers work on unary information markets with the main income sources of tips, social paying, and voluntary subscriptions. In this section, we discuss the influencer market (Figure 17.7) as an example of a ternary information market (Table 17.1, column 3). The successful influencer has two and sometimes even more sources of monetary income, namely from the platform (e.g. Google’s YouTube) and from a company working together with the influencer in order to distribute their advertising messages. The influencer produces digital content (for instance, a video on YouTube or TikTok and additionally an image with text on Instagram, Twitter, or Weibo) and hopes to allocate attention of the buyers, i.e., their fans. We have to distinguish between actual fan attention (i.e., the number of views of a single video or an image of the influencer) and expected fan attention (i.e., the number of the influencer’s followers on the service). The influencer’s first revenue stream comes from the platforms, which in turn are financed by their advertisers as in the binary market. The more there is actual fan attention, the more money the influencer will make. If influencers cooperate with YouTube, they should participate in Google’s Ads service and become Google partners, and must have more than 1,000 subscribers. About 68 percent of the video-specific Google Ads revenue goes to the influencers. In China,

Payment on information markets  357

Note: The seller, running a platform (market side 1) provides digital content, and the buyer (market side 2) pays with actual attention (e.g. on YouTube, number of viewers) and with data on followers as expected attention (e.g. on YouTube, number of followers) (payment form 1). Similar to binary information markets, on the platform, the customer-specific data become processed and sold to advertisers (market side 3), who pay with money (payment form 2) and present their ads to the buyer (e.g. a video ad on YouTube). The influencers (market side 4) monetize their actions through their magnitude of expected as well as actual attention, i.e. their amount of parasocial relationships (payment form 3) to both the platform and, additionally, as far as given, to specific advertisers (market side 5) and promote their products as service in return.

Figure 17.7

Payment on a ternary information market

Alibaba runs an own live-streaming platform named Taobao. In contrast to YouTube, Taobao does not pay its broadcasters; however, they will be paid by the tips of their fans (Guan, 2021). The second revenue option for influencers comes from companies which cooperate with them independently from the platform – and that holds true for platforms like YouTube as well as Taobao. In some of their online posts, influencers integrate advertisement for brands, be it in videos (e.g. on YouTube or TikTok), in live-videos (e.g. on Taobao), or in pictures (e.g. on Instagram; Newlands & Fieseler, 2020). The cooperation between an advertising firm and an influencer is usually arranged by a contract; the amount of payment is dependent on the number of the influencer’s followers, i.e., on the fans’ expected attention. Especially in China, we find a third path to income for influencers. Some wanghongs create their own products (e.g. fashion) and operate retailing stores on services like Taobao (Guan, 2021) – however, that is e-commerce and not i-commerce. If we were to sort all influencers of a service by income, we would see a kind of power-law or inverse-logistic distribution (Stock, 2006): one or few influencers making high profits and

358  The Elgar companion to information economics a long tail of other influencers earning much less money (Törhönen et al., 2021). So it is no surprise that micro-celebrities look for as many fans and subscribers as possible (Scheibe, 2018).

CONCLUSION Information markets exhibit a great variety of payment models including payment with noncash currencies as attention, personal data, and fan-loyalty. We applied the concept of “arity” (which is the number of different currencies occurring in an individual purchase process) as a guide to sort the different payment forms (Table 17.1). Concerning a nullary market, there is no currency at all, i.e. all products and services are free. Nullary information markets include freemium goods (trial offers, free basic goods), cross-subsidized goods, universal services, and open access. On unary markets, we find money as the only currency. There are fixed prices for single item purchases and fixed subscription prices, and there exist price discriminations at the 1st degree (price negotiations, auctions, pay-what-you-want: tipping, social paying, donations), at the 2nd degree (versioning, windowing, bundling and unbundling), and at the 3rd degree (personal, spatial, or temporal prices). Unary markets show a clear trend from prices for single items to subscription-based price models for both content and software markets. On binary information markets, we found with attention and personal data new currencies besides money. In the attention economy (e.g. on the search service Google), end users pay with their attention (here, with their search arguments and with clicking on a presented ad) and advertisers have to make a bid and pay per click after a hidden Vickrey auction. In the personal data economy (e.g. on the social networking service Facebook) end users pay with their personal data, which they voluntarily share with Facebook, and with their digital traces on Facebook and on the WWW in general. Advertisers pay, depending on their objectives, per click, per impression, or per action after a hidden first-price auction. Similar to the a priori unknown results of the auctions, advertisers additionally do not know how Google and Facebook measure the concrete quality of the ads and the quality of the landing pages. For advertisers on search engines and on social media services, the presentations of their ads have experience or credence prices (Darby & Karni, 1973). In contrast to the theory of Darby and Karni, which proposes uncertainties of product quality, on binary information markets customers are confronted with uncertainties of product prices. The price which an advertiser has to pay is unknown till the purchase is completed (experience price), while the price calculation of the platform is always unknown (credence price). A ternary information market is a binary market with fan-loyalty as additional currency. End users pay by following a micro-celebrity and by viewing his or her videos, live performances, images, etc. Sometimes, end users additionally pay with tips or subscriptions of the wanghong’s content. Some platforms (e.g. Google’s YouTube) share the revenues from the video-specific advertisers with the micro-celebrities. There are two kinds of advertisers: those who cooperate with the platforms and present their ads there (similar to the binary market), and the others who enter into a contract with the influencer. For the latter, there are price negotiations between the advertiser and the micro-celebrity. There is a rule that applies to all fan-based markets: the more parasocial or cyber-social relationships an influencer or a broadcaster has the higher are their prices.

Payment on information markets  359

REFERENCES Abdelkafi, N., Raasch, C., Roth, A., & Srinivasan, R. (2019). Multi-sided platforms. Electronic Markets, 29, 553–559. Abidin, C. (2018). Internet Celebrity: Understanding Fame Online. Bingley: Emerald. Adams, W. & Yellen, J. (1976). Commodity bundling and the burden of monopoly. The Quarterly Journal of Economics, 90(3), 475–498. Adhikari, A., Sharma, M., Basu, S., & Jha, A. K. (2022). Uniform or spatially differentiated? Pricing strategies under simultaneous and sequential decision-making in multi-market context. Journal of Retailing and Consumer Services, 64, art. no. 102832. Akerlof, G. A. (1970). The market for “lemons”: Quality uncertainty and the market mechanism. The Quarterly Journal of Economics, 84(3), 488–500. Akeroyd, J. (1991). Costing and pricing information: The bottom line. Aslib Proceedings, 43(2/3), 87–92. Alleman, J. (2018). Threat of internet platforms: Facebook, Google, etc. In 29th European Regional Conference of the International Telecommunications Society (ITS) (18 pp.). Calgary: International Telecommunications Society. Anderson, C. (2009). Free: The Future of a Radical Price. New York: Hyperion. Argyle, M. (1969). Social Interaction. New Brunswick, NJ: Aldine Transaction. Baddeley, M. (2004). Using e-cash in the new economy: An economic analysis of micropayment systems. Journal of Electronic Commerce Research, 5(4), 239–253. Balasubramanian, S., Bhattacharya, S., & Krishnan, V. V. (2015). Pricing information goods: A strategic analysis of the selling and pay-per-use mechanism. Marketing Science, 34(2), 218–234. Benway, J. P. & Lane, D. M. (1998). Banner blindness: Web searchers often miss ‘obvious’ links. Internetworking: ITG Newsletter, 1(3), 123–135. Bhargava, H. K. & Choudhary, V. (2008). When is versioning optimal for information goods? Management Science, 54(5), 1029–1035. Bichler, M. & Löbbecke, C. (2000). Pricing strategies and technologies for on-line delivered content. Journal of Organizational and End User Computing, 12(2), 4–10. Björk, B.-C. (2017). Gold, green, and black open access. Learned Publishing, 30(2), 173–175. Blount, S., Thomas-Hunt, M. C., & Neale, M. A. (1996). The price is right – or is it? A reference point model of two-party price negotiations. Organizational Behavior and Human Decision Processes, 68(1), 1–12. Boyd, C. & Sanchez, X. (2018). Breaking down the Facebook auction. Journal of Digital and Social Media Marketing, 6(2), 160–167. Braitacher, P., Mathes-Hofmann, J., & Stock, W. G. (1994). Informationen verkaufen oder Informationen verschenken? Fee or Free? In J. Hering (Ed.), Bibliotheken, Service für die Zukunft: 5. Deutscher Bibliothekskongreß – 83. Deutscher Bibliothekartag in Leipzig 1993 (pp.  71–82). Frankfurt/M.: Klostermann. Broadbent, H. E. (1981). Pricing information products and services. Drexel Library Quarterly, 17(2), 99–107. Chen, N., Elmachtoub, A. N., Hamilton, M., & Lei, X. (2020). Loot box pricing and design. In Proceedings of the 21st ACM Conference on Economics and Computation (pp. 291–292). New York: ACM. Chen, X. & Chua, A. Y. K. (2020). Reviewing the landscape of research on influencer-generated content. In 6th IEEE International Conference on Information Management (pp. 244–248). New York: IEEE. Chen, X. & Chua, A. Y. K. (2024). The role of influencer endorsements in users’ willingness to pay for knowledge products: An empirical investigation. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 379–394). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Chitilian, V. A., Netchitallo, I., Bakshi, N., Yu, J., & Patel, C. (2021). Advertising campaign scripting. Patent no. US 11,170,401 B1 (applicant Google LLC). Washington, DC: US Patent and Trademark Office. Craig, D., Lin, J., & Cunningham, S. (2021). Wanghong as Social Media Entertainment in China. Cham: Palgrave Macmillan.

360  The Elgar companion to information economics Cremer, H., Gasmi, F., Grimaud, A., & Laffont, J. J. (2001). Universal service: An economic perspective. Annals of Public and Cooperative Economics, 72(1), 5–43. Curran, K., Graham, S., & Temple, C. (2011). Advertising on Facebook. International Journal of E-Business Development, 1(1), 26–33. Cusumano, M. A. (2007). The changing labyrinth of software pricing. Communications of the ACM, 50(7), 19–22. Darby, M. R. & Karni, E. (1973). Free competition and the optimal amount of fraud. Journal of Law and Economics, 16(1), 67–88. Davenport, T. H. & Beck, J. C. (2001). The Attention Economy: Understanding the New Currency of Business. Boston, MA: Harvard Business School Press. Economic Times (2011). Between the lines: In support of piracy. The Economic Times, April 24. Eisenmann, T., Parker, G., & Van Alstyne, M. W. (2006). Strategies for two-sided markets. Harvard Business Review, 84, 92–101. Elvy, S.-A. (2017). Paying for privacy and the personal data economy. Columbia Law Review, 117(6), 1369–1459. Evans, D. S. & Schmalensee, R. (2016). Matchmakers: The New Economy of Multisided Platforms. Boston, MA: Harvard Business Review Press. Fein, C. (2013). Multidimensional journal evaluation of PLOS ONE. Libri, 63(4), 259–271. Fietkiewicz, K. J., Dorsch, I., Scheibe, K., Zimmer, F., & Stock, W. G. (2018). Dreaming of stardom and money: Micro-celebrities and influencers on live streaming services. In G. Meiselwitz (Ed.), Social Computing and Social Media: Technologies and Analytics, Part I (pp. 240–253). Cham: Springer. Filipek, D. (2010). Konsortialverträge zwischen Bibliotheken und Verlagen. Ein erfolgreiches Modell? Hamburg: Kovač. Frank, G. (2019). The economy of attention. Journal of Sociology, 55(1), 8–19. Freberg, K., Graham, K., McGaughey, K., & Freberg, L. A. (2011). Who are the social media influencers? A study of public perceptions of personalities. Public Relations Review, 37(1), 90–92. Gäbert, C. (2015). The fixed-price contract: A challenge for the software development project. In A. Holzinger et al. (Eds.), Software Technologies. ICSOFT 2014 (Communications in Computer and Information Science 555) (pp. 97–112). Cham: Springer. Gallaugher, J. M. & Wang, Y.-M. (1999). Network effects and the impact of free goods: An analysis of the web-server market. International Journal of Electronic Commerce, 3(4), 67–88. Gallaugher, J. M. & Wang, Y.-M. (2002). Understanding network effects in software markets: Evidence from Web server pricing. MIS Quarterly, 26(4), 303–327. Gans, J. S. (2012). Mobile application pricing. Information Economics and Policy, 24(1), 52–59. Geradin, D. & Katsifis, D. (2019). An EU competition law analysis of online display advertising in the programmatic age. European Competition Journal, 15(1), 55–96. Geradin, D. & Katsifis, D. (2020). “Trust me, I’m fair”: Analysing Google’s latest practices in ad tech from the perspective of EU competition law. European Competition Journal, 16(1), 11–54. Giles, D. C. (2003). Parasocial interaction: A review of the literature and a model for future research. Media Psychology, 4, 279–305. Giza, W. (2024). Asymmetric information as a market failure in retrospect. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 106–117). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Greshake, B. (2017). Looking into Pandora’s box: The content of Sci-Hub and its usage. F1000Research, 6, art. no. 541. Guan, Z. (2021). Chinese beauty bloggers: Amateurs, entrepreneurs, and platform labour. Celebrity Studies, 12(2), 326–332. Haji, R. & Stock, W. G. (2021). User settings for advertising optimization on Facebook: Active customer participation or settings blindness? Telematics and Informatics, 59, art. no. 101548. Han, X. (2021). Historicising Wanghong economy: Connecting platforms through Wanghong and Wanghong incubators. Celebrity Studies, 12(2), 317–325. Hegeman, J. & Yan, R. (2013). Multi-dimensional advertisement bidding. Patent Application no. US 2013/0124297 A2 (applicant Facebook, Inc.). Washington, DC: US Patent and Trademark Office. Heidel, A., Hagist, C., & Schlereth, C. (2021). Pricing through health apps generated data. Digital dividend as a game changer: Discrete choice experiment. PLoS One, 16(7), art. no. e0254786.

Payment on information markets  361 Hernandez, A. (2020). The many sides of a coin: Exploring the cost and value of tokens in Chaturbate. In AoIR Selected Papers of Internet Research: The 21st Annual Conference of the Association of Internet Researchers (pp. 1–4). Hinterhuber, A. & Liozu, S. M. (2020). Implementing pricing strategies. In A. Hinterhuber & S. M. Liozu (Eds.), Pricing Strategy Implementation: Translating Pricing Strategies into Results. Abingdon: Routledge. Horton, D. & Wohl, R. R. (1956). Mass communication and para-social interaction. Psychiatry: Interpersonal and Biological Processes, 19(3), 215–229. Jahan, F., Fruitwala, P., & Vyas, T. (2015). A study of working of ad auctioning by Google Adwords. Advances in Intelligent Systems and Computing, 409, 471–480. Jiao, Y., Tang, C. S., & Wang, J. (2021). Selling virtual items in free-to-play games: Transparent selling vs. opaque selling. Service Science, 13(2), 53–76. Kendall, T. A., Cohler, M. R., Zuckerberg, M. E., Juan, Y.-F., Jin, R. K.-X., Rosenstein, J. M., Bosworth, A. G., Wong, Y., D’Angelo, A., & Palihapitiya, C. M. (2018). Social advertisements and other informational messages on a social networking website, and advertising model for same. Patent no. US 9,984,392 B2 (applicant Facebook, Inc.). Washington, DC: US Patent and Trademark Office. Khamis, S., Ang, L., & Welling, R. (2017). Self-branding, ‘micro-celebrity’ and the rise of social media influencers. Celebrity Studies, 8(2), 191–208. Kocielnik, R., Keyes, O., Morgan, J. T., Taraborelli, D., McDonald, D. W., & Hsieh, G. (2018). Reciprocity and donation: How article topic, quality and dwell time predict banner donation on Wikipedia. In Proceedings of the ACM on Human-Computer Interaction, Vol. 2 (art. no. 91). New York: ACM. Kweon, H. J. & Kweon, S. H. (2021). Pricing strategy within the US streaming services market: A focus on Netflix’s price plans. International Journal of Contents, 17(2), 1–8. Larivière, V., Haustein, S., & Mongeon, P. (2015). The oligopoly of academic publishers in the digital era. PLoS ONE, 10(6), art. no. e0127502. Larnier, J. (2014). Who Owns the Future? New York: Simon & Schuster. Lehmann, S. & Buxmann, P. (2009). Pricing strategies of software vendors. Business & Information Systems Engineering, 1, art. no. 452. Lelonek-Kuleta, B., Bartczuk, R. P., & Wiechetek, M. (2021). Pay for play: Behavioural patterns of pay-to-win gaming. Computers in Human Behavior, 115, art. no. 106592. Liao, G. C., Chen, X., & Huang, J. W. (2022). Privacy-aware online social networking with targeting advertisements. IEEE-ACM Transactions on Networking, 30(3), 1312–1327. Lin, L. C.-S. (2021). Virtual gift donation on live streaming apps: The moderating effect of social presence. Communication Research and Practice, 7(2), 173–188. Linde, F. (2009). Pricing information goods. Journal of Product & Brand Management, 18(5), 379–384. Linde, F. & Stock, W. G. (2011). Information Markets. A Strategic Guideline for the I-Commerce. Berlin and New York: De Gruyter Saur. Lucking-Reiley, D. (2000). Vickrey auctions in practice: From nineteenth-century philately to twenty-first-century e-commerce. Journal of Economic Perspectives, 14(3), 183–192. Marwick, A. E. (2015). Instafame: Luxury selfies in the attention economy. Public Culture, 27(1), 137–160. McAfee, A. & Brynjolfsson, E. (2017). Machine Platform Crowd. New York: Norton. McGuigan, G. S. (2004). Publishing perils in academe: The serials crisis and the economics of the academic journal publishing industry. Journal of Business & Finance Librarianship, 10(1), 13–26. Monkman, C. D., Mehta, A., Muralidharan, O., Roat, C. M., Furrow, B., & Baltz, E. A. (2020). Multiple interaction auctions. Patent no. US 10,733,665 B2 (applicant Google LLC). Washington, DC: US Patent and Trademark Office. Natter, M. & Kaufmann, K. (2015). Voluntary market payments: Underlying motives, success drivers and success potentials. Journal of Behavioral and Experimental Economics, 57, 149–157. Newlands, G. & Fieseler, C. (2020). #dreamjob: Navigating pathways to success on an aspiring Instagram influencer. In C. Guanta & S. Rauchordiás (Eds.), The Regulation of Social Media Influencers (pp. 167–184). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Owen, B. M. & Wildman, S. S. (1992). Video Economics. Cambridge, MA: Harvard University Press. Pigou, A. C. (1932). The Economics of Welfare. 4th ed. London: Macmillan.

362  The Elgar companion to information economics Raban, D. R. & Ahituv, N. (2024). Assessing the perceived value of information in an information immersive world. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 364–378). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Raban, D. R., Barzilai, S., & Portnoy, L. (2019). The unexpected benefits of paying for information: The effects of payment on information source choices and epistemic thinking. In M. Pańkowska & K. Sandkuhl (Eds.), Perspectives in Business Informatics Research (pp. 163–176). Cham: Springer. Raban, D. R. & Włodarczyk, J. (2024). Information economics examined through scarcity and abundance. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 2–19). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Rafaeli, S. & Raban, D. R. (2003). Experimental investigation of the subjective value of information in trading. Journal of the Association for Information Systems, 4, 119–139. Ramnik, A. & Runke, K. D. (2018). User-level bidding for advertising campaigns. Patent Application no. US 2018/0218410 A1 (applicant Facebook, Inc.). Washington, DC: US Patent and Trademark Office. Rochet, J.-C. & Tirole, J. (2003). Platform competition in two-sided markets. Journal of the European Economic Association, 1(4), 990–1029. Rochet, J.-C. & Tirole, J. (2006). Two-sided markets: A progress report. The RAND Journal of Economics, 37(3), 645–667. Rowley, J. (1998). Information pricing policy: Factors and contexts. Information Services & Use, 18(3), 165–175. Ruhrberg, S. D., Kirstein, G., & Baran, K. S. (2017). User acceptance of personalized and context-specific online advertising. Open Journal of Social Sciences, 5(3), 223–232. Rusho, Y. & Raban, D. R. (2021). Join the club? Peer effects on information value perception. Journal of the Association for Information Science and Technology, 72(2), 156–172. Scheibe, K. (2018). The impact of gamification in social live streaming services. In G. Meiselwitz (Ed.), Social Computing and Social Media: User Experience and Behavior. Part I (pp.  99–113). Cham: Springer. Scheibe, K., Fietkiewicz, K. J., & Stock, W. G. (2016). Information behavior on social live streaming services. Journal of Information Science Theory and Practice, 4(2), 6–20. Scheibe, K., Meschede, C., Göretz, J., & Stock, W. G. (2018). Giving and taking gratifications in a gamified social live streaming service. In Proceedings of the 5th European Conference on Social Media (ECSM 2018), Limerick, Ireland (pp. 264–273). Reading: Academic Conferences and Publishing. Scheibe, K., Zimmer, F., Fietkiewicz, K. J., & Stock, W. G. (2022). Interpersonal relations and social actions on live streaming services: A systematic review on cyber-social relations. In Proceedings of the 55th Hawaii International Conference on System Sciences (pp.  3349–3358). Honolulu: HICSS (ScholarSpace). Seufert, E. B. (2014). Freemium Economics: Leveraging Analytics and User Segmentation to Drive Revenue. Waltham, MA: Morgan Kaufmann. Shao, G. (2009). Understanding the appeal of user-generated media: A uses and gratification perspective. Internet Research, 19(1), 7–25. Shulman, J. D. & Geng, X. (2019). Does it pay to shroud in-app purchase prices? Information Systems Research, 30(3), 856–871. Snyder, H. & Davenport, E. (1997). Costing and Pricing in the Digital Age: A Practical Guide for Information Services. New York: Neil-Schuman. Soubusta, S. (2008). On click fraud. Information – Wissenschaft und Praxis, 59(2), 136–141. Stange, M. & Funk, B. (2014). Real-time advertising. Business & Information Systems Engineering, 6, 305–308. Stiglitz, J. E. & Kosenko, A. (2024). Robust theory and fragile practice: Information in a world of disinformation. Part 1: Indirect communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 20–52). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stock, W. G. (1997). Informationsinhalte als Universaldienste. nfd. Information – Wissenschaft und Praxis, 48, 261–271. Stock, W. G. (2006). On relevance distributions. Journal of the American Society for Information Science and Technology, 57, 1126–1129.

Payment on information markets  363 Stock, W. G. (2020). N-ary information markets: Money, attention, and personal data as means of payment. Journal of Information Science Theory and Practice, 8(3), 6–14. Stock, W. G., Scheibe, K., Fietkiewicz, K. J., & Zimmer, F. (2022). Cyber social interactions: Information behavior in between social and parasocial interactions. Journal of Information Science Theory and Practice, 10(3), 15–23. Stock, W. G. & Stock, M. (2013). Handbook of Information Science. Berlin and Boston, MA: De Gruyter. Tang, Y. & Lian, Q. (2018). Competitive bidding scalers for online advertising auctions. Patent Application no. US 2015/0161679 A1 (applicant Facebook, Inc.). Washington, DC: US Patent and Trademark Office. Törhönen, M., Giertz, J., Weiger, W. H., & Hamari, J. (2021). Streamers: The new wave of digital entrepreneurship? Extant corpus and research agenda. Electronic Commerce Research and Applications, 46, art. no. 101027. Törhönen, M., Lobna, H., Sjöblom, M., & Hamari, J. (2019). Play, playbour or labour? The relationships between perception of occupational activity and outcomes among streamers and YouTubers. In Proceedings of the 52nd Hawaii International Conference on System Sciences (pp. 2558–2567). Honolulu: HICCS (ScholarSpace). Turner, C. N. (2014). E-resource acquisitions in academic library consortia. Library Resources & Technical Services, 58(1), 33–48. Varian, H. R. (1996). Differential pricing and efficiency. First Monday, 1(2). Vickrey, W. (1961). Counterspeculation, auctions, and competitive sealed tenders. The Journal of Finance, 16(1), 8–37. Wang, W. Y. & Guo, Q. (2023). Subscription strategy choices of network video platforms in the presence of social influences. Electronic Commerce Research, 23(1), 577–604. Weber, M. (1978). Economy and Society. Berkeley, CA: University of California Press. Włodarczyk, J. (2024). Information and income distribution: The perspective of information economics. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 81–104). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Wohn, D. Y., Jough, P., Eskander, P., Siri, J. S., Shimobayashi, M., & Desal, P. (2019). Understanding digital patronage: Why do people subscribe to streamers on Twitch? In CHI PLAY ’19. Proceedings of the Annual Symposium on Human-Computer-Interaction in Play (pp. 99–110). New York: ACM. Wolff, G. H. & Shen, C. (2022). Audience size, moderator activity, gender, and content diversity: Exploring user participation and financial commitment on Twich.tv. New Media & Society (online first). Yu, E.-J., Jung, C., Kim, H., & Jung, J. (2018). Impact of viewer engagement on gift-giving in live video streaming. Telematics and Informatics, 35(5), 1450–1460. Zhao, Y. (C.), Peng, X., Liu, Z., Song S., & Hansen, P. (2020). Factors that affect asker’s pay intention in trilateral payment-based social Q&A platforms: From a benefit and cost perspective. Journal of the Association for Information Science and Technology, 71(5), 516–528.

18. Assessing the perceived value of information in an information immersive world Daphne R. Raban and Niv Ahituv

INTRODUCTION Immersion in information. This concept describes the state of human society today. ● Pick up a few phrases in a foreign language? Don’t worry, just check your smartphone. ● Wish to dine out and watch a movie? Check the available recommendations. ● Need data to support an argument? Data is a few clicks away. Whether information needs are trivial or demanding, we all experience the sense of immersion, having access to and using information throughout our waking hours. The abundance of information and its unrestricted usage may divert attention from the need to assess the value of information. On one hand, information and the associated technology are essential for all modern functions, but on the other hand, unquestioned reliance on data, algorithms and interfaces may carry a host of risks from simple errors in judgment to overarching political threats. As reliance on digital output increases, the need for value assessment intensifies. Gaining a solid understanding of value assessment is important for information consumers and is likely to help information providers implement value into their tools and content. How can a person evaluate the information available to her/him while in a state of immersion? This chapter revisits the theoretical basis of information value assessment published in 1989 and examines it with a current lens (Ahituv, 1989). Enormous changes have taken place in the world of information since 1989. Updated theory needs to account for widespread phenomena such as immersion in information, disintermediation of information production, disembodiment of information generation (e.g., artificial intelligence) and movement toward openness of code, data, and content, to name a few of the major changes. The rise of user-generated data, the internet-of-things and artificial intelligence led to the strong emphasis on data and its potential. While data was always the raw material in the production of information, data has become a leading source of value, business success and economic wealth. Alongside the rise of data, the word “content” became synonymous with the earlier usage of “information” to denote meaningful and creative products of human thought. This chapter examines the premises of Ahituv’s 1989 paper, discusses meanings of value and some of the attributes that contribute to value perceptions. A description of the results of some of our empirical studies provides a step between theory and practice, and, finally, the chapter offers a revised and updated approach to the evaluation of information including an elaboration regarding the cost of data.

364

Assessing the perceived value of information in an information immersive world  365

ASSESSING THE VALUE OF INFORMATION: WHAT HAS CHANGED SINCE 1989? The Ahituv (1989) paper, entitled “Assessing the value of information: Problems and approaches”, stated three premises: 1. The value of information is assessed together with the value of the system that provides the information. Information and information system (IS) are one entity. 2. The quantity of information does not necessarily serve as an argument in the information value function. 3. The information value function is a multiple attribute function. Its arguments are characteristics of an information system (e.g., timeliness, preciseness, relevance); its results (i.e., dependent variable) relate to benefits­. In the following we discuss these premises in the context of the vast changes that have taken place since the original publication. Premise 1 – the strong movement toward openness renders this premise inaccurate. Data and content are largely separate from the systems that store, process, and exchange them. Bits are separate from atoms (Negroponte, 1996). In fact, the typical user of open information (e.g., users of Google, Wikipedia, Open Science, etc.) is oblivious to the costs involved in developing and maintaining the information system providing the information (Shapiro & Varian, 1999). Sometimes, the user may be interested in costs involved in obtaining and handling the information, but it is infrequent, particularly for individual users. The value of the system itself is, of course, substantial, however, as suggested by Nicholas Carr (2004), information systems have become so prevalent and reliable that we tend to view them as infrastructure, fairly similar to the availability of electricity. Even unique human–computer interaction elements embedded in systems cannot account for lasting value since they are quickly implemented by all relevant system providers and usually do not offer a lasting advantage (e.g., touch screen). While in 1989 systems were largely organizational and inseparable from the data they stored, nowadays, individuals are the majority of data consumers, and we consider data and content in and of themselves. The value of the system is separate. Premise 2 – in 1989, information was usually scarce which is the opposite of the current state of immersion in abundant information. We suggest considering whether information is used by people or by computers. When information is used by people, the original (1989) premise 2 is largely still true, however, it should be examined following quantification of the information overload phenomenon (Jones et al., 2001). Larger information quantity and variety is likely to be advantageous up to a certain limit. At a certain point, more is not better – people may lack time, tools, or cognitive capacities to acquire additional information. The question is how to assess the value of the information to identify when more is better, to detect the appearance of a turning point where the value becomes zero, and to reveal declining value. The advent of artificial intelligence heralded the use of data by advanced computers with minimal human intervention. When computers are semi-independent users of data, for example in machine learning, the tendency might be to assert that “more is better”. However, this carries a risk in wasting development and processing time and delaying the results by implementing a quantity-driven approach. The traditional principle of Occam’s Razor is likely to be useful even when computing power is abundant in order to deliver timely information. In

366  The Elgar companion to information economics addition, “mammoth” data processing is associated with a significant negative externality as it consumes a lot of electrical power which carries costs and an environmental effect. Another issue that may carry a quantitative effect is fake information (Stiglitz & Kosenko, 2024) – where is the tipping point when fake information starts to deliver detrimental effects? Can we assess the negative value of fake information? When fake information is revealed, is there a positive value for the user who realized it is fake? Maybe fake information should be assessed by two contradictory values: positive for the distributor; negative for the consumer of the information. Premise 3 – based on Premise 1, here too, we omit the role of the system and examine information, the intangible content or data that flow in systems. Content and data have unique characteristics differentiating them from other items offered in markets (Bates, 1988; Raban, 2007; Shapiro & Varian, 1999). For example, besides being intangible, they are readily copied, often available for free, and they address human intellect. These and other attributes may influence value assessment. The premise still holds when it comes to the dependent variable, benefits. While the word “value” has diverse meanings in various fields of research, for the purpose of information economics, an assessment of benefits is in order. Based on the vast changes in data, content, and access to information over the past three decades, we suggest a revised version of the premises so that in this day and age: 2022 Premise 1: the value of information is independent from the infrastructure systems (hardware, software, and communication) storing and conveying it. 2022 Premise 2: the quantity of information may become an argument in the information value function. The positive or negative impact of the quantity may vary according to the circumstances. 2022 Premise 3: The information value function is a multiple attribute function. Its arguments are characteristics of content and data; its results (i.e., dependent variable) relate to benefits. The discussion of the premises exposes a change in terminology. The earlier paper analyzed “information”, while the current vocabulary usually distinguishes data from content. One might safely state that information is the umbrella term for data and content combined, however, the distinction is useful for studying the respective value assessments. Apart from terminology, contemplating the three premises together suggests focusing on understanding the attributes that influence the value assessment of data and content and to consider their boundary conditions. Such focus would be the basis for advancing a current understanding of information value assessment. Our point of departure is the theory in the 1989 paper which laid out three approaches to assessing the value of information in the context of decision making: The normative approach focuses on the expected utility of a decision maker who is assumed to be a utility maximizer. The result is a theoretical analytical model (usually mathematical) that is likely to inform engineering-related decisions, however, its usefulness for describing daily information consumption value assessment is limited. Immersion in information calls for developing a practical approach to value assessment, an approach that considers some behavioral arguments. The realistic approach describes how outcomes of a decision may change depending on the use of information. This requires equal conditions for making a decision and comparing

Assessing the perceived value of information in an information immersive world  367 its outcomes when implemented without information, with partial or complete information (Ahituv et al., 1998). This is largely a hypothetical situation. Again, we find ourselves searching for a pragmatic approach. The perceived value approach does not make strong assumptions regarding decision makers. According to this approach, people have individual preferences, and they select information they perceive as likely to fulfill their expectations. Research in the area of marketing elaborated on perceived value resulting in a recommendation to assess value using an eight attribute typology (Sánchez-Fernández & Iniesta-Bonillo, 2007). While this approach may be useful for items of commerce, it seems somewhat excessive for assessing the value of information because information is acquired continuously, often in small units, it is abundant and ephemeral. Brennan et al. (2019) identified five dimensions to data value: operational impacts, replacement costs, competitive advantage, regulatory risk, and timeliness. This is useful from an organizational perspective, but less pragmatic from an individual standpoint. Behavioral economics offers a pragmatic approach to studying individual preferences and perceived value. In fact, the focus of Prospect Theory is the subjective nature of the value curve (Kahneman & Tversky, 1979). This was implemented and demonstrated in experimental studies on perceived value of commercial goods (Ariely et al., 2006; Kahneman et al., 1990; Thaler, 1980). Interestingly, behavioral economists rarely discuss the value of information and information markets, which is the focus of the present chapter. The study of information value should capture the influence of its attributes quickly. For quick assessment to take place, perceived value should be represented by a single parameter reflecting benefit (Premise 3). The single parameter can then be evaluated on a comparative basis across sources of information and contexts and help in making assessments considering the rapid rate of change in perceived value over time.

ASSESSING THE PERCEIVED VALUE OF INFORMATION: EMPIRICAL WORK Willingness-to-pay is a longstanding widely used variable for assessing perceived value or anticipated benefit (Horowitz & McConnell, 2002). In experimental settings, users select among alternatives after submitting their willingness-to-pay, which is then assessed in a comparative manner. In other words, willingness-to-pay is not taken as a representative of price; instead, it is interpreted as a relative quantitative measure across various information products or services. It is important here to distinguish between price and value. Price is determined by sellers as part of normal business practice, based on various factors such as cost. Price is usually non-negotiable – consumers make a choice whether to purchase an object at a given price. Value refers to the subjective perception of the usefulness and/or attractiveness of a product, in the present case, of information. The main advantage of the perceived value approach is that beyond its theoretical basis, it is pragmatic and can be helpful in understanding and supporting information markets. Its pragmatism derives from: (1) reflecting actual user behaviors and preferences; (2) its empirical approach can readily be implemented at scale, not only in laboratory or academic settings; (3) digital implementation enables to embed nudges and change them rapidly, when needed; (4) machine learning algorithms can capture perceived value and produce aggregate or personalized responses.

368  The Elgar companion to information economics Realizing that the perceived value approach is the more promising approach, the chapter proceeds to describe some of our empirical work for identifying attributes that influence value perception and offer a framework for value assessment.

THE PERCEIVED VALUE OF INFORMATION IN AN INFORMATION IMMERSIVE WORLD While information is abundant, immersive, cheap, and often overwhelming, its accuracy, preciseness, and age are not necessarily reliable; it might be fake or biased. Moreover, there is an apparent detachment between the cost (or value) of the system or application delivering the information and the value of the content flowing through it (Premise 1). For example, the user of WhatsApp is not interested in the cost of development and maintenance of the application but is interested in receiving and transmitting messages. This implies that the messages do carry value for the user. These conditions have led us earlier to single out the perceived value of information as the most prominent approach of measuring the value of information. In fact, the separation between the information and the information system saves some complexity and enables focus. In the following, we provide research observations emanating from information consumer behavior based either on experimental settings or on analyzing data available from actual information transactions online. To avoid repetitive descriptions of experiments, we note that the experiments mentioned below are based on the tradition of behavioral economics, mainly work related to the Endowment Effect (Thaler, 1980; Thaler et al., 1992), and usually include the following steps: ● ● ● ●

Present a scenario which requires a decision involving risk or uncertainty Offer a selection of information sources as a decision aid to reduce risk or uncertainty Implement a mechanism to buy or sell information sources for play money Record the transactions in a database for later analysis

Because we live in a world of free and abundant information, one might claim that this experimental setup resembling a market may not be representative of the real world. Indeed, this approach offers innovation which may lead in the future to dual exchanges, concurrently offering free and fee-based options for consuming information, as noted elsewhere (Raban, 2022). Imagine that government regulation would force information monopolies to offer both free and fee-based information for the consumers’ choice. Most consumers would probably choose the free products. Similar behavior is often revealed when one is asked to install a new application, where the simple version is free and a premium version is available for a fee. Most users select the free version (Anand & Gupta, 2014). However, this trend could change over time based initially on two drivers: a small number of consumers of paid information who may form a critical mass for others to join later; by forcing companies through regulation to exercise full transparency about personal data usage which finances free information. Research evidence supports the use of a market setup for studying the perceived value of information. For example, in a study where participants could use an allotted budget to choose whether to buy information or rely on free information, 6 percent selected only free sources, 38 percent played with paid information only, 44 percent played with mixed, free and paid sources, and 12 percent of the participants played without using any information (Raban &

Assessing the perceived value of information in an information immersive world  369 Koren, 2019). These findings indicate the various preferences that information consumers have. Current markets serve mainly consumers who seek free information. Another approach to relate to the claim that an experiment setup resembling a market has limited external validity is by using a different research approach. Instead of using data from laboratory experiments, some studies described below rely on data available from actual information consumption transactions providing further evidence for the consumers’ demand for various and innovative market structures. A third promising approach might be to measure the negative value of “starvation” for information or information curfew: How long can one stay with no access to routine information he/she is used to getting on a regular/immediate basis? What is one’s emotional state while experiencing information deprivation? Developing this research approach opens future opportunities for studying alternative variables as indicators of perceived value, such as time-related variables and various psychological influences. Information value perception is dynamic, changing by person and circumstance. In reference to Premise 3, a host of variables influences value perception. Variables influencing perceived value generally fall into three groups: individual user characteristics (such as prior knowledge, personal experience, preferences), characteristics of the environment where information is used (such as market structure, payment mechanisms, level of interaction, social feedback, relations of authority and social perceptions), and characteristics of the information itself (such as format, accuracy, timing, source of the information and its availability). In the following we describe some results of experimental and data-driven studies in each of these three groups. Influence of Individual Characteristics on Information Value Perception Value perception is subjective, changing based on individual differences and preferences which is why we selected to start with this perspective. A simple yet compelling case of the subjectivity of perceived value comes from the “school of thought” or disciplinary background of people. The results of an experiment show how exposing the same texts to people from different formal academic backgrounds, economics and communication, results in diverging evaluations of those texts (Gaziel Yablowitz & Raban, 2016). As mentioned earlier, endowment has been found as a substantial influencer of value perception (Kahneman et al., 1990). Yet, what is the meaning of endowment or ownership when it comes to digital information, which can easily be copied and shared, rather than sold? Information ownership usually is not transferred; rather, it is shared and multiplied by copying. It stands to reason that we should expect that the Endowment Effect does not hold for information. However, an experiment conducted using a simple business game played by 294 participants uncovered an Endowment Effect for digital information at a level comparable to that of regular market goods (Raban & Rafaeli, 2006). The effect was stronger for information presented as “original and exclusive” than for information presented as a “copy”, yet the effect was pronounced even in the latter case. Moreover, 27 percent of the information selling transactions were executed while 63 percent of the buying transactions took place. Participants, it turned out, made an effort to buy and keep information. These findings indicate that: 1. Psychological ownership is important even in the digital sphere, probably because it reflects a desire to possess information. 2. Information had no objective/realistic value in the experiment, but it had a high subjective value. 3. Under-trading took place especially in selling, suggesting a preference to buy but not sell information, a tendency to hoard information. 4.

370  The Elgar companion to information economics Possibly, when buyers become an active part of information pricing (participatory pricing), then more buying transactions take place. In another experiment, participants had to provide advice on a controversial topic – the connection between dairy products and cardiovascular health. To base their advice, groups of participants either received access to free information or had to bid to purchase information items. They also filled a questionnaire about their epistemic beliefs (Raban et al., 2019). Interestingly, while epistemic beliefs are considered to be stable, they actually changed upon the introduction of economic thinking through the need to purchase. People who were exposed only to free information accepted it “as is”, as true. People required to place a price bid became more critical and evaluative of the information. This finding suggests that a “play economy” might be used to improve information literacy. To research real information transactions, the long defunct Google Answers website presented an interesting information market environment (Rafaeli et al., 2007). This was a website devoted to answering questions for a payment, which also offered an option for free comments, and another interesting option for providing a tip, a voluntary ex-post payment. The pricing method of Google Answers resembled willingness-to-pay bids, pre and post usage of information. The Google Answers website included about 130,000 questions submitted over the course of four years, and about 52,000 paid answers. The data included: 1. Questions (with price bid). 2. Answers (paid). 3. Comments (free). 4. Tips (post-hoc). 5. Tip text (gratitude). 6. Ratings (5-star scale). Analysis of the data showed that over four years of site activity, the prices went up and the tips increased as well (Raban, 2008). Consumers were learning to pay for answers. Relating to individual characteristics that influence economic activity, findings from a study of the Google Answers Q&A site indicate how self-presentation contributes to information value perception. Data analysis revealed a mirror-effect. The more askers disclosed about themselves and wrote at length, the more the answerers did the same. This conversational social reciprocation also resulted in economic benefits seen in higher pay and higher tips and ratings provided on the Google Answers site. Another finding was that merely writing “thanks in advance” within the question was related to the provision of a tip. So an answerer who is able to detect these language cues in a question can increase his/her chances of getting a tip, just by identifying the asker’s inclination through the writing style (Raban, 2012). Providing tips in an anonymous system is still an enigma awaiting further research. Another direction awaiting research is individual characteristics influencing value perception. For example, are there gender differences? Age differences? Cultural differences? Next, we discuss the influence of the exchange environment on information value perception. Influence of Environment Characteristics on Information Value Perception An interesting phenomenon is that free and paid information coexist. This is quite a unique characteristic of information markets. Offline and online shops for various goods and services offer products for payment, however, in information markets we find an abundance of free information. In regular commerce, “free” is usually a temporary promotional gimmick given away in small quantities (Stock, 2024). In information markets, free information is the norm, and it is stable and constant, not temporary. This is especially interesting because we know from prior research and from anecdotal experience, that mixing business and pleasure is

Assessing the perceived value of information in an information immersive world  371 usually not advisable, so how does sharing information for free and selling it for payment work together? Researchers who studied motivation (Deci & Ryan, 1985; Frey & Oberholzer-Gee, 1997) call this “crowding out”. They showed that when people do things out of social incentives, social behavior is terminated with the introduction of money. Indeed, companies offering free products, such as software, need to take special measures to “educate” their customers to pay (Anand & Gupta, 2014). Dan Ariely shows a similar effect in a labor context – people will stop volunteering when money is introduced (Heyman & Ariely, 2004). So, the coexistence of free and paid information is not a trivial phenomenon, and it deserves some research attention. Another analysis of the Google Answers data mentioned earlier focused on the people providing the answers on the site. The data showed that tip, not price, is the main predictor of the number of answers provided by “sellers” of answers (Raban, 2008). Further investigation of the tips surfaced that tips were, on average, 38 percent of the price paid to frequent answer sellers, while the infrequent answerers received only 14 percent of the price as tip. This is an indication that the system evolves to beneficial selection of answerers: those who provide high-quality answers receive high tips and they tend to persist in the system, while the ones giving less helpful answers are also the ones who answer fewer questions. Tips are an interesting form of payment for information, being voluntary and paid ex post. Designers of information exchange environments may consider providing a method for the provision of tips. The number of free comments was another significant incentive for being frequent and high-quality answer sellers. The results about tips and comments indicate that the crowding out effect does not occur in information markets. To the contrary, free activity is a catalyst to paid activity in information markets. Another observation is the usefulness of tips as a payment option in the case of information. Because value perception is subjective and is based on experiencing information, tips or ex-post voluntary payments seem especially suitable to address these features of information. Furthermore, it may be that the market structure offering a combination of pre- and post-usage payment is advisable. It reduces the risk associated with unknown content while concurrently offering a post-use payment that enables to acknowledge the actual usefulness of the content. Although tips were used only in part of the transactions, they could be envisioned as a new norm in online behavior in a future setting. The extent of voluntary payment in the Google Answers website certainly passed the tipping point (pun intended), which, in network environments is crucial for success. A study based on a series of mathematical simulations using directed and undirected networks showed an evolutionary process like the one described above from field data. According to the simulations, low quality information traveling in the networks is being filtered out and reaches a small number of people. High quality information travels long and far into the network, sometimes reaching “niche corners” (Koren et al., 2016). Since information inevitably flows in networks, the network structure and affordances play an important role in understanding the spread of valuable information. Supreme Court Justice Louise D. Brandeis famously said: “Sunlight is said to be the best of disinfectants; electric light the most efficient policeman”. To paraphrase, in the current case: open bi-directional networks are the best of information disinfectants. Yet, we must keep in mind the subjective nature of the value of information. Sometimes information that spreads in networks is misleading or manipulative (Stiglitz & Kosenko, 2024), nevertheless, it spreads because a critical mass of people perceives it as valuable. In this case, the positive value of network affordances in delivering true information comes at the cost of the opportunity to disseminate fake information.

372  The Elgar companion to information economics The coexistence of free and fee-based, the exchange affordances and the network structure are all part of the environmental influences on value perception. Next, we turn our attention to characteristics of information which influence value perception. Influence of Information Characteristics on Information Value Perception Before acquiring information, we evaluate its value based on meta-data available to us (title, abstract, author, recommendations, etc.), while after using information, we are influenced by the experience of reading the content. Hence, value perception is based on either meta-data or full content. An experiment based on a knowledge game offered a digital store of information sources which were necessary for solving the knowledge questions (Raban & Mazor, 2013). Participants could submit a willingness-to-pay bid (based on meta-data) and they could provide a voluntary payment after having read the content. Of 120 participants, only 99 selected to buy information by submitting willingness-to-pay bids. Twenty-one people preferred to avoid buying information and played the game based on guessing the answers. That presented a natural opportunity to assess the realistic value of information. Indeed, people who bought information had 25 percent more correct answers. The most striking finding was a “38×38” observation – 38 percent of the participants selected the option to provide a post-reading voluntary payment, and the payment was at a rate of 38 percent of their willingness-to-pay bid, on average. This finding resonates with earlier findings from field data (Raban, 2008). From a business perspective, it indicates that merely providing the opportunity to pay after the experience of reading is desirable because it adds the value of the content to the pre-use value assigned based on meta-data. In terms of value perception, this experiment shows the respective contributions of meta-data and content. Further support for the effect of meta-data on value perceptions comes from results reported earlier regarding psychological ownership of information. The effect in that experiment was received by labeling information as “original” or “copy” (Raban & Rafaeli, 2006). Such labels are meta-data. Further evidence for the specific contribution of content to value perception comes from a study which manipulated informational cues. Planting cues describing low and high risk within texts showed how they influenced the willingness-to-pay for information (Raban & Koren, 2019). Textual cues regarding high business risk were associated with higher evaluations of content and with the purchase of a larger number of content items as compared to textual cues describing a lower level of business risk.

A FRAMEWORK FOR THE PERCEIVED VALUE OF INFORMATION Our updated premises suggest that information (content, data) should be studied independently from the system delivering it, and that value is a multi-attribute function of benefits. Based on those premises and on the findings reported here, Figure 18.1 suggests a framework accounting for the current knowledge and suggesting a few ideas for future study of the perceived value of information.

Assessing the perceived value of information in an information immersive world  373

Figure 18.1

A framework for studying the perceived value of information

The empirical findings reviewed here suggest that value-based information exchange is likely to be lively and beneficial if done in the right manner, under carefully designed circumstances accounting for individual, environment and informational aspects. In brief: ● Allow for concurrent free and fee-based information and consider offering a choice between them ● In the fee-based area, provide several payment options, pre- and post-usage ● Include social and feedback elements.

THE PERCEIVED VALUE OF INFORMATION AND DECISION MAKING It is obvious that the era of information immersion requires a thorough revisiting of the study of the value of information. Not only is the amount of information exceptionally large and rapidly increasing, but information is highly accessible. For example, managers access and use information on their own, not necessarily relying on others to provide them with data and content (Markovich et al., 2019). With face-to-face interaction giving way to digital meetings, new questions arise (Figure 18.1 – layer of proposed areas for research) as to the perceived value of information during the decision making process in teams as well as by individuals. Previous research showed that decision making and the behavior of online teams are different than face-to-face teams (Shwartz-Asher et al., 2009; Shwartz-Asher & Ahituv, 2019). Every decision making process is based on a data cycle culminating in a decision being made. The cycle can be short and based on a few data items, or complicated and involved, and based on retrieval, cleansing, and integration of big data taken from various sources. However, the stages of the Data Cycle are nearly the same for each degree of complexity, in each sector, and for each discipline. Figure 18.2 portrays the Data Cycle (Ahituv, 2019).

374  The Elgar companion to information economics

Figure 18.2

The Data Cycle (begins at the upper left corner)

We briefly describe each stage of the Data Cycle and list potential tools that can support each stage: 1. Problem definition: An initial definition of the problem, or the mission, or the purpose, for which data is required. Potential tools: formulation methods, quantitative models, qualitative approaches, mathematical tools, and the like. 2. Identifying pertinent data sources: Understanding what data are pertinent, and where they can be located. Potential tools: browsers, indices, search engines, international organizations, statistics bureaus, and the like. 3. Data collection and storing (including cleansing and backup): retrieval of data from various sources and storage in an accessible location; data validation and cleansing. Potential tools: data transfer technology – communications, clouds, database management software, data validation tools, and the like. 4. Data integration: This (very important) stage should allow the user to incorporate data from varied sources whose data definition and format were not initially compatible, nor are they synchronized. Potential tools: conversion programs, indices, meta-data tools, and the like. 5. Data mining: Selection of relevant data out of the big data. Potential tools: filters, data retrieval techniques, identification tools, AI tools, heuristics, and the like. 6. Processing and analysis: The data that were selected earlier are now screened, processed, and analyzed. Potential tools: algorithms, AI tools, machine learning, data processing programs, heuristics, and the like. 7. Visualization: Presentation of the results to the decision maker(s). Potential tools: dashboard software, graphical tools, reporting systems, interactive systems, voice, UX programs, visualization tools, and the like. 8. Learning and decision making: The final stage that is the purpose of the Data Cycle. The results are displayed to the decision makers and decisions are taken. Potential tools: decision support tools, what-if software, simulation tools, and the like.

Assessing the perceived value of information in an information immersive world  375 9. Feedback for further cycles: This stage is not always necessary. However, very often, the need to make a certain decision is repetitive, so the customer (the decision maker) can affect the usefulness and the effectiveness of the cycle by forwarding comments and changes. Potential tools: reporting systems, interactive reactions, fine tuning tools, DEVOPS tools, agile design tools, machine learning, knowledge management routines, and the like. In many cases, particularly for individual needs, the costs of using the above tools are zero or negligible. However, for many organizational or scientific problems, the costs are substantial indeed, and should be incorporated in the value assessment. In other words, the recent focus on data and the need to generate content from data raises a new interest or concern regarding costs. While previously information economics treated costs as sunk (Shapiro & Varian, 1999), the intensive use of data may be a cause for reevaluating the practice of ignoring costs on the production side. From the consumers’ perspective, research indicates that exposing production costs to consumers is likely to affect value perception (Mohan et al., 2020). Consumers exposed to cost information become more willing to pay for goods as they come to trust the sellers. Interestingly, this trust is not related to verifiability of the information provided by the sellers regarding their costs. With the increasing contribution of information to the economy, new measures of value emerge (Collis & Brynjolfsson, 2019). Information production costs, consumer surplus, positive and negative externalities are all candidates for value assessment. To conclude, we offer scenarios to support our call to assess the perceived value of information.

CONCLUSION Assessing the value of information in the information immersion environment is a major challenge. Let’s demonstrate this by asking some seemingly naïve questions: ● Suppose we don’t have Google nor Wikipedia at hand, and we want to get some information on a person or a concept. How long it would take us to obtain data that is somewhat equivalent to the data that could be retrieved from Google or Wikipedia? What would be the cost of such search? If we could know the answers, then those should be an approximation for the realistic value of information. Unfortunately, it is not likely that alternative sources of data still exist. The bridges backward have been burned. ● Suppose a top executive of a big corporation cannot access the regular dashboard that provides the continuous summary of pertinent information required to run the organization. The dashboard is assembled out of numerous sources of data (big data). How can we measure the value of the dashboard data? Can we assess the amount of money the organization loses during the downtime? Not very likely. It looks as if the perceived value, namely, a value based on a subjective judgment of the user, is the only tool for assessment. We argue that assessing the perceived value of information via a quickly measurable, single variable (willingness-to-pay) provides insight for consumers and producers of information to the extent that informs the design of information markets. Behavioral experiments indicate that while people tend to hoard information, their evaluation of information sources improves

376  The Elgar companion to information economics when they are in a marketplace setting. Implicit and explicit dialogue between sellers and buyers in information markets is helpful toward increasing the rate of transactions and the resulting satisfaction. Gradually, consumers’ value perceptions increase, and they learn to pay for information. This occurs as a result of a process of recognition of the preferred sources of information over time. Even in the brief time that passes between acquiring a source of information, reading it, and deliberating on it, value perception changes. In other words, information consumption is a learning process and so the development of information markets should follow the learning pattern. Another influence on value perception is that information transactions take place within a network. The network structure and dynamics influence the kind of information presented to users, true or fake, and it induces social influence by disclosing the behavior of others. The coexistence of free and fee-based information in the network is also a substantial influence on value perceptions. Producers may implement new approaches in information markets by considering the individual, environmental and informational influences described here, accounting for the learning processes and network structure. They may choose to innovate and offer information markets that include participatory pricing, ensure social interaction on their systems, introduce playful elements, or allow voluntary payment. As a complement or alternative, government regulation could force information monopolies to offer concurrently free and fee-based information for the consumers’ choice. Most consumers would probably choose the free products, yet, over time preferences may change for some consumers based on two drivers: (a) the formation of a critical mass, albeit small, of paying consumers may influence others; (b) by promoting government regulation requiring companies to exercise full transparency about personal data usage which finances free information. Increasing consumer awareness through such regulation may influence consumer behavior. Overall, this chapter started from the 1989 paper which recommended perceived value as the most relevant approach to assess the value of information. We updated the premises that led to this recommendation and provided an empirical approach and findings to support it. Finally, we demonstrated the relevance to decision making and to information markets. We conclude by asserting that the characteristics of information, mainly intangibility and abundance, require economic thinking that combines individual empiricism with macro-level regulation to a ensure continued vibrant and ever-expanding use of information.

REFERENCES Ahituv, N. (1989). Assessing the value of information: Problems and approaches. In J. I. DeGross, J. C. Henderson, & B. R. Konsynski (Eds.), Proceedings of the Tenth International Conference on Information Systems (pp. 314–325). Ahituv, N. (2019). What should be taught in an academic program of data sciences? Digital Presentation and Preservation of Cultural and Scientific Heritage, 9, 55–64. Ahituv, N., Igbaria, M., & Sella, A. (1998). The effects of time pressure and completeness of information on decision making. Journal of Management Information Systems, 15(2), 153–172. Anand, B. & Gupta, S. (2014). Making “freemium” work. Harvard Business Review, 92(5), 27–29. Ariely, D., Loewenstein, G., & Prelec, D. (2006). Tom Sawyer and the construction of value. Journal of Economic Behavior & Organization, 60, 1–10.

Assessing the perceived value of information in an information immersive world  377 Bates, B. J. (1988). Information as an economic good: Sources of individual and social value. In V. Mosco & J. Wasco (Eds.), The Political Economy of Information (pp.  76–94). Madison, WI: University of Wisconsin Press. Brennan, R., Attard, J., Petkov, P., Nagle, T., & Helfert, M. (2019). Exploring data value assessment: A survey method and investigation of the perceived relative importance of data value dimensions. ICEIS 2019 – Proceedings of the 21st International Conference on Enterprise Information Systems, 1 (January), 188–195. https://​doi​.org/​10​.5220/​0007723402000207. Carr, N. G. (2004). Does IT Matter? Information Technology and the Corrosion of Competitive Advantage. Boston: Harvard Business School Publishing. Collis, A. & Brynjolfsson, E. (2019). How should we measure the digital economy? Harvard Business Review, 97(6), 140–149. Deci, E. L. & Ryan, R. M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. New York: Plenum Press. Frey, B. S. & Oberholzer-Gee, F. (1997). The cost of price incentives: An empirical analysis of motivation crowding-out. American Economic Review, 87(4), 746–755. Gaziel Yablowitz, M. & Raban, D. R. (2016). Investment decision paths in the information age: The effect of online journalism. Journal of the Association for Information Science and Technology, 67(6), 1417–1429. Heyman, J. & Ariely, D. (2004). Effort for payment: A tale of two markets. Psychological Science, 15(11), 787–793. Horowitz, J. K. & McConnell, K. E. (2002). A review of WTA/WTP studies. Journal of Environmental Economics and Management, 44(3), 426–447. Jones, G. Q., Ravid, G., & Rafaeli, S. (2001). Information overload and virtual public discourse boundaries. Interact 2001. Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1990). Experimental tests of the endowment effect and the Coase theorem. Journal of Political Economy, 98(6), 1325–1348. Kahneman, D. & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291. Koren, H., Kaminer, I., & Raban, D. R. (2016). Consume, modify, share (CMS): The interplay between individual decisions and structural network properties in the diffusion of information. PLoS ONE, 11(10), e0164651. https://​doi​.org/​10​.1371/​journal​.pone​.0164651. Markovich, A., Efrat, K., Raban, D. R., & Souchon, A. L. (2019). Competitive intelligence embeddedness: Drivers and performance consequences. European Management Journal, 37(6), 708–718. Mohan, B., Buell, R. W., & John, L. K. (2020). Lifting the veil: The benefits of cost transparency. Marketing Science, 39(6), 1105–1121. Negroponte, N. (1996). Being Digital. New York: Vintage Books. Raban, D. R. (2007). User-centered evaluation of information: A research challenge. Internet Research, 17(3), 306–322. Raban, D. R. (2008). The incentive structure in an online information market. Journal of the American Society for Information Science & Technology, 59(14), 2284–2295. Raban, D. R. (2012). Conversation as a source of satisfaction and continuance in a question-and-answer site. European Journal of Information Systems, 21, 427–437. Raban, D. R. (2022). What are information markets, and why are they failing? Tel Aviv: Heinrich-Böll-Stiftung Publications. https://​il​.boell​.org/​en/​2021/​12/​24/​what​-are​-information​-markets. Raban, D. R., Barzilai, S., & Portnoy, L. (2019). The unexpected benefits of paying for information: The effects of payment on information source choices and epistemic thinking. In M. Pańkowska & K. Sandkuhl (Eds.), Perspectives in Business Informatics Research (pp. 163–176). Cham: Springer. Raban, D. R. & Koren, L. (2019). Risk as a predictor of online competitive information acquisition. Open Information Science, 3(1), 47–60. Raban, D. R. & Mazor, M. (2013). The willingness to pay for information in digital marketplaces. In A. Kobyliński & A. Sobczak (Eds.), Perspectives in Business Informatics Research (pp. 267–277). Cham: Springer. Raban, D. R. & Rafaeli, S. (2006). The effect of source nature and status on the subjective value of information. Journal of the American Society for Information Science and Technology, 57(3), 321–329.

378  The Elgar companion to information economics Rafaeli, S., Raban, D. R., & Ravid, G. (2007). How social motivation enhances economic activity and incentives in the Google Answers knowledge sharing market. International Journal of Knowledge and Learning, 3(1), 1–11. Sánchez-Fernández, R. & Iniesta-Bonillo, M. Á. (2007). The concept of perceived value: A systematic review of the research. Marketing Theory, 7(4), 427–451. Shapiro, C. & Varian, H. R. (1999). Information Rules: A Strategic Guide to the Network Economy. Boston: Harvard Business School Press. Shwartz-Asher, D. & Ahituv, N. (2019). Comparison between face-to-face teams and virtual teams with respect to compliance with directives. Journal of Service Science and Management, 12(4), 549–571. Shwartz-Asher, D., Ahituv, N., & Etzion, D. (2009). Computer-mediated group interaction processes. Proceedings of the International Conference on Complex, Intelligent and Software Intensive Systems, CISIS 2009, April (pp. 255–262). https://​doi​.org/​10​.1109/​CISIS​.2009​.124 Stiglitz, J. E. & Kosenko, A. (2024). Robust theory and fragile practice: Information in a world of disinformation. Part 2: Direct communication. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 53–80). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Stock, W. G. (2024). Payment on information markets. In D. R. Raban & J. Włodarczyk (Eds.), The Elgar Companion to Information Economics (pp. 339–363). Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Thaler, R. H. (1980). Toward a positive theory of consumer choice. Journal of Economic Behavior and Organization, 1(1), 39–60. Thaler, R. H., Kahneman, D., & Knetsch, J. L. (1992). The endowment effect, loss aversion and status quo bias. In R. H. Thaler (Ed.), The Winner’s Curse: Paradoxes and Anomalies of Economic Life. New York: Free Press.

19. The role of influencer endorsements in users’ willingness to pay for knowledge products: an empirical investigation Xiaoyu Chen and Alton Y. K. Chua

1. INTRODUCTION Digital influencers are online celebrities who have accumulated a sizeable following via social media (Chen & Chua, 2021). Considered trustworthy sources by people who follow them, they commonly monetize the trust by persuading followers to buy endorsed brands’ products or services (Wang et al., 2021). This business practice, also known as “influencer marketing”, has become prevalent as it allows sponsors to improve their brand image and the perceived value of their products (Lou & Yuan, 2019). Extant literature has focused on influencer endorsements in users’ willingness to pay for physical goods such as cosmetics and clothes. An empirical study by Lou, Tan and Chen (2019) suggests that influencer-generated content on product sharing may raise followers’ cult-like appreciation. This is because digital influencers make such content look more like organic information sharing by blending it with their online personae and lifestyles (Evans et al., 2017). Recently, digital influencers have been hired to endorse knowledge products (Chen et at., 2022). Knowledge products are different from traditional physical ones as their quality is more difficult to assess before consumption (Rusho & Raban, 2020). Consequently, users may rely heavily on the opinions of those who have purchased the products. The purchase decision-making of knowledge products requires that users perceive fair expected value from heuristics due to the quality uncertainty of information goods (Raban, 2007). However, how influencer endorsements are used as heuristics to explain users’ willingness to pay for knowledge products remains unknown. Moreover, there is some uncertainty in practice about the actual impact of digital influencers who endorse knowledge products. On the contrary, some scholars reported that influencer endorsements in knowledge products were not very effective as users might activate and carry out strategies designed to defend against influencers’ persuasive messages (Lou et al., 2019). Hence, it is necessary to examine how influencer endorsements influence users’ willingness to pay for knowledge products. For the reasons above, this chapter draws upon attachment theory and value transfer process to construct a theoretical model and empirically test the model using a scenario-based survey with 441 participants. The model examines whether perceived attractiveness of digital influencers can motivate followers’ expected value of and attachment to endorsed knowledge products, thereby increasing the probability of their willingness to pay. This chapter holds both theoretical and practical significance. On the theoretical front, the proposed model may offer a better understanding of influencer endorsements of knowledge products, which work as a heuristic for motivating users’ purchase decision-making. On the practical front, the chapter

379

380  The Elgar companion to information economics provides valuable guidelines to inform related online merchants on how to use influencer endorsements effectively in promoting the sale of knowledge products.

2.

RELATED LITERATURE AND THEORETICAL FOUNDATIONS

2.1

Willingness to Pay for Knowledge Products in Online Settings

In traditional online settings, users may receive external information to help them decide which products to buy. Prior studies have confirmed that recommender systems are significant in promoting user willingness to pay by assisting them in finding their favourite products from a large set of possible ones (Adomavicius et al., 2018; Stock, 2024). Recommendation systems can accurately estimate a user’s personal preferences for products that he or she has not yet purchased, consumed, or experienced through past purchase records. Regardless of the recommendation systems or other heuristics of the online environment, perceived value of the products is widely acknowledged as a salient factor of user willingness to pay (Raban, 2007; Raban & Ahituv, 2024). Perceived value refers to the subjective value of products revealed to a person prior to consumption (Raban, 2012). Hence, a user’s purchase intention is related to his or her value perception of a product or service. If the product or service can offer potential benefits and utility, the user recognizes its value and thus would want to purchase it (Rusho & Raban, 2021). Knowledge products are usually seen as experience goods whose value or some other attributes remain unknown until purchase (Raban, 2012). The monetization of knowledge products may depend on users’ value perception of sellers and/or creators who trade them online. Users are used to accessing knowledge-intensive content from the Internet as online public goods for free (Davidson, 2024). Hence, it takes much energy and time to shift users’ consumption habits of knowledge from a free model to a paid model (Wang et al., 2022). Moreover, knowledge products are highly customized and tend to lack “standards of practice with a broad application to products” (Fang et al., 2021, p. 1). Nowadays, there is a wide variety of knowledge products for sale online, including paid Q&A, paid lectures, consultation with a fee, live streaming and so on (Chen & Chua, 2020; Chen et al., 2022). The prevalence of knowledge products for sale reflects users’ limited abilities to identify high-value content from the abundance of information on the Internet (Raban & Włodarczyk, 2024). Users may not know whether a particular piece of online information is able to address their information needs or whether to rely on a certain information source to reduce a sense of information overload. To solve such concerns, some digital platforms have launched a premium service to stimulate users’ willingness to pay. For example, one of the most well-known Q&A websites in China, Zhihu, offers the Zhihu Live service, which allows contributors to hold live talks and collect entrance fees from participating listeners (Wang et al., 2022). In this chapter, the willingness to pay is defined as the extent to which a user would like to purchase endorsed knowledge products.

The role of influencer endorsements  381 2.2

Influencer Endorsements and Value Transfer Process

Digital influencers are generally online celebrities who have gained fame through the textual and visual narration of their interests, thoughts, and personal lifestyles. These non-traditional celebrities draw a niche group of people but exert a significant impact on such followers’ decision-making (Lou & Yuan, 2019). Given their persuasive effect, marketers and advertisers often hire digital influencers or collaborate with them to recommend their brands (Jiménez-Castillo & Sánchez-Fernández, 2019). Existing literature has documented how influencer endorsements contribute to the formation and development of individual attitudes and behaviours (Jiménez-Castillo & Sánchez-Fernández, 2019; Lou & Yuan, 2019; Wang et al., 2021). For example, one study on YouTube influencers showed that once they reviewed and recommended luxury brand products in vlogs, perception of and purchase intention for the endorsed brands were significantly enhanced among users who watched these vlogs compared with those who did not (Lee & Watkins, 2016). Some research focuses on the actual influence of digital influencers in light of some evidence that popularity does not necessarily indicate a higher impact on following. One experimental study of Instagram influencers with a large fan base found that endorsing a variety of product categories lowers the uniqueness of the advertised brands and dampens consumer attitudes toward them (De Veirman et al., 2017). Influencer marketing literature suggests that perceived attractiveness of influencers is important in shaping users’ purchase intentions. When users are attracted to an influencer – who works as a responsible third party that recommends and endorses a product or service (Lou & Yuan, 2019) – perceived attractiveness of the product or service may nudge purchase intention (Evans et al., 2017). For the purpose of this chapter, perceived attractiveness of influencers is defined as the extent to which digital influencers who endorse knowledge products appeal to followers. The literature on celebrity endorsement has provided a powerful explanation of the underlying relationship between endorser effects and audience purchase intention: a value-transfer approach (Hung et al., 2011). Specifically, celebrities promote brand purchase intention primarily through celebrity endorsers’ perceived attractiveness and expected value of the brand (Lou & Yuan, 2019; Ohanian, 1990). Users’ purchase decision-making process may benefit from opinions and suggestions from influential peers who share high-quality and credible information on social networks. Extant literature suggests that electronic word of mouth (eWOM) from digital influencers positively affects users’ perceived value of the endorsed products (Lou et al., 2019). In addition to shaping opinions, digital influencers interact with followers and facilitate them to form high-value expectations during the interaction process (Lou, 2022). In this chapter, the expected value of knowledge products is defined as the degree of potential benefit and utility of endorsed knowledge products. 2.3

Attachment Theory

The concept of “attachment” was initially proposed to explain how an individual was emotionally connected to another (Bowlby, 2012). In early related literature (e.g., Ainsworth et al., 1978; Holmes, 2000; Trinke & Batholomew, 1997), scholars hold that the desire to make an emotional attachment to a particular target serves as a basic need, including children’s attachment to their parents, the adulthood with kinships, friendships, and romance (Wan et

382  The Elgar companion to information economics al., 2017). Some scholars argue that attachment may lead to motivational and behavioural outcomes (e.g., Aron et al., 2004; Feeney & Noller, 1996), such as the desire to maintain physical proximity and the willingness to invest cognitive and financial resources (Park et al., 2010). As a result, an individual attached to a particular target will be dedicated to it and strive to protect and preserve its interactions (Fedorikhin et al., 2008). According to related literature, an individual’s attachment to a particular target can lead to “strong motivational and behavioral outcomes, such as the desire to maintain physical proximity and the willingness to defend and invest cognitive and financial resources in the attachment target” (Wan et al., 2017, p. 839). For example, Ranganathan (2018, p. 641) suggested that “when someone becomes attached to an object such as a car or piece of furniture, they are more likely to handle it with care, repair it when it breaks, and postpone its replacement”. Prior research suggests that if a consumer has an attachment to a brand (including a product or service brand), such an attachment affects the consumer’s behaviour toward the brand. For instance, someone may keep brand loyalty and continuous purchasing intentions (Ranganathan, 2018). Wan et al. (2017) find that users’ monetary donation to content creators is associated positively with their attachment to their favourite online content creators. Similarly, the literature from marketing research finds that consumers’ attachment to a specific product/service positively affects their purchasing intention (Park et al., 2010). In this chapter, followers’ attachment to knowledge products is defined as the degree of followers’ affective connection to and caring for endorsed knowledge products. In summary, Table 19.1 presents the conceptual definitions of constructs in this chapter. Table 19.1

Conceptual definitions of constructs in the chapter

Construct

Definition

Reference

Willingness to pay

The extent to which a user would like to purchase endorsed

Lopes & Galletta, 2006

knowledge products Perceived attractiveness of

The extent to which digital influencers who endorsed

influencers

knowledge products appeal to followers

Montoya & Horton, 2014

Expected value of knowledge

The degree of potential benefit and utility of endorsed

Jiménez-Castillo &

products

knowledge products

Sánchez-Fernández, 2019

Attachment to knowledge products

The degree of followers’ affective connection and caring for

Ranganathan, 2018

endorsed knowledge products

3.

HYPOTHESES DEVELOPMENT AND RESEARCH MODEL

Extant literature suggests that users’ attitudes and opinions toward a target may lead to the development of their expectations (Zeithaml et al., 1993) and the formation of perceived value of the target (Weiss et al., 2008). Such phenomena also happen when users purchase a product or service. They tend to consider the trade-off between potential benefits and costs during the decision-making process, thereby forming a value expectation (Balasubramanian & Mahajan, 2001). In the context of influencer endorsements, digital influencers’ attraction may have an impact on their followers’ overall value assessment of the product (Lou & Yuan, 2019). For this research, perceived attractiveness of influencers also contributes to the formation of followers’ value expectations of knowledge products that the influencers endorsed. Hence, the first hypothesis is as follows.

The role of influencer endorsements  383 H1: Perceived attractiveness of digital influencers is positively associated with followers’ expected value of knowledge products. Users’ perceived attractiveness reflects their overall evaluation of other entities’ abilities and attributes (Montoya & Horton, 2014). It is “widely used to explain why people maintain a social relationship with other entities, and it can be further employed to predict people's attitudes and behavior” (Ladeira et al., 2018, p. 111). In terms of maintaining a social relationship, the literature has revealed that the users’ perceived attractiveness of a target enables them to develop a bond-based and stable relationship with the target (Harris et al., 2003). Hence, the second hypothesis is as follows. H2: Perceived attractiveness of digital influencers is positively associated with followers’ attachment to knowledge products. Existing literature has acknowledged the role of digital influencers in influencing users’ purchase intentions (e.g., Evans et al., 2017; Jiménez-Castillo & Sánchez-Fernández, 2019; Lou & Yuan, 2019). When they are attracted by an influencer – who works as a responsible third party that recommends and endorses a product or service – perceived attractiveness will nudge the consumers’ purchase intention (Lou et al., 2019). According to the value transfer theory (Ohanian, 1990), endorsements from digital influencers may make users perceive the value of the products or services they transited. Hence, the third hypothesis is as follows. H3: Perceived attractiveness of digital influencers is positively associated with followers’ willingness to pay for knowledge products. The existing literature has widely recognized the positive impact of perceived value of a product or service on users’ willingness to pay (Raban, 2007; Raban, 2012; Rusho & Raban, 2020). Some research indicates that perceived value and satisfaction overlap, which are both significant indicators of users’ loyalty (Mencarelli & Lombart, 2017). Scholars have found that users’ expected value of a product and service is positively related to their purchase intention in various online marketplaces (Ponte et al., 2015; Wu et al., 2014). In the context of influencer endorsements, if a follower’s expected value of knowledge products is high, it is reasonable to suggest that their willingness to pay will be high. In practice, online merchants of knowledge products have hired digital influencers and expected that they could make target followers perceive the sufficient potential value of the endorsed products. Hence, the fourth hypothesis is as follows. Followers’ expected value of knowledge products is positively associated with their H4: willingness to pay. In the marketing literature, several studies have theorized the relationship between consumers and a product or service as an emotional attachment and examined its impact on consumers’ positive online reviews, satisfaction with the brand, brand loyalty, and purchase decision-making (Vlachos et al., 2010). In brief, attachment is a significant factor in motivating consumers’ commitment, loyalty, and prosocial behaviours. When they are emotionally attached to the target objects, they are likely to spend their own resources, such as time,

384  The Elgar companion to information economics energy, and money, on them. In the information science community, the study by Choi (2013) suggests that users’ attachment to information systems positively impacts their intention to engage in information system-enabled communities. Hence, the following two hypotheses are as follows. H5: Followers’ attachment to knowledge products is positively associated with their expected value of knowledge products. H6: Followers’ attachment to knowledge products is positively associated with their willingness to pay. In summary, Figure 19.1 presents the research model explaining the role of influencer endorsements in users’ willingness to pay for knowledge products.

Figure 19.1

4.

Research model

RESEARCH METHODOLOGY

4.1 Measures Measures of the constructs were adapted from the extant literature. We also slightly modified some words and expressions to make the survey items suitable for the research context. Specifically, items of willingness to pay were adapted from Lopes and Galletta (2006); items of perceived attractiveness of influencers were from Ohanian (1990) and Tidwell et al. (1996); items of the expected value of knowledge products were from Jiménez-Castillo and Sánchez-Fernández (2019), and items of attachment to knowledge products are from Wan et al. (2017). Measures of other constructs are reported in Table 19.2. A five-point Likert scale was employed to capture respondents’ agreement with the content of the survey items, ranging from “strongly disagree” to “strongly agree”.

The role of influencer endorsements  385 Table 19.2

Constructs and survey items

Construct

Survey items

Reference

Willingness to pay

WTP1: I intend to pay for knowledge products endorsed by digital

Lopes & Galletta, 2006

(WTP)

influencers that I follow

 

WTP2: I will feel unhappy if I pay for knowledge products endorsed by

 

digital influencers that I follow (R).  

WTP3: In the future, I will likely pay for knowledge products endorsed by

 

digital influencers that I follow. Perceived attractiveness

PA1: I feel that digital influencers who endorse knowledge products are

Ohanian, 1990; Tidwell et

of influencers (PA)

attractive to me.

al., 1996

PA2: I feel that digital influencers who endorse knowledge products can draw my attention. PA3: I feel digital influencers who endorsed knowledge products catch my eyes. Expected value of

EV1: Knowledge products endorsed by digital influencers I follow have an Jiménez-Castillo &

knowledge products

acceptable quality standard.

Sánchez-Fernández, 2019

(EV) EV2: Knowledge products endorsed by digital influencers I follow are well made. EV3: Knowledge products endorsed by digital influencers I follow are positively valued. Attachment to

AT1: I feel connected personally with the knowledge products endorsed by Wan et al., 2017

knowledge products

digital influencers that I follow.

(AT) AT2: I have a close link to the knowledge products endorsed by digital influencers that I follow. AT3: To some extent, my opinion on the knowledge products endorsed by digital influencers that I follow can be shaped.

Note: (R) means a reversed-coded item.

4.2

Data Collection

Because of the lack of a sampling frame that meets these requirements, a non-probabilistic convenience sampling procedure was deemed the most suitable sampling technique for the data collection (Jiménez-Castillo & Sánchez-Fernández, 2019). We used an online survey to collect data. The online survey was compiled and implemented in a professional Chinese survey platform (www​.wjx​.cn). It starts with a brief introduction to the research purpose. The questionnaire was prefaced with the definition of digital influencers who endorse knowledge products and a note asking respondents to answer the questions based on their most frequently followed influencers. The respondent requirements were included at the beginning of the survey using filtering questions. Failing to meet the criteria implied not continuing with the remaining parts of the online survey questions. Eligible respondents of the survey satisfied three criteria. First, they were regular Internet users over 21 years old. Second, they were active followers of digital influencers who endorsed knowledge products. Third, they had paid for knowledge products endorsed by such digital influencers. The questionnaire consisted of three parts. The first part involves the survey respondents’ demographic information, including gender, age, education, occupation, monthly spending, and experience in payment for online knowledge products. The second part is the main body of the questionnaire. The respondents

386  The Elgar companion to information economics are required to fill in their perceptions of the survey items. The third part is a plain-text table where respondents may write down their feedback on the survey. 4.3

Analytical Approaches and Results

Descriptive results were generated using the descriptive statistics functions of SPSS 21. The valid sample size was 441. We have eliminated responses from those without experience using online reviews as a shopping reference. We also excluded responses whose answer time was too short or were obviously not logical. Table 19.3 summarizes the demographic information of valid respondents. Table 19.4 presents the descriptive statistics of the constructs. Structural equation modelling (SEM) technique was used in Mplus 8.6 software. The measurement model is assessed as follows. First, the reliability and convergent validity of constructs were evaluated using confirmation factor analysis (CFA). CFA results suggested that loadings for items of each construct were above the threshold of 0.7 (Chen et al., 2018; Fornell & Larcker, 1981). Convergent validity was assessed by considering the standardized loading of all the constructs as well as the average variance extracted (AVE) (see Table 19.5). The AVE for each latent construct was more significant than 0.50 (Fornell & Larcker, 1981). Second, the internal consistency of constructs was evaluated using Cronbach’s α and composite reliability (CR) (see Table 19.5). All four constructs exceeded the recommended threshold of 0.70 (Fornell & Larcker, 1981). Composite reliability of each construct was higher than the suggested cut-off of 0.70 (Fornell & Larcker, 1981). Third, to assess discriminant validity, we used the criterion that the AVE of each construct exceeded its correlations with other constructs. As shown in Table 19.5, the AVE square roots were greater than the correlations, indicating satisfactory discriminant validity of the constructs (Fornell & Larcker, 1981). In addition, past research suggests that users’ willingness to pay in an online environment is also contingent on demographics. Therefore, we include them as control variables in the follow-up data analysis to strengthen the robustness of the results. Table 19.3

Demographic information of survey respondents

Demography

Category

Frequency (%)

Gender

Male

175 (42.58%)

Female

266 (64.72%)

21–30

202 (49.15%)

31–40

184 (44.77%)

41–50

20 (4.86%)

51–60

4 (0.97%)

>60

1 (0.25%)

Below Bachelor

12 (2.92%)

Bachelor

311 (75.67%)

Master or above

88 (21.41%)

Full-time student

65 (15.82%)

Employed

339 (82.48%)

Unemployed

1 (0.25%)

Retired

1 (0.25%)

Others

5 (1.20%)

Age

Education level

Occupation

The role of influencer endorsements  387 Demography

Category

Frequency (%)

Yearly spending on knowledge products

¥1k and below

32 (7.79%)

¥1001~ ¥3k

132 (32.12%)

¥3001~ ¥5k

222 (54.01%)

¥5001 and above

25 (6.08%)

5 years

38 (9.25%)

Consumption experience of knowledge products

Note: ¥1k ≈ ​ ​160USD.

Regarding the structural model, the goodness-of-fit statistics suggested that the model fit was satisfactory (χ2(55)=117.64, p