Behavioral Finance: A Novel Approach 2020948372, 9789811229244, 9789811229251, 9789811229268

Citation preview

12071_9789811229244_tp.indd 1

4/11/20 8:12 AM

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

12071_9789811229244_tp.indd 2

4/11/20 8:12 AM

Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress Control Number: 2020948372 British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.

BEHAVIORAL FINANCE A Novel Approach Copyright © 2021 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.









ISBN 978-981-122-924-4 (hardcover) ISBN 978-981-122-925-1 (ebook for institutions) ISBN 978-981-122-926-8 (ebook for individuals)

For any available supplementary material, please visit https://www.worldscientific.com/worldscibooks/10.1142/12071#t=suppl Desk Editors: Balasubramanian Shanmugam/Karimah Samsudin

Typeset by Stallion Press Email: [email protected] Printed in Singapore

Balasubramanian - 12071 - Behavioral Finance.indd 1

20/10/2020 4:43:55 pm

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

To my loved one, Laurie, Dana and Irit

v

b3964-fm

page v

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 fmatter

Preface This book is an eclectic collection of chapters written mostly by the participants of The Third Israel Conferences in Behavioral Finance held in 2019 at the Tel Aviv-Yaffo Academic College. It follows its predecessors: Behavioral Finance: The Coming of Age and Behavioral Finance: Where do Biases Come From? which were published following the First and Second Israel Conferences in Behavioral Finance, respectively. By now, true to the well-known law of small numbers, we can detect a trend in the style and content of the chapters of these three books: They try to provide a mix of chapters that attempt to challenge the main stream behavioral finance along with traditional behavioral ideas. There is tension between behavioral finance and the reigning finance paradigm as the former challenges the efficient market hypothesis and suggests that markets might be overreacting, too volatile to be representative of the real value of the market, and suffer from other inconsistencies and biases of the investors. This tension is conspicuously resurfacing at the time of writing this Preface, amid the coronavirus (COVID-19) crisis. During the first days of the pandemic (around February 12, 2020, few weeks before it was officially declared a pandemic by the WHO), a group of our school’s students was preparing to go to a study-tour in mid-March in Europe. They voiced concern about their safety, and we had to decide how risky such a trip would be. We chose to answer this by financial reverse engineering and infer the risk of the new virus from the behavior of the stock market and from travel insurance prices (what could be better in assessing risk than insurance firms?). We found out that

vii

page vii

November 4, 2020

viii

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

Behavioral Finance: A Novel Approach

none of these institutions showed any significant sign of an approaching problem.1 But, the students were not convinced, and we canceled the tour. Why were markets and some professionals sluggish to react and can we consider it as an example of one of their flaws?2 Markets might have been taking some cues from the economic leaders. On February 18, 2020, the Acting Chairman of the Council of Economic Advisers, Tomas Philipson, told reporters: “I don’t think corona is as big a threat as people make it out to be,” and suggested the virus would not be nearly as bad as a normal flu season.3 Are these examples indications of under reactions by the stock market and of these insurance companies? It is hard to say, but they highlight the need for behavioral finance researchers to examine the extent to which the stock market and professionals’ behavior truly represent the underlying values of assets and the outcome of events. They could be the best predictors available, but we, researchers in behavioral finance, should monitor them carefully. The book consists of three parts. In Part I, we present chapters that involve the behavior of policymakers and politics. Chapter 1 by Calipha and Venezia, Prediction Markets: Do They Predict the Polls or the Election Results? The Case of the Israeli Elections in April 2019, investigates the merits of prediction markets.4 The authors show that the market studied here, although inefficient in

1

The US stock markets did not show any decline, and travel insurance prices quoted for travel between Israel and Europe on March 2, 2020, for travel in midMarch were the same as was those quoted on January 15, 2020. 2 The (US) stock market showed signs of trouble only on February 19, 2020, which in hindsight were quite late. 3 However, it is possible that Philipson reached his conclusion by reverse engineering from the markets. But economists were not the only to read the situation wrong. On March 13, 2020, Nobel laureate and Stanford Professor of Biology Michael Levitt assured Israelis that the virus is on a downturn and said that he will be surprised if more than 10 people will die from it (currently, about a month later, the number of deaths is above 200). 4 In Israel ,this issue has lately become quite pertinent since in a span of less than a year (from April 19, 2019 to March 2, 2020) the country underwent three elections as the first two elections were not decisive and did not allow the formation of a government.

page viii

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

Preface

b3964-fm

page ix

ix

the sense of exhibiting correlated returns, and being heavily influenced by polls, provided better forecasts than the polls at a very low cost. Such markets therefore seem worthwhile. The following chapters in this part examine the behavior of economic agents facing changes in regulation. Chapter 2 by McTier and Zuta, Influential CEO and Board Behavior in Reaction to a Regulatory Reform: A Quasi-Natural Experiment, explore how CEOs alter their behavior faced with new regulation, thus circumventing the intent of a regulation. It serves as a reminder to regulators to pay more attention to the behavioral aspects of the policies they try to implement. In Chapter 3, Aiming for the Real-Estate Market But Hitting the Stock Market — An Event Study Analysis of Israeli Mortgage Reforms, Lahav, Arbel and Mizrahi, provide another example of how the failure to consider behavioral aspects could lead regulators and policy makers to undesirable results. The authors show that a new regulation introduced in Israel to lower housing prices not only failed to achieve its goal but also influenced the market value of the realestate companies traded on the Tel-Aviv Stock Exchange. Plazzi, Torous and Yilmazz examine the efficacy of corporate governance by examining evidence from the AGR governance rating in Chapter 4. They show that good corporate governance is handsomely rewarded by the capital markets. Does this imply that regulators need not nudge firms to implement good corporate governance measures since firms would anyway adopt such measures in order to optimize their performance and regulation, as we have seen from the previous two chapters, could lead to unintended consequences? The next chapter, Chapter 5, by Mugerman, Nadiv and Ofir, Are Courts Biased? The Anchoring Heuristic and Judicial Decisions in Personal Bankruptcy Proceedings, implies that opposite to its predecessors, some regulations of the courts in personal bankruptcy proceedings could be worthwhile. The authors show that the current procedures of presenting the evidence to judges in these courts may cause them to fall prey to the anchoring bias, which in turn could yield grievous errors. The authors suggest a better control of these procedures. The second part of the book covers topics on capital asset markets and methodological issues. In Chapter 6, Kudryavtsev provides new evidence about the effects of mood and limited attention on reactions of investors to new information. The following two chapters examine the effects of behavior on IPO pricing. Limited attention plays a

November 4, 2020

x

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

Behavioral Finance: A Novel Approach

role also in Chapter 7 by Siev and Rothman. They suggest that this bias can explain the phenomenon of investors recognizing a US$500 million market value of a firm as a confidence threshold when investing in newly issued pharmaceutical companies. The authors postulate that firms valued above this amount attract more attention and gain greater investor confidence than do firms below this threshold. Lower-valued firms’ shares can be considered “lottery stocks,” as their IPOs ignite a period of enthusiasm until the quiet period ends, where investors’ attention to such firms gradually diminishes, and their focus moves on to their next potential lottery-like opportunity. In Chapter 8, Behavioral Characteristics of IPO Underpricing, Michel, Oded and Shaked also explore the behavior of IPOs. They discover a new bias in the pricing of IPOs. They suggest that underpricing is negatively related to the public float (the fraction of the firm sold to the public) and propose that firms do not optimize their underpricing but allocate a fixed amount for that purpose. Chapters 9, 10 and 11 examine the effects of investors’ characteristics on risk taking, investment behavior and pension decisions, respectively. In Chapter 9, Influence of Religion and Social Attitudes in Stock Market Participation, Zhou, Yu and Zhou investigate the relationship between religion and stock market participation. Based on a sample from the Chinese population, they find that compared with non-religious households, Buddhists are more likely to invest in stocks, while the opposite is true for Muslims. However, there are no significant differences in stock market participation between non-religious households and other religious households, such as Taoists, Protestants and Catholics. Furthermore, religiosity is negatively associated with the propensity to invest in stocks. Jansson, Hemlin, Sonsino and Tr¨onnberg’s Chapter 10, Investment Beliefs and Portfolio Risk-Taking — A Comparison between Industry Professionals and Non-Professionals, deals with comparing the investment beliefs of professional and non-professional Swedish investors. Investment beliefs are increasingly implemented in the asset management industry and it has been suggested that the disclosure of clear beliefs may alleviate tensions between stakeholders and investment managers for the industry’s benefit. The current comparisons expose significant differences between the beliefs of the two groups. The risk-taking decisions of non-professionals respond to their selfconfidence and affective state, while the professionals only respond

page x

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

Preface

b3964-fm

page xi

xi

to their investment beliefs and personal risk-attitude. Yaakobi and Kallir’s Chapter 11, Boys Don’t Cry? The Emotional Effects of Poor Financial Savings Decisions among Males and Females, empirically examines the psychological effects of suboptimal financial saving decisions. The authors present empirical evidence showing that the pension and provident suboptimal decisions are associated with severe negative thoughts. Moreover, these emotions are stronger for males than females. Part II concludes with two methodological papers. Chapter 12, by Sonsino, Lahav and Levkowitz, Separating Accuracy from Forecast Certainty: A Modified Miscalibration Measure, in which they advance a new method to forecast uncertainty and demonstrate its efficacy. Chapter 13, Optimal Contracts with IntraPrincipal Conflicts and the Ubiquity of Earnings Management, by Chidambaran, Sarath and Zheng deals with conflicts of interest between managers and shareholders and how to design contracts that would alter managers’ behavior so as to minimize or eliminate such conflicts. A perfect alignment between the interests of these groups is impossible, but the authors show that despite their usual bad reputation, contracts involving earnings management could be designed that help to make both long and short-term shareholders better-off on average. In Part III, we consider some important behavioral aspects of pensions and savings. In Chapters 14 and 15, the form of pension withdrawals (annuities vs. lump sum withdrawals) is investigated, and the authors of these chapters explore to what extent behavioral and/or psychological factors influence such decisions at the expense of pure economic reasoning. Selivansky, Leiser and Spivak, in Chapter 14, examine retirement savings withdrawal preferences in Israel and the psychological and demographic factors that affect them. Unlike findings from other countries, most of the participants in their sample preferred annuity over a lump sum withdrawal. Lower confidence in the system and higher personal self-confidence were significantly associated with lump sum withdrawals preference. The authors suggest that the preference for an annuity stems from a default setting: pension savings in Israel have long been defined benefits (DB), that is, a fixed pension which to the recipient is more similar to an annuitized defined contribution (DC) system than to a lump sum withdrawal. Hurwitz and Sade’s Chapter 15, Smokers’ Life Expectancy and Annuitization Decisions, studies the life expectancy

November 4, 2020

xii

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

Behavioral Finance: A Novel Approach

perceptions of smokers. They show that smoking does not significantly affect health perceptions. They find that even though smokers’ life expectancy is lower than that of non-smokers by more than nine years and although the insurance pricing mechanism means that smokers would be offered the same annuity as nonsmokers (all else equal), smokers do not prefer the lump sum option. This finding is consistent with the hypothesis that they are unrealistically optimistic about their own life expectancy. Itzhak Venezia

page xii

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 fmatter

About the Editor Itzhak Venezia is a Professor of Finance at the Academic College of Tel Aviv-Yaffo, and the Chairman of the MBA program and of the Finance Area for MBA studies. He holds the Sanger Chair of Banking and Risk Management (emeritus) at the Hebrew University, Jerusalem, Israel, where he taught prior to assuming his current position. Professor Venezia is the editor-in-chief of the Lecture Notes Series in Finance, and the editor of the books Behavioral Finance: Where do Investors Biases Come From? and Behavioral Finance: The Coming of Age. He also authored the book Lecture Notes in Behavioral Finance. Professor Venezia has published numerous papers in leading journals such as the Journal of Finance, Journal of Economic Theory, Journal of Banking and Finance, Management Science, and is the joint-editor of the book Bridging the GAAP: Recent Advances in Accounting and Finance. He has taught as a visiting professor at Yale University, the University of California, Los Angeles, Rutgers University, and Northwestern University. Professor Venezia’s research currently concentrates on Behavioral Finance, where he contributes profoundly to the better understanding of the disposition effect, herding, the differences in biases between amateurs and professionals, and other issues. He holds a PhD from the University of California, Berkeley.

xiii

page xiii

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 fmatter

List of Contributors Abigail Hurwitz, The Hebrew University of Jerusalem, Israel, The College of Management, Academic Studies, Israel and The Wharton School of the University of Pennsylvania. Alberto Plazzi, University of Lugano, Lugano, Switzerland and Swiss Finance Institute, Switzerland. Aliza Mizrahi, Ben-Gurion University of the Negev, Be’er-Sheva, Israel. Allen Michel, Boston University, Boston, Massachusetts, USA. Amir Levkowitz, Ben-Gurion University of the Negev, Be’er-Sheva, Israel. Andrey Kudryavtsev, Yezreel Valley Academic College, Emek Yezreel, Israel. Avia Spivak, Ben-Gurion University of the Negev, Be’er-Sheva, Israel. Bharat Sarath, Rutgers University, New Jersey, USA. Brian McTier, The University of Texas at San Antonio, San Antonio, Texas, USA. Carl-Christian Tr¨ onnberg, Gothenburg Research Institute (GRI), University of Gothenburg, G¨oteborg, Sweden. David Leiser, Ben-Gurion University of the Negev, Be’er-Sheva, Israel. Doron Sonsino, College of Law and Business (CLB), Ramat-Gan, Israel and Ben-Gurion University of the Negev, Be’er-Sheva, Israel. xv

page xv

November 4, 2020

xvi

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

Behavioral Finance: A Novel Approach

Erez Yaakobi, Ono Academic College, Kiryat Ono, Israel. Ido Kallir, Ono Academic College, Kiryat Ono, Israel. Israel Shaked, Boston University, Boston, Massachusetts, USA. Itzhak Venezia, Tel Aviv-Yaffo Academic College and the Hebrew University of Jerusalem, Israel. Jacob Oded, Tel Aviv University, Tel Aviv, Israel. Jinwen Yu, Wuhan University, Wuhan, China. Lingyi Zheng, The Hong Kong Polytechnic University, Hong Kong. Magnus Jansson, Gothenburg Research Institute (GRI), University of Gothenburg, Gothenburg, Sweden. Moran Ofir, Interdisciplinary Center (IDC), Herzliya, Israel. N. K. Chidambaran, Fordham University, New York, USA. Neta Nadiv, Interdisciplinary Center (IDC), Herzliya, Israel. Omer Selivansky, Israel Democracy Institute, Jerusalem, Israel. Orly Sade, The Hebrew University of Jerusalem, Israel. Rachel Calipha, The Academic College of Tel Aviv-Yaffo, Israel. Sara Arbel, Ben-Gurion University of the Negev, Be’er-Sheva, Israel. Shlomith D. Zuta, The Academic College of Tel Aviv-Yaffo, Tel Aviv, Israel. Smadar Siev, Ono Academic College, Haifa, Israel. Sven Hemlin, Gothenburg Research Institute (GRI), University of Gothenburg, Gothenburg, Sweden. Tiran Rothman, WIZO Academic College, Haifa, Israel. Umit Yilmaz, University of Lugano and Swiss Finance Institute, Lugano, Switzerland. Walter Torous, MIT, Cambridge, Massachusetts, USA.

page xvi

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

List of Contributors

page xvii

xvii

Yang Zhou, Wuhan University, Wuhan, China. Yaron Lahav, Ben-Gurion University of the Negev, Be’er-Sheva, Israel. Yevgeny Mugerman, Bar-Ilan University, Ramat Gan, Israel. Zhiping Zhou, Tongji University, Shanghai, China.

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 fmatter

Acknowledgments The financial support of the Tel Aviv Yaffo Academic College in funding the Third Israel Conference in Behavioral Finance is gratefully acknowledged. I also thank Professor Shlomo Biderman, President and Professor Israel Borovich, Dean of the School of Business and Economics, at the Tel Aviv Yaffo Academic College for their encouragement and support. Itzhak Venezia Tel Aviv-Yaffo Academic College, Israel May 2020

xix

page xix

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

page xxi

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 fmatter

Contents Preface About the Editor List of Contributors Acknowledgments

Part I. Chapter 1.

vii xiii xv xix

Behavioral Aspects of Policy Making Prediction Markets: Do They Predict the Polls or the Election Results? The Case of the Israeli Elections in April 2019

1

3

Rachel Calipha and Itzhak Venezia Chapter 2.

Influential CEO and Board Behavior in Reaction to a Regulatory Reform: A Quasi-Natural Experiment

19

Brian McTier and Shlomith D. Zuta Chapter 3.

Aiming for the Real-Estate Market But Hitting the Stock Market — An Event Study Analysis of Israeli Mortgage Reforms

51

Yaron Lahav, Sara Arbel and Aliza Mizrahi Chapter 4.

What You See Is What You Get But Do Investors Reward Good Corporate Governance When They See It? Alberto Plazzi, Walter Torous and Umit Yilmaz xxi

73

November 4, 2020

11:51

xxii

Chapter 5.

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

page xxii

Behavioral Finance: A Novel Approach

Are Courts Biased? The Anchoring Heuristic and Judicial Decisions in Personal Bankruptcy Proceedings

99

Yevgeny Mugerman, Neta Nadiv and Moran Ofir

Part II. Chapter 6.

Investor Behavior and Methodological Novelties Psychological Aspects of Stock Price Drifts Following Analyst Recommendation Revisions

129

131

Andrey Kudryavtsev Chapter 7.

The Critical Impact of Firms’ Market Value on Investor Behavior Following Pharmaceutical IPOs

155

Smadar Siev and Tiran Rothman Chapter 8.

Behavioral Characteristics of IPO Underpricing

179

Allen Michel, Jacob Oded and Israel Shaked Chapter 9.

Influence of Religion and Social Attitudes in Stock Market Participation

209

Yang Zhou, Jinwen Yu and Zhiping Zhou Chapter 10.

Investment Beliefs and Portfolio Risk-Taking — A Comparison between Industry Professionals and Non-Professionals Magnus Jansson, Sven Hemlin, Doron Sonsino and Carl-Christian Tr¨ onnberg

239

November 4, 2020

11:51

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-fm

Contents

Chapter 11.

Boys Don’t Cry? The Emotional Effects of Poor Financial Savings Decisions among Males and Females

page xxiii

xxiii

267

Erez Yaakobi and Ido Kallir Chapter 12.

Separating Accuracy from Forecast Certainty: A Modified Miscalibration Measure

283

Doron Sonsino, Yaron Lahav and Amir Levkowitz Chapter 13.

Optimal Contracts with Intra-Principal Conflicts and the Ubiquity of Earnings Management

303

N. K. Chidambaran, Bharat Sarath and Lingyi Zheng

Part III. Chapter 14.

New Directions for Pensions and Retirement Decisions

323

Preferences for Annuities in Israel and Their Psychological Determinants

325

Omer Selivansky, David Leiser and Avia Spivak Chapter 15.

Smokers’ Life Expectancy and Annuitization Decisions Abigail Hurwitz and Orly Sade

349

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Part I

Behavioral Aspects of Policy Making

1

page 1

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0001

Chapter 1

Prediction Markets: Do They Predict the Polls or the Election Results? The Case of the Israeli Elections in April 2019 Rachel Calipha∗ and Itzhak Venezia†

Abstract Prediction markets have been found to provide relatively accurate and inexpensive means for forecasting various events including elections results. However, since in most political elections, there exist public polls that predict the results as well as betting houses that publish odds on the various candidates and since voters obtain information from these sources, the marginal contribution of prediction markets is uncertain and needs to be assessed empirically. In this chapter, we investigate the correlations between poll results and prediction markets results throughout the election campaign in Israel prior to the April 2019 elections and compare their accuracy. We show that although correlated and influenced by the polls, prediction markets have some value. Given their minimal costs of operation, such markets are worthwhile. Keywords: Prediction markets, political stock markets forecasting, polls, Israeli elections

∗ †

The Academic College of Tel Aviv-Yaffo, Israel; [email protected]. The Academic College of Tel Aviv-Yaffo, Israel; [email protected]. 3

page 3

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Behavioral Finance: A Novel Approach

4

1.

Introduction

Prediction markets have been suggested as an accurate and inexpensive method for predicting various unknowns. The premise of such markets is that they would be more accurate than polls or than asking people for their opinions because in such markets, participants have financial stakes at being right. The respondents to polls on the other hand do not stand to lose or gain anything from being wrong or right (except maybe the pride) and hence will be less motivated to think carefully and provide honest opinions.1 It is also well known that in previous elections in Israel, some respondents to polls have deliberately and strategically misrepresented their opinions.2 Whereas participants in prediction markets have more incentives to perform well than participants in polls, they usually are not a representative sample of likely voters.3 Thus, the balance of advantages and disadvantages of one method vs. the other is not a priori certain. The question of which method is superior, or whether it is optimal to use the combination of both methods, thus remains to be resolved empirically. Thus far, the jury is hung.4 In their study of several major polling organizations, Berg et al. (2008), found that prediction markets outperformed polls in 9 of 15 cases. They report an average poll error of 1.91% and lower prediction market 1

However, Hvide et al. (2019) claim that self-reporting (a method similar to polls) are often more advantageous than experiments with small stakes (a method similar to prediction markets). 2 Schneider (2019) reports in Globes (an important Israeli newspaper) that senior advisers to the Likud party told the media how down the stretch in the campaigns of 2013 and 2015 orders were given to supply the pollsters with false answers in order to arouse complacency in the opponents’ side. This was given as a possible explanation why in all the elections of the last decade polls down estimated the votes for the Likud. Polls preceding the September 2019 elections became more aware of this problem than before the April 2019, which possibly explains their better performance in the latter elections which took place September 17, 2019. 3 According to Berg et al. (2008), in the Iowa Election Markets the participants were over whelmingly male, well educated, had high income and young. 4 The division of opinion seems to be correlated with the main field of the researchers. As suggested by Dana et al. (2019), psychologists typically prefer forecasts based on self-reports, whereas economists are much more likely to base them on behavior (prediction markets).

page 4

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Prediction Markets: Do They Predict the Polls or the Election Results?

page 5

5

errors of 1.49% or 1.58% (according to the two measures of accuracy they used), thus indicating an edge for the prediction markets.5 Arrow et al. (2008, p. 878) were even stronger in favor of prediction markets and called “to clear away regulatory barriers that were never intended to inhibit socially productive innovation [prediction markets].” On the other hand, Atanasov et al. (2016, p. 1) suggest that “prediction polls with proper scoring feedback, collaboration features, and statistical aggregation are an attractive alternative to prediction markets for distilling the wisdom of crowds.” Dana et al. (2019) found evidence that markets outperform self-reports if the latter are just averaged. However, the reverse is true if the self-reports team-prediction polls when the self-reports’ forecasts were statistically aggregated using temporal decay, differential weighting based on past performance, and recalibration. The combination of both methods these authors suggest is even better. In the Israeli elections of April 2019, prediction markets were held side by side to public polls. The accuracy of these two methods is compared in this chapter, but this is not the main question we ask. Given that polls are a standard fixture of elections and that they are not likely to be replaced, the question we pose is whether the prediction markets provide any predictive power above the polls. Throughout the elections campaign, participants in the prediction markets get a great amount of information from polls published in the media. This strikes another puzzle: How well the markets would have performed in the absence of polls? And why would participants in such markets believe that they can outperform the polls; after all, the polls are conducted by professional statisticians who invest in their design and execution considerable amounts of money and have access to superior information sources than most participants in the prediction markets. The pollsters in Israel are known to be ethical 5

These results however are not overwhelmingly in favor of prediction markets. Assuming that the results in the various cases are independent, we calculated that the null hypothesis that the two methods are equally good cannot be rejected by a non-parametric test, since under these assumptions, the probability that markets outperform polls in 9 or more cases out of 15 is 15%. We could not calculate the significance of the difference between the percentage errors of the two methods since the authors provided only averages but not standard deviations.

November 4, 2020

6

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Behavioral Finance: A Novel Approach

and show no bias in favor of any politician (or party) and are highly motivated to perform well. The three largest TV stations in Israel, each employing a different polling firm, fiercely compete on their accuracy. Being more accurate is important to the TV station that employs the polling firm, since it provides the station with prestige and improves their ratings; the pollsters’ reputation and hence, their future jobs potential depend on accuracy as well.6 The pollsters are therefore not less motivated to perform well than the participants in prediction markets. Proponents of prediction markets often cite the potential of each of their participants to learn from the behavior of the other participants and to aggregate efficiently information from all sources. However, the participants in the markets usually have access to sources of information inferior to those of the pollsters and hence the rational participant does not have any reason to believe that either himself/herself or his/her competitors know any better than the polls or is able to process information better than the pollsters. Besides strategies of outwitting the other participants in the market, overconfidence bias, or some other behavioral biases, it is hard to imagine any reason why participants do not assume the polls to be their best estimates, but in such cases prices might not be efficient predictors of the voting results. Participants may not only rationally herd but also in this case the market might become inefficient (see, e.g., Bickhchandani et al., 1992; Froot et al., 1992). Participants may believe that pollsters just use simple averages of their findings in their forecasts, in which case, their forecasts would be inferior to those obtained by markets. Whereas the aforementioned generally is true for most days prior to the elections, in Israel, prediction markets have an advantage over polls in the last three days prior to elections day. By law, the media are forbidden to publish the results of polls that have been taken in these days. The media can publish the results of polls taken on elections day and on the three days before it, only after the closing of the ballot stations on elections day. Prediction

6 Mina Tsemach, the chief statistician of a veteran and respectable polling firm resigned following the poor performance of her firm’s predictions in the elections of April 2019. Her partner attributes their failings to deliberate false reports of Likud voters (Shalita, 2019).

page 6

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Prediction Markets: Do They Predict the Polls or the Election Results?

page 7

7

markets can provide information in the three days prior to the elections, which public polls are prohibited from providing. The chapter is constructed as follows. In Section 2, we explain how the prediction markets was constructed and operated. We also provide some institutional information about the Israeli elections process. In Section 3, we present tests of the efficiency of the prediction markets. In Section 4, we compare the accuracy of the prediction markets vs. the polls and in Section 5, we analyze the correlation between them. Concluding remarks are provided in the Section 6.

2. 2.1.

Methodology, Participants and Data Subjects

Around 129 MBA and BA students from Bar Ilan University and 93 from the Academic College of Tel Aviv-Yaffo registered for the market (90 women and 132 men) with an average age of about 30 years. Only about 100 of those registered traded more than once. 2.2.

The Market and Its Operation

Each participant received 60 ILS to participate in the market that operated continuously between February 18, 2019 to April 9, 2019 on a platform provided by the consulting firm “Darebiz.”7 At each point, the participants could either buy or short-sell futures of each of the 13 largest parties that ran for the Israeli parliament in the elections to be held on April 9, 2019 (the parties included in our market were the 13 largest ones: KacholLavan, Likud, Ha’avoda, YahadutHaTorah, Hadash-Taal, HayminHachadash, IhudHayamin, Meretz, Shas, Zehut, Kulanu, Israel Beytenu and Raam-Balad). The underlying asset in this market was the number of seats the party will receive in the “Knesset” (the Israeli parliament) out of the 120 seats available, and each seat is worth 1 ILS. The elections date served as the expiration date of the futures. At each point the trading participants could see on the screen of the trading platform the previous sell 7

This firm also paid the subjects for their participation and their profits from their trading.

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

page 8

Behavioral Finance: A Novel Approach

8

and buy prices, the changes from the previous trades and the result of the latest published poll. The trader could submit a buy/sell order and Darebiz’s (private) algorithm calculated an equilibrium price. In accuracy comparisons between the predictions of our market and those of the polls, we used the poll results provided by “Darebiz,” which were based on the latest results published each evening by one of the three largest TV stations operating in Israel who offered predictions every day throughout the campaign. We defined the predictions of our market, on each day of the elections campaign, to be the average prices paid for the futures of the party during the day. The profits/losses of each participant from each future held on expiration were the difference between the prices of the futures and the official number of seats the party obtained in the elections.8 3.

Tests of Market Efficiency

Although it is not our main goal in this chapter, we found it interesting to examine whether our prediction market is efficient.9 We chose to only test for correlation between returns (weak form efficiency) and decided that further tests (semi strong or strong efficiency tests) would be needed only if no correlation would be detected. Since correlation was detected, no further tests were needed. A finding of inefficiency would indicate a flaw in this market as past returns do not incorporate all the information. If such a market were to beat the polls this would constitute as an even worse sign for the polls, since it would imply that an efficient market could trample them even stronger. We first ran autocorrelation tests for the returns of each party separately. We ran for each of the 13 parties a regression of the form: Rt = a + bRt−1 8

(1)

The official results of the elections were declared only about a month after the elections because of the complex system of counting the votes and appeals of the results raised by some parties alleging foul play or mistakes in the counting. 9 In this section, efficiency is defined in the classical weak form of efficiency in finance (i.e., prices should reflect all information about previous returns). One could define, alternatively, the market as being efficient if it provides better estimates than its alternative (polls) at lower costs.

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

page 9

Prediction Markets: Do They Predict the Polls or the Election Results?

9

where the returns on the futures on day t are defined by: Rt = (Pt − Pt−1 )/Pt−1

(2)

and Pt denotes the average price of the future during day t.10 The results of running these regressions are provided in Table 1. We observe from Table 1 that for all parties the returns are negatively correlated indicating inefficiency. The negative correlation was not statistically significant for all parties. However, a non-parametric test would reject the null hypothesis that the market is efficient. Under this hypothesis there is a 50% chance that the slope of the regression is negative and hence, the chance that for all parties the slope of the regression is negative is (1/2)13 a probability well below the critical value of 0.001 needed to reject the null hypothesis. The returns on the two largest parties turned out significantly negative, indicating inefficiency even for the more active parties. Whereas the autocorrelation between returns turned out significant, its economic significance is unclear and the possibility of arbitrage to take advan tage of this inefficiency was not realistic. The R2 s were also small, indicating that the association between successive returns is quite Table 1:

The estimated parameters of Eq. (1)

Ihud Israel Parameter Hadash Ha’avoda Hayamin beytenu B N R2

Parameter

−0.09 42 0.01

Meretz

Ra’amBalad

Shas

0.06 49 0.00

−0.15 42 0.00

−0.07 49 0.01

B N R2 Note: ∗ p < 0.1, 10

−0.28∗ 42 0.08

−0.16 49 0.00

∗∗

p < 0.05 and

∗∗∗

Kachol Lavan Kulanu

Likud

0.14∗∗∗ −0.54∗∗∗ −0.26∗ −0.32∗∗∗ 47 42 49 43 0.18 0.29 0.07 0.18 Haymin Hacha dash −0.08 49 0.01

YahadutHa Torah −0.03 49 0.00

Zehut −0.05 ∗∗∗ 42 0.38

p < 0.01.

Qualitatively similar results to all our tests were obtained also when we used end of day prices.

November 4, 2020

10

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Behavioral Finance: A Novel Approach

scattered. When all parties were considered in one panel data regression, we also obtained statistically significant negative temporal correlation between the returns. All the aforementioned findings taken together imply that technically prices in the market did not reflect all information.

4.

The Accuracy of the Prediction Markets vs. the Polls

In column 1 of Table 2, we present the latest prices of the predictions market prior to the elections; in column 2 the latest predictions of the polls (which by law were taken three days before the elections) and in column 3 the official results of the elections are listed. The raw results are presented in Panel A and the precision measures in Panel B. The predictions market fared better than the polls on average. From Panel B we observe that the average absolute errors and the mean squared errors of the predictions market are 2.46 and 10.36, respectively, compared with 3.00 and 16.23 of the polls. The standard deviations of the prediction market were also lower. The average errors of the prediction market turned out positive, as the prices of this market were not calibrated to sum up to 120. Since the last polls were taken three days before the elections, whereas the market continued trading until the last day, it seems that we imposed a handicap on the polls compared to the market. We therefore also compared the accuracy of the latest polls to that of the market prices determined on that day (in column 6 of Panel A). We observe by comparisons of columns 2 and 6 of Panel B of Table 2, that the market prices three days prior to the elections also provided more accurate predictions on average than the polls (for instance an average absolute error of 2.65 for the market compared to 3 for the polls). The extent to which the aforementioned differences in accuracy are consequential is a matter of opinion and of the political reality. In this regard, we note that the differences in favor of the prediction market are higher for the bigger parties (about 4 seats difference for the largest party and 1.3 for the second largest). Such differences are

page 10

1

Predictions Market

2.46 10.36 3.327 0.384

Errors, Predictions Market

2

3

4

28 28 11 6 6 6 6 5 5 6 4 4 4

35 35 6 8 6 0 5 4 8 0 4 5 4

−2.99 −5.67 3.75 −1.35 0.55 6.39 1 1.53 −2.51 5.1 0.19 −1.01 −0.01

Polls

Trading Prices 3 Days before the Elections

3 16.23 4.192 -0.077

2.65 10.64 3.37 0.42

Errors, Polls 5 −7 −7 5 −2 0 6 1 1 −3 6 0 −1 0

Trading Prices 3 Days before the Elections 6

33.30 29.67 10.25 5.44 6.91 6.08 6.41 5.63 5.25 5.40 4.26 3.23 3.61

b3964-ch01

Panel B: Precision Measures Average Absolute Error Mean squared Error Std. Dev Average Error

Official Results

Behavioral Finance: A Novel Approach – 9in x 6in

Panel A: Summary Statistics KacholLavan 32.015 Likud 29.335 Havoda 9.75 YehadutHaTorah 6.65 Hadash-Taal 6.55 HayminHachadash 6.395 IhudHayamin 6 Meretz 5.525 Shas 5.495 Zehut 5.1 Kulanu 4.19 Israel Beytenu 3.995 3.995 Raam-Balad

Last Poll Predictions

10:10

Last Trading Price

November 4, 2020

Party Name

The predictions of the market and the polls and comparisons of their accuracy

Prediction Markets: Do They Predict the Polls or the Election Results?

Table 2:

11 page 11

November 4, 2020

12

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Behavioral Finance: A Novel Approach

quite material in the Israeli parliament and could mean the difference between deadlock or a big win (or loss) for one of the leading parties.

5.

The Correlation between the Prediction Markets Prices and the Polls

The participants of the prediction markets and the polls were predicting the same unknown, and if they used similar sources of information for their predictions this would suffice to create a correlation between them. This correlation would be magnified if market participants used the polls in their predictions. Since the market participants observe the polls but not vice versa, it is plausible that the polls would affect the markets, and that causality, if any, would run only in one direction. To what extent do market participants employ information other than the polls? Some indication can be glimpsed from intraday price variation in the prediction markets. During the day, no new polls results were provided by the TV stations and hence, variation of prices during the days could roughly imply that market participants received new information beside polls information.11 The lower the intraday price variation, the lower, we would suspect, the non-polls information the market participants received. A trend of the intraday standard deviation over time would then indicate whether the relative influence of the polls diminished or increased over time. To examine this trend, we regressed the intraday daily standard deviation on time. A negative slope of this regression would indicate that these standard deviations diminished as the time to elections neared. In Table 3, we present the average intraday standard deviation of prices for each of the parties and their evolutions over time, i.e., the slopes of the aforementioned regressions. We observe in this table that for all parties the standard deviations of intraday

11 Intraday variation could result from other reasons. However, if during the day there is no one uniform source of information for all market participants (such as polls) and each participant receives information from his/her source, this would give rise to an increased intraday variation.

page 12

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Prediction Markets: Do They Predict the Polls or the Election Results?

page 13

13

Table 3: Average daily standard deviation of prices and their evolutions over time

Party Hadash-Taal Havoda IhudHayamin Israel Beytenu KacholLavan Kulanu Likud Meretz Raam-Balad Shas HayminHachadash YehadutHaTorah Zehut

Change in Std. Dev. as the Elections were Nearing

Average Daily Standard Deviation of Prices

−0.002 −0.016 −0.005 −0.022 −0.012 −0.006 −0.003 −0.009 −0.007 −0.010 −0.014 −0.013 −0.006

0.198 0.273 0.092 0.348 0.501 0.252 0.507 0.220 0.169 0.179 0.275 0.257 0.218

prices diminished over time.12 Also, a panel random effect regression of standard deviation as a function of time provided a significant negative correlation of −0.01(p < 0.01, N = 626). These results are consistent with the hypothesis that as time to elections neared, the effect of the information unrelated to polls that the market participants received diminished. In what follows, we analyze the correlation between the market prices and the polls results. We ran quite a few regressions to study these correlations. However, because the results of all different regressions point to similar conclusions, we present the results of just one of them.13 In this example, we ran a panel data regression with random effect where the average price is the dependent variable, and the independent variables are the lagged average price and the polls forecast. The coefficients of these variables (0.78 and 0.22, 12

We can invoke an argument, same as in Table 1, that the fact that all slopes are negative implies a significant negative relationship between the variables even if each slope individually is not significantly negative. 13 The results of the other regressions could be obtained from the authors upon demand.

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Behavioral Finance: A Novel Approach

14

respectively) were positive and statistically significant (p < 0.01, N = 607). The correlation between the polls and market prices is striking, but not surprising since the market participants could observe (and probably did) the results of the polls. We thus also explored whether there was a Granger causality between these variables. We ran such causality tests for each party separately and for the whole sample. For the whole sample we tested the null hypotheses that all lags of polls do not cause prices, and vice versa. The null hypotheses were rejected by a Chi squared test (Chi2 = 26.63 for polls affecting prices and 21.69 for causality in the other direction). Whereas the hypothesis that polls affect prices is plausible, the reverse is not, and it may be due to using too many lags by the program and the only thin variation of the variables over time. When running the tests for each party separately, the results, presented in Table 4, were slightly different.We observe from Table 4 that when studied separately, there is Granger causality between the polls and market prices for six parties, which agrees with our intuition F Values for Granger causality tests

Table 4:

Party

N

F Values for Testing That Polls Granger Cause Prices

Hadash-Taal Havoda IhudHayamin Israel Beytenu KaholLavan Kulanu Likud Meretz Ra’am-Balad Shas HayminHachadash YehadutHaTorah Zehut

39 46 39 43 39 46 38 46 39 46 46 46 37

7.102∗∗∗ 1.066 6.850∗∗∗ 4.162∗∗ 0.997 8.084∗∗∗ 0.568 0.727 3.589∗∗ 0.227 2.699∗ 0.189 0.464

Note: ∗p < 0.1,

∗∗

p < 0.05,

∗∗∗

p < 0.01.

F Values for Testing That Prices Granger Cause Polls 0.33 0.9817 5.820∗∗∗ 1.85 0.373 1.251 1.148 0.847 0.176 1.154 1.438 0.096 5.922∗∗∗

page 14

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Prediction Markets: Do They Predict the Polls or the Election Results?

page 15

15

slightly more than the results from the causality tests applied to the whole population simultaneously. For the finale, we provide in Table 5, the data that explain the title of this chapter. We compare the closeness of the market prices to the polls vis-` a-vis their proximity to the official results of the elections. From Table 5, it follows that the average absolute difference between the market and the polls was 0.79, whereas the average absolute difference between the market and the official results was almost three times higher, 2.46. This result, together with the former evidence about the influence of the polls on the prediction market prices, raise the concern that the prediction market participants did not do much beyond imitating the polls. However, their work was not in vain since it had a positive marginal value as their accuracy was superior to that of the polls. Table 5: results

Proximity of market prices to the polls and to the official

Party Name (Arranged by Size)

KacholLavan Likud Havoda YehadutHaTorah Hadash-Taal HayminHachadash IhudHayamin Meretz Shas Zehut Kulanu Israel Beytenu Raam-Balad

Last Trade Price

32.01 29.33 9.75 6.65 6.55 6.39 6.00 5.53 5.49 5.10 4.19 3.99 3.99

Last Polls “Forecast”

28 28 11 6 6 6 6 5 5 6 4 4 4

Official (True) Results

35 35 6 8 6 0 5 4 8 0 4 5 4

Errors, Predictions Market

Difference between the Predictions and the Polls

PricesOfficial

Prices-Polls

−2.99 −5.67 3.75 −1.35 0.55 6.39 1.00 1.53 −2.51 5.10 0.19 −1.01 −0.01

4.01 1.33 −1.25 0.65 0.55 0.39 0.00 0.53 0.49 −0.90 0.19 −0.01 −0.01

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Finance: A Novel Approach

16

6.

b3964-ch01

Conclusions

The prediction markets we studied had some flaws. It showed weak market inefficiency, and the prices determined in this market were heavily influenced by the polls, and for many parties we could conclude they were Granger caused by the polls. Moreover, the predictions were on average closer to the polls than to the election results. However, the predictions were more accurate on average than the polls, and this superior accuracy was notable for the parties that mattered (in the government formation). Since running this market costs very little compared to the millions of dollars spent on polls, and since it adds value, it seems that adding such a market to polls is worthwhile. As this chapter is based just on one elections campaign, we look forward to more papers that would study the interaction of polls and prediction markets and would shed more light on the merits of these markets in national political campaigns. References Arrow, K. J., Forsythe, R., Gorham, M., Hahn, R., Hanson, R., Ledyard, J. O., and Neumann, G. R. (2008). The promise of prediction markets. Science, 320, 877–878. Atanasov, P., Rescober, P., Stone, E., Swift, S. A., Servan-Schreiber, E., Tetlock, P., and Mellers, B. (2016). Distilling the wisdom of crowds: Prediction markets vs. prediction polls. Management Science, 63 (3), 691–706. Berg, J., Forsythe, R., Nelson, F., and Rietz, T. (2008). Results from a dozen years of election futures markets research. Handbook of Experimental Economics Results, 1, 742–751. Bickhchandani, S. D., Hirshleifer, D., and Welch, I. (1992). A theory of fads, fashion, custom, and cultural changes as informational cascades. Journal of Political Economy, 100, 992–1026. Dana, J., Atanasov, P., Tetlock, P., and Mellers, B. (2019). Are markets more accurate than polls? The surprising informational value of “just asking”. Judgment and Decision Making, 14 (2), 135–147. Froot, K. A., Scharfstein, D., and Stein, J. C. (1992). Herd on the street: Informational inefficiencies in a market with short-term speculation. Journal of Finance, 47, 1461–1484.

page 16

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch01

Prediction Markets: Do They Predict the Polls or the Election Results?

page 17

17

Hvide, H. K., Lee, J., and Odean, T. (2019). Easy money, cheap talk, or spuds: Inducing risk aversion in economics experiments. Working Paper, The Peder Sather Center for Advanced Study at UC Berkeley. Schneider, T. (2019). A chronicle of non-reliability: How a gap of 36 mandates between the polls and the final election results was created. Globes, September 12(in Hebrew). Shalita, C. (2019). Why they lied only to the polls of channel 12? Today it is clear. Globes, September 12(in Hebrew).

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0002

Chapter 2

Influential CEO and Board Behavior in Reaction to a Regulatory Reform: A Quasi-Natural Experiment Brian McTier∗ and Shlomith D. Zuta†

Abstract We investigate the impact of a new regulation, Amendment 20, on executive compensation in Israel. The Amendment, effective as of 2013, is aimed at controlling rapidly increasing executive compensation, as well as the perceived disconnect between executive pay and firm’s financial performance. The concentrated ownership structure in Israel provides us with the opportunity to examine the value of control as manifested in the implementation of the Amendment. The analysis is conducted using hand-collected data on firms traded on Tel-Aviv Stock Exchange in a 2-year window around 2012, the year in which the Amendment was signed into law. We find evidence that chief executive officers (CEOs) classified as related parties in firm’s annual financial statements are able to impact the structure of their post-amendment compensation. Our evidence is consistent with CEOs who are related parties, compared to CEOs who are not related parties, influencing the design of their compensation so that its bonus component increases, leading to an increase in their total cash compensation. However, we are unable to find evidence that this increase stems from an increase in salary. In this way, companies with influential CEOs seem to shift the composition of pay ∗ †

The University of Texas at San Antonio, USA; [email protected]. The Academic College of Tel Aviv-Yaffo, Israel; [email protected].

19

page 19

November 4, 2020

20

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach from cash salaries to cash bonuses in response to the Amendment, visibly complying with the intent of the Amendment, however, leaving the CEO better off. We also find evidence for the ability of powerful chairmen to temper the bonus growth of non-influential CEOs. Last, we find no evidence that equity-based compensation increased, suggesting that the bonding of CEO pay to stock-market performance did not increase. Keywords: Compensation, compensation committee, control, board of directors, regulation

1.

Introduction

This chapter explores the impact of a new government regulation on executive compensation in Israel. The regulation, Amendment 20 to Israel Companies Law, was signed into law in 2012 and became effective in 2013. Amendment 20 followed a substantial increase in executive compensation in Israel in the years prior to its enactment. During these years it became evident that executive compensation was loosely tied, at best, to the financial performance of the firm, causing public outrage and media criticism and providing the impetus for regulation. Similar events took place throughout the world and led to a worldwide wave of regulation following the financial crisis of 2008. The most prominent example is the Dodd–Frank Wall Street Reform and Consumer Protection Act in the US, signed into law by President Obama in 2010. Similar provisions whereby shareholders vote on executive compensation plans, referred to as “say-on-pay” provisions, were introduced in other countries, such as Switzerland, Germany and Australia. This wave of regulation supplemented existing regulation on compensation, such as the Companies Act of 2006 in the United Kingdom and the Omnibus Reconciliation Act of 1993 in the US, which apparently were in adequate. In Israel, one of the ramifications of the 2008 financial crisis was the Social Protest of 2011, which received extensive media coverage. One of the central points of the Social Protest, especially amplified by the financial press, was executive pay. From The Marker, March 23, 2012 (Editorial): “Executive compensation 2011: while rank-andfile employees’ salaries are not enough to carry them through the month, executives are celebrating. . .. In spite of the Social Protest, senior executives still receive excessive compensation, regardless of

page 20

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 21

21

firm performance, and with no shame.” Some data were provided by Calcalist, April 2, 2013 (by Naama Sikoler): “In 2012, the average salary of a CEO in a public company was 2.7 million NIS — an increase of 330% since 1994. The average salary of the 100 highestpaid CEOs of firms traded on the Tel Aviv stock exchange was 6.7 million NIS — an increase of 963%. As a comparison, the average salary increased in the same period by 146%.” These quotes highlight the two central issues in the intense debate about CEO compensation: whether it is excessive (especially when compared to the compensation of rank-and-file employees), and whether it is sufficiently tied to performance. The public outrage and media scrutiny, as well as regulation worldwide, were the driving forces for regulatory action in Israel. In 2012, Israel Securities Authority (ISA) signed into law Amendment 20 to Companies Law that came into effect in 2013. From a speech by Shmuel Hauser, Chairman of ISA, January 7, 2013: “The main problem with executive compensation:a lack of correlation between CEO contribution to firm’s financial performance and the size of CEO compensation/bonus. This problem must be resolved for the benefit of investors, the capital market and the economy as a whole. . . Last month Amendment 20 to Israel Companies Law regarding executive compensation became effective. . . It came into being in order to cope with the phenomenon of the significant increase in executive compensation over the past few years. This increase is bothersome, especially in light of ISA’s findings regarding executive pay in public companies: a lack of correlation between compensation size and firm’s financial performance, and a lack of correlation between compensation size and executive contribution to the firm.”1 The Amendment requires the establishment of a compensation committee, comprised solely of independent members of the board of directors. The committee’s role is to formulate a compensation policy 1

Some important data from the speech: “During the period 2003–2011, the average compensation increased by 127%, including during the height of the 2008 crisis. During this period, the market value of an average firm in the Tel-Aviv 100 index increased by 40%. In an international comparison, taking into account the size of the firms, the average compensation of an executive in an Israeli company is higher than the average compensation of an executive in most developed countries, including the US, UK and Europe.”

November 4, 2020

22

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

valid for three years, taking into account the goals and strategy of the firm in tying executive compensation to performance on a long-term basis. The compensation policy should also take into consideration factors such as firm size and type of activities, executive skills and education, and the pay gap between the executive and rank-and-file employees. The ratio of the performance-based component of the compensation package (such as bonuses and option awards) to its fixed component (such as salary and benefits), as well as its retirement benefits if applicable, should also be taken into account. The committee should review the compensation policy every three years. The approach taken in Amendment 20 is to regulate through intervention in decision-making and approval procedures rather than through direct intervention in determining the compensation. This approach is in the spirit of the “say-on-pay” provisions common throughout the world, modified to incorporate characteristics of the Israeli corporate governance environment. Our study explores the impact of the Amendment on the level and structure of executive compensation. We focus on the value of control manifested in the implementation of the regulation in the Israeli market, a market that lends itself to such an investigation due to its concentrated ownership structure. We undertake this endeavor by examining separately the effect of the regulation on subsamples formed according to the power and influence of chief executive officers (CEOs) and chairmen of the board of directors. Our analysis is conducted using hand-collected data on a sample of firms traded on Tel Aviv Stock Exchange (TASE) during a 2-year window around 2012, the year in which the Amendment was signed into law. We find that CEOs classified as related parties in the firm’s annual financial statements (henceforth related or influential CEOs) influence the compensation structure design of their post-amendment compensation so that, compared to non-influential CEOs, the bonus component increases, leading to an increase in total cash compensation without an increase in salary. We find evidence for the ability of chairmen classified as related parties (henceforth related or influential chairmen) to temper the bonus of non-influential CEOs and we find no evidence that stock-based compensation increased or became a more important component of CEO pay, regardless of CEO influence.

page 22

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 23

23

This chapter extends the existing literature on executive pay regulation in examining a new regulation in a corporate governance setting different from the one in the US. Unlike most existing corporate governance research focusing on environments characterized by dispersed ownership, our research is conducted on companies traded on the Israeli capital market, characterized by concentrated ownership and the existence of powerful business groups, as described in OECD (2011) and Kosenko (2008).2 Almost all public companies in Israel have dominant controlling shareholders, some of whom take an active part in appointing directors.3 As such, CEOs and chairmen that are classified as related parties would have significant influence over the board of directors. Also, it can be argued that this setup presents a limitation in generalizing our results to the US. However, Israel is a developed market and a member of the OECD with powerful regulators and advanced legislation, which follows most regulatory and accounting standard changes adopted in the developed countries.4 That being the case, we believe that our results offer valuable insights into influence and behavioral response to regulation. An important goal of this research is to shed light not only on the effectiveness of Amendment 20 in Israel but also on the general effectiveness of regulation that puts the burden of decision on the shoulders of boards, as opposed to regulation that mandates explicit limitations on pay. In addition, the value of control in public companies is explored in the context of executive pay.

2

See also the analysis in Khanna and Yafeh (2007) regarding the ownership structure throughout the world. 3 See Justice Department background and explanations to Amendment 20 to Israel Companies Law, November 8, 2012. 4 For example, since 2009 Israel has been following a slightly modified version of SOX, suitable for its corporate governance environment. It is also worthwhile pointing out that throughout the years a few committees were appointed in order to examine various aspects of Israeli corporate governance and develop recommendations for improvement. One of the most important of these committees was the Goshen Committee, appointed in 2005 to focus on the improvement of director independence, the workings of audit committees and accountability, as well as the establishment of a court specializing in Corporate and Securities Law. Its recommendations were adopted in 2007.

November 4, 2020

24

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

The remainder of this chapter is organized as follows. Section 2 provides some background and reviews of the existing literature. Section 3 describes our data and methodology. We then present our results in Section 4. Section 5 concludes.

2.

Background and Literature Review

The topic of executive compensation and the different types of regulation implemented in order to curb it or to modify its structure have attracted the attention of the public, media and academics starting in the 1990s. For example, Balsam (2002) and Bebchuk and Grinstein (2005), among others, document a dramatic increase in pay. However, unlike the overwhelmingly negative sentiment of the public and media, academics hold a wide range of opinions, as can be seen from the written testimonies of Bebchuk (2007) and Kaplan (2007) in the Hearing before the Committee on Financial Services at the US House of Representatives (2007).5 The bulk of the early literature focused on regulated industries. Examples include Joskow et al. (1996) on the electric utility industry and Hubbard and Palia (1995) on the banking industry.6 Joskow et al. (1996) explore the regulated electric utility industry and find that CEO pay is lower when the environment in which the firm operates favors consumers over investors. They also provide evidence of political pressure constraining CEO compensation. Hubbard and Palia (1995) examine the effect of deregulating the banking industry on CEO pay, and find higher levels of pay and a stronger pay-performance relationship in the deregulated banking markets. A later body of literature examines the effect of the tax code provision Omnibus Reconciliation Act of 1993, the first large-scale attempt to regulate executive pay across industries. The provision limits the tax deductibility of the salary paid to the top executives 5

See also Edmans and Gabaix (2016) for an overview of the theoretical models on executive compensation, and Edmans et al. (2017) for a survey the theoretical and empirical literature on the topic. 6 See also Geddes (1997) and Wolfram (1998) on the electric utility industry and Barro and Barro (1990) on the banking industry.

page 24

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 25

25

to US$1 million each, starting in 1994, while not affecting the components of the compensation that are contingent upon performance, such as bonuses and stock option awards. The findings of Rose and Wolfram (2000) cast doubt on the efficacy of the provision in constraining overall executive compensation. They find that while firms near the US$1 million cap limit their salary increase, they seem to increase the performance-based components of the compensation, so that the effect of the provision on the overall pay level is unclear. This is similar to our finding that companies with influential CEOs shift the composition of pay from cash salaries to cash bonuses in response to the Amendment, visibly complying with the intent of the Amendment, however, leaving the CEO better off. Rose and Wolfram (2002) extend the previous analysis and conclude that the provision had little impact on total compensation levels. Perry and Zenner (2001) document increases in all components of the compensation package after 1993 with the largest increase in performance-based components, so that the sensitivity of compensation to performance increased as a result of the provision. Unlike previous work, Balsam and Ryan (2007) focus on CEOs hired after the imposition of the 1993 act. They argue that a firm hiring a new CEO has a better opportunity to restructure the compensation package awarded to the CEO relative to a situation where the CEO is already in place. Indeed, they find evidence for the efficacy of the provision for newly appointed CEOs, documenting that the increase in salary normally associated with the hiring of a new CEO has been mitigated and that the bonus paid to the CEO is more responsive to firm performance for CEOs appointed after 1994. Considering post-2008 literature, the debate among scholars centered on whether restrictions on executive pay are beneficial to shareholders, and if so — under which conditions and to what extent. Cebon and Hermalin (2015) develop a theoretical model and derive conditions under which external restrictions imposed on executive pay enhance welfare and benefit shareholders. However, many scholars document distortions caused by such restrictions. Dittmann et al. (2011) analyze, theoretically and empirically, several proposals aimed at restricting CEO pay. They demonstrate that many of the restrictions have unintended consequences, such as higher average compensation, higher rewards for mediocre performance and higher risk-taking incentives. Larcker et al. (2012) describe and refute some

November 4, 2020

26

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

myths associated with “say-on-pay.” An extreme position was taken by Jensen and Murphy (2018), who analyze the different pay regulations, ranging from the first disclosure rules in the 1930s to the 2018 Trump tax rules. Their conclusion is that the vast majority of these regulations damaged economic efficiency, going as far as to assert that “The best way the government can fix executive compensation is to stop trying to fix it.” Consequently, they advocate undoing the existing regulation that they view as damaging. A recent study on China by Bae et al. (2019) analyzes pay restrictions imposed on centrally administered state-owned firms in 2009 and reaches the conclusion that limiting CEO pay backfired. They show that while the restrictions caused a significant drop in CEO compensation, they also led to an increase in CEOs consumption of perks. This distortion in incentives caused a significant drop in the performance of the firms adopting the restrictions. In addition to Amendment 20 analyzed in this chapter, regulators in Israel passed in 2016 a law that restricts total executive compensation in three industries — banking, insurance and investment firms — to 35 times the salary of the lowest paid employees. This restriction translates to an upper limit on total pay of $2.53 million ILS a year (approximately US$716,901 at the time of writing), representing a significant pay limit for these executives. Abudy et al. (2017) examine the short-term market reaction to the announcement of the regulation on a sample consisting of 20 firms in the aforementioned three industries. They document significantly positive abnormal announcement returns in these industries, implying that the pay restriction appears to benefit shareholders, at least in the short term. The research question we address in this chapter is whether Amendment 20 has been effective in achieving its goals, namely, decreasing the overall level of compensation and strengthening the tie between executive pay and firm performance. Another issue explored in this chapter is the value of control in the context of executive pay. We explore this issue by examining the effect of the regulation on subsamples, differentiated by CEO and chair influence. The Israeli market lends itself to such an investigation because of its concentrated ownership structure, with most companies having controlling shareholders. This structure could potentially give rise to ties among directors and between directors, controlling shareholders and CEOs

page 26

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 27

27

along multiple dimensions. Such ties might hamper director independence, as has been suggested in previous literature (see, e.g., Brick et al., 2006; Coles et al., 2014; Hwang and Kim, 2009; Shivdasani and Yermack, 1999). For example, in Shivdasani and Yermack (1999), when the CEO is involved in the selection of directors these directors are less likely to monitor the CEO aggressively, an evidence the authors interpret as demonstrating how CEO influence serves as an important determinant of the firm’s corporate governance structure. It is worth mentioning that in Israel there is no orderly procedure for the selection of directors, and it is not uncommon for directors to be proposed by controlling shareholders. It has often been argued, especially during a recent wave of derivative lawsuits involving large family-controlled companies in Israel, that independent directors are not always truly independent. Our findings are consistent with this notion regarding independence of directors, as the differences in bonus growth and compensation composition between CEOs classified as related parties in the firm’s annual financial statements and CEOs not classified as related parties suggest that influential CEOs were able to favorably influence their compensation packages with some seemingly independent board members on the compensation committee.

3.

Data and Methodology

Our study uses hand-collected data on firms traded on TASE in a 2-year window around 2012, the year in which the regulation was signed into law. We compare each CEOs average compensation in the 2-year period prior (pre-amendment) to the introduction of the regulation (2010–2011) to their average compensation in the 2-year period following (post-amendment) the introduction of the regulation (2013–2014), excluding 2012 from the study.7 We excluded 2012 because some companies might have implemented the amendment 7

In case a firm has an observation in at least one of the years 2010–2011 and at least in one of the years 2013–2014, we include it in our sample. If a firm does not have any observation for the period before or the period after the regulation became effective, we delete it from our sample.

November 4, 2020

28

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

early, while others might have used the last window of opportunity to award excessive compensation just before the implementation of the law. From The Marker, December 3, 2012 (by Shelly Appelberg): “. . . Israel Securities Authority is warning public companies in Israel not to try to perform a last-minute coup and approve compensation packages to senior executives in the next few days, before Amendment 20 comes into effect in 12.12.2012. . . On 27.11, the Israel Corporation renewed the contracts of CEO NirGil’ad and Chairman Amir Elstein and granted each options worth 50 million NIS, and additional options worth 13.3 million NIS were granted to Avisar Paz, the Chief Financial Officer.” The companies in our sample were members of the index TA-100, an index comprising the largest 100 companies traded on TASE. Each company in our sample was a member of the index at the end of at least one of the years 2009–2014.8 As is customary, firms subject to a special regulatory regime such as banks, insurance companies and cross-listed firms (firms listed in TASE and in another exchange, usually in the US or Europe) are excluded from our sample. We use two data sources: Super-Analyst, a commercial database and Maya, the online reporting system of ISA. A number of accounting items, namely, long-term and current assets, long-term and current liabilities and net income, are retrieved from Super-Analyst. All other variables we use are hand collected from one of two parts of Maya online system — either from annual financial statements that each company files with ISA or from announcements made by firms. Compensation data are hand collected from annual statements filed by firms with ISA. It is mandatory for firms to include information in their annual financial statements regarding compensation for their five highest paid employees. Though formally the CEO is not necessarily one of the five highest paid employees, in practically all cases his or her compensation is disclosed in the statements. The information provided by firms includes a breakdown of total compensation to its components: a fixed compensation (that includes salary, management fees, consulting fees and benefits), bonuses, stock-based

8

TASE started publishing historical data regarding index composition in June 30, 2010, so the index with respect to the end of 2009 is from that date.

page 28

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 29

29

compensation and other components such as retirement benefits.9 We also collect, for each CEO, whether he or she is classified as a related party. A related party is defined as a person or entity that holds 5% or more of the shares in a firm or is able to appoint at least one director or the CEO. However, the CEO and members of the board of directors are defined as related parties, regardless of their holdings in the firm. Because the CEO is a related party by definition, the classification of a CEO as a related party beyond being a CEO is also reported by the firm, according to ISA guidelines. We rely on these classifications. Data on the board of directors are also collected since board members are instrumental in determining compensation policy. Board data include the number of independent board members and the number of board members with financial expertise. ISA requires companies to classify directors as independent or not independent and as having or not having financial expertise and report it in the financial statements, according to ISA guidelines. A director is classified as independent if he or she has no ties to the company, its management and its controlling shareholders. A director cannot be classified as independent if he or she is a family member of controlling shareholders or officers, an employee of the company or has business relationships with the company. A director with accounting and financial expertise is defined as one whose education, experience and skills provide him or her with tools to comprehend business and accounting issues and enable him or her to fully understand the financial statements of a company. Again, we rely on these classifications. In addition, we retrieve for each firm the percentage of shares of the company held by the CEO, the largest (controlling) shareholder,

9

Some benefits, such as leasing contracts, are sometimes reported under the fixed component, lumped together with salary, and at other times they are reported under other components. Most Israeli companies grant options using the capital-based method. Under the capital-based method, option compensation is marked-to-market once, at issuance, and the company recognizes the expense over future years as the options vest, according to the pre-determined schedule. Some companies use the liability-based method and grant phantom options that are marked-to-market each year as the options vest. However, phantom options are the exception to the rule.

November 4, 2020

30

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

and by all related parties.10 These variables capture the concentrated ownership structure of the Israeli economy, as described in OECD (2011) and Kosenko (2008). The number of subsidiaries was hand collected from annual financial statements to proxy for firm complexity. We use a dummy variable, Related CEO, indicating whether the CEO is a related party. This variable takes the value 1 if the CEO is a related party. Likewise, we use a dummy variable, Related Chair, indicating whether the chairperson is a related party, again taking the value 1 if the chairperson is a related party. Some studies use a control variable indicating whether the CEO is also the chairperson of the board of directors. While the CEO–chairperson duality is a common practice in the US, in Israel it is generally not permitted in a public company, except if approved by a majority of disinterested shareholders in a shareholder meeting. As a result, this duality is extremely rare in Israel and is thus irrelevant to our study. We also include firm size, proxied by log of assets, in our multivariate regression analysis since the size of the firm should affect the size of compensation: the larger the firm, the more it can afford to pay its CEO. To proxy for complexity, we use the square root of the total number of consolidated subsidiaries. One can also argue that the larger or more complex the firm, the more effort the CEO needs to exert. We use return on assets (ROA)to proxy for firm profitability. ROA should also be a key input in the considerations of the compensation committee when deciding on CEO compensation.11 This results in 298 observations with 61 unique companies and 85 unique CEOs. Over our sample period, 2010–2014, companies replaced CEOs. The years in which a company replaced its CEO were not included in our sample for that company since there are likely to be unusual payments for that year for both CEOs, for example, a retirement bonus for the retiring CEO or a signing bonus for the 10

If the CEO holds less than 1% we define it as zero. Originally, we considered leverage as an additional control variable. We excluded it from our final analysis due to its high correlations with Log Assets (r = 0.68) and ROA (r = −0.52) as including it might have induced multicollinearity. 11

page 30

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 31

31

new CEO. In addition, data on CEOs were not always provided by these firms since they are required to provide information only for their five highest paid employees and a CEO who served only part of the year might not be included in this group. Out of the 61 firms in our sample, four firms had high management turnover. These firms replaced CEOs in both pre-amendment years (2010 and 2011) or in both post-amendment years (2013 and 2014). As a result, these firms were not included in our tests of pre- vs. post-amendment compensation. Our final sample includes 253 observations with 57 unique companies and 78 unique CEOs. To reduce concerns of endogeneity, in our regression tests, we use lags of all of our explanatory variables (including control variables) to the pre-amendment period (2011). Table 1 provides the names and definitions of the variables that we present summary information for or that are included in our regression models. To examine the impact of the implementation of Amendment 20, we construct nine compensation variables. Six of them are related to the components of compensation. Total compensation is reported as such in the firm’s annual financial statements, as are the fixed component, bonus and stock-based compensation. To capture the “bird in hand” component, we add the fixed and the bonus compensation components to construct the variable cash compensation component of CEO compensation. The bonus is essentially the short-term cash incentive component of the compensation and it is based, according to the guidelines to the compensation committee as outlined in Amendment 20, on a few criteria such as accounting measures of firm performance. In addition, the committee uses as a benchmark the bonus paid to CEOs at a group of firms that the committee, usually with the help of professional advisors, defines as “similar firms.” Amendment 20 does not contain strict guidelines to the committee regarding the exact criteria, their relative weights or the choice of similar firms, thus leaving some leeway in these choices. We also construct the variable Incentive Comp as the sum of the two performance-based components of compensation: the bonus and the stock-based component. The other three compensation measures are compensation ratios — the ratio of fixed component of the compensation to total

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

32

Behavioral Finance: A Novel Approach Table 1:

Variable Name Total Comp Fixed Comp Bonus Comp Equity Comp Cash Comp Incentive Comp Fixed Comp Ratio Bonus Comp Ratio Cash Comp Ratio Log Assets Complexity ROA Expert Dir IndepDir Related CEO Related Chair CEO Shares Related Shares Controlling Shares

b3964-ch02

Variable definitions Variable Description

Total reported compensation. Fixed component of compensation (salary, management fees, consulting fees and benefits). Total reported cash bonus compensation. Total stock-based compensation. Fixed Comp plus Bonus Comp. Bonus Comp plus Equity Comp. Ratio of Fixed Comp to Total Comp. Ratio of Bonus Comp to Total Comp. Ratio of Cash Comp to Total Comp. Natural log of total assets. Natural log of total number of consolidated subsidiaries plus 1. Ratio of Net Income to Total Assets. Ratio of number on board with financial expertise to the number of board members. Ratio of number of independent board members to the number of board members. Dummy variable equaling one if CEO is related else zero. Dummy variable equaling one if the chair is related else zero. Percentage of shares held by the CEO. Percentage of shares held by related parties. Percentage of shares held by controlling parties.

compensation, the ratio of bonus compensation to total compensation, and the ratio of the cash components to total compensation. For most of the tests reported below, we examine the growth or change in the given compensation measure from the pre- to post-amendment period. Table 2 provides descriptive statistics for our variables. Examining the ownership characteristics in Table 2, two interesting points emerge. First, the mean percentage of shares held by the CEO is 7.43% whereas the median percentage of shares is zero. This is a reflection of the fact that CEOs in Israel typically do not accumulate a significant stake in the firm merely by being CEOs.

page 32

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

page 33

Influential CEO and Board Behavior in Reaction Table 2:

33

Summary statistics

Obs

Mean

Managerial Influence: Related CEO (%) Related Chair (%)

283 283

25.80 81.27

Compensation: Total Comp (millions) Fixed Comp (millions) Bonus Comp (millions) Equity Comp (millions) Cash Comp (millions) Incentive Comp (millions) Fixed Comp Ratio (%) Bonus Comp Ratio (%) Cash Comp Ratio (%)

283 283 283 283 283 283 283 283 283

4.45 2.11 1.35 0.99 3.46 2.34 59.92 24.76 84.69

Median

SD

Min

Max

3.45 2.05 0.67 0.05 2.99 1.34 57.86 20.28 97.97

3.15 0.93 1.81 1.99 2.07 2.81 25.79 23.28 20

0.40 23.49 0.00 6.78 0.00 11.70 −0.67 13.57 0.40 13.01 0.00 20.12 0.00 100.00 0.00 100.00 .69.53 17 115.44

Compensation Growth Comparing Mean 2013–2014 to Mean 2010–2011 FirmLevel Compensation Total Comp Growth (%) 57 9.67 −10.26 93.83 −69.59 649.98 53.65 −100.00 309.57 Fixed Comp Growth (%) 57 12.10 7.13 Bonus Comp Growth (%) 52 68.15 −0.15 418.76 −100.00 2923.54 Equity Comp Growth (%) 55 −14.38 0.00 70.08 −100.00 278.05 85.33 −64.67 537.28 Cash Comp Growth (%) 57 20.36 7.95 21.72 −92.16 57.00 Δ Fixed Comp Ratio (%) 57 4.07 4.65 Δ Bonus Comp Ratio (%) 57 2.67 1.02 93.27 23.88 −57.00 Δ Cash Comp Ratio (%) 57 6.74 0.00 18.45 −23.77 68.64 Governance: Indep Dir (%) Expert Dir (%)

283 283

37.41 61.13

33.33 60.00

14.35 18.40

7.14 22.22

75.00 100.00

Ownership: CEO Shares (%) Related Shares (%) Controlling Shares (%)

283 283 283

7.34 74.23 58.12

0.00 74.93 60.00

16.86 12.52 15.55

0.00 30.29 9.26

88.40 100.00 100.00

Firm Characteristics: Total Assets (billions) LogAssets

283 283

12.87 15.41

4.81 15.39

22.86 1.45

0.06 11.02

135.96 18.73

(Continued )

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

34

Table 2:

ROA (%) Complexity

(Continued )

Obs

Mean

Median

SD

Min

Max

283 283

4.93 2.10

3.47 2.20

7.70 0.94

−7.81 0.00

66.23 4.87

Notes: Related CEO and related chair are dummy variables equaling one if the CEO or chair is related, respectively, else zero. Total, Fixed, Bonus, and Equity-Based Comp are the annual amounts reported in the company’s financial statements. Cash Comp is the sum of Fixed Comp and Bonus Comp. Incentive Comp is the sum of Bonus Comp and Equity Based Comp. Fixed Comp Ratio is the ratio of annual Fixed Comp to annual Total Comp. Bonus Comp Ratio and Cash Comp Ratio are calculated in the same manner. Total Comp Growth is the growth rate in total comp for a given company (CEO) from 2010–2011 to 2013–2014. Fixed, Bonus, Equity, Cash and Incentive Comp Growth are similarly calculated, Δ Fixed Comp Ratio is the change in the Fixed Comp Ratio from 2010–2011 to 2013–2014 for each company (CEO). Δ Bonus Comp Ratio and Δ Cash Comp Ratio are calculated in the same manner. Indep Dir and Expert Dir are the ratios of number of independent directors and number of members on board with financial expertise, respectively, to the total number of directors on the board. CEO, Related, and Controlling Shares are the percentage of shares held by the CEO, related parties, and the controlling (largest) shareholder, respectively. Log Assets is the natural log of Total Assets; ROA is ratio of Net Income to Total Assets; and Complexity is the natural log of the total number of consolidated subsidiaries plus 1.

Thus, if a CEO is a related party, it means that he or she was a related party prior to his or her appointment as a CEO. Second, the mean percentage of shares held by related parties (as defined earlier) is 74.23% whereas the mean percentage shares held by controlling shareholders is 58.12%. This is because in many cases related and controlling shareholders are the same parties. In Table 3 we provide a breakdown of our sample according to industry. As can be seen, the largest sector in our sample is Real Estate & Construction, followed by Investment & Holding companies.12 12

Recall that in the sample selection process we exclude banks, insurance companies, cross-listed firms and Gas & Oil partnerships. Thus, the single company in the Financial Services sector in our sample is not a bank or an insurance company.

page 34

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Influential CEO and Board Behavior in Reaction Table 3:

35

Industry statistics

Real Estate & Construction Investment Holding Wholesale & Retail Trade Food Fashion & Clothing Chemistry, Rubber & Plastics Communication & Media Electronics & Optics Financial Services Information Services Metals & Construction Materials Software and Internet Timber, Paper & Printing

4.

page 35

#

%

22 11 6 5 3 2 2 1 1 1 1 1 1

38.6 19.3 10.5 8.8 5.3 3.5 3.5 1.8 1.8 1.8 1.8 1.8 1.8

57

100

Results

We start with analyzing the correlations between our variables. These correlations are presented in Table 4. We observe a negative and significant (at the 10% or better level) correlation between related CEO and related chair.13 This is because in most of our firms there is either a related CEO or a related chair but not both. Most of our sample firms (47 out of 57) have related chairs, and in 41 of these firms the CEO is not related. That is, related chairs tend to appoint professional unrelated CEOs. A total of 14 firms in our sample had related CEOs, and in only 6 of those the chair was also related. There were only two firms in our sample where neither the chair nor the CEO were related. Related CEO is positively and significantly correlated with Total Comp Growth, Bonus Comp Growth, Cash Comp Growth and Incentive Comp Growth. However, it is unrelated to Equity Comp Growth. This is consistent with the positive relationship to Bonus Comp Growth driving positive relationship with both Cash Comp Growth 13

To reduce the physical size of our correlation table, we use a single asterisk (∗) to represent statistical significance at the 10% level or better.

Correlations between selected variables Δ Bonus Comp Ratio

Δ Cash Comp Ratio

10:10

Δ Cash Incentive Fixed Bonus Equity Fixed Total Comp Comp Comp Comp Comp Comp Related Related Comp CEO Chair Growth Growth Growth Growth Growth Growth Ratio

Indep Expert Log Dir Dir Assets

ROA

Behavioral Finance: A Novel Approach

b3964-ch02

0.498∗

Behavioral Finance: A Novel Approach – 9in x 6in

∗

Related −0.510 Chair Total Comp 0.282∗ −0.042 Growth Fixed Comp 0.136 0.010 0.742∗ Growth Bonus Comp 0.233∗ −0.308∗ 0.536∗ −0.375∗ Growth Equity Comp 0.010 0.065 0.314∗ 0.148 −0.191 Growth 0.905∗ 0.697∗ 0.532∗ −0.149 Cash Comp 0.240∗ −0.087 Growth 0.256∗ −0.301∗ 0.555∗ −0.357∗ 0.977∗ −0.049 0.364∗ Incentive Comp Growth −0.322∗ 0.206 −0.544∗ 0.073 −0.730∗ −0.192 −0.473∗ −0.753∗ Δ Fixed Comp Ratio 0.177 −0.195 0.315∗ −0.144 Δ Bonus 0.711∗ −0.349∗ 0.551∗ 0.586∗ −0.676∗ Comp Ratio −0.150 −0.009 −0.233∗ −0.101 Δ Cash 0.074 −0.667∗ 0.156 −0.096 0.302∗ Comp Ratio

November 4, 2020

36

Table 4:

page 36

November 4, 2020 10:10

−0.110 −0.032 0.130 −0.077 −0.126 −0.122 0.007 0.449∗ −0.182 −0.149 −0.050 0.094 −0.151 0.041 −0.038 −0.025 0.402∗ 0.410∗ −0.053 −0.006

−0.113 0.141 0.046 −0.230∗ 0.086 −0.029 −0.178 −0.052 0.163 −0.175 −0.016 0.129 0.353∗ −0.041 −0.065

−0.100 −0.217 −0.139 −0.163 −0.019

−0.075 −0.315∗ 0.012 −0.060 −0.102

0.261∗ 0.111 0.149 0.047 0.104 −0.423∗ 0.048 −0.120 0.115 −0.201

Notes: Related CEO and related chair are Dummy variables equaling one if the CEO or chair is related, respectively, else zero. Total Comp Growth is the growth rate in total comp for a given company (CEO) from 2010–2011 to 2013–2014. Fixed, Bonus, Equity, Cash, and Incentive Comp Growth are similarly calculated, Δ Fixed Comp Ratio is the change in the ratio of Fixed Comp to Total Comp from 2010–2011 to 2013–2014 for each company (CEO). Δ Bonus Comp Ratio and Δ Cash Comp Ratio are calculated similarly. Indep Dir and Expert Dir are the ratios of number of independent directors and number on board with financial expertise, respectively, to the total number of directors on the board. Log Assets is the natural log of Total Assets; ROA is ratio of Net Income to Total Assets; and Complexity is the natural log of the total number of consolidated subsidiaries plus 1. ∗ Represents significance at the 10% or better.

Behavioral Finance: A Novel Approach – 9in x 6in

−0.078 0.034 −0.034 0.068 −0.292∗ 0.374∗ 0.134 −0.218∗ 0.053 0.127

Influential CEO and Board Behavior in Reaction

Indep Dir Expert Dir LogAssets ROA Complexity

b3964-ch02

37 page 37

November 4, 2020

38

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

and Incentive Comp Growth. Related CEO is also negatively and significantly related to Fixed Comp Ratio. Taken together with the positive relationship with Bonus Comp Growth, this is consistent with a shift from salary to bonus. Overall, this evidence suggest that Total Comp rose faster for influential CEOs, relative to non-influential CEOs, because of increases in the bonus component. For related chair, we find a negative and significant correlation with Bonus Comp Growth and with Incentive Comp Growth, but no significant correlation with Total Comp Growth or Equity Comp Growth. This is consistent with the notion that influential chairs are able to temper CEOs bonuses. These findings regarding a higher increase in bonuses for the CEO when the CEO is influential as opposed to a lesser increase in bonuses when the chair is influential highlight the agency problem associated with CEOs. CEOs would have chosen to increase their bonuses, which they are able to do when they are influential, whereas with influential chairmen, they are not able to do so and bonus pay increases significantly less compared to CEOs where the chair is unrelated. This is consistent with our hypothesis that bonus increases for related CEOs were due to related CEOs’ ability to impact their post-amendment compensation. In untabulated findings, for each of our nine compensation measures comparing post- to pre-amendment period, we examine the difference between a given measure for a related vs. unrelated CEO. We do the same for the difference between the measures for related vs. unrelated chair, as well as for the case where both CEO and chair are related and for the case where only the CEO is related. For the first set, related CEOs, we see that the average growth in Total Comp for related CEOs exceeded unrelated CEOs by 61%. We also see that the growth in Bonus Comp for related CEOs exceeded unrelated CEOs by 229%. A similar result is observed for Cash Comp Growth difference (47% faster rate). The sign is negative (16% lower ratio) for the Δ Fixed Comp Ratio. For the most part, the results of these means tests are insignificant, most likely due to the sample size, but they nevertheless support the findings for the Total Comp Growth and Bonus Comp Growth discussed in our correlation analysis. The higher Cash Comp Growth and lower Δ Fixed Comp Ratio support the notion that influential CEOs used their influence to pay themselves more cash

page 38

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 39

39

compensation, essentially shifting their pay to bonuses from salary as well as increasing their total pre- to post-amendment compensation at a faster rate than unrelated CEOs. Turning to our second set of results, related chairs, we observe that most of the differences are negative though, again, insignificant. CEO Bonus Comp Growth rose more slowly in companies with influential chairs than in companies with unrelated chairs. The same holds for Cash Comp Growth and Incentive Comp Growth. These findings, taken together, suggest that the main difference between influential and unrelated chairs is that CEOs in companies with influential chairs were unable to increase bonuses. Overall, univariate findings support the finding regarding Bonus Comp Growth from the correlation analysis. Influential chairs temper CEO bonuses following the amendment. Considering the third set, pertaining to the case where both CEO and chair are related, CEOs total compensation grew faster through the fixed component (40%), leading to a 107% greater growth in total comp compared to CEOs in the other subcategories. The difference is negative for the Bonus Comp Growth. This implies that when both the CEO and chair are influential, greater growth in compensation (74%) came from the fixed component of compensation rather than from the bonus component. The fourth set, when the CEO is strongest (i.e., the CEO is related and the chair is unrelated), the growth in the bonus component (compared to CEOs in other configurations) is very high (407%). We also observe a decrease in the Cash Comp Ratio and an increase in the Bonus Comp ratio relative to all other CEOs. These findings, taken together with the finding in the third set that firms with both influential CEOs and chairs shifted less to bonus compensation, indicate that most powerful CEOs shift their pay from the fixed component to the bonus component. For the most part, in the untabulated means tests the growth rate in equity compensation was negative and there was not much difference in those growth rates across our CEO and chair power subsamples. Thus, the means tests continue to support our main findings that influential CEOs shifted their pay from fixed compensation to bonus compensation while not increasing their bonding to shareholders with greater use of equity-based compensation. Table 5 contains four Panels, each presenting regression results for one of our subsamples. Panel A presents the results for Related

10:10

Impact of CEO and chair influence on post-amendment compensation

Panel A: Influential CEO

Complexity Indep Dir Expert Dir (Intercept)

Bonus Comp Growth

Equity Comp Growth

Cash Comp Growth

Incentive Comp Growth

Δ Fixed Comp Ratio

Δ Bonus Comp Ratio

Δ Cash Comp Ratio

0.430 (0.1309) −0.193∗∗ (0.0460) −3.309∗ (0.0933) 0.335∗∗ (0.0170) −0.728 (0.4195) 0.106 (0.8754) 2.557∗ (0.0914) 0.257 57

0.058 (0.7264) −0.078 (0.1621) 0.343 (0.7627) 0.270∗∗∗ (0.0013) −0.259 (0.6219) 0.018 (0.9635) 0.785 (0.3702) 0.223 57

2.398 (0.1057) 0.005 (0.9917) −6.068 (0.5341) −0.730 (0.3206) 5.615 (0.2390) −1.099 (0.7590) 0.590 (0.9380) 0.099 52

0.026 (0.9036) 0.025 (0.7337) −0.117 (0.9374) 0.002 (0.9870) −0.944 (0.1797) 1.861∗∗∗ (0.0007) −1.320 (0.2564) 0.236 55

0.321 (0.2215) −0.168∗ (0.0610) −3.382 ∗ (0.0646) 0.217∗ (0.0898) −0.439 (0.5986) −0.514 (0.4113) 2.825∗∗ (0.0452) 0.230 57

2.214∗ (0.0582) 0.016 (0.9672) −4.059 (0.6066) −0.475 (0.4114) 4.374 (0.2347) 0.711 (0.7939) −1.127 (0.8533) 0.109 56

−0.150∗∗ (0.0327) 0.028 (0.2263) 0.889∗ (0.0657) 0.019 (0.5703) 0.005 (0.9814) −0.083 (0.6136) −0.388 (0.2904) 0.172 57

0.086 (0.2627) −0.026 (0.3141) −1.025∗ (0.0552) −0.048 (0.1943) 0.010 (0.9677) −0.271 (0.1402) 0.711∗ (0.0825) 0.163 57

−0.065 (0.2788) 0.002 (0.9071) −0.136 (0.7406) −0.029 (0.3113) 0.015 (0.9371) −0.354 ∗∗ (0.0154) 0.323 (0.3075) 0.146 57

b3964-ch02

R2 Obs.

Fixed Comp Growth

Behavioral Finance: A Novel Approach – 9in x 6in

LOG Assets ROA

Total Comp Growth

Behavioral Finance: A Novel Approach

CEO Related

November 4, 2020

40

Table 5:

page 40

November 4, 2020

Panel B: Influential Chair

ROA Complexity Indep Dir Expert Dir (Intercept) R2 Obs.

0.043 −4.207∗∗ (0.8245) (0.0182) −0.087 0.063 (0.1148) (0.8920) 0.385 −10.574 (0.7396) (0.2742) 0.272∗∗∗ −0.710 (0.0012) (0.3174) −0.269 5.757 (0.6089) (0.2125) 0.022 −1.168 (0.9557) (0.7360) 0.888 3.941 (0.2853) (0.5778) 0.222 0.157 57 52

Equity Comp Growth

Cash Comp Growth

Incentive Comp Growth

Δ Fixed Comp Ratio

Δ Bonus Comp Ratio

Δ Cash Comp Ratio

0.066 (0.7904) 0.018 (0.7993) −0.044 (0.9769) 0.001 (0.9927) −0.939 (0.1818) 1.862∗∗∗ (0.0007) −1.272 (0.2546) 0.237 55

−0.270 (0.3780) −0.183∗∗ (0.0383) −3.723∗∗ (0.0478) 0.243∗ (0.0593) −0.571 (0.4956) −0.484 (0.4423) 3.352∗∗ (0.0136) 0.219 57

−2.993∗∗ (0.0247) 0.016 (0.9665) −7.213 (0.3623) −0.348 (0.5408) 3.558 (0.3251) 1.065 (0.6910) 1.778 (0.7576) 0.136 56

0.130 (0.1164) 0.035 (0.1312) 1.053∗∗ (0.0370) 0.007 (0.8427) 0.068 (0.7622) −0.097 (0.5626) −0.635∗ (0.0764) 0.136 57

−0.142 (0.1081) −0.026 (0.2989) −1.195∗∗ (0.0270) −0.039 (0.2815) −0.037 (0.8781) −0.262 (0.1484) 0.845∗∗ (0.0284) 0.185 57

−0.012 (0.8625) 0.010 (0.6301) −0.142 (0.7361) −0.032 (0.2652) 0.031 (0.8708) −0.359∗∗ (0.0151) 0.211 (0.4856) 0.126 57

Behavioral Finance: A Novel Approach – 9in x 6in

LOG Assets

−0.163 (0.6262) −0.227∗∗ (0.0201) −3.541∗ (0.0842) 0.364∗∗ (0.0111) −0.874 (0.3419) 0.143 (0.8355) 3.282∗∗ (0.0265) 0.225 57

Bonus Comp Growth

10:10

Chair Related

Fixed Comp Growth

Influential CEO and Board Behavior in Reaction

Total Comp Growth

b3964-ch02

(Continued)

41 page 41

(Continued )

ROA Complexity Indep Dir Expert Dir

R2 Obs.

0.355

−1.098

Equity Comp Growth

Cash Comp Growth

Incentive Comp Growth

0.126

0.491

−0.700

Δ Fixed Comp Ratio −0.086

Δ Bonus Comp Ratio

Δ Cash Comp Ratio

−0.034

−0.120

(0.0344) −0.196∗∗

(0.1119) −0.067

(0.6028) −0.224

(0.6895) 0.026

(0.1722) −0.176∗∗

(0.6847) −0.198

(0.3844) 0.039

(0.7482) −0.036

(0.1424) 0.003

(0.0344) −2.435 (0.2148) 0.345∗∗ (0.0119) −0.631 (0.4751) 0.082 (0.9012) 2.542∗ (0.0790) 0.289 57

(0.2109) 0.737 (0.5177) 0.267∗∗∗ (0.0011) −0.181 (0.7245) −0.003 (0.9946) 0.557 (0.5037) 0.260 57

(0.6459) −7.005 (0.4924) −0.717 (0.3428) 5.377 (0.2720) −0.601 (0.8702) 4.571 (0.5494) 0.050 52

(0.7124) −0.004 (0.9978) 0.006 (0.9581) −0.934 (0.1837) 1.847∗∗∗ (0.0007) −1.355 (0.2311) 0.238 55

(0.0441) −2.865 (0.1232) 0.227∗ (0.0746) −0.398 (0.6325) −0.523 (0.4014) 2.924∗∗ (0.0334) 0.236 57

(0.6140) −4.574 (0.5821) −0.434 (0.4690) 3.875 (0.3085) 1.076 (0.7034) 2.629 (0.6687) 0.044 56

(0.1010) 0.810 (0.1139) 0.012 (0.7256) 0.024 (0.9150) −0.089 (0.6029) −0.567 (0.1294) 0.106 57

(0.1570) −1.073∗ (0.0530) −0.043 (0.2504) −0.023 (0.9252) −0.262 (0.1587) 0.889∗∗ (0.0291) 0.143 57

(0.8802) −0.263 (0.5274) −0.031 (0.2798) 0.001 (0.9945) −0.351 ∗∗ (0.0154) 0.322 (0.2923) 0.163 57

b3964-ch02

(Intercept)

0.819∗∗

Bonus Comp Growth

Behavioral Finance: A Novel Approach – 9in x 6in

LOG Assets

Fixed Comp Growth

Behavioral Finance: A Novel Approach

CEO and Chair Related

Total Comp Growth

10:10

Panel C: Combined CEO and Chair Influence

November 4, 2020

42

Table 5:

page 42

November 4, 2020

Panel D: CEO Sole Influence

ROA Complexity Indep Dir Expert Dir (Intercept)

4.696∗∗

−0.005

−0.201

(0.9878) −0.237∗∗ (0.0152) −3.351 (0.1020) 0.359∗∗ (0.0125) −0.848 (0.3566) 0.140 (0.8390) 3.301∗∗ (0.0287)

(0.3226) (0.0102) −0.094∗ 0.048 (0.0826) (0.9169) 0.541 −10.332 (0.6369) (0.2776) 0.281∗∗∗ −0.904 (0.0008) (0.2026) −0.278 5.282 (0.5921) (0.2466) 0.024 −0.920 (0.9502) (0.7878) 1.045 0.410 (0.2139) (0.9540)

Equity Comp Growth −0.051 (0.8493) 0.020 (0.7849) −0.068 (0.9638) 0.005 (0.9624) −0.951 (0.1759) 1.863∗∗∗ (0.0007) −1.236 (0.2792)

Cash Comp Growth 0.093 (0.7758) −0.195∗∗ (0.0283) −3.513∗ (0.0628) 0.231∗ (0.0747) −0.527 (0.5316) −0.489 (0.4404) 3.302∗∗ (0.0176)

Incentive Comp Growth

Δ Fixed Comp Ratio

3.856∗∗∗ −0.162∗ (0.0062) 0.045 (0.9022) −7.555 (0.3270) −0.619 (0.2696) 4.201 (0.2341) 1.030 (0.6932) −1.290 (0.8224)

(0.0658) 0.035 (0.1290) 1.070∗∗ (0.0323) 0.017 (0.6182) 0.045 (0.8396) −0.094 (0.5738) −0.517 (0.1504)

Δ Bonus Comp Ratio 0.160∗ (0.0898) −0.026 (0.2852) −1.197∗∗ (0.0262) −0.050 (0.1746) −0.012 (0.9604) −0.266 (0.1414) 0.731∗ (0.0606)

Δ Cash Comp Ratio −0.002 (0.9795) 0.009 (0.6589) −0.127 (0.7639) −0.033 (0.2622) 0.033 (0.8629) −0.360 ∗∗ (0.0151) 0.213 (0.4886)

b3964-ch02

(Continued)

Behavioral Finance: A Novel Approach – 9in x 6in

LOG Assets

Bonus Comp Growth

10:10

Only CEO Related

Fixed Comp Growth

Influential CEO and Board Behavior in Reaction

Total Comp Growth

43 page 43

November 4, 2020

44

10:10

Table 5:

(Continued)

Panel D: CEO Sole Influence Bonus Comp Growth

Equity Comp Growth

Cash Comp Growth

0.221 57

0.236 57

0.176 52

0.236 55

0.208 57

Incentive Comp Growth 0.178 56

Δ Fixed Comp Ratio

Δ Bonus Comp Ratio

Δ Cash Comp Ratio

0.152 57

0.190 57

0.126 57

b3964-ch02

Notes: Related CEO and related chair are Dummy variables equaling one if the CEO or chair is related, respectively, else zero. CEO and chair related is a dummy variable equaling one if the CEO and the chair are related, else zero. Only CEO Related is a dummy variable equaling one if the CEO is related and the chair is not, else zero. Total Comp Growth is the growth rate in total comp for a given company (CEO) from 2010–2011 to 2013–2014. Fixed, Bonus, Equity, Cash, and Incentive Comp Growth are similarly calculated, Δ Fixed Comp Ratio is the change in the ratio of Fixed Comp to Total Comp from 2010–2011 to 2013–2014 for each company (CEO). Δ Bonus Comp Ratio and Δ Cash Comp Ratio are calculated similarly. Indep Dir and Expert Dir are the ratios of number of independent directors and number on board with financial expertise, respectively, to the total number of directors on the board. Log Assets is the natural log of Total Assets; ROA is ratio of Net Income to Total Assets; and Complexity is the natural log of the total number of consolidated subsidiaries plus 1. Independent variables are calculated at the beginning of the period while the dependent variable is calculated at the end of the period. ∗∗∗ , ∗∗ and ∗ denote significance at the 0.01, 0.05 and 0.1 level.

Behavioral Finance: A Novel Approach – 9in x 6in

Fixed Comp Growth

Behavioral Finance: A Novel Approach

R2 Obs.

Total Comp Growth

page 44

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 45

45

CEOs to examine the impact of CEO influence on post-amendment compensation changes. It can be observed that related CEOs had greater Total Comp Growth and Bonus Comp Growth than unrelated CEOs by 43% and 240%, respectively. These results are marginally significant at the 11% and 13%, respectively. It can also be observed that related CEOs had the composition of their compensation shift (more so, compared to unrelated CEOs) from fixed to non-fixed (incentive) compensation by 15%. This result is significant at the 5% level. Overall, regressions results support our correlation and univariate results. Following the enactment of Amendment 20, there was a shift in the compensation of influential CEOs from fixed compensation to bonus compensation. The greater growth rate in incentive compensation, defined as bonus plus stock-based compensation, can be attributed to the increase in bonus compensation. Further, the point estimate on the coefficient on equity-based pay is both statistically and economically in significant providing no evidence that influential CEOs increased their bonding to shareholders post-amendment through equity-based pay. Panel B contains the results for related chairs. It can be seen that in firms with related chairs, CEOs had lower Bonus Comp Growth, significant at the 5% level. The positive and negative changes in Fixed Comp Ratio and Bonus Comp Ratio, respectively, although only marginally significant (at the 12% and 11% level, respectively), are consistent with a lesser shift to bonus pay in companies with related chairs relative to those where the chair is not related. Overall, regression results support our earlier correlation and univariate results that influential chairs temper CEO bonus increases following the amendment. Panel C presents the results for the firms where both CEO and chair are related. In this case, CEOs total compensation grew faster through the fixed component (36%), leading to 82% greater growth in Total Comp compared to CEOs in the other subcategories. The 49% greater growth in the cash compensation supports the notion that the increase came from the fixed component of compensation and less so from bonuses. Panel D presents the results for the firms where only the CEO is related. We can see that in this case, when CEO influence is strongest (i.e., the CEO is related and the chair is not), the growth in the bonus

November 4, 2020

46

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

component (compared to other CEOs) is very high (470%). The intuition is that the most influential CEOs shift their pay from the fixed component to the bonus component. This is further supported by the decrease in the Cash Comp Ratio and increase in the Bonus Comp ratio relative to all other CEOs.

5.

Concluding Remarks

This chapter explores the impact of a new regulation, Amendment 20, on executive compensation in Israel. The Amendment is intended to curb excess CEO compensation and tie it to firm performance. We investigate whether the regulation met its goals, differentiating between the behavior of CEOs who are able to influence their postamendment compensation and those who do not have this ability. The concept of influence is based on whether the CEO is a related party as reported in the firm’s annual financial statements, deriving from the concentrated ownership structure in Israel and the notion that related parties might have an impact on the selection of seemingly independent directors. The analysis is conducted using handcollected data on firms traded on TASE in a 2-year window around 2012, the year in which the Amendment was signed into law. We find that the most influential CEOs are able to impact the redesign of their post-amendment compensation by shifting it away from the fixed component of compensation (mostly salary) to the bonus component of the compensation. We also document that influential chairmen are able to temper the bonus of unrelated CEOs. Also, we are unable to find any meaningful increases in equity-based compensation post-amendment. This is consistent with influential CEOs increasing their incentive-based pay, not by connecting it to market prices but rather by increasing their bonus cash compensation. Clearly, given the media and public outrage it would have been difficult, even for influential CEOs, to increase their salary or benefits. The bonus component is somewhat flexible, unlike an incentive component contingent upon market prices, and yet passes the hurdle of public and media scrutiny. In this way, companies with influential CEOs are able to shift the composition of pay from cash salaries to cash bonuses in response to the Amendment, visibly complying with the intent of the Amendment, however, leaving the CEO better off.

page 46

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 47

47

Acknowledgments We wish to thank Keren Kibovich of Keren Kibovich Consulting & Management and Moshe Peress, Head of Accounting Consulting Services at PwC Israel for insightful discussions. Excellent Research assistantship by Moti Barashi is highly appreciated.We would also like to thank Kobi Avramov of Tel-Aviv Stock Exchange for useful information. We acknowledge the financial support from the Academic College of Tel Aviv-Yaffo. References Abudy, M., Amiran, D., Rozenbaum, O., and Shust, E. (2017). Do executive compensation contracts maximize firm value? Evidence from a quasi-natural experiment. Columbia Business School Research Paper No. 17–69. Appelberg, S. (2012). Israel Securities Authority: Firms are performing a last-minute coup in top management compensation packages just before the law comes into effect. The Marker, December 3. Bae, K.-H., Gong, Z., and Tong, W. H. S. (2019). Restricting CEO pay backfires: Evidence from China. Working Paper. Balsam, S. (2002). An Introduction to Executive Compensation. San Diego, CA: The Academic Press. Balsam, S. and Ryan, D. H. (2007). Limiting executive compensation: The case of CEOs hired after the imposition of 162 (m). Journal of Accounting, Auditing, & Finance, 22, 599–621. Barro, J. R. and Barro, R. J. (1990). Pay, performance, and turnover of bank CEOs. Journal of Labor Economics, 8, 448–481. Bebchuk, L. A. (2007). Written testimony submitted by Professor Lucian A. Bebchuk. In Empowering Shareholders on Executive Compensation: H.R. 1257, the Shareholder Vote on Executive Compensation Act, 110–10, 65–73. U.S. House of Representatives. Hearing before the House Committee on Financial Services, 110th Congress. Bebchuk, L. and Grinstein, Y. (2005). The growth of executive pay. Oxford Review of Economic Policy, 21, 283–303. Brick, I. E., Palmon, O., and Wald, J. K. (2006). CEO compensation, director compensation, and firm performance: Evidence of cronyism? Journal of Corporate Finance, 12 (3), 403–423. Cebon, P. and Hermalin, B. E. (2015). When less is more: The benefits of limits on executive pay. The Review of Financial Studies, 28, 1667– 1700.

November 4, 2020

48

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch02

Behavioral Finance: A Novel Approach

Coles, J. L., Daniel, N. D., and Naveen, L. (2014). Co-opted boards. The Review of Financial Studies, 27 (6), 1751–1797. Dittmann, I., Maug, E., and Zhang, D. (2011). Restricting CEO pay. Journal of Corporate Finance, 17, 1200–1220. Editorial (2012). Executive compensation 2011: While rank-and-file employees’ salaries are not enough to carry them through the month, senior executives are celebrating. The Marker, March 23. Edmans, A. and Gabaix, X. (2016). Executive compensation: A modern primer. Journal of Economic Literature, 54 (4), 1232–1287. Edmans, A., Gabaix, X., and Jenter, D. (2017). Executive compensation: A survey of theory and evidence. National Bureau of Economic Research, Working Paper No. 23596. Geddes, R. R. (1997). Ownership, regulation, and managerial monitoring in the electric utility industry. Journal of Law and Economics, 40, 261–288. Hauser, S. (2013). Speech as the Chairman of Israel Securities Authority, Executive Compensation Conference, Israel Securities Authority, January 7 (in Hebrew): http://www.isa.gov.il/%D7%94%D7%95%D7 %93%D7%A2%D7%95%D7%AA%20%D7%95%D7%A4%D7%A8%D7 %A1%D7%95%D7%9E%D7%99%D7%9D/178/Pages/ExecutiveCompensationConf.aspx Hubbard, G. and Palia, D. (1995). Executive pay and performance: Evidence from the U.S. banking industry. Journal of Financial Economics, 39 (1), 105–130. Hwang, B. H. and Kim, S. (2009). It pays to have friends. Journal of Financial Economics, 93 (1), 138–158. Jensen, M. C. and Murphy, K. J. (2018). The politics of pay: The unintended consequences of regulating executive compensation. Working Paper. Joskow, P. L., Rose, N. L., and Wolfram, C. D. (1996). Political constraints on executive compensation: Evidence from the electric utility industry. Rand Journal of Economics, 27 (1), 165–182. Justice Department background and explanations to Amendment 20 to Israel Companies Law, November 8, 2012 (in Hebrew): https://www. justice.gov.il/Pubilcations/News/Pages/ScharBchirim.aspx. Kaplan, S. N. (2007). Testimony of Steven N. Kaplan: Are US CEOs overpaid? In Empowering Shareholders on Executive Compensation: H.R. 1257, the Shareholder Vote on Executive Compensation Act, 110–10, 120–47. U.S. House of Representatives. Hearing before the House Committee on Financial Services, 110th Congress. Khanna, T. and Yafeh, Y. (2007). Business groups in emerging markets: Paragons or parasites? Journal of Economic Literature, 45 (2), 331–372.

page 48

November 4, 2020

10:10

Behavioral Finance: A Novel Approach – 9in x 6in

Influential CEO and Board Behavior in Reaction

b3964-ch02

page 49

49

Kosenko, K. (2008). Evolution of business groups in Israel: Their impact at the level of the firm and the economy. Bank of Israel Discussion Paper (in Hebrew). Larcker, D. F., McCall, A. L., Ormazabal, G., and Tayan, B. (2012). Ten myths of “say on pay.” Rock Center for Corporate Governance at Stanford University Closer Look Series: Topics, Issues and Controversies in Corporate Governance No. CGRP–26, Stanford Graduate School of Business. Organisation for Economic Co-operation and Development (OECD) (2011). Corporate Governance in Israel, 2011. Paris: Corporate Governance, OECD Publishing. Perry, T. and Zenner, M. (2001). Pay for performance? Government regulation and the structure of compensation contracts. Journal of Financial Economics, 62, 453–488. Rose, N. L. and Wolfram, C. (2002). Regulating executive pay: Using the tax code to influence chief executive officer compensation. Journal of Labor Economics, 20, S138–S175. Rose, N. L. and Wolfram, C. (2000). Has the “million-dollar cap” affected CEO pay? The American Economic Review, 90 (2), 197–202. Shivdasani, A. and Yermack, D. L. (1999). CEO involvement in the selection of new board members: An empirical analysis. Journal of Finance, 54 (5), 1829–1853. Sikoler, N. (2013). When CEO pay increases by 330%. Calcalist, April 2. Wolfram, C. (1998). Increases in executive pay following privatization. Journal of Economics and Management Strategy, 7, 327–361.

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0003

Chapter 3

Aiming for the Real-Estate Market But Hitting the Stock Market — An Event Study Analysis of Israeli Mortgage Reforms Yaron Lahav∗ , Sara Arbel† and Aliza Mizrahi†

Abstract The massive real-estate price increases experienced in Israel over the last several years have elicited in policymakers the realization that they need to take action to reduce housing demand and prevent the Israeli housing market from collapsing. As a result, during May and October 2010, the Bank of Israel stepped in and increased the effective rate on mortgages and lowered the number of qualified applicants. In announcing the new regulations, the Bank of Israel’s main objective was to halt rising demand and to prevent further growth of the housing bubble. We use event study analysis to show that not only did the new regulations have no effect on housing prices but they in fact also markedly influenced the market value of the real-estate companies traded on the Tel-Aviv stock exchange. Keywords: Announcements, event study, real estate, reforms, regulations

∗

Corresponding author. Assistant Professor of Finance, Department of Business Administration, Guilford Glazer Faculty of Business and Management, Ben-Gurion University of the Negev, P.O. Box 653, Beer-Sheva 84105, Israel; [email protected]. † Ben-Gurion University of the Negev, P.O. Box 653, Beer-Sheva 84105, Israel; [email protected] (Sara Arbel), [email protected] (Aliza Mizrahi). 51

page 51

November 4, 2020

52

1.

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

Introduction

Since 2008, when the most recent real-estate bubble in the US burst, its long-term impacts have been felt by both the US and global economies, forcing policymakers around the world to focus on realestate prices. Many economists interpret real-estate price increases as a signal that market participants are behaving irrationally, and the inevitable outcome of such price bubbles in other parts of the world will be similar to what happened in the US during 2008. A country currently experiencing such a real-estate bubble is Israel. Ironically, real-estate prices in Israel have been skyrocketing since 2007, the year that symbolized the end of the real-estate bubble in the US. After a long period of decreasing prices, 2007 showed a 3% increase in Israeli real-estate prices, followed by 11%, 20% and 14% increases in the following years.1 As always, experts are in dispute regarding the drivers for this real-estate bubble, but possible reasons can be the relatively low interest rates in Israel, combined with the fact that the Israeli economy was one of the few that remained stable during the crisis. Interestingly, the reason for this economic stability in Israel is attributed to the strength of its banking system. In a sense, this system and its structure has an important role in this chapter. Several attempts have been made by Israeli policymakers to eliminate this phenomenon. The Israeli government,2 on the one hand, amended existing laws and enacted ad hoc regulations to increase the supply side of the market. One example is a temporary order (effective from January 1, 2011 to December 31, 2012) that exempted the seller of an apartment from paying betterment tax.3 Another example includes Israeli government’s intentions to double property tax on empty properties and to fine any real-estate company found to have been delaying construction deadlines.4

1

Source: The ICBS. Since most actions involved changes in taxes or fees, the Ministry of Finance (MoF) was involved. 3 Proposed real-estate taxation (appreciation and acquisition), law amendment no. 71. 4 As published by The Marker, an Israeli financial newspaper, on March 18, 2012 (http://www.themarker.com/realestate/1.606460). 2

page 52

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 53

53

In contrast, the Bank of Israel (BoI), through its Banking Supervision Department, focused on decreasing the demand side by issuing new regulations intended to limit the ability of potential buyers to apply for mortgages and to increase the effective mortgage rates for those still able to apply.5 This chapter investigates the two most important and meaningful announcements made by the head of the Banking Supervision Department (i.e., the Supervisor of Banks)6 at the time, Rony Hizkiyahu. These announcements were the second and third out of a total of four made by the banks supervisor between February 2010 and April 2011. Published in February 2010, the first of the four announcements involved the formation of “purchasing groups,” such that potential home buyers could become the managers of their own apartment building construction site. The second announcement, from May 2010, increased the minimum down-payment required by buyers from 30% to 40% of the property value (as such lowering the maximum loan-to-value ratio — or LTV — to 60%). The third, announced on October 24, 2010, imposed significant limitations on potential home buyers by increasing mortgage costs for almost all applicants. The fourth announcement was made in April 2011, and it dictated that the maximum portion of the mortgage that could be taken with an adjustable rate be limited to one-third (and in some cases even one-fourth). Derived from the short-term interest rate determined by the Chairman of the BoI, this rate was at an all-time low (0.5%) towards the end of 2009, and it increased only slightly (0.75% by early 2014).7 These interest values attracted many potential buyers, both from Israel and overseas. By forcing the public to take mortgages with fixed rates, the BoI expressed its concerns about the relatively high risks home buyers take upon themselves and their seeming total ignorance of the virtual guarantee that future interest rates will increase. Indeed, the higher the portion of the mortgage

5

In Israel, the banks are the only mortgage providers. This gives the Supervisor of Banks the power to control the mortgage market by regulating the banks as mortgage providers. 6 Because the Banking Supervision Department is a department in the BoI, many attribute these announcements to Stanley Fischer, the governor of the BoI. 7 Average interest on deposits in the BoI. Source: The BoI.

November 4, 2020

54

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

subject to an adjustable rate, the greater the potential increase in monthly payments as a result of rising mortgage rates. On the supply side of the market, several real-estate companies engage in building, selling, marketing and investing in real estate for both the private and the business sectors. The 15 largest of such companies, which are traded on the Tel-Aviv Stock Exchange (TASE), form the Real Estate 15 index (RE15), a representation of the real-estate market index. As of the beginning of 2011, all 15 real-estate companies included in the RE15 were also included in the TA100, and two of the companies were included in the TA25 (the index of the 25 largest companies in Israel). The value of all RE15 company shares held by the public at the beginning of 2011 was 16.7 billion NIS,8 about 4% of the value of the TA100 shares held by the public.9 In this chapter, we examine the effects of the BoI’s second and third announcements on both the real estate and the stock markets. In particular, we are interested in learning whether the issuance of these announcements achieved BoI objectives, specifically preventing further price increases, or perhaps they caused “collateral” damage by diminishing the market values of real-estate companies. The importance of this distinction between the real estate and the stock markets is in understanding the connection between information processed by the public and that processed by investors. Real-estate company value should be determined by each company’s business operation and by the condition of the market. If the public trusts that an action taken by the regulator will affect the real-estate market, then we should expect the real-estate market to react, assuming that people tend to follow their beliefs. If these expectations indeed exist, we should expect the market values of real-estate companies to fall because the public’s belief that real-estate prices will go down should lead to investors believing that real-estate company profits will decline accordingly. If, on the other hand, real-estate prices do not react to the regulation, it means that the banks supervisor’s announcements had no effect on the public’s beliefs. In this case, we would expect investors’ beliefs to follow suit and real-estate

8 9

New Israeli Shekels (the Israeli currency). Source: TASE website.

page 54

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 55

55

stock prices to remain unchanged. As we shall see, although the BoI announcements did not affect housing prices (hence, the beliefs of home buyers were not affected), the third announcement elicited in investors a belief that an adverse outcome was inevitable. In contrast to investors, potential homeowners did not take this announcement as seriously as expected (by the BoI). In other words, although it was aiming for the real-estate market, the BoI missed and hit the stock market instead. Although this chapter investigates a relatively small market, it has some important international implications, especially in light of the attempt of policymakers around the globe to get “back on track” in terms of macroeconomic policy. First, there is a considerable amount of research that focuses on the effectiveness of monetary policy on real-estate prices. Before the economic crisis, economists believed that central banks should focus on inflation targeting (Giavazzi and Mishkin, 2006). Following the most recent housing bubble and its consequences, economists now tend to believe monetary policy by itself (merely central banks focusing on targeted inflation rate) is not enough to establish financial stability because this does not take into account (and therefore does not provide answers to) systemic risks. Allen and Rogoff (2010) argue that in addition to monetary policy enforced by the central bank, “macro-prudential” policy should also be used. This leads to the second point, namely who should execute this “macro-prudential” policy. While the straightforward answer should be that the government is responsible for macroeconomic conditions, this may not be clear because the regulated markets are financial. In Israel, the BoI took initiative after realizing that the government is not doing enough in this regard. Such action on behalf of the BoI is equivalent to a physician going to work without his toolbag. One reason for this unharmed intervention on behalf of the chairman of the BoI may be his recognition that the impact of the financial sector on economic stability is much greater than any other sectors.10 Third, there is the aspect of the type of regulations that a government chooses. Taylor (2010) argues that the US government, acting

10

An argument raised by Mishkin (2011).

November 4, 2020

56

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

as the financial regulator, became less “rule-based” and more “interventionist” prior to the economic crisis of 2008. It appears that the policy implication, and not only policy itself, are important to enforce regulations. The story presented in this chapter is one more example that “ad hoc” announcements, although easier to implement politically, may not be efficient. One reason may be the fact that such regulations are hard to predict, and when the market cannot adopt to changes fast (the real-estate market, for example), regulations simply do not work. Another reason may be that ad hoc regulations may signal some short-term distortion (the formation of a bubble, for example)that the government is trying to eliminate. As real-estate companies do not share the same interest, they are not in a hurry to cooperate.

2.

Literature Review on Event Studies

This chapter relies and builds on the existing research that tests the impact of regulations, reforms and announcements on financial variables. The literature on event studies can be traced back to the work of Dolley (1933), who examine how stock splits affect the prices of stocks. Since then, the methodology has been used fairly extensively.11 The next milestone in methodological development occurred during the 1970s, when researchers started questioning the validity of the statistical assumptions under which the event study methodology was used. Patell (1976) and Boehmer et al. (1991) suggest ways to account for event-induced variance of the observed stock returns. Brown and Warner (1980, 1985) conduct several event studies to investigate how the event affected sample volatility (variance), the cross-sectional dependence of stock prices (i.e., their tendency to co-move), and the auto-correlation in daily excess returns. Schwert (1981) also provides the theoretical background for using event study analyses when the event is represented by regulations. In addition to applying the event study methodology to measure the effect on stock prices of internal events in stock markets (see, for example, Jain, 1987; Cooper et al., 2001; Tumarkin and Whitelaw,

11

We refer the readers to a review by MacKinlay (1997).

page 56

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 57

57

2001; Li and McNally, 2003; Chen and Siems, 2004; Agrawal et al., 2006), many researchers have used it to test the effect of regulations or deregulations on stock prices. Henry (2000) shows, using event study analysis, that the stock market return reacts positively to market liberalization that allows the trading of foreign stocks in the market. Johnson et al. (2000), Carow and Heron (2002) and Howe and Jain (2004) use event study analysis to show how stock market prices are affected by new legislation. Carow and Kane (2002) use the methodology to measure both the costs and the benefits of deregulations for US financial institutions. With respect to monetary effects, Bomfim (2003) uses event study analysis to study the effect of policy and interest rate announcements on stock prices both before and after the announcements, while Miyajima and Yafeh (2007) show how different companies were affected by banking crises in Japan. Even sports have been examined: Veraros et al. (2004) show how the decision to stage the 2004 Olympic Games in Greece affected the stock exchange of Greece vs. that of Italy (as a country that lost the Olympic bid). Their research followed a similar study by Berman et al. (2000) on the 2000 Sydney Olympics. Finally, Jackson and Madura (2007) use event study analysis to show that the market relies less on profit warnings when market participants have quicker access to information. Thus, the literature reflects the wide extent to which event study analysis has been used and provides numerous examples that can be either replicated or adopted as a theoretical platform. The current study applies the methodology in a manner that is similar (and in one case even identical) to the procedure used in previous research.

3.

The BoI Announcements

Real-estate prices in Israel have been increasing rapidly since the second half of the 2000s, parallelly decreasing mortgage prices as interest rates reached an all-time low. An analysis of the average annual interest rate for mortgages in Israel by year shows that it has been decreasing constantly since the end of 2006 (Figure 1).12 Such

12

Source: The BOI.

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

58

Average Annual Mortgage Rates (in percentage)

8 7 6 5 4 3 2 1 0

Figure 1:

Average annual mortgage rate

Source: The BoI.

low mortgage rates attract both potential buyers who cannot afford homes when interest rates are higher and also real-estate investors who interpret them as an opportunity to make a cheap real-estate investment. But as mortgage prices in Israel have declined, the price of housing, as reflected in the Housing Price Index (HPI), has increased rapidly since about 2007 (Figure 2).13 This trend, however, does not seem to have affected the rate of new home building, which grew tremendously during the same period. Annual data published by the Israeli Central Bureau of Statistics (ICBS) on the creation of new homes show that the numbers of new homes built in Israel annually from 2008 to 2010 are 13,482, 16,258 and 17,917, respectively. These steady, yearly increases were countered slightly in 2011 when the number of new homes built decreased to 15,118, but a further reduction in that number was not expected in 2012. Naturally, these developments attracted the attention of Israeli policymakers. Worried about the long-term ramifications, they have 13

Source: ICBS.

page 58

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 59

59

Percentage Change in Housing Prices 100.00% 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Figure 2:

Annual percentage change in housing prices

Source: ICBS.

suggested plans of action and their opinions have been published by the media. Meanwhile, the BoI — the sole regulator of the Israeli banking system — focused its efforts on the Israeli banks and their function in the housing market as the only mortgage providers. Accordingly, it announced four new regulations between February 2010 and April 2011. All the announcements concerned mortgages, but the ultimate goal was to reduce the prices in the housing market. The first announcement of February 17, 2010, involved the formation of “purchasing groups.” A purchasing group comprises a party of potential home buyers (organized either by one of them or by a professional organizer) who, instead of buying their homes individually from a contractor, organize to build their own apartment building. Accordingly, the group members become financially responsible for their own project. Together they locate the property, find contractors and assume responsibility for every aspect of the project from start to finish. Their management of the project includes, but is not limited to, the purchase of the property and all building materials and negotiations with suppliers. For the home buyers, this approach facilitates marked reductions in the marketing costs and markups

November 4, 2020

60

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

traditionally earned by contractors and enables them to reap the corresponding benefit of much cheaper homes. In addition to these benefits, however, a purchasing group also assumes additional risks, such as price increases in material inputs and poor management skills due to inexperience, among others. In recognition of the tremendous risks taken by people who decide to organize in purchasing groups, the BoI changed the definition of the loans they received from the banks. Beginning the day of the announcement, such loans were no longer classified as mortgages. Instead, the banks will provide these loans as if the borrower is a contractor, and therefore, the typical terms and conditions associated with mortgages no longer apply to these new loans. Indeed, the rates are higher and loan durations are much shorter, usually only a few years (vs. the not uncommon payback time of 30 years). The BoI’s second announcement, made on May 24, 2010, required banks to allocate additional funds to doubtful accounts for every new,approved mortgage with a loan-to-value (LTV) higher than 60%. Essentially, this approach increases the costs of risky mortgages, i.e., those with LTV higher than 60%, which incur greater expenses for the banks, and as such, they are more expensive for the borrowers. By enacting this regulation, the BoI emphasized the risk that banks take when providing mortgages with low down payments. But the most important announcement was published on October 25, 2010, and it applied to mortgages larger than 800,000 NIS with LTV greater than 60% and with at least 25% of the loan carrying adjustable interest, i.e., the vast majority of housing mortgages.14 The policy change entailed in the announcement was meant to control the increased risk associated with such mortgages. To offset that risk, lending banks are now required to hold more cash — today 100% of the loan value vs. 35% prior to the regulation — in their reserves. Thus with this announcement, the BoI aimed to increase the effective prices of these mortgages in an effort to reduce the housing demand. The last announcement, published in April 2011, required that at least two-thirds of any amount borrowed as mortgage be subject to a fixed interest rate. This announcement applies to all mortgages

14

Unfortunately, we do not have numerical data to support this statement.

page 60

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 61

61

and in fact expresses BoI concerns about the risk borne by home buyers when they assume that the interest rate will remain low in the long run. As purchasing groups constitute only a small portion of the demand for mortgages,15 it is assumed that the intention of the first announcement was not to affect the demand for housing, but rather, to reduce or eliminate the phenomenon of purchasing groups. In addition, since fixed rates were low as well, the fourth announcement did not have a significant impact on housing demand and therefore on the housing market. We have therefore decided to focus on the second and third announcements in our study.

4. 4.1.

Data and Methodology Event Study Methodology

As mentioned, event study is used to measure the effect of some event on the return of a known index, such as stock prices. Accordingly, the use of the event study methodology is based on a model that captures return behavior. Using this model, actual post-event behavior of that index return is compared to its expected (i.e., the expected values of the index return assuming no event). The first stage of an event study analysis is to determine the timing of the event and consequently the event window. As simple as that sounds, in many cases events may start earlier than defined. For the simplest case, take profit announcements. It is possible that some informed investors already took action several days before the official announcement, and therefore, prices before the announcement already incorporate the new information. This eventuality can be tested by starting the event window several days before the announcement. Indeed, researchers usually test several event windows during an event study. In the second stage, the researcher determines an estimation window, which represents a period of time that occurred prior to the event and during which the tested index was not affected by the 15

Approximately 600 per year in Tel Aviv during the years 2006–2008, according to the State Revenue Administration.

November 4, 2020

62

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

page 62

Behavioral Finance: A Novel Approach

event. Data pooled from this time period are used to estimate the parameters of the model discussed before. During the third stage, the calibrated model is used to estimate return behavior during the event window, assuming that the event would not have taken place. A measure called abnormal return (AR) is calculated as the difference between the value of the actual return and the estimated value using the model for every time period in the event window. AR is then used to estimate a statistic to test the hypothesis that there is no significant difference between the actual and estimated values of the returns. In our research, we separately tested the behavior of two indexes — the HPI, published monthly by the ICBS, and the RE15, a daily index of the average return of the 15 real-estate companies discussed in Section 1. Accessible via the ICBS website, the HPI is based on transactions data collected by the state of Israel16 and summarizes the monthly average percentage change in housing prices. The first year covered by the index is 1994. The RE15 index has been published by the TASE since 2005 and includes the 15 stocks with the highest market values in the real-estate industry. Because all the companies are heavily involved in the Israeli real-estate business, they should be affected by developments in this local market. 4.2.

Housing Price Index

To characterize HPI behavior and its response to BoI announcements, we used the following model: ΔHPI t = α + β1 × CP It + β2 × rt + β3 × πt + εt

(1)

where ΔHPI t is the percentage change in the real17 HPI in month t compared to t − 1, CP It is the percentage change of the consumer price index, rt is the short-term interest rate determined monthly by the BoI, πt is the unemployment rate and εt is the error term. We use the short-term interest rate instead of the long-term rate because the short-term rate is the one published monthly, announced on the 16 According to Israeli law, all real-estate transactions are made and registered by the state of Israel. Therefore, the HPI includes virtually all transactions. 17 The HPI is published in nominal terms. We divided the percentage change by the percentage change of the CPI to measure the HPI in real terms.

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 63

63

news and therefore available to the public. In fact, we could not find data on long-term mortgage rates on any publically available data source. A potential buyer would receive data on long-term rates for the first time only in a scheduled appointment with a bank representative when applying for a mortgage. Even then, the rates offered are not final and are subject to negotiation (which the potential home buyer is not always aware of). Banks put extreme effort so that offers to potential home buyers would not be used as reference for offers from other banks.18 In sum, we assume that the change in housing prices is affected by variables that reflect inflation, the perceived price of mortgages (short-term interest rate) and a measure for economic condition (unemployment rate). We expect β1 to be positive, because high inflation should drive investors away from financial assets (that usually pay nominal returns) towards investments in real assets like real estate (that preserve real returns). We expect β2 to be negative because interest rates represent actual mortgage prices, an assumption that implies a negative relationship. Finally, we also expect β3 to be negative because a high unemployment rate is associated with bad macroeconomic conditions, which lower the demand for investment in general and real estate in particular. To check for structural break we implement the event study methodology.19 We defined our estimation window from March 199520 to December 2009, intentionally not including months from 2010 to ensure that the announcements from that year would not affect Eq. (1) parameter estimation. The values of the parameters from Eq. (1) are presented in Table 1. As the table shows, all parameters are significant in the 5% level and have the expected signs. We then used our findings to calculate ARt : ARt = ΔHP It − (α + βˆ1 × CP It + βˆ2 × rt + βˆ3 × πt )

(2)

18 However, we did contact the BoI and asked for the data. Results using longterm instead of short-term rates were the same for the period of six months after the events. 19 We could not use the traditional tests for structural break like the Chow test, Wald test or the classic F test because of insufficient number of post-event samples. 20 The calculation of the HPI was changed starting that month, so we excluded prior data.

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

64

Table 1:

Results of Regression 1: The HPI

CPI Interest Rate Unemployment Rate Intercept R sq. adjusted Number of observations

Coef.

St. Err.

82.625∗∗∗ −0.045∗∗ −0.152∗∗∗ 1.951∗∗∗ 0.264 178

13.491 0.018 0.043 0.503

Notes: This table reports coefficients of a pooled OLS regression expressed by Eq.(1). The dependent variable is the change in the HPI in real terms. The independent variables are: percentage change in CPI, the interest rate and the unemployment rate. Note: Coefficients marked with ∗∗∗ and ∗∗ are statistically significant at the 1% and 5% level, respectively. Table 2:

CARs for event periods — the housing prices May Announcement

CAR T

October Announcement

1–5 months

1–6 months

1–12 months

−0.40 −0.201

0.71 0.294

(4.82) (1.413)

Note: This table reports the CARs of the HPI.

Next, we tested the effects of the announcements of May and October on several event periods, as described in Table 2. Unlike typical event studies that test short-term effects, we assumed that it may take more time for the housing market to respond. We therefore focused our tests on the five-month period that followed the May announcement (these five months were the longest time interval we could test before the October announcement). As for the October announcement, we focused both on the six-month and on the 12-month periods that began one month after the announcement (November 2010).21 21

When announcing the new exemption from betterment tax between January 1, 2011 and December 31 2012, the Israeli minister of the treasury, Yuval Steinitz, expected to see effect on housing prices within six months at the most (as published by Calcalist, a local Israeli financial newspaper, on March 31, 2011: http:// www.calcalist.co.il/real estate/articles/0,7340,L-3519504,00.html).

page 64

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 65

65

To form the statistical measure, we used the approach presented in Brown and Warner (1985): CARi √ J= ¯ S(ARt ) m where

 S(ARt ) = (ARt − AR)/(n − 1)

(3)

(4)

t

1 ARt n t  ARi CARi = AR =

(5) (6)

i

where t represents months in the estimation window, n is the number of months in the estimation window (in our case n = 182), i is the months in the event window and m is the number of months in the event window. The values of CARi are provided in Table 2, along with their T values. As can be seen from the table, there is not even one event window with housing prices significantly lower than expected. We therefore conclude that, up to five months after the May announcement and up to one year after the October announcement, neither of the BoI announcements affected housing prices. 4.3.

Real-Estate Stock Index

To characterize RE15 behavior and its response to the BoI announcements, we used the following market model: ΔRE15t = γ + δ × T A100t + εt

(7)

where ΔRE15t is the daily return of the RE15 index and TA100t is the daily return of the market index, the latter of which includes the 100 stocks with the highest market values traded on the TASE and is considered the local market index. This market model is widely used in event studies that test the effects of events on stocks. The parameter δ reflects the sensitivity of the RE15 index return to changes

November 4, 2020

66

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

in the return of the TA100 index. We naturally expect δ to be positive. Our estimation window for the May announcement includes 80 trade days that end on the trade week prior to the announcement. The estimation window for the October announcement includes the 74 trade days from July 1, 2010, to the last trading day of the week prior to the announcement. We did not use data prior to July 1 because we did not want to include daily returns that may have been affected by the May announcement. The parameter values are presented in Table 3. As the table shows, the parameters δ1 and δ2 are significant at the 5% level, and each has the expected sign. We then used our findings to calculate ARt and CARt independently for the 15-day, 22-day and 30-day periods after the announcement. As Table 4 shows, both the 22-day and 30-day periods after the October announcement exhibit lower returns than expected, and both are significant at the 5% level. We therefore conclude that the October announcement of the banks supervisor did affect RE15 returns, implying that the real-estate companies were affected by this announcement. The May announcement, on the other hand, did not have any significant effect on the return of the RE15 index. Table 3: Results of Regression 2: The real-estate stock index Coef.

St. Err.

May Announcement δ1 Intercept R sq. adjusted Number of observations October Announcement

1.051∗∗∗ 0.001

δ2 Intercept R sq. adjusted Number of observations

1.112∗∗∗ −0.001∗∗∗ 0.713 74

0.076 < 0.001

80 0.079 0.001

Notes: This table reports coefficients of a pooled OLS regression expressed by Eq. (7). The dependent variable is the return on the RE15 index. The independent variable is the daily return on the TA100 index. Note: ∗ 10% level; ∗∗ 5% level: ∗∗∗ 1% level.

page 66

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms Table 4:

page 67

67

CARs for event periods — the real-estate stock index May Announcement

October Announcement

1–15 days 1–22 days 1–30 days 1–15 days 1–22 days 1–30 days CAR T

0.02 0.579

0.01 0.191

0.03 0.497

−4.23 −1.926

−6.31 −2.371

−5.59 −1.799

Note: This table reports the CARs of the RE15. Bold figures are significant in 5% level of confidence.

5.

Discussion

During 2010 and 2011, the Supervisor of Banks in Israel directed four announcements at the Israeli real-estate market. According to statements made by policymakers, at least two of these were expected to be effective. Their collective aim was to halt the increase in realestate prices in the Israeli market, but the results show that they, in fact, had no effect on housing prices. In this section, we try to elucidate why the housing market did not react as anticipated. First, a potential reason for this outcome is the lack of firm response by the Israeli government. Surprisingly (and maybe consequently) the BoI was the first agency to take initiative and respond to the situation with financial regulations, rather than waiting for the government to respond with housing regulations. The poor (or even lack of) response on behalf of the government was not for a lack of government bodies with authority in housing-related matters. Over 90% of Israeli land is managed by the Israel Land Administration (ILA), a government agency empowered by the law to protect, supervise, plan, develop and manage this land. In addition, the Ministry of Construction and Housing (MoCH) has the authority to decide, among other things, how many real-estate projects will be executed and for what reasons. The Ministry of Internal Affairs (MoIA) has the power to determine the level of property taxes and allocation of municipal resources. Finally, the MoF effect on both the housing and mortgage markets is significant, as the body responsible on the Israeli Tax Authority. But for whatever reason the first authority to respond was the body in charge of monetary policy, the BoI, which through its bank supervisory department issued new regulations for the banks. In a world where the options for borrowing have become progressively easier and more accessible, responding to a market issue with limitations

November 4, 2020

68

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

on the ability to borrow (and therefore to purchase housing) may be considered a mild response. In other words, we argue that the BoI’s decision to act in the mortgage market in order to influence housing prices turned out to be ineffective, because the BoI does not have the proper instruments to enforce such policy. The BoI attempt to affect housing prices is the same as scratching the right ear with the left hand. While heroic, it did not have the desired effect because it was not the “right hand.” Second, the market structure may have weakened the regulator’s effectiveness. With no data to support this argument, it is possible that the combination of rising Israeli housing prices with decreasing prices in other housing markets enhanced the attractiveness of the Israeli housing market to investors, especially foreign investors. These buyers do not apply for mortgages, at least not in the local market. If this is the case and these foreign investors determine the market prices de facto (by having a significantly higher income and a greater willingness to pay), then not only will an intervention in the mortgages market fail to provide a solution but it will also hurt more than help those meant to be protected by the regulations (i.e., young middle class families). Third, there may indeed be a housing bubble in the Israeli market. And when confronted with a bullish market, any attempt to intervene may effect a reaction that is the opposite of what is desired. In other words, attempts to halt rising market prices can signal investors that policymakers expect prices to increase in the future (otherwise, why intervening?). There may be other reasons, but on top of all, our analysis show that the announcements and their corresponding regulations had no effect on real-estate prices. This outcome motivated us to analyze the effect these announcements had on the stock market. Our findings show that the correlation between the RE15 and TA100 is relatively high (Figure 3) (beta is approximately 1.00). In contrast to the realestate investors, who did not believe the announcement would affect real-estate prices, investors in the stock market apparently took the second announcement of the BoI seriously. Intuitively, stock market investors should believe that the announcement will influence the stock market only if they believe that real-estate investors believe that, too. And finally, we also conclude that stock market investors overreacted to the third announcement published by the BoI.

page 68

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 69

69

120.0% 100.0% 80.0% 60.0% 40.0% 20.0%

RE15 0.0%

TA100 -20.0%

Figure 3:

Time Values of RE15 and TA100

Source: The TASE.

6.

Conclusion

This chapter adds to the literature on investors’ reactions to news or regulations. These studies usually exploit event study analysis to find ARs (either positive or negative) in response to the news or regulation. In this chapter we showed that while the housing market did not react to the regulations, stock market investors did. We found that the real-estate stocks index responded negatively to the second announcement, and this manifested as negative ARs of more than 6% in 22 days. We also discuss the possible reasons for the respective behaviors of potential home buyers and of stock market investors, emphasizing the overreaction of the latter. In fact, if real-estate prices are indeed determined by real-estate investors (i.e., the ones with greater willingness to pay), then we may have two groups of investors that share different expectations. Outside the scope of this research, this question should be explored further. Moreover, it raises an additional question — can the type of investment influence how expectations are formed? Or is it simply that stock market investors overreact to news?

November 4, 2020

10:11

70

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

Behavioral Finance: A Novel Approach

References Agrawal, M., Kishore, R., and Rao, R. H. (2006). Market reactions to E-Business outsourcing announcements: An event study. Information and Management, 43, 861–873. Allen, F. and Rogoff, K. (2010). Asset prices, financial stability and monetary policy. Paper presented at Swedish Riksbank Workshop on Housing Markets, Monetary Policy and Financial Stability, November 12. Berman, G., Brooks, R., and Davidson, S. (2000). The Sydney Olympic Games announcement and Australia stock market reaction. Applied Economic Letters, 7, 781–784. Boehmer, E., Musumeci, J., and Poulsen, A. B. (1991). Event-study methodology under conditions of event-induced variance. Journal of Financial Economics, 31, 253–272. Bomfim, A. N. (2003). Pre-announcement effects, news effects, and volatility: Monetary policy and the stock market. Journal of Banking and Finance, 27, 133–151. Brown, S. J. and Warner, J. B. (1980). Measuring security price performance. Journal of Financial Economics, 8 (3), 3–31. Brown, S. J. and Warner, J. B. (1985). Using daily stock returns: The case of event studies. Journal of Financial Economics, 14 (1), 205–258. Carow, K. A. and Heron, R. A. (2002). Capital market reactions to the passage of the Financial Services Modernization Act of 1999. The Quarterly Review of Economics and Finance, 42, 465–485. Carow, K. A. and Kane, E. J. (2002). Event-study evidence of the value of relaxing long-standing regulatory restraints on banks, 1970–2000. The Quarterly Review of Economics and Finance, 42, 439–463. Chen, A. H. and Thomas, S. (2004). The effects of terrorism on global capital markets. European Journal of Political Economy, 20, 349–366. Cooper, M. J., Dimitrov, O., and Rau, P. R. (2001). A rose.com by any other name. Journal of Finance, 56 (6), 2371–2388. Dolley, J. C. (1933). Characteristics and procedure of common stock splitup. Harvard Business Review, April, 316–326. Giavazzi, F. and Mishkin, F. (2006). An evaluation of Swedish Monetary Policy 1995–2005. Reports from the Riksdag 2006/07:RFR 1, Committee on Finance. Henry, P. B. (2000). Stock market liberalization, economic reform, and emerging market equity prices. Journal of Finance, 55 (2), 529–564. Howe, J. S. and Jain, R. (2004). The REIT Modernization Act of 1999. Journal of Real Estate Finance and Economics, 28 (4), 369–388. Jackson, D. and Madura, J. (2007). Impact of regulation fair disclosure on the information flow associated with profit warnings. Journal of Economics and Finance, 31 (1), 59–74.

page 70

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch03

An Event Study Analysis of Israeli Mortgage Reforms

page 71

71

Jain, P. C. (1987). The effect on stock price of inclusion in or exclusion from the S&P 500. Financial Analysts Journal, 43 (1), 58–65. Johnson, Marilyn F., Kasznik, R., and Nelson, K. K. (2000). Shareholder wealth effects of the Private Securities Litigation Reform Act of 1995. Review of Accounting Studies, 5, 217–200. Li, K. and McNally, W. (2003). The decision to repurchase, announcement returns and insider holdings: A conditional event study. The ICFAI Journal of Applied Finance, 9 (6), 55–70. MacKinlay, A. C. (1997). Event studies in economics and finance. Journal of Economic Literature, 35 (1), 13–39. Mishkin, F. S. (2011). Monetary policy strategy: Lessons from the crisis. NBER Working Paper No. 176755, February. Miyajima, H. and Yafeh, Y. (2007). Japan’s banking crisis: An event-study perspective. Journal of Banking & Finance, 31, 2866–2885. Patell, J. M. (1976). Corporate forecasts of earnings per share and stock price behavior: Empirical Test. Journal of Accounting Research, 14 (2), 246–276. Schwert, W. G. (1981). Using financial data to measure effects of regulation. Journal of Law and Economics, 24 (1), 121–158. Taylor, J. B. (2010). Getting back on track: Macroeconomic policy lessons from the financial crisis. Federal Reserve Bank of St. Louis Review, 92 (3), 165–176. Tumarkin, R. and Whitelaw, R. F. (2001). News or noise? Internet postings and stock prices. Financial Analysts Journal, 57 (3), 41–51. Veraros, N., Kasimati, E., and Dawson, P. (2004). The 2004 Olympic Games announcement and its effect on the Athens and Milan stock exchanges. Applied Economics Letters, 11, 749–753.

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0004

Chapter 4

What You See Is What You Get But Do Investors Reward Good Corporate Governance When They See It?

Alberto Plazzi∗ , Walter Torous† and Umit Yilmaz‡

Abstract Poor corporate governance facilitates unreliable financial reporting. The AGR governance rating is based on the premise that a more accurate assessment of corporate governance can be formulated by taking this output of corporate governance into account. We document that time series variation in a firm’s AGR rating reliably forecasts measures of firm operating performance. A long/short strategy based on the AGR rating generates a risk-adjusted return of approximately 5% per year but, consistent with learning by the market, this abnormal performance has been declining over time. Most of this return differential originates with firms having poor corporate governance. Keywords: Corporate governance, operating performance, abnormal stock price performance

∗

Universit` a della Svizzera Italiana (USI-Lugano) and Swiss Finance Institute, Via Buffi 13, 6900 Lugano, Switzerland; [email protected]. † CRE and Sloan School of Management, MIT, Cambridge, MA 02138, USA; [email protected]. ‡ Universit` a della Svizzera Italiana (USI-Lugano) and Swiss Finance Institute, Via Buffi 13, 6900, Lugano, Switzerland; [email protected]. 73

page 73

November 4, 2020

74

1.

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

Behavioral Finance: A Novel Approach

Introduction

Academic research as well as numerous commercial endeavors attempt to quantify the effectiveness of a firm’s corporate governance. The resultant metrics allow users to investigate potential links between corporate governance and a firm’s subsequent operating performance and its stock price behavior. In a seminal contribution, Gompers et al. (2003) propose the so-called G score which enumerates the number of anti-takeover measures adopted by a firm and document that a portfolio long shares of firms with strong shareholder rights (five or fewer measures and labeled as “democracies”) while shorting shares of firms with weak shareholder rights (fourteen or more measures and labeled as “dictatorships”) generates abnormal returns of 8.5% per year over the sample period 1990 to 1999. Subsequently, Bebchuk et al. (2009) put forward the E or entrenchment index which relies on a subset of six of the twenty-four provisions considered in the G score that are most highly correlated with firm value and stockholder returns. They find that buying a portfolio of stocks of firms with non-entrenched management (a zero E index score) and selling short a portfolio of stocks of firms with entrenched management (a five or six E index score) earned abnormal returns of 7% annually over the 1990 to 2003 sample. Although both the G and E measures have been widely cited and used, unfortunately the effects of governance on a firm’s operating performance and stock returns remain unresolved. For example, Johnson et al. (2009) argue that the significant abnormal return spreads earned by strategies relying on these measures is a statistical artifact of ignoring industry clustering. That is, industry instead of governance is the source of variation in returns across these governance portfolios. Once statistical tests are properly adjusted for these industry effects, Johnson et al. (2009) find that the significance of the abnormal return spreads based on either the E or G measures is eliminated. Relatedly, Bebchuk et al. (2013) put forward a learning explanation for the reduced profitability of portfolios that load on these corporate governance measures and argue that while this profitability may be waning, the link between governance and firm performance has been stable over time. Furthermore, statistical inference when dealing with corporate governance measures based on takeover

page 74

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

What You See Is What You Get

b3964-ch04

page 75

75

provisions is complicated by the fact that these metrics show limited time variation. This then makes it difficult to establish a causality link between corporate governance and firm performance because the effects of corporate governance cannot be separated from those of other time-invariant firm characteristics, such as, for example, a firm’s culture. In addition to these academic studies, numerous commercial corporate governance measures have also been introduced. Firms such as Risk Metrics/ISS, Governance Metrics International, and The Corporate Library have provided institutional investors with the ratings of the quality of firm governance. Given their greater access to data and potentially more sophisticated models, it would be expected that these commercial measures would perform favorably against simple count scores like the G or E measures. To the contrary, Daines et al. (2010) find that the governance ratings of these commercial providers bear little empirical relation to either a firm’s subsequent operating performance, the likelihood of shareholder litigation, or the probability of financial restatements. However, they find somewhat stronger predictive evidence for MSCI’s AGR governance rating1 that uses information on financial statements in addition to observable corporate governance measures such as board structure. In contrast to other commercial corporate governance measures, Daines et al. (2010) also find that AGR has modest ability to forecast excess stock returns, at least as of December 31, 2005. As argued by Daines et al. (2010), AGR views a firm’s financials as an output of its governance. That is, poor corporate governance facilitates unreliable financial reporting by a firm’s management. Therefore, a more accurate assessment of the effects of corporate governance may be formulated by taking into account both corporate governance outputs as well as inputs. The purpose of this chapter is to investigate the links between the quality of a firm’s corporate governance as measured by its AGR score and its subsequent operating performance and stock price

1 The AGR methodology was developed by Audit Integrity. In August 2014, MSCI acquired Governance Holdings Inc. which was formed by the merger of Audit Integrity, Governance Metrics International, and The Corporate Library. MSCI is now responsible for AGR ratings.

November 4, 2020

76

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

Behavioral Finance: A Novel Approach

behavior. No systematic analysis of the AGR metric is available in the literature. To fill this void, we rely on an extensive database of AGR scores that ranks approximately 8,300 firms over the January 1997 to December 2011 sample period. This comprehensive panel data set affords us the opportunity to more carefully examine the links by which corporate governance impacts firm performance. The AGR metric differs importantly from the G and E measures because a firm’s AGR score exhibits non-trivial variation over time. For example, for the poorly governed firms that fall in the bottom 10% of the AGR distribution in a given month, only about 35%, on average, remain in this decile twelve months later. Similarly, about 40% of firms in the top 15% of AGR scores remain in this group after twelve months. We find little or no correlation between AGR scores and G and E measures. These findings confirm that AGR scores capture a dimension of governance that is distinct from that reflected by the number of anti-takeover provisions in place. In light of this evidence, we ask whether corporate governance as captured by AGR is associated with higher future firm operating performance. We find that higher AGR-rated firms are indeed characterized by better future operating performance as measured by the firm’s Return on Assets (ROA). In particular, a 10-point increase in AGR rating is accompanied by a 0.15% expected increase in that firm’s 2-year ahead industry-adjusted ROA. In the cross-section, the effect is more pronounced for firms in the lower AGR decile, i.e., poorly managed firms. We also find that over time, the relation between AGR and operating performance is actually stronger in more recent years. Importantly, these results hold even after controlling for firm fixed effects and time-varying firm characteristics. While our tests cannot completely address the endogeneity of a firm’s corporate governance structure, they increase the odds in favor of a causal link that runs from changes in a firm’s governance to future profitability. We also construct a portfolio that is long the stocks of better governed firms (i.e., Conservative, top 15% of AGR scores) and short the stocks of poorly governed firms (i.e., Very Aggressive, bottom 10% of AGR scores). We find that this portfolio delivers an abnormal return or alpha as large as 50 basis points per month when benchmarked against the 5-factor Fama and French (2014) model augmented by momentum as well as two accounting factors that load on the accrual and earnings surprise “anomalies”. This result

page 76

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

What You See Is What You Get

b3964-ch04

page 77

77

is robust to alternative portfolio formation strategies (value- versus equal-weighted, monthly versus annual rebalancing) and holds whether we use excess or industry-adjusted returns. Most of the profitability of this strategy originates with the portfolio of stocks having low AGR scores which delivers a negative and significant alpha of about 40 basis points. Interestingly, we find that the premium for governance has been declining almost monotonically over time. This pattern is consistent with the learning argument of Bebchuk et al. (2013) who find that the alpha associated with trading strategies based on the G and E measures has been waning in recent years. Overall, these findings suggest that firms with poor corporate governance as captured by a low AGR score tend to be subsequently characterized by abnormally low returns and poor operating performance. Since a low AGR score also reflects an increased likelihood of shareholder litigation and financial restatements, our paper also contributes to the recent literature on unethical corporate behavior. For example, Biggerstaff et al. (2015) document that CEOs who backdate their options are more likely to engage in corporate misbehavior and to induce an unethical corporate culture. This behavior eventually results in more value-destroying acquisitions, more extensive reliance on accounting manipulations, and lower stock returns. Corporate fraud and misbehavior may ultimately undermine investors’ trust in financial markets and have an overall detrimental effect on stock market participation (Giannetti and Want, 2016). We contribute to this discussion by showing that a firm’s AGR score aggregates valuable warning signals and reliably allows investors to identify firms at risk of corporate fraud.

2.

The AGR Metric

A firm’s AGR score measures the overall risk that the firm engages in fraudulent or misleading accounting and governance activities. Using publicly available information, MSCI’s objective is to discriminate between fraudulent and non-fraudulent firms. To do so, it ranks firms by their AGR scores and then groups them ranging from Very Aggressive (bottom 10%) to Conservative (top 15%), with the bulk of firms being classified as Aggressive (25%) or Average (50%).

November 4, 2020

78

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

Behavioral Finance: A Novel Approach

Being proprietary, the exact algorithm by which a firm’s AGR score is calculated is not publicly available. However, in general, the following five risk categories are considered: (i) corporate governance, (ii) high risk events, (iii) revenue recognition, (iv) expense recognition, and (v) asset-liability valuation. Within each of these risk categories, multiple issues (or “games”) are tabulated. There can be as many as twenty-five issues per category. For example, issues within the revenue recognition category include high operating revenues, large accounts receivables, large inventory, and small unearned revenues. Each issue, in turn, is measured by one or more metrics. For example, corporate governance metrics include the percentage of board directors who are officers, incentive compensation over total compensation for both the firm’s CEO and CFO, and the ratio of CFO to CEO total compensation. These metrics are the fundamental ingredients of an AGR score. In particular, firms that exhibit extreme values in these measures are hypothesized to be at higher risk of fraudulent accounting and governance activities. To that end, each metric is examined for unusual behavior according to (i) an industry comparison (number of interquartile ranges from the industry median), (ii) it’s one year change (percentage change from previous year), as well as (iii) it’s volatility (variance over the previous eight quarters). Only if a firm’s particular metric exhibits unusual behavior, defined to be in the corresponding extreme 20% of all observed values, is that metric included in a firm’s AGR score. A firm’s AGR score is then constructed as a weighted average of its extreme metrics. The weight assigned to a particular extreme metric value is determined by its importance in detecting fraudulent behavior. In particular, this weight is given by the estimated odds ratio associated with whether extreme values of the metric explain particular examples of fraudulent behavior. The scores are then transformed to fit a curve with the above predefined percentile cutoffs corresponding from Very Aggressive, with a minimum AGR score of 1, to Conservative firms, with a maximum AGR score of 100.

3.

Data and Descriptive Statistics

We rely on a comprehensive database of AGR scores that ranks approximately 8,300 firms during the January 1997 to December

page 78

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

What You See Is What You Get

b3964-ch04

page 79

79

2011 sample period. AGR scores are generally updated after the public release of new quarterly or yearly financial statements. Because of this, changes in AGR scores may occur at any point in calendar time. In our analyses, we rely on monthly observations and only update AGR scores at the end of the month following a score change. We apply this lag to ensure that our regression results are not subject to any potential look-ahead biases. Once a firm’s AGR score is updated, we retain this score until it is updated once again. We match our AGR dataset to the Center for Research in Security Prices (CRSP) dataset using a firm’s CUSIP number. Balance sheet and other fundamental data are collected from COMPUSTAT. The original sample consists of 611,838 firm-month observations. Following the literature, we apply a series of filters to these data. First, we retain only stocks with a CRSP share code equal to 10 or 11, thereby eliminating companies incorporated outside the US, trusts, closed-end funds, and REITs. Next, we remove dual-class shares owing to their peculiar governance structure (see Gompers et al., 2010). Finally, we remove stocks with a price lower than $1 (“penny stocks”), and drop observations with a monthly return greater than 300% (16 observations) to avoid exceptionally high returns that may exert undue influence on our results. These filters leave us with a reference sample of 529,833 firm-month observations on 7,189 firms. Johnson et al. (2009) point out that particular care must be paid to control for industry composition when investigating corporate governance. In what follows, based on the findings of Johnson et al. (2009), we use the finer 3-digit SIC code to take into account industry clustering. Table 1 examines the characteristics of the firms in our sample. In particular, the first set of rows report summary statistics for four measures of firm value and operating performance that have been related to corporate governance metrics in prior research, e.g., Daines et al. (2010) and Bebchuk et al. (2013). These include return on assets (ROA), Tobin’s Q, net margin, and 3-year sales growth. All measures are industry-adjusted by subtracting the median value of the corresponding measure for all firms with non-missing COMPUSTAT data in the same industry in that fiscal year. The subsequent set of rows displays analogous statistics for the control variables that we use and that have also been relied upon in the prior literature: the market value of equity (Market Value); total assets (Assets); the ratio of capital expenditures to total assets (CAPEX/Assets); the

Summary statistics of characteristics of AGR-rated firms

Mean Conservative

Mean Very Aggressive

Difference

Median

ROA, Ind. Adj. Tobin’s Q, Ind. Adj. Net margin, Ind. Adj. 3-Year Sales growth

48,139 42,497 48,001 39,285

0.005 0.448 −0.154 0.245

0.004 0.052 0.006 0.003

−0.002 −0.040 0.074 −0.087

−0.002 0.345 −0.088 0.135

−0.007 0.477 −0.339 0.438

0.008 −0.107∗ 0.257∗∗∗ −0.303∗∗∗

Market Value (in millions of $) Log (Market Value) Assets (in millions of $) Log (Assets) CAPEX/Assets Leverage R&D/Sales

48,320 48,320 48,352 48,352 45,600 46,244 24,384

3,601 6.233 6,984 6.506 0.050 0.188 4.923

423.135 6.048 615.498 6.422 0.031 0.129 0.049

−0.126 −0.172 −0.099 −0.135 0.014 −0.055 −0.007

1,181 5.714 2,398 6.155 0.049 0.163 2.590

8,877 6.626 26,397 6.932 0.046 0.206 16.247

−7.688∗∗∗ −0.892∗∗∗ −24.053∗∗∗ −0.769∗∗∗ 0.003∗ −0.044∗∗∗ −13.642

G Score E Index

7,002 7,002

9.192 2.416

9.000 2.000

−0.002 0.015

9.142 2.510

9.067 2.416

0.075 0.095

b3964-ch04

Note: We provide the following summary statistics for characteristics of AGR-rated firms measured at year end: the number of firm-year observations for a characteristic; mean value of a characteristic; median value of a characteristic; the Pearson correlation of a characteristic with the firm’s last AGR score available that year; the mean value of a characteristic for Conservative firms whose AGR score ranked in the top 15% of AGR scores that year; the mean value of a characteristic for Very Aggressive firms whose AGR scores ranked in the bottom 10% of AGR scores that year; the difference in mean values between characteristics of Conservative and Very Aggressive firms. See the Appendix for the definition of the variables. For the G score and E Index, the summary statistics are based on values observed in the years 1998, 2000, 2002, 2004, and 2006. For the other variables, the sample period is 1997–2011.

Behavioral Finance: A Novel Approach – 9in x 6in

Mean

Behavioral Finance: A Novel Approach

Obs.

10:11

Corr. with AGR

November 4, 2020

80

Table 1:

page 80

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

What You See Is What You Get

b3964-ch04

page 81

81

debt-to-assets ratio (Leverage); and the ratio of R&D expenses to sales (R&D/Sales). The construction of these variables is detailed in the Appendix. For each year, we record all variables as of the fiscal year ending on or before December and match them to the firm’s AGR score as of December of that year. Thus, a firm whose COMUSTAT last fiscal year entry for 2005 is recorded on May 2006 would be matched with its AGR score as of December 2005. We provide statistics for both levels and logs of equity and total assets as the log values will be used in our regressions to account for the high skewness of these variables. At the median, the industry-adjusted characteristics of the firms in our sample are close to zero. This implies that AGR-rated firms are fairly representative of the universe of CRSP companies across each industry. The contemporaneous correlations of the firm performance measures with AGR, reported in the fourth column, indicate that highly rated firms tend to have, on average, lower Tobin’s Q, higher net margins, and lower sales growth in the past three years. The correlations are, however, quite modest and never exceed 0.10 in absolute value. Turning to the controls, highly rated AGR firms appear to be significantly smaller in size, whether measured by equity or total asset value, and to be less levered.2 In our subsequent analyses, we report results using all AGR-rated firms as well as restricting attention to only those firms in the two extreme groupings, ‘Conservative’ and ‘Very Aggressive’. For this reason, the last three columns of Table 1 report the average values of the performance measures and controls for these two groups and their differences. To properly account for potential time-series and cross-sectional correlations, statistical significance for the differences are based on doubly-clustered standard errors at both the year and firm level. The signs of the differences are consistent with the correlations reported across all firms. In particular, ‘Conservative’ firms tend to be characterized by higher net margins, lower sales growth in the prior 3 years, smaller size, and lower leverage. There is also some evidence, albeit economically more modest, that ‘Conservative’ firms have lower Tobin’s Q. We note that the large difference in size

2

These results are clearly not independent, as leverage is on average positively correlated with size.

November 4, 2020

82

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

Behavioral Finance: A Novel Approach

is partly due to the impact of outliers. Median differences in Market Value and Assets appear less dramatic at $306 million and $363 million, respectively. Nevertheless, taken together, the evidence indicates that the two sets of firms appear ex-ante quite dissimilar. This suggests that including the controls in our analysis is key to ensuring that the AGR classification is not merely reflecting observables such as firm size. The last two rows of Table 1 relate the AGR scores to the G and E measures. We focus on the data from the 1998, 2000, 2002, 2004, and 2006 Investor Responsibility Research Center (IRRC) publications that overlap with our sample period, and contrast the G and E measures with the last AGR score available for a firm in each of these years. The combined dataset consists of about 7,000 observations on about 2,400 firms. As a preliminary, we note that the mean (median) market capitalization of these firms is much larger, at about $7.5 billion ($1.5 billion), when compared to the corresponding results presented in Table 1. These values are comparable to those reported in Core et al. (2006) and highlight that corporate governance measures based on anti-takeover provisions tend to be available for larger firms. Turning to their correlations with corresponding AGR scores, they are nearly zero for the G score and slightly positive for the E index. This implies that the type of corporate governance information contained in AGR scores is distinct from that conveyed by indices based on anti-takeover provisions. If anything, firms in the ‘Conservative’ segment of AGR scores feature somewhat higher G and E measures as compared to the ‘Very Aggressive’, but the differences are not significant.

4.

The AGR Metric and Operating Performance

In this section, we correlate AGR scores to measures of firm value and operating performance. Our goal is to investigate whether corporate governance, as captured by a firm’s AGR score, is a reliable predictor of future operating performance. A distinct feature of our estimation approach is that, given the documented time-variation in the AGR score of a given firm, we are able to include firm fixed effects in our regression framework. This is in contrast with much of the previous research that relies on corporate governance indices that exhibit little or no time variation.

page 82

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

What You See Is What You Get

4.1.

b3964-ch04

page 83

83

Return on Assets (ROA)

We begin by analyzing the relation between AGR scores and ROA. In the corporate governance literature, this measure has been used by, among others, Core et al. (2006) and Daines et al. (2010). Under the hypothesis that good governance results in more value-enhancing decisions, we expect a positive relation between AGR scores and operating performance. In Panel A of Table 2, we estimate pooled regressions of contemporaneous and future ROA on AGR scores. To capture the direct and indirect effects of governance, we investigate this relation both with and without including cross-sectional controls. As discussed above, ROA is industry-adjusted by its median value in the same three-digit SIC code industry. Future ROA is computed as the firm’s ROA in fiscal year t + 2, thereby avoiding any potential overlap with the timing of our dependent variables.3 All regressions include both year and industry fixed effects and ROA is expressed in percentage terms. Column (1) of Table 2 shows that, consistent with Table 1, the contemporaneous relation between AGR and ROA is slightly negative, but not statistically significant. After controlling for firm characteristics (column (2)), however, we see that AGR scores appear to be positively and significantly related to ROA. In the next two columns, we include firm fixed effects. This amounts to asking whether variation in AGR scores for the same firm correlates with variation in its ROA. The loading on AGR is now much smaller at 0.012, and is significant only at the 10% level. In sum, there is some evidence that firms with higher AGR scores display better current performance than otherwise comparable firms.4

3

This implies that since a great majority of firms report financials in December of each year, we use the last available AGR score in, say, 2003 to predict ROA computed as of December 2005. 4 The reduction in the number of observations (Obs.) when adding controls other than AGR is mainly due to R&D/Sales often missing. When excluding R&D/Sales, the number of observations in specification (8) increases to 30,200, and the loading on AGR is smaller at 0.008 but again statistically significant (t-ratio of 2.03). Given the significance of R&D/Sales, we decided against excluding it from the set of regressors.

Dep. Var.

Panel A: AGR score AGR score

ROA(t + 2) (4)

(5)

(6)

(7)

(8)

−0.002 (0.004)

0.056∗∗∗ (0.008) 3.729∗∗∗ (0.818) 1.633∗ (0.856) 7.224 (6.175) −10.469∗∗∗ (1.982) −0.004∗ (0.002) −3.344∗∗ (1.314) −4.319∗∗∗ (0.685)

−0.003 (0.003)

0.012∗ (0.006) 1.743∗∗ (0.789) 3.641∗∗∗ (1.184) 1.926 (4.750) −6.165∗∗ (2.403) −0.001 (0.001) 4.081∗∗∗ (1.268)

0.011∗∗∗ (0.003)

0.010∗∗∗ (0.003)

0.015∗∗ (0.006) −2.264∗∗∗ (0.680) 2.275∗∗ (1.084) 3.425 (4.846) 0.139 (2.330) −0.001∗∗ (0.000) 5.466∗∗∗ (1.046)

0.624∗∗∗ (0.012)

0.030∗∗∗ (0.005) −0.880 (0.562) 2.530∗∗∗ (0.576) 6.685∗ (3.753) 0.910 (1.471) −0.003∗∗∗ (0.001) 1.672∗∗ (0.831) −1.188∗∗∗ (0.359) 0.593∗∗∗ (0.015)

Yes Yes No 22,094 0.191

Yes Yes Yes 47,896 0.695

Yes Yes No 36,116 0.004

Yes Yes No 16,640 0.520

Yes Yes Yes 36,116 0.729

Yes Yes Yes 16,640 0.752

Log (Market Value) Log (Assets) CAPEX/Assets Leverage R&D/Sales Log (Tobin’s Q) Delaware ROA, Ind. Adj. Time Fixed Effects Industry Fixed Effects Firm Fixed Effects Obs. R2

Yes Yes No 47,896 0.002

Yes Yes Yes 22,094 0.725

b3964-ch04

(3)

Behavioral Finance: A Novel Approach – 9in x 6in

(2)

Behavioral Finance: A Novel Approach

(1)

10:11

Variable (t)

ROA(t)

November 4, 2020

AGR score and operating performance: ROA

84

Table 2:

page 84

0.522∗ 1.201∗∗ : (0.303) (0.536) 0.516∗ 1.765∗∗∗ (0.309) (0.556) 0.074 1.379∗∗ (0.350) (0.654) No Yes Yes Yes Yes Yes Yes Yes 48,139 22,214 0.665 0.698

0.399 (0.287) 0.772∗∗∗ (0.275) 0.987∗∗∗ (0.312) No Yes Yes No 36,551 0.465

0.241 (0.509) 1.247∗∗ (0.493) 2.376∗∗∗ (0.563) Yes Yes Yes No 16,942 0.494

0.580∗∗ (0.288) 0.769∗∗∗ (0.296) 1.027∗∗∗ (0.345) No Yes Yes Yes 36,661 0.706

0.461 (0.519) 0.729 (0.536) 1.276∗∗ (0.622) Yes Yes Yes Yes 16,996 0.728

b3964-ch04

Note: We tabulate OLS estimates of the pooled annual regression of return on assets (ROA) on the AGR score and additional cross-sectional controls. Panel A uses AGR scores and Panel B uses AGR groups. AGR(t) is the last rating available in year t. In specifications (1)–(4), the dependent variable is the contemporaneous ROA, ROA(t). In specifications (5)–(8), the dependent variable is the ROA in fiscal year t + 2, ROA(t + 2). ROA is computed as ratio of Operating Income After Depreciation in the current fiscal year to Assets at the end of the prior fiscal year. For a given firm-year, ROA is then adjusted by subtracting the median ROA in the industry, as defined by its three-digit SIC code. Delaware is a dummy that equals one for firms incorporated in Delaware. t-statistics based on heteroskedasticity-robust standard errors clustered at the firm level are reported in parentheses. All regressions include a constant term, whose coefficient is omitted. Sample period is January 1997 to December 2011.

What You See Is What You Get

Controls Time Fixed Effects Industry Fixed Effects Firm Fixed Effects Obs. R2

3.016∗∗∗ (0.661) 5.198∗∗∗ (0.685) 5.593∗∗∗ (0.805) Yes Yes Yes No 22,214 0.185

Behavioral Finance: A Novel Approach – 9in x 6in

Conservative

1.398∗∗∗ (0.413) 1.592∗∗∗ (0.423) 0.519 (0.467) No Yes Yes No 48,139 0.043

10:11

Average

November 4, 2020

Panel B: AGR groups Aggressive

85 page 85

November 4, 2020

86

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

Behavioral Finance: A Novel Approach

In columns (5)–(8) of Table 2, the dependent variable is now future ROA. Here the results indicate that AGR scores represent a reliable predictor of future firm profitability. The loadings are 0.011 (no controls) and 0.030 (with controls) when excluding firm fixed effects, both significant at the 1% level. The significance and magnitude of this relation are preserved when including firm fixed effects, implying that time-series variation in AGR scores for the same firm is indeed capturing future operating performance. To put these numbers in perspective, the 0.015 estimate in column (4) implies that a one-standard deviation increase in AGR is associated with an expected increase in the firm’s future ROA of about 0.42%. In the bottom Panel of Table 2, we present estimates in analogous regressions where now the AGR score is replaced by dummies for firms in the “Aggressive”, “Average”, and “Conservative” groups of AGR scores. For contemporaneous ROA, we see that the relation with these AGR groups is not clear. For example, it is humped in specifications (1) and (4), increasing in specification (2), and decreasing in specification (3). This is in contrast with columns (5)–(8), where the loadings on AGR groups are monotonically increasing. From the estimates in column (6) that include time and industry fixed effects, we see that better firm governance is accompanied by improved ROA in the three groups of 0.241%, 1.247%, and 2.376%, respectively. A similar pattern is observed when including firm fixed effects, although the estimates are generally lower and are significant only for the ‘Conservative’ group.

5.

Stock Returns and AGR Scores

We next investigate whether corporate governance, as measured by AGR, is priced in stock returns. Initial but limited work by Daines et al. (2010) suggests that a significant spread can be earned by going long well governed (high AGR) firms and shorting poorly governed ones (low AGR). The panel nature of our dataset allows us to investigate the returns obtained when loading on AGR over an extended period of time. In particular, as firms’ AGR scores change over time, our analysis will be able to isolate the extent to which stock returns’ react to changes in governance as opposed to

page 86

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

What You See Is What You Get

page 87

87

company-specific attributes that otherwise cannot be controlled for with a single snapshot view of governance. 5.1.

Portfolio Performance Regressions

We first analyze the performance of AGR-sorted portfolios. The portfolio formation is in the spirit of Fama and French (1993) and Hirshlefier et al. (2012). As mentioned previously, for this analysis we restrict our attention to the post-2004 period to ensure that our results represent that of an implementable trading strategy. Specifically, at the end of each month starting in January 2005, we group firms into thirds based on their end-of-month market capitalization (from Small to Large) and, separately, by the four AGR groups (from Very Aggressive to Conservative). The intersection of these size and AGR groups yields twelve portfolios, ranging from S&VA (Small & Very Aggressive) to L&C (Large & Conservative). We compute each corresponding portfolio return in the subsequent month as, alternatively, the value-weighted (VW) or equal-weighted (EW) average returns to stocks within the portfolio, where the weights in the former case equal the relative market capitalization of a firm’s stock as of the formation date. The portfolios are subsequently rebalanced every month. We construct returns to a given AGR group as the simple average across portfolios with different sizes.5 Similarly to Gompers et al. (2003) and Bebchuk et al. (2009), we also investigate the performance of a portfolio that is long better governed, high AGR stocks (Conservative) and short stocks in the bottom AGR group (Very Aggressive), AGRp = (S&C + M&C + L&C)/3 − (S & VA + M & VA + L & VA)/3. To assess whether AGR-based portfolios produce average returns that cannot be attributed to exposure to well-known risk factors, we rely on the following performance attribution model: rp,t = α + β1 RMRFt + β2 SMBt + β3 HMLt + β4 UMDt + β5 RMW + β6 CMA + β7 ACCt + β8 SUEt + p,t 5

So, for example, the Very (S&VA+M&VA+L&VA)/3.

Aggressive

portfolio

is

constructed

(1)

as

November 4, 2020

88

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

Behavioral Finance: A Novel Approach

where rp,t is the return to a given AGR-sorted portfolio, p = {V ery Agg, Agg, Avg, Cons}, in excess of the 3-month T-Bill rate. The first four regressors are the standard Fama and French (1993) factors measuring zero-investment returns for exposure to market risk (RMRF), size (SMB), and book-to-market ratio (HML), plus the momentum portfolio UMD as constructed by Fama and French (1996). This is the benchmark model used in prior studies that investigate the performance of governance-sorted portfolios. Given the nature of AGR, however, we augment this list with four additional factors capturing trading strategies (risk factors) that are either related to operating performance or based on accounting information. The first two additional factors, RMW and CMA, have been recently proposed by Fama and French (2014) based on the evidence that profitability and investment have a significant role in explaining the cross-section of expected returns. RWA is constructed as the difference between the returns on diversified portfolios of stocks with robust and weak profitability. CMA is defined as the difference between the returns on diversified portfolios of the stocks of low and high investment firms.The final two factors are the accrual factor (ACC) of Hirshlefier et al. (2012), constructed as the return difference to portfolios of stocks with low versus high accruals, and a portfolio based on standardized earning surprises (SUE). By including these two additional factors, we attempt to differentiate the corporate governance information contained in AGR from exposure to previously documented accounting-based “anomalies”.6 Panel A of Table 3 presents the OLS estimates of the time-series regression, Eq. (1), for portfolios ranging from Very Aggressive to Conservative as well as for the AGRp portfolio that is long Conservative firms and short Very Aggressive firms. The left-hand side of the Table is based on value-weighted returns, while the right-hand side is based on equal-weighted returns. As can be seen, the unexplained average monthly return or Alpha for the value-weighted Very Aggressive group is −0.451% and is significant at the 1% level. For the equal-weighted Very Aggressive portfolio, its Alpha is slightly

6

The ACC and SUE factors are characterized by the highest monthly Sharpe Ratio of 0.17 and 0.10, respectively, during our sample period. This too underlines the importance of including them in the analysis.

page 88

Performance analysis of AGR sorted portfolios VW

Very Agg

Agg

Avg

EW Cons

Cons–Very Agg

Very Agg

Agg

Avg

Cons

Cons–Very Agg

0.988

0.991

0.987

0.49

0.991

0.987

0.515

−0.023 0.144 0.177 (−0.32) (2.23) (2.68) −0.077 −0.093 −0.096 (−2.91) (−4.11) (−4.68) 0.529 0.492 0.430 (16.47) (16.89) (11.14) 0.131 0.139 0.100 (4.27) (4.28) (2.39)

0.580 (3.73) −0.079 (−1.61) −0.009 (−0.09) −0.070 (−0.99)

b3964-ch04

−0.403 (−2.32) −0.017 (−0.38) 0.438 (4.97) 0.170 (2.53)

0.986

89

0.562 (3.27) −0.083 (−1.38) 0.044 (0.50) −0.066 (−0.81)

0.964

Behavioral Finance: A Novel Approach – 9in x 6in

0.963

10:11

R

Panel B: Industry-adjusted returns 0.132 Alpha −0.431 −0.043 0.126 (−2.46) (−0.62) (2.19) (1.87) RMRF 0.017 −0.045 −0.074 −0.066 (0.34) (−2.50) (−3.31) (−3.54) SMB 0.371 0.458 0.449 0.415 (4.30) (20.49) (18.44) (13.53) 0.193 0.147 0.148 0.127 HML (2.70) (6.29) (4.04) (3.45)

What You See Is What You Get

Panel A: Excess returns Alpha −0.451 0.004 0.155 0.090 0.541 −0.337 0.013 0.141 0.179 0.516 (−2.74) (0.05) (2.60) (1.26) (3.03) (−2.07) (0.16) (2.23) (2.51) (3.25) RMRF 1.043 1.005 0.960 0.940 −0.102 1.022 0.976 0.952 0.911 −0.111 (22.26) (44.91) (50.04) (38.51) (−1.55) (24.25) (30.53) (55.02) (48.25) (−2.10) 0.627 0.667 0.624 0.117 0.626 0.766 0.750 0.657 0.031 SMB 0.508 (6.87) (18.86) (21.63) (14.67) (1.29) (8.60) (16.56) (22.26) (14.11) (0.34) HML 0.230 0.130 0.146 0.173 −0.057 0.171 0.145 0.154 0.145 −0.026 (3.56) (4.34) (3.69) (3.76) (−0.69) (2.79) (3.59) (4.35) (2.69) (−0.37) 0 .025 .219 0 −0.203 −0.158 −0.107 −0.027 0.176 UMD −0.195 −0.112 −0.083 (−4.62) (−3.95) (−4.15) (1.07) (5.63) (−4.60) (−5.50) (−5.14) (−0.93) (4.39) RMW −0.333 −0.055 −0.104 −0.052 0.281 −0.411 −0.174 −0.131 −0.116 0.294 −0.97) (−2.04) (−0.84) (2.24) (−3.76) (−2.53) (−2.50) (−1.75) (2.37) (−3.00) ( CMA −0.066 −0.121 −0.128 −0.062 .004 0 −0.090 −0.162 −0.154 −0.036 0.054 (0.54) (−0.51) (−2.15) (−2.24) (−1.04) (0.03) (−0.81) (−2.41) (−2.77) (−0.48) ACC −0.064 −0.045 0.015 0.037 0.101 0.088 0.047 0.073 0.016 −0.072 (−0.50) (−0.63) (0.30) (0.77) (0.74) (0.64) (0.56) (1.52) (0.31) (−0.47) SUE 0.060 −0.024 −0.020 −0.065 −0.125 0.057 −0.023 −0.028 −0.040 −0.098 −0.67) (−1.71) (−1.89) (0.93) (−0.92) (−1.09) (−0.95) (−1.22) (1.09) (−0.92) ( 2

November 4, 2020

Table 3:

(Continued) page 89

VW

UMD RMW

SUE 2

R

−0.032 (−1.81) −0.043 (−0.94) −0.043 (−0.81) −0.064 (−1.04) −0.008 (−0.39)

0.588

0.798

Cons

Cons–Very Agg

Very Agg

Agg

Avg

Cons

Cons–Very Agg

−0.024 0.046 0.170 −0.127 −0.053 −0.035 0.011 0.138 (−1.54) (2.61) (4.63) (−3.33) (−2.81) (−2.40) (0.47) (3.62) −0.045 −0.030 0.266 −0.343 −0.129 −0.092 −0.064 0.279 (−1.01) (−0.57) (2.31) (−3.20) (−2.63) (−1.99) (−1.24) (2.42) −0.018 0.028 0.045 −0.039 −0.058 −0.043 0.052 0.091 (0.98) (−0.32) (0.60) (0.40) (−0.33) (−1.01) (−0.75) (0.77) −0.013 0.049 −0.042 0.168 0.035 0.080 0.042 −0.126 (−0.24) (1.05) (−0.32) (1.15) (0.45) (1.38) (0.73) (−0.92) −0.016 −0.007 −0.107 0.098 −0.008 −0.011 0.033 −0.064 ( −0.56) (−0.22) (−1.67) (1.70) (−0.39) (−0.48) (0.96) (−0.87) 0.817

0.761

0.444

0.607

0.799

0.827

0.708

0.476

Note: At the beginning of each month, stocks are grouped into Very Aggressive, Aggressive, Average, and Conservative groups based on their AGR score reported at the end of the prior month. We compute value-weighted (VW) and equalweighted (EW) returns to these portfolios as well as the portfolio which goes long shares of Conservative firms and short share of Very Aggressive firms. The groups are then rebalanced every month. We report estimates of the following regression: rp,t = α + β1 RMRFt + β2 SMBt + β3 HMLt + β4 UMDt + β5 RMW + β6 CMA + β7 ACCt + β8 SUEt + p,t

b3964-ch04

The risk factors are the following zero-investment portfolios: the three Fama and French (1993) factors capturing exposure to the market (RMRF), size (SMB), and book-to-market (HML); the momentum factor UMD of Fama and French (1996); RWA is the difference between the returns on diversified portfolios of stocks with robust and weak profitability; CMA is the difference between the returns on diversified portfolios of the stocks of low and high investment firms; CMA is the accrual factor of Hirshleifer, Hou, and Teoh (2012); SUE is a portfolio based on standardized earning surprises. t-statistics based on Newey and West (1987) standard errors with 3 lags are reported in parentheses. Alpha is the intercept estimate and measures the monthly abnormal return, in percentage terms. Panel A uses returns in excess of the 3-month T-bill rate, while Panel B uses industry-adjusted returns. The sample period is January 2005 to December 2011.

Behavioral Finance: A Novel Approach – 9in x 6in

ACC

−0.124 (−3.23) −0.296 (−2.66) −0.017 (−0.13) 0.091 (0.67) 0.100 (1.77)

Avg

Behavioral Finance: A Novel Approach

CMA

Agg

10:11

Very Agg

EW

November 4, 2020

(Continued)

90

Table 3:

page 90

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

What You See Is What You Get

b3964-ch04

page 91

91

lower at −0.337% but still significant at the 5% level. Thus, firms in the low AGR group deliver significant average risk-adjusted returns. High AGR firms also tend to outperform the benchmark model. However, the corresponding Alpha is quite small in absolute value, ranging from 0.090% (Conservative in the value-weighted case with a t-stat of 1.26) to 0.179% (Conservative in the equal-weighted case with a t-stat of 2.51). Taken together, the spread between Conservative and Very Aggressive firms for both value-weighted and equalweighted portfolios is approximately 50 bps per month and is highly significant. This spread originates primarily from the group of Very Aggressive firms suggesting that the AGR score is particularly able to identify poorly managed firms. It is also noteworthy that this spread is greater when portfolio returns are computed in value-weighted rather than equal-weighted terms, and so is not the result of loading on smaller, and potentially more difficult to short shares. Looking across risk exposures, we see that Very Aggressive firms load significantly and positively on the market, size, and book-tomarket ratio risk factors. Interestingly, the Very Aggressive portfolio loads negatively on the momentum and RMW factors, confirming the view that these firms have recently experienced negative returns and suffered weak profitability. Across all portfolios and specifications, we find very limited exposures to the two accounting factors, the sole exception being the value-weighted Conservative portfolio with a 0.065 loading on the earnings surprise factor, albeit only marginally significant with a t-statistic of −1.71. In addition, Conservative firms appear to load only on the market and book-to-market ratio factors. The returns of the AGRp portfolio that goes long Conservative firms while shorting Very Aggressive firms are only weakly related to the Fama and French (1993) factors but load positively and significantly on momentum and negatively but not significantly on the SUE factor. In Panel B of the Table, we repeat this empirical analysis but now using industry-adjusted returns. Industry-adjusted returns are obtained by subtracting from each stock return its corresponding value-weighted three-digit SIC code industry return. Accounting for this industry effect now increases the AGR spread to 68 bps (valueweighted) and 71 bps (equal-weighted). As before, the large bulk of this spread is attributable to low AGR firms underperforming with respect to their peers. Interestingly, industry-adjusted Alphas

November 4, 2020

92

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

page 92

Behavioral Finance: A Novel Approach

are now monotonically increasing across AGR groups. Overall, the significance of the risk factor exposures is similar to that documented in Panel A. In sum, the strategy of going long firms with high AGR scores while shorting firms with low AGR scores delivers a positive return even after adjusting for an extensive set of risk factors. These results are consistent with AGR capturing an important dimension of the effectiveness of corporate governance that is not fully reflected in contemporaneous stock valuations. 5.2.

Fama and MacBeth Regressions

As an alternative to AGR-sorted portfolios, we further evaluate the robustness of our findings by estimating monthly firm-level regressions using the approach of Fama and MacBeth (1973). This crosssectional framework has the benefit of allowing the inclusion of a relatively large number of firm characteristics that is impractical to do in the time-series portfolio approach. We find that AGR’s economic and statistical significance persists even when accounting for these additional dimensions of risk. Specifically, each month t from January, 2005 to December, 2011 we estimate the following cross-sectional model:  Xi,t + i,t+1 ri,t+1 = γ0,t + γ1,t

(2)

where ri,t+1 is the return to stock i at the end of month t + 1, and Xi,t is a collection of firm-specific control variables that are observed   at the end of month t. The average coefficient, γ 1 = 1/T t γ1,t , measures the expected return (risk premium) to a zero-cost portfolio that loads on a given characteristic. Our interest is in the premium for AGR. Since AGR scores are hypothesized to be increasing in the effectiveness of governance, we expect a positive estimate of this premium reflecting positive returns to better governed firms. The comprehensive list of the conditioning variables in Xi,t follows from Hirshlefier et al. (2012), and is based on evidence in the prior asset pricing and accounting literatures. These controls include the market beta (β) estimated on the prior 60-month period; log market capitalization; log book-to-market ratio; the stock return in month t (Ret(t)) and the cumulative return in months t − 12 through t − 1 (Ret(t-12:t − 1)); idiosyncratic volatility (iv ), as measured by the

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

What You See Is What You Get

page 93

93

square root of average squared residuals from a 3-factor Fama and French (1993) model estimated using daily returns in month t, following Ang et al. (2006); the value of the accrual Accrual, computed as in Hirshlefier et al. (2012); and the most recent standardized earning surprise, SUE. In order to maintain comparability of the results across specifications, we restrict the sample to firms with at least 60 months of available return data. Table 4 reports the average coefficients for the regression (2), along with their corresponding time-series t-statistics. Five different specifications of Xi,t are explored. In the first column, the AGR score enters alone as a determinant. The corresponding coefficient is positive at 0.006, and is strongly significant with a t-statistic of 2.93. In the second column, we add a first set of control variables. The coefficient on AGR is now slightly lower at 0.005, but with a larger t-statistic of 4.11. For the other regressors, we note that the weak relation between expected returns and market betas and bookto-market is consistent with Boyer et al. (2010). In specification (3), we see that the coefficient on AGR remains stable at 0.005 when including accrual and earning surprises. Finally, we run a kitchen-sink regression in which we include all controls first excluding (column (4)), and then including (column (5)), the AGR score. In this specification, the loading on AGR remains positive at 0.005 and is significant at the 1% level with a t-statistic of 3.86. Among other factors, momentum and accrual stand out as the most robust predictors. The statistical significance of AGR goes hand in hand with its economic significance. The 0.005% monthly premium in Table 4 implies that an average difference of 83 points between the Very Aggressive and Conservative groups translates into a monthly average return differential of 0.415%, or about 5% annually. 5.3.

Time-Series Variation in AGR Premium

From an asset pricing perspective, the fact that well-governed firms persistently deliver higher risk-adjusted returns than poorly governed firms is puzzling. If differences between firms’ future governance, and hence performance, are already incorporated in current valuations, it should not be possible to generate abnormal profits by trading on governance metrics.

November 4, 2020

10:11

94

Behavioral Finance: A Novel Approach – 9in x 6in

Control AGR β log (MV) log (Book-to-Market)

Ret(t − 12:t − 1) iv Accrual SUE

page 94

Behavioral Finance: A Novel Approach Table 4:

Ret(t)

b3964-ch04

Fama-MacBeth regressions (1)

0.006 (2.93)

(2) 0.005 (4.11) 0.179 (0.75) −0.083 (−1.43) −0.002 (−0.03) −2.871 (−4.02) 0.032 (0.12) −5.373 (−0.91)

(3)

(4)

0.005 (2.79)

−1.952 (−3.64) 0.019 (1.15)

0.163 (0.69) −0.100 (−1.76) 0.013 (0.22) −2.896 (−4.06) 0.028 (0.11) −6.224 (−1.05) −1.851 (−3.98) 0.025 (2.26)

(5) 0.005 (3.86) 0.180 (0.76) −0.088 (−1.52) 0.006 (0.11) −2.905 (−4.07) 0.018 (0.07) −5.736 (−0.98) −1.763 (−3.79) 0 .023 (2.09)

Note: We provide the results of the OLS cross-sectional regression of stock returns:  Xi,t + i,t+1 ri,t+1 = γ0,t + γ1,t on various combinations of the regressors in X. AGR is the AGR score at the end of month t. MV and Book-to-Market denote, respectively, the log of stock market capitalization and book-to-market ratio at the end of the prior fiscal year. Ret(t) is the return in month t, while Ret(t − 12:t − 1) denotes the cumulative return in months t − 12 through t − 1. β is the slope coefficient in the regression of excess stock returns on a constant and RMRF estimated on the 60-month period ending in month t. Idiosyncratic volatility iv is measured by the square root of average squared residuals from a 3-factor Fama and French (1993) model estimated using daily returns in month t as in Ang et al. (2006). Accrual is the most recent accrual. SUE is the most recent earnings surprise. The regressions are estimated separately each month from January 2005 to December 2011. We tabulate average coefficients with the corresponding t-statistic based on Newey and West (1987) standard errors with 3 lags in parentheses. Returns are expressed in percentage.

A possible explanation for the documented profitability of our AGR trading strategy is that it reflects a slow adjustment towards equilibrium expected returns. As argued by Bebchuk et al. (2013),

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

What You See Is What You Get

page 95

95

if this is the case, we should detect a decline in abnormal returns as market participants begin to more aggressively trade on AGR even as we have observed that the relation between AGR and firm operating performance has actually become stronger with time. To investigate this possibility, we explore the time variation in the AGR premium. Specifically, Figure 1 displays the 24-month trailing average of the slope of the AGR metric from the full specification of the Fama– MacBeth regression (column 5 of Table 4). The plot reveals a distinct downward trend in the premium. From a peak of approximately 80 basis points (monthly) during the 2005 to 2006 time period, the premium declines almost monotonically and actually turns negative in 2009 before returning to zero by the end of the sample. Thus it appears that the declining profitability of corporate governancebased trading strategies, first evidenced by Bebchuk et al. (2013), also extends to the AGR score.

0.008

0.006

0.004

0.002

0

−0.002

−0.004

−0.006 2006

2008

2009

2010

2012

Figure 1: Time variation in AGR premium Note: We estimate monthly OLS Fama-MacBeth cross-sectional regression of stock returns of the AGR metric and firm-level controls, using the model in column 5 of Table 4. The Figure displays the trailing 24-month average of the slope on the AGR metric. The sample period is January 2005 to December 2011.

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Finance: A Novel Approach

96

6.

b3964-ch04

Conclusions

Does the market reward well-governed firms? How can we identify these well-governed firms? Several academic papers have relied on anti-takeover provisions and other governance inputs to identify weakly governed firms. Johnson et al. (2009) find that while the G and E measures have some correlation with poor operating performance and low Tobin Q values, they do not generate excess returns when industry clustering is accounted for. Similarly, commercial governance rankings have largely struggled in spite of their much better datasets and sophisticated models. This chapter focuses on MSCI’s AGR governance rating that, unlike other governance ratings, relies on both governance outputs as well as inputs. We document that over the 1997–2011 sample period, a firm’s AGR rating is economically and statistically related to future operating performance as measured by either ROA, sales growth, Tobin’s Q, or net margin. The rating is especially valuable in tracking the performance of firms in the bottom AGR decile, that is, firms with poor corporate governance. In addition to operating performance, we investigate whether loading on AGR ratings generates abnormal stock returns. We document that a long-short portfolio that goes long better governed firms while shorting poorly governed ones delivers approximately a 5% risk-adjusted return even after controlling for an extensive set of risk factors. However, consistent with learning by the market, this abnormal performance has been declining over time. Taken together, our results confirm that corporate governance does systematically affect a firm’s operating performance and its stock price behavior.

Appendix: Variables Construction From the CRSP database, we obtain the Market Value of equity (Size) as the natural log of the product of number of shares outstanding (item shrout) times the absolute value of price per share (prc) times 1,000. From the CRSP/Compustat merged database, we construct the following variables:

page 96

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

What You See Is What You Get

b3964-ch04

page 97

97

Assets Total assets (at ). Debt Sum of financial debt in current liabilities (dlc) and long-term financial debt (dltt). Return On Assets (ROA) Ratio of Operating Income After Depreciation in the current fiscal year (oiadp) to Assets at the end of the prior fiscal year. Leverage Ratio of the sum of long-term financial debt (dltt) plus long-term debt due in one year (dd1 ) to Assets. Tobin’s Q Numerator is the sum of Assets and Market Value of Equity minus the sum of Book Value of Equity and Deferred Taxes (txdb). Denominator is Assets. 3-Year Sales growth Ratio between total sales in current fiscal year (sale) to the total sales of three fiscal years ago. Net Margin Ratio between Income Before Extraordinary Items (ib) to Sales. CAPEX/Assets Ratio of Capital Expenditures (capx ) to Assets. R&D/Sales Ratio of R&D expenses (xrd) to Sales. Following Daines, Gow, and Larcker (2010), ROA is winsorized to have an absolute value not greater than one. All other variables are winsorized at the 2.5% and 97.5% to reduce the impact of outliers. References Ang, A., Hodrick, R. J., Xing, Y., and Zhang, X. (2006). The cross-section of volatility and expected returns. The Journal of Finance, 61, 259–299. Bebchuk, L., Cohen, A., and Ferrell, A. (2009). What matters in corporate governance? Review of Financial Studies, 22, 783–827. Bebchuk, L., Cohen, A., and Wang, C. C. Y. (2013). Learning and the disappearing association between governance and returns. Journal of Financial Economics, 108, 323–348. Biggerstaff, L., Cicero, D., and Puckett, A. (2015). Suspect CEOs, unethical culture, and corporate misbehavior. Journal of Financial Economics, forthcoming. Boyer, B., Mitton, T., and Vorkink, K. (2010). Expected idiosyncratic skewness. Review of Financial Studies, 23, 169–202.

November 4, 2020

98

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch04

Behavioral Finance: A Novel Approach

Core, J., Guay, W., and Rusticus, T. (2006). Does weak governance cause weak stock returns? An examination of firm operating performance and investors’ expectations. Journal of Finance, 61, 655–687. Daines, R. M., Gow, I. D. and Larcker, D. F. (2010). Rating the ratings: How good are commercial governance ratings? Journal of Financial Economics, 98, 439–461. Fama, E. F. and French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3–56. Fama, E. F. and French, K. R. (1996). Multifactor explanations of asset pricing anomalies. Journal of Finance, 51, 55–87. French (2014). A five-factor asset pricing model, Working paper. Fama, E. F. and MacBeth, J. D. (1973). Risk, return and equilibrium: Empirical tests. Journal of Political Economy, 81, 607–636. Giannetti, M. and Want, T. Y. (2016). Corporate scandals and household stock market participation. Journal of Finance, forthcoming. Gompers, P., Ishii, J., and Metrick, A. (2003). Corporate governance and equity prices. Quarterly Journal of Economics, 118, 107–155. Gompers, P. A., Ishii, J., and Metrick, A. (2010). Extreme governance: An analysis of dual-class firms in the united states. Review of Financial Studies, 23, 1051–1088. Hirshleifer, D., Hou, K., and Teoh, S. H. (2012). The accrual anomaly: Risk or mispricing? Management Science, 58, 320–335. Johnson, S., Moorman, T., and Sorescu, S. (2009). A reexamination of corporate governance and equity prices. Review of Financial Studies, 22, 4753–4786. Newey, W. K. and West, K. D. (1987). A simple positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica, 55, 703–708.

page 98

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0005

Chapter 5

Are Courts Biased? The Anchoring Heuristic and Judicial Decisions in Personal Bankruptcy Proceedings Yevgeny Mugerman∗ , Neta Nadiv† and Moran Ofir‡

Abstract This chapter examines the seminal heuristic of anchoring and adjustment and its effects on personal bankruptcy proceedings. Using a unique and detailed database of bankruptcy files we analyze the effect of the official receiver’s recommendation on court decisions. The official receiver in bankruptcy proceedings is appointed by a judicial authority and is required to bring before the court any relevant information needed in order to reach a judicial decision. As part of her responsibilities, the official receiver is required to submit a financial report, which serves as the basis for the court’s proposal for the debtor’s payment plan. This chapter sets out the main factual infrastructure for determining the payment order under bankruptcy proceedings and should include information relevant to the court’s discretion. The richness of the data allows us to investigate the impact of the receiver’s recommendation on court final decisions. We find that overall, the receiver’s recommendation serves as an anchor to the judges, and, moreover, that deviations from this recommendation by the court are extremely rare. Notably, this outcome ∗

Assistant Professor, Finance Department, Bar-Ilan University; [email protected]. Clinical Professor, Harry Radzyner School of Law, Interdisciplinary Center (IDC), Herzliya; [email protected]. ‡ Assistant Professor, Harry Radzyner School of Law, Interdisciplinary Center (IDC), Herzliya; mofi[email protected]. †

99

page 99

November 4, 2020

100

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach differs dramatically from that of corporate proceedings. Since personal bankruptcy proceedings do not allow for substantive oversight, which examines the plausibility of the actions or recommendations that the receiver seeks, there is no pure rational explanation for this finding. Keywords: Bankruptcy, anchoring, judicial decisions, behavioral finance

1.

Introduction

Despite its importance, research into the field of personal insolvency remains limited in scope and receives little acknowledgment. Consequently, even though personal bankruptcy rules have diverse repercussions which affect the lives of many private individuals in substantial ways,many fundamental questions regarding their purpose and likely impact remain largely unaddressed. Studies in the field of bankruptcy and insolvency typically investigate organizations rather than individuals; although some of the basic assumptions of the process are similar, personal insolvency is a completely different legal procedure. In this chapter, we analyze the effect of the official receiver’s recommendation on court decisions in personal bankruptcy cases. The official receiver in bankruptcy proceedings is appointed by a judicial authority and is required to bring before the court any relevant information needed in order to reach a judicial decision. As part of her responsibilities, the official receiver is required to submit a financial report, which serves as the basis for the court’s proposal for the debtor’s payment plan. This report sets out the main factual infrastructure for determining the payment order under bankruptcy proceedings and should include information relevant to the court’s discretion. Using a unique and rich data set which consists of proprietary bankruptcy proceedings cases obtained from one of the aforementioned official receivers, we analyze the judicial decision process. Our data set includes all the relevant cases from this receiver, for a total of 82 personal bankruptcy proceedings. For each case we coded a long list of variables and estimated several regression models to analyze the effect of the recommendation made by the official receiver on the judicial decisions in these cases.

page 100

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 101

101

Our main finding is that the judicial decision at the closing of a case was strongly affected by the numerical “anchor” set by the official receiver in her recommendation. This effect was found to be statistically significant even after controlling for the fixed effect of the judge presiding over the case. It is important to note that the receiver’s recommendation is non-binding, and indeed, in corporate insolvency proceedings, the court’s final decision tends to ignore this recommendation. This finding places our chapter among the growing body of research which analyzes judicial decision-making from a behavioral perspective. This body of research suggests that judges, prosecutors and other legal professionals are prone to heuristics and behavioral biases which affect their decisions in judicial processes and other legal activities. We show that personal bankruptcy judges are prone the seminal heuristic of anchoring and adjustment, and specifically, that they are affected by the recommendation of the official receiver in these cases. The chapter is structured as follows: first, we describe in detail the Israeli personal insolvency system, focusing on judicial specialization and the role of the official receiver in bankruptcy proceedings. Then, we review the literature on the effect of heuristics and biases on judicial decisions, and specifically the literature focusing on bankruptcy judges. We then describe our data and methodology, and present our results. Finally, we end the chapter with some concluding remarks.

2.

The Israeli Personal Insolvency System

The field of insolvency proceedings for individuals is one that has received little attention, despite the breadth and importance of the implications of individual bankruptcy, as well as the substantial percentage of the population that might expect to undergo such proceedings at some point in their lives. The insolvency procedure offers a way to make collective arrangements, which are often needed due to an individual’s inability to fully pay off his or her debts. As part of the proceedings, all the debtor’s assets are examined, and the proceeds are distributed equally among the creditors and in accordance with priority as prescribed by law. The debtor’s financial capabilities, after deduction of expenses, usually determine the terms of the

November 4, 2020

102

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

payment plan ordered by the court, which the debtor must meet. At the end of the procedure, debtors will be entitled to a discharge of their remaining debts, and permitted to turn a new page in their financial life. Israel’s insolvency proceedings for individuals are based on English law, which was originally legislated at the time of the British Mandate, based on the Bankruptcy Act, 1914 (c.59) [Eng.] (Boshkoff, 1982). Despite being based on English law, where insolvency proceedings are similar to continental law practices, Israeli bankruptcy proceedings have not kept up with later reforms in English law, which have made debt discharge more accessible and increasingly resemble the American model (The Enterprise Act, 2002 (c.40) Part 10 [Eng.]). The Bankruptcy Ordinance of 1936 formed the legislative framework for matters concerning individual insolvency in Israel, with the next amendment to the ordinance in 1980, and the recent legislation of the Insolvency and Economic Rehabilitation Law, which was passed in 2018 (hereinafter: “New Insolvency Law”). The new legislation codified existing insolvency laws and repealed outdated ordinances, reforming insolvency proceedings for both individuals and corporations, with a focus on giving greater consideration to the debtor’s recovery and re-integration into economic life, while avoiding infringement upon the debtor’s basic dignity. The law has a clearly stated objective, both in its title and in a specific provision within the law, of facilitating the economic recovery of the debtor in order to grant that debtor a second chance. As expressed in its explanatory notes, the new law has four purposes: to bring about the economic recovery of the debtor; to maximize the amount of debt repaid to creditors; to carry out proceedings in a manner that maintains the basic dignity of the debtor; and to increase the certainty and consistency of the law, all while reducing the bureaucratic burden placed on the debtor (Bill and Explanatory Notes 2016, p. 594). As a general matter, personal insolvency and bankruptcy laws tend to balance two, often related, fundamental policy goals: to provide creditors with a mechanism for facilitating their individual or collective recovery from defaulting debtors, and to provide debtors with a form of relief from their indebtedness and related burdens. In order to obtain these goals, the legal procedure in Israel is structured in such a way as to incorporate professionals who balance the

page 102

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 103

103

interests of both sides, debtor and creditor. The presence of professionals serves a two-fold purpose: first, they provide professional support in the field of insolvency and economic examination to the judiciary; second, the professionals assist in determining the economic solution to be imposed on the debtors and their creditors, while simultaneously considering the public interest throughout the insolvency proceedings. The title of the professional charged with these duties in Israeli courts is the official receiver. In Israel, the official receiver operates as part of an administrative agency subordinate the Ministry of Justice, and is tasked with administering the bankruptcy system. Since insolvency proceedings involve multiple sides, whose interests are occasionally mutual, but more often conflicting, it is appropriate to hold the supervision of these proceedings to a high standard, and place it in the hands of an impartial professional. In this capacity, the official receiver serves as a trustee for the debtor’s assets.

3.

Judicial Specialization

The trend of judicial specialization in courts in a manner that dictates judicial professionalism is not new (Baum, 1977), and it is customary to distinguish two different forms of judicial specialization: the specialization of an entire court, and the specialization of a judicial subject within the general court system, whether by creating a targeted division for certain legal field or by repeatedly directing cases concerning a similar subject matter to the same individual judges (Baum, 1991; Case et al., 1984). Another option for determining the specialization of judges, as suggested by Revesz (1990), is to consider the pre-judicial experience the judges have, as well as the experience they have accumulated in the courts for cases on a certain subject over a long period of time. Baum (2011) emphasizes the importance of neutrality, rather than ideology, when considering the procedure of instituting a specialized court. In his opinion, a judicial system with expertise will improve the quality of judicial decisions and the efficiency of the process and promote consistency in legal results. Two additional advantages exist beyond those put forth by Baum — the development of human capital, and the promotion of judicial doctrines (Dreyfuss,

November 4, 2020

104

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

1990; Jordan, 1981). Studies are being conducted on the advantages and disadvantages of judicial specialization in certain legal areas (for example see Aran and Ofir, 2020), and a growing number of states are seeking judicial specialization (European Commission for the Efficiency of Justice (CEPEJ), 2008), whether by instituting a specialized court for the field, a department within a court, or assigning cases concerning a specific subject to the same judges, creating ad hoc specialization (Cheng, 2008). The official receiver is the dominant force in Israeli bankruptcy proceedings partly because there is no other repeat player in the bankruptcy system with the same degree of expertise, organization, or power. Furthermore, in contrast with the US legal system, there are no specialized bankruptcy courts in Israel, and hence no specialized bankruptcy judges (Shuchman, 1978). In Israel, it is customary to assign cases in the particular field of insolvency to specific judges, thus creating ad hoc professionalization (Cheng, 2008). The Israeli system has no designated court for insolvency proceedings, and until the new Insolvency Law provided specific guidelines regarding the jurisdiction of such proceedings, the bankruptcy cases would be heard at general district courts. The new law transfers this authority to the magistrate courts, which are a lower level of general court. Regarding the situation in Israel, see Nadiv’s study (2020) on judicial specification for insolvency in general courts in Israel. One tends to consider the field of insolvency, which involves both corporate and individual interests, as justifying a similar examination regarding judicial specialization. This idea is supported by the observations of Professor Hann (2018), who demonstrated that the main legal challenge in the field of insolvency arises from the complicated and diverse relations between different fields of law in each debtor’s case. As such, the “ordinary” rules of the legal systems are not appropriate, and judges must exercise particular sensitivity to the various preferences at hand. As will be further elaborated upon, the process in Israel must tackle the conflicting preferences of the debtors and their creditors, alongside the preference of the official receiver. Hann is not alone in making this claim, and as early as 1999, the Levin Committee proposed to reform the field of insolvency proceedings in Israel in a “skinny” legislative arrangement, one that would allow judges to exercise broad discretion (The Public Committee on the Examination of Bankruptcy and Corporate Liquidation, 1999). This

page 104

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 105

105

chapter attempts to analyze the suggestions put forth by the official receiver and demonstrate that they serve as an anchor to judicial decisions. Our study not only shows that judges have made limited use the discretion afforded to them but also that the legislature in Israel has chosen to pursue an alternative system, and the new Insolvency Law significantly reduces the authority and involvement of the judiciary in the process. This is despite the abundance of evidence pointing to the fact that judicial involvement, and even more so judicial specialization, can contribute to a reduction in contradictory case laws or incorrect interpretations of precedent (Damle, 2005). One of the studies regarding specialized bankruptcy courts in the United States examined the relationship between specialization of courts and the preferences and ideologies of the involved parties (Howard et al., 2014), and another study, by Rachlinski et al. (2006), considered the influence of judicial biases on the judges in light of the specialization effect without examining the existence of specialization as such. One main conclusion to be drawn from the aforementioned studies is that judicial specialization, which may accumulate over time spent adjudicating such cases, has direct and significant influence on the results of insolvency proceedings (Posner, 2008; Rachlinski et al., 2006; Ashenfelter et al., 1995). Our study focuses on the official receivers, upon whom judges rely heavily. In the absence of any serious competition for influence, the views and perceptions of the official receiver shape, to a significant extent, the legal culture of Israeli bankruptcy law (Efrat, 2003).

4.

The Role of the Official Receiver as a Professional Player

The two objectives typically found at the core of bankruptcy proceedings are by their very nature characterized by structural tension. The objective of maximizing the debtor’s pool of assets to the advantage of the creditors is naturally at odds with the objective of discharging debts that are not in that debtor’s power to repay. The official receiver stands at the crossroads between these two objectives, as a professional tasked with providing a payment plan based on the highest monthly payment order the debtor is capable of paying under the given circumstances. In adjudicating such a payment

November 4, 2020

106

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

order, it is necessary to establish balance between the two conflicting sets of interests, and the appropriate settlement between them often requires both sides to compromise. In order to fulfill this role, the receiver must be committed to the process itself, as opposed to either of the parties. The receiver must serve as an extension of the court to allow the process to be carried out properly, and the receiver’s professional recommendation is essential to the judicial process. However, Israeli Supreme Court has stated that while the receiver’s position, naturally, is rather strong, the court need not bind itself to the receiver’s recommendation, which should be considered to be of an advisory nature rather than a binding one (HCJ case 6021/06). The Supreme Court also stressed that the district court presiding over the case must exercise independent discretion, while paying attention to a wide variety of considerations. By virtue of the statutory authority the receiver possesses, as an objective body fulfilling a public function in accordance with the law, the primary role of the receiver is to aid the court and provide expert opinions. As stated in Article 68 of the Bankruptcy Ordinance, the legislature determined that courts must treat the official receiver’s documents as evidence in the proceedings. As with any evidentiary document, there is no reason for the court to neglect the proper review of such document, as may be necessary. The role of the court as a supervising authority should be distinguished from its role as a deciding authority, in which professionals state their expert opinion. Insolvency or bankruptcy regimes can potentially reduce the incidence of household over-indebtedness exante and reduce the private and social costs of over-indebtedness when it occurs. They can ensure this by requiring the expedient distribution of a certain portion of a debtor’s assets to creditors, while prioritizing and applying distributive logic. Legislation may also, alternatively, provide for payment of some portion of a debtor’s wage income over a period of time (Ramsay, 2000), depending on the model adopted in the country in question. In general, bankruptcy proceedings are structured in one of two ways: the most common method is a repayment plan across a set number of years, while the other method is a “fresh start” mechanism involving an offer of debt relief or discharge (Cohen, 1982; McCoid, 1996). In Israel, the bankruptcy model examines

page 106

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 107

107

the economic abilities of the debtor. After submitting a petition for insolvency, the court presents a receiving order, and a representative of the receiver acts as an extension of the court charged with evaluating the economic situation and various assets of the debtor. Once the court is convinced that the debtor is insolvent, it will execute a bankruptcy order, initiating the economic payment plan, including, inter alia, the realization of assets for the creditors and the formulation of a payment plan for the debtor to be executed over four years and six months. At the end of the process, any remaining debts owed by the debtor are discharged. The possibility of debt discharge has not always been an integral part of Israeli insolvency proceedings. While even the Bankruptcy Ordinance itself determined that a debtor may be discharged, in contrast with other legal systems, the Ordinance does not allow an automatic grant of discharge as a custom, unlike, for instance, United States law, which allows it under Chapter 7 of the Insolvency Act (1986). Article 63 of the Ordinance allows for discharge as an exception only, and Efrat discovered in his research that Article 63 is almost never used in practice (Rafi, 2003). The court noted, in one case, before the new Insolvency Law amendment, that release from debts is to be allowed only in exceptional cases, emphasizing: “A review of the case law indicates that in the cases for which the court found it appropriate to accept the request for immediate discharge of debts were cases in which the official receiver supported the request or cases in which there were highly exceptional circumstances which justified granting the request” (DC case 5147/04). The new Insolvency Law strengthens the role of the official receiver, as a professional charged with examining the debtor’s capabilities, to develop the economic plan for repayment of the debt. In this regard, the receiver, who has historically had the role of collecting payment from the debtor and thoroughly examining the debtor’s assets for the benefit of covering debts, now additionally has the role of acting in the interest of the debtor’s economic recovery. The receiver must maintain a delicate balance between opposing economic interests, almost a constitutional balance, between the rights of the debtor and rights of the creditors, further providing legitimacy the claim that it is appropriate for the courts to act

November 4, 2020

108

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

as the decisive authority in such proceedings (Explanatory Notes, pp. 596, 658).

5.

Different Judicial Approaches — What Shapes and Impacts the Legal Proceedings?

The Israeli perspective determines the payment order while considering the debtor’s entire disposable income, with the deduction of reasonable household expenses. The permitted expenses are not fixed and are determined based on the family unit of the debtor, including how many family members live with the debtor and are dependent on him or her, and taking into account the specific personal circumstances of each individual debtor. Setting a payment plan and examining assets is the heart of the judicial process in insolvency cases. The plan is a judicial decision, the basis of repayment to the creditors, in effect an adjudication between the sides, balance between the debtor interested in meeting their obligations but unable, and the creditors, who are interested in being repaid as much of the debt as possible. Before the new Insolvency Law was passed, a governmental committee named the Harris Committee (2015) was convened. The Committee subsequently submitted its recommendations, which indicated that the recommendations made by professionals for monthly payment orders vary by hundreds of Israeli shekels between cases, even when presented with similar economic circumstances. The different approaches taken with regard to determining the debtor’s economic capabilities, coupled with the lack of clear guidelines on the subject, have led to inconsistencies across recommendations for monthly payments submitted to the courts, and as a result there are significant gaps in different payment orders determined by courts in bankruptcy cases. Concerningly, the consequence of this is that it is highly probable that the same court will reach different decisions for similar sets of circumstances on separate occasions. Without clear guidance, different approaches have been developed for examining a debtor’s economic capabilities, both regarding earning capacity and actual income, as well as for evaluating the debtor’s economic conduct and his or her necessary expenses to maintain a decent living standard. These are further exacerbated by

page 108

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Are Courts Biased?

page 109

109

the differences between different individual professionals (Shuchman, 1978). The greater the number of bankruptcy petitions, the more difficult it is to ensure equitable treatment, and the more pressing the need for clear, standardized criteria for the demands of the bankruptcy process. In order to comprehend the scope of insolvency proceedings in Israel, we reviewed annual reports published by Insolvency Commissioner of the Ministry of Justice (2013 till 2018). In the last six years, an average of 52,000 individual bankruptcy cases were heard by Israeli district courts. This number highlights the difficulty faced by the judiciary and official receiver in achieving equality regarding monthly payments orders. Also noteworthy is the annual increase in caseload as a result of the number of new petitions filed each year, which seems to be increasing at a much higher rate than the rate at which existing cases are closed:

Year

New Cases (N)

Open Cases (N)

2018 2017 2016 2015 2014 2013

19,854 21,231 19,277 16,491 14,497 12,131

65,628 62,417 55,657 50,296 43,185 36,721

An incorrect calculation of the monthly payment amount is likely to harm either the debtor or the creditors. Setting a monthly payment order that is too high often results in further harm to the economic survival of the debtor and his or her dependents, as well hurting the debtor’s basic human dignity. An excessive monthly payment may result in the debtor once again being unable to meet obligations, in effect exposing the debtor to the possible cancellation of the process or, alternatively, forcing the debtor to turn to other players for economic aid, such as family and friends, despite the fact that those individuals are not required by law to cover the payments imposed upon the debtor. On the other hand, an insufficient monthly payment order harms creditors, as well as violating the economic principle

November 4, 2020

110

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

which states that debtors must act to the best of their abilities to ensure maximal repayment of their obligations as a condition for being discharged from remaining debts. This complex issue results in parties regularly appealing to the Supreme Court, seeking either to reduce or to increase the monthly payment imposed upon the debtor by court order (these cases were heard in the Supreme Court until the new law came into effect, following which the jurisdiction was transferred to the magistrate courts, making district courts the appellate courts). The ultimate result is an overwhelming number of petitions to the courts, costing the judicial system considerable time and resources, diverting them away from more productive uses as courts are forced to correct errors and attempt to foster consistency and standardization between all bankruptcy cases being adjudicated. In light of all this, and in consideration of the absence of appropriate standards for structuring professional discretion in determining payment plans, the courts are in a significant position for judicial review upon the economic recommendations placed before them. It follows that the monthly payment order is often determined based on intuition and in an ad hoc manner. The official receiver reviewing the debtor’s reported household expenses makes an examination based on intuitive discretion, basing this decision on his or her personal experience and worldview, in order to determine which of the debtor’s expenses are reasonable and which are not. For bankrupt individuals in similar economic situations, each official receiver seems to make different decisions also in relation to the calculation of family members’ incomes and determining the recommended monthly payment. This variation between official receivers’ recommendations corroborates the need for judicial review and supervision. The number of judges adjudicating insolvency cases in Israel is small, and there are ten times as many official receivers as there are such judges. It follows, therefore, that a given judge may be exposed to the opinions of multiple official receivers, and is therefore the best equipped to ensure consistency between decisions. This is not a critique at the micro level, directed towards any specific recommendation made by any particular official receiver, but rather a critique at the macro level of the general manner in which insolvency proceedings are carried out, coupled with a desire to establish standardization to the

page 110

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 111

111

greatest extent possible. The system is in dire need of clear and consistent rules and a simplified procedure for determining monthly payments; the role of the court cannot be restricted to that of a mere “manager” of the process. To summarize, the official receiver’s almost total control in shaping the legal culture in personal bankruptcy procedures in Israel is, among other things, due to the unmatched deference given to official receivers by judges. By contrast, in Canada and the United States, the equivalents of the official receiver are not routinely involved in investigating the debtor’s financial affairs and collecting on the outstanding obligations on behalf of creditors, and, instead, they limit their activities to facilitating the bankruptcy process (Ramsay, 2000; Sullivan et al., 1989). Contrary to what may be expected, judges seem to have little interest, and limited time available, to devote to each case. Therefore, they tend to defer to a great extent to the official receiver’s recommendations relating to debtors’ bankruptcy petitions. Efrat (2003) reached similar conclusions in his research, finding that judges closely adhered to the official receiver’s recommendation on whether to approve the debtors’ petition for bankruptcy protection. In his study, he found that in 96.9% of cases the court approved the debtor’s bankruptcy application and issued a receiving order based on the recommendation of the official receiver (N = 185). His study also demonstrated how judges follow the official receiver’s position regarding debtors’ applications for a reduction of monthly payments. With the insolvency institution constantly changing in favor of the debtor, with the goal of attaining a point of balance between debtor and creditors, we would expect to see the court performing its nature role of judicial review. However, although approximately 20 years have passed since Efrat’s research, our study shows that official receivers still enjoy almost exclusive power to influence results, playing a critical role in shaping the legal culture of the Israeli bankruptcy system. The changes made by new Insolvency Law, which place the bankruptcy procedure under the responsibility of the official receiver (titled the “commissioner” in the law), essentially guarantee that the entire process is conducted outside the court itself, further underlining the importance of this study (New Insolvency Law, Article 266).

November 4, 2020

112

6.

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

Heuristics and Biases in Judicial Decisions

There is a growing body of research which analyzes judicial decision making from a behavioral perspective. This body of research suggests that judges, prosecutors and other legal professionals are prone to heuristics and behavioral biases which affect their decisions in judicial processes and other legal activities. Heuristics and biases were shown to exist in various parts of the legal procedure, such as the hearing process, the ruling process, and the sentencing process. The list of heuristics and biases found to affect judicial processes and decisions includes hindsight bias, confirmation bias, conjunction fallacy, anchoring and adjustment and many more. Overall, this body of literature suggests that irrelevant factors, that should not affect the legal procedure and the legal judgment, might cause systemic biases in judicial decisions, and judges are often unaware of this fact. In the following paragraphs we will briefly describe some prominent examples from the literature. Rassin et al. (2010) found that confirmation bias affects judges, lawyers and police officers. This confirmation bias causes decision makers to interpret and analyze information in a manner consistent with their prior beliefs and assumptions. In court, this may lead to biased decisions and verdicts as judges are affected by confirmation bias when presented with evidence during trials. Harley (2007) found that hindsight bias affects judges, particularly in liability cases. Hindsight bias occurs when decision makers evaluate events after their occurrence and judge them as being more probable and predictable than they really were before in advance. The hindsight bias refers to the difference between foresight and hindsight — although an event is less predictable before it occurs, after it occurs decision makers cannot ignore information on the its occurrence and assign it a higher probability than beforehand. Judges, essentially by definition, always evaluate events in hindsight, and as a result reach biased decisions and judgments. Fox and Birke (2002) found that lawyers are often affected by the conjunction fallacy. When the conjunction fallacy occurs, a more detailed description of an event leads to a higher estimate of the event’s probability. In other words, decision makers wrongfully believe that events described in greater detail are more probable than those whose description is less detailed. Although this research was

page 112

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Are Courts Biased?

page 113

113

conducted on lawyers rather than judges, Peer and Gamliel (2013) argue that judges might be prone to the conjunction fallacy as well. As Tversky and Kahneman (1974) found, anchoring and adjustment refers to the process of assimilating a numeric estimate towards a previously considered standard. Enough and Mussweiler (2001) found that judges are affected by anchoring bias in their judicial decisions. More specifically, they found that in criminal sentencing decisions that require a numeric decision on years of imprisonment, for example, judges are affected by numeric anchors such as minimal sanctions stipulated by regulations or numeric sanctions recommended by the prosecution. Guthrie et al. (2001) found that judges rely on misleading heuristics in some circumstances. Specifically, they found that judges rely on misleading numeric anchors when assessing damage awards, are sensitive to the presentation of a settlement as a gain or as a loss, are prone to the hindsight bias in assessing the potential outcome on appeal, inappropriately disregard base rate information relevant to decisions, and suffer from self-serving bias in assessing their own likelihood of being overturned on appeal. As judges are only human, they make the same kinds of mistakes as the rest of us when making judgments, and are prone to behavioral biases that might affect their judicial decisions. Awareness of the heuristics and biases affecting judicial decisions is the first necessary step towards debias. Our chapter is part of the body of literature which specifically attempts to identify effect of behavioral biases on judges, and contributes to the process of raising awareness and eliminating biases.

7.

Heuristics and Biases in Bankruptcy Proceedings

Bankruptcy judges constantly make decisions involving risk assessments, numerical values of fees, debts and incomes. When they decide in a personal bankruptcy case, they risk imposing losses on the creditors. When they decide on a debtor’s payment plan, they are dealing with numerical values. Each personal bankruptcy case can be framed as involving gains or losses to the creditors. Therefore, there is a good reason to predict that bankruptcy judges are no less prone to biases than generalist judges.

November 4, 2020

114

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

In an experimental study, Rachlinski et al. (2006, 2007) found that bankruptcy judges relied on irrelevant anchors and were influenced by the framing effect. However, at the same time, they found that judges resisted the effects of the omission bias, were unaffected by the race of the debtors, were indifferent to a debtor’s apology, and were not affected by a transient emphasis on their own morality. When comparing bankruptcy judges to generalist judges, they found no evidence that specialization had any beneficial effect on the extent to which judges are prone to behavioral biases. Therefore, they argue that there is no evidence that specialization in bankruptcy cases allows judges to develop better cognitive skills. Whereas Rachlinski et al. (2006, 2007) used a lab experiment to test the effect of behavioral biases on bankruptcy judges, our study is a field research based on court cases. We have collected and coded data from proprietary bankruptcy cases, and performed an econometric analysis in order to test whether behavioral bias is a significant factor affecting judges’ “real-life” decisions.

8.

Anchoring and Adjustment

The anchoring and adjustment effect is one of the most frequently tested behavioral heuristics. Highly robust, it also has a variety of implications for the decision-making process. As Tversky and Kahneman (1974) explain in their seminal paper on anchoring, which offered the first description of this heuristic, decision makers conduct estimates by starting from an initial value that is adjusted to yield the final answer, but the adjustments are typically insufficient: “different starting points yield different estimates, which are biased toward the initial values” (p. 1128). In their paper, Tversky and Kahneman (1974) asked the subjects of the experiment comparative and absolute questions regarding the percentage of African nations in the United Nations (UN). In the comparative question, subjects indicated whether the percentage of African nations is higher or lower than a numeric anchor. The numeric anchor was determined according to the spin of a roulette wheel. Then, the subjects were asked an absolute question regarding their personal estimate of the actual percentage of African nations in the UN. They found that the absolute judgments were assimilated

page 114

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 115

115

to the provided numeric anchor value. The mean estimate of the percentage of African nations in the UN among subjects who received the high numeric anchor was 45%, compared to 25% for subjects who received the low numeric anchor. A significant number of studies demonstrate the prevalence of the anchoring heuristic described by Tversky and Kahneman (e.g., Plous, 1989; Chapman and Johnson, 1999; Epley and Gilovich, 2001; Mussweiler and Englich, 2005; McElroy and Dowd, 2007; and more recently Hurwitz et al., 2020; see Furnham and Boo, 2011, for literature review). Most of the existing research papers are experimental studies, conducted with university students in laboratory settings with a list of questions that the participants may not have used in natural situations. A small number of studies have had the participants face real-life settings, and also showed the heuristic to be robust (e.g., Ariely et al., 2003; Englich et al., 2005; Critcher and Gilovich 2008; and more recently Mugerman et al., 2016; Mugerman and Ofir, 2020; Camanho et al., 2020). Regarding the magnitude of the anchoring heuristic, the literature shows that the higher the ambiguity, and the lower the familiarity, relevance or personal involvement with the problem, the stronger the anchoring effect (Van Exel et al., 2006). In addition, the literature shows that the informational relevance of values may affect decision makers’ susceptibility to the anchoring effect (Hastie et al., 1999; Marti and Wissler, 2000; Englich et al., 2005). More specifically, Strack and Mussweiler (1997) show that anchor values which are similar or identical in judgmental dimensions to the estimates yield a significant effect on the magnitude of anchoring. As previously mentioned, judges have also been found to be affected by anchors in their judicial decisions. Enough and Mussweiler (2001) found that judges are prone to the anchoring heuristic in criminal sentencing decisions where numerical anchors are set by law or recommended by prosecutors. Ebbesen and Koneani (1981) showed that anchors affect judicial decisions when they are presented in court by attorneys and probation officers. Moreover, there is evidence that anchoring affects the court ruling even when the judges have previously declared that the anchor is irrelevant to the judicial decision. Englich et al. (2006) found that the anchor affects judges even when the numeric anchor is provided by a journalist, considered to be an irrelevant source in criminal cases.

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

116

In addition, several studies compared the effect of anchoring on specialized judges to its effect on generalist judges. Overall, these studies found that the decisions of specialized judges are also susceptible to anchoring (Northcraft and Neale, 1987; Mussweiler et al., 2000; Rachlinski et al., 2006, 2007). 9.

Data

Our main source of data consists of proprietary bankruptcy proceedings cases data obtained from one of the aforementioned official receivers. Our data set includes all the relevant cases from the aforementioned official receiver, amounting to 82 bankruptcy proceedings in total. It is worth noting that the assignment of cases to the different official receivers is done at random, with the geographical location of the office being the only determining factor. The official receiver who provided us with our data handles cases which are presented before two district courts, one of which is located in the city of Tel Aviv (Tel Aviv District), and the other in the city of Lod (Central District). Table 1 demonstrates the descriptive statistics of our main explanatory variables. It should be stressed that the aforementioned data is similar to the general data presented in the following report, published by the Israeli Ministry of Justice committee for reviewing the repayment plan in bankruptcy proceedings (in Hebrew). The report covers all personal bankruptcy proceedings conducted in Israel during the period of time concerned: https://www.justice.gov.il/Units/Apotro posKlali/PressRoom/Documents/haris.pdf. The following figures shed additional light on the descriptive statistics of our main explanatory variables. Figure 1 shows the distribution of debtors by gender, showing that over 60% of them are male. Figure 2 shows the age of the debtor at the start of bankruptcy proceedings. The mean age is 47.2 and the median 44.5, with ages ranging from individuals in their 30s to some in their 70s and 80s. Figure 3 displays the monthly income of the debtors. Figure 4 shows the distribution of debtors’ monthly expenses, and Figure 5 shows their discretionary income, calculated as the difference between income and expenses. Figure 6 shows the monthly sum the debtor was ordered to direct towards repayment of the debt according

page 116

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

page 117

Are Courts Biased?

Table 1: Variable Name Age

Duration

Income Expenses Discretionary income Court ruling

Repayment gap

Gender Alimony

117

Descriptive statistics

Mean

Age at start of proceedings (in years) Duration of proceedings (in months) (in NIS) (in NIS) Income – Expenses (in NIS) Sum to be repaid as per court ruling (in NIS) Discretionary income – court ruling (in NIS) Male = 1, Female = 0 Yes = 1, No = 0

47.2

44.5

12.32

82

26.4

23.5

11.82

82

8,732.7 8,103.5 629.2

7,523.5 7,191.0 364.5

4,229.44 3,912.95 1,208.39

82 82 82

555.43

82

1,117.25

82

787.7

Median

Standard Deviation

Description

500

−158.5

−210.5

Observations

0.61

82

0.134

82

39.0%

61.0%

Male

Figure 1:

Female

Gender

November 4, 2020

118

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

Figure 2:

Age at start of proceedings

Figure 3:

Income (NIS)

to the court ruling, and finally, Figure 7 shows the repayment gap, that is, the difference between the debtors’ discretionary income and the repayment amount determined by the court ruling. All data are in Israeli Shekels (NIS). Figures 3–7 paint a clear picture, showing that the repayment gap of most households is negative, i.e., payments are typically higher

page 118

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

Figure 4:

Figure 5:

b3964-ch05

page 119

119

Expenses (NIS)

Discretionary income (NIS)

than the household’s ability to pay. There is a significant positive correlation between the repayment sum dictated by the court order and the ability to pay, but no significant correlation between the court order and the size of the debt. Additional control variables used were as follows: objection by the sides involved in the proceeding (objection by the official receiver

November 4, 2020

120

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

Figure 6:

Figure 7:

Court ruling (NIS)

Repayment gap (NIS)

and/or by the other lenders), whether the debtor acted in good faith (hiding sources of income, etc.), as well as controls for the district over which the court has jurisdiction.

page 120

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Are Courts Biased?

10.

page 121

121

Methodology

Our behavioral hypothesis is that the judicial decision at the closing of a case will be affected by the numerical “anchor” set by the official receiver in her recommendation. Furthermore, this effect should be statistically significant even after controlling for the fixed effect of the judge. It is important to note that the receiver’s recommendation is non-binding, and indeed, in corporate insolvency proceedings, the court’s final decision tends to ignore this recommendation. At a more formal level, we estimate the following model: court rulingn,t = α + β1 × debt sizen,t + β2 × malen,t + β3 × agen,t + β4 × durationn,t + β5 × d alimony n,t + β6 × or objectionn,t + β7 × lender objectionn,t + β8 × d good faithn,t + β9 × recommendationn,t + β10 × D TLVn,t + β11 × incomen,t + β12 times expensesn,t + JFE + YFE + εn,t where court rulingn,t represents the sum the debtor in case n in time period t was ordered to pay by the court, debt sizen,t represents the amount of money owed by debtor n in time period t, malen,t is a dummy variable which equals 1 if debtor n in time period t is male or 0 if female, agen,t is the age of debtor n at time period t, durationn,t represents the total duration of time of since the opening of case n in time period t, d alimonyn,t is a dummy variable which equals 1 if debtor n is required to pay alimony in time period t or 0 otherwise, or objectionn,t is a dummy variable which equals 1 if the official receiver made an objection in case n in time period t or 0 otherwise, lender objectionn,t is a dummy variable which equals 1 if a lender made an objection in case n in time period t or 0 otherwise, d good faithn,t is a dummy variable which equals 1 if debtor n is assessed to be acting in good faith in time period t or 0 otherwise, recommendationn,t is the repayment sum recommended by the official receiver in case n in time period t, D TLVn,t is a dummy variable which equals 1 if case n was handled at the Tel Aviv District Court and 0 if it was handled at Lod District Court in time period t, incomen,t represents the income of debtor n in time period

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

122

t, expensesn,t represents the expenses of debtor n in time period t, JFE represents the judge fixed effects, YFE represents the year fixed effects, and εn,t represents the error for case n in time period t. 11.

Results

Table 2 demonstrates the results of the econometric analysis of our data. Table 2:

Econometric analysis (1)

Male

19.02755 (27.97072)

Debt size Income Expenses Age

−0.0209249 (0.0126891) 0.01892 (0.0134388) 2.661254∗∗ (1.14476)

Duration Alimony Tel Aviv −0.9612278 (29.6265) Recommendation 0.9899148∗∗∗ (0.0289271) Official receiver objection 87.04031 (62.28738) Lender objection 20.43053 (83.49135) YFE N JFE N Observations 82

Good faith

(2) 7.419018 (27.65711) −0.0000103 (0.0000166) −0.0228712∗ (0.0123006) 0.0202106 (0.0130878) 2.487318∗∗ (1.130443) −3.20512∗∗∗ (1.124543) −31.9281 (38.86238) 21.41419 (28.71843) −7.580939 (28.69448) 1.000835∗∗∗ (0.029594) 121.8078∗∗ (61.32412) −10.49014 (85.33625) N N 82

(3) −18.44351 (34.6189) −0.0000143 (0.0000197) 0.0243291 (0.015264) 0.0219491 (0.0157765) 2.411882 (1.447491) −3.862484∗∗ (1.470965) −29.66003 (56.49839) −2.9523 (132.9336) 0.6291457 (35.28654) 0.986587∗∗∗ (0.0432833) 164.5334∗∗ (78.42942) 10.20195 (115.6619) Y Y 82

Note: ∗ , ∗∗ , ∗∗∗ represent statistical significance levels of 10%, 5% and 1%, respectively.

page 122

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 123

123

The results point to a strong and statistically significant positive correlation between the official receiver’s recommendation and the final ruling in the case. As evident in the table, this correlation persists after controlling for the remaining variables in the case, and remains stable at a 1% significance level both with and without controlling for additional variables including the size of the debt, the duration of the case, the presence of alimony payments and the jurisdiction to which the case belongs. It also persists both with and without taking into account the fixed effects of both the judge and the period. It is worth mentioning a few other variables which appear to have a significant correlation with the court ruling, which were not investigated in-depth in this chapter but may warrant further examination. The duration of the proceedings appears to have a strong negative correlation with the court ruling. The age of the debtor seems to be significant at the 5% level, although this significance disappears once judge and jurisdiction fixed effects are factored into the model. Finally, and perhaps most interestingly, once all variables are taken into account, an objection made by the official receiver also appears to be strongly positively correlated with the court ruling, another finding which may potentially indicate that judges show some form of bias towards the official receiver.

12.

Conclusion

Our study analyzes the effect of the official receiver’s recommendation on court decisions, examining the heuristic of anchoring and adjustment and its effects on personal bankruptcy proceedings. We have set out the main factual groundwork for determining the payment order in bankruptcy proceedings, based on the payment plan recommended by the official receiver. Due to the absence of clear guidelines, we found significant gaps in different payment orders determined by courts in bankruptcy cases. Unlike other countries, in Israel, the official receiver has a main role in shaping the personal bankruptcy procedures. We examined the effect of the official receiver’s recommendations on “real-life” judicial decisions. Although the receiver’s recommendation is nonbinding, the results point to a strong and statistically significant

November 4, 2020

10:11

124

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

positive correlation between the official receiver’s recommendation and the final ruling in the case. As a result, we determine that the judicial decision at the closing of a case is in fact affected by the numerical “anchor” set by the official receiver in her recommendation. More specifically, we found that not only does the receiver’s recommendation serve as an anchor to the judges but also that deviations by the court from this recommendation are exceedingly rare. It is important to note that our finding is specific to personal insolvency cases, as in corporate bankruptcy proceedings, the court’s final decision often disregards the official receiver’s recommendation. All the findings described previously provide us with additional evidence that judges are affected by anchors in their judicial decisions, and that as a result official receivers enjoy almost exclusive power to influence the results, of bankruptcy proceedings, playing a critical role in shaping the legal culture of the Israeli bankruptcy system. This finding has a significant impact on the outcome of personal insolvency cases and on debtors’ probability of financial rehabilitation. In addition, it sheds light on the efficiency of judicial proceedings in personal bankruptcy cases, by comparing the costs associated with the legal process to the added value of the “anchored” judicial decisions.

References Annual reports, Insolvency Commissioner of the Ministry of Justice (2013). Available at: https://www.gov.il/he/departments/publication s/reports/summary activities 2013. Annual reports, Insolvency Commissioner of the Ministry of Justice (2014). Available at: https://www.gov.il/he/departments/publications /reports/summary activities 2014. Annual reports, Insolvency Commissioner of the Ministry of Justice (2015). Available at: https://www.gov.il/he/departments/publications /reports/summary activities 2015. Annual reports, Insolvency Commissioner of the Ministry of Justice (2016). Available at: https://www.gov.il/he/Departments/publication s/reports/summary activities 2016. Annual reports, Insolvency Commissioner of the Ministry of Justice (2017). Available at: https://www.gov.il/he/Departments/publications /reports/summary activities 2017.

page 124

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 125

125

Annual reports, Insolvency Commissioner of the Ministry of Justice (2018). Available at: https://www.gov.il/BlobFolder/reports/summary activi ties 2018/he/SummaryReport2018.pdf. Aran Y. and Ofir, M. (2020). The effect of specialized court over time. In S. Ranchordas and Y. Roznai (Eds.), Time, Law and Change: An Interdisciplinary Study, Ch. 7. Oxford, Hart Publishing, 167–188. Ariely, D., Loewenstein, G., and Prelec, D. (2003). “Coherent arbitrariness”: Stable demand curves without stable preferences. The Quarterly Journal of Economics, 118, 73–105. Ashenfelter, O., Eisenberg, T., and Schwab, S. J. (1995). Politics and the judiciary: The influence of judicial background on case outcomes. The Journal of Legal Studies, 24, 257. Bankruptcy Act, 1914 c.59 [Eng.] Bankruptcy Ordinance 5696-1936, SH 566, p. 21 [Isr.] Bankruptcy Ordinance amendment 5740-1980, SH 34, p. 639 [Isr.] Baum, L. (1977). Judicial specialization, litigant influence, and substantive policy: The Court of Customs and Patent Appeals. Law & Society Review, 11, 823–827. Baum, L. (1991). Specializing the federal courts: Neutral reforms or efforts to shape judicial policy? Judicature, 74, 217–218. Baum, L. (2011).Specializing the Courts. Chicago: The University of Chicago Press Books, p. 32. Boshkoff, D. G. (1982). Limited, conditional, and suspended discharges in an American bankruptcy proceedings. University of Pennsylvania Law Review, 131, 69–79. Camanho, N., Levine-Schnur, R., and Farber, T. (2020). Dealing with an anchoring bias in the mortgage market: Regulatory approach. In R.Levine-Schnur (Ed.), Measuring the Effectiveness of Real-Estate Regulation, Ch. 10 Berlin: Springer. Case, T. H. and Miller, S. R. (1984). An appraisal of the Court of Appeals for the federal circuit. Southern California law Review, 57, 301–306. Chapman, G. B. and Johnson, E. J. (1999). Anchoring, activation, and the construction of values. Organizational Behavior and Human Decision Process, 79, 1–39. Cheng, E. K. (2008). The myth of the generalist judge: An empirical analysis of opinion specialization in the Federal Courts of Appeals. Stanford Law Review, 61, 519. Cohen, J. (1982). The history of imprisonment for debt and its relation to the development of discharge in bankruptcy. The Journal of Legal History, 3(153), 156–157. Critcher, C. R. and Gilovich, T. (2008). Incidental environmental anchors. Journal of Behavioral Decision Making, 21, 241–251.

November 4, 2020

126

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

Damle, S. V. (2005). Specialize the judge, not the court: A lesson from the German constitutional court. Virginia Law Review, 91, 1267, 1275– 1279. DC 5147/04 Hiller Itay v. Official Receiver (Nevo Legal Database, December 26, 2004). Department of Justice, The committee for Repayment Plan Examination In bankruptcy proceeding (Harris Committee, November 8, 2015). Available at: https://www.justice.gov.il/Units/ApotroposKlali/PressRoom /Documents/haris.pdf. Dreyfuss, R. C. (1990). Specialized adjudication. BYU Law Review, 64, 377–409. Ebbesen, E. B. and Koneani, V. J. (1981). The process of sentencing adult felons: A casual analysis of judicial decisions.In B. S. Dennis (Ed.), The Trial Process. Berlin: Springer, p. 413. Efrat, R. (2003). Legal culture and bankruptcy: A comparative perspective. Bankruptcy Development Journal, 20, 351. Englich, B., Mussweiler, T., and Strack, F. (2005). The last word in court — A hidden disadvantage for the defense. Law and Human Behavior, 29, 705–722. Englich, B., Mussweiler, T., and Strack, F. (2006). Playing dice with criminal sentences: The influence of irrelevant anchors on experts’ judicial decision making. Personality and Social Psychology Bulletin, 32, 188. Enough, B. and Mussweiler, T. (2001). Sentencing under uncertainty: Anchoring effects in the courtroom. Journal of Applied Social Psychology, 31, 1535. Epley, N. and Gilovich, T. (2001). Putting adjustment back into the anchoring and adjustment heuristic: Differential processing of self-generated and experimenter-provided anchors. Psychological Science, 12, 391–396. European Commission for the Efficiency of Justice (CEPEJ) (2008). European judicial systems, Edition 2008 (data 2006). Available at: https://rm.coe.int/european-judicial-systems-edition-2008-data-2006efficiency-and-qualit/16807477ba. Fox, C. R. and Birke, R. (2002). Forecasting trial outcomes: Lawyers assign higher probability to possibilities that are described in greater detail. Law and Human Behavior, 26, 159. Furnham, A. and Boo, H. C. (2011). A literature review of the anchoring effect. The Journal of Socio-Economics, 40, 35–42. Guthrie C., Rachlinski, J. J., and Wistrich, A. J. (2001). Inside the judicial mind. Cornell Law Review, 86, 777. Hann, D. (2018), Insolvency, 1, 14 (sec. ed.) Harley, E. M. (2007). Hindsight bias in legal decision making. Social Cognition, 25, 48.

page 126

November 4, 2020

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

Are Courts Biased?

b3964-ch05

page 127

127

Hastie, R., Schkade, D. A., and Payne, J. W. (1999). Juror judgment in civil cases: Effects of plaintiff’s requests and plaintiff’s identity on punitive damages awards. Law and Human Behavior, 23, 445–470. HCJ 6021/06. Dganit Figon v. Official Receiver [2009]. Howard, R. M. and Brazelton, S. (2014). Specialization in judicial decision making: Comparing bankruptcy panels and federal district court judges. American Bankruptcy Institute Law Review, 22, 407. Hurwitz, A., Sade, O., and Winter, E. (2020). Unintended consequences of minimum annuity laws: An experimental study. Journal of Economic Behavior and Organization, 169, 208–222. Insolvency and Economic Rehabilitation Law, 5778-2018, SH 2708, p. 310 [Isr.] Jordan, E. R. (1981). Specialized courts: A choice? Northwestern University Law Review, 76, 745–747. Marti, M. W. and Wissler, R. L. (2000). Be careful what you ask for: The effect of anchors on personal injury damages awards. Journal of Experimental Psychology: Applied, 6, 91–103. McCoid, J. C. (1996). Discharge: The most important development in bankruptcy history. The American Bankruptcy Law Journal, 70, 163– 167. McElroy, T. and Dowd, K. (2007). Susceptibility to anchoring effects: How openness-to-experience influences responses to anchoring cues. Judgement and Decision Making, 2, 48–53. Mugerman, Y., Ofir, M., and Wiener, Z. (2016). How do homeowners choose between fixed and adjustable rate mortgages. Quarterly Journal of Finance, 6(4), 1–21. Mugerman, Y. and Ofir, M. (2020). Anchoring and adjustment in the mortgage market: A regulatory experiment. In Levine-Schnur R. (Ed.), Measuring the Effectiveness of Real-Estate Regulation, Ch. 9. Berlin: Springer. Mussweiler, T. and Englich, B. (2005). Subliminal anchoring: Judgmental consequences and underlying mechanisms. Organizational Behavior and Human Decision Processes, 98, 133–143. Mussweiler, T., Srtack F., and Pfeiffer, T. (2000). Overcoming the inevitable anchoring effect: Considering the opposite compensates for selective accessibility. Personality and Social Psychology Bulletin, 26, 1142. Nadiv, N. (2020). From overwhelmed courts to specialized courts, the narrative and facts in insolvency proceedings. IDC Law Review, forthcoming. Northcraft, G. B. and Neale, M. A. (1987). Experts, amateurs, and real estate: An anchoring and adjustment perspective on property pricing decisions. Organizational Behavior and Human. Decision Process, 39, 84.

November 4, 2020

128

10:11

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch05

Behavioral Finance: A Novel Approach

Peer, E. and Gamliel, E. (2013). Heuristics and biases in judicial decisions. Court Review: The Journal of the American Judges Association, 49, 114. Plous, S. (1989). Thinking the unthinkable: The effects of anchoring on likelihood estimates of nuclear war. Journal of Applied Social Psychology, 19, 67–91. Posner, E. A. (2008). Does political bias in the judiciary matter?: Implications of judicial bias studies for legal and constitutional reform. University of Chicago Law Review, 75, 853. Rachlinski, J. J., Guthrie, C., and Wistrich, A. J. (2006). Inside the bankruptcy judge’s mind. Boston University of Law Review, 86, 1227. Rachlinski, J. J., Guthrie, C., and Wistrich, A. J. (2007). Heuristics and biases in bankruptcy judges. Journal of Institutional and Theoretical Economics, 163, 167. Ramsay, I. (2000). Market imperatives, professional discretion and the role of intermediaries in consumer bankruptcy: A comparative study of the Canadian trustee in bankruptcy. American Bankruptcy Law Journal, 74(399), 406, 438–453. Rassin, E., Earland, A., and Kuijpers, I. (2010). Let’s find the evidence: An analogue study of conformation bias in criminal investigations. Journal of Investigative Psychology and Offender Profiling, 7, 231. Revesz, R. L. (1990). Specialized courts and the administrative lawmaking system. University of Pennsylvania Law Review, 138, 1111–1160. Shuchman, P. (1978). Field observations and archival data on execution process and bankruptcy in Jerusalem. American Bankruptcy Law Journal, 52(341), 355–356. Strack, F. and Mussweiler, T. (1997). Explaining the enigmatic anchoring effect: Mechanisms of selective accessibility. Journal of Personality and Social Psychology, 73, 437–446. Sullivan, T. A., Warren, E., and Westbrook, J. L. (1989). As We Forgive Our Debtors: Bankruptcy and Consumer Credit in America. Oxford: Oxford University Press, p. 26. The Enterprise Act 2002 (c.40) part 10 [Eng.]. The Insolvency and Economic Rehabilitation Bill and Explanatory Notes (2016). 5776–2016, HH 1027, p. 592 [Isr.]. The Public Committee on the Examination of Bankruptcy and Corporate Liquidation (1999). Corporate Recovery (January 22). Tversky, A. and Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131. Van Exel, N., Brouwer, W., van den Berg, B., and Koopmanschap, M. (2006). With little help from an anchor: Discussion and evidence of anchoring effects in contingent valuation. Journal of Socio-Economics, 35, 836–853.

page 128

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Part II

Investor Behavior and Methodological Novelties

129

page 129

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0006

Chapter 6

Psychological Aspects of Stock Price Drifts Following Analyst Recommendation Revisions

Andrey Kudryavtsev∗

Abstract In this chapter, I sketch the picture emerging from two studies of mine dealing with psychological aspects of immediate and longer-term stock price reactions to analyst recommendation revisions. These studies demonstrate that: (i) if on the day of a recommendation upgrade (downgrade), the respective stock’s abnormal return is negative (positive), then the stock’s cumulative abnormal returns for the period of up to half a year following the recommendation upgrade (downgrade) tend to be higher (lower), and (ii) cumulative abnormal stock returns during half a year following analyst recommendation upgrades (downgrades) tend to be higher (lower) if the latter are published before holidays. I attribute these findings to the effects of investor inattention to news and mood, respectively. Keywords: Analyst recommendation revisions, behavioral finance, holiday effect, investor inattention, mood maintenance hypothesis, stock price drifts

∗

Yezreel Valley Academic College, Israel; [email protected].

131

page 131

November 4, 2020

132

1.

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

Introduction

The role of information in the modern world cannot be overestimated. Financial markets represent one of the fields where the crucial importance of information is obvious. Possessing relevant information makes the difference between gains and losses, and therefore, market investors put considerable efforts looking for it and trying to correctly interpret it. This chapter deals with stock price dynamics after analyst recommendation revisions, which represent an important means of transmitting company-specific information to market investors. There is an extensive literature examining the immediate and longer-term effects of analyst recommendation revisions on stock prices. The general conclusions arising from this literature are that analyst recommendation upgrades tend to be surrounded by abnormally high stock returns, while the downgrades tend to be accompanied by abnormally low stock returns (e.g., Francis and Soffer, 1997; Jegadeesh et al., 2004; Jegadeesh and Kim, 2010). Moreover, initial recommendation revisions are documented to be followed by systematic price drifts lasting up to one month for recommendation upgrades and up to six months for recommendation downgrades (e.g., Womack, 1996; Brav and Lehavy, 2003; Gleason and Lee, 2003). The dominant explanation for the existence of the post-recommendation price drifts is based on investor inattention to company-specific information resulting in underreaction to news (e.g., Peng and Xiong, 2006; Hirshleifer et al., 2011). In my studies, described in this chapter, I focus on two psychological aspects that may potentially affect the magnitude of the post-recommendation price drifts. First, I hypothesize that if on the day when a recommendation revision with respect to a stock has been issued, the sign of the stocks abnormal market-adjusted return was opposite to the direction of the revision, then it means that investors have probably been, for some reasons, especially inattentive to the revision, underreacting to it. Therefore, since recommendation revisions in general are shown to contain important investment information, I expect that following this specific revision, the postrecommendation stock price drift in the direction of the revision should be more pronounced. Second, I suggest that if a recommendation revision is issued on a trading day before a holiday, then,

page 132

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Psychological Aspects of Stock Price Drifts

page 133

133

in order not to undermine their positive pre-holiday mood, investors, or at least a part of them, may tend to “postpone influential trading decisions until the holidays are over”, and thus, to underreact to the recommendation revision, leading to stronger post-recommendation price drift. Employing an extensive database of stock recommendations revisions, I find supportive evidence for both hypotheses. The rest of the chapter is structured as follows. In Section 2, I review the literature on analyst recommendations and recommendation revisions. In Section 3, I present the literature dealing with the holiday effect and its financial implications. In Sections 4 and 5, I sketch and analyze some of the main findings of my studies dealing with the inattention effect and holiday effect on postrecommendation stock returns. In Section 6, I discuss the results and their potential implications for further research.

2.

Analyst Recommendations and Recommendation Revisions

Financial analysts have been argued to serve as an important information intermediary in capital markets and key players in shaping corporate information environment (e.g., Lang and Lundholm, 1996; Healy and Palepu, 2001; Beyer et al., 2010). They are expected to improve efficiency of the market by providing previously unknown information to the market through their recommendations (e.g., Grossman, 1995; Frankel et al., 2006). These recommendations represent an analyst’s expert opinion about a specific stock. Respectively, a recommendation revision can be defined as the difference between an analysts’ current recommendation and her previous one regarding the same stock (Boni and Womack, 2006). Recommendation revisions represent the focus of a large body of literature in finance and are documented to be more informative than the recommendation levels regarding the subsequent stock price reaction (e.g., Francis and Soffer, 1997; Jegadeesh et al., 2004; Jegadeesh and Kim, 2010). The overall conclusion of the literature dealing with analyst recommendation revisions seems to be that they contain useful investment information for investors. Stickel (1995) concludes that brokerage house recommendation changes influence stock prices.

November 4, 2020

134

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

Short-term price reaction is found to be a function of the strength of the recommendation; the size of the recommended firm; the contemporaneous earnings forecast revisions; the magnitude of the change in recommendation; the reputation of the analyst; and the size of the brokerage house. The first three factors are associated with price changes that appear to be permanent, representing information effects, while the last three factors convey temporary, price pressure effects. Womack (1996) analyzes new (revisited) buy and sell recommendations of stocks by security analysts at major US brokerage firms and documents significant, systematic discrepancies between pre-recommendation prices and eventual values. The initial return at the time of the recommendation is economically and statistically significant, especially for the recommendation downgrades, even though few recommendations coincide with new public news or provide previously unavailable facts. Green (2006) finds evidence that early access to stock recommendations provides brokerage firm clients with incremental investment value. After controlling for transaction costs, purchasing (selling short) following upgrades (downgrades) results in positive and significant two-day returns. Short-term profit opportunities persist for two hours following the pre-market release of recommendation changes. Broad literature investigates the sources of differential stock price reactions to analyst recommendations and recommendation revisions. Mikhail et al. (2004) investigate whether security analysts exhibit persistence in their stock picking ability and find that analysts whose recommendation revisions earned the most (least) excess returns in the past continue to outperform (underperform) in the future. Loh and Mian (2006) report that analysts who possess more accurate earnings forecasts at the time of the recommendation issue more profitable stock recommendations. Sorescu and Subrahmanyam (2006) document that low strength recommendation changes by analysts from reputable brokerages are associated with more return persistence. Similarly, Loh and Stulz (2011) demonstrate that a recommendation is more likely to generate a sizable stock reaction if it comes from a leader analyst. Michaely and Womack (2006) and Kecskes et al. (2010) show that stock recommendations with concurrent same-direction earnings forecast revisions lead to higher stock price reactions and are more profitable. Jegadeesh and Kim (2010) argue that recommendations that move away from consensus have stronger effects on stock prices. Li et al. (2015) find that analyst

page 134

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

Psychological Aspects of Stock Price Drifts

b3964-ch06

page 135

135

recommendations play an important role in generating the momentum effect. Previous researchers document an incomplete reaction to analyst recommendations resulting in a predictable price drift (e.g., Elton et al., 1986; Brav and Lehavy, 2003; Gleason and Lee, 2003). This price drift is stated to last up to one month for recommendation upgrades and up to six months for recommendation downgrades (e.g., Womack, 1996; Barber et al., 2001). Even though an immediate price reaction to stock recommendations is in line with the notion of efficient capital markets, a predictable post-recommendation price drift is confronting the prevailing theory of semi-strong form of market efficiency by Malkiel and Fama (1970), which states that investors should not be able to profit from the publicly available information, including analyst recommendations. The magnitude of the post-recommendation price drifts may differ for different types of recommendations and different groups of stocks. Womack (1996) finds that the drifts for the sell recommendations are larger and more longlived than those for the buy recommendations. Barber et al. (2001) state that the price drifts are more significant for smaller companies. Stickel (1995) detects that recommendation revisions by larger brokerage houses generate larger subsequent price drifts. Predictable post-recommendation drifts beg the question as to why the information is not fully incorporated in the stock price when the recommendation is released. One could potentially appeal to short-sale constraints (e.g., Diether et al., 2002; Nagel, 2005) as an explanation of why a negative drift follows downgrades, but it is not so obvious why one should observe an underreaction to upgrades. Barber et al. (2001) offer the explanation that markets are semistrong inefficient so that stock returns are predictable based on public information, like stock recommendations. Yet the dominating explanation for the existence of the postrecommendation price drifts is based on investors’ inattention leading to underreaction to news. Theoretical models by Peng and Xiong (2006) and Hirshleifer et al. (2011) suggest that investors’ attention constraints lead to “category learning,” so that investors focus more on market-wide and industry-wide information rather than on firm-specific information. This implies that investors could underreact to firm-specific information such as analysts’ firm-specific stock recommendations. In line with these models, Loh (2010) empirically

November 4, 2020

136

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

demonstrates that investors tend to underreact to news about firms that are not attention grabbing, as proxied by their low prior stock turnover. Respectively, if investors temporarily neglect the information contained in stock recommendations, then a predictable drift follows when they gradually incorporate this information. Gavriilidis et al. (2016) continue Loh’s (2010) line of reasoning, but concentrate on attention grabbing recommendations, proxied by abnormally high event-day trading volumes, rather than on attention grabbing firms. They suggest that recommendations that are accompanied by high attention generate consistently more pronounced post-announcement drifts than otherwise similar announcements, and further show that this effect mainly stems from upgrades rather than downgrades.

3.

Holiday Effect: Psychological Background and Financial Implications

The holiday, or pre-holiday effect, refers to the documented fact that stock returns typically exhibit consistent patterns around holidays, showing systematically high returns on days prior to major holidays. The effect has been initially examined in the context of the US stock market. In their seminal study, Lakonishok and Smidt (1988), analyzing a 90 year dataset, observe that the average pre-holiday rate of return equals 0.22%, compared with a regular daily rate of return of less than 0.01%. This means that pre-holiday returns are about times larger than returns on normal days, with about 63.9% of all pre-holiday returns being positive. Likewise, Ariel (1990) detects that over the period 1963–1982, the average pre-holiday returns in the US are 10 times higher than returns over the remaining days of the year. Parametric and non-parametric tests indicate that these differences are statistically significant. Similarly, Pettengill (1989) finds that returns on days immediately preceding holidays are unusually high, especially for small firms. Kim and Park (1994) likewise document the holiday effect using market indicators from all the major US stock exchanges. Brockman (1995), Brockman and Michayluk (1997) and Brockman and Michayluk (1998) demonstrate the resilience of the holiday effect, reporting about its persistence across market types (auction vs. dealer) and size portfolios. Hirshleifer et al. (2016) show

page 136

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

Psychological Aspects of Stock Price Drifts

b3964-ch06

page 137

137

that there is a pre-holiday cross-sectional seasonality at the level of individual stocks, wherein stocks that historically have earned higher pre-holiday returns on average earn higher pre-holiday returns for the same holiday over the next 10 years. The holiday effect outside the US has been also thoroughly analyzed. It has been documented in different countries, precluding the possibility that it reflects the idiosyncratic market characteristics of any specific exchange. Cadsby and Ratner (1992) consider Canada, Japan, Hong Kong and Australia from 1962 to 1989 and test for local holidays, US holidays and joint (local-US) holidays using market indices from each country. The results point out that there exist significant holiday effects in all of the sample markets, with the highest returns appearing on days just prior to joint holidays. Barone (1990) documents a strong holiday effect in the Italian stock market, with an average return of 0.27% vs. an average non-holiday return of −0.01%. In a broader study, Agrawal and Tandon (1994) examine the holiday effect in seventeen national markets, and find significant pre-holiday strength in 65% of them. Marrett and Worthington (2009) detect the holiday effect for Australian stock market, the magnitude of the former being higher in the retail industry. Dodd and Gakhovich (2011) demonstrate that the holiday effect is present in emerging Central and East European markets, being more pronounced in the earlier years of financial market operations. The magnitude and statistical significance of pre-holiday returns may vary on specific holidays. Returns prior to religious holidays tend to be higher than returns before other holidays. Chan et al. (1996) report significant holiday effects before cultural holidays in Asia. Namely, they demonstrate that in India there is a holiday effect before Hindu holidays; in Malaysia significant returns precede Islamic New Year and Vesak; Singapore market exhibits abnormal returns before Chinese New Year; and in Thailand small companies have significant abnormal returns before Chinese New Year. In New Zealand, the most significant positive returns are registered before the Easter holidays (Cao et al., 2009). Bley and Saad (2010) find significantly positive returns before the Middle Eastern religious holidays in the Middle East. The previous studies suggest a number of potential explanations for the existence of the holiday effect. The first one is the potential relationship between this effect and other calendar anomalies, such

November 4, 2020

138

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

as the monthly effect, the day-of-the–week effect and the turn-ofthe-year effect (e.g., Lakonishok and Smidt, 1988; Liano et al., 1992). These studies indicate that the high returns observed on pre-holidays are not a manifestation of other calendar anomalies. Another explanation emphasizes the existence of a link between the holiday effect and the small firm effect, since the former is more pronounced for small firms (e.g., Pettengill, 1989; Keef and Roush, 2005; Marrett and Worthington, 2009). Yet another explanation of the holiday effect is based on a set of different and systematic trading patterns. Keim (1989) argues that the pre-holiday return may be, at least partially, explained by movements from the bid to the ask price. Ariel (1990) states that pre-holiday strength can be attributed to short-sellers who desire to close short but not long positions in advance of holidays or, simply, to some clientele which preferentially buys (or avoids selling) on pre-holidays. Yet, arguably, the leading and the most promising explanation for positive abnormal returns prior to public holidays lies in investor psychology (e.g., Brockman and Michayluk, 1998; Vergin and McGinnis, 1999). This explanation stems from two psychology-based facts: first, that anticipation of holidays is associated with rising investors’ mood (e.g., Frieder and Subrahmanyam, 2004; Bergsma and Jiang, 2015), and second, that people in good mood tend to believe in more positive outcomes (e.g., Kavanagh and Bower, 1985; Thaler, 1999). Following this line of reasoning, this cohort of studies suggests that investors tend to buy stocks before holidays because of “holiday euphoria” and “high spirits”, which cause them to expect positive returns in the future. An additional, less known and much less reported aspect of the holiday effect refers to the stock trading volumes before holidays. Meneu and Pardo (2004) report lower than usual abnormal trading volumes and higher than usual bid-ask spreads before public holidays, indicating that on these days, stocks tend to be less liquid. Similarly, Cao et al. (2009) document that the daily de-trended trading volumes on pre-holiday trading days are generally lower than on “regular” trading days, and subsequently conclude that investors may not be able to capture abnormal returns prior to holidays due to the low trading volume. Dodd and Gakhovich (2011) find similar results for Central and Eastern European markets.

page 138

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Psychological Aspects of Stock Price Drifts

page 139

139

Potential explanation for lower pre-holiday trading activity also emanates from investors’ psychology and is based on the Mood Maintenance Hypothesis (MMH) (Isen, 1984, 2000), which is a documented psychological pattern suggesting that people are highly motivated to maintain their positive mood states. Psychological literature indicates that people tend to be concerned with the fact that detailed information processing might undermine pleasant mood states, and therefore, in line with the MMH, positive mood may be associated with less critical thinking and reduced information processing (Mackie and Worth, 1989; Kuykendall and Keating, 1990; Erber and Tesser, 1992; Schwarz, 2001). In the context of the holiday effect, this means that before holidays, investors, who strive to maintain their positive mood, may be less willing to make trading decisions, which are associated with information processing, and therefore, trade less.

4.

4.1.

Effect of Investor Inattention on Stock Price Reactions to Recommendation Revisions Research Hypothesis and Data Description

As shown in Section 2, financial literature concludes that on the one hand, recommendation revisions contain important company-specific information, but on the other hand, investor inattention leads to underreaction to recommendation revisions and to systematic postrecommendation price drifts. The previous researchers employ different proxies for investor inattention and demonstrate how the latter are connected to the immediate price reactions and to the subsequent stock returns. In this study, rather than looking for external factors that may serve as proxies for inattention, I employ the actual event-day stock price reactions to recommendation revisions as a proxy for how inattentive the investors have been. I hypothesize that if on the day when a recommendation revision with respect to a stock has been issued, the sign of the stock’s abnormal return was opposite to the direction of the revision, then it may be assumed that investors have been, for some reasons, especially inattentive to the revision, and therefore, the post-recommendation stock price drift in the direction of the

November 4, 2020

140

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

revision should be more pronounced. In other words, I expect that if for some reasons, investors do not pay attention to a recommendation revision and do not incorporate the new important companyspecific information in the respective stock’s price, then they “will have to” fully incorporate the information in the price during the subsequent period, creating a relatively stronger price drift. I formulate the testable hypothesis as follows: Hypothesis 1: If on the day of a recommendation upgrade (downgrade), the respective stock’s abnormal return is negative (positive), then the stock’s cumulative abnormal returns following the recommendation upgrade (downgrade) should be higher (lower). I collect the sample of stock recommendations from the Thomson Financials I/B/E/S database for the period from 2003 to 2017. I/B/E/S stock recommendations are coded in integers from 1 (for Strong Buy) to 5 (for Strong Sell). I focus on recommendation revisions, that is, on the differences between the current and the most recent recommendation levels, since prior research confirms that recommendations changes are more informative than mere levels (e.g., Boni and Womack, 2006; Jegadeesh and Kim, 2010). I define the day of a recommendation revision as the event day (Day 0), except when a revision falls on a non-trading day. In the latter case, the event day is defined as the trading day following the day the recommendation was updated. Similarly to Li et al. (2016), I exclude from the sample recommendation initiations (first-time recommendations of an analyst on a stock) and re-initiations (new recommendations issued by an analyst on a stock after more than a year from her previous recommendation on the same stock). Furthermore, following Loh (2010) and in order to be sure that a stock prices reaction to a recommendation revision was not partially driven by a contemporaneous earnings announcement of the same company, I remove from the sample recommendation revisions that had been issued in the three-day window centered around the I/B/E/S quarterly earnings announcement dates. Finally, I drop the stocks with share prices below US$1.00. These filtering rules yield the working sample of 77,894 recommendation upgrades and 87,342 recommendation downgrades. I merge the I/B/E/S recommendations data with daily stock price data for all NYSE, AMEX and NASDAQ common stocks

page 140

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Psychological Aspects of Stock Price Drifts

page 141

141

from the Center for Research in Security Prices (CRSP).1 In addition, for each recommendation revision, I match the respective company’s market capitalization, as recorded on a quarterly basis at http://ycharts.com/, for the closest preceding announcement date. 4.2.

Event-day Inattention Effect on Stock Price Drifts Following Analyst Recommendation Revisions

I calculate daily abnormal stock returns (ARs) using Market Model Adjusted Returns (MMAR)2 For estimating the postrecommendation stock price dynamics, I employ ARs for Day 1 and cumulative ARs (CARs) for Days 2–21, Days 2–63 and Days 2–126, roughly corresponding to one month, three months and six months after the revision, respectively.3 Furthermore, in order to test Hypothesis 1, I divide both recommendation upgrades and downgrades into two subsamples: (i) recommendation revisions accompanied by Day-0 ARs (AR0) of the opposite sign to the direction of the revision, and (ii) recommendation revisions accompanied by AR0, which either equals zero or has the same sign as the direction of the revision Table 1 comprises of both subsamples, average ARs and CARs over the specified periods following recommendation revisions, as well as the respective AR/CAR differences between the subsamples and their statistical significance. The results corroborate the existence of the event-day inattention effect on stock price drifts following recommendation revisions, indicating the following: • Both recommendation upgrades and downgrades accompanied by event-day ARs of the opposite sign are followed by price 1

The two data sets are merged based on either CUSIP or exchange tickers combined with the requirement that the period these identifiers are used in the data sets overlap. 2 Alternatively, I calculate ARs using Market Adjusted Returns (MAR) — return differences from the market index, and the Fama–French three-factor model. The results (available upon request from the author) remain qualitatively similar to those reported in Section 4. 3 I choose to analyze post-recommendation periods of one, three and six months following, for example, Loh (2010) and Gavriilidis et al. (2016).

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

142

Table 1: Abnormal stock returns following recommendation revisions accompanied by event-day ARs in the expected and the opposite direction

Days

Average AR/CAR Following Recommendation Revisions, % (2-Tailed p-Values)

Relative

Upgrades

to Event 1 2–21 2–63 2–126

AR0 < 0 ∗∗ 0.18

AR0 ≥ 0

Downgrades Difference

AR0 > 0

AR0 ≤ 0

Difference

∗∗ 0.17

∗∗∗ −0.32

∗ −0.07

∗∗ −0.25

(4.56%)

0.01 (54.12%)

(3.98%)

(0.87%)

(7.45%)

(1.42%)

∗∗∗ 0.71

∗∗∗ 0.19

∗∗∗ 0.52

∗∗∗ −0.90

∗∗∗ −0.27

∗∗∗ −0.63

(0.03%)

(0.54%)

(0.12%)

(0.00%)

(0.38%)

(0.02%)

∗∗∗ 0.92

∗∗∗ 0.14

∗∗∗ 0.78

∗∗∗ −1.29

∗∗∗ −0.41

∗∗∗ −0.88

(0.00%)

(0.88%)

(0.00%)

(0.00%)

(0.02%)

(0.00%)

∗∗∗ 1.10

∗∗ 0.09

∗∗∗ 1.01

∗∗∗ −1.66

∗∗∗ −0.58

∗∗∗ −1.08

(0.00%)

(1.89%)

(0.00%)

(0.00%)

(0.00%)

(0.00%)

Note: Asterisks denote 2-tailed p-values: ∗ p < 0.10;

∗∗

p < 0.05;

∗∗∗

p < 0.01.

drifts in the direction of the revision, whose magnitude continuously increases as the post-event window is expanded. For Days 2–126, average CARs following upgrades (downgrades) reach 1.10% (−1.66%). • In line with Hypothesis 1, stock price drifts following recommendation revisions accompanied by the opposite-sign event-day ARs are significantly more pronounced than those following recommendation revisions accompanied by the same-sign event-day ARs. Average CAR differences between the two subsamples of recommendation revisions increase for longer post-event windows, and for Days 2–126, reach 1.01% (−1.08%) for upgrades (downgrades). Furthermore, I analyze the magnitude of the postrecommendation stock price reversals and of AR differences between the two AR0 conditions, by firm size (market capitalization) and by historical volatility of stock returns. The motivation for this analysis is based on the findings by Baker and Wurgler (2006), who argue that stocks of low capitalization, growth stocks, and highly volatile stocks are especially likely to be disproportionately sensitive to broad waves of investor sentiment. I split the samples of events into three roughly equal parts by the firms’ market capitalization reported for

page 142

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Psychological Aspects of Stock Price Drifts

page 143

143

the end of the quarter preceding each recommendation revision and by the standard deviation of stock returns over Days −251 to −1.4 In line with Baker and Wurgler (2006), the event-day inattention effect on the price drifts following both recommendation upgrades and downgrades is stronger for: (a) Low capitalization stocks: For example, for the post-event window 2–126, average CAR differences between the subsamples of upgrades (downgrades) accompanied by the opposite-sign and the same-sign event-day ARs equal 1.34% (−1.31%) for low capitalization stocks, compared to 0.82% (−0.87%) for high capitalization stocks. (b) More volatile stocks: For example, for the post-event window 2–126, average CAR differences between the subsamples of upgrades (downgrades) accompanied by the opposite-sign and the same-sign event-day ARs equal 1.27% (−1.25%) for high volatility stocks, compared to 0.86% (−0.91%) for low volatility stocks.5 Finally, I check the persistence of the event-day inattention effect on post-event price drifts, controlling for additional company-specific and event-specific factors. To do so, separately for recommendation upgrades and downgrades, I run the following regressions for postevent days 1, 2–21, 2–63 and 2–126: ARit /CARit = β0 + β1 OPPOSITEi + β2 MCapi + β3 Betai + β4 SRVolati + β5 Magnitudei + β6 Experiencei + εit

(1)

Where ARit /CARit is the abnormal/cumulative abnormal stock return following event i for the event or post-event window t (Days 0, 1, 2–21, 2–63 or 2–126); OPPOSITEi is the dummy variable, taking the value 1 if the sign of AR0i is opposite to the direction of 4

The sample partition approach by both market capitalization and historical stock volatility is similar to the one employed by Kliger and Kudryavtsev (2010). 5 Detailed results with respect to the event-day inattention effect for different stock groups classified by market capitalization and historical volatility are available upon request from the author.

November 4, 2020

144

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

the recommendation revision, and 0 otherwise; MCapi is the natural logarithm of the firm’s market capitalization corresponding to event i, normalized in the cross-section; Betai is the estimated Market Model beta for event i, calculated over the Days −251 to −1 and normalized in the cross-section; SRVolati is the standard deviation of stock returns over the Days −251 to −1 corresponding to event i, normalized in the cross-section; Magnitude i is the number of categories changed in the revision; and Experience i is the natural logarithm of number of years that the analyst providing recommendation revision i exists in I/B/E/S prior to the revision, normalized in the cross-section. Table 2 depicts the regression coefficients and their significance. The most notable result refers to the regression coefficients on OPPOSITE, which are significantly positive (negative) for recommendation upgrades (downgrades), indicating that the postrecommendation price drifts are significantly more pronounced for recommendation revisions accompanied by the opposite-sign eventday ARs, and suggesting once again that if investors do not immediately react to relevant company-specific information, then the respective reaction follows later. Importantly, the event-day inattention effect on post-recommendation price drifts remains significant even after controlling for additional factors affecting post-event ARs, and its magnitude increases for longer post-recommendation time windows.

5.

5.1.

Holiday Effect on Stock Price Reactions to Analyst Recommendation Revisions Research Hypothesis and Data Description

In line with the literature presented in Section 3, which documents less intense trading activity before holidays, I hypothesize that if a recommendation revision is issued on a trading day before a holiday, then, in order not to undermine their positive pre-holiday mood, investors, or at least a part of them, may be less willing to process significant company-specific information and make influential trading decisions, and therefore, may react relatively more weakly to the recommendation revision. In other words, I expect that investors

page 144

November 4, 2020

Table 2: Multifactor regression analysis of stock returns following analyst recommendation revisions: Dependent variables — Stock ARs/CARs for different post-event windows

10:12

Coefficient Estimates, % (2-Tailed p-Values)

Intercept OPPOSITE MCap Beta SRVolat

AR1

0.02 (17.46%) ∗∗

0.18 (1.45%) ∗∗

−0.10 (3.91%)

CAR (2, 21)

CAR (2, 63)

CAR (2, 126)

AR1

∗∗∗ 0.25 (0.01%)

∗∗∗ 0.12 (0.35%)

∗∗∗ 0.11 (0.30%)

∗ −0.10 (1.76%)

∗∗∗

∗∗

∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗

∗∗∗

∗∗∗

∗∗∗

0.53 (0.00%) ∗∗

−0.18 (1.59%)

0.77 (0.00%) ∗∗

−0.24 (1.45%)

1.00 (0.00%) ∗∗∗

−0.27 (0.85%) ∗

−0.26 (1.27%) ∗

−0.05 (15.72%)

−0.07 (11.03%)

∗ 0.10 (7.12%)

∗∗ 0.13 (3.94%)

∗∗ 0.14 (3.25%)

∗∗ 0.17 (1.96%)

∗ −0.09 (6.34%)

0.03 (28.62%)

∗

0.07 (5.18%)

∗

0.09 (5.34%)

Note: Asterisks denote 2-tailed p-values: ∗ p < 0.10;

∗∗

−0.15 (3.88%)

∗∗

−0.18 (2.80%)

∗∗

−0.21 (2.05%)

∗∗

−0.02 (37.48%)

−0.05 (13.56%)

∗

−0.08 (7.40%)

∗∗

∗

−0.02 (45.16%)

−0.02 (29.94%)

∗

∗

0.11 (4.75%) 0.08 (5.33%) ∗∗

p < 0.05;

∗∗∗

−0.06 (7.69%)

−0.12 (4.68%)

b3964-ch06

Experience

0.09 (6.23%)

∗∗∗

−0.03 (21.49%)

−0.01 (66.32%)

∗

∗∗

−1.09 (0.00%) 0.33 (0.86%)

0.09 (8.91%)

0.09 (6.89%)

∗∗

−0.87 (0.00%)

−0.63 (0.00%)

0.27 (2.24%)

0.07 (13.67%)

∗

−0.64 (0.00%)

−0.50 (0.00%)

CAR (2, 126)

0.21 (3.17%)

0.06 (14.50%)

0.01 (45.53%)

−0.37 (0.03%)

CAR (2, 63)

0.08 (6.15%)

0.04 (18.72%)

Magnitude

CAR (2, 21)

Behavioral Finance: A Novel Approach – 9in x 6in

Explanatory Variables

Recommendation Downgrades

Psychological Aspects of Stock Price Drifts

Recommendation Upgrades

−0.08 (6.71%)

p < 0.01. 145 page 145

November 4, 2020

146

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

may tend to “postpone important their decisions until the holidays are over,” and thus, to underreact to recommendation revisions issued before holidays, leading to stronger post-recommendation (and post-holiday) price drifts. Respectively, I formulate the following hypothesis: Hypothesis 2: Cumulative abnormal stock returns following analyst recommendation upgrades (downgrades) should be higher (lower) if the latter are published before holidays. I employ the same working sample, and distinguish between preholiday and regular (published on other days of the year) recommendation revisions. US holidays examined include President’s Day, Martin Luther King Jr. Day, Good Friday, Memorial Day, Independence Day, Labor Day, Thanksgiving Day, Christmas and New Year’s Day.

5.2.

Holiday Effect on Stock Price Drifts Following Analyst Recommendation Revisions

Table 3 comprises average ARs and CARs for the same postrecommendation periods as in the previous section, following pre-holiday and regular recommendation revisions, as well as the respective AR/CAR differences and their statistical significance. The results corroborate the existence of the holiday effect on stock price reactions to recommendation revisions, indicating that stock price drifts following pre-holiday recommendation revisions are significantly more pronounced. The magnitude of these drifts continuously increases during the six-month post-recommendation period, reaching average CARs of 1.45% (−2.03%) for Days 2–126 following preholiday recommendation upgrades (downgrades), compared to 0.14% (−0.72%) following regular upgrades (downgrades) over the same post-recommendation period.6

6 It should be noted that pre-holiday recommendation revisions make up slightly less than 3% of the study’s working sample. Still, 2,168 (2,416) pre-holiday recommendation upgrades (downgrades) accumulated over the 15-year sampling period are sufficient for obtaining statistically significant results.

page 146

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Psychological Aspects of Stock Price Drifts

page 147

147

Table 3: Abnormal stock returns following pre-holiday and regular recommendation revisions Average AR/CAR Following Recommendation Revisions, % (2-Tailed p-Values) Days Recommendation Upgrades Recommendation Downgrades Relative to Event Pre-holiday Regular Difference Pre-holiday Regular Difference ∗∗

∗∗

∗∗∗

∗∗

∗∗

∗∗∗ 0.58 (0.11%)

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗∗

1

0.24 (3.65%)

0.02 (28.71%)

0.22 (3.87%)

2–21

***0.86 (0.02%)

∗∗∗ 0.28 (0.21%)

∗∗∗

∗∗∗

2–63 2–126

1.16 (0.00%) ∗∗∗

1.45 (0.00%)

0.16 (0.57%) ∗∗

0.14 (1.13%)

1.00 (0.00%) 1.31 (0.00%)

Note: Asterisks denote 2-tailed p-values:

∗∗

−0.36 (0.64%) −1.13 (0.00%) −1.75 (0.00%) −2.03 (0.00%)

p < 0.05;

∗∗∗

−0.12 (2.18%) −0.42 (0.23%) −0.58 (0.00%) −0.72 (0.00%)

−0.24 (1.54%) −0.71 (0.00%) −1.17 (0.00%) −1.31 (0.00%)

p < 0.01.

Furthermore, I analyze the dynamics of the post-event stock price reversals and of AR differences between pre-holiday and regular recommendation revisions, by firm size (market capitalization) and by historical volatility of stock returns. Similarly to Section 4, I document that the holiday effect on price drifts following both recommendation upgrades and downgrades is stronger for: (a) Low capitalization stocks: For example, for the post-event window 2–126, average CAR differences between pre-holiday and regular upgrades (downgrades) equal 1.84% (−1.98%) for low capitalization stocks, compared to 0.75% (−0.73%) for high capitalization stocks. (b) More volatile stocks: For example, for the post-event window 2–126, average CAR differences between pre-holiday and regular upgrades (downgrades) equal 1.60% (−1.72%) for high volatility stocks, compared to 0.97% (−0.88%) for low volatility stocks.7

7

Detailed results with respect to the holiday effect for different stock groups classified by market capitalization and historical volatility are available upon request from the author.

10:12

Coefficient Estimates, % (2-Tailed p-Values) Recommendation Upgrades

HOLIDAY MCap Beta SRVolat

0.01 (25.64%) ∗∗

0.23 (1.87%) ∗∗

−0.11 (3.84%)

CAR (2, 21)

CAR (2, 63)

CAR (2, 126)

AR1

∗∗∗ 0.26 (0.04%)

∗∗∗ 0.14 (0.41%)

∗∗∗ 0.13 (0.45%)

∗ −0.12 (1.65%)

∗∗∗

∗∗

∗∗

∗∗∗

∗∗∗

∗∗∗

∗∗

∗∗∗

∗∗∗

∗∗∗

0.59 (0.00%) ∗∗

−0.18 (2.31%)

1.02 (0.00%) ∗∗

−0.23 (1.97%)

1.30 (0.00%) ∗∗

−0.26 (1.24%) ∗

−0.24 (1.74%)

−0.06 (12.40%)

−0.07 (10.67%)

∗∗ 0.12 (4.21%)

∗∗ 0.14 (3.52%)

∗∗ 0.19 (1.85%)

∗ −0.08 (6.82%)

0.02 (34.91%)

∗

0.06 (6.28%)

∗

0.08 (5.17%)

Note: Asterisks denote 2-tailed p-values: ∗ p < 0.10;

∗∗

−0.16 (3.77%)

∗∗

−0.20 (2.61%)

∗∗

∗

−0.23 (1.94%)

∗∗

−0.02 (41.02%)

−0.05 (11.46%)

−0.09 (6.82%)

∗∗

∗

0.07 (5.48%)

−0.01 (65.34%)

−0.03 (31.07%)

−0.06 (8.11%)

∗

∗∗

∗∗∗

0.12 (4.67%)

p < 0.05;

−0.13 (4.24%) −0.09 (6.43%)

b3964-ch06

Experience

∗∗

−0.04 (16.75%)

∗ 0.09 (8.19%)

0.10 (5.64%)

∗∗

−1.33 (0.00%) 0.31 (1.45%)

−0.01 (75.68%)

∗

−1.18 (0.00%)

−0.70 (0.00%)

0.26 (2.76%)

0.10 (8.56%)

0.08 (7.48%)

∗∗

−0.56 (0.00%)

CAR (2, 126)

0.20 (3.88%)

0.06 (14.27%)

∗

−0.72 (0.00%)

CAR (2, 63)

0.07 (7.27%)

0.05 (16.75%)

0.01 (47.61%)

−0.40 (0.05%)

∗

0.03 (21.02%)

Magnitude

CAR (2, 21)

Behavioral Finance: A Novel Approach – 9in x 6in

Intercept

AR1

Recommendation Downgrades Behavioral Finance: A Novel Approach

Explanatory Variables

November 4, 2020

148

Table 4: Multifactor regression analysis of stock returns following analyst recommendation revisions: Dependent variables — Stock ARs/CARs for different post-event windows

p < 0.01.

page 148

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Psychological Aspects of Stock Price Drifts

page 149

149

Finally, and similarly to the previous section, I check the persistence of the holiday effect on stock reactions to analyst recommendation revisions, controlling for additional company-specific and event-specific factors. To do so, separately for recommendation upgrades and downgrades, I run the following regressions for the event day and for post-event windows 1, 2–21, 2–63 and 2–126: ARit /CARit = β0 + β1 HOLIDAYi + β2 MCapi + β3 Betai +β4 SRVolati + β5 Magnitudei + β6 Experiencei + εit (2) where HOLIDAYi is the dummy variable, taking the value 1 if the event i takes place immediately before a public holiday, and 0 otherwise. Table 4 reports the regression coefficients and their significance. The results provide additional support for Hypothesis 2. The main result is that the regression coefficients on OPPOSITE are significantly positive (negative) for recommendation upgrades (downgrades), implying that post-recommendation price drifts are significantly more pronounced for pre-holiday recommendation revisions. Once again, the magnitude of the effect persists after accounting for additional relevant variables, and its magnitude increases as the post-recommendation period is expanded. 6.

Discussion and Potential Directions for Further Research

In this chapter, I have briefly described the results of two of my studies focusing on psychological effects on stock price reactions to analyst recommendation revisions and subsequent price drifts. From the findings of the two studies arises that initial stock price reactions to recommendation revisions may be insufficient and deviate from full rationality, creating a premise for systematic price drifts in the cases when, for some reasons, the sign of the event-day abnormal stock return is opposite to the direction of the revision or when the revision is issued immediately before a public holiday. These findings may have a number of important practical implications, especially in what concerns the possibility for predicting stock returns following recommendation revisions.

November 4, 2020

10:12

150

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

In the last decades, a vast number of psychological effects on stock market investors’ behavior have been documented and became an integral part of the financial literature. Yet, human psychology is so complex and multi-dimensional that there is probably still enough space for the new generations of researchers in the field of behavioral finance. The issue of immediate and longer-term stock price reactions to analyst recommendation revisions is only one of the numerous issues that still need a lot of work to be done. Following the findings presented in this chapter, I would suggest a few potential directions for further research concentrating on this topic: • To analyze combined effects of investor inattention and pre-holiday mood: It might be interesting to check if post-recommendation price drifts following pre-holiday recommendation revisions are especially pronounced in the cases when investors are especially inattentive to the revisions. • To repeat the analysis separately for different groups of investors: If the data for individual investors are available, it would be useful to compare the behavior of different categories of investors from the point of view of their reactions to recommendation revisions in the light of the inattention and the holiday effects. • To introduce additional psychological factors: Stock price reactions to recommendation revisions may be affected by various psychological factors, so that including at least some of them in the empirical analysis may provide some additional insights. • To perform similar analysis during periods of financial crises: It might be important to check if during periods of major financial distress, there are systematic changes in stock price reactions to recommendation revisions, in general, and in the magnitude of the inattention and the holiday effects on the post-recommendation returns, in particular. • To perform similar analysis employing intraday stock data: It might be interesting to analyze the intraday dynamics of both psychological effects. References Agrawal, A. and Tandon, K. (1994). Anomalies or illusions? Evidence from stock markets in eighteen countries. Journal of International Money and Finance, 13 (1), 83–106.

page 150

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

Psychological Aspects of Stock Price Drifts

b3964-ch06

page 151

151

Ariel, R. A. (1990). High stock returns before holidays: Existence and evidence on possible causes. Journal of Finance, 45 (5), 1611–1626. Baker, M. and Wurgler, J. (2006). Investor sentiment and the cross-section of stock returns. Journal of Finance, 61 (4), 1645–1680. Barber, B., Lehavy, R., McNichols, M., and Trueman, B. (2001). Can investors profit from the prophets? Security analyst recommendations and stock returns. Journal of Finance, 56 (2), 531–563. Barone, E. (1990). The Italian stock market: Efficiency and calendar anomalies. Journal of Banking and Finance, 14 (2), 483–510. Bergsma, K. and Jiang, D. (2015). Cultural New Year holidays and stock returns around the World. Financial Management, 45 (1), 3–35. Beyer, A., Cohen, D. A., Lys, T. Z., and Walther, B. R. (2010). The financial reporting environment: Review of the recent literature. Journal of Accounting and Economics, 50 (2–3), 296–343. Bley, J. and Saad, M. (2010). Cross cultural differences in seasonality. International Review of Financial Analysis, 19 (4), 306–312. Boni, L. and Womack, K. L. (2006). Analysts, industries, and price momentum. Journal of Financial and Quantitative Analysis, 41 (1), 85–109. Brav, A. and Lehavy, R. (2003). An empirical analysis of analysts’ target prices: Short-term informativeness and long-term dynamics. Journal of Finance, 58 (5), 1933–1968. Brockman, P. (1995). A review and analysis of the holiday effect. Financial Markets, Institutions and Instruments , 4 (5), 37–58. Brockman, P. and Michayluk, D. (1997). The holiday anomaly: An investigation of firm size versus share price. Quarterly Journal of Business and Economics, 36 (3), 23–35. Brockman, P. and Michayluk, D. (1998). The persistent holiday effect: Additional evidence. Applied Economics Letters, 5 (2), 205–209. Cadsby, C. and Ratner, M. (1992). Turn of the month and pre holiday effects on stock returns: Some international evidence. Journal of Banking and Finance, 16 (3), 497–509. Cao, X., Premachandra, I. M., Bharba, G. S., and Tang, Y. P. (2009). Firm size and the pre-holiday effect in New Zealand. International Research Journal of Finance and Economics, 32, 171–187. Chan, M., Khanthavit, A., and Thomas, H. (1996). Seasonality and cultural influences on four Asian stock markets. Asia Pacific Journal of Management, 13 (2), 1–24. Diether, K. B., Malloy, C. J., and Scherbina, A. (2002). Differences of opinion and the cross-section of stock returns. Journal of Finance, 57 (5), 2113–2141. Dodd, O. and Gakhovich, A. (2011). The holiday effect in Central and Eastern European financial markets. Investment Management and Financial Innovations, 8 (4), 29–35.

November 4, 2020

152

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

Elton, E. J., Gruber, M. J., and Grossman, S. (1986). Discrete expectational data and portfolio performance. Journal of Finance, 41 (3), 699–713. Erber, R. and Tesser, A. (1992). Task effort and the regulation of mood: The absorption hypothesis. Journal of Experimental Social Psychology, 28 (4), 339–359. Francis, J. and Soffer, L. (1997). The relative informativeness of analysts’ stock recommendations and earnings forecast revisions. Journal of Accounting Research, 35 (2), 193–211. Frankel, R., Kothari, S. P., and Weber, J. (2006). Determinants of the informativeness of analyst research. Journal of Accounting and Economics, 41 (1), 29–54. Frieder, L. and Subrahmanyam, A. (2004). Nonsecular regularities in returns and volume. Financial Analysts Journal, 60 (4), 29–34. Gavriilidis, K., Herbst, P., and Kagkadis, A. (2016). Investor attention to stock recommendations. Working Paper, University of Stirling. Gleason, C. and Lee, C. (2003). Analyst forecast revision and market price discovery. The Accounting Review, 78 (1), 193–225. Green, C. T. (2006). The value of client access to analyst recommendations. Journal of Financial and Quantitative Analysis, 41 (1), 1–24. Grossman, S. (1995). Dynamic asset allocation and the informationally efficiency of markets. Journal of Finance, 50 (5), 773–778. Healy, P. M. and Palepu, K. G. (2001). Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature. Journal of Accounting and Economics, 31 (1–3), 405–440. Hirshleifer, D., Jiang, D., and Meng, Y. (2016). Mood beta and seasonalities in stock returns, Working Paper, IVEY Business School. Hirshleifer, D., Lim, S. S., and Teoh, S. H. (2011). Limited investor attention and stock market misreactions to accounting information. Review of Asset Pricing Studies, 1 (1), 35–73. Isen, A. M. (1984). Toward understanding the role of affect in cognition. In R. S. Wyer and T. K. Srull (Eds.), Handbook of Social Cognition, Hillsdale, NJ: Erlbaum, pp. 179–236. Isen, A. M. (2000). Positive affect and decision making. In M. Lewis and J. M. Havieland (Eds.), Handbook of Emotions, Vol. 2. London: Guilford, 417–435. Jegadeesh, N., Kim, J., and Krische, S. D. (2004). Analyzing the analysts: When do recommendations add value? The Journal of Finance, 59 (3), 1083–1124. Jegadeesh, N. and Kim, W. (2010). Do analysts herd? An analysis of recommendations and market reactions. Review of Financial Studies, 23 (2), 901–937.

page 152

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

Psychological Aspects of Stock Price Drifts

b3964-ch06

page 153

153

Kavanagh, D. J. and Bower, G. H. (1985). Mood and self-efficacy: Impact of joy and sadness on perceived capabilities. Cognitive Therapy and Research, 9(5), 507–525. Kecskes, A., Michaely, R., and Womack, K. L. (2010). What drives the value of analysts’ recommendations: Earnings estimates or discount rate estimates? Working Paper, Darthmouth College. Keef, S. P. and Roush, M. L. (2005). Day-of-the-week effects in the preholiday returns of the Standard & Poor’s 500 stock index. Applied Financial Economics, 15 (1), 107–119. Keim, D. B. (1989). Trading patterns, bid-ask spreads and estimated security returns: The case of common stocks at calendar turning points. Journal of Financial Economics, 25 (1), 57–67. Kim, C. W. and Park, J. (1994). Holiday effects and stock returns: Further evidence. Journal of Financial and Quantitative Analysis, 29 (1), 145–157. Kliger, D. and Kudryavtsev, A. (2010). The availability heuristic and investors’ reaction to company-specific events. The Journal of Behavioral Finance, 11 (1), 50–65. Kuykendall, D. and Keating, J. P. (1990). Mood and persuasion: Evidence for the differential impact of positive and negative states. Psychology and Marketing, 7 (1), 1–9. Lang, M. H. and Lundholm, R. J. (1996). Corporate disclosure policy and analyst behavior. Accounting Review, 71 (4), 467–492. Lakonishok, J. and Smidt, S. (1988). Are seasonal anomalies real? A ninety year perspective. Review of Financial Studies, 1 (4), 403–425. Li, F., Lin, C., and Lin, T. C. (2016). The 52-week high stock price and analyst recommendation revisions. Working Paper, University of Hong Kong. Li, K., Lockwood, J., Lockwood, L. J., and Uddin, M. R. (2015). Analyst optimism and stock price momentum, Working Paper. Liano, K., Marchand, P. H., and Huang, G. C. (1992). The holiday effect in stock returns: Evidence from the OTC Market. Review of Financial Economics, 2 (1), 45–54. Loh, R. K. (2010). Investor inattention and the underreaction to stock recommendations. Financial Management, 39 (3), 1223–1252. Loh, R. K. and Mian, G. M. (2006). Do accurate earnings forecasts facilitate superior investment recommendations? Journal of Financial Economics, 80 (2), 455–483. Loh, R. K. and Stulz, R. M. (2011). When are analyst recommendation changes influential? Review of Financial Studies, 24 (2), 593–627.

November 4, 2020

154

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch06

Behavioral Finance: A Novel Approach

Mackie, D. M. and Worth, L. T. (1989). Processing deficits and the mediation of positive affect in persuasion. Journal of Personality and Social Psychology, 57 (1), 27–40. Malkiel, B. G. and Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. Journal of Finance, 25 (2), 383–417. Marrett, G. K. and Worthington, A. C. (2009). An empirical note on the holiday effect in the Australian stock market. Applied Financial Letters, 16 (17), 1769–1772. Meneu, V. and Pardo, A. (2004). Pre-holiday effect, large trades and small investor behavior. Journal of Empirical Finance, 11 (2), 231–246. Michaely, R. and Womack, K. L. (2006). What are analysts really good at? Working Paper, Dartmouth College. Mikhail, M. B., Walther, B. R., and Willis, R. H. (2004). Do security analysts exhibit persistent differences in stock picking ability? The Journal of Financial Economics, 74 (1), 67–91. Nagel, S. (2005). Short sales, institutional investors and the cross-section of stock returns. Journal of Financial Economics, 77 (2), 277–309. Peng, L. and Xiong, W. (2006). Investor attention, overconfidence, and category learning. Journal of Financial Economics, 80 (3), 563–602. Pettengill, G. N. (1989). Holiday closings and security returns. Journal of Financial Research, 12 (1), 57–67. Schwarz, N. (2001). Feelings as information: Implications for affective influences on information processing. In L. L. Martin and G. L. Clore (Eds.), Theories of Mood and Cognition: A User’s Guidebook. Mahwah, NJ: Erlbaum, 159–176. Sorescu, S. and Subrahmanyam, A. (2006). The cross section of analyst recommendations. Journal of Financial and Quantitative Analysis, 41 (1), 139–168. Stickel, S. E. (1995). The anatomy of the performance of buy and sell recommendations. Financial Analysts Journal, 51 (5), 25–39. Thaler, R. H. (1999). The end of behavioral finance. Financial Analyst Journal, 55 (6), 12–17. Vergin, R. C. and McGinnis, J. (1999). Revisiting the holiday effect: Is it on holiday? Applied Financial Economics, 9 (3), 477–482. Womack, K. L. (1996). Do brokerage analysts recommendations have investment value? Journal of Finance, 51 (1), 137–167.

page 154

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0007

Chapter 7

The Critical Impact of Firms’ Market Value on Investor Behavior Following Pharmaceutical IPOs

Smadar Siev∗ and Tiran Rothman†

Abstract This chapter analyzes stock return behavior following initial public offering (IPO) events in the pharmaceutical sector and examines factors that could have an impact on this behavior. The results of the research indicate a positive Cumulative Average Abnormal Return (CAAR) of 3.03% in the 20 days following the IPO until the end of the quiet period for all firms under examination, and a decline of tens of percent in the 18 months post-IPO. When the sample is divided into two subsamples according to firm size, a market value (MV) of US$500 million can be identified as a threshold for positive or negative post-IPO yields. Companies with an MV below this threshold experience a positive but not significant CAAR in the first 20 days post-IPO and a significant negative CAAR from day 31 onwards. In contrast, companies above this US$500 million threshold show a significant positive CAAR 20 days post-IPO, followed by a consistent increase in CAAR for the next few months.The results also indicate that MV, IPO proceeds, shareholder dilution and clinical phases are critical factors determining ∗ Ono Academic College, Faculty of Business Administration, Haifa Campus, Port Gate Building, Port 32, Haifa 3303201, Israel; [email protected]. † WIZO Academic College, School of Management, Haganim 21, Haifa 31090, Israel; [email protected].

155

page 155

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

156

post-IPO returns. In conclusion, we suggest that investors recognize a US$500 million market value of a firm as a confidence threshold when investing in newly issued pharmaceutical companies. We postulate that firms valued above this amount attract more attention and gain greater investor confidence than do firms below this threshold. Lower-valued firms shares can be considered “lottery stocks,” as their IPO ignites a period of enthusiasm until the quiet period ends, where after investors’ attention to such firms gradually diminishes, and their focus moves on to their next potential lottery-like opportunity. Keywords: IPO, pharmaceutical companies, financial markets, behavioral finance, market value

1.

Introduction

The pharmaceutical industry discovers, develops and produces medications and medical devices. According to its revenues and capitalization, it is one of the world’s top five industries, with total annual revenues of over US$700 billion, most of which are generated by multinational pharmaceutical giants that have been dominating the industry for decades. A new drug can take on average 12–14 years to develop, at a cost of between US$1.3 and US$1.6 billion. Out of more than 10,000 drug discovery trials, only one will eventually lead to a new drug coming to the market. With the rapid development of biotechnology over the last decades,the industry has been changing and creating space for lesser-sized pharmaceutical firms. The Jumpstart Our Business Startups (JOBS)Act (detailed in the following section) has facilitated access to the capital market for small firms. As a result, a growing number of smaller pharmaceutical companies are seeking to raise public capital through IPOs. 1.1.

Focus of the Study

This chapter focuses on pharmaceutical firms that issued initial public offerings (IPOs) in the United States between January 2013 and December 2018. Its purpose is to clarify if and how the new JOBS Act, had an effect on firms’ stock returns during the 18 months postIPO. The first part of the chapter analyzes the Cumulative Average Abnormal Return (CAAR) of the stocks,and the second part examines factors that may affect these stock returns. Some of these factors are well documented, such as company market value or IPO proceeds,

page 156

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

The Critical Impact of Firms’ Market Value

page 157

157

while others are specific to pharmaceutical companies, such as drug regulatory status, firms’ therapeutic area and more.

1.2.

The JOBS Act and IPO Regulatory Periods in the United States

The JOBS Act, enacted in the United States in April 2012, was designed to help revitalize the IPO market by providing a series of regulatory, accounting and disclosure easements for Emerging Growth Companies (EGC). EGCs are characterized by annual gross revenues of less than US$1 billion over the year prior the IPO. Dambra et al. (2015) “estimate that the JOBS Act has led to 21 additional IPOs annually, a 25% increase over pre-JOBS levels.” In addition, offerings of EGC firms increased by 53% following enactment of the law in comparison to 10% for non-EGC firms. Of these, biotechnology and pharmaceutical firms had the greatest increase in activity as they were more likely to take advantage of the act’s risk reduction provisions which permit firms to file their IPO confidentially while making overtures to qualified institutional buyers. The IPO regulatory process is divided into a number of specified periods. The first, the pre-filing period, begins when a firm chooses an underwriter and ends when the firm files a registration statement with the Securities and Exchange Commission (SEC). The second waiting/pre-effective period or the quiet period, begins when the company files a registration statement with the SEC and ends when the registration statement is declared effective. During this waiting period, the laws limit the information a firm and related parties can release to the public. In addition, investment bankers and underwriters are not permitted to release any analyst coverage, including buy or sell recommendations, during this period. Once the quiet period expires, analyst coverage is released to the public. This quiet period can last as few as 10 days, but in many cases, investment bankers will require a quiet period of 25 days to fulfill their legal requirement to deliver a prospectus to the SEC. The third period, the post-effective period, begins when the registration statement is declared effective by the SEC. In the fourth, or lock-up period, major shareholders are prohibited from selling their shares. Lock-up periods usually last between 90 and 180 days following the IPO. Once the lock-up period ends, most trading restrictions are removed.

November 4, 2020

158

1.3.

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

The Clinical Journey from the Lab to the Shelf

Products development is a critical element in the work and potential success of pharmaceutical firms. In general, the stages of drug development research are the research project stage, the pre-clinical stage and clinical Phases I, II and III. The research project stage is the stage of choosing a molecule, such as a gene or protein, which has pharmacological or biological activity likely to be therapeutically useful. The pre-clinical stage is aimed at determining the dosage that can be safely administered to people during the clinical phases. The clinical stages, lasting an average of six to seven years, involve testing with humans to ensure that the drug is effective and safe to use. A drug must meet success criteria at each stage before moving on to the next one. During Phase I, the main goals are to assess safety and tolerability, and to explore how the drug interacts in the body. The main goals of Phase II are to evaluate a drug’s effectiveness in patients, to further explore its safety, and to determine the optimal dose. Studies during this phase are usually carried out in hospitals and involve a small number of patients who are already suffering from a serious illness or who have exhausted all other existing treatments. Phase III is the final step before regulatory approval by the Food and Drug Administration (FDA) and is the most expensive. During this phase, large studies are conducted involving 500–5,000 or more patients to determine a drug’s added value, effectiveness and safety. If candidates administered the drug in Phase III clearly benefit from it and the drug’s risk level is acceptable, the company can file a New Drug Application (NDA) with the FDA requesting regulatory approval to market the drug. After receiving approval, the company can move to the market stage in which it manufactures and markets the drug. 1.4.

Common Causes of Mortality in the United States

Clearly, pharmaceutical firms have a strong incentive to find solutions for serious medical issues plaguing the population. The major causes of death in the United States have remained fairly consistent over the past five years, and 10 factors are responsible for about 74% of all deaths. According to the Center for Disease Control and Prevention (CDC), in 2017, heart disease was the leading cause of death for

page 158

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

The Critical Impact of Firms’ Market Value

page 159

159

both men and women, accounting for 23.5% of all deaths. The second leading cause of death was cancer, accounting for about 21.3% of all deaths in 2017. 1.5.

Shares of Small Pharmaceutical Firms: Lottery Type Stocks?

Lottery stocks have been characterized in the literature as stocks with features similar to those of a lottery ticket: their purchase offers a high chance of a small loss but a small chance of a big profit (Markowitz, 1952). These shares were quantitatively characterized in a study by Kumar (2009) as having low prices, high idiosyncratic bias and high idiosyncratic volatility. Thus, buying a share in a pharmaceutical firm during its initial stages can be likened to buying a lottery ticket, where there is a small chance of great success (as previously noted, fewer than one out of 10,000 drug discovery trials result in a new drug coming to the market) and a large chance of losing all or part of the investment, which will be reflected in the fall in the share price. Kumar (2009) finds that lottery-type stocks underperform and that stock price is “one of the defining characteristics of lottery-type stocks because, like lotteries, if investors are searching for cheap bets, they should naturally gravitate toward low-priced stocks. Thus, stock price is likely to be an important characteristic of stocks that might be perceived as lotteries” (p. 1899). 1.6.

Stocks Returns Post-IPO and Factors Affecting these Returns

Studies involving IPOs have covered a wide range of issues and those most relevant to this chapter address share performance up to three years following firms’ IPOs. Jain and Kini (1994) showed low performance of firms for up to three years following the IPOs and Loughran and Ritter (1995) reported that IPO stocks yielded an average of 5% over a one-year post-IPO period, compared to 12% for a comparably-sized non-IPO benchmark stocks. In a seminal paper, Ritter and Welch (2002) investigated the long-term performance of IPOs and found that the three-year average market-adjusted return on IPOs was a negative 23.4%. In contrast, a study conducted

November 4, 2020

160

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

by Goergen et al. (2009) on IPOs in France and Germany issued during the period 1996–2000 found no significant abnormal returns (ARs). Researchers have also been puzzled about declines in returns close to the expiration of IPO lock-up periods, and some studies have concluded that the market reacts negatively to lock-up period expirations. The research of Ofek (2000), conducted in the United States from 1996 through 1998, found an abnormal negative return during the lock-up expiration period, as well as a 1%–3% drop in the stock price and a 40% increase in volume of trading 180 days after the IPO. Examining IPOs in the United States from 1988 through 1997, Bradley et al. (2001) (see also, Brav and Gompers, 2003; Field and Hanka, 2001) observed negative ARs of approximately 2% near the time of the lock-up period’s expiration. Komenkul and Kiranand (2017) found positive and significant CAAR of 5.57%, 36 months post-IPO in Association of Southeast Asian Nations (ASEAN) countries between 1986 and 2014. Malaysia and Singapore present the highest and lowest CAARs of 57.25% and −39.4%, respectively, three years post-IPO. Thakor et al. (2017) distinguish between pharma and biotech companies. Their findings indicate that, for the period 1980–2015, the biotech and pharma sectors produced the lowest and highest annualized mean returns of 6% and 14%, respectively. The biotech sector was also characterized by the lowest Sharp index during this period. Thakor et al. (2017) confirm that almost all biotech companies are loss-making enterprises. Previous research has analyzed a wide range of factors that have an impact on IPOs’ long-term performance: initial return, underwriter reputation, the existence of venture capital (VC) backing, financial ratios, size and many more. Other studies focusing the pharma sector have also analyzed factors like R&D expenses and the number of patents. This chapter refers mainly to studies analyzing factors relevant to its research focus, including: firm size, IPO proceeds, dilution percentage and the number of products clinically tested by the firm that resulted in mixed findings. A study conducted by Durukan (2002) analyzed stocks’ performance for three years post-IPOs on the Istanbul Stock Exchange between 1990 and 1997. Privatization, firm size and gross proceeds were found to have positive effects on returns, while shareholder dilution was found to have

page 160

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

The Critical Impact of Firms’ Market Value

b3964-ch07

page 161

161

a negative effect. Gao et al. (2006) suggested that a greater divergence of opinion among investors and investor sentiment were sources of long-term performance. Firm size and IPO proceeds were found to be irrelevant variables in explaining long-term excess returns. In contrast, a study conducted by Goergen et al. (2007), using a United Kingdom dataset of IPOs issued between 1991 and 1995, found firms’ size and multinationality at the time of the IPO to have a positive impact on long-term performance. Higher issuing costs, firms’ profitability prior to the IPO and higher shareholder dilution were found to have negative effects on returns. The age of the firm and the reputation of the underwriter were found to be irrelevant with respect to returns. Chan and Lo (2011) suggest that firms with credit ratings present significantly less initial underpricing in comparison to firms without credit ratings and that credit-rated firms do not exhibit abnormal long-term performance. These results indicate that increased disclosure contributes to price corrections in the short term. Thomadakis et al. (2012) explored Greek IPOs between 1994 and 2002 and found that the factors affecting long-run performance were ownership concentration, board classification and issuance during a pronounced “hot period” IPO wave.Firm market size was found to be an irrelevant factor in terms of long-term performance. Regarding biopharma firms’ IPOs, Higgins et al. (2011) explored the factors affecting the IPO proceeds during the two time periods of 1989–1992 and 1996–2000, and found that “firms with an affiliated Nobel prize winner succeeded in raising the value of their firms by more than $30 million compared to firms without a Nobel laureate during the first period.” The affiliation with a Nobel Prize laureate lost its significance as a signal of value in the second period. The effect of dilution was negative in both time periods, but nevertheless dropped by about half between the two periods. The number of products a firm was testing in clinical phases had a positive significant effect in the first period only. More recent studies, such as that of Gorry and Useche (2017) suggest that a firm working on a drug with an orphan drug designation could be expected to experience higher proceeds at the IPO date. In addition, the effect of an orphan drug designation is stronger than patent applications or later-stage drugs compounds. Higher valuations were also related to the VC role, underwriters’ reputations and R&D expenses. The number of drugs undergoing at least Phase II of clinical testing and the number of

November 4, 2020

162

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

patents applied for in a four-year window prior to the IPO were not statistically significant. 2.

Data and Analysis

Average firm market value in our database (Table 2) is US$537 million. We chose a rounded market value of US$500 million as a separator threshold. Companies above this threshold value will be referred below as large companies and those valued below the threshold will be referred to as small companies. 2.1.

Research Goals and Hypotheses

As described before, the quiet period of an IPO process expires 10–25 days after an IPO is priced and opens for trading. The launching of coverage by the underwriters on that day can have a significant impact on the stock price. The goal of this research was to investigate CAAR behavior from the IPO date to the end of the quiet period and thereafter. We hypothesize that an upward trend can be expected in CAAR during the quiet period and a downward trend when the quiet period expires. The expected increase in CAAR during the quiet period can be attributed to the natural excitement generated by promotion immediately following the IPO. The later downward trend can be explained, in part, by the publication of reports about the company or its sector and future forecasts by affiliated analysts. Dividing the sample into two subsamples according to firm size, it is anticipated that large-sized firms will enjoy better performance because they are likely to have more experience, more available resources and a more extensive product portfolio. The presence of these factors is likely to enhance a large firm’s potential for future success as well as attract greater attention from investors. The following hypotheses were formulated to reflect these expectations: (H1) Quiet Period and Stock’s Return: The natural excitement generated from promotion about the new IPO should result in a positive CAAR from the IPO date until the end of the quiet period. At the end of this period, when coverage by underwriters and their affiliated analysts begins, this initial excitement diminishes. As a result, the stock will then experience a negative CAAR.

page 162

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

The Critical Impact of Firms’ Market Value

b3964-ch07

page 163

163

(H2) Stock Returns and Market Capitalization: Large-sized firms’ shares should show higher yields than do shares of small-sized firms, due to higher investor attention to large-sized firms and greater certainty about such firms’ future success. There are multiple factors that are known to have an effect on returns,such as risk, firm size and Book-to-Market Ratio (Fama and French, 1992). In the context of IPOs, other possible factors could prove relevant, some of which will be examined in this chapter. Due to the long period of time required to develop a drug, financial resources are critical. Naturally, the larger the firm, the more resources it has and the greater likelihood it has of reaching the market with a product. Accordingly, we assume first, that market value will have a positive impact on returns, with a greater impact for large companies. Second, we posit that the variable of the IPO proceeds could have two contrasting effects. On the one hand, the greater the proceeds, the more diluted the existing shareholders become, which would lead to a negative impact on returns. However, when the proceeds are greater in proportion to a firm’s value after the offering, the firm’s ability to continue its operations improves, thus positively affecting its returns. Accordingly, the following additional hypotheses were formulated: (H3) Market Value: Market value should have a positive impact on returns. The impact will be higher for large companies. (H4) IPO Proceeds: The proceeds from the IPO should have two opposing effects. The sheer amount of the proceeds will have a negative impact on returns. However, as a percentage of the firm’s market value, the impact of the proceeds will be positive. Drug development is a long and expensive process (see Section 1.3). We hypothesize that advanced regulatory stages, such as Phase III and the market stage, will have a positive impact on returns as the company moves closer to sales or is already selling. In contrast, the earlier development stages are expected to have a negative impact on returns due to the large sums of money required until a product reaches the market, if at all. It should be noted that these stages are not mutually exclusive, as a company can be working on several products in different therapeutic areas and at different regulatory stages. We also have hypothesized that the total number of products and the number of products at each regulatory stage would have a positive impact on the returns due to higher future

November 4, 2020

164

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

sales potential. As described in Section 1.4, cardiovascular disease and cancer are responsible for approximately 45% of deaths in the United States in recent years. We assume that engaging in research and producing as many products as possible in these areas will result in a positive impact on returns. Based on these assumptions, the following hypotheses were formulated: (H5) Regulatory phase: Advanced drug development stages such as Phase III and the market stage should have a positive impact on returns. The earlier stages should have a negative effect. (H6) Number of products: A larger total number of products and number of products at each regulatory stage should have a positive impact on returns. (H7) Therapeutic area: Developing drugs in the areas of cardiology and oncology as well as the larger number of products in these areas should have a positive impact on returns. Most of the firms in our database are in the early stages of drug development, and are therefore characterized by significant uncertainty. Investors will tend to treat such stocks as lottery stocks (see Section 1.5), in which there is a small probability of huge rewards in the event that the firm is able to move to the next stage and a large probability of small losses if the firm fails to continue on to the next stage. If the profit potential is not realized after a short holding period, the investor will dispose of this share. We assume that small firms’ shares are more likely to be perceived as lottery shares, and this is reflected in lower trading volume and lower share price.1 Accordingly, the following research hypothesis was formulated: (H8) Lottery type stocks: Small firms’ shares are perceived as lottery stocks and therefore will underperform and experience lower trading volume and lower stock prices.

1

Due to the fact that new issues are being addressed in this chapter, and given the short period after the IPO that is being examined, volatility and idiosyncratic bias cannot be examined.

page 164

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

The Critical Impact of Firms’ Market Value Table 1:

page 165

165

IPOs per year

Year

Total No. of IPOs

No. of Pharma IPOs

No. of Pharma IPOs (In %)

2013 2014 2015 2016 2017 2018 Total

248 312 200 128 210 258 1,356

30 70 49 29 50 82 310

12 22 24 23 24 32 23

2.2.

Data and Method

Our initial database consisted of all the pharmaceutical companies that issued IPOs in the United States from January 2013 to December 2018, 96% of which were issued on the Nasdaq and the rest on the New York Stock Exchange (NYSE). We excluded firms that became private or were merged into or acquired by others from the time of the IPO until 18 months following the IPO. Our final database consisted of 310 firms. The Evaluate Pharma2 database was used to extract the issue date, products count according to therapeutic area and regulatory stage. We extracted the issue date, price and amount of money raised from the Nasdaq site.3 Closing price and trading volume were retrieved from Yahoo Finance.4 Market capitalization was calculated for December 31 of the IPO year by multiplying the number of shares appearing in the firms’ profit and loss statement by the stock price on that day. The result was confirmed with the value appearing on the stockraw.com website. Table 1 displays IPO statistics in the United States. The proportion of pharmaceutical firms among all the IPOs increased consistently from 12% in 2013 to 32% in 2018. Table 2 presents descriptive statistics of the market value of the firms in our database. A prominent feature of these firms is their relatively low market capitalization, which averaged US$539 million, 2 Evaluate Pharma database is one of the top global pharma databases: http:// www.evaluate.com/. 3 https://www.nasdaq.com/market-activity/ipos. 4 https://finance.yahoo.com/.

November 4, 2020

166

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach Table 2: Market value statistics as of December 31 of the IPO year (US$ Million) Average Std. Dev. Max Min Median

538.7 931.9 11,528.2 1.3 296.1

in comparison to the average market value of other companies issued in those years, which was US$1,891 million.5 We calculated CAAR for the 18 months following the IPO and conducted a number of sets of regressions. All calculations were performed for the entire sample and for the subsamples of large and small firms. 2.3.

CAAR Analysis

The event study approach was employed to examine market reaction to IPO events. The actual date of the IPO was marked as t = 0 and the daily stock prices were applied for the period t = 0, . . . , 375 (18 months or 375 trading days post-IPO), to calculate daily logarithmic returns. Two return benchmarks, the IXJ Healthcare Index, and the S&P 500 Market Index were utilized. The AR was calculated by subtracting the benchmark returns from the stock returns. CAAR was calculated by aggregating daily ARs and averaging across all the firms in the database. As no stock prices exist prior to the IPO, conditional returns using the market model were not calculated. In addition, normalized trading volumes were computed as a proxy for market attention. For each firm in the sample, the natural logarithm of the daily trading volume throughout the period t = 0, . . . , 375 was recorded, and each observation was normalized by subtracting the mean and dividing by the standard deviation 5

According to Jay R. Ritter, University of Florida, Warrington, Department of Finance website: https://site.warrington.ufl.edu/ritter/files/IPOs2019Statistics Mar10 2020.pdf (Table 1(a)).

page 166

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

The Critical Impact of Firms’ Market Value

b3964-ch07

page 167

167

calculated over the period. Then, the average across all firms for each day relative to the IPO date was calculated. 2.4.

CAAR Results and Discussion

The CAAR results for the entire sample and the two subsamples are displayed in Table 3. Panel A displays CAAR for selected time periods during the 18 months post-IPO, Panel B contains a display of the CAAR and the normalized trading volumes, and Panel C presents the trading volumes for the entire sample and for the two subsamples. CAARs were calculated for the two benchmarks of market and sector indices. As the CAAR results relative to these two benchmarks are similar, only the results for the sector index are displayed. As shown in Panel A of Table 3, the CAAR for the entire sample over the first 20 trading days post-IPO was positive, significant, and equals 3.03% (t = 2.54). About 20 trading days following the IPO, performance began diminishing quickly. Around 100 trading days post-IPO, CAAR = −7.01% (t = −2.08); and 375 trading days postIPO, CAAR = −43.94% (t = −6.91). These results are consistent with previous literature and support hypothesis (H1). When analyzing the subsamples, the overall picture changes dramatically.With respect to small companies, CAAR for the first 20 trading days post-IPO was positive yet not significant, with CAAR = 1.31% (t = 1.23). About 50 trading days post-IPO, CAAR was negative and significant, at −7.01%, (t = −2.79). At 100 trading days post-IPO, CAAR was −16.81%, (t = −4.4). Finally, 375 trading days post-IPO CAAR was −62.2% (t = −7.98). The results for large firms reveal a completely different picture. After 20 trading days, CAAR was positive and significant, with CAAR = 6.81%, (t = 2.9). After 50 trading days, CAAR = 10.17%, (t = 2.81); CAAR reached its peak of 15.93% on day 105 post-IPO and began to decline from that point onward. After one year, the CAAR was 7.34% (t = 0.92) until it disappeared completely 315 days post-IPO. As (H2) posited, large firms performed better than did small ones. However, 18 months after the IPO, both small and large firms presented negative CAARs. Therefore, hypothesis (H2) proved correct only with respect to the first year post-IPO. Panel B shows that the CAAR decline was consistent from day 20 onward for small firms but much more volatile for large ones.

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

168 Table 3:

Post-IPO CAARs, from 1 to 375 days

The Entire Market Market Days Sample Value < US$500 Million Value > US$500 Million Relative to Event CAAR, % t-statistic CAAR, % t-statistic CAAR, % t-statistic Panel A: CAAR Results for Selected Time Periods Post-IPO 1–10 0.77 0.84 −0.11 −0.12 2.70 1–20 3.03 2.54 1.31 1.23 6.81 1–50 −1.63 −0.73 −7.01 −2.79 10.17 1–100 −7.01 −2.08 −16.81 −4.4 14.53 1–150 −15.53 −3.71 −27.70 −5.57 11.19 1–200 −21.84 −4.56 −37.53 −6.68 12.64 1–250 −25.80 −5.05 −40.88 −6.64 7.34 1–375 −43.94 −6.91 −62.26 −7.98 −3.49

1.5 2.9 2.81 2.82 1.81 1.74 0.92 −0.36

Panel B: CAAR Graphic Display

Panel C: Trading Volumes

Trading Volumes 60,00,000 50,00,000 40,00,000 30,00,000 20,00,000 10,00,000

1 19 37 55 73 91 109 127 145 163 181 199 217 235 253 271 289 307 325 343 361

0

The EnƟre Sample

Big

Small

Note: Table 3 presents CAAR and trading volume results for selected time periods following the IPO. The entire sample contains 310 firms. The subsample of small firms with a market value MVUS$500 million contains 97 firms (31%).

page 168

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

The Critical Impact of Firms’ Market Value

page 169

169

In terms of trading volume, the IPO day was characterized by the highest trading volume, which was 15 times higher than the average trading during the entire measuring period, reflecting the great excitement immediately after the IPO. Trading volume declined significantly after this day. As shown in Panel C, the trading volume of large companies was on average 2.6 times greater than the trading volume of small ones throughout the 18 month post-IPO period. 2.5.

Regressions Equations

According to the hypotheses in Section 2.1, the extended regression equations were:6 Return(time period)/AR(time period) = β0 + β1 Ln(MV) + β2 Ln(Prcds) + β3 Prcds(%) + β4 Year2013 + .. + β8 Year2017 + β9 Is RP + β10 Is PC + β11 Is I + β12 Is II + β13 Is III + β14 Is Mrkt + β15 Prd RP + β16 Prd PC + β17 Prd I + β18 Prd II + β19 Prd III + β20 Prd Mrkt + β21 TPrd + β22 Is Onco + β23 Is Crdio + β24 Prd Onco + β25 PrdCrdio The explained variable was return and AR alternatively. AR was calculated for two benchmarks:the S&P500 Market Index and the Pharma Sector Index IXJ. The explanatory variables were: • Ln (MV) represents the natural logarithm of a firms’ market value. • Ln (Prcds) represents the natural logarithm of the amount of money raised in the IPO; Prcds (%) represents the amount raised as a percentage of the firm’s market value. • Years 2013–Year 2017 are dummy variables for the issued years. Year 2013 receives 1 and 0 otherwise and so on. 6

As trading starts on the IPO date, parameters such as risk, Book-to-Market Ratio, volatility and more could not be measured over a time period before the event and therefore are not included.

November 4, 2020

170

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

• RP represents Research Project; PC represents Pre-Clinical; I, II and III refers to Phases I through III, respectively; Mrkt represents the market stage. The next set of variables beginning with “Is” is a set of dummy variables for the firm’s drug regulatory stages. The dummy variable receives 1 if the firm has products at this stage and 0 otherwise. • The next set of variables from Prd RP to Prd Mrkt are the number of products at each regulatory stage.It should be noted a firm can have several products in different regulatory stages. • T Prd represents the total number of products for a firm; ◦ Is Onco/Is Crdio is a dummy variable that receives 1 if the company has oncology or cardiology products and 0 otherwise. ◦ PrdOnco/PrdCrdioare the number of products in the field of oncology and cardiology, respectively. These variables were measured on the day of issue. The number of observations in the regression section (Section 2.5) is lower than that in the CAAR section (Section 2.4), due to the partial availability of data. We performed three sets of OLS regressions that differed within the time period of the explained variable at the points of 6, 12 and 18 months post-IPO date. We conducted these regressions for the entire sample and for the subsamples of large and small firms. 2.6.

Regression Results and Discussion

Results for the sector index only are being presented due to a great similarity in results for AR for the two benchmark indices. The results of the reduced models are presented in Table 4. Panels A, B and C of Table 4 present data for the 6, 12 and 18-months points after the IPO, respectively. We will refer to Table 4 for the results of the return variable because of the similarity of results for the AR and the return. Market value was found to have a positive impact on returns over time. The coefficient for large firms is larger than for small ones and its effect increased over time. Coefficient values were 0.7 for large firms vs. 0.23 for small firms six months post-IPO, 0.61 for large firms vs. 0.16 for small firms 12 months post-IPO, and 5.47 for large firms

page 170

The Entire Sample Return

Regressions results

Market Value < US$500 million Return

AR to Sector

Market Value > US$500 million Return

AR to Sector

10:12

0.26 212

0.46(0.66) 0.70(0.00) −0.92(0.00)

0.36(0.73) 0.68(0.00) −0.89(0.00)

0.41(0.00) 0.12(0.24) −0.01(0.92) 0.43(0.04) 0.22(0.19) −0.17(0.03)

0.31(0.02) 0.04(0.69) 0.03(0.77) 0.40(0.05) 0.18(0.27) −0.17(0.03)

−0.11(0.68) 0.06(0.81) −0.58(0.02) 0.09(0.85) −0.11(0.65) −0.46(0.00) −0.27(0.08)

−0.21(0.42) −0.01(0.98) −0.50(0.04) −0.02(0.97) −0.13(0.60) −0.45(0.00) −0.25(0.10)

0.55(0.01) 0.24 145

0.57(0.01) 0.22 145

−1.23(0.00) 0.16(0.01)

−1.28(0.00) 0.16(0.01)

1.01(0.54) 0.61(0.03) −1.04(0.00)

0.82(0.62) 0.62(0.03) −1.02(0.00)

0.73(0.00) 0.35(0.02) −0.07(0.69) 0.72(0.01)

0.55(0.00) 0.25(0.09) 0.03(0.85) 0.7(0.02)

−0.12(0.76) 0.57(0.1) −0.34(0.33) −0.05(0.94)

−0.28(0.46) 0.48(0.17) −0.24(0.50) −0.01(0.99)

0.37 67

0.34 67

(Continued)

b3964-ch07

−0.27(0.35) 0.24(0.00) −0.25(0.00)

171

Panel B: Twelve Months Post-IPO Intercept −0.77(0.06) −0.84(0.04) Ln (MV) 0.40(0.00) 0.40(0.00) Ln (Prcds) −0.42(0.00) −0.41(0.00) Prcds(%) 0.09(0.08) 0.09(0.08) Year 2013 0.45(0.01) 0.27(0.14) Year 2014 0.43(0.00) 0.33(0.03) Year 2015 −0.15(0.38) −0.05(0.77) 0.50(0.09) 0.49(0.09) Year 2016

−0.2(0.48) 0.23(0.00) −0.26(0.00)

Behavioral Finance: A Novel Approach – 9in x 6in

−0.19(0.48) 0.41(0.00) −0.48(0.00) 0.07(0.04) 0.14(0.25) 0.1(0.33) −0.09(0.41) 0.23(0.25) 0.07(0.60) −0.25(0.00)

The Critical Impact of Firms’ Market Value

Panel A: Six Months Post-IPO Intercept −0.13(0.63) Ln (MV) 0.42(0.00) Ln (Prcds) −0.49(0.00) Prcds(%) 0.07(0.04) Year 2013 0.25(0.05) Year 2014 0.18(0.08) Year 2015 −0.14(0.21) Year 2016 0.27(0.18) Year 2017 0.1(0.47) Is RP −0.25(0.00) Is I Is Mrkt Adj R Square 0.28 Observations 212

AR to Sector

November 4, 2020

Table 4:

page 171

Market Value > US$500 million

AR to Sector

Return

AR to Sector

Return

AR to Sector

0.46(0.02) 0.33(0.15) 0.10(0.05) 0.13(0.13) 0.23 208

0.43(0.03) 0.33(0.15) 0.10(0.04) 0.14(0.12) 0.19 208

0.22(0.34) 1.1(0.00) 0.11(0.03)

0.2(0.4) 1.12(0.00) 0.11(0.03)

0.73(0.05)

0.71(0.06)

0.27 144

0.21 144

0.27 64

−2.09(0.00) 0.27(0.01)

−2.19(0.00) 0.27(0.01)

0.1(0.08) 1.15(0.00) 0.46(0.11) −0.1(0.75) 0.77(0.07) 0.05(0.88) 0.32(0.04) 2.25(0.00)

0.1(0.06) 0.89(0.01) 0.41(0.14) 0.03(0.92) 0.69(0.09) 0.07(0.86) 0.29(0.06) 2.31(0.00)

0.42 119

0.38 119

−12.98(0.01) 5.47(0.00) −6.22(0.00) 37.53(0.01) 1.03(0.19) 0.63(0.4) 0.19(0.81) 1.28(0.23) 0.93(0.23)

−12.92(0.01) 5.32(0.00) −6.00(0.00) 36.33(0.01) 0.79(0.30) 0.67(0.37) 0.34(0.66) 1.21(0.25) 0.95(0.21)

−0.92(0.08)

−0.88(0.08)

0.15 54

0.13 54

b3964-ch07

Panel C: Eighteen Months Post-IPO −1.82(0.00) Intercept −1.72(0.00) Ln (MV) 0.22(0.00) 0.22(0.00) Ln (Prcds) Prcds(%) Year 2013 1.08(0.00) 0.82(0.02) Year 2014 0.61(0.06) 0.57(0.07) Year 2015 0.07(0.83) 0.2(0.55) Year 2016 0.8(0.08) 0.73(0.11) Year 2017 0.41(0.28) 0.4(0.28) Is I 0.65(0.04) Is Mrkt 0.61(0.06) Prd III 0.3(0.04) 0.3(0.04) Adj R Square 0.19 0.15 Observations 173 173

0.24 64

Behavioral Finance: A Novel Approach – 9in x 6in

Return

10:12

Market Value < US$500 million

Behavioral Finance: A Novel Approach

Year 2017 Is Mrkt Prd II Prd III Adj R Square Observations

(Continued)

November 4, 2020

The Entire Sample

172

Table 4:

page 172

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

The Critical Impact of Firms’ Market Value

b3964-ch07

page 173

173

vs. 0.27 for small firms 18 months post-IPO. These results validate hypothesis (H3). The IPO proceeds as measured by Ln(Prcds) adversely affected returns in the first six months after the IPO. Its effect on large companies was 3.5 times greater than on small ones: −0.92 vs. −0.26. Following this period, the effect of the IPO proceeds primarily affected the large firms, and its impact increased with time, from −1.04, 12 months post-IPO to −6.22, 18 months post-IPO. Prcds(%), the amount raised as a percentage of the company’s MV, had a positive impact on returns. Up to one year after the IPO, the coefficient appeared significant for the entire sample. However, 18 months after the IPO, the coefficient was 375 times higher for large firms than for the small ones: 37.53 vs. 0.1. Examining the aggregate effect of the IPO proceeds, in the first year, it was negative for the entire sample. However, after 18 months, the direction reversed and the effect was positive for the subsample of large firms. We suggest that the negative effect in the first year after the IPO is attributable mainly to shareholder dilution. Over time, if the money raised is used in a way that contributes to the prosperity of the firm, the effect becomes positive, as reflected in the subsample of large firms. This finding supports hypothesis (H4). The years variables were tested as a group. Some of the coefficients were significant, probably due to differences in market performance during the sample years. With regard to development stages, being in a research project stage was found to have a negative effect on returns, but only in the first six months post-IPO. Being in Phase I had a negative impact on large firms in the first six months after the IPO, but a positive impact on small firms 18 months after the IPO. We suggest that firms that are in Phase I at the time of the IPO may achieve greater advances after a year and a half. Being at the market stage during the first six months after the IPO seems relevant only for small firms, and has a positive effect. One year post-IPO this effect remains positive and becomes significant for the entire sample, probably because of the impact of the increased returns for the small firms. About 18 months post-IPO, being in the market stage still has a positive effect on small firms, but has a surprisingly negative effect for the large ones.

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

174

For small companies, the effect increased over the years from 0.55, 6 months post-IPO to 1.1, 12 months post-IPO and 2.25, 18 months post-IPO. This rise in the effect of being in the market stage can most likely be attributed to increasing sales over time. These results validate hypothesis (H5). The total number of products was found to be irrelevant in explaining returns and ARs and therefore is not included in the reduced model. Consequently, the first half of hypothesis (H6), positing that the total number of products can have a positive effect on returns,was invalidated. The number of products in Phase II was found to have a positive effect on the one-year return after the IPO for the entire sample and for the small firms. The number of products in Phase III was found to have a positive effect on the entire sample at 12 and 18 months post-IPO. In view of the time that passed from the date of the IPO, it is likely that some of these products ultimately reached the market stage, which would account for the positive effect observed. Therefore, the second half of hypothesis (H6), positing that the number of products at each regulatory stage would have a positive effect on returns,was confirmed. Next, it was found that engaging in the therapeutic areas of cancer and heart disease had no effect on returns and ARS. Consequently, hypothesis (H7) was rejected. Lastly, to examine hypothesis (H8), we compared the average share price for each subsample at different points in time after the IPO. The results are shown in Table 5. The results displayed in Table 5 indicate that, at any given point in time, small firms’ stocks are characterized by lower stock prices than are large firms’ stocks. In addition, as was presented in Table 3, small firms under perform and have lower trading volumes than do Table 5:

Day Day Day Day Day

0 100 200 250 375

Average share price

MV < US$500 Million

MV > US$500 Million

P -value of Diff

15.3 15.1 13.9 13.4 13.7

21.7 24.7 26.0 26.3 27.6

< 0.01 < 0.01 < 0.01 < 0.01 < 0.01

page 174

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

The Critical Impact of Firms’ Market Value

b3964-ch07

page 175

175

large firms. We therefore conclude that small firms can be perceived as lottery stocks, thus confirming hypothesis (H9).

3.

Summary and Conclusions

This chapter analyzed the CAAR behavior of pharma firms that went public after the JOBS Act was enacted,and investigated the factors that could influence returns. In general, the JOBS Act aimed to facilitate small firms’ access to the capital market and boost job creation. Market Value has been shown to be a critical predictor of the success of a pharmaceutical firm in the short term after the issuance. A value of US$500 million was found to be a confidence threshold in investors’ willingness to buy and hold a share. Companies above this threshold gained investors’ confidence as reflected in their higher trading volumes and positive CAARs in the year following the IPO. Firms valued below this threshold might be perceived as lottery stocks that investors sell at a loss a short time after their purchase, and thus exhibit negative CAARs a few months after the IPO. The IPO ignites an initial period of enthusiasm which rises until the end of quiet period, whereupon investors’ attention to small-sized firms gradually diminishes, as they seek their next lottery-like opportunity. In spite of the success of the JOBS Act in increasing the proportion of pharmaceutical companies among all new offerings, in the short term, the consequences of IPOs for small pharma firms was a substantial loss to their shareholders. As suggested by Zingales (1995), Mello and Parsons (1998) and Dambra et al. (2015), an IPO can be a first step towards a future sale. This seems particularly relevant to small pharma firms whose acquisition by an established, asset-rich firm is likely to be the best option to support the drug development process until its successful completion. Regarding other factors effecting returns, it was found that shareholder dilution had a negative effect in the post-IPO year, but reversed its direction 18 months post-IPO. The negative effect of being in early stages of research is likely to be due to the inherent uncertainty in the process of developing a drug. Lastly, a firm’s engagement in the areas of oncology and cardiology was found to be irrelevant for its stocks’ returns.

November 4, 2020

10:12

176

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch07

Behavioral Finance: A Novel Approach

References Bradley, D. J., Jordan, B. D., Yi, H. C., and Roten, I. C. (2001). Venture capital and IPO lockup expiration: An empirical analysis. Journal of Financial Research, 24 (4), 465–493. Brav, A. and Gompers, P. A. (2003). The role of lockups in initial public offerings. The Review of Financial Studies, 16 (1), 1–29. Chan, K. C. and Lo, Y. L. (2011). Credit ratings and long-term IPO performance. Journal of Economics and Finance, 35 (4), 473–483. Dambra, M., Field, L. C., and Gustafson, M. T. (2015). The JOBS Act and IPO volume: Evidence that disclosure costs affect the IPO decision. Journal of Financial Economics, 116(1), 121–143. Durukan, M. B. (2002). The relationship between IPO returns and factors influencing IPO performance: Case of Istanbul Stock Exchange. Managerial Finance, 28 (2), 18–38. Fama, E. F. and French, K. R. (1992). The cross-section of expected stock returns. The Journal of Finance, 47 (2), 427–465. Field, L. C. and Hanka, G. (2001). The expiration of IPO share lockups. The Journal of Finance, 56 (2), 471–500. Gao, Y., Mao, C. X., and Zhong, R. (2006). Divergence of opinion and long-term performance of initial public offerings. Journal of Financial Research, 29 (1), 113–129. Goergen, M., Khurshed, A., and Renneboog, L. (2009). Why are the French so different from the Germans? Underpricing of IPOs on the Euro new markets. International Review of Law and Economics, 29 (3), 260–271. Goergen, M., Khurshed, A., and Mudambi, R. (2007). The long run performance of IPOs: Can it be predicted? Managerial Finance, 33, 401–419. Gorry, P. and Useche, D. (2017). Orphan drug designations as valuable intangible assets for IPO investors in pharma-biotech companies (No. w24021). National Bureau of Economic Research. Jain, B. A. and Kini, O. (1994). The post-issue operating performance of IPO firms. The Journal of Finance, 49 (5), 1699–1726. Higgins, M. J., Stephan, P. E., and Thursby, J. G. (2011). Conveying quality and value in emerging industries: Star scientists and the role of signals in biotechnology. Research Policy, 40(4), 605–617. Komenkul, K. and Kiranand, S. (2017). Aftermarket performance of health care and biopharmaceutical IPOs: Evidence From ASEAN countries. INQUIRY. The Journal of Health Care Organization, Provision, and Financing, 54, 004695801772710. Kumar, A. (2009). Who gambles in the stock market? The Journal of Finance, 64 (4), 1889–1933.

page 176

November 4, 2020

10:12

Behavioral Finance: A Novel Approach – 9in x 6in

The Critical Impact of Firms’ Market Value

b3964-ch07

page 177

177

Loughran, T., Ritter, J. R., and Rydqvist, K. (1995). Initial Public Offerings: International Insights. Urbana-Champaign: Center for International Business Education and Research, College of Commerce and Business Administration, University of Illinois. Markowitz, H. (1952). The utility of wealth. Journal of Political Economy, 60, 151–215. Mello, A. S. and Parsons, J. E. (1998). Going public and the ownership structure of the firm. Journal of Financial Economics, 49 (1), 79–109. Ofek, E. (2000). The IPO lock-up period: Implications for market efficiency and downward sloping demand curves. Working Paper Series 99-054, New York University. Ritter, J. R. and Welch, I. (2002). A review of IPO activity, pricing, and allocations. The Journal of Finance, 57 (4), 1795–1828. Thakor, R. T., Anaya, N., Zhang, Y., Vilanilam, C., Siah, K. W., Wong, C. H., and Lo, A. W. (2017). Just how good an investment is the biopharmaceutical sector? Nature Biotechnology, 35 (12), 1149. Thomadakis, S., Nounis, C., and Gounopoulos, D. (2012). Long-term performance of Greek IPOs. European Financial Management, 18 (1), 117–141. Zingales, L. (1995). Insider ownership and the decision to go public. The Review of Economic Studies, 62 (3), 425–448.

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0008

Chapter 8

Behavioral Characteristics of IPO Underpricing Allen Michel∗ , Jacob Oded† and Israel Shaked∗

Abstract Earlier studies document positive first-day return for initial public offerings (IPOs), commonly interpreted as underpricing of the issue. The empirical evidence also indicates that IPO underpricing is negatively related to the public float (the fraction of the firm sold to the public). One possible explanation for this relation is that firms allocate a fixed amount of money for underpricing, and set an issue price accordingly — a behavioral characteristic. But, if indeed firms allocate a fixed amount of money to underpricing, then this underpricing should diminish in the public float. Using a sample of IPOs between 1996 and 2008, we provide empirical evidence that indeed the relation between underpricing and the public float is non-linear. Specifically, the higher the public float, the less the decrease of underpricing in the public float. Moreover, in our regression analysis, regressing underpricing on the reciprocal of the public float provides the best fit. As we show, this result is consistent with firms allocating a fixed amount of money for underpricing. This finding is important because it helps predict underpricing and has implications for firms, investors and regulators. Keywords: IPO, equity issuance, underpricing, public float

∗ Allen Michel and Israel Shaked are at the Questrom School of Business, Boston University, USA; [email protected] (Allen Michel), [email protected] (Israel Shaked). † Jacob Oded is at the Coller School of Management, Tel Aviv University, Israel; [email protected].

179

page 179

November 4, 2020

180

1.

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Finance: A Novel Approach

Introduction

Earlier studies document a positive average first-day return for IPOs which is commonly interpreted as underpricing of the issue. Several explanations have been offered in the literature for this underpricing, including compensation for risk, for adverse selection (information), and for the “winner’s curse.” It has also been suggested that underpricing is good for the issuing firms as it attracts investors. An empirical regularity associated with underpricing is that underpricing is negatively related to the public float (PF), i.e., to the fraction of shares sold to the public in the initial public offering (IPO) (e.g., Habib and Ljungqvist, 2001; Bradley and Jordan, 2002; Loughran and Ritter, 2004). One possible explanation for this negative relation is that firms allocate a fixed amount of money to underpricing and set an issue price accordingly.1 The reasons for fixing the amount are likely behavioral. This is because standard motivations such as compensating new investors for risk would imply a positive relation between the PF and dollar spending on underpricing. In this chapter, we consider this behavioral approach and its implications. We argue that if indeed firms allocate a fixed amount of money to underpricing, then not only will underpricing decrease in the issue size but also underpricing will be diminishing in the PF. Using a sample of 1907 IPOs between 1996 and 2008 we find that the decrease in underpricing is indeed diminishing in the PF. That is, the higher the PF, the less the decrease in the first-day return in the PF. Moreover, regression analysis suggests that among several possible negative relations, the best fit is found when regressing underpricing is on the reciprocal of the PF. As we show, this relation is consistent with the assumption of firms allocating a fixed amount for underpricing. The fit is not as good under linear or quadratic regressions, supporting the conjecture that firms allocate a fixed amount of money to underpricing. Despite the extensive existing literature about underpricing the reasons for underpricing are still unclear. In particular, the high variation in underpricing is hard to explain. Our findings are important

1

Of course, this amount is proportional to the firm’s size, and should be thought of as a fraction of the pre-IPO firm value.

page 180

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

b3964-ch08

page 181

181

because they have the potential to explain this variation. Understanding and being able to predict the level of underpricing is important to firms, investors and regulators. We start by showing that the assumption that firms allocate a fixed amount of money to underpricing results in a linear relation between underpricing and the reciprocal of the PF. We then investigate the empirical relation between underpricing and the PF, using a sample of 1907 IPOs that took place between 1996 and 2008. We first partition the sample into five PF trenches by PF size and show that underpricing is lower, the higher the PF trench, consistent with the earlier literature. Furthermore, this relation seems to be non-linear. We then turn to regression analysis. We regress underpricing on the PF controlling for other IPO characteristics (primary-tosecondary shares issued ratio and overallotment), for operating performance variables and for firm size and leverage. While the relation is generally negative and significant, when we regress underpricing on the reciprocal of the public float (1/PF) the relation is positive and substantially more significant. The relation is less significant when we regress underpricing on the square of the public float PF 2 . In sum, the best fit is found when underpricing is regressed on 1/PF. The rest of this chapter is organized as follows: Section 2 reviews the relevant literature. In Section 3, our hypothesis is developed. Section 4 describes data and methodology, and Section 5 reports our results. Section 6 concludes.

2.

Related Literature

It is well known that IPOs are, on average, underpriced. That is, the first-day abnormal return is positive. Earlier work documenting this underpricing includes Logue (1973), Ibbotson and Jaffe (1975), Rock (1986), Tinic (1988). Ibbotson and Ritter (1995) as well as Ritter and Welch (2002) surveyed this literature. Ritter (2003) raises a number of relevant issues such as the stage of a company’s lifecycle at which it is optimal to undertake an IPO, why IPO volume varies so dramatically across time and countries, and whether the supply of shares available to the public in an IPO affects the share price. One empirical regularity that evolved in later years is that the first-day

November 4, 2020

182

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Finance: A Novel Approach

return is negatively related to the PF (Habib and Ljungqvist, 2001; Bradley and Jordan, 2002; Ofek and Richardson, 2003; Loughran and Ritter, 2004). While our approach is suggesting that the negative relation is the result of a behavioral property — firms allocating a fixed amount for underpricing — there are several other reasons suggested in the literature. In particular, the negative correlation between the size of the PF and the first-day return of IPO stock might be explained by the downward sloping demand curve and information effects. Consider first the effects of a downward sloping demand curve. Bradley and Jordan (2002) show that share overhang, which they define as the ratio of retained shares to the public shares (and which is inversely related to the PF), predicts first-day returns for the period 1990– 1999. Habib and Ljungqvist (2001) provide similar findings for IPO size measured by gross proceeds.2 Second, consider the effects of information. Because in an IPO, ownership is being transferred from informed insiders to uninformed outsiders, the standard signaling arguments apply; that is, the larger the PF, the more negative the information revealed and hence the smaller the first-day return. For the signaling argument, see, for example, Leland and Pyle (1977) and for an application to IPOs, see Grinblatt and Hwang (1989). Chemmanur (1993) shows that the first-day return is related to information asymmetry and induces information production after the IPO to reduce information asymmetry. In general, the literature documents negative abnormal long-run returns in the post-IPO period (Ritter and Welch, 2002). Interestingly, a number of authors suggest that these negative returns are the result of measurement errors (e.g., Brav, 2000) and sample period selection (e.g., Ritter and Welch, 2002). In addition, once firm size and growth benchmarks are accounted for, there is only modest evidence of under-performance (see Brav and Gompers, 1997; Gompers and Lerner, 2003). While this literature criticizes the manner in which long-run returns are measured, our focus is on the impact of the PF on the returns, rather than on the return values, per se. As discussed earlier, our findings suggest that, generally, the firstday return decreases as the PF increases. We measure this relation 2

The coefficient of the PF is found to be negative (in their Table 2, p. 449).

page 182

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

b3964-ch08

page 183

183

against different benchmarks and different horizons and find that this negative and significant relation persists. Stoughton and Zechner (1998) show that ownership structure influences first-day returns through incentive and control considerations. However, they do not consider the impact on post-IPO longrun returns. Lowry (2003) finds that long-run returns are negatively correlated with the size of the IPO. Michel et al. (2014) consider the relation between the PF and long-run returns of IPO firms (up to 3 years after the IPO). They find a U-shape relation between the PF and long-run abnormal returns. Specifically, for low levels of PF, higher PF is associated with lower long-run returns; while for high levels of PF, higher PF is associated with greater long-run returns. They offer an agency costs explanation: at low levels of PF increasing, the PF exacerbates agency problems because it reduces their motivation due to lower ownership more than it enhances disciplining insiders, while at high levels of PF the ability of outsiders (new public investors) to govern and monitor increases with the PF. Several studies have investigated whether pre-IPO firm performance factors (e.g., net income, operating income) contribute to IPO underpricing. For example, Kao et al. (2009) show that IPO firms that report better pricing-period accounting performance have poorer first-day stock returns. Other papers study whether underpricing is related to institutional investor holdings. Aggarwal et al. (2002) investigate allocations to institutional investors in IPOs and find they are positively correlated with first-day returns, and that institutional allocation in underpriced issues is in excess of that explained by book-building theories alone. Jenkinson and Jones (2004) find that this ability to receive superior allocations in good IPOs is the result of institutional investors’ good relations with the investment banks. Aggarwal (2003) documents the involvement of institutional investors in flipping activity immediately after the IPO. Fernando et al. (2004) find that higher priced IPOs show a higher fraction of institutional investment.

3.

Hypothesis Development

We consider a firm with N0 shares outstanding that wishes to issue Nn new shares. Furthermore, the firm (old shareholders) wants to

November 4, 2020

184

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

page 184

Behavioral Finance: A Novel Approach

spend a fraction c of the pre-issue firm value on underpricing. Given the pre-IPO firm value, this results in a fixed amount C being spent. We further assume the transaction itself does not create or destroy value, and that only primary shares are issued. The analysis assuming secondary shares are issued is similar and yields similar results. Empirically (as is also the case in our sample), the vast majority of shares issued are primary shares. Let P0 and P1 denote share value before and after the issue, respectively, and let S denote the issue price. Underpricing is defined as the percentage increase in the post-issue price P1 relative to the issue price S, or (P1 − S)/S, and note that positive underpricing implies S < P1 . By law of value conservation, the terminal value is the sum of the initial firm value plus the funds the new investors provide to the firm, or P0 × N0 + S × Nn = P1 × (N0 + Nn )

(1)

Assuming original shareholders lose a fixed amount that represents a fraction c of the per-IPO value on underpricing, then C = cN0 P0 and N0 × (P0 − P1 ) = C

(2)

Because this is a zero sum game, C will also be the new shareholders gain. Nn × (P1 − S) = C

(3)

Combining (2) and (3), we can write Nn × (P1 − S) = N0 × (P0 − P1 ) = C and note it must be the case that S < P1 < P0 . We can now rewrite (3) as   C 1 (4) =C P1 − S = Nn Nn Define the PF as the ratio of the shares issued to total outstanding shares after the issue.

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Characteristics of IPO Underpricing

page 185

185

Then PF =

Nn (Nn + N0 )

(5)

which can be rearranged to       1 1 1 1 ∗ − = Nn N0 PF N0 Substitution of (6) into (4) gives:       1 1 1 − P1 − S = C N0 PF N0    1 − PF C P1 − S = N0 PF

(6)

or (7)

Using (2) we can write  P1 = P0 −

C N0

 (8)

and using (4) and (8) we can write     C C − S = P0 − N0 Nn     1 1 + = P0 − C N0 Nn   N0 + Nn = P0 − C N0 Nn and using (5) we can rearrange to     1 C S = P0 − N0 PF We can write underpricing as (P1 − S) UP = = S



P1 S

 −1

(9)

November 4, 2020

186

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

page 186

Behavioral Finance: A Novel Approach

and using (7) and (9) we can write  UP =

C N0



1−P F PF



    P0 − NC0 P1F  1−P F

=  P0 N0 P F  1 − PF C  1−P F =  P0 N0PPFF −C P F ∗C

 1−P F =

C

P0 N0 P F − C

Upon rearrangement we can write 1 − PF U P =  P 0 N0 PF − 1 C Consider the denominator of (10). Assuming approximate (10) as

(10)  P 0 N0 C

P F  1 we can

K 1 − PF K(1 − P F ) = −K U P =  P 0 N0 = PF PF PF C

(11)

where K = P0CN0 . Equation (11) gives a linear relation between UP in 1 P F . Thus, our analysis suggests that if firms dedicate a fixed amount C for underpricing, the underpricing should be linearly related to the reciprocal of the PF.3 Numerical example: Let N0 = 10 and P0 = 10, and suppose the firm chooses to allocate C = 10 to underpricing, and suppose first that the firm decides to issue Nn = 5 shares. Using (2) it must set 3

To see that the approximation is OK, consider the average PF at  30%  about and UP at about 18%, then using (10) gives a value of about 5 for P0CN0 PF.

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Characteristics of IPO Underpricing

page 187

187

an issue price S such that that is P1 = 9,

10 × (10 − P1 ) = 10,

and using (3) this issue price is such that 5 × (9 − S) = 10,

or S = 7.

5 Note that S < P1 < P0 as 7 < 9 < 10, and P F = 15 = 33%. Underpricing is P1 − S = 9 − 7 = 2, and in percentage 27 = 28.6%. Equivalently, using (7), underpricing is      1 C C − P1 − S = N0 PF N0      1 10 10 − = 3 − 1 = 2, = 10 0.33 10

and in percentage, (P1S−S) = 27 = 28.6%. Now suppose, instead that the firm wants to issue Nn = 10 shares. Then using (2) it must set an issue price such that 10 × (10 − P1 ) = 10,

that is still P1 = 9

and using (3) that issue price is 10 × (9 − S) = 10,

or S = 8.

Note that still S < P1 < P0 as 8 < 9 < 10, and P F = 10 20 = 55%. Underpricing is P1 − S = 9 − 8 = 1, and in percentage 18 = 12.5%. Equivalently, using (7) underpricing is           1 10 C 1 10 C = − − = 2−1 = 1, P1 −S = N0 PF N0 10 0.5 10 and in percentage, (P1S−S) = 18 = 12.5%. Figure 1 depicts a graph of underpricing as a function of the PF. While the data are noisy, as can be observed, the figure does suggest that the relation is decreasing and that the decline is diminishing, consistent with our hypothesis.

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

page 188

Behavioral Finance: A Novel Approach

188

PF vs Underpricing 8 7 6

Underpricing

5 4 3 2 1 0 0

0.2

0.4

-1

0.8

1

1.2

PF

Figure 1:

4.

0.6

Public Float vs. Underpricing

Data and Methodology

Our firm-level data come from the merger of three databases. The IPO sample was obtained from the Securities Data Company (SDC) database for the period January 1, 1996 through December 31, 2008.4 Data for calculating IPO characteristics (primary and secondary shares issued, number of outstanding shares before and after the IPO) are taken from the SDC database and from filings with the Securities and Exchange Commission (SEC) (424B filings and their updates, and first financial statements to the SEC, i.e., 10Q and 10K forms). Price data are taken from the Center for Research in Security Prices (CRSP), and operating performance data are taken from 4 We start in 1996 because before then data availability for this study are limited. 2008 was used as the final year of the sample for several reasons. First, in the years immediately following 2008 there were very few IPOs. But possibly the most compelling reason is that we found numerous errors in the SDC database and as described in Section 4.2 we had to manually reconstruct the entire database. Because of this lengthy and time-consuming process, it was decided to reasonably limit the size of the database, cutting the database off at the beginning of the great recession beginning in 2009.

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

b3964-ch08

page 189

189

Compustat. CRSP and Compustat data are obtained through the WRDS platform. From the initial sample of all IPOs in the SDC for the period of study we removed all utility and financial firms (see, for example, Field and Karpoff, 2002), resulting in a sample size of 2,119 firms. Of this sample, 52 firms were missing price data on CRSP and 39 firms were missing Compustat data. Following earlier studies (e.g., Loughran and Ritter, 1995; Eckbo et al., 2007), 74 ADRs were also dropped, resulting in 1,954 firms. Due to missing and erroneous institutional investor data, 47 additional firms were dropped, resulting in the final sample of 1,907 firms. 4.1.

IPO Characteristics

We first measure three characteristics of all the IPOs in our sample that are likely to be related to underpricing. The first is our variable of interest, the PF, which is the ratio of the number of shares sold to the public in the IPO to the number of shares outstanding right after the issue, that is, the fraction of shares that is transferred to the public in the IPO (e.g., Bradley and Jordan, 2002; Michel et al., 2020A).5 The other two are IPO characteristics that have been shown to be related to underpricing and for which we control. One characteristic is the ratio of primary shares issued to total shares sold in the IPO (PRIM). Primary shares are new shares that are issued to the public in the IPO. The proceeds from the sale of primary shares become part of the firm’s cash assets. The second characteristic is the ratio of the rest of the shares sold to the total shares sold in the IPO. These shares are known as secondary shares and are existing shares that are sold in the IPO by pre-IPO shareholders (the entrepreneur, angel investors, institutions, etc.). Unlike primary shares, the sale of secondary shares does not raise funds for the company. Both primary and secondary share sales, however, reduce original shareholder ownership (e.g., Habib and Ljungqvist, 2001; Ljungqvist and Wilhelm,

5

Bradley and Jordan (2002) consider share overhang, which is the ratio of retained-to-sold shares. Share overhang is, closely, the reciprocal of the PF (specifically, overhang is (1 − P F )/P F ).

November 4, 2020

190

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Finance: A Novel Approach

2003; Brau et al., 2007; Michel et al., 2020B).6 Thus, the variable PRIM measures the nature of the funds involved in the IPO in terms of raising funds for the firm vs. transferring funds to pre-issue shareholders.7 The other IPO characteristic for which we control is overallotment (OVERALLOT), which is the ratio of overallotment shares to total shares issued. SDC’s New Issue database reports intended number of shares for sale as filed with the SEC, but we have found this data to be erroneous. Moreover, it does not distinguish between primary and secondary shares (see Habib and Ljungqvist, 2001). Thus, after obtaining the sample of IPO firms for the sample period, we turned to retrieving data on shares from the prospectus and registration forms directly (424B4 and S1 filings, respectively). These filings include pre-IPO information and the firm’s intentions regarding the sale of primary and secondary shares (see also Loughran and Ritter (2004) on retrieving primary and secondary shares from 424B4 forms). However, when we cross-checked this data with the first financial reports published after the issue (10Q and 10K forms), we learned that often the intentions declared in the prospectus and registration statement do not reflect what eventually happened. That is, firms often increase or reduce the number of primary and secondary shares sold after they have filed the 424B4 and S1 forms. Thus, we have corrected the registration forms’ data using the 10Q and 10K forms (the first report after the issue) to reflect what actually happened. The 424B4 filings are updated several times before the offering. In fact, we found that even the last form filed often reports intended primary and secondary share figures that are significantly different from what was

6

Habib and Ljungqvist (2001) and Ljungqvist and Wilhelm (2003) consider the “participation ratio” and “dilution ratio” of the secondary/pre-IPO and primary/pre-IPO shares, respectively. Most firms, however, sell both primary and secondary shares, and both ratios reflect reduction of original shareholder ownership (though the sale of secondary shares is more ownership reducing). Given that our focus is ownership structure and operating performance, we prefer instead to consider PF, which captures reduction in original shareholder ownership, and primary/total sold, which captures the fund-raising nature of the IPO. 7 The variables PF and PRIM include overallotment shares. However, the correlation of these variables with overallotment is low, and excluding overallotment shares does not alter our qualitative findings.

page 190

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

b3964-ch08

page 191

191

eventually sold. Nevertheless, we still use the prospectus to get information such as pre-issue shares, which is sometimes missing in the financial reports (10Q and 10K) and which we need for calculating the IPO characteristics. Table 1 reports summary statistics of the IPO sample by year and for the complete sample (bottom row). Column (1) reports the number of IPOs per year over the sample period 1996–2008, and column (2) reports these as a fraction of the total sample. Columns (3) and (4) report the average primary and secondary shares sold as a fraction of the post-IPO outstanding shares, respectively. Column (5) reports averages of PF, the PF, defined as the ratio of shares sold in the IPO to total post-issue outstanding shares. This is the sum of columns (3) and (4). Column (5) indicates that the PF has been stable at about 30% over the years. Column (6) reports averages of PRIM, which is the ratio of primary shares issued to total shares sold. PRIM is also the complement to unity of the fraction of secondary shares sold out of total shares sold (that is, PRIM + secondary divided by total = 1). Column (6) suggests that the vast majority of shares issued are primary shares, although over the years the fraction of primary shares issued has decreased slightly from about 95% to about 90%, and the fraction of secondary shares issued has increased accordingly. Column (7) reports OVERALLOT, the average of overallotment shares as a fraction of total shares issued, and indicates it averages 8.15%. We note that overallotment shares can be either primary shares or secondary shares or both. Accordingly, they are included in Columns (3)–(5), (although they are also reported separately in Column (7)). 4.2.

Control Variables and Underpricing

In addition to controlling for the IPO characteristics, primary-tosecondary ratio and overallotment, we control also for operating performance. We measure operating performance using four different common measures.8 The first three are operating income-based measures: operating return on assets, ROA, measured as operating

8

The literature commonly uses these measures or similar variations. See, for example, Jain and Kini (1994), Mikkelson et al. (1997), Grullon and Michaely (2004), Cornett et al. (2007) and Gu and Hackbarth (2013).

21 15 6 15 9 2 3 3 7 6 6 7 1 100

(5)

PF = Public Float = Total Primary Shares Secondary Shares Shared Issued/Post-IPO Sold/Post-IPO Sold/Post-IPO Shares Shares Shares Outstanding Outstanding Outstanding 0.2998 0.2964 0.2761 0.2359 0.2159 0.2581 0.2469 0.2874 0.2520 0.3144 0.2768 0.2495 0.2781 0.2683

0.0205 0.0337 0.0243 0.0187 0.0098 0.0296 0.0524 0.0866 0.0561 0.0526 0.0530 0.0523 0.0591 0.0422

0.3203 0.3301 0.3005 0.2546 0.2257 0.2877 0.2993 0.3740 0.3081 0.3670 0.3298 0.3019 0.3373 0.3105

(6)

(7)

PRIM = OVERALLOT = Primary Shares Actual Issued/Total Overallotment to Shares Sold Total Shares Issued 0.9450 0.9218 0.9138 0.9504 0.9676 0.9313 0.8896 0.8272 0.8643 0.8695 0.8663 0.8483 0.7974 0.8917

0.0670 0.0847 0.0732 0.0891 0.0919 0.0860 0.0803 0.0805 0.0754 0.0832 0.0926 0.0804 0.0749 0.0815

b3964-ch08

Note: This table describes summary statistics of the IPO sample by year and for the complete sample (bottom row). Column (1) reports the number of IPOs per year over the sample period 1996–2008 and column (2) reports this number as a fraction of the total sample. Columns (3) and (4) report the average of primary and secondary shares sold as a fraction of the post-IPO outstanding shares, respectively. Column (5) reports averages of the PF variable (public float), which is the ratio of total shares sold to total post-issue outstanding shares. This column is the sum of columns (3) and (4). Column (6) reports averages of PRIM, which is the ratio of primary shares issued to total shares sold. PRIM is also the complement to one of the fraction of secondary shares sold out of total shares sold. This column is also the ratio of column (3) to column (5). Column (7) reports averages of OVERALLOT, which is the ratio of overallotment shares to total shares issued. (In the table, PF includes OVERALLOT.)

Behavioral Finance: A Novel Approach – 9in x 6in

404 295 111 277 181 39 49 48 133 115 115 124 16 1,907

% (of total)

(4)

Behavioral Finance: A Novel Approach

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Total/Average

Number of IPOs

(3)

10:13

Year

(2)

November 4, 2020

(1)

Summary statistics of IPO characteristics

192

Table 1:

page 192

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

b3964-ch08

page 193

193

income before depreciation divided by end-of-year total assets, calculated using Compustat data items OIBDPQ and ATQ; operating return on assets-less-cash, RO ALC, calculated using Compustat data items OIBDPQ, ATQ and CHEQ; and the ratio of operating performance to revenue, ROS, calculated using Compustat data items OIBDPQ and REVQ. Here we use quarterly Compustat data for the construction of these variables for better resolution relative to the IPO date (see also Lie, 2005). That is, Compustat quarterly variables OIBDPQ and REVQ are aggregated yearly relative to the quarter of the IPO. The fourth operating performance measure is market-to-book (MtoB), which is the ratio of the market value of the firm’s equity to the book value of the firm’s equity. Here, market value is calculated as number of shares outstanding (Compustat data item CSHOQ) times share price (Compustat data item PRCCQ), and for book value we use Compustat data item SEQQ.9 ,10 We include two additional control variables in the analysis, size and leverage, as follows: For size we use lnM V , the natural log of the firm’s equity calculated as number of shares outstanding (Compustat data item CSHOQ) times share price (Compustat data item PRCCQ). Leverage (LEVER) is calculated as the ratio of long-term debt (Compustat data item DLTTQ) to total assets (Compustat data item ATQ) at the last quarter prior to the IPO.11 All variables are winsorized at 0.5% (on each side, high and low values). In Table 2, Columns (1) through (4) report summary statistics of the performance and control variables as well as the variables used to calculate them. Columns (5) through (8) of Table 2 report first-day abnormal return statistics by year and for the complete sample. The statistics for na¨ıve returns (not tabulated) are similar.

9

Our definition of MtoB follows that used in Jain and Kini (1994). Other studies calculate MtoB as market value of equity plus book value of debt divided by book assets (e.g., Lie, 2005). 10 For more on our four measures and their applicability see, for example, Barber and Lyon (1996), Jain and Kini (1994), Grullon and Michaely (2004) and Lie (2001, 2005). 11 Some studies include the current portion of long-term debt when calculating leverage. This variable is similar to the one we use because it tends to be small relative to long-term debt.

Summary statistics of operating performance and first-day returns (underpricing) Operating Performance (2)

Average Median

Stdev

N

1157.8 1.2122 0.3353 239.10 1672.3 131.92 1443.6 0.5763 1.9719 11.267

1862 1862 1850 1402 1860 1858 1810 1388 1388 1342

(5)

(6)

(7)

Average (%) Median (%) Stdev (%) 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Total

17.78 14.91 28.11 62.86 45.86 37.26 9.65 12.76 10.20 10.17 9.69 14.74 20.12 25.56

10.73 9.80 7.48 36.51 23.70 16.78 9.87 9.47 6.92 5.58 6.03 7.27 −0.19 11.26

31.09 20.80 67.99 85.04 84.19 90.39 15.20 15.36 25.39 16.29 28.05 23.29 119.84 54.99

(8) N 404 295 111 277 181 39 49 48 133 115 115 124 16 1907

b3964-ch08

Note: This table describes summary statistics of the IPO sample’s operating performance variables and Day 1 abnormal returns. Column (1)–(4) report the Average, Median, Stdev, and sample size of the complete sample for market value, leverage, and for operating performance and variables used to calculate them. The performance variables are taken from the end of the year prior to the IPO where available. If the previous year is unavailable, the first quarter after the IPO is used. Columns (5)–(8) report the Average, Median, Stdev of Day 1 abnormal returns, and sample size for the IPOs in each of the sample years. Abnormal returns are calculated using a model that is based on the three Fama-French factors and the Carhart (1997) momentum factor. The bottom line reports these statistics for the complete sample.

Behavioral Finance: A Novel Approach – 9in x 6in

579.52 252.45 5.5772 5.5312 0.2462 0.0951 37.058 3.1610 294.30 35.429 23.564 5.2910 293.13 39.153 0.0927 −0.1208 −0.6114 0.1020 −2.4056 0.0712

(4)

Behavioral Finance: A Novel Approach

MV LN of MV LEVERQ OIBD ATQ CHEQ REV OIBD/ATQ OIBD/(ATQ-CHEQ) OIBD/REV

(3)

10:13

(1)

Day 1 Abnormal Returns

November 4, 2020

194

Table 2:

page 194

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Characteristics of IPO Underpricing

4.3.

page 195

195

First-day Return and the PF

Table 3 reports first-day return statistics. Panel 3A reports results for the complete sample. Columns (1)–(3) report, respectively, the mean, median and standard deviation of the first-day na¨ıve return. Columns (4)–(6) report these statistics for the first-day abnormal Table 3:

First-day return and the public float

Panel 3A: First-day returns — complete sample Na¨ıve Return

Complete Sample

Abnormal Return

(1)

(2)

(3)

(4)

(5)

(6)

N

Mean (%)

Median (%)

Std (%)

Mean (%)

Median (%)

Std (%)

1907

25.12

11.54

45.36

25.56

11.26

54.99

Panel 3B: First-day returns sorted by public float Na¨ıve Return (1) Public Float Range (%) 0–20 20–40 40–60 60–80 80–100

N 405 1145 286 46 25

(2)

Mean Median (%) (%) 47.30 21.22 12.71 13.25 8.42

20.39 10.67 8.72 10.08 3.93

(3) Std (%) 68.64 37.68 18.36 19.18 17.80

Abnormal Return (4)

(5)

Mean Median (%) (%) 46.79 21.83 13.91 13.25 8.37

20.73 10.61 8.33 8.32 4.48

(6) Std (%) 79.33 48.48 30 19.40 17.56

Note: This table reports the average first-day returns to IPO investors defined as the returns from the offer price to the close price on the day of the IPO. Panel 3A reports the first-day returns for the complete sample of 1907 IPOs. Columns (1)–(3) describe the na¨ıve returns earned by IPO investors on the first-day (mean, median, and standard deviation, respectively). Columns (4)–(6) describe the abnormal returns calculated using a model that is based on the three Fama-French factors and the Carhart (1997) momentum factor. Panel 3B reports the average firstday return as in Panel 2A for different public float size groups in the sample of 1907 IPOs. The public float percentage ranges are 0–20%, 20–40%, 40–60%, 60–80% and 80–100%. To control for outliers, returns are winsorized at 0.5% (both panels).

November 4, 2020

196

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Finance: A Novel Approach

returns where abnormal return (henceforth “abnormal return” or “alpha”) is calculated using a 4-factor model which includes the three Fama-French (1993) factors and the Carhart (1997) momentum factor. The mean and median returns (na¨ıve and abnormal) are investigated for holding periods of one day (the offer price through the end of the day of the IPO). As indicated in Column (1), the mean na¨ıve return is 25.12%. The mean abnormal return earned on day 1 (Column (4)) is 25.56%. The return distributions are highly skewed, with median values for the na¨ıve and abnormal return distributions of 11.54% and 11.26%, respectively, reflecting the skewness of the return data. The first-day positive abnormal IPO returns presented here are consistent with earlier findings (see, for example, Ibbotson and Ritter, 1995; Ritter, 2003; Ritter and Welch, 2002).12 Panel 3B illustrates the relation between the percentage of firm ownership sold to the public (the PF) and alpha on the day of the IPO. We sort the sample of IPOs into five deciles by PF percentage. Each row in the panel represents a different range of PF in ascending order. Columns (1) and (2) of Panel 3B report the mean and median first-day na¨ıve return for the different PF ranges, and Column (3) reports the standard deviations. Columns (4)–(6) report these results for the 4-factor adjusted abnormal returns (alpha). Focusing on the 4-factor adjusted abnormal return, the results show that for the day of the IPO, in general, the higher the PF, the lower the abnormal return. Specifically, on the day of the IPO, the mean alpha (Column 4) in the first range (0%–20%) is 46.79% and declines to 21.83% in the range 20%–40%, it continues to decline and reaches 8.37% in the range 80%–100%. As can be observed, the decline when moving to a higher PF trench is generally diminishing. Similarly, a generally decreasing trend is observed for the median alpha as the PF percentage increases. The skewness of the results can be observed in each of the categories by considering the relationship between mean and median values.

12 While Ritter and Welch (2002) report simple one-day returns, we also report abnormal returns. However, for one-day return calculations, the difference between na¨ıve returns and abnormal returns is insignificant, as also reflected in Table 3.

page 196

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

5.

b3964-ch08

page 197

197

Regression Results

Table 4 reports regression results where the dependent variable is the first-day return (underpricing) in percentage and the explanatory variable is the public float PF in percentage with different controls. That is, in Table 4 we include the PF in the regressions in the traditional manner as was included in earlier studies (Habib and Ljungqvist, 2001; Bradley and Jordan, 2002). 13 In regression (1), only assets and leverage in the year ending just before the issue are controlled for. In regression (2), the ratio of primary-to-total shares issued (PRIM), and overallotment (OVERALLOT) are also added to the regression. In regressions (3)–(5), we include controls also for operating performance. The operating performance measures are OIBD/ATQ, OIBD/(ATQ-CASH) and OIBD/REV. As Table 4 shows, PF is negative and significant at the 1% level in all these regressions except regression (5) where we measure operating performance with OIBD/REV. In regressions (6)–(10), we repeat the analysis in regressions (1)–(5) but control for firm size using post-issue market value instead of pre-issue total assets. As the table shows, in none of regressions (5)–(10) is PF significant. In sum, while there is some support for the negative relation between the first-day return and the PF, once we control for market value, the relation becomes insignificant. These results confirm our observation in Table 3 that increasing the PF percentage generally leads to a reduction in the marketadjusted abnormal return on day 1. This negative relation between PF and first-day return is consistent with Bradley and Jordan (2002, Table 2), who document a positive relation between first-day returns and share overhang (the ratio of retained shares to shares sold, which is inversely related to the PF). Similar findings appear in Habib and Ljungqvist (2001, Table 3), who also report a negative and significant relation between leverage and underpricing, consistent with our findings. Table 5 repeats the analysis in Table 4 but now the explanatory variable is 1/P F instead of PF. That is, the table reports the results

13

These studies actually control for share overhang with is the complement to 1 of the PF, but in terms of the analysis this is equivalent.

ln ATQQ−1

0.3152 (6.776) −0.4198∗∗∗ (−4.679) 0.0061 (0.751)

(3) ∗∗∗

0.3208 (3.539) −0.5254∗∗∗ (−5.787) −0.0011 (−0.138)

(4) ∗∗∗

0.4194 (4.046) −0.5498∗∗∗ (−5.342) −0.0073 (−0.731)

(5) ∗∗∗

0.4322 (4.176) −0.5618∗∗∗ (−5.458) −0.0127 (−1.334)

ln MVQ1

(7) ∗∗∗

−0.4049 −0.5835 (−3.049) (−7.885) −0.1800 0.0126 (−1.593) (0.137) 0.1135∗∗∗ (8.309) 0.1596∗∗∗ (14.332)

(8) ∗∗∗

−0.6294 (−5.606) −0.0645 (−0.669)

0.1392∗∗∗ (11.883)

−0.0042 (−0.162) 0.0122∗ (1.697)

−0.0739∗ −0.0764∗ (−1.824) (−1.886) −0.1624∗∗ −0.1454∗ (−2.074) (−1.862)

(9) ∗∗∗

−0.4260 (−3.276) −0.1438 (−1.297)

0.1198∗∗∗ (8.878) −0.0577∗∗∗ (−2.396)

(10) ∗∗∗

−0.4049∗∗∗ (−3.049) −0.1800 (−1.593)

0.1133∗∗∗ (8.440)

0.1135∗∗∗ (8.309)

−0.4088 (−3.136) −0.1669 (−1.502)

0.0004 (0.056) 0.0008 (0.728) −0.1481∗∗∗ −0.1824∗∗∗ −0.1544∗∗∗ (−3.747) (−5.173) (−4.409) 0.0236 0.0787 (0.302) (1.183)

−0.1306∗∗∗ −0.1396∗∗∗ (−3.410) (−3.632) −0.0140 0.0169 (−0.180) (0.218)

0.0008 (0.728) −0.1481∗∗∗ (−3.747) 0.0236 (0.302)

b3964-ch08

OIBD/ ATQYear−1 OIBD/(ATQ− CHEQ)Year−1 OIBD/ REVYear−1 LEVERQYear−1 −0.1276∗∗∗ −0.0905∗∗∗ (−3.352) (−2.438) PRIM −0.0957 (−1.407)

(6) ∗∗∗

Behavioral Finance: A Novel Approach – 9in x 6in

PF

(2) ∗∗∗

Behavioral Finance: A Novel Approach

Intercept

10:13

First-day return as a function of IPO characteristics and pre-IPO operating performance using PF (1)

November 4, 2020

198

Table 4:

page 198

1.7985∗∗∗ (8.582) 0.1777 1379

1.2487∗∗∗ (5.629) 0.2136 1298

0.2158 1814

1.3668∗∗∗ (7.240) 0.2377 1814

1.2559∗∗∗ (5.877) 0.2167 1359

1.2618∗∗∗ (5.892) 0.2133 1359

1.2487∗∗∗ (5.629) 0.2136 1298

Behavioral Finance: A Novel Approach – 9in x 6in

Note: This table reports results of regressions in which the dependent variable is the four-factor-adjusted abnormal return for the first day after the IPO. The independent variables are IPO characteristics and pre-IPO operating performance, where “pre-IPO” is defined as the year ending in the last reported quarter just prior to the IPO (Y 0). PF is the public float measured as the ratio of shares sold in the IPO to shares outstanding after the IPO. ln ATQ is the natural log of ATQ, which is the Compustat quarterly total asset variable from the last quarter of the year prior to the IPO. ln MV is firm size, calculated as the natural log of the firm’s market value of equity. Here, market value of the firm’s equity is calculated as number of shares outstanding (Compustat data item CSHOQ) times share price (Compustat data item PRCCQ). OIBD/ATQ is operating return on assets calculated as operating income before depreciation divided by end-of-year total assets (where OIBD is calculated by aggregating Compustat quarterly data items OIBDPQ over four fiscal quarters). OIBD/(ATQ-CHEQ) is operating return on assets-less-cash, calculated using Compustat data items OIBDPQ, ATQ and CHEQ; OIBD/REV is operating performance to revenue ratio, calculated using Compustat data items OIBDPQ and REVQ. LEVERQ is leverage calculated as the ratio of long term debt (Compustat data item DLTTQ) to total assets (Compustat data item ATQ) in the last quarter of the relevant year. PRIM is the ratio of primary shares issued to total shares floated to the public in the IPO. OVERALLOT is the ratio of overallotment shares to total shares issued. All regressions include year fixed effects. t-statistics are reported in parentheses. Coefficients marked with ∗∗∗ , ∗∗ and ∗ are statistically significant at the 1%, 5% and 10% level, respectively.

10:13

0.1282 1849

1.8056∗∗∗ (8.606) 0.1760 1379

Behavioral Characteristics of IPO Underpricing

Adj R2 N

1.9836∗∗∗ (10.647) 0.1800 1849

November 4, 2020

OVERALLOT

b3964-ch08

199 page 199

November 4, 2020

200

10:13

Table 5: First-day return as a function of IPO characteristics and pre-IPO operating performance using (1/PF)

1/PF ln ATQQ−1

0.0595 (1.476) 0.0363∗∗∗ (6.930) 0.0009 (0.114)

ln MVQ1 OIBD/ ATQYear−1 OIBD/(ATQ− CHEQ)Year−1 OIBD/ REVYear−1 LEVERQYear−1 −0.1212∗∗∗ (−3.221) PRIM

(3)

(4)

(5)

(6) ∗∗∗

−0.0137 0.0783 0.0837 −0.5038 (−0.179) (0.891) (0.955) (−4.848) ∗∗∗ ∗∗∗ ∗∗∗ 0.0404 0.0373 0.0375 0.0152∗∗∗ (7.847) (6.523) (6.565) (2.475) −0.0058 −0.0125 −0.0171∗ 0.1086∗∗∗ (−0.721) (−1.243) (−1.802) (7.944) 0.1522∗∗∗ (13.466) −0.0042 (−0.163) 0.0106 (1.479) 0.0008 (0.701) −0.0875∗∗∗ −0.0748∗ −0.0774∗ −0.1461∗∗∗ (−2.389) (−1.864) (−1.930) (−3.726) −0.0657 −0.1211 −0.1042 0.0290 (−1.000) (−1.602) (−1.381) (0.387)

(7) ∗∗∗

−0.5788 (−9.922) 0.0089∗ (1.659)

0.1297∗∗∗ (11.030)

(8)

(9)

(10)

∗∗∗

∗∗∗

∗∗∗

−0.6472 −0.4991 −0.4978 −0.5038∗∗∗ (−7.306) (−4.906) (−4.882) (−4.848) 0.0142∗∗∗ 0.0147∗∗∗ 0.0153∗∗∗ 0.0152∗∗∗ (2.624) (2.450) (2.546) (2.475)

0.1136∗∗∗ (8.430)

0.1078∗∗∗ (8.044) −0.0580∗∗∗ (−2.415)

0.1086∗∗∗ (7.944)

−0.0002 (−0.023)

−0.1739∗∗∗ −0.1459∗∗∗ (−4.968) (−4.208) 0.0628 (0.980)

−0.1274∗∗∗ −0.1371∗∗∗ (−3.361) (−3.603) −0.0156 0.0184 (−0.209) (0.248)

0.0008 (0.701) −0.1461∗∗∗ (−3.726) 0.0290 (0.387)

Behavioral Finance: A Novel Approach

Intercept

(2)

Behavioral Finance: A Novel Approach – 9in x 6in

(1)

b3964-ch08 page 200

0.1403 1849

1.8487∗∗∗ (8.843) 0.1842 1379

1.8418∗∗∗ (8.817) 0.1855 1379

1.2884∗∗∗ (5.801) 0.2158 1298

0.2170 1814

1.4198∗∗∗ (7.546) 0.2404 1814

1.3027∗∗∗ (6.092) 0.2192 1359

1.3052∗∗∗ (6.091) 0.2158 1359

1.2884∗∗∗ (5.801) 0.2158 1298

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

Note: This table reports results of regressions in which the dependent variable is the four-factor-adjusted abnormal return for the first day after the IPO. The independent variables are IPO characteristics and pre-IPO operating performance, where “pre-IPO” is defined as the year ending in the last reported quarter just prior to the IPO (Y 0). PF is the public float measured as the ratio of shares sold in the IPO to shares outstanding after the IPO. Here we use the fraction 1 over PF. ln ATQ is the natural log of ATQ, which is the Compustat quarterly total asset variable from the last quarter of the year prior to the IPO. ln MV is firm size, calculated as the natural log of the firm’s market value of equity. Here, market value of the firm’s equity is calculated as number of shares outstanding (Compustat data item CSHOQ) times share price (Compustat data item PRCCQ). OIBD/ATQ is operating return on assets calculated as operating income before depreciation divided by end-of-year total assets (where OIBD is calculated by aggregating Compustat quarterly data items OIBDPQ over four fiscal quarters). OIBD/(ATQ-CHEQ) is operating return on assets-less-cash, calculated using Compustat data items OIBDPQ, ATQ and CHEQ; OIBD/REV is operating performance to revenue ratio, calculated using Compustat data items OIBDPQ and REVQ. LEVERQ is leverage calculated as the ratio of long term debt (Compustat data item DLTTQ) to total assets (Compustat data item ATQ) in the last quarter of the relevant year. PRIM is the ratio of primary shares issued to total shares floated to the public in the IPO. OVERALLOT is the ratio of overallotment shares to total shares issued. All regressions include year fixed effects. t-statistics are reported in parentheses. Coefficients marked with ∗∗∗ and ∗ are statistically significant at the 1% and 10% level, respectively.

10:13

Adj R2 N

1.9940∗∗∗ (10.786) 0.1921 1849

November 4, 2020

OVERALLOT

b3964-ch08

201 page 201

November 4, 2020

202

10:13

Table 6:

First-day return as a function of IPO characteristics and pre-IPO operating performance using P F 2

PF2 ln ATQQ−1

(3) ∗

0.2154 0.1552 (5.615) (1.878) ∗∗∗ −0.3156 −0.3936∗∗∗ (−3.096) (−3.834) 0.0085 0.0029 (1.052) (0.367)

(4) ∗∗∗

0.2543 (2.688) −0.4432∗∗∗ (−3.802) −0.0032 (−0.317)

(5) ∗∗∗

0.2643 (2.798) −0.4547∗∗∗ (−3.899) −0.0087 (−0.912)

ln MVQ1 OIBD/ ATQYear−1 OIBD/(ATQ− CHEQ)Year−1 OIBD/ REVYear−1 LEVERQYear−1 −0.1427∗∗∗ −0.1101∗∗∗ (−3.759) (−2.973) PRIM −0.0538 (−0.792)

(6) ∗∗∗

(7) ∗∗∗

−0.4664 −0.5778 (−4.050) (−9.263) −0.1735 0.0024 (−1.400) (0.024) 0.1172∗∗∗ (9.013) 0.1592∗∗∗ (14.787)

(8) ∗∗∗

−0.6569 (−6.741) −0.0479 (−0.456)

0.1410∗∗∗ (12.641)

−0.0077 (−0.294) 0.0113 (1.563)

−0.0915∗ (−2.261) −0.1241 (−1.583)

−0.0940∗∗∗ (−2.323) −0.1064 (−1.361)

(9) ∗∗∗

−0.4763 (−4.225) −0.1358 (−1.117)

0.1230∗∗∗ (9.572) −0.0585∗∗∗ (−2.428)

(10) ∗∗∗

−0.4664∗∗∗ (−4.050) −0.1735 (−1.400)

0.1170∗∗∗ (9.150)

0.1172∗∗∗ (9.013)

−0.4674 (−4.135) −0.1573 (−1.289)

0.0002 (0.026) 0.0008 (0.709) −0.1524∗∗∗ −0.1817∗∗∗ −0.1566∗∗∗ (−3.886) (−5.186) (−4.510) 0.0327 0.0842 (0.423) (1.282)

−0.1340∗∗∗ −0.1435∗∗∗ (−3.526) (−3.765) −0.0065 0.0258 (−0.085) (0.337)

0.0008 (0.709) −0.1524∗∗∗ (−3.886) 0.0327 (0.423)

Behavioral Finance: A Novel Approach

Intercept

(2) ∗∗∗

Behavioral Finance: A Novel Approach – 9in x 6in

(1)

b3964-ch08 page 202

0.1224 1849

1.7624∗∗∗ (8.367) 0.1676 1379

1.7542∗∗∗ (8.337) 0.1690 1379

1.2213∗∗∗ (5.551) 0.2132 1298

0.2158 1814

1.3560∗∗∗ (7.232) 0.2376 1814

1.2341∗∗∗ (5.824) 0.2164 1359

1.2365∗∗∗ (5.822) 0.2130 1359

1.2213∗∗∗ (5.551) 0.2132 1298

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

Note: This table reports results of regressions in which the dependent variable is the four-factor-adjusted abnormal return for the first day after the IPO. The independent variables are IPO characteristics and pre-IPO operating performance, where “pre-IPO” is defined as the year ending in the last reported quarter just prior to the IPO (Y 0). PF is the public float measured as the ratio of shares sold in the IPO to shares outstanding after the IPO, and has been squared for this regression. ln ATQ is the natural log of ATQ, which is the Compustat quarterly total asset variable from the last quarter of the year prior to the IPO. ln MV is firm size, calculated as the natural log of the firm’s market value of equity. Here, market value of the firm’s equity is calculated as number of shares outstanding (Compustat data item CSHOQ) times share price (Compustat data item PRCCQ). OIBD/ATQ is operating return on assets calculated as operating income before depreciation divided by end-of-year total assets (where OIBD is calculated by aggregating Compustat quarterly data items OIBDPQ over four fiscal quarters). OIBD/(ATQ-CHEQ) is operating return on assets-less-cash, calculated using Compustat data items OIBDPQ, ATQ and CHEQ; OIBD/REV is operating performance to revenue ratio, calculated using Compustat data items OIBDPQ and REVQ. LEVERQ is leverage calculated as the ratio of long term debt (Compustat data item DLTTQ) to total assets (Compustat data item ATQ) in the last quarter of the relevant year. PRIM is the ratio of primary shares issued to total shares floated to the public in the IPO. OVERALLOT is the ratio of overallotment shares to total shares issued. All regressions include year fixed effects. t-statistics are reported in parentheses. Coefficients marked with ∗∗∗ and ∗ are statistically significant at the 1% and 10% level, respectively.

10:13

Adj R2 N

1.9500∗∗∗ (10.423) 0.1716 1849

November 4, 2020

OVERALLOT

b3964-ch08

203 page 203

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Finance: A Novel Approach

204

when the dependent variable is the first-day return (underpricing) in percentage and the explanatory variable is the reciprocal of the PF with different controls. Again, in regression (1) only assets and leverage in the year before the issue are controlled for. In regression (2), the ratio of primary-to-total shares issued (PRIM) and overallotment (OVERALLOT) are also added; in regressions (3)–(5), we also include controls for operating performance. The operating performance measures are OIBD/ATQ, OIBD/(ATQ-CASH) and OIBD/REV, and in regressions (6)–(10) we repeat the analysis in regressions (1)–(5) but control for firm size using post-issue market value instead of pre-issue total assets. As can be observed 1/P F is positive and significant in all regressions, including regressions (5)– (10), where it was insignificant in Table 4, where the explanatory variable was PF. In sum, Table 5 suggests that 1/P F provided a better fit for the relation between the first-day return and the PF, consistent with our hypothesis. We next repeat the analysis reported in Tables 4 and 5 considering a quadratic relation (i.e., another non-linear relation), between underpricing and the PF. Specifically, we regress underpricing UP on the PF and the square of the PF (PF and P F 2 , respectively) with different controls. Table 6 reports our results. As the table shows the relation is negative, consistent with underpricing diminishing in the PF. However, as in Table 4 where a linear relation is investigated, here too the relation holds only in regressions (1)–(5), where we control for assets but not in regressions (6)–(10), where we instead control for market value. This suggests that the model in Table 5 that relates underpricing to 1/P F provides a better fit than a quadratic relation. Together, the findings in Tables 4–6 suggest that the reciprocal of the public float 1/P F provides the best fit. This is consistent with the hypothesis that firms dedicate a fixed amount of money for underpricing.

6.

Conclusion

We investigate the relation between first-day return of IPOs (underpricing) and the PF. While the earlier literature documents a negative relation, we further show that this relation is well approximated by assuming firms allocate a fixed amount (a fraction of

page 204

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

b3964-ch08

page 205

205

their pre-IPO value) for underpricing. As a result, in our regression analysis a positive and linear relation between underpricing and the reciprocal of the public float (i.e., 1/P F ) provides a better fit than a linear or a quadratic relation with the PF, consistent with the behavioral characteristic that firms allocate a fixed amount for underpricing. Acknowledgment We thank Itzik Venezia (the editor). We are grateful to Dan Negovan for excellent research assistance. Financial support from Back Bay Management Corporation, Henry Crown Institute of Business Research and Jeremy Coller Foundation is gratefully acknowledged. References Aggarwal, R. (2003). Allocation of initial public offerings and flipping activity. Journal of Financial Economics, 68, 111–135. Aggarwal, R., Prabhala, N., and Puri, M. (2002). Institutional allocation in initial public offerings: Empirical evidence. Journal of Finance, 57, 1421–1442. Barber, B. and Lyon, J. (1996). Detecting abnormal operating performance: The Empirical power and specification of statistical tests. Journal of Financial Economics, 41, 359–399. Bradley, D. and Jordan, B. (2002). Partial adjustment to public information and IPO underpricing. Journal of Financial and Quantitative Analysis, 37, 595–615. Brau, J., Li, M., and Shi, J. (2007). Do secondary shares in the IPO process have a negative effect on aftermarket performance? Journal of Banking and Finance, 31, 2612–2631. Brav, A. (2000). Inference in long-horizon event studies. Journal of Finance, 55, 1979–2016. Brav, A. and Gompers, P. (1997). Myth or reality? The long-run underperformance of initial public offerings: Evidence from venture and non venture capital-backed companies. Journal of Finance, 52, 1971–1821. Carhart, M. (1997). On persistence in mutual fund performance. Journal of Finance, 52, 57–82. Chemmanur, T. (1993). The pricing of initial public offerings: “A dynamic model with information production.” Journal of Finance, 48, 285–304.

November 4, 2020

206

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch08

Behavioral Finance: A Novel Approach

Cornett, M., Marcus, A., Saunders, A., and Tehranian, H. (2007). The impact of institutional ownership on corporate operating performance. Journal of Banking and Finance, 31, 1771–1794. Eckbo, E., Masulis, R., and Norly, O. (2007). Security offerings. In E. Eckbo (Ed.), Handbook of Corporate Finance: Empirical Corporate Finance, Chapter 6. North-Holland: Elsevier, 233–373. Fama, E. and French, K. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3–56. Fernando, C., Krishnamurthy, S., and Spindt, P. (2004). Are share price levels informative? Evidence from the ownership, pricing, turnover and performance of IPO Firms. Journal of Financial Markets, 7, 377–403. Field, L. and Karpoff, J. (2002). Takeover defenses of IPO firms. Journal of Finance, 57, 1857–1889. Grinblatt, M. and Hwang, C. (1989). Signaling and the pricing of new issues. Journal of Finance, 44, 393–420. Gompers, P. and Lerner, J. (2003). The really long-term performance of initial public offerings: The pre-Nasdaq evidence. Journal of Finance, 58, 1355–1392. Grullon, G. and Michaely, R. (2004). The information content of share repurchases. Journal of Finance, 59, 651–680. Gu, L. and Hackbarth, D. (2013). Governance and equity prices: Does transparency matter?” Review of Finance, 17, 1989–2033. Habib, M. and Ljungqvist, A. (2001). Underpricing and entrepreneurial wealth losses in IPOs: Theory and evidence. Review of Financial Studies, 14, 433–458. Ibbotson, R. and Jaffe, J. (1975). “Hot issue” markets. Journal of Finance, 20, 1027–1042. Ibbotson, R. and Ritter, J. (1995). Initial public offerings. In R. A. Jarrow, V. Maksimovic and W. T. Ziemba (Eds.), Handbooks in Operations Research and Management Science, Vol. 9, Chapter 30. North Holland, Amsterdam: Elsevier. Jain, B. and Kini, O. (1994). The post-issue operating performance of IPO firms. Journal of Finance, 49, 1699–1726. Jenkinson, T. and Jones, H. (2004). Bids and allocations in European IPO bookbuilding. Journal of Finance, 59, 2309–2338. Kao, J., Wu, D., and Yang, Z. (2009). Regulations, earnings management, and post-IPO performance: The Chinese evidence. Journal of Banking and Finance, 33, 63–76. Leland, H. and D. Pyle (1977). Information asymmetries, financial structure, and financial intermediation. Journal of Finance, 32, 371–387. Lie, E. (2001). Detecting abnormal operating performance: Revisited. Financial Management, 30, 77–91.

page 206

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Characteristics of IPO Underpricing

b3964-ch08

page 207

207

Lie, E. (2005). Operating performance following open market share repurchase announcements. Journal of Accounting and Economics, 39, 411–436. Ljungqvist, A. and Wilhelm, W. (2003). IPO pricing in the dot-com bubble. Journal of Finance, 58, 723–752. Logue, D. (1973). On the pricing of unseasoned equity issues: 1965–1969. Journal of Financial and Quantitative Analysis, 8, 91–103. Loughran, T. and Ritter, J. (1995). The new issues puzzle. Journal of Finance, 50, 23–51. Loughran, T. and Ritter, J. (2004). Why has IPO underpricing changed over time? Financial Management, 33, 5–37. Lowry, M. (2003). Why does IPO volume fluctuate so much? Journal of Financial Economics, 67, 3–40. Michel, A., Oded, J., and Shaked, I. (2014). Ownership structure and performance: Evidence from the public float in IPOs. Journal of Banking and Finance, 40, 54–61. Michel, A., Oded, J., and Shaked, I. (2020A). Institutional investors and firm performance: Evidence from IPOs. North American Journal of Economics and Finance, 51, 1010099, 1–19. Michel, A., Oded, J., and Shaked, I. (2020B). What determines institutional investors’ holdings in IPO firms? Forthcoming in International Review of Finance. Mikkelson, W., Parch, M., and Shah, K. (1997). Ownership and operating performance of companies that go public. Journal of Financial Economics, 44, 281–307. Ofek, E. and Richardson, M. (2003). DotCom mania: The rise and fall of Internet stock prices. Working Paper, NBER. Ritter, J. (2003). Investment banking and securities issuance. In G. Constantinides, M. Harris, R. Stulz (Eds.), Handbook of the Economics of Finance. North Holland: Amsterdam. Ritter, J. and Welch, I. (2002). A review of IPO activity, pricing, and allocations. Journal of Finance, 57, 1795–1828. Rock, K. (1986). Why new issues are underpriced. Journal of Financial Economics, 15, 187–212. Stoughton, N. and Zechner, J. (1998). IPO-mechanisms, monitoring and ownership structure. Journal of Financial Economics, 49, 45–77. Tinic, S. (1988). Anatomy of IPOs of common stock. Journal of Finance, 43, 789–822.

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0009

Chapter 9

Influence of Religion and Social Attitudes in Stock Market Participation Yang Zhou∗ , Jinwen Yu† and Zhiping Zhou‡

Abstract We investigate the relationship between religion and stock market participation. Using data for a representative sample of the Chinese population, we find that compared with non-religious households, Buddhists are more likely to invest in stocks, while Muslims are less likely to invest in stocks. In contrast, the differences in stock market participation between non-religious households and other religious households, such as Taoists, Protestants and Catholics, are statistically insignificant. Furthermore, religiosity is negatively associated with the propensity to invest in stocks. Social attitudes turn out to be helpful in explaining the higher stock market participation of Buddhist households. Instrumental variable estimations show that the effects of being Muslim and Buddhist on stock market participation are likely causal. Keywords: Religion, stock market participation, social attitudes

∗

Department of Finance, School of Economics and Management, Wuhan University, Wuhan, Hubei 430072, China; [email protected]. † Department of Finance, School of Economics and Management, Wuhan University, Wuhan, Hubei 430072, China; [email protected]. ‡ Corresponding author. Department of Economics and Finance, School of Economics and Management, Tongji University, Shanghai 200092, China; [email protected]. 209

page 209

November 4, 2020

210

1.

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

Introduction

The relationship between religion and risk-taking is being increasingly studied by economists. Miller and Hoffman (1995) find a positive correlation between religiosity and risk aversion. Diaz (2000) shows that religious people living in Las Vegas gamble less. Noussair et al. (2013) find that more religious people, as measured by church membership or attendance, are more risk averse with respect to financial risks. In addition to the link between individual religiosity and risk-taking, a number of recent papers rely on measures of religiosity at the county or regional level to study whether religion also influences organizational risk-taking behaviors. Hilary and Hui (2009) find that firms in more religious counties display lower levels of risk exposure. Adhikari and Agrawal (2016) show that banks headquartered in more religious areas are more reluctant to take risks. Chen et al. (2016) find that local governments in counties with a higher degree of religiosity are more conservatively managed. Overall, previous research has consistently shown that religion is negatively related to risk-taking. Nonetheless, almost all previous studies consider Western religions only. It is still unclear whether and how Eastern religions affect risk-taking. The two main groups of world religions are Western religions (such as Protestantism, Catholicism, Judaism and Islamism) and Eastern religions (such as Buddhism, Hinduism and Taoism), and they have been found to differ substantially from each other (Oxtoby and Amore, 2010; Hopfe and Woodward, 2011). For example, Western religions often teach the existence of a single supernatural God that their followers worship and believe they can communicate with and emphasize exclusivity, claiming to be the one and only correct spiritual path. Hence, being irreligious represents risk-taking behavior in Christian and Muslim societies. By contrast, in Buddhist and Hindu societies, where religions typically preach a philosophy of life (e.g., detachment and understanding), put the emphasis on personal behavior rather than organizational affiliation, and tend to be non-exclusive, not participating in the mainstream religion does not necessarily constitute risk-taking behavior. This implies that believing in Eastern religions may not make individuals risk-averse (Miller, 2000). Given the major distinctions in ideology between these two groups of religions, it is highly likely that their

page 210

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation

page 211

211

respective adherents have different risk appetite and display different risk-taking behaviors in the stock market. In this chapter, we examine whether individuals differ in their willingness to invest in stocks depending on religious affiliations. We are particularly concerned with the distinctions in the effects on stock market participation between Western religions and Eastern religions. In our view, China offers a good arena to study these questions, since it is characterized by religious diversity, with deep roots in Buddhism and Taoism and a rapid growth in the population believing in Islam, Protestantism and Catholicism in recent years. Specifically, we use data from the China General Social Survey (CGSS), which includes questions on religious affiliation, stock market participation and a wide range of demographic and socioeconomic information at the individual level. We find clear evidence that compared with non-religious households, Buddhists are more likely to invest in stocks, whereas Muslims are less likely to invest in stocks. On average, being Buddhist increases the probability of stock market participation by 2.2%, while being Muslim lowers the probability by 4.1%. In contrast, the differences in stock market participation between non-religious households and other religious households, such as Protestants and Catholics, are statistically insignificant. This goes against the recent evidence that the adherents of Western religions are more risk averse than non-religious individuals (Miller and Hoffmann, 1995; Noussair et al., 2013). A possible explanation for this contradiction is that the effects of religion on household financial behaviors differ depending on some country-specific factors, such as history and culture. Therefore, our results cast doubt on the validity of generalizing country-specific studies on the economic effects of religious beliefs. In addition, we use the frequency of attendance at religious services as a proxy for religiosity and find a negative correlation between religiosity and stock market participation. More religious Muslim and Protestant households exhibit lower willingness to participate in the stock market. We also find that the social attitudes of households differ across religious affiliations. Buddhists are less trusting and have a lower sense of fairness than non-religious households, while Muslims display a higher level of happiness. With respect to other religious denominations, Taoists are more trusting than non-religious households. Protestants are less trusting, and have a lower sense of fairness

November 4, 2020

212

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

and higher level of happiness. Moreover, we show that the different views on social beliefs can partially explain the higher probability of participating in the stock market of Buddhist households. Finally, we establish causality with an instrumental variable approach. We use the density of religious sites in the province of residence to instrument the households’ religious affiliations. The instrumental variable estimations reveal that the effects of being Muslim and Buddhist on stock market participation are likely causal. The contributions of this chapter are two-fold. First, we are among the first to study the relation between Eastern religions and household risk-taking behaviors. Interestingly, we find that Buddhists are more likely to invest in stocks, which stands in sharp contrast to the positive relation between Western religions and risk-aversion documented in the literature (Miller and Hoffmann, 1995; Hilary and Hui, 2009). Second, we try to address the problem of causality between religion and stock market participation, which is a challenge for much of the research in this field (Guiso et al., 2003; Renneboog and Spaenjers, 2012). By applying instrument variable approach, we find evidence that being Buddhist and Muslim have causal effects on household stock market participation decision. In addition to the literature on religion and risk-taking, our empirical results also add to the literature on the factors influencing household stock market participation. Prominent examples include participation costs (Abel, 2001), trust (Guiso et al., 2008), happiness (Delis and Mylonidis, 2015), cognitive abilities (Christelis et al., 2010), investors’ IQ (Grinblatt et al., 2011), social interaction (Hong et al., 2004), usage of internet (Bogan, 2008), health status (Rosen and Wu, 2004), financial literacy (Van Rooij et al., 2011), prenatal environment (Cronqvist et al., 2015), labor market experiences (Kn¨ upfer et al., 2017) and religion (Renneboog and Spaenjers, 2012). The closest to our chapter is Renneboog and Spaenjers (2012), who examine the interactions between religion, economic attitudes, and household saving and investment behaviors. While their paper takes an important step showing how religion affects household financial decision-making, it has two limitations. First, they only consider adherents of Western religions. It remains unclear whether their results can be generalized to those of Eastern religious. Second, they only estimate the correlations, leaving the issue of causality unresolved. We extend their paper by

page 212

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation

page 213

213

incorporating Eastern religions into analysis and addressing endogeneity concerns. The remainder of this chapter is organized as follows. Section 2 sketches religion in China. Section 3 describes our data set and variables. Section 4 presents the regression models and discusses the results and Section 5 concludes the chapter.

2.

Religion in China

As an old Eastern country with rich culture and civilizations, China has long been a cradle and host to a variety of religious traditions of the world. For example, Taoism, an indigenous religion in China, can be traced to nearly 1,900 years ago. Historically, religions, especially Buddhism and Taoism, have been deep-rooted from generation to generation and played a crucial role in shaping Chinese culture, such as aesthetics, politics, literature, philosophy, medicine and a lot of other aspects. After the founding of the People’s Republic of China in 1949, religious activities were largely reduced due to the imposition of the Communist ideology, albeit never reduced to zero (Yang, 2010). The Cultural Revolution even led to a policy of elimination of religions; a great number of religious sites were destroyed. This policy relaxed considerably in the late 1970s at the end of the Cultural Revolution and more tolerance of religious expression has been permitted ever since. From legal perspective, the Constitution of China in 1982 guarantees “freedom of religion” for citizens. Religious festivals are held, traditional funerals and burial rituals are restored, destroyed images and shrines have been rebuilt, priests have reappeared to organize rituals and congregations meet to worship. In recent years, the government’s attitude towards religious activities has become unprecedentedly friendly, as the Communist Party leaders realize that religion helps fill a vacuum created by the country’s breakneck growth and fight against corruption and see religion as an integral part of the Chinese culture. As a consequence, religion has experienced a significant revival and has become popular again in China. Five main religions in China are Buddhism, Taoism, Islam, Protestantism and Catholicism, which are formally and institutionally recognized by the government. Buddhism and Taoism have been

November 4, 2020

214

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

the most influential religions for thousands of years and still have a significant presence in modern China (Overmyer, 2003). Although accurate statistics on China’s current religious population are hard to find, various surveys1 have found that Buddhists and Taoists constitute 5%–10% and 0.2%–1% of the total population of China, respectively. In contrast, the other three Western religions have a shorter history, but have gained momentum in recent years. Survey data show that the shares of Chinese population believing in Islam, Protestantism and Catholicism are 2%–4%, 1.5%–2.5% and 2%–4%, respectively. 3.

Data and Variables

Our source of data is the CGSS, which is the youngest in the world General Social Survey family, but the earliest nationally representative continuous survey project run by an academic institution in Mainland China. The survey is conducted jointly by the Renmin University of China and the Hong Kong University of Science and Technology and is aimed to monitor the changing relationship between social structure and quality of life in both urban and rural China. In particular, the CGSS records detailed information about households’ religion, which enables us to revisit the relationship between religion and stock market participation in Chinese context. This is the primary reason we choose the CGSS data to pursue our research. Launched in 2003, the CGSS is an annual repeated cross-sectional survey and its data are publicly available until 2013 wave.2 We restrict our sample to the data from 2010, 2012 and 2013 waves, because only those waves collect information on both households’ 1

See, for example, the Chinese General Social Survey (CGSS) jointly conducted by the Renmin University of China and the Hong Kong University of Science and Technology, the Chinese Family Panel Studies (CFPS) conducted by the Institute of Social Science Survey (ISSS) of Peking University, and the Chinese Spiritual Life Survey directed by the Purdue University’s Center on Religion and Chinese Society. 2 The data are downloadable from http://www.chinagss.org/index.php? r=index/index&hl=en. However, the data from 2004, 2007 and 2009 waves are not available online.

page 214

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation

page 215

215

religion and stock market participation. Given the consistency in the questionnaire design, we pool these three-year data to generate a larger cross-sectional data set. To deal with the concern that households’ willingness to participate in the stock market may vary systematically over time, we will control for year fixed effects. The CGSS 2012 and 2013 cover 29 out of 31 provinces in Mainland China, with Hainan and Tibet excluded, while the CGSS 2010 covers all 31 provinces. The three waves provide 34,986 observations in total. Nevertheless, the number of observations used in the empirical analysis is lower due to missing information for specific questions and obvious measurement errors. In what follows, we outline the variables used in empirical analysis. An overview of all variables is provided in Table 1. Following Christelis et al. (2010), we construct two measures of stock market participation as dependent variables. The first measure is the dummy variable Direct participation, which equals one if the respondent holds stocks directly. The second measure is the dummy variable Total participation, which equals one if the respondent holds stocks directly or indirectly through mutual funds. In addition to the participation dummy, previous studies typically also employ the fraction of wealth invested in stocks as dependent variable (Van Rooij et al., 2011; Renneboog and Spaenjers, 2012). We are not able to construct such a portfolio share variable, because the CGSS does not provide information on the monetary value of stock investments. However, this does not seem to cause a big loss, because individuals can lie more easily about the monetary level of their stock investments than about the stock ownership, thereby introducing larger measurement errors. The explanatory variable for our analysis is religion. We measure religion primarily by religious affiliations. Specifically, we make distinctions between Buddhist, Taoist, Muslim, Catholic, Protestant and Other religion. This categorization is motivated by the relative importance of religions in China; the last category contains Jews, Hindus and other smaller religious groups. All of these religious affiliations are defined as dummy variables. The main focus of our study is on the first five categories. In a broader sense, they can be divided into two groups, Eastern religions (Buddhist and Taoist) and Western religions (Muslim, Catholic and Protestant), and there are noticeable differences between them. For example, Western religions typically

November 4, 2020

10:13

216

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Finance: A Novel Approach Table 1:

Variable

Definition of variables

Description

Stock market participation Direct participation Holding stocks directly Total participation Holding stocks directly or through mutual funds Religion Buddhist Taoist Muslim Catholic Protestant Other religion Attendance

Social attitudes Trust

Religious Religious Religious Religious Religious Religious

denomination denomination denomination denomination denomination denomination

Frequency of attendance at religious services

“Generally speaking, do you agree that most people can be trusted?”

Sense of fairness

“Generally speaking, do you agree that we have a fair society?”

Happiness

“Generally speaking, how happy would you say you are?”

Control variables Age Ln(Income+1) Male

b3964-ch09

Year of survey – year of birth Ln(respondent’s income+1) Sex of the respondent

Values

Holder = 1; other = 0 Holder = 1; other = 0

Buddhist = 1; other = 0 Taoist = 1; other = 0 Muslim = 1; other = 0 Catholic = 1; other = 0 Protestant = 1; other = 0 {Jew, Hindu, etc.} = 1; other = 0 Never = 1; less than once a year = 2; once or twice a year = 3; several times a year = 4; once a month = 5; twice or three times a month = 6; almost once a week = 7; once a week = 8; several times a week = 9 Totally disagree = 1; disagree = 2; neutral = 3; agree = 4; totally agree = 5 Totally disagree = 1; disagree = 2; neutral = 3; agree = 4; totally agree = 5 Very unhappy = 1; happy = 2; neutral = 3; happy = 4; Very happy = 5

Male = 1; female = 0 (Continued)

page 216

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation Table 1: Variable

Description Marital status Length of education that the respondent receives

Health

Subjective assessment of health status

Homeowner Internet Hukou Communist Party

Home ownership Use of internet Hukou status Membership in Communist Party of China

Instrumental variables Buddhist temples Provincial density of Buddhist temples

Taoist temples Protestant churches

217

(Continued )

Married Education

Mosques

page 217

Provincial density of Mosques Provincial density of Taoist temples Provincial density of Protestant churches

Values Married = 1; other = 0 Primary school = 6; junior high school = 9; high school = 12; technical secondary school = 13; junior college = 15; college = 16; graduate school = 19; Poor = 1; not so good = 2; fair = 3; good = 4; excellent = 5 Homeowner = 1; other = 0 User = 1; other = 0 Rural Hukou = 1; other = 0 Member = 1; other = 0

Number of Buddhist temples/area of province3 Number of Mosques/area of province Number of Taoist temples/ area of province Number of Protestant churches/area of province

preach the existence of a single supernatural God that their followers worship and believe they can communicate with and emphasize salvation to overcome uncertainty and anxiety, while Eastern religions often teach a “philosophy of life” and tend to be non-exclusive. The major ideological differences between these two groups of religions make it possible that their respective adherents have different risk appetite and exhibit different risk-taking behavior in the stock 3

Area of province is measured in terms of square kilometers.

November 4, 2020

218

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

market. Further, we also use the frequency of attendance at religious services to capture religiosity, which is labeled by Attendance. We also consider a wide range of household background characteristics as control variables, starting with the variable Age and the dummy variables Male and Married. The variable Education captures the length of education that the household receives. It is quantified by assigning the age of 6 years to a primary school education, 9 years to a junior high school education, 12 years to a senior high school education, 13 years to a vocational school education, 15 years to a junior college education, 16 years to a college level and 19 years to a graduate school education and beyond. The natural logs of the household’s income is labeled by Ln(Income + 1). To capture the potential nonlinear dependence, we also include the Age2 and Ln(Income + 1)2 in the regressions. Since homeownership, health status and use of internet have shown to affect household stock market participation in the literature (Cocco, 2005; Rosen and Wu, 2004; Bogan, 2008), we control for the dummy variable Homeowner, the continuous variable Health and the dummy variable Internet. A hukou is a record in a government system of household registration required by law in China. Because of its entrenchment of social strata, especially between rural and urban residency status, the hukou system is often regarded as a caste system of China. Similarly, the membership in Communist Party of China makes a distinction between the members and non-members with respect to social status, as the former enjoy some implicit benefits, such as advantages in job hunting and promotion in state-owned enterprises. Therefore, we also include the dummy variables Hukou and Communist Party as control variables. Finally, we include interactions of province and year fixed effects. Table 2 reports summary statistics for the variables outlined earlier. Around 5.7% of the households hold stocks directly and 7.3% of the households hold stocks directly or through mutual funds. This implies that Chinese households are reluctant to participate in the stock market as compared to those in the countries with more developed financial markets.4 The largest religious group is

4

The literature documents higher rates of stock market participation for households in the US and some European countries. For example, Guiso et al. (2008)

page 218

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

page 219

Influence of Religion and Social Attitudes in Stock Market Participation Table 2: Variable

Summary statistics St. dev.

Min

Stock market participation Direct participation 0.057 Total participation 0.073

0.231 0.260

Religion Buddhist Taoist Muslim Catholic Protestant Other religion Attendance

0.056 0.002 0.024 0.002 0.019 0.002 1.404

Social attitudes Trust Sense of fairness Happiness Control variables Age Ln (Income + 1) Male Married Education Health Homeowner Internet Hukou Communist Party Instrumental variables Buddhist temples Mosques Taoist temples Protestant churches

Mean

219

Max

Obs.

0 0

1 1

34790 34790

0.230 0.047 0.154 0.050 0.137 0.156 1.397

0 0 0 0 0 0 1

1 1 1 1 1 1 9

34979 34979 34979 34979 34979 34979 34921

3.421 3.015 3.777

1.048 1.066 0.856

1 1 1

5 5 5

34945 34926 34871

48.266 10.211 0.499 0.792 8.760 0.760 0.933 0.170 0.533 0.115

16.123 1.339 0.500 0.406 4.748 1.097 0.250 0.376 0.499 0.319

17 0 0 0 0 1 0 0 0 0

97 16.118 1 1 1 5 1 1 1 1

34981 30654 34986 34955 34976 34965 34810 34192 34980 34875

0.004 0.001 0.003 0.004

0.006 0.004 0.012 0.007

0 0 0 0

0.029 0.018 0.066 0.028

31 31 31 31

Note: The table reports the summary statistics (mean, standard deviation, minimum, maximum and number of observations) for the variables used in this study. All variables are defined in Table 1.

show that for the US, the average rate of participation through both direct and indirect stockholding is 48.9%. Renneboog and Spaenjers (2012) show that 13% of the Dutch households own stocks.

November 4, 2020

220

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

page 220

Behavioral Finance: A Novel Approach

the Buddhists, which account for 5.6% of our sample. What follows are Muslims and Protestants. Taoists and Catholics are the smallest religious groups among the five religions under consideration. 4. 4.1.

Empirical Analysis Baseline Results

We first investigate the relationship between the religious affiliations and stock market participation. For this purpose, we run the following multivariate models Yi = α + β  Xi + δ Ci + εi

(1)

where Yi captures the stock market participation variable for household i. We consider two measures of stock market participation, namely Direct participation and Total participation, respectively. Given the nature of our dependent variables, we use probit to estimate the models. Xi are the dummy variables for religious affiliations (e.g., Buddhist) and Ci is a vector of control variables (e.g., Age). Note that interactions of province and year fixed effects are also included in Ci . α is the intercept and εi is the error term for household i. We report standard errors clustered at the provincial level to adjust for the potential correlation of observations within the same cluster. We are mainly interested in the vector of coefficients β, which not only captures the relationship between the religious affiliations and stock market participation but we will also discuss δ, which are the coefficients on the control variables. Table 3 presents the regression results from estimating Eq. (1) with Direct participation as dependent variables. In Columns (1)–(6), we first consider each religious denomination in separate models and then in Column (7) include all religious denominations jointly. As explained before, each model also includes a constant, year dummies and province dummies, but those results are not reported for reasons of conciseness. Holding all controls at the mean, the marginal effects are calculated and reported instead of the estimated coefficients. At the bottom of these tables, we show the number of observations and pseudo R-squared.

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

page 221

Influence of Religion and Social Attitudes in Stock Market Participation Table 3:

221

Religion and direct participation Direct Participation

(1) Buddhist

(2)

(3)

(4)

(6)

(7)

−0.009 (0.016) 0.005∗∗∗ (0.001) −0.000∗∗∗ (0.000) 0.000

0.021∗∗∗ (0.008) 0.000 (0.023) −0.041∗ (0.022) 0.007 (0.024) −0.003 (0.014) −0.008 (0.016) 0.005∗∗∗ (0.001) −0.000∗∗∗ (0.000) −0.013

0.022 (0.008)

Taoist

0.001 (0.024)

Muslim

−0.041∗ (0.022)

Catholic

0.008 (0.024) −0.002 (0.014)

Protestant Other religion Age

0.005∗∗∗ 0.005∗∗∗ (0.001) (0.001) Age2 −0.000∗∗∗ −0.000∗∗∗ (0.000) (0.000) Ln (Income + −0.014 0.001 1) (0.010) (0.036) Ln (Income + 0.002∗∗∗ 0.002 1)2 (0.001) (0.001) Male −0.006∗ −0.004 (0.003) (0.003) Married −0.007 −0.009 (0.006) (0.007) Education 0.006∗∗∗ 0.006∗∗∗ (0.001) (0.001) Homeowner 0.001 0.002 (0.008) (0.008) Health −0.002 −0.002 (0.003) (0.003) Internet 0.032∗∗∗ 0.031∗∗∗ (0.004) (0.004) Hukou −0.057∗∗∗ −0.058∗∗∗ (0.012) (0.013) Communist −0.004 −0.005 Party (0.007) (0.007) Observations Pseudo R2

(5)

∗∗∗

27515 0.261

26010 0.262

0.005∗∗∗ 0.005∗∗∗ 0.005∗∗∗ (0.001) (0.001) (0.001) −0.000∗∗∗ −0.000∗∗∗ −0.000∗∗∗ (0.000) (0.000) (0.000) 0.002 0.001 0.003 (0.036) 0.002

(0.036) 0.002

(0.036) 0.002

(0.035) 0.002

(0.010) 0.002∗∗∗

(0.001) −0.004 (0.003) −0.009 (0.007) 0.006∗∗∗ (0.001) 0.002 (0.008) −0.002 (0.003) 0.030∗∗∗ (0.004) −0.057∗∗∗ (0.012) −0.005 (0.007)

(0.001) −0.004 (0.003) −0.008 (0.007) 0.006∗∗∗ (0.001) 0.003 (0.008) −0.002 (0.003) 0.031∗∗∗ (0.004) −0.058∗∗∗ (0.012) −0.005 (0.007)

(0.001) −0.004 (0.003) −0.008 (0.008) 0.006∗∗∗ (0.001) 0.002 (0.008) −0.002 (0.003) 0.031∗∗∗ (0.004) −0.058∗∗∗ (0.013) −0.005 (0.007)

(0.001) −0.004 (0.003) −0.009 (0.007) 0.006∗∗∗ (0.001) 0.001 (0.008) −0.003 (0.003) 0.030∗∗∗ (0.004) −0.058∗∗∗ (0.012) −0.005 (0.007)

(0.001) −0.005∗ (0.003) −0.008 (0.006) 0.006∗∗∗ (0.001) 0.000 (0.008) −0.002 (0.003) 0.031∗∗∗ (0.004) −0.055∗∗∗ (0.012) −0.005 (0.007)

26719 0.263

26017 0.262

26503 0.264

26598 0.262

29636 0.265

Note: The table reports marginal effects of the probit models, with direct participation as dependent variables. The model also includes a constant term and interactions of province and year fixed effects, but their coefficients are suppressed for brevity. Standard errors (in parentheses) are clustered at the provincial level. Coefficients marked with ∗∗∗ and ∗ are statistically significant at the 1% and 10% level, respectively.

November 4, 2020

222

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

It is clear from Table 3 that different religious denominations are associated with different propensity to hold stocks directly. When separately compared with non-religious households, Buddhists are 2.2% more likely to directly participate in the stock market, while Muslims are 4.1% less likely to invest in stocks, suggesting that these two religions have economically sizable effects on the adherents’ willingness to directly participate in the stock market. In contrast, the differences in direct stock market participation between non-religious households and other religious households, such as Taoists, Protestants and Catholics, are statistically insignificant. These results remain almost unchanged when introducing all the religious denominations jointly. The decision to directly hold stocks is also correlated with many of the demographic and background variables included in our analysis. The combination of the positive coefficient on Age and the negative coefficient on Age2 implies a bell-shaped pattern in the effect of age, which is in line with the life-cycle hypothesis. More education and internet usage are associated higher probabilities of direct stock ownership, which is consistent with the literature (Bogan, 2008; Cole et al., 2014). Households with rural hukou exhibit more cautious investment behaviors. This might be due to the scarcity of the financial service in the rural areas of China that raises the participation costs and hinders stock investment. Table 4 shows the results with total participation as dependent variables. Obviously, the same pattern emerges as in Table 3: Buddhists and Muslims have different propensity to invest in stocks from non-religious individuals, although the magnitude of the effects is slightly higher. With respect to control variables, in addition to those significant in the case of direct stock market participation, Ln(Income + 1)2 and Male are also statistically significant. As indicated by the positive coefficient on Ln(Income+1)2 , there are income effects: households with high incomes are more likely to own stocks. Female households are more likely to participate in the stock market, which goes against the traditional view that females are more conservative than males in making investment decisions (Barber and Odean, 2001; Almenberg and Dreber, 2015). The higher propensity of Buddhist households to invest in stocks stands in stark contrast with the negative relationship between Western religions and risk-taking documented in the literature (Miller and

page 222

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

page 223

Influence of Religion and Social Attitudes in Stock Market Participation Table 4:

223

Religion and total participation Total Participation

(1) Buddhist

(2)

(3)

(4)

(5)

(6)

(7)

−0.016 (0.020) 0.006∗∗∗ (0.001) −0.000∗∗∗ (0.000) −0.016

0.025∗∗∗ (0.008) 0.058 (0.042) −0.042∗∗ (0.019) 0.023 (0.027) −0.005 (0.013) −0.015 (0.020) 0.007∗∗∗ (0.001) −0.000∗∗∗ (0.000) −0.019∗

(0.023) (0.023) (0.023) (0.023) (0.022) 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗

(0.011) 0.003∗∗∗

0.026∗∗∗ (0.009)

Taoist

0.060 (0.043)

Muslim

−0.042∗∗ (0.019)

Catholic

0.023 (0.027) −0.004 (0.013)

Protestant Other religion Age

∗∗∗

0.006 (0.001) Age2 −0.000∗∗∗ (0.000) Ln (Income + −0.021∗ 1) (0.011) Ln (Income + 0.003∗∗∗ 2 1) (0.001) Male −0.010∗∗∗ (0.003) Married −0.003 (0.007) Education 0.008∗∗∗ (0.001) Homeowner 0.003 (0.009) Health −0.003 (0.003) Internet 0.036∗∗∗ (0.004) Hukou −0.064∗∗∗ (0.011) Communist 0.000 Party (0.007) Observations Pseudo R2

27515 0.261

∗∗∗

0.006 (0.001) −0.000∗∗∗ (0.000) −0.016

∗∗∗

0.006 (0.001) −0.000∗∗∗ (0.000) −0.015

∗∗∗

0.006 (0.001) −0.000∗∗∗ (0.000) −0.016

∗∗∗

0.006 (0.001) −0.000∗∗∗ (0.000) −0.015

(0.001) −0.007∗∗∗ (0.003) −0.005 (0.007) 0.008∗∗∗ (0.001) 0.004 (0.009) −0.004 (0.003) 0.033∗∗∗ (0.003) −0.065∗∗∗ (0.011) −0.001 (0.006)

(0.001) −0.007∗∗∗ (0.003) −0.005 (0.008) 0.008∗∗∗ (0.001) 0.004 (0.009) −0.004 (0.003) 0.034∗∗∗ (0.003) −0.063∗∗∗ (0.012) 0.000 (0.006)

(0.001) −0.007∗∗∗ (0.003) −0.005 (0.008) 0.008∗∗∗ (0.001) 0.004 (0.009) −0.004 (0.003) 0.032∗∗∗ (0.003) −0.066∗∗∗ (0.012) 0.000 (0.006)

(0.001) −0.007∗∗∗ (0.003) −0.005 (0.008) 0.008∗∗∗ (0.001) 0.004 (0.009) −0.004 (0.003) 0.034∗∗∗ (0.003) −0.065∗∗∗ (0.012) 0.000 (0.006)

(0.001) −0.008∗∗∗ (0.003) −0.005 (0.007) 0.008∗∗∗ (0.001) 0.003 (0.009) −0.004 (0.003) 0.034∗∗∗ (0.003) −0.065∗∗∗ (0.011) 0.000 (0.006)

(0.001) −0.009∗∗∗ (0.003) −0.005 (0.007) 0.008∗∗∗ (0.001) 0.003 (0.009) −0.004 (0.003) 0.035∗∗∗ (0.004) −0.062∗∗∗ (0.011) −0.001 (0.006)

26010 0.262

26719 0.263

26017 0.262

26503 0.264

26598 0.262

29636 0.265

Note: The table reports marginal effects of the probit models, with direct participation as dependent variables. The model also includes a constant term and interactions of province and year fixed effects, but their coefficients are suppressed for brevity. Standard errors (in parentheses) are clustered at the provincial level. ∗∗∗ , ∗∗ and ∗ denote significance at the 1%, 5% and 10% level, respectively.

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

page 224

Behavioral Finance: A Novel Approach

224

Hoffmann, 1995; Renneboog and Spaenjers, 2012; Noussair et al., 2013). This indicates that the effects on risk-taking may differ substantially across religions, especially between Western and Eastern religions. Another interesting observation is that in China, being Catholic or Protestant does not have any effect on stock ownership. This contradicts previous findings that Protestant and/or Catholic households are more risk averse than non-religious ones. Since all of these papers focus on households in Western countries, this finding implies that the economic effects of religious beliefs are likely to depend on the cultural and historical factors of the specific country under consideration and highlights the need for caution in generalizing the results of the studies on religion. 4.2.

Religiosity

We further examine the relationship between the religiosity and household risk-taking behavior in the stock market. Since people who attend religious services more often are more exposed to religious principles, we use the frequency of attendance at religious services as a proxy for the dimension of religiosity (Guiso et al., 2003; Noussair et al., 2013). The estimation models are specified as follows: Yi = α + θZi + β  Xi + δ  Ci + εi ,

(2)

where Zi is the religiosity of household i, which is measured by the frequency of attendance at religious services and θ is the associated coefficient. Other variables are the same as defined in the previous sections. Further, as the effects of religiosity on the stock market participation decision may differ across religions, we add interaction terms between the religiosity variable Zi and the religious affiliation dummies Xi to Eq. (2). Yi = α + μ Zi × Xi + θZi + β  Xi + δ Ci + εi .

(3)

Table 5 shows the estimation results. As shown in Columns (1) and (3), higher frequency of attendance at religious gatherings is related to lower probability of participation. Furthermore, we can see from Columns (2) and (4) that the interaction terms are significantly negative for Muslim and Protestant but insignificant for other religion dummies, indicating that more religious Muslim and

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation Table 5:

(1) Attendance

−0.003∗∗ (0.002)

Buddhist*Attendance Taoist*Attendance Muslim*Attendance Catholic*Attendance Protestant*Attendance Other religion* Attendance

Muslim Taoist Catholic Protestant Other religion Observations Pseudo R2

225

Religion, attendance and stock market participation Direct Participation

Buddhist

page 225

(2) 0.024 (0.015) 0.002 (0.012) 0.014 (0.034) −0.033 ∗∗∗ (0.015) 0.017 (0.020) −0.016 (0.011)

Total Participation (3) −0.002∗ (0.001)

(4) 0.026 (0.017) −0.002 (0.014) 0.041 (0.029) −0.027∗∗∗ (0.010) 0.022 (0.019) −0.036∗∗∗ (0.013)

0.029∗∗∗ (0.009) −0.032 (0.021) 0.007 (0.022) 0.017 (0.023) 0.012 (0.019) 0.000 (0.015)

0.011 (0.017) 0.282∗∗∗ (0.038) −0.174∗∗∗ (0.042) −0.016 (0.036) 0.036∗ (0.019) −0.018 (0.017) −0.002 (0.003)

0.030∗∗∗ (0.010) −0.036∗∗ (0.018) 0.063 (0.042) 0.029 (0.027) 0.005 (0.018) −0.009 (0.019)

0.003 (0.019) −0.004 (0.030) 0.008 (0.060) −0.017 (0.032) 0.061∗∗∗ (0.015) −0.022 (0.028) 0.001 (0.003)

29594 0.266

29594 0.267

29594 0.278

29594 0.278

Note: The table reports marginal effects of the probit models, with both direct participation and total participation as dependent variables. The model also includes a constant term, interactions of province and year fixed effects, and all previously used variables, but their coefficients are suppressed for brevity. Standard errors (in parentheses) are clustered at the provincial level. ∗∗∗ , ∗∗ and ∗ denote significance at the 1%, 5% and 10% level, respectively.

November 4, 2020

226

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

Protestant households are less likely to invest in stocks. This is consistent with the positive relationship between religiosity and riskaversion documented in Noussair et al. (2013), but contradicts the higher stock market participation of churchgoers found in Hong et al. (2004). Theoretically, attending religious services may have two distinct effects on stock market participation. On the one hand, it is positively associated with the degree of religious beliefs that may affect their attitudes towards financial risk-taking. On the other hand, it provides an opportunity to experience and organize social interaction among members of the community, and thus has a clear social aspect. Because social interaction has been found to have a positive effect on stock market participation (Hong et al., 2004; Brown et al., 2008), the negative correlation between the attendance of religious services and stock market participation is likely to be driven by religious beliefs themselves. Unfortunately, because the CGSS does not ask questions about the degree of beliefs, we are not able to empirically test this conjecture. 4.3.

Social Attitudes

In this section, we investigate how social attitudes are correlated with religious affiliations and whether they serve as channels through which religion affects stock market participation. Specifically, we consider three dimensions of social attitudes, including trust, sense of fairness and happiness. The attitudes of religions toward the common good and social interactions may shape the adherents’ social beliefs and influence their social behaviors. Previous literature has extensively studied the effects of religions on various aspects of social attitudes. The evidence on the relation between religion and the level of trust is rather mixed. Alesina and La Ferrara (2002) find that religious beliefs do not affect trust. In contrast, Guiso et al. (2003) show that the level of trust is positively associated with religion and is mostly affected by religious participation, but not by being brought up religiously. Arru˜ nada (2010) reports that Protestants are more trusting of anonymous counterparties. With respect to sense of fairness, Guiso et al. (2003) find that religious people are more likely to believe in the fairness of the market. Moreover, much research has been done to explore the relation between religion and happiness, but no consensus has been reached. For example, Mookerjee and Beron

page 226

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation

page 227

227

(2005) provide support for a positive association, while Francis et al. (2003) find no evidence for such a relationship. In turn, social attitudes may affect stock market participation. Guiso et al. (2008) find that less trusting individuals are less likely to buy stocks. Delis and Mylonidis (2015) show that happiness leads to a lower probability of investing in stocks and its effect is economically more sizable than trust. For sense of fairness, however, little has been done to explore its influence on the ownership of stocks. To elicit information about social attitudes, we use three questions asked in the CGSS: (1) “Generally speaking, do you agree that most people can be trusted?” (2) “Generally speaking, do you agree that society treats everyone fairly?” (3) “Generally speaking, how happy would you say you are?” For all three questions, individuals could answer in one of the five ways: (1) totally disagree; (2) disagree; (3) neutral; (4) agree; (5) totally agree. Based on these questions, we construct three categorical variables, namely Trust, Sense of fairness and Happiness, to measure the general trust, sense of fairness and happiness of the respondent.5 It is clear from Table 2 that there is variation in social attitudes, which will allow us to investigate the interplay between religious affiliations, social attitudes and stock market participation decision. We estimate the following models to investigate to what degree religion is associated with differences in social attitudes Si = α + β  Xi + δ Ci + εi

(4)

where Si is the social attitude variable for household i (Trust, Sense of fairness and Happiness), while Xi and Ci are the same religion dummies and control variables as before. We estimate the models using ordered probit, depending on the nature of dependent variables. Again, we report standard errors clustered at the provincial level. 5

Guiso et al. (2008) use a question similar to the first question to measure the general trust. Delis and Mylonidis (2015) use a question similar to the second question to measure the general happiness.

November 4, 2020

228

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

page 228

Behavioral Finance: A Novel Approach

The coefficients β measure the correlations between religious background and social attitudes. Table 6 shows that different religious denominations are related to different social attitudes. Our interests mainly lie in the coefficients on the Buddhist and Muslim dummies, as they have been found to have significant effects on household stock market participation decisions. The results in the Trust model indicate that Buddhist households are less trusting. Holding all controls constant at the mean, a calculation of the marginal effects (not reported) shows that Buddhists have a chance to “totally agree” with the statement “most people can be trusted” that is 1.45% lower that of non-Buddhists. At the same time, we also see a significantly negative coefficient on Buddhist in the Sense of fairness model: Buddhists are less likely to “totally agree” that “we have a fair society.” Being Buddhist decreases the probability by 0.67%. The results for Happiness show that Muslim households display a higher level of happiness than non-religious counterparts. The coefficient on Muslim implies a 11.88% lower likelihood of finding themselves “very happy.” Moreover, other religious denominations are also correlated with social attitudes. Taoists are more trusting than non-religious households. Protestants are less trusting, and have a lower sense of fairness and higher level of happiness. Given the correlation between religious affiliations and social attitudes found earlier, we now want to identify which social attitudes may explain why religious people have different willingness to participate in the stock market. Following Renneboog and Spaenjers (2012), we expand model (1) with the social attitude variables as follows: Yi = α + γ  Si + β  Xi + δ Ci + εi

(5)

We are interested in how the coefficients on the religious affiliation dummies change after controlling for the social attitudes. We first add each economic attitude to model (1) separately, and then include all social attitudes jointly. As before, we consider both Direct participation and Total participation as dependent variables. All previously used control variables are included in the estimations, but the coefficients on these variables are not reported for reasons of conciseness. Table 7 shows that sense of fairness and happiness are significantly negatively correlated with the decision to invest in stocks, while trust

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation Table 6:

Buddhist Taoist Muslim Catholic Protestant Other religion Age Age2 Ln (Income + 1) Ln (Income + 1)2 Male Married Education Homeowner Health Internet Hukou Communist Party Observations Pseudo R2

Religion and social attitudes Trust

Sense of Fairness

Happiness

(1)

(2)

(3)

−0.087∗∗∗ (0.033) 0.263∗∗ (0.130) 0.125 (0.080) −0.050 (0.115) −0.152∗∗∗ (0.048) −0.138 (0.168) 0.017∗∗∗ (0.003) 0.000∗ (0.000) −0.003 (0.017) 0.000 (0.001) −0.008 (0.017) 0.034∗∗ (0.017) 0.004 (0.003) 0.139∗∗∗ (0.026) 0.051∗∗∗ (0.012) 0.018 (0.024) 0.164∗∗∗ (0.029) 0.161∗∗∗ (0.028)

−0.079∗∗ (0.037) 0.072 (0.077) 0.278 (0.248) −0.008 (0.134) −0.083∗ (0.049) −0.117 (0.159) −0.008∗∗ (0.004) 0.000∗∗∗ (0.000) −0.028 (0.019) 0.002 (0.001) −0.024 (0.015) −0.033 (0.020) −0.010∗∗∗ (0.004) 0.174∗∗∗ (0.029) 0.080∗∗∗ (0.013) −0.024 (0.026) 0.237∗∗∗ (0.026) 0.082∗∗∗ (0.023)

0.031 (0.049) −0.009 (0.209) 0.527∗∗∗ (0.146) −0.178 (0.116) 0.098∗∗ (0.048) −0.042 (0.077) −0.042∗∗∗ (0.003) 0.001∗∗∗ (0.000) −0.121∗∗∗ (0.027) 0.014∗∗∗ (0.002) −0.138∗∗∗ (0.014) 0.304∗∗∗ (0.027) 0.008∗∗ (0.003) 0.191∗∗∗ (0.042) 0.238∗∗∗ (0.012) 0.003 (0.025) 0.112∗∗∗ (0.030) 0.157∗∗∗ (0.022)

29760 0.016

29744 0.016

29694 0.050

Note: The table reports coefficients of the ordered probit models, with social attitudes as dependent variables. The model also includes a constant term and interactions of province and year fixed effects, but their coefficients are suppressed for brevity. Standard errors (in parentheses) are clustered at the provincial level. ∗∗∗ , ∗∗ and ∗ denote significance at the 1%, 5% and 10% level, respectively.

page 229

229

Direct Participation

Trust

−0.001 (0.001)

Happiness Buddhist Taoist Muslim Catholic Protestant Other religion

29617 0.265

0.018∗ (0.010) 0.000 (0.023) −0.040∗ (0.021) 0.008 (0.024) −0.003 (0.014) −0.009 (0.016) 29601 0.267

29550 0.265

0.000 (0.001) −0.005∗∗ (0.002) −0.001 (0.002) 0.018∗ (0.010) 0.000 (0.022) −0.040∗ (0.021) 0.008 (0.024) −0.003 (0.014) −0.009 (0.016) 29512 0.266

(5) −0.001 (0.001)

0.025∗∗∗ (0.008) 0.058 (0.042) −0.042∗∗ (0.019) 0.022 (0.027) −0.005 (0.013) −0.015 (0.020) 29617 0.278

(6)

−0.006∗∗ (0.002) 0.020∗∗ (0.011) 0.059 (0.041) −0.041∗∗ (0.018) 0.022 (0.027) −0.005 (0.013) −0.015 (0.019) 29601 0.279

(7)

−0.003∗ (0.002) 0.024∗∗∗ (0.009) 0.058 (0.041) −0.041∗∗ (0.019) 0.021 (0.027) −0.005 (0.013) −0.015 (0.020) 29550 0.278

(8) 0.001 (0.001) −0.006∗∗ (0.003) −0.002 (0.002) 0.018∗ (0.011) 0.059 (0.041) −0.041∗∗ (0.018) 0.022 (0.027) −0.005 (0.013) −0.015 (0.019) 29512 0.279

Note: The table reports marginal effects of the probit models, with both direct participation and total participation as dependent variables. The model also includes a constant term, interactions of province and year fixed effects, and all previously used variables, but their coefficients are suppressed for brevity. Standard errors (in parentheses) are clustered at the provincial level. ∗∗∗ , ∗∗ and ∗ denote significance at the 1%, 5% and 10% level, respectively.

b3964-ch09

Observations Pseudo R2

0.021∗∗∗ (0.008) 0.001 (0.023) −0.041∗ (0.022) 0.007 (0.024) −0.003 (0.014) −0.009 (0.016)

−0.003∗ (0.002) 0.021∗∗∗ (0.008) −0.001 (0.022) −0.040∗ (0.022) 0.007 (0.024) −0.003 (0 .014) −0.008 (0.016)

(4)

Behavioral Finance: A Novel Approach

−0.005∗∗ (0.002)

(3)

Behavioral Finance: A Novel Approach – 9in x 6in

Sense of fairness

(2)

10:13

(1)

Total Participation

November 4, 2020

Religion, social attitudes and stock market participation

230

Table 7:

page 230

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation

page 231

231

is not significant, which contradicts the evidence documented in the literature (Guiso et al., 2008). When including all social attitudes, we only find evidence in favor of a role for sense of fairness. This can be attributed to the households’ gambling behavior: when they feel that they are in general unfairly treated, they may take a gamble on the stock market in order to improve financial situation, which is consistent with the prevalent preference for lottery-type stocks in China’s stock market documented in Zheng and Sun (2013). Comparing Tables 3 and 4 reveals that adding a sense of fairness reduces the magnitude and significance of the coefficients on Buddhist. This suggests that sense of fairness is helpful in explaining the higher propensity to invest in stocks by Buddhist households. For Muslims, however, the results are less convincing. Although the coefficient is slightly lower in a couple of cases, the significance level remains the same as in the baseline model. 4.4.

Causality

It is possible that the relationship between the religion and stock market participation is simply correlation rather than causality. Identifying the causal effect of religion on the stock market participation decision is an empirical challenge because of the endogeneity problem, which is a problem of much of the research in this field (Guiso et al., 2003; Renneboog and Spaenjers, 2012). This endogeneity can arise for a couple of reasons, such as reverse causality, omitted variables and measurement errors. For example, it is not clear whether the religious beliefs affect households’ willingness to participate in the stock market, as we claim, or some confounding factors affect both decisions to invest in the stock market and to become religious. To resolve the endogeneity problem, we adopt an instrumental variable (IV) strategy. Valid instruments should influence household financial behavior only through their impact on religious affiliations. To this end, we utilize the density of religious sites in the province of residence to instrument the variables of interests. On the one hand, it is unlikely to be correlated with current personal characteristics and other potentially confounding factors that influence stock market participation decisions. On the other hand, it represents the religious tradition of the province that affects individuals’ chance of being religious. The data on the number of religious sites at the provincial level

November 4, 2020

232

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

page 232

Behavioral Finance: A Novel Approach

are taken from the Spatial Explorer of Religion,6 which is constructed jointly by the University of Michigan, Purdue University and Wuhan University and allows us to explore spatial distribution of religious sites in China. Because the Spatial Explorer of Religion does not provide the information on the number of Catholic churches, we end up with four instruments and therefore exclude Catholics from the following causality analysis. The summary statistics of the instrumental variables can be found in Table 2. A noteworthy feature of our model is that both of the dependent variable (stock market participation) and endogenous regressors (religious affiliations) are dummy variables. Angrist (2001) examines the estimation strategies in this type of setting and shows that as long as the focus of estimation is on causal effects rather than regression parameters, the conventional two-stage least squares (2SLS) models are no less appropriate. Therefore, we use 2SLS to estimate our model. The first stage for the IV estimation can be written as: Xi = α + ϕ IV + δ Ci + εi

(6)

where IV is a vector of the instrumental variables, namely the provincial density of religious sites, and ϕ is the associated coefficient. We expect ϕ to be positive, because higher density of religious sites represents stronger religious tradition and is therefore associated with higher religiosity of the residents. The second stage is the same as Eq. (1) except that Xi is replaced by Xi∗ , the fitted value of Xi from the first stage. Table 8 reports the results of a 2SLS estimation with each religious dummy separately instrumented. As shown in the upper panel of Table 8, the coefficients on the instrumental variables (Buddhist temples, Taoist temples, Mosques, Protestant churches) in the first-stage regression are positive and highly significant. This is consistent with our expectation that people living in provinces with stronger religious tradition are more likely to be religious. Moreover, the F-statistics are high and above the value recommended to satisfy the “non-weak” instrument criteria established by Staiger and Stock (1997). Thus, we conclude that the

6

The data can be found from http://chinadataonline.org/religionexplorer/ index.html.

Religion and stock market participation: 2SLS estimation

Buddhist

(2)

(3)

(4)

0.035∗∗∗

(7)

(8)

(0.010) 1.856 (1.245)

−0.031∗∗∗ (0.008)

−0.048∗∗∗ (0.014)

−0.021 (0.646)

Protestant 27515 0.116

26010 0.010

26719 0.111

26503 0.113

Buddhist

Taoist

Muslim

Protestant

−0.161 (0.163) 27515 0.143

26010 0.011

26719 0.139

26503 0.133

First stage results Buddhist temples

4.657∗∗∗ (0.359)

Taoist temples

1.584∗∗ (0.809)

Mosques

15.371∗∗∗ (1.547)

Protestant churches 31.85

16.13

41.44

17.64

233

Note: The table reports estimated coefficients of the 2SLS models, with both direct participation and total participation as dependent variables. Religious denominations are instrumented with the provincial density of religious sites. The first-stage estimated coefficients on the instrumental variables are reported in the lower part of the table. The model also includes a constant term, interactions of province and year fixed effects, and all previously used variables, but their coefficients are suppressed for brevity. Standard errors (in parentheses) are clustered at the provincial level. ∗∗∗ , ∗∗ and ∗ denote significance at the 1%, 5% and 10% level, respectively.

b3964-ch09

First-stage F -statistic

1.172∗∗∗ (0.458)

Behavioral Finance: A Novel Approach – 9in x 6in

1.536 (3.968)

Muslim

Observations Pseudo R2

(6)

0.018∗

(0.012) Taoist

(5)

10:13

(1)

Total Participation

Influence of Religion and Social Attitudes in Stock Market Participation

Direct Participation

November 4, 2020

Table 8:

page 233

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

234

presence of a weak instrument is not an issue. The estimates in the second-stage regressions show that consistent with the benchmark regression results, the coefficients on Buddhist are significantly positive and those on Muslim significantly negative in both regressions on Direct participation and Total participation. These results indicate that the significant effects of being Buddhist and Muslim on stock market participation are likely causal.

5.

Conclusion

Although the past research has established the importance of religion for household financial behavior, almost all studies consider Western religions only (see, for example, Renneboog and Spaenjers, 2012; Noussair et al., 2013) and leave the causality issue unresolved. Given the significant differences between Western religions and Eastern religions, this study revisits the relationship between religion and stock market participation by also taking into account Eastern religions and addressing the endogeneity concerns. Using Chinese data from the CGSS, we find that the effects of religious affiliations on household stock market participation decisions are observed only for Buddhist and Muslim households: being Buddhist increases the probability of stock market participation by 2.2%, while being Muslim lowers the probability by 4.1%. In contrast, no significant differences in stock market participation are found between non-religious and other religious households. In addition, religiosity, as measured by the frequency of attendance at religious services, is negatively correlated with the probability of participation. More religious Muslim and Protestant households exhibit lower propensity to invest in stocks. We also find that different religious affiliations are associated with different social attitudes. In particular, compared with non-religious households, Buddhists are less trusting and have a lower sense of fairness, while Muslims display a higher level of happiness. Moreover, the higher stock market participation of Buddhists can be partially explained by differences in social attitudes. Finally, instrumenting religious affiliations with the density of religious sites in the province of residence, we find that the effects of

page 234

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation

page 235

235

being Muslim and Buddhist on stock market participation are likely causal.

References Abel, A. (2001). The effects of investing social security funds in the stock market when fixed costs prevent some households from holding stocks. American Economic Review, 91, 128–148. Adhikari, B. K. and Agrawal, A. (2016). Does local religiosity matter for bank risk-taking? Journal of Corporate Finance, 38, 272–293. Alesina, A. and La Ferrara, E. (2002). Who trusts others? Journal of Public Economics, 85 (2), 207–234. Almenberg, J. and Dreber, A. (2015). Gender, stock market participation and financial literacy. Economics Letters, 137, 140–142. Angrist, J. D. (2001). Estimation of limited dependent variable models with dummy endogenous regressors: Simple strategies for empirical practice. Journal of Business Economics and Statistics, 19, 2–16. Arru˜ nada, B. (2010). Protestants and Catholics: Similar work ethic, different social ethic. The Economic Journal, 120 (547), 890–918. Barber, B. M. and Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common stock investment. Quarterly Journal of Economics, 116, 261–292. Brown, J. R., Ivkovic, Z., Smith, P. A., and Weisbenner, S. (2008). Neighbors matter: Causal community effects and stock market participation. Journal of Finance, 63 (3), 1509–1531. Bogan, V. (2008). Stock market participation and the internet. Journal of Finance and Quantitative Analysis, 43 (1), 191–211. Chen, Y., Murgulov, Z., Rhee, S. G., and Veeraraghavan, M. (2016). Religious beliefs and local government financing, investment, and cash holding decisions. Journal of Empirical Finance, 38, 258–271. Christelis, D., Jappelli, T., and Padula, M. (2010). Cognitive abilities and portfolio choice. European Economic Review, 54 (1), 18–38. Cocco, J. (2005). Portfolio choice in the presence of housing. Review of Financial Studies, 18 (2), 1553–1604. Cole, S., Paulson, A., and Shastry, G. K. (2014). Smart money? The effect of education on financial outcomes. Review of Financial Studies, 2 7(7), 2022–2051. Cronqvist, H., Previtero, A., Siegel, S., and White, R. E. (2015). The fetal origins hypothesis in finance: Prenatal environment, the gender

November 4, 2020

236

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Behavioral Finance: A Novel Approach

gap, and investor behavior. The Review of Financial Studies, 29 (3), 739–786. Delis, M. and Mylonidis, M. (2015). Trust, happiness, and households’ financial decisions. Journal of Financial Stability, 20, 82–92. Diaz, J. (2000). Religion and gambling in Sin-City: A statistical analysis of the relationship between region and gambling patterns in Las Vega residents. The Social Science Journal, 37, 453–458. Francis, L. J., Ziebertz, H. G., and Lewis, C. A. (2003). The relationship between religion and happiness among German students. Pastoral Psychology, 51 (4), 273–281. Grinblatt, M., Keloharju, M., and Linnainmaa, J. (2011). IQ and stock market participation. Journal of Finance, 66 (6), 2121–2164. Guiso, L., Sapienza, P., and Zingales, L. (2003). People’s opium? Religion and economic attitudes. Journal of Monetary Economics, 50 (1), 225–282. Guiso, L., Sapienza, P., and Zingales, L. (2008). Trusting the stock market. Journal of Finance, 63 (6), 2557–2600. Hilary, G. and Hui, K. W. (2009). Does religion matter in corporate decision making in America? Journal of Financial Economics, 93, 455–473. Hong, H., Kubik, J. D., and Stein, J. C. (2004). Social interaction and stock-market participation. Journal of Finance, 59 (1), 137–163. Hopfe, L. M. and Woodward, M. R. (2011). Religions of the World, 12th edition. Upper Saddle River, NJ: Pearson Prentice Hall. Kn¨ upfer, S., Rantapuska, E. H., and Sarvim¨ aki, M. (2017). Formative experiences and portfolio choice: Evidence from the Finnish Great Depression. The Journal of Finance, 72 (1), 133–166. Miller, A. S. (2000). Going to hell in Asia: The relationship between risk and religion in a cross-cultural setting. Review of Religious Research, 40, 5–18. Miller, A. S. and Hoffmann, J. P. (1995). Risk and religion: An explanation of gender differences in religiosity. Journal for the Scientific Study of Religion, 63–75. Mookerjee, R. and Beron, K. (2005). Gender, religion and happiness. The Journal of Socio-Economics, 34 (5), 674–685. Noussair, C. N., Trautmann, S. T., Van de Kuilen, G., and Vellekoop, N. (2013). Risk aversion and religion. Journal of Risk Uncertainty, 47(2), 165–183. Overmyer, D. L. (2003). Religion in China today: Introduction. The China Quarterly, 174(2), 307–316. Oxtoby, W. and Amore, R. (2010). World Religions: Eastern Tradition, 3rd edition, Oxford University Press, Cambridge.

page 236

November 4, 2020

10:13

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch09

Influence of Religion and Social Attitudes in Stock Market Participation

page 237

237

Renneboog, L. and Spaenjers, C. (2012). Religion, economic attitudes, and household finance. Oxford Economic Paper, 64 (1), 103–127. Rosen, H. S. and Wu, S. (2004). Portfolio choice and health status. Journal of Financial Economics, 72 (3), 457–484. Staiger, D. and Stocks, J. (1997). Instrumental variables regression with weak instruments. Econometrica, 65 (3), 557–586. Van Rooij, M., Lusardi, A., and Alessie, R. (2011). Financial literacy and stock market participation. Journal of Financial Economics, 101 (2), 449–472. Yang, F. (2010). Religion in China under Communism: A shortage economy explanation. Journal of Church and State, 52 (1), 3–33. Zheng, Z. and Sun, Q. (2013). Lottery-like stock trading behavior analysis:Evidence from Chinese A-share stock market. Economic Research Journal (in Chinese, Jing Ji Yan Jiu), 5, 128–140.

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0010

Chapter 10

Investment Beliefs and Portfolio Risk-Taking — A Comparison between Industry Professionals and Non-Professionals

Magnus Jansson∗ , Sven Hemlin†, Doron Sonsino‡ and Carl-Christian Tr¨ onnberg†

Abstract Investment beliefs, serving as a bridge between high-level objectives and practical decision-making, are increasingly implemented in the investment industry. The present web-based study compares the beliefs of Swedish professional (N = 64) and non-professional (N = 278) investors, testing the links between investment beliefs and portfolio risk-taking in both samples. The results expose significant differences between the beliefs of professionals and others, also showing that the portfolio risktaking of non-professionals is susceptible to self-confidence and emotional effects while the professionals respond to investment beliefs and risk ∗ Corresponding author. Magnus Jansson, Gothenburg Research Institute (GRI). The School of Business, Economics & Law. The University of Gothenburg, P.O. Box 603, SE-40530 G¨ oteborg, Sweden; [email protected]. † Carl-Christian Tr¨ onnberg and Sven Hemlin are from the Gothenburg Research Institute (GRI), The School of Business, Economics & Law, The University of Gothenburg, P.O. Box 603, SE-40530 G¨ oteborg, Sweden; [email protected], [email protected]. ‡ Doron Sonsino is at the College of Law and Business (CLB), Ramat-Gan, Israel, and adjunct at the Economics department of Ben-Gurion University; [email protected].

239

page 239

November 4, 2020

240

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach attitude. The results confirm that disclosure of investment beliefs may reduce tensions between stakeholders and investment managers for the industry’s benefit. Keywords: Investment beliefs, portfolio risk-taking, pensions industry, dual processing theory, emotions

1.

Introduction

Investment beliefs are attracting increased interest in the asset management industry (The Investment Officer; August 8, 2019). Ambachtsheer (2007) concisely defines investment beliefs as beliefs regarding valuations and how financial markets function. Raymond (2008) similarly speaks of a broad set of beliefs regarding returngenerating processes. Several authors (Gray, 2009; Koedijk and Slager, 2011; Chambers et al., 2012; Rozanov, 2015) adopt the term investment philosophy to address the collection of investment beliefs adopted by the asset management firm. Woods and Urwin (2010), more practically, refer to working assumptions regarding the investment world that underlie and inform the decision-making of the firm. A July 2005 survey finds investment beliefs statements in public documents of 18 leading asset management firms with more than US$5,500 billion under management (Slager and Koedijk, 2007; exhibit 2). More recent industry surveys (Pensions and Investments; July 23, 2012) reveal that statements of investment beliefs are becoming widespread, with pension funds across the globe displaying a list of investment beliefs in their websites and official documents. The specific attributes composing the investment beliefs of investment management organizations vary between providers, but few themes constantly recur (Koedijk et al., 2010). The time horizon of investment is one of the recurrent themes. Asset managers may be long-term inclined, believing that long-term investment enhances value development, or relatively short-term focused claiming that rapid adjustment is essential for achieving competitive return. The Minnesota State Board of Investment (SBI) investment beliefs, for example, emphasize that SBI is a long-term investor, but the ability to pay benefits on a year-to-year basis is a key consideration in cases where short-term liquidity can be sacrificed for long-term return (http://mn.gov/sbi/documents/SBI%20Investment%20Belie fs.pdf). Other common dimensions of investments beliefs deal with

page 240

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

b3964-ch10

page 241

241

market rationality, the risk-return tradeoff, the ability to control risk and the advantage of diversification (see the Ontario Teachers’ Pension Plan beliefs at https://www.otpp.com/investments/perfor mance/investment-strategy/our-beliefs). Slager and Koedijk (2007) additionally emphasize the role of organizational norms and societal concerns. Indeed, many pension funds refer to these broader perspectives in recently advertised investment beliefs (see the discussion of investment beliefs at the Principles for Responsible Investment site https://www.unpri.org/). Rook (2012) illustrates that senior US and UK investment managers share common beliefs regarding sustainable investment, climate change and resource scarcity. Investment beliefs are often separated from investment policy, arguing that the mission and beliefs of the asset management firm should underlie its more concrete policy (Clark and Urwin, 2008; Fraser and Jennings, 2010; Woods and Urwin, 2010). Lydenberg (2011) claims that investment belief statements should serve as a bridge between high-level goals and practical decision-making, helping trustees and fiduciaries clarify their views on the nature of the financial markets through which they must operate. Few papers argue that the mere existence of published investment beliefs links with superior performance. Ambachtsheer (2007) connects the strong precrisis performance of the Harvard endowment Management Company (HMC) and the Ontario Teachers’ Pension Plan (OTPP) to home-grown coherent investment beliefs. Koedijk and Slager (2009) analyze the published reports of 40 pension funds and assets management firms managing more than US$10.3 billion. They find links between having published investment beliefs and improved riskreturn ratios. A 2018 survey among 18 global asset management companies relatedly suggests that the funds with high added value maintain their investment beliefs longer (FCLTglobal; August 30, 2018). The extent to which investment beliefs are shared by professional investors and their beneficiaries, however, has not yet received due consideration. Shared investment beliefs and a common understanding of how value is created may improve the ability of pension managers to meet their fiduciary obligations and decrease the risk of conflicts between the pension fund beneficiaries and their agents in the service provider chain (Johnson and de Graaf, 2009). The first purpose of this chapter is to investigate to what extent industry

November 4, 2020

242

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

professionals and non-professional investors share the same investment beliefs. A web-survey is administered to professionals employed by Swedish pension funds and asset management firms (N = 64) and to a convenience sample of non-professionals composed of students and private investors (N = 278). In addition to characterizing the beliefs of respondents in both groups, the survey administered two stylized asset allocation problems to test to what extent the portfolio risk-taking of professionals and others is affected by their investment beliefs. To control for psychological factors that often link with the willingness to take financial risk, the survey also elicits measures of individual propensity to take risk (Grable, 2000; Weber et al., 2002; Dohmen et al., 2011), self-confidence (Doran et al., 2010) and affective state (Lucey and Dowling, 2005). The non-professionals additionally rate their knowledge in finance, while the professionals evaluate the risk-taking norms in their organizations. The results of the survey point at significant differences between the financial beliefs of industry professionals and non-professional investors, also showing that the determinants of portfolio risk-taking vary between the two groups although their portfolios do not differ significantly. The professionals exhibit stronger belief in long-termed investment, the risk-return tradeoff and the premium for expertise, and their portfolio risk-taking is affected by their beliefs and their personal risk attitudes. The portfolio risk-taking of private investors, however, strongly responds to self-confidence and their transitory mood. The results connect with research proposing that the decisionmaking of professionals is generally more quantitative and deliberative, while non-professionals are more susceptible to psychological bias and emotional affects (Loewenstein et al., 2001; Slovic et al., 2005). By way of interpretation, the comparisons indeed confirm that publication of investment beliefs may bridge probable gaps between more calculated trustees and emotional savers, for the industry’s benefit (Slager and Koedijk, 2007; Johnson and De Graaf, 2009). The remainder of the chapter is organized as follows. The next section reviews the research on how expertise and emotions may influence the willingness to take stock market risk. We then briefly introduce the investment beliefs examined in the current survey, leading to the main research questions. The method, results and discussion sections follow in order.

page 242

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Investment Beliefs and Portfolio Risk-Taking

2.

page 243

243

Expertise, Emotions and Portfolio Risk-Taking

The willingness to take financial risk has shown to be affected by both objective and subjective measures of domain knowledge. Wang (2009) finds that objective knowledge, subjective knowledge and financial risk-taking positively correlate in a sample of N = 524 internet participants, with subjective knowledge mediating the objective knowledge and risk-taking link. Tang and Baker (2016) similarly find that objective knowledge and self-esteem positively affect the willingness to take financial risk among few thousand participants in a longitudinal US survey. The effect of self-esteem is both direct and indirect, through subjective financial knowledge. Hadar et al. (2013) argue that investors who perceive themselves as knowledgeable are more confident and their stronger confidence drives them to accept more risks. Manipulations that increase consumers’ knowledge while decreasing their subjective knowledge decreased the willingness to invest in financial assets. While professionals’ knowledge in topics related to their expertise naturally exceeds the knowledge of non-professionals (McKenzie et al., 2008; Baˇcov´a et al., 2017), direct comparisons between the financial risk-raking of professionals and others are scarce and the results are inconclusive. Holzmeister et al.’s (2020) recent survey of N = 2213 finance professionals and N = 4559 non-professionals from nine countries does not find a difference in the willingness of professionals and non-professionals to invest in diverse return distributions. The results of smaller sample studies sometimes discover larger or smaller willingness to invest among the professionals. Lambert et al. (2012) compares the survey investment decisions of loan officers (N = 20) and students (N = 64), confirming that investments increase with perceived knowledge in both samples, although the loan officers’ investments are about half smaller. In contrast, Thoma et al. (2015) find stronger willingness to take financial risk among N = 53 banking industry workers compared to N = 57 non-banking respondents. Risk perception studies generally argue that experts tend to have a more specific and quantitative understanding of risk, while the risk perception of lay persons is more likely to be influenced by emotional factors (Loewenstein et al., 2001; Slovic et al., 2005). The differences in risk perception and risk-taking of experts and non-experts are

November 4, 2020

244

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

frequently discussed in light of the theory of dual processing systems (Sloman, 1996; Stanovich and West, 2000; Kahneman, 2003). According to the theory, human thinking can be broadly categorized as either fast, affective and intuitive or as slow, deliberative and analytical. Experts’ perception of risk tends to be influenced more by the deliberative analytic system (risk as analysis), while lay people’s risk perception is affected more by the intuitive and emotionally triggered system (risk as feeling; Slovic and Peters, 2006). Indeed, Diacon (2004) finds that investment advisors (N = 41) exhibit weaker aversions to loss and complexity, while displaying more trust in providers and regulators compared to lay consumers (N = 123). However, other studies, including the recent comprehensive Holzmeister et al. (2020) nine countries comparison, could not point at systematic differences between the risk perceptions of finance professionals and others (see also Olsen, 1997; Sachse et al., 2012). The impact of emotions on the behaviors of investors and traders has been illustrated in diverse studies with professional and nonprofessional participants (Lucey and Dowling, 2005). In experimental studies, induced positive affect led participants to overestimate the probability of a financial reward (Nygren et al., 1996) and exhibit stronger risk appetite when trading experimental assets (Butler and Cheung, 2019). Negative affect brought parallel opposite consequences (Kuhnen and Knutson, 2011; Aldrovandi et al., 2017). The results regarding the susceptibility of finance professionals to emotions, however, are mixed again. Some studies prove that longer industry experience reduces the impact of emotions on trading behavior (Fenton-O’Creevy et al., 2011, 2012), while others illustrate that professionals strongly respond to affect in intraday trading although their emotional response leads to suboptimal performance (Lo et al., 2005; Locke and Mann, 2009). In Coates and Herbert (2008), day traders morning testosterone levels predict daily profitability, while cortisol levels rise with the volatility of individual returns and implied market-volatility estimates. The evidence regarding financial risk-taking among professional and non-professional investors is, in summary, inconclusive. While the literature appears to agree that objective and subjective knowledge bring stronger willingness to take financial risk, the results of comparisons between the willingness of professionals and others to take financial risk are mixed. Similarly, while dual processing research

page 244

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

b3964-ch10

page 245

245

predicts that professionals would exhibit stronger rationality in financial decision, empirical studies show that the susceptibility of finance professionals to emotional bias varies, depending on contexts and tasks. This chapter contributes to these lines of research by comparing the stock market risk-taking of Swedish professional and nonprofessional investors in stylized survey investment scenarios, and exploring the roles of investment beliefs as well as proclaimed knowledge and psychological attributes on financial risk-taking in the two samples. 3.

Investment Beliefs

Koedijk and Slager (2009) identified 12 types of investment beliefs categorized into 4 groups: beliefs concerning the financial market (e.g., importance of risk diversification); beliefs about the investment process (e.g., long-termed investment vs. frequent portfolio adjustment); organizational beliefs, that is, how does the management promote investment effectiveness (e.g., importance of team decisions); and beliefs about sustainability and corporate governance (e.g., good corporate governance results in higher earnings). For the present study, we devise six investment beliefs’ indices referring to major characteristics of the capital markets and possible ways to achieve high returns and control risk. Each index is constructed from three items, where the respondents rank their agreement with each item in a 1–5 Likert scale. The next paragraphs discuss the six indices, outlining the motivation for including these indices in a comparative study of the beliefs and financial risk-taking of industry professionals and others (see also Jansson et al., 2018). Appendix 1 presents the items used to derive each index. Rationality: The first index deals with belief in the rationality or efficiency of the financial markets. The efficient market hypothesis (Fama, 1970) asserts that the stock market instantly reflects all relevant news and the stock return process is a random walk (Malkiel, 2003). Investors who believe in market efficiency may be less affected by short-termed sentiments and show inclination for long-term investment rather than constant pursuing of mispricing opportunities. It is unclear, however, whether professionals should rank higher in their belief in market rationality, as by the common

November 4, 2020

246

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

view the professionals are those who instantly react to mispricing, keeping the markets efficient (Ericsson et al., 2005). Diversification: Another principle of modern finance is that diversification of risks is essential for achieving the best risk-return ratios (Markowitz, 1952; Sharpe, 1964; Lintner, 1965). Our second index measures investors’ belief in the merits of diversification. Again, as professionals may emphasize particular asset selection and response to mispricing opportunities (cf., Morrin et al., 2002), we are uncertain if their belief in diversification will be stronger. It is interesting yet to test if such belief links with the willingness to take stock market risk within the two samples. Time horizon: In a survey of 180 investment managers and top executives, Guyatt (2005) found that long-term investment is considered the best way to improve performance. Over 30% of Guyatt’s respondents stated that the most effective method to improve portfolio performance is to lengthen the investment horizon. The third item in our investment beliefs accordingly deals with the optimal time horizon of investment. Again, the more conservative view endorsing long-term investment can be challenged by investors who believe in asset picking and riding mispricing opportunities. Expertise: The efficient market hypothesis also relates to beliefs about the virtues of financial expertise. Believers in the efficiency of the financial market might de-emphasize the importance of expertise, arguing that professional investors are in no better position to predict future asset prices than lay people (Malkiel, 2003, 2005). Opponents of the efficient market hypothesis oppositely argue that the markets frequently appear inefficient as investors are irrational and cognitively biased (Barberis and Thaler, 2005; Barberis, 2018). The inefficiencies may generate mispricing opportunities that the expert asset manager may skillfully exploit. Risk-return: Another basic principle of finance theory is that taking more risk is essential for achieving higher average return. Again, however, the theoretical principle is not uniformly accepted. While empirical research identifies risk factors that consistently generate excess return (Fama and French, 1992, 1993), behavioral studies show that lay investors intuitively categorize risky companies as bad investments, expecting that the stocks of high-risk firms would bring

page 246

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

b3964-ch10

page 247

247

disappointing return (Ganzach, 2000; Shefrin, 2001). We accordingly anticipate that the professionals’ belief in the risk-return link will be stronger. Risk-control: Published investment beliefs typically include explicit statements regarding the financial risk management of the organization. Investment belief 5 of the Australian ACT government, for instance, asserts that risks should be viewed both qualitatively and quantitatively with particular focus given to the nature and likelihood of extreme events (https://www.act.gov.au). The Dutch PMT, to take a different example, asserts that risk management is designed to achieve the objective of generating the required excess returns within the framework of tight risk control (https://www.bpmt.nl/). While professional experience can increase the confidence of investors in the manageability of risks, it is oppositely possible that long acquaintance with the volatile markets would diminish such belief. The risk-control index is included in our list to compare the beliefs of professionals and others with this respect and test if belief in the controllability of financial risks correlates with portfolio risk-taking.

4.

Research Questions

The literature review motivates the following four main research questions: (1) Do professional and non-professional investors share similar investment beliefs? (2) Given the mixed findings regarding the portfolio risk-taking of professional and non-professional investors, can we find significant differences between the risk-taking of two such samples in an original survey assignment? (3) Do the investment beliefs of the responders reflect in their portfolio risk-taking? (4) Can we support the hypothesis that the portfolio risk-taking of professionals is more rational while the portfolio risk-taking of non-professionals responds to psychological attributes?

5. 5.1.

Method Sample and Procedure

Data were collected by means of a web-based survey constructed in Qualtrics. A link to the survey was distributed by email to Swedish

November 4, 2020

248

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

professional and non-professional investors and the participants could respond within a week. Participation was voluntary and provided no compensation. The professionals were fund managers and financial analysts employed at investment firms that are members of the Swedish Investment Fund Association (Fondbolagens F¨ orening). An initial contact was made with the CEOs or heads of investment in each firm. The aim of the research was presented in a formal letter with a request to provide contact information for fund managers and financial analysts who may participate in the survey. Of the 28 investment firms approached, 17 accepted the invitation. After receiving the contact information of potential respondents, we sent a welcome letter describing the study by email. A week later, the potential respondents received another email with a link to the survey. The survey closed on September 9, 2016, one week after sending the emails with the link. In total, 137 questionnaires were distributed and 64 were returned completed, representing a response rate of 46.7%. The professional sample is predominantly male, with 60 males and only 4 females, and the mean age is about 41 years. A total of 23 respondents were employed by major Swedish public pensions funds (AP-fonderna) and 41 were employed by private investment banks or mutual fund companies. Four of the 64 respondents had a top management position such as CEO, 9 were senior investment officers, 40 were fund managers and 11 were financial analysts. Respondents’ average work experience in the financial industry was 15 years. The non-professional convenience sample consisted of 278 individuals (156 males, 122 females, mean age 48.3 years) with varying experience in the stock market. The participants were recruited by advertising the survey among Gothenburg University students and among members of the Swedish association for small private shareholders (Aktiespararna). The response rate was 19.1%. For this chapter, we treat the non-professional sample as single unit, ignoring some differences between the students and the private investors (details will be provided on request). The self-rated financial knowledge of the 278 non-professional participants in a five-point scale (ranging from 1 = “No knowledge at all” to 5 = “Very good knowledge”) averaged at M = 2.05, with standard deviation SD = 1.05.

page 248

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

5.2.

b3964-ch10

page 249

249

Questionnaire and Measures

To test for psychological effects on risk-taking, the survey elicited indices of financial risk preference, financial self-confidence and affective state (see Appendix 1 for the items composing each index). Respondents’ attitude to financial risk was measured using four items in the spirit of Barclays psychometric scale of investor attitudes (Weber et al., 2013). An index of self-confidence in financial domains was similarly obtained using four items measured on a five-point Likert scale. Affective state was measured using the 10-items short version of the Positive and Negative Affect Scale (PANAS; Thompson, 2007), but since negative affect did not link with the main variables, we only discuss the positive index henceforth. The professional participants were also presented with four items dealing with the risktaking norms in their organization (see the organizational risk norm items in Appendix 1). Our main dependent variable, portfolio risk-taking, was measured by the percentage that respondents allocate to stocks in the two allocation scenarios presented in Appendix 2. In the first scenario, the participants divided an investment budget between six international stock indices and two bond portfolios assuming investment periods of 3 months and 2 years. The second scenario presents six alternative splits between large cap Swedish stocks and 2-years bonds, asking the participants to select their preferred allocation for 3-months and 2-years periods. Portfolio risk-taking was measured by the average proportion allocated to stocks in the four assignments. Short-run and long-run portfolio risk-taking were defined based on the 3-months and 2-years allocations. The allocation tasks presented to the professionals adopted a fund management cover, asking the respondents to assume they have free mandate to allocate their fund’s capital between few alternatives. The non-professionals were presented with identical scenarios, but the cover was personal asking the respondents to decide on the allocation of their savings. Since the two allocation problems only referred to large stock indices or large cap stocks, the portfolio risk-taking measure does not necessarily represent the readiness to invest in particular stocks, but measures the willingness to invest in large indices or stocks when the alternative consists of almost risk-free governmental bonds.

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

6.1.

page 250

Behavioral Finance: A Novel Approach

250

6.

b3964-ch10

Results Investment Beliefs, in General

The mean ratings of all respondents (N = 342) for each investment belief are presented in Table 1. The responders appear to perceive the stock market as irrational rather than rational (N = 85 with rationality index > 3 vs. N = 169 with rationality < 3; p < 0.01) and appear to have low confidence in the controllability of financial risks (N = 100 with risk-control index > 3 vs. N = 190 with index < 3; p < 0.01). The other investment beliefs are clearly supported, with the mean ratings significantly exceeding the midpoint of the Likert scale. The belief in long-term investment horizon is the strongest (N = 217 with horizon > 3 vs. N = 71 with horizon < 3), but financial expertise is more controversial (N = 172 showing expertise index > 3 vs. N = 108 with expertise < 3). A total of 10 significant correlations between investment beliefs were identified. Belief in the rationality of the market positively correlates with belief in expertise, dismissing the point of view that markets are efficient to the extent that the industry has no productive role. Diversification, risk-return and risk-control mutually correlate revealing that investors may concurrently believe in index investing and risky return chasing. Table 1: Investment beliefs and their correlations (complete sample; N = 342)

1. 2. 3. 4. 5. 6.

Rationality Diversification Time horizon Expertise Risk-return Risk-control

M

(SD)

2.8∗∗ 3.3∗∗ 3.4∗∗ 3.2∗∗ 3.3∗∗ 2.8∗∗

(0.6) (0.6) (0.7) (0.7) (0.7) (0.6)

α

1

0.71 0.61 0.09 0.62 −0.02 0.65 0.17∗∗ 0.73 0.15∗∗ 0.64 0.24∗∗

2

3

4

5

0.08 0.12∗ 0.15∗∗ 0.17∗∗

−0.10 −0.12∗ −0.05

0.18∗∗ 0.21∗∗

0.18∗∗

0

Note: The left columns present the mean (M), standard deviation (SD) and Cronbach’s alpha (α) for each investment belief. The asterisks at the left most column report the results of testing the hypothesis that the investment belief equals 3. The right panel shows the correlations between the investment beliefs. Significant correlations are shaded. ∗ = p < 0.05, ∗∗ = p < 0.01 throughout the chapter.

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Investment Beliefs and Portfolio Risk-Taking

page 251

251

Table 2: Investment beliefs of professionals and non-professionals Group Professionals (N = 64) Rationality Diversification Time horizon Expertise Risk-return Risk-control

2.7 3.3 3.7 3.7 3.6 2.8

(0.6) (0.7) (0.6) (0.7) (0.7) (0.6)

Non-professionals (N = 278) 2.8 3.3 3.3 3.0 3.3 2.8

(0.6) (0.6) (0.7) (0.6) (0.7) (0.6)

T -stat −0.8 −0.2 3.5∗∗ 6.9∗∗ 4.2∗∗ 0.4

Note: The table presents the mean investment belief for each group (with the standard deviation in brackets) and the results of a T -test for comparing the two samples. The shaded rows mark the dimensions where the two groups significantly differ. ∗∗ p < 0.05.

6.2.

Do Professionals and Non-Professionals Share the Same Investment Beliefs?

Table 2 tests the extent to which professionals (N = 64) and nonprofessionals (N = 278) endorse similar investment beliefs. The table shows that both groups have little confidence that the stock market is rational. They also roughly share similar beliefs in the virtues of diversification and the controllability of financial risks. The two groups however strongly differ in time horizon, expertise and riskreturn. The professionals score higher in each of these three dimensions, displaying stronger belief in the role of the industry and the possibility to chase return by long term investment and calculated risk-taking. 6.3.

The Portfolio Risk-Taking of Professionals and Non-Professionals

The portfolio risk-taking statistics are presented in Table 3. The mean allocation to stocks is 43% and the spread is very low with standard deviations around 3.8% in both samples. The hypothesis that the professionals or the non-professionals allocate 50% of their funds to stocks is easily rejected as only few respondents allocate more than 50% to the stock alternatives. Both groups allocate more

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

page 252

Behavioral Finance: A Novel Approach

252

Table 3: Portfolio risk-taking of professionals and non-professionals Group

Portfolio risk-taking 3-months 2-years T -stat

Sample (N = 342)

Professionals (N = 64)

Non-professionals (N = 278)

43 (3.9) 46 (4.9) 41 (3.8) 21.5∗∗

43 (3.8) 44 (4.9) 41 (3.5) 7.8∗∗

43 (3.9) 46 (4.9) 41 (3.8) 20.2∗∗

T -stat 1.6 2.3∗ 0.3

Note: The method of the table is similar to the one of Table 2. ∗ p < 0.1; ∗∗ p < 0.05.

to stocks in the 3-months allocation tasks compared to the longer 2-years assignments. The professionals take less risk when constructing the 3-months portfolios, but the differences disappear in the 2years allocation. 6.4.

Do Investment Beliefs Affect Portfolio Risk-Taking?

The left panel of Table 4 shows the results of regressing the portfolio risk-taking of professionals and non-professionals on the six investment beliefs. The regressions are separately run for each group and a standard F -test is used to test the equality of the estimated coefficients. The six beliefs interestingly explain more than 17% of the variance in the risk-taking of professionals while explaining less than 8% of the variance in the risk-taking of non-professionals, but drawing conclusions from this comparison is dangerous since the samples are very different in size and the convenience sample of non-professionals is quite diverse. A more puzzling result is a reversed effect of expertise and time horizon on the portfolio risk-taking of the respondents in the two groups. While the risk-taking of the professionals increases with belief in expertise and decreases with belief in time horizon, the effects of these two beliefs on the risk-taking of the non-professionals are just opposite. The reversed effects are robust and results similar to those in Table 4 emerge when the dependent variable is the 3-months or the 2-years portfolio risk-taking. Iterated removal of insignificant effects reconfirms the puzzle, as illustrated at the right panel of Table 4. The reversed effect of expertise can be tentatively

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

page 253

Investment Beliefs and Portfolio Risk-Taking Table 4:

Investment beliefs effects on the portfolio risk-taking Regressions on All Six Investment Beliefs Prof (N = 64)

Rationality Diversification Time horizon Expertise Risk-return Risk-control Intercept R2

253

Results after Iterated Removal of Insignificant Effects

Others Prof Others (N = 278) F -test (N = 64) (N = 278)

−0.05 (0.80) −0.16 (0.83) −1.46

−0.20 (0.38) 0.40 (0.38) 0.81∗

(0.80)

(0.36)

1.54∗

−1.17∗∗

(0.70) 0.71 (0.82) −2.50∗∗

(0.39) 0.00 (0.36) −0.40

(0.97) 47.3∗∗ (5.0) 17.3%

(0.40) 44.6∗∗ (2.5) 7.9%

F -test

0.87 0.55 0.01 0.01

−1.42

0.86∗∗

(0.78)

(0.35)

1.40∗

−1.23∗∗

(0.63)

(0.38)

0.01 0.00

0.45 0.05 0.63

−2.09∗∗ (0.83) 48.5∗∗ (4.4) 16.2%

44.3∗∗ (1.8) 7%

0.40

Note: The method of the table is explained in Section 6.4. Shaded rows denote the beliefs by which the two samples significantly differ. ∗ p < 0.1; ∗∗ p < 0.05.

explained by smaller inclination of non-professionals who believe in financial expertise to take stock market risk independently. Alternatively, it is possible that non-professional investors who believe in expertise may dislike the large stock indices and large cap stocks that are being offered in our allocation scenarios. The reversed effect of time horizon may be ad hoc attributed to different views regarding the investment periods of 3 months and 2 years implemented in the survey. 6.5.

Psychological Determinants of Portfolio Risk-Taking

The two horizontal panels of Table 5 examine the psychological attributes of the professional and non-professional (other) respondents, showing the simple correlation of each attribute with

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

254

b3964-ch10

page 254

Behavioral Finance: A Novel Approach

Table 5: Personal attributes and their correlations with portfolio risk-taking Panel A: Professionals (N = 64) M (SD) 1. 2. 3. 4.

Financial risk preference Financial self-confidence Positive affect Organizational risk norms

4.3 2.8 3.6 2.4

1

(1.3) (0.8) 0.12 0.25∗ (0.7) (0.7) −0.02

2

3

4

0.27∗ 0.22∗

0.09

0.06 0.08 0.00

0.09 0.10 0.05

0.07 0.11 0.04 4

Portfolio risk-taking 3-months risk-taking 2-years risk-taking

43 (3.8) 44 (4.9) 41 (3.5)

0.28∗ 0.26∗ 0.24

Panel B: Others (N = 278)

M (SD)

1

2

3

1. 2. 3. 4.

2.9 2.2 3.2 2.0

(1.5) (0.8) (0.8) (1.1)

0.51∗∗ 0.39∗∗ 0.51∗∗

0.46∗∗ 0.65∗∗

0.47∗∗

43 (3.9) 46 (4.9) 41 (3.8)

0.27∗∗ 0.16∗∗ 0.35∗∗

0.34∗∗ 0.22∗∗ 0.41∗∗

0.35∗∗ 0.27∗∗ 0.37∗∗

Financial risk preference Financial self-confidence Positive affect Self-rated Knowledge

Portfolio risk-taking 3-months risk-taking 2-years risk-taking

Note: The method of the table is explained in Section 6.5. ∗ p < 0.1;

∗∗

0.36∗∗ 0.28∗∗ 0.38∗∗ p < 0.05.

portfolio risk-taking in the two samples. The professionals of the current convenience sample score significantly higher in financial risk preference and self-confidence and also exhibit more positive affect than the non-professionals (p < 0.01 in all three comparisons). As anticipated, self-rated knowledge positively correlates with the portfolio risk-taking of the non-professionals, but the effect is quantitatively mild (mean risk-taking 44.5 for those that rank their knowledge at 4–5 compared to 41.2 for those that rank their knowledge at 1–3; p < 0.01).1 More interestingly, the correlations between each of the personal attributes (risk preference, self-confidence and positive 1

N = 92 of the non-professionals rated their knowledge at 1–2, N = 94 at 3 and N = 92 at 4–5. The portfolio risk-taking of the respondents with self-rated knowledge 4 or 5 exceeds the risk-taking of those with knowledge 1, 2 or 3. The low knowledge sample is predominantly female while the high knowledge sample is predominantly male. Regressions of portfolio risk-taking on gender, knowledge

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

b3964-ch10

page 255

255

affect) and portfolio risk-taking are always positive and highly significant for the non-professionals, while only risk-receptiveness correlates with the portfolio risk-taking of the professionals (compare the two correlation panels in Table 5). 6.6.

Summarizing Regressions

Table 6 reports the results of OLS and logistic regressions that test the response of portfolio risk-taking to investment beliefs and personal characteristics concurrently. Investment beliefs are concisely measured using standardized factors that were extracted in principal component analysis.2 For the logistic regressions, respondents are classified as high risk-takers when their portfolio risk-taking is at least 45%. The results of the OLS and the logistic regressions similarly reconfirm that investment beliefs affect the portfolio risktaking of professionals and non-professionals, but the risk-taking of the professionals is only affected by their individual risk preference while the non-professionals strongly responds to self-confidence and positive affect. The results are robust and similar bottom-line conclusions emerge when investment beliefs are measured more closely using the indices of Table 1, when the sample of non-professionals is split depending on self-rated knowledge, when gender is taken into account and in other robustness analyses. 7.

Discussion

While the literature on the determinants of investors’ inclination to take stock market risk is large and diverse (cf., Calvet and and an interaction term suggest that gender or the interaction does not affect risktaking when stated knowledge is controlled. The results reported in the main text are robust to splitting the non-professionals to low knowledge and high knowledge groups. 2 The factors were separately extracted for the two samples. The formulas are 0.20 × rationality − 0.09 × diversification − 0.40 × horizon + 0.63 × expertise − 0.57 × risk-return + 0.14 × risk-control, for the professionals (eigen value 1.05), and −0.37 × rationality − 0.17 × diversification + 0.22 × horizon − 0.39 × expertise − 0.33 × risk-return − 0.38 × risk-control, for the non-professionals (eigen value 1.21). The factors correlations with portfolio risk-takings are 0.27 (p < 0.01) for the professionals and 0.20 (p < 0.01) for the non-professionals.

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

256

Table 6:

Dependent Variable

Summarizing regressions

OLS Regressions

Logistic Regressions

Portfolio Risk-Taking

1risk-taking≥45

Professionals (N = 64)

Others (N = 278)

Professionals (N = 64)

Others (N = 278)

Investment beliefs

1.11∗ (0.47)

0.82∗∗ (0.22)

0.77∗ (0.34)

0.59∗∗ (0.15)

Financial risk attitude

0.82∗ (0.36)

—

0.59∗ (0.27)

—

Self-confidence

−0.20 (0.61)

0.91∗∗ (0.36)

−0.45 (0.42)

0.60∗∗ (0.24)

Positive affect

0.21 (0.65)

0.82∗∗ (0.29)

−0.45 (0.47)

0.40∗ (0.19)

16%

0.56∗ (0.27) 22%

Self-rated knowledge R2

—

0.32 (0.18) —

Note: The method of the table is explained in Section 6.6. Financial risk-attitude does not affect the risk-taking of the non-professionals when the other covariates are controlled. The logistic regressions account for self-rated knowledge although the effect is not-significant. The results are robust to removal of the knowledge variable. ∗ p < 0.1; ∗∗ p < 0.05.

Sodini, 2014), the influence of investment beliefs on financial risktaking has not been previously explored. The current study attempts to fill this gap. Building on previous investment beliefs research by Ambachtsheer (2007), Slager and Koedijk (2007) and Koedijk and Slager, (2009), we investigate the endorsement of six major beliefs concerning the capital markets by samples of professional and nonprofessional investors. The comparisons reveal that professionals and non-professionals share a common perception of the stock market as irrational rather than rational, and show similar doubt regarding the possibility to control market risks. The professionals, however, exhibit stronger belief in long-termed investment, the risk-return principle and the premium for expertise. The professionals’ stronger belief in the virtues of long investment horizon and their stronger endorsement of the risk-return tradeoff

page 256

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

b3964-ch10

page 257

257

are not surprising. Professionals can be expected to endorse the principles of modern portfolio more strongly than non-professionals, while behavioral studies prove that non-professionals exhibit diverse types of bias in the perception of risk and its correlation with return (Ganzach, 2000; Duxbury and Summers, 2004; Ricciardi, 2008; Wang et al., 2011). The gap between the more rational views of professionals and possibly biased views of non-professionals currently reflect in weaker endorsements of the horizon and risk-return beliefs by the non-professionals. We are neither surprised by the strong general belief in the role of expertise in the financial market and the professionals’ stronger confidence in financial expertise. Previous research has shown that professionals and non-professionals alike expect more skillful decision from experts compared to lay investors (T¨orngren and Montgomery, 2004; Peterson et al., 2015). It is argued that for this reason nonprofessionals search for the advice of investment experts (Huber et al., 2010) and put their trust and money in the hands of fund managers (Carlander et al., 2013). The stronger belief of experts in the merits of financial expertise may then unsurprisingly follow from motivational or professional identity concerns (Cohn et al., 2017), while the more sceptical view of the non-professionals can be attributed to emotional factors such as the cognitive dissonance that arises in delegation of investment funds (Chang et al., 2016). Pension funds research proposes that the display of coherent investment beliefs correlates with stronger performance of leading asset management companies (Ambachtsheer, 2007; Koedijk and Slager, 2009; FCLTglobal; August 30, 2018). It has also been argued that clear statements of investment beliefs may alleviate tensions between the agents at the savings industry chain and the principal investors (Johnson and de Graaf, 2009). In addition to pointing at significant differences in major investment beliefs of professionals and non-professionals, the results of our survey confirm the hypothesis that the non-professionals’ risk-taking decisions are susceptible to psychological effects, while the professionals are relatively immune to these same effects (Loewenstein et al., 2001; Slovic and Peters, 2006). The results thus suggest that effective use of investment beliefs may decrease the hazard of conflict in periods of booms or busts where investors worry about too conservative or too risky allocation of their savings (cf., Siev¨anen, 2012). The publication and delivery of clearly

November 4, 2020

258

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

stated beliefs, to take another perspective, may decrease tensions on concrete matters such as the choice between passive index-tracking or more active asset selection by the pension fund’s investment officers (Walden, 2015). Acknowledgments This work was supported by a grant from the Torsten S¨ oderberg foundation under grant E31/13. Appendix 1: The Main Items of the Survey Investment beliefs: Rationality The stock market tends to react emotionally* The stock market is irrational* The stock market has a short-term perspective* Investment beliefs: Diversification Spreading your risks usually gives you a higher return to a lower risk Spreading your risks is usually the most efficient way of achieving a high return in relation to a given risk Spreading your risk reduces your chance of a very high return* Investment beliefs: Time horizon Long-term investments contribute to better value development of the invested capital than short-term investments There is a positive link between a long-term perspective and high return relatively to the risk taken By having a short-term perspective and quickly adjusting to the market, you as investor will have better conditions for achieving a high return on your assets* Investment beliefs: Expertise Professional investors are better at judging financial risk than lay people It is impossible for anyone, including professional investors, to forecast the value development of financial assets such as stocks*

page 258

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Investment Beliefs and Portfolio Risk-Taking

page 259

259

Professional investors have no better chance of achieving high stock market returns than random chance* Investment beliefs: Risk and return There is a strong positive connection between risk-taking and return. Investments with higher risk are more likely to give a high return In the long term, risk-taking leads to lower return on investments* You have to take financial risks to achieve high returns Investment beliefs: Risk-control Stock market risk can to a large extent be controlled Stock market risk can to a large extent be predicted Risks affecting the stock market cannot be reliably estimated* Financial risk preference It is likely I would invest a significant sum in a high-risk investment I am a financial risk-taker Even if I experienced a significant loss on an investment, I would still consider making risky investments I like to take financial risks Financial self-confidence I am significantly better than most investors in selecting assets that deliver a high return I think it is almost impossible to predict future returns on stocks* I am skilled in predicting how the financial market will develop in a time horizon of 6 months I am skilled in predicting how the financial market will develop in a time horizon of 2 years Positive affective state (PANAS, Thompson, 2007) To what extent do you, at this moment, feel: Active/Alert/Attentive/Determined/Inspired Organizational risk norms Taking financial risks is encouraged in our organization

November 4, 2020

10:14

260

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

In our organization, risk control is subordinate to achieving a high return In our organization, it is acceptable to lose money in a quest for high returns In our organization, risk-taking should always minimized∗ ∗ Inverted variable Appendix 2: The Portfolio Risk-Taking Scenarios The table shows the portfolio risk-taking scenarios as presented to the professional investors. The non-professionals’ scenarios were identical, except for referring to the personal savings of the respondents instead of referring to the fund’s capital. Scenario 1 Imagine that you have an open mandate to allocate your fund’s capital between the following assets. How would you allocate the capital assuming investment horizons of 3 months and 2 years? You can allocate between 0% and 100% of the capital to each asset, but the total sum in each column should be 100%.

Assets Swedish treasury bonds with duration of 2 years Swedish treasury bonds with duration of 10 years Dow Jones, the American stock exchange in New York OMX, the Swedish stock exchange in Stockholm Nikkei, the Japanese stock exchange in Tokyo BSE, the Indian stock exchange in Bombay Hang Seng, the Chinese stock exchange in Hong Kong Total

Investment for 3 Months

Investment for 2 Months

%

%

%

%

%

%

%

%

%

%

%

%

%

%

100%

100%

page 260

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Investment Beliefs and Portfolio Risk-Taking

page 261

261

Scenario 2 Imagine that you have an open mandate to allocate your fund’s capital. What would be your preferred mix of bonds and stocks for an investment horizon of 3 months and 2 years? Choose one option in each column. Balance bonds between and stocks

Investment for 3 months

Investment for 3 years

100% Swedish bonds with a duration of 2 years 80% Swedish bonds with a duration of 2 years and 20% Swedish stocks (Large cap) 60% Swedish bonds with a duration of 2 years and 40% Swedish stocks (Large cap) 40% Swedish bonds with a duration of 2 years and 60% Swedish stocks (Large cap) 20% Swedish bonds with a duration of 2 years and 80% Swedish stocks (Large cap) 100% Swedish stocks (Large cap)

References Aldrovandi, S., Kusev, P., Hill, T., and Vlaev, I. (2017). Context moderates priming effects on financial risk taking. Risks, 5(1), 18. Ambachtsheer, K. (2007). Should (could) you manage your fund like Harvard or Ontario teachers. In K. Ambachtsheer (Ed.), Pension Revolution. A Solution to the Pension Crisis, Chapter 15. Hoboken: John Wiley & Sons. Baˇcov´a, V., Dudekov´a, K., Kostoviˇcov´a, L., and Bal´ aˇz, V. (2017). Financial planning for retirement in Young Adults: Interaction of professional experience, knowledge, and beliefs. Studia Psychologica, 59(2), 84. Barberis, N. C. (2018). Psychology-based models of asset prices and trading volume (No. w24723). National Bureau of Economic Research. Barberis, N. and Thaler, R. H. (2005). A survey of behavioral finance. In R. H. Thaler (Ed.), Advances in Behavioral Finance, Vol. II. Princeton: Princeton University Press.

November 4, 2020

262

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

Butler, D. J. and Cheung, S. L. (2019). Mind, body, bubble! Psychological and biophysical dimensions of behavior in experimental asset markets. In G. Fodter (Ed.), Biophysical Measurement in Experimental Social Science Research. Cambridge: Academic Press, pp. 167–196. Calvet, L. E. and Sodini, P. (2014). Twin picks: Disentangling the determinants of risk-taking in household portfolios. The Journal of Finance, 69(2), 867–906. Carlander, A., Peterson, D., Gamble, A., G¨ arling, T., Johansson, L. O., and Holmen, M. (2013). Choices of savings options related to investor-skill beliefs and trust in banks. Journal of Financial Services Marketing, 18(2), 121–136. Chambers, D., Dimson, E., and Ilmanen, A. (2012). The Norway model. The Journal of Portfolio Management, 38(2), 67–81. Chang, T. Y., Solomon, D. H., and Westerfield, M. M. (2016). Looking for someone to blame: Delegation, cognitive dissonance, and the disposition effect. The Journal of Finance, 71(1), 267–302. Clark, G. L. and Urwin, R. (2008). Best-practice pension fund governance. Journal of Asset Management, 9(1), 2–21. Coates, J. M. and Herbert, J. (2008). Endogenous steroids and financial risk taking on a London trading floor. Proceedings of the National Academy of Sciences, 105(16), 6167–6172. Cohn, A., Fehr, E., and Mar´echal, M. A. (2017). Do professional norms in the banking industry favor risk-taking? The Review of Financial Studies, 30(11), 3801–3823. Diacon, S. (2004). Investment risk perceptions: Do consumers and advisers agree? International Journal of Bank Marketing, 22(3), 180–199. Dohmen, T., Falk, A., Huffman, D., Sunde, U., Schupp, J., and Wagner, G. G. (2011). Individual risk attitudes: Measurement, determinants, and behavioral consequences. Journal of the European Economic Association, 9(3), 522–550. Doran, J. S., Peterson, D. R., and Wright, C. (2010). Confidence, opinions of market efficiency, and investment behavior of finance professors. Journal of Financial Markets, 13 (1), 174–195. Duxbury, D. and Summers, B. (2004). Financial risk perception: Are individuals variance averse or loss averse? Economics Letters, 84 (1), 21–28. Ericsson, K. A., Andersson, P., and Cokely, E. T. (2005). The enigma of financial expertise: Superior and reproducible investment performance in efficient markets. Florida State University. Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. Journal of Finance, 25 (2), 383–417. Fama, E. F. and French, K. R. (1992). The cross-section of expected stock returns. The Journal of Finance, 47 (2), 427–465.

page 262

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

b3964-ch10

page 263

263

Fama, E. F. and French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33 (1), 3–56. FCLTglobal; August 30, 2018. https://www.fcltglobal.org/news/blog/arti cle/2018/08/30. Fenton-O’Creevy, M., Lins, J. T., Vohra, S., Richards, D. W., Davies, G., and Schaaff, K. (2012). Emotion regulation and trader expertise: Heart rate variability on the trading floor. Journal of Neuroscience, Psychology, and Economics, 5 (4), 227. Fenton, O’Creevy, M., Soane, E., Nicholson, N., and Willman, P. (2011). Thinking, feeling and deciding: The influence of emotions on the decision making and performance of traders. Journal of Organizational Behavior, 32 (8), 1044–1061. Fraser, S. P. and Jennings, W. W. (2010). Examining the use of investment policy statements. The Journal of Wealth Management, 13 (2), 10–22. Ganzach, Y. (2000). Judging risk and return of financial assets. Organizational Behavior and Human Decision Processes, 83 (2), 353–370. Grable, J. E. (2000). Financial risk tolerance and additional factors that affect risk taking in everyday money matters. Journal of Business and Psychology, 14 (4), 625–630. Gray, J. (2009). Rethinking investment beliefs in a time of crisis: The calming hand of philosophy. Rotman International Journal of Pension Management, 2 (1), 6–11. Guyatt, D. (2005). A Summary of the Findings of a Survey Into: “Investment Beliefs Relating to Corporate Governance and Corporate Responsibility.” Bath: University of Bath. Hadar, L., Sood, S., and Fox, C. R. (2013). Subjective knowledge in consumer financial decisions. Journal of Marketing Research, 50 (3), 303–316. Holzmeister, F., Huber, J., Kirchler, M., Lindner, F., Weitzel, U., and Zeisberger, S. (2020). What drives risk perception? A global survey with financial professionals and lay people. Management Science, Forthcoming. Pensions and Investments; July 23, 2012. Asset managers honing focus on investment beliefs. Huber, J., Kirchler, M., and St¨ ockl, T. (2010). The hot hand belief and the gambler’s fallacy in investment decisions under risk. Theory and decision, 68 (4), 445–462. Jansson, M., Tr¨onnberg, C. C., and Hemlin S. (2018). The occurrence and importance of pension fund managers’ investment beliefs. A web survey and critical incident study. Journal of Finance and Investment Analysis, 7 (4), 1–17.

November 4, 2020

264

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

Johnson, K. L. and de Graaf, F. J. (2009). Modernizing pension fund legal standards for the twenty-first century. Rotman International Journal of Pension Management, 2 (1), 44–51. Kahneman, D. (2003). Maps of bounded rationality: Psychology for behavioral economics. American Economic Review, 93(5), 1449–1475. Koedijk, K. and Slager, A. (2009). Do institutional investors have sensible investment beliefs? Rotman International Journal of Pension Management, 2(1), 12–19. Koedijk, K. and Slager, A. (2011). Investment Beliefs — A Positive Approach to Institutional Investing. London: PalGrave Macmillan. Koedijk, K. C., Slager, A., and Bauer, R. (2010). Investment beliefs that matter: New insights into the value drivers of pension funds. Available at SSRN 1603262. Kuhnen, C. M. and Knutson, B. (2011). The influence of affect on beliefs, preferences, and financial decisions. Journal of Financial and Quantitative Analysis, 46 (3), 605–626. Lambert, J., Bessi`ere, V., and N’Goala, G. (2012). Does expertise influence the impact of overconfidence on judgment, valuation and investment decision? Journal of Economic Psychology, 33, 1115–1128. Lintner, J. (1965). The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets. The Review of Economics and Statistics, 47 (1), 13–37. Lo, A. W., Repin, D. V., and Steenbarger, B. N. (2005). Fear and greed in financial markets: A clinical study of day-traders. American Economic Review, 95 (2), 352–359. Locke, P. R. and Mann, S. C. (2009). Daily income target effects: Evidence from a large sample of professional commodities traders. Journal of Financial Markets, 12(4), 814–831. Loewenstein, G. F., Weber, E.U., Hsee, C. K., and Welch, N. (2001). Risk as feelings. Psychological Bulletin, 127 (2), 267–286. Lucey, B. M. and Dowling, M. (2005). The role of feelings in investor decision-making. Journal of Economic Surveys, 19 (2), 211–237. Lydenberg, S. (2011). Investment Belief Statements. Harvard: Harvard Initiative for Responsible Investment (IRI). Malkiel, B. G. (2003). The efficient market hypothesis and its critics. Journal of Economic Perspectives, 17 (1), 59–82. Malkiel, B. G. (2005). Reflections on the efficient market hypothesis: 30 years later. Financial Review, 40 (1), 1–9. Markowitz, H. M. (1952). Portfolio selection. The Journal of Finance, 7 (1), 77–91. McKenzie, C. R., Liersch, M. J., and Yaniv, I. (2008). Overconfidence in interval estimates: What does expertise buy you? Organizational Behavior and Human Decision Processes, 107 (2), 179–191.

page 264

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Investment Beliefs and Portfolio Risk-Taking

b3964-ch10

page 265

265

Morrin, M., Jacoby, J., Johar, G. V., He, X., Kuss, A., and Mazursky, D. (2002). Taking stock of stockbrokers: Exploring momentum versus contrarian investor strategies and profiles. Journal of Consumer Research, 29(2), 188–198. Nygren, T., Isen, A., Taylor, P., and Dulin, J. (1996). The influence of positive affect on the decision rule in risk situations: Focus on outcome rather than probability. Organizational Behavior and Human Decision Processes, 66 (1), 59–72. Olsen, R. A. (1997). Investment risk: The experts’ perspective. Financial Analysts Journal, 53 (2), 62–66. Peterson, D., Carlander, A., Gamble, A., G¨arling, T., and Holmen, M. (2015). Lay people beliefs in professional and na¨ıve stock investors’ proneness to judgmental biases. Journal of Behavioral and Experimental Finance, 5, 27–34. Raymond, D. M. (2008). Investment beliefs. In F. J. Faboozi (Ed.), Handbook of Finance, Vol. 1. New Jersey: Wiley. Ricciardi, V. (2008). The psychology of risk: The behavioral finance perspective. In F. J. Faboozi (Ed.), Handbook of Finance, Vol. 1. New Jersey: Wiley. Rook, D. P. (2012). How can we know if investors are coherently linking sustainability concepts? Journal of Sustainable Finance and Investment, 2(3–4), 198–221. Rozanov, A. (2015). Public pension fund management: Best practice and international experience. Asian Economic Policy Review, 10 (2), 275–295. Sachse, K., Jungermann, H., and Belting, J. M. (2012). Investment risk — The perspective of individual investors. Journal of Economic Psychology, 33 (3), 437–447. Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. The Journal of Finance, 19 (3), 425–442. Shefrin, H. (2001). Do investors expect higher returns from safer stocks than from riskier stocks? The Journal of Psychology and Financial Markets, 2 (4), 176–181. Siev¨ anen, R. (2012). Responsible investment by pension funds after the financial crisis. In B. Scholtens et al. (Eds.), Responsible Investment in Times of Turmoil. Dordrecht: Springer, pp. 93–112. Slager, A. and Koedijk, K. (2007). Investment beliefs: Every asset manager should have them. The Journal of Portfolio Management, 33 (3), 78–83. Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological Bulletin Journal, 119 , 3–22. Slovic, P. and Peters, E. (2006). Risk perception and affect. Current Directions in Psychological Science, 15 (6), 322–325.

November 4, 2020

266

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch10

Behavioral Finance: A Novel Approach

Slovic, P., Peters, E., Finucane, M. L., and MacGregor, D. G. (2005). Affect, risk, and decision making. Health Psychology, 24 (4S), S35. Stanovich, K. E. and West, R. F. (2000). Individual differences in reasoning: Implications for the rationality debate? Behavioral and Brain Sciences, 23 (5), 645–665. Tang, N. and Baker, A. (2016). Self-esteem, financial knowledge and financial behavior. Journal of Economic Psychology, 54, 164–176. The Investment Officer; August 8, 2019. https://www.ai-cio.com/though t-leadership/mer-investment-beliefs-governance-bedrock-investmentcommittees/. Thoma, V., White, E., Panigrahi, A., Strowger, V., and Anderson, I. (2015). Good thinking or gut feeling? Cognitive reflection and intuition in traders, bankers and financial non-experts. PloS One, 10(4), e0123202. Thompson, E. R. (2007). Development and validation of an internationally reliable short-form of the positive and negative affect schedule (PANAS). Journal of Cross-Cultural Psychology, 38 (2), 227–242. T¨orngren, G. and Montgomery, H. (2004). Worse than chance? Performance and confidence among professional and laypeople in the stock market. The Journal of Behavioral Finance, 5 (3), 148–153. Walden, M. L. (2015). Active versus passive investment management of state pension plans: Implications for personal finance. Journal of Financial Counseling and Planning, 26 (2), 160–171. Wang, A. (2009). Interplay of investors’ financial knowledge and risk taking. The Journal of Behavioral Finance, 10 (4), 204–213. Wang, M., Keller, C., and Siegrist, M. (2011). The less you know, the more you are afraid of: A survey on risk perception of investment products. Journal of Behavioral Finance, 12 (1), 9–19. Weber, E. U., Blais, A. R., and Betz, N. E. (2002). A domain-specific riskattitude scale: Measuring risk perceptions and risk behaviors. Journal of Behavioral Decision Making, 15 (4), 263–290. Weber, M., Weber, E. U., and Nosic, A. (2013). Who takes risks when and why: Determinants of changes in investor risk taking. Review of Finance, 17 (3), 847–883. Woods, C. and Urwin, R. (2010). Putting sustainable investing into practice: A governance framework for pension funds. Journal of Business Ethics, 92 (1), 1–19.

page 266

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0011

Chapter 11

Boys Don’t Cry? The Emotional Effects of Poor Financial Savings Decisions among Males and Females Erez Yaakobi∗ and Ido Kallir†

Abstract Previous research examined the causes for suboptimal financial decisions. However, scant research has been devoted to examine the psychological effects of suboptimal financial decisions. This chapter empirically examines the emotional outcomes of pension and provident suboptimal financial decisions. Study 1 reveals that the word pension elicits many negative associations. Study 2 reveals that suboptimal pension and provident decisions lead to greater accessibility of negative thoughts. Moreover, the greater accessibility of negative thoughts following suboptimal financial decisions was found to be more salient among male then female participants. In this chapter, we provide theoretical as well as practical applications of the current findings. We also suggest future research directions that could be conducted to better enable capture this important topic. Keywords: Pension, provident, suboptimal financial decisions, emotions

1.

Introduction

The importance of savings and having sufficient income for one’s retirement as well as for the shorter period is crucial. Moreover, most ∗ †

Ono Academic College, Kiryat Ono, Israel; [email protected]. Ono Academic College, Kiryat Ono, Israel; [email protected]. 267

page 267

November 4, 2020

268

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

households know very little about their pension plan or finance in general (Lusardi and Mitchell, 2014). However, the effects of these suboptimal savings decisions regarding both long term and more short term remains understudied. In this chapter, we empirically examined the psychological effects of suboptimal financial savings decisions. We first present the characteristics of pension and provident funds. Then we present findings regarding gender differences regarding financial decisions. After presenting the literature review, we present the two studies that we have conducted to reveal the emotional effects associated with savings and the emotional effects of suboptimal financial saving decisions. We conclude by discussing the theoretical as well as the practical contribution of our findings and suggest future research directions.

2.

Pension Funds and Provident Funds

In more and more economies, households are required to handle their own pension portfolio and make important and sometimes irreversible decisions. The pension system in Israel is a good example, since households are required to make substantial financial decisions regarding their pension plan when individuals are still at the start of their careers. These financial decisions impacts the wealth level of the households after retirement and their level of insurance coverage in case of premature death or incapacity during the employment period. In Israel, in addition to having to decide about their pension funds, many households are asked to choose an additional, short-term savings tool that for historical reasons also provides tax deductions: Provident funds. Provident funds are financial instruments that grant capital gains tax (CGT) and personal income tax (PIT) exemptions, but it is locked for withdraw only for six years. Some households see it as part of their long savings, but most households use it for midlife financial projects such as family vacations, buying a new car or paying part of the mortgage. With pensions and social security, expected income accounting for half of the wealth at retirement and evidence that individuals with pensions save more in other forms, one would expect to find that knowledge of pensions and social security would influence

page 268

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

The Emotional Effects of Poor Financial Savings Decisions

page 269

269

retirement saving. Gustman et al. (2012), however, found no evidence that knowledge of pensions and social security were related to non-pension, non-social security wealth, numeracy or that it plays an intermediate role in the numeracy–wealth relationship. Chan and Stevens (2008) found that well-informed individuals are five times more responsive to pension incentives than the average individual, when level of knowledge is disregarded. Gerardi et al. (2013) measured financial literacy and cognitive ability in a survey of subprime mortgage borrowers and matched them to objective, detailed administrative data on mortgage characteristics and payment histories. They found that the relationship between numerical ability and mortgage default was robust to controlling for a broad set of sociodemographic variables, and was not driven by other aspects of cognitive ability. Gerardi et al. (2013) also found no support for the hypothesis that numerical ability impacts mortgage outcomes through the choice of the mortgage contract. Rather, the results suggest that individuals with limited numerical ability default on their mortgage due to behavior unrelated to the initial choice of their mortgage. In a study that examined whether cognitive abilities are related to differences in trajectories for key economic outcomes as individuals move towards and through their retirement, Banks et al. (2010) found that individuals with lower numeracy had significantly lower wealth trajectories both pre- and post-retirement than their more numerate counterparts. The pension plan in Israel differs in some aspects from both the European and the American pension plans that will be described here. 2.1.

The Israeli Pension Plan

Although there was no mandatory pension plan in Israel until 2008, the entire public sector and large sections of the private sector provided pension plans to their employees. Before 1995 the typical pension plan was either budgeted by the government or was a Defined Benefit plan, mostly subsidized by the employer. These Defined Benefit plans guarantee an annuity for an employee upon the employee’s retirement. The payment in a Defined Benefit pension plan is determined by a formula that can include the employee’s income, years of employment, age at retirement and other factors. Since 1995, the pension system has undergone a series of substantial

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

270

changes. It became an entirely Defined Contribution plan, was completely privatized. As of 2008 it is mandatory for all employees up to the average salary. The current pension scheme is a semi-mandatory, tax exempt DC fund. The pension scheme allowance is two-fold: up to the average salary ∼ (US$36,000) the scheme is mandatory. Above US$36,000–90,000, it is PIT free, thus providing a tax-free annuity based on the individual’s tax bracket. At the minimum age of 60 or after retirement (most employees, men and women retire at 67) this annuity can be withdrawn, tax free. Though the scheme is mandatory, the funds are private and compete against each other. There are nine such funds. A typical scheme is as follows: (a) (b) (c) (d)

The The The The

employee contributes 6% of her gross salary. employer endows an additional 14.83%. endowment is PIT free. endowment is locked until retirement.

The pension annuity is the second pillar of the old-age scheme. The first pillar is the social security flat annuity. The social security annuity is considered among the lowest in the Organisation for Economic Co-operation and Development (OECD) countries. For an average household, it is expected to provide less than 30% of its postretirement income. The rest is expected to be covered by the pension scheme. Pension funds are therefore treated seriously. The short term, six years provident funds are as said, unique Israeli systems. Financially, a provident fund functions just like a mutual fund, with tax exemptions. The plan is not mandatory but rather a “beneficiary arrangement” between the employer and employee. The employee contributes 2.5%, the employer adds 7.5% of the gross salary, up to US$ 55,000. The endowment is exempt from PIT and the entire sum is exempt from the CGT. In 2017, 42%1 of employees had a provident fund. Though the entire endowment can be saved until retirement and beyond, exempt from CGT, 77% of all households withdraw it after exactly six years and only less than 9% save it for more than 10 years2 Households see it as a short to medium term savings tool that is not 1 2

Israeli treasury data. According to the funds’ own data.

page 270

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

The Emotional Effects of Poor Financial Savings Decisions

page 271

271

part of the pension plan. Employees use it, for example,3 to pay for (part of) their mortgage (27%), down payments on apartments (45%), buy a new car (25%), expensive vacations (19%), family celebrations such as weddings (13%) or to cover loans (13%). All in all, provident funds are used to improve the standard of living. As people around the world make poor pension decisions or at least do not plan their pensions with full attention, despite its immense effects on both themselves as well as their countries, it is important to examine the psychological effects of these inaccurate decisions. We posit that in addition to the financial effect of poor financial decisions, people who make suboptimal financial decisions lead to augment peoples’ negative emotions following their bad decisions. In this chapter, we claim that suboptimal financial decisions and specifically, suboptimal pension and provident financial decisions, evoke negative thoughts. Thus, we claim that these suboptimal financial decisions will be relevant for both long-term funds that should affect one’s older ages (i.e., pension) as well as short-term savings (i.e., provident funds). There are many reasons that can serve as a rationale for our claim. First, pension funds are aimed to provide financial support to people of older ages when financial burdens are more salient. Making suboptimal decisions leading to lower amount of money when growing older should lead people to more negative feelings than when making more qualified pension decisions. Moreover, pension is mostly associated with negative events such as a premature death or incapacity during the employment period. These negative events appear both on the web as well as among agents who sell pension funds. One of the motivators they often use is to provide people with a notion of fear regarding old age as a burden on their children, health difficulties and other negative effects of one’s retirement. Thus, it is reasonable to argue, that when people feel that they made suboptimal pension financial decisions it would lead them to have more negative emotions than when making better financial decisions with regard to pension. However, most of these reasons are also associated with

3

Taken from the authors’ 2014 survey of 250 respondents. Percentages total more than 100% since more than one answer could be marked.

November 4, 2020

272

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

shorter savings, such as for a period of 6 years as in the provident funds. This is possibly because one has made a suboptimal decision that will possibly lead to having less money when withdrawing the money, which may as well lead to negative emotions. 3.

Gender Differences in Financial Decisions

There is a large body of literature presenting the gender differences in financial decision-making. Women make less risky financial decisions (Croson and Gneezy, 2009); tend to make moderate investments (Willows and West, 2015) and if they hold asset portfolios, those tend to more conservative (Barber and Odean, 2001). Powell and Ansic (1997) show that females are more risk averse than males regardless of familiarity framing, costs of investing or level of ambiguity. They also specify that males and females adopt different strategies in financial decision environments but these strategies have no significant impact on their actual results. The literature suggests the following three possible explanations for the gender gap: (a) Women lack proper financial literacy (Goldsmith and Goldsmith, 1997, 2006; Powell and Ansic, 1997). (b) Women are by nature more risk averse (Byrnes et al., 1999; Barber and Odean 2001; Charness and Gneezy, 2012). This can be particularly true for long-term decisions and when risk taking might be actually recommended (Huang and Kisgen, 2013). Watson and McNaughton (2007) show that women chose financial retirement portfolios with lower percentage of stocks compared to men. (c) Men are more confident, therefore, they more easily make their financial decision (Barber and Odean, 2001; Odean, 1999). Overconfidence differences were more evident among singles (Odean, 1999). 3.1.

Overview of the Current Studies

For better capture of these hypotheses we conducted two studies. In the first study, students in a financial MBA course were asked to mention their associations regarding the notion of pension. In the second experiment, participants were randomly divided into two

page 272

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

The Emotional Effects of Poor Financial Savings Decisions

page 273

273

groups: one making a financial decision by choosing one alternative out of four whereas only one was a value-maximizing choice. The other group completed that same task with the same identical financial information but the fund was presented as a pension fund. We hypothesized that those participants who chose a suboptimal fund were associated with greater accessibility of negative thoughts than participants who made value-maximizing financial choices. We also examined whether significant differences would be found when making suboptimal pension in comparison to suboptimal provident decisions. This comparison was more an exploratory analyses of possible pension vs. provident suboptimal financial decisions on emotions. 3.2.

Study 1

In this study, we examined what are the free associations that comes into ones’ mind when thinking about pension (provident fund). We asked students that took the “Pension and Long-term Finance” course to answer the following questions. These questions were asked in three classes, with a total of 112 respondents. All respondents were MBA “Pension and Long-term Finance” course students who were asked individually to respond to two questions: (1) What is the first association that comes to your mind for the term “pension fund?” (2) What is the first association that comes to your mind for the term “provident fund?” The questions were asked one by one with a 20-second pause to answer the first question. We found that 76 (68%) respondents attributed associative negative value to the word “pension”. Of those, 23 referred to old age, 31 to fear of deprivation, 12 to illness and 10 directly to death. Of the 36 who did not view pension as negative, 19 referred to the opportunity not to work, 8 to freedom and 9 used neutral words. In contrast, 65 respondents (58%) used positive words about savings in a provident fund. The words “vacation” “new car” and “shopping” led the list. About 40 (35%) simply wrote “I do not have” in this context. About 12 (10%) respondents used financial jargon such as “yield” or “management fees” compared to only 2 that did so in response to the word “pension”. These findings supported our claim that the notion of pension by itself is associated with negative thoughts. To further empirically examine our main hypothesis that suboptimal financial decisions lead to greater accessibility of negative-related words we conducted our

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

274

second experiment. As noted we also examined the possible moderating role of types of financial decision (pension vs. provident) on the effects of suboptimal financial decisions on emotions. This was done in a more exploratory manner. 3.3.

Study 2

Study 2 was conducted to examine whether the accessibility of negative-related thoughts will be higher in participants who made suboptimal financial decision than participants who made valuemaximizing financial decisions. We also examined the possible moderating roles of types of fund (pension vs. provident funds) on the effects of making suboptimal financial decisions on the accessibility of negative emotions. 3.3.1.

Method

Participants: One-hundred and fourteen Israeli undergraduates (58% males and 42% females, ranging in age from 19 to 53 with a mean of 32.05; 50% of them were single, 44% married, 5% divorced and 1% widow; 91% of them were working at the time of the experiment and 47% of them related to their work as their intended career; their number of children ranged from 0 to 4) with a median of 2. The participants participated in the study without a monetary reward. Materials and procedure: Participants were invited to participate in a study of personality and social psychology in groups of 4–20 participants. They were randomly divided into two experimental conditions (pension or provident). The subjects received a set of questionnaires. Subjects were asked to answer the questions in the order in which they appear. One group received a questionnaire on pension fund and the second group received a provident fund questionnaire. The subjects were given a set of financial and other information that would assist them in making a certain economic decision. The decision related to the choice of a pension fund or the selection of a provident fund out of four alternatives. The subjects received a standard explanation taken from the economic media briefly explaining what a pension fund is or what a provident fund is, including the standard insurance components and the possible uses of the fund. Thereafter, detailed information appeared, and afterwards they were

page 274

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

The Emotional Effects of Poor Financial Savings Decisions

page 275

275

asked to choose one of the four alternatives provided. The information includes the management fees collected from the periodic deposits and the accrual, the historical yield and the fund’s Sharpe ratio. In addition, we provided “white noise” information, about the size of the funds, for example, when they were all of the same size and the number of people in the workplace who chose them was also of the same size that were fully balanced as well between participants to overcome alternative explanations for possible effects for this “white noise” information. The information provided in each scenario was built in accordance with the level of informativity that can be obtained from the information systems operated by the Ministry of Finance. In these questionnaires, the respondents were asked to choose one of the four options. The information about the various funds were carefully arranged so that each questionnaire and each version always had one fund that is strictly dominant over the other funds. The only difference between the information provided to the pension and provident funds was the label and short similar information. This was done to enable deducing the conclusion that the label of pension itself led to the results. Following one of these presented scenarios of financial decision (pension or provident), all participants had to decide which of the four alternatives they choose. In the research questionnaire we provided the participants only with parameters that they could have received if they were required to make a real decision and would have faced financial information from the government decision support systems. 3.3.2.

Example for purchasing a pension fund scenario

You are asked to choose a pension fund out of four possible pension funds. The purpose of a pension fund is to provide a living allowance in old age. In addition, the pension fund provides insurance for disability that causes work disability, as well as an insurance for the family in case of early death. It is clear that an optimal savings for pension is critical in order to improve the allowance in old age and to maintain insurance coverage for you and your family against work disability and early death risks. For the sake of convenience, we will

November 4, 2020

276

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

mark the funds as A, B, C and D. Please take a few minutes to think, and then answer the questionnaire attached in the end and tell us which of the funds you chose. 3.3.3.

Example for a pension fund alternative

Pension fund A has managed assets volume of 9.3 billion NIS. Around 30% of the fund’s assets are invested in designated bonds, and between 26% and 29% of the assets are invested in shares. About 16% of the employees chose this fund. The management fees from the accrual are 0.25%, and the management fees from deposits are 2.4%, which constitute a discount of about 55%. The financial fiveyears’ yield is 4.36%. The value of the Sharp index, which measures the efficiency of the fund is 0.84. 3.3.4.

Example for purchasing a provident fund scenario

You work for a large company that offers its employees a provident fund as part of the employment benefits. You have a choice between four provident funds, which were examined by a team of experts on behalf of the management of the company’s employees’ committee. We would like to remind you that as of six years from its commencement date, you can use the provident fund for any purpose: a new car, a vacation abroad, and you can even repay a part of your mortgage with it. Your provident fund is expected to accumulate an amount of 100,000–120,000 NIS in a period of six years, which means that this is a significant economic decision. Following are various data about the four funds. For the sake of convenience, we will mark them as A, B C and D. Please take a few minutes to think, and then answer the questionnaire attached in the end and tell us which of the funds you chose, and why. 3.3.5.

Example for a provident fund alternative

Approximately 21% of the company’s employees chose fund A. Fund A has managed assets volume of 2.5 billion NIS. The risk level of fund A is defined as medium. About 24% of the fund’s assets are invested in shares. The value of the Sharpe index, which is the index for quality of financial performance and weighs the yield in relation to the financial risk, is 1.17. The management fees from the

page 276

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

The Emotional Effects of Poor Financial Savings Decisions

page 277

277

deposits are zero, and the management fees from the accrual are 0.44% of the accrual and constitute a discount of 78% of the management fee ceiling. The average annual yield over the past five years is 5.07%. After making a decision, participants’ accessibility of negativerelated thoughts was assessed by using the Hebrew-revised version of the word completion task, originally devised in English by Greenberg et al. (1994) and successfully used in Hebrew by Mikulincer and Florian (2000) on an Israeli sample. The original task was built to assess the participants’ accessibility of death-related thought. Here we revised the target words as will be presented later to words that could be either neutral of negative (instead of death) related words. This is a common technique to examine death anxiety in the Terror Management literature. In this study, the task consisted of 20 Hebrew word fragments that participants were asked to complete with the first word that came to mind by filling in one missing letter. A total of 8 of the 20 Hebrew fragments could be completed to form either neutral or negative-related Hebrew words. For example, participants saw the Hebrew fragment TACH and could complete it with the Hebrew word PTACH (open) or with the negative emotional word METACH (stress) (in Hebrew, mentioning the word M is enough for completing the word stress). The possible negative Hebrew words were: stress, sad, shame, sorrow, pain, failure, jealous and anxious. The dependent measures were the number of negative-related Hebrew words (0–8) completed by each participant.

4.

Results and Discussion

To examine whether gender moderates the effects of the quality of the financial decision on the accessibility of negative-related thoughts, we implemented Hayes’ (2013) PROCESS macro (Models 1). We also conducted a preliminary analyses including types of funds (pension vs. provident) as an additional variable to examine the exploratory hypothesis of a possible moderating role of types of funds on suboptimal financial decisions on emotions by implementing Hayes’ (2013) PROCESS macro (Models 3) for examining a three-way interaction of financial decisions’ quality (suboptimal vs. value-maximizing) × type of fund (pension vs. provident) × gender (male vs. female). As

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

Number of negative words (Z score)

278 1.5 1 0.5 0 –0.5

Male Female

–1 –1.5 –2

Figure 1:

Low Quality

High Quality

Quality of financial decision vs. gender.

no significant results were found in these analyses we only report the results of the simple and two-way interactions of quality of financial decision × gender on accessibility of negative emotions. The analysis revealed significant results for the quality of decision (β = −0.75, t = −2.45, p = 0.016) but not for gender (β = 0.15, t = 0.72, p = 0.471). Moreover, a significant two-way interaction between the quality of the financial decision and gender is (β = 0.50, t = −2.39, p = 0.018). Specifically, participants who made valuemaximizing financial decision completed significantly less words as negative than participants who made suboptimal financial decisions. No differences were found between males and females in the number of negative words completed. To probe the essence of the two-way interaction, the financial quality decision (value-maximizing vs. suboptimal) × gender was analyzed. The results revealed that whereas male participants who made suboptimal decisions showed greater accessibility of negativerelated thoughts, than when making value-maximizing decisions, β = −0.26 p < 0.05, female participants did not show differences in their accessibility of negative-related thoughts in the two qualities of financial decisions, β = 0.24 p = 0.14 (Figure 1). 5.

General Discussion

Taken together, in this chapter we discuss and provide empirical evidence that the pension and provident suboptimal decisions are

page 278

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

The Emotional Effects of Poor Financial Savings Decisions

page 279

279

associated with greater negative thoughts. Moreover, we show that this is more evident among males than females. Thus, in addition to Lusardi and Mitchell’s (2014) claim that most households know very little about their pension plan or finance in general, it could be that this avoidance may also follow people’s suboptimal they may have made in their past financial decisions. This can hint on a possible mechanism that should be empirically examined in future research one of the possible reasons for not paying enough cognitive thoughts on savings. Our findings that males were more affected by making suboptimal financial decisions than females are consistent with previous findings showing gender differences in financial decision-making. As suggested, these can be due to women’s greater tendency of being risk aversive than males (Powell and Ansic, 1997), women’s tendencies for less risky financial decisions (Croson and Gneezy, 2009), their tendency to make moderate investments (Willows and West, 2015) or more conservativeness (Barber and Odean, 2001). Our findings could be also explained by previous different strategies males and females use, such as men are claimed to be more confident, therefore, they more easily make their financial decisions (Barber and Odean, 2001; Odean, 1999). All these possible explanations should be empirically examined in future research. Our empirical findings make theoretical as well as practical contributions. These findings contribute to the economic field of investment decision-making. It also contributes to the household finance decision literature and to the psychology literature and specifically to the emotional and behavioral finance by empirically examining the effects of poor financial savings decisions. Our findings have key practical contribution as well. One of the key implications may be that, as this research found that suboptimal financial savings decisions augment one’s accessibility of negative thoughts, it leads to the importance of implementing mechanisms that buffer these negative emotions on the one hand, and on the other hand to the importance of applying mechanisms leading to behavioral changes that give a solution to the source of these negative emotions, namely leading people to change their funding choices they made in the past. As in Israel, the ability to move from one fund to another has no fee, cognitive, emotional and behavioral mechanisms aimed to lead people to act and change their suboptimal fund choices, it may lead to more positive emotions and well-being.

November 4, 2020

280

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

The psychology literature has developed various mechanisms that enable overcoming the resistance to change which will be of great relevance for the current purposes as well. In a nutshell, it is important to identify the reasons underling the resistance such as sank cost, technological difficulties or other barriers and provide solutions to overcome these barriers. Our findings also highlights the importance of providing better education to people or consulting people for the information needed for make better financial savings decisions. In addition, it could be that framing savings financial decisions in a more positive way (e.g., time for leisure activities, more freedom to actualize one’s dreams etc.) in contrast to associating them with more negative events (e.g., financial burdens) may lead to greater wellbeing and eliminate or at least lower negative emotional outcomes of poor financial savings decisions. Future research should examine this hypothesis. The current findings could be implemented by insurance companies that can borrow from the current results and reframe the salespersons’ guidelines in a way that put less emphasis on the association of savings with burdens but as a way to fulfill one’s desires. Thus, it should be also efficient to frame financial savings decisions as a way of saving money for achieving a better life in the future that can lead to better financial resources for more qualitative life when getting old (pension) and even for achieving more wealthy life (pension and provident). In the same way framing pension as savings for a better future life should also be a more efficient presentation of pension investments. Future research should examine these hypotheses as well. The current research could also hint on more fruitful actions that can be taken by nations for facilitating better policy decisions that should contribute to effectively manage the limited resources that are so needed nowadays. In addition, it can also contribute at the micro level by enabling identify actions that can be implemented by consultants that can buffer negative emotions from suboptimal financial decisions and constitute a more effective buffering mechanisms. In addition, it is interesting to examine whether better financial knowledge should moderate the current research results and facilitate better accurate pension financial decisions.

page 280

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

The Emotional Effects of Poor Financial Savings Decisions

6.

page 281

281

Conclusion

In this chapter, we present a novel direction for the psychological effects of poor financial savings decisions. Specifically, we reveal that suboptimal financial savings decisions lead to negative emotions even in the unconscious level which are more salient among males than females. We also provide some initial directions that should be empirically examined in future research for achieving a better understanding of the mechanisms that are supposed to encourage people to save their money that can also lead to better financial decisions and wellbeing. Our results provide evidence for the psychological effects of suboptimal decisions that may also be responsible for the deficits in these financial decisions. Finally, we have tried to provide further guidelines to facilitate better decisions. We believe that this first research step can lead to a fruitful empirical examination of the suggested novel direction that integrates the interdisciplinary fields of psychology and economics in the future. References Banks, J., O’Dea C., and Oldfield, Z. (2010). Cognitive function, numeracy and retirement saving trajectories. The Economic Journal, 120 (548), 381–410. Barber, B. M. and Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common stock investment. The Quarterly Journal of Economics, 116 (1), 261–292. Byrnes, J. P., Miller, D. C., and Schafer, W. D. (1999). Gender differences in risk taking: A meta-analysis. Psychological Bulletin, 125 (3), 367. Chan, S. and Stevens, A. H. (2008). What you don’t know can’t help you: Pension knowledge and retirement decision-making. The Review of Economics and Statistics, 90 (2), 253–266. Charness, G., and Gneezy, U. (2012). Strong evidence for gender differences in risk taking. Journal of Economic Behavior & Organization, 83 (1), 50–58. Croson, R. and Gneezy, U. (2009). Gender differences in preferences. Journal of Economic Literature, 47 (2), 448–474. Gerardi, K., Goette, L. and Meier, S. (2013). Numerical ability predicts mortgage default. Proceedings of the National Academy of Sciences, 110, 11267–11271.

November 4, 2020

282

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch11

Behavioral Finance: A Novel Approach

Goldsmith, E. and Goldsmith, R. E. (1997). Gender differences in perceived and real knowledge of financial investments. Psychological Reports, 80 (1), 236–238. Goldsmith, R. E. and Goldsmith, E. B. (2006). The effects of investment education on gender differences in financial knowledge. Journal of Personal Finance, 5 (2), 55–69. Greenberg, J., Pyszczynski, T., Solomon, S., Simon, L., and Breus, M. (1994). Role of consciousness and the specificity to death of mortality salience effects. Journal of Personality and Social Psychology, 67, 627–637. Gustman, A. L., Steinmeier, T. L. and Tabatabai, N. (2012). The growth in social security benefits among the retirement-age population from increases in the cap on covered earnings. Social Security Bulletin, 72(2), 49–61. Huang, J. and Kisgen, D. J. (2013). Gender and corporate finance: Are male executives overconfident relative to female executives? Journal of Financial Economics, 108 (3), 822–839. Lusardi, A. and Mitchell, O. S. (2014). The economic importance of financial literacy: Theory and evidence. Journal of Economic Literature, 52 (1), 5–44. Mikulincer, M. and Florian, V. (2000). Exploring individual differences in reactions to mortality salience: Does attachment style regulate terror management mechanisms? Journal of Personality and Social Psychology, 79, 260–273. Odean, T. (1999). Do investors trade too much? American Economic review, 89 (5), 1279–1298. Powell, M. and Ansic, D. (1997). Gender differences in risk behaviour in financial decision-making: An experimental analysis. Journal of Economic Psychology, 18 (6), 605–628. Watson, J. and McNaughton, M. (2007). Gender differences in risk aversion and expected retirement benefits. Financial Analysts Journal, 63 (4), 52–62. Willows, G. and West, D. (2015). Differential investment performance in South Africa based on gender. International Business & Economics Research Journal, 14 (1), 221–236.

page 282

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0012

Chapter 12

Separating Accuracy from Forecast Certainty: A Modified Miscalibration Measure Doron Sonsino∗ , Yaron Lahav† and Amir Levkowitz‡

Abstract Interval forecasting tasks are commonly used to test for forecastoverconfidence. Pointing at deficiencies of the methodology, we advance a modified assignment, where subjects provide point predictions and assess the likelihood of return falling within small intervals around their estimates. The difference between the subjective likelihood assessments and the realized hit rates is advanced as an improved forecast-overprecision measure. Over three incentivized studies, 163 of 195 participants overestimate their hit rates, and a closer look at the data illustrates that inaccuracy and excessive certainty act as distinct sources of overprecision. Applications where the adapted task may prove more powerful than standard interval forecasting are discussed. Keywords: Interval forecasting, forecast-accuracy, trading

∗

miscalibration,

overconfidence,

Corresponding author. Doron Sonsino is at the College of Law and Business, Ramat Gan, Israel and adjunct at the Economics Department of Ben-Gurion University; [email protected]. † Yaron Lahav is from the Guilford Glazer Faculty of Business and Management (GGFBM) at Ben-Gurion University; [email protected]. ‡ Amir Levkowitz is a doctoral student at GGFBM; [email protected]. 283

page 283

November 4, 2020

284

1.

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

Introduction

The false belief in the accuracy of subjective assessments is considered an especially persistent form of judgmental overconfidence (Alpert and Raiffa, 1982; Klayman et al., 1999; Moore et al., 2015). Moore and Healy (2008) term the excessive certainty in beliefs overprecision, separating it from other facets of overconfidence such as overestimation of individual abilities or overplacement relatively to others. Diverse studies propose that overprecision contaminates the decision of traders (Daniel and Hirshleifer, 2015), executives (Malmendier and Taylor, 2015) and entrepreneurs (Astebro et al., 2014). Daniel et al. (1998), for a specific example, develop a model where overprecision explains the persistence of non-profitable trading, generating predictability in stock returns. Finance experiments and surveys commonly use confidence interval tasks to test for forecast-overprecision (for diverse examples see Glaser and Weber, 2007; Oberlechner and Osler, 2012; BenDavid et al., 2013; Sonsino and Regev, 2013; Fellner-R¨ ohling and Kr¨ ugel, 2014; Broihanne et al., 2014; Merkle, 2017; Grosshans and Zeisberger, 2018). The confidence level is exogenously provided and the forecaster submits lower and upper bounds for the target return. The participants are faced with several such intervalproduction assignments and the collection of intervals is utilized to derive the forecast-overconfidence measures.1 While the results of these studies uniformly support the overconfidence hypothesis, the overprecision metrics derived from forecast intervals show inconclusive statistically-weak results in empirical tests of the hypothesized overprecision-trading links (Glaser and Weber, 2007; Fellner-R¨ ohling and Kr¨ ugel, 2014; Broihanne et al., 2014; Merkle, 2017). This chapter points at deficiencies that arise with using interval forecasting tasks to derive overprecision measures, advancing a more direct method that avoids the obstacles. Essentially, the alternative approach diverts from interval-production to interval-evaluation (Winman et al., 2004). Subjects are asked to provide a point estimate (F ) for the target return first, and then assess the likelihood of

1

The terms forecast-overconfidence, forecast-overprecision miscalibration are used interchangeably henceforth.

and

forecast-

page 284

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Separating Accuracy from Forecast Certainty

page 285

285

return falling within a fixed length interval (F ±δ) around their point forecast. Essentially, the subjects assess the likelihood of exhibiting smaller than δ errors when forecasting the returns on designated stocks. Overprecision is supported if subjects overestimate the hit rates of the small-error intervals around their point predictions. We report the results of three incentivized experiments, applying the altered approach on N = 195 competent subjects. Utilizing the modified task to illustrate that inaccuracy and excessive-certainty complementarily contribute to forecast-overconfidence, we argue that the modified task may prove more powerful than standard interval forecasting in testing the hypothesized overprecision-trading links. 2.

Motivating Discussion

Figure 1 illustrates the SIFT as employed in diverse overconfidence studies (similar formats are used, for instance, in Glaser and Weber, 2007; Oberlechner and Osler, 2012; Ben-David et al., 2013). The forecaster provides a point prediction and 95% confidence limits for some future return, so that the interval between the limits represents a 90% confidence interval for the target return. In typical surveys and experiments, the forecaster is faced with a collection of such forecasting assignments. Calibration is tested when the returns R are realized, and the term HIT is used for cases where r falls within the submitted interval. The difference between the exogenous confidence level and the actual hit rate, addressed as the miscalibration score, is utilized as an applicable forecast-overprecision metric. If the hit — Submit a median prediction for the return on StockName in April–June, 2018 (the second quarter of 2018) ______

— With probability of 95%, I believe that the return on StockName in the second quarter of 2018 will be smaller than _______

— With probability 95%, I believe that the return on StockName in the second quarter of 2018 will be larger than ________

Figure 1:

The Standard Interval Forecasting Task (SIFT)

November 4, 2020

286

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

rate of 90% confidence intervals, for instance, is only 50% then the forecaster appears to overestimate the precision of the submitted interval forecasts by 40%, in line with forecast-overprecision. The hit rates of forecast intervals, however, jointly depend on the accuracy of the intervals and their length. High miscalibration scores may emerge when the forecasts are inaccurate or when the intervals are too narrow (see Klayman et al., 1999; Juslin et al., 2007 for related judgmental psychology discussions).2 If the target return, for a formal example, is normally distributed around μt with standard deviation (volatility) σt , then underestimation of the volatility by 25% would decrease the hit rate to 78%, while 50% discount of σt would cut the hit rate to 59%. Hit rates of 78% (59%), however, alternatively emerge when the point estimate is 0.84 (1.4) standard deviations from the true mean. Indeed, in McKenzie et al. (2008), IT experts provide more accurate, but shorter, confidence intervals for quantities related to their expertise. The contradicting effects cancel out and the experts’ miscalibration scores are similar to those of non-expert students. Forecasting studies more generally show that domain knowledge and expertise affect the accuracy of forecasts and the hit rates of prediction intervals (Lawrence et al., 2006; Hyndman and Athanasopoulos, 2018). The confounding knowledge and accuracy effects cast doubt on using miscalibration scores as meaningful measures of excessive forecast-certainty. As an alternative approach, recent studies use the confidence intervals to estimate the perceived volatility of the forecasted return. The perceived volatility estimates are contrasted with empirical benchmarks and the difference (empirical volatility minus perceived volatility) is adopted as a direct measure of overprecision. This alternative method, however, suffers from the weakness of resting on explicit assumptions regarding the stochastic process of stock returns. Indeed, the papers employing this methodology use a wide range of models to approximate the perceived and empirical volatilities. Graham and Harvey (2001), Glaser et al. (2013) and Merkle 2

We separate between financial forecasting studies where the prediction targets are future returns or prices and judgmental psychology studies where subjects estimate hidden quantities such as the population of given cities. Interval-evaluation tasks were tested in few judgmental psychology studies (see Section 3), but we are not aware of financial forecasting studies employing the methodology.

page 286

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Separating Accuracy from Forecast Certainty

page 287

287

(2017) use the Keefer and Bodily (1983) approximations to derive the standard deviation of the forecasted return. Oberlechner and Osler (2012) assume GARCH processes and use option implied volatilities to derive the empirical benchmarks. Sonsino and Regev (2013) compare the length of the forecast intervals with the realized spreads in recent histories, illustrating that the extent of overprecision may strongly vary with the length of history window selected for the comparison. The sensitivity to statistical assumptions again raises the concern that the proposed overprecision measures cannot be used to effectively rank investors in terms of relative forecast-certainty. Based on the lengths of the intervals, an investor that submits a [10%, 30%] interval, would be classified as less confident than one that submits a [0%, 10%] prediction, but the ranking may overturn if the heteroskedasticity of return processes is taken into account. On top of these problems, few preceding overconfidence studies point at nonsensical results, questioning the internal validity of the standard confidence interval assignment. Teigen and Jørgensen (2005; experiment 3) find that 90% confidence intervals are only marginally longer, exhibiting similar hit rates as 50% intervals. In Langnickel and Zeisberger (2016) the lengths and hit rates of 90%, 60% and 30% confidence intervals are almost identical. Added to the accuracy-length confound and the statistical problems in deriving volatility estimates, these paradoxical results motivate a search for alternative improved measures of forecast-overprecision.

3.

The Forecast Accuracy Assessment Task (FAAT)

The modified task, illustrated in Figure 2, consists of three steps: Step 1: The subject submits a median forecast F for the target return. The instructions explain that the median is a point estimate, positive or negative, such that the provider assigns 50% likelihoods to larger or smaller returns. Step 2: The instructions guide the subject to construct a fixed length interval around F , adding and subtracting a given margin δ from the median. Step 3: The subject provides a likelihood assessment CONF of the interval [F − δ, F + δ], estimating the likelihood of return falling

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

288

— Submit a median prediction for the return on StockName in April–June, 2018 (the second quarter of 2018) _______

— What, in your opinion, is the probability that the return on StockName in the second quarter of 2018 would fall in a range of plus or minus 5% from the median?

The next diagram may help you develop your estimate:

[lower bound__________

median forecast _______________upper bound]

Add 5% to the median prediction and fill in the “upper bound” box. Subtract 5% from the median prediction and fill in the "lower bound" box.

Submit the probability you assign to quarterly April–June 2018 StockName return within the interval between the lower and upper bounds (i.e., in a range of plus or minus 5% from your median forecast): _____ (between 0% and 100%)

Figure 2:

The Forecast Accuracy Assessment Task (FAAT)

within the 2δ interval centered at F . If, for example, the median forecast is 7% and the margin δ is 5%, then the subject estimates the likelihood of the [2%, 12%] interval. Likelihood assessments can take any value between 0% and 100%. In the studies described next, the subjects are faced with a sequence of such FAATs. Again, the term HIT is used for cases where r falls within the given interval, and the difference between the average likelihood that the subject assigns to calibration (CONF ) and the average hit rate (HIT ) represents our modified overconfidence measure: OC = CONF − HIT . If the CONFs, for example, are 70%, 50%, 90%, 70% while the realized return falls within the interval in only 1 of the 4 cases, OC is 45%. The subject overestimates the hit rate by 45%, in line with the overprecision hypothesis.

page 288

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Separating Accuracy from Forecast Certainty

b3964-ch12

page 289

289

While the overconfidence scores derived from the FAATs may appear essentially similar to the miscalibration scores derived from SIFTs, the competing measures are fundamentally different. In FAAT, a hit is recorded when the realized return (r) falls within the fixed length prediction interval; i.e., when F − δ ≤ r ≤ F + δ or r − δ ≤ F ≤ r + δ. The hit rate therefore only depends on the accuracy of the median forecasts. As the overconfidence score is derived by subtracting the realized hit rate from the average likelihood that subjects assign to the respective intervals, FAAT fixes the accuracylength concern raised for the standard miscalibration measure. The impact of accuracy and confidence on miscalibration can be separately assessed. Subjects may classify as overconfident for showing low accuracy, exaggerated forecast certainty, or both (see Sections 4 and 5 for examples). Such clean separation is impossible when miscalibration is measured using standard intervals. In addition, FAAT improves on the standard interval forecasting in few methodological/technical aspects: — In FAAT, forecast-confidence is elicited in a familiar 0–100 percentile scale. Intuitively, probabilistic assessments are more natural and easier than quantile assessments. Abbas et al. (2008), for instance, compare two methods for approximating the beta distributions of random variables. The method that is based on likelihood assessments shows higher consistency rates than a parallel method based on quantile assessments. It is also ranked as easier and more popular among the subjects. Similar results emerge in Wallsten et al. (2016).3 — Assuming the experiment is incentivized by randomly picking one of the assignments as a payment task (Cubitt et al., 1998; Hey and Lee, 2005), the likelihood assessment of the FAAT is easier to incentivize compared to the quantile assessments of the standard task. A loss function for the α quantile of a distribution X, for 3

In Abbas et al. (2008) and Wallsten et al. (2016) the probability-based methods also produce more accurate estimates of the true probability distributions, but this aspect of the results is contended in other studies (e.g., Bansal and Palley, 2017).

November 4, 2020

290

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

example, is |x−q|·(α·1{x>q} +(1−α)·1{x≤q} ), where q is the elicited α quantile and x is the realized X, while a quadratic scoring rule (QSR) for the likelihood p of event E is p2 · 1E + (1 − p)2 · 1E C (Gneiting and Raftery, 2007). The quadratic score takes only two values, while the loss function for the quantile varies with x. — FAAT avoids the statistical problems that arise with deriving perceived volatility estimates and comparing with empirical benchmarks. Overconfidence is measured directly with no need for exploring the (perceived) stochastic process of returns. Methods similar to FAAT were utilized in few psychology overconfidence studies (see Murphy and Winkler, 1974, for an early example). A common result of these scarce studies is that miscalibration decreases in interval evaluation compared to interval production. In Winman et al. (2004), for example, the miscalibration rate decreases from 34% to 15%, when distinct subjects judge the likelihood of 90% intervals produced by their peers. Teigen and Jørgensen (2005; experiment 2) ask subjects to add and subtract 25% of their point assessments and estimate the likelihood of the resulting intervals; the confidence estimates and actual hit rates almost agree. SpeirsBridge et al. (2010) alternatively advance a four-step procedure where experts submit a point prediction, lower and upper bounds, and then assess the likelihood of the resulting interval. Again, miscalibration significantly decreases compared to standard intervalproduction tasks. Given this consistent evidence, we utilize FAAT to test the robustness of overprecision, in addition to advancing it as an improved measure of forecast-overconfidence. Finally, returning to the Moore and Healy (2008) taxonomy of overconfidence, note that as FAAT measures overprecision in terms of the distance between perceived and actual accuracy rates, OC may be classified as a hybrid overprecision-overestimation measure. The tendency to overestimate one’s skills or knowledge was witnessed in diverse economic studies (e.g., Bhandari and Deaves, 2006; Neyse et al., 2016; Bergu, 2019). Bhandari and Deaves (2006), for example, study a survey where Canadian pension plan members are presented with two multiple choice questions regarding the past performance of stocks and bonds and report how certain they are in each answer. On average, confidence exceeded the correct choice rate by 22%, showing

page 290

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Separating Accuracy from Forecast Certainty

page 291

291

that the survey participants overestimate their knowledge of past returns. The current FAAT adopts a comparable approach to test for forecast-overprecision.

4.

Study 1

4.1.

Method

The questionnaire of Study 1 consisted of eight FAATs. The prediction stocks were members of TA-25, the 25 leading stocks of the Tel-Aviv exchange, and the prediction period was the last quarter of 2016. We used distinct stocks, from different sectors, to decrease the correlation between the target returns.4 In the first four assignments, the subjects did not receive information except for the stock’s name. The next four tasks were preceded by charts of the daily price trends and trading volumes in the first six months of 2016 (see Web Supplement A).5 The questionnaire was distributed in class to MBA students within four days, in early August 2016. The instructions were presented verbally and distributed in print, and the participation time was not effectively constrained. Surfing the Web and page turning were forbidden. To incentivize the FAATs we used random task selection combined with binarized scoring rules. The random task selection is employed to motivate subjects for independent work in each task (Cubitt et al., 1998; Hey and Lee, 2005; see Murad et al., 2016 for recent application). Binarized scoring rules are adopted to prohibit the bias that personal risk attitudes may impose on experimental responses (Hossain and Okui, 2013; Harrison et al., 2014). The probability of winning a fixed prize of 100 NIS (about US$25) decreased with the absolute prediction error (|F-r|) when the payment task was median forecasting. A standard QSR was used to derive the winning probability in the CONF assignments. The payment task was randomly drawn at the end of each session and used in January 2017 to determine the eligibility for the 100 NIS payoff. The students were invited to supervise the process and keep record of the

4 5

Indeed, the realized quarterly returns ranged between −22% and +11%. The Web supplements are available at http://www.bgu.ac.il/∼ sonsinod/.

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

292

draws.6 The next section reports the results for the N = 72 students (60% males) that completed the eight FAATs with no errors. 4.2.

Results

On average, the subjects assigned 75% likelihood to the fixed length intervals around their median forecasts, while the actual hit rate was only 55%. Almost 90% of the participants (64 of 72) showed CONF > HIT , so the hypothesis that subjects are as likely to show over- or underconfidence is easily rejected (p < 0.01 in a sign-test or a Wilcoxon signed-ranks test). Closer look at the data suggests that low accuracy and exaggerated forecast-confidence equally contribute to miscalibration. The Spearman correlation between OC and CONF is similar in magnitude to the correlation between OC and HIT and the hypothesis that the two correlations are equal in absolute value cannot be rejected (ρ(OC, CONF ) = 0.54; ρ(OC, HIT ) = −0.64; p = 0.48 by Hotelling–Williams test for dependent correlations). The disjoint confidence-accuracy effects on miscalibration clearly show in direct inspection of the data. At the OC = 50 level, for instance, subject 53 exhibited ultimate confidence CONF = 100 with about average HIT = 50, while subject 11 showed about average CONF = 75 with half smaller than average HIT = 25 (for convenience, we omit the % sign when discussing CONF , HIT and OC). The average confidence of the N = 8 underconfident (OC < 0) subjects, to take another perspective, ranged between 20 and 80, while their hit rates varied between 50 and 87.5. A comparison between the four tasks with history charts (HC) and the four preceding tasks where historical information was not provided (NHC) reveals that miscalibration is about half smaller in the HC tasks (mean OC 13 compared to 27; p < 0.01). Still, 2/3 of the subjects exhibit overconfidence in the HC condition, illustrating that overprecision persists when forecasters are faced with graphs showing strong price volatility. 6

The probability of winning the 100 NIS was 100% for prediction errors smaller than 1%, 98% for errors smaller than 2%, etc. The QSR was 100 × [1 − (1 − P )2 ], where P is the likelihood assigned to the realized event (hit or miss). When drawing the payment assignment, we also drew a 1–100 threshold. Subjects received the 100 NIS when their (payment task) winning probability exceeded the threshold.

page 292

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Separating Accuracy from Forecast Certainty

5.

b3964-ch12

page 293

293

Study 2

Study 2 was distributed as a take-home incentivized survey to a convenience sample (Ferber, 1977) consisting of participants in a professional preparation course to the Israeli SEC exams (N = 43), members of an Internet stock-trading forum (N = 29), and MBA students (N = 25). The questionnaire consisted of three slightly altered FAATs where subjects submit their best point prediction F for the target return and separately assess the likelihood of return exceeding F + δ1 and the likelihood of return smaller than F − δ2. The increments δ1 and δ2 were equal (10% and 5%) in two tasks, but asymmetric δ1 = 11% and δ2 = 6% in the third task. The likelihood CONF that subjects assign to the [F − δ2, F + δ1] interval was calculated by subtracting the tail events’ likelihoods from 100%; e.g., if the subject assigns 25% likelihood to return larger than F + δ1 and 30% likelihood to return smaller than F − δ2, then CONF = 45. One prediction stock repeated in all the questionnaires. The other two stocks were randomly drawn from the 150 largest stocks of the Tel-Aviv exchange, to preclude the bias that unrepresentative task selection may bring (Gigerenzer et al., 1991). The data were collected in March–April 2015 and the prediction period was the six months starting in May 1, 2015. As the equality of CONF , HIT and OC across the three subsamples could not be rejected, we report the results for the complete sample of N = 97 subjects (see Web Supplement B for the sample specific results).7 Surprisingly, more than 1/3 of the participants (N = 36 of 97) submitted at least one tail-likelihood strictly exceeding 50% (again, the violation rates did not differ significantly across the three subsamples). Intuitively, a large tail probability may represent a statement of lack of confidence in the submitted point forecast. Formally, the submission of tail likelihoods exceeding 50% connects to violations of set inclusion (monotonicity) documented in diverse psychology studies. Slovic et al. (1976), for example, ask subjects to estimate the likelihood of three events regarding a character Tom: (a) Tom will select journalism as his College major (b) but quickly become

7

Study 2 was run before Study 1 and motivated the simpler version of FAAT as presented in Figure 2.

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

page 294

Behavioral Finance: A Novel Approach

294

unhappy with his choice (c) and switch to engineering. The average likelihood that subjects assign to (a) alone was 0.21; the likelihood of (a) and (b) almost doubled to 0.39; and the average likelihood of the conjunction (a)–(c) slightly increased further to 0.41 (see Tversky and Koheler, 1994 for more examples). In the current application, subjects that violate monotonicity appear as if assigning 50% probability to the event R ≥ F , while assigning larger probability to smaller events such as R ≥ F + δ. The results for the complete sample and for the subjects that did not violate monotonicity are summarized in Table 1. The mean CONF was 49 for the complete sample (N = 97) and 62 for the subjects that did not violate monotonicity (N = 61). The respective HIT rates were 19 and 18, with 78% (90%) of the participants showing CONF > HIT . Overconfidence is stronger in the common task (mean OC of 41, with 75 of the 97 subjects showing OC > 0), but it is still highly significant in the randomly drawn assignments (mean OC of 24, with 69 of the 97 subjects showing OC > 0 when the common task is ignored). The hypothesis OC = 0 is rejected at p < 0.01 for each of the subsamples: SEC exams students, Web-forum traders and MBAs. Again, the correlations between OC and CONF (0.74 for N = 97 and 0.60 for N = 61) are similar in magnitude to the negative correlations between OC and HIT (−0.63 and −0.75). The overconfidence scores of subjects 83 and 113, for brief example, are almost similar (15.33 and 16.67), but subject 113 is low in confidence (15.33) and calibration (0), while subject 83 shows CONF = 83.33 and HIT = 66.67.

Table 1:

Study Study Study Study Study

1 2 2 3 3

Summary of results

(N = 72) – full sample (N = 97) – restricted sample (N = 61) – FAAT (N = 26) – SIFT (N = 24)

CONF

HIT

OC

% (OC > 0)

75 49 62 64 90

55 19 18 26 40

20 30 44 38 50

89 78 90 88 100

Note: This table presents the mean CONF , HIT and OC for each sample. % (OC > 0) is the proportion showing overconfidence. Confidence is exogenously fixed at 90% in SIFT.

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Separating Accuracy from Forecast Certainty

6.

b3964-ch12

page 295

295

Study 3

To compare the FAAT to the SIFT, we ran an exploratory smallsample study where N = 50 finance undergraduate students were randomly assigned to one of the two formats. The questionnaires consisted of six forecasting tasks and the incentivization method was similar to the one of the preceding experiments (see Web Supplement C for details).8 In the questionnaires with standard tasks, subjects submitted their median forecast and 95% confidence limits for the target return, as illustrated in Figure 1. The questionnaires with FAATs used the same six prediction targets and fixed margins of 5% around the median, as shown in Figure 2. The prediction targets included two domestic stocks, two index-linked certificates, and two leading stocks from the Paris and NYSE exchanges. The data were collected in March 2018 and the forecasting interval was the second quarter (April–June) of 2018. N = 26 students completed the FAAT questionnaires and N = 24 students completed the SIFT version. The results are summarized at the bottom panel of Table 1. On average, the 90% confidence intervals submitted by the SIFT subjects were 20.2% long, about twice longer than the FAAT 10% intervals. The hit rate of the 90% intervals was close to 40, generating an overconfidence score of about 50. The mean CONF of the subjects receiving the FAAT questionnaires was 64. The hit rate of the shorter FAAT intervals was only 26, so that the mean overconfidence score was 38 for the FAAT subjects. While the miscalibration score in FAAT is 12 points smaller, a Pitman permutation test could not reject equality (two-tailed p = 0.16). Overprecision, moreover, is clearly evident in both groups, with all 24 SIFT subjects showing HIT smaller than 90%, and 23 of the 26 FAAT subjects exhibiting HIT < CONF . While the sample size of Study 3 is too small for drawing general conclusions, we note that extending the comparative analysis requires extensive effort since the results may strongly 8

Since the incentivization of quantile assessments is relatively complex, the SIFT instructions (and, for symmetry, Study 3’s FAAT instructions) skipped the details regarding the incentivization of the quantile (likelihood) assessments. The instructions explained that we skip details for brevity but the method is designed so that truthful reporting maximizes the chances of winning the fixed NIS prize. Students were invited to send an email for details.

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

296

depend on the format of the tasks (Juslin et al., 1999) and the CONF/δ parameters assumed in defining the SIFT/FAAT, respectively (Teigen and Jørgensen, 2005; Speirs-Bridge et al., 2010).9

7.

Discussion

The vast psychology research on overconfidence proves the robustness of miscalibration, illustrating that it is hard to debias subjects from overtrusting their private assessments (cf., Alpert and Raiffa (1982) for an early report of failed debiasing attempts; Moore et al. (2015) for a recent survey). The results of the three current studies indeed reveal that the exaggerated confidence in subjective forecasts persists when prospective investors judge the likelihood of exhibiting prediction errors smaller than pre-assigned levels. Over three studies, 163 of 195 finance competent subjects overestimate their hit rates, in line with the forecast-overprecision hypothesis. By separating forecast-accuracy from forecast-certainty, the threestep FAAT improves on standard interval forecasting, showing that inaccuracy and excessive certainty act as distinct sources of forecastmiscalibration. As point forecasting and likelihood assessments are easier and more natural to subjects than quantile assessments (Abbas et al., 2008; Wallsten et al., 2016), FAAT may produce less noisy and more meaningful measures of forecast-overprecision. The FAAT overconfidence score may therefore prove more powerful in testing hypothesized links between forecast-overconfidence and aspects of financial decision (Skala, 2008). Empirical studies illustrate that intuitive proxies for overconfidence such as partition by gender (Barber and Odean, 2000), managers’ inclination to hold on to their stock options (Malmendier and Tate, 2008), or traders’ risk exposure in terms of position size (Forman and Horton, 2019) link with exaggerated non-profitable trading. The micro level, experimental or survey-based evidence regarding the overprecision-trading correlations, however, is inconclusive and statistically weak (cf., Biais et al., 2005; Glaser and Weber, 2007; Deaves et al., 2009; Nosic and Weber, 2010; Broihanne et al., 2014; Fellner-R¨ohling and Kr¨ ugel,

9

See Teigen and Jørgensen (2005) Studies 2–3 for specific examples.

page 296

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Separating Accuracy from Forecast Certainty

b3964-ch12

page 297

297

2014; Merkle, 2017). By separating forecast-accuracy from forecastcertainty the FAATs of the present study open possibility for more efficient tests of the links between trading propensity/profitability and forecast-certainty, accuracy and overconfidence. Acknowledgments Studies 1 and 2 were conducted while Doron Sonsino was at the College of Management Academic Studies (COMAS) in Israel and funded by the research authority of COMAS. We thank participants at the Warwick University 2016 FUR conference, the Tucson ESA 2016 meetings, the SPUDM 2017 Technion conference, the IAREP 2017 COMAS meetings, the Vienna ESA 2017 meetings and seminars at the Gothenburg Research Institute, Ben-Gurion University and the Technion for their helpful comments. We particularly thank David Budescu and Nigel Harvey for directing us to related psychology research on interval evaluation. References Abbas, A. E., Budescu, D. V., Yu, H. T., and Haggerty, R. (2008). A comparison of two probability encoding methods: Fixed probability vs. fixed variable values. Decision Analysis, 5 (4), 190–202. Alpert, M. and Raiffa, H. (1982). A progress report on the training of probability assessors. In D. Kahneman, P. Slovic, and A. Tversky (Eds.), Judgment under Uncertainty: Heuristics and Biases. Cambridge: Cambridge University Press. Astebro, T., Herz, H., Nanda, R., and Weber, R. A. (2014). Seeking the roots of entrepreneurship: Insights from behavioral economics. Journal of Economic Perspectives, 28 (3), 49–70. Bansal, S. and Palley, A. (2017). Is it better to elicit quantile or probability judgments? A Comparison of direct and calibrated procedures for estimating a continuous distribution. Working Paper, Kelly School of Business Research Paper No. 17–44. Barber, B. M. and Odean, T. (2000). Trading is hazardous to your wealth: The common stock investment performance of individual investors. The Journal of Finance, 55 (2), 773–806. Ben-David, I., Graham, J. R., and Harvey, C. R. (2013). Managerial miscalibration. The Quarterly Journal of Economics, 128 (4), 1547–1584.

November 4, 2020

298

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

Bergu, K. (2019). Overconfidence and trading behavior: Does there exist a causal link? Working Paper. Bhandari, G. and Deaves, R. (2006). The demographics of overconfidence. The Journal of Behavioral Finance, 7 (1), 5–11. Biais, B., Hilton, D., Mazurier, K., and Pouget, S. (2005). Judgmental overconfidence, self- monitoring, and trading performance in an experimental financial market, The Review of Economic Studies, 72 (2), 287–312. Broihanne, M. H., Merli, M., and Roger, P. (2014). Overconfidence, risk perception and the risk-taking behavior of finance professionals. Finance Research Letters, 11 (2), 64–73. Cubitt, R. P., Starmer, C., and Sugden, R. (1998). On the validity of the random lottery incentive system. Experimental Economics, 1 (2), 115–131. Daniel, K. and Hirshleifer, D. (2015). Overconfident investors, predictable returns, and excessive trading. The Journal of Economic Perspectives, 29 (4), 61–87. Daniel, K., Hirshleifer, D., and Subrahmanyam, A. (1998). Investor psychology and security market under-and overreactions. The Journal of Finance, 53 (6), 1839–1885. Deaves, R., L¨ uders, E., and Luo, G. (2009). An experimental test of the impact of overconfidence and gender on trading activity. Review of Finance, 13, 555–575. Fellner-R¨ ohling, G., and Kr¨ ugel, S. (2014). Judgmental overconfidence and trading activity. Journal of Economic Behavior and Organization, 107, 827–842. Ferber, R. (1977). Research by convenience. Journal of Consumer Research, 4 (1), 57–58. Forman, J. and Horton, J. (2019). Overconfidence, position size, and the link to performance. Journal of Empirical Finance, 53, 291–309. Gigerenzer, G., Hoffrage, U., and Kleinb¨olting, H. (1991). Probabilistic mental models: A Brunswikian theory of confidence. Psychological Review, 98 (4), 506–528. Glaser, M., Langer, T., and Weber, M. (2013). True Overconfidence in Interval Estimation Based on a New Measure of Miscalibration. Journal of Behavioral Decision Making, 26, 405–417. Glaser, M. and Weber, M. (2007). Overconfidence and trading volume. The Geneva Risk and Insurance Review, 32 (1), 1–36. Gneiting, T. and Raftery, A. E. (2007). Strictly proper scoring rules, prediction, and estimation. Journal of the American Statistical Association, 102 (477), 359–378.

page 298

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Separating Accuracy from Forecast Certainty

b3964-ch12

page 299

299

Graham, J. R. and Harvey, C. R. (2001). Expectations of Equity Risk Premia, Volatility and Asymmetry from a Corporate Finance Perspective (No. w8678). National Bureau of Economic Research. Grosshans, D. and Zeisberger, S. (2018). All’s well that ends well? On the importance of how returns are achieved. Journal of Banking & Finance, 87, 397–410. Harrison, G. W., Mart´ınez-Correa, J., and Swarthout, J. T. (2014). Eliciting subjective probabilities with binary lotteries. Journal of Economic Behavior and Organization, 101, 128–140. Hey, J. D. and Lee, J. (2005). Do subjects separate (or are they sophisticated)? Experimental Economics, 8 (3), 233–265. Hossain, T. and Okui, R. (2013). The binarized scoring rule. The Review of Economic Studies, 80 (3), 984–1001. Hyndman, R. J. and Athanasopoulos, G. (2018). Forecasting: Principles and Practice. Melbourne: OTexts. Juslin, P., Wennerholm, P., and Olsson, H. (1999). Format dependence in subjective probability calibration. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25 (4), 1038–1052. Juslin, P., Winman, A., and Hansson, P. (2007). The na¨ıve intuitive statistician: A na¨ıve sampling model of intuitive confidence intervals. Psychological Review, 114 (3), 678–703. Keefer, D. L. and Bodily, S. E. (1983). Three-point approximations for continuous random variables. Management Science, 29 (5), 595–609. Klayman, J., Soll, J. B., Gonzalez-Vallejo, C., and Barlas, S. (1999). Overconfidence: It depends on how, what, and whom you ask. Organizational Behavior and Human Decision Processes, 79 (3), 216–247. Langnickel, F. and Zeisberger, S. (2016). Do we measure overconfidence? A closer look at the interval-production task. Journal of Economic Behavior and Organization, 128, 121–133. ¨ Lawrence, M., Goodwin, P., O’Connor, M., and Onkal, D. (2006). Judgmental forecasting: A review of progress over the last 25 years. International Journal of Forecasting, 22 (3), 493–518. Malmendier, U. and Tate, G. (2008). Who makes acquisitions? CEO overconfidence and the market’s reaction. Journal of Financial Economics, 89 (1), 20–43. Malmendier, U. and Taylor, T. (2015). On the verges of overconfidence. Journal of Economic Perspectives, 29(4), 3–8. McKenzie, C. R., Liersch, M. J., and Yaniv, I. (2008). Overconfidence in interval estimates: What does expertise buy you? Organizational Behavior and Human Decision Processes, 107 (2), 179–191.

November 4, 2020

300

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch12

Behavioral Finance: A Novel Approach

Merkle, C. (2017). Financial overconfidence over time: Foresight, hindsight, and insight of investors. Journal of Banking and Finance, 84, 68–87. Moore, D. A. and Healy, P. J. (2008). The trouble with overconfidence. Psychological Review, 115 (2), 502–517. Moore, D. A., Tenney, E. R., and Haran, U. (2015). Overprecision in judgment. In G. Keren and G. Wu (Eds.), The Wiley Blackwell Handbook of Judgment and Decision Making. Hoboken: John Wiley & Sons Inc., 182–209. Murad, Z., Sefton, M., and Starmer, C. (2016). How do risk attitudes affect measured confidence? Journal of Risk and Uncertainty, 52 (1), 21–46. Murphy, A. H. and Winkler, R. L. (1974). Probability forecasts: A survey of National Weather Service forecasters. Bulletin of the American Meteorological Society, 55 (12), 1449–1452. Neyse, L., Bosworth, S., Ring, P., and Schmidt, U. (2016). Overconfidence, incentives and digit ratio. Scientific Reports, 6, 23294. Nosic, A. and Weber, M. (2010). How riskily do I invest? The role of risk attitudes, risk perceptions, and overconfidence. Decision Analysis, 7(3), 282–301. Oberlechner, T. and Osler, C. (2012). Survival of overconfidence in currency markets. Journal of Financial and Quantitative Analysis, 47 (01), 91–113. Skala, D. (2008). Overconfidence in psychology and finance — An interdisciplinary literature review. Bank i Kredyt, 39 (4), 33–50. Slovic, P., Fischhoff, B., and Lichtenstein, S. (1976). Cognitive processes and societal risk taking. In H. Jungermann and Gerard de Zeeuw (Eds.), Decision Making and Change in Human Affairs. Springer: Netherlands, 7–36. Sonsino, D. and Regev, E. (2013). Informational overconfidence in return prediction — more properties. Journal of Economic Psychology, 39, 72–84. Speirs-Bridge, A., Fidler, F., McBride, M., Flander, L., Cumming, G., and Burgman, M. (2010). Reducing overconfidence in the interval judgments of experts. Risk Analysis, 30 (3), 512–523. Teigen, K. H. and Jørgensen, M. (2005). When 90% confidence intervals are 50% certain: On the credibility of credible intervals. Applied Cognitive Psychology, 19 (4), 455–475. Tversky, A. and Koehler, D. J. (1994). Support theory: A nonextensional representation of subjective probability. Psychological Review, 101 (4), 547–567.

page 300

November 4, 2020

10:14

Behavioral Finance: A Novel Approach – 9in x 6in

Separating Accuracy from Forecast Certainty

b3964-ch12

page 301

301

Wallsten, T. S., Shlomi, Y., Nataf, C., and Tomlinson, T. (2016). Efficiently encoding and modeling subjective probability distributions for quantitative variables. Decision, 3 (3), 169–189. Winman, A., Hansson, P., and Juslin, P. (2004). Subjective probability intervals: How to reduce overconfidence by interval-evaluation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30 (6), 1167–1175.

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0013

Chapter 13

Optimal Contracts with Intra-Principal Conflicts and the Ubiquity of Earnings Management N. K. Chidambaran∗ , Bharat Sarath† , and Lingyi Zheng‡

Abstract The agency paradigm is primarily concerned with compensation contracts that align the interests of top management with those of shareholders, but such alignment may be imperfect. In particular, if shareholders differ in their time-preferences, it will be impossible to find a contract that fully aligns manager interests with those of both long-term and short-term shareholders. In this study, we examine the relationship between the inter-temporal weighting on a CEO’s contract and shareholder welfare in a setting where shareholders liquidate choose to cash in their holdings in different time-periods. We show that the managerial compensation contract will typically create incentives for the manager to trade-off short-term price increases with long-term value creation and it is possible to design a second-best contract that is Pareto optimal. This analysis shows that the impact of earnings management is often mischaracterized as Pareto suboptimal even though it may be a requisite to make both long and short-term shareholders better-off on average. Keywords: Compensation duration, investor horizon, real earnings management, myopia, investment distortions

∗

Fordham University, USA; [email protected]. Rutgers University, USA; [email protected]. ‡ The Hong Kong Polytechnic University, Hong Kong; polyu.edu.hk. †

303

lingyi.zheng@

page 303

November 4, 2020

304

1.

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Behavioral Finance: A Novel Approach

Introduction

There is a perception that managers of public corporations are fixated on short-term performance goals rather than long-term shareholder value (Graham et al., 2006). The blame is usually ascribed to the use of equity incentives that result in an unhealthy focus on the current stock price at the expense of long-term performance. The goal of this chapter is to examine a formal economic framework where the use of equity incentives and the resulting short-termism is consistent with economic efficiency. We show that the existence of multiple shareholder types, some of whom have a preference for short-term payoffs, makes it efficient to boost short-term cash flows at the expense of damaging total cash flows. This result is related to the conclusion of Bolton, Scheinkman, and Xiong (2006) where the presence of speculators is sufficient to ensure that the efficient equilibrium involves some level of “myopia”. Agency theory Jensen and Meekling (1976) is the central paradigm for studying CEO compensation and it is formalized through Principal–Agent models (Holmstrom, 1979, Gjesdal, 1982, Grossman and Hart, 1983). Under an assumption that the Agent faces a private externality, usually characterized as effort aversion, the Principal faces the problem of maximizing payoff through the design of an optimal contract. The payoff associated with the optimal contract is called the “second-best” as opposed to the “first-best” that is attained if there is no private externality for the agent. We use an extension of the Principal–Agent model where a diversity of investors contract with a single agent. The shareholders differ in their time-preferences regarding the cash generated by the company. This intra-principal conflict over the value of payoffs affects their preferences over the action choices of the agent. In more detail, we focus on a horizon conflict across shareholders that creates differences over the contract that is considered optimal for the agent and examine whether this will result in an equilibrium that would encourage the CEO to “front-end” cash flows. In any framework involving multiple participants, it is impossible, in general, to come up with a single “social welfare” function that simultaneously maximizes the utility of all participants (Arrow, 1951) although it can be established in special cases (Wilson, 1968). In our context, if shareholders differ in their time-preferences, it

page 304

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Optimal Contracts with Intra-Principal Conflicts

b3964-ch13

page 305

305

is impossible to find a single contract that fully aligns manager’s interests with those of both long-term and short-term shareholders. Therefore, for any given incentive contract, there will typically be some loss (at least for one shareholder type) relative to the contract that would be considered optimal for that shareholder type. We will identify the contract that minimizes the average loss across all shareholders. Since any other contract will involve a greater total loss than the minimum average loss contract, it follows that at least one of the shareholder types must be worse off under any alternative contract, and hence, that the minimum average loss contract is Pareto optimal. We use a two-period principal-agent model where there is trading and a market clearing price in each period. There are two types of shareholders: short-term and long-term. Short-term shareholders wish to liquidate their holdings at the end of the first period whereas long-term shareholders wish to liquidate in the second-period. The CEO has perfect information regarding future cash flows in both periods but investors face uncertainty in the first-period and become fully informed only at the end of the second-period. The CEO’s compensation depends on both the first-period stock price and the secondperiod stock price. The CEO chooses a decision variable that can shift the cash flows to the first-period and her decision is determined by the compensation weights assigned to the first and second-period stock prices. In this setting, we examine the characteristics of the resulting (Pareto efficient) incentive contract and the resulting action and payoffs. We establish that: (i) the second-best contract weights the CEO’s short and long-term compensation in exactly the same proportion as the mix of short and long-term shareholders; (ii) this contract creates incentives to increase first-period cash flows even though total cash flows may be reduced and (iii) the board might implement a sub-optimal contract based on their own preferences without any pressure from the manager. Empirical applications of the agency paradigm often test whether equity-based incentives lead to CEO action choices that shift cash flows from later periods to earlier periods even though the total cash flows across all periods are reduced. Such CEO behavior is usually termed as “managerial myopia” in the finance literature (Stein, 1989) and as “earnings management” in the accounting literature (Sloan, 1996). In our model, the decision choice resting with the CEO has exactly these “negative” characteristics. The next step in many

November 4, 2020

306

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Behavioral Finance: A Novel Approach

empirical papers is an imputation that equity-based compensation increases “agency costs”. Viewed from the perspective of our model, or for that matter, any Principal-Agent model, this is a mischaracterization. The simple point is that only second-best payoffs can actually be observed in practice. Since agency cost is defined as the difference between first- and second-best payoffs to the shareholders, its measurement involves a counterfactual (i.e., what the payoff might have been in first-best). It is difficult to imagine an experiment that can convincingly measure the difference between first and second-best without lapsing into untestable assumptions about first-best payoffs. In contrast, the measurement of costs of contract sub-optimality, that is the difference in payoffs between the contract offered to the CEO and second-best, is empirically feasible. Inter-firm comparisons of the short and long-term components of the CEO’s compensation could lead to predictions about the loss in payoffs resulting from the failure to properly weight short and long-term compensation (Divya Anantharaman et al., 2014). We incorporate this idea into our model and allow the CEO to be compensated both on the short and longterm stock prices and focus our analysis on contract sub-optimality Despite much academic criticism, the pure equity component of the average CEO’s pay increased from 35% to 52% over the period 2009–2018.1 In addition, the use of stock-based incentives as a potential motive for CEO misconduct was decisively rejected by the legal system. Ever since the enactment of the PSLRA in 1995, US courts have ruled that the combination of opportunity and motive required to plead scienter cannot be based purely on the existence of compensation schemes that create incentives to run up the stock price.2 The question of whether any compensation could be considered excessive was the subject of a rare public debate between Justices Easterbrook and Posner of the Seventh circuit of appeals (Harris vs Jones 2009). While Justice Easterbrook held that excess compensation is an unlikely possibility given market forces, Justice Posner held that 1

https://corpgov.law.harvard.edu/2019/04/16/2019-u-s-executive-compensati on-trends/. 2 For example, the Fifth circuit of appeals ... held that allegations concerning incentive compensation could not support an inference of scienter because “the vast majority of corporate executives receive such compensation”; https://www. jdsupra.com/legalnews/fifth-circuit-affirms-dismissal-of-62157/.

page 306

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Optimal Contracts with Intra-Principal Conflicts

b3964-ch13

page 307

307

the market for CEO compensation was lapsing into cronyism and functioning poorly.3 Justice Posner’s opinion cited many academic studies including (Bebchuk et al., 2002) that asserted directors often acted in the interest of CEO’s rather than shareholders.4 The case was appealed to the Supreme Court that reaffirmed precedent, and ruled that allegations of excess compensation should be treated on a case-by-case basis rather than relying on abstractions.5 We incorporate this theme into our analysis by focusing on factors related to the board of directors that could lead to sub-optimal contracts. Accounting literature distinguishes between the effects of equitybased compensation as it relates to “accrual earnings management” and “real earnings” management. Accrual Earnings management refers to the reporting of choices that do not change the underlying cash flows whereas real earnings management does change the underlying cash flows (Gunny, 2010). The CEO’s decision variable in our model can be interpreted as either type of earnings management by assuming that total cash flows across both periods stay the same irrespective of the CEO’s choice (accrual earnings management) or that they decrease (real earnings management). From a theoretical perspective, the economic rationale for either type of earnings management raises thorny problems of investor rationality. In a rational expectations equilibrium, the probability distribution of expected future cash flows would be the same as the distribution of realized cash flows. Therefore, deferring some portion of the CEO compensation to the future would result in the elimination (or at least, reduction) of decision choices that reduce future value. It is important to point out the difference between our framework and that adopted in Stein (1989) where earnings management takes place even though the investors are not deceived by the CEO’s actions. That paper uses an elegant model where investors are not fooled by myopic behavior but managers are locked into inefficient decisions because the Nash equilibrium where

3

See Easterbrook’s opinion at https://www.delawarelitigation.com/uploads/fil e/int69.PDF. 4 See Posner’s opinion at https://www.delawarelitigation.com/uploads/file/int 6A.PDF. 5 https://supreme.justia.com/cases/federal/us/559/335/.

November 4, 2020

308

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Behavioral Finance: A Novel Approach

managers are not myopic is unsustainable.6 However, it is unclear whether a Nash equilibrium is the appropriate vehicle for describing investor–manager–shareholder interactions since these interactions are repeated over long periods of time. The no-manipulation equilibrium can be sustained through pre-commitment which is a fundamental assumption of the Principal-Agent model. In the specific context of our model, the no-manipulation equilibrium is always the dominant choice if the first-period stock price rationally anticipates the second-period stock price. Therefore, we make an assumption common in accounting research that investors do not fully anticipate the true correlation of cash flows across periods, and overestimate the persistence of first-period cash flows Bernard and Thomas (1989). This behavioral assumption of limited rationality is necessary to support any model of earnings management that is consistent with the underlying assumptions of both the Principal-Agent model and the empirical literature on earnings management. In our model, we combine ideas from finance, accounting and law to construct a framework that conforms broadly with observed empirical features. Shareholders are assumed to have heterogeneous time preferences. They would like to offer equity incentives that align the CEO’s horizon with their own time-horizon by changing the compensation weights associated with the first and second-period stock prices. The greater the alignment of the contract with one group of shareholders, the higher the wealth transfer (through CEO’s decisions) to those shareholders. The stock price in the first period, though endogenously determined, can be influenced by the CEO through earnings management because investors overestimate the persistence of first period cash flows. In this setting, we identify the contract that maximizes the wealth of the average shareholder and show that the weights on each period’s stock price exactly mirrors the proportion of each type of shareholder. However, it is possible that sub-optimal contracts are offered that reduce the total payoffs while increasing payoffs to one or the other group. We also show how a small group of directors chosen from a pool of applicants that contains both long and short-term horizon candidates

6

“The Nash approach clearly exposes the fallacy inherent in a statement such as since managers can’t systematically fool the market, they won’t bother trying” Stein, (1989).

page 308

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Optimal Contracts with Intra-Principal Conflicts

b3964-ch13

page 309

309

may not proportionately represent the mix of shareholders, that is the proportion of directors with long or short-horizon leanings will be different from the average across shareholders even if the pool of applicants for the board has exactly the same composition as that of shareholders. This finding contrasts with arguments of cronyism or other types of collusion between directors and the CEO as in Bebchuk et al. (2002). Our results provide a reason for the widespread prevalence of equity earnings management/myopia similar to Bolton et al. (2006). Rewarding CEO’s for the short-term stock price increases the value to short-term shareholders and reduces the value to long-term shareholders. The optimal contract is one that aligns the CEO’s contract horizon with the average shareholder horizon. Unlike the findings in some empirical studies, misalignment of the CEO’s horizon in either direction results in sub-optimality, that is, increasing the weight on the long-term stock price may actually generate inefficiency relative to the shareholder’s perceptions of firm value. At least within the context of our model, failing to manage earnings can result in a loss of aggregate value across long and short-term shareholders. However, for earnings management to have pricing consequences, it is necessary that investors are not fully rational and overestimate the persistence of first period cash flows. 2. 2.1.

Model Framework Model Time Line

At t = 0, β fraction of the firm’s shareholders are short-term (ST) and (1 − β) fraction of shareholders are long-term shareholders (LT). At t = 1, short-term shareholders liquidate their holdings and the shares are purchased by a new set of short-term shareholders (NS). Earnings of the firm are reported and distributed at t = 1 and t = 2 and the manager can manage the level of reported earnings through real earnings management at t = 0 to influence shareholders’ trading decision so that the market-clearing price at t = 1 and the long-term value of the firm at t = 2 as well.7 Figure 1 shows the time line and the details of our two-period, three-date model framework. 7 We use real earnings management instead of accrual earnings management since real earnings management can affect firm’s cash flows and therefore

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Behavioral Finance: A Novel Approach

310 β given; α set

eˆ1 reported p1

eˆ2 reported p2

t=0 e observed by CEO CEO chooses a

t=1 ST Shareholders sell New Shareholders buy

t =2 Firm Liquidated

Figure 1:

2.2.

b3964-ch13

Two-period Model

Model Parameters

At t = 0, shareholders set the manager’s compensation contract. The primary conflict between managers and shareholders stems from the distribution of cash flows across periods. Some shareholders wish to exit at t = 1 and would like the first-period price, p1 , to be as high as possible. Other shareholders plan to hold till the second period and their payoffs depend on the price at t = 2, denoted by p2 . We assume that the contract implemented for the CEO may reflect a compromise across different types of shareholders in terms of weighting the first period and second period prices. Specifically, the parameter α, that determines the fraction of the manager’s compensation that is based on the short-term value of the firm and (1 − α), fraction is based on the long-term value of the firm. The CEO’s decision variable is denoted by a ∈ [0, 1]. The choice of a affects the reported income as well as the actual realized income across the two periods. For simplicity, we assume that the total income across the two periods is unaffected by the level of a although the model can easily accommodate an effect of a on the total twoperiod income. Our setting allows us to focus on the issue of intertemporal shifting of cash flows rather than the effect on total cash flows. In other words, the manager chooses an investment/reporting strategy that favors either the short-term or long-term shareholders based on the contract that has been implemented. At t = 0, absent any action from the manager, the firm will generate e˜ as earnings in each period. e˜ is distributed according to the normal N (E[e], σe ). The manager privately observes the actual level of earnings e, but shareholders only know the distribution e˜. At t = 1, depending on

long-term value while accrual earnings management only shifts the accrual part of earnings without directly influencing firm’s cash flows.

page 310

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Optimal Contracts with Intra-Principal Conflicts

page 311

311

the choice of a, the firm reports earnings, eˆ1 = e + a, a higher than the permanent earnings e, The increase in reported earnings may be viewed either as a change in project selection so cash flows get frontended (Real Earnings Management) or purely as a shift of accruals (Accrual Earnings Management) or a combined effect. While the manager observes e and a, shareholders update their beliefs about t = 2 earnings and price based on eˆ1 = e + a. It is at this point that we invoke a behavioral updating rather than using a Bayesian-rational framework. We denote the expectation by eˆ2 (a|e) and assume that it is always increasing in a although the realized second period cash flows are in fact decreasing in a. As noted in the introduction, it is impossible to fit the empirical literature on earnings management (or myopia) without assuming some form of investor irrationality. In our framework, we assume that investors systematically underestimate the long-term consequences of managerial actions that increase short term earnings. More formally, at t = 1 shareholders expect that the earnings of the firm at t = 2 to be, E[eˆ2 |a, e] = eˆ2 (a|e) > e where

 ∂ 2 eˆ2 (a|e) ∂ˆ e2 (a|e)  ∂ˆ e2 (a|e) > 0; (ii) < 0; (iii) =0 (i) ∂a ∂a2 ∂a a=1

(1)

The expected second-period earnings eˆ2 (a|e) are increasing in the first-period earnings but the effect of the earnings management parameter, a, has a decreasing marginal effect. Investors take into account the possibility that higher eˆ1 could be because of higher true earnings e or due to higher earnings management a. Nevertheless, they fail to adequately cancel out all the effects of a except at the extreme value a = 1. As we have noted in the introduction, some behavioral irrationality is needed in order to sustain earnings management in equilibrium. The specific choice is consistent with the empirical explanations advanced in Bernard and Thomas (1989) and Sloan (1996). Short-term shareholders trade at t = 1 and sell their holdings to a new set of short-term investors who hold shares till t = 2. The market clearing price, p1 , will be a function of the expectations for future earnings, the number of shares traded and the risk exposure to future prices. In particular, the first period price depends on the expectations of future earnings eˆ2 (a|e) (which typically differs from

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

page 312

Behavioral Finance: A Novel Approach

312

the actual value e − a). At t = 2, all uncertainty is resolved and the values of e and a become common knowledge.8 Actual earnings, e − a, are distributed to shareholders and the future realizations of earnings are known with certainty to be e − a. Therefore, p2 = K ×e2 = K(e−a), which is the capitalized value of the firms earnings under certainty. Within the limits of our model, the new shareholders cannot impose a penalty at t = 2 if a is positive. 2.3.

Shareholder Trading at t = 1

At t = 1, short-term investors sell all their shares to new investors. The supply of shares is equal to βQ, where Q is the number of shares outstanding and β is the fraction owned by short-term shareholders. New shareholders maximize their expected utility by setting prices appropriately for the quantity traded. The trade price maximizes investors’ expected utility of expected future cash flows and clears the market. Investors’ expectation of future earnings eˆ2 (a|e), and the conjectured liquidation value E[p2 ] = K × eˆ2 (a|e), will determine the market clearing price at t = 1. All expectations are formed after observing the reported earnings eˆ1 = e + a. New shareholders buy shares QN S at a price p1 and expect proceeds of E[p2 ] = K × eˆ2 (a|e) when the firm is liquidated at t = 2. The expected utility of the new shareholders (N S), is E[UN S (−QN S p1 + QN S K eˆ2 )] = E[UN S {QN S K(eˆ2 − p1 )}] (2) Using a Taylor series expansion of Eq.(2) at K eˆ2 = p1 , we obtain E[UN S (0)] + U  (0)QN S E[(K + 1)eˆ2 − p1 ] +

 (0)Q2 UN S NS E[(K + 1)eˆ2 − p1 ]2 2

(3)

e2 − p1 ] = (K + 1)E[ˆ e2 ] − p1 Substituting E[UN S (0)] = 0, E[(K + 1)ˆ and E[(K + 1)ˆ e2 − p1 ]2 = σ 2 + ((K + 1)E[ˆ e2 ] − p1 )2 in Eq. (3) 8

This assumption is consistent with a two-period model. The extension to multiperiods is discussed in Section 3.1.

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Optimal Contracts with Intra-Principal Conflicts

page 313

313

we get,  UN ˆ2 − p1 ) + S (0)QN S (E[K e

 (0)Q2 UN S NS 2

(σ 2 + (E[(K + 1)ˆ e2 ] − p1 )2 )

(4)

Investor’s choose to trade such that their expected utility is maximized. The first order condition for the maximization with respect to the quantity of shares traded is:   e2 ] − p1 ) + (σ 2 + (KE[ˆ e2 ] − p1 )2 )UN UN S (0)(KE[ˆ S (0)QN S = 0.

(5) U  (0)

) as γ. The market clearing Denote the risk aversion level (− UNS  NS (0) ∗ price p1 for quantity of shares traded QN S , therefore, is given by, p∗1

= E[K eˆ2 ] −

1 γQNS

−



( γQ1NS )2 − 4σ 2

(6)

2

Simplifying further, p∗1 = E[K eˆ2 ] −

   1 2 2 2 1 − 1 − 4σ (QN S ) γ 2γ QN S

(7)

Markets clear when the quantity of shares purchased by new shareholders is equal to the number of shares sold by the initial short-term shareholders, i.e., when QN S = βQ. Therefore, p∗1 = E[K eˆ2 ] −

  1  1 − 1 − (2γβ Q σ)2 2γβ Q

(8)

Equation (7) shows that the market-clearing price, p∗1 , is related to investors’ expectation about the terminal value of the firm and a risk adjustment factor. The second term in the equation is strictly positive except when there is no risk (σ 2 = 0) or the purchasing shareholders are risk-neutral (γ = 0), when it is equal to zero. We next turn to the manager’s incentive maximization problem given the optimal p∗1 as a solution to the shareholders’ investment problem.

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

page 314

Behavioral Finance: A Novel Approach

314

Compensation Maximization

2.4.

The CEO’s expected compensation is assumed to be a function of prices p1 and p2 : C = Et=0 [αp1 + (1 − α)p2 ]. Note that the CEO faces no uncertainty. Therefore, risk-aversion does not play a role in the CEO’s decision. The CEO chooses the level of a to maximize compensation as given by her payoff function. The optimal choice of a maximizes her compensation, subject to shareholder trading at t = 1 and the market clearing price pˆ∗1 . As the CEO faces no uncertainty over e or a, the optimization program is max a

C

=

αˆ p1 (a|e) + (1 − α)K(e − a)

subject to: p1 = p∗1 (a|e)

(9)

Substituting for p1 = p∗1 (a|e), we get, C(a|e) = max {αK eˆ2 (a|e) − (1 − α)K(e − a)} a    1 − 1 − (2σβQγ)2 −α γ βQ

(10)

Differentiating the objective function with respect to a, we get, ∂ˆ e2 (a|e) ∂C(a|e) = αK + (1 − α)K = 0 ∂a ∂ˆ a

(11)

Therefore, the optimal choice of a is such that, 1 ∂ˆ e2 (a|e) = K( − 1) ∂a α

(12)

The optimal choice of earnings management, a, is reached as follows. An increase of a unit in a increases the expected second period price e2 ] by the amount K ∂E[ˆ and decreases the actual second period ∂a price by K. The trade-off between increasing the first period price by inducing optimistic expectations and the negative impact on the second period price due to the cost of earnings management depends on the relative weight of each period in the managers compensation captured by the parameter α. If α is very low, the right-hand side of the FOC will be very high. Given the concavity of eˆ2 (a|e)

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

page 315

Optimal Contracts with Intra-Principal Conflicts

315

in a, and assuming the slope of eˆ2 in a is finite, this will drive the corner solution a = 0 for low values of α. On the other hand, the opposite corner solution a = 1 will obtain only if α = 1 as in this case, the right-hand side will be 0 and a = 1 as per (i), (ii) and (iii). To summarize, the solution in (12) is quite intuitive in that the CEO will not engage in any form of manipulation if their short-term compensation proportion is very low whereas they will manipulate to the maximum if all their compensation is based on short-term outcomes. The benchmark result in (12) will be used to compare the weight α that maximizes shareholder welfare. As noted in our earlier discussion, given some proportion of short-term shareholders, there is a role for weighting short-term performance in addition to long-term performance. 2.5.

Optimal Compensation Horizon

Any attempt at deriving an optimal contract weight α involves a trade-off between short and long-term shareholders. Increasing α increases a and, thereby, p∗1 and the payoff to short-term shareholders. On the other hand, it decreases p2 and the payoff to long-term shareholders. Therefore, some criterion is needed to combine the interests of both groups into a single function. We choose the simplest and maximize the per-share value across both groups of shareholders. This is easy to derive as given below: The value of the shares held by short-term shareholders is equal to the value they receive for selling their shares at t = 1 for p∗1 whereas long-term shareholders obtain the value p2 . Note that there is no residual uncertainty at t = 2 and we can effectively assume that the long-term shareholders are risk-neutral. Short-term shareholder Value = βQp∗1 and Long-term shareholder Value = (1 − β)Qp2 (13) Normalizing by Q, the shareholder value on a per-share basis is: Shareholder Value = βp∗1 + (1 − β)p2

(14)

November 4, 2020

316

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

page 316

Behavioral Finance: A Novel Approach

Any level of α results in a choice of a as per (12) and maximizes the value: V alue M ax = [αp∗1 + (1 − α)p2 ]

(15)

Comparing it follows immediately that setting α = β results in the optimal contract. The difference between α and β is the degree of misalignment between the horizons of the shareholders and managers. We note that the shareholders (in totality) become worse off irrespective of whether α > β or α < β, i.e., a misalignment in either direction results in the manager’s optimal choice of a differing from the level of a that maximizes shareholder value. If shareholders have different time preferences, the CEO’s decisions can lead to intra-shareholder transfers. So the question arises as to whether there are Pareto optimal contracts where one shareholder group cannot be made better off without making the other shareholder group worse off. The contract that balances the short and long-term compensation in the same proportion as that of short and long-term shareholders is Pareto efficient in this sense. As it maximizes the total value of the firm, any change must leave one or (both) of the shareholder groups worse off. Indeed, it is the only Pareto efficient contract even though it does not maximize the terminal value of the firm. This finding suggests that the criticism of shorttermism and “myopia” may not be fully justified from an economic perspective. The next step in our analysis is to see why misalignment between the preferences of the compensation committee and those of the average shareholder may arise simply due to chance without any recourse to collusion or capture. We will also quantify the probability of misalignment as a function of the shareholder mix. 3.

Sub-optimal Contracting

Shareholders elect members of the board and those members choose an optimal contract for the CEO. We consider a setting where each board member acts honestly in the belief that their own horizons are what the shareholders want as well. We assume also that the CEO is indifferent in terms of own preferences and does not try to influence

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Optimal Contracts with Intra-Principal Conflicts

page 317

317

the inter-temporal weighting of her contract. In other words, the CEO accepts the contract as offered and maximizes her own payoffs through her choice of actions. In this setting, we examine different possibilities for why the contract given to the CEO may differ from the horizon preferences of an “average” shareholder. The main assumption that we make is that the probability that a board member will reflect a short or long-horizon perspective is the same as the mix of shareholders. Specifically, we assume that if β of the shareholders have a short-horizon, then the probability that the director will also have a short-horizon is also β. In other words, we assume that the directors are chosen randomly from a pool that reflects the shareholders time-preferences.9 Given this assumption, we develop the parameters of the contract that will be offered to the CEO. For simplicity, we assume that the contract will be drawn up and offered by a three-member compensation committee. Given the assumption that the directors are drawn from a pool reflecting the shareholders’ interests, we have the following table of probabilities: Composition of Compensation Committee Short-term Shareholders Long-Term Shareholders Probability

0 3 (1 − β)3

1 2 3β(1 − β)2

2 1 3β 2 (1 − β)

3 0 β3

In particular, we examine the probability that the board will be dominated by minority shareholder preferences. Since the table is symmetric in β and (1 − β), we examine the setting where β ≥ 12 . In this case, the probability that the board of three will contain two or more long-term shareholders is given as:  ΠL

9

1 β≥ 2



  1  = P rob LT S ≥ 2 β ≥ = 3β(1 − β)2 + β 3 (16) 2 

This is an assumption that may not hold in firms where some founding member has undue influence or in firms dominated by a family. However, this assumption is a reasonable starting point for deriving board characteristics.

November 4, 2020

318

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Behavioral Finance: A Novel Approach

Differentiating and solving the first-order condition tddxevΠL ββ = 0, we obtain 3 − 12β + 12β 2 = 0Lraβ =

1 2

This result implies that the probability of having a board that does not reflect shareholder priorities is more likely if the shareholders are more evenly divided. This result may appear counter-intuitive but it is driven by a simple fact that the board contains three people and the likelihood that two or more of them will belong to the minority group increases if there are more directors in the pool of choice who reflect the preferences of a minority group. The main focus of the analysis here is to provide a plausible mechanism as to why the CEO’s contract may not accurately align with the shareholders’ preferences. This mechanism is simply an outcome of random choice and does not require the board members to be manipulated or otherwise suborned by the CEO in the pursuit of private interest. Rather, selection of a small board from a pool that accurately reflects the shareholders could lead to a lack of alignment between the horizon of the board and the “average” shareholder. In turn, this leads to a contract that may motivate short-termism from the CEO. 3.1.

Extensions

The main point of our analysis is to examine situations where the compensation contract offered to the CEO that maximizes the payoffs to some shareholders results in the shifting of cash flows to the first period at the expense of the second period. The analysis is based on a key assumption that purchasing shareholders at the end of the first period overestimates future cash flows due to the actions of the CEO. As discussed in the introduction, it is clear that such an assumption violates rational expectations. A more natural question is whether earnings management is plausible in a multi-period model where investors gradually adjust their expectations as they learn about the future cash flows. Such a model would involve multiple rounds of trading and Bayesian belief updates that eventually result in a revelation of the CEO’s private information. While such an extension would make the model more realistic, it would make

page 318

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Optimal Contracts with Intra-Principal Conflicts

b3964-ch13

page 319

319

it much harder to derive the optimal CEO contract and derive the simple result that the Pareto efficient compensation contract reflects the shareholder mix. The second natural extension of our model is to allow some ex-post settling up for the shareholders who purchased at t = 1. This would highlight a point made in (Dontoh, Ronen, and Sarath 2013) that such ex-post transfers punish the long-term shareholders who hold the shares at t = 2 and reward the short-term shareholders who purchased at t = 1. It does not affect the profits of the short-term shareholders who sold at t = 1 and profited from the manipulation of the stock price. In other words, while an ex-post settling up would add realism, it will not resolve the problem that the shareholders who plan to sell at t = 1 would prefer increasing the short-term value of the firm irrespective of any long-term damage that may be associated with such behavior. A third possibility is a role for riskaversion. Although the first-period price involves a risk-adjustment, it is independent of a. A more complete model will allow for conjectures about the level of a that then drive the first-period price. We have not taken this further step since we will have to assume the lack of rational expectations about the underlying distribution of a in order to drive a wedge between the expected second-period price and the actual second-period price. While allowing the beliefs about the CEO’s choice of a to be part of the investors’ expectations would be a further step in descriptive accuracy, it will eventually lead back to equilibrium solutions that are qualitatively similar to the assumptions in Equation (1).

4.

Conclusion

The use of equity-based compensation and its effect on opportunistic managerial conduct has been criticized in many studies, both in accounting and finance. Yet, this practice remains widespread and has increased over time. We argue that short-term share price compensation is a natural consequence of the fact that shareholders have heterogeneous time-preferences. CEO’s ought to respond optimally to their contract — such behavior is efficient. When shareholders do not agree on what they would like the CEO to do, a mix of shortterm and long-term equity-based incentives is necessary to align the

November 4, 2020

320

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Behavioral Finance: A Novel Approach

CEO’s decisions to conform to the interest of an “average” shareholder. So the existence of incentives to boost short-term share price cannot be construed as sub-optimal but as a natural consequence of the presence of short-horizon shareholders who intend to liquidate at the first profitable opportunity. When shareholders have conflicting time-preferences regarding the stock price, they will also disagree about the contract that is optimal for the CEO and every contract will result in a loss to at least one type of shareholder. We show, by solving a model with multiple shareholder types, that the only Pareto-Efficient contract is one where the weights on the short and long-term incentives exactly reflect the mix of short and long-term shareholders. This result suggests that CEO contracts which either over-emphasize or under-emphasize short-term share price are sub-optimal. Our findings are that there could be overall benefits through an emphasis on quarterly results and that the movement to ban quarterly reporting may be misguided. The last point of our analysis is to suggest that the misalignment between the CEO’s horizon as reflected in the equity-based contract and the shareholder preferences could arise simply due to the fact that compensation contracts are designed by a small committee elected from a pool of potential directors. We show that even if the pool of potential choices exactly reflects the composition of shareholders, the preferences of a small committee would typically fail to reflect the average preferences of the pool simply due to the laws of basic probability. However, we must note that in practice, the use of compensation consultants may be sufficient to mitigate or overcome the simple arithmetic that underlies our analysis. More generally, it must be borne in mind that theoretical models are essentially normative. We have built our model around a central behavioral assumption — that it is possible to inflate short-term share price at the expense of long-term value in a way that misleads potential investors and have shown how this could make earnings management actually benefit shareholders in the aggregate. In deriving this result we have simplified or abstracted away from many issues that would be important in a more practical context. We have discussed how we might incorporate some of these issues as extensions of our chapter. Others, such as the use of external compensation consultants, lie outside the scope of our analysis.

page 320

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Optimal Contracts with Intra-Principal Conflicts

b3964-ch13

page 321

321

Acknowledgments We thank Stan Baiman, Sasson Bar-Yosef, Suresh Govndraj, Feng Gao, Iftekhar Hasan and participants in a seminar at Tel Aviv University for their comments and suggestions. References Arrow, K.J. (1951), Social Choice and Individual Values, New York: Wiley. Second edition (1963), Yale University Press. Third edition (2012), Yale University Press. Bebchuk, L. A., Fried, J. M., and Walker, D. I. (2002). Managerial power and rent extraction in the design of executive compensation. University of Pennsylvania Law Review, 69, 751–846. Bernard, V. L. and Thomas, J. K. (1989). Post-earnings-announcement drift: Delayed price response or risk premium? Journal of Accounting Research, 27, 1–36, http://www.jstor.org/stable/2491062. Bolton, P., Scheinkman, R., and Xiong, W. (2006). Executive compensation and short-termist behavior in speculative markets. Review of Economic Studies, 73, 577–610. Anantharaman, D., Gong, G., and Fang, V. W. (2014). Inside debt and the design of corporate debt contracts. Management Science, 60, 1083– 1350. Alex, D., Ronen, J., and Sarath, B. (2013). Financial Statements Insurance. Abacus, 49:, 269–307,http://dx.doi.org/10.1111/abac.12012. Gjesdal, F. (1982). Information and Incentives: The agency information problem. The Review of Economic Studies, 49(3), 373–390, http://dx. doi.org/10.2307/2297362. Graham, J. R., Harvey, C. R., and Rajgopal, S. (2006). Value destruction and financial reporting decisions. Financial Analysts Journal, 62(6), 27–39, http://dx.doi.org/10.2469/faj.v62.n6.4351. Grossman, S. J. and Hart, O. D. (1983). An Analysis of the principal-agent problem. Econometrica, 51(1), 7–45, http://dx.doi.org/10.2307/19122 46. Gunny, K. A. (2010). The Relation Between Earnings Management Using Real Activities Manipulation and Future Performance: Evidence from Meeting Earnings Benchmarks, Contemporary Accounting Research, 27(3), 855–888,https://EconPapers.repec.org/RePEc:wly:coacre:v:27: y:2010:i:3:p:855-888. Holmstrom, B. (1979). Moral hazard and observability. The Bell Journal of Economics, 10(1), 74–91,http://www.jstor.org/stable/3003320.

November 4, 2020

322

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch13

Behavioral Finance: A Novel Approach

Jensen, M. C. and Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3. Jones V. Harris Associates L. P. Citation: 559 U.S. 335 Court: US Supreme Court Date: March 30, 2010 https://supreme.justia.com/cases/federal /us/559/335/. Sloan, R. G. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71, 289–315. Stein, J. C. (1989). Efficient capital markets, inefficient firms: A model of myopic corporate behavior. Quarterly Journal of Economics, 104. Wilson, R. (1968). The theory of syndicates. Econometrica, 36(1), 119–132, http://www.jstor.org/stable/1909607.

page 322

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Part III

New Directions for Pensions and Retirement Decisions

323

page 323

b2530   International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6

01-Sep-16 11:03:06 AM

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0014

Chapter 14

Preferences for Annuities in Israel and Their Psychological Determinants Omer Selivansky∗ , David Leiser† and Avia Spivak‡

Abstract This chapter examines retirement savings withdrawal preferences in Israel and the psychological and demographic factors that affect them, by questioning recipients in a setting where real-life pension decisions are made. Unlike findings from other countries, most of the participants preferred annuity over a lump sum withdrawal. Lower confidence in the system and higher self-confidence were significantly associated with lump sum withdrawals preference. We suggest that the preference for an annuity stems from a default setting: pension savings in Israel have long been defined benefit (DB), that is, a fixed pension which to the recipient is more similar to an annuitized defined contribution (DC) system than to a lump sum withdrawal. Keywords: Annuity puzzle, economic psychology, household finance

1.

Introduction

Nobody knows the day of their death. While being an existential problem, it is also a practical one: suppose that you have an account of US§1 million at the age of 65. How to divide it? If you take your life expectancy (say 20 years at this age) and allocate each year ∗

Israel Democracy Institute, Israel; [email protected]. Ben-Gurion University of the Negev, Israel; [email protected]. ‡ Ben-Gurion University of the Negev, Israel; [email protected]. †

325

page 325

November 4, 2020

326

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

US§50,000, there is a good chance that you will be destitute later, if you live longer than 85 years. Conversely, if you take a longer horizon, there is a good chance that you will not have consumed as much as you could, leaving an unintended bequest. This conundrum is solved by an annuity: a payment system where you receive regular installment as long as you live. This is what happens with Social Security pension and traditional pension funds (defined benefit or DB). In this fund, the insured and his employer pay into the pension throughout the working period, and the accumulated funds plus interest serve as the source for the annuities. But a person with a relatively big account can also convert her money into a stream of payments that are conditioned on her being alive. She gives the insurance company a big one time premium, and it will give her a monthly allowance as long as she lives. (“Immediate life annuity.”) It thus insures her against the risk of living too long. Roughly speaking, the insurance company distributes the money left among the survivors — the annuity is a mechanism of bequest sharing. To put numbers on this abstract notion, immediate life annuities have a relatively high payout rate. For example, a 65-year-old male currently can buy a US$100,000 annuity with a payout rate of 6.72% per year, according to New York Life insurance company. (Regular interest rates are less than half of it). This rate has an actuarial component: the annuity allows the insureds to consume all their money without undesired bequest. However, very few people buy immediate annuities. In the US in 2012, sales totaled US$7.7, as compared with US$5.3 trillion investment in individual retirement accounts (IRAs) and workplace retirement plans. This is the well-known “annuity puzzle”. The importance of this phenomenon increased with the recent changes in the global pension market that transferred responsibility for pension management to the citizens, and increased flexibility regarding pension withdrawal options. Theoretical economic analyses have demonstrated that annuitization of the entirety of a pension (Yaari, 1965) or its majority (Brown, 2007) maximizes personal welfare. Economic studies took into consideration factors that may detract from the advantages of an annuity, such as scarcity of annuity products that satisfy the need for liquidity (Davidoff et al., 2003) and for protection from inflation (Brown et al., 2007), the desire to leave an inheritance (Lockwood, 2012),

page 326

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Preferences for Annuities in Israel

b3964-ch14

page 327

327

selection bias causing individuals with low life expectancy to abstain from purchasing annuities (Warner and Pleeter, 2001), the role of the family unit as a substitute for annuity products (Kotlikoff and Spivak, 1979), and discouragement of the private market from offering annuity products by stipends received from social security (Dushi and Webb, 2004). While these studies suggest that the purchase of a whole life nominal annuity with the entirety of an individual’s retirement savings is not necessarily an option always to be preferred, Brown (2007) shows that, even when adjustments are made for the a factors, purchase of an annuity with most or all the retirement savings remains the optimum choice for most individuals. Nonetheless, demand for annuity products is meagre. In Britain, for example, only 12% of the public chose to invest their pension savings in annuity, and by contrast 54% of the public cashing their whole pension savings as a lump sum (Reichman, 2018); moreover, only 6% of households had purchased an annuity product on the open market (Inkmann et al., 2011). In the United States, only 21% of defined contribution pension funds offered annuity products, and among the members of these funds, only 6% of individuals with the option to invest part of their savings in annuity products chose to do so (Schaus, 2005). Men aged 63–67 in the United States were found to have invested a mean of only 5% of their savings in annuity products, with a median investment of 0% (Poterba, 2003). Abstention from purchasing annuities was also observed in a natural experiment in the US Army. When the Army implemented a manpower reduction plan, it gave those discharged the choice of withdrawing compensation as either a lump sum or an annuity. Although the value of the annuity was 20% greater than that of the lump sum, most servicepeople preferred to receive their compensation as a lump sum (Warner Pleeter, 2001). Owing to the difficulty of explaining this phenomenon under rational assumptions, the annuity puzzle is explained by recourse by theories and methodologies borrowed from behavioral economics (Brown, 2007). A number of scholars have examined the effects of various personal characteristics, among them psychological and sociodemographic factors (Agnew et al., 2008; Shu et al., 2013), as well as varying perceptions relating specifically to pension withdrawal and the pension system in general (Brown, 2013). This chapter seeks to

November 4, 2020

328

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

examine the annuity puzzle in Israel, and to identify the personal psychological and demographical factors that are relevant to it.

2.

Explanations of the Annuity Puzzle: Behavioral Psychological Factors

Decision-making with regard to retirement savings in general, and withdrawal of those savings in particular, is highly vulnerable to psychological biases resulting from the fact that such decisions are characterized by a great degree of uncertainty. Such decisions require knowledge and understanding of a range of complex concepts, generally do not permit a learning process to take place, and have ultimate consequences that become manifest only after the fact. The default option effect (Beshears et al., 2009) Butler and Teppa (2007) found that in Switzerland for funds where the default option was withdrawal of the pension as an annuity, an average of 73% of members chose full annuitization, while 17% preferred partial annuitization. Conversely, at funds where a lump sum was the default option, an average of only 10% of savings was annuitized. Similar results were found in the public sector in the US: the majority of employees remained with the default option set by the government, regardless of whether this behavior was financially justified.Israel is a clear case of a country that encourages annuitization. In 2008, Amendment 3 to the Control of Financial Services Law became valid with the principal aim of encouraging retirement savings. Annuitization up to a ceiling was made the only option, incentivized by substantial taxation on lump-sum withdrawal. In addition to influences exerted by the manner in which retirement funds withdrawal is structured, withdrawal preferences are influenced by personal demographic and psychological factors (Agnew et al., 2008; Shu et al., 2013). Confidence in the financial system is particularly critical because familiarity with the various elements of the system is a complex matter that requires a degree of expertise (Giddens, 2013). The financially uneducated person tends to form a naive index of confidence in the various financial systems reliant principally on emotions that these systems arouse and reflecting personal values and norms (Hyde et al., 2007). Gardner and Wadsworth (2004) showed that lack of

page 328

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Preferences for Annuities in Israel

b3964-ch14

page 329

329

confidence in companies managing retirement savings and in financial institutions in general was one of the most influential factors informing participants’ preference to abstain from purchasing annuity products. Brown and Mitchell (2007) similarly demonstrated that individuals preferred exchanging stipends for a one-time lump sum payment if they were inclined to assess that those stipends were likely to be reduced. Financial literacy and numerical ability are associated with financial behavior in a number of areas, including the field of pensions (Lusardi, 2008). Studies of the relationship between an individual’s degree of financial literacy and numerical ability, and a preference for annuitization over a lump sum have arrived at contradictory conclusions. Some studies found that the greater a participants’ level of financial literacy, the more they preferred an annuity over a lump sum (Cappelletti et al., 2013; Brown and Mitchell, 2007; Schreiber and Weber, 2016). Other studies, however, observed the opposite relation, which may be explained by a preference on the part of individuals with low levels of financial literacy to abstain from reinvesting their savings and to their lack of confidence in their ability to preserve their savings (Agnew and Szykman, 2011; Mottola and Utkus, 2007). Self-confidence may influence decision-making with regard to the manner in which a person withdraws their pension. Payment of a lump sum means that the administration of the savings is transferred from the pension fund to the individual. For the individual to assent to such a development, they must believe themselves capable of appropriately managing these important funds. A high level of self-confidence may contribute to active retirement planning, and thus investment of resources in learning and understanding the pension problem and deep familiarity with impending choices and their consequences (Neukam and Hershey, 2003). Conversely, excessive confidence may result in imprudent decisions and unprofessional management of retirement savings (Kahneman et al., 2005). Lown and Robb (2011) demonstrated that individuals who assessed their ability to manage their savings as “very good” tended prefer annuitization less than those who described their ability to manage their savings as merely “good”, or worse. Individual discounting rate It is likely that individuals who tend to place greater weight on the present than on the future will tend

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

330

to prefer a lump sum, as this option permits utilization of savings in the immediate term, whereas the benefits of annuitization pertain to the long term, and indeed, Weber and Schreiber (2016) found that those with high present discount rates preferred a lump sum to annuitization. This chapter seeks to examine withdrawal preferences in Israel, in real life situation — the sample is of real people who come to the pension fund’s offices and are about to decide on their pension plans. This is the first study on Israeli preferences. We wish to assess the demographic and psychological factors that influence demand for annuity products in the Israeli market. We investigate the following demographic factors: age, number of children, income level and education. The psychological factors to be examined are: time preference, level of financial literacy and numerical literacy, level of self-confidence and level of confidence in the financial system. 3.

Survey

3.1.

Method

The variables in the study were examined by means of an anonymous questionnaire exploring withdrawal preferences and factors informing them. 3.1.1.

Participants

The study was conducted from December 2013 to February 2014, in the reception area of Menora-Mivtachim, the largest pension product management company in Israel. The experiment took place in the reception area of the company during 31 days, sampling a total of 297 fund members, who had come to conduct business concerning their retirement savings. The ages ranged 42–67 (median age = 59). The questionnaire was prepared in three versions, Hebrew, Russian and Arabic. A total of 203 participants completed the questionnaire in Hebrew, 76 completed the Russian version and 17 the Arabic one. To avoid interfering gender effects, and in recognition of the differences between the processes undergone by men and women in withdrawing retirement savings, only men were sampled. As a means of incentivizing fund members to participate in the study, invest

page 330

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Preferences for Annuities in Israel

b3964-ch14

page 331

331

resources in completing the questionnaire, each participant who completed the questionnaire received a thermal cup bearing the emblem of Ben-Gurion University as a gift in appreciation of his cooperation. The average time required to complete the questionnaire was approximately 15 minutes. This study is unique in the method employed to collect data. Most prior studies of the annuity puzzle relied on widely distributed internet-based questionnaires. In the present study, data were sampled in a location where the participants were making real-world decisions concerning their retirement savings, by experimenters who interacted individually with the participants and were available to explain the content of the questionnaire to them. In studying preferences pertaining to withdrawal of retirement savings, it is important to administer a questionnaire face-to-face and, in a situation resembling that in which decisions regarding retirement savings are made. The withdrawal of accrued retirement savings is a one-time event with significant psychological ramifications. It may be feared that participants asked through an internet-based questionnaire about how they would prefer to withdraw retirement savings may fait to be emotionally and cognitively engaged, and risk giving superficial or even frivolous responses. Studies of the annuity puzzle utilizing internet-based questionnaires may therefore be flawed. To counter these problems, we conducted our study in the reception area of the participants pension funds management company, with appropriate props (table, poster). By so doing, we hope to have encourage the participants to answer our questions as an extension of their business in that locale and situation. 3.1.2.

Questionnaire

The dependent variable in our study is the preferred form of retirement fund withdrawal, viz., annuitization or a lump sum. The preferred mean of pension withdrawal was measured using a process constructed by Schreiber and Weber (2016). First, we asked how they would prefer to withdraw their retirement savings (as a stipend received once per month beginning at the time of retirement, or as a one-time lump sum). They were then requested to allocate a given sum they would wish to withdraw on a one-time, immediate payout basis (on a scale of 0%–100% with increments

November 4, 2020

332

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

of 10%) and receiving the reminder as an annuity, using the pension conversion factor commonly employed in Israel, viz., 200. This allocation was repeated for three sums to be considered as retirement funds — 800,000 ILS, 1.1 million ILS, and 1.4 million ILS (1 million ILS equal about US$300,000). To prevent an order effect, the amounts were presented to half of the participants in ascending order and to the other half in descending order. We then averaged the three percentages. And used this as a measure of the percent of savings that a participant preferred to receive as an immediate payout was obtained as an average of the results of the three tasks. Independent variables were measured as follows: Level of confidence in the system was measured using three questions. It first asked participants to indicate to what degree they believed that the Israeli pension system at present provided financial security during retirement. The second question requested them to indicate to what degree they believed that the Israeli pension system in the future would provide financial security during retirement. The third question asked them to indicate to what degree they considered the companies administering retirement funds to be dependable. Degrees of confidence were expressed on a scale of 1–5 (1 — very low, 5 — very high). The three values were then totaled, resulting a new measure in a scale of 3–15 with high internal consistency (Cronbach’s alpha of 0.8). Self-confidence with regard to personal management of retirement savings was measured based on a Lown and Robb (2011). The participants were asked to grade their ability to personally manage their retirement savings by choosing one of five statements, ranging from I have no ability whatsoever to manage my account on my own to I would be able to manage my account very well. The responses were coded on a scale of 1–5, (1 representing the lowest degree of selfconfidence). Temporal preference was measured by presenting six hypothetical questions. Each question asked participants to choose which of two tax refunds they would prefer to receive, with one option to be paid soon and the other paid later with interest. This method has been used in a number of studies and a variety of contexts to assess

page 332

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Preferences for Annuities in Israel

b3964-ch14

page 333

333

payout scheme type (Harrison et al., 2002; Meier and Sprenger, 2013; Schreiber and Weber, 2016). Level of financial literacy and numerical abilities measured using a financial literacy questionnaire based on the work of Lusardi and Mitchell (2007), composed of five questions of varying difficulty that were translated and adapted by Carmel et al., (2014). Participants were also asked the three questions of the Cognitive Reflection Test (CRT) (Frederick, 2005). Reason for visiting the office of the pension fund company was identified by having each participant indicate their reason for being there. They checked the appropriate answer out of six: withdrawal of severance package, taking out a loan, receiving a stipend, changing investment plans, obtaining general information and joining the fund. We derived a binary variable out of the answer, according to whether they were there to withdraw severance pay, as this reason is the principal option available to individuals who wish to withdraw some part of their retirement fund up front. Demographic factors collected were age, number of children, monthly salary bracket and education. The questionnaire also included an assessment of whether retirement savings were perceived in terms of desires or in terms of needs. The measure obtained from this task contributed little to explaining withdrawal preferences, and will not be discussed further. 3.2. 3.2.1.

Results Withdrawal references

Asked whether they would prefer to receive their pension investments as a fixed stipend throughout the remainder of their lives or as a one-time lump sum, 68% of participants responded that they would prefer to receive their savings in the form of a fixed life annuity. When asked to allocate various sums between an immediate payout and a fixed life annuity, participants exhibited a marked preference in favor of annuitization rather than an immediate payout. Figure 1 shows the percentage of participants who selected each of the various withdrawal options when asked to allocate savings between an immediate payout and an annuity. Participants showed a clear preference for annuitization over an immediate payout, as expressed by the fact

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

334 45%

Percentage of parƟcipants

40%

39%

35% 30% 25%

20%

20% 15% 10% 5%

14% 11% 5%

3%

3%

3%

1%

0%

0%

Percentage of savings allocated as a lump sum

Figure 1: option

Percentage of participants that selected each withdrawal

that participants allocated an average of 72% of their savings to a fixed annuity. The median participant preferred to withdraw 87% of his savings as an annuity, 78% of participants preferred to withdraw the bulk of their retirement savings as an annuity, and 39% of participants preferred to annuitize the entirety of their retirement savings. Nevertheless, a significant number of participants (14%) preferred to withdraw the entirety of their retirement savings as a lump sum. These findings are at variance with those of a range of studies performed in various parts of the world. In most studies where participants were asked to report their withdrawal preferences, an immediate payout was preferred. We also found that most participants who preferred most of their withdrawal up front wanted all of it in a lump sum. 4.

Factors Influencing Withdrawal Preferences

Factors with a bearing on pension withdrawal preferences were examined using two models. The first model examined the factors that affect the binary choice: receiving retirement savings as a lump sum vs investing them in a fixed life annuity. The model was estimated by

page 334

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Preferences for Annuities in Israel

page 335

335

logistic regression, with a preference for annuitization of retirement savings coded as 0, and a lump sum as 1. The second model was used to examine the factors influencing the average percentage of savings allocated to an immediate payout in the three tasks that had participants allocate retirement savings of varying amounts between an annuity and an immediate payout. This model was estimated by OLS. Table 1 shows the results of the estimates for the two models. The variables found to be significant for the first model are degree

Table 1:

Models for estimating pension withdrawal preferences

Estimation Method Age Salary Education level Num of children Withdrawal of severance Confidence in the system Level of self confidence Financial literacy and numeracy Temporal preference Constant Num of obs R-squared Correctly Classified

(1) Binary Choice: Annuity or Lump Sum

(2) Percentage of Savings Allocated as a Lump Sum

Logistic Regression

OLS Regressions

0.0296 (0.0247) −0.00780 (0.110) 0.159 (0.145) 0.0336 (0.110) 0.582∗ (0.310) −0.652∗∗∗ (0.200) 0.156 (0.126) 0.0295 (0.0976) −0.0278 (0.0789) 1.679 (1.858) 219 — 71%

−1.139∗∗∗ (0.394) 0.496 (1.547) −0.880 (1.921) 0.108 (1.507) 4.434 (4.570) −6.976∗∗ (2.871) 4.194∗∗ (1.788) 1.246 (1.460) −0.661 (1.039) 101.7∗∗∗ (29.80) 219 0.1 —

Note: Robust standard errors in parentheses. ∗∗∗ ,∗∗ and ∗ indicate significance at the 1%, 5% and 10% level, respectively.

November 4, 2020

336

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

of confidence in the system (with a significance level of 1%) and the reason for coming to the office of the insurance company (with a significance level of 10%). It was found that as the degree of confidence in the system decreased, as well as when participants had come with the purpose of withdrawing severance, they tended to exhibit a stronger preference in favor of receiving their savings as an immediate payout. The variables found to be significant for the second model are degree of confidence in the system (5%), level of self-confidence (5%) and age (1%). High levels of confidence in the system and selfconfidence were found to be associated with a preference for annuitization, while each additional year of age was found to correlate with a decrease of 1.14% in the proportion of retirement savings that the participant preferred to receive as an immediate payout. The predictive value of the two models is modestly high. The logistic regression model classifies 71% of cases, with 68% of participants reporting that they would prefer to annuitize their savings. The explained variance of the second model is minimal, as reflected by the low R2 value (0.1). We are left to conclude that the personal characteristics studied reflect only a small part of the factors that influence withdrawal preferences. The variables found to have a significant effect are as follows: Confidence in the system: The sole factor found to be significant in both models is the individual’s degree of confidence in the system. This result is unsurprising: it stands to reason that the greater an individual’s confidence that the system will take good care of his savings, the more willing he will be to leave his money in it. Coming to withdraw severance pay: When offered a binary choice, participants who had come to withdraw severance pay tended to exhibit a stronger preference for a lump sum over an annuity than those present for other reasons. The relation was found to be significant only in the model examining a binary choice, and only marginally (P = 0.065). This is surprising, as withdrawing severance pay is in fact a withdrawal of retirement savings as an immediate payout. It may be that the discrepancy manifested because some participants came to withdraw severance due to circumstances that compelled them to do so, contrary to their fundamental preferences. Alternately, the disparity may have resulted from the distinction

page 336

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Preferences for Annuities in Israel

b3964-ch14

page 337

337

in Israel between severance pay and provident fund contributions. The participants may have perceived the severance component — available for withdrawal as cash under specific conditions such as losing one’s job — as distinct from retirement savings per se, which are to be paid only upon retirement. Participants who came to withdraw severance may not have realized their withdrawal affects retirement funding, and saw themselves as withdrawing that part of their savings that was designated for withdrawal at the moment without impacting retirement savings. Level of self-confidence: High levels of self-confidence were found to be associated with a preference for an immediate payout. The link between high levels of self-confidence and withdrawal preferences was found to be significant in the model examining the percentage of savings that participants preferred to receive as an immediate payout. Age: As the age of participants increased, they were more likely to prefer an annuity to an immediate payout. Several variables were found to relate to withdrawal preferences, but with scant predictive value. All the models used in previous studies, which sought to identify the personal characteristics that influence withdrawal preferences, explained only a small portion of the variance, with an R2 < 0.1. Several variables not found to be significant in the present study had be found to be significant in previous studies. All told, it appears that the influence of personal characteristics and character traits on the pension withdrawal decision is limited. 5.

Discussion

This study examined pension withdrawal preferences and the factors influencing them, by running two statistical models. It was found that withdrawal preferences in Israel differ from those documented in other countries, as expressed by the tendency to prefer annuitization over receipt of a lump sum. The only factor that was significant in both models examined is the individual’s degree of confidence in the system, an indication of the great importance of public confidence and expectations concerning the pension system. Some additional factors were weakly associated with withdrawal preferences:

November 4, 2020

338

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

the reason for coming to the office of the insurance company, level of self-confidence and age. Summarizing, in keeping with the literature, this study, as its predecessors, personal characteristics contribute but little to explain withdrawal preferences (e.g., Brown et al., 2008; Brown, 2013; Schreiber and Weber, 2016). Strikingly, the overall withdrawal preference revealed in the present study is at odds with that found in previous studies.We find a clear preference for annuitization over a lump sum, in marked distinction to the minimal demand for annuity products found in many studies of the annuity puzzle. For instance, Brown et al. (2007) found that about two-thirds of their Health and Retirement Study (HRS) sample would prefer to withdraw half of their social security income in cash (see also Agnew et al., 2008; Schreiber and Weber, 2016). Yet even among participants who came to the company intending to withdraw severance, most also preferred annuitization over a lump sum. Inasmuch as a substantial proportion of our sample had come in order to withdraw severance, our sample was likely more skewed than the general population toward a preference for a lump sum to annuitization, so that the contrast with past studies is all the more surprising. We suggest a possible explanation for the discrepancy, which stems from changes in the Israeli pension system that took place about 20 years ago. The main changes were mandatory savings, and the transition from a defined benefit (DB) to a defined contribution (DC) system. DB pension is not expressed in terms of accumulated funds, but rather in accumulated rights — a certain percentage of salary to be received during retirement. A regular monthly inflow fits therefore the perspective of people in the age bracket we studied, and annuity corresponds to this. To be sure, there will always be a percentage of the population who do not trust the system, and demand their money now. This explains why the most important predictor of the preference for annuities is trust in the system. The DC pension has an opposite philosophy. It is based on the accumulation of money by the individual. Throughout the working years, the sum saved is regularly stressed. However, upon retirement it is automatically converted into a monthly payment, an annuity. In fact, there is no other realistic option. The accumulation perspective is in line with our finding that younger participants tend to prefer

page 338

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Preferences for Annuities in Israel

page 339

339

lump sum over annuity, in contrast to older participants. Accordingly, we should find an increase in the demand for lump sum payment of retirement savings in the future, as cohorts who thought in terms of DC all their lives reach retirement age.1 Supplementary data Summary statistics

Variable Preferred form of retirement fund withdrawal Dichotomic choice between annuity and a lump-sum (0 — annuity, 1 — lump sum) Percent of savings that a participant preferred to receive as a lump-sum Demographics Age Income (1 − 5)∗ Education (1 − 5)∗∗ Number of children Percent of participants who came for severance package withdrawal Psychological factors Level of confidence in the system Self-confidence measure Num of correct answers in the financial literacy questionnaire Number of correct answers in the test Temporal preference measure

Num of Obs

Mean

297

32%

297

28%

13%

34%

297 275 292 276 280

59 3.84 2.87 2.9 45%

60 4 1.27 3

6.3 1.46 3 1.36 50%

269 286 297

8.24 2.94 1.72

8 3 1

2.87 1.27 1.42

297 286

0.33 8.12

0.67 8

0 2.1

Median

Std. Error

47%

Note:∗ Income level was coded as follows: 1 = income of 0–2,000 ns, 2 = 2,001–4,000 ns, 3 = 4,001–6,000 ns, 4 = 6,001–8,000 ns, 5 = 8,001–10,000 ns, 6 = more than 10,000 ns.∗∗ Education level was coded as follows: 1 = elementary school, 2 = secondary school, 3 = diploma, 4 = BA, 5 = MA or PhD.

1 Taxation rules promulgated in 2008 prefer annuities to lump sum withdrawal in other vehicles of long-term saving, other than pension funds. The persons questioned here belong by our choice to a pension fund, Mivtachim-Menora, hence those changes in rules hardly have any impact on them.

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

340

b3964-ch14

Behavioral Finance: A Novel Approach

Questionnaire Demographics Age: Martial status: Single / Married / Divorced / Widowed Number of children: Please circle: What is your monthly gross salary? (0–2000 nis / 2001–4000 nis / 4001–6000 nis / 6001–8000 nis / 8001– 10000 nis / More / Refuses to answer) What is the highest educational qualification that you have completed (Elementary / high school / Diploma / BA / Graduate)? How you would estimate your ability to manage your pension savings all by yourself? (I cannot manage my account myself / I would be having a hard time to do so / I’m able to do so mediocrely / I’m able to do it well / I’m able to manage my account very well) Why did you come today to Menora-Mivtachim? (Withdrawal of severance package / Receiving a loan / Receiving a pension / Investment track changes / For general information / enrollment freelance or employee) If you came to withdraw your severance package, what portion of the total available amount you would like to withdraw? (The entire amount / part of the amount / not sure / I Came for a different purpose / I do not want to answer) The reason why my pension savings are managed by MenoraMivtachim is: (Due to the choice of the organization where I work or worked in the past / by my own choice / I do not know)

page 340

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Preferences for Annuities in Israel

page 341

341

To what extent do you believe that the Israeli pension system currently provides financial security at retirement? (Not at all / Slightly / Moderately / Much / Very much) To what extent do you believe that the Israeli pension system in the future will provide financial security at retirement? (Not at all / Slightly / Moderately / Much / Very much) Please specify from one to five how much you perceive the companies managing the pension as fair? (1 — very unfair, 5 — very fair)

Main Questionnaire Part 1 Suppose you are given a choice between two options to withdraw your retirement funds: 1 — Withdrawal of a lifetime monthly guaranteed annuity starting at retirement. 2 — Withdrawal of the pension savings as a lump sum which will be transferred to your bank account at once. Which of these two options would you prefer (please check)? [] Option 1: an annuity [] Option 2: a lump sum • Assume that during your working years you accumulated pension savings of 1.4 million ns. Suppose you could decide on the balance between the amount that will be allocated as a fixed monthly life annuity, and the size of the amount to be received as a lump sum, how would you divide the amount between the two options? Please select the preferred option:

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

342

The LumpSum Withdrawal 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

None 140,000 280,000 420,000 560,000 700,000 840,000 980,000 1,112,000 1,260,000 1,400,000

Monthly Annuity 7,000 6,300 5,600 4,900 4,200 3,500 2,800 2,100 1,400 700 None

• Assume that during your working years you accumulated pension savings of 1.1 million ns. Suppose you could decide on the balance between the amount that will be allocated as a fixed monthly life annuity, and the size of the amount to be received as a lump sum, how would you divide the amount between the two options? Please select the preferred option: The LumpSum Withdrawal 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

None 110,000 220,000 330,000 440,000 550,000 660,000 770,000 880,000 990,000 1,100,00

Monthly Annuity 5,500 4,950 4,400 3,850 3,300 2,750 2,200 1,650 1,100 550 None

• Assume that during your working years you accumulated pension savings of 800,000 ns. Suppose you could decide on the balance between the amount that will be allocated as a fixed monthly life annuity, and the size of the amount to be received as a lump sum, how would you divide the amount between the two options? Please select the preferred option:

page 342

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

page 343

Preferences for Annuities in Israel

The LumpSum Withdrawal 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

343

Monthly Annuity

None 80,000 160,000 240,000 320,000 400,000 480,000 560,000 640,000 720,000 800,000

4,000 3,600 3,200 2,800 2,400 2,000 1,600 1,200 800 400 None

Part 2 Assume that you could withdraw some of your pension savings at retirement. Please mark from 1 to 6 how much it is likely that you will use some of the money for the following purposes (1 indicates that it is very unlikely, and 6 indicates that it is very likely):

How Likely is it that you will Spend your Pension Savings on that: Very Unlikely I would like to upgrade something (e.g., to replace my car with a new one, renovate my house or buy new home appliances) Purchase a long-term care insurance and/or a long-term health insurance Do something special for myself to celebrate the retirement (e.g., to travel abroad) Invest the money or repay old debts Allow myself to keep enjoying the things I used to do before retirement (e.g., to eat out, buy good food for Shabbat, spoil the grandchildren or go out drinking with friends) Open savings plans for my children or grandchildren

Very Likely

1

2

3

4

5

6

1

2

3

4

5

6

1

2

3

4

5

6

1 1

2 2

3 3

4 4

5 5

6 6

1

2

3

4

5

6

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

344

Part 3 The purpose of the next part is to explore your personal preferences, so there are no right or “wrong” answers. Please answer the following questions. A. Suppose you deserve an immediate tax refund of 10,000 ns. Alternatively, you have the option to receive a higher refund of 10,300 ns which will be paid in 10 months. Circle the type of return you would prefer. 1. An immediate refund of 10,000 ns. 2. A deferred refund of 10,300 ns in 10 months. B. Assume the options are: 1. An immediate refund of 10,000 ns. 2. A deferred refund of 11,000 ns in 10 months. C. Assume the options are: 1. An immediate refund of 10,000 ns. 2. A deferred refund of 13,000 ns in 10 months. D. Assume now that you can get the refund either in 18 months or in 28 months. The refund that will be received in 18 months would be of 10,000 ns. Alternatively, you have the option to get a refund of 10,300 ns that will be received in 28 months. Circle the type of refund you would prefer. 1. A deferred refund of 10,000 ns in 18 months. 2. A deferred refund of 10,300 ns in 28 months. E. Assume the options are: 1. A deferred refund of 10,000 ns in 18 months. 2. A deferred refund of 11,000 ns in 28 months. F. Assume the options are: 1. A deferred refund of 10,000 ns in 18 months. 2. A deferred refund of 13,000 ns in 28 months.

page 344

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Preferences for Annuities in Israel

page 345

345

Part 4 Please circle the correct answer to each of the following 6 questions, or mark “do not know” Q1: Assume a friend inherits 10,000 today and his sibling inherits 10,000 3 years from now. Who is richer because of the inheritance? 1. 2. 3. 4.

My friend. His sibling. They are equally rich. Do not know.

Q2: Suppose you had 100 in a savings account and the interest rate was 4% per year. After 10 years, how much do you think you would have in the account if you left the money to grow? 1. 2. 3. 4.

Less than 140 ns. Exactly 140 ns. More than 140 ns. Do not know.

Q3: Which of the following statements is correct? 1. Once one invests in a mutual fund, one cannot withdraw the money in the first year. 2. Mutual funds can invest in several assets, for example invest in both stocks and bonds. 3. Mutual funds pay a guaranteed rate of return which depends on their past performance. 4. Do not know. Q4: Buying a company stock usually provides a safer return than a stock mutual fund. True or false: 1. True. 2. False. 3. Do not know. Q5: The main reason for the purchase of insurance is: 1. Provide high investment returns for myself. 2. Improve your standard of living by filing false claims.

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

346

3. Protect you against loss occurred recently. 4. Prevent you from absorbing catastrophic loss. 5. Do not know. Part 5 (CRT Test) Please solve the following three riddles: A bottle and a cork cost 1.10 ns in total. The bottle costs 1.00 ns more than the cork. How much does the cork cost? If it takes 5 machines 5 minutes to make 5 widgets, how long would it take 100 machines to make 100 widgets? Minutes In a lake, there is a patch of lily pads. Every day, the patch doubles in size. If it takes 48 days for the patch to cover the entire lake, how long would it take for the patch to cover half of the lake? Days Thank you for your cooperation!

References Amendment 3 to the Provident Funds Law (2008). www.knesset.gov.il/ privatelaw/data/17/3/291 3 3.rtf Agnew, J. R. and Szykman, L. (2010). Annuities, financial literacy and information overload. Pension Research Council WP, 33. Agnew, J. R., Anderson, L. R., Gerlach, J. R., and Szykman, L. R. (2008). Who chooses annuities? An experimental investigation of the role of gender, framing, and defaults. The American Economic Review, 98 (2), 418–422. Beshears, J., Choi, J. J., Laibson, D., and Madrian, B. C. (2009). The importance of default options for retirement saving outcomes: Evidence from the United States. In Jeffrey R. Brown et al. (Eds.), Social Security Policy in a Changing Environment. University of Chicago Press, pp. 167–195. Brown, J. R. (2007). Rational and behavioral perspectives on the role of annuities in retirement planning (No. w13537). National Bureau of Economic Research.

page 346

November 4, 2020

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

Preferences for Annuities in Israel

b3964-ch14

page 347

347

Brown, J. R., Casey, M., and Mitchell, O. S. (2007). Who values the social security annuity. Evidence from the Health and Retirement Study. NBER Retirement Research Center Paper No. NB 07-02. Brown, J. R., Kapteyn, A., Luttmer, E. F., and Mitchell, O. S. (2013). Decision complexity as a barrier to annuitization. NBER Working Paper Series No. 19168. Butler, M. and Teppa, F. (2007). The choice between an annuity and a lump sum: Results from Swiss pension funds. Journal of Public Economics, 91 (10), 1944–1966. Cappelletti, G., Guazzarotti, G., and Tommasino, P. (2013). What determines annuity demand at retirement. The Geneva Papers on Risk and Insurance-Issues and Practice, 38 (4), 777–802. Carmel, E., Carmel, D., Leiser, D., and Spivak, A. (2015). Facing a biased adviser while choosing a retirement plan: The impact of financial literacy and fair disclosure. Journal of Consumer Affairs, 49 (3), 576–595. Davidoff, T., Brown, J. R., and Diamond, P. A. (2003). Annuities and individual welfare (No. w9714). National Bureau of Economic Research. Dushi, I., and Webb, A. (2004). Household annuitization decisions: Simulations and empirical analyses. Journal of Pension Economics and Finance, 3 (2), 109–143. Frederick, S. (2005). Cognitive reflection and decision making. The Journal of Economic Perspectives, 19 (4), 25–42. Gardner, J. Wadsworth, M. (2004). Who would buy an annuity? An empirical investigation. An Empirical Investigation Watson Wyatt Technical Paper, (2004-4). Giddens, A. (2013). The Consequences of Modernity. John Wiley & Sons. Harrison, G. W., Lau, M. I., and Williams, M. B. (2002). Estimating individual discount rates in Denmark: A field experiment. The American Economic Review, 92 (5), 1606–1617. Hyde, M., Dixon, J., and Drover, G. (2007). Assessing the capacity of pension institutions to build and sustain trust: A multidimensional conceptual framework. Journal of Social Policy, 36 (3), 457. Inkmann, J., Lopes, P., and Michaelides, A. (2011). How deep is the annuity market participation puzzle? Review of Financial Studies, 24 (1), 279–319. Kahneman, D., Odean, T., and Barber, B. (2005). Privatized pensions: An irrational choice. Global Agenda Magazine. Kotlikoff, L. J. and Spivak, A. (1979). The family as an incomplete annuities market. Journal of Political Economy, 89, 372–391. Lockwood, L. M. (2012). Bequest motives and the annuity puzzle. Review of Economic Dynamics, 15 (2), 226–243.

November 4, 2020

348

10:15

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch14

Behavioral Finance: A Novel Approach

Lown, J. M. and Robb, D. K. (2011). Attitudes toward immediate annuities: Overcoming the annuity puzzle. Journal of Consumer Education, 28, 44–60. Lusardi, A. (2008). Financial literacy: n essential tool for informed consumer choice? (No. w14084). National Bureau of Economic Research. Lusardi, A. and Mitchell, O. (2007). Financial literacy and retirement preparedness: Evidence and implications for financial education. Business Economics, 42 (1), 35–44. Meier, S. and Sprenger, C. D. (2013). Discounting financial literacy: Time preferences and participation in financial education programs. Journal of Economic Behavior & Organization, 95, 159–174. Mottola, G. R. and Utkus, S. P. (2007). Lump sum or annuity? An analysis of choice in DB pension payouts. Vanguard Center for Retirement Research, 30 (89), 2. Neukam, K. A. and Hershey, D. A. (2003). Financial inhibition, financial activation, and saving for retirement. Financial Services Review, 12 (1), 19. Poterba, J. M. (2003). Employer stock and 401 (k) plans. The American Economic Review, 93 (2), 398–404. Schaus, S. (2005). Annuities make a comeback. Journal of Pension Benefits: Issues in Administration, 12 (4), 34–38. Schreiber, P. and Weber, M. (2016). Time inconsistent preferences and the annuitization decision. Journal of Economic Behavior & Organization, 129, 37–55. Shu, S. B., Zeithammer, R., and Payne, J. (2013). Consumer preferences for annuities: Beyond NPV. Working Paper. Warner, J. T. and Pleeter, S. (2001). The personal discount rate: Evidence from military downsizing programs. American Economic Review, 33–53. Yaari, M. E. (1965). Uncertain lifetime, life insurance, and the theory of the consumer. The Review of Economic Studies, 32(2), 137–150.

page 348

November 4, 2020

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

c 2021 World Scientific Publishing Company  https://doi.org/10.1142/9789811229251 0015

Chapter 15

Smokers’ Life Expectancy and Annuitization Decisions Abigail Hurwitz∗ and Orly Sade†

Abstract We examine subjective life expectancy perceptions of Israeli smokers by investigating the results of an online survey of a representative sample of 963 Israeli residents aged 50–70 years. Our results suggest that smokers are overoptimistic regarding their subjective life expectancy, a fact that is expected to influence the decision-making process in general, and financial decisions in particular. Indeed, our results are consistent with the results obtained by Hurwitz and Sade (2019) with regard to annuitization decisions. Keywords: Self-assessed life expectancy, smoking, financial decisions, time preferences

1.

Introduction

Past studies suggest that people who are seeking to make important decisions such as how much to save and consume over their lifespan ∗ The Hebrew University of Jerusalem, Israel; The College of Management, Academic Studies, Israel; The Wharton School of the University of Pennsylvania, USA; [email protected]. † The Hebrew University of Jerusalem, Israel; [email protected].

349

page 349

November 4, 2020

350

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Behavioral Finance: A Novel Approach

will devote some thought to how long they will actually live. In addition, people are expected to take into account personal characteristics that could affect survival outcomes (such as health and their parents’ longevity). Hence, it is not surprising that it is of interest for researchers to learn more about individuals’ estimations of expected longevity. Hamermesh (1985) document that individuals’ estimations of their subjective survival probabilities are coherent, useful for prediction and conformed to actuarial tables. However, Hamermesh (1985) also reports that people tend to overvalue their parents’ survival information when considering their own futures. Subsequent researchers have shown that people do exhibit systematic biases when predicting their own life expectancies. In particular, Elder (2013) documents that individuals appear to overstate mortality rates at relatively young ages and understate mortality rates at older ages. Wu et al. (2013) conclude that although peoples’ subjective life expectancies are close to life table data, they differ by age cohort. Subjective survival probabilities are closely related to financial decisions. Bloom et al. (2006) suggest that those who believe they would live longer than average also tend to save more. Hurd et al. (2004) find that low subjective probabilities of survival predicts earlier retirement in those individuals. Using Dutch data, Teppa and Lafourcade (2013) also determine a positive relation between self-life expectancy and demand for annuities. While cigarette smoking is the leading preventable cause of death in Israel and in the Western world, it is still a widely used recreational activity (Wang, 2014). According to a 2014 report from the Israeli Ministry of Health, 19.8% of Israeli adults aged 21 years and over were “current cigarette smokers”. This ratio was higher for men (27.3%) than for women (12.6%).1 The Israeli Ministry of Health2 provides information about the mortality differences between smokers and non-smokers in selected countries (including Israel). It is evident that there is a gap of more than nine years in life

1 2

Ministry of Health report on smoking in Israel 2014, published in May 2015. Ministry of Health report on smoking in Israel 2013, published in May 2014.

page 350

November 4, 2020

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Smokers’ Life Expectancy and Annuitization Decisions

page 351

351

expectancy between the two groups.3 Given its unique characteristics, it is of interest to learn more about perceived life expectancy of smokers. Given the above, the aim of this chapter is to study life expectancy perceptions of smokers. Our investigation is conducted in Israel. To investigate life expectancy perceptions of individuals in Israel, we obtained the results of an online survey of a representative sample of 963 Israeli residents aged 50–70 years. The results support the conjecture that smokers experience self-illusions regarding health and life expectancy. We provide several robustness tests to support these findings, and in particular we show that smoking (controlling for a current medical condition) does not significantly affect health perceptions. Our chapter is closely related and contributes to the study by Hurwitz and Sade (2019) that investigates Israeli smokers’ annuitization decisions. Their study determines that although a smoker’s life expectancy is nine years less than a nonsmoker’s, insurance pricing mechanisms still offer smokers the same annuity as nonsmokers (all else equal). Nevertheless, smokers still do not prefer the lump sum option. This chapter provides an explanation that is consistent with Hurwitz and Sade’s (2019) empirical findings and conjectures, namely, that smokers are optimistic about their own life expectancy. In what follows, we first elaborate on our survey structure. Second, we present the data and report the empirical results with regard to life expectancy and health perception. Finally, given our results we hypothesis the relationship to annuity purchase and relate it to recent findings and last we conclude.

2.

A Survey on Life Expectancy and Long-Term Savings Decisions

To investigate life expectancy perceptions of Israelis, we explored responses to an online survey conducted on 1,000 representative residents aged 50–70 years in March 2015. We focus on these ages as we aim to relate our findings to withdrawal decisions that are made at 3

For further elaboration of this literature see Hurwitz and Sade (2019).

November 4, 2020

352

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Behavioral Finance: A Novel Approach

retirement. Given the complexity of these choices, we survey an agerelated representative sample.4 After missing values were removed, the final sample was 963 respondents.5 The mean age of the respondents was 58.17 years (Median of 58, Standard Deviation of 5.45)6 and 40.6% were male; hence our survey had a higher concentration of females than that of the general population. The marital status of the respondents was varied: 73.4% were married, 16.9% divorced and 3.3% widowed.7 On average our sample was more educated than the general population of Israel: 0.2% of participants had less than a high school diploma, 22% had a high school diploma and 76.5% had higher education (including college, graduate school and other higher education, such as rabbinical studies).8 Of the respondents, 88% believed that their health was good or very good, 17.4% reported that they currently smoked and 31.5% reported that they had smoked in the past (for this survey, a person was defined as a “smoker” if she or he smoked more than three cigarettes per day).9 Women made up 58% of the smokers, 53%

4 For further discussion about the choice of age representative sample see Hurwitz et al. (2020). 5 The survey was conducted by Sarid: Research Services and Training, using an online panel of registered potential participants with a wide distribution age who registered voluntarily. In exchange for participation, respondents earned points that could be converted into money or vouchers. 6 This is lower than the average age of the retirees in the insurance corporation data, because we deliberately addressed our life expectancy questions to people prior to retirement. 7 According to data from the Israeli CBS for 2012, the family status of Israeli citizens over the age of 50 was 68% married (which is also close to the proportion of married people in the data we received from the insurance company), 13% divorced and 15% widowed. Please note that the CBS data also include citizens over age 70. 8 According to the CBS social questionnaire, only 26.4% of the population had an academic education, while 3% had studied in a rabbinical school (Yeshiva). The fact that the survey participants had, on average, more education than the general population can be explained by the choice to conduct the survey online for respondents 50 years or older. 9 According to a Ministry of Health report on smoking in Israel 2014, published May 2015, 27.3% of men and 12.6% of women in the adult population smoked.

page 352

November 4, 2020

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Smokers’ Life Expectancy and Annuitization Decisions

page 353

353

of the former smokers and 62.4% of nonsmokers. Of the smokers, 69.1% had a college degree, compared to 78% for former smokers and non-smokers (the difference is significant with a 0.03 significance level), implying that smokers were less educated than former smokers and non-smokers. With regard to income, 64.3% of smokers reported that they earned more than the average income in the general population, compared to 65.7% and 55.7% of former smokers and nonsmokers, respectively (these differences are not statistically significant). 2.1.

Survey Structure

Our survey asked respondents questions related to life expectancy, demographic details, long-term saving decisions and self-assessments of health. The median time participants took to complete the survey was 6.5 min. Because the focus of our research was to obtain life expectancy perceptions, we asked several questions that are well accepted in the academic literature on financial economics. The specific questions and explanations for our choices follow: (1) In your opinion, what is the current life expectancy in Israel (respond for your own gender)? This question was not intended to assess subjective life expectancy. Rather, it was designed to explore peoples’ perception of life expectancy of others (in the general population). We did not ask about conditional life expectancy at the specific age of the respondent, since we were interested in asking a clear and relatively simple question that would not confuse the respondents. (2) Do you expect your own life expectancy to be lower, identical to, or higher than the average life expectancy you have mentioned above? Comparing the respondent’s life expectancy to the life expectancy of the general population was used, for example, by Beshears et al. (2014), who asked respondents how much longer they expected to live relative to others their age.

November 4, 2020

354

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Behavioral Finance: A Novel Approach

2.2.

Survey Results

2.2.1.

Life expectancy

Respondents’ answers to the question about life expectancy in the general population for their own gender appear in Table 1. Respondents gave reasonable answers (81.1 years for men and 82.54 years for women) comparing to the actual life expectancy at birth in Israel according to the CBS in 2013 (80.3 years for men and 83.9 years for women) and to life expectancy at age 58 years10 (82.5 years for men and 85.1 years for women). Regarding the expectation themselves living longer, shorter or the same as the general population, 34.2% of the respondents believed they would live longer than average, whereas 52.7% thought that they would live the same length as the average person and 12.9% suspected they would have a shorter life expectancy. The median response from our respondents was the belief that they would live as long as the average person in the general population. Our results are very similar to the results reported by Beshears et al. (2014), who conducted two surveys of US respondents. Our finding that only 12.9% of respondents thought that they had a shorter than average life expectancy is not necessarily an indication of overoptimism in this sample. Beshears et al. (2014) noted that a proportion of respondents projecting a relatively long life could be a result of the sample being more educated than average in the population, since longevity is positively correlated with education (Cutler et al., 2011). 2.2.2.

Smoking status and health condition

We asked respondents to assess their health condition by rating their own health on a scale of 1 (very good health) to 4 (poor health). Eight people refused to answer and were omitted from this analysis. Smokers, former smokers and nonsmokers gave similar assessments of their health (Median of 2 in all categories). Using the nonparametric test for the difference between the medians of the different groups (smoker vs. former smoker; smoker vs. non-smoker),

10

The average age in our survey.

page 354

November 4, 2020

Projected Life Expectancy In the general population In the general population (male) In the general population (female) Self Self (male) Self (female)

N

Mean (SD)

Min

Max

Mdn

% Same as as Meana

% Longer than Meanb

963 391 572 963 391 572

81.96 (4.45) 81.11 (3.98) 82.54 (4.67)

40 65 40 1 1 1

100 100 96 3 3 3

82 81.1 82.54 2 2 2

52.7 47.5 56.5

34.2 40.3 29.7

b3964-ch15

Note: Projected life expectancy in the general population refers to the respondents’ perception of life expectancy in the general population measured. Self refers to respondents’ beliefs in their own life expectancy being longer (value 3), equal to (value 2) or shorter (value 1) in comparison to the life expectancy they estimated for the general population. N = number of participants (total and separately for males and females). Actual life expectancy at birth in Israel, according to the Central Bureau of Statistics, was 80.3 years for men and 83.9 years for women in 2013. a Percentage of respondents who thought they would live as long as the average person in the general population. b Percentage of respondents who thought they would live longer than the average person in the general population.

Behavioral Finance: A Novel Approach – 9in x 6in

Life Expectancy in ears

10:16

Projected life expectancy

Smokers’ Life Expectancy and Annuitization Decisions

Table 1:

355 page 355

November 4, 2020

356

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Behavioral Finance: A Novel Approach

we fail to reject either of the null hypotheses that the median values are equal. With regard to self-assessed health, we investigate the relation between smoking and health perception, controlling for both socioeconomic variables (such as age, number of children, gender, marital status, education and income) and related health conditions (such as smoking in the present or the past, participation in extreme sports and the age of parents at death). Smoking is not significantly related to health perception or self-assessed life expectancy, as reported in Table 2. Results hold for different specifications; with all other variables remaining equal, the average smoker in our survey did not perceive her- or himself as any less healthy than a non-smoker. 2.2.3.

Smoking status and life expectancy

Next, we investigate smokers’ perceptions regarding their own life expectancies. Smokers and non-smokers gave similar ratings of their own life expectancy compared to the average life expectancy of the population. Because smoking is negatively correlated with longevity, then on average, life expectancy, conditional on the individual being a smoker, is indeed expected to be lower than life expectancy of the general population. Further, life expectancy conditional on the individual being a non-smoker, is expected to be higher than life expectancy of the general population. Thus, we would expect on average, if smokers are rational and all else remains equal, that they will anticipate a shorter life expectancy than people in the general population11 Of the smokers in our data set, 57% believed they would live as long as the average person, 22% believed they would live longer than 11

We can illustrate this with a numerical example. Assume that in our economy there are only two agents. They are identical with one difference: One is a smoker and the other is a non-smoker. Statistics have indicated that in this economy in the past, smokers lived 75 years and non-smokers 85 years. Assuming an equal number of agents of each type (in order to simplify), the average age of this population is 80. If our two agents are rational in respect to their life expectancy predictions, the smoker should expect to live below the average and the nonsmoker should expect to live above the average. Hence, the smoker is considered optimistic if he or she expects to live as long as or above the average.

page 356

November 4, 2020

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

page 357

Smokers’ Life Expectancy and Annuitization Decisions

357

Table 2: Life and health perception in the survey. Dependent variables: Health perception and life perception. Health Perception

Variable Age Children Male Marital status Single Married Divorced Widowed Smoker Former smoker Education level High school Postsecondary Unknown Extreme sport activities Age of father at death Age of mother at death High income

Life Perception

(1)

(2)

(3)

(4)

Ordered Probit Coefficient

Ordered Logit Coefficient

Ordered Probit Coefficient

Ordered Logit Coefficient

0.0188∗∗ (0.00814) −0.0210 (0.0273) 0.0933 (0.0813)

0.0336∗∗ (0.0141) −0.0386 (0.0480) 0.159 (0.141)

0.00169 (0.00793) −0.00692 (0.0267) 0.227∗∗∗ (0.0792)

0.00277 (0.0135) −0.0218 (0.0452) 0.385∗∗∗ (0.136)

0.892∗ (0.458) 0.798∗ (0.427) 0.671 (0.434) 0.971∗∗ (0.475) 0.162 (0.109) 0.0198 (0.0899)

1.555∗∗ (0.771) 1.399∗ (0.715) 1.178 (0.728) 1.710∗∗ (0.795) 0.276 (0.189) 0.0354 (0.157)

−0.0534 (0.405) 0.176 (0.372) −0.152 (0.380) 0.326 (0.424) −0.130 (0.297) −0.132 (0.0867)

−0.261 (0.717) 0.207 (0.655) −0.365 (0.668) 0.377 (0.737) −0.228 (0.500) −0.220 (0.149)

−0.609 (0.804) −0.825 (0.802) −0.778 (0.887) −0.0700 (0.199) 0.00200 (0.00129) 0.00130 (0.00116) −0.291∗∗∗ (0.0834)

−1.144 (2.071) −1.542 (2.069) −1.387 (2.178) −0.0604 (0.348) 0.00400∗ (0.00225) 0.00251 (0.00202) −0.497∗∗∗ (0.145)

−6.276 (93.14) −6.071 (93.14) −5.941 (93.14) −0.0699 (0.190) 0.000248 (0.00124) −0.000735 (0.00112) 0.0342 (0.0810)

−16.34 (639.4) −15.98 (639.4) −15.66 (639.4) −0.161 (0.323) 0.000679 (0.00213) −0.00103 (0.00193) 0.0599 (0.139) (Continued )

November 4, 2020

10:16

358

Behavioral Finance: A Novel Approach – 9in x 6in

(Continued )

Health Perception

Variable

(2)

(3)

(4)

Ordered Probit Coefficient

Ordered Logit Coefficient

Ordered Probit Coefficient

Ordered Logit Coefficient

−0.504∗∗∗ (0.0656) −0.169

−0.881∗∗∗ (0.115) −0.300

(0.153)

(0.260)

Health perception and smoking (interaction)

Constant cut 2 Constant cut 3 Observations Pseudo R2

Life Perception

(1)

Health perception

Constant cut 1

page 358

Behavioral Finance: A Novel Approach Table 2:

Physical problems

b3964-ch15

1.379∗∗∗ (0.0899) 0.965 (1.038) 3.077∗∗∗ (1.044) 4.252∗∗∗ (1.045) 955 0.1714

2.434∗∗∗ (0.172) 1.599 (2.356) 5.271∗∗ (2.367) 7.602∗∗∗ (2.369) 955 0.1680

−8.079 (93.14) −6.385 (93.14)

955 0.0781

−19.56 (639.4) −16.68 (639.4)

955 0.0785

Note: Standard errors in parentheses. Main explanatory variables are gender, marital status, smoking, education, income and parent’s age at death (N = 955). ∗∗∗ p < 0.01, ∗∗ p < 0.05 and ∗ p < 0.1.

the average and only 21% believed they would have a shorter life than the average person in the population. In summary, this means that 79% of smokers believed they would have the same or even a longer lifespan than the general population (Figure 1). This proportion is significantly higher than the group of smokers who believed that they would have a shorter life expectancy compared to the general population (21%). We used the percentage of smokers who indicated a life expectancy equal to or greater than average (79%) as an optimism proxy for smokers. Further, we would expect non-smokers to estimate their life expectancy as greater than the average. Hence, as a conservative optimism proxy estimate, we used the percentage that indicates a

November 4, 2020

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Smokers’ Life Expectancy and Annuitization Decisions

page 359

359

Figure 1: Proportion of smokers who rated their own life expectancy as above, the same as or below the life expectancy they judged as average for the population

life expectancy greater than the average (39%). The optimism proxy for smokers was significantly higher than for non-smokers. To reinforce these results, we calculate the median projection of self-assessed life expectancy in the two groups. The median projection of life expectancy for both smokers and nonsmokers is exactly 2 (the difference from the value 2 is not statistically significant using a one-sample Wilcoxon median test for both populations), meaning that at least half of the participants believed they would live as long as or longer than the average person in the population. To test for robustness, and to address the concern that our sample was more educated than the population, we took a subsample from the full survey population in which the proportions of education levels were comparable to the education proportions in the Israeli population, as published in the CBS data. After taking 30 subsamples of 100 observations each from the population, we tested for the median projections of smokers and nonsmokers from the 30 different subsamples. The median of the 30 median projections was 2, for both smokers and nonsmokers. We did this again for other subsamples of the 300 examinees, and we find that for each subsample, the projection of smokers was not statistically different from 2,

November 4, 2020

360

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Behavioral Finance: A Novel Approach

meaning that smokers did in fact believe that in general, their life expectancy would be similar to average, indicating overoptimism. It could also be claimed that conditional on living to or surpassing an age of 50–70 years, smokers should be optimistic regarding their life expectancy. Nevertheless, as mentioned before, mortality from smoking is higher at older ages (as reflected in life insurance policy pricing).12 This result is consistent with data from the 2006 landmark US Health and Retirement Study (Khwaja et al., 2006),13 and other studies showing that heavy smokers were overoptimistic regarding their life expectancy (Schoenbaum, 1997) and more receptive to supportive information about smoking (denial of the smoking– cancer link) than to non-supportive information (affirmation of the smoking–cancer link; Brock and Balloun, 1967). In addition, the results are consistent with previous literature regarding smokers’ health perceptions. For instance, only 29% of American smokers reported that their own risk of heart attack was higher than average, and only 40% thought that their risk of cancer was higher than average (Ayanian and Clearly, 1999). The results are also consistent with a study showing optimism among people at risk for Huntington’s Disease (Oster et al., 2013), as well as a study showing that older people tend to be more optimistic regarding their life expectancies(Heimer et al., 2019). As an additional robustness test, shown in Table 2, we investigate the impact of self-assessed health and smoking on self-assessed life expectancy, controlling for both socioeconomic variables and related health conditions. The results of the robustness test are consistent with our claim that smokers do not update their expectations and

12

The gap in prices between smokers and non-smokers in Israeli insurance policies increases with age. 13 Our findings are consistent with and contribute to the medical and psychological academic literature. McKenna et al. (1993) argued that smokers consider themselves less likely than non-smokers to be hurt by smoking-related diseases. Masiero et al. (2015) suggested that this bias could result from the illusion of control and from the need to preserve good self-esteem. Slovic (1998) argued that optimism bias is likely to be greater when people think that signs of vulnerability will appear early, and people think that an absence of present symptoms means they are safe from future risks.

page 360

November 4, 2020

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Smokers’ Life Expectancy and Annuitization Decisions

page 361

361

indicate that neither smoking nor past smoking influences perceived life expectancy. 3.

Survey Findings and Annuity Decisions

Given our findings that smokers experience self-illusions regarding health and life expectancy, the next natural question to ask is whether this plays a significant role in their financial choices, especially those closely linked to longevity estimations. One important decision that should be affected by subjective life expectancy is an annuitization decision. Indeed, as our survey results suggests, Hurwitz and Sade (2019) documented annuitization results that are consistent with smokers’ self-illusion with regard to their life expectancy. By using data from an Israeli insurance corporation (and controlling for several related variables) they find that, smokers, as compared to non-smokers, do not prefer the lump sum option, despite the life expectancy gap and the fact that in Israel, annuity prices are the same regardless of health and smoking status. 4.

Conclusion

Even though the literature finds a close relationship between smoking and medical conditions, smokers do not perceive themselves as having a shorter lifespan, meaning that smokers experience self-illusions regarding their own life expectancy. This observation led us to a further investigation on the self-life expectancy perception of smokers and its effect on their financial decisions. We obtained the results of an online survey of 1,000 Israeli residents, aged 50–70 years old, that asked questions related to life expectancy estimations, demographics, long-term savings choices and a self-assessment of health. The survey results suggested that smokers believe they will live the average lifespan. To conclude, our results suggest that smokers might be overoptimistic regarding their subjective life expectancy, a fact that is expected to influence the decision-making process in general, and financial decisions in particular. We leave for further research the investigation of different nudges that may be able to mitigate this tendency.

November 4, 2020

10:16

362

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Behavioral Finance: A Novel Approach

Acknowledgments We have benefited from comments from Doron Avramov, Shlomo Benartzi, Cameron Ellis, Luigi Guiso, Eyal Lahav, Olivia S. Mitchell, Francois Pannequin, Oded Sarig, Kirill Shakhnov, Meir Statman, Federica Teppa, Steve Zeldes, the editor and participants at the 2016 Financial Literacy Research and Practice conference at Ben-Gurion University; the 2016 IRMC, Jerusalem; the 2016 Incentives and Behavior Change Conference at Tel Aviv University; the 2016 ESA International Meeting at the Hebrew University; the 2016 Research in Behavioral Finance Conference (RBFC), Amsterdam; the 2016 CEAR Conference — Risk Literacy — Methods and Applications, Naples; the 10th Financial International Forum, Paris, 2017; the Second Israel Behavioral Finance Conference, Subjective Probability, Utility and Decision Making Conference (SPUDM) 2017, Technicon, August 2017; The Annual Netspar International Pension Workshop, Leiden, January 2018; the Seminar on Aging, Retirement and Pensions: Trends, Challenges and Policy, Ashkelon, March 2018; the CEAR/MRIC Behavioral Insurance Workshop 2018, Munich, December 2018; and seminar participants at the EIEF Rome; Hebrew University; IDC; Tel Aviv University; University of Vienna; The Wharton School; The College of Management Academic Studies, and MLA. This received financial support from the National Insurance Institute of Israel, the Kruger Center at the Hebrew University (Sade), the Israeli Ministries of Finance and Science (Hurwitz) and the Research Authority of the College of Management Academic Studies, Rishon Lezion, Israel (Hurwitz). Sade thanks the Stern School of Business at New York University for support and hospitality. Hurwitz wishes to thank the Bogen fellowship for supporting her research.

References Ayanian, J. Z. and Cleary, P. D. (1999). Perceived risks of heart disease and cancer among cigarette smokers. JAMA, 281 (11), 1019–1021. Beshears, J., Choi, J. J., Laibson, D., Madrian, B. C., and Zeldes, S. P. (2014). What makes annuitization more appealing? J Public Economics, 116, 2–16.

page 362

November 4, 2020

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Smokers’ Life Expectancy and Annuitization Decisions

page 363

363

Bloom, D. E., Canning, D., Moore, M., and Song, Y. (2006). The effect of subjective survival probabilities on retirement and wealth in the United States (No. w12688). National Bureau of Economic Research. Brock, T. C., and Balloun, J. L. (1967). Behavioral receptivity to dissonant information. Journal of Personality and Social Psychology, 6 (4), 413–428. Cutler, D. M., Lange, F., Meara, E., Richards-Shubik, S., and Ruhm, C. J. (2011). Rising educational gradients in mortality: he role of behavioral risk factors. Journal of Health Economics, 30 (6), 1174–1187. Elder, T. E. (2013). The predictive validity of subjective mortality expectations: Evidence from the health and retirement study. Demography, 50 (2), 569–589. Hamermesh, D. S. (1985). Expectations, life expectancy, and economic behavior. The Quarterly Journal of Economics, 100 (2), 389–408. Heimer, R. Z., Myrseth, K. O. R., and Schoenle, R. S. (2019). YOLO: Mortality beliefs and household finance puzzles. The Journal of Finance, 74 (6), 2957–2996. Hurd, M. D., Smith, J. P., and Zissimopoulos, J. M. (2004). The effects of subjective survival on retirement and social security claiming. Journal of Applied Econometrics, 19 (6), 761–775. Hurwitz, A. and Sade, O. (2019). An investigation of time preferences, life expectancy, and annuity versus lump sum choices: Can smoking harm long-term saving decisions? Journal of Economic Behavior & Organization, forthcoming. Hurwitz A., Sade O., and Winter, E. (2020). Unintended consequences of minimum annuity laws: An experimental study Journal of Economic Behavior & Organization, 169, 208–222. Khwaja, A., Sloan, F., and Salm, M. (2006). Evidence on preferences and subjective beliefs of risk takers: he case of smokers. International Journal of Industrial Organization, 24 (4), 667–682. Masiero, M., Lucchiari, C., and Pravettoni, G. (2015). Personal fable: Optimistic bias in cigarette smokers. International Journal of High Risk Behaviors and Addiction, 4 (1), e20939. McKenna, F. P., Warburton, D. M., and Winwood, M. (1993). Exploring the limits of optimism: he case of smokers’ decision making. British Journal of Psychology, 84 (3), 389–394. Oster, E., Shoulson, I., and Dorsey, E. (2013). Limited life expectancy, human capital and health investments. American Economic Review, 103 (5), 1977–2002. Schoenbaum, M. (1997). Do smokers understand the mortality effects of smoking? Evidence from the Health and Retirement Survey. American Journal of Public Health, 87 (5), 755–759.

November 4, 2020

364

10:16

Behavioral Finance: A Novel Approach – 9in x 6in

b3964-ch15

Behavioral Finance: A Novel Approach

Slovic, P. (1998). Do adolescent smokers know the risks? Duke Law Journal, 47, 1133–1141. Teppa, F. and Lafourcade, P. (2013). Can longevity risk alleviate the annuitization puzzle? Empirical evidence from survey data. Working Paper. Available at: http://www.aeaweb.org/aea/2014conference/ program/retrieve.php. Wang, Y. (2014). Dynamic implications of subjective expectations: Evidence from adult smokers. American Economic Journal, 6 (1), 1–37. Wu, S., Stevens, R., and Thorp, S. (2013). Die young or live long: Modeling subjective survival probabilities. ARC Centre of Excellence in Population Aging Research Working Paper, 19.

page 364