Advances in Longitudinal Data Methods in Applied Economic Research: 2020 International Conference on Applied Economics (ICOAE) (Springer Proceedings in Business and Economics) [1st ed. 2021] 303063969X, 9783030639693

Citation preview

Springer Proceedings in Business and Economics

Nicholas Tsounis Aspasia Vlachvei  Editors

Advances in Longitudinal Data Methods in Applied Economic Research 2020 International Conference on Applied Economics (ICOAE)

Springer Proceedings in Business and Economics

More information about this series at http://www.springer.com/series/11960

Nicholas Tsounis • Aspasia Vlachvei Editors

Advances in Longitudinal Data Methods in Applied Economic Research 2020 International Conference on Applied Economics (ICOAE)

Editors Nicholas Tsounis Laboratory of Applied Economics, Department of Economics University of Western Macedonia Kastoria, Greece

Aspasia Vlachvei Laboratory of Applied Economics, Department of Economics University of Western Macedonia Kastoria, Greece

ISSN 2198-7246 ISSN 2198-7254 (electronic) Springer Proceedings in Business and Economics ISBN 978-3-030-63969-3 ISBN 978-3-030-63970-9 (eBook) https://doi.org/10.1007/978-3-030-63970-9 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

This year conference is co-organised by the Hellenic Open University (HOU) and the Department of Economics of the University of Western Macedonia, Greece. Unfortunately, due to the coronavirus pandemic the conference has not taken place in Heraklion, Crete, where it would have been hosted by the Hellenic Open University at Archanes after the kind invitation by Prof. George Agiomirgianakis who is also co-chair of the conference, but it is a virtual conference. The aim of the conference is to bring together economists from different fields of Applied Economic Research in order to share methods and ideas. The topics covered include: • • • • • • • • •

Applied Macroeconomics Applied International Economics Applied Microeconomics including Industrial Organisations Applied work on International Trade Theory including European Integration Applied Financial Economics Applied Agricultural Economics Applied Labour and Demographic Economics Applied Health Economics Applied Education Economic

All papers presented in ICOAE 2020 and published in the conference proceedings were peer reviewed by anonymous referees. In total, 54 works were submitted from 13 countries while 40 papers were accepted for presentation and publication in the conference proceedings. The acceptance rate for ICOAE 2020 was 74%. The full text articles will be published online by Springer in the series “Springer Proceedings in Business and Economics” The organisers of ICOAE 2020 would like to thank: • The Scientific Committee of the conference for their help and their important support for carrying out the tremendous work load organising and synchronising

v

vi

•

•

• •

Preface

the peer-reviewing process of the submitted papers in a very specific short period of time. The anonymous reviewers for accepting to referee the submitted conference papers and submit their reviews on time for the finalisation of the conference programme. The keynote speaker, Dr. Giovanni Cerulli from the Research Institute on Sustainable Economic Growth, National Research Council of Italy, for accepting to present his work on the Covid-19 outbreak. The organising committee for its help for the success of the conference. Dr. Eirini Arvanitaki, Mr. Gerassimos Bertsatos, Mr. Lazaros Markopoulos, and Mr. Stelios Angelis for secretarial and technical support. Kastoria, Greece

Nicholas Tsounis

Kastoria, Greece

Aspasia Vlachvei

Contents

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Milan Christian de Wet From Clubs to Communities. From Tourists to International Friends. Crisis Legacy in Music Organizations with Revenue Management and Relationship Marketing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Angela Besana and Annamaria Esposito

1

31

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real Options: Amazon’s Acquisition of Whole Food . . . . . . . . . . . . . . . . . . ˇ Andrejs Cirjevskis

43

Raising Rivals’ Costs When the Downstream Firms Compete in Stackelberg Fashion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jacek Prokop and Adam Karbowski

57

Comparing Five Generational Cohorts on Their Sustainable Food Consumption Patterns: Recommendations for Improvement Through Marketing Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Irene Kamenidou, Spyridon Mamalis, Ifigeneia Mylona, and Evangelia Zoi Bara

69

The Effect of Budgetary Policies on the Economy Activity in Algeria: A Markov Switching Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Touitou Mohammed

81

Structure of Bond Pension Funds During Decreasing Yield Curves . . . . . . . Mário Papík

95

Extracting Common Factors from Liquidity Measures with Principal Component Analysis on the Polish Stock Market . . . . . . . . . . . 109 Joanna Olbrys and Elzbieta Majewska

vii

viii

Contents

The Mechanism of Political Budget Cycles in Greece . . . . . . . . . . . . . . . . . . . . . . . . 123 George Petrakos, Konstantinos Rontos, Chara Vavoura, and Ioannis Vavouras Examination of Business Interest in Level of Complexity of Facial Biometric Technology Implementation in Slovakia . . . . . . . . . . . . . . . . . . . . . . . . . . 135 Michal Budinský and Janka Táborecká-Petroviˇcová Innovation and Sales Growth Among Heterogeneous Albanian Firms: A Quantile Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 Blendi Gerdoçi and Sidita Dibra Quantitative Analysis of Inequalities at ICT Sector in Visegrad Countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 Tatiana Corejova, Roman Chinoracky, and Alexandra Valicova Does Government Spending Cause Investment?: A Panel Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Nihal Bayraktar An Exploratory Study of Fans’ Motivation in Albanian Football Championship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 Julian Bundo and Mirdaim Axhami Evaluation of Knowledge in Accounting of Regional Economic University Students . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 Ivana Koštuˇríková and Markéta Šeligová Corporate Governance Disclosure in Slovak Banks . . . . . . . . . . . . . . . . . . . . . . . . . . 211 Janka Grofˇcíková, Katarína Izáková, and Dagmar Škvareninová Implementation of Critical Reflection Analysis in Teaching and Learning Focused on Developing Critical Thinking Skills . . . . . . . . . . . . . 233 Lenka Theodoulides and Gabriela Nafoussi (Kormancová) Comparison of Methods of Poverty Rates Measurement . . . . . . . . . . . . . . . . . . . . 249 Anna Saczewska-Piotrowska ˛ The Role of Strategic Agility and Economic Environment’s Friendliness-Hostility in Explaining Success of Polish SMEs . . . . . . . . . . . . . . . 267 Tomasz Sikora and Ewa Baranowska-Prokop Macroeconomic Determinants of NPLs Using an Extended Sample and Dominance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 George Sfakianakis, George M. Agiomirgianakis, and George Manolas Aspects of Financial Accounting and Managerial Accounting Outputs in Connection with the Decision-Making Processes of Accounting Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 Markéta Šeligová

Contents

ix

Movies Performance: Empirical Evidence from Italy . . . . . . . . . . . . . . . . . . . . . . . 311 Anna Maria Bagnasco Patterns of Knowledge Creation in European Regions: An Analysis by the Phases of the EU-Enlargements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325 Thomas Baumert Influence of Economic Sanctions: Empirical Evidence for Iran and Russia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 Anton Filipenko, Olena Bazhenova, Roman Stakanov, and Ihor Chornodid Corporate Governance and Its Association with Audit Opinion: The Case of Greece. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357 Georgia Boskou, Maria Tsipouridou, and Charalambos Spathis CO2 Emissions, Energy Consumption, Economic Growth, Trade, and Urbanization in Greece . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 Pavlos Stamatiou, Chaido Dritsaki, and Dimitrios Niklis Does the Time-Driven ABC Method Apply in a Construction Company . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 N. Kartalis, Ath Patsios, I. Velentzas, G. Broni, G. Charitoudi, G. Panoy, and G. Kiriakoylis Economic Crisis Predictors Revisited in Preparation for the COVID-19 Aftermath . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409 Demosthenes Georgopoulos, Theodore Papadogonas, and George Sfakianakis Should Market Makers Hedge with Realised or Implied Volatility? . . . . . . . 421 Alexis Levendis, Pierre Venter, and Eben Maré Stress Testing Option Sensitivities in a Stochastic Market . . . . . . . . . . . . . . . . . . 431 Alexis Levendis, Pierre Venter, and Eben Maré Firm Performances and the Onset of Shocks in India . . . . . . . . . . . . . . . . . . . . . . . 445 Elangovan Avinash Filter or No Filter? An Instagram View on Modern Visual Culture . . . . . . . 459 Aikaterini Stavrianea, Evangelia Besleme, and Irene (Eirini) Kamenidou The Neoclassical Approach for Measuring Total Factor Productivity: The Case of the Greek Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 469 Thomas Siskou and Nicholas Tsounis Consumers’ Motives for Visiting Social Media Brand Pages and Social Media Advertisements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493 E. Iliopoulou and A. Vlachvei

x

Contents

Factors Affecting e-Marketing Adoption and Implementation in Food Firms: An Empirical Investigation of Greek Food and Beverage Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 Ourania Notta and Afroditi Kitta An Application of Differential Equations on Anthropogenic Climate Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527 Gerassimos Bertsatos, Soultana Moustakli, Zacharoula Kalogiratou, and Theodoros Monovasilis Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach Milan Christian de Wet

Abstract Identifying optimal models to forecast economic cycles has been a point of great consideration in literate. A key point of debate in the literature is whether linear or non-linear models perform best at forecasting economic cycles. The literature largely forces on the forecasting of business cycles, and very limited work has been done on financial cycle forecasting. Given the proven destructiveness of financial cycles, the ability to accurately forecast future financial cycle movements in an economy could aid policymakers in managing such cycles. This article evaluates the forecasting performance of both the non-linear Markov RegimeSwitching Autoregressive methodology and Smooth Transition Autoregressive methodology relative to the benchmark ARIMA model in forecasting the aggregate South African financial cycle over different time horizons. A fixed window rolling forecast approach is followed, whereby the performance of forecasting the aggregate South African financial cycle 3-steps forward, 6-steps forward, 12-steps forward, 18-steps forward and 24-steps forward is evaluated. The findings indicate that the linear ARIMA model outperforms the non-linear MSMV-AR and LSTAR models at forecasting short periods ahead such as 3–6 months ahead. However, both the MSMV-AR and LSTAR models outperform the ARIMA model, given a longer time horizon such as 12–24 months. Hence, to forecast the aggregate South African financial cycle 3–6 months ahead policymakers should use an ARIMA. However, the MSMV-AR and LSTAR models should be used to forecast the aggregate South African financial cycle 12–24 months ahead. Keywords Non-linear forecasting · Financial cycles · Markov switching · STAR · Regime shifts

M. C. deWet () University of Johannesburg, Johannesburg, South Africa e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_1

1

2

M. C. de Wet

1 Introduction Financial crises have plagued the economies for the past four decades, of which the 2007/2008 financial crisis was the most severe. The financial crisis of 2007 triggered the most severe global economic contraction since the great depression. The financial crisis of 2007 had a negative economic impact on almost all advanced economies and the majority of emerging economies in the world, with only a few exceptions. Furthermore, the financial crisis in Japan during the 1980s and 1990s had severe implications for the Japanese economy. Emerging markets, in particular, have been plagued by disruptive financial crises, such as the Mexican financial crisis that began in 1994, the Asian crisis that began in 1997 and the Argentinian crisis that began in 2001 to name a few. All these examples proved to be a result of an unsustainable financial build-up of some sort. This signals the need for policymakers to improve their understanding of financial conditions and to obtain a means to improve their ability to manage cyclical fluctuations and the effect of such fluctuations. For the most part, before 2007, policymakers neglected the role played by financial factors and the potentially disruptive impact fluctuations in these factors might have on the real economy (Borio, 2014; Strohsal, Proano, & Wolters, 2019). A possible reason why policymakers neglected financial conditions is that policymakers largely believed that financial conditions are driven by real economic conditions, and not the other way around (Borio, Drehmann, & Xia, 2018). Therefore, the need to understand the state of financial conditions in an economy seemed less important. Over the past four decades, however, financial crises around the world have proven otherwise. The 2007/2008 financial crisis caught the attention of policymakers, resulting in the realization that the old school of thought, which places all the attention on the real economy, is not completely accurate and effective. The 2007/2008 financial crisis provided clear evidence that financial factors of an economy could be disconnected from the real economy, and extremely disruptive when in disequilibrium. Therefore, financial factors could not merely be monitored and managed through simply monitoring and managing real economic conditions. Hence, specific attention needs to be paid to the cyclicality and cyclical state of financial variables in an economy to fully gauge the financial state of an economy and thereby employing effective management in this regard. After the 2007 financial crisis, policymakers need to re-evaluate how they consider financial factors and how they should go about monitoring and managing financial factors in such a way that extensive financial build-ups and financial disequilibrium are mitigated or avoided. This is to avoid potential future damage that could be caused by such build-ups and disequilibria. The ability to accurately forecast financial cycle movements could aid in this regard. The value of information that timely and accurately indicates the future development of economic cycles is vast. The importance of such information has led to extensive research on developing means to forecast future moves in aggregate economic cycles. Therefore, the forecasting of economic cycles has received a

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

3

significant amount of attention since the initial groundwork on cyclical research by Burns and Mitchell (1946). The main reason for this interest is the major economic benefits that lie within the ability to accurately and timely forecast future cyclical fluctuations. The ability to timely and accurately predict future turning points in economic cycles will afford policymakers a means to make more timely policy decisions relating to cyclical fluctuations, and other economic participants to better position themselves for future fluctuations. From cyclical forecasting research, a large number of forecasting models have been suggested, implemented and tested on a wide range of economic cycles, of which business cycles are by far the economic cycle subject to the most research. Major debates in the literature and practice prevail around which models perform best at forecasting economic cycles. Empirically, a lot of mixed results on this topic exists, indicating that optimality varies from cycle to cycle. For example, the debate between linear and non-linear forecasting models. Botha, Greyling, and Marais (2006) groups econometric models to forecast economic cycles into two major groups, namely linear models and non-linear models. Botha et al. (2006) states that one of the major debates in the literature regarding the forecasting of economic cycles is around the forecasting performance of linear and non-linear models as a means to forecast economic cycles. The debate thus considered whether forecasting models should allow for non-linearities amongst the relational dynamics between a financial cycle under analysis and corresponding endogenous and exogenous explanatory variables during different cyclical phases. Furthermore, within each of these broad two groups, a range of models exists, and even within each group, a lack of consensus exists on which models perform optimally at forecasting economic cycles. Varying viewpoints and empirical results make it challenging for policymakers, as well as other economic participants, such as asset managers, business managers and risk managers, to depend on a given set of forecasts, hindering optimal decisionmaking and actions from these various parties. This limitation is amplified when considering financial cycles, given that very limited research has been done to identify optimal models to forecast financial cycles. This study will contribute to the body of knowledge on economic cycles forecasting through identifying a model that performs optimally at forecasting the aggregate South African financial cycle. Henceforth, this study will be structured as follows. Firstly, a literature overview is provided in Sect. 2. Then a data description will be provided, and the methodological approach to be followed in this study will be stipulated in Sect. 3. In Sect. 4, the results will be presented, and a corresponding discussion of results will be provided. Finally, the conclusion will be provided in Sect. 5.

2 Literature Overview Despite very limited to no work done on forecasting aggregate financial cycles, the body of knowledge on research done on forecasting business cycles, as well as

4

M. C. de Wet

various financial variables, is vast. Even though not an exhaustive list, examples of empirical work done in this field are the work done by Laubscher (2019), Nyberg (2018), Wai, Kun, Ismail, and Karim (2015), Singh (2012), Baharumshah and Liew (2006), Botha et al. (2006), Teräsvirta, Van Dijk, and Medeiros (2005), Marcellino (2005), Moolman (2004), Crawford and Fratantoni (2003), Sarantis (2001), Rech (2002), Tkacz (2001) and Clements and Krolzig (1998). Given the lack of literature on financial cycle forecasting, the literature on business cycle forecasting will be utilized as an empirical base. Several researchers such as Botha et al. (2006), Teräsvirta et al. (2005), Moolman (2004), Sarantis (2001), and Clements and Krolzig (1998) consider the linear autoregressive integrated moving average (ARIMA) as a benchmark model for research on forecasting performance. The forecasting with linear econometric models typically utilizes linear co-movements of economic elements with the forecasted variable under consideration (Teräsvirta et al., 2005). Writes that linear forecasting models include econometric models such as ARIMA, classic linear multiple regression models, probit and logit regression models and vector autoregressive (VAR) models. Linear models, such as linear regression models, assume linear relational dynamics between a cyclical measure and given explanatory variables, thus not accounting for any asymmetries. As a result, linear models do not account for any cyclical asymmetries. However, economic cycles do not typically evolve in linear fission, but often exhibits asymmetrically sharp movements during cyclical downturns relative to upturns (Bouali, Nasr, & Trabelsi, 2016; Nyberg, 2018). A large amount of empirical research work exists which supports this statement (Baharumshah & Liew, 2006; Clements & Krolzig, 1998; Crawford & Fratantoni, 2003; Moolman, 2004; Sarantis, 2001; Singh, 2012; Teräsvirta et al., 2005; Wai et al., 2015). For example, it is found that variables have a much harsher reaction to cyclical contractions relative to a cyclical upturn, showing that asymmetries do exist (Balcilan, Gupta, & Miller, 2015; Botha et al., 2006; Moolman, 2004). By assuming cyclical symmetry when forecasting, forecasting accuracy might be suboptimal, leading to the need for non-linear models as a means to forecast economic cycles (Botha et al., 2006; Moolman, 2004; Nyberg, 2018). Bouali et al. (2016) stated that non-linear models used for economic cycle modelling typically allow regime changes and are therefore capable of accommodating relational asymmetries across cyclical regimes. Non-linear models employed to capture such asymmetries include Markov regime-switching (MRS) models and a range of smooth transition autoregressive (STAR) models (Teräsvirta et al., 2005). Provided the range of possible models that can be used as a means to forecast economic cycles, it is of interest to identify an optimal forecasting model and have been the focal point of several empirical studies. Within this strand of research, the forecasting performance of various models forecasting a range of different variables have been conducted. Studies range from forecasting broad-based economic conditions such as business cycles to forecasting specific variables such as oil, house prices, currencies, and equity prices. A study conducted by Clements and Krolzig (1998) is one of the first studies set out

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

5

to comprehensively identify an optimal forecasting method for business cycles. Clements and Krolzig (1998) considered the forecasting ability of the Markov Regime-Switching Autoregressive model (MS-AR), the Smooth Transition Autoregressive (STAR) model and several traditional linear models at forecasting the US business cycle. Out of the two non-linear models, Clements and Krolzig (1998) found that the MS-AR model, encompassing of three different regimes and four lagged autoregressive terms generally outperforms the SETAR model in estimating key elements of the US business cycle. They further found that both the MS and STAR models outperform linear models such as the popular autoregressive moving average (ARMA) model. This finding concurs with the findings by researchers such as Wai et al. (2015), Crawford and Fratantoni (2003) and Sarantis (2001). Nyberg (2018) compared the forecasting ability of the non-linear Markov Switching model to that of the linear Vector Autoregressive model and found that the non-linear Markov Switching model is superior at forecasting both the US business cycle and US interest rates. The study by Li et al. (2005) addressed the same, however, focusing on not only industrialized economies but also on newly industrialized economies and developing economies. They considered the performance of Markov-switching models in modelling the business cycles of the USA, Japan, Taiwan, South-Korea, Malaysia and Indonesia. Li et al. (2005) found that the MS model sufficiently captures the growth and contraction periods in the business cycles of the industrialized and developing countries’ business cycles. Contrary to this finding though, they found that the implemented MS model does not sufficiently capture the business cycle growth and contraction regimes of the newly industrialized countries that they analysed. Li et al. (2005) argue that the shift to industrialization for these countries caused structural breaks, resulting in the ineffectiveness of the MS model in identifying business cycle regime shifts. Li et al. (2005) therefore adjusted for structural breaks by dividing the business cycles into two distinct periods, a pre-industrialized period and a post-industrialized period. They found evidence that an MS two-regime two AR lag approach effectively identifies growth and contraction periods in the business cycle of the newly industrialized countries under analysis. Teräsvirta et al. (2005) compared the forecasting performance of linear autoregressive models and the non-linear STAR model at forecasting a range of macroeconomic time series. Teräsvirta et al. (2005) found evidence that the STAR model predominantly outperforms linear autoregressive methodology. However, studies by Marcellino (2005), Rech (2002) and Tkacz (2001) disagree with Teräsvirta et al. (2005), generally found no clear evidence that the non-linear models perform better than a linear autoregressive approach at forecasting economic variables. Singh (2012) conducted a study that aims to determine whether non-linearities exist in the economic growth cycles of OECD countries. Evidence of non-linearity will indicate that non-linear models might be necessary to effectively model economic growth rate cycles. Singh (2012) found evidence through employing a range of STAR family models that the hypothesis of linearity in the economic growth cycle of almost every country under analysis could be rejected, and therefore, characteristics of non-linearities do exist in the economic growth cycles

6

M. C. de Wet

of these economies. This implies that non-linear models might be more accurate in modelling economic growth cycles than linear models. Baharumshah and Liew (2006) conducted a study, aiming to determine whether the traditional AR model or Exponential Smooth Transition Autoregressive (ESTAR) model best performs in modelling and forecasting the path of East Asian currencies. In this study, they found that the non-linear parameters of each currency pair were statistically significant, providing evidence that the long-run equilibrium in all the Asian currency pairs under analysis follows a non-linear path. Furthermore, they found that the residual variance ratio of the ESTAR model applied to each currency pairs relative to the residual variance ratio of the AR model that corresponds to that the ESTAR model is below one. This indicates that for each currency pair, the ESTAR model renders a lower variance in the residual and therefore the ESTAR model proved to result in lower forecasting errors (Baharumshah & Liew, 2006). Balcilan et al. (2015) analysed the modelling and forecasting performance of the ESTAR and STAR non-linear models versus the linear AR model on US house prices. They found that given a long-time horizon, non-linear models perform better than linear models in point forecasting the underlying financial variable. Balcilan et al.’s (2015) findings on forecasting short-term price moves do however not conform to the findings above. However, Balcilan et al. (2015) found no evidence that nonlinear models perform better at forecasting house prices over a short-time horizon relative to linear models. Balcilan et al. (2015) also found little evidence that the non-linear models implemented in their study outperform the linear models when it comes to density forecasting, regardless of the time horizons. This indicates that the ability of the non-linear models to forecast the probability distribution of the underlying financial variable, relative to linear models, is limited. In comparing the modelling and forecasting performance of the two non-linear models, Balcilan et al. (2015) found that the Logistic Smooth Transition Autoregressive model (LSTAR) outperforms that of the ESTAR model. Relating to the cyclical movements of US house prices, Crawford and Fratantoni (2003) found that regime-switching models do better in depicting realized house price patterns, relative to ARIMA and GARCH family models. They argued that regime-switching models can effectively be utilized to create a cyclical framework for historic house price time-series data because these models identify the turning points, amplitude, and frequency of cyclical moves better. Yet, corresponding to the findings by Balcilan et al. (2015), Crawford and Fratantoni (2003) found no clear evidence that regime-switching models perform better in point forecasting US house prices, relative to the linear ARIMA model. Based on the contradicting findings in empirical literature considered, it is not clear that non-linear models will necessarily outperform linear models at forecasting economic variables. This is because some researchers found that non-linear models outperform linear models, and other researchers found no such evidence. It is thus not apparent that non-linear models will necessarily outperform linear models at forecasting South African economic cycles. This necessitates research that specifically focusses on South African economic cycles.

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

7

Extensive research has been done on the South African business cycle, for example, the work by Laubscher (2019), Botha et al. (2006), Boshoff (2005) and Moolman (2004). Moolman (2004) modelled the South African business cycles through a Markov regime-switching model. Botha et al. (2006) added to the study conducted by Moolman (2004), testing whether the South African business cycle is best estimated and forecasted through linear or non-linear models. Du Plessis (2004) considered the South African Business cycle and how dependent this cycle is on its own duration. The study by Botha et al. (2006) contributes to the debate of whether linear or non-linear models perform best in estimating business cycles. Moreover, by focusing on the South African business cycle, the findings of this study are of particular importance to this study. Botha et al. (2006) found that non-linear models outperform linear models in forecasting the South African business cycle. Furthermore, Botha et al. (2006) found that the ESTAR model is the most effective model out of all the non-linear models to forecast the South African business cycle. The findings by Moolman (2004) support the findings by Botha et al. (2006). Based on the mean absolute percentage error (MAPE) statistic and the square root of the mean squared error (RMSE) statistic, Moolman (2004) concluded that the Markov regime-switching model performed much better than the linear AR(4) model in terms of turning point prediction accuracy. The findings by Botha et al. (2006) and Moolman (2004) are insightful because the findings in the broader body of literature are inconclusive on whether linear or non-linear models perform best at forecasting business cycles. This indicates that non-linear models perform best in the South African context. Boshoff (2005) touched on the topic of South African financial cycles by estimating how well these variables aided as leading indicators for the South African business cycle. However, this study did not go through the process of determining the optimal variables to include in the aggregate South African financial cycle and chosen a few ad hoc financial variables to represent the South African financial cycle. Furthermore, the study by Boshoff (2005) did not forecast the South African financial cycle and only used the cycle to predict the business cycle. Very little to no research work has been done in an attempt to accurately forecast an aggregate South African financial cycle through different methods and thereby comparing the performance of the range of available forecasting methods to forecast the aggregate South African financial cycle. Given the proven role that changes in aggregate financial conditions have on the future economic path, timely and accurately forecasting cyclical fluctuations in financial conditions are essential to policymakers. Considering the large number of forecasting methods available in the literature, research on determining optimal ways to forecast a methodically constructed aggregate South African financial cycle will aid policymakers to better estimate future cyclical fluctuations in aggregate South African financial conditions. This research will attempt to address the research gap by comparing the forecasting performance of an optimally estimated linear ARIMA model to the forecasting performance of the non-linear MS-AR and STAR models at forecast aggregate South African financial cycles over different forecasting periods.

8

M. C. de Wet

3 Data Description and Methodology This study will utilize the aggregate South African financial cycle constructed by de Wet (2020), and the aggregate financial cycle constructed by de Wet (2020) will be used as the aggregate South African financial cycle in this study. The cyclical measure constructed by de Wet (2020) consists of 59 variables which in turn reflect 7 major financial components. The financial components reflected by this measure are South African credit conditions, South African property market conditions, South African interest rate conditions, balance sheet conditions of the South African financial sector, South African equity market conditions, South African economic confidence levels, and South African foreign financial positions. The aggregate South African financial measure, to be forecasted in this paper, is thus a single aggregate measure that reflects the cyclical movement of seven key South African financial components. de Wet (2020) made use of a principal component analysis and a dynamic factor model to aggregate 59 variables into a single measure. Furthermore, de Wet (2020) employed a Christiano Fitzgerald filter as a cyclical extraction method. The aggregate South African financial cycle typically expands as credit levels and asset prices increase (de Wet, 2020). Furthermore, the aggregate South African financial cycle typically expands due to an aggregate bank balance sheet expansion but contracts due to increase in interest rate (de Wet, 2020). This measure ranges from January 1975 to January 2017, and the data frequency is monthly. de Wet (2020) implemented a Dynamic Factor model to construct a single aggregate South African financial conditions index and made use of a Christiano Fitzgerald bandpass filter to extract the cyclical component from the aggregate conditions index to provide an aggregate South African financial cycle measure. The aggregate South African financial cycle from 1975 to 2017 is depicted in Fig. 1. The cyclicality over time is clear, and the cyclical measure provides a smooth representation of the cyclicality in aggregate South African financial conditions over time. A number of models will be used to forecast the aggregate South African financial cycle. The forecasting performance of each model will be compared to determine the most accurate model to forecast the aggregate South African financial cycle, provided varying forecasting timeframes. As done by a number of researchers such as Botha et al. (2006), Teräsvirta et al. (2005), Moolman (2004) Sarantis (2001) and Clements and Krolzig (1998), the forecasting performance of the linear ARIMA model will be used as a benchmark. The aggregate South African financial cycle will then be forecasted with the non-linear MS-AR model estimated to achieve objective three. In addition to the non-linear MS model, the aggregate South African financial cycle will be forecasted with an optimal STAR family model. The forecasting results of these non-linear models will be compared to that of the ARIMA model to determine whether accounting for non-linearities improves the performance of forecasting aggregate South African financial cycles. A fixed window rolling forecasting approach will be followed, and the study will consider the performance of each model to forecast the aggregate South African

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

9

4

Cyclical magnitude

3 2 1 0 -1 -2 -3 -4 1975

1980

1985

1990

1995

2000

2005

2010

2015

Time (years) Source: de Wet (2020) Fig. 1 The aggregate South African financial cycle. Source: de Wet (2020)

financial cycle over different time horizons. According to Clark and McCracken (2009), a fixed window rolling forecasting approach is an approach that uses a fixed estimation sample size of a time series to forecast a continuum of a fixed nstep ahead forecasts over time. The alternative to a fixed window rolling forecast is a recursive forecasting approach (Clark & McCracken, 2009). With such a forecasting approach, the estimation sample is heterogenic over time and grows as time progresses. The argument against such a forecasting approach is that the data points in the estimation sample can become irrelevant and redundant as the estimation sample grows (Clark & McCracken, 2009). The heterogenic forecasting sample also hinders the ability to compare the forecasting results of various models and various timeframes. This motivates why a fixed window rolling forecasting approach will be adopted in this study. The consensus in the literature is to use a third of the total number of observations in a time series as an estimation sample, leaving two-thirds of the total series to forecast (Clark & McCracken, 2009). The estimation outputs will thus be based on a third of the full data set. This forecasting process is best explained by the hand of an example. Assume a time series consists of 105 monthly closing price observations, dating from 31 January 2010 to 31 October 2018, and a 4-step ahead forecast is conducted. The fixed estimation sample will consist of 35 observations, thus approximately 2 years and 11 months. The estimation period for the first 4step ahead forecasted point will range from 31 January 2010 to 31 December 2012. Thus, the model under consideration will be estimated with data points ranging from 31 January 2010 to 31 December 2012. The 4-step ahead forecasted data point will represent a forecasted data point for 30 April 2013. The model under consideration will then be re-estimated for the next 4-step ahead forecasted data point with the estimation period ranging from 28 February 2010 to 31 January 2013. The 4-step ahead forecasted data point will represent a forecasted data point for 31 May 2013.

10

M. C. de Wet

Note how the estimation sample size remains 35 months, and the forecasted data points remain 4-step ahead forecasted values. This process will be repeated, and a 4-step ahead forecasted data point for each month up to 31 October 2018 will be obtained. This will result in a continuous series with data points that are forecasted four-steps ahead. In this study, fixed window rolling forecasts with ARIMA, MS-AR and STAR models will be done 1-step ahead, 3-steps ahead, 6-steps ahead, 12-steps ahead, 18-steps ahead and 24-steps ahead. The forecasting performance of each model with different forecasting time horizons will be analysed and compared. This will indicate which model performed best at forecasting the aggregate South African financial cycle and whether different models perform better at forecasting different forecasting time frames.

3.1 ARIMA Methodology The autoregressive integrated moving average (ARIMA) model where AR refers to autoregressive terms, I refer to the level of integration of a given variable, and MA the moving average of a given variable which measures the stochastic white noise error of the model as an amalgamation of previous errors. The ARIMA model is a very well-known statistical method and widely used in literature. An ARIMA (p, d, q) indicates that p number of autoregressive should be included in the model, the variables need to be differenced d amount of times for the variable to be stationary and q number of moving average terms needs to be included in the model (Brooks, 2019). The ARIMA model, where the variable has been differenced d amount of times to be stationary, can be written in the following linear equation according to Brooks (2019): ASAFCt = c + ∅1 ASAFCt−1 + ∅p ASAFCt−p + β1 μt + βq μt−q

(1)

where ASAFCt is the dependent variable at time t, c represents the intercept of the model, ∅p is the coefficients of the various autoregressive terms of the dependent variable ASAFCt , and β q represents the coefficients of the various moving average term μt . According to Gujarati and Porter (2009), a one-period ahead forecast of the dependent variable ASAFCt can be forecasted with the ARIMA models depicted in (1) as follows: ASAFCt+1 = c + ∅1 ASAFCt + ∅p ASAFCt−p + β1 μt + βq μt−q

(2)

To determine I, the integration level of dependent variable Yt , a Dickey-Fuller unit root will be conducted to determine at which level variable Yt is stationary as suggested by Gujarati and Porter (2009). Furthermore, as suggested by Gujarati and Porter (2009), the optimum number of autoregressive terms to include in the model will be done by making use of the partial autocorrelation function (PACF).

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

11

This function indicates the partial correlation of the dependent variable with its own lags by controlling lag values of relatively shorter lags when considering relatively longer lags (Gujarati & Porter, 2009). The optimal amount of moving average terms to be included in the model will be determined through the autocorrelation function (ACF), which differs from the PACF by not controlling for relatively shorter lags when considering longer lags (Gujarati & Porter, 2009). These measures will indicate how many AR terms and MA terms to include to estimate an optimal ARMA model. However, Gujarati and Porter (2009) state that conclusions based on these measures can be subjective and therefore suggested additional measures such as the Akaike Information Criterion (AIC), Schwartz Criterion (SC) and Hannan Quinn Criterion (HQC) to determine the optimal ARMA model. Henceforth, in addition to ACF and PACF, the AIC, SC and HQC will be used to determine the optimal ARMA model.

3.2 Non-linear Models In this section, a methodological outline will be provided for both the MS-AR method and the STAR method to be employed in this study. These models make it posable for autoregressive parameters in a model to change over time (Botha et al., 2006).

3.2.1

MS-AR Models

In this study, four variants of the MS-AR model will be considered with various laglengths. The variations are an MS-AR model with a fixed mean and fixed variance, an MSM-AR model with a regime-dependent mean and fixed variance, an MSV-AR model with a fixed mean and regime-dependent variance and an MSMV-AR model with a regime-dependent mean and regime-dependent variance. Two extensions to the STAR model will be considered in this study, namely the logistic smooth autoregressive (LSTAR) model and an exponential transition function, which will result in the estimation of an exponential smooth autoregressive (ESTAR) model. Both these extensions will be considered, and the optimal version will be used to forecast aggregate South African financial cycles. An optimal MS-AR model to estimate the aggregate South African financial cycle will be identified based on the Akaike Information Criterion (AIC), Schwartz Information Criterion (SIC) and Hannan Quinn Criterion (HQC). The model with the lowest information criterion value will be selected as the optimal model, and further analysis will be based on the results rendered by the optimal model (Brooks, 2019). This selection approach is similar to the selection approach followed by Tastan and Yildirim (2008). Furthermore, it will be assumed that the aggregate South African financial cycle has two states, namely an expansion and a contraction.

12

M. C. de Wet

The MS-AR model with a fixed mean and fixed variance is specified as follows: yt = βs1 (yt−1 + yt−2 + · · · + yt−k ) + βs2 (yt−1 + yt−2 + · · · + yt−k ) + εt (3) where, st ∈ {1, 2} signifies the regime state under consideration, i.e. state one and two, k signifies the optimal lag length, εt is a non-state-dependent error term and xt is a vector of explanatory variables. Equation (3) can be restated to accommodate for a regime-switching mean and in this setting can be re-specified as follows (Tastan & Yildirim, 2008): yt = cts + βs1 (yt−1 + yt−2 + · · · + yt−k ) + βs2 (yt−1 + yt−2 + · · · + yt−k ) + εt (4) where Cts is a state-dependent, s, intercept, allowing for a state-dependent mean. Lastly, (3) can be restated to accommodate for a regime-switching mean and a regime-switching variance and in this setting can be re-specified as follows: yt − μst = cts + βs1 (yt−1 + yt−2 + · · · + yt−k − μst−1 ) + βs2 (yt−1 + yt−2 + · · · + yt−k − μst−2 ) + εt

(5)

Assuming that St is a first-order Markov process meaning that the current regime is a function of the preceding regime St − 1 , then the transition probabilities of progressing from one regime to another regime can be stated as follows (Tastan & Yildirim, 2008):

pij = Pr (St = j |St−1 = i) ,

n 

pij = 1, ∀i, j ∈ {1, 2, . . . , n)

(6)

j =1

Thus for a cycle that exhibits two states, an expansion and a contraction, where St = {1, 2}, respectively, the transition matrix is as follows (Tastan & Yildirim, 2008):  P =

p11 p12 p21 p22

 (7)

where each entry, p11 depicts the conditional probability of remaining in an expansion once in an expansion, p12 depicts the conditional probability of moving from an expansion to a contraction, p21 depicts the conditional probability of moving from a

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

13

contraction to an expansion and p22 depicts the conditional probability of remaining in a contraction once in a contraction.

3.2.2

STAR Models

Following Teräsvirta et al. (2005), the STAR model is specified as follows:   st = β0 + β’1 wt + ∅0 + ∅1 wt ∂ (st−d ) + ε

(8)



where wt = (1, st − 1 , . . . , st − k ) , consisting of k lags of the aggregate South African financial cycle, ∂ is a transition function and st − d is the transition variable and d is the delay parameter. Furthermore, ∅1 = (∅1 , ∅2 , . . . , ∅p )’ and  β 1 = (β 1 , β 2 , . . . , β k ) are parameter vectors. The transition function can be normal, logistic or exponential (Sarantis, 2001). A normal transition function will result in the estimation of a standard STAR model, a logistic transition function will result in the estimation of a logistic smooth autoregressive (LSTAR) model and an exponential transition function will result in the estimation of an exponential smooth autoregressive (ESTAR) model. Modelling a STAR model requires three procedures (Teräsvirta et al., 2005). The first procedure is to estimate an autoregressive model and determine the optimal number of autoregressive lags which will become k in (8). The second procedure is to establish whether the variable under consideration exhibits nonlinear characteristics within a STAR set-up. If the variable under analysis exhibits non-linear characteristics, the use of the non-linear STAR model to forecast such variables will be justified (Sarantis, 2001). If the variable under analysis does not exhibit non-linear characteristics, modelling and forecasting with a non-linear model will be pointless and a linear model, such as an AR model, will be stuffiest. Thirdly, a sequence of nested tests will be conducted to determine whether the LSTAR model or the ESTAR model is optimal. To carry out process two and three, the following auxiliary regressions will be estimated as specified by Teräsvirta et al. (2005): 2 3 Vt = β0 + β1 wt + β2 wt st−d + β3 wt st−d + β4 wt st−d + μt

(9)

where Vt is the residuals of the linear AR(4) model specified in (1). A range of d values will be considered, and the axillary model will be reiterated for each value. Sarantis (2001) suggested an integral of 1  d ≥ 6. The smooth threshold linearity test will be conducted to determine whether the South African financial cycle has non-linear characteristics. The null hypothesis for the smooth threshold linearity test is as follows: H0 : β2 = β3 = β4 = 0

14

M. C. de Wet

Rejecting the null hypothesis will indicate that the variable under analysis does exhibit non-linear characteristics. To establish whether the ESTAR model or the LSTAR model is optimum, the Terasvirta sequential test will be conducted as suggested by Teräsvirta et al. (2005). The null hypotheses for this test are as follows: H0 : β3 = 0 H0 : β2 = 0 | β3 = 0 H0 : β1 = 0 | β3 = β2 = 0 The selection rule states that if H0 : β 1 = 0  β 3 = β 2 = 0 has the smallest pvalue, then an ESTAR model should be used. If any of the other null hypotheses have the lowest p-value, then an LSTAR model should be estimated. Furthermore, the optimum lag number for d will be established based on the d with the lowest p-value, thus the most significant d. Estimation of the STAR, LSTAR and ESTAR models will provide two sets of beta coefficients for each threshold variable (Teräsvirta et al., 2005). The first set of coefficients will indicate the relationship between a given threshold variable and the aggregate South African financial cycle during linear periods in the aggregate financial cycle. The second set of coefficients will indicate the relationship between a given threshold variable and the aggregate South African financial cycle during non-linear periods in the aggregate financial cycle (Teräsvirta et al., 2005). Linear periods refer to consecutive periods in the aggregate South African financial cycle where no cyclical regime change occurred, for example, a pivot from an expansion to a contraction phase. On the other hand, non-linear periods refer to periods where the aggregate South African financial cycle series pivoted from one cyclical regime to another, for example moving from an expanding to a contracting regime (Teräsvirta et al., 2005). Thus, the set of beta coefficients for threshold variables during non-linear periods indicate the relational dynamics between the aggregate South African financial cycle and a corresponding threshold variable during a period where a cyclical change took place.

LSTAR The LSTAR model allows asymmetrical adjustments to the non-linear process. Therefore, if non-linear shifts in the aggregate South African financial cycle does not occur symmetrically during different regimes, then the LSTAR model will be more effective in forecasting turning points in the aggregate South African financial

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

15

cycle (Teräsvirta et al., 2005). Teräsvirta et al. (2005) state that the LSTAR model follows the following form: 



∂ (st−d ) = 1 + exp −γ

K

−1 (st−d − ck )

,

(10)

k=1

γ > 0, c1 ≤ · · · ≤ cK where γ measures the speed of moving from one regime to the other. Teräsvirta et al. (2005) write that the most common value for k in the transition function γ > 0, c1 ≤ . . . ≤ cK is k = 1. Teräsvirta et al. (2005) explain that if k = 1, then parameters ∅ + λG(st − d, h ; γ , c) change monotonically from ∅ to ∅ + λ meaning that as λ converge to zero, the LSTAR model develops into a linear AR model and becomes a two-regime TAR model if the closer c1 converges to 1.

ESTAR According to Enders (2004), the ESTAR model is symmetrical if ASAFCt − 1 = c, allowing the model to approximate gravitational attraction, making the model optimal if the series depicts various levels of autoregressive decay at cyclical pivot points. Enders (2004) states that the STAR model can be transformed to an exponential form by making

γ = 1 + exp( − s(st−0 − c)2 y>0

(11)

Since λ is constant, movements of γ towards 0 or ∞ make the ESTAR model a AR(p) process, whereas divergence of λ from 0 or ∞ makes the process non-linear.

3.3 Determining Forecasting Performance To determine which model performs the best at forecasting the aggregate South African financial cycle, five information criteria will be used, namely root mean square error (RMSE), mean absolute error (MAE), the mean absolute percentage error (MAPE) and Theil’s U statistic. These performance measures are often used in literature, such as the work by Wai et al. (2015), Nyberg (2018), Baharumshah and Liew (2006), Botha et al. (2006) and Moolman (2004), as a means to determine the forecasting performance of models. It is argued, however, that forecasting performance measures should be used in combination to get a consensus view, and not in isolation. This will increase the validity of one’s findings, hence all five

16

M. C. de Wet

measures will be used. The formulas for each measure are as follows, as seen in Brooks (2019):   T 1   2 yt,s − ft,s , RMSE =  n

(12)

t=T1

where yt, s is the actual value at time t and ft, s is the forecasted value at time t. Furthermore, n represents the total number of observations in the time series.  T  100   y − ft,s  MAPE =  Y , n t,s

(13)

t=T1

     n−1 ft,s −yt,s 2  t=1 y  t−1 Theil s U1 statistic =   2  n−1 ft,n −yt,s t=1

(14)

yt−1

where yt − 1 is the actual value one period prior to the period under consideration and Ft, n represents the naïve forecasted value. Given that RMSE and MAPE indicate the error of a given forecasting process, the lower the values generated by RMSE and MAPE, the better the forecasting accuracy (Brooks, 2019). The average of each measure over time will be considered. The Theil’s U statistic provides a slightly different measure than the other four measures considered in this study. It indicates whether the forecasting model performs better than a naïve forecasting approach. A Theil’s U coefficient smaller than one indicates that the errors rendered by the forecasting model are lower than that of a naïve forecasting approach and are thus superior to a naïve forecasting approach (Brooks, 2019). On the other hand, if the Theil’s U coefficient is larger than one, then a naïve forecasting approach outperforms the forecasting ability of the forecasting model under consideration (Brooks, 2019). The Theil’s U statistic will therefore not be used to compare the various forecasting models to each other, but rather to determine whether the model under consideration is more effective than using a naïve forecasting approach. Provided that a rolling forecasting approach will be taken for the various n-ahead forecasting time horizons, the average of the various forecasting performance measures over time, for a given forecasting horizon, will be considered.

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach Table 1 Phillip-Perron unit root test and the Breakpoint Augmented Dickey-Fuller unit root test results

17

Breakpoint Augmented Dickey-Fuller test results T-statistic Augmented Dickey-Fuller test statistic −8.519 Test critical value at a 99% level −3.445 Test critical value at a 95% level −2.868 Test critical value at a 90% level −2.570 Phillips-Perron test results T-statistic Phillips-Perron test statistic −4.269 Test critical value at a 99% level −3.445 Test critical value at a 95% level −2.867 Test critical value at a 90% level −2.570 Source:Author’s calculation

4 Results, Findings, and Discussion Firstly, the results rendered by the ARIMA model will be considered, followed by the results rendered by the STAR model. Then the results from the MS-AR model will be considered, followed by the forecast performance evaluation.

4.1 ARIMA Model Outputs First, the order of integration, I, of the aggregate South African financial cycle will be established. Research has shown that the standard Augmented Dickey-Fuller unit root test is often sub-par when working with a time series that exhibits cycles and regime-switching properties (Nelson, Piger, & Zivot, 2000). Nelson et al. (2000) suggest using the Phillip-Perron unit root test or a breakpoint Augmented DickeyFuller unit root test which allows for endogenous probabilistic trend fluctuations in a series when testing a cyclical series for stationarity. Therefore, the Phillip-Perron unit root test and the Breakpoint Augmented Dickey-Fuller unit root test will be used to test the level of integration of the aggregate South African financial cycle. Table 1 depicts the results rendered by these tests. These tests indicate whether the aggregate South African financial cycle is stationary and if not, how many times the series needs to be differenced to be stationary. This will indicate the level of integration (I). In absolute terms, both the Breakpoint Augmented Dickey-Fuller test statistic and the Phillips-Perron test statistic are larger than the critical value at a 99% confidence level. Thus, the null hypothesis of a unit root can be rejected, and it could be concluded that the aggregate South African financial cycle is stationary at level. The order of integration (I) is thus 0, and the series does not need to be differenced. Given that the I in the ARIMA model is now established, the optimal number of AR

18

M. C. de Wet

Table 2 ACF and PACF for the aggregate South African financial cycle series Autocorrelation .|***** .|*** .|**** .|*** .|. | .|. | .|. | .|. | .|. | .|. |

Partial correlation .|***** ***|. .|*** **|. .|. | .|. | .|. | .|. | .|. | .|. |

Number of lags 1 2 3 4 5 6 7 8 9 10

AC 0.916 0.820 0.871 0.822 0.113 0.045 0.030 0.040 0.006 0.020

PAC 0.986 −0.771 0.767 −0.698 0.055 0.046 0.037 0.036 0.013 0.003

Prob 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Source:Author’s calculation

and MA terms must now be determined. A correlogram will be used to display the ACF and PACF of the aggregate South African financial cycle series. The rule of thumb is that the ACF and the PACF should die down to zero if the time series is stationary. Furthermore, spikes in the ACF on the correlogram indicates the optimal MA specification and spikes in the PACF on the correlogram indicates the optimal AR specification. Table 2 depicts ACF and PACF outputs and the corresponding correlogram. Based on the correlogram depicted in Table 2, both the ACF and PACF die down to zero fairly quickly, indicating that the aggregate South African financial cycle series is stationary. This corresponds with the results rendered by the ADF and PP unit root tests. Based on the correlogram depicted in Table 2, the ACF function spikes at the first four lags and then dies out after the first three lags. Furthermore, the PACF spikes up to four lags and then dies down. This indicates that an AR(4)MA(4) specification will be optimal. Table 3 depicts the AR(4)MA(4) model outputs and based on these outputs, it can be seen that all the AR and MA terms are significant at a 99% confidence level. The model depicted in Table 3 has an adjusted R-squared of 0.851, meaning that the explanatory variables in the model explained 85.1% of the change in the aggregate South African financial cycle. Given this high explanatory power, the AR(4)MA(4) model will be used to forecast the aggregate South African financial cycle, and the forecasting performance of this model will be assessed. As done by a number of researchers such as Botha et al. (2006), Teräsvirta et al. (2005), Moolman (2004) Sarantis (2001) and Clements and Krolzig (1998), the forecasting performance of the ARMA model will be used as a benchmark. The forecasting performance of non-linear models will be compared to that of the ARMA model to determine whether accounting for non-linearities improves the performance of forecasting the aggregate South African financial cycle.

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach Table 3 ARIMA results

Variable Coefficient C 0.062 AR (1) 3.909 AR (2) 4.271 AR (3) −3.834 AR (4) −0.962 MA (1) 1.086 MA (2) −0.167 MA (3) −0.540 MA(4) 0.092 Adjusted R-squared: 0.851

19

P-value 0.607 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.001*** 0.000*** 0.000***

*, ** and *** denote statistical significance at a 99%, 95% and 99% confidence level, respectively, based on p-values Source:Author’s calculation Table 4 Outputs rendered by the STAR model

Coefficient P-value Variables during the linear phase AR(1) 0.998 0.000*** AR(2) 0.942 0.009*** AR(3) −1.081 0.000*** AR(4) −0.713 0.001*** Variables during non-linear phase AR(1) −0.727 0.021** AR(2) −0.923 0.011** AR(3) 0.343 0.027** AR(4) −0.161 0.033** Non-threshold variable C −0.004 0.013** Threshold: 2.471 and P-value of threshold: 0.000*** Adjusted R-squared: 0.906 *, ** and *** denote statistical significance at a 99%, 95% and 99% confidence level, respectively, based on p-values Source:Author’s calculation

4.2 STAR Model Outputs The results in the previous section indicated that four AR lag terms are optimal and therefore the base STAR(n) model will be a STAR(n) model with four lags, STAR(n). The results for the standard STAR(4) are presented in Table 4. The top part of the table depicts the coefficients of each AR term exhibited during linear periods in aggregate South African financial cycles. Based on the P-values of each AR term, all four AR terms are statistically significant.

20 Table 5 Smooth threshold linearity test results

M. C. de Wet Null hypothesis H04: b1 = b2 = b3 = b4 = 0 H03: b1 = b2 = b3 = 0 H02: b1 = b2 = 0 H01: b1 = 0

P-value 0.000*** 0.005*** 0.009*** 0.002**

*, ** and *** denote statistical significance at a 99%, 95% and 99% confidence level, respectively, based on p-values Source:Author’s calculation

The middle part of the table depicts the coefficients of each AR term exhibited during non-linear periods in aggregate South African financial cycles. Based on the P-values of each AR term, all four AR terms are statistically significant during non-linear periods. The lower part of the table depicts the coefficient of the non-threshold variable, the transition threshold value, and the adjusted R-squared. The threshold value is 2.471, indicating that the probability of a cyclical change increases significantly when the cycle reaches the 2.471 level. Also, the adjusted Rsquared of 0.906 indicates that 90.6% of the variance in the aggregate South African financial cycle is explained by the variables in the model. This is an improvement in the linear ARIMA model which had an adjusted R-squared of 85.1%. This indicates that the consideration of non-linearity improves the explanatory power of a model. From this STAR(4) model, a smooth threshold linearity test is conducted within the STAR(4) model set-up to determine whether aggregate South African financial cycles exhibit non-linear characteristics. Furthermore, a Terasvirta sequential test is conducted, as suggested by Teräsvirta et al. (2005), to determine whether the transition function is a normal, logistic or exponential. If the transition function is normal, then the STAR model is the optimal smooth transition model; if the transition function is logistic, then the LSTAR model is the optimal smooth transition model and if the transition function is exponential, then the ESTAR model is the optimal smooth transition model. Once this has been determined, the P-values for a range of delay factors, estimated as part of the linearity test, are considered to determine the optimal d in (8) and (9). These tests will aid in selecting the optimal smooth transition model, which in turn will ultimately be used to forecast aggregate South African financial cycles. Table 5 depicts the results rendered by the smooth threshold linearity test. All four hypotheses in Table 5 can be rejected at a 99% confidence interval. The 2 and w s 3 , results in Table 5, therefore, indicate that the beta coefficients of wt st−d t t−d the two non-linear measures in the auxiliary regression, represented by (9), are highly significant (Teräsvirta et al., 2005). The aggregate South African financial cycle, therefore, does exhibit non-linear characteristics, making it appropriate to model the aggregate South African financial cycle with the STAR methodology. Table 6 represents the results rendered by the Terasvirta sequential tests. Provided that H2: b2 = 0 | b3 = 0 is significant at a 99% confidence level, as depicted in Table 6, and has a lower p-value than H1: b1 = 0 | b2 = b3 = 0, a

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach Table 6 Terasvirta sequential test results

Null hypothesis H3: b3 = 0 H2: b2 = 0 | b3 = 0 H1: b1 = 0 | b2 = b3 = 0

21 P-value 0.035** 0.000*** 0.009***

*, ** and *** denote statistical significance at a 99%, 95% and 99% confidence level, respectively, based on p-values Source:Author’s calculation Table 7 P-values for the range of delayed factors considered in this study based on linearity test

d lag 1 2 3 4 5 6

P-value 0.046** 0.003*** 0.018** 0.025*** 0.028** 0.037**

*, ** and *** denote statistical significance at a 99%, 95% and 99% confidence level, respectively, based on pvalues Source:Author’s calculation

logistic transition function is optimum. The aggregate South African financial cycle will thus be modelled and forecasted by the LSTAR model. Finally, Table 7 depicts the P-values of a range of delay factors derived from the linearity test. Based on the results in Table 7, d = 2 is the most significant given that d = 2 has the lowest p-value. The delay factor for the STAR model in this study will thus consist of two lags. Given the results considered, one now knows that there are non-linearities in the aggregate South African financial cycle and that the transition function exhibits logistic characteristics. Furthermore, the optimum d in (8) and (9) is established. This allows for the estimation of an optimal smooth transition model which is an LSTAR(4) model with a two-lag delay factor. Table 8 depicts the results estimated by the LSTAR(4) model. All the threshold variables during a linear period in the aggregate financial cycle are significant at a 99% confidence level, provided that each variable exhibited a pvalue smaller than 0.01. The beta coefficients indicate that, during a linear period, a one unit increase in the aggregate South African financial cycle one, two, three and four periods back will lead to a 0.878, 0.599, −0.757, −0.640 unit change in the current value of the aggregate South African financial cycle, respectively. Thus, an increase in the aggregate South African financial cycle in time t1 will typically have a positive impact on the aggregate South African financial cycle in the following two periods but then have a negative impact three and four periods ahead. During a non-linear period, a one unit increase in the aggregate South African financial cycle one, two, three and four periods back will lead to a −0.421, −0.337, 0.248 and −0.454 unit change in the current value of the aggregate South African

22

M. C. de Wet

Table 8 Outputs rendered by the LSTAR model Coefficient Variables during the linear phase AR(1) 0.878 AR(2) 0.599 AR(3) −0.757 AR(4) −0.640 Variables during non-linear phase AR(1) −0.421 AR(2) −0.337 AR(3) 0.248 AR(4) −0.454 Non-threshold variable C −0.006 Absolute threshold value: 2.786 and P-value of threshold: 0.000*** Adjusted R-squared: 0.936

P-value 0.000*** 0.000*** 0.000*** 0.000*** 0.020** 0.042** 0.030** 0.022** 0.014**

*, ** and *** denote statistical significance at a 99%, 95% and 99% confidence level, respectively, based on p-values Source:Author’s calculation

financial cycle, respectively. This mostly indicates that an inverse relationship exists between the aggregate South African financial cycle at time t1 and lags of itself, except for the third AR lag term. The cyclical series is in a transition period during a non-linear phase, thus an inverse relationship between the current period and the preceding periods is expected. Consider now the threshold value and Adjusted Rsquared depicted in the bottom part of Table 8. The absolute threshold value is 2.786, indicating the typical level at which aggregate South African financial cycles reach a pivot point. This indicates that the probability of a cyclical change increases significantly when aggregate South African financial cycles reach the 2.786 or −2.786 level. Also, the adjusted R-squared of 0.936 indicates that the threshold variables in the model perform well in explaining movements in the aggregate South African financial cycle.

4.3 MS-AR Model Outputs The results from the various criterion for the various MS-AR-type models are presented in Appendix. The HQC, AIC and SIC prove to be the lowest for the MSMV(2)-AR(3) model. Thus, given that all the three information criteria are lowest for the MSMV(2)-AR(3) model, the optimal specification to model the aggregate South African financial cycle is an MS model with three AR lags, a regimedependent mean and a regime-dependent variance with two states. Hence, the results of the MSMV(2)-AR(3) model will further be considered, and the MSMV(2)-AR(3)

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach Table 9 Estimation output of the MSMV(2)-AR(3) model

23

Variable MSMV(2)-AR(3) μs1 0.099*** μs2 −0.162*** β 1s1 AR(1) 1.966*** β 2s1 AR(2) 1.162*** β 3s1 AR(3) −1.105*** β 1s2 AR(1) 2.063*** β 2s2 AR(2) 1.753*** β 3s2 AR(3) −0.986** σ s1 −3.661*** σ s2 −5.787*** Transition matrix parameters P11-C 2.974*** P21-C −3.198*** Typical duration (in months) Expanding phase 49.31 Contracting phase 35.79 Transition probabilities p11 0.951 p12 0.049 p22 0.891 p21 0.109 ** and *** denote statistical significance at a 95% and 99% confidence level, respectively, based on p-values Source:Author’s calculation

model will be used to forecast the aggregate South African financial cycle. The outputs from the MSMV(2)-AR(3) model are depicted in Table 9. In the MSMV(2)-AR(3) model, the mean, μs , and variance, σ s , and cyclical persistent terms, β 1s AR(1), β 2s AR(2) and β 3s AR(3), depend on the unobservable Markov state variable that may assume two values, st ∈ {1, 2}. Firstly, consider the results from the regime-dependent means, μs1 and μs2 . The regime-dependent means of both regimes, μs1 and μs2 , are statistically significant at a 99% confidence level and have opposite signs. Thus, the point estimates of the regime-dependent means are statistically different from each other. This provides evidence that supports the assumption that two distinct regimes characterize the aggregate South African financial cycle. This justifies the use of an MSMV(2)-AR(3) model that accounts for a regime-dependent mean. The regime-dependent mean in regime 1, μs1 , is positive, and the regimedependent mean in regime 2, μs2 , is negative. Given that μs1 > μs2 , the evidence is provided that one can interpret regime 1 as the expanding regime and regime 2 as the contracting regime (Tastan & Yildirim, 2008). Secondly, consider the variance parameter, σ s1 , and σ s2 . Both these parameters are statistically significant at a 99% confidence level with different magnitudes. In absolute terms σ s1 < σ s2 , as stated by Tastan and Yildirim (2008), indicates that there is volatility asymmetry

24

M. C. de Wet

between regimes. These results indicate that volatility is lower during an expanding phase relative to the volatility in a contracting phase. This result was expected and corresponds to empirical literature providing evidence that cyclical contractions in both financial conditions, i.e. the 2007 financial crisis, and real economic conditions, i.e. business cycle contractions, are more violent and harsh relative to expansions (Tastan & Yildirim, 2008; McQueen & Thorley, 1993). These asymmetries provide additional justification for the use of non-linear MS-AR methodology that accounts for asymmetries and accounts for a regime-dependent variance. The transition matrix parameters, P11-C and P21-C, in Table 9 are both statistically significant at a 99% confidence level and have opposite signs. The positive P11-C and negative P21-C signifies that increases in the aggregate South African financial cycle are associated with higher probabilities of remaining in the expanding regime, lowering the transition probability out of regime 1 and increasing the transition probability from regime 2 into regime 1. Furthermore, the results from this model indicated that an aggregate South African financial cycle expansion lasts approximately 49.31 months, thus 4 years and 1.31 months and a contraction last approximately 35.79, thus 3 years. These results thus indicate that an expansion in the aggregate South African financial cycle has a longer duration than a contraction. The aggregate South African financial cycle thus exhibits a level of durational asymmetry. The various models employed to estimate the aggregate South African financial cycle will now be used to forecast the aggregate South African financial cycle, and the forecasting performance of the various models will be compared to identify the optimal model to forecast the aggregate South African financial cycle.

4.4 Forecasting Performance Evaluation In this section, the rolling forecasting performance of the linear AR(4)MA(4) model, the LSTAR(4) model and the MSMV(2)-AR(3) model will be established and compared. A fixed window rolling forecasts with each model will be done 1-step ahead, 3-steps ahead, 6-steps ahead, 12-steps ahead, 18-steps ahead and 24-steps ahead. The RMSE, MAPE and Theil U1 coefficient are considered to identify the model with the best forecasting performance, given the different forecasting horizons. Table 10 depicts the forecasting performance measures rendered by each of the three models for the various forecasting time horizons. Consider the results for a three-step and six-step forward forecasting horizon in Table 10. These are the two shortest forecasting time horizons in this study, and based on the RMSE and MAPE the linear ARIMA model produced the most accurate forecasts for this time horizon. The benchmark ARMIA model thus outperforms the non-linear models at forecasting short periods. This corresponds to the findings by Balcilan et al. (2015) who found that non-linear models typically do not outperform standard ARIMA models in forecasting short periods ahead. A possible explanation for this can be that aggregate South African financial cycles exhibit

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

25

Table 10 Forecasting performance measures RMSE Three-steps forward AR(4)MA(4) 7.86E-05a MSMV(2)-AR(3) 0.014 LSTAR(4) 0.010 Six-steps forward AR(4)MA(4) 0.001a MSMV(2)-AR(3) 0.027 LSTAR(4) 0.038 Twelve-steps forward AR(4)MA(4) 0.312 MSMV(2)-AR(3) 0.109 LSTAR(4) 0.064a Eighteen-steps forward AR(4)MA(4) 0.315 MSMV(2)-AR(3) 0.046a LSTAR(4) 0.297 Twenty-four-steps forward AR(4)MA(4) 1.307 MSMV(2)-AR(3) 0.132a LSTAR(4) 0.818

MAPE

Theil U1 coefficient

0.070a 3.890 3.366

2.64E-05a 0.005 0.004

0.116a 6.943 8.629

0.000a 0.009 0.013

44.634 19.027 11.547a

0.104 0.012a 0.037

45.978 12.749a 43.968

0.105 0.020a 0.101

107.760 15.730a 86.382

0.433 0.045a 0.271

Source:Author’s calculation a Indicates the lowest measure and thus the model with the best forecasting performance

linearities over short periods. In other words, aggregate South African financial cycles do not reach a cyclical turn every 3–6 months, thus not exhibiting nonlinearities. Therefore, by accounting for non-linearities does not improve forecasting accuracy, as such non-linearities are seldom over short periods. Furthermore, the simplicity of ARIMA modelling and forecasting can attribute to the forecasting accuracy of such models (Balcilan et al., 2015; Crawford & Fratantoni, 2003). In combination, the simplicity of ARIMA forecasting and the possible non-linearities exhibited by aggregate South African financial cycles over short periods can explain why the ARIMA model outperforms MS-AR and LSTAR models over a 3 and 6step forecasting horizon. On the other hand, based on the RMSE and MAPE, both the non-linear MSMV(2)-AR(3) and LSTAR(4) models produced more accurate forecasts for 12-, 18- and 24-steps ahead forecasting time horizon than the linear AR(4)MA(4) model. Based on the RMSE and MAPE, the forecasting accuracy of the linear AR(4)MA(4) model deteriorates drastically as the forecasting time horizon increases. The mean absolute percentage error is as high as 107.760% for a 24-period ahead forecast, indicating how inaccurate the AR(4)MA(4) model becomes at forecasting aggregate South African financial cycles. This corresponds to the findings of a large number of researchers such as, but not limited to, the work done by Wai et al. (2015), Baharumshah and Liew (2006), Botha et al. (2006), Teräsvirta et al. (2005),

26

M. C. de Wet

Moolman (2004), Crawford and Fratantoni (2003) and Clements and Krolzig (1998) who found evidence that forecasting gains can be generated by exploiting non-linear structures offered by STAR and MS-AR models. Forecasting 12-steps ahead, the LSTAR(4) model generated the most accurate forecasts according to the RMSE and MAPE. The Theil U1 coefficient indicates that the MSMV(2)-AR(3) model performs slightly better at forecasting aggregate South African financial cycles 12-steps ahead. It might, therefore, be useful to use both the LSTAR and the MSMV(2)-AR(3) models to forecast the aggregate South African financial cycle 12-steps ahead. According to all three performance measures, the MSMV(2)-AR(3) model outperforms the LSTAR (4) and AR(4)MA(4) models at forecasting the aggregate South African financial cycle 18- and 24-steps ahead. These are the longest forecasting time horizons considered in this study. This shows that the MSMV(2)-AR(3) model renders the most accurate forecasts of the aggregate South African financial cycle given a longer-term time horizon. This corresponds to the work done by Clements and Krolzig (1998) who found evidence that MS-AR models outperform STAR models at forecasting economic variables. The Theil U1 coefficient indicates that the linear AR(4)MA(4) model as well as the non-linear MSMV(2)-AR(3) and the LSTAR(4) models outperforms the naïve forecasting approach over all time horizons.

5 Conclusion The aim of this article was to identify the best model to forecast aggregate South African financial cycles over various time periods, specifically distinguishing between the forecasting performance of linear vs non-linear models. The forecasting performance of the linear ARIMA model was used as a benchmark and the forecasting performance of the LSTAR and MS-AR models was measured relative to that of the ARIMA model. The RMSE, MAPE and Theil U coefficient were used as forecasting performance measures because these measures are widely accepted and used in forecasting literature to establish and compare the forecasting performance of various models to one another. The results rendered by the LSTAR model indicate that the threshold level of the aggregate South African financial cycles is 2.891 in absolute terms. This indicates that the aggregate South African financial cycle tends to reach a pivot point when it reaches a level of 2.891 or −2.891. Therefore, the probability of a cyclical change increases considerably when the aggregate South African financial cycle reaches 2.891 or −2.891. Furthermore, the results rendered by the smooth threshold linearity test indicate that non-linearities are present in the aggregate South African financial cycle series. This corresponds to the research done by Singh (2012). This indicates that the modelling of aggregate South African financial cycles with linear models might be suboptimal and that it might be necessary to account for non-linearities in the modelling of aggregate South African financial cycles. It also indicates that the use of non-linear models to forecast aggregate South African financial cycles might improve forecasting accuracy. It was therefore expected that the non-linear MS-AR and LSTAR models will outperform the benchmark ARIMA model.

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

27

Interestingly, evidence indicates that the linear AR(4)MA(4) model outperforms the LSTAR(4) and MSMV(2)-AR(3) models at forecasting the aggregate South African financial cycle three- and six-steps ahead. Thus, given a short forecasting horizon, no forecasting gains are achieved by accounting for non-linearities. However, for longer forecasting time horizons the non-linear MSMV(2)-AR(3) and the LSTAR(4) models outperform the linear AR(4)MA(4) model. Thus, as the forecasting time horizon increases, forecasting gains are achieved by exploiting the non-linear structure of the LSTAR and MSMV-AR models. Furthermore, evidence is found that the MSMV(2)-AR(3) model outperforms the LSTAR(4) model at forecasting aggregate South African financial cycles 18- and 24-steps ahead. Policymakers and other economic participants which are exposed to the aggregate South African financial cycle should use the AR(4)MA(4) model to forecast the aggregate South African financial cycle 3–6 months ahead. However, the MSMV(2)-AR(3) model should be used to forecast the aggregate South African financial 12–24 months ahead.

A.1 Appendix: Selection Criterion for Various MS-AR Models

MS(2)-AR(1) MS(2)-AR(2) MS(2)-AR(3) MS(2)-AR(4) MSM(2)-AR(1) MSM(2)-AR(2) MSM(2)-AR(3) MSM(2)-AR(4) MSV(2)-AR(1) MSV(2)-AR(2) MSV(2)-AR(3) MSV(2)-AR(4) MSMV(2)-AR(1) MSMV(2)-AR(2) MSMV(2)-AR(3) MSMV(2)-AR(4)

HQC −0.804 −6.369 −9.002 −7.908 −1.724 −9.443 −8.442 −5.885 −1.207 −7.384 −9.384 −5.639 −1.212 −9.021 −10.374a −5.546

AIC −0.823 −6.395 −9.034 −7.947 −1.747 −9.478 −8.482 −5.914 −1.233 −7.422 −9.422 −5.671 −1.234 −9.057 −10.448a −5.572

Source:Author’s calculation a Optimal model based on criterion

SIC −0.774 −6.329 −8.951 −7.848 −1.689 −9.387 −8.377 −5.840 −1.167 −7.321 −9.323 −5.589 −1.177 −8.966 −10.206a −5.506

28

M. C. de Wet

References Baharumshah, A. Z., & Liew, V. K. (2006). Forecasting performance of exponential smoothing transition autoregressive exchange rate models. Open Economics Review, 17, 235–251. Balcilan, M., Gupta, R., & Miller, S. M. (2015). The out-of-sample forecasting performance of nonlinear models of regional housing prices in the US. Applied Economics, 47(22), 2259–2277. Borio, C. (2014). The financial cycle and macroeconomics: What have we learnt? Journal of Banking and Finance, 45, 182–198. Borio, C., Drehmann, M., & Xia, D. (2018). The financial cycle and recession risk. Bank of International Settlements Quarterly Review, 32(4), 59–71. Boshoff, W. H. (2005). The properties of cycles in south African financial variables and their relation to the business cycle. South African Journal of Economics, 73(4), 694–709. Botha, I., Greyling, L., & Marais, D. J. (2006). Modelling the business cycle of South Africa: A non-linear approach. Journal of Studies in Economics and Econometrics, 30(2), 103–126. Bouali, M., Nasr, A. B., & Trabelsi, A. (2016). A nonlinear approach for modeling and forecasting US business cycles. International Economic Journal, 30, 39–74. Brooks, C. (2019). Introductory econometrics for finance (4th ed.). Cambridge: Cambridge University Press. Burns, A. F., & Mitchell, W. C. (1946). Measuring business cycles. New York: NBR. Clark, E., & McCracken, M. W. (2009). Improving forecast accuracy by combining recursive and rolling forecasts. International Economic Review, 50(2), 363–395. Clements, M. P., & Krolzig, H. M. (1998). A comparison of the forecast performance of Markovswitching and threshold autoregressive models of US GNP. Econometric Journals, 1, 47–75. Crawford, G. W., & Fratantoni, M. C. (2003). Assessing the forecasting performance of regimeswitching, ARIMA and GArch models of house prices. Real Estate Economics, 32(2), 223–243. De Wet, M. C. (2020). Modelling the South African financial cycle. Johannesburg: University of Johannesburg (PhD thesis). Du Plessis, S. (2004). Stretching the South African business cycle. Economic Modelling, 21(4), 685–701. Enders, W. (2004). Applied econometric time series (2nd ed.). New Jersey: Wiley. Gujarati, D. N., & Porter, D. C. (2009). Basic econometrics (5th ed.). New York: McGraw Hill Inc.. Laubscher, P. (2019). Forecasting business cycles in South Africa. In S. Smirnov, A. Ozyildirim, & P. Picchetti (Eds.), Business cycles in BRICS. Societies and political orders in transition. Cham: Springer. Li, M.L., Lin, H.W. & Hsiu-Hua. (2005). The performance of the Markov-switching model on business cycles identification revised. Applied Economic Letters, 12, 513–520. Marcellino, M. (2005). Instability and non-linearity in the EMU. In C. Milas, P. Rothman, & D. van Djik (Eds.), Non-linear analysis of business cycles. Amsterdam: Elsevier. McQueen, G., & Thorley, S. (1993). Asymmetric business cycle turning points. Journal of Monetary Economics, 31, 341–362. Moolman, E. (2004). A Markov switching regime model of the south-African business cycle. Economic Modelling, 21, 631–646. Nelson, C. R., Piger, J., & Zivot, E. (2000). Markov regime-switching and unit root tests. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.248028 Nyberg, H. (2018). Forecasting US interest rates and business cycle with a nonlinear regime switching VAR model. Journal of Forecasting, 37(1), 1–15. Rech, G. (2002). Forecasting with artificial neural network models. SEE/EFI Working Paper Series in Economics and Finance, 491. Sarantis, N. (2001). Nonlinearities, cyclical behaviour and predictability in stock markets; international evidence. International Journal of Forecasting, 17, 459–482. Singh, T. (2012). Testing nonlinearities in economic growth in the OECD countries: And evidence from SETAR and STAR models. Applied Economics, 44, 3887–3908.

Forecasting the South African Financial Cycle: A Linear and Non-Linear Approach

29

Strohsal, T., Proano, C. R., & Wolters, J. (2019). Characterizing the financial cycle: Evidence from a frequency domain analysis. Journal of Banking & Finance, 106, 514–526. Tastan, H., & Yildirim, N. (2008). Business cycle asymmetries in Turkey: An application of Markov-switching autoregressions. International Economic Journal, 22(3), 315–333. Teräsvirta, T., Van Dijk, D., & Medeiros, M. C. (2005). Linear models, smooth transition autoregressions, and neural networks for forecasting macroeconomic time series: A re-examination. International Journal of Forecasting, 21, 755–774. Tkacz, G. (2001). Neural network forecasting of Canadian GDP growth. International Journal of Forecasting, 17, 57–69. Wai, P. S., Kun, S. S., Ismail, M. T., & Karim, S. A. A. (2015). Model performance between linear vector autoregressive and Markov switching vector autoregressive models on modelling structural changes in time series data. In International Symposium on Mathematical Sciences and Computing Research (pp. 379–383).

From Clubs to Communities. From Tourists to International Friends. Crisis Legacy in Music Organizations with Revenue Management and Relationship Marketing Angela Besana and Annamaria Esposito

Abstract The prompt and cost-effective segmentation of audiences and stakeholders is, today, as essential as revenue diversification in US symphony orchestras and opera houses. The lack of resources was particularly heavy during crisis years and as from 2008. Fundraisers of these music organizations engaged both with clubs and communities. At the same time, marketing officers explored new audiences and their segmentation. Relationship marketing was a pivotal strategy, in order to enhance stakeholders’ engagement and loyalty. The crisis legacy allowed these organizations to survive with revenue management and relationship marketing. The purpose of this study is a profiling of a sample of 120 USA symphony orchestras and opera houses, with different marketing and fundraising. Thanks to a k-means cluster analysis of diversified revenues, expenses and gains from 2008 to 2015, the paper will separate this sample into two poles, according to average variations of economic performances and to the focus on relationship marketing in the whole period. One pole grew as concerns relationships, revenue management and diversification. The other pole was affected by diminishing intensity of marketing and increasing fundraising and, as a consequence, retrenchment of some revenues except for contributions. Relationship marketing was in this cluster supported by volunteers. This pole profited by the highest increase in gains. Keywords Economics · Marketing · Classical music · USA · Cluster analysis

A. Besana () · A. Esposito Dipartimento di Business, Law, Economics and Consumer Behaviour Business, Diritto, Economia e Consumi, International University of Languages and Media, Libera Università di Lingue e Comunicazione IULM, Milan, Italy e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_2

31

32

A. Besana and A. Esposito

1 Introduction Since 2008, the beginning of the latest financial and real crisis, US symphony orchestras and opera houses have struggled to survive in a very competitive scenario, with revolutionary implications for their strategies. During the crisis, marketing and fundraising have not always revealed themselves as efficient strategies in a very uncertain climate. Revenues have been falling and revenue diversification has been not easily implemented and it has been often eluded. Single ticket and group sales have not fully compensated the drop of subscriptions (Besana, 2012; Besana & Esposito, 2019; Pompe, Tamburri, & Munn, 2019; Voss, Voss, Yair, & Lega, 2016). Besides, contributed and investment incomes have not easily recovered after years of fluctuations. As a consequence, orchestras and opera houses have been innovatively thinking about their marketing and fundraising, with attention to new segments and stakeholders and with a different and versatile implementation thanks to social media. Since these hard times, the audience development has concerned both local communities and tourists on the marketing side (Besana & Esposito, 2019; Poon & Lai, 2008). On the fundraising side, donors’ exploitation has not more concerned clubs, corporations and grant-making foundations, and also national and international friends (Cancellieri & Turrini, 2016; Kemp & Poole, 2016; Pompe & Tamburi, 2016). Orchestras and opera houses usually engage with their local communities thanks to education and entertainment programmes: performances, musical activities, rehearsals and concerts in offices halls of private and public buildings and other events, deepening the experience of orchestral music and music education for communities who would typically not otherwise engage with the music organization (League of American Orchestras, 2009; Ravanas, 2007, 2008; Tamburri, Munn, & Pompe, 2015). Tourists are included in the audience development, especially as for guided tours in mostly well-known North American cities, with an ad hoc marketing of bundles of attractions and hotels (Besana & Esposito, 2019). As a crisis legacy for revenue management and diversification, the ‘public good content’ has been stressed by the Fundraiser, asking for private philanthropy, worldwide sponsorships and donations, the ‘creative and experience content’ has been stressed by the Marketing Expert, instead, at the best supported by ICTs and social media. Fundraising may be thought full-grown and mature, as any segments of foundations (from corporate to community, from independent to family ones), corporate donors and philanthropists, they have been cashed for decades. While grants, donations, and sponsorships, they are the main share of revenues (Besana, 2012; Besana & Esposito, 2019), international friends might be the frontier of fundraising, whose goals can match with marketing ones, when the tourist can get on to the international friend. Above all, social media have revealed themselves as leading and innovative tools for more than one decade, in order to increase both marketing-oriented and

From Clubs to Communities. From Tourists to International Friends. Crisis. . .

33

fundraising-oriented audiences and segment them (Esposito, 2016; Roberts, 2014; Van Bree, 2009). Thanks to a k-means cluster analysis of diversified revenues, expenses and gains from 2008 to 2015, the paper will separate this sample into two poles, according to average variations of economic performances and to the focus on relationship marketing in the whole period. One pole grew as concerns relationships, revenue management and diversification. The other pole was affected by diminishing intensity of marketing and increasing fundraising and, as a consequence, retrenchment of some revenues except for contributions. Relationship marketing was in this cluster supported by volunteers. This pole profited by the highest increase in gains.

2 Relationship Marketing in US Opera Houses and Symphony Orchestras: From Engagement to Loyalty Inside and Outside Relationship marketing is a strategy designed to foster stakeholders’ engagement and loyalty (Berry, 1995; Christopher, Payne, & Ballantyne, 1991) as it is essential and indispensable to deal with different groups of stakeholders (Peck, 1996) and with the wide range of connections between these groups (Christopher et al., 1991; Doyle, 1995; Gummesson, 1996). During the latest crisis, diminishing resources push music organizations to differentiate, segment and maximize their relations with past and new stakeholders, with the commitment of employees, whose roles are clearly separated, either for marketing or fundraising. Without a trade-off in the allocation of resources for marketing and fundraising, fundraisers and marketing officers must focus on building effective relationships among and with staff and volunteers, in order to enhance their engagement and ultimately to improve their performance (Bussell & Forbes, 2006; Kumar & Pansari, 2016). Volunteers, coming from different sectors of civic society, can act as fundraisers, soliciting grants and contributions of money and goods and services from potential donors. They can also be in charge of some general activities related to the accomplishment of the organization mission. On the one hand, for instance, orchestras and opera houses can offer to participate in the effort to improve lives through programmes, on the other hand, staff and volunteers can deliver ushering, ticket taking and other services, administrative support in the office; organizing and executing fundraising events and special events (like anniversaries, dinners with main conductors, videomaking of concert halls); promoting main and ancillary activities to their friends, and in addition they could be involved in facilitating experiences that make performances truly memorable. Managing relationship marketing (Berry, 1995; Das, 2009) is important to define strategies able to attract, maintain, and enhance long-lasting relationships with paid and especially unpaid staff over time. In fact, relationship can become the salient

34

A. Besana and A. Esposito

attribute linking staff and volunteers to the organization and, in the same time, the pillar of the original motivation of volunteers (Arnett, German, & Hunt, 2003). From this standpoint, relationship marketing allows orchestras and opera houses to understand who staff and volunteers are and what drives them in their activities, allowing organizations to identify the strategies more appropriate to manage each of them. In addition, relationship marketing helps to meet the needs of stakeholders, to provide them with information directly suited to their interests, values and visions about the organizations (Bussell & Forbes, 2006, 2007), and to spread an inside-awareness of the social values which lead the achievement of the mission (Andreasen, Kotler, & Parker, 2003). Adopting a marketing approach allows non-profit to generate stakeholders’ trust and commitment (Hussain, Rawjee, & Penceliah, 2014), and to exploit strategically valuable sustainable resources and capabilities, among employees and volunteers. Furthermore, according to Scholars (Colbert, 2001; Hill, O’Sullivan, & O’Sullivan, 1995; Radbourne & Fraser, 1996), the more the organization learns about and monitors the stakeholders’ needs, preferences, attitudes and concerns, the more their satisfaction and commitment levels grow. In the same direction, if the mission is properly internally communicated, it can intercept the stakeholders’ needs and commitment, leading orchestras and opera houses to reach sustainable management.

3 Method 990 Forms of the fiscal years 2015 and 2008 are analysed for the IRS—USA Internal Revenue Service—category ‘A69-Symphony Orchestra’ and ‘A6A-Opera’: 100 organizations for every category, from the highest to the lowest total income. These reports can be downloaded from the Guidestar website www.guidestar.org and the main websites of organizations themselves.1 The sample sums up to 158 organizations, whose available 990 Forms were downloadable at www.guidestar.org or their websites. According to 990 Forms Glossary and Accounting Standards, contributions for US not-for-profit organizations include the direct public support from individuals, grant-makers, foundations, sponsors like corporations and the Government grant for projects of public interest. In the accounting lines of revenues, contributions can be summed up with the programme service revenue, which is money for sold and rendered services. Contributions and programme service revenue are 85% of total revenues of the here-investigated sample. Next to them, ancillary revenues come from interests and

1 The

Guidestar website collects 990 Forms of USA Not-For-Profits. Not-For-Profits are listed for the relevancy to the keyword. 990 Forms report Statements of Revenues and Expenses and Financial Statements of USA Not-For-Profits.

From Clubs to Communities. From Tourists to International Friends. Crisis. . .

35

gains of financial assets, sales of assets, rental income, fundraising from special events and other residual revenues. The composition of revenues will be here investigated for 2015–2008 percent change of main categories: Contributions with the target of the willingnessto-donate, Programme Service Revenue with the target of willingness-to-pay, Investment Income and Other Revenue. Expense categories include the following: Programme Service Expense related to marketing and management of the core business; Fundraising expense and Management and general expense, a miscellaneous cost that is not related to the previous accounting lines. Next to revenue and expense categories, the (Net) Gain or Loss of the year as the difference—positive or negative between revenues and costs is also here investigated. For expense categories and gains (or losses), 2015–2008 percentage changes will be, at the same time, calculated in order to focus on trends of economic performances during and soon after the crisis times. In order to gain the impact of employees (like fundraisers and marketing officers) and volunteers on economic performances, the ratio of volunteers/employees was added as concerns 2015s data. At the end of the crisis, this ratio is meaningful in order to show how much volunteers and employees have led, and they are leading engagement and loyalty from the inside of the organization, fundraising and marketing from the outside of the organization. All the monetary data are, first of all, filed in Excel, and 2015–2008 percentage variations are calculated for main items: programme service revenues, contributions, investment income, other revenue, programme service expense, management and general expense, fundraising expense, gain or loss. Secondly, together with volunteers’/employees’ ratios, these variations are clustered in order to obtain meaningful groups with relevant and separating features. We have adopted the K-means clustering as an iterative follow-the-leader strategy.

4 Key Findings and Discussion: The Fundraiser and the Revenue Manager on the Stage of US Classical Music K-means clustering of the above-mentioned sample of 158 is significant for 120 organizations, which are divided into two clusters. Average performances of two clusters are shown in Table 1. Membership of clusters is reported in Appendix. The most crowded cluster (74 organizations) is the Fundraiser who shows a very important increase in the fundraising expense, +17.45% and as a consequence, in contributions, +18.42%. Save for a very modest increase in the programme service expense (+0.87%), any other item is decreasing. Gains are consistently increasing,

36

A. Besana and A. Esposito

Table 1 The crisis legacy for 120 USA symphony orchestras and opera houses (average 2015– 2008% change) Contributions Programme service revenue Investment income Other revenue Programme service expense Management and general expense Fundraising expense Gain or loss Volunteers/employees

Fundraiser −74 organizations +18.49 −13.81 −28.20 −31.70 +0.87 −11.94 +17.45 +46.69 9.15

Revenue manager −46 +12.82 +21.99 +75.85 +50.69 +13.26 +44.82 +40.17 +9.50 1.04

Source: Elaboration with Jump Statistics Software

+46.69%. Volunteers are here essential, nine times employees. They play the role of fundraisers, calling for donations, sponsorships and grants. Social media are mature gatekeepers, in order to promote special events, campaigns, community empowerment. Communities are engaged with plentiful programmes in private and public buildings where orchestra and opera officers tell their histories, seasons, special events and where rehearsals, edutainment and fundraising campaigns take place. Main organizations of big cities like Chicago, Dallas and Los Angeles are included in this cluster. Nevertheless, middle-sized and small American towns are included, too, where performances are together with lunch, dinner, coffee and ‘gelato’ timing, behind the scenes, with intimate concerts, music and wellness programmes. Revenue diversification is significant in the second cluster, the Revenue Manager. The investment income increase of +75.85% is here matching with increasing other revenue, programme service revenue and contributions, these ones not so high as in the other pole of the Fundraiser. Expenses are increasing and so are gains, but not with the same percentage of the Fundraiser. Some of these organizations count on employees, who show proficiency in investing in financial markets as well as partnershiping with donors’ clubs and international friends worldwide. This cluster includes giants like the Met. Several middle-sized and small opera houses and symphony orchestras are here included, whose halls see music travellers of different music genres and whose social media marketing is well-developed in order to engage citizens and tourists, donors, clubs, sponsors and international supporters. Plan Your Visit is provided with information about parking, accessibility, hotels, attractions, programmes notes and comments. Tourists are not more a frontier for marketing in these organizations.

From Clubs to Communities. From Tourists to International Friends. Crisis. . .

37

5 Conclusion Today marketing and fundraising of US symphony orchestras and opera houses, they both include diversified tactics and strategies: audiences and philanthropists are investigated as for their willingness-to-pay and willingness-to-donate. Tourists do not more represent the frontier of their flexible subscriptions. Social media are levers of all their audiences and stakeholders. As a matter of fact, thanks to a price strategy that implies discrimination both for audiences and philanthropists, symphony orchestras and opera houses are connecting with communities, and they are emphasizing relationships, whose performances are particularly meaningful for the Fundraiser cluster. When frontiers of marketing are open to new, innovative and international segmentation and they include tourists, the Revenue Manager is the prevailing profile. Considering the latest 10 years, it can be confirmed that marketing is as essential as fundraising. Revenue management implies the key consideration that fundraising performances can support or compensate marketing ones, especially when marketing of the place is maximizing occupancy of the houses. Focus on several and multiple stakeholders is the main objective, and marketing is separated from fundraising so that the location can maximize occupancy and revenues. Research limitations refer, first of all, to the opportunity of the here-investigated organizations to develop new targets like tourists, when US destinations are very different as for tourism: some of them are very attractive as they refer to main opera houses and concert halls like the Met and Carnegie in New York, while some of them remain attractions for business travellers, who are not mainly concerned with opera and classical music as the most important motivation of their journey. Secondly, the here-investigated period was a matter of a real and financial crisis, whose implications hit the whole US economy in spite of skills and marketing efforts of opera and symphony managers. The general lack of resources was deeply affecting all not-for-profit organizations. Managerial implications imply that managers of these music organizations should continually stress the importance of the selection of new targets. Above all, their segmentation should, at the same time, exploit their willingness-to-pay on the fundraising side (international friends) and on the marketing side (tourists). Social media will facilitate these segmentation and exploitation as they result efficient and pervasive communication channels.

A.1 Appendix A.1.1 Cluster Fundraiser ARIZONA OPERA COMPANY—PHOENIX BALTIMORE OPERA COMPANY INC—BALTIMORE

38

A. Besana and A. Esposito

BANGOR SYMPHONY ORCHESTRA—BANGOR BOSTON LYRIC OPERA COMPANY—BOSTON CALIFORNIA SYMPHONY ORCHESTRA INC—WALNUT CREEK CANTON SYMPHONY ORCHESTRA ASSOCIATION—CANTON CHEYENNE SYMPHONY SOCIETY INC—CHEYENNE CHICAGO SINFONIETTA—CHICAGO CHICAGO SYMPHONY ORCHESTRA—CHICAGO CINCINNATI OPERA ASSOCIATION INC—CINCINNATI CINCINNATI SYMPHONY ORCHESTRA—CINCINNATI COBB SYMPHONY ORCHESTRA—MARIETTA DALLAS SYMPHONY ASSOCIATION—DALLAS DES MOINES METRO OPERA INC—INDIANOLA DETROIT SYMPHONY ORCHESTRA INC—DETROIT ERIE PHILHARMONIC INC—ERIE FAIRFAX SYMPHONY ORCHESTRA—FAIRFAX GREAT FALLS SYMPHONY ASSOCIATION INC—GREAT FALLS GREATER AKRON MUSICAL ASSOCIATION INC—AKRON GREENSBORO SYMPHONY ORCHESTRA INC—GREENSBORO JACKSONVILLE SYMPHONY ASSOCIATION INC—JACKSONVILLE JOHNSTOWN SYMPHONY ORCHESTRA—JOHNSTOWN LANCASTER SYMPHONY ORCHESTRA—LANCASTER LENAWEE SYMPHONY ORCHESTRA SOCIETY INC—ADRIAN LONG BEACH SYMPHONY ASSOCIATION—LONG BEACH LOS ANGELES OPERA COMPANY—LOS ANGELES LYRIC OPERA OF CHICAGO—CHICAGO MADISON OPERA—MADISON MEMPHIS ORCHESTRAL SOCIETY INC—MEMPHIS MOBILE SYMPHONY INC—MOBILE MUSIC CENTER OF SOUTH CENTRAL MI—BATTLE CREEK NASHVILLE OPERA ASSOCIATION—NASHVILLE NASHVILLE SYMPHONY ASSOCIATION—NASHVILLE NEVADA OPERA ASSOCIATION—RENO NEW HAVEN SYMPHONY ORCHESTRA INC—NEW HAVEN NEW JERSEY SYMPHONY ORCHESTRA—NEWARK NEW MEXICO SYMPHONY ORCHESTRA—ALBUQUERQUE NEW ORLEANS OPERA ASSOCIATION—NEW ORLEANS NEW YORK CITY OPERA INC—NEW YORK NEWBERRY OPERA HOUSE FOUNDATION—NEWBERRY NORTH ARKANSAS SYMPHONY ORCHESTRA—FAYETTEVILLE OMAHA SYMPHONY ASSOCIATION—OMAHA OPERA BIRMINGHAM—BIRMINGHAM OPERA COLORADO—DENVER OPERA OMAHA—OMAHA OPERA SAN JOSE INCORPORATED—SAN JOSE OPERA SOUTHWEST—ALBUQUERQUE

From Clubs to Communities. From Tourists to International Friends. Crisis. . .

39

OREGON SYMPHONY ASSOCIATION—PORTLAND PALM BEACH OPERA INC—WEST PALM BEACH PENSACOLA OPERA INC—PENSACOLA POCKET OPERA INC—SAN FRANCISCO PORTLAND OPERA ASSOCIATION—PORTLAND ROANOKE SYMPHONY ORCHESTRA—ROANOKE SACRAMENTO PHILHARMONIC ORCHESTRA ASSOCIATION INC— SACRAMENTO SAINT LOUIS SYMPHONY ORCHESTRA—SAINT LOUIS SAN ANTONIO OPERA—SANT ANTONIO SAN DIEGO SYMPHONY ORCHESTRA ASSOCIATION—SAN DIEGO SANTA BARBARA SYMPHONY ORCHESTRA ASSOCIATION—SANTA BARBARA SARASOTA OPERA ASSOCIATION INC—SARASOTA SEATTLE OPERA—SEATTLE SHEBOYGAN SYMPHONY ORCHESTRA INC—SHEBOYGAN SPRINGER OPERA HOUSE ARTS ASSOCIATION INC—COLUMBUS STOCKTON SYMPHONY ASSOCIATION INC—STOCKTON SYRACUSE OPERA COMPANY INC—SYRACUSE THE ATLANTA OPERA INC—ATLANTA THE HENDERSONVILLE SYMPHONY ORCHESTRA INC—HENDERSONVILLE THE OPERA ASSOCIATION OF CENTRAL OHIO—COLUMBUS TOLEDO OPERA ASSOCIATION—TOLEDO TRAVERSE SYMPHONY ORCHESTRA—TRAVERSE CITY TULSA OPERA INC—TULSA VIRGINIA OPERA ASSOCIATION—NORFOLK WHATCOM SYMPHONY ORCHESTRA—BELLINGHAM WICHITA GRAND OPERA INC—WICHITA WILLIAMSPORT SYMPHONY ORCHESTRA—WILLIAMSPORT

A.1.2 Cluster Revenue Manager ALBANY SYMPHONY ORCHESTRA INC—ALBANY BERKELEY SYMPHONY ORCHESTRA—BERKELEY BOSTON YOUTH SYMPHONY ORCHESTRA INC—BOSTON BUFFALO PHILHARMONIC ORCHESTRA SOCIETY INC—BUFFALO CHATTANOOGA SYMPHONY AND OPERA ASSOCIATION—CHATTANOOGA DAYTON PHILHARMONIC ORCHESTRA ASSOCIATION—DAYTON DUBUQUE SYMPHONY ORCHESTRA—DUBUQUE EL PASO SYMPHONY ORCHESTRA ASSOCIATION INC—EL PASO

40

A. Besana and A. Esposito

EUGENE SYMPHONY ASSOCIATION INC—EUGENE FLORENTINE OPERA COMPANY INC—MILWAUKEE GLENS FALLS SYMPHONY ORCHESTRA INC—GLEN FALLS HAWAII OPERA THEATRE—HONOLULU HOUSTON GRAND OPERA ASSOCIATION INC—HOUSTON KALAMAZOO SYMPHONY ORCHESTRA—KALAMAZOO KANSAS CITY SYMPHONY—KANSAS CITY KENTUCKY OPERA ASSOCIATION—LOUISVILLE KNOXVILLE SYMPHONY SOCIETY INC—KNOXVILLE LYRIC OPERA OF KANSAS CITY—KANSAS CITY METROPOLITAN OPERA ASSOCIATION INC—NEW YORK MONTEREY COUNTY SYMPHONY ASSOCIATION INC—CARMEL OPERA AMERICA INC—NEW YORK OPERA CAROLINA—CHARLOTTE OPERA COMPANY OF PHILADELPHIA—PHILADELPHIA OPERA IN THE HEIGHTS—HOUSTON OPERA NORTH—LEBANON PEORIA SYMPHONY ORCHESTRA FOSTER ARTS CENTER—PEORIA PIEDMONT OPERA INC—WINSTON SALEM PITTSBURGH OPERA INC—PITTSBURGH PORTLAND MAINE SYMPHONY ORCHESTRA—PORTLAND QUAD CITY SYMPHONY ORCHESTRA ASSOCIATION—DAVENPORT RHODE ISLAND PHILHARMONIC ORCHESTRA & MUSIC SCHOOL— EAST PROVIDENCE SAN FRANCISCO OPERA ASSOCIATION—SAN FRANCISCO SANTA ROSA SYMPHONY ASSOCIATION—SANTA ROSA SEATTLE YOUTH SYMPHONY ORCHESTRAS—SEATTLE SHREVEPORT OPERA—SHREVEPORT SKYLIGHT OPERA THEATRE—MILWAUKEE SOUTH BEND SYMPHONY ORCHESTRA ASSOCIATION INC—SOUTH BEND SYMPHONY SOCIETY OF SAN ANTONIO TACOMA OPERA ASSOCIATION—TACOMA THE LOUISVILLE ORCHESTRA INC—LOUISVILLE THE LOUSIANA PHILHARMONIC ORCHESTRA—NEW ORLEANS THE MINNESOTA OPERA—MINNEAPOLIS THE STAMFORD SYMPHONY ORCHESTRA INC—STAMFORD UTAH FESTIVAL OPERA COMPANY—LOGAN WEST SHORE SYMPHONY ORCHESTRA—MUSKEGON WOODLAND OPERA HOUSE INC—WOODLAND

From Clubs to Communities. From Tourists to International Friends. Crisis. . .

41

References Andreasen, A. R., Kotler, P., & Parker, D. (2003). Strategic marketing for non-profit organizations. Upper Saddle River: Prentice Hall. Arnett, D. B., German, S. D., & Hunt, S. D. (2003). The identity salience model of relationship marketing success: The case of nonprofit marketing. Journal of Marketing, 67(2), 89–106. Berry, L. L. (1995). Relationship marketing of services—Growing interest, emerging perspectives. Journal of the Academy of Marketing Science, 23(4), 236–245. Besana, A. (2012). Alternative resources: Revenue diversification in the not-for-profit USA symphony orchestra. The Journal of Arts Management, Law and Society, 42(2), 79–89. Besana, A., & Esposito, A. (2019). Fundraising, social media and tourism in American symphony orchestras and opera houses. Business Economics, 54, 137–144. Bussell, H., & Forbes, D. (2006). Developing relationship marketing in the voluntary sector. Journal of Nonprofit & Public Sector Marketing, 15(1–2), 151–174. Bussell, H., & Forbes, D. (2007). Volunteer management in arts organizations. International Journal of Arts Management, 9(2), 17–28. Cancellieri, G., & Turrini, A. (2016). The phantom of modern opera: How economics and policies affect the programming strategies of opera houses. International Journal of Arts Management, 18(3), 25–36. Christopher, M., Payne, A., & Ballantyne, D. (1991). Relationship marketing. Oxford: Butterworth Heinemann. Colbert, F. (2001). Marketing culture and the arts. Montreal: HEC Press. Das, K. (2009). Relationship marketing research (1994-2006): An academic literature review and classification. Marketing Intelligence & Planning, 27(3), 326–363. Doyle, P. (1995). Marketing in the new millennium. European Journal of Marketing, 29(13), 23– 41. Esposito, A. (2016). Web e social media per il fundraising. In A. Besana & A. Esposito (Eds.), Economia, marketing e comunicazione per il nonprofit (pp. 149–170). Milan: McGraw-Hill. Gummesson, E. (1996). Mega and non relationships in relationship marketing. Irish Marketing Review, 9, 9–16. Hill, E., O’Sullivan, C., & O’Sullivan, T. (1995). Creative arts marketing. Oxford: Butterworth Heinemann. Hussain, S. B., Rawjee, V. P., & Penceliah, S. (2014). The applicability of relationship marketing at non-profit organizations: A developing country perspective. Problems and Perspectives in Management, 12(4), 155–160. Kemp, E., & Poole, S. M. (2016). Arts audiences: Establishing a gateway to audience development and engagement. The Journal of Arts Management, Law and Society, 46(2), 53–62. Kumar, V., & Pansari, A. (2016). Competitive advantage through engagement. Journal of Marketing Research, 53(4), 497–514. League of American Orchestras. (2009). Audience demographic research review. Retrieved July, 2017, from www.americanorchestras.org Peck, H. (1996). Towards a framework for relationship marketing—The six markets model revisited and revised, in 2001—A vision for the next 25 years. In Proceedings of Marketing Education Group Annual Conference. University of Strathclyde. Pompe, J., & Tamburi, L. (2016). Fiddling in a vortex: Have American orchestras squandered their supremacy on the American cultural scene? The Journal of Arts Management, Law and Society, 46(2), 63–72. Pompe, J., Tamburri, L., & Munn, J. (2019). Marketing strategies for performing arts audiences: Characteristics of ticket purchasers. Journal of Nonprofit and Public Sector Marketing, 32(5), 453–464. Poon, J. P. H., & Lai, C. A. (2008). Why are non-profit performing arts organisations successful in mid-sized US cities? Urban Studies, 45(11), 2273–2289. Radbourne, J., & Fraser, M. (1996). Arts management—A practical guide. Sydney: Allen & Unwin.

42

A. Besana and A. Esposito

Ravanas, P. (2007). A quiet revolution: The metropolitan opera reinvents client relations management. International Journal of Arts Management, 9(3), 79–87. Ravanas, P. (2008). Hitting a high note: The Chicago symphony orchestra reverses a decade of decline with new programs. New services and new prices. International Journal of Arts Management, 10(2), 68–78. Roberts, N. (2014). Social media for orchestra. Association of California Symphony Orchestras. Retrieved from https://www.acso.org/acso%20socmedia%20wkshop_final.pdf Tamburri, L., Munn, J., & Pompe, J. (2015). Repertoire conventionality in major US symphony orchestras: Factors influencing management’s programming choices. Managerial and Decision Economics, 36, 97–108. Van Bree, M. (2009). Orchestras and new media: A complete guide. Communications, social media, culture. Retrieved from http://mcmvanbree.com/projects Voss, Z. G., Voss, G. B., Yair, K., & Lega, K. (2016). Orchestra facts: 2006–2014. Retrieved July, 2017, from www.americanorchestras.org

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real Options: Amazon’s Acquisition of Whole Food ˇ Andrejs Cirjevskis

“Amazon buying Whole Foods is incredibly interesting, highly strategic, and definitely not standard” Josh Chapman, Toptal Finance Expert (Clarence-Smith, 2020)

Abstract Acquisition-based dynamic capabilities have become well established as a new imperative for organizing M&A processes. However, understanding the full benefits and possible limits of real options applications to measure a dynamic capability-based (managerial) synergy remains a challenge. The paper draws on real options theory to describe some of these benefits and limits to value a synergy in highly strategic and not standard M&A deals. The acquisition of Whole Foods by Amazon makes it possible to combine two streams of research on dynamic capabilities and real options in a cohesive whole. More specifically, the author develops three propositions to justify the role of dynamic capabilities as antecedents of success or failures of M&A deals and to demonstrate real options application to measure synergies of M&A deals. Keywords Merger and acquisition · Dynamic capabilities · Synergy · Real options

1 Introduction: Purpose, Motivation, and Originality “Synergies do not magically materialize. By definition, they are possibilities, not certainties” (Ficery, Herd, & Pursche, 2007, p. 35). While there is some evidence of synergy in the aggregate across all acquisitions, most mergers fail in delivering any

ˇ A. Cirjevskis () RISEBA University of Applied Sciences in Business, Arts and Technology, Riga, Latvia e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_3

43

44

ˇ A. Cirjevskis

synergy (Damodaran, 2005, p. 47). This paper aims to justify the role of dynamic capabilities as antecedents of success or failures of M&A deals and to demonstrate real options application to measure managerial synergies in M&A deals. In the current paper, the author argues that the intersection between dynamic capabilities frameworks and real options theory can shed light on the antecedents of successes and failures of M&A deals. The interaction between dynamic capabilities and real option valuation enables the acquirer to elect and exercise those options that have a high probability to provide managerial synergies and let expire the options that have low probability. The paper develops three propositions as follows. The probability to exercise a real option in the M&A deal can be measured by exploring similarities and complementarity of the dynamic capabilities of acquirers and targets. The managerial synergies are provided by the successful integration of the dynamic capabilities of an acquirer and a target. Such type of synergy can be assessed and measured by real option application. The motivation for this research is as follows. First, the majority of papers on the synergetic effects of M&A deals typically focus on a particular type of synergy (Loukianova, Nikulin, & Vedernikov, 2017), while the current paper proposes a model that accounts for the cumulative simultaneous effect of different types of operating, financial, and dynamic capabilities-based synergies. Second, even though the dynamic capabilities framework and its empirical applications (Capron & Anand, 2007; Teece, 2007, 2011) make dynamic capabilities more visible, the real option application making dynamic capabilities measurable in the M&A deals. The originality of this research is an application of the real option pricing theory to recent Amazon’s acquisition of Whole Foods to measure the synergies as an added value to the acquirer’s shareholders. The author selected Amazon’s acquisition of Whole Food due to following reasons. The synergy is reflected in additional value created by unifying the companies. For the M&A deal to be successful, this value of a newly merged company should be larger than the value of the stand-alone companies before M&A (Loukianova et al., 2017). So, what does Amazon hoped to gain with this merger? This paper analyzes Amazon acquisition’s antecedents through the lenses of dynamic capabilities framework and real options theory. The paper has the following structure. The first section introduces the concept of dynamic capabilities as antecedents of successful M&A deals, synergies that arise from an M&A deal, and discusses the applicability of the real options approach for their assessment. The role of dynamic capabilities in the M&A deal is discussed in terms of abilities to integrate two merging companies in search of synergies. The sections are devoted, respectively, to develop three propositions which can be justified empirically by analysis of recent Amazon’s acquisition of Whole Food case. The following section provides an application of the real options theory to Amazon’s dynamic capabilities in the case of Whole Food acquisition. The method was used ex-post to find synergy values in a recent Amazon’s M&A deals (2017– 2018) and produced sound results. On average, the author found that the option premiums exceeded the actual takeover premium suggesting that, from an option pricing point of view, this

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real. . .

45

acquisition was not overpaid. At the end of the paper, the author discusses theoretical and managerial contributions. In conclusion subchapter, the author highlights the research limitations and future works.

2 Key Literature Review Dynamic capabilities are the renewing and regenerative capabilities that enable firms to change their operating processes incrementally and radically. Real options valuation provides an appropriate platform for firms to measure managerial flexibilities. “The two distinct concepts of dynamic capability and real options have received notable attention from strategic management scholars in recent years. These two streams of research in strategic management literature are certainly not mutually exclusive” (Jahanshahi & Nawaser, 2018, p. 395). Nevertheless, there are many differences between real options theory and dynamic capability framework like the difference in the origin, in the aims, and in the context of usage, there are many similarities within two concepts. Both are necessary for managing changes, both are created by managers, and both are new and growing concepts (Jahanshahi & Nawaser, 2018). Dynamic capabilities are necessary to exploit real options opportunities, whereas real options are necessary to evaluate opportunities (Jahanshahi & Nawaser, 2018).

2.1 Exploring Dynamic Capabilities in Merger and Acquisition Deals The recent scientific discussion in the field of strategic management broadly favors the idea of dynamic capabilities to overcome potential rigidities of organizational capability building (Schreyogg & Kliesch-Eberl, 2007). “The theoretical and practical importance of developing and applying dynamic capabilities to sustain a firm’s competitive advantage in complex and volatile external environments has catapulted this issue to the forefront of the research agendas of many scholars” (Zahra, Sapienza, & Davidsson, 2006, p. 917). Stefano et al. argue that despite the exceptional rise in interest and influence of dynamic capabilities, criticisms of the dynamic capabilities’ perspective continue to mount (Stefano, Peteraf, & Verona, 2014). Common concerns are related to a lack of consensus on basic theoretical elements and limited empirical progress (Stefano et al., 2014). Specific capabilities that have been identified and studied involve research and development (Helfat, 1997), product innovation (Danneels, 2002), ambidextrous organizational structures (O’Reilly & Tushman, 2013), network responsiveness (Kleinbaum & Stuart, 2014), and human capital management (Chatterij & Patro, 2014).

46

ˇ A. Cirjevskis

However, there are only a few pieces of research on specific dynamic capabilities that have been identified and studied involving mergers and acquisitions. Teece argues that it might be “because assets are bundled together often tightly linked inside incumbent firms, it may be difficult to obtain assets in the desired configurations through asset purchase or sale in mergers and acquisitions” (Teece, 2007). What is more, there is no consensus, how to measure synergies in merger and acquisition deals created by dynamic capabilities. “Studies give clear empirical evidence that complementarities are a significant factor for M&A success” (Bauer & Matzler, 2014, p. 272). Through the interaction of complementary characteristics, value creation does not only derive from cost savings, but the value is also created by a growing turnover and market share thanks to dynamic capabilities (Kleinbaum & Stuart, 2014). Complementarity has been studied in terms of top management team complementarity (Kleinbaum & Stuart, 2014), technological complementarity (Makri, Hitt, & Lane, 2010), strategic and market complementarity (Kim & Finkelstein, 2009), or product complementarity (Wang & Zajac, 2007). However, the study in terms of complementarity of dynamic capabilities in M&A is still waiting for researchers. It is especially true for an application of real options to measure added value created by dynamic capabilities. In the recent ˇ publications, Cirjevskis argues and demonstrates (2017, 2019) that a dynamic capabilities framework (Teece, 2007, 2011) is useful to business analyses of M&A deal to identify similarities and complementarity between the dynamic capabilities of an acquirer and a target. Therefore, Proposition 1 The higher the degree of similarities and complementarity between the dynamic capabilities of an acquirer and a target, the higher the probability to exercise of a real option on an acquisition of this target.

2.2 Exploring Synergies in M&A as Market Value-Added A combined company can achieve synergistic benefits by generating economies of scale and scope through assets consolidation, combining sales operations, sharing information, distribution channels, and eliminating redundant operation sources (Alhenawi & Krishnaswani, 2015; Capron, 1999). Although synergies have been under intense interest and study for decades, there is still no common ground on what appropriate way for categorizing synergy items. Trautwein’s (1990) efficiency theory distinguishes three main categories of synergies: operational, financial, and managerial. Managerial synergies refer to gains that the bidder can achieve in a situation in which the acquiring company’s management has super knowledge and acquisitionbased capabilities (Bosecke, 2009, p. 27; Trautwein, 1990). These knowledge and acquisition-based capabilities can be hugely advantageous regarding the future of

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real. . .

47

the combined company and vital for acquirer management (Goold & Campbell, 1998). Capron and Anand (2007) named those as acquisition-based dynamic capabilities. In this vein, dynamic capabilities (superior knowledge and capabilities of the acquiring company’s management) can generate dynamic capabilities-based (managerial) synergies. Synergies in an acquisition are a function of strategic similarity, complementarities, and transferability of dynamic capabilities in the M&A deals. Merging companies generate managerial synergies by working closely together and executing tasks through an iterative knowledge-sharing process. However, there is no single way how to identify, validate, and value the potential of dynamic capabilities-based synergy, and valuation can be done in using different approaches. If the acquirer wants to ensure a successful value creation process, the application of appropriate measurement tools is essential. In recent publications, ˇ scholars (Cirjevskis, 2017, 2019) provided the practice-driven model that bridges the dynamic capabilities framework with building blocks of the business model canvas (Osterwalder & Pigneur, 2009). The presented methodology is encouraging to analyze the importance and strengths of acquisition-based dynamic capabilities and to measure the degree of similarities, complementarity, and transferability of dynamic capabilities of an acquirer and a target that is useful to the current research. Thus, Proposition 2 Managerial synergies in M&A deals are provided by the degree of similarities, complementarity, and transferability of the dynamic capabilities of an acquirer and a target.

2.3 Measuring Dynamic Capabilities-Based Synergies in M&A with a Real Option To account for managerial flexibility connected with an M&A deal that is reflected in different future potential strategic alternatives, several authors (see, e.g., Baldi & Trigeorgis, 2009) have proposed embedding a real options perspective in the valuation framework. Acquisitions sometimes open up possibilities that would not have been available otherwise, and these opportunities are difficult to convert into expected cash flows. The real options argument is heavily dependent upon two concepts—the learning that occurs by being in a new market and the more informed decisions that flow from the learning (Damodaran, 2005). Thus, there are various strategic managerial real options embedded in mergers and acquisitions. Real options analysis provides a technique for incorporating and valuing synergies that generate by dynamic capabilities in mergers and acquisitions. In this vein, the value of synergies of dynamic capabilities-based synergies in M&A can be measured with a real options valuation technique. The incorporation of real options into the synergy valuation measures managerial flexibility arising from M&A deals (Loukianova et al., 2017). Dunis and Klein (2005) argue that synergies can be viewed as a real option value employing input variables for the

48

ˇ A. Cirjevskis

European and/or American call option with Black Scholes Option Pricing Model and/or accordingly with Binominal Option pricing model. The share price (So) is proxied by the sum of capitalization of merging companies before the deal’s announcement. The exercise price (E) is proxied by the sum of the future market capitalization of merging companies in 1 year if the merger would not be consummated. Thereby, the future capitalization of two separate companies can be calculated employing a discounted free cash flow method. Cash flow is, in theory, the free cash flow, but in practice, it is proxied by EBITDA. Therefore, the exercise price is the hypothetical future market value without the merger or theoretical market value calculated by using revenue and EBITDA multiples. The volatility (σ ) of share price can be obtained from the V-Lab APARCH Volatility Analysis (NYU Stern, 2019) or by direct observation. Assuming semiefficient markets that incorporate publicly available new information promptly, the calculation of the standard deviation of the acquirer stock price return was started the week after the announcement. Duration (T) getting synergy is managerial anticipation of when dynamic capabilities-based synergies would be fully realized in terms of the year following completion of the merger or acquisition. For time to maturity, 1 year was assumed for the deal of Amazon-Whole Food. This was due to data availability and the assumption that efficient markets should have well anticipated potential long-term merger gains within this period even if accounting data might not reflect any benefits in this short period due to integration costs. The US dollars was chosen as the reference currency for Amazon Whole Food. The risk-free rate (rf) is a long-term government bond yield of an acquirer’s country (Dunis & Klein, 2005). Therefore, Proposition 3 Dynamic capabilities-based synergies in M&A deals can be measured by real options application using BSOPM and BOPM. To test the internal and external validity of the proposed propositions, it was applied to a recent case of dynamic capabilities-based M&A deal in the grocery retail industry: Amazon’s acquisition of Whole Food in 2017.

3 Illustrative Case Study Amazon’s Acquisition of Whole Food Most acquisitions are carried out to acquire these target firm’s capabilities; how is the Amazon acquisition of Whole Foods in 2017 different?

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real. . .

49

3.1 Illustration of Acquisition-Based Dynamic Capabilities of Amazon.com Justification of Proposition 1 The higher the degree of similarities and complementarity between the dynamic capabilities of an acquirer and a target, the higher the probability to exercise of a real option on an acquisition of this target. The persistence of existing dynamic capabilities depends on the impetus for change (sensing), the strength of the perceived need to change (seizing), and the managerial capacity to integrate and recombine resources (transforming) as desired (Teece, 2007; Zahra et al., 2006). Zahra et al. (2006) argue that the lack of success to solve a problem with current capabilities triggers the development and use or acquire new dynamic capabilities. The research has explored the selected dynamic capabilities of the target company and the acquirer’s company. There are several similarities between the dynamic capabilities of Amazon and Whole Food. Both companies are sensing market demands and seizing external opportunities. However, their transforming capabilities need to be mutually complemented. To grab for more grocery market share, Amazon should learn to sell food offline (Kowitt, 2018). On the other hand, having a variety of niche products with a high price charged, the growth of Whole Foods had slowed because competitors began to offer organic foods at a lower price. From 2013 to 2016 Whole Foods lost nearly half its market value (Helmore, 2017). That is why just a few days after the merger, Amazon dropped prices by as much as 43% of a range of Whole Food products (Garfield, 2017). Amazon has limited knowledge and experience in the offline retail environment. That is why, for Amazon Fresh to be successful, the company needed to acquire more expertise in perishable grocery procurement. It made the probability to exercise the real option of the acquisition of Whole Foods as very high. Justification of Proposition 2 Managerial synergies in M&A deals are provided by the degree of similarities, complementarity, and transferability of the dynamic capabilities of an acquirer and a target. How acquisition-based dynamic capabilities contribute to reduce cost, to create a new revenue stream, to deliver a new value proposition, and therefore provide a managerial synergy by adding market value-added of the acquirer? The acquisition-based dynamic capabilities helped Amazon to provide managerial synergies. Amazon sensed that Whole Foods would provide broad access to retail outlets in a great location across the USA. Amazon seized a high-end brand name of Whole Foods and affluent buyers of Whole Foods. The justification of managerial synergies in this deal is as follows. A grocery is a category that people buy almost every day. However, Whole Foods had only a 1.2% share in this market that was dominated by Walmart with a 14.5% share, and even Amazon could not get more than 0.2% market share in the last 10 years (Thomas, 2017). Having exploited big data strategy, Amazon can create a

50

ˇ A. Cirjevskis

daily habit among buyers to order groceries from its app and to make them loyal clients who are highly profitable for the corporation. It also can reduce Amazon’s supply chain management’s costs due to higher purchasing and bargaining power. With Whole Food and Go stores, Amazon operates 603 physical stores in 60 cities in the USA giving customer delivery in as far as an hour on the thousand organic products from Whole Foods (Redman, 2018). Having added grocery dynamic capabilities through the acquisition of Whole Foods, Amazon offers clickand-collect service for quality-conscious buyers, enjoys higher operating profit, and offers stores as points to pick up other online orders. Therefore, Amazon can transform its customer value proposition, deliver new value to the buyers of both companies, and capture new added value for shareholders. Justification of Proposition 3 Dynamic capabilities-based synergies in M&A deals can be measured by real options application using BSOPM and BOPM. On June 16, 2017, Amazon.com announced that it would purchase Whole Foods Market for a total of $13.7 billion. The capitalization of Amazon was $478.6 bn; the capitalization of Whole Food was $13.8 bn (Pillars of Wall Street, 2017). The exercise price (E) is the combined hypothetical future market value after 1 year without a merger. The hypothetical future market value of the separated entities (target and acquirer) after 1 year has been calculated using EV/Revenues (Enterprise Value) and EV/EBITDA (Enterprise Value/Earnings before Interest, Taxes, Depreciation, and Amortization) multiples. Having used Amazon revenues $142.6 bn in 2017 and EV/Revenues multiple 3.3 (Pillars of Wall Street, 2017), the hypothetical future market value of Amazon without the acquisition has been estimated as $ 470.6 bn. Having used Whole Food EBITDA $ 1.3 bn in 2017, and EV/EBITDA multiple 11.1 (Pillars of Wall Street, 2017), the hypothetical future market value of Whole Food without the merger has been estimated as $ 14.3 bn. Therefore, the cumulated hypothetical future market value of the target and the acquirer after 1-year equals (E) $ 484.9 bn. The risk-free rate of return (rf) in 2017 has been defined as Long-Term Government Bond Yields (10 years) for the USA which was 2.16% (YCharts, 2020). Expected volatility (σ ) has been determined based on historical volatilities for 3 years. Following the AlphaQuery report (AlphaQuery, 2020), the volatility (σ ) of Amazon after an announcement of the acquisition was assumed as 25.25%. Time to expiration in years (T) equals 1 year with five-time steps (one step is about 2 months) for the Binominal Option pricing model. The option premium as a competence-based synergies result has been calculated using an Excel spreadsheet. Results are given in Tables 1, 2, 3, and 4 as follows. According to the Black-Sholes Option pricing model (BSOPM), the value of the real option (call option value as synergies value) equals $ 50 bn as shown in Table 1. According to the Binominal Option pricing model (BOPM) equals $52.7 bn as shown in Tables 3 and 4. Therefore, the expected market value of Amazon, Inc. is the cumulated future market value of target and acquirer before the announcement (So) $ 478.6 bn plus

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real. . . Table 1 Black Scholes option pricing model (in $ bn)

51

Real Options valuation Black-Scholes The cumulated market value of target and acquirer before the announcement (So) Hypothetical future market of the separated entities forecast before the merger (K)

484.90

The risk-free rate of return (Rf) in 2017

2.16%

Time to expiration in years (T) The volatility of future share price Amazon (σ) in July of 2017 after the announcement

1 25.50%

d1

0.161

d2

-0.094

Value of the call option (C) = Synergies

Table 2 Recombining binomial lattice parameters

478.60

50.4

Real options binomial option pricing model Time increment (years) 0.20 Up factor (u) 1.121 Down factor (d) 0.892 Risk-neutral probability (p) 0.490

Table 3 Binominal option pricing model: a lattice of the underline values of Amazon after the acquisition (in $ bn) 0

1

2

3

4

5 846.46 755.23

673.83 601.21 536.41 478.60

673.83 601.21

536.41 478.60

427.02

536.41 478.60

427.02 380.99

427.02 380.99

339.93

339.93 303.30 270.61

synergies $52 bn equals $ 530.6 bn. Takeover premium is the difference between the market price $13.8 billion (or estimated value $14.3 billion) of a company and the actual price paid to acquire it ($13.4 billion), expressed as a percentage (2.8– 3.0%). The premium represents the additional value of owning 100% of a company in a merger or acquisition and is also known as the control premium. The control premium is the additional benefit an acquirer receives (compared to an individual shareholder) from having full control over the business. Therefore, the author found that the option premium significantly exceeded the actual takeover premium suggesting that, from an option pricing point of view, those acquisitions provided significant dynamic capabilities-based synergies. Put simply, the acquisition was able to generate significant value-added for the acquirer’s shareholders. “In most acquisitions, even those where synergy is real and creates

ˇ A. Cirjevskis

52

Table 4 Binominal option pricing model. Real options lattice: a value of Amazon synergies of the acquisition (in $ bn) 0

1

2

3

4

5 361.56 272.42

193.11 130.11 84.19 52.7

188.93 118.40

70.58 40.70

22.92

51.51 25.16

12.29 6.00

0.00 0.00

0.00

0.00 0.00 0.00

value, the acquiring firm’s stockholders get little or none of the benefits from synergy” (Damodaran, 2005, p. 41), due to biased evaluation process; managerial hubris (pride), and a failure to plan for synergy. But it is not a case of Amazon’s acquisition of Whole Food! Firms that like Amazon are disciplined when “making acquisitions and stay focused are better able to deliver promised synergy benefits. Synergy is difficult to deliver but it is not impossible to create” (Damodaran, 2005, p. 44).

4 Finding and Discussion Jahanshahi and Nawaser (2018) argue that study on a real option and dynamic capabilities suggest future research on many open questions. “Future research can test this relationship in the project and firm-level” (Jahanshahi & Nawaser, 2018, p. 400). The current paper contributes to this scientific discussion. The current paper justified the role of dynamic capabilities as antecedents of success or failures of M&A deals and to demonstrate real options application to measure managerial synergies in M&A deals. Testing empirically this relationship, current paper enriches our knowledge about how organizations can benefit from real option and redefine dynamic capabilities framework to the heart of strategic management. Therefore, the paper contributes and demonstrates how acquisition-based dynamic capabilities provide managerial synergies. This is the major theoretical contribution of the current paper. Whole Food is an attractive platform for Amazon for the transformation of an industry. Therefore, the first and second propositions have been justified empirically. Having advanced future research designs for real option valuation, Trigeorgis and Reuer (2017, p. 57) argue “we would encourage the use of real option with a greater focus on the individual project level of analysis, . . . on individual real

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real. . .

53

Fig. 1 The relationship among developed propositions

option cases.” Therefore, the current research also contributes to the real options theory in strategic management. Regarding managerial contribution, the proposed approach to value M&A synergy can be identified and assessed in the pre-merger due diligence process. According to Bruner (2004, pp. 326–327), the synergy areas can help in the overall transaction process and strategy by revealing interdependencies and value creation potential. The current research points out that the real option application provides an adequate practical approach for synergy valuation. Therefore, the third proposition has been justified quantitatively with an application of BSOPM and BOPL techniques. To sum up theoretical and managerial contribution, the relationship among developed proposition is given in Fig. 1. Figure 1 illustrates the likely relationships among the main construct presented in the paper, with dynamic capabilities shown as an antecedent of managerial synergies. Acquirers need to absorb and to integrate dynamic capabilities of targets and convert M&A deal into value creation process which can be evaluated using real options application. The proposed approach to value M&A synergy (Fig. 1) can be used by firms before an M&A deal in the due diligence process.

5 Conclusion, Limitation, and Future Work When some dynamic capabilities are missing, a company has the option to develop them internally or purchase them from outside. The current paper contributes to theory and practice by empirically illustrating how this logic works in the M&A process. “This partnership presents an opportunity to maximize value for Whole Foods Market’s shareholders, while at the same time extending our mission and bringing the highest quality, experience, convenience, and innovation to our

54

ˇ A. Cirjevskis

customers” (Whitten, 2017). Whole Foods provides Amazon with an incredible platform for the transformation of the grocery industry (HBS Working Knowledge, 2017). The current paper also demonstrates the limitation of the real option application to measure a dynamic capabilities-based synergy. It is difficult to validate the synergetic effect of one isolated acquisition deal when several acquisitions happen within the anticipation of the duration of getting synergy. Time to maturity 1 year was assumed for the deal of Amazon-Whole Food, namely form the end of June 2017 till the end of June 2018. This was the assumption that efficient markets should have a well-anticipated potential long-term merger gain within this period. Real option application provided forecast on the total market capitalization of Amazon 1 year after, namely $ 529.6 bl. However, the real market capitalization of Amazon after 1 year was $ 805.72 bn on 27.06.2018 (YCharts, 2020). The differences can be explained by exploring many M&A deals of Amazon within the period June 2017 to June 2018 which would have provided synergies and added much more market value to the merging organization (Wikipedia, 2020). In this vein, more research is needed to justify the developed proposition. Moreover, the paper, being of an exploratory and interpretive, raises several opportunities for future research, both in terms of theory development and findings validation. Three propositions discussed in the paper can be used to generate several hypotheses for further empirical testing using a broader sample and quantitative research methods. Certainly, the testing of the propositions presented here should help determine the applicability of real options valuation to the M&A deals and bring this emerging theory closer to the dynamic capabilities’ framework.

References Alhenawi, Y., & Krishnaswami, S. (2015). Long-term impact of merger synergies on performance and value. The Quarterly Review of Economics and Finance, 58, 93–118. AlphaQuery. (2020). Amazon.com, Inc (AMZN). Retrieved January 28, 2020, from https:// www.alphaquery.com/stock/AMZN/volatility-option-statistics/90-day/iv-mean Baldi, F., & Trigeorgis, L. (2009). Assessing the value of growth option synergies from business combinations and testing for goodwill impairment. Journal of Applied Corporate Finance, 21(4), 115–124. Bauer, F., & Matzler, K. (2014). Antecedents of M&A Success: The role of strategic complementarity, cultural fit, and degree and speed of integration. Strategic Management Journal, 35(2), 269–291. Bosecke, K. (2009). Value creation in merger, acquisition, and alliances (p. 203). Wiesbaden: Gabler. Bruner, R. (2004). Applied mergers and applications. New York: Wiley. Capron, L. (1999). The long-term performance of horizontal acquisitions. Strategic Management Journal, 20(11), 987–1018. Capron, L., & Anand, J. (2007). Acquisition-based dynamic capability. In C. E. Helfat, S. Finkelstein, W. Mitchell, M. Peteraf, H. Singh, D. Teece, & S. Winter (Eds.), Dynamic capabilities: Understanding strategic change in organizations (pp. 80–99). London: Blackwell.

Measuring Dynamic Capabilities-Based Synergies in M&A Deals with Real. . .

55

Chatterij, A., & Patro, A. (2014). Dynamic capabilities and managing human capital. The Academy of Management Perspectives, 28(4), 395–408. ˇ Cirjevskis, A. (2017). Acquisition based dynamic capabilities and reinvention of a business model: Bridging two perspectives together. The International Journal Entrepreneurship and Sustainability, 4(4), 516–525. ˇ Cirjevskis, A. (2019). The role of dynamic capabilities as drivers of business model innovation in mergers and acquisitions of technology-advanced firms. Journal of Open Innovation: Technology, Market, and Complexity, 5(12), 1–16. Clarence-Smith, T. (2020). Amazon vs. Walmart: Bezos goes for the jugular with whole foods acquisition. Retrieved January 28, 2020, from https://www.toptal.com/finance/mergers-andacquisitions/amazon-vs-walmart-acquisition-strategy Damodaran, A. (2005). The value of synergy. New York University—Stern School of Business. Retrieved January 16, 2020, from https://papers.ssrn.com/sol3/papers.cfm?abstract_id=841486 Danneels, E. (2002). The dynamic of product innovation and firm competence. Strategic Management Journal, 23, 1095–1121. Dunis, C. L., & Klein, T. (2005). Analyzing mergers and acquisitions in European financial services: An application of real options. European Journal of Finance, 11(4), 339–355. Ficery, K., Herd, T., & Pursche, B. (2007). Strategy in action. Accenture. Retrieved January 28, 2020, from http://www.firstcalladvisors.com/files/TheSynergyEnigma_4.pdf Garfield, L. (2017). Whole Foods CEO: Amazon saved us from the ‘whole paycheck’ trap. Retrieved August 19, 2019, from https://www.businessinsider.com/whole-foods-ceo-amazonreputation-lower-prices-2017-9 Goold, M., & Campbell, A. (1998). Desperately seeking synergy. Harvard Business Review, 76(5), 130–143. HBS Working Knowledge. (2017). Amazon-whole foods deals is a big win for consumers. Retrieved January 17, 2018, from https://www.forbes.com/sites/hbsworkingknowledge/2017/ 06/17/amazon-whole-foods-deal-is-a-big-win-for-consumers/#706c71347232 Helfat, C. E. (1997). Know-how and asset complementarity, and dynamic capability accumulation: The case of R&D. Strategic Management Journal, 18(5), 339–360. Helmore, E. (2017). Hard times for whole foods: People say it’s for pretentious people. I can see why. The Guardian. Retrieved August 19, 2020, from https://www.theguardian.com/business/ 2017/apr/29/whole-foods-hard-times-retail Jahanshahi, A. A., & Nawaser, K. (2018). Is real options reasoning a cause or consequence of dynamic capability? Strategic Change, 27(4), 395–402. Kim, J. Y., & Finkelstein, S. (2009). The effects of strategic and market complementarity on acquisition performance: Evidence from the U.S. commercial banking industry, 1989–2001. Strategic Management Journal, 30(6), 617–646. Kleinbaum, M., & Stuart, T. E. (2014). Network responsiveness: The social structural microfoundations of dynamic capabilities. The Academy of Management Perspectives, 28(4), 353– 367. Kowitt, B. (2018). How Amazon is using whole foods in a bid for total retail domination. Retrieved August 19, 2020, from https://fortune.com/longform/amazon-groceries-fortune-500/ Loukianova, A., Nikulin, E., & Vedernikov, A. (2017). Valuing synergies in strategic mergers and acquisitions using the real options approach. Investment Management and Financial Innovations, 14(1), 236–247. Makri, M., Hitt, M. A., & Lane, P. J. (2010). Complementary technologies, knowledge relatedness, and invention outcomes in high technology mergers and acquisitions. Strategic Management Journal, 31(6), 602–628. NYU Stern. (2019). Koninklijke Ahold Delhaize NV APARCH Volatility Analysis. Retrieved September 30, 2019, from https://vlab.stern.nyu.edu/analysis/VOL.AD:NA-R.APARCH O’Reilly, C., & Tushman, M. (2013). Organizational ambidexterity: Past, present, future. The Academy of Management Perspectives, 27(4), 324–338. Osterwalder, A., & Pigneur, Y. (2009). Business model generation. Self-published.

56

ˇ A. Cirjevskis

Pillars of Wall Street. (2017). Deal of the week: Amazon to buy whole foods for $13.4B. Retrieved January 28, 2020, from https://pillarsofwallstreet.com/wp-content/uploads/2013/02/6.16.17Amazon-to-Buy-Whole-Foods-for-13.4B.pdf Redman, R. (2018). Amazon’s sales jump in Q3, but not at physical stores. Retrieved August 19, 2020 from https://www.supermarketnews.com/online-retail/amazon-s-sales-jump-q3-notphysical-stores Schreyogg, G., & Kliesch-Eberl, M. (2007). How dynamic can organizational capabilities be? Towards a dual-process model of capability dynamization. Strategic Management Journal, 28, 913–933. Stefano, G. D., Peteraf, M., & Verona, G. (2014). The organizational drive train: A road to integration of dynamic capabilities researches. The Academy of Management Perspectives, 28(4), 307–327. Teece, D. J. (2007). Explicating dynamic capabilities: The nature and micro-foundations of (sustainable) enterprise performance. Strategic Management Journal, 28, 1319–1350. Teece, D. J. (2011). Dynamic capabilities: A guide for managers. Retrieved December 29, 2018, from http://iveybusinessjournal.com/publication/dynamic-capabilities-a-guide-for-managers/ Thomas, L. (2017). Don’t worry, Wal-Mart; Amazon buying Whole Foods is just a ‘drop in the bucket’. Retrieved August 19, 2020, from https://www.cnbc.com/2017/06/21/dont-worry-walmart-amazon-buying-whole-foods-is-just-a-drop-in-the-bucket.html Trautwein, F. (1990). Merger motives and merger perspectives. Strategic Management Journal, 11, 283–295. Trigeorgis, L., & Reuer, J. J. (2017). Real options theory in strategic management. Strategic Management Journal, 38(1), 42–63. Wang, L., & Zajac, E. J. (2007). Alliance or acquisition? A dyadic perspective on interfirm resource combinations. Strategic Management Journal, 28(13), 1291–1317. Whitten, S. (2017). Whole foods stock rockets 28% on $13.7 billion Amazon takeover deal. Retrieved February 17, 2018, from https://www.cnbc.com/2017/06/16/amazon-is-buyingwhole-foods-in-a-deal-valued-at-13-point-7-billion.html Wikipedia. (2020). List of mergers and acquisitions by Amazon. Retrieved January 28, 2020, from https://en.wikipedia.org/wiki/List_of_mergers_and_acquisitions_by_Amazon YCharts. (2020). 10 year treasury rate. Retrieved January 27, 2020, from https://ycharts.com/ indicators/10_year_treasury_rate Zahra, S. A., Sapienza, H. J., & Davidsson, P. (2006). Entrepreneurship, and dynamic capabilities: A review, model, and research agenda. Journal of Management Studies, 43, 917–955.

Raising Rivals’ Costs When the Downstream Firms Compete in Stackelberg Fashion Jacek Prokop and Adam Karbowski

Abstract The aim of this paper is to investigate the raising rivals’ cost effect when the downstream competition follows Stackelberg pattern. We find that the raising rivals’ cost effect occurs in such a market setup, and the effect is asymmetric. It means that the leader can via the unit R&D investment raise the overall marginal cost of the follower from 1/3 to 1/2, depending on the R&D externalities. The follower can, in turn, via the unit R&D investment raise the overall marginal cost of the leader from 1/6 to 1/2, depending on the R&D spillovers in the industry. We also find that under downstream Stackelberg competition, the larger R&D spillovers lead to the smaller R&D investment as well as the profit of the upstream firm, but the profits of the downstream Stackelberg follower are increasing with the larger R&D externalities. The behavior of output, prices of intermediate and final products as well as the profit of the downstream Stackelberg leader are nonmonotonic with respect to the research externalities. Comparing with the downstream monopoly, we conclude that downstream duopolists jointly invest more in R&D than does a downstream monopolist as long as the R&D spillovers are not very high. The profit of the downstream monopolist is higher than the joint profit of the downstream Stackelberg duopolists for any level of the R&D spillovers. However, the profit of the upstream firm is significantly lower when the downstream market is captured by a monopolist. Keywords R&D investment · Vertical relations · Stackelberg competition · Downstream monopoly JEL Classification: L1, L2, O3

J. Prokop () · A. Karbowski Department of Business Economics, SGH Warsaw School of Economics, Warsaw, Poland e-mail: [email protected]; [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_4

57

58

J. Prokop and A. Karbowski

1 Introduction Geroski (1992), Harabi (2002) or, more recently, Ge, Hu, and Xia (2014) observe that vertical R&D investments may perform better than the horizontal ones. Moreover, vertical R&D investment is a more frequent mode of operations than the horizontal R&D investment (Arranz & de Arroyabe, 2008; Dai, Zhang, & Tang, 2017; Ge et al., 2014; Karbowski & Prokop, 2019). However, as Banerjee and Lin (2003) notice, theoretical industrial organization papers on firms’ R&D concentrate on horizontal relations between enterprises. The works on firms’ R&D in vertical setting are quite scarce (Arranz & de Arroyabe, 2008; Atallah, 2002; Dai et al., 2017; Ishii, 2004; Karbowski, 2019; Manasakis, Petrakis, & Zikos, 2014; Steurs, 1995; Xu, Liang, Duan, & Xiao, 2015) and usually they only compare firms’ R&D investments under different R&D regimes, i.e., R&D competition, R&D cooperation, research joint venture (RJV) competition, and RJV cooperation. Kamien, Muller, and Zang (1992) distinguished four different regimes of R&D mentioned above (see also Capuano & Grassi, 2019; Karbowski & Prokop, 2019). In R&D competition, firms decide on their R&D investments unilaterally in order to maximize their individual economic profits. In R&D cooperation, firms coordinate their R&D investments, but compete in the final product market, in order to maximize the sum of their economic profits. In RJV competition, firms behave as in the R&D competition, but the results of R&D works are fully shared (knowledge sharing occurs). In RJV cooperation, firms share their knowledge and at the same time coordinate their R&D investments in order to maximize the sum of their overall profits. It is worth noticing that firms’ vertical R&D investment decisions are particular due to the existence of both positive and negative R&D externalities (Arrow, 1962; Geroski, 1995; Glaeser, Kallal, Scheinkman, & Shleifer, 1992; Jacobs, 1969; Marshall, 1890; Porter, 1990; Romer, 1986). In horizontal process R&D, the investments made by one firm lead to the reduction of the manufacturing costs of the rivals (d’Aspremont & Jacquemin, 1988; Kamien et al., 1992; Kamien & Zang, 2000). In vertical case, the investments made by a downstream firm also reduce the manufacturing costs of the rivals via R&D spillovers, but at the same time increase the demand for an input (component), allowing the upstream enterprise to raise the input price (Banerjee & Lin, 2003; Karbowski, 2019). The rise in component price exerts a negative impact on the manufacturing costs of the downstream rivals. This ‘raising rivals’ costs’ effect is often used by downstream firms to gain a cost advantage over the downstream competitors (Banerjee & Lin, 2003; Karbowski, 2019). The raising rivals’ cost effect mentioned above has been observed for vertical structures with downstream Cournot competition (Banerjee & Lin, 2003; Karbowski, 2019). Such a structure is however observed not very often in real world market setups. Usually, we downstream enterprises face various asymmetries, and this leads to the formation of the downstream leader and downstream followers (Karbowski & Prokop, 2018).

Raising Rivals’ Costs When the Downstream Firms Compete in Stackelberg Fashion

59

The aim of this paper is to investigate the raising rivals’ cost effect in the R&Dinvesting industry, when the downstream competition follows Stackelberg pattern. For simplicity, we focus on the case of a downstream duopoly and one upstream supplier. The paper is organized as follows. In the next section, we model the behavior of firms in the non-integrated supply chain, when the downstream firms compete according to Stackelberg model. For comparison, Sect. 3 considers the case of a downstream monopoly. Conclusions follow.

2 Stackelberg Competition Consider an industry composed of one upstream firm, U, and two downstream firms, denoted 1 and 2. The upstream firm supplies an intermediate good to the downstream firms at the price w. We normalize the costs of the upstream firm to zero (Banerjee & Lin, 2003). The downstream firms manufacture q1 and q2 units of a homogeneous final product, respectively. The production of each unit of the final good requires one unit of the intermediate good purchased from the upstream firm. The market demand for the final product is given as a linear price function p = a − q1 − q2 ,

(1)

where p denotes the market price, Q = q1 + q2 is the volume of total production of the industry, while a (a > 0) is a given market parameter. Each of the downstream companies is characterized by a linear function of the total manufacturing costs     Ci qi , xi , xj = c − xi − βxj qi ,

(2)

where c (c < a) is a given parameter of an initial efficiency of firm i, xi denotes the amount of R&D investments made by the company i, and xj denotes the amount of R&D investments made by the competitor. Parameter β (0 ≤ β ≤ 1) determines the size of R&D externalities, i.e., the benefits for a given company obtained as a result of research undertaken by the competitor. Higher level of β means that the R&D investments made by one company allow the competitor to reduce the manufacturing costs by a greater amount for free (Prokop & Karbowski, 2013). Let w be the price of the intermediate good. The costs of the R&D investments have a form of quadratic function γ

xi2 , 2

(3)

60

J. Prokop and A. Karbowski

where γ (γ > 0) is a given parameter. The entry barriers to the industry are viewed as too high for new enterprises to enter. We assume that in this industry one company, say firm 1, plays the role of the Stackelberg leader, and the other one, say firm 2, is the follower. Thus, firm 1 is the first to set the level of its supply (q1 ), and firm 2, given the production level set by the leader, decides about its own output level (q2 ). The game proceeds in two stages. At the first stage, both companies simultaneously and independently decide about their levels of R&D investments (xi ). These decisions affect the function of total manufacturing costs of each firm. At the second stage, the companies compete in the final product market according to the Stackelberg leadership model (Karbowski & Prokop, 2018). Consider the profit of the follower firm at the second stage of the game for a given amount of R&D investments, x1 and x2 π2 = (a − Q) q2 − (c − x2 − βx1 + w) q2 − γ

x22 . 2

(4)

For a given output level of the leader (q1 ), the follower maximizes its own profit by setting the production level at 1 (a − c − w − q1 + βx1 + x2 ) . 2

(5)

Taking into account the follower’s reaction given by (5), the leader maximizes its own profit, with a given size of x1 and x2 x12 . 2

(6)

1 (a − c − w + (2 − β) x1 + (2β − 1) x2 ) . 2

(7)

π1 = (a − Q) q1 − (c − x1 − βx2 − w) q1 − γ The optimal production volume of the leader is given by

q1 =

Substituting (7) into (5), we obtain the optimal output level of the follower

q2 =

1 (a − c − w + (3β − 2) x1 + (3 − 2β) x2 ) . 4

(8)

Given the R&D investments and the price of the intermediate good, w, the production levels q1 and q2 given by (7) and (8) constitute the Nash-Stackelberg equilibrium. The derived demand for the intermediate product is thus

Raising Rivals’ Costs When the Downstream Firms Compete in Stackelberg Fashion

Q=

61

1 (3a − 3c − 3w + (2 + β) x1 + (1 + 2β) x2 ) , 4

(9)

1 (3a − 3c − 4Q + 2x1 + βx1 + (1 + 2β) x2 ) . 3

(10)

or equivalently, w=

The upstream firm sets the price of the intermediate good at the monopoly level by maximizing π U = w · Q w∗ =

1 (3a − 3c + (2 + β) x1 + (1 + 2β) x2 ) . 6

(11)

Thus, the equilibrium aggregate output is given as

Q∗ =

1 (3a − 3c + (2 + β) x1 + (1 + 2β) x2 ) . 8

(12)

Note, that overall marginal cost of a downstream firm 1 is

w ∗ + c − x1 =

1 (3 (a + c) + (−4 + β) x1 + (1 + 2β) x2 ) . 6

(13)

Thus, 1 ∂ (w ∗ + c − x1 ) = − (4 − β) , ∂x1 6

(14)

1 ∂ (w ∗ + c − x1 ) = (1 + 2β) . ∂x2 6

(15)

and

The above result may be summarized as the lemma below. Lemma 1 (a) A unit reduction in the leader firm’s marginal cost decreases its overall marginal cost by 16 (4 − β). (b) A unit reduction in the follower firm’s marginal cost raises the overall marginal cost of the leader firm by 16 (1 + 2β). The overall marginal cost of a downstream firm 2 is w ∗ + c − x2 =

1 (3 (a + c) + (2 + β) x1 + (−5 + 2β) x2 ) . 6

(16)

62

J. Prokop and A. Karbowski

Thus, ∂ (w ∗ + c − x2 ) 1 = (2 + β) , ∂x1 6

(17)

1 ∂ (w ∗ + c − x2 ) = − (5 − 2β) . ∂x2 6

(18)

and

Based on the above result, we formulate the next lemma. Lemma 2 (a) A unit reduction in the leader firm’s marginal cost raises the overall marginal cost of the follower firm by 16 (2 + β). (b) A unit reduction in follower firm’s marginal cost decreases its overall marginal cost by 16 (5 − 2β). We may restate the above lemmas as the following property. Property (a) A unit reduction in the leader firm’s marginal cost decreases its overall marginal cost by 16 (4 − β) and raises the overall marginal cost of the follower firm by 1 6 (2 + β). (b) A unit reduction in follower firm’s marginal cost decreases its overall marginal cost by 16 (5 − 2β) and raises the overall marginal cost of the leader firm by 1 6 (1 + 2β). Substituting for w* given by (11) into (7) and (8), we obtain the quantities offered by the downstream firms

q1 =

q2 =

1 (3 (a − c) + (10 − 7β) x1 + (−7 + 10β) x2 ) , 12

1 (3 (a − c) + (−14 + 17β) x1 + (17 − 14β) x2 ) . 24

(19)

(20)

The corresponding profits are π1 =

 1  100 − 140β + 49β 2 − 144γ x12 − 2 (−10 + 7β) x1 (3 (a − c) 288  + (−7 + 10β) x2 ) + (3 (a − c) + (−7 + 10β) x2 )2 , (21)

Raising Rivals’ Costs When the Downstream Firms Compete in Stackelberg Fashion

π2 =

63

1  9(a − c)2 + (14 − 17β)2 x12 − 6 (a − c) (−17 + 14β) x2 576   + 289 − 476β + 196β 2 − 288γ x22 + 2 (−14 + 17β) x1 (3 (a − c)  + (17 − 14β) x2 ) . (22)

At the R&D stage each firm chooses xi to maximize its profit. The equilibrium R&D investments for each downstream firm are given, respectively, by

x1 = −

  (a − c) (−10 + 7β) 17 − 31β + 14β 2 − 12γ , −170 − 259β 3 + 98β 4 + 978γ − 576γ 2 + 12β 2 (6 + 49γ ) − 7β (−37 + 216γ )

(23)

x2 = −

  (a − c) (−17 + 14β) 10 − 17β + 7β 2 − 6γ . −170 − 259β 3 + 98β 4 + 978γ − 576γ 2 + 12β 2 (6 + 49γ ) − 7β (−37 + 216γ )

(24) The profit of each downstream firm is  2  (a − c)2 100 − 140β + 49β 2 − 144γ 17 − 31β + 14β 2 − 12γ γ

π1 = −  2 , 2 170 + 259β 3 − 98β 4 − 978γ + 576γ 2 − 12β 2 (6 + 49γ ) + 7β (−37 + 216γ )

(25)   2 (a−c)2 289−476β+196β 2 −288γ 10−17β+7β 2 −6γ γ

π2 = −  2 . 2 170+259β 3 −98β 4 −978γ +576γ 2 −12β 2 (6+49γ ) +7β (−37+216γ ) (26) The upstream supplier’s profit is  2 1728(a−c)2 9−16β+7β 2 −6γ γ 2

πU =  2 . 170+259β 3 −98β 4 −978γ +576γ 2 −12β 2 (6 + 49γ ) +7β (−37 + 216γ ) (27) Table 1 provides the results of numerical analysis of equilibrium for the parameters a = 100, c = 10, γ = 10, and various levels of parameter β.

64

J. Prokop and A. Karbowski

Table 1 Equilibrium in the case of downstream Stackelberg duopoly for a = 100, c = 10, γ = 10, and β ∈ [0, 1] β 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

x1 1.931650 1.794700 1.657680 1.520630 1.383640 1.246790 1.110150 0.973840 0.837945 0.702571 0.567823

x2 1.594080 1.466880 1.338310 1.208610 1.077960 0.946556 0.814571 0.682166 0.549493 0.416697 0.283912

q1 23.1798 23.1575 23.1304 23.0982 23.0607 23.0176 22.9687 22.9139 22.8530 22.7861 22.7129

q2 11.2523 11.2837 11.3097 11.3307 11.3469 11.3587 11.3661 11.3694 11.3688 11.3645 11.3565

p 65.5678 65.5589 65.5599 65.5711 65.5924 65.6237 65.6652 65.7167 65.7781 65.8495 65.9306

w 45.9096 45.9215 45.9201 45.9052 45.8768 45.835 45.7797 45.7111 45.6292 45.5340 45.4259

π1 249.996 252.029 253.768 255.202 256.326 257.132 257.618 257.781 257.620 257.135 256.327

π2 113.910 116.563 118.954 121.081 122.943 124.540 125.871 126.937 127.741 128.283 128.566

πu 1580.77 1581.59 1581.49 1580.47 1578.51 1575.64 1571.84 1567.13 1561.51 1555.01 1547.63

Source: own calculations

Table 1 shows that with the larger level of research spillovers, the R&D investments as well as the profits of the upstream firm are falling, but the profits of the downstream Stackelberg follower are increasing. The behavior of output, prices of intermediate and final products as well as the profit of the downstream Stackelberg leader are nonmonotonic with respect to the research externalities. The highest profit of the downstream Stackelberg leader is achieved when the R&D spillovers are at the level β = 0.7.

3 Successive Monopoly In this section, for comparative purposes, we investigate the firms’ R&D investments, product prices and quantities, as well as firms’ profits under downstream monopoly which eliminates the raising rivals’ cost effect. The profit of a downstream monopolist is πM = (a − qM ) qM − (c − xM + w) qM − γ

2 xM . 2

(28)

The optimal production volume of the downstream monopolist is given by qM =

1 (a − c − w + xM ) . 2

(29)

Given the R&D investment xM and the price of the intermediate good w, the production level of the intermediate product is determined by (29). Thus, the price of the intermediate product could be expressed by

Raising Rivals’ Costs When the Downstream Firms Compete in Stackelberg Fashion

w = a − c − 2qM + xM .

65

(30)

The upstream firm sets the price of the intermediate good at the monopoly level by maximizing π U = w · qM w =

1 (a − c + xM ) . 2

(31)

Then, the quantity of the intermediate product as well as the quantity of the output offered by the downstream firm is given by  qM =

1 (a − c + xM ) . 4

(32)

The profit of the downstream monopolist is πM =

 1  2 . (a − c)2 + 2 (a − c) xM + (1 − 8γ ) xM 16

(33)

At the R&D stage, the downstream monopolist chooses xM to maximize its profit. Thus, the equilibrium investments are obtained as xM =

a−c . −1 + 8γ

(34)

The profit of the downstream monopolist is πM =

(a − c)2 γ . −2 + 16γ

(35)

The upstream supplier’s profit is πU =

8(a − c)2 γ 2 (1 − 8γ )2

(36)

.

From Tables 1 and 2, it follows that downstream duopolists jointly invest more in R&D than does a downstream monopolist as long as the R&D spillovers are not very high (β ≤ 0.8). When β is getting closer to 1.0, the downstream monopolist invests more than the downstream duopolists together. Table 2 Equilibrium in the case of downstream monopoly for a = 100, c = 10, and γ = 10 xM 1.13924

qM 22.7848

Source: own calculations

pM 77.2152

wM 45.5696

πM 512.658

πu 1038.3

66

J. Prokop and A. Karbowski

The profit of the downstream monopolist is higher than the joint profit of the downstream Stackelberg duopolists for any level of the R&D spillovers. However, the profit of the upstream firm is significantly lower when the downstream market is captured by a monopolist.

4 Conclusions In this paper, we investigated the raising rivals’ cost effect in the supply chain, when the downstream competition follows Stackelberg pattern. The raising rivals’ cost effect occurs in such a market setup, and the effect is asymmetric. It means that the leader can via the unit R&D investment raise the overall marginal cost of the follower from 1/3 to 1/2, depending on the R&D externalities in the industry. The follower can, in turn, via the unit R&D investment raise the overall marginal cost of the leader from 1/6 to 1/2, depending on the R&D spillovers in the industry. Under downstream Stackelberg competition, the larger R&D spillovers lead to the smaller R&D investment as well as the profit of the upstream firm, but the profits of the downstream Stackelberg follower are increasing with the larger R&D externalities. The behavior of output, prices of intermediate and final products as well as the profit of the downstream Stackelberg leader are nonmonotonic with respect to the research externalities. The highest profit of the downstream Stackelberg leader is achieved when the R&D spillovers are at the level β = 0.7. Comparing with the downstream monopoly, we can conclude that downstream duopolists jointly invest more in R&D than does a downstream monopolist as long as the R&D spillovers are not very high (β ≤ 0.8). When β is getting closer to 1.0, the downstream monopolist invests more than the downstream duopolists together. The profit of the downstream monopolist is higher than the joint profit of the downstream Stackelberg duopolists for any level of the R&D spillovers. However, the profit of the upstream firm is significantly lower when the downstream market is captured by a monopolist. Acknowledgment This research was supported by National Science Centre, Poland (grant number UMO-2017/25/B/HS4/01632).

References Arranz, N., & de Arroyabe, J. (2008). The choice of partners in R&D cooperation: An empirical analysis of Spanish firms. Technovation, 28, 88–100. Arrow, K. (1962). Economic welfare and the allocation of resources for invention. In The rate and direction of inventive activity: economic and social factors. Princeton, MA: UMI. Atallah, G. (2002). Vertical R&D spillovers, cooperation, market structure, and innovation. Economics of Innovation and New Technology, 11, 179–209.

Raising Rivals’ Costs When the Downstream Firms Compete in Stackelberg Fashion

67

Banerjee, S., & Lin, P. (2003). Downstream R&D, raising rivals’ costs, and input price contracts. International Journal of Industrial Organization, 21, 79–96. Capuano, C., & Grassi, I. (2019). Spillovers, product innovation and R&D cooperation: A theoretical model. Economics of Innovation and New Technology, 28, 197–216. d’Aspremont, C., & Jacquemin, A. (1988). Cooperative and noncooperative R&D in Duopoly with spillovers. American Economic Review, 78, 1133–1137. Dai, R., Zhang, J., & Tang, W. (2017). Cartelization or cost-sharing? Comparison of cooperation modes in a green supply chain. Journal of Cleaner Production, 156, 159–173. Ge, Z., Hu, Q., & Xia, Y. (2014). Firms’ R&D cooperation behavior in a supply chain. Production and Operations Management, 23, 599–609. Geroski, P. (1992). Vertical relations between firms and industrial policy. Economic Journal, 102, 138–151. Geroski, P. (1995). Do spillovers undermine the incentive to innovate? In S. Dowrick (Ed.), Economic approaches to innovation. Aldershot: Edward Elgar. Glaeser, E., Kallal, H., Scheinkman, J., & Shleifer, A. (1992). Growth in cities. Journal of Political Economy, 100, 1126–1152. Harabi, N. (2002). The impact of vertical R&D cooperation on firm innovation: An empirical investigation. Economics of Innovation and New Technology, 11, 93–108. Ishii, A. (2004). Cooperative R&D between vertically related firms with spillovers. International Journal of Industrial Organization, 22, 1213–1235. Jacobs, J. (1969). The economy of cities. New York: Random House. Kamien, M., Muller, E., & Zang, I. (1992). Research joint ventures and R&D cartels. American Economic Review, 82, 1293–1306. Kamien, M., & Zang, I. (2000). Meet me halfway: Research joint ventures and absorptive capacity. International Journal of Industrial Organization, 18, 995–1012. Karbowski, A. (2019). Greed and fear in downstream R&D games. Decyzje, 32 (in press). Karbowski, A., & Prokop, J. (2018). R&D activities of enterprises, product market leadership, and collusion. Proceedings of Rijeka Faculty of Economics Journal of Economics and Business, 36, 735–753. Karbowski, A., & Prokop, J. (2019). The impact of vertical R&D cooperation on market performance of firms. Entrepreneurial Business and Economics Review, 7, 73–89. Manasakis, C., Petrakis, E., & Zikos, V. (2014). Downstream research joint venture with upstream market power. Southern Economic Journal, 80, 782–802. Marshall, A. (1890). Principles of economics. London: Macmillan. Porter, M. (1990). The competitive advantage of nations. London: Macmillan. Prokop, J., & Karbowski, A. (2013). R&D cooperation and industry cartelization. The Economics Discussion Paper Series, 2013-41, The Kiel Institute for the World Economy. Romer, P. (1986). Increasing returns and long-run growth. Journal of Political Economy, 94, 1002– 1037. Steurs, G. (1995). Inter-industry R&D spillovers: What differences do they make? International Journal of Industrial Organization, 13, 249–276. Xu, L., Liang, D., Duan, Z., & Xiao, X. (2015). Stability analysis of R&D cooperation in a supply chain. Mathematical Problems in Engineering, Article ID 409286.

Comparing Five Generational Cohorts on Their Sustainable Food Consumption Patterns: Recommendations for Improvement Through Marketing Communication Irene Kamenidou, Spyridon Mamalis, Ifigeneia Mylona, and Evangelia Zoi Bara

Abstract Contemporary food consumption patterns are regarded as one cause of environmental deprivation, so new sustainable food consumption patterns are vital for this world’s future. This paper presents the results of field research that studies the sustainable food consumption patterns of five generational cohorts, namely Generation Z, Generation Y, Generation X, Baby Boomers, and the Silent Generation. Specifically, it explores if the subjects of the five generational cohorts (N = 1561) have adopted or are willing to adopt sustainable food consumption behavior. Additionally, it investigates differences amongst the generational cohort’s sustainable food consumption behavior. Keywords Sustainable food consumption · Generations · Generational cohorts · Sustainability · Communication · Marketing · Consumer behavior JEL Classification: M31, M37, M38

1 Introduction Climate change and pressures exerted on the environment due to the long-standing human activity such as extensive land, and water exploitation has resulted in degradation of ecosystems (Siardos & Koutsouris, 2011; UNDP, 2012, p. 1), thus raising the issue of sustainability. The concepts of sustainability and sustainable development were introduced in 1987 by the Brundtland Commission of the United

I. Kamenidou () · S. Mamalis · I. Mylona · E. Z. Bara Department of Management Science and Technology, School of Business and Economics, International Hellenic University, Agios Loukas, Kavala, Greece e-mail: [email protected]; [email protected]; [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_5

69

70

I. Kamenidou et al.

Nations, while sustainable consumption was introduced in 1992 in the United Nations Conference on Environment and Development (UNCED), Agenda 21 (United Nations, 1992, p. 18). Since then, the intention of achieving a sustainable world through different parameters such as sustainable development, sustainable tourism, and sustainable consumption is consistently in focus. In 1994, the Oslo Roundtable on Sustainable Production and Consumption Symposium defined sustainable consumption as “the use of goods and services that respond to basic needs and bring a better quality of life, while minimizing the use of natural resources, toxic materials and emissions of waste and pollutants over the life cycle, so as not to jeopardize the needs of future generations” (Oslo Roundtable on Sustainable Production and Consumption, 1994). Meulenberg (2003) asserts that sustainable consumption is a practice based on the individuals’ decision-making process, taking into account the social responsibility of the consumer and their personal needs and desires. Sustainable food consumption behavior (SFCB) is a critical topic of sustainable consumption and is of high interest due to its multiple impacts (Reisch, Eberle, & Lorek, 2013). Moreover, due to its importance, continuous emerging research has focused on consumers and households’ SFCB (Grebitus, Steiner, & Veeman, 2012; Han & Hansen, 2012; Hunecke & Richter, 2017). While many academic papers are focusing on sustainable food consumption (SFC) (Reisch et al., 2013; Thøgersen, 2017; Vanhonacker, Van Loo, Gellynck, & Verbeke, 2013), there is no single accepted definition (Annunziata & Scarpato, 2014; Reisch et al., 2013). Nonetheless, SFC is a result of sustainable food choices and sustainable food diets (U.K. Parliament, 2011). According to FAO (2010), “Sustainable Diets are those diets with low environmental impacts which contribute to food and nutrition security and healthy life for present and future generations. Sustainable diets are protective and respectful of biodiversity and ecosystems, culturally acceptable, accessible, economically fair and affordable, nutritionally adequate, safe, and healthy; while optimizing natural and human resources.” As Meybeck and Gitz (2017, p. 1) state, a sustainable diet “combines two different perspectives: a nutrition perspective, focused on individuals, and a global sustainability perspective, in all its dimensions: environmental, economic and social.” SFC has been connected to diverse consumption elements, such as consuming local products and decreasing the consumption of meat and processed products (Redman & Redman, 2014). SFC also incorporates the increase of fruit and vegetable consumption (Carlsson-Kanyama & González, 2009), and generally consuming food that has a small ecological, carbon, and water footprint. Having all the above in mind, the focus of this paper is to examine the SFCB of five generational cohorts in Greece, namely Generation Z, Generation Y, Generation X, Baby Boomers, and the Silent Generation. Additionally, it explores existing differences in the SFC of these generational cohorts. Studying these generational cohorts is of great significance since they reflect almost the total adult population, and therefore this study explores the SFCB that the country engages in.

Comparing Five Generational Cohorts on Their Sustainable Food Consumption. . .

71

2 Literature Review SFC has been extensively studied, although it is a relatively new subject of research (after the 1994 Oslo Roundtable on Sustainable Production and Consumption). SFC has been studied from different points of view, such as fair trade (Tanner & Wölfing Kast, 2003), animal welfare (Clonan, Holdsworth, Swift, & Wilson, 2010), the environmental impact of food production, marketing, and consumption (Iribarren, Hospido, Moreira, & Feijoo, 2010). Consumer behavior studies on SFC focus on meat consumption reduction (de Bakker & Dagevos, 2012), or/and consumption of locally produced products (Wallgren & Höjer, 2009). Also, there is a bulk of studies focusing on the consumption of organic foods as means of SFC patterns (e.g., Thøgersen, 2010; Vittersø & Tangeland, 2015); waste behavior (Clemente, PérezSánchez, Ribal, Sanjuán, & Escobar, 2013), attitudes, intentions, involvement, perceptions, and motives towards/for employing SFCB (Jackson, 2005; PickettBaker & Ozaki, 2008; Vermeir & Verbeke, 2006), or barriers to adopting an SFCB (Robinson & Smith, 2002). Moreover, consumer characteristics and their effect on SFC have been studied too (Diamantopoulos, Schlegelmilch, Sinkovics, & Bohlen, 2003; Vanhonacker et al., 2013; Verain et al., 2012), providing conflicting results as to their impact on SFCB. Besides these themes, segmentation analysis based on SFC (De Maya, López-López, & Munuera, 2011; Vanhonacker et al., 2013) has also been studied. The above consists of only a few areas of research that regards SFC behavior. Regarding generational cohorts and SFC, the majority of the studies focus on Generation Y (or Millennials), studying food in general (Happonen, 2016; Mohd Suki & Mohd Suki, 2015), organic foods (Leong & Paim, 2015; Thambiah, Khin, Muthaiyah, & Yen, 2015), or organic wine (Pomarici, Amato, & Vecchio, 2016). As to Generation Z, few studies deal with SFC in general (Kamenidou, Mamalis, Pavlidis, & Bara, 2019; Su, Tsai, Chen, & Lv, 2019), though, no study was found that focused solely on Baby Boomers, Generation X, or the Silent Generation. However, a handful of studies focus on multiple generations simultaneously, such as Generation Z and Y (Zalega, 2019), or five cohorts, such as Vilceanu, Grasso, and Johnson (2019), with data for the latter from the Simmons OneView Research NHCS survey.

3 Methodology This research targets five generational cohorts in Greece: Generation Z, Generation Y, Generation X, the Baby Boomers, and the Silent Generation. The generational cohort theory states that people born in the same year range, at similar places, and have lived through the same life-changing events during their coming years (17– 23 years old) belong in the same generational cohort (Mannheim, 1952). Previous researchers (e.g., Williams & Page, 2011) have intensively studied each cohort’s

72

I. Kamenidou et al.

consumer characteristics. Since researchers do not agree on the year range of each cohort, this study follows the time range adopted by Kamenidou, Mamalis, Pavlidis, and Bara (2018) because it refers to the same country. Field research was realized for this study. The field research questionnaire was adopted from an extensive literature review and had qualitative follow-up research to verify the items for the research in Greece. Items regarding the question of engaging in specific SFC patterns were stemmed from the studies of Tobler, Visschers, and Siegrist (2011) and Vanhonacker et al. (2013), while three items came from the qualitative research implemented. The scale used was adopted from Kamenidou et al. (2019), employing a seven-point Likert-type scale. The ranking used was as follows: 1 = I am not doing this, and I am not willing to do it ever, 2 = I am not doing this, and I am not willing to do so in the far future (long run), 3 = I am not doing this, and I am not willing to do this in the near future, 4 = I am not doing this and will think about it in the future if I will do it (neutral), 5 = I am doing this occasionally (whenever I can), 6 = I am doing this already very often, 7 = I am already doing this, and I consider that I will continue doing it forever. Additionally, socioeconomic, and demographic questions of participants were also included. The questionnaire was forwarded via different roots (Facebook, e-mail, printed handouts, and personal interviews) depending on the targeted age, employing a non-probability sampling method. By this procedure, 1561 valid questionnaires were gathered and considered appropriate for the analysis. The statistical package SPSS ver.24 for windows was used, applying descriptive analysis, reliability, factor analysis with varimax rotation, One-Way ANOVA, model, and multiple comparisons of means. Ethical approval: There are no ethical issues involved in the processing of the questionnaire data used in the study. The necessary consents have been obtained by the persons involved, and the anonymity of the participants has been secured. All procedures performed in studies involving human participants were in accordance with the ethical standards of the International Hellenic’s University research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

4 Results: Discussion 4.1 Sample Profile As to gender, 47.3% were males and 52.7% females. Regarding generational cohorts, 24.1%, 27.4%, 15.5%, 20.1%, and 12.9% belonged in Generation Z, Generation Y, Generation X, Baby Boomers, and the Silent Generation, respectively. Also, 40.2% were single, 48.6% married, and 11.2% were divorced or widowed. As to education, 13.6%, 45.3%, and 12.2% had elementary education, secondary education, and post-secondary, respectively. Additionally, 29.0% had a graduate or

Comparing Five Generational Cohorts on Their Sustainable Food Consumption. . .

73

postgraduate education. As to the participants’ profession, 28.8% were employees (federal and private), 17.6% were businessmen, 3.1% are laborers, 21.2% were on a pension, and 29.4% were dependent on others (students, housekeepers, and unemployed). Lastly, as to participants’ income (net family income per month), 5.1% are considered with a very low income ( 1). Therefore, three factors were extracted (KMO = 0.904; BTS = 11,155,288; df = 105; p = 0.000). These factors account for 61.8% of Total Variance (T.V.), and the reliability coefficient alpha value of the total scale (15 items) is Cronbach α = 0.897, which is considered satisfactory (Spector, 1992). The first factor is named “Abstaining food consumption patterns,” explaining 23.0% of T.V. This construct consists of six items, four of which are patterns that avoid specific consumption behavior. This factor has a mean factor score (MFS) MFS = 4.28 (StD = 1.23), and the reliability coefficient alpha value of the factor is α = 0.848. The second factor is named “Obtaining sustainable food consumption patterns.” It explains 22.8% of T.V. and comprises of five items that are directly associated with the (desired) consumption patterns of sustainable food products. This factor has MFS = 4.74 (StD = 1.21), and the reliability coefficient alpha value of the factor is α = 0.848 too. Lastly, the third factor is named “Consuming meat protein substitutes.” This factor explains 16.0% of T.V. and consists of three items, referring to ways for substituting protein intake from meat. This factor has MFS = 3.11 (StD = 1.39), and the reliability coefficient alpha value of the factor is α = 0.774. The three MFS of the abovementioned extracted factors were used as new variables for further analysis.

4.4 Hypotheses Testing The main hypothesis of this research is that generational cohorts differ in SFC behavior. Hence, three separate hypotheses arise from this broad hypothesis, which is formed due to the three factors/constructs extracted from factor analysis. These three hypotheses are defined as follows: Hypothesis N.1 H1 = There are differences between generational cohorts and their SFC behavior, and specifically in terms of abstaining food consumption patterns. This hypothesis is presented statistically as: H10 = There are no statistically significant differences between the generational cohorts and abstaining food consumption patterns (a = 0.05).

Comparing Five Generational Cohorts on Their Sustainable Food Consumption. . .

75

Table 2 ANOVA tests between SFC constructs and generational cohorts SFC constructs Abstaining food consumption patterns Obtaining sustainable food consumption patterns Consuming meat protein substitutes

Sum of squares df Mean square F Sig. Between groups 78.735 4 19.684 13.379 .000 Within groups 2289.291 1556 1.471 Total 2368.026 1560 Between groups 65.315 4 16.329 11.513 .000 Within groups Total Between groups Within groups Total

2206.890 2272.205 10.606 2984.909 2995.516

1556 1560 4 1556 1560

1.418 2.652 1.918

1.382 .238

Source: The authors

H11 = There are statistically significant differences between the generational cohorts and abstaining food consumption patterns (a = 0.05). Hypothesis N.2: H2 = There are differences between generational cohorts and their SFC behavior, specifically in SFC patterns. This hypothesis is presented statistically as: H20 = There are no statistically significant differences between the generational cohorts and obtaining SFC patterns (a = 0.05). H21 = There are statistically significant differences between the generational cohorts and obtaining SFC patterns (a = 0.05). Lastly, hypothesis N.3 is as follows: H3 = There are differences between generational cohorts and their SFC behavior, and specifically in terms of consuming meat protein substitutes, which is presented as follows: H30 = There are no statistically significant differences between the generational cohorts and consuming meat protein substitutes (a = 0.05). H31 = There are statistically significant differences between the generational cohorts and consuming meat protein substitutes (a = 0.05). Table 2 presents the results of the hypotheses testing the One-Way ANOVA model. In all cases, the three constructs of SFC behavior were the dependent variable, and the generational cohorts were the independent variable. Concerning the first hypothesis, the One-Way ANOVA test revealed significant differences between generational cohorts [F (4.1556) = 19.684, p < 0.001]. Consequently, the null hypothesis is rejected. As regards the second hypothesis, the One-Way ANOVA test also unveiled significant differences between generational cohorts [F (4.1556) = 16.329, p < 0.001]. Henceforth, the null hypothesis is rejected too. Lastly, referring to the third hypothesis, One-Way ANOVA did not reveal significant differences between generational cohorts [F (4.1556) = 2.652, p > 0.05]. As a result, the null hypothesis cannot be rejected.

76

I. Kamenidou et al.

Table 3 Tuckey B test between SFC and generational cohorts Constructs of SFC Abstaining food consumption patterns Obtaining sustainable food consumption patterns

Gen Z 3.99b

Gen Y 4.15b

Gen X 4.52a

Baby boomers 4.41a

Silent generation 4.6a

4.51b

4.58b

4.96a

4.90a

5.01a

Source: The authors

As the data analysis of this study reveal that in two cases (abstaining food consumption patterns and obtaining SFC patterns), at least two cohorts differ in their SFC behavior.

4.5 Multiple Comparison of Means Multiple comparisons of means were conducted to investigate in-depth which generational cohort differs from others in the above two cases where the null hypothesis was rejected. Table 3 presents the results of the multiple comparisons of means for the two constructs, using the post hoc Tuckey B comparisons test. In Table 3, each row with different letters beside the mean scores reveals significant differences, starting with “a” for the highest mean score. Therefore, numbers with the same letters in a row reveal no statistical differences. In respect of the first construct of SFC behavior, “Abstaining food consumption patterns” Tuckey’s B test indicated that the mean scores for the older cohorts are significantly higher as compared to the younger ones (Generation Z and Y). The oldest generational cohort (Silent Generation) is the one that seems to be more in line with these tactics as compared to the other cohorts. One reason could be that due to their pension, which is relatively quite low (due to cut-offs implemented by TROIKA, after the economic crisis started). Thus, they are not willing to pay a premium price for imported food or food with a high price due to a more luxurious or better-designed package. In this way they abstain from nonSFC patterns. This suggestion is in line with Ang, Leong, and Kotler (2000), who assert that consumers come to be more rational and purchase the necessary goods under turbulent economic conditions. Additionally, Kamenidou, Rigas, and Priporas (2018), in their research on a sample of 1305 households in Greece, found that the economic crisis raised food insecurity issues, whereas 65.6% of the households that participated in the study considered that the economic crisis had an impact on their food access. On the other hand, the younger generations (Generations Z and Y) are not prone to abstain from non-SFC patterns. This behavior could be due to the influence of the products’ marketing cues, the probable significance of peers or they may not be ecologically conscious to the point that they will likely change their consumption patterns.

Comparing Five Generational Cohorts on Their Sustainable Food Consumption. . .

77

Regarding the second construct, the findings suggest that the older generations are more prone to SFC behavior. At the same time, the younger generational cohorts (Z and Y) are less willing to do so. These findings contradict findings of previous researchers that found that the younger generations (specifically, Generation Y) care about the environment and are more willing to proceed to an SFCB (Gurau, 2012; Ntanos, Michalis, & Ntanos, 2014; Organic Trade Association, 2016), while the findings do align with those of Kamenidou et al. (2019), regarding the Generation Z cohort.

5 Conclusions: Realizations—Limitations of the Study and Directions for Further Research This research aimed to investigate the SFC behavior that the Greek generational adult cohorts hold, seeking differences regarding the generational cohorts and SFC behavior. This research revealed that the older generational cohorts are those prone to SFC behavior, while the younger ones do not seem to be engaged in this type of conduct. Additionally, differences were found regarding SFC behavior and generational cohorts, specifically for the first and second constructs of SFC patterns (Abstaining food consumption patterns and Obtaining SFC patterns, respectively). Marketing communication through contemporary communication channels will provide awareness to younger generational cohorts for the benefits of SFC behavior. For example, the use of digital marketing could be a way for younger cohorts to understand the benefits of SFC behavior. Social media allow marketers to directly engage consumers in the creative process as they are active participants instead of passive recipients (Thackeray, Neiger, Hanson, & McKenzie, 2008). The development of social media marketing provides marketers with new opportunities, as they can communicate with millions of young consumers and inform them about their products and their benefits (Duffett, 2017). Social media can be used as a powerful marketing tool for younger consumers (Mulero & Adeyeye, 2013). This research has some restrictions that should be mentioned, such as a relatively small sample per cohort, justified by the fact that this research is self-funded. A larger sample in future research would be desired to validate the findings of this research. Additional items measuring SCF behavior could be added since the items included in the questionnaire were validated from previous literature and qualitative research. Even though this research has the abovementioned limitations, it is considered significant because it provides an insight into the SFC behavior regarding five generational cohorts simultaneously, where a scarcity of studies exists.

78

I. Kamenidou et al.

References Ang, S. H., Leong, S. M., & Kotler, P. (2000). The Asian apocalypse: Crisis marketing for consumers and businesses. Long Range Planning, 33(1), 97–119. Annunziata, A., & Scarpato, D. (2014). Factors affecting consumer attitudes towards food products with sustainable attributes. Agricultural Economics/Zemedelska Ekonomika, 60(8), 353–363. Carlsson-Kanyama, A., & González, A. D. (2009). Potential contributions of food consumption patterns to climate change. The American Journal of Clinical Nutrition, 89(5), 1704S–1709S. Clemente, G., Pérez-Sánchez, M., Ribal, J., Sanjuán, N., & Escobar, N. (2013, July). Influence of agro-food waste on sustainable food consumption. In LCM 2013. Clonan, A., Holdsworth, M., Swift, J., & Wilson, P. (2010, March). U.K. consumers priorities for sustainable food purchases. In 84th Annual Conference of the Agricultural Economics Society (pp. 29–31). de Bakker, E., & Dagevos, H. (2012). Reducing meat consumption in today’s consumer society: Questioning the citizen-consumer gap. Journal of Agricultural and Environmental Ethics, 25(6), 877–894. De Maya, S. R., López-López, I., & Munuera, J. L. (2011). Organic food consumption in Europe: International segmentation based on value system differences. Ecological Economics, 70(10), 1767–1775. Diamantopoulos, A., Schlegelmilch, B. B., Sinkovics, R. R., & Bohlen, G. M. (2003). Can sociodemographics still play a role in profiling green consumers? A review of the evidence and an empirical investigation. Journal of Business Research, 56, 465–480. Duffett, R. G. (2017). Influence of social media marketing communications on young consumers’ attitudes. Young Consumers, 18(1), 19–39. https://doi.org/10.1108/YC-07-2016-00622 FAO. (2010). Food and Agriculture Organization of the United Nations. Definition of sustainable diets. Paper presented at: International Scientific Symposium: Biodiversity and Sustainable Diets United Against Hunger, 3–5 November 2010, Rome, Italy. Retrieved July 26, 2018, from http://www.fao.org/ag/humannutrition/28507-0e8d8dc364ee46865d5841c48976e9980.pdf Grebitus, C., Steiner, B., & Veeman, M. (2012). Personal values and decision making: Evidence from environmental footprint labeling in Canada. American Journal of Agricultural Economics, 95(2), 397–403. Gurau, C. (2012). A life-stage analysis of consumer loyalty profile: Comparing generation X and millennial consumers. Journal of Consumer Marketing, 29(2), 103–113. Han, Y., & Hansen, H. (2012). Determinants of sustainable food consumption: A meta-analysis using a traditional and a structural equation modelling approach. International Journal of Psychological Studies, 4(1), 22. Happonen, V. (2016). Differences in consumer perspectives how sustainable food consumption is perceived and signalled in the prevailing consumer culture: Among Generation Y. Retrieved July 30, 2019, from https://jyx.jyu.fi/bitstream/handle/123456789/50690/ URN%3aNBN%3afi%3ajyu-201607023457.pdf?sequence=1&isAllowed=y Hunecke, M., & Richter, N. (2017). Mindfulness, construction of meaning, and sustainable food consumption. Mindfulness, 10, 446–458. Iribarren, D., Hospido, A., Moreira, M. T., & Feijoo, G. (2010). Carbon footprint of canned mussels from a business-to-consumer approach. A starting point for mussel processors and policy makers. Environmental Science & Policy, 13(6), 509–521. Jackson, T. (2005). Motivating sustainable consumption. Sustainable Development Research Network, 29, 30. Kamenidou, I. E. C., Mamalis, S., Pavlidis, S., & Bara, E. Z. (2018). A multi-generational approach of destination components in actual destination choice. In Tourman 2018 Conference Proceedings: “In search of excellence in tourism, travel & hospitality” (pp. 442–448). Rhodes, Greece, 25–28 October 2018.

Comparing Five Generational Cohorts on Their Sustainable Food Consumption. . .

79

Kamenidou, I. C., Mamalis, S. A., Pavlidis, S., & Bara, E. Z. G. (2019). Segmenting the generation Z cohort university students based on sustainable food consumption behaviour: A preliminary study. Sustainability, 11(3), 837. Kamenidou, I., Rigas, K., & Priporas, C. V. (2018). Household behaviour on food security during an economic crisis. In G. Mergos & M. Papanastassiou (Eds.), Food security and sustainability (pp. 243–261). Cham: Palgrave Macmillan. Leong, T. P., & Paim, L. (2015). Mediating effects of intention on the factors affecting organic food products consumption among Chinese generation y in Malaysia. International Journal of Business Research and Management, 6(1), 1–19. Mannheim, K. (1952). The problem of generations. In P. Kecskemeti (Ed.), Karl Mannheim: Essays (pp. 276–322). London: Routledge. Meulenberg, M. T. G. (2003). Consument en burger, betekenis voor de markt van landbouwproducten en voedingsmiddelen [Consumer and citizen, meaning for the market of agricultural products and food products]. Tijdschrift voor Sociaal Wetenschappelijk onderzoek van de Landbouw, 18(1), 43–54. Retrieved September 10, 2019, from http://library.wur.nl/WebQuery/ wurpubs/fulltext/217867 Meybeck, A., & Gitz, V. (2017). Sustainable diets within sustainable food systems. Proceedings of the Nutrition Society, 76(1), 1–11. Mohd Suki, N., & Mohd Suki, N. (2015). Does religion influence consumers’ green food consumption? Some insights from Malaysia. Journal of Consumer Marketing, 32(7), 551–563. Mulero, O., & Adeyeye, M. (2013). An empirical study of user acceptance of online social networks marketing. SACJ, 50, 6–14. Ntanos, A., Michalis, S., & Ntanos, S. (2014). Millennial consumers’ perceptions on the organic products. In 9th eRA International Scientific Conference (pp. 26–35). Piraeus. Organic Trade Association. (2016). Millennials and organic: A winning combination. Retrieved December 4, 2019, from https://www.ota.com/news/press-releases/19256 Oslo Roundtable on Sustainable Production and Consumption. (1994). Part 1—the imperative of sustainable production and consumption. Retrieved December 18, 2019, from http:// enb.iisd.org/consume/oslo004.html Pickett-Baker, J., & Ozaki, R. (2008). Pro- environmental products: marketing influence on consumer purchase decision. Journal of consumer marketing. Pomarici, E., Amato, M., & Vecchio, R. (2016). Environmental friendly wines: A consumer segmentation study. Agriculture and Agricultural Science Procedia, 8, 534–541. Redman, E., & Redman, A. (2014). Transforming sustainable food and waste behaviours by realigning domains of knowledge in our education system. Journal of Cleaner Production, 64, 147–157. Reisch, L., Eberle, U., & Lorek, S. (2013). Sustainable food consumption: An overview of contemporary issues and policies. Sustainability: Science, Practice and Policy, 9(2), 7–25. Robinson, R., & Smith, C. (2002). Psychosocial and demographic variables associated with consumer intention to purchase sustainable produced foods as defined by the Midwest food Alliance. Journal of Nutrition Education and Behaviour, 34(6), 316–325. Siardos, G., & Koutsouris, A. (2011). Sustainable agriculture and development (3rd ed.). Thessaloniki: Zygos. Spector, P. E. (1992). Summated rating scale construction: An introduction. Sage University Paper Series No.82, on quantitative applications in the social sciences. Beverly Hills, CA: Sage. Su, C. H. J., Tsai, C. H. K., Chen, M. H., & Lv, W. Q. (2019). U.S. sustainable food market Generation Z consumer segments. Sustainability, 11(13), 3607. Tanner, C., & Wölfing Kast, S. (2003). Promoting sustainable consumption: Determinants of green purchases by Swiss consumers. Psychology & Marketing, 20(10), 883–902. Thackeray, R., Neiger, B. I., Hanson, C. L., & McKenzie, J. F. (2008). Enhancing promotional strategies within social marketing programs: Use of Web 2.0 social media. Health Promotion Practice, 9(4), 338–343. Thambiah, S., Khin, A. A., Muthaiyah, S., & Yen, Y. Y. (2015). Organic food consumption among Generation Y in Malaysia: A conceptual. Journal of Applied Sciences, 15(3), 570–575.

80

I. Kamenidou et al.

Thøgersen, J. (2010). Country differences in sustainable consumption: The case of organic food. Journal of Macromarketing, 30(2), 171–185. Thøgersen, J. (2017). Sustainable food consumption in the nexus between national context and private lifestyle: A multi-level study. Food Quality and Preference, 55, 16–25. Tobler, C., Visschers, V. H., & Siegrist, M. (2011). Eating green. Consumers’ willingness to adopt ecological food consumption behaviours. Appetite, 57(3), 674–682. U.K. Parliament. (2011). Sustainable food. Session 2010-11, 6th April 2011. Retrieved from https:/ /publications.parliament.uk/pa/cm201011/cmselect/cmenvaud/writev/food/m22.htm UNDP-United Nations Development Programme. (2012). The future we want. Biodiversity and ecosystems–driving sustainable development. Biodiversity and ecosystems global framework 2012-2020. Retrieved January 18, 2016, from http://www.undp.org/content/dam/undp/library/ Environment%20and%20Energy/biodiversity/UNDP-Biodiversity-and-Ecosystems-GlobalFramework-2012-2020.pdf United Nations Sustainable Development United Nations Conference on Environment & Development Rio de Janerio, Brazil, 3 to 14 June. (1992). AGENDA 21. Retrieved from https:// sustainabledevelopment.un.org/content/documents/Agenda21.pdf Vanhonacker, F., Van Loo, E. J., Gellynck, X., & Verbeke, W. (2013). Flemish consumer attitudes towards more sustainable food choices. Appetite, 62, 7–16. Verain, M., Bartels, J., Dagevos, H., Sijtsema, S. J., Onwezen, M. C., & Antonides, G. (2012). Segments of sustainable food consumers: A literature review. International Journal of Consumer Studies, 36(2), 123–133. Vermeir, I., & Verbeke, W. (2006). Sustainable food consumption: Exploring the consumer “attitude–behavioural intention” gap. Journal of Agricultural and Environmental Ethics, 19(2), 169–194. Vilceanu, M. O., Grasso, O., & Johnson, K. (2019). Bridging the Gap Between Public Opinion Research and Consumer Marketing Research: Insights into US Shoppers of Organic Foods. Association of Marketing Theory and Practice Proceedings 2019, Georgia Southern University, Digital Commons Georgia Southern. https://digitalcommons.georgiasouthern.edu/amtpproceedings_2019/7/ Vitters, G., & Tangeland, T. (2015). The role of consumers in transitions towards sustainable food consumption. The case of organic food in Norway. Journal of Cleaner Production, 92, 91–99. Wallgren, C., & Höjer, M. (2009). Eating energy—Identifying possibilities for reduced energy use in the future food supply system. Energy Policy, 37(12), 5803–5813. Williams, K. C., & Page, R. A. (2011). Marketing to the generations. Journal of Behavioral Studies in Business, 3, 1–17. Zalega, T. (2019). Sustainable consumption in consumer behaviour of young Polish consumers. Studia Ekonomiczne, 383, 82–107.

The Effect of Budgetary Policies on the Economy Activity in Algeria: A Markov Switching Approach Touitou Mohammed

Abstract This study aims to highlight the effect of budgetary policies on the economy activity in Algerian for the period 1986 to 2014. Contrary to prior studies, our analysis opts for a recent and robust technical setting within the framework of Markov switching models. The findings indicate that a deterioration of the balance is favorable to the activity; this situation describes a Keynesian nature (Budget deficit helps boost economic activity). The improvement of budget balance improves economic activity. This is the classic nature (according to the Classics, the financing of activity by the budget deficit has a negative effect on economic activity) according to the terminology of the study. The findings could help policymakers to establish efficient economic decisions to boost the economy. Favorable economic repercussions may result from an effective policy coordination in decision-making. Keywords MS-VAR approach · Budgetary policy · Expansion · Recession Jel Classification: E52, C32, C87, E32

1 Introduction Budgetary policy is an instrument of economic policy. Keynesian theory states that it can stimulate aggregate demand and revive a stagnant economy. On the other hand, the classics (neoclassical) suggest that an expansionary budgetary policy has no positive effect on economic activity: deficits are harmful and lead to higher rates interest. Thus, according to the theory of rational expectations, agents make expectations about the taxes they will have to pay in the future, which results in a fall in private demand and supply and consequently a slowdown in activity.

T. Mohammed () Faculty of Economics and Business, University of Algiers 3, Road Ahmed Oueked, Dely Brahim, Algiers, Algeria © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_6

81

82

T. Mohammed

Moreover, empirical studies on the subject conducted in the 1990s have called into question the importance of the Keynesian fiscal multiplier: these studies have concluded that expansionary fiscal policies could have negative effects on consumption, investment, and interest rates. In general, the work on structural VAR modeling remains limited: Hénin & N’Diaye1 (2001) show that in this work, innovation on the budget variable contributes only 6–20% of the variance of the forecast error of GDP. Other arguments suggest that the interactions between fiscal variables and macroeconomic activity may be neither symmetrical nor homogeneous over time. Also, of all the above, is it not justified to reformulate the interactions between the budget deficit and the economic activity as dependent on the budgetary regime and the conjunctural conditions? The work therefore consists of determining, by using the Markov-Switching Vector Autoregressive (MS-VAR) approach, the effect of budgetary policies on economic activity in Algeria. In addition, it is a question of identifying whether the budgetary policies adopted in the context of Algeria have a classic effect or a Keynesian effect. This study will be conducted by using quarterly data for the period 1986–2014. To achieve this objective, we assume that the effect of budgetary policies on the economic activity of Algeria would depend on economic conditions (expansion or recession) and/or the budgetary regime (deficit or balance of the budgetary balance).

2 An Empirical Review of the Literature In recent years the economic literature has experienced the birth of a new empirical analysis approach, namely VAR modeling. This approach is ranked among the main works of modern macroeconomics. This methodology developed by Sims (1980) has been essentially applied in the analyzes related to the effects of monetary shocks, while little work has been done on the effects of budgetary shocks. It is recently that other theoretical and empirical literatures have emerged that have focused on the analysis of the effects of fiscal policy. The VAR Structural Approach (SVAR), which refers to the pioneering work of Blanchard and Perotti (2002), simulates a public spending shock on the US economy and finds a positive response from private consumption and US GDP (the multiplier being 0.9 or 1.29 according to the estimation method). On the other hand, a tax revenue shock negatively affects private consumption and US GDP. They also say that private investment and imports react negatively following a shock on public spending. Biau and Gerard (2004), using quarterly data for the period 1978T1–2003T4, have shown that a structural shock of public expenditure in France has a positive

1 Both

authors use a methodology that is not very common in the literature to measure the effect of budgetary policies on activity, namely the MS-VAR approach. This methodology will be presented in this work.

The Effect of Budgetary Policies on the Economy Activity in Algeria: A. . .

83

impact on the short-term overall demand (1.4 euro) in France because the positive effects of the shock on private demand, particularly on consumption and private investment. The effect is decreasing and thus regains, from the second year onward, its long-run equilibrium level due to the adjustment of economic activity to higher prices. However, economic activity reacts negatively following a structural shock to tax revenues; this is explained by the resulting contraction of private consumption expenditure. This effect fades rapidly from the second quarter to return to its longterm level. Mihov and Fatas (2001) compare the results of a VAR model estimated for the United States to the predictions of a neoclassical model. It follows from their study that any increase in State consumption expenditure financed by higher taxes implies a reduction in household consumption because of the decrease in their income. On the other hand, their VAR model shows that a shock on overall public spending encourages private sector investment and promotes employment. Mountford and Uhlig (2002), using SVAR modeling, find that a shock on tax revenues has negative effects on private consumption and US GDP, while a public expenditure shock does not reduce consumption in United States, but has the effect of crowding out private investment. Badinger (2006) provides in his analysis an assessment of the effects of discretionary budgetary policy in Austria during the period 1983Q1 to 2002Q4. It shows that tax shocks have a negative effect on GDP, consumption, and investment. Public expenditure shocks have a positive effect on production and private consumption, but decline largely after a few years, resulting in a cumulative response close to zero. These results are difficult to reconcile with the neoclassical prediction that public spending negatively affects private consumption. In addition, tax shocks have larger effects than spending shocks in some specifications. This could be due to the fact that taxes do not only affect output through the income effect but also distortion prices, a point emphasized in neoclassical models. Afonso and Sousa (2009), from a Bayesian structural VAR, analyze the empirical data of the United States, the United Kingdom, Germany, and Italy, respectively, for the periods 1970T3-2007T4, 1964T2-2007T4, 1980T3-2006T4, and 1986T22004T4. The results of this article can be summarized as follows: governmentspending shocks generally have a small effect on GDP but do not translate into significant effects on private consumption. They affect private investment and have no significant impact on the price level and the average cost of refinancing the debt. Also, they have a positive but weak impact on monetary aggregates. On the other hand, government revenue shocks have a positive (albeit lagged) effect on GDP and private investment as a result of budgetary consolidation; but they have no impact on the price level. According to Reinhart (2010), the significant use of budgetary measures to counter the global financial crisis of 2007–2009 has further boosted the debate over the size of the budgetary multiplier. The state of public finances is rather estimated by using the stock of current public debt and/or the size of budget deficits. For Corsetti, Meier, and Müller (2012), a global shock to public spending causes an increase in debt that induces, over time, a systematic reduction in the

84

T. Mohammed

government’s future spending. This is because the real interest rates are failing to rise in response to higher public spending. The study by Hénin and N’Diaye (2001) complements the traditional VAR approach by taking into account the asymmetric nature of the responses of the activity to an increase or decrease in the public deficit. For four major countries (France, the United States, the United Kingdom, and Canada), the authors estimate a Markov-based VAR model of regimes in order to assess the extent to which the fiscal multiplier can depend on the conjunctural situation (expansion or recession). The findings of the study confirm the uncertainties noted by previous studies on the Keynesian effects of budgetary policy. The authors conclude non-Keynesian effects of budgetary policy in the majority of cases. Our review of the literature in this section has shown that GDP responses to short-run fiscal policy shocks can be positive or negative depending on the country. In addition, three channels of transmission have been identified: the public consumption expenditure channel, the investment channel and the inflation rate channel. The present study draws on the study by Hénin and N’Diaye (2001) to analyze the effect of budgetary policy on economic activity in Algeria. We will consider an estimation of a reduced equation of activity as a function of the primary balance, and according to expansion and recession conjunctural regimes then of cross regimes, associating with this cyclical typology two budgetary policy: “improvement of the balance” or “deterioration of the balance.”

3 Methodology It should be recalled that the purpose of this work is to evaluate the extent to which the effects of budgetary policy on activity depend on the economic phase considered (expansion or recession) and/or the budgetary regime in terms of balance or deficit. Thus, we first introduce the MS-VAR specifications selected for this study. We will then present the strategy of the empirical study and finally the stationarity of the series.

3.1 Presentation of the Model MS-VAR models, Markov-Switching Vector Autoregressive, introduced by Hamilton (1989) are used to model time series subject to discrete and occasional regime changes. The different regimes, of a discreet nature, are stochastic and unobservable. They thus allow the implementation of nonlinear dynamic models complex. The specification adopted is therefore that proposed by Hamilton (1989) where the transition between the regimes follows a Markov process.

The Effect of Budgetary Policies on the Economy Activity in Algeria: A. . .

85

Thus, if yt measures the cyclical gap of the activity, bt the primary budgetary balance, the model represents the economy by the following equation: yt = μs(t) +

k  i=1

ϕi,s(t) yt−i +

k 

λi,s(t) bt−i + δs(t) εt

(1)

i=1

It is assumed that the perturbations εt are independent, identically distributed and follow the normal center reduced. However, the regimes are heteroscedastic; the constant term depends on the state of the regime. There is a change in the autoregressive component of the cyclical gap; the impact of the fiscal balance on activity depends on the phase of the cycle. st being the state of the economy representing the expansion or recession (st = {0, 1}), the assumptions above allow specifying the parameters μs(t) , ϕi, s(t) , λi, s(t) and δ s(t) as follows: μs(t) = α0 + α1 st ϕi,s(t) = ϕi,0 + ϕi,1 st λi,s(t) = λi,0 + λi,1 st with λ< in the Keynesian case: a reduction in the balance has a stimulation on demand and > in the classical case. δs(t) = δ0 + δ1 st We admit that the continuation s(t) is a Markovian process of the first order. Pr ob (st /st−1 , st−2 . . . ) = Pr ob (st /st−1 ) Pr ob (st = 1/st−1 = 1) = P Pr ob (st = 0/st−1 = 1) = 1 − P Pr ob (st = 0/st−1 = 0) = q Pr ob (st = 1/st−1 = 0) = 1 − q

86

T. Mohammed

The matrix P of the probabilities of transition is then written as follows:  P =

P (1 − q) q (1 − P )



The probability of leaving the regime “1” is constant and equal to 1 − p. The average life of this regime is 1/(1 − P). The average duration of the regime “0” is 1/(1 − q). The state of symmetry is given by the condition p = q; the two regimes in this case are then persistent. The effect of initial conditions on the state of the system fades over time. The probability of observing the regime “1,” independently of the intermediate states, tends to a constant π , equal to the probability of being previously in this regime and to remain there (pπ ) plus the probability of being previously in “0” and getting out [(1 − q)(1 − π )] hence π = (1 − q)(2 − − p − q) also interpretable as the unconditional probability of the regime “1.”  we have :

π 1−π



 =p

π 1−π



The estimation of the model is based on the probability of realization of a regime conditionally to the previous states of the system. st not being observable, this probability must be expressed conditionally to the observation of past achievements of variable yt ; this requires the use of a filtering algorithm. The estimation of the model is done by maximizing the likelihood of the observations, the state of the system constituting a latent variable whose value is predicted by the BaumLindgren-Hamilton-Kim filter. We can then consider a multivariate dynamics of an interest variable. We can therefore associate with the previous model the equation of the budgetary balance bt : bt = vs(t) +

k  i=1

i,s(t) yt−i +

k 

ξi,s(t) bt−i + δgs(t) ηt

(2)

i=1

In the context of this study, the model (1) conditioning the impact of the balance on activity by the cyclical phase can thus be extended to the hypothesis of a double conditioning of the coefficient λi , on the one hand, by the economic situation (E or R, for expansion or recession) and, on the other hand, the existing budgetary situation characterized by the opposition between budget balance (bb) and budget deficit (bd). We will thus designate the four regimes of this model generalized by (Ebb, Edb, Rbb, Rbd) of the regime “Ebb” expansion of activity and balanced budget to the regime “Rbd” combining recession and budget deficit. The identification in these terms of the regimes results from the empirical application of the algorithm Esperance-Maximization (E.M.).

The Effect of Budgetary Policies on the Economy Activity in Algeria: A. . .

87

3.2 Strategy of the Empirical Study The empirical exploration strategy adopted for the study, as in Hénin and N’Diaye (2001), consists to estimate successively the three variants of the (1) specified in the framework of a bivariate VAR. Concretely, it is about estimating the following: – A linear model, on a single regime on the set of available samples; – A Markovian regime change model with two conjunctural regimes (expansion and recession, noted E and R); – A model in the context of an MS-VAR of the joint dynamics of the cyclical gap (1) and the balance (2) with the four conjunctural regimes (Ebb, Ebd, Rbb, Rbd). – The criterion of identification in posteriority (Hénin & N’Diaye, 2001) is used to identify the regimes. The estimation is made by using the OX statistical software developed by KROLZIG, H-M. (1998).

3.3 The Data of the Study In our econometric analysis, we use quarterly GDP and primary balance data (as a percentage of GDP). These data are spread over the period 1986–2014. They come from the National Bank (NB), the World Bank (WB), the Ministry of Finance (MF), and the National Statistics Office (NSO). The activity variable used is a measure of the cyclical gap, i.e., the difference between the logarithm of GDP and the stationary logarithmic component of GDP resulting from the application of the Hodrick-Prescott filter with the weighting coefficient λ = 1600 habitual in step quarterly. These annual data were quarterly by the method Goldenstein and Khan (1976).

3.4 Analysis of Series Stationarity 3.4.1

The Economic Gap and Budget Balance in Algeria

The graphical representation of the two series indicates that they are high chances of being stationary at level (Graph 1). Dickey Fuller’s unit root test confirms this apprehension. In fact, with regard to the economic gap, the ADF statistic = −75.7 is lower than the critical values for all the thresholds considered as can be seen in Table 1. Concerning the Budget Balance, the ADF statistic = −3.98 is also lower than the critical values for all the thresholds considered (Table 2).

88

T. Mohammed

10 5

0 -5

-10 86

88

90

92

94

96

98

00

02

04

Budget balance

06

08 10 12

14

Economic gap

Graph 1 Evolution of the Economic gap and Budget Balance in Algeria between 1986 and 2014 Table 1 Dickey Fuller’s test on the series of economic gap of Algeria ADF test statistic

−75.72561

1% Critical Valuea 5% Critical value 10% Critical value

−2.7628 −1.8945 −1.7052

Augmented Dickey-Fuller test equation Dependent variable: D(EC) Method: Least squares Sample (adjusted): 1986:4 2014:4 Included observations: 113 after adjusting endpoints Variable EC(−1) D(EC(−1)) D(EC(−2))

Coefficient −4.949555 1.971461 0.988272

Std. error 0.042651 0.030282 0.017576

t-statistic −92.60252 65.10434 56.22970

Prob. 0.0000 0.0000 0.0000

Source: Authors a MacKinnon critical values for rejection of hypothesis of a unit root

3.4.2

Determination of the Optimal Delay of the VAR

We can consider at most (p = 2) delays; it is usually sufficient, given the high number of parameters to be estimated and the problem of the MS-VAR model. Hénin and N’Diaye (2001) systematically set the autoregressive dimension of the models in their study to two delays. To retain the optimal delay of the VAR, we estimated two models: the first to one delay and the second to two delays. For each model, we calculated the criteria of Akaike (AIC), Schwarz (SC) as well as log likelihood (LV). The following table shows the results obtained.

The Effect of Budgetary Policies on the Economy Activity in Algeria: A. . .

89

Table 2 Dickey Fuller’s test on the budget balance series in Algeria ADF test statistic

−3.984562

1% 5% 10%

−2.6826 −1.9745 −1.8263

Critical valuea Critical value Critical value

Augmented Dickey-Fuller test equation Dependent Variable: D(SB) Method: Least Squares Included observations: 113 after adjusting endpoints Variable SB(−1) D(SB(−1)) D(SB(−2))

Coefficient −0.261416 0.024199 0.187342

Std. error 0.065035 0.096733 0.094162

t-statistic −4.019593 0.250162 1.989568

Prob. 0.0001 0.8029 0.0491

Source: Authors a MacKinnon critical values for rejection of hypothesis of a unit root Number of delays P=1 P=2

LV −456.2156 −428.9826a

AIC 8.2820 8.0172a

SC 8.7594 8.6893a

a Index the order p to retain according to the criterion used

The three criteria used lead us to retain a VAR process (2); the log likelihood y is greater, and the AIC and SC statistics are smaller than the autoregressive vector process of order 1.

4 Results and Discussion Empirical results for data are shown in Table 3. The first column reports the model’s OLS estimate on all data: from first quarter of 1986 in the fourth quarter of 2014, a total of 144 observations. These observations are classified in a single regime. From the values read from the table (t-student = 0.17 for SB-1; t-student = 0.13. for SB2), the student test allows us to say that we cannot reject the assumption that the coefficients of the primary balance delayed by one and two quarters are zero. The economic gap then follows an autoregressive process of order 2 AR (2). The second column of the table presents the results of the MS-VAR model with two conjunctural regimes. Regime 1 is identified as the recession regime, and Regime 2 is identified as expansion (The constant relative to the regime 1 is lower than the constant relative to the regime 2; moreover, it is significantly different from 0). In recession, activity is on average lower ((−1.33 (−1.18)) or 2.5 percentage points of GDP) at its level of expansion. Fifty-two points (45% of observations) are classified in expansion, and 62 points (55% of observations) are classified in recession.

Model Equation Variable Const Standard error t-statistic EC_1 Standard error t-statistic EC_2 Standard error t-statistic SB_1 Standard error t-statistic SB_2 Standard error t-statistic Unconditional Probability Number of observations

Table 3 Results

114

0.006660 (0.19309) (0.03449) −0.496123a (0.08393) (−5.91137) −0.498338a (0.08407) (−5.92776) 0.009895 (0.05635) (0.17559) 0.007598 (0.05643) (0.13464)

Linear sample Complete

Cyclical regimes 1(R) 2(E) −1.183865a 1.339212 0.4363 1.4221 (−2.7137) 0.9417 −0.693589a −0.242870 0.1233 0.2569 −5.6263 −0.9453 −0.482446a −0.392531a 0.1311 0.1283 −3.6787 −3.0584 −0.249367a −0.203767 0.1242 0.2902 −2.0078 −0.7022 0.046020 0.010049 0.0902 0.0865 0.5104 0.1162 0.54474 0.455263158 62.1 51.9 1(Rbb) −1.475928a 0.1327 −11.1212 −0.961307a 0.0981 −9.7990 −1.002547a 0.0165 −60.6295 0.025484 0.0172 1.6784b −0.008738 0.0181 −0.4829 0.2785 34.4

Cross regimes 2(Ebd) 1.255163a 0.5890 2.1311 −0.038648 0.1093 −0.3535 −0.035282 0.0607 −0.5809 −0.001066 0.1187 −0.0090 0.009956 0.0358 0.2781 0.3364 38.8 3(Ebb) 1.383350a 0.5794 2.3875 −0.045355 0.1631 −0.2781 −0.061812 0.1503 −0.4113 −0.019714 0.1450 −0.1360 0.004068 0.0994 0.0409 0.0482 8.7

4(Ebd) 1.064932 1.0603 1.0044 −0.150496 0.1962 −0.7671 −0.148356 0.1298 −1.1432 0.005375 0.2246 0.0239 −0.001305 0.0647 −0.0202 0.3370 32.1

90 T. Mohammed

12.38 R1 0.9192 0.0808

−428.9826 8.0172

Regime 1 Regime 2

−544.1527 9.7746

12.06 R2 0.0829 0.9171

Source: OX Software a The coefficient is significantly different from zero at the 5% threshold b The coefficient is significantly different from zero at the 10% threshold

Log vraisemblance AIC

Duration of Regime Probability of Transition

1.02 R1 Reg 1 Reg 2 Reg 3 Reg 4 0.01907 0.5325 0.07457 0.3739

2.26 R2 0.4409 0.557 0.00167 0.0004034 −261.2273 5.7057

1.42 R3 0.4618 0.01098 0.294 0.2333

2.92 R4 0.3045 0.0005371 0.03773 0.6572

The Effect of Budgetary Policies on the Economy Activity in Algeria: A. . . 91

92

T. Mohammed

The transition probability matrix shows that the probability of going out of one regime to another is very low (about 8% in both cases), and the probability of persisting in one state is 92%. The durations in each regime are approximately the same (12 quarters). The analysis of the estimated coefficients by regime reveals that in a period of recession (1), the balance coefficient is now significant to one delay; it is negative and apprehensive, therefore a positive effect; a deterioration of the balance is favorable to the activity; this situation describes a Keynesian nature (Budget deficit helps boost economic activity). The consideration of the economic phase allows to highlight a budgetary effect (Keynesian effect). The third column of the table shows the results of the cross plans (four latent regimes). The identification of the different regimes as the following: – We have E (Yt/s = 2, 3) > E (Yt/s = 1.4), so expansion and budget balance (Ebb) and expansion and budget deficit (Ebd) belong to {2,3} – Recession and balanced budget (Rbb) and Recession and budget deficit (Rbd) belong to {1,4} – Moreover, the expectation of the budget balance is greater in the balanced budget than in the budget deficit E (Gt/s = 3, 4) > E (Gt/s = 1, 2) – Expansion of Balanced Budget (Ebb) and Recession of Balanced Budget (Rbb) belong to {3,4} – The expansion and budget deficit (Ebd) and the recession and budget deficit (Rbd) belong to {1,2}. – It is easy to deduce that: 1. 2. 3. 4.

Represents recession and budget deficit (1 = Rbd) Represents expansion and budget deficit (2 = Ebd) Represents expansion and balanced budget (3 = Ebb) Represents recession and balanced budget (4 = Rbb)

The reduction of the AIC of this model compared to that with two Markov regime changes shows that this specification is not over-parsed. In a recession situation and a balanced budget, the coefficient on the budget balance is significantly different from 0 at the 10% threshold, it is now positive; the improvement of budget balance improves economic activity. This is the classic nature (according to the Classics, the financing of activity by the budget deficit has a negative effect on economic activity) according to the terminology of the study. The number of observations in expansion regime and balanced regime (8 observations) is relatively low unlike other regimes (more than 32 observations).

5 Conclusion The purpose of this study was to show that the effect of fiscal policies on the activity of the Algerian economy depends on the cyclical situation and/or the budgetary regime. The methodology used is that of the Markov regime change models (MSVAR).

The Effect of Budgetary Policies on the Economy Activity in Algeria: A. . .

93

The results show that in a recession, the balance coefficient is significant at a delay; in contrary to the same coefficient when applying a linear regression on the model; it is negative and ready for the following conclusion: an improvement in the balance is unfavorable to the activity; this situation therefore describes a Keynesian nature. The consideration of the conjunctural phase makes it possible to highlight a Keynesian effect. In a recession and a balanced budget, the coefficient on the budget balance with one delay is significantly different from 0 and becomes positive: the improvement in the budget balance improves economic activity. This is classic nature according to the terminology of the study. This work illustrates the contribution of the Markovian regime change model and accounts for the various forms of regime-dependent effects. The MS-VAR methodology highlighted specific effects that did not emerge when applying linear regression to our model. The MS-VAR methodology also has limitations, especially when short-lived macroeconomic samples are applied.

References Afonso, A., & Sousa, R. M. (2009, janvier). The macroeconomic effects of fiscal policy. European Central Bank Working Papers Series, 991. Badinger, H. (2006, novembre). Fiscal shocks, output dynamics and macroeconomic stability: An empirical assessment for Austria. JEL, C22, C32, E60. Biau, O., & Gerard, E. (2004, September). Budgetary policy and economic dynamics in France: The VAR structural approach. Afse. Blanchard, O., & Perotti, R. (2002). An empirical characterization of the dynamic effects of changes in government spending and taxes on output, forthcoming. Quarterly Journal of Economics, 117(4), 1329–1368. Corsetti, G., Meier, A., & Müller, G. J. (2012). What determines Government spending multipliers? International Monetary Fund (IMF) WP, 12/150, June. Goldstein, M., & Khan, M. (1976). Large versus small price changes and the demand for imports. IMF Staff Papers, 23, 200–225. Hamilton, J. D. (1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica, 57(2), 357–384. JSTOR, www.jstor.org/stable/1912559. Hénin, P., & N’Diaye, P. (2001). The effect of fiscal policies on activity: A function of cyclical conditions and the budgetary regime. Economy and Forecast, 147. Mihov, I., & Fatas, A. (2001, April). The effects of fiscal policy on consumption and employment: Theory and evidence. CEPR Discussion Paper Series, 2760. Mountford, A., & Uhlig, H. (2002). What are the effects of fiscal policy shocks? mimeo, février. Reinhart, C. M., & Kenneth, S. R. (2010). Growth in a time of debt. American Economic Review, 100(2), 573–578. https://doi.org/10.1257/aer.100.2.573 Sims, C. A. (1980). Macroeconomics and reality. Econometrica, 48(1), 1–48. JSTOR, www.jstor.org/stable/1912017

Structure of Bond Pension Funds During Decreasing Yield Curves Mário Papík

Abstract Pension system in Slovakia has been transformed from a single-pillar system to a three-pillar system since 2004. Following outbreak of financial crisis between years 2008 and 2009, pension sector has been subjected to several regulatory changes in order to protect retirement savings. Bond pension funds are currently the most dominant element in pension system with 70% of total assets allocated in this kind of pension funds. Aim of this manuscript is to analyse the structure of assets owned by bond pension funds and to identify key portfolio components that impact returns of these funds. This relationship has been analysed on sample of all five bond pension funds operating in Slovakia for period from 2009 to 2017. Relationship between individual components of portfolio and portfolio returns has been described by two linear regression models with mixed effects. Statistically significant variables have been identified as bonds evaluated at fair value with maturity less than 3 months and bonds evaluated at fair value with maturity longer than 5 years, both denominated in Euros. This manuscript has showed that bond pension portfolios, that are evaluated mainly at fair value and with long maturity, are prone to increased interest rate risks in than in the past. Higher interest rate risk could have negative impact on pension’s savings in the future. Keywords Pension funds · Asset allocation · Yield curve · Bonds

1 Introduction Pension system reforms have taken place at the breakthrough of millennium in many Central and Eastern European countries. Demographic changes such as reduced fertility of population and longer life span have resulted in necessity of pension system reforms comparable to reforms already conducted in Western

M. Papík () Comenius University, Bratislava, Slovakia © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_7

95

96

M. Papík

countries (Mitková, 2016). Negative trends in demography and complicated labour migration from foreign countries make Slovakia no different than other Central and Eastern European countries (Saxunová & Chorvatoviˇcová, 2018). Pay-as-yougo (PAYG) pension system has been applied in post-communist countries whilst Western countries have reformed their pension systems to multi-pillar private pension systems. In 2018, assets managed by pension funds in Western countries represented 198.6% of GDP in Denmark, 173.3% in the Netherlands and 161% in Iceland. In contrast, amount of assets managed by pension funds relative to GDP is around 11% in Slovakia. Among all post-communist countries, Estonia reports highest amount of assets managed by pension funds relative to GDP (17%) whilst average OECD amount is approximately 50% (OECD, 2019). Majority of pension system resources in Slovakia (over 70%) are allocated in bond pension funds that guarantee nominal value of portfolio. Participation of contributors in this type of pension fund is result of Amendment no. 137/2009. During financial crisis in 2008 and 2009, all contributors have been transferred from growth funds (currently named equity funds) to conservative (currently named bond funds) funds. Contributors could have still chosen type of fund on individual basis, even though, but most of them remained in conservative pension funds as assigned by the government regulations. Slovak pension funds are legally limited in their ownership and can only own money market instruments and debt securities. As interest rates in the European Union are close to zero, Exchange Traded Fund (ETF) or various bond funds might be an alternative to these financial market instruments. Guaranteed pension funds achieve lower returns in long run in comparison with returns of equity pension funds. Returns of bond pension funds consist mainly of revenues from fixed payments (coupons and interest), revaluation differences of securities and differences arising from foreign exchange revaluation. As there is currently no academic research focusing on significant portfolio components of these institutions, aim of this manuscript is to analyse structure of assets owned by bond pension funds and to identify key portfolio components impacting financial returns of these pension institutions. This manuscript is divided into six chapters. First chapter includes introduction and second chapter includes literature review. Chapter three formulates aim of this manuscript. Methodology is included in fourth chapter and fifth chapter includes results of analysis. The last chapter describes conclusions of this study.

2 Literature Review Study of Chovancová, Hudcovský, and Kotasková (2019) has proved that returns of pension funds are significantly correlated with bond indices. This proves assumption that fixed income securities are the main component of pension funds. This

Structure of Bond Pension Funds During Decreasing Yield Curves

97

dependence is even stronger than that on stock indices. This relationship may also depend on the fact that bond funds, mixed funds and part of equity funds in Slovakia contain very high proportion of bond components in portfolios. Naczyk and Domonkos (2016) have compared impact of legislation changes on allocation of assets in pension funds. Their research has showed slight increase in government and domestic bonds during post-crisis years in Slovakia. This research has also showed decline of stock and foreign securities in portfolios of these funds. Changes of portfolio structure may be result of amendments to the pension savings system implemented during this period. Government interventions also negatively affect performance of pension funds in Croatia (Draženovi´c, Hodži´c, & Maradin, 2019). Similarly, Witkowska, Kompa, and Mentel (2019) has showed that every change in legislation equals to increase in portfolio volatility during the postgovernment intervention period. Papík (2017) has showed that eight portfolio components have high representation in Slovak equity and mixed funds. These components are: bonds at market value up to 1 year, up to 5 years and more than 5 years, followed by shares and stocks, both denominated in Euros, US dollars and Czech crowns and deposits in banks in Euros. Statistically significant variables impacting returns of pension funds are bonds at market value over 5 years in Euros, stocks at market value in Euros, stocks at market value in US dollars, stocks at market value in Czech crowns, shares at market value in Euros and shares at market value in US dollars. This research has also shown that returns of equity and mixed funds also largely depend on equity securities. Research conducted by Louton, McCarthy, Rush, Saraouglu, and Sosa (2015) on composition of US pension funds has confirmed positive relationship between US equities and real estate, US equities and other assets, non-US equities and real estate, and non-US equities and other assets. Amir and Benatzi (1998) have proved existence of relationship between equities and expected rate of return of pension funds in the US. The lower the volume of equity securities, the lower the expected rate of return. This study has also showed that returns of pension funds do not depend on volume of assets these funds own. In contrast to Amir and Benatzi research, Li and Klumpes (2013) have showed that size of pension funds, in terms of value of managed assets, impacts expected rate of return the UK. Studies focused on impact of accounting standards on structure of pension fund portfolios are separate group of studies. Amir, Guan, and Oswald (2010) has showed once accounting standards IFRS 17 (Insurance contracts) and IAS 19 (Employee Benefits) have been applied, there is decline in ownership of equity securities in UK pension funds. Equity securities have been replaced by debt securities. This shift could be result of increases in funding levels, effective tax rates, financial leverage and investment horizons that have taken place during same period as transition to these accounting standards. Similar results have also been achieved by Anantharaman and Chuk (2017). Their study has proved that after implementation of IAS 19R, risk-taking in pension investments in Canadian companies has decreased.

98

M. Papík

Burke, Chen, and Eaton (2016) has proved that asset evaluation through markto-market (MTM) method brings financial transparency. On the other hand, MTM evaluation has negative impact on reduction of earnings persistence and comparability outweighs. This leads to increase in net costs for future financial reporting. Slovak Accounting Standards (SAR) require bond securities to be evaluated at market value or amortized cost. If bonds be evaluated at market value (MV), their price depends on movement of yield curve. If bonds are evaluated at amortized cost (AC), they are evaluated at effective interest rate.

3 The Aim of the Manuscript Aim of this manuscript is to analyse structure of assets owned by bond pension funds and to identify key portfolio components impacting financial returns of these institutions. This relationship has been analysed on a sample of all five bond pension funds currently operating in Slovakia. Data for individual pension funds have been collected for period from 2009 to 2017. Individual components of pension fund portfolios, relationship among these components and financial revenues have been described through linear regression model with mixed effects.

4 Research Methodology Data have been collected for all Slovak bond pension funds between 2009 and 2017 to fulfil the aim of this manuscript. Dataset includes selected financial information from balance sheet, income statement and notes to financial statements that have been prepared in accordance with Slovak Accounting Standards. Bond pension funds included in dataset have been managed by pension management fund companies Allianz, AXA, Poštová banka, NN and VÚB Generali. Company Aegon has been excluded from this research as its merger with NN meant all historical data is no longer publicly available. Two linear models with mixed effects have been developed to analyse relationship between asset accounts (mixed effects) and revenues from financial investments of pension funds. Developed models have also analysed random slopes. In the first model random slope is represented by time factor, and in the second model, random slope is represented by time and particular pension fund management company (Field, Miles, & Field, 2012). First developed model (1) with random slope time has following form: Yi = β1 X1,t + · · · + βk Xk,t + δ1 T1 + · · · + δt Tt + εt

(1)

where Yit is a dependent variable representing revenues from financial investments, i.e. sum of interest revenue, interest expense, net profit or loss from financial

Structure of Bond Pension Funds During Decreasing Yield Curves

99

operations, exchange rate differences arising from financial assets. Index t denotes specific point in time from 2009 to 2017. Variable Xk, t is an independent variable which represents percentage of financial assets to total assets of pension fund. Variable β k represents coefficients of regression line for respective independent variable, and k denotes number of independent variables tested. For bond pension funds k equals 13 variables. Variable Tt is a dummy variable of binary character expressing individual years during the analysed period from 2009 to 2017. δ t is a random effect factor for binary time variable. The last variable εt reflects value of corresponding residual component of linear regression model with mixed effects for particular time t. Second developed model (2) with random slopes time and pension fund management company has the following form: Yit = β1 X1,it + · · · + βk Xk,it + δ1 T1 + · · · + δt Tt + γ1 E1 + · · · + γn En + εit (2) Variables are same as variables in model (1). Index i denotes specific entity— pension fund management company. Variable En is a dummy binary independent variable specific to each entity. γ n represents random effect factor for binary dummy variable. Variable n then represents total number of institutions considered in each of financial sectors. For pension fund sector, it has been five pension fund management companies. Coefficients of models (1) and (2) have been estimated by statistical software R studio. Linear regression model with mixed effects package has been used for estimation. Final model has been tested for absence of multi-collinearity. Accuracy of model has been verified by chi-square of goodness-of-fit test, the log-likelihood ratio function, Akaike Information Criteria, and Bayesian Information Criteria. Estimated values have been tested by Shapiro-Wilk test of normal data distribution (Field et al., 2012).

5 Results of Analysis Returns of actively managed bond pension funds have been guaranteed. Therefore, pension funds have mainly owned safe assets during the analysed period. These pension funds have been regulated by Amendment no. 137/2009 and by Amendment no. 252/2012. Government regulations restrict ownership of bond pension funds to bonds, financial investments, and transactions designed to restrict foreign exchange and interest rate risks. Therefore over 75% of assets allocated in pension funds are debt securities and 20% are bank deposits (National Bank of Slovakia, 2010). Newly introduced feature of bond pension funds are ETF securities that copy yield of bond indices in order to increase profitability. Their representation among all assets has reached 6% in 2017 and has been used in half of bond pension funds (National Bank of Slovakia, 2018).

100

M. Papík

Fig. 1 Euribor with 1 year, 3 years, and 5 years maturity. Source: Own calculation based on Eurostat data

Interest rates in Eurozone have decreased firstly after outbreak of financial crisis between years 2008 and 2009, and then following outbreak of the debt crisis in Europe in 2011. Intention of European Central Bank (ECB) to decrease interest rates has been stimulation of economic growth. Ever since, interest rates have been decreasing until nowadays. Nowadays, 3-year interest rates reach negative values and 5-year interest rates are slightly above 0. Evolution of the 1, 3, and 5-year Euribor rates is shown in Fig. 1. Extension of bond maturities is caused by decreasing interest rates as price of debt securities increases with decreasing interest rates. This situation is depicted in Table 1. Table 1 lists asset components of bond pension funds in each year. As already mentioned, residual maturity of bonds has increased by more than 2 years. This increase in residual maturity has also caused increase in modified duration indicator, which increased to 3.4 for bond portfolios. Number of bonds held to maturity (evaluated by amortized costs) has increased to almost 15% since 2012. Benefit of holding bonds to maturity in a portfolio is no change in value of these bonds in case market interest rates change. Therefore, in the event of an unexpected upward shift of yield curve, value of bond portion of portfolio would not decrease. Growth of this component in asset structure of pension funds suggests that funds are trying to protect themselves from eventual interest rate risks. Volume of government debt securities owned by pension funds has decreased by 10%, to current 50% of total assets. This decline can be explained by increased interest of pension fund management in bonds of financial and non-financial institutions. Volume of bonds issued by non-financial institutions is comparable to that of financial institutions reaching up to 25% of total assets owned by pension funds. The most commonly held foreign currency throughout analysed period has been US dollar, Polish zloty, and Czech crown.

Structure of Bond Pension Funds During Decreasing Yield Curves

101

Table 1 Bond pension funds financial assets Year Market value Bond securities EUR Up to 1 month Up to 3 months Up to 6 months Up to 1 year Up to 2 years Up to 5 years Over the 5 years USD Up to 5 years Over the 5 years CZK Up to 5 years Over the 5 years PLN Up to 5 years Over the 5 years Shares EUR USD Stocks EUR Amortized cost Bond securities Up to 1 month

2009 2010 2011 2012 2013 2014 2015 2016 2017 71.78% 73.06% 74.11% 62.45% 70.60% 73.96% 73.35% 79.51% 72.75% 71.78% 69.98% 76.49% 61.86% 65.13% 68.87% 67.37% 71.56% 64.28% 71.78% 69.98% 76.49% 61.86% 64.93% 68.10% 66.46% 70.00% 63.37% 0.51% 3.27% 3.36% 2.01% 0.01% 0.01% 0.02% 0.00% 0.00% 19.64% 1.71%

1.48%

0.04%

0.03%

0.04%

0.04%

0.00%

0.00%

0.00%

2.66%

0.47%

0.02%

0.03%

0.03%

0.00%

0.00%

33.23% 30.12% 16.43% 10.68% 0.45%

0.02%

0.02%

0.00%

0.00%

0.00%

1.54%

1.44%

0.70%

0.30%

0.03%

8.98%

9.50%

9.46%

1.34%

15.95% 19.28% 25.50% 25.09% 24.63% 27.15% 21.03% 15.94% 11.32% 2.45%

6.59%

17.55% 14.09% 38.43% 39.31% 43.88% 53.36% 51.75%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.20% 0.02%

0.45% 0.20%

0.58% 0.18%

0.96% 0.57%

0.29% 0.18%

0.00%

0.00%

0.00%

0.00%

0.18%

0.26%

0.39%

0.39%

0.11%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.03% 0.03%

0.03% 0.03%

0.07% 0.03%

0.07% 0.03%

0.00%

0.00%

0.00%

0.00%

0.00%

0.00%

0.00%

0.04%

0.04%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.00% 0.00%

0.29% 0.22%

0.31% 0.08%

0.53% 0.00%

0.55% 0.00%

0.00%

0.00%

0.00%

0.00%

0.00%

0.06%

0.22%

0.53%

0.55%

0.00% 0.00% 0.00% 0.00% 0.00% 0.00%

0.00% 0.00% 0.00% 0.00% 0.00% 0.00%

0.00% 0.00% 0.00% 0.00% 0.00% 0.00%

0.43% 0.43% 0.00% 0.16% 0.16% 7.15%

0.53% 0.53% 0.00% 4.94% 4.94% 5.09%

0.13% 0.13% 0.00% 4.96% 4.96% 7.70%

0.37% 0.37% 0.00% 5.60% 5.60% 9.76%

1.55% 1.41% 0.14% 6.39% 6.39% 10.32%

2.34% 2.00% 0.34% 6.00% 6.00% 14.54%

0.00% 0.00% 0.00% 7.15% 5.09% 7.70% 9.76% 10.32% 14.54% 0.00%

0.00%

0.00%

0.00%

0.00%

0.01%

0.01%

0.00%

0.00%

(continued)

102

M. Papík

Table 1 (continued) Up to 5 years Over the 5 years Loans and receivables Short-term receivables EUR USD Long-term receivables EUR Cash and cash equivalents EUR PLN USD

0.00%

0.00%

0.00%

0.53%

0.00%

0.00%

0.27%

0.00% 0.00%

0.00%

0.00%

0.00%

6.62%

5.09%

7.69%

9.47%

10.32% 14.54%

20.88% 24.83% 22.89% 27.30% 20.82% 17.11% 14.20% 9.14% 11.49% 20.48% 22.97% 21.43% 26.66% 19.93% 11.05% 4.82%

5.32% 8.18%

20.38% 21.30% 21.43% 26.66% 19.93% 10.18% 4.81% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.01% 0.40% 1.86% 1.46% 0.64% 0.89% 6.06% 9.38%

5.32% 7.26% 0.00% 0.00% 3.82% 3.31%

0.40% 1.86% 1.46% 0.64% 0.89% 6.06% 9.38% 3.82% 3.31% 7.33% 2.11% 2.41% 3.01% 3.49% 1.24% 2.14% 0.49% 0.77%

7.33% 0.00% 0.00%

2.11% 0.00% 0.00%

2.41% 0.00% 0.00%

3.01% 0.00% 0.00%

3.46% 0.00% 0.02%

1.16% 0.00% 0.08%

2.13% 0.00% 0.01%

0.46% 0.76% 0.00% 0.01% 0.03% 0.00%

2015 0.75% 0.55%

2016 2.24% 1.70%

Source: Own calculation based on annual financial statements Table 2 Bond pension funds’ returns and profits Returns Profit before income taxes

2009 1.10% 1.10%

2010 0.88% 0.87%

2011 1.12% 1.12%

2012 1.75% 1.49%

2013 0.60% 0.32%

2014 3.44% 2.84%

2017 1.52% 1.10%

Source: Own calculation based on annual financial statements

Interest revenue, received coupons, differences arising from purchase or sale of securities and exchange rate differences are main revenue sources for bond pension funds. As Table 2 clearly shows, development of returns of pension funds has been oscillating with increasing trend depending on situation on financial markets. Government deregulations for pension funds have been main source of revenue growth. Consequently, pension funds have been able to claim various payments related to management of these funds. Despite unfavourable trends during financial crisis, nominal value of bond pension funds has gradually increased since crisis period. Figure 2 provides closer look at profits of pension funds generated during financial year. Stagnation of profits is clear during crisis throughout 2009–2012. Profit decline in 2013 is caused by increased costs associated with merger of growth, balanced and conservative funds into one guaranteed bond fund. Pension fund of management company VÚB Generali has been the only fund that has generated accounting loss in this period. Profit levels oscillated with slightly increasing trend

Structure of Bond Pension Funds During Decreasing Yield Curves

103

Fig. 2 Coplot graph for returns. Source: Own calculation in R Studio

for all the analysed funds. Except for pension fund managed by VÚB Generali pension fund management company, all coplot charts in Fig. 2 show that profit trends for all the analysed bond funds have been following similar trends. Results for linear model with mixed effects (1) included in Table 3 show that only bonds valued at market value with maturity up to 3 months (t (43.79) = 1.788; p < .1) and bonds valued at market value with maturity more than 5 years (t (36.47) = 5.155; p < .05), both denominated in Euros, have statistically significant effect on profits of Slovakia pension funds. Coefficients of both variables are positive, indicating their positive impact on analysed variable. Results are also confirmed by calculated confidence intervals for statistically significant variables. Statistically significant variables have achieved values higher than zero on confidence intervals 95%. Higher percentage of these assets in pension funds’ portfolios is reflected in the highest profit among other variables. Coefficients of tested variables and results of further analysis are shown in Table 3. Random effects from Table 3 show that, during first half of analysed period, pension funds have generated higher profits than population average during whole analysed period. At the end of analysed period (2015–2017), results have been lower than the population average. This has been reflected in developed model with negative coefficient for particular random effect. Graphical interpretation with comparison of fitted vs. residuals indicates homoscedasticity of developed model as seen in Fig. 3. Values of developed model are normally distributed as seen on the left part of Fig. 3a. All estimated values in Fig. 3a have approximately same standard deviation. The Shapiro-Wilk test of normality has not rejected null hypothesis (W = 0.97241; p = 0.35) that residual values are of normal distribution. These results are also confirmed visually in right part of the Fig. 3b normal Q-Q figure. This figure shows that even existence of outliers does not affect normality of developed model. Coefficient of determination is 0.65, and therefore this model is able to describe up to 65% of analysed data for guaranteed bond pension funds.

104

M. Papík

Table 3 Results of linear model with mixed effects (1) Fixed effects Bonds at MV up to 1 month in Euros Bonds at MV up to 3 months in Euros Bonds at MV up to 6 months in Euros Bonds at MV up to 1 year in Euros Bonds at MV up to 2 years in Euros Bonds at MV up to 5 years in Euros Bonds at MV more than 5 years in Euros Stocks at MV in Euros Shares at MV in Euros Bonds at AC up to 1 month in Euros Bonds at AC more than 5 years in Euros Receivables Cash Random effect (time) 2009 2010 2011 2012 2013 2014 2015 2016 2017 R-squared Residual deviance Degrees of freedom residuals Akaike Inf. Criterion Bayesian Inf. Criterion Log-likelihood ratio test Shapiro-Wilk norm. test (p value)

Estimate 0.018

Std, error 0.026

df 36.512

t value 0.662

Pr(>|t|) 0.512

P value

0.033

0.018

43.789

1.788

0.081

.

0.011

0.076

41.785

0.147

0.884

−0.003

0.015

43.418

−0.206

0.838

0.005

0.020

37.402

0.262

0.795

−0.007

0.011

44.392

−0.618

0.540

0.035

0.007

36.468

5.155

0.000

0.015 −0.055 −0.008

0.025 0.040 0.017

44.435 33.773 41.569

0.609 −1.367 −0.484

0.546 0.181 0.631

0.006

0.007

44.617

0.876

0.386

0.035 0.018

0.039 0.026

44.990 36.512

0.898 0.662

0.374 0.512

0.0023 0.0049 0.0045 0.0074 −0.0104 0.0136 −0.0084 −0.0005 −0.0032 0.65 −304.80 31 −276.80 −251.51 152.4 0.353

Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Source: Own calculation in R Studio based on annual financial statements

***

Structure of Bond Pension Funds During Decreasing Yield Curves

105

Fig. 3 (a) Fits vs. residuals plot and (b) normal Q–Q plot for commercial banks’ model. Source: Own calculation in R Studio

Analysis has further extended developed linear model (1) with mixed effects by random effect into another model (2). Another random effects are pension fund management companies. Results of variance analysis by ANOVA function in R studio statistic have showed that there is no statistically significant difference (SD = −306.1; χ 2 (1) = 1.299; p = 0.2544) between linear model with mixed effect with time as only random effect (1) and linear model with mixed effects with time and pension fund management companies as random effects (2). Hypothesis that pension fund management company affects profits of this company has been rejected. Profits of pension funds are hence affected by both volume of 3-monthdenominated market value bonds denominated in Euros and bonds valued at more than 5 years’ market value also denominated in Euros. Profits of pension funds are also affected by time factor that affects all pension funds and can be interpreted as situation on financial markets.

6 Conclusions This manuscript has shown that key components of portfolio with significant effect on returns of bond pension funds are bonds valued at market value with maturity of 3 months denominated in Euros and bonds valued at market value with maturity of more than 5 years also denominated in Euros. Developed model has showed that revenues of pension funds depend on time factor. Time factor, in this manuscript, represents macroeconomic trends. Ability to manage portfolio of bond pension funds therefore does not affect their returns. This leads to a suggestion that contributors might rather consider macroeconomic trends and focus on structure of individual pension funds than selecting a fund by its managing institution.

106

M. Papík

Finding that pension funds in Slovakia have been influenced by external trends is in accordance with other studies showing correlation between returns of pension funds and bond indices. Dominant portfolio components of bond pension funds, like of equity and mixed pension funds, are bonds with longer times to maturity evaluated at market value. Non-guaranteed pension funds in Slovakia therefore, whilst having different investment strategy, contain significant portion of portfolio comparable to portfolio of guaranteed pension funds. Possible limitation of this manuscript is relatively small data sample. Even though data sample represents entire market of bond pension funds in Slovakia, it still consists of only five bond funds. Another limitation of this research could be Slovak Accounting Standards. Slovak Accounting Standards are very similar to IFRS, but still have minor differences and local specifics that could lead to differences when compared with global pension funds. Future studies should address factor of accounting standards that does have significant impact on portfolio management, specifically in the process of bond evaluation. With increasing interest rates, evaluation of bonds at market value rather than evaluation at amortized cost might have impact on returns of portfolios. Both portfolio managers and contributors should consider this finding in their future investment decisions.

References Amendment no. 137/2009: ktorým sa mení a dop´lˇna zákon cˇ . 43/2004 Z. z. o starobnom dôchodkovom sporení a o zmene a doplnení niektorých zákonov v znení neskorších predpisov. Retrieved February 26, 2020, from http://www.socpoist.sk/ext_dok-137-2009-z-zp-32911/ 54480c Amendment no. 252/2012: ktorým sa mení a dop´lˇna zákon cˇ . 461/2003 Z. z. o sociálnom poistení v znení neskorších predpisov a ktorým sa menia a dop´lˇnajú niektoré zákony. Retrieved February 26, 2020, from http://www.socpoist.sk/ext_dok-12-z252/56750c Amir, E., & Benatzi, S. (1998). The expected rate of return on pension funds and asset allocation as predictors of portfolio performance. The Accounting Review. Amir, E., Guan, Y., & Oswald, D. (2010). The effect of pension accounting on corporate pension asset allocation. Review of Accounting Studies, 15, 345. https://doi.org/10.1007/s11142-0099102-y Anantharaman, D., & Chuk, E. (2017). The economic consequences of accounting standards: Evidence from risk-taking in pension plans. The Accounting Review. https://doi.org/10.2139/ ssrn.2629277 Burke, Q. L., Chen, P. C., & Eaton, T. V. (2016). An empirical examination of mark-to-market accounting for corporate pension plans. Journal of Accounting and Public Policy., 36, 34. https:/ /doi.org/10.1016/j.jaccpubpol.2016.11.001 Chovancová, B., Hudcovský, J., & Kotasková, A. (2019). The impact of stocks and bonds on pension fund performance. Journal of Competitiveness, 11(2), 22–35. https://doi.org/10.7441/ joc.2019.02.02 Draženovi´c, B. O., Hodži´c, S., & Maradin, D. (2019). Efficiency of mandatory pension funds: Case of Croatia. South East European Journal of Economics and Business, 14(2), 82–94. Field, A., Miles, J., & Field, Z. (2012). Discovering statistics using R. United Kingdom. London: Sage.

Structure of Bond Pension Funds During Decreasing Yield Curves

107

Li, Y., & Klumpes, P. J. (2013). Determinants of expected rate of return on pension assets: Evidence from the UK. Accounting and Business Research, 43, 3. https://doi.org/10.1080/ 00014788.2012.685286 Louton, D., McCarthy, J., Rush, S., Saraouglu, H., & Sosa, O. (2015). Tactical asset allocation for US pension investors: How tactical should the plan be? Journal of Asset Management, 16, 427. Mitková, L. (2016). Europe’s ageing population and the gender pension gap. In Európska ekonomická integrácia v kontexte aktuálneho vývoja a výziev pre cˇ lenské štáty Európskej únie (pp. 68–80). Wolter Kluver. Naczyk, M., & Domonkos, S. (2016). The financial crisis and varieties of pension privatization reversals in Eastern Europe. Governance, 29, 167. https://doi.org/10.1111/gove.12159 National Bank of Slovakia. (2010). Analýza slovenského finanˇcného sektora za rok 2009. Bratislava: Národná banka Slovenska. National Bank of Slovakia. (2018). Analýza slovenského finanˇcného sektora za rok 2017. Bratislava: Národná banka Slovenska. OECD. (2019). Pension markets in focus 2019. Retrieved from www.oecd.org/daf/fin/privatepensions/pensionmarketsinfocus.htm Papík, M. (2017). Composition of equity and mixed pension funds in Slovakia. In Oeconomia Copernicana. Faculty of Economic Sciences and Management. Toru´n: University in Toru´n. Saxunová, D., & Chorvatoviˇcová, L. (2018). Management of labour force movement applied in Slovakia. Social and Economic Revue, 16(2), 35–43. Witkowska, D., Kompa, K., & Mentel, G. (2019). The effect of government decisions on the efficiency of the investment funds market in Poland. Journal of Business Economics and Management, 20(3), 573–594. https://doi.org/10.3846/jbem.2019.9861

Extracting Common Factors from Liquidity Measures with Principal Component Analysis on the Polish Stock Market Joanna Olbrys and Elzbieta Majewska

Abstract According to the literature, the principal component analysis (PCA) can be utilized to extract common features of a set of economic variables. Therefore, the aim of this research is to assess a possibility of using the PCA to extract common components of liquidity across a sample of equities, and from a set of liquidity measures on the Polish stock market. The database contains the group of 86 WSE-listed companies. Seven liquidity proxies, namely percentage relative spread, percentage realized spread, percentage price impact, percentage order ratio, modified version of the Roll estimator, modified turnover and modified Amihud measure, are utilized in the study. The PCA results reveal that common latent factors in liquidity estimates exist on the Polish stock market. Moreover, the robustness analyses confirm the evidence of common sources in liquidity variation within the whole sample period and three subsamples: the pre-crisis, crisis and post-crisis periods. To the best of the authors’ knowledge, all empirical findings reported here are novel and have not been presented in the previous literature. Keywords PCA · Liquidity measure · High-frequency data · Daily data · Warsaw stock exchange JEL codes: C10, C58, G01, G10, G12

J. Olbrys () Bialystok University of Technology, Bialystok, Poland e-mail: [email protected] E. Majewska University of Bialystok, Bialystok, Poland e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_8

109

110

J. Olbrys and E. Majewska

1 Introduction Financial markets liquidity is especially difficult to explore as it is unobservable. Moreover, some researchers emphasize that liquidity measures capture only noisy information about liquidity (e.g. Chen, 2005). Since Chordia, Roll, and Subrahmanyam (2000), the market microstructure literature, has shifted its focus from the assessment of liquidity of individual equities towards analyses of common determinants and components of liquidity across securities. According to the literature, the principal component analysis (PCA) has widespread applications as it unveils simple underlying structures in complex data sets using analytical solutions from linear algebra. It can be utilized to extract common features of a set of economic variables. Therefore, one of possible applications of the PCA could be an extraction of latent factors from a set of liquidity measures across securities on equity markets. The vast majority of research on common components of liquidity with the PCA concerns developed markets. Hasbrouck and Seppi (2001) analyse the U.S. stock market, and they use the PCA to show that common factors exist in order flows and returns of the 30 equities in the Dow Jones Industrial Average (DJIA). Chen (2005) stresses that although there are various liquidity proxies which are substantially different in theory, they share common sources of liquidity variation on the New York Stock Exchange (NYSE). The author examines seven liquidity proxies and employs the PCA to extract the first principal component, which captures 62% of standardized liquidity variance. Korajczyk and Sadka (2008) utilize the asymptotic PCA to extract the common, systematic components of liquidity across a large sample of stocks and from a set of liquidity measures on the NYSE. They assess commonality in liquidity with principal components of liquidity proxies as latent factors. Foran, Hutchinson, and O’Sullivan (2015) investigate liquidity commonality and pricing in the United Kingdom (UK) stock market. They apply the asymptotic PCA on the sample of equities to extract market or systematic liquidity factors. Von Wyss (2004) conducts the research on liquidity in the Swiss Exchange. Among other things, he provides and interprets the results of the PCA for each stock, separately. In general, the studies concerning commonality in liquidity on emerging markets in the world are scarce, and, in particular, they do not utilize the PCA to identify latent factors in liquidity. For example, B˛edowska-Sójka (2019) and Olbry´s (2019) deeply explore commonality in liquidity on the Warsaw Stock Exchange (WSE) by using different econometric methods and various liquidity proxies, but they do not employ the PCA. Therefore, the main goal of this study is to assess a possibility of using the PCA to extract common components of liquidity across a sample of equities, and from a set of liquidity estimates on the Polish emerging stock market. To the best of the authors’ knowledge, all empirical findings presented here are novel and have not been reported in the literature thus far.

Extracting Common Factors from Liquidity Measures with Principal. . .

111

The remainder of this study is organized as follows. Section 2 presents the methodological background of the PCA. In Sect. 3, liquidity/illiquidity proxies used in this study are described in detail. Section 4 contains data description and discusses the empirical results concerning common sources of liquidity variation on the WSE. The last section recalls the main findings, concludes and indicates further directions of the research.

Nomenclature %RS %RealS %PI %OR LR MAmih MRoll MT WSE

Percentage relative spread Percentage realized spread Percentage price impact Percentage order ratio The Lee and Ready (1991) trade side classification algorithm Modified Amihud measure Modified version of the Roll estimator Modified turnover The Warsaw Stock Exchange

2 Principal Component Analysis: Methodological Background The most common mathematical factor model is principal component analysis (PCA) (Brooks, 2019). The PCA is a multivariate non-parametric technique that analyses a data matrix in which observations are described by several intercorrelated quantitative variables (Abdi & Williams, 2010). The main idea is to reduce the dimensionality of a data set containing a large number of interrelated variables, while retaining as much as possible of the variation present in the data. Observable input variables are transformed into new unobservable (latent) variables called principal components, which are linear combinations of original variables. Consecutive principal components are uncorrelated with each other and are calculated in a way that maximizes variance that has not been explained by preceding principal components. Consequently, a few first components contain the vast majority of information that is present in the data set. Briefly, the PCA constructs common factors to maximize explanatory power within a set of related variables (Hasbrouck & Seppi, 2001). Suppose that X is a vector of N random variables, and a covariance (or correlation) matrix Ω of elements of vector X is known. The first step of the PCA is to look for a linear function α1T x of the elements of the vector X (called the first principal component) having a maximum variance and given by (1):

112

J. Olbrys and E. Majewska

z1 = α1T X = α11 x1 + α12 x2 + · · · + α1N xN =

N 

α1j xj ,

(1)

j =1

where α 1 = [α 11 , α 12, · · · , α 1N ]T is a vector of N factor loadings. The next step is to look for a linear function z2 = α2T x (called the second principal component) which is uncorrelated with z1 and has maximum variance, and the procedure is then repeated. The k-th derived variable is the k-th principal component, and it is given by (2): zk = αkT X, k = 1, 2, . . . , N,

(2)

where α k is an eigenvector of Ω corresponding to its k-th largest eigenvalue λk . Furthermore, if α k is chosen to have unit length, then var(zk ) = λk (Jolliffe, 2002, pp. 2–4). It is important to note that the PCA is scale-dependent. The principal components of a covariance matrix and those of a correlation matrix are different. In applied research, the PCA of a covariance matrix is useful only if the variables are expressed in commensurable units. When variables are measured along different scales or they differ significantly in terms of variability, it is advisable to calculate the components from the sample correlation matrix, which is analogous to standardizing all the variables prior to calculation (e.g. Jackson, 1991; Jolliffe, 2002). The goal of the PCA is to extract the majority of important information from a data matrix. The question is, how many components need to be considered? One of the standard procedures is the Kaiser (1958) criterion. For the correlation matrix used as input for the PCA, the Kaiser criterion suggests that only the principal components with an eigenvalue larger than unity, which means larger than the average eigenvalue, should be used (e.g. Abdi & Williams, 2010; von Wyss, 2004).

3 Liquidity Proxies Utilized in the Study This section presents seven liquidity proxies that are utilized in the research. Five out of them (i.e. percentage relative spread, percentage realized spread, percentage price impact, percentage order ratio, and modified version of the Roll estimator) are approximated using high-frequency (intraday) data. Furthermore, two liquidity measures are calculated from low-frequency (daily) data. These estimates are as follows: the modified daily turnover and the modified version of daily Amihud liquidity/illiquidity proxy.

Extracting Common Factors from Liquidity Measures with Principal. . .

113

3.1 Liquidity Proxies Derived from High-Frequency Data High-frequency financial data are useful in analysing a variety of topics related to trading processes. To estimate various liquidity/illiquidity proxies using intraday data, it is important to recognize the side initiating the transaction and to distinguish between buyer- and seller-initiated trades. The WSE is as an order-driven market with an electronic order book, but information about the best bid and ask price is not publicly available. As a consequence, researchers rely on indirect classification rules to infer the initiator of a trade. Various classification procedures of this type are presented in the literature, but the Lee and Ready (1991) algorithm (LR) remains the most frequently used.1 Table 8 in the Appendix describes details concerning the LR procedure. In this study, the LR algorithm is employed, as Olbry´s and Mursztyn (2015) confirmed that the LR performs quite well for data from the WSE. The empirical results turn out to be robust to the choice of the sample and do not depend on a firm’s size.2 In this research, five alternative liquidity/illiquidity proxies derived from intraday data are employed. Three of them, namely percentage realized spread, percentage proxy for price impact, and percentage order ratio, are supported by a trade side classification procedure (Olbry´s & Mursztyn, 2017). Moreover, two other estimates based on high-frequency data are utilized in the study: percentage relative spread and the modified version of the Roll (1984) estimator for the effective spread. Table 1 contains definition of these five liquidity estimates, and it needs some comments. Percentage Relative Spread is a measure of illiquidity because a high value of this indicator denotes low liquidity, while a low value indicates high liquidity. The value of daily percentage relative spread is calculated as a volume-weighted average of percentage relative spreads computed over all the trades within a day (Olbry´s, 2019). Percentage Realized Spread is a temporary component of the effective spread, and it is sometimes referred to as a price reversal component since a dealer makes a profit only if the price reverses (e.g. Huang & Stoll, 1996). Daily percentage realized spread value is calculated as a volume-weighted average of percentage realized spreads computed over all trades within a day, and it is defined as equal to zero when all transactions within a day are unclassified (Olbry´s & Mursztyn, 2017). Percentage Price Impact is usually defined as the increase (decrease) in the quote midpoint over a time interval beginning at the time of the buyer(seller-)-initiated trade. This is a permanent component of effective spread, e.g. (Goyenko, Holden, & Trzcinka, 2009). Daily proxy of percentage price impact is calculated as a volumeweighted average of percentage price impact estimates computed over all trades within a day. Moreover, it is defined as equal to zero when all transactions within a day are unclassified (Olbry´s & Mursztyn, 2017). Percentage Order Ratio is utilized as an indicator of imbalance in daily orders. The value of daily order ratio is defined to be equal to zero in the following two 1 For a brief literature review concerning various trade classification rules, see, e.g., Olbry´s and Mursztyn (2015). 2 For details about the C++ program for the LR procedure, see Olbry´s and Mursztyn (2015, p. 48).

114

J. Olbrys and E. Majewska

cases: (1) when all of the transactions within a day are unclassified, or (2) when the total volume of daily trading, in the denominator, is equal to zero (Olbry´s, 2019). The modified version of the Roll estimator for the effective spread is estimated using trade-to-trade price changes within a day, e.g. (Stoll, 2000). The original version of the Roll estimator is well-defined only for negative first-order serial covariance of price changes, which is not guaranteed in practice. Corwin and Schultz (2012) stress that the serial covariance of price changes is frequently positive and, in such cases, researchers could do one of the following things: (1) treat the observation as missing, (2) set the Roll estimator to zero, or (3) multiply the serial covariance by negative one, estimate the spread, and multiply the spread by negative one to obtain a negative spread estimate. In this study, the second proposition is used (see Table 1). Table 1 Definition of liquidity proxies derived from high-frequency data Liquidity proxy

Definition   200· PtH −PtL H L Pt +Pt

1

Percentage relative spread

%RSt =

2

Percentage realized spread

%RealSt =



%PIt =

200 · ln 200 · ln

⎧ ⎪ ⎨ 200 · ln

Pt Pt+5 , when trade t Pt+5 Pt , when trade t

mid Pt+5

is classified as buyer − initiated is classified as seller − initiated

, when trade t is classified as buyer − initiated

Ptmid Ptmid mid , when Pt+5

3

Percentage price impact

4

N Percentage %OR = 100 · n=1 Vn order ratio The modified  √ version of the 200 · − cov (Rt , Rt−1 ), when cov (Rt , Rt−1 ) < 0 Roll estimator MRollt = 0, when cov (Rt , Rt−1 ) ≥ 0

5

⎪ ⎩ 200 · ln

trade t is classified as seller − initiated

    m   i=1 VBuyi − kj =1 VSellj 

Abbreviations: PtH PtL are the high and low prices of a stock at time t, respectively, Pt is the transaction price of a stock at time t, approximated by the closing price, Pt+5 is the closing price of the fifth trade after trade t for a stock, P H +P L

mid t t P is the midpoint price at time t for a stock, 2 t m = i=1 VBuyi is the daily cumulative volume of trading related to transactions classified as buyer− initiated for a stock, k j =1 VSellj is the daily cumulative volume of trading related to transactions classified as seller−  initiated for a stock, N n=1 Vn is the daily cumulative volume of trading for all transactions for a stock, Rt is the logarithmic ultra − short rate of return of a stock at time t For all measures, the multiplier 100 converts units to percent Source: Authors’ elaboration based on Roll (1984), Olbry´s (2019) and Olbry´s and Mursztyn (2019)

Extracting Common Factors from Liquidity Measures with Principal. . .

115

Table 2 Definition of liquidity proxies derived from low-frequency daily data

1 2

Liquidity proxy

Definition

The modified version of the Amihud measure

MAmiht =

The modified version of daily turnover



 log 1 +

 MTt = log 1 +

|rt | Vt



, when Vt = 0

0, when Vt = 0 

1 30 · k=1 log 1 + − 30

Vt NSOt

Vt−k NSOt−k

Abbreviations: rt is the simple rate of return of stock on day t, Vt is the trading volume of stock on day t, NSOt is the number of shares outstanding of stock on day t. Source: Authors’ elaboration based on Amihud (2002), Karolyi, Lee, and van Dijk (2012) and Olbry´s (2019).

3.2 Liquidity Proxies Derived from Low-Frequency Daily Data Some measures are especially often advocated in the literature to provide empirical study in liquidity/illiquidity effects in low-frequency data. In this study, two proxies based on daily data are calculated: the modified version of the Amihud (2002) illiquidity measure and the modified version of daily turnover. Details are presented in Table 2, but they require a few comments. In the literature, the Amihud measure is usually calculated monthly or for other periods (e.g. Fong, Holden, & Trzcinka, 2017; Foran et al., 2015; Goyenko et al., 2009), but in this paper, daily values of this proxy are estimated (Olbry´s, 2019). The modified version of daily turnover is computed in logs and de-trend with a 30-day moving average to account for non-stationarity. The moving average is calculated for the available data over the past 30 trading days. It is important to note that using the modified version of this indicator disentangles day-of-the-week effects from daily turnover.

4 Data Description and Empirical Results on the WSE In the present study, two data samples are utilized. The first sample consists of daily data from January 2, 2005, to December 30, 2016, for the group of 86 WSEtraded companies. It contains 3005 trading days for each stock. The second sample contains high-frequency data rounded to the nearest second for the same period and the same group of equities. Both daily and intraday data sets include the opening, high, low and closing prices, and volume for each security over one unit of time. All details concerning the database are presented in the papers (Olbry´s, 2019; Olbry´s & Mursztyn, 2018), but it is important to note that a large number of the WSElisted firms unveil a substantial non-trading problem which affects the database content (Nowak & Olbry´s, 2016). This study is the continuation of the research on commonality in liquidity on the WSE presented in the paper (Olbry´s, 2019), and therefore the database is the same.

116

J. Olbrys and E. Majewska

To verify the robustness of the empirical findings, the calculations are conducted both for the whole sample and over three consecutive subsamples, each of equal length (436 trading days), e.g. (Olbry´s & Mursztyn, 2019, p. 185): 1. The pre-crisis period (from September 6, 2005, to May 31, 2007), 2. The Global Financial Crisis (GFC) period (from June 1, 2007, to February 27, 2009), 3. The post-crisis period (from March 2, 2009, to November 19, 2010). The GFC period on the WSE is precisely defined based on the paper (Olbry´s & Majewska, 2015).

4.1 Descriptive Statistics and Some Properties of Liquidity Proxies Time Series on the WSE The properties of liquidity proxies used in this research have been thoroughly investigated and reported in some recent papers. Olbry´s and Mursztyn (2018, 2019) have explored distributional properties, linear and non-linear dependences, as well as stationarity of daily time series of seven liquidity proxies that are utilized in the present study. The results have revealed some stylized facts and typical features of the analysed time series. Moreover, Olbry´s (2018) has tested stability of correlations between four liquidity estimates derived from high-frequency data within the whole sample period and three sub-periods: the pre-crisis, crisis and post-crisis periods, which are presented above in this section. Although the results have been not homogenous, they have indicated that liquidity measures seem to capture various sources of market liquidity. In fact, six liquidity proxies used in this study, namely (1) percentage relative spread, (2) percentage realized spread, (3) percentage price impact, (4) percentage order ratio, (5) modified version of the Roll estimator, and (6) modified Amihud measure, approximate illiquidity. Therefore, before the PCA calculations, the series are multiplied by negative one to obtain variables that are increasing alongside with liquidity of individual stocks. Only the modified version of daily turnover measures liquidity, while the remaining six estimates are transformed to be liquidity proxies. In the next step, the aggregate liquidity proxies are calculated as the cross-sectional averages of the corresponding individual-stock liquidity measures for the group of 86 equities within the whole sample period. Table 3 contains basic descriptive statistics for seven cross-sectional liquidity proxies.

Extracting Common Factors from Liquidity Measures with Principal. . .

117

Table 3 Basic descriptive statistics for cross-section of liquidity measures within the whole sample period from January 2005 to December 2016 (3005 trading days for each variable) 1 2 3 4 5 6 7

Liquidity proxy %RS %RealS %PI %OR MRoll MT MAmih

Mean −0.12926 −0.23750 0.08743 −38.54558 −0.67543 0.00043 −0.00413

Average std. dev. 0.03325 0.13818 0.08968 3.65778 0.17567 0.00042 0.00631

Max −0.05286 0.39079 0.65915 −19.66027 −0.26632 0.00501 −0.00004

Min −0.31897 −1.04905 −0.45153 −52.36167 −1.89240 −0.00086 −0.05014

Notes: Liquidity proxies are defined in Tables 1 and 2. ‘Mean’ is the cross-sectional average of the time series averages. ‘Average Std. Dev.’ is the cross-sectional standard deviation of the time series means of each measure Source: Own calculations Table 4 The PCA results from the seven liquidity proxies (the whole sample period) Principal component analysis Liquidity proxy 1 %RS 2 %RealS 3 %PI 4 %OR 5 MRoll 6 MT 7 MAmih Eigenvalue % variance explained % cumulative variance explained

PC1 0.235 0.661 −0.626 −0.060 0.315 −0.067 −0.086 2.014 28.77% 28.77%

PC2 −0.662 0.204 −0.342 −0.105 −0.521 0.304 0.172 1.326 18.94% 47.71%

PC3 −0.045 −0.005 −0.005 −0.583 −0.321 −0.617 −0.416 1.301 18.59% 66.30%

PC4 0.186 −0.018 0.042 −0.572 0.078 −0.048 0.792 0.965 13.79% 80.10%

PC5 0.240 −0.056 0.069 −0.514 0.023 0.719 −0.391 0.714 10.19% 90.29%

PC6 0.631 0.130 −0.029 0.231 −0.721 0.044 0.097 0.577 8.24% 98.53%

PC7 0.122 −0.708 −0.696 0.015 0.012 −0.018 0.0003 0.103 1.47% 100%

Notes: Liquidity proxies are defined in Tables 1 and 2. PC1–PC7 denote principal components. Eigenvalues larger than unity (PC1–PC3) are marked in bold Source: Own calculations

4.2 Common Components of Liquidity on the WSE In this subsection, the PCA empirical results from seven liquidity proxies for the group of 86 companies on the WSE are reported. In the PCA, input liquidity variables are converted into new (latent) uncorrelated liquidity variables called principal components, which are linear combinations of original variables. To get standardized factors, the correlation matrix is used as input for the PCA. Table 4 reports seven principal components PC1–PC7 corresponding to the first to smallest eigenvalues, the loadings of the principal components with the respective variance that the corresponding eigenvectors explain, and the cumulative variance explained within the whole sample period. The three eigenvalues that are larger than unity represent more than 66% of total liquidity variation. According to the Kaiser

118

J. Olbrys and E. Majewska

0.8

PC1

PC2

PC3

0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 %RS

%RealS

%PI

%OR

MRoll

MT

MAmih

Fig. 1 The first three eigenvectors of the PCA within the whole sample period from January 2005 to December 2016 (based on Table 4)

(1958) criterion, these three principal components are sufficient to substitute the seven liquidity proxies utilized in this study. Figure 1 presents the first three eigenvectors of the PCA within the whole sample period that are reported in Table 4. The factor PC1 with the highest eigenvalue explains about 28.8% of total variance. It combines rising liquidity of the %RS, %RealS, and MRoll with declining liquidity of the remaining measures. Factor two (PC2), which explains an additional 18.94% of variance, combines rising liquidity of the %RealS, MT, and MAmih with declining liquidity of the remaining proxies. Factor three (PC3) is responsible for another 18.59% of variance and captures declining liquidity of all liquidity estimates.

4.3 Robustness Analyses Tables 5, 6 and 7 report principal components PC1–PC7 corresponding to the first to smallest eigenvalues over three consecutive subsamples: the pre-crisis (Table 5), crisis (Table 6) and post-crisis (Table 7) periods. The results are homogenous because first three eigenvalues (PC1–PC3) are larger than unity in the case of all investigated periods. Moreover, the cumulative variance explained is almost the same within all considered sub-periods. Specifically, the first three eigenvalues PC1–PC3 represent more than 67% (the pre-crisis period), 66% (the crisis period) and 68% (the post-crisis period) of total liquidity variation, respectively.

Extracting Common Factors from Liquidity Measures with Principal. . .

119

Table 5 The PCA results from the seven liquidity proxies (the pre-crisis period) Principal component analysis Liquidity proxy 1 %RS 2 %RealS 3 %PI 4 %OR 5 MRoll 6 MT 7 MAmih Eigenvalue % variance explained % cumulative variance explained

PC1 −0.045 −0.703 0.694 0.089 −0.093 −0.070 −0.008 1.920 27.42% 27.42%

PC2 0.709 0.024 0.145 −0.163 0.596 −0.232 −0.201 1.504 21.48% 48.91%

PC3 −0.078 0.024 −0.030 −0.582 −0.400 −0.656 −0.252 1.327 18.95% 67.86%

PC4 0.093 −0.009 0.044 −0.320 0.057 −0.122 0.932 0.960 13.72% 81.58%

PC5 0.243 0.124 −0.040 0.715 −0.253 −0.567 0.166 0.765 10.92% 92.50%

PC6 −0.622 −0.084 −0.097 0.112 0.639 −0.419 0.010 0.454 6.48% 98.98%

PC7 0.185 −0.694 −0.695 0.003 −0.023 −0.007 0.009 0.071 1.02% 100%

PC6 −0.400 −0.094 0.024 −0.480 0.595 0.434 −0.240 0.626 8.94% 98.59%

PC7 0.097 −0.707 −0.700 0.013 0.014 −0.019 0.017 0.098 1.41% 100%

Notes: Notation like in Table 3 Table 6 The PCA results from the seven liquidity proxies (the crisis period) Principal component analysis Liquidity proxy 1 %RS 2 %RealS 3 %PI 4 %OR 5 MRoll 6 MT 7 MAmih Eigenvalue % variance explained % cumulative variance explained

PC1 −0.152 −0.694 0.678 0.078 −0.172 0.001 0.012 1.935 27.65% 27.65%

PC2 0.332 −0.059 0.104 −0.486 0.100 −0.643 −0.464 1.425 20.36% 48.00%

PC3 −0.584 0.075 −0.175 −0.410 −0.666 −0.087 −0.061 1.323 18.90% 66.91%

PC4 0.009 −0.029 0.054 −0.418 0.187 −0.252 0.850 0.902 12.89% 79.79%

PC5 −0.597 0.019 −0.072 0.429 0.357 −0.571 −0.026 0.690 9.86% 89.65%

Notes: Notation like in Table 3 Table 7 The PCA results from the seven liquidity proxies (the post-crisis period) Principal component analysis Liquidity proxy 1 %RS 2 %RealS 3 %PI 4 %OR 5 MRoll 6 MT 7 MAmih Eigenvalue % variance explained % cumulative variance explained Notes: Notation like in Table 3

PC1 −0.344 −0.580 0.520 0.323 −0.375 0.158 −0.068 2.109 30.14% 30.14%

PC2 0.420 −0.374 0.473 −0.346 0.353 −0.463 0.017 1.635 23.35% 53.49%

PC3 0.047 −0.057 0.078 0.112 0.192 0.266 0.932 1.019 14.56% 68.05%

PC4 −0.455 0.042 −0.084 −0.477 −0.445 −0.503 0.325 0.867 12.39% 80.44%

PC5 −0.012 0.092 −0.090 0.729 0.075 −0.661 0.099 0.699 9.99% 90.42%

PC6 0.694 0.093 0.029 0.064 −0.703 0.004 0.105 0.579 8.26% 98.69%

PC7 −0.119 0.708 0.695 0.016 0.004 0.021 −0.017 0.092 1.31% 100%

120

J. Olbrys and E. Majewska

5 Conclusion The goal of this research was to employ the PCA to extract common components of liquidity across a broad sample of equities, and from a set of liquidity measures on the Polish stock market. Seven liquidity proxies, namely percentage relative spread, percentage realized spread, percentage price impact, percentage order ratio, modified version of the Roll estimator, modified turnover, and modified Amihud measure, were utilized. The empirical PCA results reveal that first three principal components PC1–PC3 seem to capture common sources of liquidity variation on the WSE. The findings are robust to the sub-period choice. Therefore, one of possible directions for further investigation could be to assess commonality in liquidity on the WSE with these three principal components of liquidity proxies as latent factors in econometric models (e.g. Foran et al., 2015; Korajczyk & Sadka, 2008). To the best of the authors’ knowledge, no such research has been conducted for the Polish equity market so far. According to the recent literature, commonality in liquidity on the WSE is rather weak and robust to the choice of a liquidity measure and a subsample period (e.g. B˛edowska-Sójka, 2019; Olbry´s, 2019). Therefore, the main aim of the further study could be to confirm or deny this evidence. Acknowledgements The contribution of the first named author was supported by the grant ‘Comparative research on commonality in liquidity on the Central and Eastern European stock markets’ from the National Science Centre, Poland, No. 2016/21/B/HS4/02004.

A.1 Appendix Table 8 presents details concerning the Lee-Ready (LR) procedure. The midpoint price Ptmid at time t is calculated as the arithmetic mean of the best ask price Pt (a) t (b) . Considering that the bid and the best bid price Pt (b) at time t: Ptmid = Pt (a)+P 2 Table 8 The Lee and Ready (1991) procedure for inferring the initiator of a trade Conditions Stage I Trade is classified as buyer-initiated if Pt > Ptmid If Pt = Ptmid , then: Stage II Trade is classified as buyer-initiated if Ptmid > Pt−1 . When Ptmid = Pt−1 , the decision is taken based on the sign of the last non-zero price change Pt − k If Pt > Pt − k , then trade is classified as buyer-initiated If Pt < Pt − k , then it is classified as seller-initiated Source: Olbry´s and Mursztyn (2017)

Trade is classified as seller-initiated if Pt < Ptmid

Trade is classified as seller-initiated if Ptmid < Pt−1 .

Extracting Common Factors from Liquidity Measures with Principal. . .

121

and ask prices are not made public on the WSE, the midpoint price Ptmid at time t is approximated by the arithmetic mean of the low price PtL and the high price PtH at time t, which approximate the best ask price and the best bid price, respectively. The transaction price Pt at time t is approximated by the closing price. The opening trade is treated as being unclassified according to the LR procedure (Olbry´s & Mursztyn, 2017).

References Abdi, H., & Williams, L. J. (2010). Principal component analysis. Wiley Interdisciplinary Reviews: Computational Analysis, 2(4), 433–459. Amihud, Y. (2002). Illiquidity and stock returns: Cross-section and time-series effects. Journal of Financial Markets, 5(1), 31–56. B˛edowska-Sójka, B. (2019). Commonality in liquidity measures. The evidence from the Polish stock market. Hradec Economic Days, 9(1), 29–40. Brooks, C. (2019). Introductory econometrics for finance (4th ed.). Cambridge: Cambridge University Press. Chen, J. (2005). Pervasive liquidity risk and asset pricing. Job Market Paper, Columbia University. Chordia, T., Roll, R., & Subrahmanyam, A. (2000). Commonality in liquidity. Journal of Financial Economics, 56, 3–28. Corwin, S. A., & Schultz, P. (2012). A simply way to estimate bid-ask spreads from daily high and low prices. Journal of Finance, 67(2), 719–759. Fong, K. Y. L., Holden, C. W., & Trzcinka, C. (2017). What are the best liquidity proxies for global research? Review of Finance, 21, 1355–1401. Foran, J., Hutchinson, M. C., & O’Sullivan, N. (2015). Liquidity commonality and pricing in UK. Research in International Business and Finance, 34, 281–293. Goyenko, R. Y., Holden, C. W., & Trzcinka, C. A. (2009). Do liquidity measures measure liquidity? Journal of Financial Economics, 92, 153–181. Hasbrouck, J., & Seppi, D. J. (2001). Common factors in prices, order flows, and liquidity. Journal of Financial Economics, 59(3), 383–411. Huang, R. D., & Stoll, H. R. (1996). Dealer versus auction markets: A paired comparison of execution costs on NASDAQ and the NYSE. Journal of Financial Economics, 41, 313–357. Jackson, J. E. (1991). A user’s guide to principal components. New York: Wiley. Jolliffe, I. T. (2002). Principal component analysis. Springer series in statistics (2nd ed.). New York: Springer. Kaiser, H. F. (1958). The VARIMAX criterion for analytic rotation in factor analysis. Psychometrika, 23, 187–200. Karolyi, G. A., Lee, K.-H., & van Dijk, M. A. (2012). Understanding commonality in liquidity around the world. Journal of Financial Economics, 105(1), 82–112. Korajczyk, R., & Sadka, R. (2008). Pricing the commonality across alternative measures of liquidity. Journal of Financial Economics, 87(1), 45–72. Lee, C. M. C., & Ready, M. J. (1991). Inferring trade direction from intraday data. Journal of Finance, 46(2), 733–746. Nowak, S., & Olbry´s, J. (2016). Direct evidence of non-trading on the Warsaw Stock Exchange. Research Papers of Wroclaw University of Economics, 428, 184–194. Olbry´s, J. (2018). Testing stability of correlations between liquidity proxies derived from intraday data on the Warsaw Stock Exchange. In K. Jajuga, H. Locarek-Junge, & L. Orlowski (Eds.), Contemporary trends and challenges in finance. Proceedings from the 3rd Wroclaw

122

J. Olbrys and E. Majewska

International Conference in Finance. Springer Proceedings in Business and Economics (pp. 67–79). Cham: Springer. Olbry´s, J. (2019). Intra-market commonality in liquidity. New evidence from the Polish stock exchange. Equilibrium. Quarterly Journal of Economics and Economic Policy, 14(2), 251–275. Olbry´s, J., & Majewska, E. (2015). Bear market periods during the 2007–2009 financial crisis: Direct evidence from the Visegrad countries. Acta Oeconomica, 65(4), 547–565. Olbry´s, J., & Mursztyn, M. (2015). Comparison of selected trade classification algorithms on the Warsaw Stock Exchange. Advances in Computer Science Research, 12, 37–52. Olbry´s, J., & Mursztyn, M. (2017). Measurement of stock market liquidity supported by an algorithm inferring the initiator of a trade. Operations Research and Decisions, 27(4), 111– 127. Olbry´s, J., & Mursztyn, M. (2018). On some characteristics of liquidity proxy time series. Evidence from the Polish stock market. In N. Tsounis & A. Vlachvei (Eds.), Advances in time series data methods in applied economic research. Springer Proceedings in Business and Economics (pp. 177–189). Cham: Springer Nature Switzerland AG. Olbry´s, J., & Mursztyn, M. (2019). Alternative estimators for the effective spread derived from high-frequency data. In W. Tarczy´nski & K. Nermend (Eds.), Effective investment on capital market. Springer Proceedings in Business and Economics (pp. 177–188). Cham: Springer Nature Switzerland AG. Roll, R. (1984). A simple implicit measure of the effective bid-ask spread in an efficient market. Journal of Finance, 39(4), 1127–1140. Stoll, H. S. (2000). Friction. Journal of Finance, 55(4), 1479–1514. von Wyss, R. (2004). Measuring and predicting liquidity in the stock market. Dissertation Nr. 2899, University of St. Gallen.

The Mechanism of Political Budget Cycles in Greece George Petrakos, Konstantinos Rontos, Chara Vavoura, and Ioannis Vavouras

Abstract Extended empirical research has established the existence of political budget cycles in Greece but remains agnostic about the mechanism which generates them. In this paper we contribute to the literature by investigating precisely this mechanism of the creation of political budget cycles using data from the Greek economy for the last four decades (1980–2018). We find that it is via the manipulation of public expenditure rather than through the handling of public revenue that opportunistic politico-economic behaviour arises. We go on to build a novel empirical model linking government spending and revenue and estimate that, in years of general elections, public expenditure rises by around 2.2% of GDP. This level is not typical of a developed economy. Still, our finding is robust to various specifications of our model, both linear and non-linear, and hints towards a severe decline in the underlying political culture of the country. We conclude that, in the case of Greece, future fiscal rules aiming to suppress the political budget cycles phenomenon should target the control of pre-election transfer payments instead of resorting to tax increases.

G. Petrakos Department of Public Administration, Panteion University of Social and Political Sciences, Athens, Greece e-mail: [email protected] K. Rontos Department of Sociology, University of the Aegean, Mytilene, Greece e-mail: [email protected] C. Vavoura Department of Economics, University of Athens, Athens, Greece e-mail: [email protected] I. Vavouras () Panteion University of Social and Political Sciences, Athens, Greece e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_9

123

124

G. Petrakos et al.

Keywords Political budget cycles · Political fiscal cycles · Politico-economic models · Public revenue · Public expenditure JEL Classification: D72, E62, H62

1 Introduction This paper identifies the mechanism generating political budget cycles (PBCs), a term used to refer to the jump of budget deficits during election years. This phenomenon is generally interpreted as being triggered by the government’s pursuit of re-election, and it is played out when incumbents pursue opportunistic fiscal policies before general elections so as to appear competent and offer voters the illusion of economic prosperity (Rogoff, 1990; Rogoff & Sibert, 1988). Specifically, either because they are facing a myopic electorate with a decaying memory of past events or taking advantage of informational asymmetries that exist between them and rational constituents, politicians may choose to maximise their own voting function (Nordhaus, 1975) instead of behaving benevolently by maximising a social welfare function as is commonly assumed by traditional macroeconomic theory (Theil, 1956; Tinbergen, 1975). There are effectively two possible mechanisms that could generate PBCs, either through excessive public spending or via taxation policies aiming at a suboptimal level of tax revenue (Alesina, 1987, 1988), each with very different social welfare implications. Our paper contributes to the literature by tracing the steps taken in order for PBCs to be created. PBCs on the revenue side take place mainly in the form of a direct and indirect tax rate reduction, whereas PBCs on the expenditure side materialise mostly through an increase in transfer payments. Hence, the former tends to be of practical concern to relatively wealthier and, therefore, less numerous potential voters. As a result, we argue that revenue-side PBCs seem less likely, and for the additional reason that they require a considerably more significant time lag between the reduction in tax rates and the corresponding reduction in tax revenue and timing is of critical importance in increasing the effectiveness of pre-electoral fiscal manipulation. To test our hypothesis that it is expenditure manipulation that gives rise to PBCs, we propose a novel empirical model to examine how the political cycles affect public spending and public revenue. To confidently construct our model, we look into the connection between elections and the dynamic evolution of the two aforementioned macroeconomic variables. There is an important, although nuanced, relationship between current expenditure, past expenditure and current revenue. How the latter two shape the former has been a matter of an extended debate which is key in the PBCs literature. For example, de Haan and Klomp (2013) consider expenditure to be a function of its lagged values, whereas Shi and Svensson (2006), among others, take the approach that the level of expenditure of a given year is influenced not only by its own past values but also by the level of current revenue.

The Mechanism of Political Budget Cycles in Greece

125

Our focus is on Greece during the period between 1980 and 2018. It is well recognised that, over the past four decades, the Greek economy has been characterised by severe PBCs both at the national (De Haan & Klomp, 2013; Vavoura, Vavouras, Rontos, & Petrakos, 2019) and at the municipal level (Chortareas, Logothetis, & Papandreou, 2016).1 And although the existence of the cycles has been repeatedly investigated, the path of their occurrence has not yet been confidently described. This paper is, to our knowledge, the first to explore whether the generating mechanism of the PBCs in Greece is expenditure or revenue-driven and settle on a causal link between public expenditure and public revenue. Our work relates to the broad political business cycles literature2 and is particularly relevant to the emerging new empirical literature emphasising the size limiting effect of fiscal rules on PBCs (Bonfatti & Forni, 2019; Gootjes, de Haan, & JongA-Pin, 2019; Vavoura et al., 2019). The ability of governments to create PBCs is found to be decreasing in the level of economic and social development, the quality of institutions and the transparency of the political process (De Haan & Klomp, 2013). Most crucially, PBCs appear to be reduced in the presence of fiscal rules. We contribute to this literature by using our framework to derive interesting policy implications which could help redesign such rules more effectively. The structure of our paper is organised as follows. In Sect. 2, we present our methodology and discuss our findings. In Sect. 3 we conclude.

2 Methodology and Results In order to explore the mechanism that generates PBCs, we follow an empirical approach based upon annual administrative data from 1980 to 2018, following the standard PBCs literature (Chortareas et al., 2016; De Haan & Klomp, 2013; Sakurai & Menezes-Filho, 2011; Shi & Svensson, 2006; Veiga & Veiga, 2007). We use six variables and follow the definition of Eurostat for variables (1)–(5): 1. The general government total expenditure (GGE), as percent of GDP. 2. The general government total expenditure of the previous year (GGE-1) as percent of GDP. 3. The general government total revenue (GGR) as percent of GDP.

1 De

Haan and Klomp (2013) use data for the period 1975–2005, Vavoura et al. (2019) use data that cover the period 1980–2017, while Chortareas et al. (2016) use data that cover the period 1985–2004. 2 See, for example, Rogoff and Sibert (1988), Rogoff (1990), Alesina, Cohen, and Roubini (1997), Persson and Tabellini (2000), Brender and Drazen (2005, 2008), Shi and Svensson (2006) and Bonfiglioli and Gancia (2013). Notice that political budget cycles materialise through an increase in the budget deficits as opposed to political business cycles that occur when politicians in power exploit the short-run Phillips curve by increasing the rate of inflation, in order to reduce unemployment levels. For a review of the transition from the political business cycle to political budget cycle, see Efthyvoulou (2012).

126

G. Petrakos et al.

4. The general government total revenue of the previous year (GGR-1) as percent of GDP. 5. The growth rate of the real total GDP (TYGR) since a declining growth may put additional pressures on incumbent politicians in power to increase public deficit before elections. 6. Election (ELEC), a dichotomous variable taking the value of 1 in the years of general elections in Greece and 0 otherwise. To test whether PBCs are driven by an increase in expenditure or a decrease in revenue, we use three models. In Model (1) we look into the formation of public revenue. Let y1i denote the observed annual GGR, considered as the response variable on the ith time segment (i = 1,2, . . . ,37), covering the period between 1980 and 2018. Let z1i = y1i−1 denote the observed values of the 1-year lag of the GGR (GGR-1 or variable Z1 ), and let z2i denote the value of Z2 , corresponding to the total GDP growth rate (TYGR) variable. Also, let eleci denote the observed values of the dichotomous variable ELEC. In this context, revenue-side PBCs would yield a statistically significant coefficient of the election variable. We assume that the mean of the response variable can be modelled as a linear combination of both the quantitative and dichotomous variables in the following way: E (y1i ) = β0 +

2 

βj zj i + γ eleci

(1)

j =1

Model (1) presents a very good overall goodness of fit as the R2 = 95.1% adj = 94.66%), while the F-test indicates overall statistical significance (F = 219.78, df = 3, p-value = 0.00) and the DW statistic equals 2.29536. The coefficients of the model are statistically significant except for the elections variable, and they are reported in Table 1. We conclude that PBCs are not created through a drop in public revenue triggered by a decrease in tax rates before a general election. PBCs should, therefore, arise via a boost in public expenditure. As a result, public revenue is found to be best approximated by a model (Model 1.1) that includes only GGR-1 and TYGR as regressors. Model (1.1) presents a very good overall goodness of fit as the R2 = 95.1% (R2 adj = 94.82%). The economic intuition behind Model (1.1) is that public revenue is persistent, slowly adjusting overtime. Moving on to Model (2), we now test the hypothesis that PBCs are in fact generated on the expenditure side. Let y2i denote the observed annual GGE, considered as the response variable on the ith time segment (i = 1,2, . . . ,37), covering the period between 1980 and 2018. Let x1i = yi−1 denote the observed (R2

Table 1 Regression coefficients of Model (1)

Variables Constant GGR-1 TYGR ELEC

Coefficient 2.74 0.9472 −0.1613 −0.001

Standard error 1.47 0.0375 0.0741 0.55

p-value 0.071 0.000 0.037 0.999

VIF 1.01 1.01 1.02

The Mechanism of Political Budget Cycles in Greece

127

values of the 1-year lag of the GGE (GGE-1 or variable X1 ). In addition, let x2i denotes the value of X2 , corresponding to TYGR variable and x3i the value of X3 corresponding to the GGR. Finally, let eleci denote the observed values of the dichotomous measure ELEC. We start off by assuming that the mean of the response variable can be modelled as a linear combination of both the quantitative and dichotomous variables in the following way: E (y2i ) = β0 +

3 

βj xj i + γ eleci

(2)

j =1

Model (2) presents a very good overall goodness of fit as the R2 = 84.5% adj = 82.6%), while the F-test indicates overall statistical significance (F = 44.93, df = 4, p-value = 0.00) and the DW statistic equals 2.24878. The coefficients of the model are statistically significant and economically meaningful and they are reported in Table 2 (Fig. 1). (R2

Table 2 Regression coefficients estimation of Model (2)

Fig. 1 Diagnostics of Model (2)

Variables Constant GGE-1 TYGR GGR ELEC

Coefficient 10.3 0.580 −0.255 0.236 2.261

Standard error 3.01 0.121 0.127 0.124 0.920

p-value 0.002 0.000 0.053 0.065 0.019

VIF 3.91 1.07 3.80 1.02

128

G. Petrakos et al.

As Table 2 indicates, during years of general elections, public spending increases by 2.261% of GDP, thus implying that Greece is characterised by severe PBCs. This result goes in the direction of Vavoura et al. (2019) and de Haan and Klomp (2013), at the national level, and Chortareas et al. (2016) at the municipal level in documenting the existence of PBCs in Greece. Quantitatively, the PBCs that we document are disproportionately high in relation to other developed economies, which tend to have cycles that vary from well below 1% of GDP to insignificant (De Haan & Klomp, 2013). This result is a strong indication that, over the last four decades, Greece has had the socio-political characteristics of a developing economy rather than those of a developed one. In comparison to the works of Andrikopoulos, Loizides, and Prodromidis (2004), who analyse the period between 1970 and 1998 and find no evidence on the existence of PBCs, and de Haan and Klomp (2013) who report small PBCs during 1975–2005, we conclude that PBCs have become an increasingly pressing matter for the Greek economy. However, VIF values for GGE-1 (3.91) and GGR (3.80) indicate multicollinearity issues among predictors. Preliminary analysis of the relationship between the two variables results in a significant Pearson Correlation Coefficient of 0.857 and a Spearman Correlation Coefficient of 0.780. These results mean the inclusion of current revenue as well as past expenditure as explanatory variables, in line with the literature following Shi and Svensson (2006), causes bias. Analysing further the relationship between revenue and expenditure (Fig. 2), a strong and sharp linear positive relationship for the smaller values of GGR (35) the relationship between the two variables is neither strong nor sharp. The economic intuition behind Fig. 2 is that when public expenditure as percentage of GDP is relatively low, then it is mainly financed by public revenue and so the two variables are positively linked. For increased levels of spending though, public revenue as percentage of GDP, which is not associated with high fluctuations, becomes relatively less important in financing public expenditure, compared to borrowing, and, hence, the strong relation between revenue and expenditure disappears. In Table 2, we also report that the coefficient of the GDP growth rate (TYGR) is equal to −0.255% of GDP, meaning that during times of economic slowdown, corresponding to decreasing growth rates, governments tend to respond with expansionary fiscal policies. However, the p-value of (TYGR) is as high as 0.065, which seems counterintuitive. One way to understand this is that the effect of economic growth on current expenditure is partly absorbed by the inclusion of the lagged variable GGE-1. Given the multicollinearity problems of including both GGE-1 and GGR, we move on to the examination of Model (3). In Model (3), we drop GGR and consider (2) but having as predictors only the variables GGE-1, TYGR and ELEC, as follows: E (y2i ) = β0 +

2  j =1

βj xj i + γ eleci

(3)

The Mechanism of Political Budget Cycles in Greece

129

Fig. 2 The relationship between general government expenditure and revenue Table 3 Regression coefficients estimation of Model (3)

Variables Constant GGE-1 TYGR ELEC

Coefficient 10.24 0.778 −0.235 2.210

Standard error 3.12 0.066 0.132 0.955

p-value 0.002 0.000 0.083 0.027

VIF 1.06 1.06 1.01

Model (3) presents a very good overall goodness of fit as the R2 = 82.77% adj = 81.25%), while the F-test indicates overall statistical significance (F = 54.45, df = 4, p-value = 0.00), but the DW statistic equals 2.65171 indicating some autocorrelation issues. The coefficients of the model are reported in Table 3 (Fig. 3). From Model (3), which follows de Haan and Klomp (2013), we get that the effect of the PBCs is very robust to different specifications of the model. Besides, a linear model without GGR, including only lagged GGE (GGE-1) and the GDP growth rate (TYGR) as regressors, as well as the election dummy, does not suffer from multicollinearity but is susceptible to autocorrelation. Finally, bearing in mind that the diagnostics figures (residuals vs fitted) show evidence of nonlinearity, we formulate our final model. Model (4) is a novel empirical model in which we replace GGE-1 with the combination of GGR and GGR2 : (R2

130

G. Petrakos et al.

Fig. 3 Diagnostics of Model (3) Table 4 Regression coefficients estimation of Model (4)

E (y2i ) = β0 +

Variables Constant TYGR GGR GGR2 ELEC

3 

Coefficient −43.6 −0.614 4.123 −0.045 2.180

Standard error 11.5 0.139 0.640 0.009 0.944

2 βj xj i + β4 x3i + γ eleci

p-value 0.001 0.000 0.000 0.000 0.027

(4)

j =2

Model (4) presents a very good overall goodness of fit as the R2 = 86.31% adj = 84.7%), while the F-test indicates overall statistical significance (F = 53.60, df = 4, p = 0.00). The coefficients of the model are statistically significant and economically meaningful, and they are reported in Table 4 (Fig. 4). Model (4) shows that when exploring the generating mechanism of PBCs, linear models are inadequate. When the lagged value of public expenditure is removed from the model, all our variables become statistically significant at the 5% level. Moreover, the growth rate becomes a crucial determining factor of public expenditure, just as intuition dictates. Regarding the relationship between current expenditure and current revenue, we find that: (R2

The Mechanism of Political Budget Cycles in Greece

131

Fig. 4 Diagnostics of Model (4)

∂GGE ∼ = 4.123 − 0.09GGR ∂GGR

(5)

Equation (5) implies that the effect of current revenue on current expenditure changes sign (from positive to negative) at a level of GGR ∼ = 45.8. This value is similar to the one in Fig. 2 (bearing in mind that Fig. 2 shows the relation between GGR and GGE-1). Finally, it is striking that regardless of the approach (linear vs non-linear) and the variables we include as regressors, the magnitude of PBCs is robust, and it amounts to a 2.2% of GDP increase in public expenditure. From the analysis presented above, we can conclude that, over the last four decades, the Greek economy has been characterised by severe PBCs of about 2.2% of GDP. These cycles are of alarming magnitude, given that PBCs in developed countries lie between well below 1% of GDP and insignificant (De Haan & Klomp, 2013) and indicate a severe decline in the underlying political culture of the country. These PBCs have largely driven the excessive and persistent public deficits and the resulting debt crisis that devastated the Greek economy for the last 10 years and cost the country more than one-fourth of its GDP per capita (Vavouras, 2019). As a result of the crisis, the European Union implemented strict fiscal rules, mainly in the form of the excessive deficit procedure under the corrective arm of the Stability and Growth Pact. Such fiscal rules have been found to suppress politicians’ ability to create PBCs (Bonfatti & Forni, 2019; Gootjes et al., 2019; Vavoura et al., 2019). By highlighting the mechanism that generates PBCs, we add to the literature by deriving economic policy implications useful for the design of effective fiscal rules.

132

G. Petrakos et al.

In particular, our results show that, in the case of Greece, fiscal rules aiming to limit the impact of PBCs should target the control of pre-election transfer payments rather than resorting to tax increases.

3 Conclusions In this paper, we show that PBCs in Greece arise due to a sharp increase of around 2.2% of GDP in the level of public expenditure during election years, and the magnitude of these cycles is robust to different specialisations of the model describing the evolution of public expenditure. As a result, we find that, over the last 40 years, the Greek economy has been characterised by severe PBCs which are not characteristic of a developed but rather a developing economy. Our contribution is twofold. First, our work adds to the literature on the effect of fiscal rules on limiting PBCs because we can use our findings that PBCs are expenditure-oriented to justify the imposition of fiscal rules aiming at the reduction of transfer payments, rather than the ones focused on increasing tax rates in order to tackle opportunistic budgetary behaviour on the side of the Greek governments. Second, on a technical level, we move away from the linear model, currently the workhorse of the PBCs empirical literature, leveraging on the fact that public revenue and public spending appear to have a non-linear relationship. We end up with a quadratic model linking the two variables which works well for our limited dataset. However, the next task of research within this model would be to explore even more sophisticated relationships between our key macroeconomic variables with the use of larger and richer datasets.

References Alesina, A. (1987). Macroeconomic policy in a two party system as a repeated game. Quarterly Journal of Economics, 102(3), 651–678. Alesina, A. (1988). Credibility and policy convergence in a two-party system with rational voters. American Economic Review, 78(4), 796–805. Alesina, A., Cohen, G. D., & Roubini, N. (1997). Political cycles and the macroeconomy. Cambridge: MIT Press. Andrikopoulos, A., Loizides, I., & Prodromidis, K. (2004). Fiscal policy and political business cycles in the EU. European Journal of Political Economy, 20(1), 125–152. Bonfatti, A., & Forni, L. (2019). Fiscal rules to tame the political budget cycle: Evidence from Italian municipalities. European Journal of Political Economy, 60, 101800. Bonfiglioli, A., & Gancia, G. (2013). Uncertainty, electoral incentives and political myopia. The Economic Journal, 123(568), 373–400. Brender, A., & Drazen, A. (2005). Political budget cycles in new versus established democracies. Journal of Monetary Economics, 52(7), 1271–1295. Brender, A., & Drazen, A. (2008). How do budget deficits and economic growth affect reelection prospects? Evidence from a large panel of countries. American Economic Review, 98(5), 2203– 2220.

The Mechanism of Political Budget Cycles in Greece

133

Chortareas, G., Logothetis, V., & Papandreou, A. A. (2016). Political budget cycles and reelection prospects in Greece’s municipalities. European Journal of Political Economy, 43, 1–13. De Haan, J., & Klomp, J. (2013). Political budget cycles and election outcomes. Public Choice, 157(1/2), 245–267. Efthyvoulou, G. (2012). Political budget cycles in the European Union and the impact of political pressures. Public Choice, 153(3/4), 295–327. Gootjes, B., de Haan, J., & Jong-A-Pin, R. (2019). Do fiscal rules constrain political budget cycles? DNB Working Papers 634, Netherlands Central Bank, Research Department. Nordhaus, W. D. (1975). The political business cycle. Review of Economic Studies, 42(2), 169–190. Persson, T., & Tabellini, G. (2000). Political economics: Explaining economic policy. Cambridge: MIT Press. Rogoff, K. (1990). Equilibrium political budget cycles. American Economic Review, 80(1), 21–36. Rogoff, K., & Sibert, A. (1988). Elections and macro-economic policy cycles. Review of Economic Studies, 55(1), 1–16. Sakurai, S. N., & Menezes-Filho, N. (2011). Opportunistic and partisan election cycles in Brazil: New evidence at the municipal level. Public Choice, 148(1–2), 233–247. Shi, M., & Svensson, J. (2006). Political budget cycles: Do they differ across countries and why? Journal of Public Economics, 90(8–9), 1367–1389. Theil, H. (1956). On the theory of economic policy. American Economic Review, Papers and Proceedings, 46(2), 360–366. Tinbergen, J. (1975). On the theory of economic policy (6th ed.). Amsterdam: North-Holland. Vavoura, C., Vavouras, I., Rontos, K., & Petrakos, G. (2019). Electoral cycles and public deficits in Greece. Presentation at the Conference “Social Sciences Today: Dilemmas and Perspectives beyond the Crisis”, Mytilene, June 2019 (forthcoming in Greek). Vavouras, I. with the cooperation of Vavoura C. (2019). Economic policy (3rd ed.) Athens: Papazisis Publishers (in Greek). Veiga, L. G., & Veiga, F. J. (2007). Political business cycles at the municipal level. Public Choice, 131(1–2), 45–64.

Examination of Business Interest in Level of Complexity of Facial Biometric Technology Implementation in Slovakia Michal Budinský and Janka Táborecká-Petroviˇcová

Abstract Increase in the requirement of still more demanding customers forces businesses to improve their processes, especially in meeting their needs with the highest efficiency. In order to reach this aim, one of the possibilities is utilization of new technologies, what improves whole business performance at all. It is not a secret that biometric technologies are spreading quickly within distinctive sectors and their popularity increases. Related to this, utilization range of these technologies in private sectors and the required level of their complexity is an interesting issue for examination. The main aim of this paper is to reveal the level of complexity of facial biometric technology implementation, which businesses operating in Slovakia would be interested in. Notably, we devoted our attention to the identification of certain level of complexity among business, we focused on identification of possible relationship between complexity level and added value/usefulness or AIDA/STDC model. Within our research was proved assumption that at least half of businesses in Slovakia are interested in some form of facial biometric technology implementation. In addition, our research confirmed dependence between the required level of implementation complexity and distinctive factors. Based on the results were formulated several managerial implications. This paper contains partial results of complex research focused on the examination of market potential of facial biometric technology implementation in Slovakia. Keywords Biometric technologies · Facial biometrics · Market potential

1 Introduction Competition on a market is increasing on its intensity recently, while almost all business processes are slightly changing and become even more difficult to be managed. If businesses want to plan their activities and strategies properly, they M. Budinský · J. Táborecká-Petroviˇcová () Faculty of Economics, Matej Bel University, Banská Bystrica, Slovakia e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_10

135

136

M. Budinský and J. Táborecká-Petroviˇcová

need to obtain more and more precise data about their customers. For this purpose, serve them CRM systems, loyalty programs, or marketing information systems. These traditional tools are useful, but slowly they are not enough to secure advanced requirements of customers. While the trend of gathering, processing, and analyzing data directs to the utilization of the newest technologies. Notably, biometric technologies are becoming more often utilized for the purposes of customer data gathering, even though they were initially developed for public sector and security enhancement. Therefore, it is not a surprise that day-by-day are released more references to biometrics or biometric technologies utilization, their importance, or questionability of its usage. In terms of this, we consider interesting to investigate what certain level of implementation complexity of facial biometric technology, would be businesses operating in Slovakia interested in.

2 Theoretical Background on Facial Biometric Technology First mention related to the origin of biometrics is connected with the Chinese culture of fourteenth century, where vendors stamped footprints and palm of children on a paper, in order to distinguish them. Lately, in 1880s Bertillon and Clerk draw up a method for identification of criminals, based on a multiple body measurement, which was lately took over by police and used in crime investigation. Thus, we may see that biometrics itself is not a novelty and its evolution has undergone several steps of its development. In addition, in the last decade, it significantly influenced the fundamental perception of security or property protection (Jones, Williams, Hillier, & Comfort, 2007). Different forms of biometric systems are used around the world in many distinctive areas. We can see their application in public sphere, banking and insurance sphere, and also in various commercial stores abroad. Biometrics is becoming a natural part of our life, which is proven by increased amount of biometric systems we are getting in a touch recently. Usage of biometric technologies opens huge opportunity for businesses in obtaining and processing various data about customers (Jones et al., 2016). Interesting perspective was brought by Heracleous and Wirtz (2006) where they understand biometrics as a tool for authentication or identification of persons based on physiological features. In addition, they identified that main advantage of biometrics lies in higher practicability and security against some identifier you know, such as entry passwords. Moreover, it is more practical and secure than something you have, such as ID cards, token, or some card keys. Thus, there is no possibility of lost, forgetting, or copying biometrics features, especially in a case of multi-biometric technologies. In general, there arise a number of possibilities for biometrics usage in public sector. Governments around the whole world utilize distinctive biometric systems for better identification of persons. It is not a surprise that leader in this field is the USA, where government itself develops systems for fingerprint identification, electronic passports, or many others. On the other hand, for example, Great Britain

Examination of Business Interest in Level of Complexity of Facial Biometric. . .

137

implemented ID cards with biometric features or within European Union integration of biometric passports is becoming a standard in the field of person identification. However, the space for implementation of these systems is much broader (Acharya & Kasprzycki, 2010). Taking a look at examples of successful implementation of biometric technology in commercial sphere, we can mention for example Japanese stores, where faces of customers are recorded and encoded based on biometric system and later shared among 115 stores in order to prevent thefts and robberies. Software-developing company based in Nagoya introduced this system in 2013 as prevention against thefts. According to the law, businesses involved in a system can share this data, but they are not allowed to provide them to the third parties (Davis West, 2018). Biometrics at all represents sophisticated technology, which is able to identify unique physical and behavioral feature with the main purpose of its backward identification (Shyam & Singh, 2016). Moreover, different commercial systems and algorithms were developed in the past because interest for face recognition systems increased in the context of immigration in many countries, especially on places such as airports and ports, where facial recognition systems were implemented after the US terrorist attacks on September 11 (Shin, Kim, Lee, Shin, & Choi, 2008). Moreover, according to Rouse (2018), facial biometric technologies belong to that category of biometric software which is able to map characteristics of a person’s features mathematically. These technologies are based on advanced algorithms which compare digital images or live captures with database face-prints for the purpose of verification. Related to the studied issue, we consider necessary to briefly introduce AIDA model of purchase behavior, as we will investigate its potential relationship with the required level of complexity. This model describes degree of customer readiness within purchase behavior. Firstly, we are talking about attention, which is also called as “awareness” and refers to if consumers find a product as an attractive. Consequently, when product has attention, it is very important to keep it, what is sometimes more difficult. In this step, we are talking about creation of interest. Next, after attention and interest were developed, desire for product is required to be established. Finally, the last step is about calling a consumer for action, thus to purchase a product/technology. It means that, based on the stage of AIDA model among consumers, their perception of new technology market potential is different (Suggett, 2017). In addition, recently a new model of consumer buying process identification appeared that is very similar to the AIDA, but goes a little bit further. Specifically, we are talking about “See-Think-Do-Care” (STDC) model, which is a partially analogy of AIDA model. Within STDC model, a very fist stage “See” refers to a group of potential customers that currently gained basic information about product/service and create wide audience. Next stage, “Think” is represented by those customers that have an interest in product or consider their purchase. Following stage of “Do” is connected with customers definitely decided about purchase of the product/service or those who already made purchase. The last step of this model, “Care,” refers to the customers who made at least two purchases of

138

M. Budinský and J. Táborecká-Petroviˇcová

the product, and the aim of business is to care about them (Hart, 2018). In general, we can see an analogy between these two models, especially in the first two phases. Further in our research we used these concepts to formulate respective question in the survey.

3 Methodology of Research The main aim of this paper is to reveal the level of complexity of facial biometric technology implementation, which businesses operating in Slovakia would be interested in. Notably, we devoted our attention to the identification of certain level of complexity among business, later we focused on identification of possible relationship between complexity level and added value/usefulness or AIDA/STDC model. In order to investigate these issues, we performed primary research. In relation to fulfillment of main aim, we have formulated hypothesis H1 . H1 : We assume that at least half of businesses will be interested in some form of facial biometric technology implementation in their business.

For the purpose of this study, we realized quantitative research, where questioning was applied as a data gathering method and as a tool we used a questionnaire survey.1 In this research, we approached businesses of relevant size and from different industry, but operating in Slovakia. For the purpose of data collection process, we used two different forms. Firstly, students from Faculty of Economics at Matej Bel University participated as field researchers in this survey who contacted businesses across Slovakia to fill up questionnaire with adequate representatives. Secondly, after the process of data collection from students terminated, we approached businesses in Slovakia by email. For this purpose, an online version of our questionnaire was created in Google Docs application. In order to approach as many businesses as was possible, we decided to utilize access to database of contacts through analytical portal. This process lasts 3 months from September to November of 2019. Within our questionnaire, we utilized questions with semantic differential statements, basic optional questions, or level of agreement statements, enhanced by seven-point Likert scale to ensure sufficient range of options. For verification of hypotheses and statistical testing of various relationships within the study, we utilized correlation analysis, especially Spearman correlation coefficient. In the context of our research, we focused on businesses with at least ten employees, therefore small, medium, and large businesses, because potential of this technology in micro-businesses is very small, and we do not expect their interest in facial biometric technologies implementation. Finally, from data collection process, we were able to reach 521 complete answers that follow representativeness 1 Results

presented in this paper are part of the research solved within the national grant VEGA 1/0488/20 Market, marketing, legislative and ethical aspects of biometric technologies utilisation in commercial sector (2020–2022, project leader: Janka Taborecka-Petrovicova).

Examination of Business Interest in Level of Complexity of Facial Biometric. . .

139

3,26%

17,47%

79,27%

10 - 49 employees

50 - 249 employees

more than 250 employees

Fig. 1 Distribution of businesses according to the number of employees

criteria (research sample representative by two signs—size of business/number of employees and region/business residence). Within our research, we collected 521 complete answers, while the structure of these respondents can be divided into distinctive groups according to the identification criteria. We decided to distribute these respondents according to the criteria: business activity, ownership, average annual revenues, business size, and region. Our research samples were 189 of manufacturing businesses (36.28%), 62 of wholesalers (11.90%), 47 of retailers (9.02%), and 223 businesses from service (42.80%). Further from total amount, 371 of businesses (71.21%) are owned solely by Slovak owners, 19 of businesses (3.65%) are owned mostly by Slovak owners, 11 of respondents (2.11%) are half owned by Slovak and half owned by foreigners, 32 businesses (6.14%) are mostly owned by foreign owners, and finally 88 of business (16.89%) are solely owned by foreign owners. Later, our research sample consists of 236 businesses (45.30%) with average annual revenues less than 2 million A C, 180 businesses (34.55%) with average annual revenues between 2 and 10 million A C, 75 of businesses (14.40%) with average annual revenues between 11 and 50 million A C, and 30 of businesses (5.76%) with average annual revenues higher than 50 million A C. Research sample consists of 79.27% (413) of small business with 10– 49 employees, 17.49% (91) of medium-size businesses with 50–249 employees, and finally 3.26% (17) of large businesses with more than 250 employees. Distribution is reflected in Fig. 1.

4 Results and Discussion Firstly, we focused on revealing interest of respondents in the level of complexity of facial biometric technology implementation into the business processes. Through our research, we found out that 51.63% (269) of respondents do not have interest

140 Fig. 2 Distribution of businesses by interest in level of complexity implementation

M. Budinský and J. Táborecká-Petroviˇcová

6,91%

3,26%

38,20%

no interest advanced implementation

51,63%

basic implementation complex solution

in any form of facial biometric technology implementation. On the other hand, 38.20% (199) of respondents are interested in single/basic form of implementation of facial biometric technology. In addition, in our research sample, 6.91% (36) of respondents were interested in advanced functionalities of this technology and its implementation to the business. Finally, answers of the respondents revealed that 3.26% (17) of businesses have an interest in complex solution of facial biometric technology and its integration to all business processes. Distribution of respondents according to their interest in the level of complexity of facial biometric technology is depicted in Fig. 2. Even though the amount of not interested respondents is relatively high and represent half of the approached businesses, it is still positive finding that almost half of the addressed respondents would have interest in at least little form of facial biometric technology implementation. This fact revealed us interesting information about the potential of this technology on a market as well as this issue is still very young in Slovakia. Related to the required level of complexity of facial biometric technology implementation, we need to verify our hypothesis H1 , where we assume that at least half of businesses will be interested in some form of face biometric technology implementation in their business. For this purpose, we utilized Binominal test via SPSS program and statistically tested this assumption. The results of binominal test showed us that as p-value (0.483) is not lower than alpha (0.05) we do not reject 0 hypothesis, thus it means that there is exactly 50% of businesses that are interested in some form of implementation of this technology. As 50% fall under the scope of our prediction (at least 50%), we can conclude that the hypothesis H1 was confirmed. The result of binomial test is reflected in Table 1. It is important to mention that this question as well as answers on this question serve us as filter of our respondents. In this point, we can divide our respondents into two groups—those who do not have interest in any form of facial biometric technology implementation and those who have interest in any form of facial biometric technology implementation. For further processing of our paper, 48.47% (252) of respondents represent the basis on which we will focus, when identifying

a Exact

Interested in at least single form of implementation No interest Total

N 252

269 521

Category 2

1

results are provided instead of Monte Carlo for this test

Level of complexity

Binomial test

Table 1 Confirmation of Hypothesis H1

0.52 1.00

Observed prop. 0.48

Test prop. 0.50

Exact sig. (two-tailed) 0.483

Exact sig. (two-tailed) .483a

Examination of Business Interest in Level of Complexity of Facial Biometric. . . 141

142

M. Budinský and J. Táborecká-Petroviˇcová

Table 2 Correlations between AIDA/STDC model and level of complexity Correlations Spearman’s rho

Stage of AIDA model

Level of complexity

a Correlation

Correlation coefficient Sig. (two-tailed) N Correlation coefficient Sig. (two-tailed) N

Stage of AIDA model 1000 . 521 .282a .000 521

is significant at the 0.01 level (two-tailed)

their expectations, purchase intention, relationships between variables, and revealing potential of facial biometric technology on a market. Finally, we took a look also at relationship between AIDA/STDC model and required level of complexity of facial biometric technology implementation. For this purpose, we used combination of previous mentioned AIDA and STDC models according to Suggett (2017) and Hart (2018). In our research, we found out that 35 (6.72%) of respondents are before stages of AIDA model, thus they have never heard about facial biometric technologies before. On contrary, majority of respondents, together 450 (86.37%), expressed that until now, they have just heard about facial biometric technologies, thus they belong to the first stage of this model. Next, only 15 (2.88%) of respondents are in second stage of AIDA/STDC model, hence they have heard about this technology and consider its purchase. On the other hand, only four respondents (0.77%) are in third stage, which refers to decision-making process about which facial biometric technology to purchase. Finally, we were surprised that in our research sample, 17 businesses (3.26%) already use facial biometric technology in their business, therefore they are fourth stage of this model. In this case, results of statistical test showed us that there is statistically significant relationship between these variables with direct medium-strong dependence (Spearman’s rho = 0.282), as p-value (0.00) is lower than alpha (0.01). This finding means that, with increased stage of AIDA model among businesses, their interest for more complex solution of this technology rises. We can interpret it the way that during the “journey” through particular stages of AIDA/STDC businesses are acquiring more detailed and specific information; they become more educated and hence they can see a full potential of this technology. Results of correlation analysis are shown in Table 2. When discussing opinions of businesses on facial biometric technology, we were interested in overall added value perceived by respondents. We found out that only 3.45% (18) of respondents perceive high added value for their business from implementation of facial biometric technology, followed by 6.14% (32) who perceive relatively high added value. Only 22.26% (116) of businesses perceived added value above average, while 38.77% (286) of businesses stated that their expected added value is below average. This fact could be caused due to not widespread information about this technology, or due to different associations with

Examination of Business Interest in Level of Complexity of Facial Biometric. . .

143

Table 3 Correlations between added value and level of complexity Correlations Spearman’s rho Added value

Correlation coefficient Sig. (two-tailed) N Level of complexity Correlation coefficient Sig. (two-tailed) N

a Correlation

Added value 1000 . 521 −.572a .000 521

Level of complexity −.572a .000 521 1000 . 521

is significant at the 0.01 level (two-tailed)

usage of this technology (probably price, difficulty of its implementation, trade-off between costs and benefits, etc.). In the next step, we were interested in perceived added value by businesses in relation to complexity. Based on statistical test of Spearman correlation coefficient, we found out that there is statistically significant relationship with indirect mediumstrong dependence (Spearman’s rho = −0.572) between perceived added value and level of complexity of technology implementation because p-value (0.00) is lower than alpha (0.01). Results are presented in Table 3. Therefore, according to the character of questions and statements, businesses which perceive higher added value from implementation of facial biometric technology are interested in more complex form of technology implementation. This correlation seems to be logic because required level of complexity should rise with increased added value perception as a result of positive attitude. While considering added value, we were interested also in overall perception of businesses about usefulness of this technology. In order to reveal this issue, we used simple question based on seven-point Likert scale in order to provide enough space for respondents to express their opinion. Based on information from graph above, we can conclude that perceived usefulness of facial biometric technology among business in Slovakia is relatively low. Specifically, only 5.57% (29) of respondents see this technology as completely usable for their business, and 5.76% (30) of respondents see technology as relatively usable. On the other hand, 23.99% (125) of respondents perceive facial biometric technology as relatively useless, while 14.78% (77) of all respondents perceive this technology as completely useless for their business. Finally, 18.62% of business were indecisive in this question. In addition, in examination of this question, we have revealed relationship between perceived usefulness of facial biometric technology and required level of complexity by businesses. According to the Spearman correlation coefficient, statistically significant relationship between these variables with indirect mediumstrong dependence (Spearman’s rho = −0.589) was confirmed, as well as p-value (0.00) is lower than alpha (0.01). Table 4 represents the results of correlation analysis. This means that, according to the character of questions and statements, with increased level of usefulness perceived by businesses, increases also their interest for more complex implementation of this technology.

144

M. Budinský and J. Táborecká-Petroviˇcová

Table 4 Correlations between level of complexity and usefulness Correlations Spearman’s rho Level of complexity Correlation coefficient Sig. (two-tailed) N Usefulness Correlation coefficient Sig. (two-tailed) N a Correlation

Level of complexity 1000 . 521 −.589a .000 521

Usefulness −.589a .000 521 1000 . 521

is significant at the 0.01 level (two-tailed)

Since we are interested in examination of market potential of facial biometric technology, we considered as necessary to address also the purchase intention of businesses operating in Slovakia. Measuring purchase intention of respondents in case of such high-end facial biometric technology is not very easy through questionnaire, due to possible lack of information background necessary for decision-making of respondents. In order to reveal purchase intention of business in Slovakia to facial biometric technology, we applied procedure of Heijden and Verhagen’s (2004; In: Kwon, 2012). In this case respondents were asked to express their opinion on seven-point Likert scale to three statements (no purchase intention, purchase intention, and indecisive). Consequently, their answers have been averaged and classified into three categories. Values from interval represented those who did not agree with statements (purchase intention was not proved). We revealed that among 58.59% (150) of respondents were not identified purchase intention related to facial biometric technology. However, 31.64% (81) of respondents showed purchased intention for facial biometric technology. The rest of respondents 9.77% (25) did not show any marginal results because their mean values were equal to 4. Finally, we took a look at relationship between required level of complexity of facial biometric technology implementation and business’s purchase intention. Apparently, we expected that with increased level of complexity of implementation by businesses, purchase intention would be higher. According to the results of Spearman correlation test, statistically significant relationship between these variables with indirect medium-strong dependence (Spearman’s rho = −0.443) was approved because p-value (0.00) is lower than alpha (0.01). Results are shown in Table 5. Therefore, we can say that businesses which are interested in more complex form of implementation of facial biometric technology showed higher purchase intention for this technology. Related to the required level of complexity by interested businesses, we found out that majority 38% of them demand single/basic form of implementation of this technology. Thus, they do not intend to expand its usage in all business processes, but only to individual activities. At first, based on these findings, we recommend businesses commercializing facial biometric technology to be close to the customer

Examination of Business Interest in Level of Complexity of Facial Biometric. . .

145

Table 5 Correlations between purchase intention and level of complexity Correlations Purchase intention Spearman’s rho Purchase intention Correlation coefficient 1000 Sig. (two-tailed) . N 252 Level of complexity Correlation coefficient −.443a Sig. (two-tailed) .000 N 252 a Correlation

Level of complexity −.443a .000 252 1000 . 521

is significant at the 0.01 level (two-tailed)

and to offer them what they look for. Therefore, they should create a simple option/package for businesses within implementation of this technology, where they could choose for example 1 or 2 functionalities that will be activated according to their preferences. On the other hand, we suggest to create the packages that would contain advanced up to the complex implementation option for businesses and to present these packages on personal meeting in order to attract customers. From sales point of view, we know that sometimes customer do not know exactly what he wants, until you show it to him. Thus, in personal meetings, advanced forms of implementation of this technology can also be presented as a further possibility or extension opportunity. This may attract them and broaden their perspective about usage of this technology and create real interest for more complex solutions. We should not forget that there are also specific/individual customers that are more demanding and could have specific requirements for complex solutions. Hence, individual approach to these customers is inevitable, in order to satisfy their needs and expectations.

5 Conclusion Results of this paper proved that among half of businesses operating in Slovakia prevail interest for implementation of facial biometric technology at least in single form. Further, study confirmed dependence between required level of implementation complexity and distinctive factors, such as higher complexity required when among businesses prevail higher added value from usage of this technology or see higher usefulness. In addition, positive relationship between reference to a particular stage of purchase behavior according to AIDA model and required level of complexity was approved. Next practical contribution of our paper represents identification of interest for certain level of complexity of facial biometric technology implementation among businesses in Slovakia. These findings are valuable for businesses developing and commercializing this technology on a market because they are concerned about

146

M. Budinský and J. Táborecká-Petroviˇcová

what specific form of implementation would be their potential customers interested in. This data provides a picture about level of customer demand and enable businesses to adapt technology as well as its functionalities in compliance with their requirements.

References Acharya, L., & Kasprzycki, T. (2010). Biometrics and government. Library Parliament, Canada. Retrieved April 21, 2018, from http://www.lop.parl.gc.ca/content/lop/researchpublications/0630-e.pdf Davis West, J. (2018). 21 amazing uses for face recognition—Facial recognition use cases. Retrieved December 10, 2018, from https://www.facefirst.com/blog/amazing-uses-for-facerecognition-facial-recognition-use-cases/ Hart, C. (2018). See, think, do, care: A new way to communicate your SEO strategy. Search Engine Journal. Retrieved December 10, 2018, from https://www.searchenginejournal.com/seo-guide/ see-think-do-care-seo-strategy/ Heracleous, L., & Wirtz, J. (2006). Biometrics: The next frontier in service excellence, productivity and security in the service sector. Managing Service Quality, 16(1), 12–22. Jones, P., Williams, P., Hillier, D., & Comfort, D. (2007). Biometrics in retailing. International Journal of Retail & Distribution Management, 35(3), 217–222. Jones, C., et al. (2016). The future of retail engagement lies in biometrics. Retrieved December 10, 2018, from http://www.icon-uk.net/biometrics-in-retail Kwon, E. S. (2012). Exploring consumers’ attitudes and behavior toward product placement in television shows, 2012. Media Studies—Theses. Paper 4. Rouse, M. (2018). Facial recognition. Retrieved December 10, 2018, from https:// searchenterpriseai.techtarget.com/definition/facial-recognition Shin, Y., Kim, J., Lee, Y., Shin, W., & Choi, J. (2008). Formal implementation of a performance evaluation model for the face recognition system. Journal of Biomedicine and Biotechnology, p. 1–10. Shyam, R., & Singh, Y. N. (2016). Multialgorithmic frameworks for human face recognition. Journal of Electrical and Computer Engineering, 2016(9), 9. Suggett, P. (2017). Get to know and use, AIDA. Retrieved January 27, 2018, from http:// www.thebalance.com/get-to-know-and-use-aida-39273

Innovation and Sales Growth Among Heterogeneous Albanian Firms: A Quantile Approach Blendi Gerdoçi and Sidita Dibra

Abstract This study contributes to the stream of research that critically questions the relationship between innovation and firm’s growth performance. Using the 2019 World Bank Group enterprise survey data, Ordinary least square (OLS) and Quantile regression (QR) have been employed to examine the effect of various measures of innovation on the sales growth of Albanian firms. The two-regression analysis offer inconsistent results. OLS study results show that the adoption by firms of new processes is the only innovation measure that positively affects sales growth. Controversially, the more nuanced QR results show that the impact of innovation on sales growth is significant only for those firms located at the 90th percentiles. Product innovation and internal R&D appear to be the drivers of high-growth firms’ performance. Surprisingly, process innovation and external R&D have a negative impact on the growth performance of such firms. For the rest of the quantiles, the results show that innovation does not affect sales growth. Our study results show that innovation explanatory power is weak and noteworthy only for high-growth firms. Keywords Innovation · Firm growth · Quantile regression · Albania · Transition countries

1 Introduction In his seminal work, The Theory of Economic Development (1934), Joseph Schumpeter argued that innovation is an essential driver of economic growth. Schumpeter’s theory, built around the innovative entrepreneur, views markets as a dynamic arena where heterogeneous firms compete with each other by introducing new products, processes, and ways of doing business. Theoretical models rooted in the Schumpeterian thought which take a macroeconomic perspective (e.g.,

B. Gerdoçi · S. Dibra () Faculty of Economy, University of Tirana, Tirana, Albania e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_11

147

148

B. Gerdoçi and S. Dibra

Aghion, Bloom, Blundell, Griffith, & Howitt, 2005; Nelson & Winter, 1982) have contributed to the view of innovation as an engine of growth. At a micro-level, innovation is often associated with increases in productivity and competitiveness and, consequently, the firm’s growth. Empirical research focused on developed economies has found that sales growth is positively linked to investment in R&D and patents (Demirel & Mazzucato, 2013), product innovation (McKelvie, Brattström, & Wennberg, 2017; Na & Kang, 2019; Roper, 1997), process innovation (Cohen & Klepper, 1996), or a combination of different innovation measures (Bianchini, Pellegrino, & Tamagni, 2018). In developing economies, innovation consists of imitating products and processes designed elsewhere, mainly Vivarelli (2014). Furthermore, essential firm characteristics such as size do not appear to be a vital determinant of sales growth rate. Some research suggests a small, negative relationship (Becchetti & Trovato, 2002; Bottazzi, Coad, Jacoby, & Secchi, 2005). On the contrary, firm age has been found to affect sales growth with younger firms featuring higher growth rates (Coad, Segarra, & Teruel, 2013), or moderating the relationship between innovation and growth (Coad, Segarra, & Teruel, 2016). The innovation, strategy, and marketing research have examined the mechanisms of the innovation-firm growth nexus investigating a variety of innovation types, including inputs (internal and external R&D), outputs (product and process innovation), and intermediate outputs (patents). Most scholars agree that R&D spending leads to the introduction of new products (Cooper & Kleinschmidt, 1987; Krishnan & Ulrich, 2001; Parisi, Schiantarelli, & Sembenelli, 2006) that in turn enables firms to increase customer satisfaction (Johannessen, Olsen, & Lumpkin, 2001) and gain market share early (Banbury & Mitchell, 1995). As a result, firms can improve financial performance and grow faster. In contrast, process innovation is more production-oriented compared to the sales or customer-oriented nature of product innovation (Hervas-Oliver, Sempere-Ripoll, & Boronat-Moll, 2014). It is mainly related to new processing machines or IT equipment (Edquist, 2001), and it is predominantly focused on the reduction of production cost (Papinniemi, 1999) and labor costs (Coad & Rao, 2011). Furthermore, process and product innovation can be interrelated. For example, process innovation can increase product quality (Damanpour, 1991). Finally, patented innovations may lead to increased profit margins or sales (Brouwer & Kleinknecht, 1999) although mainly in conjunction with other innovation mechanisms (Cohen, Nelson, & Walsh, 2000). However, in general, empirical research exploring the relationship between innovation and growth has had mixed results (Audretsch, Coad, & Segarra, 2014; Coad, 2009). Some studies have found a positive relationship (e.g., Deeds, 2001; Roper, 1997; Yasuda, 2005), others found no significant results (e.g., Bottazzi, Dosi, Lippi, Pammolli, & Riccaboni, 2001; Geroski & Mazzucato, 2002; Lööf & Heshmati, 2006), even for persistent innovators (Guarascioa & Tamagni, 2019), while a few yielded negative results (e.g., Freel & Robson, 2004). Firm sales growth rates appear to be unpredictable and random (Guarascioa & Tamagni, 2019), and the explanatory power of innovation determinants is extremely weak (Geroski, 2000).

Innovation and Sales Growth Among Heterogeneous Albanian Firms: A. . .

149

Some scholars have been motivated by these inconsistent findings to use novel statistical techniques such as quantile regression (QR) in an attempt to account for the heterogeneous impact of innovation on firms’ sales growth (Coad & Rao, 2008; Santi & Santoleri, 2017; Segarra & Teruel, 2014). Whether innovation is measured as a composite index (Bianchini et al., 2018; Coad & Rao, 2008) or separate indicators are used (Segarra & Teruel, 2014), the results suggest that innovation is of significant importance for high-growth firms only. This paper aims to test the innovation-firm growth relation in the context of a developing economy, using a heterogeneous sample of firms operating in different sectors based on the 2019 World Bank Group enterprise survey data. The study focuses on how different firm-specific characteristics such as size, age, patenting, internal and external R&D, product, and process innovation affect sales growth. The paper is structured as follows: Section 2 presents data and the model used. Section 3 contains linear and quantile regression analysis. Section 4 includes discussions, implications at the policy level, and limitations.

2 Method and Data 2.1 Data The survey data were obtained from the World Bank Group (WBG) enterprise survey dataset, collected between January and May 2019. The final sample for this research comprises 272 firms randomly selected using three levels of stratification: industry, firm size, and region (World Bank Albania, 2019). All cases whose responses were classified as accurate or somewhat accurate, and cases with missing data, more than 20% have been removed from the sample.

2.2 Main Variables and Measurement Sales growth (dependent). Following Bianchini et al. (2018), sales growth was measured as the log-difference between the natural logarithm of sales for the 2018 fiscal year with the logarithm of sales for the previous year. Independent variables. The independent variables in this study are (1) process innovation, (2) product innovation, (3) external research and development (R&D), (4) internal research in R&D, and ownership of patents. All these variables are binary. Managers were asked whether any of the above has been introduced or used by the firm during the last 3 years. Control variables. Size (natural logarithm of the number of employees) and age (natural logarithm and years since foundation) are the two controls in this study.

150

B. Gerdoçi and S. Dibra

2.3 Model and Estimation This study analysis starts by investigating the relationship between innovation and sales growth using ordinary least square (OLS) regression. This analysis represents the baseline. However, since OLS estimates are calculated based on the average effect of the independent variables on the outcome, they provide an incomplete picture. Given that sales growth distributions are fat-tailed and skewed, the results will be biased also (Coad & Rao, 2008). Instead, the QR approach provides more robust results when the error term is not normally distributed (Buchinsky, 1998). Moreover, it allows us to identify potential asymmetric effects and variations in the coefficient estimates of the high-growth firm (the top quantiles) vs. low-growth firms (the bottom quantiles).

2.3.1

Correlations and Collinearity Statistics

Table 1 shows the correlations among the variables. The coefficients between some of the independent variables are significant, as expected. However, the majority are weak or moderate. No multicollinearity-related issues are found as the variance inflation factor (VIF) values are well below the threshold of 5 (values are around 1).

3 Results The results for OLS and QR estimation are reported in Tables 2 and 3. The OLS results show that only process innovation has a significant effect on sales growth, while the other variables do not. In contrast to the baseline analysis, QR results, at all quantile levels, except for the 90% quantile, are not significant (one exception for size at the 50% quantile, although the results are not robust). At the 90% quantile, the coefficients of innovation are first, much larger, and second, significant for the most part. Product innovation and internal R&D have a positive effect on sales growth while process innovation and external R&D, surprisingly, have a negative impact. The evidence here suggests, therefore, that investments in innovation activities have a significant contribution to the firm’s sales growth performance only for high-growth firms. The analysis of the pseudo R2 results depicts an interesting picture. Innovation measures account for around 6% only for slow-growth and high-growth firms. For the other quantiles, pseudo R2 figures are around 1–2% suggesting a weaker contribution of innovation in explaining growth performance. The pseudo R2 appears to be U-shaped when moving from lower to higher quantiles. However, despite these nuances, data suggest that very little of the sales growth variance is explained by innovation.

Log diff. gr. 1 .052 .080 −.018 .131* −.024 −.015 .043

**p < 0. 01, *p < 0.05

Variables Log diff. growth Ln (size) Ln (age) Product inn. Process inn. Internal R&D External R&D Patents

Table 1 Bivariate correlations Ln (age)

1 −.077 .070 −.027 −.043 −.006

Ln (size)

1 .219** .065 .277** .020 .096 .118* 1 .079 .157** .119* .099

Product innovation

1 .019 .051 .229**

Process innovation

1 .368** .216**

Internal R&D

1 .248**

External R&D

1

Patents

Innovation and Sales Growth Among Heterogeneous Albanian Firms: A. . . 151

152

B. Gerdoçi and S. Dibra

Table 2 OLS and QR estimation: the coefficient and standard error on innovation measures reported for the full sample and 10, 25% quantiles Variables Ln (size) Ln (age) Product innovation Process innovation R&D internal R&D external Patents Constant R2/pseudo R2 N. of observation

Full sample Coef. .111 −.009 −.019 .262** −.099 −.025 .089 −.413 0.024 272

S. E. .101 .041 .111 .129 .162 .301 .202 .289

Q10 Coef. .072 .003 .050 .308 −.407 −.631 .168 −.808 0.0527 272

S. E. .108 .276 .229 .224 .657 .446 .184 .925

Q25 Coef. .022 −.009 .021 .081 .040 −.518 .032 −.180 0.0157 272

S. E. .026 .073 .063 .059 .067 .512 .054 .175

***p < 0. 01, **p < 0.05, *p < 0.1 Table 3 QR estimation: the coefficient and standard error on innovation measures reported for 50, 75, and 90% quantiles Variables Ln (size) Ln (age) Product innovation Process innovation R&D internal R&D external Patents Constant R2/pseudo R2 N. of observation

Q50 Coef. .023* −.032 .011 .035 .049 .002 −.040 .056 0.0133 272

S. E. .009 .036 .037 .029 .044 .181 .047 .104

Q75 Coef. .013 −.053 .0501 .0192 .0357 −.073 .010 .283 0.0109 272

S. E. .014 .047 .041 .029 .128 .143 .063 .116

Q90 Coef. .0324 −.041 .198 −.143 .213 −.412 −.007 .383 .0614 272

S. E. .033 .065 .088*** .053*** .167* .184** .172 .199

***p < 0. 01, **p < 0.05, *p < 0.1

Overall, the results show that the innovation measures that have a strong and significant relationship with sales growth are product and process innovation. These results are in line with Vivarelli (2014) discussion, who suggested that these types of innovations are more suitable for firms operating in transition and developing economies.

4 Discussion Since the seminal work of Joseph Schumpeter (1934), innovation has been widely recognized as a source of competitive advantage and growth for firms. However,

Innovation and Sales Growth Among Heterogeneous Albanian Firms: A. . .

153

empirical research has struggled to confirm this theoretical axiom. As suggested by Coad (2009), returns from innovation are highly uncertain because of firms’ heterogeneity. Besides, they are difficult to determine due to challenges in measuring innovation. Additionally, standard regression analyses are inappropriate to investigate innovation-sales growth relations since the distribution of the latter is usually highly skewed. QR allows determining the effects of innovation measures over the entire conditional growth rate distribution (ibid). This study aims to analyze the innovation determinants of firm’s sales growth by accounting for the heterogeneity using both OLS and QR. First, the study results show that the strength of the causal link between innovativeness and firm growth is overestimated by the theory, at least for developing countries. These results do not come as a surprise considering the average low level of innovativeness and expenditure in R&D in these countries. However, independent from context, they are in line with other studies conducted in developed economies (e.g., Coad, 2009; Geroski, 2000) or even middle-income countries (e.g., Santi & Santoleri, 2017) that suggest that firms’ growth is very idiosyncratic. Second, the QR analysis results show that innovation contributes to sales growth only for fast-growth firms as suggested by other empirical studies (e.g., Coad & Rao, 2008; Santi & Santoleri, 2017; Segarra & Teruel, 2014). More specifically, product innovation and internal R&D have a positive effect on sales growth. The result related to the negative impact of process innovation and external R&D comes as a surprise although in line with other studies (e.g., Na & Kang, 2019). The presence of lagged effects might be a possible explanation. Since this study is crosssectional and has not controlled for lagged growth, it is difficult to capture the effects in time of different innovation activities. Future research needs to investigate this relationship more in-depth. Third, on a methodological note, OLS baseline analysis results and QR results depict entirely different pictures of the innovation-sales growth relation. While OLS regression results suggest that process innovation positively affects sales growth, the QR results show that this effect is either non-significant or negative. It can be argued that when dealing with highly skewed growth rates, it is necessary to use more parsimonious regression techniques such as QR. From a policy-making viewpoint, our results suggest that state-owned agencies should promote innovation among firms with high-growth potential. This approach is used by some donor development programs (e.g., some EU-funded programs) that are already doing through competitive grants, technical support, and other mechanisms. A one-size-fits-all solution does not produce the aimed results in terms of growth and, consequently, employment. Several limitations bound the study. First, external and internal R&D are measured as binary variables since data on R&D expenditure are missing almost completely. As suggested by Coad and Rao (2008), R&D expenditures, despite their limitations, can capture better the phenomenon of firm-level innovation. Second, this study is cross-sectional. Longitudinal data might shed some light on the effects of the sequential adoption of innovation activities and their lagged impact. Finally, this study focuses on the relationship of different measures of innovation on sales

154

B. Gerdoçi and S. Dibra

growth, neglecting the compounding effect of different innovation activities (see Bianchini et al., 2018; Coad et al., 2016). Future research can use composite ‘innovativeness’ indexes.

References Aghion, P., Bloom, N., Blundell, R., Griffith, R., & Howitt, P. (2005). Competition and innovation: An inverted-U relationship. The Quarterly Journal of Economics, 120(2), 701–728. https:// doi.org/10.3386/w9269 Audretsch, D. B., Coad, A., & Segarra, A. (2014). Firm growth and innovation. Small Business Economics, 43(4), 743–749. https://doi.org/10.1007/s11187-014-9560-x Banbury, C. M., & Mitchell, W. (1995). The effect of introducing important incremental innovations on market share and business survival. Strategic Management Journal, 16(Special issue), 161–182. https://doi.org/10.1002/smj.4250160922 Becchetti, L., & Trovato, G. (2002). The determinants of growth for small and medium sized firms. The role of the availability of external finance. Small Business Economics, 19(4), 291–306. https://doi.org/10.1023/A:1019678429111 Bianchini, S., Pellegrino, G., & Tamagni, F. (2018). Innovation complementarities and firm growth. Industrial and Corporate Change, 27(4), 657–676. https://doi.org/10.1093/icc/dty008 Bottazzi, G., Coad, A., Jacoby, N., & Secchi, A. (2005). Corporate growth and industrial dynamics: Evidence from French manufacturing. Applied Economics, 43(1), 103–116. https://doi.org/ 10.1080/00036840802400454 Bottazzi, G., Dosi, G., Lippi, M., Pammolli, F., & Riccaboni, M. (2001). Innovation and corporate growth in the evolution of the drug industry. International Journal of Industrial Organization, 19(7), 1161–1187. https://doi.org/10.1016/S0167-7187(01)00068-6 Brouwer, E., & Kleinknecht, A. (1999). Innovative output, and a firm’s propensity to patent: An exploration of CIS micro data. Research Policy, 28(6), 615–624. https://doi.org/10.1016/S00487333(99)00003-7 Buchinsky, M. (1998). Recent advances in quantile regression models: A practical guideline for empirical research. Journal of Human Resources, 33(1), 88–126. https://doi.org/10.2307/ 146316 Coad, A. (2009). The growth of firms: A survey of theories and empirical evidence. Cheltenham: Edward Elgar Publishing. Coad, A., & Rao, R. (2008). Innovation and firm growth in high tech sectors: A quantile regression approach. Research Policy, 37(4), 633–648. https://doi.org/10.1016/j.respol.2008.01.003 Coad, A., & Rao, R. (2011). The firm-level employment effects of innovations in high-tech US manufacturing industries. Journal of Evolutionary Economics, 21(2), 255–283. https://doi.org/ 10.1007/s00191-010-0209-x Coad, A., Segarra, A., & Teruel, M. (2013). Like milk or wine: Does firm performance improve with age? Structural Change and Economic Dynamics, 24(1), 173–189. https://doi.org/ 10.1016/j.strueco.2012.07.002 Coad, A., Segarra, A., & Teruel, M. (2016). Innovation and firm growth: Does firm age play a role? Research Policy, 45(2), 387–400. https://doi.org/10.1016/j.respol.2015.10.015 Cohen, W. M., & Klepper, S. (1996). Firm size and the nature of innovation within industries: The case of process and product R&D. The Review of Economics and Statistics, 78(2), 232–243. https://doi.org/10.2307/2109925 Cohen, W., Nelson, R., & Walsh, J. (2000). Protecting their intellectual assets: Appropriability conditions and why US manufacturing firms patent (or not). NBER working paper no 7552. https://doi.org/10.3386/w7552

Innovation and Sales Growth Among Heterogeneous Albanian Firms: A. . .

155

Cooper, R., & Kleinschmidt, E. (1987). New products: What separates winners from losers? Journal of Product Innovation Management, 4(3), 169–184. https://doi.org/10.1016/07376782(87)90002-6 Damanpour, F. (1991). Organizational innovation: A metaanalysis of effects of determinants and moderators. Academy of Management Journal, 34(3), 555–590. https://doi.org/10.2307/256406 Deeds, L. (2001). The role of R&D intensity, technical development and absorptive capacity in creating entrepreneurial wealth in high technology start-ups. Journal of Engineering and Technology Management, 18(1), 29–47. https://doi.org/10.1016/S0923-4748(00)00032-1 Demirel, P., & Mazzucato, M. (2013). Innovation and economic performance (industrial and financial): Recent results and questions for future research. Innovation and finance. Routledge. Edquist, C. (2001). Innovation policy: A systemic approach. In D. Archibugi & B. A. Lundvall (Eds.), The globalising learning economy (pp. 219–238). Oxford: Oxford University Press. Freel, M., & Robson, P. (2004). Small firm innovation, growth and performance evidence from Scotland and Northern England. International Small Business Journal, 22(6), 561–575. https:/ /doi.org/10.1177/0266242604047410 Geroski, P. A. (2000). The growth of firms in theory and practice. In N. Foss & V. Mahnke (Eds.), New directions in economic strategy research. Oxford: Oxford University Press. Geroski, P., & Mazzucato, M. (2002). Learning and the sources of corporate growth. Industrial and Corporate Change, 11(4), 623–644. https://doi.org/10.1093/icc/11.4.623 Guarascioa, D., & Tamagni, F. (2019). Persistence of innovation and patterns of firm growth. Research Policy, 48(6), 1493–1512. https://doi.org/10.1016/j.respol.2019.03.004 Hervas-Oliver, J. L., Sempere-Ripoll, F., & Boronat-Moll, C. (2014). Process innovation strategy in SMEs, organizational innovation and performance: A misleading debate? Small Business Economics, 43(4), 873–886. https://doi.org/10.1007/s11187-014-9567-3 Johannessen, J. A., Olsen, B., & Lumpkin, G. T. (2001). Innovation as newness: What is new, how new, and new to whom? European Journal of Innovation Management, 4(1), 20–31. https:// doi.org/10.1108/14601060110365547 Krishnan, V., & Ulrich, K. (2001). Product development decisions: A review of the literature. Management Science, 47(1), 1–21. https://doi.org/10.1287/mnsc.47.1.1.10668 Lööf, H., & Heshmati, A. (2006). On the relationship between innovation and performance: A sensitivity analysis. Economics of Innovation and New Technology, 15(4/5), 317–344. https:// doi.org/10.1080/10438590500512810 McKelvie, A., Brattström, A., & Wennberg, K. (2017). How young firms achieve growth: Reconciling the roles of growth motivation and innovative activities. Small Business Economics, 49(2), 273–293. https://doi.org/10.1007/s11187-017-9847-9 Na, K., & Kang, Y.-H. (2019). Relations between innovation and firm performance of manufacturing firms in Southeast Asian emerging markets: Empirical evidence from Indonesia, Malaysia, and Vietnam. Journal of Open Innovation: Technology, Market, and Complexity, 5(4), 98. https://doi.org/10.3390/joitmc5040098 Nelson, R., & Winter, S. (1982). An evolutionary theory of economic change. Cambridge: The Belknap Press of Harvard University Press. Papinniemi, J. (1999). Creating a model of process innovation for reengineering of business and manufacturing. International Journal of Production Economics, 60–61, 95–101. https://doi.org/ 10.1016/S0925-5273(98)00146-7 Parisi, M., Schiantarelli, F., & Sembenelli, A. (2006). Productivity innovation and R&D: Micro evidence for Italy. European Economic Review, 50(8), 2037–2061. https://doi.org/10.1016/ j.euroecorev.2005.08.002 Roper, S. (1997). Product innovation and small business growth: A comparison of the strategies of German, UK and Irish companies. Small Business Economics, 9(6), 523–537. https://doi.org/ 10.1023/A:1007963604397 Santi, C., & Santoleri, P. (2017). Exploring the link between innovation and growth in Chilean firms. Small Business Economics, 49(2), 445–467. https://doi.org/10.1007/s11187-016-9836-4 Schumpeter, J. (1934). The theory of economic development. Cambridge: Harvard University Press.

156

B. Gerdoçi and S. Dibra

Segarra, A., & Teruel, M. (2014). High-growth firms and innovation: An empirical analysis for Spanish firms. Small Business Economics, 43(4), 805–821. https://doi.org/10.1007/s11187014-9563-7 Vivarelli, M. (2014). Innovation, employment and skills in advanced and developing countries: A survey of economic literature. Journal of Economic Issues, 48(1), 123–154. https://doi.org/ 10.2753/JEI0021-3624480106 World Bank Albania. (2019). Enterprise Survey (ES) World Ban Ref. ALB_2019_ES_v01_M. Retrieved March, 2020, from https://www.enterprisesurveys.org/portal/login.aspx Yasuda, T. (2005). Firm growth, size, age and behavior in Japanese manufacturing. Small Business Economics, 24(1), 1–15. https://doi.org/10.1007/s11187-005-7568-y

Quantitative Analysis of Inequalities at ICT Sector in Visegrad Countries Tatiana Corejova, Roman Chinoracky, and Alexandra Valicova

Abstract The information and communication technology sector significantly influences business models, companies or processes. It is an integral part of the economy and a part of key innovations. It is also the essence and bearer of the great economic paradoxes of today. Rapid advances in ICT create opportunities to gain market advantage and evoke challenges in relation to consumption, distribution, allocation of factors of production, evaluation of efficiency and effectiveness. Inequalities between market players occur in each market and relate to the unequal distribution of income, assets or access to scarce resources throughout society. In the quantitative analysis of the ICT sector, we focus on economic disparities and inequalities between different categories of business entities. To identify inequalities in the ICT sector, the procedures used to quantify income inequalities are used. The results of the study of the ICT sector in the V4 countries show significant differences in the shares of individual size categories of companies in total turnover, total assets and intangible fixed assets. This indicates inequalities, the magnitude of which is reflected in both the Lorenz curves and the Gini coefficients. The results of research in the V4 countries confirmed the dynamic changes in the level of concentration as well as the reduction of inequalities between different size categories of companies in the market. Keywords Concentration · Gini coefficient · inequalities · ICT sector · Lorenz curve · Visegrad countries

T. Corejova () · R. Chinoracky · A. Valicova University of Zilina, Zilina, Slovakia e-mail: [email protected]; [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_12

157

158

T. Corejova et al.

1 Introduction In each country, the ICT sector represents an important and strategically important area for the state, companies, households and citizens-customers. The development of ICT has a direct and fundamental impact on the development of the entire infrastructure and economy of each country. ICTs are the driving force behind the development of the economy and business. The ICT sector has continued to experience enormous dynamic development worldwide in recent decades, which has a significant impact on the development of the economy and society as a whole. At present, the difficulties of the multifaceted and far-reaching changes in the electronic communications sector that have triggered the transition from a monopoly to a competitive environment still persist. In the area of services, where voice services in fixed networks predominated, new services of fixed and mobile networks dominate. Following the creation of a transparent competitive environment, conditions have been created for new market entrants, stabilization and gradual market growth. The current nature of communication platforms, especially the Internet, opens the door to the integration of the world economy and brings challenges and challenges for individual countries. An important and inseparable attribute of the ICT sector’s offer is sufficiently secured data protection, fulfilling the customer’s vision and helping to build trust in the new digital environment (Chinoracky, 2019; Pólya et al., 2011). The development of the ICT sector is reflected in new concepts such as the digital economy. The digital economy as an umbrella term is used to describe markets that focus on digital technologies. Typically, these are electronic transactions with information goods and services. In the study “Harnessing the digital economy for developing countries” (Dahlman, Mealy, & Wermelinger, 2016), the digital economy is said to be a combination of several universal technologies and economic and social activities carried out by people over the Internet and related technologies (Ristvej, Lacinak, & Ondrejka, 2020). It includes the physical infrastructure on which digital technologies are based, accessible devices, applications that use accessible devices and the functions they provide. The digital economy has permeated virtually all aspects of modern life, including retail, transport, education, agriculture and benefits consumers, businesses and governments (Dahlman et al., 2016). Tsyganov and Apalkova (2016) define the term digital economy as a paradigm of the global information society. In general, it is a model of post-industrial development of the world economy, which is based on the use of technological platforms such as the Internet and other electronic devices and the creation of a set of financial and economic relations in the system of production, distribution, exchange and consumption of goods and services in world markets. Production, manufacturing and services related to digital technologies fall under the ICT sector. The ICT sector is the core of the digital sector (Bukht & Heeks, 2017). The following sections are devoted to comparison of Visegrad countries or V4 that include The Czech Republic, Hungary, Poland and The Slovak Republic, by the structure of ICT industry and level of inequalities among different size of

Quantitative Analysis of Inequalities at ICT Sector in Visegrad Countries

159

companies. The study builds on previous analysis (Corejová, 2013; Majercakova, 2014; Stofkova & Stofko, 2016) of the regional disparities as well as tendencies of the ICT industry (Chatfield, Borsella, Mantovani, Porcari, & Stahl, 2017; Kim & Kim, 2018). For the comparison of different countries conditions related to the ICT sector, its conditions, consolidation and development, it is necessary to include more characteristics or features (Madudova, Corejova, & Valica, 2018; Wannapan & Chaiboonsri, 2018). So, this evoke the research questions: is it possible to use the concept of evaluation of income inequalities for investigation of conditions at the ICT market? How can we identify the differences? Can we use the methods of concentration evaluation? And what indicators can we use? And can we identify the trends and causes of the inequalities at the ICT market?

2 Theoretical Background The issue of market inequalities, regional differences and regional development is a frequently discussed topic today on a national and international scale (Hudec, 2007). Various methods and procedures are used to understand and quantify inequalities and regional differences. The reasons that cause inequalities between market players and interregional differences are diverse and specific to a particular sector or region. Inequalities between market players occur in each market and relate to the unequal distribution of income, assets or access to scarce resources throughout society. The share of large enterprises in the total volume of production, the size of turnover, the range of assets and their individual types expresses the level of concentration (Mas, Robledo, & Peréz, 2012; Šíbl et al., 2002). The Herfindal index (HHI), the values of which belong to the interval (0, 1>), is most often used to measure the concentration rate in a given industry. Regional disparities express differences and inequalities in signs, phenomena or processes. We define them as “a consequence of regional development, when regional development in specific historical conditions can lead to uneven development of regions, resulting in a number of inequalities: social, economic, cultural, infrastructural, inequalities in living conditions, living standards, etc., which may lead to regional polarization (quantitative and qualitative)” (Kožiak, 2008). The geographical and social development of inequalities is a manifestation of hierarchical organization (Lauko et al., 2014). Inequalities can be a source of social and regional tension or can threaten the ecological balance and stability of society. Viturka (2008) states that inequalities arise from the developmental and hierarchical differentiation of social systems. Výrostová (2010) distinguishes three types of regional disparities: • economic disparities—relate to the difference in the quality and quantity of regional output • social disparities—relate to the income or standard of living of the population • territorial disparities (physical disparities)—relate to geographical or natural conditions

160

T. Corejova et al.

Due to research questions concerning the situation in the ICT sector in other parts of the work, we will focus on economic disparities and inequalities between different categories of business entities. To identify inequalities in the ICT sector, we will use modified procedures used in the quantification of income inequalities, namely the Lorenz curve and the Gini coefficient. The Lorenz curve shows a standard uneven distribution of wealth in a population. The perfect distribution of wealth in society means that there are no inequalities and everyone has the same wealth. In this case, the curve is a straight line with a 45◦ angle in the standard xy coordinate system. The absolute inequality in the distribution of wealth, that is, wealth owned by one person in the whole society, copies the axes of the graph. In reality, the Lorenz curve is in the range below the 45◦ curve and above the graph axes. The further the curve deviates from the line of absolute equality, the greater the income inequality in society. The magnitude of the deviation of the Lorenz curve from absolute equality expresses the degree of inequality (Šíbl et al., 2002). Gini’s coefficient reflects numerically the Lorenz curve. It is calculated as twice the area between the ideal curve and the actual Lorenz curve. The coefficient takes values in the interval < 0, 1>, i.e. from absolute equality to absolute inequality (Byrtusova, 2015). The closer the index is to 1, the greater is the inequality in society.

3 Data and Methodology In identifying differences in the ICT sector within the V4 countries, we used the international standard classification of economic activities (ISIC, 2008). It is used to break down data that are linked to an economic entity as a statistical unit. It is one of the tools used to implement various statistics such as outputs, inputs to the production process, capital formation and financial transactions of economic entities. Data from OECD, Finstat and Amadeus databases (Amadeus, 2020; Finstat, 2020; OECD, 2018, 2019) are used. Business entities included in the assessment of inequalities and disparities in the ICT sector in the V4 countries were divided into four size groups marked as follows: small companies, medium companies, large companies and very large companies (Amadeus, 2020; Finstat, 2020). The classification characteristic was the volume of the company’s total assets. When calculating the sample of entities operating in the ICT sector according to (Amadeus, 2020, Finstat, 2020), a confidence interval of 95% and a maximum error margin of ±5% were required. The division of companies into size groups in percentage terms is shown in Table 1. The research objectives were focused on: 1. determining the degree of inequality in the ICT sector between four categories of companies in each of the V4 countries

Quantitative Analysis of Inequalities at ICT Sector in Visegrad Countries

161

Table 1 Percentage representation of companies by size categories Category of companies by size in % All Very large Large Medium Small

Czech Republic 100 3.72% 24.78% 60.00% 11.50%

Slovakia 100 2.36% 20.35% 64.90% 12.39%

Hungary 100 3.50% 13.83% 70.67% 11.25%

Poland 100 10.13% 23.45% 61.16% 5.25%

2. quantification of the degree of inequality in the ICT sector in the V4 countries 3. identification of inequality rate trends Within these objectives, we assume that there are significant inequalities between individual categories of companies in the ICT sector in terms of their market shares and that the most significant inequalities are reflected in intangible fixed assets. At the same time, we start from the fact (Synek et al., 2003) that the advantages of large companies in the ICT sector include capital strength, which leads to the possibility of investing in intangible fixed assets and consequently to higher labour productivity. We verify the research assumptions using the market concentration indicator and by quantifying inequalities and disparities on the basis of three selected indicators, namely: • operating revenue or turnover in thousands A C include net sales, other operating revenues and stock variations and do not include VAT • total assets in thousands A C • intangible fixed assets in thousands A C represent formation expenses, research expenses, goodwill, development expenses and all other expenses with a longterm effect

4 Results Based on the available data, we processed the characteristics of the ICT sector for four countries in the period of 2015–2018 in terms of the development of inequalities between the four size categories of companies. Table 2 shows the market concentration levels as assessed by the Herfindal Concentration Index (HHI). Data on turnover, total assets and intangible assets were used to assess the level of concentration. The results show that it is possible to use this index to evaluate the concentration level. The index of shares in intangible fixed assets shows the highest level of concentration in all countries although in the given period 2015–2018 the level of concentration decreased with the exception of Hungary. According to the indicators of companies’ shares in turnover, there was a slight increase in shares in two countries and a decrease in two. However, there is no stable trend of reducing the concentration of large and very large companies in the ICT market.

162 Table 2 Level of concentration (HHI) by selected indices in V4 countries in 2015–2018

T. Corejova et al. Indices Operating revenue Total assets Intangible fixed assets Operating revenue Total assets Intangible fixed assets Operating revenue Total assets Intangible fixed assets Operating revenue Total assets Intangible fixed assets

2015 2016 Czech Republic 0.458 0.447 0.586 0.520 0.745 0.687 Slovakia 0.355 0.378 0.338 0.355 0.671 0.550 Hungary 0.377 0.384 0.542 0.536 0.512 0.592 Poland 0.439 0.424 0.534 0.495 0.744 0.644

2017

2018

0.443 0.530 0.669

0.439 0.545 0.698

0.378 0.350 0.452

0.380 0.356 0.461

0.381 0.522 0.549

0.387 0.540 0.546

0.419 0.490 0.500

0.420 0.481 0.675

Table 3 Share of very large companies on the ICT market by selected indices in V4 countries in 2015–2018 Indices Operating revenue Total assets Intangible fixed assets Operating revenue Total assets Intangible fixed assets Operating revenue Total assets Intangible fixed assets Operating revenue Total assets Intangible fixed assets

2015 Czech Republic 61.15% 74.22% 85.78% Slovakia 37.92% 35.88% 68.22% Hungary 50.50% 70.84% 68.22% Poland 58.82% 69.56% 85.41%

2016

2017

2018

59.04% 68.23% 81.88%

58.45% 69.24% 80.57%

57.08% 70.63% 82.50%

44.62% 36.02% 74.99%

44.63% 38.40% 71.43%

44.28% 38.54% 71.22%

51.64% 70.28% 74.99%

50.57% 68.99% 71.43%

51.33% 70.63% 71.22%

56.48% 65.86% 78.12%

55.57% 65.19% 64.15%

55.32% 64.24% 80.48%

There are significant differences between the numbers of size categories of companies (see Table 1). It can be assumed that it is important to examine inequalities in the share of total turnover, total assets and intangible assets (Table 3). To determine the level of inequality according to the three mentioned indicators, we used the Lorenz curve. Figure 1 shows the inequalities in the intangible fixed assets.

Quantitative Analysis of Inequalities at ICT Sector in Visegrad Countries

163

Fig. 1 Lorenz curves for intangible fixed assets

After determining the Lorenz curves, the Gini coefficients were calculated for all indicators and countries (see Table 4). It is clear from the values of the coefficient that in the course of 2015–2018, inequalities between the size categories of companies in the ICT market decreased in all indicators, while the largest decrease

164 Table 4 Gini coefficients by selected indices in V4 countries in 2015–2018

T. Corejova et al. Country

CZ SK H PL CZ SK H PL CZ SK H PL

Gini coefficient 2015 2016 2017 2018 By operating revenue (turnover) 0.742 0.587 0.586 0.583 0.653 0.512 0.513 0.513 0.689 0.522 0.521 0.529 0.773 0.568 0.563 0.566 By total assets 0.806 0.627 0.631 0.640 0.611 0.456 0.463 0.469 0.502 0.621 0.615 0.625 0.820 0.612 0.610 0.605 By intangible fixed assets 0.860 0.699 0.693 0.706 0.918 0.697 0.629 0.708 0.770 0.652 0.630 0.627 0.905 0.687 0.691 0.686

was recorded between 2015 and 2016. However, the values of the coefficient show values above 0.4, which indicates inequalities in the distribution.

5 Discussion and Conclusion The results of the study of the ICT sector in the V4 countries show significant differences in the shares of individual size categories of companies in total turnover, total assets and intangible fixed assets. This indicates inequalities, the magnitude of which is reflected in both the Lorenz curves and the Gini coefficients. Between 2015 and 2018, we record a slight decrease in the level of concentration in the Czech Republic and Poland (Table 2), while the highest decrease in concentration is reflected in the indicator of intangible fixed assets, in Czech Republic by 0.047 points, in Poland even by 0.069. The highest decrease in the level of concentration according to the share of intangible fixed assets was recorded in Slovakia even by 0.21 points. These values correspond to the conclusions of Ha Thi Thu Le, Quyen Thi Mai Dao, Van-Chien Pham and Duong Thuy Tran (2019). The ICT sector has an impact on technological innovation due to the types of innovation. Cooperation with other companies has a positive impact on the creation of new ICT companies’ technological innovation (Kim & Kim, 2018). ICT companies such as start-ups are perceived through innovation activity and also through the concept of openinnovation. Use of the term “open innovation” has been promoted in particular by Henry Chesbrough (2003, 2007; Delgado-Verde, Martín-de-Castro, & Navas, 2011). The concept of open innovation is also associated with a change in the perception of intellectual property and related rights (Chesbrough & Bogers, 2014).

Quantitative Analysis of Inequalities at ICT Sector in Visegrad Countries

165

The decrease in the Gini coefficient for intangible fixed assets ranged from 0.143 in Hungary to 0.245 in Poland. Overall, a decrease in the Gini coefficient was recorded in all indicators, except Hungary in the case of total assets. The values indicate a decrease in inequalities between individual size categories of companies in the ICT sector. The highest decrease in inequalities was recorded in all indicators between 2015 and 2016. The main causes of changes in the ICT sector include differences resulting from access to new forms of business models associated with start-ups, crowdsourcing methods, changes in licensing options, speed of innovation and, above all, strengthening cooperation between different entities and network effects. However, there are also reasons for the different degree of distribution of ownership of and access to networks. These are linked to the market position of the incumbents and the ways in which they are regulated. The research assumptions were confirmed on the basis of the achieved results. The ICT sector is developing dynamically in the V4 countries, which is reflected in changes in the level of concentration and in the reduction of inequalities between different size categories of companies in the market. Changes in the economy associated with digitization and digital transformation, as well as lifestyle changes with the widespread use of electronic devices, play an important role in reducing inequalities in the ICT sector. Acknowledgement This contribution was undertaken as a part of the research project 1/0152/18 VEGA Business models and platforms in the digital environment.

References Amadeus. (2020). Data update: 13/03/2020. Bureau van Dijk. A Moody’s Analytics Company. Retrieved March 13, 2020, from https://amadeus.bvdinfo.com/version-2019919/ home.serv?product=AmadeusNeo, n0 3061 Bukht, R. and Heeks, R. (2017). Defining, conceptualising and measuring the digital economy. In Development informatics. 2017. Retrieved October 4, 2019, from https:// diodeweb.files.wordpress.com/2017/08/diwkppr68-diode.pdf Byrtusova, A. (2015). Use of the index methods of income inequality for assessing the security level. In Economics and management N. 2/2015. Brno: University of Defence in Brno. Retrieved May 1, 2020, from https://www.unob.cz/eam/Documents/Archiv/EaM_2_2015/ Byrtusova.pdf Chatfield, K., Borsella, E., Mantovani, E., Porcari, A., & Stahl, B. C. (2017). An investigation into risk perception in the ICT industry as a Core component of responsible research and innovation. Sustainability, 9(8), 1424. Chesbrough, H. (2003). The era of open innovation. MIT Sloan Management Review, 44(3), 35–41. Chesbrough, H. (2007). Business model innovation: It’s not just about technology anymore. Strategy & Leadership, 35(6), 12–17. https://doi.org/10.1108/10878570710833714 Chesbrough, H., & Bogers, M. (2014). Explicating open innovation: Clarifying an emerging paradigm for understanding innovation. In H. Chesbrough, W. Vanhaverbeke, & J. West (Eds.), New frontiers in open innovation (pp. 3–28). Oxford: Oxford University Press.

166

T. Corejova et al.

Chinoracky, R. (2019). Hodnotenie potenciálu digitálnych technológii pri transformacii podnikov, hospodárstva a spoloˇcnosti. Project to dissertation exam. Žilinská univerzita v Žiline, FPEDAS, UNIZA, 2019. Corejová, T. (2013). Indikátory sektora IKT a ich komparácia v rámci V4. In: Rozvoj Euroregiónu Beskydy VII : poznávanie a diagnostika spoloˇcensko - ekonomických podmienok v regiónoch : medzinárodná vedecká konferencia : zborník : Žilina, 28. február 2013. - Žilina: Žilinská univerzita, 2013. - ISBN 978-80-554-0767-8. - S. 31–37. Dahlman, C., S. Mealy and M. Wermelinger (2016), Harnessing the digital economy for developing countries, OECD Development Centre Working Papers, no. 334, OECD Publishing, Paris, https://doi.org/10.1787/4adffb24-en Delgado-Verde, M., Martín-de-Castro, G., & Navas, L. (2011). Social capital, relational capital and technological innovation. Empirical evidence in Spanish high and medium-high technology manufacturing firms. Cuadernos De Economía Y Dirección De La Empresa, 14(4), 207–221. https://doi.org/10.1016/j.cede.2011.04.001 Finstat. (2020). Retrieved April 29, 2020, from https://www.finstat.sk/ Le, H. T. T., Dao, Q. T. M., Pham, V.-C., & Tran, D. T. (2019). Global trend of open innovation research: A bibliometric analysis. Cogent Business & Management, 6(1), 1633808. To link to this article:. https://doi.org/10.1080/23311975.2019.1633808 Hudec, O. (2007). Regionálne inovaˇcné systémy: strategické plánovanie a prognózovanie (p. 204). Košice: Ekonomická fakulta Technickej univerzity v Košiciach. isbn:978-80-8073-964-5. ISIC (2008). International Standard Industrial Classification of All Economic Activities (ISIC), Rev.4. United Nations. New York: United Nations. 2008. ISBN 978-92-1-161518-0. Retrieved December 12, 2019, from https://unstats.un.org/unsd/publication/seriesM/seriesm_4rev4e.pdf> Kim, S. Y., & Kim, E. (2018). How intellectual property management capability and network strategy affect open technological innovation in the Korean new information communications technology industry. Sustainability, 10(8), 2600. Kožiak, R. (2008). Zmierˇnovanie regionálnych disparít prostredníctvom regionálnej politiky. 1.vyd. Banská Bystrica: EF Univerzita Mateja Bela,2008. 138 s. ISBN 978-80-8083-573-6. Lauko, V. et al. (2014) Regionálne dimenzie Slovenska. Bratislava:UK, 2014. 524 s. ISBN 97880-223-3725-0. Madudova, M., Corejova, T., & Valica, M. (2018). Economic Sustainability in a wider context: Case study of considerable ICT sector sub-divisions. Sustainability, 10(7). https://doi.org/ 10.3390/su10072511. July 2018, MDPI. Majercakova, M. (2014). Vplyv regulácie na vývoj roamingových cien operátorov v Slovenskej ˇ a Ceskej republike. In: IPoCC - international postal and e-communications conference, proceedings, Pardubice, September 18th-19th, 2014. ISBN 978-80-86530-94-9. - CD-ROM, s. 132-137. Mas, M.; Robledo, J. C. and Peréz, J. (2012) ICT sector definition transition from NACE rev. 1.1 to NACE rev. 2. Retrieved October 16, 2019, from http://publications.jrc.ec.europa.eu/repository/ bitstream/JRC77364/jrc77364.pdf> OECD (2019). Measuring the digital transformation: A roadmap for the future. OECD (2018) Key ICT indicators. Retrieved October 17, 2019, from http://www.oecd.org/internet/ ieconomy/oecdkeyictindicators.htm Pólya, A. et al. (2011) Informatika v 21.storoˇcí. 1.vyd. Bratislava: EKONÓM, 2011. 360 s. ISBN 978-80-225-3285-3. Ristvej, J., Lacinak, M., & Ondrejka, R. (2020). On smart city and safe city concepts. Mobile Networks and Applications, 17, 836. https://doi.org/10.1007/s11036-020-01524-4. Retrieved from: https://link.springer.com/article/10.1007/s11036-020-01524-4 Stofkova, Z. and Stofko, S. (2016). Some indicators of quality of life in a globalized world. In: 16th Int. Scientific Conference on Globalization and its Socio-Economic Consequences. Rajecke Teplice, Slovakia. Oct 05–06, 2016, PTS I-V. Pages 2127–2134 Synek, M., et al. (2003). Manažérska ekonomika (p. 466). Grada Publishing a.s. isbn:80-247-0515X. Šíbl. D. et al. (2002). Vel’ká ekonomická encyklopédia. SPRINT. P. 967. ISBN 80-89085-04-0

Quantitative Analysis of Inequalities at ICT Sector in Visegrad Countries

167

Tsyganov, S. and Apalkova, V. (2016). Digitálna ekonomika: nová paradigma globálnej informaˇcnej spoloˇcnosti. In Ekonomické rozhl’ady. 2016. roˇc. 45., cˇ is. 3. s 295–311. Viturka, M. (2008) Disparity v regionálním rozvoji. [online]. Brno: Centrum výzkumu konkurencieschopnosti cˇ eskej ekonomiky, 2008. 18 s. ISSN 1801-4496. Retrieved October 13, 2019, from https://www.researchgate.net/publication/268361825_Disparity_v_regionalnim_rozvoji Výrostová, E. (2010) Regionálna ekonomika a rozvoj. 1.vyd. Bratislava: Iura Edition spol s.r.o., 2010. 352 s. ISBN 978-80-8078-361-7. Wannapan S., Chaiboonsri C. (2018). An analysis of ICT sectors and economic growth: Evidence from ASEAN countries. In: Tsounis, N., Vlachvei, A. (eds) Advances in panel data analysis in applied economic research. ICOAE 2017. Springer proceedings in business and economics. Springer, Cham. ISBN 978-3-319-70054-0, DOI: https://doi.org/10.1007/978-3-319-700557_14

Does Government Spending Cause Investment?: A Panel Data Analysis Nihal Bayraktar

Abstract Government spending has increased recently in almost all countries to ease the severely negative economic impacts of the current health crisis. Similar expansionary fiscal and monetary policies were observed during other economic downturns too. The effectiveness of expansionary policies, especially in terms of their effects on investment, has been discussed widely. Thanks to the possible multiplier effect of higher government expenditures, it is expected that government spending would generate a higher amount of income and therefore consumption in economies. At some point, this higher government spending with glowing economic activities is expected to increase private investment—an important item to promote job creations and improvements in production. Although one of the direct or indirect expected outcome of higher government spending is larger investment, many empirical studies in the literature cannot observe this positive expected effect of government spending on investment. As a result, even the necessity of increased government expenditures during economic crisis has been questioned. In this paper, the causal and correlation relationship between government spending and investment is investigated in a panel data setting to better evaluate the importance of higher government spending during economic downturns. The findings show that country classifications based on income, time periods covered in the analysis, measures of government spending and investment, and the time lag of government policies can make a difference. There are cases where government spending highly significantly causes private investment, and high correlations between two variables are observed. Therefore, accurate evaluations of the impacts of government spending on investment may require detailed data analysis. Keywords Government spending · Investment · Business cycles · Panel data · Causality · Correlation

N. Bayraktar () Penn State University—Harrisburg, Middletown, PA, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_13

169

170

N. Bayraktar

JEL codes: E2, E3, E6, H5

1 Introduction Many countries have been dealing with one of the most serious economic crises of our modern history. Economic activities are expected to decline globally due to partial economic closings and lockdowns as a result of the covid-19 health crisis. Some economists name this crisis as crisis by design because governments deliberately fully or partially closed their economies to ease the pressure on the health system. However, even without lockdowns, the spreading virus was going to deteriorate labor force so badly that the economy was going to close at some point anyway. Whether this crisis is named crisis by design or real crisis, the response of governments has been similar as businesses have been temporarily or permanently closed and unemployment rates have jumped. Almost all countries announced expansionary monetary policies. Central banks cut their target interest rates, increased loans to banks and businesses, and engaged in open market operations. For example, the Federal Reserve System of the U.S. has announced open-ended open market purchases and creating loans for the first time in its history in March 2020. In many advanced and developing countries, expansionary monetary policies have been accompanied by easy fiscal policies. Government spending has jumped as new stimulus packages have been announced to help consumers, unemployed people, and businesses. Tax collection deadlines have been extended, and relaxed tax policies have been announced for closed businesses and unemployed people. Expansionary government policies are commonly used in many countries during crisis to partially replace declining private consumption and private investment. The logic behind this higher government spending is its multiplier effect. Governments make more transfers and spend more money during crisis to create income for some group of people and businesses. These people and businesses start spending this additional income and generate even more income for some other groups of people and businesses. Through this mechanism, government spending is expected to generate higher incomes and private expenditures in the economic system. As the economy expands with this stimulus from higher government spending, total private investments start increasing at some points, as firms face with better profit opportunities. Therefore, firms start to expand their businesses, open new businesses, purchase more machines, equipment, and tools. This larger amount of investment is expected to be the real push for economic expansion and then lower unemployment rates through job creations. The link between unemployment and investments is solid. Even if the economy starts to expand, it takes much longer to lower unemployment rates because of cautious and hesitant actions of businesses to undertake expensive investments before they are sure about their profitability. Therefore, it is important to support higher investment by firms so that they can produce more and create more jobs to lower unemployment. In this

Does Government Spending Cause Investment?: A Panel Data Analysis

171

regard, one of the roles of higher government spending, directly or indirectly, is to promote investment to accomplish solid economic improvements and rising economic prosperity with better job opportunities. The issue in this story is that, in addition to these expected positive effects of higher government expenditures on private consumption and investment, there might be undesired negative effects too. Higher government spending means higher government debt and larger government budget deficit. This may lead to higher future taxes and interest rates, both of which can crowd out the private sector. In this process, inflation can also be an issue. Additionally, as governments cut back their spending to the reasonable levels over time, countries may face a new set of economic crises with declining economic activities. Therefore, in order to better evaluate the impact of higher government spending, we need to consider both positive and negative effects to understand the net impact on economies. It should be always remembered that higher government expenditures may not necessarily benefit the economy all the time, and this may potentially lead to debt crisis, new economic crisis, and inflation problems in the future. Given the undeniable importance of higher total investment for improvements in the economic activities and unemployment numbers, it is essential to understand the nature of the role and the impact of government spending on private investment and other major macroeconomic indicators during global crisis. In this paper, the aim is to run some analysis to understand the causal and correlation relationship between government spending and private investment in select high-, middle-, and low-income countries between 1970 and 2018 as well as, more specifically, during and after the 2008 global economic and financial crisis. The aim of the paper is to evaluate how effective government spending is in terms of its impacts of private investment in different economies, especially during economic downturns. For such analysis, it is essential to consider both cross section and time dimensions. Therefore, a panel dataset is used in analysis, and panel Granger causality tests are calculated and compared across different time periods, definitions, time lags, and country groups. The findings show that country classifications based on income, time periods covered in the analysis, measures of government spending and investment, and the time lag of government policies make a difference in terms of the effectiveness of government spending on private investment. There are cases where government spending highly significantly causes private investment with high correlations. Therefore, accurate evaluations of the impacts of government spending on investment may require detailed data analysis. The findings of the paper are also helpful to evaluate the expected effects of higher government spending on investment during the current global health and subsequent economic crisis.

172

N. Bayraktar

2 Effects of Government Spending and Literature Review The expenditure definition of gross domestic production (GDP), a major measure of economic development, consists of private consumption, private investment, government spending, and net exports (i.e., exports minus imports): GDP = Private consumption+Private investment+Government spending+net exports (1)

On the one hand, in many advanced economies the share of private consumption is the largest, while the share of private investment is relatively low. On the other hand, in developing countries the share of private investment can be relatively higher than what we observe in advanced economies. Positive or negative impacts of government spending on private consumption and private investment can be observed anytime, but this link gets even more important during economic downturns and crisis. Governments start to follow expansionary monetary and fiscal policies to stimulate weak economic conditions. As economies start to experience problems, people tend to lose their jobs and incomes. At the same time, businesses start to face a lower demand for their products and subsequent lower profits. When a recession hits, two components of GDP, namely private consumption and private investment, fall sharply. However, drops and fluctuations in private investment are expected to be even deeper than the ones in consumption. For example, private consumption in the United States, on the one hand, is expected to drop by −4.2% in 2020 but rise by 3.7% in 2021 (EIU, May 2020 https://country. eiu.com/united-states). Based on the same data source, private investment, on the other hand, is forecasted to drop by −12% in the United States in 2020 and fall again by −1.1% in 2021. Therefore, investment is more sensitive to declining economic conditions, and it takes longer to recover it. This fact increases the importance of rising investment for economic improvement. During economic crisis periods, the main aim of expansionary government policies, such as higher government spending, is to temporarily replace declining private consumption and private investment to keep the level of GDP as high as possible, as it can be seen in (1). With increasing consumer confidence, the first component that is expected to start to rise is private consumption. Then increasing private consumption improves demand conditions for businesses and their profit expectations. As a result, firms start to invest more to open new businesses or expand their businesses and create new jobs. These expected positive results of higher government spending sound perfect on paper, but these positive effects in reality depend on how responsive consumers and firms are to changing government spending and how long it takes to see these responses. The deeper the economic crisis, the longer it takes to see the positive impacts of higher government spending on private consumption and investment. The impact of government spending is measured by a multiplier which is a function

Does Government Spending Cause Investment?: A Panel Data Analysis

173

of marginal propensity to consume or save. For example, if consumers and firms that collected government payments and transfers do not spend these funds but save them instead, government spending cannot grow in multiple terms. If and only if consumers and firms spend the income that they received from the government, it would be a source of additional income for some other groups of people and firms. As a result, with the help of this multiple-income creation process (i.e., a multiplier higher than one), GDP can increase beyond the initial rise in government spending. The other way of saying is that a unit increase in government spending can generate more than one unit rise in GDP. In addition to this positive effect of higher government spending, there are also negative effects, named crowding-out effects of higher government spending. Increasing government spending may use resources that are taken away from the private sector. Larger budget deficits due to higher government spending can increase government debt. In order to finance these larger deficits, governments need to borrow more and more funds in financial markets. This process makes less funds available for the private sector and then increases interest rates with an increasing demand for funds by governments. Because it is not easy for the private sector to compete with governments that are mostly considered to have lower risks, firms’ cost of borrowing rises considerably with increasing interest rates. As a result, firms are expected to cut their investments or stop making investments altogether. The declining share of private investment in GDP may also lower the level of GDP. Therefore, the net effect of higher government spending on investment would be much lower than expected when the negative crowding-out effect is added to the positive multiplier effect. Another set of negative effects of higher government spending on the private sector is related to again increasing budget deficits, caused not only by increasing government spending but also by declining tax revenues or tax adjustments for a quicker economic recovery. For example, given the suddenness of the shocks, as a response to the declining economy due to the health crisis and subsequent economic lockdowns, deficits are getting larger with extremely large stimulus packages and dropping tax revenues. Governments must finance the deficits with a large amount of new government debts. Much higher public debt mean that governments may need more resources to pay them back, such as much higher taxes in the future. Therefore, higher expected taxes can put additional pressure on firms and lower their investments. This can be an additional source of negative effects of higher government expenses. Rising expected inflation rates due to large budget deficits can be another source of problems which would affect firms and their investments in a negative way. In many countries where governments’ borrowing capacity can be lower for some reasons such as high risks, they may need to print money to close their deficits. This additional money in the system, combined with higher demand for goods and services, can create perfect conditions for increasing prices and inflation. Right now, economists do not have immediate concerns about higher inflation because there are more serious problems at this stage, such as sharply declining GDP and rapidly climbing unemployment rates, which have not been seen in decades.

174

N. Bayraktar

However, ultimately when the time arrives to pay back government debts and to close government deficits, inflation issues will be discussed more. Firms are already aware of this possibility, and this may delay their current investment decisions—a single most important item to create jobs and increase GDP. Therefore, some people fear that the current economic downturn may lead to back-to-back crisis: debt and inflation crisis may follow the current economic and financial crisis. Therefore, it is critical to closely watch the price stability to evaluate the possibility of additional crisis (Mackintosh, 2020). This discussion shows that there is no clear answer on whether higher government spending is beneficial for highly anticipated improvements in private investments. Although there are other studies focusing on the effects of government spending on investment in the literature, there is no clear consensus on theorical and empirical impacts of government expenditures on private investment. Given the importance of understanding the responsiveness of investment to changing levels of government spending during crisis periods, the aim of this paper is to study the causal link between two variables in a panel data setting under different definitions of variables, time periods, time lags, and country groups. This systematic analysis would be the contribution of this paper to the literature, as many empirical studies claim that the impact of government spending on private investment is negative, and a few show a positive link (see Sen ¸ and Kaya (2014) for a detailed literature review). Some related studies are as follows. Furceri and Sousa (2011) investigate the impact of government spending on private consumption and investment. They use a panel sample of 145 countries from 1960 to 2007. Their findings indicate that there are important crowding-out effects, and government spending negatively affect both private consumption and investment. Similarly, Afonso and Jalles (2015) investigate the relevance of fiscal items for private and public investments. They use a panel dataset of 95 developed and developing countries for the period of 1970–2008. They find a negative effect of government expenditure on private investment; only government health spending has a positive and significant impact on private investment. In the literature, there are also empirical studies focusing on specific country groups. Alesina, Ardagna, Perotti, and Schiantarelli (2002) investigate a panel of OECD countries and want to understand the nature of the effects of fiscal policy on investment. Their empirical outcomes show a large negative effect of public spending on business investment. Laopodis (2001) focuses on Greece, Ireland, Portugal, and Spain, and investigate the impacts of military and nonmilitary public expenditures on gross private investment by using cointegration and error-correction analysis. The findings show that in some cases public spending stimulates investment. Afonso and Sousa (2009) show that government spending negatively affects private investment in the USA, the UK, Germany, and Italy. Similarly, Afonso and Sousa (2011) show that government spending crowds-out private investment by using a time-series data for Portugal for the period of 1979– 2007. Barro and Redlick (2011) cannot find any clear evidence of a multiplier effect in the USA, including the impact on private investment.

Does Government Spending Cause Investment?: A Panel Data Analysis

175

In the literature, not all studies support the negative link between government expenses and private investment. Positive effects of government spending on private investment can be observed in some countries, but they are mostly conditional on definitions and components of government spending. For example, Bayraktar and Moreno-Dodson (2015) show that the impact of government spending can be positive, given that expenses are productive. Their dataset covers eight fastgrowing developing countries and eight random developing countries. Similarly, Bayraktar and Fofack (2011) show that government capital spending, but not total government spending, has a positive effect on private capital accumulation in sub-Saharan countries. Bayraktar and Fofack (2018) present the importance of government spending on education investment and growth. Bayraktar and MorenoDodson (2018) study the importance of government expenses for economic growth and investment in sub-Saharan Africa. Sen ¸ and Kaya (2014) analyze empirically the impacts of government spending on private investment in Turkey for the period of 1975–2011. Their findings indicate that government current spending and interest payments negatively affect private investment, whereas government capital spending has a positive impact on investment. The contribution of my paper to these earlier studies will be that my dataset will cover earlier years as well as the years after 2008, corresponding to one of the deepest global economic crises. It is important to analyze these years because it can be seen clearer how government spending can determine investment during a global crisis period. The analysis of these years can give better idea on what should be expected with current higher government expenses. Also, another contribution of my paper to this literature is that the main interest will be the identification of the possible causal relationship between government spending and investment in different country groups based on their income levels. A larger number of countries are included in my dataset. An additional contribution of my paper is that most empirical studies take expenditure and investment data in percent of GDP in their analysis. However, with business cycle fluctuations, GDP as well as investment and government spending may all change at similar amounts. In this case, it may not be easy to capture the impacts of fluctuations in these variables when they are reported in percent of GDP. Therefore, growth rates of investment and government spending are calculated and used in my analyses, in addition to the definitions in percent of GDP, to better evaluate the possible impacts of fluctuations in government spending, especially during crisis periods.

3 Data Information The analysis presented below uses mainly two variables taken from the World Bank’s World Development Indicators and the IMF’s World Economic Outlook. The dataset covers 30 low-income countries, 59 middle-income countries, and 39 high-income countries. The list of countries is presented in the Appendix table. The period covers 1970–2018.

176

N. Bayraktar

1. Gross capital formation by the private sector consists of outlays on additions to the fixed assets of the economy plus net changes in the level of inventories. Fixed assets include land improvements (fences, ditches, drains, and so on); plant, machinery, and equipment purchases; and the construction of roads, railways, and the like, including schools, offices, hospitals, private residential dwellings, and commercial and industrial buildings. Inventories are stocks of goods held by firms to meet temporary or unexpected fluctuations in production or sales, and “work in progress.” It is in constant 2010 local currency units and also in percent of GDP. 2. General government expenditure includes all government current and capital expenditures for purchases of goods and services (including compensation of employees). It is in constant 2010 local currency units and also in percent of GDP. The government spending and investment series in growth rates and in percent of GDP are presented in Fig. 1. As explained in the literature review above, most empirical studies on the impact of government spending on investment introduce these variables in their analysis in percent of GDP. The left-hand-side panels of Fig. 1 show the variables in percent of GDP. The differences across country groups are clear. The middle-income group has the highest share of investment in GDP. It fluctuates around economic cycles, but it has an increasing trend between 1970 and 2018. The high-income group had the second highest level of investment in percent of GDP until 2008. Throughout the years, it fell from around 28% on average in 1970 to 22% in 2018. The average value of investment in percent of GDP in low-income countries was the lowest one for many years, when compared to other country groups. Although the series in the low-income group fluctuates the most, it has a clear upward trend. It increased from 13% to nearly 25% of GDP between 1970 and 2018 and has passed the share of investment in the high-income group since 2008. The lower panel on the left-hand side of Fig. 1 presents government spending in percent of GDP across different income groups. The following trends are observed. The high-income group clearly has the highest share of government spending in percent of GDP. While the share in the middle-income countries has been increasing throughout the years, the series for the low-income group presents sharp fluctuations from 1 year to another between 1970 and 2018. The right-hand-side panel in Fig. 1 shows the same series in growth rates instead. These growth rates are calculated based on the series in constant 2010 local currency units. These alternative measures in growth rates can capture fluctuations in government spending and investment better than the measures in percent of GDP. The reason is that as GDP fluctuates, the ratios of government spending and investment to GDP change as well—even if their levels are stable. Therefore, real growth rates of these two variables can give a better view of what changes are actually observed in them. When the series in percent of GDP on the left are compared to the series in growth rates on the right, it can be seen how sharp fluctuations in these variables are in growth rates. With the measures in percent of GDP, these fluctuations are mostly disguised by fluctuating GDP, as can be seen

Does Government Spending Cause Investment?: A Panel Data Analysis

177

Private investment (annual % growth)

Private investment (% of GDP) 35

30

30

25 20

25

15

20

10 5

15

1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017

0

10

-5

5

-10 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018

0

Low income

Middle income

-15 -20 Low income

High income

Government spending (% of GDP)

Middle income

High income

Government spending (annual % growth) 20

20 19 18 17 16 15 14 13 12 11 10

15 10 5

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 Low income

Middle income

High income

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018

0 -5 -10 Low income

Middle income

High income

Fig. 1 Government spending and private investment. Source The author’s calculations based on World Development Indicators and World Economic Outlook

how smooth series are on the left-hand side of Fig. 1. For each income group, the fluctuations are strong, but they are the sharpest for the low-income countries. For the high-income group, variations are relatively modest. These fluctuations in the growth rates show that if the analyses are based only on the variables in percent of GDP, a significant amount of information would be lost. Similarly, studies taking the averages over time to smooth the series also eliminate valuable shorter-term information. Another observation from Fig. 1 is that the government spending and investment series are fairly different across country income groups. Therefore, empirical studies combining countries from different income groups or focusing on only one income group may ignore important information that can be gained by comparison of groups. While changes in government spending may not have an important effect on investment in some countries, it can be more significant for a different group of countries and for different periods. Therefore, conclusions on the effectiveness of government spending for investment may require considerations of short-term fluctuations, country income groups, and different measures of government spending and investment, as well as different time periods.

4 Panel Causality Analysis There are many empirical studies indicating that government spending cannot determine investment. The purpose of this section is to show that actually this link between government spending and investment depends on which measures are used

178

N. Bayraktar

Table 1 Correlation coefficients

(in % of GDP) G(−1) and G and INV INV

(growth rate in constant LCU) G(−2) and INV

G(−1) and G and INV INV

1970–2018 Low income 0.178** 0.121* −0.027 0.255*** 0.187* * Middle income 0.131* 0.151* 0.022 0.452*** 0.408* ** High income −0.54*** 0.100 −0.231*** −0.341*** 0.007 2008–2018 Low income 0.24*** 0.718*** 0.251*** 0.141* 0.671*** Middle income −0.855*** −0.607*** −0.578*** −0.524*** 0.324*** High income −0.831*** −0.333*** −0.407*** −0.732*** −0.198**

G(−2) and INV −0.050 −0.029 −0.033 −0.090 0.081 0.459***

Note: G stands for governments pending and INV stands for private investment. (−1) stands for a variable lagged one period sand (−2) stands for a variable lagged two periods.*, **,*** stand for the level of significance at 10%, 5%, and 1%, successively

(in percent of GDP or growth rates), how countries are classified (low-, middle-, and high-income groups), which time lags are used (impact of one-period and twoperiod lags of government spending on investment), and which time periods are analyzed (between 1970 and 2018 versus during and after the 2008 global economic crisis). The analysis in this section is based on Granger causality tests for panel datasets for the classifications listed above. The aim is to understand whether government spending actually causes private spending in different income groups, in different periods, and using different measure of government spending and investment. This causality analysis can be extended to regression analysis in the future to better evaluate the impact of government spending on private investment. Before the causality test results are presented, the correlation coefficients between government spending and investment are presented in Table 1 to give initial idea on the nature of the relationship between two variables. While calculating these correlation coefficients, different lagged values of government spending are introduced to understand whether it takes time for government spending to be linked with investment. Investment is the component of GDP with largest fluctuations, especially during economic downturns. Therefore, it takes time to improve this item because firms can be hesitant to undertake expensive investments until they are more confident about expected returns from investments and wait until expected returns can cover high costs of investment. Therefore, the impact of government spending on investment may take a couple years, as well as some immediate impacts. As can be seen in Table 1, most correlation coefficients are significant at the 1% significance level because of large numbers of observations in each group. While, in the first panel of the table, the government spending and investment series are presented in percent of GDP, they are in growth rates of real values in the last

Does Government Spending Cause Investment?: A Panel Data Analysis

179

columns of the table. One common observation from the table is that the correlation coefficients are much higher on the right side, indicating that the growth rates instead of measures in percent of GDP can capture fluctuations and therefore correlations better. The focus is on two time periods. One period is between 1970 and 2018. It covers almost 50 years, and many recession and economic improvements are observed during this period. When the time period is long, estimate coefficients and correlation calculations will be based on averages across many years. Thus, we may lose significant information in this process because the link between government spending and investment may be more important in shorter time periods, especially during economic crisis. Therefore, the analysis is repeated for the years specifically corresponding to an economic downturn period (the 2008 global economic crisis) to observe changes in the relationship between government expenditure and private investment. When we check the results for the 1970–2018 period with the measures in percent of GDP in Table 1, it can be seen that the correlation between government spending (G%GDP) and investment (INV%GDP) is positive and weakly statistically significant for the low- and middle-income groups. This is true for the correlations between the same period variables as well as between the current value of INV%GDP and the lagged-one-period value of G%GDP. On the other hand, the correlation between G%GDP and INV%GDP is negative and highly significant for the high-income group. This is the negative link between government spending and investment observed in many other empirical studies. In high-income countries, governments generally start spending more money when the economy is not doing well, and consumption and investment are declining. Given the continuous importance of government spending on average in economies of the low- and middle-income groups, we do not observe this negative link between G and INV for these groups of countries. No significant correlation is observed between the one-period lagged value of G%GDP and the current value of INV%GDP in the high-income group. We observe similar correlations when the growth rates are used instead of the measures in percent of GDP. When the focus is on the period corresponding to the years after the 2018 global crisis, the correlation between government expenses and investment becomes fairly higher than the ones calculated by using observations from 50 years. Almost all correlations are statistically significant. For low-income countries, the correlation coefficients are positive and highly significant. This positive link indicates government spending is important for these countries’ investment and development. On the other hand, for middle-income and high-income countries the correlation coefficients are significantly negative for each correlation calculated with the measures in percent of GDP. Interestingly, for the same groups of countries, the correlation coefficient gets positive and significant when we consider the link between the lagged values of the growth rate of government spending and the current value of the growth rate of investment. Based on this result, it takes 2 years to see the positive effect of government spending on investment in high-income countries, while it takes 1 year to see this positive impact in the middle- and low-income countries. These positive correlations are observed may be because fluctuations are

180

N. Bayraktar

better captured by the growth rates instead of the measures in percent of GDP; we cannot see these positive effects with the measures in percent of GDP. The additional point is that in panel data settings the impact of government spending can be better evaluated when the focus is on periods where many countries have been affected at the same time, instead of focusing on a very long time period to see the link. Government spending can be more important for private spending during deeper economic crisis periods and when the impact is global. This can be one explanation why we observe higher correlations when the time period is shorter. After the analysis of the correlation coefficients, we can focus on the causal link between government spending and investment. To check whether any dual causality relationship is observed, a set of pairwise Granger panel causality tests are preformed (Granger, 1969). Table 2 reports the results for low-income countries. In the causality analysis, similar to the correlation calculations in Table 1, alternative measures of government spending and investment (in percent of GDP or in growth rates), alternative time periods (1970–2018 versus 2008–2018), and different lags are used. Because the impact of government spending on private investment is expected to be seen in a couple of years, causality tests are calculated with 1 lag, 2 lags, and 3 lags. The test results are systematically more significant when government spending and investment is measured in growth rates in the low-income group. For this set of countries, the causality moves from government spending toward private investment. The causality hypotheses moving from investment to government spending are all rejected, meaning that no causality is observed moving from investment to government spending for low-income countries. The strongest causality is observed with one lag, but the causality tests with two lags also give significant results. When we use the causality link between the growth rates of the variables, the causality test results are significant with three lags, indicating that government spending causes private investment. Table 3 shows the same set of causality test results for the middle-income group. In almost each test, government spending causes investment at high significance levels. The correlations between the variables were also high for these countries as reported in Table 1. For the middle-income group, we observe several examples of dual causality, meaning that government spending causes investment and, at the same time, investment causes government spending. The dual causality is more apparent for the period after 2008 and when the variables are measured in percent of GDP. This dual causality may lead to an endogeneity problem in regression analysis. On the other hand, when the variables are measured in real growth rates, we do not see a dual causality issue except one test result. The other interesting observation in middle-income countries is that the causality effect of government spending on investment is strongest with two lags, while the causality effect is weakest with three lags. In Table 4, the causality test results for high-income countries are reported. Many empirical studies for this group of countries fail to find any significant impact of government spending on investment. The findings indicate that, if any casualty exists, it moves from government spending toward investment, not the other way around, for this set of countries. For the 1970–2018 period and with the measures

1.757 1.101 1.413 1.573

Reject H0 Fail to reject H0

Fail to reject H0 Fail to reject H0

1.511 0.787

Fail to reject H0 Fail to reject H0 2008–2018 2.146 1.782

2.257 0.721

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

3.216 0.982

Reject H0 Fail to reject H0

G and INV (in % of GDP) Test result F-statistic 1970–2018

0.232 0.143

0.082* 0.471

0.032** 0.071*

0.191 0.723

0.022** 0.731

0.007*** 0.523

p-value of F-statistic

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

1.726 0.918

4.127 1.012

2.616 1.482

2.123 0.919

3.611 1.015

4.112 1.611

0.082* 0.551

0.001*** 0.515

0.009*** 0.201

0.032* * 0.551

0.005*** 0.511

0.001*** 0.114

G and INV (growth rate in constant LCU) Test result F-statistic p-value of F-statistic

Note: Only low-income countries are included in the analysis. There are 30 countries and the time period is 1970–2018. The list of countries is given in the Appendix. The panel dataset is unbalanced. The causality test is the one named Granger panel data test. The null hypothesis of the causality tests are given in ne first column. The dependent variables of the causality regressions are given in bold in the null hypothesis. In the first panel, G and INV are measured in % of GDP, and in the second panel G and INV are measured as growth rates of these variables in constant LCU. The test results are presented in the second and fifth columns. The test statistics and their p-values are reported in the columns number 3, 4, 6 and 7. For each test, we reject the null hypothesis if the p-value of the test is less than alpha, meaning that we observe Granger causality moving from X to Y (X Granger Causes Y) with 1-alpha probability. If we fail to reject the null hypothesis (the p-value of the test is higher than alpha), it means that this test does not confirm any causally issue moving from X to Y (X does not Granger Cause Y). The definitions of the abbreviations are as follow: G stands for government spending and INV stands for private investment. *, **, and *** stand for significance levels of 10%, 5%, and 1%, successively

Lag = 1 G does not Granger cause INV INV does not Granger cause G Lag = 2 G does not Granger cause INV INV does not Granger cause G Lag = 3 G does not Granger cause INV INV does not Granger cause G

Lag = 1 G does not Granger cause INV INV does not Granger cause G Lag = 2 G does not Granger cause INV INV does not Granger cause G Lag = 3 G does not Granger cause INV INV does not Granger cause G

Null Hypothesis (H0 )

Table 2 Low-income countries: pairwise Granger panel causality tests Does Government Spending Cause Investment?: A Panel Data Analysis 181

4.134 1781 1.699 1.891

Reject H0 Reject H0

Reject H0 Reject H0

1.211 0.891

Fail to reject H0 Fail to reject H0 2008–2018 3.134 2.153

1.861 1.511

Reject H0 Fail to reject H0

Reject H0 Reject H0

1.961 1340

Reject H0 Fail to reject H0

Gand INV (in % of GDP) Test result F-statistic 1970–2018

0.091* 0.062*

0.001* ** 0.07*

0.007* ** 0.03**

0.421 0.611

0.06* 0.191

0.052* 0.313

p-value of F-statistic

Fail to reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Reject H0

Reject H0 Fail to reject H0

1511 1.311

3.987 1.342

3.563 1.021

1.715 1.320

4.132 1714

2.080 1.411

0.191 0.331

0002* ** 0.311

0.005* ** 0.501

0.091* 0.321

0.001* ** 0.09*

0.042** 0.233

G and INV (growth rate in constant LCU) Test result F-statistic p-value of F-statistic

Note: My middle-income countries are included in the analysis. There are 59 countries and the time period is 1970–2018. The list of countries is given in the Appendix. The panel dataset is unbalanced. The causality test is the one named Granger panel data test. The null hypothesis of the causality tests are given in the first column. The dependent variables of the causality regressions are given in bold in the null hypothesis. In the first panel, G and INV are measured in % of GDP, and in the second panel, G and INV are measured as growth rates of these variables in constant LCU. The test results are presented in the second and fifth columns. The test statistics and their p-values are reported in the columns number 3, 4, 6 and 7. For each test, we reject the null hypothesis if the p-value of the test is less than alpha, meaning that we observe Granger causality moving from X to Y (X Granger Causes Y) with 1-alpha probability. If we fail to reject the null hypothesis (the p-value of the test is higher than alpha), it means that this test does not confirm any causality issue moving from X to Y (X does not Granger Cause Y). The definitions of the abbreviations are as follow: G stands for government spending and INV stands for private investment. *, **, and *** stand for significance levels of 10%, 5%, and 1% successively

Lag = 1 G does not Granger cause INV INV does not Granger cause G Lag = 2 G does not Granger cause INV INV does not Granger cause G Lag = 3 G does not Granger cause INV INV does not Granger cause G

Lag = 1 G does not Granger cause INV INV does not Granger cause G Lag = 2 G does not Granger cause INV INV does not Granger cause G Lag = 3 G does not Granger cause INV INV does not Granger cause G

Null Hypothesis (H0 )

Table 3 Middle-income countries: pairwise Granger panel causality tests 182 N. Bayraktar

1.976 1.034 1.741 1.432

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

1.713 1.556

Reject H0 Fail to reject H0 2008–2018 2.123 1.543

1.612 1.233

Fail to reject H0 Fail to reject H0

Reject H0 Fail to reject H0

1.511 1.211

Fail to reject H0 Fail to reject H0

G and INV(in %of GDP) Test result F-statistic 1970–2018

0.08* 0.212

0.051* 0.499

0.032** 0.161

0.09* 0.151

0.112 0.411

0.191 0.421

p-value of F-statistic

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Fail to reject H0 Fail to reject H0

Reject H0 Fail to reject H0

Reject H0 Fail to reject H0

2.011 1.651

3.751 1.211

4.110 1.313

1.613 1.656

1.912 1.013

1.814 1.612

0.04** 0.101

0.003*** 0.421

0.001*** 0.325

0.112 0.101

0.052* 0.513

0.061* 0.112

G and INV (growth rate in constant LCU) Test result F-statistic p-value of F-statistic

Note: Only high-income countries are included in the analysis. There are 39 countries and the time period is 1970–2018. The list of countries is given in the Appendix. The panel dataset is unbalanced. The causality test is the one named Granger panel data test. The null hypothesis of the causality tests are given in the first column. The dependent variables of the causality regressions are given in bold in the null hypothesis. In the first panel, G and INV are measured in % of GFP, and in the second panel, G and INV are measured as growth rates of these variables in constant LCU. The test results are presented in the second and fifth columns. The test statistics and their p-values are reported in the columns number 3, 4, 6 and 7. For each test, we reject the null hypothesis if the p-value of the test is less than alpha, meaning that we observe Granger causality moving form X to Y (X Granger Causes Y) with 1-alpha probability. If we fail to reject the null hypothesis (the p-value of the test is higher than alpha), it means that this test does not confirm any causality issue moving from X to Y (X does not Granger Cause Y). The definitions of the abbreviations are as follow: G stands for government spending and INV stands for private investment. *, **, and *** stand for significance levels of 10%, 5%, and 1%, successively

Lag = 1 G does not Granger cause INV INV does not Granger cause G Lag = 2 G does not Granger cause INV INV does not Granger cause G Lag = 3 G does not Granger cause INV INV does not Granger cause G

Lag = 1 G does not Granger cause INV INV does not Granger cause G Lag = 2 G does not Granger cause INV INV does not Granger cause G Lag = 3 G does not Granger cause INV INV does not Granger cause G

Null Hypothesis (H0 )

Table 4 High-income countries: pairwise Granger panel causality tests Does Government Spending Cause Investment?: A Panel Data Analysis 183

184

N. Bayraktar

in percent of GDP, we do not observe any causality moving from government spending toward investment except with three lags. On the other hand, we observe government spending causes investment with one lag and two lags when the variables are measured in real growth rates. The growth rate measures can capture the fluctuations better, therefore it might be easier to see a causal relationship. It is a totally different story when the focus is on the period during and after the global crisis of 2008. In each lag specifications and measurements of government spending and investments, government spending Granger causes investments. When government expenses are most needed (i.e., during global financial downturns), the importance of government spending and its causal links toward investments become more significant. Similar to the analysis with the larger period, we do not observe any causality moving from private investment toward government spending in the high-income group. Overall, the link between government spending and investment depends on many factors even in the simplest sets of analysis. The causality and correlation results change from one group of countries to the other; from one measure of government spending to the other; and from one time period to another (existence of global crisis and the depth of crisis can make a difference). The link will also depend on the time lags. Therefore, the link between government spending and investment is too complex to make general conclusions with one-dimensional analysis.

5 Conclusions and Policy Recommendations One of the biggest roles during economic crisis is played by governments, and they generally start to follow easy fiscal and monetary policies. A typical prescription advises them to spend more money, lower taxes, and increase money supply. Especially, during severe and unexpected economic crisis, as it is happening during the current health crisis, it gets extremely difficult to follow targeted spending. Given the emergency of the situation, governments may end up spending huge amount of money without carefully considering where it goes and what might be the ultimate impact. However, this study shows that the high multiplier effect is possible, and still government spending can improve consumption, production, and ultimately investment in most cases with or without time lags. While evaluating the effect of government spending, we should consider the income level of countries, how to measure the variables, which periods are analyzed and how long it may take to see the impacts. Given the basic results presented in this paper, it is not totally accurate to identify government spending as insignificant for private investment. The answer depends, and the link might be stronger during severe crisis. The study in the future can be extended to consider how regression results may change based on classifications used in this paper and by taking into account different sets of control variables to better understand whether public spending is really significant for private investment.

Does Government Spending Cause Investment?: A Panel Data Analysis

185

A.1 Appendix: List of Countries

Low income Afghanistan Benin Burkina Faso Burundi Central Afr. Rep. Chad Congo, Dem. Rep Eritrea Ethiopia Gambia Guinea Guinea-Bissau Haiti Liberia Madagascar Malawi Mali Mozambique Nepal Niger Rwanda Sierra Leone Somalia South Sudan Syria Tajikistan Tanzania Togo Uganda Yemen

Middle income Albania Algeria Angola Argentina Armenia Azerbaijan Bangladesh Bolivia Botswana Brazil Bulgaria Cabo Verde Cambodia Cameroon China Colombia Comoros Congo, Rep. Costa Rica Cote d’Ivoire Dominican Rep. Ecuador Egypt El Salvador Equatorial Guinea Fiji Gabon Georgia India Indonesia

Jordan Kenya Malaysia Mauritania Mexico Moldova Morocco Myanmar Namibia Nicaragua Nigeria Pakistan Paraguay Peru Philippines Romania Russia Senegal South Africa Sri Lanka Sudan Thailand Tunisia Turkey Ukraine Uzbekistan Vietnam Zambia Zimbabwe

High income Australia Austria Belgium Canada Chile Croatia Cyprus Czech Rep. Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Israel Italy Japan Korea, South Kuwait Latvia Lithuania Netherlands New Zealand Norway Panama Poland Portugal

Singapore Slovak Republic Slovenia Spain Sweden Switzer land United Kingdom United States Uruguay

References Afonso, A., & Jalles, J. T. (2015). How does fiscal policy affect investment? Evidence from a large panel. International Journal of Finance & Economics, 20(4), 310–327. Afonso, A., & Sousa, R. M. (2009). The macroeconomic effects of fiscal policy. European Central Bank Working Paper 991.

186

N. Bayraktar

Afonso, A., & Sousa, R. M. (2011). The macroeconomic effects of fiscal policy in Portugal: A Bayesian SVAR analysis. Portuguese Economic Journal, 10(1), 61–82. Alesina, A., Ardagna, S., Perotti, R., & Schiantarelli, F. (2002). Fiscal policy, profits, and investment. American Economic Review, 92(3), 571–589. Barro, R. J., & Redlick, C. J. (2011). Macroeconomic effects from government purchases and taxes. The Quarterly Journal of Economics, 126(1), 51–102. Bayraktar, N., & Fofack, H. (2011). Capital accumulation in sub-Saharan Africa: Income-group and sector differences. Journal of African Economies, 20(4), 531–561. Bayraktar, N., & Fofack, H. (2018). A model for gender analysis with informal production and financial sectors. Journal of African Development, 20(2), 1–20. Bayraktar, N., & Moreno-Dodson, B. (2015). How can public spending help you grow? An empirical analysis for developing countries. Bulletin of Economic Research, 67(1), 30–64. Bayraktar, N., & Moreno-Dodson, B. (2018). Public expenditures and growth in a Monetary Union: The case of WAEMU. Eurasian Journal of Economics and Finance, 6(1), 107–132. Furceri, D., & Sousa, R. M. (2011). The impact of government spending on the private sector: Crowding-out versus crowding-in effects. Kyklos, 64(4), 516–533. Granger, C. W. J. (1969, July). Investigating causal relations by econometric models and crossspectral methods. Econometrica, 37, 424–38. Laopodis, N. T. (2001). Effects of government spending on private investment. Applied Economics, 33(12), 1563–1577. Mackintosh, J. (2020, March 29). After Coronavirus, we will have to reckon with the debt. The Wall Street Journal. Retrieved from https://www.wsj.com/articles/after-coronavirus-we-willhave-to-reckon-with-the-debt-11585494002 Sen, ¸ H., & Kaya, A. (2014). Crowding-out or crowding-in? Analyzing the effects of government spending on private investment in Turkey. Panoeconomicus, 61(6), 631–651.

An Exploratory Study of Fans’ Motivation in Albanian Football Championship Julian Bundo and Mirdaim Axhami

Abstract Football continues to excite and attract more and more people to most countries. This phenomenon that is now bypassing all the borders of the world has been starting up in Albania in the last few years. As one of the most attractive sports which have long roots in Albanian history, football is trying to give important signals, both in terms of the game and for the participation of the fans attending the events in the stadium. Based on the overview of the football championship, we verified that more and more fans particularly female and youngest people are heading to the stadiums to enjoy the football game. With the beginning of the twenty-first century, FIFA and EUFA have changed their program, giving particular importance to increase participation in football activities regardless of gender and age. Even football fans like in other sports are influenced by their motivation which leads them to behave or act in different ways. Precisely, the study aims to examine football fans’ motivation factors of attendance using a nation-wide representative sample (N = 768). Using exploratory factor analysis, a group of motivation factors is analyzed, testing constructs that explain attendance. The study identifies three main factors affecting attendance of Albanian football fans, considered as entertainment, tradition, and group involvement. Entertainment is identified as the most important factor. MANOVA was later used to identify differences in motivation factors between age groups and participation in an organized fan group (support group/ultras). Keywords Motivational factors · Football fan attendance · Albania championship

J. Bundo () · M. Axhami Faculty of Economy, Department of Marketing and Tourism, University of Tirana, Tirana, Albania © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_14

187

188

J. Bundo and M. Axhami

1 Introduction In the last decade of the twentieth century, the first changes for football clubs in terms of marketing took place, implementing their marketing plans and strategies, thus developing programs that would increase the value of their offer for satisfy their own fans and attracting new ones (Beccarini & Ferrand, 2007). In this context, the sport marketing industry has been grown significantly, and from year to year her image around the world has created all the ingredients to increase the participation for the new generation on these events (Ramadhan, Ramadhan, Yusuf, Yahya, & Febriani, 2019). Very quickly the football marketers began to apply Kotler principles; according to him “achieving organizational objective depends on determining the needs and wants for target market, and delivering all the satisfactions and desired more efficiently and effectively than competitors do” (Beccarini & Ferrand, 2007). These changes created all the possibilities until football marketers understood the relationship that football fans are looking for by offering them the right stimuli from the event’s experience (Draper, 2002, p. 13). According to Know and Trail (2001) fan motivation is therefore defined as a set of attitudes and purchases toward players and teams. The importance that football has ensured in the various countries of the world by motivating more and more people has not left indifferent even the fans of this sport in Albania. As the most aimed sport, football takes place every year in Albania organized in two competitions: National and European football Championships. In the National Football Championship, “Kategoria Superiore” is the elite division, composed from 10 clubs, with 36 matches in total (teams meet each other four times during the season, in four stages). In Albanian Cup, football clubs from every division are involved, in a play-off type of competition. Both competitions start at the beginning of September till on the last week of May, with different agenda and organization. According to UEFA Country Ranking Albania is placed on the 36th position among 55 countries. In addition to internal competitions, there are two more competitions organized annually by UEFA (Union of European Football Associations), which are as follows: All the best teams from each European country are involved in these competitions. The winner clubs of UCL automatically participate on the competitions for the International Cup for clubs held generally on November every year. There are two more quadrennial competitions held every 4 years with 2 years apart from each other: FIFA World Cup is the most iconic football event in the world, 32 national teams from all over the world are involved in this competition (France won the last Cup at “Russia 2018”). UEFA Europe Cup is the most important football event for the old continent, and 24 European national teams are involved in this competition (Portugal won the last Cup at “EURO 2016” in France). These events are held during the June–July of every calendar, attracting the interest of hundreds of millions of fans and spectators. This huge organization and consumption worldwide create the

An Exploratory Study of Fans’ Motivation in Albanian Football Championship

189

best conditions for marketers, to analyze and evaluate in Albania, the motives and reasons among their fans for football attendance. From this point of view, we will try to understand how the variables like gender, age, and club influence the motivation and fans’ behavior.

2 Literature Review Generally being motivated means doing something that feeling inspirited to act (Ryan & Deci, 2000). Basically, motivation is an important tool for understanding consumer behavior (Shank, 1999). For James and Ridinger (2002), a significant number of people who attend sport events consider themselves sport fans. According to Agas, Georgakarakou, Mylonakis, and Panagiotis (2012), fans means expresses enthusiasm, passion, and eagerness and even moving beyond reason. Fans are a very important patrimony for all sports, for their support, energy, and emotional enthusiasm toward players and clubs through their cheering and singing (Wiid & Cant, 2015). Motivations generate behavior due to the satisfaction or enjoyment generated during the events or game in the stadium (Dieci, 1971). For Sloan (1989), most of fan behaviors fulfill social or psychological needs. Several models and scales for measuring fan’s motivations in attending sport activities in the stadium have been developed (Correia & Esteves, 2007; Hunt, Bristol, & Bashaw, 1999; Izzo, Walker, Munteanu, Piotrowski, & Neulinger, 2014; Mehus, 2005; Wann, 1995). Based on the use of different scales and studies by researchers, it is constantly understood that the motivations are very dynamic, even their characteristics are traceable by the motivations of the fans (Kim & Chalip, 2004) . . . . . . For Dunning (1999) fans had the pleasure to follow sports events not only to maintain social contact with the rest of the fans but also for the emotions they experience through sports. Based on the study conducted by Mehus (2005), motivation for euphoria was more important than social one for fans. The results of study by Correia and Esteves (2007) for fans in Portugal suggest that connection with their clubs’ entertainment and being in the group were the strongest motivation factors. The study for Korea and Japan League, respectively, suggests that both countries were very motivated to attend sports events in group, where Japanese fans preferred to be accompanied with family members and Korean fans were motivated to be with friends (Won & Kitamura, 2007). In most of Eastern countries, entertainment and socializing influence motivate fans to attend football game in stadium, even attendance in football matches is lower due the fact that structure is less developed compared with rest of the Western countries (Izzo et al., 2014). As the most used and common motivation scale deriving their conclusion from a decade-long series of studies, Wann, Melnick, Russel, and Pease (2001) identified eight motives for attending sporting events through Sport Fan Motivation Scale. The SFMS has been used in one previous study for examining the impact of intrinsic and extrinsic motivation factors among football fans in Albania. For the purpose of our study how variables like gender, age, and club influence fan motivation will be based on SFMS

190

J. Bundo and M. Axhami

model composed of eight factors (Beccarini & Ferrand, 2007): 1—eustress (positive and stimulative), 2—self-esteem enhancement, 3—escape, 4—entertainment, 5— economic motivation, 6—esthetics, 7—group affiliation (need for belongingness), and 8—family socialization. Sport marketers and researchers must evaluate carefully the important factors that influence fans to attend sport events. However, it is not easy to understand the factors of diverse attendance, for the fact that people’s behavior is not measured by a single motive or factor, even they have different profiles (Mashiach, 1980). Gender differences in sports have attracted the attention of various scholars from different fields including law, economics, history, sports sciences, psychology, communications, and sociology, due to the fact that it is practiced almost in all societies (Deaner, Balish, & Lombardo, 2015). As the number of sports that see woman as protagonists has growing rapidly, their fan base is growing constantly using sport also like a tool for supporting women rights. According to Dieztz-Uhler (2002), both men and women sports fans are equal, which does not mean that both of them have the same reasons for attending events or game in the stadium. Meanwhile both of them aim to participate in sports events and have different individual motives, which means that men are motivated to show that they have better ability than other men, while woman are motivated to participate in sports to improve their looks (Apostolou, 2015). According to, men and women have very different attitude and mental system, that is why men are more interested towards the game itself, instead women are more inclined towards family occasion, and they consider the events as their educational leisure (Correia & Esteves, 2007). Scholars suggests that women participate less in sports than men because they have less time and are more involved with housework compared to men (Gantz & Wenner, 1991). However, the differences based on the sexual differences for the motivations and participation in football matches depend also from the role that women and men have in their society. The characteristics on motivations for football fans based on age differences in football events take full life on the 16 and 18 years. Over the age of 18, every young person is free to think and act in full autonomy up to the age of 65, and once they retire the desire to be active in sport events begins to be less active with the decrease in their incomes (Correia & Esteves, 2007). Both men and women tend to participate more frequently in football events when they are young, at the same time they participate less when they get older (Apostolou, 2015). For clubs and marketers, it is very important that the age group between 18 and 65 years should be motivated to attend the game every week in the stadium.

3 Methodology The conduction of questionnaire was distributed across 11 football matches, to ensure the representation of all football fans. As observed, typically, fans take their seat on the stadium about 30 min before the match (on average). The conduction of

An Exploratory Study of Fans’ Motivation in Albanian Football Championship

191

the questionnaire started at about 40 min before the match. Based on the literature, questions capturing motivation factors for attendance were composed in a five-point Likert scale (1 = not important to 5 = very important). Pretesting and piloting were conducted to validate the relevance and full understanding of the motives included. The pretesting was conducted with university students who were avid football fans. The second pretesting was conducted during a championship match. Although no important changes were made, small changes and fine tunings resulted in an improved questionnaire. Finally, the questionnaire was piloted during two championship matches, after which it was validated and ready for conduction. A total of 768 interviews were conducted. The composition of the sample is highly influenced by the actual attendance in Albanian stadiums. About 96% of the sample is composed of male football fans, while only 4% of female football fans. To present a better picture of the situation, 22 out of 36 women interviewed were in Tirana, the capital city. The absence of women attendance in Albanian stadiums is much due to the primitivity and vulgarity still occurring, and on the other hand to the marginalized role of women in the Albanian society, which is mostly manifested in the regions rather than in the capital city. However, this stands as a general observation, an assumption, as there is no particular study to explore this issue. On the other hand, the sample covers all age categories: 18–24 years old (27%), 25– 34 years old (24%), 35–44 years old (12%), 45–54 years old (16%), 55–64 years old (16%), 65 years old, and over (5%). The purpose of the analysis is to identify key constructs of the motivation factors that affect the attendance using exploratory factor analysis. Exploratory factor analysis is generally used to assist in the development of measurement instruments by determining the dimensionality of a set of measured variables and to determine the specific measured variables that best reflect the conceptual dimensions underlying the set of measured variables (Fabrigar & Wegener, 2011). To conduct the analysis, principal axis factoring estimation is used. Multivariate analysis of variance (MANOVA) was used to explore whether there are differences in the motivation factors between age groups and between organized football fans (ultras) and non-organized football fans. According to Hair, Black, Babin, Anderson, and Tatham (2018), if the ratio of the largest groups is more than 1.5, a test for equality of variance is in precondition to the analysis. Conducting the Box test, the equality of variance was concluded for both groups. The literature suggests differences in motivation factors even between gender; however, the sample size of only 4% female does not leave much room for a multivariate analysis of variance.

4 Results Table 1 displays the sample’s descriptive statistics (mean and standard deviation) of each of the original reasons that explain the motivation to attend a football match. These values provide an initial sense of the primary trends that might be underlined in this data. The highest value is “to closely support the team” (4.63),

192 Table 1 Descriptive statistics

J. Bundo and M. Axhami

To closely support the team To have fun To be part of a group To break away from the daily routine To be with the family in the stadium To be with friends in the stadium To closely see the football players To express my emotions Is the most followed sport It is a family tradition I have played football To dress in our team’s uniform

Mean 4.63 4.62 2.64 4.07 2.44 4.44 3.61 4.22 4.59 3.44 3.65 3.59

Std. dev 0.87 0.83 1.69 1.25 1.64 1.02 1.53 1.23 0.91 1.63 1.53 1.57

closely followed by “to have fun” (4.62), and “it’s the most followed sport” (4.59). The reasons with the highest means appear quite diverse, as the first one purely relates to fan’s presence to give support, the second one reveals the pure enjoyment of watching a football match, while the third the following of a popular trend. The lowest values are “to be with the family in the stadium” (2.44), closely followed by “to be part of a group” (2.64). The two reasons with the lowest means offer an individualistic taste of the football fans. But looking further, the reason “to be with the friend in the stadium” (4.44) shows a significantly high mean, which inclines the motivation toward a certain group. To determine the number of common factors, Kaiser’s criteria (eigenvalue above 1) and Scree test are employed. The appropriate number of factors corresponds to the number of eigenvalues prior to the last major drop in the plot of eigenvalues from the reduced correlation matrix (Fabrigar & Wegener, 2011). Basically, both approaches boil down to choosing components that have eigenvalues greater than 1. Examining Fig. 1, the Scree plot, and produced eigenvalues, it shows a departure from linearity with a three-factor result. Hence, the Scree test results that three factors should be analyzed. The KMO values is 0.80, above the cut-off point of 0.5, showing sampling adequacy and suitability for EFA. Bartlett’s Test of Sphericity shows significance at p < 0.01, which confirms that our sample has patterned relationships. Using varimax rotation method to provide easily interpretable results, Table 3 shows the rotated factor loadings. The pattern matrix contains coefficients seen as standardized regression coefficients, in which common factors predict measured variables. The results presented in Table 3 show the existence of three factors. The first factor contains six different measured variables, the second and the third contain three each. Considering the methodological notes of Tabachnick, Fidell, and Ullman (2007), as a general rule of thumb, using an alpha level of 1%, a rotated factor loading for a sample size of at least 300 would need to be at least 0.32 to be considered statistically meaningful, while lower values might also be considered for much larger samples. The sample size of the study is 768, which makes a strong

An Exploratory Study of Fans’ Motivation in Albanian Football Championship

193

Fig. 1 Scree test criterion

case for the statistic validity of even factor loadings as “to break away from the daily routine” (0.301), or “to be part of a group” (0.313). Looking further into the factor loading, cross loadings do not constitute of a problem for the results. The only cross loading is in the variable “to closely see the football players,” which mostly loads at the third factor but has a significant load also on the first factor. The difference between the primary and the secondary loading is lower than 0.1, which enables the loading to be interpreted in both factors. However, considering the large nature of the first factor, the factor loading will only be interpreted for the third factor, putting also caution to the interpretation. Analyzing the results, while considering the literature about the motivation factors, as well as “reading between the lines,” the factors are named as follows: (1) entertainment motivation factor; (2) tradition motivation factor; (3) group involvement motivation factors. The entertainment motivation factor, relating to Factor 1, is composed of the pure enjoyment, emotion, and satisfaction of watching a football match, involving their favorite team. The tradition motivation factor relates to the long history of football and the continuous impact it has had in generations, which are still being passed on other generation—covered with a glimpse of nostalgia about the child memory of playing football and wanting to represent your favorite team. The group involvement motivation factor relates to football as a shared interest within a group. The factor which seems to prevail is the entertainment motivation factor. Out of the motivation sources pertaining to Factor 1, “to have fun” shows the highest coefficient, hence showing the strongest correlation. Other factor loadings with strong correlations to Factor 1 are “is the most followed sport,” “to express my emotions” as well as “to closely support the team.” On the other hand, the lowest

194

J. Bundo and M. Axhami

Table 2 Rotated factor loading matrix To closely support the team To have fun To break away from the daily routine To be with friends in the stadium To express my emotions Is the most followed sport It is a family tradition I have played football To dress in our team’s uniform To be part of a group To be with the family in the stadium To closely see the football players

Factor 1 0.530 0.638 0.301 0.481 0.554 0.590 0.156 0.107 0.237 0.080 0.035 0.319

Factor 2 0.144 0.076 0.049 0.138 0.072 0.183 0.508 0.646 0.454 0.120 0.078 0.126

Factor 3 −0.007 0.045 0.157 0.165 0.264 0.014 0.241 0.034 0.329 0.313 0.606 0.382

Table 3 Univariate comparisons of motivation factors by age groups Motivational type means 18– 24 years Overall old 25–34 Entertainment 4.42 4.34 4.29 Tradition 3.56 3.64 3.41 Group 2.89 3.07 2.92 involvement

35–44 4.50 3.47 3.06

45–54 4.59 3.59 2.77

55–64 4.50 3.70 2.62

65+ 4.51 3.45 2.63

F 4.505 1.330 3.877

p 0.000 0.249 0.002

coefficient is that of “to break away from the daily routine,” hence showing the weakest correlation. Out of the motivation sources pertaining to Factor 2, “I have played football” shows the strongest correlation, supporting the case of nostalgia and self-affiliation with the played sport as a strong motivator for attendance. Out of the motivation sources pertaining to Factor 3, “to be with the family in the stadium” shows the strongest correlation, showing a family centric and group approach motivation for attendance in a football match (Table 2). Further, MANOVA was used to determine whether the motivation factors differ for age groups. The MANOVA analysis showed strong statistical significance (Wilks’ lambda 0.921, F 4.229) at the .00 level on all criteria, indicating that the motivation factors differ across age groups. Hence, a univariate ANOVA was used to further explore the differences. The results are displayed in Table 3. The results show that age differences are significant in entertainment motivation factors (p < 0.05) and group involvement motivation factors (p < 0.05), while no significant differences are found in tradition motivation factors (p > 0.05). Looking further, the age group of 45–54 years old appear more inclined to entertainment motivation factors, while the age group of 18–24 years old appear less inclined. On the other hand, the age group of 18–24 years old appear more inclined to group involvement motivation factors, while the oldest group (65+ years old) appear less inclined.

An Exploratory Study of Fans’ Motivation in Albanian Football Championship

195

Table 4 Univariate comparisons of motivation factors by participation in an organized fan club

Entertainment Tradition Group involvement

Motivational type means Overall Part of a group 4.42 4.4233 3.56 3.93 2.89 3.45

Not part of a group 4.4239 3.44 2.72

F 0.000 24.597 67.148

p 0.991 0.000 0.000

MANOVA was used to determine whether the motivation factors differ between football fans which are part of an organized fan club (supporters’ group or ultras). Out of 768 interviews, 186 (32%) declared themselves to be part of an organized fan group. The MANOVA analysis showed strong statistical significance (Wilks’ lambda 0.897, F 29.116) at the .00 level on all criteria, indicating that the motivation factors differ between organized and non-organized football fans. A univariate ANOVA was again used to further explore the differences. The results show that differences in motivation factors between fans that are part of an organized group and those who are not are most significant in group involvement motivation factors (p < 0.05) and tradition motivation factors (p < 0.05), while no significant differences are found in entertainment motivation factors (p > 0.05)—quite obvious given the very similar means. Considering differences, organized football fans are more motivated by tradition and by group involvement factors, while in the second case the difference is considerable. In both cases, the results are aligned with the expectations—football fans become part of a group to be jointly involved in supporting the team and on the other hand these groups are known for their tradition nuances (Table 4).

5 Conclusions The analysis of the motivation factors of attendance in a football match is crucial to this stage of development of Albanian football. The EFA resulted in 12 items measuring three motivation factors. The entertainment motivation factor is the most important one to explain attendance in a football match. This shows that Albanian football fans primarily attend a football match to purely entertain themselves, express emotion and satisfaction of watching a football match, involving their favorite team. Entertainment stands out as the usual motivation for attending most sport activities, besides the results show that two more other factors influence attendance. The tradition motivation factor blends the passing through generations of this tradition and the nostalgia of fans playing themselves. The group involvement stands also as a motivation factor for attendance, confirming football as a shared interest within a group. The analysis considered also differences in motivation factors between age groups and participation or not in an organized fan group. The results suggest that there are significant differences in motivation factors, both

196

J. Bundo and M. Axhami

between age and participation in a fan group. Some older groups (45–54 years old) are more motivated to attend matches for entertainment, while other younger age groups (18–24 years old) are more motivated to attend matches for group involvement reasons. On the other hand, entertainment is equally perceived by organized fans and non-organized ones, but organized fans were more motivated to attend for traditional and group involvement factors.

References Agas, K., Georgakarakou, C., Mylonakis, J., & Panagiotis, A. (2012). Traveling abroad internal and external motives towards different sports fan types. International Journal of Business and Management, 7(3), 111–126. Apostolou, M. (2015). The evolution of sports: Age-cohort effects in sport participation. International Journal of Sport and Exercise Psychology, 13(4), 359–370. Beccarini, C., & Ferrand, A. (2007). Factors affecting soccer club season ticket holders’ satisfaction: The influence of club image and fans’ motives. European Sport Management Quarterly, 6(1), 1–22. Correia, A., & Esteves, S. (2007). An explanatory study of spectators’ motivation in football. International Journal of Management and Marketing, 2(5/6), 572–590. Deaner, R. O., Balish, S. H. M., & Lombardo, M. P. (2015). Sex differences in sport interest and motivation: An evolutionary perspective. Evolutionary Behavioral Sciences, 10(2), 73–97. Dieci, E. L. (1971). Effects of externally mediated rewards on intrinsic motivation. Journal of Personality and Social Psychology, 18(1), 105–115. Dietz-Uhler, B., End, C. M., Harrick, E. A., & Jacquemotte, L. (2002). Identifying with winners: A reexamination of sport fans’ tendency to BIRG. Journal of Applied Social Psychology, 32(5), 1017–1030. Draper, P. (2002). Marketing brands to fan. Brand Strategy, 162, 13–14. Dunning, E. (1999). Sport matters: Sociology studies of sport. Violence and civilization. London and New York, NY: Routledge. Fabrigar, L. R., & Wegener, D. T. (2011). Exploratory factor analysis. New York: Oxford University Press. Gantz, W., & Wenner, L. A. (1991). Men, women, and sports: Audience experiences and effects. Journal of Broadcasting & Electronic Media, 35(2), 233–243. Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2018). Multivariate data analysis (8th ed.). Upper Saddle River, NJ: Prentice Hall. Hunt, K. A., Bristol, T., & Bashaw, R. E. (1999). A conceptual approach to classifying sports fans. Journal of Services Marketing, 13(6), 439–452. Izzo, M. G., Walker, W. J., Munteanu, C., Piotrowski, P., & Neulinger, A. (2014). Exploring soccer fans’ consumer motives in four Eastern European countries. Journal of International Business Research, 13(2), 50–65. James, J., & Ridinger, L. (2002). Female and male sport fans. A comparison of sport consumption motives. Journal of Sport Behavior, 25(3), 1–19. Kim, M. S., & Chalip, L. (2004). Why travel to the FIFA World Cup? Effects on motives, background, interest and constraints. Tourism Management, 25(6), 695–707. Know, H., & Trail, G. (2001). Sport fan motivation: A comparison of American students and international students. Sport Marketing Quarterly, 10(2), 147–155. Mashiach, A. (1980). A study to determine the factors which influence American spectators to see the Sumer Olympics in Montreal, 1976. Journal of Sport Behavior, 3(1), 16–28. Mehus, I. (2005). Sociability and excitement motives of spectators attending entertainment sport events: Spectators of soccer and ski-jumping. Journal of Sport Behaviour, 28(4), 333–350.

An Exploratory Study of Fans’ Motivation in Albanian Football Championship

197

Ramadhan, D., Ramadhan, S., Yusuf, A., Yahya, M., & Febriani, F. (2019). Factors affecting attendee’s motivation to come to an international sport events. Retrieved from https:// www.researchgate.net/publication/330511983_FACTORS_AFFECTING_ATTENDEE’S_ MOTIVATION_TO_COME_TO_AN_INTERNATIONAL_SPORT_EVENT Ryan, R. M., & Deci, E. L. (2000). Intrinsic and extrinsic motivations: Classic definitions and new directions. Contemporary Educational Psychology, 25(1), 54–67. Shank, M. D. (1999). Sport marketing: A strategic perspective. Upper Saddle River, NJ: Prentice Hall, Inc.. Sloan, L. R. (1989). The motives of sports fans. In J. H. Goldstein (Ed.), Sports, games and play: Social and psychology viewpoints (2nd ed., pp. 175–240). Hillsdale, NJ: Erlbaum Associates. Tabachnick, B. G., Fidell, L. S., & Ullman, J. B. (2007). Using multivariate statistics (Vol. 5). Boston, MA: Pearson. Wann, D. L. (1995). Preliminary validation of the sport fan motivation scale. Journal of Sport and Social Issues, 19(4), 377–396. Wann, D. L., Melnick, M. J., Russel, G. W., & Pease, D. G. (2001). Sport fans: The psychology and social impact of spectators. New York: Routledge. Wiid J. A., & Cant, M.C. (2015). Sport fan motivation: Are you going to the game? International Journal of Academic Research in Business and Social Sciences, 5(1), 383–398. Won, J., & Kitamura, K. (2007). Comparative analysis of sport consumer motivations between South Korea and Japan. Sport Marketing Quarterly, 16(2), 93–105.

Evaluation of Knowledge in Accounting of Regional Economic University Students Ivana Koštuˇríková and Markéta Šeligová

Abstract At present, education is considered not only as an integral part of everyone’s life but also as an important element of their personal development. Therefore, most young people in developed countries seek to acquire professional knowledge by graduating from university. The aim of this paper is to present the results of the research focused on the basic knowledge of accounting issues of regional economic university students and their dependence on aspects of university education. Research shows that almost one-third of students achieved very good knowledge and 43% of students showed good accounting skills. Only 9% of students had insufficient knowledge of accounting issues. Using the Kruskal–Wallis test, the dependence of the accounting knowledge of SBA students on aspects of university study was researched. Dependence on two aspects, the study field and students’ work experience, has been demonstrated. This was confirmed by Pearson’s Chi-square test. The intensity of dependence was then researched using Pearson contingency coefficient. Based on the resulting values, it was possible to conclude that this is not a close dependence. Keywords Education · University studies · Work experience · Kruskal · Wallis test · Pearson Chi-square test · Pearson coefficient JEL Classification: I21, M41

I. Koštuˇríková () · M. Šeligová Department of Finance and Accounting, School of Business Administration in Karviná, Silesian University in Opava, Karviná, Czech Republic e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_15

199

200

I. Koštuˇríková and M. Šeligová

1 Introduction The importance of education is constantly increasing. The educational level of the population is increasingly influencing the prosperity and economic growth of individual states. Education including lifelong learning is closely related to the issue of competitiveness of companies and regions. In this context, professional education as a system for collecting skills and competencies has an essential role to respond to the rapidly changing labor market needs, modernizing and upgrading industries and the development of the structure of job opportunities. Gradual globalization, technical progress, innovation, and also changing social values, the diversification of human resources, or the use of the natural environment, among others, are the driving forces of the constantly changing business environment. Education develops the personality of a person who will be equipped with knowledge, skills, and competences not only for personal life but also for the performance of a profession or work activity. The process of economic globalization and the Czech Republic’s membership in the European Union require additional new knowledge and competencies from the accounting profession, but at the same time it opened the way for professional accountants to pursue a career at the international level and the opportunity to do business on a much larger scale. The professional qualification of an accountant is a necessary condition for the quality performance of any of a number of possible accounting specializations at all levels.

2 Aim and Methodology The aim of the article is to evaluate the level of knowledge of students from School of Business Administration in Karviná (SBA) in the field of accounting and to find out whether there is a dependence of their accounting knowledge on the form and degree of their study, on the field of study and last but not least on their work experience. In order to achieve the objective of the article, a research question was determined: “How do university aspects affect student knowledge of accounting issues?” As an aspect of university education were determined the form of study (full-time versus part-time), the degree (bachelor’s versus followup master’s), study field (Accounting and Taxes versus other economic fields), and finally the work experience of students. The non-parametric Kruskal–Wallis test was used to answer this research question. According to Gravetter and Wallnau (2007), the Kruskal–Wallis test is a generalization of the nonparametric Mann–Whitney test for more than two groups compared. It is used to compare two or more independent samples of the same or different size. Like the Mann–Whitney test, it does not test the agreement of specific parameters, but the conformity of the sample distribution functions of the compared sets, with the key assumption being the independence of the observed values. The

Evaluation of Knowledge in Accounting of Regional Economic University Students

201

Kruskal–Wallis test shows that at least one sample stochastically dominates another sample. Because it is a nonparametric method, the Kruskal–Wallis test does not assume a normal residue distribution, unlike an analogous one-way analysis of variance. The main idea of the Kruskal–Wallis test is that, under the null hypothesis, the pooled values from all the samples are so well mixed that the average order corresponding to each sample is similar. To calculate the test, we reorder all observations by size (as if they were from a single selection) and assign them to the order values. The test statistics of Kruskal–Wallis test have the following form (1).

Q=

k T 2 12 i − 3 (n + 1) i=1 ni n (n + 1)

(1)

We reject the null hypothesis at the significance level α if the realization of the test statistic Q is greater than the critical value (quantile) corresponding to the significance level α. Based on the acquired nominal data, the researched dependence was also verified using the Chi-Square tests. Empirically determined frequencies (nij ) were compared with the theoretical, or expected (eij ) frequencies, which represent the associated frequencies expected under the assumption of independence of the variables. The magnitude of the differences between empirical and theoretical frequencies is assessed using test statistics (2). 2 c  r   nij − eij χ = eij 2

(2)

i=1 j =1

For the verification of the research question using Pearson’s chi-quadrate, the SPSS statistical program was used to calculate the level of statistical significance (i.e., p-value). The achieved level of statistical significance was examined at a significance level α = 0.05. If the p-value is less than 0.05, the effect of university studies’ aspects on the perception of the importance of the accounting profession is demonstrated. In such a case, i.e., the demonstration of the effect between variables, the intensity (tightness) of the dependence was further examined using the Pearson contingency coefficient (3).  C=

χ2 n + χ2

(3)

The closer the coefficient is to 1, the more intensity of the dependence is between the characters, the closer the coefficient is to 0, the less intensity of the dependence is.

202

I. Koštuˇríková and M. Šeligová

3 Education and Accounting Profession There is a constant talk about the need for education, training, and knowledge. According to Beneš (2019), knowledge, i.e., information and their thought processing, brings immediate effects, especially those that generate, among other things, economic profit. Organized knowledge, also known as intellectual capital, is used to create the wealth of a company (Vodák & Kucharˇcíková, 2011). Amstromg (2002) defines intellectual capital as the stocks and flows of knowledge that are available in an organization. According to the Ministry of Education, Youth and Sports (2019), the content of education of future generations must be comprehensively connected with education in the areas of digital technologies, which permeate the entire process and spectrum of teaching in all areas of regional education, due to the progressive technological development. The aim of higher education is to provide students with appropriate professional qualifications, to prepare them for research work, to participate in lifelong learning, to contribute to the development of civil society and to develop international, especially European, cooperation as an essential dimension of all activities. As stated by Veteška and Tureckiová (2008), today lifelong learning is perceived as a necessary process leading to active employment and employment in the labor market. All Member States of the European Union, including the Czech Republic, address this key aspect and determine an effective framework to support structural reforms. Education reforms are taking place not only in Europe but worldwide. For example, Donghui (2012) dealt with education reform in Chinese universities. Education plays a key role in shaping current and future economic growth, as annual labor costs increase significantly with higher levels of education. This development is also influenced by the quality of education; however, this impact is manifested only with a certain delay (MEYS, 2015). Brožová (2003) emphasizes education as a prevention against unemployment in the context of higher adaptability to changing market demands. Riddell and Song (2011) investigate the causal effects of education on employment and unemployment, with particular focus on the extent to which education improves re-employment outcomes among unemployed workers. The essence of professional education is the creation and maintenance of harmony between subjective and objective qualifications. A subjective qualification is a set of competencies acquired during life with the potential possibility of use for the performance of a certain activity. Objective qualifications are demands for the performance of a specific profession (Mužík, 1998). ˇ According to Tokarˇcíková, Kucharˇcíková, and Durišová (2015), universities have a significant role in education, which are important centers of knowledge. Students should acquire professional knowledge and should be able to apply this knowledge in solving global problems. Boccanfuso, Larouche, and Trandafir (2015) deal with the issue of the importance of university level and the impact of quality professional education on the employment of skilled workers.

Evaluation of Knowledge in Accounting of Regional Economic University Students

203

The importance of the knowledge of the accountant can be defined in connection with the characteristics of the accountant profession, which is listed in the database of the National System of Occupations of the Czech Republic. This database identifies the following positions under the term “accountant”: auditor, principal accountant in the business sector, payroll accountant, professional accountant, fixed asset accounting officer, stockholding officer, accounting and auditing specialist, independent accountant in the business sector and accounting methodology specialist in the business sphere (MLSA, 2017). Jaworska (2016) analyzed the role of accounting specialists and the tasks they perform based on social responsibility, as well as the directions of its change. Gibassier, Rodrigue, and Arjaliès (2015) dealt with the role of accountants in integrated reporting. The professional accountants were defined by Suddaby, Gendron, and Lam (2009). The International Federation of Accountants is committed to the development of the accounting profession, contributing to the adoption and implementation of highquality international ethical standards for accountants (IFAC, 2018). Other international organizations such as the Association of Chartered Certified Accountants and Institute of Management Accountants inquire into the causes of the changes that are creating the environment for businesses and professional accountants (ACCA and IMA, 2012). Professional accountants can effectively stabilize business (Šipková, 2013). In the Czech Republic, the Chamber of Certified Accountants deals with professional accountants. This institution focused on the issue of improving the position of certified accountants on the labor market, increasing their prestige and further education within the international project “Human Resources Development in the Area of Certified Accountants” (CACR, 2014). According to a survey by the Chamber of Certified Accountants, more than 69% of professionally qualified financial accountants in the Czech Republic believe that their level of qualification should be determined by the market itself, and 71% of the respondents believe that entrepreneurs do not currently fully appreciate the risks associated with the wrong choice of accountant. Employers often mistakenly perceive accountants only as a non-profit expense item, while a qualified, knowledgeable, and experienced accountant can provide management with not only accounting but also tax and legal information that is relevant to its economic decision-making (CACR, 2013).

4 Accounting Knowledge of SBA Students The accounting knowledge of full-time and part-time students from the School of Business Administration in Karviná was researched in an internal project “New trends and specifics of accounting in the context of legislative changes in the Czech Republic.” The structure of respondents is shown in Table 1. Marking of SBA students’ knowledge in the field of accounting was based on the standard classification at European universities—The European Credit Transfer and

204

I. Koštuˇríková and M. Šeligová

Table 1 The number of respondents according to aspects of university study

Number of respondents Full-time students Part-time students Bachelor’s degree students Follow-up master’s degree students Students of the Accounting and Taxes field Students of the other fields

252 457 521 188 178 531

Source: Own processing Table 2 Accounting knowledge of students

Level of knowledge Very good Good Sufficient Fail

ECTS grades A–B C–D E F

Number of students 225 303 113 68

Source: own processing

full-time form

27,78%

part-time form

44,84%

33,92%

0%

20%

Very good (A-B)

17,86%

41,58%

40% Good (C-D)

60% Sufficient (E)

9,52%

14,88% 9,63%

80%

100% Fail (F)

Fig. 1 Accounting knowledge of students according to the form of study. Source: Own processing

Accumulation System. ECTS is a tool for increasing the effectiveness of studies and courses and contributes to improving the quality of university education (European Commission, 2017). In Spain, for example, Cañibano (2008) addressed the process of adapting Spanish universities to the Bologna Principles. For a clearer view, the students’ knowledge was rated as “very good” in case of grading A and B, “good” in case of grading C and D, “sufficient” in case of grading E and “fail” in case of grading F (Table 2). Almost one-third of students achieved very good knowledge, and 43% of students showed good accounting skills. Only 9% of students had insufficient knowledge of accounting issues. In the part-time form of study, more students had very good knowledge than in the full-time form; on the other hand, full-time students had more good knowledge. In both forms of study, the lack of knowledge in the field of accounting manifested itself in approximately the same percentage of students as shown in Fig. 1.

Evaluation of Knowledge in Accounting of Regional Economic University Students

205

bachelor degree

32,26%

40,32%

16,67% 10,75%

follow-up master´s degree

31,55%

43,59%

15,68% 9,18%

0% Very good (A-B)

20%

40%

60%

Good (C-D)

80%

Sufficient (E)

100% Fail (F)

Fig. 2 Accounting knowledge of students according to the degree of study. Source: Own processing

AT field

40,68%

other fields

38,98%

28,76%

0%

20% Very good (A-B)

43,98%

40% Good (C-D)

11,86% 8,47%

17,29%

60% Sufficient (E)

80%

9,96%

100% Fail (F)

Fig. 3 Accounting knowledge of students according to the field of study. Source: Own processing

If we compare students of different degrees of study, namely the bachelor’s degree and the follow-up master’s degree, their knowledge in the field of accounting is remarkably similar. As can be seen from Fig. 2, more students have a good knowledge at the bachelor’s degree (about 3% points) than at the master’s level, where more than 1.5% points of students demonstrated insufficient knowledge compared to bachelor’s degree students. Students of the field of Accounting and Taxes (AT) should have the greatest awareness of accounting issues in comparison with students of other fields. This premise is evidenced by Fig. 3, which shows that the very percentage of students in the AT field demonstrated very good knowledge about accounting issues. The least students in this field also had insufficient accounting knowledge. Figure 4 shows that students who have work experience in the field of economics have the most good knowledge of accounting issues (42.42%). In the case of students without work experience, this level of knowledge was achieved by only 24% of respondents. Students who have work experience but in the noneconomic

206

I. Koštuˇríková and M. Šeligová

working in the economic field working in the non-economic field

24,03% 0%

20%

Very good (A-B)

16,84%

45,70%

26,12%

not working

12,12% 7,20%

38,26%

42,42%

20,78%

44,81% 40% Good (C-D)

60%

11,34%

10,39%

80%

Sufficient (E)

100% Fail (F)

Fig. 4 Accounting knowledge of students according to work experience. Source: Own processing Table 3 The influence of the form of university study on the level of accounting knowledge

Test statistics Kruskal–Wallis H df Asymp. sig.

2.135 1 0.144

Source: Own processing in the SPSS program Table 4 The influence of university degree on the level of accounting knowledge

Test statistics Kruskal–Wallis H df Asymp. sig.

0.109 1 0.741

Source: Own processing in the SPSS program

field have the most good knowledge (45.7%). This group, on the other hand, has the highest percentage of students with insufficient knowledge of accounting (11.34%).

5 Results and Discussion It was also determined by a questionnaire survey how students of both full-time and part-time forms of study at the School of Business Administration perceive the accounting profession in the Czech Republic. In the survey, 1035 students were approached, and the questionnaire filled 709 people. The non-parametric Kruskal–Wallis test was used to determine the influence of the form and degree of university study, as well as fields of study and, last but not least, the work experience of SBA students on the level of their accounting knowledge. The results are shown in Tables 3, 4, 5 and 6.

Evaluation of Knowledge in Accounting of Regional Economic University Students Table 5 The influence of the field of study on the level of accounting knowledge

Test statistics Kruskal–Wallis H df Asymp. sig.

207

8.564 1 0.003

Source: Own processing in the SPSS program Table 6 The influence of work experience on the level of accounting knowledge

Test Statistics Kruskal–Wallis H df Asymp. sig.

21.567 2 2.074E − 05

Source: Own processing in the SPSS program Table 7 The influence of work experience on the level of accounting knowledge Chi-Square tests Field of study Work experience of students No. of valid cases

Pearson Chi-Square Pearson Chi-Square

Value 10.214a 24.711a 709

df 3 6

Asymp. sig. 0.017 3.862E − 04

Source: Own processing in the SPSS program a – 0 cells (0,0%) have expected count less than 5

As can be seen from the results of Tables 3 and 4, the form of study (full-time versus part-time) and the degree of study (bachelor’s versus follow-up master’s) do not affect students’ level of knowledge, as significance values are greater than 0.05. In contrast, in the field of study, the dependence of students’ accounting knowledge on this aspect has already been proved, as shown in Table 5. Students “work experience had the most significant effect on students” knowledge of accounting, as shown in Table 6. Pearson’s chi-square test was performed to verify the dependence of the level of accounting knowledge on the field of study and on the student’s work experience (Table 7). Based on the detected levels of statistical significance, where all significance values are less than 0.05, it can be stated with 95% probability that the level of knowledge of SBA students depends on their field of study and their work experience. The contingency coefficients were also calculated to determine the intensity of the dependence, and since their value is close to 0, no close dependence can be concluded (Table 8).

208

I. Koštuˇríková and M. Šeligová

Table 8 Intensity of dependence of student knowledge level Symmetric measures Field of study Work experience of students No. of valid cases

Contingency coefficient Contingency coefficient

Value 0.119 0.184 709

Appr. sig. 0.017 3.862E − 04

Source: Own processing in the SPSS program

6 Conclusion Education develops the personality of a person who will thus be equipped with knowledge, skills, and competences not only for personal life but also for the performance of a profession or work activity. The process of economic globalization and the Czech Republic’s membership in the European Union require further new knowledge and competencies from the accounting profession, but at the same time it opened the way for professional accountants to pursue a career at the international level and the opportunity to do business on a much larger scale. The professional qualification of an accountant is a necessary condition for the quality performance of any of a number of possible accounting specializations at all levels. The aim of the article was to assess the accounting knowledge of full-time and part-time students from the School of Business Administration in Karviná. The results of the questionnaire survey show that almost 32% of students have very good knowledge of accounting and almost 43% of respondents have demonstrated good knowledge in this area. Using the Kruskal–Wallis test, the dependence of the accounting knowledge of SBA students on aspects of university study was researched. Dependence on two aspects, the study field and students’ work experience, has been demonstrated. This was confirmed by Pearson’s Chi-square test. The intensity of dependence was then researched using Pearson contingency coefficient. Based on the resulting values, it was possible to conclude that this is not a close dependence. Acknowledgments This research was financially supported by the Ministry of Education, Youth and Sports within the Institutional Support for Long-term Development of a Research Organization in 2020.

References ACCA and IMA. (2012). 100 drivers of change for the global accountancy profession. London: ACCA. [online]. Retrieved February 6, 2020, from https://www.imanet.org/insightsand-trends/the-future-of-management-accounting/100-drivers-of-change-for-the-globalaccountancy-profession?ssopc=1 ˇ Amstromg, M. (2002). Rízení lidských zdroj˚u. Praha: Grada Publishing. Beneš, Z. (2019). Vzdˇelání je povinnost, vzdˇelanost je výzva. [online]. Retrieved March 16, 2020, from https://archiv.ihned.cz/c1-66469150-vzdelani-je-povinnost-vzdelanost-je-vyzva

Evaluation of Knowledge in Accounting of Regional Economic University Students

209

Boccanfuso, D., Larouche, A., & Trandafir, M. (2015). Quality of higher education and the labor market in developing countries: Evidence from an education reform in Senegal. World Development, 74, 412–424. Brožová, D. (2003). Spoleˇcenské souvislosti trhu práce. Praha: Sociologické nakladatelství. CACR. (2013). Na vyšší odbornou kvalifikaci úˇcetních by mˇel více tlaˇcit trh, myslí si témˇerˇ dvˇe tˇretiny úˇcetních profesionál˚u. [online]. Retrieved April 26, 2020, from https://www.komoraucetnich.cz/cze/informace-komory/media/media_zpravy/page:6 CACR. (2014). Rozvoj lidských zdroj˚u v oblasti certifikovaných úˇcetních. [online]. Retrieved April 26, 2020, from https://www.komora-ucetnich.cz/cze/informace-komory/projekt Cañibano, L. (2008). Higher education in ‘business administration’ in Spain: Adapting to the European area of higher education. Revista Española de Financiación y Contabilidad, 37(139), 589–597. Donghui, Z. (2012). Tongshi education reform in a Chinese university: Knowledge, values, and organizational changes. Comparative Education Review, 56(3), 394–420. European Commission. (2017). ECTS users’ guide 2015. [online]. Retrieved April 26, 2020, from https://publications.europa.eu/en/publication-detail/-/publication/da7467e6-845011e5-b8b7-01aa75ed71a1 Gibassier, D., Rodrigue, M., & Arjaliès, D. L. (2015). From share value to shared value: Exploring the role of accountants in developing integrated reporting in practice. New York: IMA-ACCA. [online]. Retrieved March 16, 2020, from https://www.imanet.org/insights-and-trends/externalreporting-and-disclosure-management/share-value-to-shared-value?ssopc=1 Gravetter, F. J., & Wallnau, L. B. (2007). Statistics for the behavioral sciences (7th ed.). Belmont, CA: Thomson Wadsworth. IFAC. (2018). Handbook of the international code of ethics for professional accountants. New York: IFAC. [online]. Retrieved February 6, 2020, from https://www.ifac.org/system/files/ publications/files/IESBA-Handbook-Code-of-Ethics-2018.pdf Jaworska, E. (2016). The role of professional accountants in socially responsible enterprises. Finance, Financial Markets, Insurance, 2(80), 123–131. ˇ MEYS. (2015). Dlouhodobý zámˇer vzdˇelávání a rozvoje vzdˇelávací soustavy Ceské republiky na období 2015-2020. [online]. Retrieved April 26, 2020, from http://www.msmt.cz/file/ 35188_1_1/ ˇ MEYS. (2019). Dlouhodobý zámˇer vzdˇelávání a rozvoje vzdˇelávací soustavy Ceské republiky na období 2019-2023. [online]. Retrieved April 26, 2020, from http://www.msmt.cz/file/50917/ download MLSA. (2017). Národní soustasva povolání. [online]. Retrieved April 26, 2020, from https:// nsp.cz/hledat Mužík, J. (1998). Profesní vzdˇelávání dospˇelých. Praha: Codex Bohemia. Riddell, W. C., & Song, X. (2011). The impact of education on unemployment incidence and re-employment success: Evidence from the U.S. labour market. Labour Economics, 18(4), 453– 463. Šipková, K. (2013). Accountants can effectively stabilize business. CFO World. [online]. Retrieved March 16, 2020, from https://cfoworld.cz/financni-sluzby/katerina-sipkova-accaucetni-mohou-ucinne-stabilizovat-byznys-2620 Suddaby, R., Gendron, Y., & Lam, H. (2009). The organizational context of professionalism in accounting. Accounting, Organizations and Society, 34(3–4), 409–427. ˇ Tokarˇcíková, E., Kucharˇcíková, V., & Durišová, M. (2015). Education of students of the study program informatics in the field of corporate social responsibility. Periodica Polytechnica Social and Management Sciences, 23(2), 106–112. Veteška, J., & Tureckiová, M. (2008). Kompetence ve vzdˇelávání. Praha: Grada Publishing. Vodák, J., & Kucharˇcíková, A. (2011). Efektivní vzdˇelávání zamˇestnanc˚u (2nd. ed.). Praha: Grada Publishing.

Corporate Governance Disclosure in Slovak Banks Janka Grofˇcíková, Katarína Izáková, and Dagmar Škvareninová

Abstract Effective corporate governance is essential for proper functioning of the banking sector and the economy as a whole. Banks play an important role in the economy by channelling savings and depositors’ funds into activities that support entrepreneurship and help drive economic growth. From this point of view, safety and reliability of banks are key to financial stability, and therefore the way they do business is crucial to economic health. We assume that the significance of the banking sector and of systemically important banks, in particular, should also be reflected in more accurate reporting on corporate governance compared to other banks. The aim of our paper is to examine the compliance of information reported under the current legislation in the Slovak Republic with the principles of the G20/OECD on corporate governance, which were accepted by the Slovak Association of Corporate Governance. Specifically, we focus on those principles that are outlined in Part V. Disclosure and Transparency. Through the research we examine accuracy of reporting in the banks on the Slovak financial market, and subsequently we determine their order. We also compare reporting in systemic and nonsystemic banks, and for the issuers of securities and banks whose securities are not issued on the regulated market in Slovakia. The research is done through Friedman test, Wilcoxon Signed Ranking Test, Kruskal–Wallis test and Kendall’s coefficient of concordance. Our findings indicate that reporting on corporate governance is more accurate in the systemic banks and issuers of securities in comparison with other banks. Keywords Corporate governance · Disclosure · Slovak banks

J. Grofˇcíková () · K. Izáková · D. Škvareninová Faculty of Economics, Department of Finance and Accounting, Matej Bel University in Banská Bystrica, Banská Bystrica, Slovakia e-mail: [email protected]; [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_16

211

212

J. Grofˇcíková et al.

1 Introduction Banks play a crucial role in the economy through providing funds from savers and depositors to activities that support business and help to drive the economic growth. Thus, security and reliability of banks are the basis for financial stability, and therefore the way they do business is a key to the economic health of society. Weaknesses in governance in banks, which play an important role in the financial system, can lead to the transmission of problems throughout the banking sector and the economy as a whole. Therefore, effective corporate governance (CG) is inevitable for the proper operation of the banking sector and the economy as a whole, taking also the fact that the European economy is financed mainly through banks into consideration. The Capital Requirements Directive IV (CRD IV) and the Capital Requirements Regulation (CRR) are the measures that were approved in order to strengthen the financial system at the G-20 level in 2009, and are reflected in the outputs of the Basel Committee on Banking Supervision (the so-called Basel III concept). The Basel III concept consists of the three pillars. The first pillar determines the quantitative requirements of banks, the second pillar sets the qualitative requirements for management, including also the so-called corporate governance, which is the system of banks governance. And finally, the third pillar defines the requirements for market discipline and transparency of disclosure. The guidelines of the Basel Committee are based on the principles of CG published by the Organization for Economic Cooperation and Development (OECD). The aim of the generally accepted and used OECD Principles is to assist governments in their efforts to evaluate and improve their CG frameworks and to provide guidance to financial market participants and regulators.

2 Literature Review The concept of CG has been studied by several authors, such as Cadbury (1992), Shleifer and Vishny (1997), Millstein (1998) and others. Basically, these aforementioned authors pay attention mostly to non-financial corporations. In the year 1992, the first Code of Corporate Governance was developed by Adrian Cadbury. For the first time he executed the Code in which principles of CG were incorporated. The term CG refers to the system by which companies are managed and controlled (Cadbury, 2002). Musa, Musová, and Debnárová (2014) extends this statement as he says that the application of CG principles has, in addition to the impact on innovation potential, also an impact on corporate performance. Impact of CG on the financial performance and corporate social responsibility of insurance companies was also investigated by Grofˇcíková and Izáková (2019) and Grofˇcíková, Izáková, and Škvareninová (2019).

Corporate Governance Disclosure in Slovak Banks

213

With regard to systemic importance, there are several studies that examine these issues in the area of banking sector, such as Adams (2012), Beltratti and Stulz (2012). These authors pay attention to bank governance before and after the period of financial crisis. Besides these, more economists and authors deal with the specific issue of CG in banks. In her studies, Marcinkowska (2014) presents an objective analysis of the concept of CG. Her research results are critical of banking practice, as well as legal solutions and supervisory authorities. The findings of Andreás and Vallelado González (2008) indicate that the composition and size of Bank Boards are related to the capability of directors to monitor and supervise their management, and that larger and less-independent Boards may prove more effective in monitoring and advising, and thus they may create a higher value. Anginer et al. (2018) examine whether CG to shareholders leads to greater risk in larger banks compared to their smaller counterparts, as larger banks generally benefit from greater protection of the financial safety net due to their “too big to fail” status. At the same time, they also deal with the question whether shareholder-friendly CG may increase risks of banks in the countries with more generous financial safety nets. Brandao-Marques, Correa, and Sapriza (2020) conclude that in the field of banking, the presence of moral hazard which is induced by government support may lead to reduction of bank complexity and thus this may strengthen market discipline. Various elements of CG that are related to bank performance in the period of crisis are analysed by Aebi, Sabato, and Schmid (2012). The impact of higher capital requirements imposed by Basel III regulatory reforms and its macroeconomic impacts are monitored by Fidrmuc and Lind (2020). The transmission of shocks in global and domestic banking sectors is analysed by Dungey, Flavin, and LagoaVarela (2020). Their studies also provide some findings in terms of the Euro area. Another reason why CG is becoming increasingly important, the banking sector included, is that with advances in communications technology, detailed information on individual corporations and their national governance frameworks is now available on computer screens, and thus it may intensify the level of their public control and evaluation. International institutional investors are of a high importance in terms of applying the same security and return tests worldwide, in the countries in which the funds are placed. Convergence of this kind may generally raise CG standards as these investors seek and expect the same levels of government efficiency, transparency, accountability and financial integrity in various places of their investments. Basically, banks that are responsible for allocating these funds face the same problems as institutional investors. They have the same responsibility for efficiency and integrity as they manage and control the companies they finance. As a result, and regardless of their source, the funds will be used by the companies around the world that meet internationally recognized CG standards (Claessens, 2003).

214

J. Grofˇcíková et al.

3 Characteristics of the Slovak Banking Sector There are 12 commercial banks1 and 15 branches2 of foreign banks operating in Slovakia as of 1 April 2020. According to the Act on Banks no. 483/2001 Collection of Laws, commercial banks are defined as legal entities with the seat in the territory of the Slovak Republic, established as joint-stock companies that accept deposits and provide loans in addition to other banking activities. The development of the assets of the Slovak financial sector in the last decade has been most significantly influenced by the development in the banking sector, which manages more than 70% of the assets of the entire financial sector in Slovakia. In 2018, two-thirds of bank assets were loans, of which 75% were provided by the four 1 , 10% of loans were provided by medium ˇ largest banks (VUB, SLSP, TB, CSOB) and small banks, and slightly over 6% by foreign banks. During the period under review, the total liabilities of credit institutions were mostly affected by the loans and deposits that were received in the range of 75 – 79%. Credit claims account for the largest share of total assets of credit institutions (as on 31 December 2018 it was 81.5%, while in the year 2010 the share was 69%), and securities other than shares and share certificates (as on 31 December 2018 it was 12. 9%, while in 2010 this share represented 25% of total assets). Development in the retail loan market is considered to be the factor that significantly influenced the stability of the Slovak banking sector. With regard to domestic credit receivables, the volume of which amounted to 59.4 billion EUR in 2018, in comparison with 36.7 billion EUR in 2012, retail loans accounted for a stable 93 – 94% share. As we can see in Fig. 1, these are the households and self-employed (sole proprietors) that represent the highest rate of retail deposits and loans. Slovakia ranked among the countries with the highest growth of retail loans in the European Union, which were achieving a year-on-year growth for almost three years from 2012. In December 2014, this year-on-year growth was 13.5% (subsequently, the year-on-year growth decreased to 11.7% in 2017, and to 10.3% in 2018, in the 1 Ceskoslovenská ˇ

ˇ obchodná banka, a. s. (The Czechoslovak Trade Bank, hereinafter only CSOB); ˇ ˇ ˇ CSOB stavebná sporitel’ˇna, a. s. (CSOB Building Savings bank, abbreviation used CSOB SS); OTP Banka Slovensko, a. s. (OTP Bank Slovakia, OTP); Poštová banka, a. s. (Post bank, PB); Prima banka Slovensko, a. s. (Prima bank Slovakia, abbrev. used Prima); Privatbanka, a. s. (Privat); Prvá stavebná sporitel’ˇna, a. s. (First Building Savings bank, PSS); Slovenská sporitel’ˇna, a. s. (Slovak Savings Bank, SLSP), Slovenská záruˇcná a rozvojová banka, a. s. (Slovak Guarantee and Development Bank, SZRB); Tatra banka, a. s. (Tatra bank, TB); Všeobecná úverová banka, a. s. (General Credit Bank, VÚB); Wüstenrot stavebná sporitel’ˇna, a. s. (Wüstenrot Building Saving bank, Wüstenrot). 2 Citibank Europe plc (abbreviation used—Citibank); Fio banka, a. s. (Fio bank, Fio), J&T Banka, a. s. (J&T bank, J&T); Komerˇcní banka, a. s. (Commerz bank, KB); mBank S.A. (mBank); Oberbank AG; Raiffeisen Centrobank AG (Raiffeisen), UniCredit Bank Czech Republic and Slovakia, a. s. (Unicredit); BKS Bank AG (BKS); BNP Paribas Personal Finance SA (BNP); ˇ ˇ Ceskoslovenské úvˇerní družstvo (Czechoslovak credit cooperative, CSUD); Cofidis SA (Cofidis); Commerzbank AG (Commerz); ING Bank N.V. (ING); KDB Bank Europe Ltd. (KDB).

Corporate Governance Disclosure in Slovak Banks

215

Fig. 1 Total volume of retail deposits and loans in the balance sheets of Commercial Banks in the Slovak Republic at the end of the period (in mil. EUR)

amount of 3.4 billion EUR with retail loans included). In 2016, the share of loans to households in their disposable income was the highest in the Slovak Republic among the 11 countries of Central and Eastern Europe. Considering this fact we can assume that this share was the only one growing, and its value doubled from the beginning of the crisis. In addition to the rapid growth in the volume of loans, the volume of non-performing loans was also growing. In order to support financial stability, in the year 2014 the National Bank of Slovakia issued recommendations with regard to macro-prudential policy on the risks associated with developments in the retail credit market. It was not until 2019 that the year-on-year growth of retail loans slowed to 7.9% in December. Together with the October (7.8%) and November (7.7%) growth rates, this has been the lowest recorded value since the inception of mortgage banking in Slovakia (NBS, 2020). According to the tenth Allianz Global Wealth Report, 2019, which analysed the financial assets and debts of households in 53 countries worldwide in 2018, Slovakia is the only country among 11 countries in Central and Eastern Europe where household debts exceed savings per capita by more than 1000 Euros. The debt of the average Slovak is 7437 Euros, while the average in the region is 4868 Euros. Furthermore, the average amount of savings for the average Slovak is 6255 Euros, while the average for the region is 11,337 Euros. At the same time, retail deposits in the SR reversed the slowdown trend and increased by 4.8% in 2017 and by 6.9% in 2018. That time the volume of non-term deposits as well as savings deposits was growing, while term deposits at least stabilized the pace of decline. In the year 2015, the debts of Slovaks began to exceed their savings, when the debt of a Slovak reached the value of 5570 Euros. While comparing all the monitored countries in that particular period, we can see that the Slovak households reported

216

J. Grofˇcíková et al.

up to 64% of their savings deposited on retail deposits (in comparison with 45% on average in Eastern European countries). With regard to the fact that majority of commercial banks operate in Slovakia as subsidiaries of their foreign parent banks, when it comes to their corporate governance they refer to their financial groups on their websites. One of them is ˇ ˇ for example CSOB. As this bank is part of the CSOB Financial group, it follows the rules of administration and management of joint-stock companies which are based on the principles set by the OECD. Similarly, SLSP in its annual report states that the principles of CG are applied in accordance with the commitment of the parent company SLSP Erste Group Bank, which was declared in 2003 and in which it voluntarily undertook to comply with the Austrian CG Code. At the same time, SLSP, a member of the Slovak Association of Corporate Governance (SACG hereinafter), subscribed to the Code of Corporate Governance in Slovakia, which was issued by this association. In terms of responsible business, VÚB, as a member of the VÚB Group, follows the rules and policies of the parent company Intesa Sanpaolo. In its annual report, Tatra bank, as part of the Raiffeisen Banking Group, states that CG is applied in compliance with the CG Code in Slovakia, which is publicly available on the website of the SACG. Similarly, OTP bank agrees with an improvement of the level of CG, and subsequently this bank adopted the CG Code in Slovakia (The Code of Corporate Governance in Slovakia, 2016). At the same time the bank issued a statement on CG pursuant to Act no. 431/2002 Collection of Laws on Accounting, Section 20 paragraph 6. In its annual report, Prima bank informs that it complies with the Code of Governance, which is available on the Bratislava Stock Exchange (BCPB hereinafter) website. However, in its annual report, Post bank states only the information on the social responsibility of the bank (which is also documented by the aforementioned banks), but this report does not include any information on its CG. All in all, the authors Bhimani and Soonawalla (2005) recommend that CG and social responsibility should be perceived as two sides of the same coin. However, in this paper we do not intend to discuss this issue in more details as we primarily pay attention to the specific issues of CG.

4 Methodology The banking sector is one of the most important sectors of every economy, and it is not otherwise in Slovakia. As of 31 December 2019, a total of 12 commercial banks1 had their registered offices in Slovakia. In addition, 15 branches2 of foreign banks operate in the Slovak Republic. With regard to the availability of the examined data, our basic set consists of a total of 23 banks, of which 12 are with the seat in the Slovak Republic and 11 banks are the branches of foreign banks. The objective of our paper is to explore the compliance of information reported under applicable legislation in the Slovakia with G20/OECD Corporate Governance

Corporate Governance Disclosure in Slovak Banks

217

Principles, as these have also been accepted by the (SACG) Slovak Association Corporate Governance. In terms of the particular period, our research will pay attention to the reporting years from 2015 to 2019. We have made this selection due to the fact that the year 2015 is the year of the latest modifications of the guideline G20/OECD and Basel principles for banks affecting corporate governance reporting. On the other hand, this particular selection is related to the fact that this year the European Central Bank started to identify systemically important banks of individual countries as well as specify the conditions for their capital requirements. In Slovakia, five systemically important banks were designated under the Decision of the NBS No. 4 of 26 May ˇ 2015. These are CSOB, Post bank, SLSP, Tatra bank and VÚB (abbreviation see footnote 1). In the case of all of these banks, it is the Central Bank which determines the amount of the countercyclical capital cushion on a quarterly basis in accordance with Section 33 letter g. With effect from 20 May 2020, this amount shall be set at a rate of 1.5%. Information about financial situation and corporate governance of banks is drawn from their annual reports for the years 2015–2019 that are published on the websites of these individual banks. We investigate the scope, quality and structure of the published information on corporate governance. The extent and quality of the published information is quantified by a score from 0 to 2, where 0 means the information is not published, 1 means the information is only partially published, 2 means the information is published in details as requested. Within the period of these years, we monitor changes in the scope, quality and structure of the reported information through the corporate governance disclosure index, which we calculate separately for each bank and for every individual principle of corporate governance. The corporate governance disclosure index is quantified as follows: 1. CGDI: as the ratio of the number of points obtained (numerator) and the total number of points obtained by all banks (denominator). The total sum of the indices is 1; 2. CGDI(max) : as the ratio of the number of points obtained (numerator) and the maximum number of points that the bank might have obtained (denominator). The maximum number of points that the bank might have obtained is equal to the product of the number of examined principles and the maximum number of points, i.e. 27 principles * 2 points = 54 points. When calculating the index for each principle, the maximum number of points is 46 (23 banks * 2 points). Its total value in Table 1 is expressed as an average. We assume that the importance of systemically important banks should also be reflected in more accurate reporting of corporate governance in comparison to other banks. Through the selected financial indicators as well as the Friedman and Wilcoxon tests (H0 : μ0 = μ1 ; H1 : μ0 = μ1 ), we will verify the position of systemically important banks in the Slovak financial market and determine the order of information published in accordance with the examined principles of corporate governance. We examine the disclosure of 27 information in accordance with the G20/OECD

218

J. Grofˇcíková et al.

Table 1 Corporate governance disclosure Principles CGDI(max) Average CGDI

p1 0.848 1.696 0.073

p2 0.804 1.609 0.069

p3 0.478 0.957 0.041

p4 0.652 1.304 0.056

p5a 0.370 0.739 0.032

p5b 0.391 0.783 0.034

p5c 0.478 0.957 0.041

Principles CGDI(max) Average CGDI

p6a 0.630 1.261 0.054

p6b 0.130 0.261 0.011

p7 0.630 1.261 0.054

p8a 0.326 0.652 0.028

p8b 0.304 0.609 0.026

p9a 0.565 1.130 0.049

p9b 0.413 0.826 0.036

Principles CGDI(max) Average CGDI

p9c 0.217 0.435 0.019

p10 0.413 0.826 0.036

p11a 0.304 0.609 0.026

p11b 0.087 0.174 0.008

p11c 0.370 0.739 0.032

p12a 0.478 0.957 0.041

p12b 0.391 0.783 0.034

Principles CGDI(max) Average CGDI

p12c 0.348 0.696 0.030 ˇ CSOB

p12d 0.348 0.696 0.030 ˇ CSOB SS

p12e 0.304 0.609 0.026

p13 0.478 0.957 0.041

p14 0.370 0.739 0.032

p15 0.457 0.913 0.039

Average 0.429 0.858 0.037

B 0.630 1.259 0.061

B 0.426 0.852 0.041

OTP B 0.500 1.000 0.048

Post bank B 0.500 1.000 0.048

Prima B 0.259 0.519 0.025

Privat B 0.185 0.370 0.018

PSS B 0.556 1.111 0.053

Bank B/BFB CGDI(max) Average CGDI

SLSP B 0.815 1.630 0.078

SZRB B 0.370 0.741 0.036

TB B 0.407 0.815 0.039

VUB B 0.593 1.185 0.057

Wüstenrot B 0.296 0.593 0.029

KB BFB 0.926 1.852 0.089

Bank B/BFB CGDI(max) Average CGDI

mBank BFB 0.926 1.852 0.089

Oberbank BFB 0.111 0.222 0.011

BKS BFB 0.926 1.852 0.089

Raiffeisen BFB 0.204 0.407 0.020

Commerz BFB 0.019 0.037 0.002

J&T BFB 0.741 1.481 0.071 ˇ CSÚD

Bank B/BFB CGDI(max) Average CGDI

ING BFB 0.704 1.407 0.068

KDB BFB 0.130 0.259 0.012

Citibank BFB x x x

FIO BFB x x x

UniCredit BFB x x X

Bank B/BFB CGDI(max) Average CGDI

BFB 0.056 0.111 0.005

BNP BFB 0.111 0.222 0.011

Cofidis BFB x x x

Average x 0.452 0.858 0.043

CGDI corporate governance disclosure index, B bank, BFB branch of a foreign bank; for bank abbreviation, see footnotes 1 and 2

Principles of corporate governance, Chapter V. Disclosure and transparency. Their characteristics are presented in chapter 5.4. With Kendall’s coefficient of concordance, we verify the agreement in the distribution of the investigated principles (H0 :

Corporate Governance Disclosure in Slovak Banks

219

W = 0; H1 : W = 0). Through the Kruskal–Wallis test (H0 : μ0 = μ1 ; H1 : μ0 = μ1 ), we will verify the scope and quality of reported information that is published by the systemic banks and issuers of securities which are admitted to trading on the regulated market of the Slovak Republic. The formulated hypotheses will be verified at the significance level α = 0.05.

5 Empirical Results and Discussion 5.1 Legislation for Reporting of Information by Banks in Slovakia The reporting of information by banks and foreign bank branches in Slovakia is governed by several legislative standards. The primary legal regulation of the activities of banks in Slovakia is the Act on Banks No. 483/2001 Collection of Laws. Pursuant to Section 6 of this Act, the activities of banks and branches of foreign banks are subject to supervision by the NBS. In terms of supervision of banks, the NBS reviews and evaluates the organization of management, the allocation of responsibilities, the strategies adopted, the systems and procedures in place to carry out authorized banking activities, the information flows and risks to which banks may be exposed and at the same time verifies their sufficient coverage by equity. Article 37 of the Act on Banks determines the information that banks are obliged to publish on their websites and on their premises. This information must be published in the Slovak language in a comprehensible and written form. The content of the annual report is primarily determined by Section 20 of Act No. 431/2002 Collection of Laws on Accounting. The annual report shall contain the financial statements for the accounting period for which it is prepared and the auditor’s report on these financial statements, unless a special regulation provides otherwise. An entity’s annual report shall present a true and fair view and shall be audited within 1 year of the end of the reporting period. In addition, the information to be published by the Bank in its Annual Report is supplemented by the Act on Banks in Section 37 paragraphs 6 and 9. If the published information is incomplete or deviates significantly from the current position of the bank, then the bank is obliged to publish the information on the corrective measures without delay. For the purposes of supervision and in accordance with Section 37 paragraph 14 letter (c) to (e) on the Act on Banks, the NBS issued the Measure of NBS no. 16 of 2 September 2014 on the disclosure of information by banks and branches of foreign banks. This Measure was later amended by NBS Measure no. 13 of 20 October 2015. The exact structure and content of information in the form of statements that banks are obliged to submit to the NBS are defined by NBS Measure no.

220

J. Grofˇcíková et al.

3/2017, NBS Measure no. 13/2017 and NBS Measure no. 17/2014. However, this information is not freely available to the public, and it is not disclosed in such details in the financial statements and annual reports. This information is used only for the needs of NBS to regulate and supervise the banking sector. Reporting on CG is governed by Section 20 paragraphs 6–8 on the Accounting Act. The corporate governance statement, as a separate part of the annual report, is required to list those companies that have issued securities and have been admitted to trading on a regulated market. Pursuant to that amendment, the content of this declaration is as follows: (a) a reference to the CG code that applies to it, or which the company has decided to adhere to, and the information where the CG code is publicly available, (b) all relevant information on management methods and the information it is available, (c) information on deviations from the CG code and the reasons for such deviations, or information on non-application of any CG code and the reasons for which this code has not been applied by the bank, (d) a description of the main internal control and risk management systems in relation to the financial statements, (e) information on the activities of the General Meeting, its powers, a description of the rights of shareholders and the implementation procedure, (f) information on the composition and activities of the company’s bodies and their committees.

5.2 Principles of Corporate Governance for Banks The basic documents that regulate the information required for CG reporting include the G20/OECD Principles of corporate governance as of the year 2015, as well as the Corporate governance principles for banks issued in July 2015 by the Basel Committee on Banking Supervision. The main objective of bank governance should be to sustainably secure the interests of stakeholders in a manner consistent with the public interest. The Guidelines reinforce the responsibilities of banks’ managing authorities for supervision and for risk management in particular. The implementation of these principles should be proportionate to the size, complexity, structure, economic significance, risk profile and a business model of the bank. Among stakeholders, particularly retail banks, the shareholders’ share would be secondary to depositors’ interests. The G20/OECD principles of CG consist of six individual chapters: (1) Ensuring the basis for an effective CG framework, (2) The rights and equitable treatment of shareholders and key ownership functions, (3) Institutional investors, stock markets, and other intermediaries, (4) The role of stakeholders in CG, (5) Disclosure and transparency, and (6) The responsibilities of the board. Each of the chapters contains a list of supporting sub-principles, supplemented by explanatory notes.

Corporate Governance Disclosure in Slovak Banks

221

Corporate governance principles for banks (2015) determine the allocation of powers and responsibilities of the governing bodies and top management of the bank with an emphasis on the bank’s strategy and objectives, control functions, employee selection and supervision, day-to-day operations, protection of depositors’ interests and taking the interests of other stakeholders into account, with an emphasis on aligning the bank’s corporate culture, activities and behaviour in line with the expectation that the bank will operate in a safe and reliable manner, with integrity and in accordance with applicable laws. The guidelines of the Basel Committee are based on the principles of CG published by the OECD. The OECD Principles aim to assist governments in their efforts to evaluate and improve their CG frameworks and provide guidance to financial market participants and regulators. The Principles for enhancing Corporate governance, presented by the Basel Committee in October 2010, have represented a continuous development of the Committee’s long-standing efforts to promote good CG practices in banking organizations. The 2010 Principles aimed to reflect on the key impacts of the global financial crisis that started in 2007 and to improve the way banks are able to manage themselves and at the same time the way of supervision of particular bodies related to this critical area. One of the main objectives of the 2015 modification was to strengthen the Board’s accountability in the area of collective supervision and risk management. Another important goal was to emphasize key elements of risk management, such as risk culture, willingness to undertake risk, and their relationship to the bank’s risk capacity. This amended guidance also defines the specific roles of the Board, the Board’s risk committees, senior management and control functions, including the CRO (chief risk officer) and the internal audit. Another key emphasis is put on the strengthening of overall controls and bank balances. A higher risk focus and a supportive management framework include identifying the responsibilities of different parts of the organization for dealing with and managing risk. The CG principles for banks include the following 13 principles: (1) Board’s overall responsibilities, (2) Board qualifications and composition, (3) Board’s structure and practices, (4) Senior management, (5) Governance of group structures, (6) Risk management function, (7) Risk identification, monitoring and controlling, (8) Risk communication, (9) Compliance, (10) Internal audit, (11) Compensation, (12) Disclosure and transparency, (13) The role of supervisors.

5.3 Implementation of CG Principles of Banks into Slovak National Legislation Some of the OECD principles for CG as well as the Basel principles of CG for banks were already part of the national legislation of the Slovak Republic at the time of their codification. However, others were implemented into the legislation consequently. Later, many of these legal norms have been amended, which has

222

J. Grofˇcíková et al.

resulted in a complicated mixture of norms, regulations and measures that are used to govern companies in their administration and management (by their government). Apparently, the principles of CG are extensively implemented in the Commercial Code no. 513/1991 Collection of Laws. The Commercial Code is the primary legal norm in Slovakia that regulates the business of joint-stock companies. In accordance with the principles of the Act on Banks, a joint-stock company is the only legal form that is permitted for banks in Slovakia to do their business. The Commercial Code incorporates relevant amendments to the rights of shareholders to the management and control of the company and also to their equal treatment. This Code also regulates the rights of other stakeholders, related party transactions, disclosure and transparency obligations, as well as the rights, duties and responsibilities of the management bodies of joint-stock companies. The rights and obligations of institutional investors and the rules for ensuring a credible financial market were reflected in the amendment to Act No. 566/2001 Collection of Laws on Securities and Investment Services; and in the Act 203/2011 Collection of Laws on collective investments; as well as in Act no. 429/2002 Collection of Laws on the Stock Exchange. The implementation of selected principles of CG was also reflected in the Act no. 483/2001 Collection of Laws on Banks. In Section 34 paragraph 5, this Act regulates the obligation of securities managers, including banks mostly, to conduct investment transactions only at a price and conditions favourable to the client, and with professional care, unless otherwise stated in the client’s order or requirements. Sections 23 and 24 of the Banking Act also regulate the rights and obligations of the bank’s governing bodies and the organization of internal control and audit processes to ensure compliance with binding legislation and the functionality and effectiveness of the bank’s management and control system, as well as to ensure the bank’s safety and health in order to increase the value of the bank’s shares or its permanent profit. The prohibition of trading in confidential information is regulated by Section 25 paragraph 3 of the Act on Banks. In addition to the prohibition on misusing information acquired in connection with the performance of one’s duties or a position to gain undue advantage to oneself or to another, this Section also restricts the parallel functions in the exercise of which such misuse could occur. The selected shareholders’ rights related to large ownership shares are regulated by Section 28 of the Act on Banks. Section 37 of the Act on Banks lists the information that banks are required to disclose. These requirements are comparable to the OECD recommendations, particularly in part V that is related to Disclosure and Transparency. Disclosure of this information will be the subject of our further research.

Corporate Governance Disclosure in Slovak Banks

223

5.4 Reporting of Recommended Information on CG in the Annual Reports of Banks in Slovakia In compliance with The OECD Principles, Part V. Disclosure and transparency, the companies are requested to disclose the following information: 1. Audited financial statements (p1), 2. Information on financial performance and financial situation (p2), 3. Information on transactions of individual parts that form a company, contingent liabilities and off-balance sheet transactions (p3), 4. Goals and objectives of the company (p4), 5. Non-financial information including business ethics (p5a), environmental commitments (p5b) and social policy (p5c), 6. Significant ownership shares and information on entities with a significant influence on the company’s control (p6a), information on share ownership by members of the Board of Directors and members of the Supervisory Board (p6b), 7. Information on transactions with related parties, related parties’ transactions, including the terms of such transactions and their incorporation by their severity and conditions (p7), 8. A statement of remuneration in the company, including information on the remuneration of members of the governing bodies and senior management (p8a) and information on the connection between their remuneration and the long-term performance of the company (p8b), 9. Information about the members of the company’s bodies (p9a), including their qualifications, selection procedure and membership in other bodies and executive functions (p9b), and a statement as to whether the company considers them to be independent (p9c), 10. Foreseeable risk factors specific to the sector and location (p10), 11. Information concerning employees, i.e. relations between management and employees, including remuneration (p11a), collective bargaining, employee representation mechanisms (p11b) and other stakeholders (i.e. creditors, suppliers, local communities) (p11c), 12. Information on the internal organization of bodies (p12a) and strategies in the field of CG, including the content of the CG code (p12b), the procedure and processes through which it is implemented (p12c); the statement should include information on the division of powers between the shareholders, management and members of the company’s bodies (p12d), as well as the individual functions and competencies of the CEO and the Chairman of the Board of Directors (p12d), 13. Information about the audit of financial and non-financial reporting by an independent, competent and qualified audit firm (p13), 14. Information about the audit committee or other committees that supervise the internal audit activities and relations with the external auditor (p14), 15. Information about the channels through which the company disseminates information in such a way as to ensure that its users have equal, timely and affordable access to relevant information (p15).

224

J. Grofˇcíková et al.

When reporting information on CG, companies in Slovakia primarily follow the provisions of relevant laws. The information they disclose in detail is thus consistent with the information in the financial statements. Other pieces of information, which are in the form of recommendations only, are provided through general statements that are not basically specific enough in order to obtain a more accurate overview. The same procedure is followed in the case of companies that operate in the banking sector. The objective of our research is to find out to what extent the banks comply with the recommendations on disclosure of information on CG that are set by the OECD, the SACG, as well as the BCPB. With regard to Section 20, paragraphs 6–8 on the Accounting Act, issuers of securities are obliged to publish a Statement of Administration and Management, which is available on the websites of the SACG and the BCPB. In this Statement, companies are obliged to follow the “comply or explain” principle. In accordance with this principle, they are expected to explain to what extent there is compliance with the principles of these institutions, or otherwise explain the reasons for noncompliance. According to the information available on the website of the Central Register of Regulated Information and the website of the BCPB, the issuers of securities that are actively traded on the regulated market of the BCPB as of 26 April 2020 are the following banks with the seat in the Slovakia (for bank abbreviation, see footnote 1): • Issuers of shares—OTP, Tatra bank and VÚB, ˇ • Issuers of bonds—CSOB, SLSP, Tatra bank and VÚB. As aforementioned, all of these banks, with the exception of OTP bank, are locally systemically important banks NBS Decision no. 4 of 26 May 2015. Considering this, this is the main reason why we carry out the research in order to find out how these banks publish the information on the administration and management of companies. With regard to the branches of foreign banks that operate on the Slovak market, there is only UniCredit, whose bonds are actively traded on the regulated market of the Slovak Republic. Furthermore, the content of published information of public interest entities, including also banks and branches of foreign banks, is regulated by the Accounting Act in Section 20 paragraphs 9–15. The results of the survey on reporting information, which is set by the G20/OECD CG Principles in Part V. Disclosure and transparency in banks with the seat in the Slovak Republic and branches of foreign banks, are presented in Table 1. A description of individual principles is provided in the text above. The quality and scope of reporting information did not change between banks during the years under review. Banks mostly copied information on corporate governance to new annual reports from previous years. Year-on-year changes of the data in Table 1 are not relevant. We have come to the following findings: there is the highest average number of points in the case of reporting audited financial statements (p1) and related information on the financial situation and performance of banks (p2), disclosure of objectives (p4), ownership interests (p6a), transactions with related parties (p7) and

Corporate Governance Disclosure in Slovak Banks

225

information on members of management bodies (p9a). In rare cases, banks report information on the independence and ownership interests of members of the bodies (p9c, p6b) and information on collective bargaining and employee representation mechanisms (p11b). The total value of CGDI(max) shows that the average quality and scope of information reported by banks is 45.2%. In the case of individual principles, it is 42.9%. Systemic banks and issuers of securities report information on average at 58.9%, other banks on average at 41.4%. These results indicate that the systemic banks and issuers of securities are aware of their legal obligations, and thus disclose the information accurately and precisely. The average score of an individual principle is 0.858 points. Since we rated the disclosure of information on a scale from 0 to 2, this result indicates that the recommended information is published quite generally by the banks. For the principles p1, p2, p3, p4, p5c, p6a, p7, p9a, p12a, p13 and p15, there is a higher rating than the average per principle. This means that the banks pay more attention to reporting these pieces of information in comparison to their overall average. ˇ The following banks received a higher than average rating1 : CSOB, OTP, PB, PSS, SLSP, VUB, J&T, KB, BFB, BKS and ING. The order of reporting on CG principles was verified through the Friedman test (H0 : μ0 = μ1 ; H1 : μ0 = μ1 ). The test result (χ 2 = 162.416; Sig. = 0.000) indicates that at the significance level α = 0.05, there are differences in reporting of individual principles, i.e. we reject the null hypothesis. According to Kendall’s coefficient of concordance (W = 0.272, Sig. = 0.000), we can see a weak congruence in reporting principles (H0 : W = 0; H1 : W = 0). The results of the Wilcoxon Signed Ranks Test are shown in Table 2. Reporting of the examined principles can be divided into three statistically significant groups (α = 0.05). The first group consists of the principles p1, p2, p4, p6a, p7, p9a, p5c, p9a, p5, and the principle p13. On the one hand, banks most often report the information, which is recommended to report, in accordance with these principles. On the other hand, we can see that the information according to the principles of p3, p15, p12a, p10, p9b, p12b, p5b, p14, p5a, p12d, p11c, p12c, p8a, p8b, p12e, p11a, p9c is reported less frequently and to a lesser extent. The third statistically significant group of information consists of information on the ownership shares of members of bodies (p6b) and on collective bargaining (p11b). These are the pieces of information reported only exceptionally. In this part of our paper, we will pay attention to reporting of information on CG by both the banks that are systemically important for Slovakia as well as by issuers of shares on the regulated market. Firstly, we will verify the order of importance of banks with the seat in Slovakia. In order to meet this objective, we gained the information on the amount of assets, equity, sales, profit after tax, the volume of receivables from clients, the volume of liabilities to clients from the annual reports within the years 2015–2019, and at the same time we calculated the indicators of indebtedness, ROA, ROE and ROS. We arranged this data in descending order for each bank (from the highest to the lowest value). We indicated the statistics of the Friedman test (χ 2 = 303.815; Sig. = 0.000) and Kendall’s coefficient of concordance (W = 0.690, Sig. = 0.000).

226

J. Grofˇcíková et al.

Table 2 Wilcoxon signed ranks test statistics Z Sig.a

p1-p2 −1.000b .317

p2-p4 −.378c .705

p4-p6a −1.508b .132

p6a-p7 .000d 1.000

p7-p9a −1.732c .083

p9a-p5c −1.633b .102

p5c-p13 −1.342c .180

Z Sig.a

p13-p3 −2.236b .025

p3-p15 −.707c .480

p15-p12a −.378c .705

p12a-p10 −1.000b .317

p10-p9b −.707b .480

p9b-p12b .000d 1.000

p12b-p5b −1.000b .317

Z Sig.a

p5b-p14 −1.134c .257

p14-p5a −.816b .414

p5a-p12d −.333c .739

p12d-p11c −.447b .655

p11c-p12c .000d 1.000

p12c-p8a −.378b .705

p8a-p8b −.577b .564

Z Sig.a

p8b-p12e −.447c .655

p12e-p11a −1.134c .257

p11a-p9c −1.897b .058

p9c-p6b −2.000b .046

p6b-p11b −.378c .705

a Asymp.

Sig. (two-tailed) on positive ranks c Based on negative ranks d The sum of negative ranks equals the sum of positive ranks b Based

Table 3 Wilcoxon signed ranks test statistics Post bank– PSS–Post ˇ ˇ CSOB bank CSOB–VUB Z −1.464b −3.611b −2.282b Sig.a .143 .000 .022 ˇ ˇ Prima–PSS Privat–Prima OTP–Privat CSOB SSWüstenrot–SZRB SS–OTP SZRB–CSOB b b b b b Z −2.861 −1.347 −.846 −1.179 −.108 −.128b a Sig. .004 .178 .398 .238 .914 .898 TB–SLSP −4.300b .000

VUB–TB −1.032b .302

a Asymp. b Based

Sig. (two-tailed) on negative ranks; for bank abbreviation, see footnote 1

Through Wilcoxon test we determined the significant ranking of individual banks in the period under our research. The results show that SLSP is the most significant ˇ bank in the Slovak Republic, TB, VÚB and CSOB belong to the second significant group, then Post bank belongs to the third group according to its significance, and PSS is the fourth, and finally, the fifth group includes other banks with their seats in ˇ Slovakia, such as Prima bank, Privat bank, OTP, CSOB SS, SZRB and Wüstenrot (Table 3). SLSP, VUB and TB are among the leaders of the Slovak banking market. This fact has also been proved by Friedman and Wilcoxon tests, through which we detected their ranking in the groups that are created in accordance with the particular financial indicators which we examined. Given the importance of systemic banks for the country’s economy, we want to find out what information on CG is reported to a greater extent and detail by these banks compared to nonsystemic banks. The basic set of 23 banks consists of 5 systemic banks and 18 other banking entities, including branches of foreign banks.

Corporate Governance Disclosure in Slovak Banks

227

Table 4 Kruskal–Wallis test statisticsa Principles p1 Mean rank—SIB 14.000 Mean rank—others 11.444 Chi-square 1.278 Asymp. sig. 0.258

p2 15.000 11.167 2.108 0.146

p3 17.700 10.417 5.282 0.022

p4 14.500 11.306 1.040 0.308

p5a 14.400 11.333 0.962 0.327

p5b 19.500 9.917 10.040 0.002

p5c 17.000 10.611 3.904 0.048

Principles p6a Mean rank—SIB 13.800 Mean rank—others 11.500 Chi-square 0.569 Asymp. sig. 0.451

p6b 10.500 12.417 0.917 0.338

p7 17.000 10.611 4.392 0.036

p8a 13.400 11.611 0.340 0.560

p8b 13.800 11.500 0.592 0.442

p9a 12.400 11.889 0.027 0.870

p9b 14.800 11.222 1.275 0.259

Principles p9c Mean rank—SIB 11.200 Mean rank—others 12.222 Chi-square 0.135 Asymp. sig. 0.713

p10 18.200 10.278 6.126 0.013

p11a 16.500 10.750 3.619 0.057

p11b 16.900 10.639 7.723 0.005

p11c 13.300 11.639 0.282 0.595

p12a 10.800 12.333 0.229 0.632

p12b 15.200 11.111 1.660 0.198

Principles p12c Mean rank—SIB 14.800 Mean rank—others 11.222 Chi-square 1.310 Asymp. sig. 0.252

p12d 12.900 11.750 0.133 0.715

p12e 13.700 11.528 0.503 0.478

p13 12.500 11.861 0.046 0.831

p14 16.300 10.806 3.006 0.083

p15 12.800 11.778 0.101 0.751

a Grouping

variable: SIB system important bank

The results of the Kruskal–Wallis test (H0 : μ0 = μ1 ; H1 : μ0 = μ1 ) are shown in Table 4. The results of Table 4 show that despite the low number of systemically important banks in the sector, the mean rank values of reporting CG information are, except for three cases (p6b, p9c, p12a), higher for other banks. Statistically significant differences in reporting are identified through the p-value (Sig < 0.05). These include the following indicators: group transactions (p3), environmental liabilities (p5b), social policy (p5c), related party transactions (p7), industry and location risk factors (p10) and collective bargaining (p11b). In all these cases, systemically important banks achieve a higher mean rank value, which means that they report and disclose this information to a statistically significantly higher and more detailed extent in comparison with nonsystemic banks. From a statistical point of view, the reporting of other information on CG cannot be characterized as more extensive and detailed for any of the groups of banks (we cannot reject the null hypothesis). Due to the aforementioned strict regulation of the information disclosed by issuers of securities admitted to the regulated market, we are subsequently interested in the differences in the reporting of the examined information on CG between these groups of banks. As of 26 April 2020, a total of six issuers of securities admitted to trading on a regulated market are operating on the Slovak banking market, of

228 Table 5

J. Grofˇcíková et al. Kruskal–Wallis test statisticsa

Principles p1 Mean rank—issuer 14.000 Mean rank—others 11.294 Chi-square 1.624 Asymp. sig. 0.203

p2 15.000 10.941 2.679 0.102

p3 17.917 9.912 7.232 0.007

p4 13.500 11.471 0.476 0.490

p5a 14.000 11.294 0.849 0.357

p5b 15.333 10.824 2.520 0.112

p5c 13.667 11.412 0.551 0.458

Principles p6a Mean rank—issuer 14.33 Mean rank—others 11.18 Chi-square 1.215 Asymp. sig. .270

p6b 10.50 12.53 1.165 .280

p7 17.00 10.24 5.581 .018

p8a 12.33 11.88 .024 .876

p8b 11.25 12.26 .131 .718

p9a 12.08 11.97 .001 .969

p9b 14.33 11.18 1.125 .289

Principles p9c Mean rank–issuer 12.42 Mean rank—others 11.85 Chi-square .047 Asymp. sig. .829

p10 16.33 10.47 3.802 .051

p11a 16.50 10.41 4.598 .032

p11b 13.83 11.35 1.374 .241

p11c 12.17 11.94 .006 .939

p12a 11.08 12.32 .170 .680

p12b 16.08 10.56 3.434 .064

Principles p12c Mean rank—issuer 15.83 Mean rank—others 10.65 Chi-square 3.119 Asymp. sig. .077

p12d 14.92 10.97 1.777 .183

p12e 14.17 11.24 1.038 .308

p13 12.50 11.82 .058 .810

p14 14.50 11.12 1.291 .256

p15 11.42 12.21 .068 .794

a Grouping

variable: issuer

which one is a branch of a foreign bank. However, we were unable to obtain its annual reports, so this bank will be excluded from the file. Thus, information from the annual reports of five issuers of securities and 17 banks that did not issue securities on the Bratislava Stock Exchange have been taken into consideration and are compared in our research (Table 5). Statistically significant differences in reporting on information that is in compliance with the principles of CG are identified through the Kruskal–Wallis test (H0 : μ0 = μ1 ; H1 : μ0 = μ1 ) in the three areas: group transactions (p3), related party transactions (p7) and management and employee relations (p11a). Taking the pvalue (Sig < 0.05) into consideration, we can reject the null hypothesis, and with regard to the value of mean rank we can state that issuers of securities disclose this information to a statistically significantly higher and more detailed extent. With regard to reporting other information, we cannot prove the same fact since we cannot reject the null hypothesis. This also applies to the information on ownership shares of members of bodies (p6b), impact of remuneration to performance (p8b), internal organization of bodies (p12a) and information channels (p15), where the mean rank values of issuers are lower than non-issuers values, but the p-value is higher than the selected level of significance (α = 0.05). Thus, despite the higher value in case of non-issuers which results in the higher mean rank value, we cannot declare these results statistically significant.

Corporate Governance Disclosure in Slovak Banks

229

6 Conclusion Currently, the concepts of corporate governance in companies draw attention of hundreds of professionals as well as general public. Their research focuses primarily on various aspects of governance and the determinants that are concerned. Our paper focused on the principles of corporate governance since we examined the scope and quality of the information which is reported on corporate governance by banks operating on the Slovak financial market. Furthermore, we also examined the extent of implementation of the G20/OECD recommendations, with an emphasis on the principles of disclosure and transparency. We determined the order of published information and subsequently compared the accuracy of the information published by the systemic banks and issuers of securities admitted to the regulated market in the Slovak Republic with other banks. Through the research we came to the following findings: (1) the local systemically important banks that are designated by the NBS maintain a long-term leading position on the Slovak financial market, (2) the systemic banks and issuers of listed securities publish the information in their annual reports with higher accuracy and more precisely, (3) the quality of published information is average, the examined banks present the information mostly in a general form (average number of points per 1 examined principle is 0.858; the value of the corporate governance index is on average 42%, or 45%), (4) the information related to financial statements, financial position and performance, objectives, ownership interests and related party transactions is reported in the highest quality and in details by the banks. Acknowledgements This paper has been supported by the Scientific Grant Agency of Slovak Republic under project VEGA No. 1/0749/18 “Research on the application of CG principles in companies in Slovakia”. The authors would like to express their gratitude to the Scientific Grant Agency of The Ministry of Education, Science, Research and Sport of the Slovak Republic for financial support of this research and publication.

References Act no. 203/2011 Collection of Laws on Collective Investments. Act no. 429/2002 Collection of Laws on Stock Exchange. Act no. 431/2002 Collection of Laws on Accounting. Act no. 483/2001 Collection of Laws on Banks. Act no. 513/1991 Collection of Laws Commercial Code. Act no. 566/2001 Collection of Laws on Securities and Investment Services. Adams, R. B. (2012). Governance and the financial crisis. International Review of Finance, 12(1), 7–38. https://doi.org/10.1111/j.1468-2443.2011.01147.x Aebi, V., Sabato, G., & Schmid, M. (2012). Risk management, corporate governance, and bank performance in the financial crisis. Journal of Banking & Finance, 36(12), 3213–3226. https:// doi.org/10.1016/j.jbankfin.2011.10.020

230

J. Grofˇcíková et al.

Allianz Global Wealth Report. (2019). Economic research Allianz SE. Munich. Retrieved February 22, 2020, from https://www.allianz.com/content/dam/onemarketing/azcom/Allianz_com/ economic-research/publications/specials/en/2019/AGWR_2019.pdf Andreás, P., & Vallelado González, E. (2008). Corporate governance in banking: The role of Board of Directors. Journal of Banking & Finance, 32(12), 2570–2580. Anginer, D., et al. (2018). Corporate governance of banks and financial stability. Journal of Financial Economics, 130(2), 327–346. https://doi.org/10.1016/j.jfineco.2018.06.011 Basel Committee on Banking Supervision. (2010). Principles for enhancing Corporate governance. Basel Committee on Banking Supervision. (2015). Guidelines. Corporate governance principles for banks. Beltratti, A., & Stulz, R. M. (2012). The credit crisis around the globe: Why did some banks perform better? Journal of Financial Economics, 105(1), 1–17. https://doi.org/10.1016/ j.jfineco.2011.12.005 Bhimani, A., & Soonawalla, K. (2005). From conformance to performance: The corporate responsibilities continuum. Journal of Accounting and Public Policy, 24, 165–174. https:// doi.org/10.1016/j.jaccpubpol.2005.03.001 Brandao-Marques, L., Correa, R., & Sapriza, H. (2020). Government support, regulation, and risk taking in the banking sector. Journal of Banking & Finance, 112(3), 105284. https://doi.org/ 10.1016/j.jbankfin.2018.01.008 Cadbury, A. (1992). Report of the committee on the financial aspects of corporate governance. London: Gee & Co. Ltd. Cadbury, A. (2002). Corporate governance and chairmanship: A personal view (p. 280). Oxford: OUP. https://doi.org/10.1093/acprof:oso/9780199252008.001.0001 Central Register of Regulated Information. Retrieved from https://ceri.nbs.sk/ Claessens, S. (2003). Corporate governance and development. IFC Corporate Governance FOCUS publication; no. 1 (p. 56). Washington, DC: World Bank Group. Dungey, M., Flavin, T. J., & Lagoa-Varela, D. (2020). Are banking shocks contagious? Evidence from the Eurozone. Journal of Banking & Finance, 112(3), 105359. https://doi.org/10.1016/ j.jbankfin.2018.07.010 Fidrmuc, J., & Lind, R. (2020). Macroeconomic impact of Basel III: Evidence from a meta-analysis. Journal of Banking & Finance, 112(3), 105359. https://doi.org/10.1016/ j.jbankfin.2018.05.017 G20/OECD Principles of Corporate Governance. (2015). Grofˇcíková, J., & Izáková, K. (2019). Impact of the CG on the financial performance of insurance companies in the Slovak and Czech Republics. In N. Tsounis & A. Vlachvei (Eds.), Advances in cross-section data methods in applied economic research. ICOAE 2019 (pp. 1–21). Springer Proceedings in Business and Economics. https://doi.org/10.1007/978-3-030-38253-7 Grofˇcíková, J., Izáková, K., & Škvareninová, D. (2019). Impact of CG on CSR in Slovak insurance companies. In N. Tsounis & A. Vlachvei (Eds.), Advances in cross-section data methods in applied economic research. ICOAE 2019 (pp. 571–592). Springer Proceedings in Business and Economics. https://doi.org/10.1007/978-3-030-38253-7 Marcinkowska, M. (2014). Corporate governance w bankach. Teoria i praktyka. Łód´z: Wydawnictwo Uniwersytetu Łódzkiego. Millstein, I. M. (1998). Corporate governance: Improving competitiveness and access to capital in global markets: A report to the OECD. https://doi.org/10.1787/9789264162709-en Musa, H., Musová, Z., & Debnárová, L. (2014). Responsibility in the corporate governance framework and financial decision making process. Procedia Economics and Finance, Elsevier, 23, 1023–1029. https://doi.org/10.1016/S2212-5671(15)00371-8 NBS. (2020). Financial Stability Reports 2004-2019. Retrieved April 13, 2020, from https:// www.nbs.sk/sk/publikacie/analyzy-slovenskeho-financneho-sektora NBS Decision no. 4 of 26 May 2015 on the designation of locally systemically important bank. NBS measure no. 13 of 12 December 2017 on the submission of annual reports and other reports by banks and branches of foreign banks for the purposes of supervision.

Corporate Governance Disclosure in Slovak Banks

231

NBS measure no. 13 of 20 October 2015 on the disclosure of information by banks and branches of foreign banks. NBS measure no. 16 of 2 September 2014 on the disclosure of information by banks and branches of foreign banks. NBS measure no. 17 of 2 September 2014 on the submission of statements by banks. NBS measure no. 3 of 20 June 2017 on the submission of statements by banks, branches of foreign banks and securities dealers for the purpose of ensuring data collection pursuant to a special regulation. Shleifer, A., & Vishny, R. W. (1997). A survey of corporate governance. Journal of Finance, 52(2), 737. Slovak Association of Corporate Governance. Retrieved from https://sacg.sk/ The Bratislava Stock Exchange. Retrieved from http://www.bsse.sk The Code of Corporate Governance in Slovakia. (2016). (Kódex správy a riadenia spoloˇcností na Slovensku 2016). Retrieved from https://sacg.sk/kodex/

Implementation of Critical Reflection Analysis in Teaching and Learning Focused on Developing Critical Thinking Skills Lenka Theodoulides and Gabriela Nafoussi (Kormancová)

Abstract Traditional approaches in teaching and learning do not seem to be satisfactory for students at higher education institution. They do not fully focus on students’ improvement in the process of reality understanding, evaluating the huge amount of data, making the decisions and taking the responsibility for their behaviour. In order to foster those learning processes in formal education, the development of the critical thinking skills became the biggest challenge for higher education institutions (HEI). Critical thinking (CT) is one of the key competences of a university-educated person. This cannot be achieved without changing the perspective of what is the role of the teacher together with the implementation of the teaching techniques which are fostering the critical thinking. Creation of the new types of relationship between teacher and students has significant impact on critical thinking. In this paper, preliminary results of the ongoing national research project conducted at the Matej Bel University in Slovakia are presented. The current level of the critical thinking skills among the students of various study programs were tested. The main aim of this paper is to identify the processes of teaching and learning with the elements of the critical thinking. The project research strategy and methodology has been developed upon the implementation of the Critical Reflection Analyses which assess and evaluates the observed processes as well as provides a room for improvement. The new approach for teaching and learning in higher education which enhance critical thinking skills of students is considered as the key project outcome. Keywords Critical reflection analysis · Critical thinking · higher education JEL Classification: I23, J24

L. Theodoulides · G. Nafoussi (Kormancová) () Department of Corporate Economics and Management, Faculty of Economics, Matej Bel University in Banská Bystrica, Banská Bystrica, Slovakia e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_17

233

234

L. Theodoulides and G. Nafoussi (Kormancová)

1 Introduction In the last years, the surveys below showed quite alarming situation about the decreasing level of skills of university graduates in Slovakia. According to the survey provided by Slovak centre of scientific and technical information in 2015, where 2671 enterprises were involved, the negative perception of graduates’ communication skills was identified (they struggle to express themselves in a coherent way or present themselves before a large audience). The young graduates have quite serious problems to use their theoretical knowledge in the business environment, and work autonomously. As the worst was evaluated their ability of taking decision, assuming the responsibility and thinking strategically. ˇ Dad’o, Kormancová, Theodoulides, and Táborecká (2018) pointed out that the collaboration between potential employers and educational institutions in Slovakia seemed to be insufficient. HEI are lacking a practical focus in their education process, thus performance of university graduates shows significant shortcomings in using information and communication technologies, implementing marketing tools as well as exhibiting their interpersonal and critical thinking skills on the satisfactory level (Purg, Lalic, Pope, 2018). In comparison with the Vanˇco’s survey from 2008, this type of skills is considerably weaker. Moreover, current graduates struggle to implement theoretical knowledge on the workplace and use them in real working situation (Vanˇco et al., 2016). In 2017, the European Commission introduced its vision for 2025 of a European Education Area in which the free movement of learners is guaranteed with the focus on common key competencies for lifelong learning, digital skills and common values (European Commission, Higher education policy, 2017). As a reaction to the current situation, the European Union in its education and training policy aims that all students need to acquire advanced transversal skills and key competences that will allow them to succeed after graduation (European Commission, Higher education policy, 2017). These skills include high-level digital competences, numeracy, critical thinking and problem-solving. There is also a strong need for flexible, innovative teaching and learning techniques designed to improve the effectiveness of education while creating more capacity for students in higher education institutions (“HEI”). As is stated in the Council of the European Union Recommendation of 22 May 2018 on key competences for lifelong learning, “In the knowledge economy, memorisation of facts and procedures is key, but not enough for progress and success. Skills, such as problem solving, critical thinking, ability to cooperate, creativity, computational thinking, self-regulation are more essential than ever before in our quickly changing society. They are the tools to make what has been learned work in real time, in order to generate new ideas, new theories, new products, and new knowledge”. Critical thinking (CT) skills have been identified as the one of the most important capabilities of university graduates by a number of employers. The critical thinking

Implementation of Critical Reflection Analysis in Teaching and Learning. . .

235

consists of evaluating what is really going on, searching for the best account to be offered and being alert to the kind of reasoning that lies behind explanations, theories and the scientific methods of investigating. One of the key features of the critical thinking is to explore questions from a variety of intellectual perspectives. Another one is to develop argument and variety of solutions. Ciulla (1996) stated that there are two main goals to teaching critical thinking at the higher education institution. The first is to develop analytical skills for finding fallacies in arguments and to explore the nature of truth and validity (educating the head). The second focuses on interpretive skills and the emotive content of language (educating the heart). Students shall be able to read between the lines of texts, become aware of their own biases, and the way in which these biases colour their understanding of the world. Students which are trained and constantly instructed to use various elements of critical thinking become socially responsible, and they shall also be able to take a responsibility for their decisions and actions. In this paper, the concept building approach and the reflective approach were introduced as they both offer significant contribution to examine critical thinking. The concept building approach is based on examining those teaching and learning processes that are essential to develop CT skills. The reflective approach represents the qualitative research strategy and is performed by using critical reflection analysis. Based on the previous research, it is argued that critical thinking skills are crucial to higher education, and the focus on their development shall be everyday practice of each teacher.

2 Critical Thinking Skills and Reflection in Education Process Results of recent surveys mentioned above demonstrate that it is necessary to change the approach of universities towards teaching process and students as well. The traditional methods of teaching seem to be focused on memorizing; they are less interactive, and do not force students to take responsibility for their behaviour. Despite the importance conveyed by the education system about developing critical thinking skills, effective efforts to put such skills into practice and to promote their training have not been noticeable so far (Noddings, 2008). More complex thinking skills are not covered by conventional teaching and assessment formats, which are still too focused on data transmission, memorization of factual information and subsequent evocation of knowledge in evaluation situations (Brady, 2008; Paul, 2005; Pithers & Soden, 2000). In sum, we can accept that critically thinking is not an innate and intuitive ability, spontaneously sprouted (Rivas & Saiz, 2010). On the contrary, it emerges from the learning-teaching process, being gradually and deliberately acquired, and assuming a previous and symbiotic mastery of a set of basic skills, such as reading comprehension, argument analysis and production, or still, search for evidence to stand for a particular point of view (Facione, 2010; van Gelder, 2005).

236

L. Theodoulides and G. Nafoussi (Kormancová)

2.1 Concept Building Approach Defining the Purpose of Critical Thinking in Higher Education Critical thinking is now generally recognized as one of the essential generic skills for developing a professional career as well as acting as a responsible citizen. It has been assumed that both formal and informal education, from the early to the university education, affect the level of mastering these skills. However, there are two basic approaches to developing CT skills in education—one is promoting a model where critical thinking skills are taught separately, the other claims that it is more effective to integrate development of critical thinking skills into teaching subject-specific knowledge and skills (Behar-Horenstein & Niu, 2011). In both cases, though, it is necessary to understand the processes that constitute critical thinking and enable students go through these processes to develop them (Lipman, 1988 in Behar-Horenstein & Niu, 2011). Some of the tasks typically associated with the HEI business students’ critical thinking focused on identifying problems, incorporating underlying assumptions, using relevant data sources, problem-solving from various perspectives, generating viable alternatives and comprehending the consequences of the suggested solutions. In order to examine the teaching process and analyse how the critical thinking is developed, the five key parameters were determined as essential to observe. They were formulated as follows: 1. 2. 3. 4. 5.

Students do not accept data and information automatically. Students have doubts on what they read or what they are told. Students suggest new solutions. Students develop good arguments. Students raise questions.

2.2 Reflective Approach Towards Development and Improvement of Critical Thinking The role of reflection is analysed by a number of scholars, and it is associated with its development in different contexts. Kolb (1984) saw the reflection as an essential element in experiential learning in the context of the cycle of the learning based upon experience in Kolb’s work. On the contrary, in work of Bound, Keogh, and Wolker (1985), the reflection was associated with the role of emotion. Reflection plays also important role in professional development, and it has been important as a subject of research (Moon, 1999). Reflection is defined as a cognitive process in which people attempt to increase their awareness of personal experiences and therefore their ability to learn from them (Gray, 2007). Dualism in any reflection process has been identified by Anseel, Lievens, and Schollaert (2009). They suggest that reflection as a dual process model

Implementation of Critical Reflection Analysis in Teaching and Learning. . .

237

of information processing and the depth of elaboration of complex data influences learning and behavioural outcomes. During the process of education reflection interventions provided by teachers are helping students to switch their mode of data, information and knowledge processing from passive (automatic) to conscious that lead to critical thinking and better learning. The reflection opens the door for effective information flow and interactions which are inevitably important to build good relations based on trust. Mutual understanding and trust between teacher and student are framed by bothways communication, exchange of information and giving–receiving feedback. This is an ongoing interaction exchange which is composed under the reflection process. Kolb’s learning cycle (1984) which emphasize the experience and practice was modified in order to describe the key activities contributed to learning process based upon the reflection. The modified learning cycle involves four stages, namely exploring (searching for ideas, changes and concepts), testing (experiments, modelling, practicing), output (performance, change, experience) and reflection (observation, consideration and assessment). Students reflect on the activity undertaken during the “reflection phase”, and share their reactions in a structured way with other members of the group. The teachers’ role as facilitator is very important during each phase of the cycle. During the process phase, teachers should be prepared to help students think critically about their experience and to help them verbalize their feelings and perceptions. Additionally, teachers’ role is to help students to conceptualize their reflections on the experience so that they can move towards clear conclusions.

3 Research Philosophy and Methodology The paper aims to identify the process of teaching and learning with the elements of the critical thinking. It will prove how the process of teaching impacts on the learning and developing the critical thinking skills of students at the higher education. It has been examined through three research assumptions which were formulated as follows: 1. Traditional teaching and learning process in higher education is lacking the focus on the development of the critical thinking. 2. The role of teacher is shifting from the directive mentorship towards the role of facilitator, coach and study guide. 3. Development of the critical thinking skills of students depends on possessing these skills by teachers and implement methods, techniques and tools focusing on critical thinking development. The assessment and evaluation of the findings were performed as the first stage of the ongoing research project conducted in Slovakia. The results reflect the current level of UMB students’ critical thinking and proposals on how their critical thinking can be developed. The project (KEGA 018UMB-4/2018 titled “Coaching approach

238

L. Theodoulides and G. Nafoussi (Kormancová)

Fig. 1 Research methodology. Source authors Research strategy

Research design

Data collection

Data analysis techniques

Research aims Research questions

Target group: UMB teachers

Interviews (feedback from teachers, best practices) Observation Questionaire

CRA – Critical reflctive analysis 1/ Reflection: teacher vs.observer 2/ Reflection: Students vs. Observers

as a new form of critical thinking development of students in higher education”) has been conducted between 2018 and 2020 and its research design consists of several steps as it is shown in Fig. 1. The target group consists of 17 different teachers on both undergraduate and graduate level and represents all academic disciplines at Matej Bel University: natural science, social and economic science, law and humanity science. The selected teachers were observed performing particular teaching activity (e.g. seminars, lectures, consultancies). The research outcome resulted in formulating the key steps and set of recommendations for teachers what is necessary to do in order to develop students’ critical thinking skills. In the first step (November 2018–April 2019), 17 observations of teachers’ classes were provided using the Critical Reflective Analysis (CRA), described in detail below. To increase the objectivity of the research, one-to-one interviews with participating teachers after each observation were arranged. The purpose of this was to evaluate together with an observer the students’ performance, at the same time, to have a feedback from teachers. Additionally, it was used for collecting the best practices of researched participants. In the second step (April 2019), an additional questionnaire has been designed to reflect the partial results of the previous step of the research. The aim of this survey was to identify the reasons why students are afraid to ask questions and

Implementation of Critical Reflection Analysis in Teaching and Learning. . .

239

actively participate in seminars. For this purpose, eight questions were formulated (to identify the barriers hindering them to ask questions). Finally, 116 participants were involved (83.6% of women; 16.4% of men), students at bachelor level, study program Business Economics and Management, at Matej Bel University, Faculty of Economics.

3.1 Implementation of the Critical Reflection Analysis The original framework of the Critical Reflective Analysis (CRA) has been developed in 2013 as the Reflection Method (ReMe) by Lena Theodoulides and Peter Jahn, from the Matej Bel University. Since that it was used to assess and evaluate different processes in several companies in Slovakia. The obtained results have been examined, and they provided important base for further improvements. The processes of reflecting and formulating improvements were conducted between managers and their followers and thus it increases the organizational performance as well as the performance of the individuals. The successful results of Reflection Method in the business world have encouraged us to test the method further. It was implemented in researching how leaders reflect on the importance of “complexity, system view, process of feedback, sharing information and knowledge” in their leadership actions. The method has been enriched by three core processes of learning, critical thinking and reflection that contribute to the construction of the critical reflection analysis (Theodoulides, Kormancova, & Cole, 2019). Critical reflection analysis (CRA) is the method that offers possibilities for identifying the key processes or parameters that can be observed, evaluated and assessed. CRA is considered as the broad method that can be utilized and generally applied for monitoring of any social process for instance teaching and learning. Its aim is to offer a solution and to provide quantitative measurements as well as the qualitative evaluation of social processes which might be difficult to measure. Therefore, critical reflection analysis is characterized as mainly qualitative method, but it is performed by using evaluation scales and ranges that are expressed in quantitative measures. The obtained results provide the important evidence for further discussion and improvements.

3.1.1

Selection of Variables

The selection of variables is the starting step in the process of implementation of critical reflection analysis. Its main aim is to figure out variables or factors that make a significant influence on the observed phenomena. The conceptual approach described on previous chapter highlighted five key parameters of the critical thinking which are essential to develop and maintain during higher education.

240

3.1.2

L. Theodoulides and G. Nafoussi (Kormancová)

Assigning the Weights of Variables

Selected variables/parameters which are essential for critical thinking have different priority or level of influence. In order to distinguish from the most influential variable up to those with the least priority, the weighting scale from 1 to 10 is suggested. The range is set up within three levels as follows: (a) Significant impact (weighting from 8 to 10)—parameters strongly influence the process, they represent the key performing criteria, greatest expected results or has the highest priority for the researched phenomena. (b) Standard impact: (weighting from 4 to 7)—parameters have standard, regular influence on the examined processes. They constantly exist in observed topic and their relevance depends on the certain conditions, circumstances or current situation in specific period. (c) Additional impact: (weighting from 1 to 3)—parameters have only substitutional influence on evaluated process. They usually appear as additional parameters to those fundamental ones or in certain crisis situations. As it is shown in Table 2 the highest weight was given to the variable “student propose new solutions”. Referring to the concept building approach and theory related to critical thinking, this parameter is the most difficult to perform and requires consistent training of critical thinking skills. The variable “student do not accept data and information automatically” has been assigned with the lowest weight as it is considered to be the most fundamental feature of critical thinking and should be practice at the beginning of the process of developing critical thinking skills. From the studied theory and practical trainings conducted in the field of critical thinking, there can be few other parameters proposed to observe. There is no constant number of variables suggested because it depends on the several conditions for instance on the number of observed processes, depth of the analysis or novelty of the observed phenomena.

3.1.3

Reflective Assessment and Evaluation of the Variables

The process of assessment and evaluation begin by creation of reflective table which consists of five zones. These zones are constant as well as the scaling. Table 1 presents measurements from 1 to 99 and qualitative evaluation of the researched variables. The outcomes and qualitative evaluation of the variables are formulated always differently depending on the expected performance and level of the variables/parameters/or observed processes. However, in education process, the assessment and evaluation of the students’ performance always varies. The focus on the critical thinking is crucial to structure in both quantitative and qualitative forms. The expected results shall be described in more specific comments and linked them with the scaling range. It is not necessary to structure the scaling and detailed qualitative comments as it is

Implementation of Critical Reflection Analysis in Teaching and Learning. . .

241

Table 1 Reflective assessment and evaluation table

Zone (verbal evaluation) Very bad (extremely poor results, serious mistakes and weaknesses, standards only formally fulfilled)

Scaling 1–29

Point range 1–9

10–19

20–29

Bad (still poor results, mistakes and weaknesses exists, positive sign of standards occurred)

30–44

30–34

35–39

40–44

Zone of indecision (acceptable expressions, standards completed partially, more evidence occurred)

45–54

45–48

49–51

52–54

Good (sticking to the standards, set of expectations are fulfilled in satisfactory level, positive achievements)

55–79

55–62

63–71

Expected outcomes and qualitative evaluation of observed variables Results do not correspond with goals, expectations and requirements Extremely small signs of expected performance which lead to completing required parameter Limited results and small attempt to fulfil the goals and prospects appeared Rather formal expressions of completing the proposed goals/criteria Attempts of the meaningful activities which might reach the expected results Signs of fulfilment of the expectations in the line of formal and practical compliance. Results remain with significant limitations Expressions of the expected results oscillate between formal and real compliance. In their completion an evaluation with positive results occurred Expressions of the expected results oscillate between formal and real compliance. In their overall evaluation positive signs dominated Fulfilment of criteria is in progress. In responses to observed activity positive evaluation dominated Completion of parameter oscillates around the line of standard expectations and appears on the satisfactory level Completion of the parameter oscillates above the line of set of standards. Positive achievements are recognized (continued)

242

L. Theodoulides and G. Nafoussi (Kormancová)

Table 1 (continued)

Zone (verbal evaluation)

Scaling

Point range 72–79

Very good (standards are exceeded, excellent performance beyond expectations)

80–99

80–86

87–92

93–99

Expected outcomes and qualitative evaluation of observed variables Completion of the parameter oscillates above the line of set of standards and shows its permanent potential for improvement Achieving the expectations reaches permanent superior level and shows efforts for excellent results Achieving the expectations reaches high level of standards and shows excellent results Fulfilment of the parameter highly and permanently exceeds the standards as a result of which is setting the new criteria, goals, requirements and expectations

Source: Authors based on Theodoulides and Jahn (2013)

presented in Table 1. However, more detailed description of the expected results in the qualitative (verbal) form is required. Additionally, the alignment with the measurable assessment in a form of scaling or numerical points begins to establish a base for the reflection and feedback process which are essential elements of the critical thinking skills.

3.2 Presentation of Results The standard values of the researched parameters have been identified after the completion of the conceptual phase, and they refer to the outcomes of the current situation analysis. They were determined before class observations were conducted. Target values of the parameters represent the long-term objectives of the research team. There are more class observations planned in the near future. At the same time, the launching of training material is in the process. Finally, a few feedback sessions with the teachers involved will be organized regularly. The score evaluated by the observer is compared with the standard and target values. The existing gaps provide important information for the further improvement and possible change in the process of teaching. Observations of the academic sessions of the target group and the assessment based upon the critical reflection analysis were conducted by two research fellows. After the observation, each teacher reflected the parameters by himself/herself also

Implementation of Critical Reflection Analysis in Teaching and Learning. . .

243

Table 2 Research results based on the critical reflection analysis Critical thinking parameter/variable

Weight

Observer’s assessment

Teacher selfreflection (average)

Standard value

Target value

Students do not accept data and information automatically

6

38

53

68

88

students present doubts on what they read or what they are told

7

25

35

52

68

Students propose new solutions Students develop good

10

22

46

45

72

9

41

54

65

85

8

25

41

68

78

arguments

Students raise questions Source: Authors

by using CRA. After these two independent observations, the feedback discussion between the observer and teacher took place. The presented results in Table 2 show that all parameters assessed by observers and each teacher scored significantly lower than the standard as well as the target values. Moreover, the assessment scores of all parameters conducted by observer were assessed lower than teachers’ self-reflection scores. The lowest level measured by an observer was in the parameter “students propose new solutions”. This ability is the most difficult to develop in the academic environment. On the one hand, the measured parameter “25” requires the focus on critical and creative thinking that is associated with the invention and innovation. On the other hand, the teacher’s assessment of this parameter was 46 which might be explained by different perceptions and expectations of this parameter by teachers. Moreover, the same second lowest score 25 measured by observers was given to the two parameters “students present their doubts on what they read or what they are told” and “students are raising questions”. The measurements conducted by teachers were the lowest in the parameter “students present their doubts on what they read or what they are told” with scores on average 35, and the second lower score got the parameter “students are raising questions” with average 41 points. The findings presented in Table 2 and conducted feedback sessions highlighted some dilemmas and challenges that teachers are recently facing. Apparently, it is expected that teachers themselves understand the fundamental concept of critical thinking. Additionally, they are expected to be able to use various techniques in their teaching activities in order to enhance the critical thinking of their students. But it seemed to be a big challenge for a lot of teachers to accept the critical thinking as a part of their teaching strategy. Similarly, they need to get familiar with the key

244

L. Theodoulides and G. Nafoussi (Kormancová)

techniques, for instance, how to analyse the data and recognize fallacies, how to develop a good argument and finally asking the right questions. The conducted reflection on teaching approach and process of learning is not an easy approach to perform. The main reason for significant differences in evaluating parameters of the students’ critical thinking skills can be explained by Moon’s (1999) definition of the reflection as “form of mental processing with a purpose and/or anticipated outcome that is applied to relatively complex or unstructured ideas for which there is not an obvious solution”. Even though the critical reflection analysis provided a structure of observed parameters and specific measurements, the teachers’ approach towards developing critical thinking skills of their students remains rather subjective. Through the feedback discussions the process of our own (teachers) teaching and learning was considered as a process of metacognition. Each teacher from the research team critically reviews own behaviour during teaching process (style of communication and building relationships with students), the product of the teaching (essay, presentation, discussion or case study) and his or her engagement in self-development in critical thinking. Teachers may not be clear about the fact that critical thinking is a process and the necessity to build a strategy to satisfy those specific aims—five parameters (mentioned in Table 2). The results obtained by the implementation of the CRA, as a research method, can support the assumption that reflection is necessary to be implemented in the teaching and learning process. Furthermore, the reflection can develop the critical thinking and also contributes to the personal development. The relevance of the obtained results by critical reflection analysis has been justified by two types of information, i.e. quantitative data and verbal explanation. The comparison with the planned standard and target values generated many recommendations on how to improve teaching process in order to develop the critical thinking of students. Consequently, teachers who are the enthusiastic advocates of reflective tools and techniques (essay, journal writing, service learning, etc.) will practice critical thinking more effectively than those colleagues who do not apply reflection in their professional work. Educators frequently acknowledge that good students ask good questions (Gavett et al., 2007). Formulate a good question is rather hard not only for students (as was confirmed by our results) but also for teachers. At the same time, asking questions support the students’ critical thinking. Another part of the research aimed to identify the reasons why surveyed students do not ask questions. Based on the results from questionnaire (described above), 19.8% of respondents have never asked directly a question to his/her teacher. On the one hand, some of the reasons were quite obvious, e.g. not to study enough before a lecture/seminar (60.3%). On the other hand, some were quite surprising. Almost one-half of students (45.7%) are afraid to say something inappropriate before their classmates which consequently might be quite embarrassing for them. They often find their group in the class too big to say something relevant (32.8%). Additionally, they expect to be asked directly, and they are not willing to start the discussion themselves

Implementation of Critical Reflection Analysis in Teaching and Learning. . .

245

(15.5%). These results indicate that the first research assumption can be confirmed. It was stated that the traditional teaching and learning process in higher education is lacking the focus on the development of the critical thinking. At the same time, it is necessary to acknowledge that the development of the critical thinking skills at the university level is quite challenging if they were not developed in a sufficient way in the primary and secondary education. In the other part of the survey, students explained the situations when they would present their own opinion more willingly. One-half of them mentioned the use of more appropriate form of teaching. About one-quarter of them prefer to have more space to express themselves, given by their teachers or given by their classmates (19.8%). From these results, it is quite evident that the majority of students desire to have more interaction in the teaching process. This was quite a valuable piece of information which can help teachers to improve their teaching approach towards students in the future. This study has some limitations, e.g. the sample of students surveyed, who were just from one university. However, efforts were made to ensure that the results presented are accurate and complete as possible. The both steps of the research provided are complementary. In the first step, the behaviour of students to teachers’ performance was observed. In the second step, the reasons for this behaviour were analysed. This helped teachers to understand how students perceive their approach to teaching. Additionally, the significant research findings have been formulated and were shared among participating scholars and other colleagues interested in developing CT skills among their students. The exchange of research results and pedagogical experience can consequently support less experienced counterparts in the researched topic. The most crucial ones are as follows: (a) There is a precondition that only teachers possessing critical thinking skills are able to form critical thinking skills of their students. (b) The constant dialogue and exchange of different views between academics and students are essential for teaching and learning. (c) The creation and maintenance of informal and friendly atmosphere help to establish the trust which overcomes the biases and fear to raise questions, argue and/or participate in debate in the classroom. (d) The processes of reflection, self-reflection and feedback have significant impact on both teachers’ and students’ development of the critical thinking skills. The research results will be summarised once the project will be finished and presented via e-learning Moodle system and open access platform. This is one of the possibilities how the research outcomes will be shared and disseminated.

246

L. Theodoulides and G. Nafoussi (Kormancová)

4 Conclusion The main aim of this paper is to identify the process of teaching and learning with the elements of the critical thinking. In order to develop the critical thinking among students in HEI, the change in teaching and learning is crucial. Apparently, the role of teachers is gradually changing. They become more facilitators and study guides with the constant questioning and debating on everything rather than the directive mentors who present their opinion as the “only truth” (as was accepted in the past). The arguing, requesting of an evidence, reasoning, verifying and justifying everything shall become a key component of every lecture, seminar or discussion. The improvement of critical thinking skills of students in HEI can be done by implementation of those techniques and tools which are focused on the critical thinking, for example essays, journals, case studies, storytelling, service learning and debates. The development of critical thinking skills among students is feasible by those teachers who themselves recognise the importance of these skills. It requires teachers’ acknowledgement that critical thinking has become part of their private and professional life. Thus, it creates an environment where implementation of the critical thinking is the highest priority for every teacher and for HEI. Finally, there is a vast number of studies focusing on students’ critical thinking skills. On the contrary, the teachers’ critical skills are not in the centre of interest of researchers. It might be a subject of the future research. Acknowledgements The Grant Agency KEGA supported this research, project KEGA 018UMB4/2018 Coaching approach as a new form of critical thinking development of students in higher education.

References Anseel, F., Lievens, F., & Schollaert, E. (2009). Reflection as a strategy to enhance task performance after feedback. Organizational Behavior and Human Decision Processes, 110, 23–35. Behar-Horenstein, L., & Niu, L. (2011). Teaching critical thinking skills. In Higher education. A review of the literature. Journal of College Teaching & Learning, 8(2), 25–41. Bound, D., Keogh, R., & Wolker, D. (1985). Reflection: Turning experience into learning. London: Kogan Page. Brady, M. (2008). Cover the material: Or teach students to think? Educational Leadership, 65, 64–67. Ciulla, J. B. (1996). Ethics and critical thinking in leadership education. Journal of Leadership Studies, 3(3), 111–119. Council of the EU. (2018). Recommendation on key competences for lifelong learning. Retrieved April 23, 2020, from https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/ ?uri=CELEX:32018H0604(01)&from=EN

Implementation of Critical Reflection Analysis in Teaching and Learning. . .

247

ˇ Dad’o, J., Kormancová, G., Theodoulides, L., & Táborecká, J. (2018). Management and leadership development needs—The case of Slovakia. In D. Purg, A. B. Lali´c, & J. A. Pope (Eds.), Business and society: Making management education relevant for the 21st century (pp. 177– 202). Cham: Springer International Publishing. European Commission. (2017). Higher education policy. Retrieved April 23, 2020, from https:// ec.europa.eu/education/policies/higher-education/about-higher-education-policy_en Facione, P. A. (2010). Critical thinking: What it is and why it counts. In Insight assessment. Retrieved March 24, 2020, from http://www.insightassessment.com/home.html Gavett, E., et al. (2007). Critical thinking: The role of questions. Perspectives on Issues in Higher Education, 10(1), 3–5. Gray, D. E. (2007). Facilitating management learning: Developing critical reflection through reflective tools. Management Learning, 38, 495–517. Kolb, D. A. (1984). Experiential learning as the science of learning and development. Englewood Cliffs: Prentice Hall. Moon, J. (1999). Learning journals. A handbook for academics, students and professional development. New York: Routledge Falmer. Noddings, N. (2008). What does it mean to educate the WHOLE CHILD? Educational Leadership, 63, 8–13. Paul, R. (2005). The state of critical thinking today. New Directions for Community Colleges, 2005, 27–38. Pithers, R. T., & Soden, R. (2000). Critical thinking in education: A review. Educational Research, 42, 237–249. Rivas, S. F., & Saiz, C. (2010). Es posible evaluar la capacidad de pensar críticamente en la vida cotidiana? In H. J. Ribeiro & J. N. Vicente (Eds.), O lugar da lógica e da argumentação no ensino da Filosofia (pp. 53–74). Coimbra: Unidade I & D, Linguagem, Interpretação e Filosofia. Slovak Centre of Scientific and Technical Information. (2015). Employers’ survey. Retrieved from https://www.cvtisr.sk/buxus/docs//VS/absolvent/zamestnavatelia.pdf Theodoulides, L., & Jahn, P. (2013). Reflective method: Tool for organizational learning. Bratislava: Iura Edition. Theodoulides, L., Kormancova, G., & Cole, D. (2019). Leading in the age of innovation: Change of values and approaches. New York: Routledge Taylor & Francis Group. van Gelder, T. (2005). Teaching critical thinking: Some lessons from cognitive science. College Teaching, 53, 41–46. Vanˇco, M., Srnankova L’, Blanar, F., Slovikova, M. (2016). Analýza získania prierezových kompetencií na slovenských vysokých školách. Retrieved April 23, 2020, from https://www.minedu.sk/ data/att/10091.pdf

Comparison of Methods of Poverty Rates Measurement Anna Saczewska-Piotrowska ˛

Abstract There is a lot of poverty lines used in the world. Commonly lines employed to determine the level of global poverty are the World Bank’s poverty lines. This institution used in the past one international poverty line ($1 from 1990 and $1.25 from 2008 to 2014), but now the World Bank employs a few poverty lines to adapt them to income situation in the countries. There are used $1.9, $3.2 and $5.5 lines. Calculations of the percentage of poor people (so-called poverty rate) using these three lines give different results. Besides, each country employs its own line (so-called national lines) to determine the poverty rate in own country. To assess the agreement between international and national methods of measurement of poverty rates, Bland-Altman plots and Passing-Bablok regression were applied. The data about poverty rates in 102 countries were used in the study. The analysis was conducted for all countries and in groups of countries according to their income situation (low-income, lower-middle-income, upper-middle-income, and high-income countries). The analysis was preceded by an assessment of the strength of association between income group and poverty prevalence (Cramer’s V), and of the degree of correlation between national and international poverty rates (Spearman’s rank correlation). The study showed that national and international poverty lines are not substitutes and give different information about the poverty level. International poverty lines give information about global poverty, but they do not include regional specificity which is incorporated in national poverty lines. Keywords National poverty rates · International poverty rates · Agreement of methods JEL codes: I32, C12, C13

A. Saczewska-Piotrowska ˛ ( ) Department of Labour Market Forecasting and Analysis, University of Economics in Katowice, Katowice, Poland e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 N. Tsounis, A. Vlachvei (eds.), Advances in Longitudinal Data Methods in Applied Economic Research, Springer Proceedings in Business and Economics, https://doi.org/10.1007/978-3-030-63970-9_18

249

250

A. Sa¸czewska-Piotrowska

1 Introduction The methods of measurement of poverty may vary in different countries. The starting point for the adoption of a given method is to choose a specific definition of poverty. There is no universal method of defining and measuring poverty. All poverty definitions can be divided into three categories (Hagenaars & de Vos, 1988). The first category includes definitions according to which poverty is “having less than an objectively defined, absolute minimum.” The second group includes definitions according to which poverty is “having less than others in society.” In the third group, there are definitions according to poverty “is feeling you do not have enough to get along.” The first category includes, inter alia, definitions: basic needs approach used by Booth (1892), Rowntree (1901), Orshansky (1965), and food/income ratio definition used by Watts (1967), Love and Oja (1977). Relative deprivation with respect to various commodities proposed by Townsend (1979) is an example of a definition from the second category. Definitions from the last group have been developed by Goedhart, Halberstadt, Kapteyn, et al. (1977). According to the first category the poverty is absolute, according to the second category it is relative, and according to the third category absolute, relative, or something between. Another difference between the groups is that the first and second category defines poverty as an objective situation, and the third category as a subjective situation. The poverty may be analyzed as a unidimensional phenomenon (usually the choice between expenditure or income as an indicator of welfare, discussion about indicator presented in many publications, e.g., Hagenaars, Vos, & Zaidi, 1994) or multidimensional phenomenon (newer approach described by Sen (1997), Nolan and Whelan (2007), Alkire and Santos (2014)). Due to the availability of the data, a one-dimensional approach is usually chosen. It should be mentioned that the choice of definition of poverty is one of many decisions to be made in measuring poverty. The other choices are related to, inter alia, equivalence scale, measurement unit, measure of central tendency, poverty line. A comprehensive description of the methodology of defining and measuring poverty is presented in the literature, e.g., Atkinson, Cantillon, Marlier, et al. (2002), Haughton and Khandker (2009). Adopting a specific definition of poverty and making other decisions about measuring poverty result in different estimates of the percentage of the population whose income (or expenditure) falls below a poverty line. This percentage of the poor population is the most popular poverty measure and is known as the poverty headcount ratio (or poverty rate or incidence of poverty). International institutions use different poverty lines. Eurostat—statistical office of the European Union (EU)—uses a relative poverty measure set at “60% of the national median equivalised disposable income after social transfers” (Eurostat, 2018). The Organisation for Economic Co-operation and Development (OECD) also uses a relative measure which is set at “50% the median household income of the total population” (OECD, 2020). The World Bank uses an absolute poverty line. Original dollar-a-day line (Ravallion, Datt, & van de Walle, 1991) was based on

Comparison of Methods of Poverty Rates Measurement

251

six of the lowest national lines from 33 national poverty lines in the database—the common value of these six countries is about one US dollar (in 1985 Purchasing Power Parity—PPP). This line was used for the first time in the 1990 World Development Report (The World Bank, 1990). The updated dollar-a-day line (in 1993 PPP) was based on the median value of the ten lowest lines from the same database (Chen & Ravallion, 2001). The exact value was $1.08 a day. The next update (Ravallion, Chen, & Prem, 2009) was based on data from 74 countries, and the poverty line was calculated as the average value of the national poverty lines—the value was $1.25 a day in 2005 PPP and was used from 2008 to 2014. The latest update is a simple average of national poverty lines (the same database of 74 countries) and is set at $1.9 a day in 2011 PPP (Jolliffe & Prydz, 2015; Ferreira, Chen, Dabalen, et al., 2016; Jolliffe & Prydz, 2016; The World Bank, 2020b). This is an extreme poverty line for low-income countries. Additionally, there were set the lines for lower-middle-income countries ($3.2 a day in 2011 PPP) and uppermiddle-income countries ($5.5 a day in 2011 PPP). Detailed information about income groups is set in the next section. Countries use own poverty lines and therefore there is a difference between poverty rates calculated based on international and national poverty lines. Correlation between national and international poverty rates was studied by Gentilini and Sumner (2012), Greenstein, Gentilini, and Sumner (2014), and Wagle (2019). However, the existence of a correlation does not provide information on the comparability of methods used. This study aimed to assess the agreement between methods of calculation of poverty rates: international (used by the World Bank) and national. The analysis was conducted for all countries and in groups of countries according to their income situation.

2 Materials and Methods Data source The study was conducted based on data from the World Bank’s database (World Bank, 2020a). The data are mostly from 2015 to 2017, but in some cases, there were used data from earlier years (the oldest poverty rates are from 2011). In the analysis, there were included 102 countries divided into four income groups: low-income, lower-middle-income, upper-middle-income, and high-income groups according to gross national income (GNI) per capita (World Bank, 2020c). The condition that had to be met was the full information about all analyzed poverty rates: measured with national and international poverty lines ($1.9, $3.2, $5.5). The list of countries and the definition of income categories are presented in Table 1. The spatial differentiation of income groups is presented in Fig. 1. It can be seen that in a dataset the majority of high-income countries are from Europe (the exceptions are Chile, Panama, Uruguay, and Seychelles), and almost all countries in the low-income group are from Sub-Saharan Africa (the exceptions: Haiti, Tajikistan, Yemen). The lower-middle and upper-middle-income groups consist

252

A. Sa¸czewska-Piotrowska

Table 1 Country classification by income group Income group Low Lower-middle Upper-middle High income

GNI per capita, U.S. dollars (Atlas method) 1025 or less 1026–3995 3996–12,375 12,376 or more

Number of countries Overall Included in the study 37 20 47 33 60 36 80 13

Source: Own processing

Fig. 1 The analyzed countries according to their income groups. Source: Own processing

of countries from different continents, and there is no spatial pattern in group compositions. Statistical analysis The Cramer’s V was used to show the relationships between income groups of the country (low, lower-middle, upper-middle, high) and poverty prevalence. Due to large sample size, commonly used chi-square test was not appropriate in the analysis because in very large samples p-values go quickly to zero (Kim, 2017; Leon-Guerrero & Frankfort-Nachmias, 2014; Lin, Lucas Jr, & Shmueli, 2013). There was used interpreting the meaning of Cramer’s V was interpreted according to Rea and Parker (2014). To measure the strength of association between national and international poverty rates nonparametric tests were used (after testing for normal distribution by ShapiroWilk test) and Spearman’s rank correlations rS were calculated.

Comparison of Methods of Poverty Rates Measurement

253

Table 2 Poverty distribution by income groups

Income group Low Lowermiddle Upper middle High Total

Population (in million) 376.89 2396.24

Poor (in million) National poverty lines $1.9 line 176.45 197.92 515.49 364.30

36

2227.61

272.65

13 102

102.32 5103.06

15.03 979.62

Number of analyzed countries 20 33

$3.2 line 288.78 1108.90

$5.5 line 343.77 1777.47

31.76

160.06

548.52

0.58 594.56

1.24 1558.98

3.96 2673.72

Source: Own processing

To assess the agreement between methods of measurement of poverty rates, the techniques commonly used in medicine and chemistry were applied: Bland-Altman plots and Passing-Bablok regression. In the Bland-Altman plot, the differences between the two methods are plotted against the averages of the two methods (Bland & Altman, 1986, 1999). Horizontal lines are drawn at the mean difference and at the limits of agreement, which are defined as the mean difference ±1.96 times the standard deviation of the differences. The Passing-Bablok regression to detect the systematic difference and the proportional difference between methods was used (Passing & Bablok, 1983). All calculations and plots (Bland-Altman and Passing-Bablok results) were performed with Statistica; the cartogram was performed with MS Excel.

3 Results Conducted analysis showed that almost 980 million people live in poverty defined according to the national lines (Table 2). Measuring poverty with the international poverty lines, there are from 600 million ($1.9 line) to 2680 million ($5.5 line) poor people in the analyzed 102 countries. The prevalence of poverty is varied in income groups (Table 3). According to the national poverty lines, there are between 12.24% (in the upper-middle-income group) and 46.82% (in the low-income group) poor people. It should be noted that poverty prevalence in the high-income group is higher than in the uppermiddle-income group. Association between income group and poverty prevalence is statistically significant, and this relationship is moderate (Cramer’s V = 0.228). Considering international poverty lines, the prevalence of poverty is significantly associated with income group (Tables 4, 5, and 6). The higher is the poverty line, the stronger is this relationship (Cramer’s V from 0.415 to 0.537). It is clearly seen that the higher the poverty rates the poorer the countries are. According to a $1.9 line

254 Table 3 Poverty prevalence in income groups according to the national poverty lines (%)

A. Sa¸czewska-Piotrowska

Income group Low Lower-middle Upper-middle High

Status (%) Poor Nonpoor 46.82 53.18 21.51 78.49 12.24 87.76 14.69 85.31

Cramer’s V 0.228

p-value