Advances in Financial Economics 9781783501212, 9781783501205

Citation preview

ADVANCES IN FINANCIAL ECONOMICS

ADVANCES IN FINANCIAL ECONOMICS Series Editors: Kose John, Anil K. Makhija and Stephen P. Ferris Recent Volumes: Volume 6:

Advances in Financial Economics

Volume 7:

Innovations in Investments and Corporate Finance

Volume 8:

Coporate Government and Finance

Volume 9:

Corporate Governance

Volume 10:

The Rise and Fall of Europe’s New Stock Markets

Volume 11:

Corporate Governance: A Global Perspective

Volume 12:

Issues in Corporate Governance and Finance

Volume 13:

Corporate Governance and Firm Performance

Volume 14:

International Corporate Governance

Volume 15:

Advances in Financial Economics

ADVANCES IN FINANCIAL ECONOMICS VOLUME 16

ADVANCES IN FINANCIAL ECONOMICS EDITED BY

KOSE JOHN Charles William Gerstenberg Professor of Banking and Finance, New York University, USA

ANIL K. MAKHIJA David A. Rismiller Professor of Finance, The Ohio State University, USA

STEPHEN P. FERRIS J.H. Rogers Chair of Money, Credit and Banking Finance, University of Missouri, USA

United Kingdom
North America
Japan India
Malaysia
China

Emerald Group Publishing Limited Howard House, Wagon Lane, Bingley BD16 1WA, UK First edition 2013 Copyright r 2013 Emerald Group Publishing Limited Reprints and permission service Contact: [email protected] No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. Any opinions expressed in the chapters are those of the authors. Whilst Emerald makes every effort to ensure the quality and accuracy of its content, Emerald makes no representation implied or otherwise, as to the chapters’ suitability and application and disclaims any warranties, express or implied, to their use. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-1-78350-120-5 ISSN: 1569-3732 (Series)

ISOQAR certified Management System, awarded to Emerald for adherence to Environmental standard ISO 14001:2004. Certificate Number 1985 ISO 14001

CONTENTS LIST OF CONTRIBUTORS

vii

THE INCREASE IN CEO PAY AFTER LARGE INVESTMENTS: IS IT PURELY RENT EXTRACTION? Zhan Jiang, Kenneth A. Kim and Yilei Zhang

1

WHO CHOOSES BOARD MEMBERS? Ali C. Akyol and Lauren Cohen

43

THE GROWTH OF GLOBAL ETFS AND REGULATORY CHALLENGES Reena Aggarwal and Laura Schofield

77

OVERCONFIDENCE, CORPORATE GOVERNANCE, AND GLOBAL CEO TURNOVER Hyung-Suk Choi, Stephen P. Ferris, Narayanan Jayaraman and Sanjiv Sabherwal

101

HUMAN AND SOCIAL CAPITAL IN THE LABOR MARKET FOR DIRECTORS George D. Cashman, Stuart L. Gillan and Ryan J. Whitby

137

DUAL CLASS DISCOUNT, AND THE CHANNELS OF EXTRACTION OF PRIVATE BENEFITS Ben Amoako-Adu, Vishaal Baulkaran and Brian F. Smith

165

A CRISIS OF INVESTOR CONFIDENCE: CORPORATE GOVERNANCE AND THE IMBALANCE OF POWER Richard L. Wise

217

ABOUT THE AUTHORS

257 v

LIST OF CONTRIBUTORS Reena Aggarwal

McDonough School of Business, Georgetown University, Washington, DC, USA

Ali C. Akyol

Department of Finance, University of Melbourne, Melbourne, Victoria, Australia

Ben Amoako-Adu

Financial Services Research Centre, School of Business and Economics, Wilfrid Laurier University, Waterloo, Ontario, Canada

Vishaal Baulkaran

Faculty of Management, University of Lethbridge, Lethbridge, Alberta, Canada

George D. Cashman

Area of Finance, Rawls College of Business, Texas Tech University, Lubbock, TX, USA

Hyung-Suk Choi

Ewha School of Business, Ewha Womans University, Republic of Korea

Lauren Cohen

Harvard Business School, Boston; National Bureau of Economic Research (NBER), Cambridge, MA, USA

Stephen P. Ferris

Department of Finance, Robert J. Trulaske, Sr. College of Business, University of Missouri, Columbia, MO, USA

Stuart L. Gillan

Department of Finance, Terry College of Business, University of Georgia, Athens, GA, USA

Narayanan Jayaraman

Scheller College of Business, Georgia Institute of Technology, Atlanta, GA, USA

vii

viii

LIST OF CONTRIBUTORS

Zhan Jiang

Shanghai Advanced Institute of Finance (SAIF), Shanghai Jiaotong University, Shanghai, China

Kenneth A. Kim

Renmin Business School, Renmin University of China, Beijing, China; School of Business, State University of New York at Buffalo, Buffalo, NY, USA

Sanjiv Sabherwal

Department of Finance and Real Estate, College of Business, University of Texas at Arlington, Arlington, TX, USA

Laura Schofield

McDonough School of Business, Georgetown University, Washington, DC, USA

Brian F. Smith

Financial Services Research Centre, School of Business and Economics, Wilfrid Laurier University, Waterloo, Ontario, Canada

Ryan J. Whitby

Department of Economics and Finance, Jon M. Huntsman School of Business, Utah State University, Logan, UT, USA

Richard L. Wise

Counsellor at Law

Yilei Zhang

Department of Finance, University of North Dakota, Grand Forks, ND, USA

THE INCREASE IN CEO PAY AFTER LARGE INVESTMENTS: IS IT PURELY RENT EXTRACTION? Zhan Jiang, Kenneth A. Kim and Yilei Zhang ABSTRACT Purpose
The change in CEO pay after their firms make large corporate investments is examined. Whether the change in CEO pay is a beneficial practice or harmful practice to firms is investigated. Design/methodology/approach
A sample of firms that make large corporate investments is identified. For this sample, we identify the change in CEO pay before and after the investment, and we also measure the pay-for-size sensitivity of these investing firms. Findings
For firms that make large corporate investments, CEOs get significantly more option grants when their firms’ stock returns are negative after the investments and these investing CEOs get more option grants than noninvesting CEOs. Research limitations/implications
The present study suggests that firms may have to increase CEO pay after large corporate investments to encourage investment. However, the results may also be consistent with an agency cost explanation. Future research should try to distinguish between the two explanations.

Advances in Financial Economics, Volume 16, 1
42 Copyright r 2013 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3732/doi:10.1108/S1569-3732(2013)0000016001

1

2

ZHAN JIANG ET AL.

Social implications
The study reveals a potential way to prevent CEOs from underinvesting. Originality/value
The study provides important insights to shareholders on how to encourage CEOs to get their firms to invest, and on how to view CEO pay increases after their firms invest. Keywords: Corporate investment; CEO pay; managerial incentives

INTRODUCTION The existing literature finds that a CEO’s compensation generally increases after large investments even when these investments are followed by a stock price decline (Avery, Chevalier, & Schaefer, 1998; Bliss & Rosen, 2001; Harford & Li, 2007). Previous studies predominantly interpret this empirical evidence as an agency problem because managers are insulated from the consequence of poor performance. In previous studies, the board seems to be ineffective preventing this practice. Despite the empirical and popular focus on rent extraction as the only explanation, we offer a more benign co-existing explanation, that is, this increase in CEO pay can in some cases prevent suboptimal investment decisions. In a perfect capital market with no mispricing, the price reaction to the investment decision is entirely a reflection of the net present value (NPV): Price decreases if the investment has a negative NPV and increases if it has a positive one. Therefore, a CEO’s interest is perfectly aligned with shareholders’: the CEO suffers a loss in his or her equity holdings if he or she undertakes negative NPV investments, and vice versa. However, as many studies have pointed out, misvaluation (particularly overvaluation) is often observed prior to the investments.1 Then, the price reaction on the investment event becomes a combination not only of the NPV but also the correction of mispricing. Our study investigates how this imperfect market valuation adversely alters the CEO incentives and induces suboptimal investment decisions, particularly underinvestment. If firms are overvalued prior to a positive NPV investment, the stock price might still drop after the announcement when the correction of overvaluation dominates the positive NPV. As a result, CEO suffers a shortrun loss in his or her existing equity holdings even though the investment has a positive NPV. Then the CEO has incentive to forego this positive

Increase in CEO Pay after Large Investments

3

NPV project, which creates an underinvestment problem. To remedy this potential problem, the board can make two adjustments in the CEO pay. First, the board can grant more compensation after the investment, especially when the negative return is observed. Note that this increase in pay is not rewarding CEO for bad performance but offsetting the CEO’s wealth loss due to the correction of overvaluation. We call this compensation practice asymmetric pay-for-performance sensitivity (PPS thereafter).2 Second, the board can reward the CEO for the increase in firm size regardless of the performance. We call this compensation practice pay-for-size sensitivity (PSS thereafter). Both adjustments compensate for the CEO’s wealth loss and, therefore, reduce underinvestment problems. Not surprisingly, discouraging underinvestment might lead to perverse outcome, namely, overinvestment. If a CEO gets an excessive pay increase, he or she may have incentives to take negative NPV projects resulting in an overinvestment problem. This overinvestment problem, however, can be mitigated if the board chooses to grant equity in lieu of cash. If overinvestment were realized, these stocks and options would be likely to have lower value (or even out of the money for options) in the long run due to negative NPV. We argue that the asymmetric PPS and larger PSS can be a part of an optimal contract, particularly in the case of investments preceded by overvaluations and followed by negative return. We acknowledge that the above predictions can be consistent with other existing explanations.3 We do not attempt to draw an explicit line between optimal adjustment and excessive pay. We only argue that treating the pay increase purely as rent extraction is inappropriate in certain cases. We show that there is a possible rational explanation for the board to allow for the prevalent practice of increasing pay after large investment regardless of performance under certain circumstances. Following the hypotheses developed in our analysis, our empirical study provides support to the optimal contracting argument. We first examine different components of CEO wealth after large investments as opposed to noninvestments. This method enables us to analyze the opportunity benefits and costs of investment decisions under different scenarios. Specifically, we observe a significant wealth loss in the existing equity portfolio, a wellrecognized opportunity cost associated with large investments. However we find that the new option grants to CEOs in the investment firms have lower strike price and higher moneyness in the long-run than those in the noninvestment firms. The reason is simple: without investments, the CEO would have received option grants with overvalued strike price; while the

4

ZHAN JIANG ET AL.

correction of overvaluation following large investments can lower the stock price and strike price. This opportunity gain has not been recognized in similar studies. We further observe more option exercises by CEOs in the investment firms than those in the noninvestment firms prior to the event year, especially in the negative return investment firms. If price decline is anticipated, ex ante option exercises reduce the ex post wealth loss and therefore opportunity cost associated with the investment. Then, we study the two board adjustments corresponding to potential suboptimal investment problem. We decompose CEO incentives into PPS and PSS, Consistent with our prediction, we find a negative PPS of the option grant number when the investment are followed by negative returns but insignificant PPS when followed by positive returns. CEO gets significantly 870 more shares of option grants for every negative 1% of stock return; while for every positive 1% of stock return, CEO gets insignificantly 54 more shares of option grants. In total, CEOs in the negative return investment firms receive an average of 16,510 more shares of option grants. This negative PPS shows that firms grant more options to CEOs after negative returns to compensate their wealth loss. Otherwise CEOs might forego the positive NPV project. Furthermore, we observe a significantly larger PSS in investment firms than noninvestment firms. For every $100 million increase in firm size, investment CEO gets an average of 2,200 shares of option grants but noninvestment CEO only gets 900 shares of option grants. As a result, CEOs in the investment firms receive an average of 12.9 thousand more shares of option grants, 169.7 thousand dollars more in the value of option grants and 275 thousand dollars more in total value of compensation grant due to the increase in firm size. This large positive PSS and asymmetric PPS are consistent with our optimal contracting hypothesis. Since the above evidence can support rent extraction as well, we examine the influence of corporate governance on the CEO incentives. If the rent extraction hypothesis dominates in most cases, better governance should reduce PSS and the asymmetry in PPS. Our results are mixed with different governance measures commonly used in the literature. We do not find the higher CEO power (measure by CEO tenure and CEO board/chair duality) increases PSS, which is inconsistent with the agency explanation. However, we do find some evidence that a strong board decreases the magnitude of PSS and asymmetry of PPS. The mixed results provide support to our argument that these two explanations might co-exist. Our study contributes to the literature of the optimality explanation of prevailing compensation policy. Acharya, Kose, and Sundaram (2000)

Increase in CEO Pay after Large Investments

5

rationalize option repricing, although it has been generally criticized as a reflection of rent extraction. Rajan and Wulf (2006) argue that CEO perks are not purely managerial excess but can serve to enhance management productivity. Similarly,
and unlike previous studies that have focused on the negative side (Bliss & Rossen, 2001)
we propose an optimality explanation for the practice of increasing CEO pay around large investments in some cases. Our study also contributes to the methodology that examines CEO incentives around corporate events. We link misvaluation to CEO incentives4 and, therefore, to their investment decision. Furthermore, we include CEO’s own portfolio adjustment (i.e., option exercise),5 the dynamic moneyness of option grant as well as the quantity, and the change in board compensation policy in our analysis and demonstrate they jointly influence the investment decisions. More importantly, we decompose the incentives into PPS and PSS. Although a positive relation between the executive pay and firm size has been found in general (Bebchuk & Grinstein, 2007; Moeller, Schlingemann, & Stulz, 2004), we are the first to quantitatively link PSS with investment decisions. Our methodologies as described are not limited to large investments and can be applied to many other corporate events in which adjustments to CEO incentives are under consideration. Lastly, by using different measures of corporate governance, we show that the empirical evidence on agency explanation in the previous studies is not conclusive. For example, Grinstein and Hribar (2004) only examine the effect of CEO power on merger and acquisition bonuses, and Harford and Li (2007) rely only on CEO tenure to support agency explanation. We find that with different measures of governance, some support our optimality hypothesis and some are consistent with rent extraction explanation. The chapter proceeds as follows. In the second section, we develop a simple model and hypotheses. In the third section, we describe our sample selection and distribution. In the fourth section, we present our empirical results. The fifth section concludes.

HYPOTHESES DEVELOPMENT Consider a firm that decides to make a large investment at t. Immediately prior to the announcement at time t, the stock price is Pt. If no investment is undertaken, then the stock price stays at Pt at least for a while. On the

6

ZHAN JIANG ET AL.

other hand, if large investment is announced, the stock price changes to (Pt + δ1 + δ2). δ1 is the correction of mispricing6; it is negative (positive) if the firm is overvalued (undervalued). For most cases, during our sample period from 1993 to 2005, δ1 is negative. δ2 is the NPV of the large investment; if it is positive (negative), the investment is value-enhancing (valuedestroying). Managers are aware of the separate value of δ1 and δ2, but the board can only observe the combined effect δ1 + δ2 from the market reaction to the announcement of the investment.

Net Wealth Effect for Managers The CEO net wealth effect after a large investment is calculated as the difference in the CEO’s wealth between the case when the CEO undertakes the large investment and the case when the CEO does not. This definition includes both the opportunity cost and opportunity gain
the losses and benefits that only occur when the manager undertakes the investment. The CEO’s objective is to undertake the investment if his or her net wealth effect is positive. Note that the shareholders’ interest is only dependent on the NPV of the investment. We develop the CEO’s net wealth effect (ΔNWCEO) as follows (derivation provided in Appendix A.1):
 ΔNWCEO = ΔWt0 − ΔWt = Pt × Qs0t
Qst þ ðδ1 þ δ2 Þ × Qs0t þ Qc0t × C½Pt þ δ1 þ δ2 ; X ðPt þ δ1 þ δ2 Þ
Qct × C ½Pt ; X ðPt Þ þ βHoldings;t
1 × ðδ1 þ δ2 Þ þ Cash0t
Casht

ð1Þ

where ΔWt (ΔWt0 ) is the change in CEO wealth without (with) large investment. Pt is the price at time t without large investments, Pt − 1 is the price at time t − 1, one year prior. Qst (Qst0 ) and Qct (Qct0 ) are the number of new stock grants and option grants at time t without (with) investment. Casht (Casht0 ) is cash compensation and includes both salary and bonus without (with) investment. βHoldings,t − 1 measures the change in dollar value of the CEO’s existing equity holdings when the stock price changes by 1 dollar. C[•, X(•)] is the value for new option grants, and X(•) is the strike price, which is close to the prevailing market price at the grant date. For a quick and simple analysis, assume the board grants a constant amount of cash and equity (derivation in Appendix A.2), We find that the net wealth function of managers is measured by (δ1 + δ2) × (Qst + βHoldings,t−1).

Increase in CEO Pay after Large Investments

7

Without mispricing (δ1 = 0), the managers’ interest (i.e., δ2 × (Qst + βHoldings,t−1)) is perfectly aligned with those of shareholders. Managers will take positive NPV investments (δ2 > 0) and reject negative NPV projects (δ2 < 0). However, when δ1 is not equal to zero, a suboptimal investment problem might occur. For example, when δ1 is negative (overvaluation) and the NPV of the investment (δ2) is positive but not positive enough to offset the negative effect of overvaluation (δ1 + δ2 < 0), managers might forgo the investment project. This situation creates an underinvestment problem. On the other hand, even when the NPV of the investment project (δ2) is negative, managers might still take on the project if δ1 is positive enough (undervaluation) to offset δ2. This condition leads to an overinvestment problem.

The Existing Equity Holdings Starting this section we include two additional factors into CEOs’ wealth function: long-run effect and the CEO’s own portfolio adjustments. We examine the wealth change in existing equity holdings in this section and the wealth change in new grants in the section “The New Grants.” First, in the long run, the mispricing will be corrected eventually even without the investment; therefore the short-run opportunity cost due to the correction of mispricing becomes less significant in the long run. Assume without the investment, at future time t + n, market would recognize the fundamental value of the stock and the stock price changes to Pt + δ1 + δ3, where δ3 is the price change from t to t + n due to other exogenous factors that are not correlated with either δ1 or δ2. The difference in the longrun wealth effect for the existing holdings with or without investment is βHoldings,t
1 × δ2 (derivation in Appendices A.3 and A.4), which only depends on the NPV of the investment, not on the mispricing. Second, we include the CEO’s portfolio adjustment. Assume the firm is overvalued (δ1 < 0), the CEO would have had more time (i.e., [t, t + n]) to exercise his or her vested overvalued option holdings without undertaking any investment, which constitutes an opportunity cost. However, if the CEO anticipates the price decline after the investment, she/he can strategically increase the option exercise before the investment. Empirically we cannot directly observe the amount of option that the manager would have exercised if no investments were undertaken. However, it may be proxied by the number of vested option holdings which include both the options vested before the large investment while not yet exercised7 and the options vested right after the large investment during [t, t + n]. Based on

8

ZHAN JIANG ET AL.

the two considerations included, we developed the following empirical predictions. Hypothesis 1a. Managers in the investment firms will strategically increase the option exercise to reduce the opportunity cost associated with investment, especially when they anticipate a negative price reaction afterward (i.e., δ1 + δ2 < 0). Hypothesis 1b. The vested equity holding constitutes opportunity cost to CEOs associated with large investment. The New Grants The incentive effect from the new compensation grants, especially option grants, is much more complicated than that of the existing equity holdings. Many components related to the new grants are not only dependent on the NPV of the project itself (δ2), but also on the misvaluation (δ1) or the combined effect (δ1 + δ2). In this section, we discuss the opportunity costs and gains in the most common cases in which the combined return is negative (i.e., δ1 + δ2 < 0). In the section Potential Suboptimal Investment Decisions, we discuss the suboptimal investment problems in four possible scenarios. We arrange the manager’s wealth function as follows (derivation in Appendix A.5): ΔWt0 þ n − ΔWt þ n = ðcashÞ ðstockÞ ðoptionÞ

  β0cash × rt þ ðδ1 þ δ2 Þ=Pt − 1
βcash × rt þ 
  ðPt þ δ1 þ δ3 Þ × β0s × rt þ ðδ1 þ δ2 Þ=Pt − 1
βs × rt   þ δ2 × β0s × rt þ ðδ1 þ δ2 Þ=Pt-1 þ   β0c × rt þ ðδ1 þ δ2 Þ=Pt − 1 × δ3
βc × rt × δ3
βc × rt × δ1 þ

ðHoldingsÞ βHoldings;t − 1 × δ2 þ ðβExercise;t − 1 þ βUnvested;t − 1 Þ × δ1 þ β0Exercise;t − 1 × ðδ1 þ δ2 Þ

ð2Þ

We notice that managers suffer a reduction in cash compensation and option grants since the market reaction δ1 + δ2 is negative (i.e., β0cash × [rt + (δ1 + δ2)/Pt
1] − βcash × rt < 0 and βc0 × [rt + (δ1 + δ2)/Pt
1] × δ3 − βc × rt × δ3 < 0). At the same time, managers benefit from the decrease in option strike price after the large investment (i.e., βc × rt × δ1). Without the investment, managers are granted new options with overvalued strike price. These options are likely to be out of the money at future time t + n once

Increase in CEO Pay after Large Investments

9

the mispricing is finally corrected, no matter how many of them are granted. By undertaking the large investments, the mispricing is corrected immediately (i.e., at time t); the strike price is close to the fundamental value, which increases the moneyness of these option grants in the long run. In summary, the opportunity gain comes from the increase in moneyness of option grants while the opportunity cost is mainly from the reduction in cash grants. Therefore, we develop following hypotheses. Hypothesis 2a. The increase in long-run moneyness is higher in investment firms than in the noninvestment firms and higher in the investment firms with negative returns than investment firms with positive returns. Hypothesis 2b. Higher option grants relative to cash grants in the compensation structure increase the long-run moneyness and the opportunity gains associated with large investments.

Potential Suboptimal Investment Decisions The effect of mispricing can lead to suboptimal investment decisions, namely, either overinvestment or underinvestment. Overinvestment (underinvestment) occurs when the NPV of the investment is negative (positive) and managers do (do not) undertake it. Based on CEO net wealth function (Eq. (2)) we examine the CEO’s investment decision in the following four scenarios. Case 1. δ2 is positive and δ1 + δ2 is negative
underinvestment is possible In this case, the NPV is positive (δ2 > 0) and the firm is relatively largely overvalued (δ1 < − δ2 < 0). The magnitude of the opportunity cost associated with investment, which can be strikingly significant, comes from four sources: a reduction in cash compensation (β0cash × [rt + (δ1 + δ2)/Pt−1] − βcash × rt < 0), a reduction in the number of shares of stock grants due to a negative price effect δ1 + δ2 ((Pt + δ1 + δ3) × [βs0 ×(rt + (δ1 + δ2)/Pt−1) − βs × rt] < 0), a reduction in the number of shares of option grants (βc0 ×[rt + (δ1 + δ2)/Pt−1] × δ3 − βc × rt × δ3 < 0), and finally, the loss associated with the existing holdings, even after controlling for the option exercises prior to the event. The more negative δ1 + δ2 is, the larger the opportunity cost associated with the investment project is. On the other hand, mechanisms exist to generate opportunity gains. First, the stock grants have a higher value per share due to positive

10

ZHAN JIANG ET AL.

NPV of the investment, as measured by δ2 × βs0 ×[rt + (δ1 + δ2)/Pt
1]. Second, a long-run moneyness effect is found for new option grants as measured by βc × rt × δ1 (i.e., lower strike price and higher moneyness), which increases in the overvaluation. If the overall opportunity cost is bigger than the opportunity gain leading to a negative net wealth, it is rational for managers to reject the investment opportunity. This scenario thus creates an underinvestment problem. Case 2. δ2 is positive and δ1 + δ2 is positive
suboptimal investments are unlikely In this case, the NPV is positive and the firm is undervalued (δ1 > 0) or mildly overvalued (−δ2 < δ1 0 > −δ2). Managers benefit from an increase in cash grants and the number of shares of stock grants due to a positive combined price effect. At the same time, they suffer from a lower value for

Increase in CEO Pay after Large Investments

11

each share of stock grants in the long run. The net effect can, therefore, be either positive or negative. However, firms have more difficulty taking advantage of undervaluation (i.e., share repurchases) while making large investments, while overvalued firms are more likely to finance with overvalued equity and therefore reveals the overvaluation to the market surrounding large investments.

Board’s Adjustment in Pay Policy As previously discussed, shareholders and the board can differentiate Cases 1 and 3 from Cases 2 and 4, but they cannot differentiate Case 1 from Case 3 or Case 2 from Case 4. Among these four cases, we notice that all cases can possibly lead to overinvestment or underinvestment. However, case 1 is most likely to occur and therefore drives the adjustment of board compensation policy. Pay-For-Performance Sensitivity (PPS) Adjustment To avoid underinvestment in Case 1 (i.e., δ2 > 0 and δ1 + δ2 < 0), the board can adjust PPSs (i.e., βc0 , βs0 , and β0cash) to encourage value-enhancing investments. Because option grants are usually more frequent and larger than the restricted stock grants, we focus on the adjustment of βc. Without the adjustment (i.e., βc0 = βc), the net wealth effect of option grants βc × (δ1 + δ2)/Pt
1 is negative: Managers are punished because the stock price decline (i.e., δ1 + δ2 < 0) even when the investment has positive NPV (i.e., δ2 > 0). By reducing PPS βc0 , the board can reduce the opportunity cost to managers. Under some scenario, when rt + (δ1 + δ2)/Pt
1 is negative, the PPS βc0 has to be negative to discourage underinvestment. We develop the following hypothesis. Hypothesis 3a. The board will grant more options to managers in investment firms than in noninvestment firms, and more in investment firms with negative returns than in investment firms with positive returns. Hypothesis 3b. PPS in option grants is lower in investment firms than in noninvestment firms and is lower (or negative) in investment firms with negative returns than investment firms with positive returns. Although the adjustment is an ex post decision, it provides an ex ante incentive to prevent underinvestment. The logic is similar to the rent

12

ZHAN JIANG ET AL.

extraction hypothesis in the previous literature, when CEOs know that they won’t be punished ex post or even will be rewarded ex post, they have ex ante incentives to engage in negative NPV acquisitions. We argue when CEOs know that they would not be punished ex post due to correction of overvaluation, they would not have ex ante incentives to skip positive NPV investments. Pay-For-Size Sensitivity (PSS) Adjustment Although extensive literature links firm size and CEO pay, very few papers have directly tested the magnitude of the pay-for-size incentive in CEO compensation structure and how this pay-for-size incentive influences large investment decisions. The size change after large investment is usually significant. This, together with a larger positive PSS can provide tremendous opportunity gain to offset the negative price effect due to correction of overvaluation. We develop the following hypothesis. Hypothesis 4. The PSS is significantly positive and larger in the investment firms than noninvestment firms, especially in the firms with negative returns after a large investment.

Other Related Factors A large amount of literature suggests that managers are more likely to invest during a period of overvaluation through equity financing (Dong et al., 2006). Such practices can cause the board to be concerned that managers are overinvesting. However, taking advantage of the overvalued equity to finance investments benefits the current shareholders (Savor & Lu, 2009; Shleifer & Vishny, 2003). Thus, although overinvestment is a possibility, investment based on overvaluation should not be considered, a priori, overinvestment. Clearly, behavior biases such as hubris and CEO self-serving motives (Billett & Qian, 2008; Harford & Li, 2007; Malmendier & Tate, 2005; Roll, 1986) can also lead to overinvestment: negative NPV and negative return assuming no misvaluation. Contributing to the existing studies, our study points out that increasing CEO pay can be optimal in some cases: positive NPV but negative return due to correction of overvaluation. Our study provides a possible explanation for why the board and the shareholders would always allow for what appears to be a pure rent-extracting behavior.

Increase in CEO Pay after Large Investments

13

It is because the board cannot directly observe NPV yet they want to underinvestment.

SAMPLE FORMATION Our sample is comprised of two major types of investments over the sample period from 1993 to 2005: external investments (i.e., acquisitions) and internal investments (i.e., capital expenditures or research and development). We collect our sample of acquisitions from the Securities Data Company’s (SDC) U.S. Merger and Acquisition Database. We require the investment value to be at least 10% of the firm’s prior year-end total assets. Because the decision to make a large internal investment is inherently a lumpy one, we follow existing literature (Elsas, Flannery, & Garfinkel, 2006; Titman et al., 2004) and employ a discrete measure to construct the internal investment sample. We require that the total investment (sum of capital expenditure and research and development) in any fiscal year be 100% larger than the previous three-year average and that the investment be at least 10% of the firm’s prior year-end total assets. In addition, both types of investment firms must have available stock prices from the Center for Research and Security Prices, accounting data from Compustat, and executive compensation data from ExecuComp to be included in the sample. These screening criterions lead to 4,154 investment events on 1,663 firms. In addition, we categorize the event sample into three groups: pure internal investment, pure external investment (acquisition), and a mixed group. In the mixed group, firms have both acquisitions and large internal investments within one fiscal year. Our sample consists of 2,239 pure internal investments, 1,349 pure acquisitions, and 567 mixed events. Table 1 displays the sample distribution across years (Panel A) and industries (Panel B). We find that our investment sample clusters during the Internet bubble period from 1997 to 2000. In addition, the highest number of large investments is in the business services, computers, electronic equipment, and retail sectors. Therefore, we control for industry and year effects in all regressions. Note that both the acquisition and internal investment groups show similar year and industry distribution. Table 2 compares CEO compensation and firm characteristics between investment and noninvestment firms. The noninvestment sample includes all firms from ExecuComp with no large investments over the sample period from 1993 to 2005. They match our event sample by industry (two-digit

4,154

Total

8 2 14

32 5

Alcoholic Beverages

External investment

Aircraft Agriculture Automobiles and trucks Banking

Industry name

0 1

3 9 54

Internal investment

221 354 378 391 420 465 405 405 304 187 188 226 210

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Panel B: Sample Distribution by Industry

All investment

Year

Panel A: Sample Distribution by Year

33 6

12 13 72

Machinery Restaurants, hotel, motel

Consumer goods Insurance Measuring and control equipment

Industry name

1,348

47 77 92 104 130 137 155 123 106 100 101 96 80

External investment

Sample Distribution.

All investment

Table 1.

64 14

24 58 34

Acquisition

57 73

18 11 40

Internal investment

138 95

44 72 94

All investment

2,239

151 241 238 237 224 232 185 205 160 75 75 108 108

Internal investment

14 ZHAN JIANG ET AL.

8 7 225 38 279 16 42 4 135 148 10 173 11 17 14 38 23 1 34

36

3 190 34 80

21 44 0 49 39

17 64

3 32 39 17 2 0 42

15 51 54 72 25 2 87

29 287

39 98 4 238 215

10 531 83 436

46

16 9 51 48

Transportation Textiles Utility Wholesale

1,348

2 1 26 38 11

Total

4

14

2,239

102 13 82 19

0 3 47 62 15

20 39 25 0 234 4

14 35 8 4 66 7

Miscellaneous Business supplies Personal Services Real Estate Retail Rubber and plastic products Shipbuilding, railroad equipment Tobacco products Candy and soda Steel works, etc. Telecommunications Recreational products

12

69

4

27

Nonmetallic mining

Medical equipment

5,851

122 24 148 71

2 4 81 117 29

8

37 75 43 4 323 13

18

124

Note: The table presents the distribution of the large investment firms, external and internal investment firms over the sample period of 1993
2005 by year (Panel A) and industry (Panel B). The SDC and Compustat data sets are used to identify external and internal investments, respectively. We identify the external investment (acquisition) event by requiring that the acquisition value accounts for more than 10% of the current total assets. We identify the internal investment by requiring capital expenditure plus research and development expenditure to be 100% larger than the previous three years average and accounts for more than 10% of the current total assets. The large investment firms include all acquisition and internal investment firms with valid data. The table also reports the distribution of pure acquisition and pure internal investment firms (i.e., the firms either have acquisition or large internal investment but not both in a fiscal year). The industry classification follows Fama and French (1997).

Printing and publishing Shipping containers Business service Chemicals Electronic equipment Apparel Construction Coal Computers Pharmaceutical products Electrical equipment Petroleum and natural gas Fabricated products Trading Food products Entertainment Precious metals Defense Healthcare

Increase in CEO Pay after Large Investments 15

16

ZHAN JIANG ET AL.

Table 2.

Descriptive Statistics: CEO Firm-Related Wealth and Firm Characteristics around Large Investment.

Variables

A. CEO firm-related wealth Current compensation ($thou) Cash compensation ($thou) Option grants ($thou) Option to cash ratio Vested to total option holding B. Firm characteristics Change in total assets ($mil) Change in sales ($mil) Change in market cap ($mil) Percentage change in total asset (%) Market leverage Market to book Free cash flow

Mean

Median

Investment Firms

Noninvestment Firms

Difference

Investment Firms

Noninvestment Firms

Difference

4,128

4,044

84

2,238

2,279


41**

1,175

1,441


266***

875

1,138


263***

2,106 0.41 0.58

1,488 0.30 0.61

618*** 0.11***
0.03***

794 0.42 0.59

502 0.28 0.63

292*** 0.14***
0.04***

585

547

38

142

52

90***

57 579

40 373

17*** 206***

19 156

10 71

9*** 85***

34.35

7.42

26.93***

24.82

3.67

21.15***

0.10 4.24 0.11

0.18 2.57 0.07


0.08*** 1.67*** 0.04***

0.06 3.25 0.11

0.16 1.94 0.06


0.10*** 1.31*** 0.05***

Note: The sample of large investment firms is constructed as in Table 1. The noninvestment firms include all ExecuComp firms with no significant investment from 1993 to 2005. Asterisks on means and medians in the noninvestment firm columns indicate they are significantly different from the corresponding means and medians in the investment columns. The difference in means t-test assumes unequal variances across groups when a test of equal variances is rejected at the 10% level. The significance level of the difference in medians is based on a Wilcoxon sum-rank test. All dollar values are in 2004 dollars. Current compensation includes all compensation grants in the event year (year 0). Cash compensation is CEO salary plus bonus in the event year (year 0). Option grants is the Black
Scholes value of CEO option grant. Option to cash ratio is the ratio of Black
Scholes value of option grant relative to the sum of option grant and cash compensation. Vested to total option holding is the number of shares under exercisable options scaled by CEO’s total option holdings. Change in total assets, change in sales, and change in market capitalization are the changes in firm’s total assets (TA), sales and market capitalization (MktCap), respectively from year
1 to year 0. The following variables are measured at the prior fiscal year end: Market leverage = book value of debt/(book value of debt + market value of equity); market to book = (market value of equity + book value of debt)/total assets; free cash flow = (operating income
interest
dividend
tax)/TA. All dollar values are in 2004 constant dollars. All variables are winsorized at 5% and 95%. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

SIC code) and year, which leads to a total of 5,851 noninvestment observations on 975 firms.8 All variables are winsorized at 5% and 95% to exclude outliers. All dollar values are in 2004 constant dollars. Both mean and median values are reported. We observe that although CEOs in investment firms receive lower total compensation grants and, particularly, cash compensation, they receive much larger option grants. In fact, it is not surprising that the ratio of

Increase in CEO Pay after Large Investments

17

option grants to cash compensation is significantly higher in investment firms than noninvestment firms. This higher ratio in investment firm is consistent with Hypothesis 2b. Further, we show that the ratio of vested equity holdings relative to total option holdings is lower in investment firms. Because the vested equity holdings represent opportunity cost to investment firm CEOs, lower vested holdings imply lower opportunity costs. This univariate result is consistent with our Hypothesis 1b that CEOs with lower opportunity costs are more likely to invest. Not surprisingly, the change in firm size is more pronounced in the investment firms, whether the firm size is measured by the total assets, sales, or the market value of total assets. The size increase in the noninvestment firms is also significant although much smaller than in the investment firms
partly because the noninvestment firms from ExecuComp are larger than our sample firms. The mean (median) percentage size change is 7.4% (3.67%) for noninvestment firms and 34.4% (24.8%) for the investment firms. We also report firm characteristic variables used as control variables in regressions. Our result show that (a) investment firms are less leveraged, indicating they rely more on equity financing; (b) they have significantly higher market-to-book ratio prior to expansion, suggesting the possibility of overvaluation; and (c) they have more free cash flows within the firm, consistent with the literature (Jensen, 1986; Lang, Stulz, & Walkling, 1991).

EMPIRICAL ANALYSES Opportunity Cost and Gain of CEO Wealth around Large Investment To determine the opportunity costs and gains, we examine the change in CEO wealth surrounding large investments from year −1 to year 2. Note that both the investment event and compensation grant can occur any time during a specific fiscal year, which means the compensation can be granted prior to or after the investment in year 0.9 Therefore, measurement error will occur if we only look at the wealth change during the event year (i.e., year 0). Furthermore, Harford and Li state, “Both equity-based compensation and managers’ own career tenure considerations suggest that a long-term perspective is required to fully evaluate the success of a specific merger deal” (2007, p. 920, para. 2). In addition, it may take time for the

18

ZHAN JIANG ET AL.

market to fully realize the true NPV of the investment. Therefore, we allow a longer window over year (−1, +2) to capture the changes in CEO wealth.10 We first compare the change in CEO wealth between investment firms and noninvestment firms. Then, we split the investment sample based on the post-event stock returns. This further division corresponds to our four cases as defined in the section “Potential Suboptimal Investment Decisions”: The negative return group corresponds to Case 1 and Case 3 (δ1 + δ2 < 0), and the positive return group corresponds to Case 2 and Case 4 (δ1 + δ2 > 0). To capture the different components of opportunity costs and gains, we analyze (a) the wealth change in the existing portfolio due to price change; (b) the change in new grants, including both cash and equity grants; (c) the option exercise prior to the investments; and (d) the change in long-run moneyness of CEO option grants.11 Investment Firms vs. Noninvestment Firms In Table 3, Panel A, we show that investment firms experience a mean (median) −9.14% (−19.44%) return, significantly lower than the noninvestment firms.12 Among all investment firms, 64% of them firms experience negative returns, while 52% of the noninvestment firms also experience negative returns. The stock price decline leads to two direct losses: first, the loss in existing stock holdings, calculated as the number of shares of stock holdings in years −1 × (P2−P1), where P2 and P1 are the fiscal year-end stock price in year 2 and year −1, respectively; second, the loss in the existing option holdings (vested and unvested), calculated as the number of options in year −1 × [max(0, P2−X) − max (0, P1−X)], where X is the average strike price of the option holdings. Our results show that both components of CEO existing portfolio experience a significantly larger loss in the investment firms than in the noninvestment firms. Note this change is entirely due to the price change because we fix the number of stocks and options at year −1. Different from the existing equity portfolio, the change in new compensation grants over year (−1, +2), including cash and options, is generally positive. We find that the increase in the cash compensation is lower in the investment firms than noninvestment firms, but the increase in option grants is significantly larger in the investment firms, both in terms of grant number and grant value. Consistent with our previous discussion, cash compensation does not have a moneyness effect and could even enhance the potential for overinvestment problems. Hence, option, instead of cash, should be used as the main mechanism to curb the underinvestment problem. Finally, the total current compensation increases significantly in

800.4

1,124.7

44 −0.01 −0.01

35

578.1

31.6

68.9

1,329.7

293.4

669.4

−1,308.4

−2,211.9

175.7

−2,074.6

−3,047.1

136.5

−2,525.2

−7,759.6

Long-run moneyness change of CEO option grants Change in moneyness from −1 to 2 0.06 Change in moneyness from −1 to 3 0.09

The change in option exercise Change in the value of option exercise ($thou) % positive exercising

The change in CEO new grants Change in cash compensation−1→2 ($thou) Change in the value of option grants−1→2 ($ thou) Change in the number of option grants−1→2 (thou) Change in all new grants−1→2 ($thou)

Change in existing stock holdings
1→2 ($thou) Change in vested option holdings
1→2 ($thou) Change in unvested option holdings−1→2 ($thou)

3.91 52

Noninvestment firms

−9.14 64

Investment firms

0.07*** 0.10***

9***

751.6***

324.3***

37.3***

376.0***

−39.2***

−903.5***

−972.5***

−5,234.4***

−13.05*** 12***

Diff: Invt
Noninvt

Mean

18.2

+

+ +

−

+

130.3

+

0.03 0.07

0.00

336.4

77.4

−244.4

−299.6

−318.6

−19.44

Investment firms

−

−

−

−

− +

Predicted Sign of Diff

−0.02 −0.001

0.00

289.0

0.0

6.3

95.7

−31.8

−73.2

−5.6

−2.45

Noninvestment firms

Median

0.05*** 0.07***

0.00***

47.4**

18.2***

124.1***

−18.3***

−212.6***

−226.4***

−313.0***

−16.99***

Diff: Invt
NonInvt

Opportunity Cost and Gain of CEO Wealth: Investment vs. Noninvestment Firms.

The change in CEO existing portfolio Stock return
1→2 (%) % of negative return

Variables

Table 3. Increase in CEO Pay after Large Investments 19

0.08 0.02

Investment firms −0.03 −0.07

Noninvestment firms 0.11*** 0.09***

Diff: Invt
Noninvt

Mean

+ +

Predicted Sign of Diff 0.18 0.12

Investment firms 0.03 −0.01

Noninvestment firms

Median

0.15*** 0.13***

Diff: Invt
NonInvt

Note: The table presents the changes in the mean and median CEO compensation and CEO option exercise from before to after the investment year for investment versus noninvestment firms. Except otherwise specified, the evolution is measured over year −1 to +2. The table reports the wealth change in the CEO’s existing portfolio and the changes in the CEO compensation grant. First, change in stock price−1→2 and Stock return−1→2 are the change of stock price and stock return over year [ − 1, +2]. % negative return is the number of firms with negative Stock return as a percentage of the total number of firms. Change in existing stock holdings−1→2 is calculated as the number of shares in years −1 × (P2 − P1), where P2 and P1 are the fiscal year end price in year 2 and year −1, respectively. Change in vested option holdings−1→2 = the number of vested option holdings in year−1 × [max(0, P2 − Xv) − max(0, P1 − Xv)], where Xv is the average exercise price of vested options in year −1. Change in unvested option holdings−1→2 = the number of unvested option holdings in year −1 × [max(0, P2 − Xu) − max(0, P1 − Xu)], where Xu is the average exercise price of unvested options. Second, change in cash compensation−1→2, change in the value of option grants−1→2, change in the number of option grants−1→2, and change in current compensation−1→2 are changes of cash compensation, option grant number, option grant value, and total current compensation, respectively, over year −1 to year 2. The table also reports change in the value of option exercise = (option exercise in year −4 and −3) − (option exercise in year −2 and −1). % positive exercising is the number of firms with positive change in the value of option exercise as a percentage of the total number of firms. Finally, the table reports the long-run moneyness change of CEO option grants. Long-run moneyness is defined as the ratio of stock price at the end of year 5 after investment to the strike price of option grant in each year up to year 5. Change in moneyness from −1 to 2 is defined as long-run moneyness in year 2 minus long-run moneyness in year −1. Similarly, change in moneyness from −1 to 3, 4, 5 are defined as long-run moneyness in year 3, 4, 5 minus long-run moneyness in year −1. All variables are winsorized at 5% and 95% percentile. The significance for a mean change in the variable is based on a t-test. The significance for a median change in the variable is based on Wilcoxon signed rank test. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

Change in moneyness from −1 to 4 Change in moneyness from −1 to 5

Variables

Table 3. (Continued ) 20 ZHAN JIANG ET AL.

Increase in CEO Pay after Large Investments

21

investment firms, indicating that the noncash compensation component is the major source of wealth gain among investment firm CEOs. Panel A further shows an average increase in option exercises is $1.329 million in the investment firms, significantly higher than an average of $0.578 million in the noninvestment firms.13 Because the median option exercise is zero for both groups, we examine the percentage of CEOs who have increased their option exercise prior to the event year. Our result shows that 44% of CEOs in investment firms have, 9% higher than CEOs in noninvestment firms, significant at 1% level. The evidence is consistent with our Hypothesis 1a, a CEO can expedite exercising his or her vested inthe-money options prior to the investment to reduce the opportunity costs associated with large investments. We calculate the increase as the difference between the total option exercises during year −4 and −3 and the total option exercises during year −2 and −1. Because the CEO option exercise is usually lumpy, this method provides a better way than a single-year increase (i.e., from year −2 to year −1) to capture the abnormal increase. Next, we document an important finding on the opportunity gain associated with the large investment: a higher long-run moneyness in the investment firms than noninvestment firms. A decrease in stock price implies a lower strike price of option grants and higher moneyness in the future, since firms generally grant at-the-money options. This effect gains economic significance, considering firms grant a greater number of options after large investments. To explicitly measure this long-run effect, we introduce the variable, long-run moneyness, defined as the ratio of the stock price at the end of year 5 after the investment (P5) to the strike price of option grants in each year up to year 5 (Xt, where t = −1, 0, …, 5). Panel A reports the long-run moneyness change from year −1 (i.e., P5/X
1) to year 2, 3, 4, and 5. The positive change of long-run moneyness in the investment sample indicates that options granted after the investment are more attractive in the long-run than those granted prior to the investment. The positive change is not observed in the noninvestment sample. The above evidence is consistent with our Hypothesis 2a. The Likelihood of Firm Expansion Decisions In table 4, we run a Probit model to directly link the managerial wealth effects to the investment decisions. The dependent variable is a dummy variable that equals to 1 if the investment is undertaken, and zero otherwise. We employ four specifications: (a) the change in option exercise, (b) the change in new grants, (c) the change in long-run moneyness, and (d) all a), b) and c). Independent variables also include the changes in the

22

ZHAN JIANG ET AL.

Table 4.

Likelihood Regression.

Dependent Variable: Likelihood of Investment (1) Change in the value of option exercise

0.021 (3.05)***

(2) 0.020 (2.94)***

(3) 0.004 (0.36)

(4) 0.003 (0.23)

Change in cash compensation−1→2

0.079

−0.004

Change in the value of option grant−1→2

(2.10)** 0.035

(−0.07) 0.055

(3.40)*** Change in moneyness from −1 to 2 Change in existing stock holdings−1→2 Change in vested option holdings−1→2 Change in unvested option holdings−1→2 Log Total assets Market leverage Market to book Free cash flow Observations Pseudo R2

(3.29)*** 0.158 (1.67)*

−0.002 −0.002 −0.004 (−1.42) (−1.76)* (−2.10)** −0.009 −0.009 −0.007 (−1.73)* (−1.82)* (−0.96) −0.003 −0.007 0.011 (−0.38) (−1.05) (1.01) −0.200 −0.211 −0.208 (−12.89)*** (−13.26)*** (−7.81)*** −0.862 −0.848 −1.216 (−4.09)*** (−3.96)*** (−3.18)*** 0.090 0.083 0.089 (8.42)*** (7.69)*** (4.97)*** 5.020 5.098 6.042 (11.48)*** (11.53)*** (6.042)*** 3,738 3,688 1,383 0.20 0.20 0.22

0.188 (1.96)** −0.005 (−2.28)** −0.008 (−1.00) 0.007 (0.66) −0.230 (−8.19)*** −1.088 (−2.79)*** 0.074 (4.09)*** 6.171 (7.90)*** 1,371 0.23

Note: This table reports Probit regressions of investment decisions on the CEO wealth effect. The dependent variable is a dummy variable that equals 1 when there is a large investment, and zero otherwise. The independent variables (change in existing stock holdings−1→2 … change in moneyness from −1 to 2) are defined as in Table 3 and are in millions of dollars (except the moneyness variable). All control variables (log total assets … free cash flow) are measured at the prior fiscal year end and defined as in Table 2. All variables are winsorized at 5% and 95%. All dollar values are in 2004 dollars. All regressions include one-digit SIC industry and year dummies (not reported). The heteroscedasticity-robust z-statistics are reported in parentheses under the estimates. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

existing equity portfolio to control for the opportunity costs, and standard firm characteristics such as size, market-to-book, leverage, free cash flow following literature. All regressions include industry and year effects. In model 1, we show that the coefficient on the option exercise is positive and significant as we expected, since the ex ante option exercise reduces the

Increase in CEO Pay after Large Investments

23

opportunity cost and therefore increases the likelihood of investment decisions. This finding is consistent with Hypothesis 1a. Not surprisingly, model 2 demonstrates that an ex post increase in compensation grant, cash, and, more important, options encourages ex ante investment decisions, consistent with Hypothesis 3a. In model 3, we observe that higher long-run moneyness motivates investments, consistent with Hypothesis 2a. Note that when including a long-run moneyness variable, the sample size significantly decreases, leading to the loss of significance of some variables, such as the change in option exercise. However, the increase in option grant and longrun moneyness effect remain significant. Positive Returns vs. Negative Returns By splitting the investment sample based on the cumulative return over year (−1, +2), we obtain 1,196 positive investment events and 2,113 negative investment events. The comparison between investment firms with positive ex post returns vs. investment firms with negative ex post returns is presented in Table 5. It is not surprising that CEOs in the negative investment firms suffer wealth loss on existing portfolios while CEOs in the positive investment firms experience an increase in wealth. We notice that CEOs in negative investment firms receive a lower increase in cash compensation than CEOs in positive investment firms. It seems to suggest the interest alignment between shareholders and CEOs. However, CEOs in negative investment firms do receive larger numbers of option grants, although the grant value is lower.14 In addition, these new grants have significantly higher long-run moneyness in the future. A higher moneyness effect, together with a larger number of option grants, provides incentives for CEOs to undertake the investment even though price drops due to the correction of the firm’s overvaluation. Consistent with our prediction, CEOs in negative investment firms tend to exercise more options prior to the investment to reduce the opportunity cost. PPS and PSS To examine how the board adjusts the compensation policy, we explicitly decompose CEO incentives into PPS and PSS with the following specification: ΔW = a0 þ a1 × Return þ a2 × ΔSize þ Industry dummy þ Year dummy

ð3Þ

−0.13 −0.15 −0.03 −0.11

−

−29.4***

− − − −

−

−8*** −0.29*** −0.37*** −0.17*** −0.21***

−

−875.5***

1273.2***

+ −

+ + + +

Predicted Sign of Diff

288.0*** 628.5***

94.82*** 24772.3*** 7487.0*** 5796.2***

Diff: POS
NEG

−0.10 −0.10 0.08 −0.01

0.00

784.2

9.5

207.4 298.3

37.92 2330.1 1226.0 901.5

POS

0.10 0.17 0.24 0.21

0.00

119.1

23.1

1.7 49.8

−42.28 −3873.4 −1094.9 −1043.6

NEG

Median

−0.20*** −0.27*** −0.16*** −0.22***

0.00***

655.1***

−13.6***

205.7*** 248.5***

80.2*** 6203.5*** 2320.9*** 1945.1***

Diff: POS
NEG

Note: The table presents the changes in the mean and median CEO compensation and CEO option exercise from before to after the investment year for investment firms with positive returns over year [ − 1, +2] versus investment firms with negative returns. All variables are defined as in Table 3. The significance for a mean change in the variable is based on a t-test. The significance for a median change in the variable is based on Wilcoxon signed rank test. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

0.16 0.22 0.14 0.10

47

39

Long-run moneyness change of CEO option grants Change in moneyness from −1 to 2 Change in moneyness from −1 to 3 Change in moneyness from −1 to 4 Change in moneyness from −1 to 5

663.8

1937.0 1733.6

79.5

50.1

858.1

32.4 441.9

320.4 1,070.4

−43.41 −16695.2 −4875.5 −3823.3

NEG

Change in the value of option exercise ($thou) % positive exercising

The change in option exercise

The change in CEO new grants Change in cash compensation−1→2 ($thou) Change in the value of option grants−1→2 ($ thou) Change in the number of option grants−1→2 (thou) Change in all new grants−1→2 ($thou)

51.41 8027.1 2611.5 1972.9

POS

Mean

Opportunity Cost and Gain of CEO Wealth: Investment Firms with Positive vs. Negative Returns.

The change in CEO existing portfolio Stock return−1→2 (%) Change in existing stock holdings−1→2 ($thou) Change in vested option holdings−1→2 ($thou) Change in unvested option holdings−1→2 ($thou)

Variables

Table 5. 24 ZHAN JIANG ET AL.

Increase in CEO Pay after Large Investments

25

where dependent variable ΔW is the change in CEO compensation including change in CEO cash compensation, the change in option grant number, the change in option grant value, and the change in total current compensation grant over year (−1, +2), from model 1 to model 4 respectively. We realize that the incremental option compensation consists of two factors: first, the change in the numbers of options with respect to return and, second, the change in the unit value of the option grant with respect to return.15 Therefore, we use both the change of the option grant value (total effect) and the change of the option grant numbers (volume effect) as dependent variables. The independent variables include the cumulative stock Return over year (−1, +2) and the post-investment change in the firm size (ΔSize). The coefficient on return (α1) measures the PPS, which is denominated in thousands of dollars increase in CEO grants for each 1% increase in the stock return. The coefficient on size change (α2) measures the PSS, which is denominated in dollars of compensation increase for each $1,000 increase in firm assets. Table 6 reports results for investment versus noninvestment firms and positive investment versus negative investment firms. We test the difference in slope coefficients between subsamples based on standard dummy variable techniques.16 In all cases, we control for industry and year fixed effects and our results are robust to median regressions, which adapt to the right skewness and outliers. Asymmetric PPS First, the results show that the PPS of the incremental option grant number is significantly negative while the PPS of incremental cash compensation is significantly positive. The difference in PPS between investment and noninvestment firms is not significant because of the asymmetric PPS in the investment firms with positive versus negative returns: significantly negative in the negative return investment group but insignificant in the positive return group. For every negative 1% of stock return, CEOs are getting significantly 870 more shares of option grants. For every positive 1% of stock return, CEOs are getting insignificantly 54 more shares of option grants. This evidence supports the asymmetric PPS argument of Hypothesis 3b. The board grants higher number of options to CEOs in negative return firms, because these firms have the potential to underinvest. A negative PPS can mitigate this problem. The PPS of option value is still positive, it means that for every negative 1% of stock return, CEOs are still losing 8,200 dollars in option grants. This is because the per unit value of option grant still decline since the stock price declines. However, as we discussed

0.079 (3.31)*** [2.06]**

1,196 0.06

Change in total assets (PSS effect)

Observations Adjusted R2

2,113 0.05

0.024 (1.57)

4.433 (9.07)***

1,196 0.05

0.017 (2.33)** [ − 1.02]

0.054 (0.37) [3.71]***

POS

3,309 0.04

2,113 0.05

1,196 0.05

0.378 (4.22)*** [0.85]

8.274 (4.46)*** [ − 0.73]

−0.870 (−4.35)*** 0.024 (4.32)***

POS

3,177 0.09

0.291 (5.50)*** [2.76]***

8.202 (9.34)*** [2.59]**

Investment firms

NEG

4,234 0.05

0.009 (3.28)***

−0.271 (−4.36)***

−0.253 (−3.42)*** [0.57] 0.022 (4.98)*** [2.48]**

Noninvestment firms

Investment firms

2,113 0.05

0.247 (3.78)***

9.642 (4.30)***

NEG

4,110 0.07

0.120 (3.43)***

5.478 (7.74)***

Noninvestment firms

Change in the Value of Option Grants
1→+2

Dependent variable Change in the Number of Option Grants
1→+2

1,150 0.08

0.616 (4.18)*** [0.93]

11.742 (4.16)*** [ − 1.93]*

POS

3,177 0.09

0.470 (5.58)*** [1.74]*

14.582 (11.09)*** [0.83]

Investment firms

2,027 0.07

0.390 (3.80)***

19.307 (5.69)***

NEG

4,110 0.07

0.318 (5.28)***

12.224 (10.35)***

Noninvestment firms

Change in Total New Grants
1→+2

Note: This table reports the regressions of the CEO’s firm-related wealth change on the change of firm’s total assets and stock returns. The dependent variables are the change over year (−1, +2) of CEO’s cash compensation, option grant, and total new grants (cash and equity) in thousands of 2004 dollars, and the change in the number of option grants in thousands of numbers. CumRet is the cumulative shareholder returns over (−1, +2). Change in total assets is in millions of 2004 dollars. All regressions include one-digit SIC industry and year dummies (not reported). t-statistics are based on robust standard errors. t-statistics that test the null that the coefficients for a subsample do not differ significantly from zero are in parentheses. t-statistics that test the null that the coefficients for investment firms (positive investment firms) do not differ significantly from the corresponding coefficients for noninvestment firms (negative investment firms) are in brackets. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

2.358 (5.95)*** [ − 3.47]***

NEG

POS

CumRet (PPS effect)

4,234 0.10

0.056 (5.30)***

3.378 (16.63)***

Noninvestment firms

3,309 0.10

0.040 (3.15)*** [ − 0.54]

Change in total assets (PSS effect)

Observation Adjusted R2

3.047 (17.02)*** [ − 1.51]

Investment firms

Change in Cash Compentation−1→+2

Asymmetric PPS and PSS: Investment Firms vs. Noninvestment Firms, Positive Return Firms vs. Negative Return Firms.

CumRet (PPS effect)

Table 6. 26 ZHAN JIANG ET AL.

Increase in CEO Pay after Large Investments

27

before, in the long-run, these options have higher moneyness. Since cash grant is much less sensitive to the both return and size change than the option grant, we argue that incentive comes mainly from the option grant. Larger PSS Second, we find a significantly positive PSS effect for both cash and option compensation. This effect is also identified in both investment and noninvestment firms, and investment firms with either positive or negative returns, which is consistent with Bebchuk and Grinstein’s (2007) size argument. However, except for the PSS of cash compensation, we find the PSS of the incremental option grant (both value and number) and the total compensation grant is significantly larger in investment firms than in noninvestment firms, consistent with our Hypothesis 4. For example, for every million dollar increase in firm size, the incremental option grant in the investment firms increases 0.29 thousand, compared with the noninvestment firms, which only increase 0.12 thousand. Note that the investment firms also have larger increases in firm size. The increase in the incremental option grant due to the average (median) increase in firm size is 169.7 (41.2) thousand dollars for CEOs of investment firms, which is much larger than the 65.6 (6.2) thousand dollars of their noninvestment firm counterparts. Similarly, we observe a larger PSS in option grant number but smaller PSS in cash grants for investment firms followed by negative returns than investment firms followed by positive returns. This finding is consistent with Hypothesis 4; however the difference is not significant. In another word, in investment firms followed by positive returns, we still observe a significant PSS when underinvestment does not exist. This evidence could be interpreted as a manifestation of CEOs’ empire building and rent seeking (Bebchuk & Grinstein, 2007; Harford & Li, 2007) or, alternatively, as reward for CEOs’ talent and skill to manage a larger firm (Gabaix & Landier, 2008).17

The Influence of Governance on PPS and PSS Our analysis thus far provides some theoretical framework and empirical evidence to explain the CEO wealth effect of investment from a rational standpoint. However, the evidence is also consistent with agency problem. To examine whether rent seeking is the only explanation for the above evidence as most previous studies have argued, we investigate the governance

28

ZHAN JIANG ET AL.

structure of the investment firms and the impact of governance on the PPS and PSS. If rent seeking is indeed the only explanation, strong governance should at least reduce the magnitude of the PSS. The results are reported in Table 7. Panel A of Table 7 describes the governance structures of both investment and noninvestment firms and both positive return investment and negative return investment firms. Our results show significant differences in the governance structure between investment firms and noninvestment firms. Specifically, CEOs in investment firms usually have longer tenure, and investment firms tend to have a smaller board and a lower fraction of outside directors. These results suggest a potentially higher degree of CEO entrenchment in investment firms. However, we find a lower percentage of CEOs who are also the chair of the board in investment firms and investment firms have a lower G-index and higher institutional ownership. When investment firms are divided into positive and negative returns, the governance structure is not significantly different. We then expand the PPS/PSS regressions by adding the governance variables interacted with both stock return and firm size change. Table 7, Panel B reports the results using the change of option grant value as the dependent variable on the combined investment and noninvestment firms. We also use the change of cash compensation, option grant number, and total compensation grant as dependent variables. We run regressions separately on investment and noninvestment samples as well as positive return and negative return subsamples. Our main conclusions on the governanceinteraction variables are the same, and we observe few differences across subsamples. In Table 7, the return
governance interaction variables measure the effect of governance on PPS, and the size
governance interaction variables measure the effect of governance on the PSS. Across all specifications and all firm groups, the return
governance interaction variables are generally insignificant; the main result on PPS alone holds. This finding is quite robust, especially considering that the sample size changes across different specifications due to data availability. Next we look at the effect of governance on PSS. The main result on PSS alone is consistently positive and significant across all specifications. The lower G-index (hence higher shareholder rights and better governance) and CEO
board chair duality is correlated with a higher PSS, suggesting PSS is considered optimal by better governance; this evidence is not consistent with agency explanation. At the same time, we find that a higher percentage of outside directors leads to a lower PSS, suggesting stronger governance may help reduce PSS and

8.20 8.79 60.57 8.92 59.82

8.56 8.79 59.70 9.08 59.52

CEO Tenure Board size Outside director (%) G-index InstOwn (%)

6.46 8.00 62.50 9.00 61.77

Median

Positive return investment

6.13 8.00 62.50 9.00 62.61

Median

Investment

1,097 507 507 713 678

1,196

Return × CEO_chair

Return

6.579 (8.17)*** −0.139 (−0.13)

(1)

n

3,816 1,785 1,785 2,388 2,109

4,154

n

Effects of Governance.

8.31 8.87 61.00 8.81** 60.04

62.09

Mean

7.45*** 10.53*** 64.87*** 9.28*** 53.34***

69.29***

Mean

6.46 9.00 63.64 9.00* 63.03

Median

Negative return investment

4.96*** 10.00*** 66.67*** 9.00*** 54.74***

Median

Noninvestment

5.912 (7.21)***

(2)

10.395 (4.11)***

(3)

8.076 (2.52)**

(4)

7.050 (2.91)***

(5)

Dependent variable: Change in the value of option grants−1→+2

Panel B: Effects of Governance on Pay-For-Size and Pay-For-Performance

61.54

CEO is board chair (%)

Mean

60.93

CEO Tenure Board size Outside director (%) G-index InstOwn (%)

Mean

CEO_chair (%)

Variables

Panel A: Governance Structure of the Event Firms

Table 7.

2.004 (1.01)

(6)

1,925 1,056 1,056 1,253 1,155

2,113

n

5,227 2,620 2,620 4,075 2,665

5,851

n

Increase in CEO Pay after Large Investments 29

(Continued )

Change in total assets × InstOwn

Change in total assets × G-index

Change in total assets × Outside director

Change in total assets × Board size

Change in total assets × Tenure

Change in total assets × CEO_Chair

Change in total assets

Return × InstOwn

Return × G-index

Return × Outside director

Return × Board size

Return × Tenure

0.316 (4.65)***
0.148 (
2.02)**

(1)


0.002 (
0.63)

0.221 (5.19)***

0.096 (1.34)

(2)


0.011 (
1.28)

0.343 (3.04)***

−0.386 (
1.53)

(3)


0.897 (
4.18)***

0.801 (5.36)***


2.160 (
0.44)

(4)


0.032 (
2.97)***

0.499 (4.72)***


0.025 (
0.10)

(5)

Dependent variable: Change in the value of option grants−1→+2

Panel B: Effects of Governance on Pay-For-Size and Pay-For-Performance

Table 7.

0.000 (
0.14)

0.071 (2.06)** 0.190 (1.82)*

(6)

30 ZHAN JIANG ET AL.

7,287 0.08

6,551 0.08

3,676 0.09

3,676 0.10

4,953 0.09

3,914 0.07

Note: Panel A presents the governance structure of testing firms. CEO_chair is a dummy variable that equals 1 if CEO is the board chair, and zero otherwise. CEO Tenure is the number of years the CEO holds the position. Board size is the number of board members. Outside director is the fraction of the number of outside directors to the total number of the board. G-index is Gompers, Ishii, and Metrick (2003) governance index. InstOwn is the percentage of firm shares owned by institutions. Panel B reports the regressions of pay-for-size and payfor-performance involving the governance variable interaction terms. The dependent variable is the change of CEO option grant value form year
1 to year +2. Only the coefficients on return, size change, and interaction variables are reported. All regressions include onedigit SIC dummies and year dummies (not reported). The t-statistics in parentheses are based on White’s standard errors. ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

Observations Adjusted R2

Increase in CEO Pay after Large Investments 31

32

ZHAN JIANG ET AL.

mitigate the potential agency problem. In conclusion, by using different measures of corporate governance, we show that the empirical evidence on agency explanation in the previous studies is not conclusive. Because the results are mixed with different measures of governance, we argue that the optimal contracting and agency explanations co-exist to explain the change in compensation after large investment.

Additional Tests and Robustness Check To shed more light on the cases when optimal contracting might have more explanatory power than the agency explanation, we further split the negative return investment firms into high market-to-book (more likely to be overvalued) and low market-to-book groups. Our theory argues that more overvaluation leads to higher probability of suboptimal investment decision. Therefore the board adjustment in incentive structure should be more in the overvalued firms. We still observe the asymmetric PPS: within negative return groups, the PPS of option grant number is significantly negative in both high and low market-to-book subgroups; while within positive return groups, PPS is generally insignificant. We further find that the negative return/high market-to-book group has the highest positive PSS in option grant number, which is more than doubled than that in the negative return/low market-to-book group. Since the negative return/high marketto-book group is most likely to be over-valued, this group gets the most compensation from both PPS and PSS adjustments. This evidence is generally consistent with our model analysis. We further investigate the PPS and PSS of the investment firms prior to the events (i.e., over years (−4, −1) and find that the magnitude of both PPS and PSS is much smaller. This time-series evidence indicates that our previous results on PPS and PSS are directly related to the large investment event. Moreover, we examine the positive return and negative return groups of both investment and noninvestment firms and find that (a) for the PPS of the option grant number has the following order: PPS in negative investment firms < PPS in negative noninvestment firms < PPS in positive noninvestment firms < PPS in positive investment firms, and (b) the PSS is always higher in the investment firms prior to the investment decision. These findings strengthen our results. As a robustness check, we employ alternative measures for size, including sales and market capitalization. Our results hold for all the robustness tests. Because previous studies assume that the PPS is the only incentive

Increase in CEO Pay after Large Investments

33

from wealth, we contribute to the literature by showing that the investment decisions should depend on the combined effects of size increases, the return, PPS, and PSS.

CONCLUSION We provide an optimal contracting explanation for the large increase in CEO pay after investments. Previous studies have explained this empirical observation as evidence of managerial rent extraction. In our framework, the investment decisions are not only affected by the NPV of the project but also by the market (mis-)valuation. However, the board can find it difficult to differentiate the magnitude of the two components. If managers are rewarded for the combined effect, their utility function is not perfectly in line with shareholders. As a result, a rational manager would choose the project that maximizes his or her own utility function, which could lead to suboptimal investment decisions. To mitigate this problem, our model predicts that in some cases the board will adjust CEO pay after the large investments to achieve optimal results. We present two mechanisms that the board can apply: First, the board can adopt asymmetric PPS, specifically negative PPS in investments which are followed by negative returns, or, second, the board can apply large positive PSS after the investment. Both practices increase the CEO wealth and promote ex ante incentive for optimal investment decisions. In the empirical framework, we report that CEO wealth increases after large investments, consistent with previous studies. We further show that the increase is mainly from option grants and that managers receive even more option grants when the ex post return is negative than when the return is positive. This negative PPS can partially offset the opportunity cost associated with the correction of overvaluation and reduce the likelihood of underinvestment. Similarly, we find a larger positive PSS in the investment firms than noninvestment firms. The empirical findings are consistent with our optimal contracting hypothesis, although they do not exclude possibility of an agency explanation. Unlike previous studies, we use different measures of corporate governance to differentiate the two arguments of optimal contracting and rent extraction. Our mixed results show that the optimal contracting and rent extract hypotheses could possibly coexist. Along with proposing a rational explanation for CEO pay increases after large investments, we contribute to the literature by introducing the

34

ZHAN JIANG ET AL.

opportunity cost approach when analyzing CEO wealth change. In addition, the long-run-moneyness effect and PSS introduced in this chapter contribute to the literature which studies the dynamic features of managerial incentives.

NOTES 1. Polk and Sapienza (2009) point out market misevaluation is related to levels of corporate investment. Dong, Hirshleifer, Richardson, and Teoh (2006) find that misvaluation drives the takeover market. 2. Garvey and Milbourn (2006) first uncover an asymmetric benchmarking in CEO compensation and attribute this asymmetry to agency problem. 3. For example, PSS can be justified as rewarding a CEO’s talents and effort for managing a bigger firm, and yet an excessive pay increase can raise the concern of rent extraction. Our model analysis focuses on wealth effect only and does not include behavioral biases such as overconfidence or career concerns, but we do not rule out this possibility. 4. Forsyth, Teoh, and Zhang (2007) examine whether mispricing of a firm’s stock affects CEO equity-based compensation in a general setting. Datta, IskandarDatta, and Raman (2005) find that the market reacts more negatively to a seasoned equity offering announcement with a high executive equity-based compensation and argue that market perceives high manager-shareholder interest alignment as a clearer signal that the firm is issuing over-valued equity. 5. Evidence exists that insiders’ stock option exercise has informational content (e.g., Carpenter & Remmers, 2001; Huddart & Lang, 2003; Kyriacou & Mase, 2003). The evidence of CEO altering his or her existing holdings through option exercising in anticipation of corporate events has also been reported in the literature. For example, Cline and Fu (2007) find, on average, 1.76% of a firm’s total market capitalization is sold and exercised in the months around the seasoned equity offering. 6. One of the challenges in the research on mergers and acquisition is that announcement return reflects both the expected synergy of the merger and the signaling effect of mispricing. As shown in the literature, firms tend to finance their investment with overvalued equity (Shleifer & Vishny, 2003; Titman, Wei, & Xie, 2004). Therefore, investments can signal overvaluation to investors and lead to price correction; although, based on the argument by Myers and Majluf (1984), it is not impossible to signal the undervaluation if the investments are financed through debt instead of equity. Therefore, our model allows for both overvaluation and undervaluation. For the simplicity of the model, we assume that the mispricing is fully corrected at the investment announcements. 7. Managers can not exercise all of their vested holdings due to explicit or implicit restrictions (Cai & Vijh, 2007). 8. This match is not one-to-one. One firm in our sample might have more than one matching firms. Specifically, starting with ExecuComp dataset, we first exclude all large investment firms, and then the industries (based on two-digit SIC code)

Increase in CEO Pay after Large Investments

35

and years beyond the investment sample scope are excluded. Next, we require the non-investment firms to have available data for total asset, CEO compensation grant, option grant, and total wealth change. If we required one-to-one match, our sample size would be greatly reduced. 9. We cannot obtain the exact date for internal investment nor for the compensation grant from EXECUCOMP. 10. We conduct all the univarite analyses and regressions using (
1, 0) and (
1, +1) windows, and the major results generally hold though with lower significance. The tests can be obtained from the authors. 11. Due to the data requirement over longer window for the long-run moneyness variables, the sample size is reduced. 12. If a reduction in stock price makes the executive option deep out-of-themoney, a firm may reprice the option by resetting the exercising price to make it less or not out-of-the-money or grant a new option to replace old one. This action may lead to a larger option grant in a given year. However, we find 61 or 1.47% of expansion firms have repricing activity in event year, 18 or 1.16% repricing firms in year 1; correspondingly, 46 or 0.79% nonexpanding firms have repricing in an event year, 38 or 0.83% in year 1. Given a small fraction of repricing observations (although the expansion group has a higher percentage), it should not affect the result (deleting the repricing observations does not change our results). More importantly, repricing itself is a way to compensate executives, so including repricing observations is not inconsistent with our hypotheses. 13. A possible concern here is that when managers exercise options, it signals the market that the firm is misvalued prior to the investment. Nevertheless, we believe that CEOs can camouflage their trading behavior to some extent to capitalize on their overvalued equity holdings before the information is fully revealed. In fact, the information content of an option exercise is relatively noisy compared with a large investment because CEOs exercise their option holdings from time to time and, thus, it does not necessarily signal overvaluation. Even if we concede that exercising options may send a negative signal to the market, this possibility only biases against finding support in our empirical tests. 14. The underlying price of each option is much higher in the firms with positive ex post returns, which leads to the higher Black
Schole value of each option. 15. Change in market price changes the option terms (i.e., strike price and moneyness). 16. The dummy variable ID = 1 for investment (positive investment) firms is interacted with all right-hand side variables, and the interaction variables are added as explanatory variables. The coefficients on ID interaction variables estimate the difference in coefficients across the different subsamples. The corresponding t-statistics test the null that the coefficients for the investment (positive investment) firms do not differ significantly from the corresponding coefficients for the noninvestment (negative investment) firms. 17. We read some proxy statement of expansion firms. Some explicitly cite successful expansion as a reason to grant more to CEOs. 18. For the simplicity of notation, we combine the stocks and options for the existing portfolio. The beta for stock is basically the number of shares, but the beta for the option is related to several parameters associated with options, such as the time to maturity, volatility, and exercise price.

36

ZHAN JIANG ET AL.

19. Note that the board cannot change the sensitivity of the CEOs’ existing portfolio holdings βHoldings,t-1, but it can adjust the sensitivities of the new grants. For simplicity, we do not include a constant term, which is the minimum cash or equity grants that managers will receive regardless of the return. We argue that this simplification will not affect our model because this minimum grant is not an opportunity cost.

ACKNOWLEDGMENTS We thank Matt Billett, Jon Garfinkel, Erik Lie, Dave Mauer, John McInnis, Ashish Tiwari, and Anand Vijh for helpful comments and suggestions. We also thank seminar participants at University of Iowa, SUNYBuffalo, Penn State
Great valley, University of New Hampshire, University of Texas at El Paso, and Marquette University. All errors are our responsibility.

REFERENCES Acharya, V. V., Kose, J., & Sundaram, K. (2000). On the optimality of resetting executive stock options. Journal of Financial Economics, 57, 65
101. Avery, C., Chevalier, J. A., & Schaefer, S. (1998). Why do managers undertake acquisitions? An analysis of internal and external rewards for acquisitiveness. Journal of Law, Economics, and Organization, 14, 24
43. Bebchuk, L. A., & Grinstein, Y. (2007). Firm expansion and CEO pay. Harvard Law and Economics Discussion Paper No. 533. Harvard Law School, Cambridge, MA. Billett, M., & Qian, Y. (2008). Are overconfident managers born or made? Evidence of selfattribution bias from frequent acquirers. Management Science, 54, 1037
1051. Bliss, R. T., & Rosen, R. J. (2001). CEO compensation and bank mergers. Journal of Financial Economics, 61, 107
138. Cai, J., & Vijh, A. M. (2007). Incentive effects of stock and option holdings of target and acquirer CEOs. Journal of Finance, 62, 1891
1933. Carpenter, J., & Remmers, B. (2001). Executive stock option exercises and inside information. Journal of Business, 74, 513
534. Cline, B., & Fu, X. (2007). Executive stock option exercise and seasoned equity offerings. EFA Working Paper. Retreived from http://www.fma.org/Orlando/Papers/ExecutiveStockOption ExerciseandSeasonedEquity.pdf. Accessed on September 20, 2009. Datta, S., Iskandar-Datta, M., & Raman, K. (2005). Executive compensation structure and corporate equity financing decisions. Journal of Business, 78, 1859
1890. Dong, M., Hirshleifer, D., Richardson, S., & Teoh, S. H. (2006). Does investor misvaluation drive the takeover market? Journal of Finance, 61, 725
762.

Increase in CEO Pay after Large Investments

37

Elsas, R., Flannery, M. J., & Garfinkel, J. A. (2006). Major investment, firm financing decisions, and long-run performance. Unpublished working paper. Institute of Finance and Banking, Munich School of Management, Munich. Fama, E. F., & French, K. R. (1997). Industry costs of equity. Journal of Financial Economics, 43, 153
193. Forsyth, J., Teoh, S. H., & Zhang, Y. (2007). Misvaluation, CEO equity-based compensation, and corporate governance. Unpublished working paper. Pepperdine Graziadio School of Business, Malibu, CA. Gabaix, X., & Landier, A. (2008). Why has CEO pay increased so much? Quarterly Journal of Economics, 123, 49
100 Garvey, G., & Milbourn, T. (2006). Asymmetric benchmarking in compensation: Executives are rewarded for good luck but not penalized for bad. Journal of Financial Economics, 82, 197
225. Grinstein, Y., & Hribar, P. (2004). CEO compensation and incentives: Evidence from M&A bonuses. Journal of Financial Economics, 73, 119
143. Harford, J., & Li, K. (2007). Decoupling CEO wealth and firm performance: The case of acquiring CEOs. Journal of Finance, 62, 917
949. Huddart, S., & Lang, M. (2003). Information distribution within firms: Evidence from stock option exercises. Journal of Accounting and Economics, 35, 315
344. Jensen, M. C. (1986). Agency costs of free cash flow, corporate finance, and takeovers. The American Economic Review, 76, 323
329. Kyriacou, K., & Mase, B. (2003). The information contained in the exercise of executive stock options. Unpublished working paper. Brunel University, Uxbridge, UK. Lang, L. H. P., Stulz, R. M., & Walkling, R. A. (1991). A test of the free cash flow hypothesis. Journal of Financial Economics, 29, 315
335. Malmendier, U., & Tate, G. (2005). CEO overconfidence and corporate investment. Journal of Finance, 60, 2661
2700. Moeller, S. B., Schlingemann, F. P., & Stulz, R. M. (2004). Firm size and the gains from acquisitions. Journal of Financial Economics, 73, 201
228. Myers, S., & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics, 13, 187
221. Polk, C., & Sapienza, P. (2009). The stock market and corporate investment: A test of catering theory. Review of Financial Studies, 22, 187
217. Rajan, R. G., & Wulf, J. (2006). Are perks purely managerial excess? Journal of Financial Economics, 79, 1
33. Roll, R. (1986). The hubris hypothesis of corporate takeovers. Journal of Business, 59, 197
216. Savor, P. G., & Lu, Q. (2009). Do stock mergers create value for acquirers? Journal of Finance, 64, 1061
1097. Shleifer, A., & Vishny, R. W. (2003). Stock market driven acquisitions. Journal of Financial Economics, 70, 295
311. Titman, S., Wei, K. C. J., & Xie, F. (2004). Capital investments and stock returns. Journal of Financial and Quantitative Analysis, 39, 677
700.

38

ZHAN JIANG ET AL.

APPENDIX A.1. CEO WEALTH FUNCTION
AN OPPORTUNITY APPROACH Let the CEO be the representative manager. The CEO net wealth effect after a large investment is calculated as the difference in the CEO’s wealth between the case when the CEO undertakes the large investment and the case when the CEO does not. We define the change in CEO total wealth from t − 1 to t without large investments as ΔWt = Pt × Qst þ Qct × C½Pt ; X ðPt Þ þ Casht þ βHoldings;t
1 × ðPt
Pt
1 Þ ðA:1:1Þ and with large investments as ΔWt0 = ðPt þ δ1 þ δ2 Þ × Qs0t þ Qc0t × C ½Pt þ δ1 þ δ2 ; X ðPt þ δ1 þ δ2 Þ þ Cash0t þ βHoldings;t
1 × ðPt þ δ1 þ δ2
Pt
1 Þ

ðA:1:2Þ

where Pt is the price at time t without large investments, Pt
1 is the price at time t
1, one year prior. Qst (Qst0 ) and Qct (Qct0 ) are the number of new stock grants and option grants at time t without (with) investment. Casht (Casht0 ) is cash compensation and includes both salary and bonus without (with) investment. Because the board cannot disentangle the correction of misevaluation effect and the NPV effect, the new compensation grant with investment is proportional to δ1 + δ2 instead of δ2. βHoldings,t − 1 measures the change in dollar value of the CEO’s existing equity holdings when the stock price changes by 1 dollar.18 C[•, X(•)] is the value for new option grants, and X(•) is the strike price, which is close to the prevailing market price. The CEO wealth effect with and without new investment projects is different not only because the stock price will change (from Pt to Pt + δ1 + δ2) but also because the board will change the compensation granting policy. To capture these factors, we model the number of new equity grants without investments as Qst = βs × ðPt
Pt
1 Þ=Pt
1 = βs × rt Qct = βc × ðPt
Pt
1 Þ=Pt
1 = βc × rt Casht = βcash × ðPt
Pt
1 Þ=Pt
1 = βcash × rt

ðA:1:3Þ

39

Increase in CEO Pay after Large Investments

and with investments as   Qs0t = β0s × ðPt þ δ1 þ δ2
Pt
1Þ=Pt
1 = β0s × rt þ ðδ1 þ δ2 Þ=Pt
1  Qc0t = β0c × ðPt þ δ1 þ δ2
Pt
1 Þ=Pt
1 = β0c × rt þ ðδ1 þ δ2 Þ=Pt
1  Cash0t = β0cash × ðPt þ δ1 þ δ2
Pt
1 Þ=Pt
1 = β0cash × rt þ ðδ1 þ δ2 Þ=Pt
1 ðA:1:4Þ If the CEO decides to undertake the investment, the board may change the PPSs from βs, βc, βcash to βs0 , βc0 , β0cash. The sensitivities are defined, respectively, as the change in the number of stock grants, option grants, and the change in the dollar value of cash grants if the stock price changes by 1%.19 The CEO’s objective is to undertake the investment if the CEO’s own net wealth effect is positive, that is, when the CEO is better off undertaking the investment project than forgoing it. We develop the CEO’s net wealth effect (ΔNW) as follows:
 ΔNWCEO = ΔWt0
ΔWt = Pt × Qs0t
Qst þ ðδ1 þ δ2 Þ × Qs0t þ Qc0t × C½Pt þ δ1 þ δ2 ; X ðPt þ δ1 þ δ2 Þ
Qct × C ½Pt ; X ðPt Þ þ βHoldings;t − 1 × ðδ1 þ δ2 Þ þ Cash0t − Casht

ðA:1:5Þ

A.2. A QUICK AND SIMPLE ANALYSIS For a quick and simple analysis, if we assume the board grants the same amount of bonus and equity (both stock and options), then Qs0t = Qst, Qc0t = Qct, Cash0 t = Casht. In other words, if the PPS in the new grants each year equals zero (i.e., β0s = βs = 0, β0c = βc = 0 and β0cash = βcash = 0) or the equity grants and cash grants are kept constant each year, then ΔW0t
ΔWt = ðδ1 þ δ2 Þ×Qst þ Qc0t × C½Pt þ δ1 þ δ2 ;X ðPt þ δ1 þ δ2 Þ
Qct × C½Pt ; X ðPt Þþ βHoldings;t −1 × ðδ1 þ δ2 Þ þ Cash0t
Casht
 = ðδ1 þ δ2 Þ× Qst þ βHoldings;t − 1 þ Qct ×C ½Pt þ δ1 þ δ2 ; X ðPt þ δ1 þ δ2 Þ
Qct × C½Pt ; X ðPt Þ ðA:2:1Þ

40

ZHAN JIANG ET AL.

If for the purpose of this quick and simple analysis, we assume that the strike price X(Pt + δ1 + δ2) is equal to the current stock price Pt + δ1 + δ2; similarly, X(Pt) = Pt. This wealth equation is reduced to ΔWt0 − ΔWt = (δ1 + δ2) × (Qst + βHoldings,t − 1).

A.3. THE SHORT-RUN WEALTH EFFECT OF EXISTING HOLDINGS WITHOUT CEO OWN PORTFOLIO ADJUSTMENT Without large investments, wealth changes in the existing equity holdings: ΔWHoldings;t = βHoldings;t − 1 × ðPt
Pt − 1 Þ;

ðA:3:1Þ

With large investments, wealth changes in the existing holdings: 0 ΔWHoldings;t = βHoldings;t − 1 × ðPt þ δ1 þ δ2
Pt − 1 Þ

ðA:3:2Þ

Then the difference in wealth change in the existing holdings is shown as: 0 ΔWHoldings;t
ΔWHoldings ;t = βHoldings;t − 1 × ðδ1 þ δ2 Þ;

ðA3:3Þ

A.4. THE LONG-RUN WEALTH EFFECT OF EXISTING EQUITY HOLDINGS WITH CEO OWN PORTFOLIO ADJUSTMENTS Without large investments, wealth changes in the existing holdings:
 ΔWHoldings; t þ n = βHoldings;t − 1
βExercise;t − 1
βUnvested;t − 1 × ðPt
Pt − 1 þ δ1 þ δ3 Þ þ ðβExercise;t − 1 þ βUnvested;t − 1 Þ × ðPt
Pt − 1 þ δ3 Þ;

ðA:4:1Þ

With large investments, wealth changes in the existing holdings: 0 0 ΔWHoldings; t þ n = ðβHoldings;t − 1
βExercise;t − 1 Þ × ðPt þ δ1 þ δ2 þ δ3
Pt − 1 Þ

þ βExercise;t − 1 × ðPt
Pt − 1 þ δ3 Þ;

ðA:4:2Þ

Increase in CEO Pay after Large Investments

41

Then the difference in wealth change in the existing holdings is shown as: 0 ΔWHoldings; t þ n
ΔWHoldings;t þ n = ðβHoldings;t − 1
βExercise;t − 1
βUnvested;t − 1 Þ × δ2

þ ðβExercise;t − 1 þ βUnvested;t − 1
β0Exercise;t − 1 Þ × ðδ1 þ δ2 Þ = βHoldings;t − 1 × δ2 þ ðβExercise;t − 1 þ βUnvested;t − 1 Þ × δ1 þ β0Exercise;t − 1 × ðδ1 þ δ2 Þ

ðA:4:3Þ

where βExercise,t − 1 is the number of the vested equity holdings that are exercised without large investments. β0Exercise,t − 1 is the number of these option exercises with large investments; β0Exercise,t − 1 < βExercise,t − 1 as CEOs cannot exercise all of their vested holdings due to explicit or implicit restrictions. βUnvested,t − 1 is the number of unvested equity holdings that would have become vested during the time t to t + n if no investment is undertaken.

A.5. THE LONG-RUN WEALTH EFFECT FOR NEW GRANTS
0  0 0 ΔWGrants;t þ n − ΔWGrants;t þ n = ðPt þ δ1 þ δ3 Þ × Qst
Qst þ δ2 × Qst 0 þ Qct × C ½Pt þ δ1 þ δ2 þ δ3 ; X ðPt þ δ1 þ δ2 Þ
Qct × C½Pt þ δ1 þ δ3 ; X
ðPt Þ þ Cash0t
Cash t  = ðPt þ δ1 þ δ3 Þ × β0s × rt þ ðδ1þ δ2 Þ=Pt − 1
βs × rt  þ δ2 × β0s × rt þ ðδ1 þ δ2 Þ=Pt − 1 þ β0c × rt þ ðδ1 þ δ2 Þ=Pt − 1 × δ3
βc × rt × ðδ1 þ δ3 Þ þ β0cash × rt þ ðδ1 þ δ2 Þ=Pt − 1
βcash × rt : ðA:5:1Þ If we simply use the payoff of options to estimate the value of options and assume the strike price X(Pt + δ1 + δ2) is equal to the prevalent price at the grant date (i.e., X(Pt + δ1 + δ2) = Pt + δ1 + δ2), then at time t + 1, C [Pt + δ1 + δ2 + δ3, X(Pt + δ1 + δ2)] is equal to δ3 and C[Pt + δ1 + δ3, X(Pt)] is equal to δ1 + δ3. Eq. (A5.1) can be reduced to:  0
  0 ΔWGrants;t þn = ðPt þδ1 þδ3 Þ× β s × rt þ ðδ1 þδ2 Þ=Pt −1
βs ×rt þn
ΔWGrants;t     0 þδ2 ×β0s × rt þ ðδ1 þδ2 Þ=Pt −1  þβc × rt þ ðδ1 þδ  2 Þ=Pt −1 ×δ3 0
βc ×rt × ð δ1 þ δ3 Þ þβ × r þ ð δ þδ Þ=P ×rt t −1
β
cash  t
0 1 2
cash  0 0 þ β ×r ×rt = ðPt þδ þδ Þ×r × β −β
β ×δ þ β 1 3 t t 3 s c s c cash
β cash 
  þβ0s × ðPt þδ1 þδ3 Þ× ðδ1 þδ Þ=P þδ × r þ ð δ þδ Þ=P t −1 2 t 1 2 t −1  2 þβ0cash × ðδ1
þδ2 Þ=Pt −1  þβ0c × ðδ1 þδ2 Þ=Pt −1
×δ3
βc ×r t ×δ1 
= ðPt þδ1 þδ3 Þ×rt × β0s
βs þ β0c −βc ×rt ×δ3 þ β0cash −βcash ×rt 
  þβ0s × ðδ1 þδ2 Þ2 =Pt −1 þ ðδ1 þδ2 Þ× 1þrt þδ3 =Pt −1 þδ2 ×rt þβ0c ×½ðδ1 þδ2 Þ=Pt −1 ×δ3 
βc ×rt ×δ1 þβ0cash × ðδ1 þδ2 Þ=Pt −1 ðA:5:2Þ

42

ZHAN JIANG ET AL.

A.6. THE LONG-RUN WEALTH EFFECT FOR BOTH NEW GRANTS AND EXISTING EQUITY HOLDINGS


 ΔWt0 þ n
ΔWt þ n = ðPt þ δ1 þ δ3 Þ × rt × β0s
βs þ β0c
βc × rt × δ3 þ β0cash − βcash × rt 
  þ β0s × ðδ1 þ δ2 Þ2 =Pt − 1 þ ðδ1 þ δ2 Þ × 1 þ rt þ δ3 =Pt − 1 þ δ2 × rt þ β0c × ðδ1 þ δ2 Þ=Pt − 1 × δ3
βc × rt × δ1 þ β0cash × ðδ1 þ δ2 Þ=Pt − 1 þ βHoldings;t − 1 × δ2 þ ðβExercise;t − 1 þ βUnvested;t − 1 Þ × δ1 þ β0Exercise;t − 1 × ðδ1 þ δ2 Þ ðA:6:1Þ

Rearranging the terms, we have the following net wealth functions including both long-run effect and CEO own portfolio adjustments ΔW0t þ n
ΔWt þ n =   ðcashÞ β0cash × rt þ ðδ1 þ  δ0 2 Þ=P
t − 1
βcash × rt þ  ðstockÞ ðPt þ δ1 þ δ3 Þ × βs × rt þ ðδ1 þ δ2 Þ=P t − 1
β s × rt  þ δ2 × β0s × rt þ ðδ1 þ δ2 Þ=Pt − 1 þ 0 ðoptionÞ βc × rt þ ðδ1 þ δ2 Þ=Pt − 1 × δ3
βc × rt × δ3
βc × rt × δ1 þ ðHoldingsÞ βHoldings;t − 1 × δ2 þ ðβExercise;t − 1 þ βUnvested;t − 1 Þ × δ1 þ β0Exercise;t − 1 × ðδ1 þ δ2 Þ ðA:6:2Þ

WHO CHOOSES BOARD MEMBERS? Ali C. Akyol and Lauren Cohen ABSTRACT Purpose
To explore the importance of the board of director nomination process (that is, who nominates a given director for a position on the firm’s board) for the voting outcomes, disciplining of management, and overall monitoring quality of the board of directors. Design/methodology/approach
We exploit a recent regulation passed by the US Securities and Exchange Commission (SEC) requiring disclosure of the board nomination process. In particular, we focus on firms’ use of executive search firms versus allowing internal members (often simply the CEO) to nominate new directors to serve on the board of directors. Findings
We show that companies that use search firms to find board members pay their CEOs significantly higher salaries and significantly higher total compensations. Further, companies with search firmidentified independent directors are significantly less likely to fire their CEOs following negative performance. In addition, companies with search firm-identified independent directors are significantly more likely to engage in mergers and acquisitions (M&A) and see abnormally low returns from this M&A activity. We instrument the endogenous choice of using an executive search through the varying geographic distance of companies to executive search firms. Using this instrumental variable framework, we show search firm-identified independent directors’

Advances in Financial Economics, Volume 16, 43
75 Copyright r 2013 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3732/doi:10.1108/S1569-3732(2013)0000016002

43

44

ALI C. AKYOL AND LAUREN COHEN

negative impact on firm performance, consistent with firm behavior and governance consequences we document. Originality/value
Given the recent law passage, we are the first to directly analyze the nomination process, and show a surprisingly large predictive effect of seemingly arm’s-length nominations. This has clear implications for thinking carefully through how independence is defined in the director nomination process. Keywords: Corporate directors; executive search firms; governance; SEC regulation

The principal
agent problem in economics arises because of incomplete and asymmetric information when a principal hires an agent to pursue the principal’s interests, along with the inability to write complete contracts. For example, shareholders, the owners of a firm, hire a CEO to manage the firm on their behalf. Left unmonitored, CEOs may shirk duties and take actions that may not maximize shareholder value. However, shareholders have several tools to monitor CEOs and mechanisms to make sure that they will act in the best interests of shareholders. Perhaps the most important of these mechanisms is the board of directors. Boards are charged with two basic but crucial functions: monitoring and evaluating managers and their decisions (Fama & Jensen, 1983). In these functions, boards ensure that CEOs and other managers act in the shareholders’ best interests. How directors make their way onto board is thus of obvious importance. Evidence on how directors are identified is limited, largely due to the unavailability of data. In 2003, however, the US Securities and Exchange Commission (SEC) passed a regulation requiring companies to explain their director nomination process and to disclose the sources of all new directors (the new regulation became effective in 2004).1 These disclosures permit us to examine board power dynamics more precisely and to investigate under what circumstances firms use insiders versus outsiders to identify their directors. We can then determine whether the sources of board nominees are related to important firm characteristics that affect performance. Most important, we can investigate whether the monitoring incentives, as well as governance ability, of the newly appointed directors depend on the source of their nomination. We focus on companies’ use of executive search firms versus allowing internal members (often simply the CEO) to nominate new directors to

Who Chooses Board Members?

45

serve on the board of directors. We underscore the importance of these “nomination rights,” because under the plurality system used by most US firms, nominees often require only a single “yes” vote to win election to the board, making “nomination” the sole true determinant of board composition. We hand-collect data for S&P 1500 firms starting in the year of the regulatory change, 2004, for the five-year period through 2008. These firms brought on 5,866 new directors over this period, 4,963 of which were independent directors. Of these independent directors, 23.9% were identified through the use of an executive search firm. The only other agents who were not members of the executive team or board but also nominated independent directors were large (block) shareholders, who nominated 8.5% of independent directors. We address a number of first-order governance questions using these unique data. Because we are the first to use these data, we begin by describing both the nomination data provided and the level of detail we can obtain on each director’s nomination process. From there, we move the focus to executive search firms. These executive search firms play a key role in our analysis because they are the dominant form of (potentially) arm’s-length board nomination by firms (that is, not nominated by a related firm-party). In other words, when firms decide to nominate new board members from outside, they generally use executive search firms. We describe what executive search firms are and where they are located in the United States (including their branches). Next, we describe the firm and board characteristics of those firms (and boards) that use executive search firms to nominate directors. Three of the largest executive search firms in our sample are: (i) Korn Ferry, headquartered in California, with 21 offices throughout the United States; (ii) Spencer Stuart, headquartered in Illinois, with 15 US offices; and (iii) Egon Zehnder International, Inc., headquartered in New York, with 9 US offices. These firms seek out and provide a menu of suitable candidates from which the board then chooses to nominate new directors. We find that certain firms are much more likely to use directors identified by executive search firms. They are significantly larger, being twice as likely to be in the S&P 500 Index. They also have significantly lower sales growth and are significantly more likely to pay dividends. They are spread out evenly across all industry sectors. We next study the more interesting question of the governance and value implications of having these search firm-identified independent directors as opposed to directors nominated from within. Companies that use search

46

ALI C. AKYOL AND LAUREN COHEN

firms to identify board members pay their CEOs significantly higher salaries and higher total compensations. Further, companies with such directors are significantly less likely to fire their CEOs following negative performance. In addition, we find that companies with search firmidentified independent directors have significantly lower return on assets (ROA), are significantly more likely to engage in mergers and acquisitions (M&A), and see abnormally low returns from this M&A activity. We also seek to pin down a causal mechanism between directors hired through search firms and these negative firm outcomes. To do this, we uncover an exogenous factor that drives companies’ use of search firms: the varying geographic distance of companies to executive search firm offices. We hypothesize that companies are more likely to employ executive search firms if they are located nearby, which allows companies and search firm to have more meetings and less costly interactions regarding the slate of potential directors. As long as a portion of this geographic proximity is unrelated to characteristics that may drive firm performance, which we argue is very reasonable,2 we can use this in a two-stage least squares (2SLS) instrumental variable framework to identify the causal impact of directors hired through search firms on firm performance. Upon doing this analysis, we find first, in the first stage of the 2SLS, that geographic proximity to a search firm is a large and significant determinant of a company’s decision to use that search firm. Using solely the piece of search firm choice that is based on geographic closeness, we then show that the orthogonal piece is a significant predictor of lower profitability that we document in the chapter. This gives support to the causal interpretation of search firm directors. The remainder of the chapter is organized as follows. Section I describes the setting and related literature. Section II describes the data. Section III presents the main regression results. Section IV examines the instrument and additional results. Section V concludes.

BACKGROUND Boards of directors look after shareholders’ interests by monitoring CEOs and advising them on corporate affairs. The monitoring function is important: in its absence, CEOs may shirk duties and take actions that can reduce shareholder value. Moeller, Schlingemann, and Stulz’s (2005) results emphasize the importance of the board’s monitoring role. They report that

Who Chooses Board Members?

47

management may destroy shareholder value via acquisitions if the firm is overvalued and management has gained discretion due to the firm’s overvaluation. Left unmonitored, CEOs may also increase their perquisite consumption, which may subsequently reduce shareholder value. Yermack (2006) confirms this, finding a negative relationship between the announcements of CEO perquisite consumption and stock returns. Management may also alter the schedule of investment projects at the expense of longterm shareholder value to meet short-term expectations. Graham, Harvey, and Rajgopal (2006) report that managers are willing to delay positive net present value (NPV) projects in order to meet the market’s earnings estimates. We have witnessed epic corporate failures (such as Enron, Tyco, and Worldcom) in the past decade. Boards have been blamed for failing to adequately monitor CEOs in these and other corporate failures. In response to these and other corporate failures, regulatory agencies, such as the SEC, and the stock exchanges have passed regulations that tighten the grip on CEOs and boards to prevent similar corporate failures in the future. The absence of the monitoring function can lead to shareholder value destruction, and how a board performs this function is partly affected by how the board is composed. The literature shows that board composition has value relevance for shareholder value. Byrd and Hickman (1992), for example, find that the market’s reaction to merger announcements by firms with independent boards is less negative, implying that independent directors are safeguarding shareholder value and that CEOs are less likely to engage in valuedestructive acquisitions when boards are independent. Similarly, Brickley, Coles, and Terry (1994) report positive stock price returns to announcements of poison pill adoptions when boards are independent. CEOs are more likely to lose their job because of poor performance when boards are composed mainly of independent directors (Weisbach, 1988). Hermalin and Weisbach (1988) find that independent directors are more likely to be added to boards following poor performance. Rosenstein and Wyatt (1990) report that the market reacts significantly positively to announcements of outside director appointments, implying that board composition is important for shareholder value. Board composition arguably also affects firm performance. The literature, however, mostly finds a weak relationship between board composition and firm performance. For example, the results of Bhagat and Black (2002) do not indicate that greater board independence enhances firm performance. However, recent studies report a positive effect of board independence on firm performance. For example, Knyazeva, Knyazeva, and Masulis (2011), using local director pool as an

48

ALI C. AKYOL AND LAUREN COHEN

instrument, report positive effects of board independence on firm performance, CEO compensation, and CEO turnover. The board is not a homogenous body. Boards may comprise independent directors, gray directors, and inside directors, each class having its own incentives and desired levels of monitoring the CEO. Although an independent board is less likely to be controlled by the CEO, having a board composed of mainly independent directors does not necessarily imply effective CEO monitoring. The literature shows that board independence can be affected by, for example, board interlocks, directors’ connectedness to the CEO through social or educational ties, and the CEO’s involvement in the selection of directors. When a board has two or more members who also share seats on another firm’s board, CEO compensation tends to be higher and CEO turnover probability decreases (Fich & White, 2003). Hwang and Kim (2009) examine directors’ social ties to CEOs and show that including social ties in the definition of director independence significantly reduces board independence. They also find that boards that are not socially and conventionally independent from their CEO tend to reward the CEO with a high compensation package and are less likely to remove the CEO because of poor performance. The board’s independence and its effective monitoring ability can be hampered if the CEO handpicks directors to the board. Although some studies indicate that a friendly board is sometimes optimal (see, for example, Adams & Ferreira, 2007), Shivdasani and Yermack (1999) find that the market’s reaction to announcements of independent director appointments is significantly lower when the CEO is involved in the selection process. Shivdasani and Yermack argue that CEOs reduce boards’ monitoring intensity by being involved in the selection of directors. Every board is composed of individuals with different characteristics and skill sets. If certain directors share similar characteristics, views, or skills, however, then collectively the board’s view or its monitoring ability of the CEO will be affected. Several studies indicate that director types have meaningful effects on how boards monitor their CEOs. For example, Adams and Ferreira (2009) find that female directors allocate more time in monitoring CEOs and that, in well-governed firms, this increased monitoring may negatively affect firm performance. Masulis, Wang, and Xie (2012) study foreign directors in US firms. They find that foreign directors are poor monitors, and they report a negative relationship between the presence of foreign directors on the board and firm performance. Fich and Shivdasani (2006) show that busy directors
those with three or more

Who Chooses Board Members?

49

outside directorships
are not effective monitors: their presence is associated with poor firm performance and corporate governance. Masulis and Mobbs (2011) find positive effects of inside directors with outside directorships on shareholder value. They report that having an inside director who has an outside directorship on the board is associated with better firm performance and acquisitions and a lower likelihood of earnings management. The literature, overall, shows that board composition affects the board’s monitoring ability, which subsequently affects performance. In this respect, it is crucial to understand how directors are recruited to the board. There is limited evidence in the literature on how directors are nominated. This is partly due to the unavailability of data. Studies on director recruitment generally rely on surveys or interviews. For example, O’Neal and Thomas (1995) interview directors and report that the CEO and the chairman are influential in director selection and that personal networks of directors play an important role in identifying suitable candidates. Mace’s (1971) discussion of how directors are selected also reveals the CEO’s influence in the director selection process. Similarly, the survey results of Lorsch and MacIver (1989) illustrate the control that CEOs have in selecting directors. Shivdasani and Yermack (1999) also report results consistent with the idea that CEOs effectively handpick nominees. Our chapter differs from others that examine the selection of directors in that we identify the sources of director nominations and examine how directors recommended by executive search firms affect boards by studying CEO compensation and turnover, M&As, and firm performance. Although there is limited evidence in the literature on how search firms are involved in the director selection process, there is a small but growing literature on intermediaries such as compensation consultants. Dasgupta and Ding (2010) examine how executive search firms affect CEO compensation in a theoretical setting. They argue that the weakening of the firm’s internal labor markets has helped search firms to flourish, which in turn has increased CEO pay. Canyon, Peck, and Sadler (2009), Armstrong, Ittner and Larcker (2010), Cadman, Carter, Hillegeist (2008), and Murphy and Sandino (2010) examine the effect of compensation consultants and report that their use is associated with higher CEO pay. Rajgopal, Taylor, and Venkatachalam (2012) also report that the use of compensation consultants is associated with higher CEO pay. However, they show that firms that use compensation consultants report better operating and stock price performance. Their results contrast with that of Jensen, Murphy, and Wruck (2004), who argue that CEOs’ use of compensation consultants is an attempt to extract rents from shareholders.

50

ALI C. AKYOL AND LAUREN COHEN

Although the literature shows that CEOs remain influential in the selection of directors, their apparent involvement has been declining owing to recent governance reforms. Shivdasani and Yermack (1999) assume that CEOs are involved in the director selection process if they are on the nominating committee, if the firm has one. If there is no nominating committee, then they assume that CEO is involved in director selection by default. CEOs are no longer allowed to sit on the nominating committee, and almost every firm in our sample now has a nominating committee, making it impossible to determine CEO involvement using Shivdasani and Yermack’s methodology. The 2003 rule change, however, requires firms to describe how they identify board nominees, allowing us to determine the source of director nomination. Our study is related to papers that examine director types and how director types are associated with certain board outcomes. In this respect, we identify directors recommended by executive search firms as another director type and investigate how their presence on the board affects the CEO and the firm. Our chapter is complementary to studies that examine board independence and director characteristics and stresses the need to identify how directors are nominated to the board to determine their monitoring incentives of the CEO. We also contribute to the literature by examining how firms identify their directors after the recent corporate governance reforms. Although executive search firms have been involved in director selection for more than 50 years (Dasgupta & Ding, 2010), their involvement in this process is mostly overlooked in academic studies. We try to fill this gap in the literature and highlight the importance of intermediaries in identifying suitable board candidates.

DATA In this section, we describe our data collection process and variables and provide summary statistics. The data in this study are collected from several sources. Our primary data on directors recommended by executive search firms are hand-collected from proxy statements. On November 10, 2003, the SEC adopted a rule requiring firms to disclose more information about their director nomination process. The new disclosure requirements became effective on January 1, 2004. Because we need information on directors located by search firms, our sample period starts on January 1, 2004. To collect data on directors identified by search firms, we start with the 2006 list of S&P 1500 firms and follow our sample

Who Chooses Board Members?

51

firms from 2004 to the end of 2008. We choose the 2006 list of S&P 1500 firms because it is in the middle of our sample period, which ensures that we do not lose many firms in the sample owing to bankruptcies or mergers. Using proxy statements, we identify who recommends a director to the board or to the nominating committee. The sources of director nominations are diverse and include search firms, the CEO, the chairman, other executives of the firm, major shareholders, independent directors, and the nominating committee. We specifically look for a new director appointment (or nomination) in proxy statements. We also compare proxy statements in a given year to the previous year’s proxies to identify midyear director appointments. The information about the source of director nomination is provided either in the proxy in a special part about the firm’s director nomination process or in director biographies. Below we provide an example from Advance Auto Parts, Inc.’s 2004 proxy statement for the nomination of its new independent director, John C. Brouillard: “Our nominating and corporate governance committee retained a third-party recruiting firm, which identified Mr. Brouillard, evaluated his credentials, interviewed him and recommended him for consideration by the nominating and corporate governance committee.”

For each year that we observe a new director nomination or appointment, we obtain the source of the nomination from the proxy along with other related information about the board and CEO characteristics, which we define below. It is possible that a director may be recommended to the board by more than one party. For example, both a search firm and the CEO may recommend the same person to the nominating committee. Whenever we observe recommendations for the same person from two (or more) different parties, we give priority to insiders (for example, the CEO) and treat the recommendation as having coming from insiders. This is a conservative approach and should work against our finding significant results. In the analysis, we focus only on independent directors purely nominated by search firms. It is possible that firms are not completely forthcoming about the sources of director nominations. Companies may use vague statements like a “search firm may be used.” Below we provide an example from Affiliated Computer Services, Inc.’s 2008 proxy statement: “The Nominating and Corporate Governance Committee generally identifies director nominees through the personal, business and organizational contacts of existing

52

ALI C. AKYOL AND LAUREN COHEN directors and management. However, the Nominating and Corporate Governance Committee may use a variety of sources to identify director nominees, including thirdparty search firms and stockholder recommendations.”

We read all the statements in proxy statements about how a director is identified or how the board identifies a director, and whenever the source of the nomination is unclear, we make a conservative judgment when assigning it to a source category. Vague statements, such as the example above, are generally assigned to the nominating committee category. Our assumption is that in such cases, the nominating committee identifies the director. We label such vague statements “ambiguous appointments” and identify them with a dummy variable. In our analysis, we use the full sample of independent director appointments, which includes ambiguous appointments. Our results are robust, however, only if we restrict our sample to the appointments in which the source of the director nomination is clear. From proxy statements, we collect information about the CEO and the board. Specifically, we collect the following data: whether a nominee is recommended by a search firm, the CEO, an another insider, an independent director, a major shareholder, or the nominating committee, or joined the board through a merger. We identify the current CEO and chairman and classify each director as an independent, gray, or inside director. We determine the CEO’s starting year and calculate his or her tenure as CEO. Director gender and age are also obtained from proxy statements. We identify whether the CEO is from the founding family using proxy statements, company websites, or other Internet sources. We obtain the number of shares owned by each director and calculate his or her percentage ownership. Finally, we determine if there is an independent director on the board who owns at least 5% of the shares. Using the director nomination data from proxy statements, we construct two key variables. The first is the percentage of search firm-recommended independent directors on the board. To do this, we start with the first year of a new independent director observation and calculate the percentage of search firm-recommended directors on the board. If we observe at least one new independent director nomination in a given year, but none is recommended by a search firm, then the percentage variable is simply zero. If there is no new independent director recommended by a search firm on the board the following year (or the departure of a search firm-recommended director), we assign the search firm director percentage from the current year to that year. Once we observe a new independent director recommended by a search firm on the board or a departure of a search

Who Chooses Board Members?

53

firm-recommended independent director, we update our search firm director percentage. Our second variable is a dummy variable for the presence of at least one search firm-recommended independent director on the board. If there is at least one independent director on the board recommended by a search firm, the dummy is then equal to 1, and 0 otherwise. Again, we control for departures and new appointments. We should note that our variables probably underestimate the true number of search firm-recommended directors on the board. As noted earlier, the SEC rule came into effect in 2004, and it is only after 2004 that we are able to identify the sources of director nominations. So it is quite possible that during our sample period a board has search firm-recommended directors appointed before 2004. We then merge our search firm data with several other databases. Accounting data come from Compustat, compensation data from Execucomp, governance data from RiskMetrics, and M&A data from Securities Data Company (SDC). We construct several variables from these databases. We list and explain our variables in Appendix Table A.1. We obtain a list of executive search firms operating in the United States as of 2004 from Kennedy Information, LLC. These search firms specialize in identifying board members. The list includes the headquarter city and all the branches associated with each search firm. To construct our instrumental variable (described in more detail in Section IV), we first calculate the distance of each firm’s headquarters to every search firm office and headquarters and then create an indicator variable that equals 1 if there is a top-search firm (the first six search firms with the most offices plus the headquarters) within a 100-kilometer radius of the firm’s headquarters. We present the search firms with the most offices and some statistics in Table 1. The largest search firm by branch number is Korn/Ferry International. Korn/Ferry International has 21 offices, including its headquarters. The next search firm is Heidrick & Struggles International, Inc., with 16 offices, including its headquarters. In Panel B, we divide the search firm locations into regions. Most search firms are located in the South (169 search firms), followed by the Northeast (149 search firms). Overall, the 2004 list has 395 distinct search firms. Most search firms have no offices other than their headquarters. There are 138 such offices, and in total, there are 533 search firm offices and headquarters in the 2004 list. We present director-related summary statistics in Table 2. Panel A shows that there were 5,866 new director nominations between 2004 and 2008 and that 1,239 of these were recommended by search firms (21.1%). In total, out of 4,963 new independent director nominations, 1,184 independent director recommendations came from search firms (23.9%).

54

ALI C. AKYOL AND LAUREN COHEN

Table 1. Search Firms. Rank

Search Firm

State

Offices (Main + Branch)

CA IL IL NY IL NY IL NY IL IL NY CA NJ NY CA TX NY NY MA NJ CA

21 16 15 12 10 9 8 8 7 6 6 6 6 5 4 4 4 3 3 3 3

Number of branches

Total

Panel B: Regional distribution Northeast 113 Midwest 69 South 120 West 93

36 20 49 33

149 89 169 126

Total

138

533

Panel A: Search firms with most offices 1 Korn/Ferry International 2 Heidrick & Struggles International, Inc. 3 Spencer Stuart 4 Boyden 5 Stanton Chase International 6 Egon Zehnder International, Inc. 7 Furst Group 8 Hudson Highland Group, Inc. 9 Signium International, Inc. 10 Cook Associates, Inc. 11 CTPartners 12 Dick Wray Executive Search 13 The Cassie Shipherd Group, LLC 14 Battalia Winston International 15 Executives Unlimited, Inc. 16 The Alexander Group 17 The Elliot Group, LLC 18 CCL Medical Search 19 Levin & Company, Inc. 20 Mancino Burfield Edgerton 21 System 1 Search, Inc. Region

Number of main offices

395

Note: We provide search firm-related statistics in this table. Panel A shows the search firms with most offices based on the 2004 data obtained from Kennedy Information, LLC. The second column, State, is the state of the main office, and the number of offices (including the main office) is provided in the last column in Panel A. Panel B breaks down search firms (main office and branches) by region.

Panel B shows that, out of the all sample firms with proxies, 764 firms had at least one new director in 2004. There were 623 firms with new directors in 2008. The number of firms with new directors ranges between 623 and 764 during 2004
2008. It seems that about 46% of the firms had a

Panel C: New Director Nominations by Search Firms 2004 2005 2006 2007 2008 2004
2008

Year

Panel B: Director Nominations by Years 2004 2005 2006 2007 2008 2004
2008

Year

Panel A: All Director Nominations Total number of new director nominations Less number of new inside and gray director nominations Less number of merger related new independent director nominations Total number of new independent director nominations

Table 2.

139 120 132 120 113 624

Number of firms that used a search firm

764 694 722 653 623 3,457

275 222 260 240 242 1,239

Number of new director nominations by search firms

1,333 1,155 1,227 1,086 1,065 5,866

3,779

4,627 592 256

Others

268 210 244 229 233 1,184

Number of independent director nominations by search firms

1,153 978 1,012 899 921 4,963

Number of new independent director nominations

Nominations By

Number of new director nominations

1,184

4,963 Number of firms with new directors

1,239 55 0

search firm 5,866 647 256

Full Sample

Sample Characteristics. Who Chooses Board Members? 55

56 0.140

Others

Search firm

55 0.193

0.009 0.073 0.027 0.127 0.255 0.445 0.064

Gray (110)

0.101 0.041 0.015 0.034 0.024 0.067 0.719

Inside (537)

Director type

(0.004)*** (0.000)***

Difference

0.227 0.116 0.049 0.223 0.085 0.049 0.250

Independent (5,219)

Note: We report director nomination summary statistics in this table. The sample period is from 2004 to 2008. Panel A provides statistics on all director nominations categorized by nomination type, whereas Panel B breaks down the total number of nominations by years. We focus on search firms in Panel C and provide search firm-identified new director nominations broken down by years. Further details about the sources of director nominations are provided in Panel D. The rows are the sources of new director nominations, and the columns represent director type. Last, we provide the median age and the fraction of female director nominations by search firms in Panel E and compare the numbers to nominations by other sources (all the sources other than search firms). *** denotes statistical significance at the 1% level when comparing means and medians.

Median age Fraction of female nominees

Panel E: Nominee Characteristics by Source

Panel D: Source of Nomination Fraction recommended by a search firm Fraction recommended by the CEO Fraction recommended by other executives Fraction recommended by an independent director Fraction recommended by a major shareholder Fraction nominated because of a merger Fraction recommended by the nominating committee

Source

Table 2. (Continued ) 56 ALI C. AKYOL AND LAUREN COHEN

Who Chooses Board Members?

57

new director in a given year during our sample period and that about 84% of new directors are independent directors. Panel C provides the number of firms that used a search firm out of the firms that had a new director. For example, out of 764 firms that had a new director in 2004 (Panel B), 139 firms in 2004 used a search firm. Over the sample period, the number of firms that used a search firm was fairly stable (between 113 and 139). The next column shows the number of director nominations per year, and the last column tells how many of the new directors in Column 2 were independent directors. For example, 764 firms had a new director in 2004 (Panel B), and 139 of them used a search firm to identify 275 directors. And 268 of the 275 directors that were recommended by search firms were independent directors. We next report statistics related to the sources of director nominations. Interestingly, we find that 10.1% of all inside director nominations came via search firms, meaning that search firms recommended an executive to the board. Search firms on average recommended 22.7% of independent directors to the board (based on 5,219 independent director nominations that also include M&A related nominations). CEOs are also influential in recommending independent directors to the board: 22.3% of independent director nominations came from CEOs. Remember that not all firms in our sample are clear about how they identify their directors and that we were conservative in assigning sources (when a director nomination was not clear, we assigned it to the nominating committee). Of all independent director nominations, 25% came via the nominating committee. As a final set of statistics, we examine the median director age and the fraction of female nominees in Panel E by nomination source. Search firms appear to have recommended younger people and more females to the board. It is possible that the use of search firms is confined to specific industries. We explore this in Table 3. Table 3 provides the industry distribution of our sample firms according to the Fama-French 48 industry classification. The table provides the number of firms in each industry that used a search firm at least once during the sample period. We do not observe that search firm use is confined to certain industries. We provide summary statistics for our sample in Table 4 by using all available data. Board and governance characteristics are presented in Panel A. There are 5,277 firm-year observations for search firm director indicator (SFID indicator). In about 27.6% of the observations, we observe at least one search firm-identified independent director. On average, 6.3% of the independent directors on the board are recommended by search firms, suggesting that the use of search firms is fairly common. Conditioning on the

58

ALI C. AKYOL AND LAUREN COHEN

Table 3.

Industry Distribution of Search Firm Usage.

Fama-French 48 Industry Code and Name

# Firms

Fama-French 48 Industry Code and Name

# Firms

1
Agriculture

0

0

2
Food products 3
Candy and soda 4
Beer and liquor

8 0 4

25
Shipbuilding, railroad equipment 26
Defense 27
Precious metals 28
Non-metallic and industrial metal mining 29
Coal 30
Petroleum and natural gas 31
Utilities 32
Communication 33
Personal services 34
Business services 35
Computers 36
Electronic equipment 37
Measuring and control equipment 38
Business supplies 39
Shipping containers 40
Transportation 41
Wholesale 42
Retail 43
Restaurants, hotels, motels 44
Banking 45
Insurance 46
Real estate 47
Trading 48
Other

5
Tobacco products 6
Recreation 7
Entertainment 8
Printing and publishing 9
Consumer goods 10
Apparel 11
Healthcare 12
Medical equipment 13
Pharmaceutical products

1 3 1 4 12 7 4 11 17

14
Chemicals 15
Rubber and plastic products 16
Textiles 17
Construction materials 18
Construction 19
Steel works etc. 20
Fabricated products 21
Machinery 22
Electrical equipment 23
Automobiles and trucks 24
Aircraft

12 1 1 8 9 5 0 20 5 8 4

3 0 2 1 5 20 7 5 29 12 31 12 8 3 7 11 28 6 17 21 0 12 5

Note: This table provides industry breakdown of the sample firms that have used a search firm at least once between 2004 and 2008 to identify a new independent director. The industry categories are based on the Fama-French 48 industries classification. There are 390 different firms that used a search firm at one time between 2004 and 2008. Fama-French 48 industries codes and names are provided in the first and third columns. Number of firms in each Fama-French industry is provided in the second and fourth columns.

presence of a search firm director, the search firm director percentage on the board increases to 21.5%. The typical firm has a board size of 10, and 98.2% of the boards are independent (>50% independent directors). CEO/ chairman duality is observed in about 60% of the firm-year observations, and the average CEO ownership is 3.8% (the median is 0.9%). The typical CEO has a tenure of seven years. About 15.6% of the CEOs are related to

59

Who Chooses Board Members?

Table 4. Summary Statistics. N

Mean

Median Standard Deviation

Panel A: Board and governance characteristics SFID dummy SFID percentage SFID percentage when SFID dummy = 1 Board size GIM index CEO/chairman duality CEO’s stock ownership Retirement age dummy CEO’s tenure as CEO CEO from founding family Independent board dummy Blockholder other than CEO dummy

5,277 5,277 1,975 5,277 4,599 5,277 5,277 5,277 5,277 5,277 5,277 5,277

0.276 0.063 0.223 9.854 9.325 0.599 0.038 0.175 7.039 0.156 0.982 0.175

0.000 0.000 0.182 10.000 9.000 1.000 0.009 0.000 5.000 0.000 1.000 0.000

0.447 0.122 0.147 2.427 2.506 0.496 0.104 0.380 7.150 0.363 0.132 0.380

Panel B: CEO Compensation CEO salary (in thousands) CEO total compensation (in thousands)

5,153 5,126

797 6,041

750 3,733

416 7,527

Panel C: Firm Characteristics S&P 500 dummy Market capitalization (in millions) ROA Tobin’s Q Sales growth R&D expenditures Capital expenditures Dividend dummy Negative excess return dummy (current year) Negative excess return dummy (previous year) Stock return volatility

5,277 0.360 5,277 10,264 5,277 0.096 5,277 1.947 5,277 0.114 5,277 0.033 5,277 0.064 5,277 0.608 5,277 0.446 5,258 0.471 5,204 0.020

0 2,279 0.086 1.563 0.099 0.000 0.033 1.000 0.000 0.000 0.018

0.480 28,837 0.079 1.145 0.165 0.069 0.099 0.488 0.497 0.499 0.008

Note: We present summary statistics related to board, governance, CEO compensation, and firm characteristics in this table. SFID stands for search firm-identified independent director. All other variables are defined in the appendix. Panel A presents summary statistics for boardand governance-related variables. The next two panels, B and C, show summary statistics for CEO compensation and firm characteristics related variables. The first column, N, is the number of firm-year observations with available data. The next three columns provide the mean, the median, and the standard deviation.

the founding family. We also observe that there is an independent blockholder director on about 17.6% of the boards. The average CEO salary is $0.79 million, with the average total compensation being $5.9 million (Panel B). The average firm is large, with a market capitalization of over $10 billion (Panel C). The median firm, however, is .

60

ALI C. AKYOL AND LAUREN COHEN

much smaller, with a market value of $2.3 billion. ROA is fairly high, with 9.6% for the average firm. We observe a Tobin’s Q of 1.95, R&D expenditures of 3.3%, and sales growth of 11.4% for the average firm. About 45% (47%) of the firms in the sample had a negative excess return in the current (previous) year. Overall, the summary statistics for our sample firms are similar to the ones from previous studies. We compare firm-year observations in which we observe a search firmidentified independent director to the rest of firm-year observations in Table 5. Panel A is for board and governance characteristics. Firms that use search firms have larger boards, more independent directors, younger CEOs, lower CEO tenure, and a lower probability of a non-CEO independent director blockholder than the firms that do not use a search firm. The governance index is statistically higher for the firms that use a search firm. CEO ownership is much lower in companies that use search firms. We also note that CEOs are less likely to be related to the founding family when a search firm is used. In Panel B, we provide statistics for CEO compensation. Salary and total compensation is much higher in firm-year observations in which we observe a search firm-identified independent director. The average salary is $0.89 million when a search firm is used, and $0.76 million when not used. Similarly, total compensation is $7.45 million in firms that use a search firm, compared to $5.42 million in firms that do not. Panel C shows that firms that use a search firm are much larger firms. For example, the market value of equity is $12.1 billion when a search firm is used, and $8.9 billion when not used. We do not observe a difference with respect to ROA. However, Tobin’s Q is statistically different between firms that use a search firm and firms that do not. Sales growth, however, is statistically lower for the firms that use a search firm (9.2% vs. 12.3%). Also, fewer firms have negative excess stock price return when they have a search firm-identified independent director on the board.

RESULTS In this section we test the impacts of having a search firm-identified independent director on firm governance and behavior. Given that we are testing the impact of certain board members on firm behavior, it is best to choose governance characteristics and decisions in which the board is likely involved. We thus begin by testing the impact of directors found through search firms on CEO compensation.

61

Who Chooses Board Members?

Table 5. Summary Statistics by Search Firm Usage. SFID Dummy = 1 N

Mean Median

SFID Dummy = 0 N

Mean

Median

9.691*** 9.212*** 0.594 0.046*** 0.196*** 7.639*** 0.186*** 0.979*** 0.184***

9.000*** 9.000*** 1.000 0.011*** 0.000*** 5.000*** 0.000*** 1.000*** 0.000***

Panel A: Governance Characteristics Board size GIM index CEO/chairman duality CEO’s stock ownership Retirement age dummy CEO’s tenure as CEO CEO from founding family Independent board dummy Independent director blockholder

1,458 10.279 1,331 9.602 1,458 0.612 1,458 0.018 1,458 0.120 1,458 5.466 1,458 0.078 1,458 0.991 1,458 0.152

10.000 10.000 1.000 0.005 0.000 4.000 0.000 1.000 0.000

3,819 3,268 3,819 3,819 3,819 3,819 3,819 3,819 3,819

Panel B: CEO Compensation CEO salary (in thousands) CEO total compensation (in thousands)

1,435 893 1,430 7,501

882 5,441

3718 760*** 700*** 3696 5,475*** 3,226***

1,458 0.552 1,458 12,989 1,458 0.094 1,458 1.951 1,458 0.092 1,458 0.038 1,458 0.057 1,458 0.656 1,458 0.405

1.000 4,489 0.087 1.618 0.083 0.000 0.034 1.000 0.000

3,819 3,819 3,817 3,819 3,819 3,819 3,819 3,819 3,819

1,457

0.426

0.000

3,801 0.488*** 0.000***

1,447

0.019

0.017

3,757 0.020*** 0.019***

Panel C: Firm Characteristics S&P 500 dummy Market capitalization (in millions) ROA Tobin’s Q Sales growth R&D expenditures Capital expenditures Dividend dummy Negative excess return dummy (current year) Negative excess return dummy (previous year) Stock return volatility

0.287*** 9,224*** 0.096 1.945 0.123*** 0.031*** 0.066*** 0.590*** 0.462***

0.000*** 1,804*** 0.085 1.543** 0.105*** 0.000*** 0.033 1.000*** 0.000***

Note: This table provides summary statistics in Table 3 classified by the use of a search firm. A firm is classified as type (search firm-identified independent director (SFID) dummy = 1) if it uses a search firm to identify a new independent director at each firm-year observation, and the firm is classified as type (search firm identified independent director (SFID) dummy = 0) if it does not employ the services of a search firm to identify a new independent director. All variables are defined in the appendix. *** and ** denote statistical significance at the 1% and 5% levels when comparing means and medians.

The results of these tests are presented in Table 6. The dependent variables are the log of CEO salary or the log of total CEO compensation. We use both measures, since the board sets both of these, and part of the CEO compensation is the imputed value of options and restricted stock grants

62

ALI C. AKYOL AND LAUREN COHEN

(necessarily more uncertain), whereas salary is more certain and straightforwardly measured, given both its horizon and security structure. The independent variable of interest is SFID Percentage, which is the percentage of board members found using an executive search firm.3 We use a number of CEO-, firm-, and board-level characteristics to control for other potential determinants of the CEO’s compensation, all measured in the fiscal year prior to CEO compensation. These include, for the CEO: CEO tenure, whether the CEO is chairman of the board, whether the CEO is a member of the founding family, CEO ownership in the firm (along with ownership squared), and CEO age. For the board, these are: board size, a dummy for whether the board is majority independent, and a dummy for whether any of the independent directors are blockholders in the firm. And for the firm, these are: the Gompers, Ishii, and Metrick (2003) governance index, firm size, book-to-market, profitability, sales growth, R&D, capital expenditures, a dummy for whether the firm paid dividends the past year, past returns, stock return volatility, and a dummy for whether the firm is in the S&P500 Index.4 Last, we include industry and year fixed effects to control for any variation in compensation driven by a given industry or a given year in our sample. We also adjust all standard errors for clustering at the firm level. All the columns from Table 6 deliver the same message: Having more directors nominated by search firms is associated with significantly higher CEO compensation. SFID Percentage has a significant coefficient in each specification, for both salary and total compensation. In the full specifications, the magnitude on search firm-identified independent directors for salary (total compensation) is 0.262, t = 2.54 (0.426, t = 3.42). To give an idea of magnitude, having a one standard deviation larger percentage of the board made up of search firm-identified independent directors (roughly switching out one non-search firm director for one search firm director) results in a 1.1% higher CEO salary and 1.4% higher total compensation. Table 7 provides complementary evidence. Here we examine the performance sensitivity of CEO departures when more search firm-identified independent directors are on the board. The dependent variable in these regressions is a dummy variable equal to 1 if the CEO departed in a given year. We use the same control variables that we use in Table 6 along with two new interaction variables. The independent variable of interest is now the interaction of poor returns and percentage of search firm-identified independent directors. The hypothesis is that, given that boards with directors nominated by search firms appear to pay CEOs more, they may also be more hesitant to fire CEOs even in the face of poor performance. This is exactly what we see in Table 7. The interaction between poor returns and

63

Who Chooses Board Members?

Table 6.

CEO Compensation.

Dependent Variables

Log (Salary)

Log (Salary)

Log (TDC1)

Log (TDC1)

SFID percentage

0.301*** (0.005)

0.262** (0.011)

0.427*** (0.001)

0.426*** (0.001)

GIM index

0.018*** (0.001)

0.022*** (0.001)

S&P 500

0.077* (0.055)

0.046 (0.276)

0.074 (0.169)

0.070 (0.200)

Independent board dummy

−0.042 (0.556) 0.188** (0.029)

−0.031 (0.648) 0.199** (0.011)

−0.057 (0.681) 0.173 (0.120)

−0.015 (0.916) 0.142 (0.215)

0.063*** (0.007)

0.046*** (0.009)

0.028 (0.283)

0.026 (0.240)

Retirement age

−0.009 (0.796)

−0.004 (0.907)

−0.022 (0.655)

−0.030 (0.553)

CEO is from founding family

−0.113 (0.201)

−0.112 (0.141)

−0.238** (0.014)

−0.196** (0.045)

CEO/chair duality

0.064* (0.073) −1.169 (0.205)

0.076*** (0.008) −0.608 (0.378)

0.163*** (0.000) −1.228 (0.126)

0.141*** (0.000) −1.594* (0.076)

(CEO ownership)^2

1.852* (0.096)

1.161 (0.168)

1.645* (0.080)

2.056* (0.070)

Independent director blockholder

−0.025 (0.676)

0.049 (0.129)

−0.101* (0.055)

−0.051 (0.303)

0.093*** (0.006)

0.113*** (0.000)

0.412*** (0.000)

0.416*** (0.000)

1.312 (0.020)

1.126** (0.014)

0.547 (0.233)

0.501 (0.322)

−0.192*** (0.001)

−0.158*** (0.000)

−0.164*** (0.000)

−0.154*** (0.001)

Sales growth

−0.271* (0.058)

−0.135* (0.088)

0.000 (0.998)

−0.005 (0.963)

R&D

−0.049 (0.910)

0.043 (0.926)

0.480 (0.370)

0.636 (0.294)

Capex

−0.547*** (0.007)

−0.345*** (0.008)

−0.101 (0.679)

−0.007 (0.978)

Dividend dummy

0.035 (0.343)

0.000 (0.998)

−0.068 (0.118)

−0.058 (0.196)

Negative excess return dummy (current year)

−0.007 (0.780)

0.002 (0.952)

−0.113*** (0.000)

−0.104*** (0.002)

Log(board size) Log(CEO tenure)

CEO ownership

Log(market cap) ROA Tobin’s Q

64

ALI C. AKYOL AND LAUREN COHEN

Table 6.

(Continued )

Dependent Variables

Log (Salary)

Log (Salary)

Log (TDC1)

Log (TDC1)

Negative excess return dummy (previous year)

−0.021

−0.030

−0.113***

−0.110***

(0.313)

(0.172)

(0.000)

(0.000)

Stock return volatility

−2.237 (0.252)

−3.495 (0.139)

5.560* (0.059)

7.456** (0.040)

5.968*** (0.000) Yes Yes 5,050 0.169

5.486*** (0.000) Yes Yes 4,448 0.187

5.180*** (0.000) Yes Yes 5,050 0.399

4.836*** (0.000) Yes Yes 4,448 0.405

Constant Year fixed effects Industry fixed effects Number of observations R2

Note: This table presents the results of CEO compensation regressions. The dependent variable in the first two OLS and the logarithm of CEO salary in year t and the logarithm of total CEO compensation (TDC1) in year t in the last two models. Financial data are as of the previous fiscal year. All variables are described in the Appendix. We control for year and industry firm fixed effects in all models. Industry fixed effects are based on the Fama-French 48 industries classification. P-values are provided in parentheses and are based on heteroskedasticity robust standard errors. Standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels.

the percentage of search firm-identified independent directors is negative and significant, −1.736 (t = 2.15). This suggests that boards with directors identified by search firms are significantly less likely to fire a CEO following poor performance. In Table 8, we move on to another important decision made by boards of directors: that of the acquisitions policy of the firm. This is widely seen as one of the board’s most important functions, and it can obviously have large potential value implications (positive or negative) for the company. In Panel A of Table 8, we regress whether the firm had any M&A activity in the year on the percentage of search firm-identified independent directors on the board and controls (the same controls as used in Tables 6 and 7). Columns 1 and 2 both show that the greater the percentage of search firmidentified independent directors on a company’s board, the more M&A activity they engage in. For instance, the coefficient on SFID Percentage in Column 2 of 0.578 (t = 3.04), implies that a one standard deviation higher percentage of search firm-identified independent directors on the board is associated with an increase in the amount of M&A activity by nearly six percentage points (mean of roughly 18%), or more than a 30% increase.

65

Who Chooses Board Members?

Table 7. CEO Turnover. Dependent Variables SFID percentage

Forced Turnover

Forced Turnover

−0.343 (0.351)

−0.171 (0.640)

SFID percentage * Negative excess return dummy (current year) SFID percentage * Negative excess return dummy (previous year) GIM index All other controls Year fixed effects Industry fixed effects Number of observations R2

Yes Yes Yes 5,050 0.169

Forced Turnover

Forced Turnover

0.285 (0.599) −1.315*

0.761 (0.138) −1.736**

(0.087)

(0.032)

0.219

0.086

(0.763)

(0.907)

0.019

0.018

(0.291)

(0.307)

Yes Yes Yes 4,448 0.186

Yes Yes Yes 5,050 0.172

Yes Yes Yes 4,448 0.191

Note: This table presents the results of forced CEO turnover regressions. The dependent variable in all models is an indicator variable that is equal to 1 if there is a forced CEO turnover in year t. Financial data are as of the previous fiscal year. All other control variables from Table 6 are included but not reported for brevity. All variables are described in the Appendix. We control for year and industry firm fixed effects in all models. Industry fixed effects are based on the Fama-French 48 industries classification. p-values are provided in parentheses and are based on heteroskedasticity robust standard errors. Standard errors are clustered at the firm level. ** and * denote statistical significance at the 5% and 10% levels.

In Panel B, we explore whether this increased M&A activity appears to be value-enhancing or value-destroying for the firms with more search firmidentified independent directors. To do so, we regress acquirer merger announcement returns from the day (−2, +2) window on the percentage of search firm-identified independent directors at the acquiring firm. The mean acquirer announcement return for our sample (S&P 1500 firms) over our time period is 49 basis points (similar to that found in Masulis, Wang, and Xie (2007)). From both Columns 1 and 2, SFID Percentage is negatively related to this announcement return. In other words, the more search firmidentified independent directors on the board of the acquiring company announcing the merger, the lower the announcement return. The impact, while marginally statistically significant, is quite large economically. For instance, the coefficients in Columns 1 and 2 imply a decrease in announcement return of between 1.7% and 2.6% (from a mean of 49 basis points).

0.001 (0.861) 0.021*** (0.007)

Private target dummy

Tender offer dummy

0.000 (0.924) 0.022*** (0.007)

−0.018*** (0.006)

−0.020*** (0.003)

Relative deal size

Public target dummy

(0.072)

−0.004 (0.925) −0.017 (0.242) −0.002*

Announcement return

Yes Yes Yes 4,542 0.064

0.578*** (0.003) 0.037 (0.001)

M&A dummy

−0.025* (0.074)

−0.028 (0.472) −0.026* (0.074)

Announcement return

Yes Yes Yes 5,194 0.059

0.522*** (0.007)

M&A dummy

Mergers and Acquisitions.

−0.016 (0.257)

GIM index

SFID percentage

Panel B: Announcement Returns Constant

Variables

All other controls Year fixed effects Industry fixed effects Number of observations R2

GIM index

Panel A: M&A Probability SFID percentage

Variables

Table 8. 66 ALI C. AKYOL AND LAUREN COHEN

0.004 (0.619) 0.003 (0.826) Yes Yes Yes 1,142 0.118

Deal competed dummy

All other controls Year fixed effects Industry fixed effects Number of observations R2

Yes Yes Yes 1,018 0.132

−0.001 (0.890) −0.003 (0.858)

0.004 (0.281)

Note: This table presents the results of mergers and acquisition regressions. The dependent variable in the Probit models in Panel A is a dummy variable equal to 1 if the firm has made an acquisition in year t. The dependent variable in the Panel B is the abnormal return from the day (−2, +2) window. Abnormal returns are based on the market model with an equally weighted market index. Financial data are as of the previous fiscal year. All other control variables from Table 6 are included but not reported for brevity. All variables are described in the Appendix. We control for year and industry fixed effects in all models. Industry fixed effects are based on the FamaFrench 48 industries classification. p-values are provided in parentheses and are based on heteroskedasticity robust standard errors. Standard errors are clustered at the firm level. *** and * denote statistical significance at the 1% and 10% levels.

Hostile deal dummy

0.007* (0.092)

All cash dummy

Who Chooses Board Members? 67

68

ALI C. AKYOL AND LAUREN COHEN

The evidence in Tables 6
8 is consistent with search firm-identified independent directors being associated with significantly poorer governance of the companies of which they oversee. Importantly, we focus on precisely those decisions over which boards are unambiguously involved: setting CEO compensation, deciding on the turnover of the firm, and advising on the M&A decisions of the company. Besides simply being involved, these three board decisions are arguably among the most important that boards make from a firm-value perspective, and all appear significantly related to the presence of search firm-identified independent directors.

INSTRUMENT Although the relationships documented in Section III are suggestive, the main hurdle we face in the analysis is pinning down a causal mechanism between search firm-identified independent directors and these negative firm outcomes. To do this, we need to find an exogenous factor that drives companies to use search firms. In other words, we need to find some part of the reason why a company decides to use a search firm to appoint its directors that has nothing to do with an unobservable characteristic that might also be driving the relationships we document above (but not work through search firm-identified independent directors). To be clear, the instrument need not be the entirety of the reason why, nor even the most important part of the reason why; it simply needs to be some part of a company’s decision to use a search firm when appointing a director. Also, what we are instrumenting for is the decision whether to use a search firm. As we are comparing to other firms that are appointing new directors, the act of new director appointment is being conditioned upon, so it is only how this new director is found that is being instrumented for. We then hypothesize that distance from an executive search firm may affect companies’ use of executive search firms. The idea is that companies are more likely to employ the use of search firms if they are located near the firm, as this proximity allows the company and search firm to have more meetings and less costly interactions regarding the slate of potential directors. Again, as long as a portion of the decision of whether to use an executive search firm to find a new director is driven by distance to an executive search firm, we can use this in a two-stage least squares instrumental variable framework to identify the causal impact of search firmidentified independent directors on future firm governance behavior and performance.

69

Who Chooses Board Members?

Table 9.

Firm Performance Regression Analysis: ROA.

Variables Panel A: OLS Regressions SFID percentage

ROA −0.027** (0.040)

GIM index

ROA −0.024* (0.070) 0.001* (0.051)

All other controls Year fixed effects Industry fixed effects Number of observations R2 Variables Panel B: First Stage of 2SLS Regressions Search firm within 100 kilometers

Yes Yes Yes 5,277 0.312

Yes Yes Yes 4,599 0.327

SFID percentage

SFID percentage

0.017** (0.013)

GIM index

0.021*** (0.004) 0.000 (0.857)

All other controls Year fixed effects Industry fixed effects Number of observations R2

Yes Yes Yes 5,277 0.125

Yes Yes Yes 4,599 0.130

Variables

ROA

ROA

Panel C: Second Stage of 2SLS Regressions SFID percentage (instrumented)

−0.407*** (0.001)

GIM index

−0.335*** (0.003) 0.001* (0.050)

All other controls Year fixed effects Industry fixed effects Number of observations R2

Yes Yes Yes 5,277 0.314

Yes Yes Yes 4,599 0.328

Note: This table presents the results of ROA regressions. The dependent variable in the OLS and second stage of the 2SLS models is ROA in year t. The dependent variable in the first stage of the 2SLS models is the percentage of search firm
identified directors on the board (SFID percentage). Financial data are as of the previous fiscal year. All other control variables from Table 6 are included but not reported for brevity. All variables are described in the Appendix. We control for year and industry firm fixed effects in all models. Industry fixed effects are based on the Fama-French 48 industries classification. p-values are provided in parentheses and are based on heteroskedasticity robust standard errors. Standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels.

70

ALI C. AKYOL AND LAUREN COHEN

In Panel A of Table 9, we regress ROA on SFID Percentage along with other control variables (controlling for all other determinants and controls we have been including up to this point, including year and industry fixed effects) and find that ROA is decreasing in the percentage of search firmidentified independent directors on the board. This is in line with the evidence presented in previous tables. However, as noted above, this result is suggestive, and we need to establish that SFID Percentage affects firm performance. In Table 9, Panel B, we show the first stage of 2SLS to explore whether distance to a search firm really does affect a company’s decision to use a search firm (again controlling for all other determinants and controls we have been including up to this point, including year and industry fixed effects). The dependent variable is the regression is SFID Percentage, and the independent variable of interest is Search Firm within 100 kilometers, a categorical variable equal to 1 if the firm has an executive search firm branch within 100 kilometers, and 0 otherwise. Both columns of Table 9, Panel B, deliver the same message: having an executive search firm near the company results in the company being significantly more likely to use an executive search firm when replacing directors. The magnitude of this effect is large. Companies with a search firm nearby have 34% more directors found by executive search firms (coefficient of 2.1%, t = 2.97, from a mean of 6.3%).5 In the second stage, we then use this piece of the search firm choice that is based on geographic closeness (the instrumented piece). The results are in Table 9, Panel C. We examine the impact of search firm-identified independent directors on company profitability (ROA). From Table 9, Panel C, the orthogonal piece of search firm directors is a significant predictor of lower profitability, consistent with the results provided in Section III. This 2SLS procedure and result give support to the causal interpretation of search firm-identified independent directors on firm behavior and performance.6 In an unreported table, we repeat the analysis in Table 9 with Tobin’s Q and obtain similar conclusions.

CONCLUSION Studying boards and understanding how they are constructed is a first-order question facing corporate governance, since boards arguably represent the most important device through which agency problems are mitigated between managers and shareholders. We exploit a recent regulation passed by the SEC to explore the nomination of board members to US

71

Who Chooses Board Members?

publicly traded firms. In particular, we focus on firms’ use of executive search firms versus simply permitting internal members (often simply the CEO) to nominate the new directors to serve on the board. We show that companies that use search firms to find board members pay their CEOs significantly higher salaries and significantly higher total compensations. Further, companies with search firm-identified independent directors are significantly less likely to fire their CEOs following negative performance. In addition, we find that companies with search firm-identified independent directors are significantly more likely to engage in M&A and that they see abnormally low returns from this M&A activity. We instrument the endogenous choice of using an executive search firm when choosing directors through the varying geographic distance of companies to executive search firms. We hypothesize that companies are more likely to use executive search firms if they are located near the firm, as this proximity allows the company and search firm to have more meetings and less costly interactions regarding the slate of potential directors. We show that in a two-stage least squares instrumental variable framework the causal impact of search firm-identified independent directors on firm performance, consistent with firm behavior and governance consequences we also document. The future of corporate governance hinges on our understanding of the inner workings of boards as agency cost-moderating devices. Our study pushes forward the understanding of boards’ roles in governing corporations, specifically enhancing our understanding of power dynamics within the board. However, more needs to be done in this vein, as any push toward optimal governance frameworks needs to have design implications for monitor (that is, board director) choice.

NOTES 1. See http://www.sec.gov/rules/final/33-8340.htm for more about the new disclosure regulation. 2. We in fact hand-collect the founding dates of search firm branch office locations and run tests in which we solely examine the decisions of companies that were established decades after the search firm offices to deal with the reverse causality of search firm office location. 3. We run all tests in the chapter using a dummy variable version of search firm director board membership instead of the continuous version. The magnitude and significance of the results are nearly unchanged and are available on request. 4. Given that our sample consists of all firms in the S&P 1500 Index, this dummy is a distinction between the 500 and 1500 index constituents.

72

ALI C. AKYOL AND LAUREN COHEN

5. We also use continuous versions of this variable-distance to a search firm and log(distance to a search firm) and both yield the same large and significant relationship. We choose the dummy variable specification for its ease of interpretation. 6. One remaining worry is that the executive search firms may simply locate where they think they’ll generate business from existing firms. To address this, we actually hand-collect the founding dates of search firm branch office locations and run tests in which we solely examine the decisions of companies that were established after (mean of over 10 years after) the search firm offices to deal with the reverse causality of search firm office location. The coefficient in the second stage of the 2SLS on instrumented search firm percentage is −0.758 (t = 3.32), even larger than that in Table 9.

ACKNOWLEDGMENT We thank Rene´e Adams, Chris Malloy, Ron Masulis, Patrick Verwijmeren, and David Yermack, seminar participants at the University of Adelaide, the University of Melbourne, York University, and participants at the 2012 University of Delaware Corporate Governance Symposium for helpful comments and suggestions. We also thank Mark Spencer Wallis and Sonya Lai for excellent research assistance. We are grateful for funding from the National Science Foundation and the University of Melbourne.

REFERENCES Adams, R. B., & Ferreira, D. (2007). A theory of friendly boards. Journal of Finance, 62, 217
250. Adams, R. B., & Ferreira, D. (2009). Women in the boardroom and their impact on governance and performance. Journal of Financial Economics, 94, 291
309. Armstrong, C., Ittner, C. D., & Larcker, D. F. (2010). Economic characteristics, corporate governance, and the influence of compensation consultants on executive pay levels. Stanford University Working paper. Bhagat, S., & Black, B. (2002). The non-correlation between board independence and longterm firm performance. Journal of Corporation Law, 27, 231
274. Brickley, J. A., Coles, J. L., & Terry, R. L. (1994). Outside directors and the adoption of poison pills. Journal of Financial Economics, 35, 371
390. Byrd, J. W., & Hickman, K. A. (1992). Do outside directors monitor managers? Evidence from tender offer bids. Journal of Financial Economics, 32, 195
221. Cadman, B., Carter, M. E., & Hillegeist, S. (2008). The role and effect of compensation consultants on CEO pay. University of Pennsylvania Working paper. Canyon, M. J., Peck, S. I., & Sadler, G. V. (2009). Compensation consultants and executive pay: Evidence from the United States and the United Kingdom. Academy of Management Perspective, 23, 43
55.

Who Chooses Board Members?

73

Dasgupta, S., & Ding, F. (2010). Search intermediation, internal labor markets, and CEO pay. Hong Kong University of Science and Technology Working paper. Fama, E. F., & Jensen, M. C. (1983). Separation of ownership and control. Journal of Law and Economics, 26, 301
325. Fich, E. M., & Shivdasani, A. (2006). Are busy boards effective monitors? Journal of Finance, 61, 689
724. Fich, E. M., & White, L. J. (2003). CEO compensation and turnover: The effects of mutually interlocked boards. Wake Forest Law Review, 38, 935
959. Gompers, P. A., Ishii, J. L., & Metrick, A. (2003). Corporate governance and equity prices. Quarterly Journal of Economics, 118, 107
155. Graham, J. R., Harvey, C. R., & Rajgopal, S. (2006). Value destruction and financial reporting decisions. Financial Analysts Journal, 62, 27
39. Hermalin, B., & Weisbach, M. (1988). The determinants of board composition. Rand Journal of Economics, 19, 589
606. Hwang, B.-H., & Kim, S. (2009). It pays to have friends. Journal of Financial Economics, 93, 138
158. Jensen, M., Murphy, K. J., & Wruck, E. G. (2004). Remuneration: Where we’ve been, how we got to here, what are the problems, and how to fix them. Harvard University Working paper. Knyazeva, A., Knyazeva, D., & Masulis, R. W. (2011). Effects of local director markets on corporate boards. University of Rochester Working paper. Lorsch, L. W., & MacIver, E. (1989). Pawns or potentates: The reality of America’s corporate boards. Boston, MA: Harvard Business Press. Mace, M. L. (1971). Directors: Myth and reality. Cambridge, MA: Harvard University Press. Masulis, R. W., & Mobbs, S. (2011). Are all inside directors the same? Evidence from the external directorship market. Journal of Finance, 66, 823
872. Masulis, R. W., Wang, C., & Xie, F. (2007). Corporate governance and acquirer returns. Journal of Finance, 62, 1851
1889. Masulis, R. W., Wang, C., & Xie, F. (2012). Globalizing the boardroom: The effects of foreign directors on corporate governance and firm performance. Journal of Accounting and Economics, 53, 527
554. Moeller, S. B., Schlingemann, F. P., & Stulz, R. M. (2005). Wealth destruction on a massive scale? A study of acquiring-firm returns in the recent merger wave. Journal of Finance, 60, 757
782. Murphy, K. J., & Sandino, T. (2010). Executive pay and “independent” compensation consultants. Journal of Accounting and Economics, 49, 247
262. O’Neal, D., & Thomas, H. (1995). Director networks/director selection: The board’s strategic role. European Management Journal, 13, 79
90. Rajgopal, S., Taylor, D., & Venkatachalam, M. (2012). Frictions in the CEO labor market: The role of talent agents in CEO compensation. Contemporary Accounting Research, 29, 119
151. Rosenstein, S., & Wyatt, J. (1990). Outside directors, board independence, and shareholder wealth. Journal of Financial Economics, 26, 175
191. Shivdasani, A., & Yermack, D. (1999). CEO involvement in the selection of new board members: An empirical analysis. Journal of Finance, 54, 1829
1853. Weisbach, M. (1988). Outside directors and CEO turnover. Journal of Financial Economics, 20, 431
460. Yermack, D. (2006). Flights of fancy: Corporate jets, CEO perquisites, and inferior shareholder returns. Journal of Financial Economics, 80, 211
242.

74

ALI C. AKYOL AND LAUREN COHEN

APPENDIX Table A.1. Variable Main independent variables Search firm
identified independent director (SFID) percentage Search firm within 100 kilometers

Control variables S&P 500 Independent board Board size CEO tenure Retirement age CEO is from founding family CEO/Chair duality CEO ownership

Governance index Independent director blockholder Firm size Tobin’s Q

R&D/Sales Sales growth Capex Dividend dummy Negative excess return dummy (current year)

Variable Definitions. Definition

The percentage of search firm-identified independent directors on the board. Dummy variable: 1 if the company has an executive search firm branch within 100 kilometers, 0 otherwise. Dummy variable: 1 if the firm is in the S&P 500 Index during the year, 0 otherwise. Dummy variable: 1 if the majority of the directors on the board are independent directors, 0 otherwise. Number of directors on the firm’s board The number of years the person has been the CEO of the company. Dummy variable: 1 if the CEO is over 61 years old, 0 otherwise. Dummy variable: 1 if the CEO belongs to the founding family, 0 otherwise. Dummy variable: 1 if the CEO of the firm is also the chairman of the board, 0 otherwise. CEO’s percentage stock ownership in his firm, including both stock and stock options that are exercisable within 60 days (as of the proxy date). Gompers et al. (2003) index. Dummy variable: 1 if there is a 5% independent director blockholder on the board, 0 otherwise. Log of market value of equity (CSHO *PRCC_F). Market value of assets (calculated as the book value of assets (AT) plus the market value of common stock (CSHO *PRCC_F) less the sum of book value of common equity (CEQ) and balance sheet
deferred taxes (TXDB)) divided by the book value of assets (AT) . Research and development expenses scaled by sales. Percentage change in sales over the previous fiscal year. Capital expenditure scaled by sales. Dummy variable: 1 if the firm pays any dividends during the year, 0 otherwise. Dummy variable: 1 if percentage change in the firm’s stock price over the year net of the change in S&P 500 is negative, 0 otherwise.

75

Who Chooses Board Members?

Table A.1. Variable Negative excess return dummy (previous year) Stock return volatility Dependent variables Salary Total compensation CEO turnover Number of M&As M&A returns ROA

(Continued ) Definition

Dummy variable: 1 if percentage change in the firm’s stock price over the previous year net of the change in S&P 500 was negative, 0 otherwise. Standard deviation of daily stock returns during the year. Log of salary (SALARY) from Execucomp. Log of total compensation (TDC1) from Execucomp. Dummy variable: 1 a forced CEO turnover is observed during the year, 0 otherwise. The number of M&As conducted by the firm during the year. Abnormal returns from the (−2, +2) window around the M&A announcements using the market model. Operating income before depreciation (OIBDP) over book value of assets (AT).

THE GROWTH OF GLOBAL ETFS AND REGULATORY CHALLENGES Reena Aggarwal and Laura Schofield ABSTRACT Purpose
Exchange traded funds (ETFs) are one of the most innovative financial products listed on exchanges. As reflected by the size of the market, they have become popular among both retail and institutional investors. The original ETFs were simple and easy to understand; however, recent products, such as leveraged, inverse, and synthetic ETFs, are more complex and have additional dimensions of risk. The additional risks, complexity, and reduced transparency have resulted in heightened attention by regulators. This chapter aims to increase understanding of how ETFs function in the market and can potentially impact financial stability and market volatility. Design/methodology/approach
We discuss the evolution of ETFs, growing regulatory concerns, and the various responses to these concerns. Findings
We find that concerns related to systemic risk and excess volatility, suitability for retail investors, lack of transparency and liquidity, securities lending and counterparty exposure are being addressed by both market participants and policy makers. There has been a shift

Advances in Financial Economics, Volume 16, 77
100 Copyright r 2013 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3732/doi:10.1108/S1569-3732(2013)0000016003

77

78

REENA AGGARWAL AND LAURA SCHOFIELD

toward multiple counterparties, overcollateralization, disclosure of collateral holdings and index holdings. Originality/value
The analysis contained in this chapter provides an understanding of the role of ETFs in the financial markets and the global economy that should be valuable to market participants, investors, and policy makers. Keywords: Exchange traded funds; regulation; synthetic ETFs; leveraged ETFs

INTRODUCTION Exchange Traded Funds (ETFs) are similar to mutual funds, but unlike mutual funds, they are listed on an exchange and trade throughout the day, similarly to stocks. ETFs may have lower expense ratios and certain tax efficiencies compared to traditional mutual funds, and they allow investors to buy and sell shares at intra-day market prices. Moreover, investors can sell ETF shares short, write options on them, and set market, limit, or stop-loss orders. The shares of ETFs often trade at market prices close to the net asset value (“NAV”) of the shares, rather than at discounts or premiums. ETFs are one of the most innovative and successful products listed on exchanges and have grown tremendously over the years. The original ETFs were simple, providing diversification benefits at a low cost and allowing intra-day trading. More recently, complex products with additional risks have been introduced, attracting the attention of regulators. Regulatory concerns have focused around the issues of systemic risk, transparency, lack of liquidity, complexity and suitability, counterparty risk, and the lending market in ETFs. Although ETFs did not cause the “Flash Crash” of May 6, 2010, the event did raise regulatory concerns about the potential role of ETFs on days of high volatility. This chapter aims to closely examine the enormous growth in market size and complexity of ETFs as well as the regulatory concerns raised by them. We also examine efforts by regulators and the industry to address some of the concerns. The rest of the chapter is organized as follows: The second section explains the types of ETFs and the recent trends in ETFs including leveraged, inverse, synthetic and actively managed ETFs; the third section

The Growth of Global ETFs and Regulatory Challenges

79

discusses regulatory concerns raised by policy makers in different countries; the fourth section summarizes the response to the regulatory concerns; the third section examines the role of ETFs in emerging markets using the case of India; and the second section concludes.

ETFS: EVOLUTION AND RECENT TRENDS ETFs are being used for active and passive strategies. They provide an alternative to derivatives and stocks when investors are looking to increase or decrease exposure. They can be used for a buy and hold strategy or for market timing purposes. Institutions, such as some pension funds, that have restrictions on investing in derivatives can invest in ETFs. Other institutions find them to be an alternative to futures that have margin requirements and expiration dates. Hedge funds can use them to take long or short positions. They are also used to temporarily park cash during transitions in investment strategy or change in management. ETFs are increasingly being used by institutional investors for both strategic and tactical purposes; they are also popular among retail investors. These products are generally bought on a commission basis, and investors pay brokerage commissions when they buy or sell. Similar to stocks that trade on an exchange, ETFs can be bought on margin. In comparison to mutual funds, the tax efficiency of ETFs arises because mutual funds need to sell shares for investors’ redemption, and this can result in capital gains. These capital gains have to be distributed to investors; hence, investors may incur taxes. ETFs don’t have to sell shares to meet redemptions. Liquidity, expense ratios, and tracking error are important factors for investors investing in ETFs. ETFs are registered with the securities commissions and are generally organized as open-end investment companies. Sometimes, they are also organized as unit investment trusts. ETF shares are purchased and redeemed directly from the fund sponsor in large blocks called “creation units.” Arbitrage activity in ETF shares is facilitated by the transparency of the ETF’s portfolio. Each day, an ETF publishes the identities of the securities in the purchase and redemption baskets, which are representative of the ETF’s portfolio. Each exchange on which the ETF shares are listed typically discloses the current value of the basket on a per share basis (“Intraday Value”) at 15-second intervals throughout the day and for index-based ETFs, disseminates the current value of the relevant index.

80

REENA AGGARWAL AND LAURA SCHOFIELD

This transparency can contribute to the efficiency of the arbitrage mechanism because it helps arbitrageurs determine whether to purchase or redeem creation units based on the relative values of ETF shares in the secondary market relative to the securities contained in the ETF’s portfolio. Arbitrage activity in ETF shares is also affected by the liquidity of the securities in an ETF’s portfolio. There are no exact figures available about ownership of ETFs. In the United States, aggregate ETF ownership is estimated to be 50% retail and 50% institutional investor; however, institutions account for more than 80% of trading activity. New ETFs are typically held entirely by institutional investors and retail ownership builds up over time as investors become familiar with the product. In contrast, in many emerging markets, ETFs are mostly owned by retail investors.

Evolution of ETFs Growth in ETF Market The first ETF was introduced in Canada in 1989 as the Toronto Index Participation Fund (TIP 35). In 1993, an ETF was introduced in the United States by State Street, the Standard and Poor’s 500 Depository Receipts (SPDR) that tracks the broad market index S&P 500. The Hong Kong Tracker Fund was the first ETF in Asia, introduced in 1999, and the first ETF in Europe was Euro STOXX 50 launched in 2001. At the end of July 2012, there were 4,593 ETFs and Exchange Traded Products (ETPs) as shown in Fig. 1. ETPs include other products such as Exchange Traded Notes (ETNs) that are debt securities. These products are backed by the credit of the issuer. Barclays Bank PLC issued the first ETN, iPath, in 2006. The ETF & ETP markets have grown from $79 billion in 2000 to more than $1.7 trillion in 2012. Investment in ETFs accounts for 40% of the total amount invested in index mutual funds in the United States. The top global ETP providers are listed in Table 1. ETFs in the United States make up 25
30% of total market volume and have topped 40% on some days.1 ETF activity has increased dramatically in the last ten years both in terms of assets and trading volume. Plain vanilla ETFs on broad market indexes account for a large percentage of the activity. ETFs domiciled in the U.S. account for almost 70% of total activity followed by Europe at 25%. The majority of global ETFs track equity indices (75%), followed by fixed income (15%), and commodities (10%). Fixed

81

The Growth of Global ETFs and Regulatory Challenges 2,000.0

3,500

1,800.0

3,000

1,600.0

2,500

1,400.0 1,200.0

2,000

1,000.0 800.0

1,500

600.0

1,000

400.0 500

200.0 – 20

00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 Ju l-1 2

–

Assets (US$ Bn)

00

01

02

03

04

05

06

07

08

09

10

11

July 2012

ETF total

74.3

104.8

141.6

212.0

309.8

412.1

565.6

796.7

711.1

1,036.0

1,311.3

1,350.9

1,541.4

ETF equity

74.3

104.7

137.5

205.9

286.3

389.6

526.5

729.9

596.4

841.6

1,053.8

1,057.4

1,193.4

ETF fixed Inc

0.1

0.1

4.0

5.8

23.1

21.3

35.8

59.9

104.0

167.0

207.3

257.7

308.2

ETF

-

0.0

0.1

0.3

0.5

1.2

3.4

6.3

10.0

25.6

45.7

31.1

34.5

ETP total commodity

5.1

3.9

4.1

6.3

9.3

15.9

32.5

54.6

61.2

119.7

171.3

173.5

181.7

79.4

108.7

145.7

218.3

319.1

428.0

598.1

851.3

772.3

1,155.8

1,482.7

1,524.5

1,723.1

# ETFs

92

202

280

282

336

461

713

1,170

1,595

1,944

2,460

3,011

3,282

# ETPs

14

17

17

18

21

63

170

371

625

750

1,083

1,210

1,311

106

219

297

300

357

524

883

1,541

2,220

2,694

3,543

4,221

4,593

ETF/ETP total

# ETF/ETP total

Fig. 1. Global ETF and ETP Asset Growth, as at the End of July 2012. Source: ETF Landscape
Industry Highlights, BlackRock (2012, July).

income ETFs are linked to money market, government, and corporate debt. Commodity ETFs are mostly on precious metals (particularly gold) because of their low storage costs and nonperishable nature. From January to October 2012, 146 new ETFs were introduced in the United States, lower than the 222 introduced in the same period the previous year.2 By October 2012, 86 closed, resulting in a net increase of 60. Exchanges in both developed and emerging markets now list ETFs. However, the tremendous growth in ETFs is limited to a few countries.

Growing Complexity of ETF Products Since the plain vanilla ETFs were developed, they have evolved over time. The initial ETFs held a basket of securities that replicated the component

82

REENA AGGARWAL AND LAURA SCHOFIELD

Table 1. Top 10 Global ETP Providers ranked by Assets, as at Year End 2012. ETP Provider

# of ETPs

Assets (US$ Bn)

% Market Share

iShares State Street Global Advisors Vanguard PowerShares/Deutsche Banke db x-trackers/db ETC Lyxor Asset Management/Soc Gen ETF Securities Van Eck Associates Corp ProShares Nomura Asset Management Others (185 providers)

607 168 81 202 294 241 297 54 138 39 2,627

711.8 333.4 230.8 76.4 49.2 40.3 29.9 28.1 22.4 20.8 302.3

38.6 18.1 12.5 4.1 2.7 2.2 1.6 1.5 1.2 1.1 16.4

Total

4,748

1,845.4

100.0

Source: ETP Landscape
Global Handbook, BlackRock, Q4 (2012).

securities of broad-based stock market indexes, such as the S&P 500. However, many of the newer ETFs are based on more specialized indexes, including indexes that are designed specifically for a particular ETF, bond indexes, and international indexes. Index-based ETFs track indexes and have specified methodologies that select component securities that are generally highly liquid. For example, the SPDR® Barclays Capital High Yield Bond ETF replicates the performance of the Barclays Capital High Yield Very Liquid Index. The underlying index is a rules-based index designed to reflect the 50 most liquid U.S. dollar-denominated high-yield corporate bonds registered for sale in the United States or exempt from registration. PowerShares offers ETFs that mirror custom-built indexes based on Intellidexes. Some of the index providers that compile and revise the indexes are affiliated with the sponsor of the ETF. They seek to track the price and yield performance of domestic and international equity securities indexes provided by an affiliate. In the United States, ETFs are registered as open-end investment companies, under the Investment Act of 1940. However, several of the features of ETFs are not consistent with the requirement of the Act that apply to mutual funds, and exemptions are needed from the SEC. Therefore, in the United States, ETFs require exemptions from the SEC before starting operations. The SEC provides these exemptions on a case-by-case basis. Instead of providing case-by-case exemption, in 2008, the U.S. SEC proposed new rules to permit ETFs to operate without the need for individual

The Growth of Global ETFs and Regulatory Challenges

83

exemptive orders (see SEC press release, March 4, 2008, “SEC Proposes to Streamline ETF Approval Process”), in order to eliminate barriers to entry and avoid delay. However, the financial crisis of 2008 has delayed action, and the proposed rules have not been implemented. Leveraged and Inverse ETFs Leveraged and inverse ETFs are relatively new types of ETFs that were introduced only in 2006.3 A leveraged ETF tracks the value of an index, a basket of stocks, or another ETF, with the additional feature that it uses leverage. Leveraged ETFs aim to achieve 2 × or 3 × long exposure. Similarly, inverse ETFs provide −1 × or −2 × short exposure. The majority of the activity is in the 2 × and −2 × leveraged products, with much smaller amounts in products with higher leverage and inverse ETFs. Leverage ETFs are popular with hedge funds. An example of a leveraged ETF, the ProShares Ultra Financial ETF (UYG), was introduced in January 2006 and offers double exposure to the Dow Jones U.S. Financial Index. The ETF invests two dollars in a basket that tracks the index, for each dollar of UYG’s net asset value. Leverage is provided by borrowing the second dollar that is invested in the index. Hence, UYG has 2 × long position. The description of the UYG ETF is stated on ProShares website: This ETF seeks a return of 200% of the return of an index (target) for a single day. Due to the compounding of daily returns, ProShares’ returns over periods other than one day will likely differ in amount and possibly direction from the target return for the same period. Investors should monitor their ProShares holdings consistent with their strategies, as frequently as daily.

Similarly, ProShares Short Financials, SEF, −1 ×, seeks a return of −100% of the target index for a single day and was started in June 2008. ProShares UltraShort Financial ETF, SKF, is a short leveraged ETF, −2 × and was first offered in January 2007. This ETF short sells a basket of stocks that track the Dow Jones U.S. Financial Index. The performance of these ETFs is shown in Fig. 2. The difference between traditional ETFs and leveraged ETFs is not simply the exposure to returns, but they are also constructed differently. In a traditional ETF, when authorized participants, for example, institutional investors buy and redeem creation units, the underlying stocks are transferred. Leverage and inverse ETFs are pre-packaged margin products and are constructed using derivatives. They are created and redeemed in cash and not by the transfer of the underlying basket of stocks. These products

84

REENA AGGARWAL AND LAURA SCHOFIELD 800 700 600 500 400 300 200 100 0

DJUSFN

SKF

UYG

SEF

Fig. 2. Dow Jones U.S. Financials Index (DJUSFN), ProShares UltraShort Financials (SKF), ProShares Ultra Financials (UYG), and ProShares Short Financials (SEF). Source: Yahoo Finance.

provide an alternative to direct short selling, and also allow access to leverage. Leveraged ETFs need to maintain a daily ratio of leverage to the benchmark. The daily rebalancing of leverage keeps the specific leverage ratio intact but implies that long-term performance of the ETF may differ significantly from the unleveraged performance of the benchmark index times the leverage ratio. Avellaneda and Zhang (2009) find leveraged ETFs to generally underperform a buy and hold leveraged strategy. Cheng and Madhavan (2009) discuss how the daily rebalancing can create volatility, particularly at the end of the day, and Militaru and Dzekounoff (2010) show that both long and short ETFs can lose money even when the underlying index is flat. This discrepancy is partially driven by the daily rebalancing of leveraged ETFs. Physical versus Synthetic ETFs Physical ETFs hold all or most of the assets in a particular benchmark index. Investors in physical ETFs receive returns from the basket of securities net of expenses and any revenue from securities lending.4 These ETFs can be fully replicated or optimized. Typically, full replication is used for blue-chip indexes in the developed markets. Optimization strategies are more common for broader indexes or for indexes tracking illiquid securities.5 Transparency is high because the portfolio composition is disclosed

The Growth of Global ETFs and Regulatory Challenges

85

regularly, and the underlying index can be easily followed. In the United States, regulatory restrictions on the use of derivatives have resulted in the continued popularity of physical ETFs. In fact, since March 2010, the SEC has not considered exemptive requests from ETFs that would make significant investment in derivatives until it completes a review to evaluate the use of derivatives. ETFs that had already obtained an exemptive relief prior to March continue to operate as usual. Most ETFs in the United States and Asia are physical, whereas synthetic ETFs have become quite popular in Europe. Almost half of the ETFs in Europe are synthetic. The first synthetic/swap-based ETF was introduced in Europe in 2001. Synthetic ETFs use futures, options and swaps to simulate the return of an underlying index unlike physical ETFs that hold assets underlying a benchmark index. Synthetic ETFs have lower tracking error because the use of derivatives in these products makes it possible to more accurately obtain the same return as the underlying. Synthetic ETFs may be needed when physical replication is not possible. For example, commodity ETFs, such as energy-related ETFs tend to be of the synthetic variety. Synthetic ETFs can allow exposure to countries such as India and China that have foreign investor restrictions or Russia that has operational issues. Synthetics can be complex; they can lack transparency and have counterparty risk as explained below. In a synthetic ETF, the ETF does not contain the securities in an index. Instead, the fund enters a total return swap agreement with the swap counterparty based on the return of the underlying index as shown in Fig. 3. The value of the swap is marked to market daily. This setup may leave investors with counterparty risk and insufficient transparency about counterparty exposure. The concern is that the counterparty to the derivatives trade may not be able to meet its obligations due to financial difficulties and may default. If the counterparty defaults, then the ETF may not perform as expected. This risk makes the issue of diversifying counterparty risk important. It also highlights the importance of obtaining high quality. There are also concerns about the lack of transparency of the reference basket. Actively Managed ETFs An ETF that does not “seek to track the performance of a market index by either replicating or sampling the index securities in its portfolio” is considered an actively managed ETF as described in SEC Concept Release: Actively Managed Exchange-Traded Funds.6 In contrast to an index-based ETF, actively managed ETFs need not seek to track a particular index; securities may be selected consistent with the investment objective, without

86

REENA AGGARWAL AND LAURA SCHOFIELD

Collateral basket Collateral basket return

Cash Cash

Cash

ETF provider

Investor Share

S&P 500 ETF Collateral basket return

Share S&P 500 return Often the same banking group

Swap counterparty

Fig. 3. Structure of a Simple Synthetic ETF. Source: Potential Financial Stability Issues Arising from Recent Trends in Exchange-Traded Funds, Financial Stability Board (2012, April 12).

actually replicating or sampling the underlying securities. There are currently more than 40 actively traded ETFs in the United States. The fee for actively managed ETFs is higher than for passive ETFs.

REGULATORY CONCERNS IN DEVELOPED MARKETS Regulators around the world have been showing concerns about the risks associated with the more complex ETFs that may have additional risks associated with their construction and performance. The major concerns are related to: • • • •

Systemic risk and excess volatility Suitability of complex ETFs for retail investors Lack of transparency and liquidity of the securities in the portfolio Securities lending of the ETF itself and the underlying securities

The Financial Stability Board (FSB), established to coordinate the development and promotion of effective regulatory, supervisory, and other financial sector policies, issued a note on ETFs on April 12, 2011.7 FSB has warned that the recent financial innovation in ETFs requires closer monitoring of potential vulnerabilities and warrants the attention of regulators. The main issues that the regulators in Europe and the United States are currently addressing are as follows.

The Growth of Global ETFs and Regulatory Challenges

87

Systemic Risk and Excess Volatility The FSB is concerned that when the same bank serves both as the provider of a synthetic ETF and the swap counterparty, investors may be exposed, if the bank defaults. The concern is that counterparty risk could be a source of contagion and systemic risk, since it entails possibilities of a bank default. Further, many ETFs are cross-listed, and hence, there is potential for contagion and systemic risk in the financial system crossing over country borders. The “Flash Crash” of May 6, 2010 put the spotlight on ETFs as discussed in Box 1.

Box 1: The Flash Crash Equity-based ETFs suffered some of the most severe price dislocations on May 6, 2010 when the Dow Jones Industrial Average plunged by almost 1000 points in 20 minutes, wiping out more than $1 trillion in market value. As a result, 21,000 trades were cancelled, 68% being ETF trades. The day started with unusual volatility with concerns about the European debt and potential Greek default. As reported in the SEC-CFTC study, by 2:30 p.m., the S&P 500 volatility index (“VIX”) was up 22.5% from the opening level and selling pressure had pushed the Dow Jones Industrial Average (“DJIA”) down about 2.5%. 1 Buy-side liquidity in the E-Mini S&P 500 futures contracts (the “E-Mini”), as well as the S&P 500 SPDR exchange traded fund (“SPY”), the two most active stock index instruments traded in electronic futures and equity markets, had fallen from the early-morning level of nearly $6 billion dollars to $2.65 billion (representing a 55% decline) for the E-Mini and from the early-morning level of about $275 million to $220 million (a 20% decline) for SPY. At 2:32 p.m., a mutual fund complex initiated a sell program to sell a total of 75,000 E-Mini contracts, worth approximately $4.1billion as a hedge to an existing equity position, using an automated execution algorithm programmed to feed orders to target an execution rate set to 9% of the trading volume calculated over the previous minute, but without regard to price or time. The Sell Algorithm based only on trading volume, and neither price nor time, executed the sell program extremely rapidly in just 20 minutes. The combined selling pressure from the Sell Algorithm, High-frequency traders and other traders drove the price of the E-Mini down approximately 3% in just four

88

REENA AGGARWAL AND LAURA SCHOFIELD

minutes from 2:41 p.m. to 2:44 p.m. During this same time crossmarket arbitrageurs who did buy the E-Mini, simultaneously sold equivalent amounts in the equities markets, driving the price of SPY also down approximately 3%. Source: CFTC-SEC Report on the Flash Crash, 2010.

There are also concerns that if a counterparty bank finances illiquid assets through swaps in the case of synthetic ETFs, there may be a liquidity mismatch between short-term liabilities and long-term funding, leading to systemic problems if there is huge liquidation of ETFs. In a synthetic ETF, a bank may sell ETF shares in exchange for cash. The cash is invested in a collateral basket, the return of which is swapped by the derivatives desk of the same bank for the return of an index (e.g., S&P 500). The ETF does not contain the securities in an index. Instead, the fund enters a total return swap agreement with the swap counterparty based on the return of the underlying index as shown in Fig. 4. The value of the swap is marked to market daily. This setup may leave investors with counterparty risk. The dual role in swap-based ETFs entering into a derivatives contract with the ETF promoter’s investment banking arm can also cause conflict of interest. FSB suggests rules on selecting collateral, screening for credit quality and liquidity, completing the valuation processes, and limiting derivatives exposure.

Authorized Participant Investment Bank Cash

Index Return +/-Swap Spread

Swap Counterparty Cash

Investment Bank (3)

Custodian

Fig. 4.

ETF

(2)

Collateral

Collateral = Swap Valuation + margin

(1)

Swap-Based ETFs with Over-Collateralized Exposure. Source: BlackRock.

The Growth of Global ETFs and Regulatory Challenges

89

Suitability for Retail Investors Several regulators around the globe have been concerned about the suitability of certain types of ETFs for retail customers. In the United States, FINRA highlighted its focus on ETFs in its 2011 Annual Regulatory and Examination Priorities Letter. They explained that this focus is a result of the complexity of these products along with a considerable increase in their number and trading volume and increased interest by retail investors. In addition to overall sales practice concerns, we have identified marketing materials that appear to omit the material risk disclosures necessary to provide a sound basis for evaluating a product as required by FINRA’s advertising rules. In this regard, FINRA is conducting targeted exams to gather information on advertising and sales literature pertaining to ETPs that are not registered investment companies.

In its 2009 Regulatory Notice 09-31, FINRA pointed out that the performance of leveraged and inverse ETFs over longer periods of time can differ significantly from their stated daily objective due to the effects of compounding. Therefore, according to FINRA, these products are unsuitable for retail investors who plan to hold them for longer than one trading session, particularly in volatile markets. As early as 2001, FINRA, in its Fall 2001 Regulatory and Compliance Alert, discussed disclosure of ETF performance. ETF returns are calculated based on NAV; however, ETF shares may trade at a discount or premium. Therefore, FINRA raised concerns that only NAV-based returns might not provide a complete picture of performance. In 2011, the U.K. Financial Services Authority (FSA) raised concerns about the suitability of leveraged ETFs for retail investors.8 The FSA plans to take a much more interventionist approach to the regulation of retail financial services. FSA believes that the previous approach of ensuring that sales processes are fair and that product disclosure is transparent has proved insufficient to protect retail customers, and a new regulatory approach, involving earlier intervention, is needed. FSA is not banning leveraged ETFs, but it wants the starting point to be that these products are unsuitable for most retail customers. Therefore, anyone promoting them would need to provide justification. In contrast with the SEC, the FSA has warned about the risks involved in leveraged ETFs but not inverse ETFs. The ETF exemptive rules proposed by the SEC in 2008 included a condition requiring each ETF to agree not to market or advertise the ETF as an open-end fund or mutual fund and to explain that ETF shares are not

90

REENA AGGARWAL AND LAURA SCHOFIELD

individually redeemable. This condition was designed to help prevent retail investors from confusing ETFs with traditional mutual funds. Similarly, the proposed rule would require each ETF relying on the rule to identify itself in any sales literature as an ETF that does not sell or redeem individual shares and explain that investors may purchase or sell individual ETF shares in secondary market transactions that do not involve the ETF. These proposed rules were shelved by the SEC due to the financial crisis of 2008. A recent article in the Wall Street Journal noted that iShares, the largest ETF manager, warned that some ETF providers are not doing enough to make their products safe.9 The European Securities and Markets Authority (ESMA) issued a discussion paper on “ESMA’s policy orientations on guidelines for Undertakings for Collective Investment in Transferable Securities (UCITS) Exchange-Traded Funds and Structured UCITS,” on July 21, 2011. ESMA determined that the regulatory regime related to UCITS ETFs is not sufficient and is examining possible measures that could mitigate the risk of some of these complex products. UCITS may put limits on the sale of complex ETFs to retail customers. One suggestion has been to divide European UCITS products into complex and noncomplex, and restrictions could be placed on the distribution of complex products to retail customers. However, as of now, there is no consensus about regulatory approaches, and the industry is proactively taking steps to address the suitability issues.

Transparency and Liquidity Regulators have expressed concerns about several aspects of insufficient transparency, including counterparty exposure, collateral, and underlying indexes that are proprietary in many cases. In addition to counterparty risk, regulators are concerned about transparency and disclosure related to counterparty exposure. Investors need to have sufficient and timely disclosure about counterparty exposure. The financial crisis of 2008 heightened concerns about the quality of collateral posted; this is another area that needs further transparency. Finally, the lack of transparency of the reference basket for complex ETFs is also of concern. Another issue is that ETFs offer on-demand liquidity to investors even though the underlying assets might not be liquid. During market meltdowns, investors may demand massive redemptions. Even if redemption is in-kind/cash, there would be issues about liquidity risk of ETF providers

The Growth of Global ETFs and Regulatory Challenges

91

and counterparties. It may be noted that UCITS has provided greater flexibility in the use of derivatives in ETFs, leading to a large number of synthetic ETFs being introduced in Europe, compared to the United States. Swap-based ETFs can be UCIT compliant if they satisfy certain conditions, such as use of eligible counterparty. Securities Lending The securities lending business in ETFs is extremely active and is a significant source of income for investors and ETF managers. One of the concerns expressed is that hedge funds are using ETFs to short stock indexes, sometimes resulting in mismatches between outstanding ETF shares, the number of shares short and the actual ownership of underlying index assets by the ETF. FSB has expressed concerns that the low margins in plainvanilla ETFs provide incentive for aggressive securities lending. Concerns about liquidity, counterparty, and collateral risk exist on the securities lending aspect of the business. Similar concerns have been expressed by the International Monetary Fund and by the Bank for International Settlements (Ramaswamy, 2011) in their notes on ETFs (IMF, 2011).

RESPONSE TO REGULATORY CONCERNS In response to regulatory concerns and after the crisis of 2008, there have been a number of moves to mitigate risks related to ETFs, including the move toward multiple counterparties and to have over collateralized swap exposure. Multiple counterparties also allows for competitive swap pricing. In order to address transparency issues, there are recommendations to provide full disclosure of collateral holdings and index holdings. Recently, swap-based ETFs have started to report collateral holdings, index holdings, swap counterparties, and swap pricing on their websites. European synthetic ETFs are UCITS and cannot have more than 10% exposure to a swap counterparty. In the funded swap ETFs, introduced in Europe in 2009, the counterparty posts collateral assets with a third party custodian. The collateral belongs to the funds, and hence, the risk of counterparty default is mitigated. In August 2011, the Hong Kong Securities and Futures Commission mandated 100% collateralization of counterparty risk when derivatives are used to replicate index performance. As of August 2011, there were

92

REENA AGGARWAL AND LAURA SCHOFIELD

49 synthetic ETFs listed in Hong Kong, 13 of them were domestic. If equity is used for collateral then coverage has to be 120%. The Commission also requires all synthetic ETFs to carry an “X” in front of the name. Collateralization levels have to be made available on the ETFs website. In Australia, no more than 10% of the ETFs net asset value can be swapped out using derivatives. Also, only authorized deposit-taking institutions or authorized foreign banks can be counterparties. BetaShares in Australia decided to convert its two synthetic ETFs into physical ETFs due to the emerging regulatory concerns. The U.S. SEC has been considering not requiring ETFs to obtain an exemptive relief if they satisfy three conditions that facilitate the arbitrage mechanism: transparency of the ETF’s portfolio, disclosure of the ETF’s Intraday Value, and listing on a national securities exchange. An ETF can rely on the proposed rule only if a national securities exchange disseminates the Intraday Value at regular intervals during the trading day. Further, in the case of ETFs that have a stated investment objective of maintaining returns that correspond to the returns of a securities index, their providers need to disclose on their website the identities and weightings of the component securities and other assets of the index. The proposed rule does not limit the types of indexes that an ETF may track or the types of securities that comprise any index. Thus, the rule does not limit the exemption to ETFs investing in liquid securities or asset. Instead, it requires ETFs to comply with the liquidity guidelines applicable to all open-end funds. The ETF should be listed on a national securities exchange, and the national securities exchange typically disseminates the Intraday Value of ETF shares at 15-second intervals throughout the trading day, thereby providing institutional investors and other arbitrageurs the information necessary to engage in ETF share purchases and sales on the secondary market and purchases and redemptions with the fund, which should help keep ETF share prices from trading at a significant discount or premium. These proposed rules were shelved by the SEC due to the financial crisis of 2008. The industry itself has become proactive and is taking steps to address the criticisms. For example, in October 2011, BlackRock recommended the following reforms:10 • • • • •

Clear labeling of product structure and investment objectives Frequent and timely disclosure for all holdings and financial exposure Clear standards for diversifying counterparties and quality of collateral Disclosure of all fees and costs paid, including those to counterparties Universal trade reporting for all equity trades, including ETFs

The Growth of Global ETFs and Regulatory Challenges

93

The intention of these recommendations is to reduce the risks and increase transparency in areas that have concerned regulators and market participants. Many ETF sponsors have voluntarily adopted many of these best practices. Regulatory mandates and voluntary reforms by market participants can ensure that ETFs will continue to be a safe and useful product for investors.

EMERGING MARKETS AND THE CASE OF INDIA Several emerging markets now trade ETFs; in addition, emerging market ETFs are also listed in foreign markets. Broad-based emerging market ETFs were introduced almost ten years ago. We use India as an example of a growing emerging market that has ETFs in order to examine the role of ETFs in emerging markets. Emerging market ETFs have grown significantly over the last decade and now investors can access almost all the MSCI emerging market countries. In 2010, there were 450 emerging market ETFs/ETPs with 869 listings on 38 exchanges from 32 countries from 94 providers with $193b in assets. MSCI Emerging Markets ETF is the largest, trading in the United States with assets greater than $45 billion.11 The appendix provides a global listing of all ETFs. ETF activity in Asia is quite limited relative to other regions. Japan, Hong Kong, Korea, and Taiwan have the most ETF activity in Asia. Among the BRIC countries of Brazil, China, India, and Russia, ETF activity is highest in China and Brazil, followed by India and barely exists in Russia. ETF activity is concentrated only in a few countries around the world. As of May 2011, there were 47 ETFs that offer Russian exposure, 69 ETFs offer exposure to Brazil, 160 offer exposure to China, and there are 43 India-related ETFs listed in the United States. In May 2011, Direxion Funds filed with the SEC to introduce nine new India-related ETFs that are not leveraged.12 Emerging Global Advisors has also filed for additional Indian ETFs with nine of them focusing on different sectors of the economy. These ETFs provide investors another option to obtain easy exposure to foreign markets. ETFs trading in emerging markets are typically of the vanilla type with synthetic or leveraged ETFs not being allowed in most emerging markets. One of the issues in emerging markets is that only stocks in broad-based indexes tend to be liquid; therefore, ETFs have been limited to broad indexes. The liquidity issue raises concerns about the spreads of ETFs.

94

REENA AGGARWAL AND LAURA SCHOFIELD

In contrast to developed markets, the ETF market in countries such as India is dominated by retail investors. Hence, securities regulators are even more inclined to be conservative in allowing complicated products. In order to trade ETFs in India, investors need demat/broking accounts, and many Indian investors do not have these accounts and therefore do not consider ETFs. Banks play a large role in the Indian financial markets and are the biggest distributors. They find it easier to sell open-end mutual funds that do not require demat accounts. They also do not want to be seen as selling stock market products for the fear of additional regulation and scrutiny. There are 31 ETFs listed in India as of September 2011, with total assets of $2 billion. They are listed on the NSE and/or BSE. Growth in Gold ETFs have seen a rising trend as shown in Fig. 5; however, other ETFs have seen a decline in activity from 2007 to 2011. Gold ETFs are backed by physical holding of gold of 99.5% purity. There is no wealth tax on gold ETFs. Of the total, 65% is invested in gold, 25% in equity, and the rest in money markets as of June 2011.13 Equity ETFs represent large cap and small cap stocks on major indexes; sector-based ETFs are mostly bankrelated, and there is also an ETF related to infrastructure. There are two international ETFs, linked to Nasdaq 100 and the Hang Sang Index. There is only one fixed income ETF, Liquid BeES, and it invests in short-term debt instruments. These are plain vanilla ETFs, as synthetic ETFs are not permitted in India. The situation is similar in most emerging markets. The expense ratios of the ETFs are typically quite reasonable, ranging from 0.50% for the local broad-based indexes to 1% for gold and international ETFs.14

6000

10000

5000

8000

Other ETFs (RHS)

Gold ETFs (LHS)

4000

6000

3000 4000

2000

2000

1000

Fig. 5.

Feb-11

Apr-11

Oct-10

Dec-10

Jun-10

Aug-10

Feb-10

Apr-10

Oct-09

Dec-09

Jun-09

Aug-09

Feb-09

Apr-09

Oct-08

Dec-08

Jun-08

Aug-08

Feb-08

Apr-08

Oct-07

Dec-07

Jun-07

Aug-07

0 Apr-07

0

Growth in Indian ETFs. Source: Mutual Fund Category Analysis, HDFC Securities (2011, June 28).

The Growth of Global ETFs and Regulatory Challenges

95

Local regulators in emerging markets have typically allowed only simple ETFs in the local market. However, foreign providers can create and list complex ETFs in the foreign market. For example, in 2010, ETF provider Direxion introduced the Direxion Daily India Bull 2 × ETF (INDL), which seeks daily investment results before fees and expenses of 200% of the price performance of the Indus India Index. Similarly, the Direxion Daily India Bear 2 × ETF (INDz) seeks −200% of the price performance of the Index. At the end of 2011, Direxion converted these 2 × ETFs to 3 ×. The underlying Indus India Index is designed to replicate the Indian equity market as a whole, through a group of 50 Indian stocks selected from a universe of the largest companies listed on NSE and BSE. Emerging market regulators have been appropriately cautious in not allowing complex ETFs. In countries such as India, trading in ETFs has been quite limited relative to the United States. and Europe. ETFs based on broad market indexes with sufficient liquidity appear to be suitable products for retail customers.

CONCLUSION ETFs have grown tremendously during the last decade and have become a significant part of the equity market activity; hence, regulators are keeping a close watch on any potential impact of these products on financial stability and market volatility. Many ETFs are cross-listed and hence contagion and systemic risk in the financial system cross over country borders. The growth of ETFs has been accompanied by innovation and complexity in some of these products. The suitability of some complex ETFs has also been of concern to regulators. In some countries, regulations do not allow complex ETFs which are more widespread in regimes where the regulatory structure is less stringent. Synthetic ETFs are more prevalent in Europe and have raised the greatest concerns. In the United States, concerns have been raised about leveraged and inverse ETFs. The industry itself has recognized the concerns about transparency and counterparty risk and has started to address them proactively. ETFs are one of the most successful products introduced on exchanges in recent years. Regulators will need to tread carefully to manage risks and yet not impose unnecessary regulation. There is little by way of data and facts concerning the risks of ETFs. Academic scholars can play a role by conducting comprehensive and unbiased analysis.

96

REENA AGGARWAL AND LAURA SCHOFIELD

NOTES 1. http://www.indexuniverse.com/sections/features/9681-etf-trading-volumesspike-amid-correction.html 2. http://www.reuters.com/article/2012/10/11/us-etflaunches-niche-idUSBRE89 A1G820121011 3. Leveraged ETFs were initially issued by Rydex in 2006. 4. At any point in time some of the ETF’s holdings may have been lent out, with the portfolio temporarily owning other assets taken as collateral. 5. Optimization strategies use a representative portfolio to mimic the index if the index consists of a very large number of stocks or has illiquid securities. The portfolio holds a subset of the index assets, which is expected to deliver the same aggregate return as the overall index. For example, S&P 500 tracker could hold just 100 shares whose performance is expected to be representative of all 500 index stocks. In some cases, the ETF may also hold securities that don’t actually belong to the index; for example, a fixed income ETF facing limited liquidity in a specific bond may choose to diversify into other bonds with very similar characteristics and expected returns (Heaton, 2011). This strategy may however entail higher tracking error. 6. U.S. SEC Concept Release IC-25258, File No. S7-20-01, May 18, 2004. 7. “Potential Financial Stability Issues Arising from Recent Trends in ExchangeTraded Funds,” FSB, 2011. 8. “Retail Conduct Risk Outlook,” Financial Services Authority, February 28, 2011. 9. “Financial News: Turmoil Raises Fears About Synthetic ETFs,” Wall Street Journal, August 14, 2011. 10. “ETFs: A Call for Greater Transparency and Consistent Regulation,” ViewPoint, BlackRock (October 2011). 11. http://etfdb.com/type/region/emerging-markets/ 12. They are: IndiaShares Fixed Income Shares, IndiaShares Mid & Small Cap Shares, IndiaShares Consumer Shares, IndiaShares Energy & Utility Shares, India Shares Financial Shares, IndiaShares Industrial Shares, IndiaShares Infrastructure Shares, IndiaShares Materials Shares, and IndiaShares Technology & Telecommunication Shares. 13. http://www.risk.net/asia-risk/news/2080294/risk-india-etfs-set-grow-indiaregulators-wary-systemic-risk 14. For details, see “Mutual Fund Category Analysis,” HDFC Securities, June 28, 2011.

ACKNOWLEDGMENTS The chapter has benefitted from the comments of Jim Angel, Rasmeet Kohli, Nirmal Mohanty, Adam Patti, and K. N. Vaidyanathan. The author would like to thank David Mann of BlackRock for helpful

The Growth of Global ETFs and Regulatory Challenges

97

discussions. Doria Xu and Sonya Chawla provided excellent research assistance. Aggarwal acknowledges support from the Robert E. McDonough endowment and the National Stock Exchange of India.

REFERENCES Avellaneda, M., & Zhang, J. (2009). Path-dependence of leveraged ETF returns. Working Paper. New York University, New York, NY. Cheng, M., & Madhavan, A. (2009). The dynamics of leveraged and inverse-exchange traded funds. Journal of Investment Management, Winter Issue, 7(4), 43–62. Heaton, C. S. (2011). What is an ETF? Retrieved from http://www.indexuniverse.eu/europe/ features-a-news/8094-what-is-an-etf.html?showall=&fullart=1&start=2 IMF. (2011). Durable financial stability, getting there from here. Global Financial Stability Report. Retrieved from http://www.imf.org/external/pubs/ft/gfsr/2011/01/ Militaru, R., & Dzekounoff, D. (2010, March 1). Trading with leveraged and inverse ETFs. Retrieved from http://www.futuresmag.com/2010/03/01/trading-with-leveraged-and-inverseetfs Ramaswamy, S. (2011). Market structures and systemic risks of exchange-traded funds, Working Paper No. 343. Bank for International Settlements, Basel, Switzerland.

98

REENA AGGARWAL AND LAURA SCHOFIELD

APPENDIX: GLOBAL ETF LISTINGS BY EXCHANGE, AS AT YEAR END 2011 The table shows the number of ETFs, number of total ETF listings, ETF assets under management in dollars (AUM), and the 20-day average dollar trading volume (ADV) for Asia Pacific, Americas and Europe, and Middle East and Africa. The statistics are also shown for each country in the region. Region/ Country Listed

Exchange

Asia Pacific Australia China

Hong Kong India

Indonesia Japan

Malaysia New Zealand Singapore South Korea Taiwan

Thailand

Australian Securities Exchange Shanghai Stock Exchange Shenzhen Stock Exchange Hong Kong Stock Exchange Bombay Stock Exchange National Stock Exchange Indonesia Stock Exchange Osaka Securities Exchange Tokyo Stock Exchange Nagoya Stock Exchange Bursa Malaysia Securities Berhad New Zealand Stock Exchange Singapore Stock Exchange Korea Stock Exchange Taiwan Stock Exchange GreTai Securities Market Stock Exchange of Thailand

# ETFs

# Total Listings

AUM (US$ Bn)

20-Day ADV (US$ Mn)

393

509

$89.2

1,123.8

31

52

$3.3

$20.0

22

22

$7.1

$160.6

15

15

$4.7

$95.4

48

77

$22.9

$165.7

2

2

$0.0

$0.0

19

19

$0.3

$5.8

1

1

$0.0

$0.0

12

12

$13.3

$39.0

77

81

$21.6

$65.0

1

1

$0.0

$0.0

4

5

$0.3

$1.4

6

6

$0.3

$0.2

31

89

$2.3

$7.5

103 15

103 18

$8.6 $4.3

$530.3 $32.3

2

2

$0.0

$0.0

4

4

$0.1

$0.4

99

The Growth of Global ETFs and Regulatory Challenges

Region/ Country Listed

Exchange

Americas Brazil Canada Chile Mexico United States

BM&F Bovespa Toronto Stock Exchange Bolsa Comercio Santiago Mexican Stock Exchange BATS Chicago Board Options Exchange (CBOE) Chicago Stock Exchange

# ETFs

# Total Listings

AUM (US$ Bn)

20-Day ADV (US$ Mn)

1,356

1,810

$993.0

$53,458.7

10 227 0 20 0 0

10 280 50 370 0 0

$1.5 $42.3 $0.0 $8.1 $0.0 $0.0

$29.5 $771.5 $0.0 $220.5 $7,597.7 $155.6

0

0

$0.0

$682.4

0

0

$0.0

$16,945.7

FINRA Alternative Display Facility (ADF) NASDAQ OMX BX

Region/Country Listed

0

0

$0.0

$651.7

NASDAQ Stock Market

82

82

$41.0

$12,646.7

NASDAQ OMX PHLX

0

0

$0.0

$1,407.1

NYSE AMEX

0

0

$0.0

$0.0

National Stock Exchange (NSX) NYSE Arca

0

0

$0.0

$148.5

1,016

1,016

$899.4

$12,200.6

Exchange

Europe, Middle East, and Africa (EMEA) Austria Wiener Borse Belgium NYSE Euronext Brusseis Botswana Botswana Stock Exchange Finland NASDAQ OMX Helsinki France NYSE Euronext Paris Germany Deutsche Boerse Boerse Stuttgart Greece Hungary Ireland Italy Netherlands Norway

Athens Exchange Budapest Stock Exchange Irish Stock Exchange Borsa Italiana NYSE Euronext Amsterdam Oslo Stock Exchange

# ETFs

# Total Listings

AUM (US$ Bn)

20-Day ADV (US$ Mn)

1,262 1 1

4,293 21 28

$268.7 0.0 0.0

$2,964.9 0.0 0.0

0

1

0.0

0.0

1

1

0.2

4.2

277 448 0

511 910 406

43.5 103.8 0.0

345.6 767.5 21.6

3 1

3 1

0.0 $0.0

0.0 $0.0

1 23 26

1 562 117

0.0 2.2 0.6

0.0 261.3 58.5

7

15

0.6

41.9

100

Region/Country Listed Poland Portugal Russia Saudi Arabia Slovenia South Africa Spain Sweden

Switzerland Turkey UAE United Kingdom

Global total

REENA AGGARWAL AND LAURA SCHOFIELD

Exchange Warsaw Stock Exchange NYSE Euronext Lisbon RTS Stock Exchange Saudi Stock Exchange Ljubljana Stock Exchange Johannesburg Stock Exchange Bolsa de Madrid Latibex NASDAQ OMX Stockholm Burgundy SIX Swiss Exchange Istanbul Stock Exchange Abu Dhabi Securities Exchange London Stock Exchange European Reported OTC

# ETFs

# Total Listings

AUM (US$ Bn)

20-Day ADV (US$ Mn)

1

3

0.0

0.2

3

3

$0.1

0.2

1 3 0

1 3 0

$0.0 0.0 $0.0

2.5 0.1 0.0

26

26

2.1

5.0

14 0 24

75 1 64

0.5 0.0 2.7

18.8 0.0 95.7

0 130 12

24 730 12

0.0 44.3 $0.2

26.9 269.2 11.1

1

1

$0.0

$0.0

258

773

67.6

572.8

0

0

3,011

6,612

461.8 1,350.9

Source: ETF Landscape
Global Handbook, BlackRock, Q4 (2011).

57,547.4

OVERCONFIDENCE, CORPORATE GOVERNANCE, AND GLOBAL CEO TURNOVER Hyung-Suk Choi, Stephen P. Ferris, Narayanan Jayaraman and Sanjiv Sabherwal ABSTRACT Purpose
To determine what role overconfidence plays in the forced removal of CEOs internationally. Design/Methodology
The study makes use of the Fortune Global 500 list. Findings
We find that overconfident CEOs face significantly greater hazards of forced turnovers than their non-overconfident peers. Regardless of important differences in culture, law, and corporate governance across countries, overconfidence has a separate and distinct effect on CEO turnover. Overconfident CEOs appear to be at greater risk of dismissal regardless of where in the world they are located. We also discover that overconfident CEOs are disproportionately succeeded by other overconfident CEOs, regardless of whether they are forcibly removed or voluntarily leave office. Finally, we determine that the dismissal of overconfident CEOs is associated with improved market performance, but only limited enhancement in accounting returns.

Advances in Financial Economics, Volume 16, 101
136 Copyright r 2013 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3732/doi:10.1108/S1569-3732(2013)0000016004

101

102

HYUNG-SUK CHOI ET AL.

Originality/Value
This study is unique with its examination of overconfidence among global CEOs rather than being limited to U.S. chief executives. It also provides insight into how overconfidence is related to national cultures, legal systems and corporate governance mechanisms. Keywords: Overconfidence; turnover; corporate governance; behavioral finance

INTRODUCTION Goel and Thakor (2008) contend that CEOs who are overconfident should have a higher likelihood of forced turnover. This occurs because overconfident CEOs overestimate their own skills and information acquisition abilities, consequently overinvesting in projects that reduce firm value. This behavior by CEOs will prod boards of directors to remove such individuals and seek a new CEO who will maximize firm value. Campbell, Gallmeyer, Johnson, Rutherford, and Stanley (2011) empirically test this conjecture and find evidence for it in a large cross-section of U.S. firms.1 But given the significant differences across countries in culture, investor protections, and corporate governance practices reported by researchers such as La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998, 1999, 2000), Stulz and Williamson (2003), Doidge, Karolyi, and Stulz (2007), and Aggarwal, Erel, Stulz, and Williamson (2009), it is not clear whether overconfident CEOs worldwide are disciplined comparable to their U.S. counterparts. This uncertainty regarding the international retention of overconfident CEOs serves as our fundamental research question: Is the forced termination of overconfident CEOs a global phenomenon or is it largely a U.S. practice? One factor influencing the disciplining of CEOs through termination is the extent of legal protections provided to minority investors by the mechanisms of corporate governance. DeFond and Hung (2004) provide evidence that strong corporate governance is important for the discipline of poorly performing managers. DeFond and Hung (2004), however, use country-level proxies for their measure of corporate governance and consequently are unable to capture variation in corporate governance across firms within the same country. Their proxies imply a homogeneity in corporate governance across firms within a country that does not exist in practice.2 Consequently, in this study, we use firm-level governance data that permits a more accurate measure of the ability of corporate governance to influence the CEO turnover decision.

Overconfidence, Corporate Governance, and Global CEO Turnover

103

The extensive literature on cross-cultural psychology (e.g., Goszczytiska, Tyszka, & Slovic, 1991; Teigen, Brun, & Slovic, 1988; Weber & Hsee, 2000; and Whitcomb, Onkal, Curley, & Benson, 1995) offers another country factor that might result in differences regarding the retention of overconfident CEOs. Specifically, these studies find that a country’s culture influences how overconfidence might influence decision-making and whether overconfidence is viewed as a positive trait. Consequently, national culture can affect the decisions made by boards regarding the retention or termination of an overconfident CEO. We show that overconfidence is a significant determinant of forced CEO turnover and that its influence is not restricted to the U.S. We find that this result holds even after controlling for the determinants of turnover established in the prior literature. We determine that overconfident CEOs face a significantly greater hazard of forced turnovers than non-overconfident CEOs. These results are robust to controls for CEO age, equity performance, corporate governance, and the country in which the firm is headquartered. Most importantly, we establish that the effect of overconfidence on CEO turnover is a separate effect, distinct from CEO demographics, firm performance, and governance, and is international in its occurrence. We also examine a follow on question regarding the nature of firm performance following forced turnover. We find that the market-adjusted stock return is significantly higher following the forced removal of a CEO. However, we do not find a significant improvement in operating performance following a forced turnover. The findings in this study advance the literature in two different ways. First, our results extend the growing literature on the effect of overconfidence on CEO corporate decision-making and firm value. Roll (1986) proposed in his seminal “hubris” theory of acquisitions that optimism and managerial overconfidence have explanatory power for corporate decisionmaking. He argues that successful acquirers might be optimistic or overconfident and thereby overstate the synergies associated with an acquisition. Malmendier and Tate (2005, 2008) provide important evidence on the implications of CEO overconfidence for corporate investments and acquisition decisions. More directly, we extend the evidence supporting forced turnover of CEOs of the U.S. firms in Campbell et al. (2011) to a global context. Our findings also provide international evidence consistent with the predictions of Goel and Thakor (2008). Further, we add to a small, but growing literature that examines international CEO turnover. Although there has been extensive research regarding CEO turnover in the U.S., few studies examine this issue internationally.3

104

HYUNG-SUK CHOI ET AL.

Of the studies that do focus on global CEO turnover, most tend to focus on a single country. For example, Kaplan (1994a) examines German firms while Kaplan (1994b) and Kang and Shivdasani (1995) study Japanese firms. More recently, Dahya, McConnell, and Travlos (2002) provide evidence on turnover among U.K. firms while Volpin (2002) focuses on Italian firms. We contribute to the literature by investigating the patterns and determinants of CEO turnover for a more comprehensive set of global firms, spanning a variety of countries. This approach allows us to better understand the relation between CEO turnover and national cultures, legal systems, and corporate governance mechanisms. The remainder of this study is organized as follows. Section 2 motivates our central research question by providing a discussion of how international differences in culture, legal origins, and corporate governance practices might produce a global relation between overconfidence and CEO retention that differs from that observed for the U.S. Section 3 describes our data and a description of how we measure overconfidence. Section 4 presents and discusses our sample summary characteristics and background empirical findings regarding the global distribution of CEO overconfidence and overconfidence patterns in CEO succession. We explicitly examine the relation between CEO turnover and overconfidence with our multivariate analysis in Section 5. We examine corporate performance changes following CEO turnover in Section 6. We provide a brief summary and discussion of our results in Section 7.

COUNTRY EFFECTS ON MANAGERIAL OVERCONFIDENCE Both the cross-cultural psychology and finance literatures suggest important reasons why there might be international effects in the relation between overconfidence and turnover. These literatures collectively suggest that differences in national cultures generate varying levels of overconfidence. Hence the findings regarding the decision-making of overconfident U.S. CEOs and their turnover might not apply with equal veracity to their international counterparts. Weber and Hsee (2000) contend that national cultures generate critical differences in four areas of judgment and decision-making. These areas, in turn, shape the extent to which individuals can be characterized as overconfident. Studies such as Phillips and Wright (1977), Wright and Phillips

Overconfidence, Corporate Governance, and Global CEO Turnover

105

(1980), and Pollock and Chen (1986) show that there are trans-cultural differences in probabilistic thinking and the quality of probability judgments, which are critical components of assessing business opportunities. Douglas and Wildavsky (1982) examine the influence of culture on both the perception of risk and its acceptability. They conclude that cultural differences in attitudes toward risk can be explained in terms of their contribution to maintaining a particular way of life. Slovic (1997) suggests that cultural differences in trust in institutions are likely to explain national differences in perceived risk. Weber and Hsee (1998) examine differences in risk preferences and conclude that members of collectivist cultures take more risk because their social networks can protect them against catastrophic outcomes. Extensive work by researchers such as Von Winterfeldt and Edwards (1986), Simon (1990), Shafir, Simonson, and Tversky (1993), Damasio (1993), and Goldstein and Weber (1995) documents significant cultural differences in the mix of the analytical and intuitive for decisionmaking. In aggregate, these studies describe how a country’s culture helps to shape the overconfidence that is exhibited by a CEO. These studies further imply that what a board of directors views as acceptable behavior by an overconfident CEO will also vary. Beyond different cultural values and norms, differences in corporate governance and the protections provided to minority investors might also effect the relation between overconfident CEOs and their likelihood of being terminated. In a series of studies, La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1997, 1999, 2000) describe how differences in a country’s legal origins result in differences in capital market development, market transparency, ability to enforce contracts, and the strength of corporate governance over insiders. These factors, in turn, affect how CEOs decide and the confidence they have in their decision-making ability. They also affect the ability of monitors such as a board of directors to discipline them.

DATA AND THE MEASUREMENT OF OVERCONFIDENCE Data and Sample Construction Fortune magazine provides an annual ranking of the 500 largest companies of the world based on revenue. We begin our sample construction by compiling these annual lists over the years 2000
2006. From these annual lists,

106

HYUNG-SUK CHOI ET AL.

we create a dataset of all non-bank firms that appear at least once in this list and the countries in which these firms are headquartered. Because of the political issues associated with the disciplining of CEOs in state-owned enterprises, we exclude such firms from our sample. For a firm in our dataset, we include all of the firm’s CEOs over our sample period. During the years when a firm is not in the Fortune Global 500 list or is on the list during 2000
2003 when CEO information was not included, the names of the CEOs are hand collected from a variety of sources. The biographical data of all the CEOs such as their date of birth, birthplace, nationality, and tenure with a firm are also hand collected from various sources such as Mergent Online, individual corporate web sites, financial statements, and other online sources. Country-level characteristics are obtained from several sources. The legal regime for countries is drawn from the classification taxonomy constructed by La Porta et al. (1998). National culture dimensions are those developed by Hofstede (2001). We consider a country to be above average on a particular cultural dimension if it has a Hofstede score above the global median score for that dimension. We obtain firm-level accounting data from the Compustat Global and Compustat North America databases. We measure the size of a firm as the log of assets at the beginning of the year and the accounting rate of return as EBIT divided by the total assets. We convert accounting data other than ratios to U.S. dollar using the exchange rates obtained from the Compustat Global database. Items measured at a specific time, such as assets, are also converted from local currency to U.S. dollar based on the exchange rate at that time. Items measured over a year, such as sales, are converted from local currency to U.S. dollar based on the 12-month average exchange rate over that year. The stock market performance of the firm is market-adjusted. The market returns for each country are proxied by the MSCI country index. All stock market data are obtained from Datastream. The corporate governance data are taken from RiskMetrics Group’s Corporate Governance Quotient (CGQ) dataset. The construction and measure of the corporate governance data are discussed in greater detail in the following section.

The Measurement of Overconfidence Malmendier and Tate (2008) use the propensity of managers of the acquiring firms to hold in-the-money equity options as their primary measure of

Overconfidence, Corporate Governance, and Global CEO Turnover

107

managerial overconfidence. Unfortunately, such option-holding data is not available for international CEOs. Thus, a comparable measure of overconfidence cannot be constructed for our sample. But Malmendier and Tate (2008) and more recently Ferris, Jayaraman, and Sabherwal (2012) and Hirshleifer, Teoh, and Low (2012) estimate an overconfidence measure based on press releases that we can calculate for our set of global firms. That is, the descriptions of CEOs as contained in public news articles can be used to measure their overconfidence. Malmendier and Tate (2008) observe that this proxy provides direct insight into the type of person classified as overconfident and its strength is its ability to measure CEO beliefs as assessed by outsiders. Consistent with Malmendier and Tate (2008), Ferris et al. (2012), and Hirshleifer et al. (2012), we measure overconfidence based on how the market perceives the confidence level of a CEO prior to turnover. Our proxy for the market’s perception is based on the Factiva database, which contains articles from global news sources. For each CEO of a firm, we record the number of articles related to the firm in Factiva during 1996
2006, but prior to the year of the individual’s departure as CEO, that refer to the CEO using the terms (a) “confident” or “confidence,” (b) “optimistic” or “optimism,” (c) “not confident,” (d) “not optimistic,” and (e) “reliable,” “cautious,” “conservative,” “practical,” “frugal,” or “steady.” We then compare the number of articles that portray a CEO as confident and optimistic to the number of articles that portray him as not confident, not optimistic, reliable, cautious, conservative, practical, frugal, or steady. That is, we classify a CEO as overconfident if a + b > c + d + e.4 We adopt the following strategy to decide whether the turnover of the CEO is forced or voluntary. We review the news releases surrounding our sample of turnover announcements. We categorize the turnover as voluntary if any one of the following reasons is stated: (a) the CEO retired; (b) the CEO was an interim CEO and this was known at the start of his/her tenure; (c) the company was acquired by another company; (d) the CEO continued on as chairman; or (e) the CEO resigned to become CEO of another company. We categorize the turnover as “forced” if any of the following is mentioned surrounding the turnover announcement: (a) accounting/financial scandal; (b) poor performance of the firm; (c) management conflicts; or (d) rumors that the CEO was removed by the board. In the following section, we present our sample summary characteristics and empirical findings regarding CEO turnover and overconfidence.

108

HYUNG-SUK CHOI ET AL.

EMPIRICAL RESULTS CONCERNING TURNOVER AND OVERCONFIDENCE Sample Characteristics Table 1 presents a series of descriptive panels regarding the 361 CEO turnovers that serve as the sample of this study. Using the criteria discussed in the immediately preceding section, we classify 106 (29.4%) of our turnovers as forced while the remaining 255 (70.6%) turnovers are assigned as other (i.e., retirements and voluntary resignations). Panel A of Table 1 provides a distribution of our sample turnovers by industry. Most of the turnovers occur in the manufacturing industry (39.9%) followed by finance (20.5%) and transportation (16.9%). The fewest turnovers occur in technology (4.7%) and services (3.3%). The percentage of total turnovers classified as forced is the highest in services (41.7%), followed by trade (35.8%) and finance (28.4%). Panel B contains a distribution of CEO turnover by year. We observe that the total number of turnovers is fairly uniformly distributed across the 7 years of our sample. The fewest turnovers occur in 2000, the first year of our sample period. The greatest number is observed in 2005, when there are 71 turnovers, representing 19.7% of our sample. The largest number of forced turnovers occurs in 2002 with 29, followed by 19 in years 2003 and 2005. Forced turnover as a means of CEO disciplining occurs most frequently in 2002, when 45.3% of all turnovers can be classified as forced. We construct a geographical distribution of our sample in Panel C where we identify the continent or country where our sample turnovers occur. Not surprisingly, we obtain nearly half of our sample (48.8%) from the U.S., Japan is next with a contribution of 13.9% of our sample. U.K. and the rest of Europe collectively contribute over a quarter of the sample. The fewest number of turnovers occurs in the Americas less the U.S. and accounts for only 3.9% of our sample. Clearly, our sample is largely drawn from the developed markets of the U.S., U.K., and the rest of Europe. Similarly, the largest number of forced turnovers occurs in the U.S. with 49, representing 46.2% of this type of turnover. Countries from continental Europe account for another 28 forced turnovers, representing 26.4% of the total forced turnover sample. We examine the influence of legal regime on the incidence of CEO turnover in Panel D. We find that nearly two-thirds of our sample turnovers occur in common law countries, reflecting the strong presence of the U.S. and U.K. firms in our sample. Among the civil law countries, the highest

Overconfidence, Corporate Governance, and Global CEO Turnover

109

Table 1. Distribution of CEO Turnovers. Industry

Total

Forced

Other

74 (20.5%)

21 (28.4%)

53 (71.6%)

Manufacturing

144 (39.9%)

39 (27.1%)

105 (72.9%)

Service

12 (3.3%)

5 (41.7%)

7 (58.3%)

Technology

17 (4.7%)

5 (29.4%)

12 (70.6%)

Trade

53 (14.7%)

19 (35.8%)

34 (64.2%)

Transportation

61 (16.9%)

17 (27.9%)

44 (72.1%)

All

361 (100.0%)

106 (29.4%)

255 (70.6%)

Panel A. Turnover by Industry Finance

Year

Total

Forced

Other

26 (7.2%)

10 (38.5%)

16 (61.5%)

2001

36 (10.0%)

8 (22.2%)

28 (77.8%)

2002

64 (17.7%)

29 (45.3%)

35 (54.7%)

2003

54 (15.0%) 56 (15.5%)

19 (35.2%) 10 (17.9%)

35 (64.8%) 46 (82.1%)

2005

71 (19.7%)

19 (26.8%)

52 (73.2%)

2006

54 (15.0%)

11 (20.4%)

43 (79.6%)

Panel B. Turnover by Year 2000

2004

Country/continent Panel C. Turnover by Country/Continent U.S. Americas other than U.S. U.K.

Total

Forced

Other

176 (48.8%)

49 (27.8%)

127 (72.2%)

14 (3.9%) 30 (8.3%)

4 (28.6%) 13 (43.3%)

10 (71.4%) 17 (56.7%)

110

HYUNG-SUK CHOI ET AL.

Table 1. Country/continent

(Continued )

Total

Forced

Other

Europe other than U.K.

66 (18.3%)

28 (42.4%)

38 (57.6%)

Japan

50 (13.9%) 25 (6.9%)

3 (6.0%) 9 (36.0%)

47 (94.0%) 16 (64.0%)

Asia other than Japan

Legal origin Panel D. Turnover by Legal Regime Civil

French German

Total

Forced

Other

41 (11.4%) 81

14 (34.1%) 19

27 (65.9%) 62

(22.4%)

(23.5%)

(76.5%)

6

3

3

(1.7%)

(50.0%)

(50.0%)

Scandinavian All civil

128

36

92

(35.5%)

(10.0%)

(25.5%)

Common Civil versus common (t-statistic) Measure

Tier

Panel E. Turnover by Hofstede Measures Power distance High Low

233

70

163

(64.5%) −0.38

(30.0%) 0.38

(70.0%)

Total

Forced

Other

37 (10.2%)

12 (32.4%)

25 (67.6%)

324

94

230

(89.8%)

(29.0%)

(71.0%)

0.43

−0.43

85 (23.5%) 276

19 (22.4%) 87

66 (77.6%) 189

(76.5%)

(31.5%)

(68.5%)

−1.61

1.61

101 (28.7%)

251 (71.3%)

t-statistic Uncertainty avoidance

High Low t-statistic

Individualism

High Low t-statistic

352 (97.5%) 9

5

4

(2.5%)

(55.6%)

(44.4%)

−1.74*

1.74*

Overconfidence, Corporate Governance, and Global CEO Turnover

Table 1.

111

(Continued )

Measure

Tier

Total

Forced

Other

Masculinity

High

310 (85.9%)

92 (29.7%)

218 (70.3%)

Low

51

14

37

(14.1%)

(27.5%)

(72.5%)

0.32

−0.32

9 (13.4%)

58 (86.6%)

t-statistic Long-term orientation

High Low

67 (18.6%) 294

97

197

(81.4%)

(33.0%)

(67.0%)

−3.20***

3.20***

t-statistic Panel F. Turnover by Governance Total

Forced

Other

First quartile (Low governance)

77 (25.0%)

17 (22.1%)

60 (77.9%)

Second quartile

72 (23.4%)

27 (37.5%)

45 (62.5%)

Third quartile

76 (24.7%)

22 (28.9%)

54 (71.1%)

Fourth quartile (high governance)

83 (26.9%)

25 (30.1%)

58 (69.9%)

GOV44

Note: This table presents the distribution of our sample CEO turnovers across various categories. Total is the total number of turnovers, Forced refers to involuntary turnovers, and Other refers to all non-forced turnovers such as retirements and voluntary resignations. Panel A reports the number of observations and the proportion (in parentheses) across industry grouping. Comparable data by year is presented in Panel B while Panel C reports by sample country. Panel D contains data by legal regime. Panel E reports turnover data by Hofstede measures. The t-statistics in Panels D and E are for the test that the proportions are equal. CEO turnover data by the firm’s corporate governance is presented in Panel F. GOV44 is a corporate governance index based on Aggarwal et al. (2009). Statistical significance at the 1%, 5%, and 10% levels is indicated by ***, **, and *, respectively.

percentage of turnovers occurs within the German civil law regime, while the fewest are in the Scandinavian civil law countries. Seventy (66%) of the forced turnovers occur within the common law countries compared to only 36 (34%) in the civil law countries. When we examine each legal regime separately, we find that 30% of all turnovers in common law countries are

112

HYUNG-SUK CHOI ET AL.

forced compared to 28% in civil law countries. These results suggest that executive disciplining is equally common across historical legal regimes. Panel E introduces the Hofstede cultural dimensions into our analysis of the international turnover, which are more fully described in the appendix. These dimensions have been used in a number of finance studies (e.g., Chakrabarti et al., 2009; Datta & Puia, 1995; Ferris et al., 2012; Gleason, Mathur, & Mathur, 2000; Kirkman, Lowe, & Gibson, 2006; Kwok & Tadesse, 2006; and Sekely & Collins, 1988) since their creation by Hofstede (1980). These measures consist of five different dimensions of a country’s culture. The power distance index captures the extent to which less powerful members of organizations and institutions within a country both accept and expect that power is distributed unequally. Individualism measures the extent to which individuals are integrated into groups within a country. Masculinity refers to the distribution of roles between genders. The uncertainty avoidance measure addresses a society’s tolerance for uncertainty and ambiguity. It indicates the extent to which that country’s culture emphasizes rules and regulations to avoid risk and to process change incrementally. The last of the Hofstede dimensions is long-term orientation and focuses on the relative cultural importance of thrift, perseverance, tradition, and satisfaction of social obligations. We observe in Panel E that the incidence of turnover is broadly consistent with the underlying traits and behaviors associated with each of these cultural dimensions. We find that total turnover is higher in those nations with a low power distance, indicating that subordinates are less accepting of an unequal power sharing. That is, CEOs are more likely to leave a firm located in a country where subordinates require greater equality in the distribution of power. Likewise, turnover is greater when the national culture de-emphasizes the long-term and focuses on the more immediate. Turnover is also more frequently observed when the cultural tendency to avoid uncertainty is low. High levels of masculinity and individualism are associated with greater CEO removal. These traits imply aggressiveness and a willingness to make difficult decisions that are often required when removing a CEO, especially an overconfident individual. The last panel in Table 1 contains our examination of corporate governance and its influence on the rates of CEO turnover. Using the corporate governance index (GOV44) developed by Aggarwal et al. (2009), we find in Panel F a suggestion that turnover is weakly associated with corporate governance.5 Specifically, we observe that for firms with weak governance as measured by their presence in the bottom quartile of the GOV44 index, there is less total turnover of CEOs when compared to firms in the top

113

Overconfidence, Corporate Governance, and Global CEO Turnover

governance quartile. Similar results occur for the percentage of forced turnovers relative to total turnover. The results, however, are not monotonic across quartiles. Our findings suggest that weak corporate governance reduces the frequency with which boards might remove CEOs, but that average or strong governance provides approximately equivalent CEO disciplining. Table 2 provides summary financial characteristics for our sample firms. Given that our firms are drawn from the Fortune Global 500 list, it is not surprising that they are large, with an average asset value in excess of $88

Table 2. Sample Profile. Mean (Median) Variable

Forced versus Other

Total

Forced

Other

Test statistic

Total assets (MM$)

88,813 (25,105)

131,045 (29,948)

70,661 (22,162)

2.58** (2.93***)

Market value of equity (MM$)

48,079 (14,934)

37,387 (20,257)

52,527 (13,550)

−0.41 (2.61***)

Total debt/Total assets

69.9% (70.9%)

69.0% (66.6%)

70.3% (71.4%)

−0.60 (−0.66)

Market to book ratio

387.0% 287.8% 428.3% (201.4%) (190.7%) (205.8%)

−0.74 (−0.44)

Fixed assets/total assets

60.8% 61.6% 60.5% (60.0%) (63.3%) (59.4%) 125.5% 124.5% 126.0% (115.7%) (112.8%) (117.4%)

0.48 (0.56) −0.23 (−0.90)

Current ratio Accounting rate of return

8.1% (6.4%)

7.3% (6.4%)

8.4% (6.4%)

−1.18 (−1.06)

Return in 1 year prior to turnover

4.1% (1.1%)

−3.8% (−6.9%)

7.4% (2.9%)

−2.29** (−2.87***)

Annualized return over tenure of CEO

13.0% (3.2%)

1.2% (−5.6%)

17.9% (5.6%)

−1.91* (−4.60***)

Note: This table presents the summary statistics for our sample of CEO turnovers. Means and medians (in parentheses) are reported. Total is the total number of turnovers, Forced refers to involuntary turnovers, and Other refers to all turnovers which are not forced such as retirements and voluntary resignations. Variables are collected from the Compustat database as of the time of the CEO turnover by matching the calendar year and month. The accounting rate of return is measured as EBIT divided by the average total assets as of the year of and 1 year prior to the turnover. t-statistics (z-statistics) for the test that the mean (median) is equal for forced and other turnovers are reported in the last column. Statistical significance at the 1%, 5%, and 10% levels is indicated by ***, **, and *, respectively.

114

HYUNG-SUK CHOI ET AL.

billion and a mean market equity capitalization of $48 billion. Our sample firms appear profitable, with a mean accounting return of 8.1%. The accounting return for the set of firms that had forced CEO departures is not statistically different from those firms with voluntary or normal departures. However, the average stock market return in the year prior to turnover is significantly lower for the firms that had forced departures (
3.8%) as compared to firms with voluntary departures (7.4%). When we measure the returns over the entire tenure of the CEO, the mean return of firms where the CEOs are forced out is 1.2%. This is significantly lower than that for firms with voluntary departures, which have an average return of 17.9%.

CEO Succession and Overconfidence An interesting issue is whether an overconfident CEO removed from office is succeeded by another overconfident CEO or do boards tend to hire less confident successors. To examine this issue more fully, we present a transition matrix in Table 3 based on the overconfidence status of both the terminated CEO and the successor. We observe that an overconfident CEO is usually followed by another overconfident CEO, regardless of the circumstances under which the CEO left. This result appears to be different from the findings of Campbell et al. (2011) that boards remove overconfident CEOs and hire less confident CEOs in an effort to maximize share value. When non-overconfident CEOs are forced out, we find that they too tend to be followed by overconfident CEOs. It is only when non-overconfident CEOs leave voluntarily or due to retirement that the successor is less likely to be overconfident. These findings seem to suggest that boards have a preference for hiring overconfident CEOs, regardless of the overconfidence of the previous incumbents or the circumstances of their removal. But that conclusion cannot be fully justified by our data. We only measure overconfidence after the fact, so what we actually document is a situation in which boards tend to hire individuals who are subsequently determined to be overconfident. In Panel B we examine the extent to which the preference for an overconfidence CEO is a global phenomenon and not driven by the behavior of U.S. firms. We find that even for non-U.S. firms, overconfident CEOs are succeeded by other overconfident CEOs, regardless of the manner in which they leave the firm. Interestingly, we find that non-overconfident CEOs who are terminated are also followed by overconfident CEOs. We do not

Overconfidence, Corporate Governance, and Global CEO Turnover

115

Table 3. Transition Matrix. Incumbent (turnover)

Successor

t-statistic

OC

Non-OC

67 (87.0%)

10 (13.0%)

9.60***

OC (other)

81 (69.8%)

35 (30.2%)

4.63***

Non-OC (forced)

14 (73.7%)

5 (26.3%)

2.28**

Non-OC (other)

26 (51.0%)

25 (49.0%)

0.14

Panel A. Aggregate Sample OC (forced)

Country

Incumbent (Turnover)

Panel B. U.S. Versus Non-U.S. Firms OC (forced)

Successor OC

32

6

(84.2%) OC (other) U.S.

Non-OC (other)

3

OC (other)

4

13

OC (other) Civil law Non-OC (forced)

4.12***

(25.0%)

8

2

2.25*

(20.0%)

15

13

Incumbent (turnover)

Panel C. Distribution by Legal Regime OC (forced)

8.08***

(10.3%)

39

(53.6%) Legal origin

(52.2%)

35

(80.0%) Non-OC (other)

−0.19

12

(75.0%) Non-OC (forced)

1.00

(33.3%)

11

(89.7%)

Non-U.S.

2.61**

(34.4%)

6

(47.8%) OC (forced)

5.71***

22

(66.7%)

0.37

(46.4%) Successor

t-statistic

OC

Non-OC

17 (85.0%)

3 (15.0%)

22 (73.3%) 7 (77.8%)

t-statistic

(15.8%)

42 (65.6%)

Non-OC (forced)

Non-OC

8 (26.7%) 3 (22.2%)

4.27*** 2.84*** 1.89*

116

HYUNG-SUK CHOI ET AL.

Table 3. Legal origin

(Continued )

Incumbent (turnover)

Successor OC

Non-OC (other)

Non-OC

13

10

(56.5%) OC (forced)

50

OC (other)

Law

Non-OC (forced) Non-OC (other)

59

27 (31.4%)

7 (70.0%)

3 (30.0%)

13

GOV44

Incumbent (turnover)

15

26 (89.7%)

OC (other) Low

30 (75.0%)

Non-OC (forced)

4 (80.0%)

Non-OC (other)

12 (46.2%)

OC (forced)

34 (82.9%)

OC (other) High

45 (68.2%)

Non-OC (forced)

6 (85.7%)

Non-OC (other)

12 (60.0%)

3.70***

1.31 −0.36

(53.6%)

Successor OC

Panel D. Distribution by Governance OC (forced)

8.60***

(12.3%)

(68.6%)

(46.4%)

0.62

(43.5%) 7

(87.7%)

Common

t-statistic

t-statistic Non-OC

3

6.89***

(10.3%) 10

3.61***

(25.0%) 1

1.50

(20.0%) 14

−0.38

(53.8%) 7

5.53***

(17.1%) 21

3.15***

(31.8%) 1

2.50**

(14.3%) 8

0.89

(40.0%)

Note: This table presents the distribution of 263 CEO turnovers and successions. Forced refers to involuntary turnovers. Other refers to all turnovers which are not forced. Hence it includes both retirements and voluntary resignations. OC refers to overconfident CEOs and Non-OC refers to CEOs who are not overconfident. In the last column, t-statistics for the equality tests are reported. ***, **, and * represent statistical significance at 1%, 5%, and 10% levels, respectively.

Overconfidence, Corporate Governance, and Global CEO Turnover

117

observe this pattern among U.S. firms. We conclude from our examination of CEO succession patterns that the practice of hiring overconfident individuals as CEOs is a common global behavior and not exclusive to the U.S. Indeed, our evidence suggests that the selection of overconfident successors is a more common practice globally than it is in the U.S. We continue our investigation of the global popularity of overconfident CEOs by examining their appointment pattern across legal regimes. That is, to what extent do differences in the set of legal rights provided to shareholders influence the kind of CEO chosen to follow an overconfident CEO? Campbell et al. (2011), for instance, argue that overconfident CEOs are often replaced by more modest CEOs in an effort to enhance share prices. In Panel C, we find again that overconfident CEOs are disproportionally followed by other overconfident CEOs, irrespective of how their appointments are terminated. Firms in both common and civil law regimes more frequently appoint overconfident CEOs to follow previous overconfident CEOs. We conclude our analysis of CEO succession by examining the effect that corporate governance might exert on the type of successor that is chosen. Goel and Thakor (2008) argue that corporate boards acting in the best interest of shareholders will terminate CEOs with excessive overconfidence. We find that the firm’s governance does not appear to have much of an effect on the preference for overconfident CEOs. Firms in the bottom quartile of the Aggarwal et al. (2009) GOV44 index appoint overconfident successors to overconfident CEOs significantly more often than they appoint nonoverconfident individuals. We observe similar behavior for firms with strong governance that are located in the top quartile of the GOV44 index values. These findings continue to show the strong preference that boards have for overconfident CEOs regardless of its composition and independence.

MULTIVARIATE ANALYSIS OF OVERCONFIDENCE AND CEO TURNOVER In this section, we introduce our multivariate analysis of overconfidence and the turnover of CEOs while controlling for other determinants of turnover suggested in both the existing literature and the research questions posed in this study. Given the nature of our data and analysis, we use the Cox semi-parametric proportional hazard model to provide this analysis.6 Table 4 contains our empirical findings. The dependent variable equals one for forced turnover, and zero otherwise. The main control variables include

Family * overconfident CEO

U.S. or Japan × overconfident CEO Family owned firms

Overconfident CEO dummy Age as of turnover High power distance dummy High individualism dummy High masculinity dummy High uncertainty avoidance dummy High long-term orientation dummy Log market capitalization Market performance Firm size Strong board governance dummy Number of articles mentioning the CEO Legal dummy for common law U.S. or Japan dummy

Variable

(0.05)

(0.03) (0.17)

(0.99)

(0.23)

(0.13)

(0.01)

(0.01)

(0.07) (0.95) (0.54)

(0.76)

−0.04 −1.53

10.97

0.80

1.06

−2.58

−0.32

−0.01 0.01 0.21

0.00

pvalue

0.79

Coefficient

Model 1

1.00

0.99 1.01 1.23

0.73

0.08

2.89

(0.82) (0.48)

0.33

(0.07) (0.85) (0.43)

(0.01)

(0.01)

(0.10)

(0.34)

(0.99)

(0.04) (0.17)

(0.04)

pvalue

0.00

−0.01 0.02 0.27

−0.37

−2.50

1.18

0.68

11.30

6E + 04

2.23

−0.04 −1.56

0.80

Coefficient

Model 2

1.39

1.00

0.99 1.02 1.31

0.69

0.08

3.25

1.98

8E + 04

0.96 0.21

2.23

(0.05) (0.32)

2.46 −0.94

(0.27)

(0.84)

(0.05) (0.73) (0.25)

(0.00)

(0.00)

(0.19)

(0.22)

(0.99)

(0.02) (0.63)

(0.06)

0.52

0.00

−0.01 0.03 0.40

−0.96

−3.24

1.03

0.95

12.95

−0.04 −0.63

1.56

Hazard Coefficient pratio value

Model 3

0.39

11.69

1.69

1.00

0.99 1.03 1.49

0.38

0.04

2.80

2.58

4E + 05

0.96 0.53

4.76

(0.64) (0.61)

−0.59

(0.58)

(0.51)

(0.07) (0.91) (0.48)

(0.03)

(0.02)

(0.11)

(0.29)

(0.99)

(0.04) (0.21)

(0.04)

−0.51

0.26

0.00

−0.01 0.01 0.24

−0.33

−2.34

1.17

0.72

11.44

−0.04 −1.40

0.87

Hazard Coefficient pratio value

Model 4

Cox Proportional Hazards Model for Forced Turnover.

0.96 0.22

2.20

Hazard ratio

Table 4.

0.55

0.60

1.30

1.00

0.99 1.01 1.28

0.72

0.10

3.21

2.06

9E + 04

0.96 0.25

2.38

Hazard ratio

−0.63

−0.51

−0.79

2.42

0.46

0.00

−0.01 0.03 0.36

−0.95

−3.08

0.97

0.99

13.24

−0.05 −0.41

1.55

(0.59)

(0.64)

(0.41)

(0.05)

(0.33)

(0.50)

(0.05) (0.78) (0.31)

(0.00)

(0.00)

(0.22)

(0.18)

(0.99)

(0.02) (0.75)

(0.07)

pvalue

Model 5 Coefficient

0.53

0.60

0.45

11.24

1.59

1.00

0.99 1.03 1.44

0.38

0.05

2.63

2.69

6E + 05

0.96 0.67

4.69

Hazard ratio

43.30

535 Yes

535

Yes

(0.00) 72

43.20

72

(0.00)

48.70

Yes

535

72

(0.00)

50.50

Yes

535

72

(0.00)

56.45

Yes

535

72

(0.00)

Note: Overconfident CEO dummy is an indicator variable with a value of one for overconfident CEOs. Age as of turnover is the age of CEO as of the turnover. Five Hofstede measures are also examined: power distance, individualism, masculinity, uncertainty avoidance, and long-term orientation. Log Market Capitalization is the logarithm of the stock market capitalization of the country in U.S. dollar terms. Market performance is the 1-year excess return of the company’s stock prior to the CEO turnover. Firm size is the log of the market value of equity of the company at the beginning of the CEO turnover year. Strong governance dummy takes a value of one if GOV44 for the firm is greater than the median GOV44, where GOV44 is the governance index used by Aggarwal et al. (2009). Number of articles mentioning the CEO is the total times the CEO was referred to in the press. Legal dummy for common law is an indicator variable for common law countries. U.S. or Japan dummy is an indicator variable for firms headquartered in the U.S. or Japan. Family owned firms is an indicator variable for firms classified as family owned.

Number of forced turnovers Number of censored observations Industry fixed effects

Overall Chi-square

120

HYUNG-SUK CHOI ET AL.

the CEO’s age, the five Hofstede cultural dimensions, equity market capitalization, equity return performance, firm size, the GOV44 index measure of corporate governance, and the number of articles mentioning the CEO. In some models, we also include additional control variables such as family ownership or a U.S./Japan indicator variable. In Table 4 we test whether overconfident CEOs face significantly greater forced turnover hazards than their non-overconfident counterparts. We find across all five model specifications that overconfident CEOs face significantly greater forced turnover hazards than their non-overconfident counterparts. All overconfidence coefficients are positive and significant with p-values of 0.065 or smaller. Depending on the model, the coefficients imply that an overconfident CEO faces 119
376% greater probability of forced turnover than a non-overconfident CEO. The average probability across the five models is 225%. In Model 1, we examine the explanatory power of various national characteristics on the likelihood of forced CEO turnover. We find that firms with higher stock market performance are less likely to forcibly remove their CEOs. Older CEOs are less likely to be removed. Goel and Thakor (2008) predict that forced turnovers are more likely to occur in firms with stronger corporate governance. To test this conjecture, we divide our sample firms into strong and weak governance portfolios based on the median value of the Aggarwal et al. (2009) GOV44 index. We then construct a binary indicator variable, strong governance, with a value of one for firms assigned to the strong governance portfolio and zero otherwise. We find that the coefficient for the strong governance indicator variable is positive but statistically insignificant. We also examine the influence of Hofstede’s five cultural dimensions. We find that long-term orientation is inversely related to forced turnover, suggesting that cultures that are more long-term oriented are less likely to experience the forced removal of their corporate CEOs. But countries whose cultures emphasize uncertainty avoidance are more likely to remove CEOs, perhaps in an effort to resolve the various operating uncertainties that might occur with an overconfident individual providing senior corporate executive leadership. Also countries with a high level of individualism and masculinity tend to remove their CEOs more often, but the coefficients are not statistically significant. We then estimate a number of alternative specifications of our basic regression equation with Models 2
5. In Model 2 we introduce a binary indicator variable for legal regime and obtain qualitatively identical results. Model 3 controls for a possible dominant country effect (i.e., Japan or U.S.) and again the results for CEO overconfidence remain significant.

Overconfidence, Corporate Governance, and Global CEO Turnover

121

Model 4 introduces the family owned indicator variable into the analysis, with results comparable to those of the preceding three models.7 The last model specification is comprehensive and includes all the control variables simultaneously. We conclude from our analysis in Table 4 that overconfident CEOs face a significantly greater hazard of turnover than non-overconfident CEOs. This suggests that overconfidence is a distinct risk to CEOs, separate from that of culture, firm performance, governance, legal regime, or family ownership. Further, we determine that the increased risk of dismissal to overconfident CEOs is global in nature and not limited to only U.S. or Japanese firms. Similar to Campbell et al. (2011), we conduct two additional tests to eliminate the possibility that our results are driven by some other dimension of CEO behavior or their performance. First, we exclude all retirements from our dataset so that we can compare only forced turnovers to voluntary turnovers. If turnovers of overconfident CEOs are voluntary, the overconfidence dummy variable should have no power to distinguish between voluntary and forced turnovers. Table 5 contains our findings for the analysis of how overconfidence can explain the difference between forced and voluntary turnover. The dependent variable equals one for forced turnovers and zero for voluntary turnovers. We structure our set of model specifications as identical to those of Table 4. Most importantly, we find across all of the five models that the overconfidence binary indicator coefficients are positive and statistically significant. The hazard ratios are also high in these models, averaging 292%. This suggests that the relative probabilities of forced turnover versus voluntary turnover for overconfident CEOs are economically large. The signs and statistical significance of the control variables are broadly comparable to those obtained in Table 4. The second method by which we eliminate the possibility that misclassification of forced turnovers explains our results is by excluding all forced turnovers. These results are presented in Table 6. Here, we compare only voluntary turnovers to non-turnovers. In these specifications, the dependent variable assumes a value of one for a voluntary turnover and zero otherwise. For none of the models do we find that overconfident CEOs are more likely to be voluntarily removed compared to their non-overconfident peers. We conclude from the results in Tables 5 and 6 that our measure of overconfidence is specifically related to forced turnovers in general. That is, we do not misclassify our sample of forced turnovers. Further, the analysis

Family × Overconfident CEO

U.S. or Japan × Overconfident CEO Family owned firms

Overconfident CEO dummy Age as of turnover High power distance dummy High individualism dummy High masculinity dummy High uncertainty avoidance dummy High long-term orientation dummy Log market capitalization Market performance Firm size Strong board governance dummy Number of articles mentioning the CEO Legal dummy for common law U.S. or Japan dummy

Variable

(0.09)

(0.00) (0.24)

(0.99)

(0.22)

(0.07)

(0.00)

(0.01)

(0.08) (0.37) (0.23)

(0.46)

0.73

−0.07 −1.56

12.47

1.04

1.39

−2.97

−0.37

−0.01 0.10 0.41

0.00

pvalue

Model 1

1.00

0.99 1.11 1.51

0.69

0.05

4.02

(0.47) (0.89)

−0.07

(0.08) (0.41) (0.25)

(0.02)

(0.00)

(0.09)

(0.22)

(0.99)

(0.00) (0.25)

(0.09)

0.00

−0.01 0.10 0.40

−0.36

−2.99

1.36

1.07

12.49

3E + 05

2.83

−0.07 −1.54

0.74

pvalue

Model 2

Coefficient

0.93 0.21

2.09

Hazard ratio

0.94

1.00

0.99 1.10 1.50

0.70

0.05

3.90

2.92

3E + 05

0.93 0.21

2.10

Hazard ratio

(0.18) (0.21)

1.69 −1.41

(0.93)

(0.47)

(0.07) (0.38) (0.20)

(0.10)

(0.00)

(0.04)

(0.17)

(0.99)

(0.00) (0.23)

(0.07)

pvalue

−0.05

0.00

−0.01 0.10 0.46

−0.47

−3.98

2.02

1.26

12.21

−0.08 −1.86

1.80

Coefficient

Model 3

0.24

5.41

0.95

1.00

0.99 1.11 1.58

0.63

0.02

7.54

3.52

2E + 05

0.93 0.16

6.06

−0.88

0.08

−0.06

0.00

−0.01 0.10 0.39

−0.33

−2.95

1.47

1.05

12.58

−0.07 −1.49

0.87

(0.47)

(0.94)

(0.91)

(0.66)

(0.08) (0.41) (0.28)

(0.03)

(0.00)

(0.07)

(0.23)

(0.99)

(0.00) (0.27)

(0.06)

Hazard Coefficient pratio value

Model 4

Model 5

0.41

1.09

0.94

1.00

0.99 1.10 1.47

0.72

0.05

4.33

2.85

3E + 05

0.93 0.23

2.39

−1.16

0.36

−1.37

1.76

−0.02

0.00

−0.01 0.11 0.44

−0.48

−4.02

2.13

1.26

12.44

−0.08 −1.74

1.94

(0.34)

(0.75)

(0.22)

(0.16)

(0.97)

(0.65)

(0.07) (0.38) (0.22)

(0.09)

(0.00)

(0.04)

(0.17)

(0.99)

(0.00) (0.26)

(0.06)

Hazard Coefficient pratio value

Cox Proportional Hazards Model for Forced Turnover versus Voluntary Turnover.

Coefficient

Table 5.

0.31

1.44

0.26

5.82

0.98

1.00

0.99 1.11 1.56

0.62

0.02

8.39

3.53

3E + 05

0.93 0.18

6.98

Hazard ratio

62.65

195 Yes

195

Yes

(0.00) 72

62.67

72

(0.00)

64.87

Yes

195

72

(0.00)

66.21

Yes

195

72

(0.00)

68.45

Yes

195

72

(0.00)

Note: Overconfident CEO dummy is an indicator variable with a value of one for overconfident CEOs. Age as of turnover is the age of CEO as of the turnover. Five Hofstede measures are also examined: power distance, individualism, masculinity, uncertainty avoidance, and long-term orientation. Log Market Capitalization is the logarithm of the stock market capitalization of the country in U.S. dollar terms. Market performance is the 1-year excess return of the company’s stock prior to the CEO turnover. Firm size is the log of the market value of equity of the company at the beginning of the CEO turnover year. Strong governance dummy takes a value of one if GOV44 for the firm is greater than the median GOV44, where GOV44 is the governance index used by Aggarwal et al. (2009). Number of articles mentioning the CEO is the total times the CEO was referred to in the press. Legal dummy for common law is an indicator variable for common law countries. U.S. or Japan dummy is an indicator variable for firms headquartered in the U.S. or Japan. Family owned firms is an indicator variable for firms classified as family owned.

Number of forced turnovers Number of censored observations Industry fixed effects

Overall Chi-square

U.S. or Japan × Overconfident CEO Family owned firms

Overconfident CEO dummy Age as of turnover High power distance dummy High individualism dummy High masculinity dummy High uncertainty avoidance dummy High long-term orientation dummy Log market capitalization Market performance Firm size Strong board governance dummy Number of articles mentioning the CEO Legal dummy for common law U.S. or Japan dummy

Variable

(0.26)

(0.49) (0.03)

(0.15)

(0.27)

(0.26)

(0.83)

(0.92)

(0.95) (0.02) (0.79)

(0.12)

0.20

0.01 −1.83

−1.89

−0.43

0.74

−0.14

0.01

0.00 −0.12 0.07

0.00

pvalue

1.00

1.00 0.88 1.07

1.01

0.87

2.10

0.65

0.15

1.01 0.16

1.22

Hazard ratio

(0.08) (0.00)

0.00 1.95

(0.81) (0.05) (0.11)

(0.10)

−0.18 0.00 −0.11 0.41

(0.71)

(0.03)

(0.01)

(0.72)

(0.36) (0.00)

(0.18)

pvalue

0.24

1.71

−1.62

−0.49

0.01 −2.71

0.24

Coefficient

Model 2

7.02

1.00

1.00 0.90 1.51

0.83

1.27

5.54

0.20

0.61

1.01 0.07

1.27

(0.09)

(0.33)

−0.63 0.69

(0.00)

(0.08)

(0.82) (0.05) (0.14)

(0.45)

(0.46)

(0.08)

(0.02)

(0.81)

(0.26) (0.02)

(0.40)

1.91

0.00

0.00 −0.11 0.39

−0.17

0.50

1.48

−1.48

−0.37

0.01 −2.39

−0.31

Hazard Coefficient pratio value

Model 3

2.00

0.53

6.76

1.00

1.00 0.90 1.47

0.84

1.65

4.39

0.23

0.69

1.01 0.09

0.74

−0.60

1.91

0.00

0.00 −0.12 0.42

−0.15

0.33

1.68

−1.61

−0.46

0.01 −2.62

0.25

(0.08)

(0.00)

(0.12)

(0.76) (0.03) (0.11)

(0.17)

(0.61)

(0.03)

(0.01)

(0.73)

(0.23) (0.01)

(0.20)

Hazard Coefficient pratio value

Model 4

0.55

6.74

1.00

1.00 0.89 1.52

0.86

1.39

5.38

0.20

0.63

1.02 0.07

1.28

Hazard ratio

−0.59

0.80

−0.72

1.87

0.00

0.00 −0.12 0.39

−0.14

0.63

1.41

−1.47

−0.32

0.02 −2.27

−0.37

(0.09)

(0.06)

(0.26)

(0.00)

(0.12)

(0.75) (0.03) (0.14)

(0.55)

(0.35)

(0.09)

(0.02)

(0.83)

(0.15) (0.03)

(0.31)

pvalue

Model 5 Coefficient

Cox Proportional Hazards Model for Voluntary Turnover versus Non-Turnover.

Model 1

Coefficient

Table 6.

0.56

2.22

0.49

6.47

1.00

1.00 0.88 1.47

0.87

1.88

4.12

0.23

0.73

1.02 0.10

0.69

Hazard ratio

33.33

340 Yes

Yes

(0.00)

340

47.88 195

(0.00)

195

50.54

(0.00)

Yes

340

195

54.52

0.16

Yes

340

195

(0.00)

(0.68)

1.18

57.95

0.10

Yes

340

195

(0.00)

(0.81)

1.10

Note: Overconfident CEO dummy is an indicator variable with a value of one for overconfident CEOs. Age as of turnover is the age of CEO as of the turnover. Five Hofstede measures are also examined: power distance, individualism, masculinity, uncertainty avoidance, and long-term orientation. Log Market Capitalization is the logarithm of the stock market capitalization of the country in U.S. dollar terms. Market performance is the 1-year excess return of the company’s stock prior to the CEO turnover. Firm size is the log of the market value of equity of the company at the beginning of the CEO turnover year. Strong governance dummy takes a value of one if GOV44 for the firm is greater than the median GOV44, where GOV44 is the governance index used by Aggarwal et al. (2009). Number of articles mentioning the CEO is the total times the CEO was referred to in the press. Legal dummy for common law is an indicator variable for common law countries. U.S. or Japan dummy is an indicator variable for firms headquartered in the U.S. or Japan. Family owned firms is an indicator variable for firms classified as family owned.

Number of voluntary turnovers Number of censored observations Industry fixed effects

Overall Chi-square

Family × Overconfident CEO

126

HYUNG-SUK CHOI ET AL.

controlling for a dominant country effect establishes that this result is a robust international phenomenon and not simply an artifact of the practices of U.S. firms.

POST-TURNOVER PERFORMANCE Having established that overconfident CEOs face greater risk of disciplinary termination, it is helpful to inquire if such actions are useful. That is, are such terminations followed by improved performance? Huson, Malatesta, and Parrino (2004) show that firms experiencing a CEO succession exhibit an increase in return on assets over the 3 years following turnover. Dezso (2007) presents evidence that firms with entrenched CEOs exhibit significantly poorer performance in the year prior to the forced turnover, and experience significantly stronger performance improvement during the 3 years following the forced turnover. Both of these studies, however, are limited to the U.S. firms. Because of the differences in national cultures, legal systems, and corporate governance practices, it is not clear that these post-turnover performance results will hold globally. Consequently, we examine corporate performance changes following CEO turnover for our global sample. In Table 7 we present the regression results from an analysis of the relation between firm performance and CEO turnover along with a number of control variables. More specifically, the dependent variable is measured as the difference between the market-adjusted annual return for 1 year following the CEO turnover and that for the year of the turnover. The primary independent variable is an indicator variable that equals one if the turnover is forced and zero otherwise. The other control variables include overconfidence, a dominant country indicator variable, corporate governance measure, family firm indicator variable, firm size, leverage, market to book ratio, asset turnover, and profitability. Table 7 contains the results from five different estimation models. The relation between performance improvement and forced turnover is positive and significant in Model 1. The control variables in this model include an indicator variable for a dominant country, the firm’s corporate governance, and a family firm indicator variable. This result suggests that a firm is more likely to enjoy performance improvement following the forced turnover of its CEO. In Model 2, we introduce an interactive term between forced turnover and the dominant country indicator variable. Models 3
5 introduce

127

Overconfidence, Corporate Governance, and Global CEO Turnover

Table 7.

Market Performance, Overconfidence, and Forced Turnover.

Explanatory Variables

Model 1

Model 2

Model 3

Model 4

Model 5

Intercept

0.02 (0.17)

0.00 (0.02)

0.05 (0.47)

0.01 (0.10)

−0.70 (−1.36)

Forced turnover

0.14* (1.71)

0.18 (1.32)

0.40** (1.99)

0.39* (1.89)

0.40* (1.72)

Overconfident CEO (OC) Forced turnover * OC U.S. or Japan Dummy

0.05

0.02

−0.01

(0.46)

(0.19)

(−0.01)

−0.31

−0.29

−0.35

(−1.38)

(−1.29)

(−1.34)

−0.01

0.01

−0.02

0.05

(−0.13)

(0.09)

(−0.25)

(0.33)

(U.S. or Japan) × Forced turnover

−0.06 (−0.33)

Strong governance Family owned firms

0.13

0.13

0.13

0.19

(1.40)

(1.42)

(1.41)

(1.48)

0.04

0.04

0.05

0.07

(0.19)

(0.21)

(2.40)

(0.31)

Market value of equity

0.05 (1.14)

Total debt ratio

0.41 (1.21) −0.01

Market to book ratio

(−0.55) −0.08

Asset turnover

(−1.13) −0.25

Profit margin

(−1.41) N Adjusted R2 (%) Industry fixed effects

216 4.05 Yes

216 4.11 Yes

216 3.90 Yes

216 4.91 Yes

171 9.84 Yes

Note: The dependent variable is the difference between the market-adjusted annual return for 1 year following the CEO turnover and that for the year of the turnover. Forced turnover is an indicator variable with a value of one if the CEO turnover is involuntary and zero otherwise. Overconfident CEO (OC) is an indicator variable with a value of one for overconfident CEOs and zero otherwise. U.S. or Japan dummy is an indicator variable for firms headquartered in the U.S. or Japan. Strong governance is an indicator variable with a value of one if GOV44 for the firm is greater than the median GOV44, where GOV44 is the governance index used by Aggarwal et al. (2009). Family owned firms is an indicator variable for firms classified as family owned. Market value of equity is the log of the dollar value of the market value of the firm equity at the end of the CEO turnover year. Total debt ratio, Market to book ratio, Asset turnover, and Profit margin are computed from the Compustat database as of the time of the CEO turnover by matching the calendar year and month. The results include the coefficient of each independent variable and the associated Wald Chi-squared statistics in parentheses. Statistical significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively.

128

Table 8.

HYUNG-SUK CHOI ET AL.

Accounting Performance, Overconfidence, and Forced Turnover.

Explanatory Variables Intercept Forced turnover

Model 1

Model 2

Model 3

Model 4

0.02** (1.99) 0.01 (1.63)

0.01 (0.96) 0.03** (2.25)

0.01 (0.87) 0.02 (1.10) 0.01

0.01 (1.32) 0.02 (1.19) 0.01

Overconfident CEO (OC)

(0.78) Forced turnover × OC U.S. or Japan dummy

(0.85)

Model 5 0.03 (0.82) 0.01 (0.85) 0.01 (0.91)

−0.01

−0.01

−0.01

(−0.46)

(−0.64)

(−0.51)

−0.01

0.00

−0.01

−0.01

(−0.73)

(0.37)

(−0.69)

(−0.66)

(U.S. or Japan) × forced turnover

−0.02 (−1.60)

Strong governance

0.00 (−0.35)

Family owned firms

0.00

0.00

(−0.28)

(−0.45)

0.00 (0.00)

0.00

0.00

0.00

0.00

(0.03)

(0.17)

(0.07)

(0.07)

Market value of equity

0.00 (−0.47)

Total debt ratio

0.00 (0.10) −0.00***

Market to book ratio

(−2.66) Asset turnover

0.00 (−0.12)

Profit margin

0.00 (−0.32)

N Adjusted R2 (%) Industry fixed effects

179 7.98 Yes

179 9.35 Yes

179 7.62 Yes

179 8.43 Yes

169 15.59 Yes

Note: The dependent variable is the difference between the annual accounting rate of return, which is operating income divided by average total assets, for 1 year following the CEO turnover and that for the year of the turnover. Forced turnover is an indicator variable with a value of one if the CEO turnover is involuntary and zero otherwise. Overconfident CEO (OC) is an indicator variable with a value of one for overconfident CEOs and zero otherwise. U.S. or Japan dummy is an indicator variable for firms headquartered in the U.S. or Japan. Strong governance is an indicator variable with a value of one if GOV44 for the firm is greater than the median GOV44, where GOV44 is the governance index used by Aggarwal et al. (2009). Family owned firms is an indicator variable for firms classified as family owned. Market value of equity is the log of the dollar value of the market value of the firm equity at the end of the CEO turnover year. Total debt ratio, Market to book ratio, Asset turnover, and Profit margin are computed from the Compustat database as of the time of the CEO turnover by matching the calendar year and month. The results include the coefficient of each independent variable and the associated Wald Chi-squared statistics in parentheses. Statistical significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively.

Overconfidence, Corporate Governance, and Global CEO Turnover

129

other combinations of control variables, with the coefficient of forced turnover statistically significant for all of these specifications. We conclude from this multivariate analysis that the firm’s marketadjusted performance significantly improves following a forced turnover. Further, the uniform insignificance of the dominant country indicator variable indicates that this performance improvement is not just a U.S. phenomenon, but appears to be a global one. As a further insight into the post-turnover performance of firms, we provide an analysis of their subsequent operating performance in Table 8. The dependent variable is measured as the difference between the annual accounting rate of return (operating income divided by the average total assets) for 1 year following the CEO turnover and that for the year of the turnover. The model specifications mirror those estimated for the market analysis. We find that across all models that the coefficient on forced turnover is positive, but statistically significant for only one of the models. We conclude that there is only weak evidence for improvement in accounting performance following CEO removal. This might be due to inflexibilities in cost or pricing structure that will take some time to correct. Our results seem to suggest that in spite of the absence of clear evidence of an immediate boost in accounting profitability, the market reacts positively to the decision to replace a CEO. The market appears to quickly capitalize the anticipated performance improvements regardless of when they appear in the accounting figures.

CONCLUSION Recent work by Goel and Thakor (2008) argues that CEOs who are overconfident should have a higher likelihood of forced turnover. Because overconfident CEOs overestimate their own skills and information acquisition abilities, they will overinvest in projects that reduce firm value. Such behavior will spur boards of directors to remove these individuals and to seek a new CEO who will maximize firm value. Campbell et al. (2011) provide confirming empirical evidence for a sample of U.S. firms. But given extensive and significant differences in national cultures, legal systems, and corporate governance practices, it is unclear whether such disciplining occurs globally. Through an examination of 361 CEO terminations occurring among the Fortune Global 500 firms during the years 2000
2006, we are able to offer

130

HYUNG-SUK CHOI ET AL.

empirical evidence on the international robustness of the Goel and Thakor (2008) model. We observe that there is global evidence showing that overconfident CEOs are disproportionately succeeded by other overconfident CEOs, regardless of whether they are forcibly removed or voluntarily leave office. But the most critical finding of this study is that overconfident CEOs face significantly greater hazards of forced turnovers than their nonoverconfident peers internationally. Regardless of the important differences in culture, law, and corporate governance across countries, overconfidence has a separate and distinct effect on CEO turnover. Overconfident CEOs appear to be at greater risk of dismissal regardless of where in the world they are located. We further find that this dismissal is associated with improved market performance, but with only limited enhancement in accounting returns. It appears that the market reacts quickly to anticipated performance improvements, regardless of near-term accounting profitability. This study provides the first international evidence on the relation between CEO overconfidence and disciplining. Its contributions reside in three areas. First, this study is an important complement to the findings provided by Campbell et al. (2011) for U.S. firms. It shows that the relation between overconfidence among CEOs and their forced removal is a global phenomenon and is not limited to the practices of U.S. firms with their focus on short-run earnings. Our results also offer evidence in favor of functional convergence in international corporate governance first described by Gilson (2000). Functional or de facto convergence occurs when institutions are sufficiently flexible to respond to market demands and no formal changes in the rules are necessary. Our evidence of the global practice of terminating overconfident CEOs suggests that countries have adopted this governance norm without the need for any formal rule adoption or regulatory intervention. Finally, our analysis underscores the importance of including behavioral considerations in understanding corporate decision-making and is consistent with recent studies such as Baker and Wurgler (2004), Doukas and Petmezas (2007), Ferris et al. (2012), Graham, Harvey, and Puri (2009), Griffin, Li, Yue, and Zhao (2009), Hirshleifer et al. (2012), Malmendier and Tate (2005, 2008, 2009), and Malmendier, Tate, and Yan (2010).

NOTES 1. Campbell et al. (2011) also recognize that there is an alternative interpretation of their results in which risky firms deliberately hire overconfident CEOs because

Overconfidence, Corporate Governance, and Global CEO Turnover

131

such individuals are more willing to assume the desired risks. They are also more likely to fire these CEOs because the inherent riskiness of the firms means that there will be more failure. 2. We observe examples of such variation even within the United States. Although the U.S. has a single, common law based legal and governance system, the firm-level governance index of Gompers, Ishii, and Metrick (2003) demonstrates considerable cross-sectional variability. 3. Among the many studies that examine CEO turnover in the U.S. are Warner, Watts, and Wruck (1988), Huson, Parrino, and Starks (2001), Lehn and Zhao (2006), and Parrino (1997). 4. We also test whether there exists a cultural effect on our measure of overconfidence. Specifically, we estimate culture-adjusted overconfidence using standardized values of the Hofstede cultural distance measure of Chakrabarti, Gupta-Mukherjee, and Jayaraman (2009). We then re-estimate our analyses using this culture-adjusted overconfidence measure. We find the results to be qualitatively identical to those using the unadjusted measure and hence do not separately report them. 5. Aggarwal et al. (2009) develop an additive corporate governance index using 44 attributes provided in the data supplied by CGQ. The 44 attributes cover four broad sub-categories: (a) Board (25 attributes), (b) Audit (3 attributes), (c) Anti-takeover (6 attributes), and (d) Compensation and Ownership (10 attributes). If a firm satisfies all 44 governance attributes GOV44 index would be equal to 100%. 6. Campbell et al. (2011) note several advantages of this approach over the more common logistic and multinomial logistic models used in the literature. First, they note that the Cox model controls for the fact that a CEO can be at risk of turnover in a given year and yet not be removed during that year. They also note advantages that the Cox model has in using the time-series information about a CEO for estimating the hazard of forced turnover that the individual faces. Finally, this approach is distribution free and requires no assumptions regarding the nature and shape of the underlying data. 7. We thank Chip Ryan for providing us with the data for family-owned U.S. firms.

REFERENCES Aggarwal, R., Erel, I., Stulz, R., & Williamson, R. (2009). Differences in governance practices between U.S. and foreign firms: Measurement, causes and consequences. Review of Financial Studies, 22, 3131
3169. Baker, M., & Wurgler, J. (2004). A catering theory of dividends. Journal of Finance, 59, 1125
1165. Campbell, T., Gallmeyer, M., Johnson, S., Rutherford, J., & Stanley, B. (2011). CEO optimism and forced turnover. Journal of Financial Economics, 101, 695
712. Chakrabarti, R., Gupta-Mukherjee, S., & Jayaraman, N. (2009). Mars-Venus marriages: Culture and cross-border M&A. Journal of International Business Studies, 40, 216
236. Dahya, J., McConnell, J., & Travlos, N. (2002). The Cadbury committee, corporate performance, and top management turnover. Journal of Finance, 57, 461
483. Damasio, A. (1993). Descartes’ error. New York, NY: Avon Books.

132

HYUNG-SUK CHOI ET AL.

Datta, D. K., & Puia, G. (1995). Cross-border acquisitions: An examination of the influence of relatedness and cultural fit on shareholder value creation in U.S. acquiring firms. Management International Review, 35, 337
359. DeFond, M., & Hung, M. (2004). Investor protection and corporate governance: Evidence from worldwide CEO turnover. Journal of Accounting Research, 42, 269
312. Dezso, C. (2007). Entrenchment and changes in performance following CEO turnover. Working paper. University of Maryland, MD. Doidge, C., Karolyi, G., & Stulz, R. (2007). Why do countries matter so much for corporate governance? Journal of Financial Economics, 86, 1
39. Douglas, M., & Wildavsky, A. (1982). Risk and culture: An essay on the selection of technological and environmental dangers. Berkeley, CA: University of California Press. Doukas, J., & Petmezas, D. (2007). Acquisitions, overconfident managers and self-attribution bias. European Financial Management, 13, 531
577. Ferris, S., Jayaraman, N., & Sabherwal S. (2012). CEO Overconfidence and International Merger and Acquisition Activity, Journal of Financial and Quantitative Analysis, forthcoming. Gilson, R. (2000). Globalizing corporate governance: Convergence of form or function. American Journal of Comparative Law, 49, 329
357. Gleason, K., Mathur, L., & Mathur, I. (2000). The interrelationship between culture, capital structure, and performance: Evidence from European retailers. Journal of Business Research, 50, 185
191. Goel, A., & Thakor, A. (2008). Overconfidence, CEO selection, and corporate governance. Journal of Finance, 63, 2737
2784. Goldstein, W., & Weber, E. (1995). Content and its discontents: The use of knowledge in decision making. In J. R. Busemeyer, R. Hastie, & D. L. Medin (Eds.), Decision making from a cognitive perspective: The psychology of learning and motivation (32, 83
136). New York, NY: Academic Press. Gompers, P., Ishii, J., & Metrick, A. (2003). Corporate governance and equity prices. Quarterly Journal of Economics, 118, 107
155. Goszczytiska, M., Tyszka, T., & Slovic, P. (1991). Risk perception in Poland: A comparison with three other countries. Journal of Behavioral Decision Making, 4, 179
193. Graham, J., Harvey, C., & Puri, M. (2009). Managerial attitudes and corporate actions. Working Paper. Duke University, NC, USA. Griffin, D., Li, K., Yue, H., & Zhao, L. (2009). Cultural values and corporate risk taking. Working paper. University of British Columbia, BC, Canada. Hirshleifer, D., Teoh, S., & Low, A. (2012). Are overconfident CEOs better innovators? Journal of Finance, 67, 1457
1498. Hofstede, G. (1980). Culture’s consequences: International differences in work-related values. Beverly Hills, CA: Sage. Hofstede, G. (2001). Culture’s consequences: Comparing values, behaviors, institutions and organizations across nations (2nd ed.). Thousand Oaks, CA: Sage. Huson, M., Malatesta, P., & Parrino, R. (2004). Managerial succession and firm performance. Journal of Financial Economics, 74, 237
275. Huson, M., Parrino, R., & Starks, L. (2001). Internal monitoring mechanisms and CEO turnover: A long-term perspective. Journal of Finance, 56, 2265
2297. Kang, J., & Shivdasani, A. (1995). Firm performance, corporate governance, and top executive turnover in Japan. Journal of Financial Economics, 38, 28
58. Kaplan, S. (1994a). Campeau’s acquisition of Federated: Post-bankruptcy results. Journal of Financial Economics, 35, 123
136.

Overconfidence, Corporate Governance, and Global CEO Turnover

133

Kaplan, S. (1994b). Federated’s acquisition and bankruptcy: Lessons and implications. Washington University Law Quarterly, 72, 1103
1226. Kirkman, B. L., Lowe, K. B., & Gibson, C. B. (2006). A quarter century of culture’s consequences: A review of empirical research incorporating Hofstede’s cultural value framework. Journal of International Business Studies, 37, 285
320. Kwok, C., & Tadesse, S. (2006). National culture and financial systems. Journal of International Business Studies, 37, 227
247. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1997). Legal determinants of external finance. Journal of Finance, 52, 1131
1150. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1998). Law and finance. Journal of Political Economy, 52, 1113
1155. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1999). Corporate ownership around the world. Journal of Finance, 54, 471
517. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (2000). Agency problems and dividend policies around the world. Journal of Finance, 55, 1
33. Lehn, K. M., & Zhao, M. (2006). CEO turnover after acquisitions: Are bad bidders fired?. Journal of Finance, 61, 1759
1811. Malmendier, U., & Tate, G. (2005). CEO overconfidence and corporate investment. Journal of Finance, 60, 2661
2700. Malmendier, U., & Tate, G. (2008). Who makes acquisitions? CEO overconfidence and the market’s reaction. Journal of Financial Economics, 89, 20
43. Malmendier, U., & Tate, G. (2009). Superstar CEOs. Quarterly Journal of Economics, 124, 1593
1638. Malmendier, U., Tate, G., & Yan, J. (2010). Managerial beliefs and corporate financial policies. Working Paper. University of California, Berkeley, CA. Parrino, R. (1997). CEO turnover and outside succession: A cross-sectional analysis. Journal of Financial Economics, 46, 165
197. Phillips, L., & Wright, G. (1977). Cultural differences in viewing uncertainty an assessing probabilities. In H. Jungerman & G. de Zeeuw (Eds.), Decision making and change in human affairs. Dordrecht, Netherlands: Reidel. Pollock, S., & Chen, K. (1986). Strive to conquer the big stink: Decision analysis in the people’s republic of China. Interfaces, 16, 31
37. Roll, R. (1986). The hubris hypothesis of corporate takeovers. Journal of Business, 59, 197
216. Sekely, W., & Collins, J. M. (1988). Cultural influences on international capital structure. Journal of International Business Studies, 19, 87
100. Shafir, E., Simonson, I., & Tversky, A. (1993). Reason-based choice. Cognition, 49, 11
36. Simon, H. (1990). Invariants of human behavior. Annual Review of Psychology, 41, 1
19. Slovic, P. (1997). Trust, emotion, sex, politics, and science: Surveying the risk assessment battlefield. In M. H. Bazerman, D. M. Messick, A. E. Tenbrunsel, & K. A. Wade-Benzoni (Eds.), Psychological perspectives to environmental and ethical issues in management. San Francisco, CA: Jossey-Bass. Stulz, R., & Williamson, R. (2003). Culture, openness, and finance. Journal of Financial Economics, 70, 313
349. Teigen, K. H., Brun, W., & Slovic, P. (1988). Societal risks as seen by Norwegian Public. Journal of Behavioral Decision Making, 1, 111
130. Volpin, P. (2002). Governance with poor investor protection: Evidence from top executive turnover in Italy. Journal of Financial Economics, 64, 61
90.

134

HYUNG-SUK CHOI ET AL.

Von Winterfeldt, D., & Edwards, W. (1986). Decision analysis and behavioral research. Cambridge: Cambridge University Press. Warner, J., Watts, R., & Wruck, K. (1988). Stock prices and top management changes. Journal of Financial Economics, 20, 461
492. Weber, E., & Hsee, C. (1998). Cross-cultural differences in risk perception but cross-cultural similarities in attitudes towards risk. Management Science, 44, 1205
1217. Weber, E., & Hsee, C. (2000). Culture and individual judgment and decision-making. Applied Psychology: An International Review, 49, 32
61. Whitcomb, K. W., Onkal, D., Curley, S. P., & Benson, G. (1995). Probability judgment accuracy for general knowledge: Cross-national differences and assessment methods. Journal of Behavioral Decision Making, 8, 51
67. Wright, G., & Phillips, L. (1980). Cultural variation in probabilistic thinking; Alternative ways of dealing with uncertainty. International Journal of Psychology, 15, 239
257.

Overconfidence, Corporate Governance, and Global CEO Turnover

135

APPENDIX: OVERVIEW OF THE HOFSTEDE CULTURAL MEASURES Sample and Construction In 1980, Geert Hofstede published Culture’s Consequences: International Differences in Work-Related Values, in which he developed his multidimensional framework for the analysis of culture. In this work, Hofstede statistically analyzes over 116,000 questionnaires collected in 1967 and 1973 from employees working in forty IBM subsidiaries around the world. Hofstede then undertakes a country-level factor analysis of these questionnaires. From this analysis, Hofstede develops four dimensions of culture. A fifth dimension, the extent of long-term orientation, was added in 1991. The Five Cultural Dimensions The five cultural dimensions identified by Hofstede are described below: Power Distance focuses on the amount of equality or inequality between people in a country. A high power distance ranking indicates that inequalities of power and wealth have been allowed to grow within the society. These societies are more likely to follow a caste system that does not allow significant upward mobility of its citizens. A low power distance ranking indicates that the society de-emphasizes the differences between citizen’s power and wealth. In these societies equality and opportunity for everyone is stressed. Individualism measures the degree to which society reinforces individual, or collective, achievement and interpersonal relationships. A high individualism ranking indicates that individuality and individual rights are paramount within the society. Individuals in these societies may tend to form a larger number of looser relationships. A low individualism ranking typifies societies of a more collectivist nature with close ties between individuals. These cultures reinforce extended families and collectives where everyone takes responsibility for fellow members of their group. Masculinity captures the extent to which society reinforces, or does not reinforce, the traditional masculine work role model of male achievement, control, and power. A high masculinity ranking indicates that the country experiences a high degree of gender differentiation. In these cultures, males dominate a significant portion of the society and power structure, with females being controlled by male domination. A low masculinity ranking

136

HYUNG-SUK CHOI ET AL.

indicates that the country has a low level of differentiation and discrimination between genders. In these cultures, females are treated equally to males in all aspects of the society. Uncertainty Avoidance reflects the level of tolerance for uncertainty and ambiguity within the society, that is, unstructured situations. A high uncertainty avoidance ranking indicates that the country has a low tolerance for uncertainty and ambiguity. This creates a rule-oriented society that institutes laws, rules, regulations, and controls in order to reduce the amount of uncertainty. A low uncertainty avoidance ranking indicates that the country has less concern about ambiguity and uncertainty and has more tolerance for a variety of opinions. This is reflected in a society that is less ruleoriented, more readily accepts change, and takes more and greater risks. Long-Term Orientation focuses on the degree the society embraces, or does not embrace, long-term devotion to traditional, forward thinking values. A high long-term orientation ranking indicates that the country prescribes to the values of long-term commitments and respect for tradition. This is thought to support a strong work ethic where long-term rewards are expected as a result of today’s hard work. Business, however, might take longer to develop in this society, particularly for an “outsider.” A low longterm orientation ranking indicates that the country does not reinforce the concept of long-term, traditional orientation. In this culture, change can occur more rapidly as long-term traditions and commitments do not become impediments to change.

HUMAN AND SOCIAL CAPITAL IN THE LABOR MARKET FOR DIRECTORS George D. Cashman, Stuart L. Gillan and Ryan J. Whitby ABSTRACT Purpose
This study examines the director labor market to better understand which director attributes are important for board service. Design/Methodology/Approach
Director level data, which includes proxies for both human and social capital, is analyzed to determine which characteristics increase the likelihood of gaining additional board appointments. Findings
We find that general skills and director connections are valued in the marketplace. Among specific director characteristics, financial expertise, holding an MBA degree, and S&P 500 experience are positively associated with gaining new board appointments. Moreover, regardless of the director’s level of expertise, highly connected individuals are more likely to obtain new appointments. Finally, from a range of characteristics, only director connections mitigate the negative consequences of serving on the boards of firms that restate their financials.

Advances in Financial Economics, Volume 16, 137
164 Copyright r 2013 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1569-3732/doi:10.1108/S1569-3732(2013)0000016005

137

138

GEORGE D. CASHMAN ET AL.

Originality/Value
While most research has analyzed the effectiveness of boards of directors as a whole, this study examines the value of individual director characteristics within the context of the labor market. Keywords: Boards of directors; human capital; social capital; director selection; director effectiveness

Due to the central role that the board of directors plays within a firm, boards have been the focus of extensive study across a variety of disciplines. While early research emphasized the effects of board size and independence, more recent research has focused on board composition in greater detail. Specifically, researchers have begun to examine how various characteristics of the directors composing the board are associated with firm outcomes. However, the results of these studies are often mixed, potentially due to the difficulty associated with linking director characteristics to relatively distal firm outcomes (Johnson, Schnatterly, & Hill, 2013). We take an alternative approach to identify the characteristics of an effective director by focusing on how a director’s business experience (human capital) and social connections (social capital) influence the likelihood that they receive an additional board seat.1 Our approach stems from the idea that the labor market motivates directors to act in the best interest of shareholders and develop reputations as experts, as described in Fama (1980) and Fama and Jensen (1983). Additionally, Hillman and Dalziel (2003) note that, while directors are motivated to effectively monitor the firm, only those with the necessary ability, human and social capital, will be effective.2 Building on this foundation we argue that by studying the association between director characteristics and labor market outcomes, we are able to identify the characteristics that the market associates with director effectiveness. Understanding the characteristics of an effective director is especially important given the influence that boards have been found to have on firm strategy.3 Additionally, boards have come under increased scrutiny during the past decade. For example, the Sarbanes Oxley Act of 2002 requires specific expertise in the form of financial experts on audit committees, and the NYSE requires fully independent audit, compensation, and nominating committees. More recently, the Dodd
Frank Act included a mandate allowing shareholders to directly nominate individuals for board seats (proxy access). While this provision has been struck down, the Securities and Exchange Commission has adopted rules allowing shareholders to

Human and Social Capital in the Labor Market for Directors

139

petition for proxy access on a company-by-company basis. In anticipation of such regulatory changes institutional investors, in particular, the California Public Employees’ Retirement Systems CalPERs and CalSTERs have jointly developed a database of potential director candidates.4 However, all of these changes are being adopted in a relative vacuum of information as to what constitutes an effective director. We explore this issue and find evidence that both the director’s business experiences and social connections increase their chances of gaining an additional board seat, suggesting that both what and who a director knows is associated with director effectiveness. In particular, our analysis indicates that S&P 500 board experience, holding an MBA degree, and having more connections increase the likelihood of gaining an additional board seat. The preference for directors with S&P 500 experience and an MBA indicates a demand for directors with transferrable or general skills. This echoes the findings of Murphy and Zabojnik (2007) and Kaplan, Klebanov, and Sorenson (2012), which find that the executive labor market prefers such transferrable or general skills. Additionally, we find evidence suggestive of the relative importance of a director’s social connections. Specifically, we find that professionally connected directors, regardless of their level of skill or expertise, are more likely to gain new board seats than less professionally connected directors. Moreover, only skilled directors who are also highly connected are more likely to receive an additional board seat. We also find evidence that only professional connections are able to mitigate the negative effects of having served on the board of a firm that restated its financials, a result consistent with that of Marcel and Cowen (Forthcoming). We do not try to answer the question of whether or not firms are choosing directors optimally, nor do we attempt to determine whether or not specific skills or attributes are over- or under-valued in the marketplace. Rather, we assume that the director labor market is relatively efficient, rewards additional board seats to effective directors, and thus studying the aggregate marketplace provides insights into the specific characteristics of the market values.

BACKGROUND AND LITERATURE REVIEW Director Labor Market Underlying our analysis is the Fama and Jensen (1983) assertion that the labor market for directors rewards individuals who possess characteristics

140

GEORGE D. CASHMAN ET AL.

that facilitate effective monitoring and advising of management. Evidence consistent with this view is provided by Brickley et al. (1999), who report that the likelihood that a retired CEO sits on his former firm’s board (or other corporate boards) following his retirement is related to the performance of his firm while he served as CEO. Similarly, Ferris et al. (2003) find that directors at firms with better performance hold more board seats. Additionally, Coles and Hoi (2003) find that directors of firms rejecting anti-takeover protections included in Pennsylvania Senate Bill 1310 gained directorships in the following 3 years. Lastly, Ertimur et al. (2010a) find that directors who implement majority vote shareholder proposals are less likely to lose board seats. In addition to rewarding effective directors, the market also appears to penalize directors for bad actions. Fich and Shivdasani (2007) find that outside directors at firms subject to fraud-related lawsuits hold significantly fewer board seats in the future. Similarly, Srinivasan (2005) finds that directors tend to hold fewer board seats when a firm they are associated with restates earnings. Moreover, Harford (2003) reports that directors at firms that are the target of hostile takeover bids hold fewer directorships following the takeover attempt. These findings suggest that the market imposes costs on directors who fail in their monitoring role. Thus, our empirical strategy is to examine the association between labor market outcomes and individual director characteristics in order to provide a market-based perspective as to the importance of human capital, what the director knows, and social capital, who the director knows. Although previous research does provide insight into this issue, the majority of the work to date examines characteristics of the board in general and links these board characteristics to potentially disparate firm outcomes. In contrast, we examine the labor market for directors in order to focus on aspects of a director’s human and social capital that are sought after by companies.

Director Characteristics Our focus is on the role that human and social capital plays in the selection of new directors. Human capital refers to a director’s experiences and skills, which Carpenter and Westphal (2001) argue is necessary to ensure that the director is able to effectively control and advise management. Additionally, Kroll et al. (2008) and McDonald et al. (2008) find evidence that firms with boards that possess acquisition experience are associated with superior

Human and Social Capital in the Labor Market for Directors

141

acquisitions. Stearns and Mizruchi (1993) and Mizruchi and Stearns (1994) find evidence that the amount and type of financing a firm uses is influenced by the presence of a director who is an employee of a financial institution. More recently, Krause, Semadeni, and Cannella (Forthcoming) examine the impact of external executive directors on the board and find that the impact of these directors is dependent on the operational efficiency of the firm, as well as the executive’s position in their home firm.5 Thus, a number of papers suggest the importance of a director’s human capital. While human capital refers to a director’s experience and skills, social capital refers to a director’s social relationships. As noted by Burt (1997) social connections can prove valuable as they provide access to information and resources.6 Carpenter and Westphal (2001) argue that social connections allow a director to learn about business practices more quickly than if he or she was forced to rely on secondary sources. Consistent with the view that social connections allow for the learning of business practices, Connelly, Johnson, Tihanyi, and Ellstrand (2011) examine how a firm’s directors’ social connections influence the firm’s decision to expand into China. They find that when a firm’s directors have social ties to individuals associated with a successful expansion into China, the firm is more likely to expand into China. Similarly, though not an examination of directors, Geletkanycz and Hambrick (1997) find evidence that a CEO’s social connections influence firm strategies. In related work, Geletkanycz, Boyd, and Finkelstein (2001) find evidence that CEOs social connections are valued in that more socially connected CEOs are generally more highly compensated. Moreover, Engelberg, Gao, and Parsons (2013) find that when a firm’s directors have fewer connections, they are willing to pay more to hire a well-connected CEO. Further, Tian et al. (2011) find evidence that both the human and social capital of the directors play an important role in how the market responds to the appointment of a new CEO. Anecdotal evidence also suggests that director connections are important. For example, Lipin (1999) highlights the importance of director’s connections for the start-up firm FirstMark: Its directors include Nathan Myhrvold, chief technology officer at Microsoft; Bert Roberts, chairman of MCI WorldCom; Washington power broker Vernon Jordan; former Secretary of State Henry Kissinger; Sir Evelyn de Rothschild, chairman of N.M. Rothschild & Sons; and Michael Price, a former partner of Lazard Freres, who signed on as co-chief executive. … The contacts have already helped win key licenses to build a so-called fixed wireless network, raised financing for the venture, and helped find strategic partners.

142

GEORGE D. CASHMAN ET AL.

DATA AND UNIVARIATE ANALYSES Our data on corporate directors come from BoardEx.7 BoardEx collects biographical information on corporate managers and directors, including an individual’s date of birth, education, and employment history. The employment history includes information on current and prior positions, directorships, affiliations with non-profits, and beginning and ending dates of each position held. Coverage for many individuals starts in 1999, and personal information for some of those individuals dates back as far as 1926. Prior to 2003, however, the number of firms and directors varies substantially from year to year as the data set was being populated. Accordingly, our analysis focuses on the period between 2003 and 2008. This is also advantageous in that it corresponds largely with the post-SOX era and thus allows us to study what others have suggested is a new regime in the labor market for directors. We capture a director’s human capital using binary variables that indicate whether or not an individual has a certain trait. There are numerous classifications within the BoardEx database. Rather than using each individually, we use the aggregate number of qualifications reported in BoardEx and then include indicators for specific types of experience that prior studies indicate are likely to influence a director’s likelihood of gaining a board seat. Our proxies for financial and legal expertise are CFA/CPA and JD, respectively. We also identify executives and CEOs of publicly traded firms to capture the fact that such individuals often seek, and are sought for, board service.8 Indicators for holding an MBA and having S&P 500 board experience proxy for general skills, while indicators for holders of an MD or PhD are designed to capture more specialized backgrounds. We also consider measures that summarize an individual’s experience over time. For example, we calculate each individual’s aggregate experience as a director and the total number of qualifications each director possesses. Since we are interested in characteristics and experience at the director level, we differentiate between directors that have specific types of experience and those that do not. An example of this is merger experience, which is defined as the cumulative number of merger transactions a director has been involved with. We also track whether or not an individual has served on the board of a firm undertaking a financial restatement. To focus on each director’s social capital we categorize an individual’s connections as either “professional” or “other.” Professional connections include common board appointments and overlapping work experience. Other connections include education networks, such as attending the same

Human and Social Capital in the Labor Market for Directors

143

university and connections through non-profit organizations or charities. For each data year we construct both a “professional” network and an “other” network. Following the social networking literature, we calculate degree, closeness, and betweeness (Sabidussi, 1966 & Freeman, 1977). Degree is the number of direct connections each director has in the respective network. While closeness and betweenness attempt to measure how centrally located a director is within each network. Although these variables have very distinct interpretations, they are also highly correlated and are sometimes difficult to interpret. We therefore use a principal components analysis to reduce the social network variables into a single “connectedness” measure for each director in each network each year. Additionally, we attempt to control for variation in experience and contacts that likely results from service on the boards of different firms. For example, serving on the board of a large firm might differ substantially from serving on the board of a small firm. Therefore, we consider the attributes of the firms at which individuals serve as directors by aggregating firm-level information for each director. Specifically, we aggregate firm size (as measured by market capitalization) for all boards that a director serves on to proxy for the overall reputation of the companies on which an individual serves. Similarly, we calculate prior year average ROE (valueweighted and industry-adjusted) as a measure of the performance of the firms where an individual is a director. We also calculate a 2-year, valueweighted market return for the firms at which a director has served. Given our focus on the director level, an issue is that many directors serve on multiple boards simultaneously. Thus, any variable used in the director level analysis must be aggregated across firms. While aggregation is straightforward for some variables (e.g., size and experience), it can pose a challenge for others (e.g., industry and CEO connectedness); Therefore, we do not attempt to aggregate some variables, such as industry, at the director level. We use CRSP to calculate market values, Compustat for firm-specific financial data, and Audit Analytics to identify firms undertaking a restatement. Our final sample is the intersection between BoardEx, CRSP, and Compustat for U.S. companies. The BoardEx sample of U.S. firms has more than 5,400 companies and 127,000 directors while Compustat has data for more than 15,000 firms over this time period. After combining the datasets, we eliminate all observations with missing data points. This results in a sample of 5,036 unique publicly traded firms, with 21,211 unique directors and 4,963 new director appointments from 2003 to 2008 (Table 1).

144

GEORGE D. CASHMAN ET AL.

Table 1.

Annual Board-Level Summary Statistics.

Year

Firms

Directors

Boards per Director

New Board Seats

2003 2004 2005 2006 2007 2008 Mean

3,661 4,444 4,688 4,621 4,427 3,872 4,286

11,676 14,005 14,479 14,667 14,206 12,797 13,638

1.84 1.74 1.73 1.70 1.68 1.68 1.73

366 583 828 1,181 1,161 844 827

25,713 5,036

81,830 21,211





4,963


Total Unique

Note: Summary information regarding the number of firms, directors, director board seats, and new board seats awarded each year for our sample of BoardEx, CRSP, and Compustat firms.

The average director holds 1.73 board seats, a value that declined somewhat from 1.84 in 2003 to 1.68 by 2008. The unconditional probability of an individual obtaining an additional board seat in our sample is approximately 6.07%.9 Table 2 provides more details regarding our measures of a director’s human and social capital. Panel A reports summary statistics for the full sample. In Panel B, we report summary statistics for two subsamples of directors: those who receive additional seats and those who do not. We also report tests of differences between the two groups. In Panel C, we compare the characteristics of individuals joining a board with the characteristics of the director they are replacing. Focusing on Table 2, Panel A, we find that approximately 23% of our directors serve on the board of an S&P 500 company, 9.7% are current executives, and 7.1% are current CEOs. In terms of other qualifications, approximately 33% are MBAs, 11.2% have a JD, 11.4% a PhD, and 11% a CPA. With respect to the other qualifications we examine 2.8% are MDs and less than 1% have a CFA. The average board member has 2.15 BoardEx qualifications, is 60 years old, sits on 1.73 boards, and has approximately 12 years of cumulative board experience. The average director has spent some 4 years serving on audit and compensation committees and 3 years on governance committees. Panel B of Table 2 provides a comparison of the characteristics of directors who receive an additional board seat during our sample period relative to those who do not. We see that directors receiving a new appointment are more likely to have S&P 500 experience, currently be an executive, have

S&P 500 (1/0) Executive (1/0) CEO (1/0) MBA (1/0) JD (1/0) PhD (1/0) CPA (1/0) CFA (1/0) MD (1/0) Male (1/0) Nationality (U.S. = 1) Age Number of qualifications Time on boards Current boards Professional network Other network Merger experience Average ROE Total market value 2-Year market return Restatements Audit Compensation Governance

Panel A: Full Sample Summary Statistics

0.228 0.097 0.071 0.328 0.112 0.114 0.110 0.007 0.028 0.906 0.583 59.938 2.149 12.255 1.725 0.739 0.004 3.630 0.057 9.059 0.358 0.074 4.265 3.907 2.835

Mean 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 60.000 2.000 7.900 1.000 0.007 0.000 2.000 0.082 0.920 0.180 0.000 3.000 3.000 2.000

Median

Table 2. Annual Director-Level Summary Statistics.

0.420 0.296 0.256 0.470 0.315 0.318 0.313 0.084 0.165 0.292 0.493 9.105 0.949 14.302 1.193 2.283 0.577 4.660 9.426 33.983 1.359 0.262 4.690 4.510 3.857

Standard Deviation

Human and Social Capital in the Labor Market for Directors 145

(Continued )

S&P 500 (1/0) 0.398 Executive (1/0) 0.122 CEO (1/0) 0.070 MBA (1/0) 0.385 JD (1/0) 0.092 PhD (1/0) 0.097 CPA (1/0) 0.124 CFA (1/0) 0.005 MD (1/0) 0.025 Male (1/0) 0.852 Nationality (U.S. = 1) 0.560 Age 57.126 Number of qualifications 2.208 Time on boards 7.412 Current boards 2.259 Professional network 2.280 Other network 0.018 Mergers 4.486 Average ROE −0.005 Total market value 18.589 2-Year market return 0.212 Restatements 0.082 Audit 4.332

Mean

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 58.000 2.000 3.100 2.000 1.465 0.000 3.000 0.093 3.041 0.161 0.000 2.000

0.489 0.328 0.255 0.487 0.289 0.296 0.330 0.072 0.157 0.355 0.496 7.608 0.945 11.101 1.321 3.003 0.733 5.508 0.988 49.787 0.706 0.275 5.083

0.218 0.095 0.071 0.325 0.113 0.115 0.109 0.007 0.028 0.909 0.585 60.108 2.146 12.548 1.693 0.646 0.003 3.578 0.061 8.483 0.367 0.074 4.261

Mean 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 61.000 2.000 8.200 1.000 −0.036 0.000 2.000 0.081 0.857 0.181 0.000 3.000

Median 0.413 0.294 0.257 0.468 0.317 0.319 0.312 0.085 0.165 0.288 0.493 9.161 0.949 14.421 1.177 2.198 0.566 4.598 9.704 32.696 1.388 0.261 4.666

Standard deviation

N = 76,867

N = 4,963

Median Standard deviation

New seat = 0

New seat = 1

0.180 0.027 −0.001 0.060 −0.021 −0.017 0.014 −0.002 −0.003 −0.057 −0.025 −2.983 0.062 −5.136 0.566 1.634 0.016 0.908 −0.066 10.106 −0.154 0.008 0.071

Difference

Mean

Panel B: Difference in Mean and Median for Directors Receiving an Additional Seat and Those that do not

Table 2.

p-value