UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Essays on empirical corporate finance Gao, Huasheng 2009

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2009_fall_gao_huasheng.pdf [ 853.27kB ]
Metadata
JSON: 24-1.0067252.json
JSON-LD: 24-1.0067252-ld.json
RDF/XML (Pretty): 24-1.0067252-rdf.xml
RDF/JSON: 24-1.0067252-rdf.json
Turtle: 24-1.0067252-turtle.txt
N-Triples: 24-1.0067252-rdf-ntriples.txt
Original Record: 24-1.0067252-source.json
Full Text
24-1.0067252-fulltext.txt
Citation
24-1.0067252.ris

Full Text

Essays on Empirical Corporate Finance by Huasheng Gao B.S., Shanghai Jiao Tong University, 2003 M.Sc. University of British Columbia, 2006  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR THE DEGREE OF Doctor of Philosophy in The Faculty of Graduate Studies (Business Administration)  The University of British Columbia (Vancouver) May, 2009 © Huasheng Gao 2009  Abstract In this thesis, I examine a few corporate finance topics, including mergers and acquisitions, CEO compensations, and corporate governance. The first paper studies the effect of managerial horizon on acquisition activities. Managers with a long horizon emphasize firms’ long-term value, whereas short-horizon executives are concerned about firms’ value in the short run. When bidding firms’ stocks are overvalued, long-horizon acquiring managers tend to acquire targets with equity to enhance the acquirers’ long-term value. However, acquiring managers with a short horizon are inclined to pay for acquisitions with cash, in attempt to hide the signal to the market of potential stock overvaluation. Moreover, short-horizon managers, relative to their long-horizon counterparts, are more likely to choose acquisition projects that the market wants to see, thus boosting the short-term stock price by catering to investor sentiment. The paper’s main predication is that acquiring firms managed by short-horizon executives have higher abnormal returns at acquisition announcements, less likelihood of using equity to pay for the transactions, and worse post-merger stock performance in the long run. I construct two proxies for managerial horizon based on the CEO’s career concern and compensation scheme, and provide empirical evidence supporting the above prediction. Moreover, I also show that long-horizon managers are more likely to initiate acquisitions as responding to high stock market valuation. The second paper examines optimal compensation contracts when executives can hedge their personal portfolios. In a simple principal-agent framework, I predict that the CEO’s pay-performance sensitivity decreases with the executive hedging cost. Empirically, I find evidence supporting the model’s prediction. Providing further support for the theory, I show that shareholders also impose high sensitivity of CEO wealth to stock volatility via compensation contracts when managers can hedge. In addition to providing higher-power contracts, shareholders increase financial leverage to resolve the executive-hedging problem. Moreover, executives with lower hedging costs hold more exercisable in-the-money options, have weaker incentives to cut dividends, and pursue fewer corporate diversification initiatives. Overall, the ability to hedge firm risk undermines executive incentive and enables managers to bear more risk, thus affecting governance mechanisms and managerial actions. The third paper investigates the causes and consequences of sharp CEO pay cuts, a phenomenon that has been mostly overlooked in the attention paid to overall rising executive pay. We find that a large CEO pay cut is not uncommon and is typically triggered by poor stock performance. Good corporate governance structures strengthen the link between poor performance and CEO pay cut. After the pay cut, the CEO can restore his pay level by reversing the poor performance. Pay cuts are only a short-term substitute for dismissal—a pay-cut CEO with continued poor performance is just as likely to be dismissed as a CEO with similar performance whose pay was not cut. On average, CEOs respond to their pay cut by curtailing capital expenditures, reducing R&D expenses, and allocating funds to reduce leverage. For most firms, performance improves and the CEO’s pay is restored. Compared to option repricing, pay cuts appear to be more effective in improving firm performance. Together, our results show that the possibility of these compensation cuts provides ex ante incentives for CEOs to exert effort to avoid poor performance and ex post incentives to improve poor performance once pay is cut. ii  Table of Contents Abstract……………………………………………………………………………………..ii Table of Contents…………………………………………………………………………… iii List of Tables………………………………………………………………………………….v List of Figures………………………………………………………………………………..vi Acknowledgement………………………………………………………………………...vii Dedication…………………………………………………………………………………..viii Statement of Co-Authorship…………………………………………………………………ix CHAPTER 1 – Introduction………………………………………………………………..…1 References………………………………………………………………………………..3 CHAPTER 2 – Market Misvaluation, Managerial Horizon and Acquisitions ………….……4 2.1 Introduction…………………………………………………………………….…….4 2.2 Main Hypothesis and Testable Implications…………………………………………7 2.2.1 Equity Issuance Channel…………………………………………………....…7 2.2.2 Catering Channel…………………………………………………………….8 2.2.3 Testable Predictions………………………………………………………….9 2.3 Proxy for Managerial Horizon……………………………………………………...9 2.4 Sample Selection and Data Description…………….…………………………….11 2.5 Empirical Results…………………………………….……………………………..16 2.5.1 Abnormal Returns during Announcement Periods….………………………..16 2.5.2 Probability of Equity Payment……………………….………………………18 2.5.3 Long-term Stock Performance Following Acquisitions….………………..…20 2.5.4 Additional Analysis on Acquisition Frequency…………….………………21 2.5.5 Alternative Explanation and Robustness Checks………………………….…23 2.6 Conclusion.……………………………….………………………………………...25 References………………………………………………………………………….....35 CHAPTER 3 –Optimal Compensation Contracts When Managers Can Hedge……………38 3.1 Introduction……………………………………………………………………….38 3.2 Background on Executive Hedging………………………………...........................42 3.3 The Model………………………………………………………..............................44 3.4 Data Construction and Sample Selection…………………………...........................51 3.4.1 Measures of CEO Incentive………………………………………………….51 3.4.2 Measures of Executive Hedging Cost………………………………………..52 3.4.3 Control Variables………………………………………..……………………53 3.4.4 Data Source………………………………………………..…………………55 3.5 Empirical Results………………………………………………………...................55 3.5.1 Summary Statistic……………………………………………………………55 3.5.2 Managerial Hedging Cost and Incentive Pay………………………………56 3.5.3 Managerial Hedging Cost and Convexity in Incentive Pay………………….62 3.5.4 Managerial Hedging Cost and Capital Structure…………………………….65 3.5.5 Managerial Hedging Cost and Option Exercising/Holding Behavior………..68 3.5.6 Managerial Hedging Cost and Corporate Dividend Policy…………………..70 3.5.7 Managerial Hedging Cost and Corporate Diversification………………….72 iii  3.6 Additional Investigation…………………………………………………………...74 3.6.1 Firm Beta and Incentive Pay…………………………………...…………….74 3.6.2 Proxy for the Ease of OTC Hedging Transactions………………………….75 3.6.3. Direct Executive Hedging Transactions and Incentive Pay…………………77 3.7 Conclusion………………………………………………………………………….78 References…………………………………………………………………………….94 CHAPTER4 –Incentive Effects of Extreme CEO Pay Cuts………………………………...98 4.1 Introduction…………………………………………………………………………98 4.2 Prior Literature and Hypothesis Development……………………………………101 4.2.1 Literature Review……………………………………………………..……101 4.2.2 Our Hypotheses…………………………………………………………….103 4.3 Sample Formation and Variable Construction…………………………………….105 4.4 Causes and Consequences of CEO Pay Cuts……………………………………111 4.4.1 Why Cut Pay?................................................................................................111 4.4.2 CEO Retention after a Pay Cut……………………………………………115 4.4.3 Firm Performance after a Pay Cut………………………………………...118 4.4.4 Corporate Policies after a Pay Cut………………………………………….119 4.4.5 CEO Pay Recovery after a Pay Cut…………………………………………122 4.5 Additional Investigation…………………………………………………………124 4.5.1 Pay Change for Top Management Team……………………………………124 4.5.2 Controlling for Business Cycle…………………………………………...124 4.5.3 Controlling for Initial Abnormal Pay……………………………………….126 4.5.4 Results from Performance-based Control Firms……………………………127 4.5.5 Pay Cut versus CEO Turnover……………………………………………128 4.5.6 Pay Cut versus Option Repricing…………………………………………129 4.6 Conclusion…………………………………………………………………………130 References……………………………………………………………………………..153 CHAPTER5 –Conclusion………………………………………………………………….156 References…………………………………………………………………………...160 Appendix: Example of CEO pay cut in our sample………………………………………..161  iv  List of Tables 2.1 Summary Statistics of Corporate Acquisitions………….……………………………27 2.2 Managerial Horizon and Merger Performance……………….………………………29 2.3 Regression Analysis on Announcement Period Abnormal Returns…….……………30 2.4 Logit Regression Analysis on the Likelihood of Stock Payment………….…………31 2.5 Regression Analysis on Long-run Stock Performance Following the Acquisitions…...32 2.6 Regression Analysis on Acquisition Frequency…….………………………………….33 3.1 Descriptive Statistic of Sample Firms………………………….………………………81 3.2 Managerial Hedging Cost and Pay-performance Sensitivity………………….……….83 3.3 Managerial Hedging Cost and Sensitivity of CEO Wealth to Stock Volatility………...85 3.4 Simultaneous Equations (3SLS): Managerial Hedging Cost, Pay-performance Sensitivity, and Sensitivity of CEO Wealth to Stock Volatility………….…………………86 3.5 Regression Analysis on Capital Structure……………….……………………………..87 3.6 Regression Analysis on Executive Option Holding………….………………………88 3.7 Regression Analysis on Corporate Dividend Policy………….………………………..89 3.8 Regression Analysis on Corporate Diversification………………….…………………90 3.9 Beta and Incentive Pay……………………………...……………….…………………91 3.10 Financial Industry and Incentive Pay……………………………...………………….92 3.11 Direct Executive Hedging Transactions and Incentive Pay…………………………..93 4.1 Sample Distribution………………………….……………………………………….131 4.2 Summary Statistics………………….………………………………………………133 4.3 What Predicts a CEO Pay Cut?.....................................................................................136 4.4 CEO Retention After a Pay Cut………….…………………………………………138 4.5 Firm Performance After a CEO Pay Cut……….……………………………………..140 4.6 Capital Expenditures After a CEO Pay Cut….……………………………………….141 4.7 R&D Expenses After a CEO Pay Cut…….…………………………………………..142 4.8 Capital Structure After a CEO Pay Cut……………….…………………………….143 4.9 CEO Pay Recovery…………….……………………………………………………144  v  List of Figures Figure 4.1 CEO Pay Around a Pay Cut…………….……………………………………..145 Figure 4.2 Firm Performance Around a Pay Cut………………………………………….147 Figure 4.3 Capital Expenditures, R&D, and Capital Structure Around a Pay Cut……….148 Figure 4.4 Top Executive (Excluding CEO) Pay Around a CEO Pay Cut………………..151 Figure 4.5 Control-adjusted Performance Measures……………………………………...152  vi  Acknowledgements The completion of my thesis concludes the five years of wonderful life at the University of British Columbia and the city of Vancouver. I am quite grateful to a number of people who supported me during my program. First of all, I thank the faculty members at the Sauder School of Business. My committee members, Hernan Ortiz-Molina, Alan Kraus, and Ralph Winter, provided invaluable input and comments. I would especially like to thank my advisor, Kai Li, for her constant support and encouragement throughout my studies and the job market. I am also grateful to Tan Wang, Adlai Fisher, Murray Carlson, Robert Heinkel, Jason Chen, and Jenny Chen for their helpful advice. Second, I thank my colleagues and friends, especially Hamed Mahmudi, Feng Zhang, Wei Zhang, Alexandra Boonekamp, David Newton, Charles Gaa, and Kyung Shim. Finally, I would like to thank my parents, Chengyu Wang and other family members, who are currently living in China. This thesis would not have been possible without their support.  vii  Dedication For my parents  viii  Statement of Co-Authorship I am the solo author of my first essay (Chapter 2) and second essay (Chapter 3). The third essay of my thesis (Chapter 4) is a jointed work with Jarrad Harford and Kai Li. My contribution to the paper includes gathering the data, doing the empirical analysis, and writing part of the final manuscript.  ix  Chapter 1 Introduction In this thesis, I examine (1) the impact of managerial horizon on acquisition decisions, (2) how managerial incentive and corporate governance mechanism depend on the opportunity that the managers have to hedge their personal portfolios, and (3) the practice of making a big cut to the CEO’s annual pay package. The first paper shows that managerial horizon significantly shapes acquisition features. Measuring managerial horizon based on their career concern and compensation structure, I find that long-horizon bidding managers have lower announcement returns, higher likelihood of using equity to pay the deals, and better post-merger performance, than do the short-horizon ones. The results support the theory of stock market driven acquisitions (Shleifer and Vishny (2003)) and the agency cost of overvalued equity advanced by Jensen (2005). The second essay examines executive incentive and compensation contracts when those managers have the opportunities to hedge their stock-based compensation. Equity-based compensation is widely regarded as a useful tool to resolve managerial agency problems. Most of current literature assumes that managers cannot hedge their stock and option awards. However, executives do hedge their incentive portfolios through various channels, including equity swaps (Bolster et al. (1996)), zero-cost collars (Bettis et al. (2001)), and selling existing shares effectively (Ofek and Yermack (2000)). In this essay, I explore the research questions: How does managerial hedging alter a manager’s incentive and how do shareholders adjust governance structure correspondingly? 1  I find that executive hedging significantly influences managers’ decisions on corporate polices: When managers can hedge, they are less likely to diversify their own firms, have weaker intention to avoid dividend, and are less eager to cash out their vested option grants. I also document evidence that shareholders increase pay-for-performance sensitivity, convexity, and financial leverage to resolve managerial agency problem in response to managerial hedging issues. The results support the idea that managerial hedging undermines the managers’ incentive but enable them to take more risk, and that shareholders take this issue into consideration when designing corporate governance structure. In the third essay, I examine the practice of making large, discrete pay cuts to CEOs, which has been overlooked in the literature of CEO compensation. This study fits within the general question of how do boards change compensation to provide CEOs incentives? The study is particularly relevant for the executive compensation reform adopted by the US government on a broad basis in the financial industry bailout. The key findings are that that (1) CEO pay cuts are commonly used by the boards following poor performance and (2) on average, the CEO can improve firm performance after a pay cut and have their pay restored to the pre-pay-cut level. Our results suggest that the pay cut practice is consistent with the optimal contracting view of executive compensation.  2  References Bettis, J. C., J. M. Bizjak,, and M. L. Lemmon, 2001, Managerial ownership, incentive contracting, and the use of zero-cost collars and equity swaps by corporate insiders, Journal of Financial and Quantitative Analysis 36, 345-370. Bolster, P., D. Chance, and D. Rich 1996, Executive equity swaps and corporate insider holdings, Financial Management 25, 14-24. Jensen, M., 2005, Agency costs of overvalued equity, Financial Management 34, 5-19. Shleifer, A., and R. Vishny, 2003, Stock market driven acquisitions, Journal of Financial Economics 70, 295-311. Ofek, E., and D. Yermack, 2000, Taking stock: Equity-based compensation and the evolution of managerial ownership, Journal of Finance 55, 1367-1384.  3  Chapter 2 Market Misvaluation, Managerial Horizon, and Acquisitions1 2.1. Introduction Corporate acquisition decisions are susceptible to acute conflicts of interest among different groups of agents, including short-term shareholders, long-term shareholders, and managers. Although this topic has been extensively studied in the existing literature, it is not yet conclusive what explains the cross-sectional variations of the gains (losses) of bidding firms by making acquisitions, nor has a definitive conclusion emerged regarding the motivations behind these acquisition decisions. The purpose of this paper is to analyze the motives and consequences of mergers and acquisitions from a managerial horizon perspective. In particular, I examine the effect of managerial horizon on acquirers’ announcement returns, methods of payment in the transactions, long-term performance following the acquisitions as well as acquisition frequency. Managers of a long horizon place more emphasis on firms’ long-term value than the short-term value; they tend to make takeover decisions to increase firms’ long-run stock price. In contrast, short-horizon managers stress firms’ short-term performance and prefer acquisitions that enhance firms’ stock value in the short run. I test the above hypothesis using two proxies for managerial horizon, and find supporting evidence. Managerial horizon determines whether the managers are more concerned with the firm’s short-run stock price or with the long-run price. It significantly shapes acquisition decisions in an inefficient stock market, where the firm’s current market value (short-run 1  A version of this chapter will be submitted for publication. Gao, H., Market Misvaluation, Managerial Horizon, and Acquisitions. 4  value) deviates from its fundamental value (long-run value). Long-horizon managers tend to exploit overvalued stock price by making equity transactions, while short-horizon executives tend to boost the near-term stock price by catering to the investor sentiments. The existing literature proposes two channels through which managerial horizon influences takeover events. The first is the equity issuance channel: Long-horizon managers use overpriced equity to acquire the target’s assets. Shleifer and Vishny (2003) provide a model of market driven acquisitions to illustrate this idea. These mergers benefit the bidders’ long-run shareholders because they cushion the subsequent drop of the overvalued stock price. One of the key assumptions in the model is that the acquiring-firm managers act in the interests of the long-term shareholders (i.e., the managers have a long horizon). Their paper also suggests that short-horizon managers may avoid using equity to acquire if equity issuance knocks down the short-term stock price by somewhat revealing the signs of overvaluation to the market. The second perspective is the catering channel. Short-horizon managers may undertake acquisitions that the market wants to see, and enhance the firm’s short-term stock price, even though these mergers may cost the shareholders in the long run. As suggested by Jensen (2005), short-horizon managers tend to make financing and investment decisions to cater to market sentiment; they are likely to follow this sentiment and make risky negative net present value (NPV) investments that the market deems profitable. The paper’s main prediction is that long-horizon acquiring managers (as opposed to short-horizon ones) experience lower abnormal returns at merger announcements, are more likely to use equity as the payment mode, and have better post-merger stock performance. 5  The key explanatory variable is managerial horizon, which I propose two proxies to measure. The first one is a dummy variable, indicating whether or not the CEO is near retirement. Career concern is a natural factor correlated with managerial horizon. A near-retirement CEO usually has little time to remain in office and therefore it is less likely for the CEO to benefit from the firm’s long-term performance (Dechow and Sloan (1991) and Gibbon and Murphy (1992)). My second proxy is the value of the CEO’s restricted stock and options that become vested during a given year as a percentage of the CEO’s total pay. As suggested by existing literature, if a CEO has a sizeable amount of incentive portfolio to be vested, she will probably cash out these equity positions within a short period of time and will consequently be more concerned about the firm’s short-run stock price (see, e.g., (Hall and Murphy (2000) and (2002)). While the first proxy measures managerial horizon from the perspective of career concern, the second one focuses on the executive compensation scheme. The empirical result is consistent with my predictions. For example, a near-retirement bidding CEO (short horizon) is associated with about 0.79 percentage points higher bidders’ announcement return, 8 percentage points smaller likelihood of using stock to acquire, and 10 percentage points worse performance during the three years after the acquisitions. I also examine the relation between managerial horizon and acquisition frequency. The results show that long-horizon CEOs tend to make more acquisitions than the short-horizon ones do, and this relation becomes stronger when the acquirer’s market valuation is high. The article proceeds as follows. I develop my hypothesis in Section 2.2. Section 2.3 6  describes the proxies for managerial horizon. In Section 2.4, I describe the data and sample construction. Section 2.5 reports the empirical results. Finally, Section 2.6 concludes.  2.2. Main Hypothesis and Testable Implications The current literature suggests that managerial horizon influences acquisitions through two channels: the equity issuance channel and the catering channel. My main hypothesis is that long-horizon bidding managers undertake acquisitions through the equity issuance channel while their short-horizon counterparts make use of the catering channel.  2.2.1. Equity Issuance Channel Shleifer and Vishny (2003), Rhodes-Kropf et al. (2005), and Dong et al. (2006) suggest that stock market misvaluation is a significant driving force for merger activities. They provide evidence that an overvalued stock market stimulates managers to undertake more acquisitions, especially stock-financed acquisitions. Long-horizon managers tend to use their overpriced equity to acquire assets from targets to enhance bidders’ long-term share price. Managerial horizon plays an important role in whether the companies take advantage of their overvalued equity. Stein (1996) shows that the impacts of market inefficiency on financing and investment decisions depend on whether managers have long or short horizons. When the stock market can only partially correct the overvaluation at the news of an equity issuance, stock-financed acquisitions benefits bidders’ shareholders in the long run but harm those shareholders in the short run. The decrease of the short-term stock price is mainly due to the price-pressure-related losses associated with an equity transaction. Stein (1996) interprets this price pressure as investors updating their beliefs 7  somewhat when they see managers undertaking an equity transaction. Therefore, short-horizon managers, who are concerned about the firm’s short-term stock price, are less likely to use equity to acquire, so that they can preserve the overvaluation in the short term. This argument can be also supported by Emery and Switzer (1999), who find that bidders tend to choose the method of payment that can produce higher expected abnormal returns. Moreover, Stein’s (1996) theory predicts that short-horizon managers in equilibrium use cash to make investment to delay the revelation of overvaluation.  2.2.2. Catering Channel Market inefficiency can be exploited not only by issuing overvalued equity, but also by making investments to cater to investor sentiment. Managers with a short horizon may make an acquisition that has a negative NPV (and avoid an acquisition that has a positive NPV) as long such a strategy increases stock price in the short run. Consistent with the above view, Polk and Sapienza (2009) show that short-horizon managers tend to make investments that can boost short-term stock prices by stimulating or catering to market optimism. Although responding to present market demands may boost the near-term stock price, it generally costs the shareholders in the long run. As stated by Stein (1996), this catering behavior temporarily distorts the firm’s financing and investment decisions and therefore misallocates resources. Consistent with this argument, Polk and Sapienza (2009) find that companies that make investments catering to market demands have low subsequent stock return. Brandenburger and Polak (1996) and Hirshleifer (1993) also argue that concern for short-term stock price may lead managers to make decisions other than those suggested by 8  their own superior information. Jensen (2005) particularly addresses the cost of catering in the acquisition market. He points out that when short-horizon managers choose the acquisition projects that cater to investor wishes, the value of the core business is usually compromised. Shareholders in the long run will be worse off, even though the short-term stock price is boosted.2  2.2.3. Testable Predictions Based on the analysis above, I propose three testable implications. Implication 1. The acquirers managed by short-horizon CEOs experience higher stock returns at acquisition announcements than do the acquirers managed by long-horizon CEOs. Implication 2. The acquirers managed by short-horizon CEOs are less likely to make equity-financed acquisitions than are the acquirers managed by long-horizon CEOs. Implication 3. The acquirers managed by short-horizon CEOs have lower long-run stock return after the acquisitions than do the acquirers managed by long-horizon CEOs.  2.3. Proxy for Managerial Horizon Managerial horizon, like many other managerial characteristics, is naturally hard to observe. The first proxy is a dummy variable indicating whether or not the CEO is about to retire. This measure is fairly intuitive since a near-retirement CEO on average has a short horizon. Similar to Gibbons and Murphy (1992), I define MH1 as a dummy which takes the value of one if the CEO is less than 62 years old, and zero otherwise. In other words, MH1 equals one if the CEO is beyond three years of reaching age 65 or older than age 65, which 2  In the literature of asset pricing, Shleifer and Vishny (1990) show that short-term stock price of short-horizon managers is higher in equilibrium than that of long-horizon managers. 9  implies a long horizon. Consistent with this view, Dechow and Sloan (1991) find that CEOs who are near retirement tend to cut the firm’s research and development (R&D) and advertising, in order to increase the near-term earnings. Gibbon and Murphy (1992) also argue that managers become more short-term oriented when they approach retirement. However, simply treating near-retirement CEOs as short-horizon managers might be problematic, since several forces could push in the opposite direction. First of all, those CEOs who survive in their positions for a long time are likely to be quite successful. These managers as a group may have a longer horizon than young managers, because taking actions that maximize long-run value should lead to greater long-term success. To address this concern, I add CEO tenure as one independent variable in the regression analysis to control for the effect of past CEO work experience. Second, a founder CEO may have a long horizon even when she is retiring. Addressing this possibility, I construct a dummy variable indicating whether or not the CEO is the firm’s founder, and use this variable as a control. Moreover, CEOs in smaller and younger firms and CEOs with lower compensation tend to be more ambitious. They may hope to achieve short-run success in hopes of being rewarded with a more lucrative position in a larger company. Responding to this possibility, I control for firm size, firm age and CEO compensation level. Finally, Graham and Narasimhan (2004) and Malmendier and Tate (2005) find that CEOs from the Great Depression cohort tend to be more conservative in assessing external markets. The interpretation of this result is that early macroeconomic experience influences individuals’ economic decisions, even much later in life. To separate the effect of my horizon proxy 10  from that of the Great Depression, I include in the regressions a dummy variable that flags the CEOs from that cohort. My second proxy for managerial horizon is the value of the CEO’s restricted stock and options that become vested during a given year normalized by the CEO’s total pay. Stock and options form the majority of a CEO’s compensation. Newly-granted stock and options are always restricted from being sold or exercised until they become vested. If the CEO has little vested equity portfolio at hand, she may not be very concerned about the firm’s near-term stock price because it does not have much direct impact on her personal wealth. In contrast, if a sizeable amount of the CEO’s incentive portfolio becomes vested, she may be more concerned about the current stock price because a high short-term price increases her proceeds when she cashes out. This notion is true especially when managers are under-diversified and risk-averse. Hall and Murphy (2000) and (2002) argue that managers, who are risk-averse individuals and hold lots of their own firms’ equity, are quite eager to sell/exercise their vested stock and options. Malmendier and Tate (2005) also suggest that those managers should minimize their vested equity holdings in order to divest themselves of idiosyncratic risk. If a CEO has a large amount of stock and options that have recently become vested, she will probably cash out these positions within a short period of time and will consequently be more concerned about the firm’s short-run stock price.  2.4. Sample Selection and Data Description The sample of acquisitions comes from the SDC U.S. Mergers and Acquisitions Database. I begin with all completed deals announced between January 1, 1993 and December 31, 2004. I only select deals whose value is no less than $10 million and that for 11  which the acquirer controls more than 50% of the target’s shares after the acquisition. Deal value is defined by SDC as the total value paid by the acquiring firm, excluding fees and expenses. I also require that bidders have available stock price from CRSP, accounting information from Compustat, and CEO compensation and age data from ExecuComp. To identify the acquisitions that may have significant impact on shareholders’ and managers’ wealth, I eliminate those in which the deal value is less than 1% of the bidder’s total assets (market value) prior to the acquisitions. I finally end up with a sample of 2,894 deals. In my sample, 481 deals are made by bidding CEOs who are age 62 or older (MH1=1). To construct my second horizon proxy MH2, I first compute an executive’s total compensation (Totalpay) in a given year as the sum of her salary, bonuses, long-term incentive plans, the grant-date value of restricted stock and the Black-Scholes value of granted options. I then calculate Value_VestingEquity(t), the value of restricted stock and options that turn to be vested in Year t, as Value_VestingEquity(t) = Unvested_Equity(t-1) +  EquityGrant(t)–Unvested_Equity(t).  The  variables  Unvested_Equity(t)  and  Unvested_Equity(t-1) are the value of unvested stock and options in Year t and t-1, respectively; EquityGrant(t) is the value of newly-granted stock and options in Year t. I then define VestingEquity as dividing Value_VestingEquity by Totalpay. The information on the CEO’s total pay, unvested equity portfolio and newly-granted stock and options is obtained from ExecuComp. The variable VestingEquity captures the value of the turn-vested stock and options as the proportion of the CEO’s total pay. Finally, I compute MH2 as MH2=1–VestingEquity. The negative sign in front of VestingEquity indicates the notion that a CEO with a bigger VestingEquity variable has a relatively shorter horizon. A 12  higher MH2 value indicates a longer horizon. The correlation coefficient between MH1 and MH2 is around 0.01, implying that these two variables are not highly correlated. The dummy Founder takes the value of one if the CEO is one of the firm’s founders, and zero otherwise. Since ExecuComp does not provide information on whether the CEO is a founder, I follow the way of Adam et al. (2005) to construct this variable. In particular, the Founder dummy is set to be zero if the firm was incorporated 64 years or more prior to the current year or if the CEO joined the company at least four years after the firm’s incorporation. Cleary, the current CEO cannot be the founder in the above two cases. For the remaining firm-year observations, I check the information about the firm’s founder from proxy statements, annual reports and internet searches. In my sample, 245 deals are made by founder CEOs of bidders. To flag the CEOs from the Great Depression cohort, I construct the Depression dummy, which equals one if the CEO is born during the 1930s or earlier, and zero otherwise. In my sample, CEOs from the Great Depression cohort undertake 171 acquisitions. Other bidders’ characteristics, like firm size, ROA, leverage, firm age and stock return, are constructed from Compustat and CRSP. All of the variables are measured at the fiscal year-end prior to the acquisition announcements. To ensure that some outliers in the data are not driving my results, I winsorize all the continuous variables at one percentage tails. Table 2.1 presents descriptive statistics of my sample. Panel A reveals that acquisitions tend to be cyclical as both the total number and median deal value of mergers closely follow the business cycle expansion over the late 1990s. The evidence suggests significant 13  time series clustering of acquisition activity, especially the clustering of stock-financed mergers. The number of stock-financed acquisitions declines sharply in the early 2000s, which coincides with the decline of overall stock market. Panel B reports the characteristics of acquiring firms. The median bidder is quite large; its annual sales volume is $919 million. Bidding firms are performing well with median market-to-book ratio (M/B) of 2.5, past-year stock return of 21.6%, and ROA of 14%. The median FirmAge is about 14 years. The median CEO is 54 years old; her tenure is six years. She is holding a VestingEquity of 0.23, implying that the value of stock and options that have just become vested is about 23% of her total annual pay. Her total annual compensation is $2,278 thousand and her ownership of the firm (inclusive of options) is 1.73%. Following standard event study methods, I estimate the three-day accumulative abnormal returns (CAR3) over the event windows (-1, 1) around the announcement date (day 0) based on the market model using CRSP value-weighted index returns. The parameters are estimated within a (-200, -60) event window relative to the announcement date. The CAR3 of the bidder is slightly positive, with a mean of 0.3% and a median of 0.2%. Following Barber and Lyon (1997) and Barber et al. (1999), I employ the control firm approach which uses control firm size, book-to-market ratio, and prior-year stock return as the benchmark for post-merger stock performance. Barber and Lyon (1997) report that the control firm approach not only eliminates the skewness bias associated with the long-term buy-and-hold abnormal returns but also yields well-specified statistics. Following Barber et 14  al. (1999) and Chen et al. (2007), I sort the population of CRSP firms into 14 size groups and 10 book-to-market groups. After determining which of the 140 groups the bidding firm belongs to at the month prior to the acquisition completion, I choose from the group the control firm that is the closest match on prior-year stock return and is not involved in any acquisition events. Then, three-year buy-and-hold returns (starting from the month after the acquisition completion) are computed for the sample firms and control firms. At last, the three-year buy-and-hold abnormal returns are the difference between sample firm returns and corresponding returns of control firms (BHAR3). The BHAR3 has a mean of –6.3% and a median of –6.5%, suggesting that the acquirer underperforms its control by a considerable magnitude. Table 2.2 reports my univariate comparison of merger performance sorted by managerial horizon; it shows that long-horizon bidding CEOs have lower announcement returns but better post-merger long-run performance. In Panel A, I sort the sample by MH1. The average CAR3 in the subsample with MH1=1 (long horizon) is about 0.1%; this number is significantly less than 0.9%, the average CAR3 in the subsample with MH1=0 (short horizon). The comparison of medians gives the same result. The difference in median CAR3 between long-horizon bidders and their short-horizon counterparts is about -0.5% and this difference is statistically significant at the 1% level. The mean BHAR3, when MH1=1, is -5%, about six percentage points bigger than the value when MH1=0; the difference is significant at the 10% level. The comparison of BHAR3 in medians provides a similar but less significant result. In Panel B of Table 2.2, the acquisition sample is sorted into quintiles based on MH2. 15  Panel B reports the means and medians of CAR3 and BHAR3 for the largest and smallest quintiles. Although the difference in CAR3 is not statistically significant, the mean BHAR3 in the biggest quintile is about 12.7 percentage points higher than that in the smallest quintile. The insignificance of the mean test in CAR3 also suggests that it is important to control for some confounding variables in the regression analysis. Overall, the univariate test provides supportive evidence that long-horizon bidders experience worse performance in the short run but better outcomes in the long run than do short-horizon bidders.  2.5. Empirical Results 2.5.1. Abnormal Returns during Announcement Periods I run several cross-sectional OLS regressions using the following model: CAR3 = δ 0 + δ1Horizon + δ 2 PastrReturn + δ3 M / B + δ 4 ROA +δ5 Firmsize + δ 6 Leverage + δ 7Tender + δ8 Depression + δ9 Founder +δ10 Ln(Totalpay) + δ11Firmage + δ12Tenure + δ13Ownership + YearDummies  (2.1)  + IndustryDummies + ε where the dependent variable is the bidder’s three-day announcement abnormal returns (in percentage). Tender is a dummy variable that equals one if the acquisition is a tender offer, and zero otherwise. Year dummies are employed to account for economy-wide shocks. Industry dummies based on the 48-industry classification of Fama and French (1997) are employed to control for industry effects. According to Implication 1, I expect δ1 to be negative. Corresponding p-values are computed based on Huber-White robust standard errors.3 3 This paper follows the classical “irrational investors approach” in the literature of behavioral corporate finance, which assumes that rational managers make corporate decisions in response to stock market mispricing (Baker et al. (2007)). Although it is difficult to directly measure stock market mispricing, I use M/B ratio to control for it following Dong et al. (2006). 16  The primary result from Table 2.3 is that the coefficients on Horizon, proxied by MH1 and MH2, are negative and significant in all of the six models. In Regression (1), I include MH1 as well as year and industries dummies as the independent variables. The coefficient on MH1 is –0.71 with the p-value of 0.019. In Regression (2), I include PastReturn, M/B, ROA, FirmSize, Leverage and Tender as additional control variables. The coefficient of MH1 is -0.71 with a p-value of 0.017. In Regression (3), I add Depression, Founder, Ln(TotalPay), FirmAge, Tenure and Ownership as additional controls. The coefficient on MH1 is -0.76 and is significant at the 5% level. CAR3 is decreased by 0.76 percentage points when MH1 changes from zero to one, compared to the sample mean of 0.3%. In Regressions (4)-(6), I substitute MH1 with MH2 to test Implication 1 under this alternative proxy for managerial horizon. The coefficients on MH2 are -0.05, -0.11, and -0.12, respectively, and all of them are significant at the 10% level, thus indicating that managerial horizon measured by MH2 is also negatively associated with bidders’ announcement returns. Taking Model 6 as another example, the coefficient on MH2 indicates that an increase in MH2 by one standard deviation (2.21) is expected to decrease CAR3 by 0.26 percentage points. Like Dong et al. (2006), I also find that bidders of high M/B ratio experience lower announcement returns. Dong et al. (2006) interpret this result to mean that overvalued bidders have more negative returns at acquisition announcements. In addition, my regression shows that larger bidders have lower CAR3, which is consistent with the size effects of acquirers documented by Moeller et al. (2004). Notably, the Founder dummy has 17  a significantly negative coefficient, indicating that a founder CEO experiences poorer announcement returns than do other CEOs. Overall, my findings support the prediction that acquiring firms managed by short-horizon CEOs experience higher abnormal returns at merger announcements.  2.5.2. Probability of Equity Payment I use logit regressions to test whether or not long-horizon CEOs are more likely to pay with equity for acquisitions. Specifically, I estimate the following model:  Pr( Allstock ) = F ( Horizon, PastReturn, M / B, ROA, FirmSize, Leverage, Tender , Depression, Founder , Ln(TotalPay ), FirmAge, Tenure, Ownership, (2.2) YearDummies, IndustryDummies) The dependent variable, Allstock, takes a value of one if the acquisition is paid only by equity and zero otherwise. The variable F denotes the logit cumulative distribution function. Based on Implication 2, the coefficients on Horizon are expected to be positive. The regression results are reported in Table 2.4, where the coefficients are estimates of the marginal effect on the probability when all of the other independent variables are at their mean value.4 Table 2.4 indicates that the probability of equity payment increases with managerial horizon. In all of the six models, the coefficients on MH1 and MH2 are positive and significant, and their economic magnitude is also quite sizeable. In Models (1)-(3), I use MH1 and some control variables. The coefficient on MH1 is 0.08 and it is significant at the 1% level in Model (1), where the control variables just include year dummies and industry dummies. This result implies that an increase of MH1 from zero to one will increase the probability of equity payment by about 8 percentage 4  I also make a subsample where acquisitions with mixed payments are excluded, and re-run the logit regressions in Table 2.4. The results remain almost completely the same. 18  points. Given the fact that the unconditional probability of an acquisition in my sample being financed all by equity is 24.5% (see Table 2.1), this marginal effect is definitely remarkable. In Model (2), I include PastReturn, M/B, ROA, Size, Leverage, Cash, and Tender as additional explanatory variables. The variable MH1 has a coefficient of 0.067 which is significant at the 1% level. I further add Depression, Founder, Ln(Totalpay), FirmAge, Tenure and Ownership in Model (3), in which the coefficient on MH1 is significant and has a value of 0.07. In Models (4)-(6), I use MH2 instead of MH1 in the regression analysis; the results are similar to those in the previous three models. Taking Model (4) for example, the coefficient on MH2 is 0.008 and the p-value is 0.042, indicating that an increase in MH2 from the mean by one standard deviation is expected to increase the probability of equity payment by about 1.8 percentage points. Of the other control variables, M/B ratio is positively associated with the use of stock. This result is consistent with the prior literature which suggests that high-valuation firms are more likely to use equity to make acquisitions (Dong et al. (2006)). Firms with higher ROA are less likely to use equity. This finding supports the premise that bidders tend to use cash to acquire when they have high cash flow (Martin (1996)). The coefficient on Leverage is negative and is significant, indicating that less levered bidders make more stock-financed acquisitions. Like prior literature, I find that a tender offer is usually associated with cash payment. Moreover, founder CEOs tend to use stock to acquire. This finding contributes to the recently increasing focus in the literature on founder CEOs, who behave quite differently from successor CEOs (Fahlenbrach (2008)). 19  The results from Table 2.4 support the prediction that long-horizon acquiring CEOs are more likely to use equity as the exchange medium than are their short-horizon counterparts.  2.5.3. Long-term Stock Performance Following Acquisitions I run cross-sectional OLS regressions to test whether or not long-horizon CEOs have better long-term stock performance following acquisitions. The following model is estimated: BHAR3 = α0 + α1Horizon + α2 PastReturn + α3M / B + α4 ROA + α5 FirmSize +α6 Leverage + α7Tender + α8 Depression + α9 Founder + α10 Ln(TotalPay)  (2.3)  +α11FirmAge + α12Tenure + α13Ownership + YearDummies + IndustryDummies + ε The dependent variable is three-year buy-and-hold abnormal returns (in percentage) after the deal completion, as defined in Section 2.4. Based on Implication 3, I expect α1 to be positive. As shown in Table 2.5, the coefficients of MH1 and MH2 are significant and positive in all of the regression models. The coefficients on MH1 are around 10, implying that an increase in MH1 from zero to one is expected to enhance BHAR3 by around 10 percentage points, which is certainly economically important relative to the unconditional average of –6.3%. The coefficients on MH2 are about 3, which indicates that BHAR3 will increase by about 6.6 percentage points when MH2 increases by one standard deviation. The coefficients on ROA are consistently positive and significant at the 5% level, suggesting that bidders’ prior-year accounting performance is positively associated with their long-run stock performance. M/B has significantly negative coefficients, which is consistent with previous literature that glamour-focused acquirers perform poorly after the 20  transactions (see, e.g., Rau and Vermaelen (1998) and Andre et al. (2004)). All of the other control variables are generally insignificant. The major conclusion from this regression analysis is that managerial horizon emerges as an important and robust determinant of long-term stock performance after the merger. The results support the prediction that a long-horizon acquiring CEO has better long-run performance following acquisitions than does a short-horizon manager.5  2.5.4. Additional Analysis on Acquisition Frequency All of the above analysis relates to either the value implications or financing choices conditional on making a merger deal. It is interesting to investigate the effect of managerial horizon on the choice of whether or not to make an acquisition at all. A manager with a short horizon might refrain from undergoing risky projects like pursuing acquisitions. In contrast, a manager with a long horizon is more likely to undertake mergers when her firm is overvalued. In this section, I examine how managerial horizon influences the manager’s decision to make acquisitions by estimating the following equation:  AcquisitionNumberit = β0 + β1Horizonit −1 + β2 Horizonit −1 × M / Bit −1 + β3M / Bit −1 +β4 PastReturnit −1 + β5 ROAit −1 + β6 FirmSizeit −1 + β7 Depressionit −1 + β8 Founderit −1 +β9 Ln(Totalpay)it −1 + β10 Firm _ Ageit −1 + β11Tenureit −1 + β12Ownershipit −1  (2.4)  +YearDummies + IndustryDummies + εit where i indexes firms and t indexes years. The dependent variable is the number of acquisitions made by a company in a certain year. The independent variables include managerial horizon proxies, M/B ratio, the interaction between horizon and M/B ratio, and other controls. The M/B variable measures the firm’s market valuation. The regression is based on data for firms in ExecuComp from 1993 to 2004. 5  Like Dong et al. (2006), I don’t control for method of payment in Table 2.3 or Table 2.5 because the payment method is treated as a choice variable in the paper. As a robust check, I find similar results after including it in the regressions. 21  In Panel A of Table 2.6, I run cross-sectional OLS regressions and compute the p-values based on robust standard errors clustered at the firm level. As shown in Panel A, long-horizon managers tend to make more acquisitions than do short-horizon CEOs, and this relation is stronger when firms have inflated market valuation. In Model (1), the coefficient on MH1 is around 0.017 and is significant at the 10% level. This result implies that a manager with a long horizon makes more acquisitions. In Model (2), I add the interaction between MH1 and M/B, MH1×M/B. This interaction term has a positive but insignificant coefficient of 0.005. I replace MH1 with MH2 in Models (3) and (4) to further examine the relation between managerial horizon and the acquisition frequency. The term, MH2×M/B, has a positive coefficient which is significant at the 1% level. This result implies that long-horizon managers are more likely to undertake acquisitions when their firm’s market valuation is high. Not surprisingly, bigger firms and firms with high stock return complete more acquisitions. Like Fahlenbrach (2008), I also find that founder CEOs are more active in acquisition activities. The finding that highly paid CEOs are associated with more acquisitions could be explained by the fact that CEOs with high compensation are usually employed by big firms. The variable FirmAge has a significantly negative coefficient, indicating that mature firms are less likely to undergo corporate restructuring, such as merger and acquisition activity. The coefficient of Depression is negative but not always significant; it suggests that CEOs born during the 1930s or earlier are making fewer takeovers. This result is also broadly consistent with Malmendier and Tate (2005)’s finding that the CEOs born during the Great Depression period are on average making less capital 22  investments. There could be omitted firm characteristics that are correlated with the horizon measures and therefore explain the differences in acquisitiveness. To address this possibility, I run firm fixed effect regressions in Panel B. The regression specification is the same with that in Panel A, except that I replace industry dummies with firm fixed effects. The regression results in Panel B are very similar to those in Panel A. The interaction MH1×M/B does not have significant coefficients in either of the two panels. This result may be due to the high correlation between MH1×M/B and M/B: The corresponding correlation coefficient is 0.88.6 Moreover, the coefficients in front of MH1 and MH2 are not always consistent with each other. They thus could indicate that the impact of managerial horizon on acquisitiveness is not as evident as that on valuation effects and payment choice. Overall, the results in Table 2.6 support, at least weakly, the view that long-horizon managers make more acquisitions than do short-horizon managers and that long-horizon managers tend to make use of high market valuation to complete acquisitions.  2.5.5. Alternative Explanation and Robustness Checks CEO age might be a proxy for CEO ability. The older is the CEO, the greater are the market’s assessments of her ability given that this CEO has survived in the labor market (Milbourn (2003)). Under this view, my horizon variable, MH1, is expected to be negatively associated with CEO ability. But this interpretation makes no predictions towards the choice of payment mode in acquisitions. Moreover, it is inconsistent with the  6  The correlation coefficient between MH2×M/B and M/B is 0.33. 23  empirical result that MH1 is positively related to long-run abnormal returns, because CEOs with high ability should not underperform low-ability managers in the long term. Another interpretation for the MH2 variable is the CEO’s equity-based compensation. A CEO with more stock-based compensation may, on average, have more vested stock and options (a lower MH2 value). Existing literature suggests that equity-based compensation alleviates managerial agency problems and influences acquisition decisions; therefore, MH2 may capture the incentive effects from equity compensation. Datta et al. (2001) show that bidding CEOs paid under more equity-based compensation schemes are associated with better announcement returns and post-acquisition long-run performance. Their results imply that MH2 is not capturing the incentive-alignment effects from stock-based pay, because MH2 is positively associated with post-merger long-run performance. Finally, managerial horizon may also be interpreted as shareholders’ horizon, if shareholders select the CEOs who have the same horizon as theirs. Gaspar et al. (2005) use share turnover by investors as a proxy for shareholders’ horizon, and show that bidding firms with short-term shareholders experience lower announcement abnormal returns as well as lower long-run performance after mergers. This finding implies that it is inappropriate to interpret managerial horizon as shareholders’ horizon, because acquiring firms managed by short-horizon CEOs have higher announcement returns than do those managed by long-horizon CEOs. My regression results are robust to a number of different specifications. For example, I obtain similar coefficients on MH1 and MH2 in all the regressions if (1) I compute firm size as Ln(Total Assets) or Ln(Market Capitalization) instead of Ln(Sales Volume), (2) I use 24  Return on Equity or Sales Growth as proxies for the firm’s accounting performance, (3) I calculate the accumulative abnormal returns at announcement during the event windows (-1, 0) and (-2, 2), (4) I use control firms only on book-to-market ratio and size as benchmarks to compute the long-run abnormal returns, (5) I construct industry dummies using two-digit SIC code, and (6) I use a probit regression model in Table 2.4.  2.6. Conclusion The major goal of this paper is to establish the relationship between managerial horizon and corporate takeover decisions. The paper analyzes how managers with different horizons make acquisitions as a response to market inefficiencies. Long-horizon managers (1) focus more on firms’ long-term value, and (2) tend to use their overvalued equity to acquire target firms in order to preserve some of the temporary overvaluation for long-run shareholders. In contrast, short-horizon managers (1) pay greater attention to firms’ short-term value, (2) prefer using cash to pay for mergers for the purpose of hiding the information about firms’ overvaluation and (3) tend to cater to investor sentiment by completing the acquisitions that the market is currently optimistic about, even at the expense of their firms’ long-run value. Specifically, I test the predictions that the acquiring firms controlled by long-horizon managers have lower abnormal returns around announcements, higher likelihood of using equity as the payment mode, and better post-merger stock performance, than do the bidders controlled by short-horizon managers. Two proxies are used to measure managerial horizon from the perspectives of managerial career concern and executive compensation. I argue that retirement and vested 25  equity portfolio are two important factors for determining a short horizon. A near-retirement CEO on average has a shorter horizon than other CEOs. Also, when the CEO has a sizeable incentive portfolio becoming vested, she will be more concerned about the short-term stock price and will thus have a relatively short horizon. The empirical evidence supports all of the above predictions. This article is intended as a first systematic examination on the influences of managerial horizon on merger and acquisition activity. It raises many interesting questions for future research as well. The effect of managerial horizon on corporate policies such as debt and equity issuance, share repurchase, dividend, and investment remains to be explored. A good deal of empirical evidence supports the view that managers make the above policies to exploit market inefficiencies. In future research, it would be interesting to explore the relation between managerial horizon and the above corporate decisions.  26  Table 2.1. Summary Statistics of Corporate Acquisitions Panel A: Distribution of Mergers and Acquisitions by Announcement Year The sample includes 2,894 completed U.S. acquisitions from January 1, 1993 to December 31, 2004 as listed by SDC, where the acquiring firm gains control of the target firm and whose deal value is at least $10 million and 1% of the bidder’s total assets (market value). Deal value is defined by SDC as the total value of consideration paid by the acquiring firm, excluding fees and expenses. The sample includes bidding firms for which stock price data is in CRSP, accounting data is in Compustat, and CEO compensation and age data is in ExecuComp. Stock (Cash) refers to the deals where the payments to the targets are all by stock (cash). Mixed refers to deals in which both cash and stock are used.  Year  Number of Acquisitions (%)  Median Deal  Method of Payment  Value ($ Millions)  Stock  Cash  Mixed  1993  23 (0.8)  43  8  5  10  1994  167 (5.8)  128  44  54  69  1995  252 (8.7)  92  95  61  96  1996  271 (9.4)  122  74  69  128  1997  316 (11)  196  100  68  148  1998  310 (10.7)  201  100  61  149  1999  337 (11.6)  241  100  69  168  2000  296 (10.2)  240  91  49  156  2001  232 (8)  120  38  57  137  2002  238 (8.2)  93  22  87  129  2003  224 (7.7)  121  18  88  118  2004  228 (7.9)  129  20  91  117  Total  2894 (100)  142  710  759  1425  27  Table 2.1, continued Panel B: Acquiring Firm Characteristics The sample includes 2,894 completed U.S. acquisitions from January 1, 1993 to December 31, 2004. VestingEquity is the value of the CEO’s restricted stock and option holding that becomes vested in a given year normalized by the CEO’s total annual pay. Totalpay ($K) is the CEO’s total annual compensation. Ownership is the fraction of the firm’s shares owned by the CEO inclusive of options. Tenure is the number of years that the CEO has served in her position. Sales ($million) refers to the firm’s annual sales volume. M/B is the ratio of market value of equity over book value of equity. ROA is the accounting return on assets, obtained as the ratio of earning before interest and taxes to total assets. Leverage is the ratio of total debt over total assets. FirmAge refers to the age of the company. PastReturn denotes the compound stock return of acquiring firms over the year prior to acquisition announcements. All of the above variables are measured at the fiscal year end prior to the acquisition announcement. CAR3 is the three-day accumulative abnormal returns over the event window (-1, 1) around the announcement date (day 0) based on the market model using CRSP value-weighted index returns. The parameters are estimated within a (-200, -60) event window relative to the announcement date. BHAR3 is the three-year buy-and-hold abnormal return following the acquisition. Matching firms are formed based on size, book-to-market ratio and stock return. Year  Mean  Std  5th Pct  Median  95th Pct  CEO Age VestingEquity TotalPay ($K) Ownership Tenure Sales M/B ROA Leverage FirmAge PastReturn CAR3 BHAR3  54.3 0.19 4748 4.14% 7.6 3005 3.9 14% 22% 17.9 34.9% 0.3% -6.3 %  7.3 2.21 7378 6.46% 6.8 5783 5.1 9% 17% 13.3 71.7% 6% 114%  42 0 444 0.16% 1 104 0.9 2% 0 2 -42.7% -9.9% -185%  54 0.23 2278 1.73% 6 919 2.5 14% 21% 14 21.6% 0.2% -6.5%  67 2.72 17661 18.2% 21 13544 11.7 29% 53% 46 157.9% 10.2% 175%  28  Table 2.2. Managerial Horizon and Merger Performance Panel A: Sorting Sample by MH1 The sample includes 2,894 completed U.S. acquisitions from January 1, 1993 to December 31, 2004. I sort all selected deals into subsamples based on MH1, where MH1, my first proxy for managerial horizon, equals one if the CEO is less than 62 years old, and zero otherwise. A CEO with MH1=1 is expected to have a longer horizon than a CEO with MH1=0. CAR3 is the three-day accumulative abnormal return around the announcement date. BHAR3 is the three-year buy-and-hold abnormal return following the acquisition. Matching firms are formed based on size, book-to-market ratio and stock return. The middle column and final column give the difference of the two means and the two medians, respectively. The tests of means are based on t-statistics; the tests of medians are based on Wilcoxon signed tests. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Mean  CAR3 BHAR3  Median  MH1=1 (1)  MH1=0 (2)  Difference (1)-(2)  MH1=1 (4)  MH1=0 (5)  Difference (4)-(5)  0.1% -5%  0.9% -11%  -0.8%*** 6%*  0.1% -6.4%  0.6% -8.6%  -0.5%*** 2.4%  Panel B: Sorting Sample by MH2 The sample includes 2,894 completed U.S. acquisitions from January 1, 1993 to December 31, 2004. I sort all selected deals into quintiles based on MH2. The variable MH2 is my second proxy for managerial horizon; it is computed as 1-VestingEquity, where VestingEquity is the value of the CEO’s restricted stock and option holding that becomes vested in a given year normalized by the CEO’s total annual pay. A CEO with a larger value of MH2 is expected to have a longer horizon. I report the means and medians for the largest and smallest quintiles. The middle column and final column give the difference of the two means and the two medians, respectively. The tests of means are based on t-statistics; the tests of medians are based on Wilcoxon signed tests. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Mean Largest MH2 (1) CAR3 BHAR3  0.2% -0.8%  Smallest MH2 (2) -0.04% -13.5%  Median Difference (1)-(2) 0.24% 12.7%**  29  Largest MH2 (4) 0.02% -5%  Smallest MH2 (5)  Difference (4)-(5)  0.2% -9%  -0.18% 4%  Table 2.3. Regression Analysis on Announcement Period Abnormal Returns The dependent variable is three-day (-1,1) accumulative abnormal returns (in percentage). The variables Tender, Depression and Founder are dummies that equal one if the deal is a tender offer, if the CEO is born during the 1930s or earlier, and if the CEO is one of the firm’s founders, respectively, and zero otherwise. Corresponding p-values from Huber-White robust standard errors are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  MH1  (1)  (2)  (3)  -0.71** [0.019]  -0.71** [0.017]  -0.76** [0.022]  MH2  -0.05* [0.098]  -0.11*** [0.006] 0.007*** [0.003] -0.09*** [0.01] 3.76** [0.02] -0.36*** [0.000] 1.49* [0.06] 0.04 [0.92]  -0.12** [0.04] 0.007*** [0.002] -0.12*** [0.01] 3.63** [0.03] -0.36*** [0.004] 1.15 [0.17] 0.07 [0.71] 0.44 [0.46] -0.81* [0.06] 0.01 [0.94] 0.001 [0.94] 0.005 [0.83] 1.59 [0.53]  Yes  Yes  Yes  Yes  Yes  0.82 [0.29] 2893 2.2%  6.89*** [0.000] 2833 3.5%  6.71*** [0.005] 2734 3.4%  0.44 [0.58] 2839 2.2%  6.37*** [0.000] 2779 3.7%  ROA Size Leverage Tender Depression Founder Ln(Totalpay) FirmAge Tenure Ownership  N Adjusted-R2  (6)  0.01*** [0.003] -0.1*** [0.004] 3.48** [0.04] -0.39** [0.001] 1.08 [0.19] -0.14 [0.73] 0.49 [0.46] -0.85** [0.04] 0.06 [0.68] -0.002 [0.84] -0.005 [0.81] 1.62 [0.51]  M/B  Intercept  (5)  0.01*** [0.004] -0.09** [0.012] 3.63** [0.027] -0.35*** [0.000] 1.11 [0.16] -0.05 [0.91]  PastReturn  Year&Industry Dummies  (4)  30  Yes 6.73*** [0.006] 2626 3.6%  Table 2.4. Logit Regression Analysis on the Likelihood of Stock Payment This table reports logit models predicting the probability of stock payment in acquisitions. The sample includes 2,894 completed U.S. acquisitions from January 1, 1993 to December 31, 2004. The dependent variable is a dummy variable that equals one if acquirers pay the targets all by stock, and zero otherwise. The coefficients are estimates of the marginal effect on the probability when all of the independent variables are at their mean value. Corresponding p-values from Huber-White robust standard errors are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  MH1  (1)  (2)  (3)  0.08*** [0.000]  0.067*** [0.000]  0.07*** [0.005]  MH2  (5)  (6)  0.008** [0.042]  0.005* [0.055] 0.000 [0.25] 0.014*** [0.000] -0.23** [0.019] 0.001 [0.85] -0.22*** [0.000] -0.16*** [0.000]  0.004* [0.056] 0.000 [0.89] 0.014*** [0.000] -0.28*** [0.007] 0.001 [0.86] -0.18*** [0.000] -0.16*** [0.000] -0.059* [0.09] 0.056** [0.04] 0.016 [0.19] -0.000 [0.98] -0.02 [0.45] -0.06 [0.66]  0.000 [0.36] 0.013*** [0.000] -0.25** [0.012] 0.003 [0.64] -0.21*** [0.000] -0.16*** [0.000]  0.000 [0.18] 0.014*** [0.000] -0.31*** [0.002] 0.004 [0.61] -0.21*** [0.000] -0.16*** [0.000] -0.009 [0.86] 0.051** [0.03] 0.01 [0.17] -0.000 [0.90] -0.01 [0.48] 0.03 [0.82]  Yes  Yes  Yes  Yes  Yes  Yes  2894 18.7%  2765 23.3%  2666 23.6%  2818 18.7%  2680 23.4%  2546 23.1%  PastReturn M/B ROA Size Leverage Tender Depression Founder Ln(Totalpay) FirmAge Tenure Ownership Year&Industry Dummies N Pseudo-R2  (4)  31  Table 2.5. Regression Analysis on Post-Merger Long-run Stock Performance The sample includes 2,894 completed U.S. acquisitions from January 1, 1993 to December 31, 2004. The dependent variable is three-year buy-and-hold abnormal returns (in percentage) after the completion date of the merger, as defined in Table 2.1. Corresponding p-values from Huber-White robust standard errors are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  MH1  (1)  (2)  (3)  9.09* [0.089]  10.89** [0.045]  11.43* [0.07]  MH2  3.31*** [0.000]  2.98*** [0.000] 0.044 [0.19] -0.93* [0.052] 65.92** [0.042] 3.53** [0.04] -3.59 [0.82] -2.76 [0.74]  3.25*** [0.000] 0.044 [0.29] -0.81* [0.1] 69.88** [0.04] 2.01 [0.35] -3.79 [0.81] -2.85 [0.75] -1.71 [0.74] -1.41 [0.83] 0.79 [0.31] 0.09 [0.96] 0.34 [0.41] -27.73 [0.51]  Yes  Yes  Yes  Yes  Yes  -0.19 [0.98] 2807 3.4%  -69.32* [0.068] 2751 3.7%  -78.72* [0.1] 2653 3.6%  13.31 [0.33] 2753 4.5%  -64.42* [0.085] 2697 4.5%  ROA Size Leverage Tender Depression Founder Ln(Totalpay) FirmAge Tenure Ownership  N Adjusted-R2  (6)  0.013 [0.69] -0.86* [0.071] 70.99** [0.026] 1.33 [0.53] -7.61 [0.62] -6.41 [0.51] 2.71 [0.71] 0.34 [0.94] 2.01 [0.16] 0.073 [0.71] 0.51 [0.14] -26.91 [0.51]  M/B  Intercept  (5)  0.018 [0.58] -0.81* [0.088] 69.27** [0.029] 3.32* [0.052] -5.87 [0.71] -4.13 [0.62]  PastReturn  Year&Industry Dummies  (4)  32  Yes -68.31 [0.11] 2550 4.3%  Table 2.6. Regression Analysis on Acquisition Frequency Panel A: Pooled OLS Regression  The dependent variable is the number of acquisitions made by a company in a given year for the ExecuComp population firms during 1993-2004. Corresponding p-values from robust standard errors clustered at the firm level are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  MH1  (1)  (2)  0.017* [0.09]  0.002 [0.94] 0.005 [0.34]  MH1×M/B  (3)  (4)  0.001 [0.85]  0.006* [0.06] 0.092*** [0.000] 0.003 [0.975] 0.044*** [0.000] -0.04 [0.22] 0.055** [0.031] 0.053*** [0.000] -0.002** [0.022] -0.002 [0.14] 0.17 [0.26]  0.001 [0.95] 0.092*** [0.000] 0.005 [0.954] 0.044*** [0.000] -0.042 [0.21] 0.056** [0.031] 0.053*** [0.000] -0.002** [0.023] -0.002 [0.14] 0.17 [0.27]  0.006* [0.054] 0.090*** [0.000] -0.01 [0.904] 0.043*** [0.000] -0.05* [0.091] 0.059** [0.023] 0.056*** [0.000] -0.002** [0.028] -0.002* [0.06] 0.14 [0.35]  -0.012 [0.19] 0.003*** [0.008] -0.29*** [0.005] 0.090*** [0.000] -0.19*** [0.001] 0.039*** [0.000] -0.059*** [0.009] 0.064*** [0.000] 0.066*** [0.000] -0.001*** [0.003] -0.002** [0.013] 0.05 [0.59]  Yes  Yes  Yes  Yes  -1.49*** [0.000] 15937 6.3%  -1.48*** [0.000] 15937 6.3%  -1.51*** [0.000] 15329 6.3%  -1.52*** [0.000] 15329 6.3%  MH2 MH2×M/B M/B PastReturn ROA Size Depression Founder Ln(Totalpay) FirmAge Tenure Ownership Year&Industry Dummies Intercept N Adjusted-R2  33  Panel B: Firm Fixed Effects Regression The regression specification is the same as that in Panel A, except that I run firm fixed effects regressions. The control variables (unreported) have similar coefficients to those in Panel A. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  MH1 MH1×M/B  (1)  (2)  0.013* [0.08]  0.003 [0.97] -0.001 [0.83]  MH2 MH2×M/B  34  (3)  (4)  0.002 [0.53]  -0.003 [0.42] 0.002* [0.072]  References Adam, R.B., H. Almeida, and D. Ferreira, 2005, Powerful CEOs and their impact on corporate performance, Review of Financial Studies 18, 1403-1432. Andre, P., M. Kooli, and J. L’Her, 2004, The long-run performance of mergers and acquisitions: Evidence from the Canadian stock market, Financial Management 33, 27-43. Baker, M., R. Ruback, and J. Wurgler, 2007, Behavioral corporate finance: A survey, in B.Espen Eckbo (ed.), Handbook of Corporate Finance: Empirical Corporate Finance, Handbooks in Finance Series, Elsevier/North Holland. Barber, B.M., and J.D. Lyon, 1997, Detecting long-run abnormal stock returns: The empirical power and specification of test statistics, Journal of Financial Economics 43, 341-372. Barber, B.M., C. Tsai, J.D. Lyon, 1999, Improved methods for tests of long-run abnormal stock returns, Journal of Finance 54, 165-201. Bertrand, M. and A. Schoar, 2003, Managing with style: The effect of managers on firm policies, Quarterly Journal of Economics 118, 1169-1208. Brandenburger, A., and B. Polak, 1996, When managers cover their posteriors: Making the decisions the market wants to see, RAND Journal of Economics 27, 523-541. Chen, X., J. Harford, and K. Li, 2007, Monitoring: Which institutions matter?, Journal of Financial Economics 86, 279-305 Datta, S., M. Iskandar-Datta, and K. Raman, 2001, Executive compensation and corporate acquisition decisions, Journal of Finance 56, 2299-2336. Dechow, P.M., and R.G. Sloan, 1991, Executive incentives and the horizon problem: An empirical investigation, Journal of Accounting and Economics 14, 51-89. Dong, M., D. Hirshleifer, S. Richardson, and S. Teoh, 2006, Does investor misvaluation drive the takeover market?, Journal of Finance 61, 725-762. Emery, G.W., and J.A. Switzer, 1999, Expected market reaction and the choice of method of payment of acquisitions, Financial Management 28, 73-86. Fahlenbrach, R., 2008, Founder-CEOs, investment decisions, and stock market performance, Journal of Financial and Quantitative Analysis, forthcoming. 35  Fama, E., and K. French, 1997, Industry costs of capital, Journal of Financial Economics 43, 153-193. Garspar, J., M. Massa, and P. Matos, 2005, Shareholder investment horizon and the market for corporate control, Journal of Financial Economics 76, 135-165. Gibbons, R., and K.J. Murphy, 1992, Optimal incentive contracts in the presence of career concerns: Theory and evidence, Journal of Political Economy 100, 468–505. Graham, J.R. and K. Narasimhan, 2004, Corporate survival and managerial experiences during the Great Depression, Duke University working Paper. Hall, B.J., and K.J. Murphy, 2000, Optimal exercise prices for executive stock options, American Economic Review 90, 209–214. Hall, B.J., and K.J. Murphy, 2002, Stock options for undiversified executives, Journal of Accounting and Economics 33, 3–42. Hirshleifer, D., 1993, Managerial reputation and corporate investment decisions, Financial Management 22, 145-160. Jensen, M., 2005, Agency costs of overvalued equity, Financial Management 34, 5-19. Malmendier, U., and G. Tate, 2005, CEO overconfidence and corporate investment, Journal of Finance 60, 2661-2700. Malmendier, U., and G. Tate, 2008, Who makes acquisitions? CEO overconfidence and the market’s reaction, Journal of Financial Economics 89, 20-43. Martin, K.J., 1996, The method of payment in corporate acquisitions, investment opportunities, and management ownership, Journal of Finance 51, 1227-1246. Milbourn, T., 2003, CEO reputation and stock-based compensation, Journal of Financial Economics 68, 233-262. Moeller, S.B., F.P. Schlingemann, and R.M. Stulz, 2004, Firm size and the gains from acquisitions, Journal of Financial Economics 73, 201-228. Polk, C. and P. Sapienza, 2009, The stock market and corporate investment: A test of catering theory, Review of Financial Studies 22, 187-217. Rau, P.R., and T. Vermaelen, 1998, Glamour, value, and the post-acquisition performance of acquiring firms, Journal of Financial Economics 49, 223-253. 36  Rhodes-Kropf, M., D. Robinson, and S. Viswanathan, 2005, Valuation waves and merger activity: The empirical evidence, Journal of Financial Economics 77, 561-603. Shleifer, A., and R. Vishny, 1990, Equilibrium short horizons of investors and firms, American Economic Review 80, 148-153. Shleifer, A., and R. Vishny, 2003, Stock market driven acquisitions, Journal of Financial Economics 70, 295-311. Stein, J.C., 1996, Rational capital budgeting in an irrational world, Journal of Business 69, 429-455.  37  Chapter 3 Optimal Compensation Contracts When Managers Can Hedge7  3.1. Introduction Equity-based compensation is widely regarded as an effective way to align managers’ interests with those of their shareholders. Most of the literature on executive compensation is built on an essential assumption that managers cannot hedge their incentive portfolios. However, executives can certainly employ a number of financial instruments to hedge the risk in their compensation packages. Bettis et al. (2001) show that there has been a huge increase in the development and use of financial derivatives to enable corporate insiders to hedge stock ownership in their firms. These hedging transactions cover about 30% of executives’ firm-specific wealth (Jagolinzer et al. (2007)). Stulz (1984, p.139) explicitly states, “It would be interesting to show how the choice of the management compensation schemes depends on the opportunities managers have to hedge.” This paper analyzes the impact of managerial hedging on executive compensation from both a theoretical and empirical perspective. First, I extend Holmstrom and Milgrom’s (1987) principal-agent model by allowing a manager to costly hedge her incentive portfolio. The access to the hedging market increases the manager’s ability to bear risk, and it decreases her incentive to exert effort. In equilibrium, shareholders should provide a higher-power contract so that the manager’s after-hedging incentive is closer to the optimal level. The central empirical prediction of the model is that the pay-performance sensitivity 7  A version of this chapter will be submitted for publication. Gao, H., Optimal Compensation Contracts When Managers Can Hedge. 38  in compensation contracts decreases alongside the manager’s hedging cost. In the empirical tests, I use two variables to measure the executive hedging cost. The first one is a dummy indicating whether the firm has listed options on the option exchanges or not. When the firm has publicly tradable options, it is easier and less costly for managers to unwind their incentive pay through derivative markets (e.g., buying put options, or short selling call options). Regarding my second proxy, for firms with traded options I measure the hedging cost using the average daily trading volume of the firms’ options. Intuitively, a high volume indicates high liquidity and active trading of the firm’s derivatives, which lowers the cost of making derivative transactions. These two proxies capture the opportunity that managers have to hedge through the public option market. Using a large set of compensation data, I then provide empirical evidence supporting the model’s prediction. To further my understanding of how managerial hedging influences compensation contracts, I examine the sensitivity of Chief Executive Officer (CEO) wealth to stock return volatility which captures the convexity of the relation between CEO wealth and stock price. As suggested by prior literature (e.g., Smith and Stulz (1985)), the optimal convexity in CEO pay is determined by the benefit of inducing the CEO to take risky value-increasing projects and the cost of compensating her to bear risk. When managers can hedge more easily, the disutility imposed by the risk in their incentive pay becomes smaller; thus, the optimal convexity should be higher. Consistent with this view, I find that a CEO with lower hedging costs receives an incentive pay of greater sensitivity to stock return volatility. Given that stock-based pay is one of a few mechanisms that can be used to discipline 39  managers, another related question is whether shareholders use other governance policies to resolve managerial agency problems in response to executive hedging. Specifically, I deal with this question by investigating corporate capital structure. As a substitute for incentive contracts, existing studies suggest that debt can be used as a powerful tool to discipline managers. Shareholders are supposed to use more debt when it is easier for their managers to unwind incentive pay, because the incentive contract is less effective in this situation (Garvey (1997)). Consistent with this view, firms are found to have higher debt levels when their managers have better hedging opportunities, and that this relation is stronger for better-governed companies. This evidence also supports the implication that shareholders use other mechanisms (besides offering higher-power contracts), such as alterations to the capital structure, to overcome the executive hedging problem. Whether or not a manager can hedge clearly influences the way she deals with her personal portfolio. To understand this issue, I examine how executives rebalance their personal portfolios when hedging transactions are possible. Given the fact that a big portion of managers’ wealth is tied to their own firms, these under-diversified and risk-averse managers have strong incentives to diversify their portfolios by unwinding their equity holdings. Therefore, they should be eager to exercise their stock options when available and in the money. However, when managers can hedge their compensation portfolios to some extent, they will suffer less disutility from bearing risk, and will be less eager to exercise their vested options (Bettis et al. (2005), Carpenter (1998), and Hemmer et al. (1996)). In other words, managers are supposed to hold more exercisable in-the-money options when they have a low hedging cost. The empirical analysis supports this view. 40  A natural extension of my study is to examine how executive hedging influences managers’ decisions on corporate policies. The basic idea is quite intuitive: Managers who can hedge are less influenced by their incentive pay. In particular, I examine the corporate dividend payout in the presence of managerial hedging. Executive stock options induce managers to reduce corporate dividends because the payment of dividends, ceteris paribus, reduces the value of call option (Lambert et al. (1989)). However, if managers have hedged their incentive portfolios, they will not have that strong motivation to avoid dividends, simply because paying dividends has less of a negative effect on their personal wealth. Empirically, I provide evidence that executive hedging weakens the negative relation between stock option compensation and dividend payout. This result also supports a broader view that managerial hedging undermines the influence of incentive pay on management decision making. In addition, this paper examines the impacts of executive hedging on corporate diversification. To the extent that firm diversification is another way for managers to reduce risk, I argue that hedging a personal portfolio and making corporate diversification are substitutes for executives to decrease the idiosyncratic firm risk they face. Consistent with this argument, I find that managers diversify their companies less when it is less costly to hedge their incentive pay. I implement an extensive additional investigation to understand better the relationship between executive hedging and compensation contracts. I first find that CEOs in the firms with higher beta are associated with weaker pay-performance sensitivity. This result is consistent with the view that higher beta causes more difficulty for managers to hedge firm 41  systematic risk, and thus, leads to lower-power contracts (Garvey and Milbourn (2003)). I then single-out the CEOs working within the financial industry. In support of the idea that CEOs in the financial industry may have better chances to implement over-the-counter (OTC) hedging transactions, these CEOs have higher pay-performance sensitivity than CEOs in other industries. Finally, I collect a sample of CEOs who actually hedged and find that those CEOs receive higher incentive pay than other CEOs do. The article proceeds as follows. Section 3.2 introduces some background and related literature. Section 3.3 presents the model and provides the empirical prediction. Section 3.4 describes the data source and sample construction. Section 3.5 reports the empirical results. An additional investigation is conducted in Section 3.6 and Section 3.7 concludes.  3.2. Background on Executive Hedging Current legal system and managerial contracts play a very limited role in governing executive hedging transactions. While it is illegal for managers to short sell their own firms’ stock, it is legal for them to buy put options as long as the amount of securities underlying the put equivalent position does not exceed the amount of underlying securities otherwise owned.8 It is also illegal for insiders to trade derivatives based on material value-relevant information (insider trading). Although some hedging transactions of executives could be correlated with insider trading (Bettis et al. (2001)), this paper concentrates on the pure hedging purpose. As summarized by Schizer (2000), although existing executive contracts and security law have put some barriers up to managerial hedging, their gaps are still significant. Bebchuk et al. (2002) suggest that executives have freedom to access the  8  See Section 16 (c) of the Securities and Exchange Act of 1934 and Rule 16c-4. 42  financial market to hedge. As pointed out by Garvey (1997), the direct bans on management hedging are costly to enforce because the securities market is sufficiently rich and liquid that the manager’s participation in hedging cannot be perfectly controlled. Business practice has long witnessed the prevalence of executive hedging activities. Puri (1997) reports that a growing number of banks are marketing derivatives to help executives hedge. In the Wall Street Journal, Simon (2000, p.C1) states “It is impossible to precisely gauge the popularity of these hedges, but derivatives specialists suggest that hundreds, perhaps even a couple of thousand, are executed each year.” Schizer (2000) points out that the growing importance of equity-based compensation is accompanied by the simultaneous increase in the derivative instruments, which managers can use to hedge. Existing studies find three forms of off-market derivatives used by executives to hedge: zero-cost collars (collar), equity swaps, and prepaid variable forward contracts (PVFs). A collar involves the simultaneous purchase of a put option and sale of a call option covering the firm’s stock. The holders are protected from the downward movement in the stock price below the strike price of the put, while they forgo the profit from the stock price appreciation above the strike price of the call. An equity swap agreement enables managers to exchange the future returns on their stock for the cash flows of another financial instrument, such as LIBOR or the S&P 500 index. Based on 89 transactions on collars and equity swaps during 1996-1998, Bettis et al. (2001) show that these transactions involve high-ranking executives and effectively reduce their ownership by 25% on average. In a PVF agreement, the executive promises to sell the firm’s shares, usually to an investment bank, at some future date in exchange for an up-front cash advance. Similar to 43  zero-cost collars, a PVF provides holders protection against depreciation of the underlying stock price. However, the holders can still benefit from the price appreciation up to a pre-determined level. Jagolinzer et al. (2007) show that an average PVF covers about 30% of the executive’s personal holdings, corresponding to around $22 million. Both the academic and the practitioner have expressed great concern over the managerial hedging issue. The Economist (1999, p.64) states “Such hedging is wholly against the spirit of the massive awards of shares and share options.” Ofek and Yermack (2000) and Antle and Smith (1986) argue that the optimal contracting model should take the managers’ freedom to hedge away the risk in their compensation into account. Despite the significant literature on executive compensation, the understanding of managerial hedging is quite limited. Jin (2002) and Garvey and Milbourn (2003) study the case in which executives can trade market indices. Jin (2002) mainly addresses the effects of firm idiosyncratic and systematic risk on compensation contracts; the latter one focuses on justifying the rare use of relative performance evaluation. My paper complements their research by examining the case when managers can diversify their firm-specific risk exposure. More importantly, this article provides an extensive empirical analysis on the influence of managerial hedging on compensation contracts and corporate policies.9  3.3. The Model This section presents a standard principal-agent model in which the manager can hedge her equity holdings at a certain hedging cost. The structure of the model follows Holmstrom and Milgrom (1987).  9  Other related theoretical work includes Acharya and Bisin (2005), Bisin et al. (2006), and Ozerturk (2006). 44  One risk-averse manager works in a firm owned by risk-neutral investors. The manager’s utility function is given by: c U ( w, a ) = − exp[−η ( w − a 2 )] 2  (1)  where w is the manager’s total wage, a is her effort level, η is the coefficient of risk aversion, and c>0 is a constant reflecting the manager’s aversion to effort. The firm’s cash flow, p, equals the manager’s effort plus noise:  p = a +ε  (2)  where ε is normally distributed with zero mean and variance σ 2 . The manager gets a compensation package in the form:  t + sp  (3)  where t is the fixed pay level and s is the performance-based component of her compensation. Equivalently, t is the manager’s fixed salary and s is her share of the firm. Similar to the classical papers of Holmstrom and Milgrom (1987, 1991), I restrict my attention to a linear contract for algebra simplification. The optimality of the linear contract is based on the critical assumption of a constant absolute risk aversion utility function. Although a linear sharing rule might not be optimal with more general preference, it is still a good approximation to the executive compensation practice. Jin (2002, Footnote 6) states, “In practice, however, the sharing rule is often close to linear because the convexity induced by CEO option holdings is negligible to the first order.” The manager has the access to the hedging market, and can trade her shares at the hedging cost of  φ 2  x 2 , where x stands for the number of shares traded and φ > 0 is a  constant capturing her cost of hedging. This model allows a very general form of hedging 45  without being limited to one particular hedging instrument. Similar to Garvey and Milbourn (2003), I am assuming a strict convex function for the hedging cost, reflecting the reasonable assumption that it is costly for managers to take additional steps to either augment or offset their exposure to their own firms’ equity. If one interprets the hedging cost as the transaction expense during the asset trading, this convexity is also consistent with the evidence that the transaction cost in financial markets is a convex function of trading size (see, e.g., Breen et al. (2002), and Korajczyk and Sadka (2004)). Another interpretation on this hedging cost can be the probability for the manager to be caught in the hedging transactions. As the manager hedges more, there is more likelihood that she will be detected by shareholders and suffer some corresponding penalty. This hedging cost function incorporates a less reasonable assumption that long positions are as costly as short positions. This is primarily for notational convenience, as we shall see that the actual choices of x are always negative in equilibrium (i.e., managers always take a short position in the hedging transactions). This notational convenience of assuming symmetric costs for purchasing and selling is similar to Garvey and Milbourn (2003).10 The sequence of the game is described as follows: Stage 1. The shareholders optimally set the compensation rule (t, s) to maximize the net-of-wage firm value, taking into account the subsequent hedging behavior and the effort of the manager.  10  Alternatively, the hedging cost can be modeled as  φ1 2  2  x when x is the sales, and  ≠ φ2 ). The equilibrium results will be the same; but the algebra is more complicated. 46  φ2 2  2  x when x is the purchase ( φ1  Stage 2. The manager trades her shares in the hedging market, where the share price reflects the rational expectation about her subsequent effort level. Stage 3. The manager chooses her effort. Both the hedging transaction and effort level are chosen to maximize the manager’s own utility. Stage 4. The firm’s cash flow is realized and the manager consumes her wealth. The manager’s wealth after hedging is: w = t + ( s + x) p − xE[ p 0 ] −  φ 2  x2  (4)  where E[ p0 ] is the expected firm value at the hedging market. Since the hedging occurs before the effort is made, E[ p0 ] reflects the expected managerial effort level a 0 (x); in particular, E[ p 0 ] = a0 ( x) . The notation a 0 ( x) indicates that the hedging market rationally infers the manager’s subsequent effort based on the number of shares hedged. By rewriting the manager’s objective in terms of her certainty-equivalent wealth, the following reformulation is obtained: c η φ Maxa , x t + ( s + x)a − a 2 − ( s + x) 2 σ 2 − x 2 − xa 0 ( x) (5) 2 2 2  First-order conditions with respect to a and x, respectively, lead to:  ca* = s + x *  (6)  a * −η ( s + x*)σ 2 − φ x * −a 0 ( x) − x  da 0 ( x) =0 dx  (7)  The rational expectation condition implies: a 0 ( x) = a* =  s + x* c  (8)  Equations (6)-(8) result in the following solutions to the manager’s optimization problem:  a* =  φ + 1/ c s c(φ + ησ 2 + 1/ c)  (9)  47  x* =  −ησ 2 s φ + ησ 2 + 1/ c  (10)  Based on Equation (9), the effort-compensation sensitivity is  da * φ + 1/ c = and it ds c(φ + ησ 2 + 1/ c)  is increasing in φ , implying that managerial effort is more sensitive to her stock-based pay when the hedging is more costly. Equation (10) shows the intuitive result that the manager hedges more when the hedging cost φ is smaller, for a given level of s. When φ → +∞ , I obtain x* = 0 , suggesting that the manager will not hedge when it is too costly. It is worth noting that the company can also provide hedging instruments to the managers, like granting put options, so that managers in equilibrium may not need to hedge via their personal account. In this situation, the manager’s personal hedging transaction is determined by both the costs of the firm providing hedging and her personal hedging costs. But empirically, it is rarely observed that the company gives put-equivalent compensation to managers. When φ = 0 , I obtain x* = −  ησ 2 s . This result means that, even in the case of ησ 2 + 1/ c  zero hedging cost, the manager will not hedge all her exposure to the firm-specific risk. Instead, she will still hold a certain number of shares, which are proportional to the initial equity given by the shareholders. The intuition for this result is as follows. The firm’s stock has a positive risk premium in equilibrium, because the optimal managerial effort level is positive. When the hedging is costless, the manager is actually making a portfolio choice between a risk free asset (risk-free rate is normalized to be zero in the paper) and the firm’s stock. She will optimally hold a number of the firm’s shares so that the marginal benefit of a higher expected return equals the marginal cost of a higher volatility of the portfolio. This 48  intuition is similar to that of Jin (2002), who studies the case in which managers can freely undo the firm’s systematic risk by trading the market portfolio. He shows that managers in equilibrium will choose to keep a positive exposure to market risk as long as the market risk premium is positive. The shareholders maximize the net-of-wage firm value, which is: Maxa ,t , s , x E[ p − t − sp ] subject to c a, x ∈ arg max a , x E{− exp[ −η ( w − a 2 )]} 2  (IC)  c 2  E{− exp[ −η ( w − a 2 )]} ≥ − exp( −η w)  (IR)  where w denotes the manager’s reservation wage. Based on Equations (9) and (10), the principal problem can be rewritten as Maxa , x ,t , s (1 − s )a − t subject to a=  φ + 1/ c s c(φ + ησ 2 + 1/ c) c 2 η  x=  −ησ 2 s φ + ησ 2 + 1/ c  (11)  φ  t + ( s + x)a − a − ( s + x) 2 σ 2 − x 2 − xa 0 ( x) = w 2 2 2 x + s a 0 ( x) = c  Substituting for the value of t in the individual-rationality constraint and maximizing with respect to s and t, I obtain the following solutions for managerial effort, the manager’s after-hedging equity holding, and the optimal compensation policy:  a* =  1 φ + 1/ c 2 c (1 + cησ )(φ + 1/ c) + cη 2σ 4  s * + x* =  s* =  (12)  φ + 1/ c (1 + cησ )(φ + 1/ c) + cη 2σ 4  (13)  2  φ + ησ 2 + 1/ c (1 + cησ 2 )(φ + 1/ c) + cη 2σ 4  (14) 49  Notably, when φ → +∞ , s* =  1 . This is Holmstrom and Milgrom’s (1987) solution 1 + cησ 2  with no executive hedging. The analysis above yields the following result:  Proposition: The optimal pay-performance sensitivity s * is decreasing in the managerial hedging cost. The derivation of this proposition is straightforward. It is easy to show from Equation (14), that  ds * ησ 2 =− < 0 . The economic intuition is dφ [(1 + cησ 2 )(φ + 1/ c) + cη 2σ 4 ]2  as follows. A decrease in the hedging cost reduces the manager’s disutility of bearing risk since she is able to hold less after-hedging stock shares (Equation (13)). In other words, managerial hedging enhances the manager’s ability to bear risk. Given the fact that an optimal contract is always about the trade-off between incentive and risk, managers with high risk-bearing ability should be given a high-power contract. Another simple interpretation is that if managers can costly undo the incentive of a compensation contract, the contract will be made to provide more incentive to start with so that it is closer to what is optimal once managers undo part of the incentive. As we see, hedging transactions undermine the effectiveness of incentive contracting. This problem is therefore interesting only if direct bans on management hedging are difficult to enforce. The model effectively assumes that managers’ participation in the hedging market cannot be perfectly controlled by shareholders. The assumption is valid because managers’ trades in their own firms’ securities are verifiable only by testimony from parties who know the managers personally (Muelbroek (1992)). In practice, the manager’s personal portfolio is not publicly disclosed; it is difficult and costly for 50  shareholders to monitor. Current studies, such as Garvey (1997) and Ofek and Yermack (2000), also provide evidence in favor of this assumption.  3.4. Data Construction and Sample Selection 3.4.1. Measures of CEO Incentive I use two complementary variables to measure the CEO’s incentive pay.  Jensen and Murphy’s (1990) Pay-Performance Sensitivity (PPS). Following Jensen and Murphy (1990), PPS is the dollar value of a CEO’s wealth change relative to $1,000 change of shareholders’ value. Although managers can receive pay-performance incentives from a variety of sources, the vast majority of these are ownership of stock and stock options. Similar to Core and Guay (1999), I compute this sensitivity as the dollar value change of stock and options held by a CEO relative to $1,000 shareholder return. For stock, the PPS is simply the fraction of the firm that the executive owns. The PPS for options is the fraction of the firm’s stock on which the options are written multiplied by the options’ delta. I use the method developed by Core and Guay (2002) to estimate option deltas. They use the Black-Scholes option valuation model as modified by Merton (1973) to adjust dividend payouts. Their method can avoid the cost and difficulty of collecting option data from various proxy statements, since it requires information from only the most recent proxy statements. More important, they show that their estimates are effectively unbiased and 99% correlated with the measures that would be obtained if the parameters of a CEO’s option portfolio were completely known.  Core and Guay’s (1999) Portfolio Equity Incentive (PEI). Following Core and Guay (1999), PEI is defined as the change in the dollar value of the CEO’s stock and option 51  holding for 1% change in the stock price. As pointed out by them, the PEI is actually equal to the PPS multiplied by the firm’s market value of equity, and divided by $100,000. Although both the two variables measure how closely the CEO’s pay is related to shareholder wealth, they differ in the underlying assumptions about what drives managerial incentives. PPS measures the CEO’s wealth change relative to the dollar value change of shareholder wealth, under the assumption that incentives increase with a manager’s fractional ownership of the firm. PEI captures the CEO’s wealth change compared to the percentage change of shareholder wealth, by assuming that incentives increase with a manager’s dollar ownership of the firm (Core and Guay (1999)). As argued by Baker and Hall (1998), the more appropriate measure depends on how CEO actions are assumed to influence shareholder value. When CEO actions primarily affect firm dollar return (such as purchasing a corporate jet), the appropriate incentive measure is PPS. In contrast, when CEO actions primarily affect firm percentage returns (such as corporate reorganization and strategic redirection), PEI is the appropriate measure for executive incentives. This paper is not intended to contribute to the debate about measuring executive incentives; for robustness purposes, I use both measures.  3.4.2. Measures of Executive Hedging Cost The key explanatory variable is the cost of managerial hedging, which can be equivalently interpreted as the hedging opportunities managers have. I propose two proxies to measure these opportunities. The first is an Option dummy, which equals one if the firm’s option is traded in at least one of the six U.S. option exchanges, and zero otherwise.11 The economic 11 The six exchanges are the American Stock Exchange, Boston Options Exchange, Chicago Board Options Exchange, International Securities Exchange, Pacific Exchange, and Philadelphia Stock Exchange. 52  intuition behind this variable is as follows. When the firm’s option is publicly tradable, managers will have better opportunities to undo their equity holding in the derivatives market, which decreases the managerial hedging cost. In other words, the managerial hedging cost will be high when Option=0, and relatively low otherwise. Moreover, this Option dummy can be largely regarded as an exogenous variable to the company itself, to the extent that the decision to list a firm’s option is made by the option exchanges and not by the firm (Mayhew and Mihov (2004)).12 The second proxy is the firm’s option trading volume, which reflects the liquidity, activeness, and development of the firm’s option in the derivative market. Intuitively, a high volume indicates that it is relatively easy and convenient to trade the firm’s options.13 Managers may have a number of ways to hedge their firm-specific risk, such as using customized derivatives or trading the competitors’ stock. This paper mainly focuses on the possibility that managers hedge through the public option market.  3.4.3. Control Variables Besides the hedging cost, I also include a set of control variables that influence compensation policies as suggested by existing literature. These controls are as follows. CEO Age. Career concern is another factor influencing managerial behavior and firms’ compensation policies. Gibbons and Murphy (1992) provide theory and evidence showing that firms will use more equity-based compensation for older CEOs. Like them, CEO age is controlled to take account of the manager’s horizon problem. Moreover, a CEO’s age 12  Unlike the stock market, where firms apply to be listed, decisions to list options are made within the exchanges. Generally, stocks are selected for option listing by committees composed of members of the exchange after soliciting feedback from the general membership. 13 As a robustness check, I also normalized this trading volume by the firm’s shares outstanding. The results are qualitatively similar. 53  may also be associated with her reputation, personal wealth, and risk aversion. Firm Risk. Optimal contracting involves the trade-off between providing incentives and risk sharing between managers and shareholders, such that incentive level should decrease with firm risk. As a common approach, I measure the firm’s risk by using the stock return variance based on the firm’s monthly returns of the past five years. Firm Size. The cross-sectional level of a CEO’s incentive compensation changes predictably with firm size (see, e.g., Baker and Hall (1998)). To control for this size effect, I compute firm size as the natural logarithm of the firm’s market value of equity. Leverage Ratio. If managers have strong incentives to maximize shareholders’ value, debt holders will demand higher risk premiums for providing capital considering the problems of risk shifting. Based on this intuition, John and John’s (1993) model predicts a negative relation between leverage and pay-performance sensitivity. Therefore, the book-value ratio of long-term debt over total assets is included in the empirical study. Market-to-Book Ratio (M/B). As suggested by numerous studies (see, e.g., Yermack, 1995), when companies have large growth opportunities, shareholders have greater difficulty evaluating managers’ decisions, and thus, should provide managers with more stock-based compensation. I use M/B to control for the firm’s growth opportunity. Cash. Hall and Liebman (1998) suggest that scarcity of cash may lead firms to substitute cash payment with equity compensation. Therefore, availability of cash holding may be an important determinant in setting executive compensation. I measure Cash as the ratio of cash and short-term investment over the firm’s total assets. Moreover, companies tend to use more equity-based compensation when the firm 54  performance is high (e.g., Core and Guay (1999)). To account for firm performance, I also include return on equity (ROE) and the firm’s annual stock return as additional controls.  3.4.4. Data Source I collect stock returns from CRSP, compensation data from ExecuComp, accounting information from Compustat, and option trading data from OptionMetrics. OptionMetrics is a comprehensive source of historical price and implied volatility data for all U.S. exchange-listed equity options, starting from January 1996. All of the monetary variables are measured in 2000-constant dollars. To mitigate the effect of outliers, I winsorize all the continuous variables at the 1% level in both tails of the distribution. The final sample consists of 13,691 CEO-year observations from 1996 to 2005, 74% of which have options traded on the U.S. option exchanges (10,123 CEO-year observations). This big proportional number is not surprising since the companies in ExecuComp are usually the 1,500 biggest U.S. public firms.  3.5. Empirical Results 3.5.1. Summary Statistics Panel A of Table 3.1 reports the firm’s characteristics. The median firm is quite large; its market capitalization of equity is $1,339 million. The sample firms are performing well, with a median M/B ratio of 2.14, ROE of 12%, and annual stock return of 10.5%. Moreover, those firms are moderately levered with a median leverage ratio of 17%, and have sizeable cash holdings, with a median Cash ratio of 5%. For the firms with publicly tradable options, the mean and median daily option trading volumes are 990 and 139 contracts, respectively, indicating that this variable is highly skewed. The median CEO is 56 years old. 55  The variable PPS has a mean of $26 per $1,000 shareholder return and a median of $7; this number is similar to that reported by Hall and Liebman (1998).14 The median PEI is $112,000 change in CEO wealth for a 1% change in stock price, and this variable is substantially skewed with an average value of $389,000. The PEI value in my sample is a little smaller than that in Core and Guay (1999). The difference is consistent with the decreasing role of options in compensation after the downturn of the stock market in the early 2000s.15 As complementing measures for managerial incentives, PPS and PEI are positively correlated; their correlation coefficient is 0.46. The correlations between the explanatory variables are reported in Panel B. With the exception of the large positive correlations among Ln(Volume), Firmsize, and Firmsize2, all of the correlations are below 0.4 in magnitude. Both Option and Ln(Volume) are positively correlated with firm size, which is consistent with the finding that large firms are more likely to have listed options and large option trading volumes (Mayhew and Mihov (2004)). This fact also suggests that it is important to control for the size effect in the regressions.  3.5.2. Managerial Hedging Cost and Incentive Pay Existing literature on CEO compensation proposes the use of ordinary least squares (OLS), median and fixed effects regressions. I perform all of the three types and find qualitatively similar results. In particular, tests on the relation between the hedging cost and incentive pay are based on the following equation:  Incentiveit = a0 + a1Optionit or Ln(Volume)it + a2 Firmsizeit −1 + a3 Firmsize2it −1 +a4CDF (Variance)it −1 + a5M / Bit −1 + a6 ROEit −1 + a7 Leverageit −1 + a8 Ageit + a9Cashit −1 +a10 Stockreturnit −1 + IndustryDummy or Firm Fixed Effects + YearDummy + εit 14 15  The PPS reported in Hall and Liebman (1998) is $25 at the mean and $5.29 at the median. Core and Guay’s (1999) sample is from 1992 to 1996. 56  (3.1)  where i indexes firms and t indexes year. The dependent variable is the pay-performance sensitivity in a CEO’s compensation package, measured by PPS and PEI. Fama and French’s (1997) 48 industry dummies and year dummies are included to control for industry and time variation in executive pay schemes. Throughout the entire empirical test, p-values for the OLS regressions are computed based on robust standard errors clustered at the firm level. Estimating positive coefficients for a1 would be consistent with the prediction that the managerial hedging cost is negatively associated with incentive pay. Table 3.2 reports the regression results, using the Option dummy as the proxy for the hedging cost. The coefficients on Option are both economically and statistically significant in all six regressions. In Regression (1), the dependent variable in this OLS model is PPS. The coefficient of Option is about 3.2 and is significant at the 1% level. This result indicates that a change of Option from zero to one is associated with an increase in PPS by $3.2 per $1,000 shareholder wealth change, compared to the median PPS of $7. The coefficients of other control variables are generally consistent with existing empirical studies. In particular, PPS tends to be higher for firms of smaller size, higher growth potential, better accounting performance, lower leverage ratio, older CEOs, and less liquidity constraints. Table 3.1 shows clearly the right skewness of the compensation data. For this reason, the median as a measure of the center of a distribution is more robust than the mean. Therefore, following Aggarwal and Samwick (1999) and Jin (2002), I use median regression to estimate PPS in Column (2). Median regression minimizes the sum of absolute deviations rather than the sum of squared deviations, and can thus increase the 57  precision of estimating executive incentives (Aggarwal and Samwick (1999)). The corresponding p-values are computed according to bootstrapped standard errors based on 20 replications. In Column (2), all the independent variables have qualitatively similar coefficients to those in Column (1). The variable Option has a coefficient of 0.76 and it is significant at the 1% level. Not surprisingly, all the median regression estimates are well below the magnitude of the OLS estimates of PPS, because of the right skewness of the compensation data.16 Although plenty of controls are included in the regression, it is still possible that the proxy for hedging cost, Option, is correlated with some unobserved firm characteristics that affect CEO compensation. To address this issue, I add firm fixed effects in Column (3). It is important to note that the inclusion of firm fixed effects can control for any other aspects of the firm that influence the managerial compensation scheme. The results from the firm fixed effects regression further demonstrate the strong positive association between Option and PPS. The coefficient on Option is around 4 and the corresponding p-value is 0.013. The economic implication of this coefficient is that, for the same firm, an increase of the Option dummy from zero to one is associated with an increase of the PPS by about $4 per $1,000 shareholder return, relative to the median PPS of $7. In Columns (4)-(6), I replace PPS with Ln(PEI) as an alternative measure of the  16  Aggarwal and Samwick (1999) and Jin (2002) also find that the estimates in OLS are bigger in magnitude than those in  median regression. For example, Aggarwal and Samwick (1999) report that OLS estimates of PPS are more than twice those obtained from the median regression. 58  executive incentive and re-do the previous three regressions. 17 The main results are unchanged: Option dummy has a strongly positive relation with executive incentives measured by PEI. For example, in the fixed effects regression reported in Column (6), the coefficient of Option is 0.1 and is significant at the 1% level. This coefficient is also economically remarkable since an increase in Option from zero to one is associated with an approximate 10% increase in PEI. The coefficients on those controls are qualitatively consistent with the findings of Core and Guay (1999). The regression results also indicate that the levels of PEI are well-explained by the regression model outlined earlier. Taking Column (4) for example, the adjusted R2 is 47%, implying that the model explains a substantial proportion of the cross-sectional variation in PEI. Overall, the results in Panel A provide evidence supporting the prediction that the pay-performance sensitivity of management compensation is negatively correlated with the managerial hedging cost. In Panel B, I use Ln(Volume) as an alternative proxy for the executive hedging cost, and repeat the earlier regression analysis in Panel A. The sample used in this panel is a sub-sample, in which the firms have available listed options in the option exchanges. Consistent with the previous panel, Panel B further supports the prediction that the hedging cost is negatively associated with pay-performance sensitivity. The regression result in Column (1) highlights that the coefficient of Ln(Volume) is 3.57 and is significant at the 1% level. This coefficient is also economically meaningful: When Ln(Volume) increases by one standard deviation (1.93), the executive is awarded an 17  The reason that natural logarithm transform is not taken for PPS is to make the results easier to compare with those in  prior literature. 59  increased PPS of $6.9 (3.57×1.93) per $1,000 shareholder return. Therefore, this result is consistent with the prediction that a lower managerial hedging cost leads to higher stock-based pay sensitivities of CEOs. In the median regression (Column (2)) and fixed effects regression (Column (3)), the coefficients of Option are 1.14 and 6.77, respectively; both of them are significant at the 1% level. Columns (4)-(6) in Panel B show that Ln(Volume) also has a significant positive relation with PEI. The coefficients on Ln(Volume) are 0.11, 0.12 and 0.13, respectively, implying that a one-standard-deviation increase in Ln(Volume) predicts an increase in PEI of about 23% (12%×1.93). Other controls have very similar coefficients to those in Panel A. Since the firms listed in the option exchanges and the ones with large option trading volumes are usually large, the Option and Ln(Volume) variables might just capture the firm size effect rather than the hedging cost effect. This concern can be ruled out for two reasons. First, Firmsize and Firmsize2 are controlled in the regressions. Therefore, the positive coefficients on Option and Ln(Volume) indicate that firms with publicly tradable options or large option trading volumes provide higher CEO incentive than other firms of similar size. I have also used sales volume and total assets to measure firm size instead of market value of equity; the coefficients on the two hedging cost proxies are quite robust. Since Firmsize and Firmsize2 are highly correlated, I have also tried to just include either of them in the regressions, and the results on Option and Ln(Volume) are largely unchanged. Second, the relation between firm size and CEO incentives depends on the measures of incentive pay. I find that firm size is negatively related to PPS but positively related to PEI; this relation is consistent with prior literature, such as Core and Guay (1999) and 60  Baker and Hall (1998). However, the coefficients on Option and Ln(Volume) are always significantly positive regardless of whether incentive pay is measured by PPS or PEI. This fact further supports the notion that the hedging cost proxies are not capturing firm size effects. Another concern with my hedging cost proxies is that they might primarily reflect firm risk, as riskier firms are more likely to be listed in option exchanges and to have larger option trading volumes (Mayhew and Mihow (2004)). This concern is not valid for the following reasons. First, the regressions have accounted for stock return volatility; that is, the effect of firm risk on incentive pay has been controlled. Second, standard principal-agent theory predicts a negative relation between firm risk and managerial incentive levels (Holmstrom and Milgrom (1987)), which implies that the coefficients on Option and Ln(Volume) would be negative if they captured firm risk. Third, as summarized by Prendergast (2002), empirical studies have failed to find any robust association between risk and executive incentive, further indicating that the positive relation between hedging cost proxies and incentive compensation is not due to the risk effect on compensation. It is possible that the hedging cost proxies reflect the financial sophistication of directors, as firms with complicated derivatives may hire finance experts as their board of directors. This possibility can be ruled out by the study of Guner et al. (2008), who study the influence of directors with financial expertise on corporate decisions and find that those directors have little influence on CEO compensation scheme. The conclusion from Table 3.2 is clear: CEOs are receiving higher-power compensation contracts when it is less costly for them to hedge their incentive portfolios. This evidence 61  supports the proposed theoretical model.  3.5.3. Managerial Hedging Cost and Convexity in Incentive Pay My earlier analysis on PPS and PEI is concerned with the slope of the relation between the CEO’s wealth and stock price. Although managing slope is important in setting CEO pay, another important aspect is the convexity in the compensation package. As documented by existing literature (Jensen and Meckling (1976) and Smith and Stulz (1985)), the convexity of the relation between stock price and CEO wealth, in addition to the slope, has to be properly designed to induce executives to make optimal corporate decisions. This convexity refers to the sensitivity of executives’ wealth to the volatility of stock return. As shown by Smith and Stulz (1985), risk-averse managers are likely to forgo risk-increasing but positive net-present-value (NPV) projects, and this risk-related agency problem can be resolved by using stock options to construct a convex relation between executive wealth and firm performance. Following the framework of Holmstrom and Milgrom (1987) and Guay (1999), the optimal convexity in CEO incentive pay is determined by the benefits of risky positive NPV projects and the cost of compensating the manager for bearing the risk. When shareholders increase the sensitivity of CEO wealth to firm risk, CEOs are less likely to pass up those risky but value-increasing investments. However, shareholders also need to increase the level of total pay to compensate those risk-averse managers for taking the risk. In equilibrium, the optimal convexity in CEO pay should decrease in the managerial aversion to risk (Coles et al. (2006) and Guay (1999)). This risk aversion effect depends on 62  the degree of diversification of the manager’s wealth portfolio and her utility function. All else being equal, good hedging opportunities enable managers to diversify risk and to be less vulnerable to stock price volatility. For this reason, the optimal sensitivity of CEO wealth to firm risk should be higher when managers can hedge more easily. Similar to Guay (1999), I define Vega as the change in dollar value of the executive’s wealth for a 0.01 change in the annualized standard deviation of stock return. Following Coles et al. (2006), Vega of the option portfolio is used to measure the total Vega of executives’ total equity portfolios because option Vega is many times higher than stock Vega. In my sample, the mean and median Vega are $112.9 thousand and $44.5 thousand, respectively. I regress Ln(Vega) on the hedging cost proxies, controlling for potential confounding variables. The regression model is specified below:  Ln(Vega)it = b0 + b1Optionit or Ln(Volume)it + b2 Firmsizeit −1 + b3 Firmsize2it −1 +b4CDF (Variance)it −1 + b5 M / Bit −1 + b6 ROEit −1 + b7 Leverageit −1 + b8 Ageit + b9Cashit −1 +b10 Stockreturnit −1 + IndustryDummy or Firm Fixed Effects + YearDummy + εit  (3.2)  Estimating positive coefficients of b1 would be consistent with the prediction that the sensitivity of CEO wealth to firm risk is higher when the CEO has better opportunities to hedge her personal portfolio. Table 3.3 reports a positive relation between Ln(Vega) and the hedging cost proxies; the relation is both statistically and economically significant. Taking Column (1) for example, I use the Option variable as the proxy for the hedging cost and run a pooled OLS regression. The coefficient on Option is 0.17 and is significant at the 1% level. This result indicates that a zero-to-one increase in Option is associated with an approximate 17% increase in Vega. Taking Column (6) as another example, I use Ln(Volume) to proxy for the 63  hedging costs in the firm fixed effects regression; the coefficient on Ln(Volume) is 0.07 and is statistically significant at the 1% level. The result is quite economically important; Vega will increase by about 13.5% when Ln(Volume) increases by one standard deviation. As suggested by Coles et al. (2006), the slope and convexity in the CEO’s compensation contract are jointly determined. In other words, shareholders choose a combination of the slope and the convexity to solve the compensation problem optimally. In this case, using simultaneous equations to estimate PPS/PEI and Vega jointly could be a more appropriate approach than estimating them separately. Following Coles et al. (2006), I use three-stage least squares (3SLS) to estimate the following simultaneous equations: Ln(Vega)it = γ 0 + γ1Ln(PEI )it + γ 2Optionit or Ln(Volume)it + γ 3 Firmsizeit −1 +γ 4 Firmsize2it −1 + γ 5CDF (Variance)it −1 + γ 6M / Bit −1 + γ 7 ROEit −1 + γ 8 Leverageit −1 +γ 9Cashit −1 + γ10 Stockreturnit −1 + γ11Ln(Cash Compensation)it + IndustryDummy  (3.3a)  +YearDummy + εit  Ln(PEI )it = δ0 + δ1Ln(Vega)it + δ2Optionit or Ln(Volume)it + δ3Firmsizeit −1 +δ4 Firmsize2it −1 + δ5CDF(Variance)it −1 + δ6M / Bit −1 + δ7 ROEit −1 + δ8 Leverageit −1 (3.3b) +δ9 Ageit + δ10Cashit −1 + δ11Stockreturnit −1 + IndustryDummy + YearDummy + εit Table 3.4 contains the system’s specifications using two hedging cost proxies. In each case, the jointly determined variables are Vega and PEI. The independent variables are generally drawn from the prior literature (Coles et al. (2006) and Guay (1999)). The variable Cash Compensation is the dollar value of the CEO’s cash pay. The coefficients on Option and Ln(Volume) are positive and both economically and statistically significant. This result supports the fact that shareholders increase both the slope and convexity in the CEO’s incentive pay when the CEO has good hedging opportunities. In Columns (1) and (2), the coefficients on Option are about 0.16 and 0.06, respectively, and both are significant at the 5% level. Controlling for other factors, a firm 64  with publicly-tradable options provides about 16% higher Vega and 6% higher PEI than a company without such options. I use Ln(Volume) as an alternative proxy for hedging costs in Columns (3) and (4). The corresponding coefficients are 0.16 and 0.12, respectively, and both are significant at the 1% level. A one-standard-deviation increase in Ln(Volume) is associated with a 31% increase in Vega and a 23% increase in PEI. Other control variables are generally consistent with the results in Coles et al. (2006). I also replace Ln(PEI) with PPS in the systems and find that the coefficients on Option and Ln(Volume) do not change qualitatively. The simultaneous equations results for PPS are omitted here for brevity. In summary, the evidence in Tables 3 and 4 supports the view that higher executive hedging costs are associated with not only lower pay-performance sensitivity but also lower sensitivity of CEO wealth to equity volatility.  3.5.4. Managerial Hedging Cost and Capital Structure Given the fact that managerial hedging undermines the efficacy of incentive contracts, how will shareholders use other mechanisms to resolve the executive incentive problem? This section addresses this problem by examining firms’ capital structure decisions. Agency theory suggests that debt mitigates the shareholder-manager agency problem by inducing lenders to monitor, reducing the free cash flow available to managers, and forcing them to maximize value when facing the threat of bankruptcy (Jensen (1986) and Stulz (1990)). For this reason, Ortiz-Molina (2007) suggests that high leverage and high-power incentive contracts can be substitutes. In a theoretical model, in which shareholders set both capital structure and compensation policy to discipline managers, Garvey (1997) shows that 65  debt is important in aligning shareholder-manager interests, especially when managers can unload their incentive contracts in a liquid secondary market. In one word, the literature on capital structure and agency problem implies that firms should experience higher debt level when their managers have better opportunities to hedge incentive pay.18 To examine this prediction empirically, I run pooled OLS regressions using the model below:  Leverageit = c0 + c1Optionit or Ln(Volume)it + c2 Democracyit +  c3Democracyit × (Optionit or Ln(Volume)it ) + c4 Firmsizeit −1 + c5Stockreturnit −1 +c6CDF (Variance)it −1 + c7 M / Bit −1 + c8 ROEit −1 + c9Tangibilityit −1  +c10 R & Dit −1 + c11 Advertisingit −1 + IndustryDummy + YearDummy + εit  (3.4)  Here, the dependent variable is the firm’s leverage ratio, following the same definition as Leverage defined in Section 3.4.3. Democracy takes the value of one if the firm’s Gompers et al.’s (2003) G-index value is less than or equal to five, and zero otherwise. Gompers et al. (2003) construct the G-index to measure governance from the perspective of firm-level anti-takeover protection. They show that better governed firms (which they call firms with “Democracy”) have a higher firm value and better performance. The variable Tangibility is the ratio of the firm’s fixed assets over total assets; R&D is the ratio of research and development (R&D) expenses to sales; Advertising is the ratio of advertising expenses over sales. Since I no longer require data availability in ExecuComp, the sample size for analyzing capital structure increases to 59,381 firm-year observations from 1996 to 2005. Among them, 17,638 observations have traded options. Available studies agree that capital structures are influenced by factors, such as firm size, asset tangibility, growth opportunities, advertising expenditure, R&D expenditure, volatility and profitability (Harris and Raviv 18  Although it could be the managers who make the capital structure decisions, the board and shareholders do have strong  influence on the financing decision (see, e.g., Klein (1998) and Guner et al. (2008)). 66  (1991)). Estimating positive coefficients of c1 would support the prediction that financial leverage is negatively associated with the executive hedging cost. Table 3.5 highlights the hedging cost as a strong determinant for capital structure decisions. In Column (1), I regress book leverage on the Option dummy, as well as control variables. The coefficient on Option is 2.19 and is significant at the 1% level. The zero-to-one increase in Option is associated with an increase in book leverage by about 2.2 percentage points, relative to the sample median of 9%. Option is then replaced with Ln(Volume) in Column (2). The coefficient of Ln(Volume) is 0.43 and is significant at the 1% level. Again, the economic impact is sizeable; book leverage will increase by about 0.81 percentage points when Ln(Volume) increases by one standard deviation. Next, I interact Democracy with Option and Ln(Volume) in Columns (3) and (4). Both interaction terms have positive and significant coefficients, indicating that better-governed firms are more likely to increase financial leverage in response to managerial hedging. Among the control variables, the coefficients of M/B are persistently negative across all of the four regressions, indicating high-growth firms use less debt. The variable Tangibility is significantly positively associated with the leverage level. The regression analysis in Table 3.5 supports the prediction that shareholders tend to use more debt when managers have better opportunities to hedge. This result also implies that, besides providing higher-power contracts, shareholders simultaneously use other mechanisms, such as capital structure, to resolve the shareholder-manager agency problem as responding to executive hedging.  67  3.5.5. Managerial Hedging Cost and Option Exercising/Holding Behavior To further support that the two hedging cost proxies do influence CEOs’ hedging behavior, I investigate executive option exercising/holding behavior in this section. CEOs usually receive large grants of stock and options of their own firms as compensation, and in the mean time, their human capital is also intimately linked to the firm performance. As they are usually prevented from unwinding their equity ownership, these under-diversified and risk-averse CEOs should be eager to exercise their in-the-money options when the vesting period expires (Hall and Murphy (2002)). As argued by Malmendier and Tate (2005), those CEOs should minimize their holdings of company equity to divest themselves of idiosyncratic risk. However, ceteris paribus, when CEOs can hedge their equity positions to a certain extent, they will be less eager to exercise their vested options, and will hold more exercisable in-the-money options, simply because the firm-specific risk can be diversified away through the hedging instruments. Consistent with this argument, Hemmer et al. (1996, p.49) state “Hedged managers do not bear the risk of holding employee stock options (ESOs) and therefore have no incentive to exercise their ESOs early to diversify their portfolio.” Other studies, such as Bettis et al. (2005) and Carpenter (1998), also suggest a positive relation between the strength of the hedging and the holding of exercisable in-the-money options.19 To test this prediction, I use three variables to measure a CEO’s ownership of exercisable 19  There are two reasons why CEOs should hedge and keep exercisable options instead of exercising them directly.  First, direct option exercising would send negative signals to investors. If managers fear this signal, they may retain their options while taking some hedging positions. Second, the market value of a “live” option usually exceeds the proceeds from exercise; managers face the tradeoff between diversification benefits and cost of early exercise. When managers can use hedging instruments, such as put options, to protect themselves from stock price downturn, they will keep their exercisable options alive. 68  in-the-money options. They are: (1) dollar value of the CEO’s exercisable in-the-money options (Opt1); (2) Opt1 as a percentage of the CEO’s total annual compensation (Opt2); and (3) the number of common shares underlying the CEO’s exercisable in-the-money options as a percentage of the firm’s shares already owned by the CEO (Opt3). Pooled OLS regressions are run using the following model:  Exercisable In − the − Money Optionit = d0 + d1Optionit or Ln(Volume)it + d2 Firmsizeit −1 +d3CDF (Variance)it −1 + d4Ownershipit −1 + d5M / Bit −1 + d6 Ageit + d7 Stockreturnit + IndustryDummy + YearDummy + ε it  (3.5)  Existing studies suggest a few variables that influence managers’ option exercising behavior, including firm size, managerial ownership, stock volatility, growth opportunities, managerial risk-aversion, and recent stock movement (Huddart and Lang (1996) and Ofek and Yermack (2000)). The Ownership variable measures the percentage of the firm’s shares owned by the CEO. Based on the implication that a manager with lower hedging costs will hold more vested in-the-money options, the coefficient d1 is expected to be positive. Columns (1)-(3) of Table 3.6 highlight a positive relationship between the Option dummy and CEOs’ exercisable in-the-money options. The dependent variable in Model (1) is Ln(1+Opt1), and the coefficient on Option is 0.4 with the p-value less than 0.001. The economic magnitude is quite large; as Option increases from zero to one, the dollar value of CEOs’ vested in-the-money options will increase by 40%. I then normalize the dollar value of options by the CEOs’ total annual compensation, and use Ln(1+Opt2) as the left-hand variable in Column (2). The variable Option has a coefficient of 0.22, which is significant at the 1% level. To examine the robustness of the results further, I employ the shares of vested in-the-money options instead of dollar values. The predicted variable in Model (3) is 69  Ln(1+Opt3); the coefficient on Option is 0.095 and is significant at the 5% level, indicating that a zero-to-one increase of Option is associated with a 9.5% increase of Opt3. Regressions (4)-(6) provide further results supporting the expected relation between executive option holding and the hedging cost, using Ln(Volume) as the proxy. All of the three regressions (except Model (5)) highlight a significantly positive relation between Ln(Volume) and executives’ holding of exercisable in-the-money options. The conclusion from Table 3.6 is quite clear. Managers are less eager to unwind their equity portfolios when they can use hedging instruments more easily. The finding also supports the claim that the two variables, Option and Ln(Volume), capture the effect of managerial hedging cost and influence managers’ hedging behavior.  3.5.6. Managerial Hedging Cost and Corporate Dividend Policy In an earlier analysis in this paper, I mainly focus on how shareholders design compensation in response to executive hedging. A natural question for extension is how the hedging influences managers’ decisions on corporate policies. In this section, I address this question by examining the effect of executive hedging on corporate dividend payments.20 Executive stock options furnish management with the incentive to reduce dividends because the value of executive stock options, like all call options, are negatively related to future dividend payments. Consistent with this hypothesis, Lambert et al. (1989) and Fenn and Liang (2001) find a strongly negative relationship between dividends and management stock options. However, in the presence of managerial hedging, managers will not have such a strong incentive to cut dividends, simply because paying dividends will have less of a 20 The influence of managerial hedging on other corporate decisions, such as investment and financing decisions, could be an interesting topic for future research. 70  negative effect on their personal wealth.21 To test this view, I run pooled OLS regressions using the following model: Dividendit +1 = e0 + e1Optionit or Ln(Volume)it + e2Optionpayit or Ln(Optionwealth)it +  e3 (Optionit or Ln(Volume)it ) × (Optionpayit or Ln(Optionwealth)it ) + e4 Firmsizeit +  e5CDF (Variance)it + e6 M / Bit + e7 ROEit + e8 Leverageit + e9Cashit + IndustryDummy +YearDummy + ε it +1  (3.6)  Here, the dependent variable is the firm’s dividend payment (Compustat Item 21) normalized by stock market capitalization. The variable Optionpay is the value of the CEO’s annual option grants as a proportion of her total annual compensation; Optionwealth is the Black-Scholes value of the CEO’s total stock options. Obviously, Optionpay is a flow variable, and Optionwealth is a level variable. I use these two variables to measure the relative importance of stock options for a CEO’s wealth. I expect the e2 coefficient to be negative, and the e3 coefficient to be positive. Since my compensation data is from 1996 to 2005, the corresponding dividend data is from 1997 to 2006. The sample in this regression consists of 17,036 firm-year observations; 11,247 observations have publicly traded options. Table 3.7 reports the results of estimating the above equation, in which I regress corporate dividend payments on the control variables plus CEOs’ option holding, as well as the latter’s interaction with the hedging cost proxies. In Column (1), I use the Option dummy to measure the hedging cost and Optionpay for the CEO’s option pay. The coefficient of Optionpay is significantly negative. The interaction term Option×Optionpay has a significantly positive coefficient. I replace Optionpay with Ln(Optionwealth) in Column (2). Similar to Column (1), I find a  negative  coefficient  on  Ln(Optionwealth),  and  a  positive  coefficient  21 Here I assume that a risk-averse manager first hedge part of her incentive portfolio and then deal with the dividend dilution impact on options. 71  on  Option×Ln(Optionwealth). Both of the coefficients are significant at the 1% level. Furthermore, I substitute Option with Ln(Volume) in Columns (3) and (4), repeat the previous two regressions, and find qualitatively similar results. The interactions, Ln(Volume)×Optionpay and Ln(Volume)×Ln(Optionwealth), are positive; both Optionpay and Ln(Optionwealth) have significantly negative coefficients. Those coefficients of the interaction terms are also economically significant. Taking Model (1) for example, Optionpay and Option×Optionpay have the coefficients of -1.22 and 0.71, respectively. The interpretation of this result is as follows. When Option=0, the partial effect of Optionpay on dividend payment is -1.22; when Option=1, the partial effect of Optionpay is reduced to -0.51 (-1.22+0.71=-0.51). Table 3.7 supports the prediction that managerial hedging weakens the negative relation between management option pay and corporate dividends. The result is also consistent with a broader idea that managers who can hedge are less influenced by their incentive pay.  3.5.7. Managerial Hedging Cost and Corporate Diversification In addition to hedging personal incentive portfolios, another way for managers to hedge risk is to diversify their companies. This section addresses the natural question of how managerial hedging costs influence corporate diversification decisions. Existing literature finds risk reduction as a strong motive for corporate diversification (Amihud and Lev (1981) and May (1995)). To the extent that diversification decreases firm risk, managers facing higher idiosyncratic risk tend to diversify the companies more. Given that managerial hedging enables the managers to unwind their wealth from firm risk, I expect that hedging personal portfolios and diversifying the firms are substitutes for 72  managers to reduce risk. In other words, when executives can hedge more easily, they will execute fewer corporate diversification initiatives. To test this prediction, I run pooled OLS regressions estimating the following model:  Corporatediversificationit = f 0 + f1Optionit or Ln(Volume)it + f 2 Firmsizeit −1  + f3 Stockreturnit −1 + f 4CDF (Variance)it −1 + f 5 M / Bit −1 + f 6 ROEit −1 + f 7Cashit −1 + IndustryDummy + YearDummy + ε it  (3.7)  I use two measures of corporate diversification: the Herfindahl Index of the concentration of sales across the various business segments and the number of reported business segments. A more diversified firm is represented by a lower value of Herfindahl Index and more segments. The regression sample consists of 52,472 firm-year observations from 1996 to 2005; 16,369 observations have options traded in the option exchanges. The means of Herfindahl Index and segment number are 0.83 and 2.01, respectively. To support the prediction that corporate diversification is positively associated with managerial hedging costs, I expect positive (negative) coefficients for f1 when the Herfindahl Index (segment number) is the dependent variable. Table 3.8 highlights the significantly positive relation between the degree of corporate diversification and executive hedging costs. In Columns (1) and (2), the dependent variable is the Herfindahl Index. The coefficients on Option and Ln(Volume) are 0.017 and 0.004, respectively; both are significant at the 10% level. These positive coefficients indicate that lower hedging costs are associated with a higher Herfindahl Index (less firm diversification). I then replace the Herfindahl Index with the natural logarithm of segment number as the dependent variable in Columns (3) and (4). The results are qualitatively similar. Other controls are generally consistent with those in Coles et al. (2006). In summary, the 73  regression analysis supports the prediction that when managers can hedge their incentive pay more easily, they initiate fewer diversification projects with their companies.  3.6. Additional Investigation 3.6.1. Firm Beta and Incentive Pay While this paper focuses on Option and Ln(Volume) as the key variables to measure hedging costs, there are also other proxies. Higher firm beta could be associated with higher hedging costs, because managers need to short sell more market portfolios to diversify the firm’s systematic risk when the firm beta is higher, ceteris paribus.22 For example, to completely diversify the market risk for a certain amount of firms’ stock, managers in the firm with the beta of two need to short sell about twice as many market portfolios as the managers whose firm beta is one. A similar argument can be found in Garvey and Milbourn (2003), whose model predicts an inverse relation between pay-performance sensitivity and firm beta under the assumption that it is costly for managers to short sell market indices. I re-do the regressions specified in Panel A of Table 3.2 by replacing Option with the firm beta, and the beta for each firm-year observation is estimated based on a simple OLS regression of the firm’s returns on returns of the CRSP value-weighted index using the preceding five years of monthly data. Consistent with the above prediction, Panel A of Table 3.9 highlights a reverse relation between firm beta and incentive pay. The coefficients of beta in all six regressions are significantly negative. Taking Column (1) for example, the coefficient in front of beta is -9.03 and it is significant at the 1% level, indicating that a one-standard-deviation increase in 22 Here, I assume a positive firm beta, which is usually true for ExecuComp firms. A similar argument can be made even if the beta is negative. 74  beta (0.72) is associated with a reduction of PPS by $6.5 per $1,000 shareholder return. An endogeneity problem may arise when simply using firm beta as the proxy for the hedging cost, because CEOs can potentially control the level of their firms’ risk. As suggested by Jin (2002), the industry average risk, which can hardly be manipulated by the CEOs, is a more robust measure of the firm’s environmental risk level. To compute the industry beta of Fama and French’s (1997) 48 industries in each sample year, I run OLS regressions of the returns of value-weighted industry portfolio on the returns of the CRSP value-weighted index using the preceding five years of monthly return. I then assign the industry beta to the firm-year observations within the same industry, and re-do the regression in Panel A of Table 3.9 by replacing the firm beta with the firm’s industry beta. Consistent with Panel A, the regression results in Panel B show that the firm’s industry beta is negatively associated with incentive pay. In short, firms with higher beta (or industry beta) impose lower-power contracts to their CEOs. The result supports the view that high beta reflects large hedging costs and, consequently, leads to low pay-performance sensitivity.  3.6.2. Proxy for the Ease of OTC Hedging Transactions Hedging through the OTC market is also an important channel for managers to diversify firms’ risk. While existing literature has indentified few variables indicating the ease of implementing OTC transactions (to my knowledge), CEOs in the financial industry probably have better access to these off-market derivatives. Hedging by OTC transactions usually involves private negotiation between corporate executives and some financial institutions (Bettis et al. (2001) and Jagolinzer et al. (2007)). Therefore, a good social 75  network with these financial organizations should facilitate the executives during the negotiation process. A CEO working within the financial industry presumably has a better personal relation with financial companies than CEOs outside the industry, which gives her better opportunities to implement the OTC hedging operations. Moreover, the CEO in the financial industry may understand the off-market derivatives better than CEOs working in other sectors; this can also increase her comfort in the OTC market. Based on this idea, I define an indicator variable, Financial Industry, equaling one if the firm is in banking, insurance, or trading industries23, and zero otherwise. In my sample of 13,691 firm-year observations used in Panel A of Table 3.2, 1,780 (13%) of them are in the financial industry. I then re-do the regressions in Panel A of Table 3.2 by replacing Option with Financial Industry (industry dummies and firm fixed effects are also excluded). Given that the Financial Industry dummy is supposed to indicate the ease of implementing OTC hedging transactions, its coefficients are expected to be positive. Table 3.10 reports that CEOs in the financial industry have greater pay-performance sensitivity than other CEOs, after controlling for the CEO and firm characteristics. The coefficients of Financial Industry are all positive and significant (except Model (1)). Taking Model (3) for example, Financial Industry has a significantly positive coefficient of 0.41, indicating that CEOs in the financial industry have 41% higher PEI than other CEOs. The result supports the argument that it is easier for CEOs in the financial industry to conduct the OTC hedging transactions; therefore, shareholders impose higher-power contracts.  23  The corresponding classifications of Fama and French’s (1997) 48 industries are 44, 45, and 47, respectively. 76  3.6.3. Direct Executive Hedging Transactions and Incentive Pay Starting in 1996, Primark/Disclosure (now part of Thomson Reuters) began to collect all the hedging transactions recorded in Table II of Forms 3, 4, and 5. Similar to Bettis et al. (2001) and Jagolinzer et al. (2007), I search in the insider trading database of Thomson Reuters the security codes “EQSWP”, “FWD” and “PUT”, which correspond to the executive transactions of equity swaps, forward sales and put options, respectively. During 1996-2005, I identify 75 ExecuComp CEO-year observations that are involved in the above transactions. To flag the hedged CEO, a dummy variable, Hedged, is constructed, which equals one if the CEO-year observation is one of those 75 observations, and zero otherwise. I re-do the regressions specified in Table 3.2, Panel A by replacing Option with Hedged. It is worth noting that some CEOs could have used hedging instruments but did not report them (Bettis et al. (2001)). This possibility will cause biases against me to detect differences in incentive pay between hedged CEOs and others. In other words, the detected differences will become more evident if I analyze a complete sample of CEOs who used hedging instruments. Panel A of Table 3.11 shows that CEOs who hedged experience significantly higher pay-performance sensitivity. In all the four columns, the coefficients of Hedged are positive and both statistically and economically significant. Taking Model (2) for example, the Hedged variable has a coefficient of 27.26 and the coefficient is significant at the 1% level, implying that the CEOs who did hedging transactions have about $27 higher wealth change per $1,000 shareholder return. Instead of the full ExecuComp population, I further use the matched-pairs sample for 77  the regression analysis in Panel B. Using a matched-pairs sample produces more efficient parameter estimates than using the full ExecuComp population would, given the very small fraction of hedged CEOs in the population (Manski and Lerman (1977)). Similar to Bettis et al. (2001), for each of the 75 hedging observations, I identify a comparison firm in the same Fama and French (1997) 48 industries with the closest market value of equity in the preceding year of the transaction. The regression results based on this match sample in Panel B show the same result: The hedged CEOs have higher pay-performance sensitivity than their comparison group. In summary, based on a sample of direct hedging transactions, I find supporting evidence that shareholders impose higher-power contracts to CEOs who can hedge.  3.7. Conclusion This paper examines the optimal executive compensation with respect to managerial hedging. The driving force behind my theoretical analysis is the notion that an executive’s actions are influenced by the cost for her to access hedging instruments. I extend previous research by showing that the hedging cost has important effects on the manager’s effort-exerting incentive and risk-bearing ability. My model predicts a negative association between pay-performance sensitivity and the managerial hedging cost. I then provide empirical evidence to support the model’s prediction. Two variables are primarily employed to measure the hedging cost. The first measure is a dummy variable indicating the availability of the firm’s options on option exchanges. I then use the firm’s option trading volume as the second proxy. In a straightforward manner, these two variables capture the ease with which one can trade the firm’s derivatives to hedge 78  idiosyncratic risk. Equivalently, the two proxies reflect the opportunities that managers have to make the hedging transactions. In addition to examining the pay-performance sensitivity, I also investigate the impact of managerial hedging on the sensitivity of CEO wealth to stock volatility. The findings support the view that shareholders increase the convexity of the relation between CEO wealth and stock return, along with increasing the slope, when managers can hedge. To deepen the understanding of the managerial hedging problem, I then examine whether shareholders use other mechanism to resolve this hedging issue, in addition to offering high-power contracts. Particularly, I address this question by investigating the capital structure decision. As a substitute for incentive pay, debt is widely suggested by available studies as a powerful way to align shareholder-manager interests. When executive hedging undermines the effectiveness of incentive compensation, shareholders are expected to increase financial leverage as an alternative way to restore executive incentive. Consistent with this argument, I document evidence that firms exhibit higher leverage ratios when it is easier for their managers to unwind the incentive contracts. This relation is found to be stronger for better-governed companies. To validate that the two proxies measure the hedging cost and influence managers’ personal trading, I further analyze executives’ option exercising/holding behavior. Existing studies suggest that managers will hold more exercisable in-the-money options when they can diversify firm-specific risk through hedging instruments. Consistent with this prediction, my analysis documents a reverse relation between the hedging cost and holdings of options that have become vested and in-the-money. 79  Furthermore, I extend this study by investigating how managerial hedging influences corporate policies. In particular, I look at corporate dividend payouts. Prior research shows that option pay induces managers to cut dividend payments. Based on the idea that managers who can hedge are less influenced by their incentive portfolios, managerial hedging is expected to undermine the negative association between option compensation and dividend payments. I then provide evidence supporting this implication. To the extent that diversifying the firm and hedging personal portfolios are substitutes for managers to reduce risk, I document evidence that managers undergo fewer corporate diversification initiatives when they have lower costs to hedge their incentive pay. Finally, I conduct an extensive additional investigation and find that CEOs receive higher-power contracts when they can hedge firm systematic risk more easily, when they can implement OTC transactions more easily, and when they have actually hedged. In summary, this study concludes five major implications that: (1) managerial hedging undermines managers’ incentive to exert effort and increases their ability to bear risk; (2) shareholders enhance both the sensitivity and convexity of the relation between CEO wealth and stock return in corresponding compensation contracts; (3) shareholders adopt higher financial leverage to overcome this executive-hedging issue in addition to providing higher-power contracts; (4) managerial hedging induces managers to delay the exercise of their option grants, and weakens the negative relation between option compensation and dividend payment; and (5) managers diversify their firms less when they can hedge their incentive pay more easily. Finally, this paper provides indirect evidence that managers tend to use public options markets to undo their incentive compensation. 80  Table 3.1. Descriptive Statistic of Sample Firms Panel A: Descriptive Statistic of Firm Characteristics The sample consists of 13,691 firm-year observations from 1996 to 2005. In the sample, 10,123 firm-year observations have their options traded on U.S. option exchanges. I obtain stock price data in CRSP, accounting data in Compustat, CEO compensation data in ExecuComp, and option trading data in OptionMetrics. MV Equity ($million) refers to the market capitalization of the equity. ROE is the accounting return of equity, obtained as the ratio of earnings before interest and taxes to the book value of common equity. Leverage is the ratio of long-term debt (book value) over total assets. M/B is the ratio of market value of equity over book value of equity. Variance is the stock return variance based on the monthly return of past five years. Cash is the ratio of cash plus short-term investment over total assets. Stockreturn is the firm’s annual stock return. Option is a dummy variable, which equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Volume is the average number of daily option contracts traded. PPS is calculated as the dollar value change of the stock and options held by a CEO for per $1,000 shareholders return. PEI ($thousand) is the sensitivity of the total value of stock and options held by a CEO to 1% change in stock price. Ln() denotes the natural logarithm transform. All the dollar-value variables are measured in 2000-constant dollars.  MV Equity ROE Leverage M/B Variance*100 Age Cash Stockreturn Option Volume Ln(Volume) PPS PEI  Mean  Std  4814 10% 19% 2.93 1.77 55.76 12% 6.5% 0.74 990 5.09 26.08 388.9  10424 20% 18% 2.62 1.67 6.9 16% 47% 0.44 2554 1.93 51.56 978.37  81  5th Pct  Median  95th Pct  153 -17% 0 0.73 0.34 44 0 -78% 0 7 2.12 0.73 5.87  1339 12% 17% 2.14 1.17 56 5% 10.5% 1 139 4.94 6.96 112.2  20796 31% 49% 7.95 5.35 67 49% 76% 1 5153 8.55 138.7 1569.7  Panel B: Correlation Matrix of Explanatory Variables The sample consists of 13,691 firm-year observations from 1996 to 2005. Firmsize is the natural logarithm of the firm’s market value of equity; other variables used in this matrix are defined in Panel A. Correlations with an absolute value greater than 0.03 are significant at the 5% level.  (1) Option (2) Ln(Volume) (3) Firmsize (4) Firmsize2 (5) Variance (6) M/B (7) ROE (8) Leverage (9) Age (10) Cash (11)Stockreturn  (1)  (2)  (3)  (4)  1 0.34 0.30 0.09 0.15 0.06 -0.02 -0.03 0.11 -0.02  1 0.58 0.59 0.18 0.26 0.02 0.00 -0.04 0.19 0.00  1 0.89 -0.38 0.24 0.26 0.12 0.10 -0.18 -0.06  -0.28 0.27 0.24 0.08 0.07 -0.10 -0.06  (5)  (6)  (7)  (8)  (9)  (10)  (11)  1 0.12 -0.11 -0.10 0.29 -0.08  1 -0.05 0.08 -0.14 0.06  1 0.06 -0.43 -0.02  1 -0.16 0.02  1 -0.03  1  1 1 0.18 -0.39 -0.19 -0.23 0.52 -0.14  82  Table 3.2. Managerial Hedging Cost and Pay-performance Sensitivity Panel A: Using Option Dummy as the Proxy for Hedging Cost The sample consists of 13,691 firm-year observations from 1996 to 2005, 10,123 of which have their options traded on U.S. option exchanges. PPS is calculated as the dollar value change of the stock and options held by a CEO for per $1,000 shareholders return. PEI is the sensitivity of the total value of stock and options held by a CEO to 1% change in stock price (in $thousand). Option is a dummy variable, which equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Option Firmsize Firmsize2 CDF of Variance M/B ROE Leverage Age Cash Stockreturn Year Dummy Industry Dummy Firm Fixed Effects Intercept N Adjusted-R2/ Pseudo R2  (1) PPS OLS  (2) PPS Median  (3) PPS Fixed Effect  (4) Ln(PEI) OLS  (5) Ln(PEI) Median  (6) Ln(PEI) Fixed Effect  3.21*** [0.000] -26.46*** [0.000] 1.11** [0.017] 42.88*** [0.002] 2.01*** [0.000] 19.92*** [0.000] -33.92*** [0.000] 1.77*** [0.000] 33.76** [0.014] 51.66*** [0.000] Yes Yes No 59.44** [0.037] 13314  0.76*** [0.005] -5.65*** [0.000] 0.16*** [0.000] 9.91*** [0.000] 0.57*** [0.000] 4.39*** [0.000] -1.59** [0.04] 0.37*** [0.000] 7.16*** [0.000] 10.91*** [0.000] Yes Yes No 19.01*** [0.000] 13314  4.07** [0.013] -51.33*** [0.000] 2.36*** [0.000] 27.94*** [0.000] 2.35*** [0.000] 16.49*** [0.000] -17.69*** [0.000] 1.14*** [0.000] 4.76 [0.41] 45.01*** [0.000] Yes No Yes 208.35*** [0.000] 13314  0.35*** [0.000] 1.15*** [0.000] -0.04*** [0.000] 0.76*** [0.000] 0.06*** [0.000] 0.54*** [0.000] -0.34** [0.04] 0.042*** [0.000] 0.55*** [0.000] 1.26*** [0.000] Yes Yes No -3.89*** [0.000] 13314  0.11*** [0.000] 1.11*** [0.000] -0.04*** [0.000] 0.75*** [0.000] 0.08*** [0.000] 0.53*** [0.000] -0.21*** [0.01] 0.044*** [0.000] 0.76*** [0.000] 1.29*** [0.000] Yes Yes No -3.97*** [0.000] 13314  0.1*** [0.000] 0.78*** [0.000] -0.02*** [0.000] 0.21*** [0.002] 0.07*** [0.000] 0.34*** [0.000] -0.64*** [0.000] 0.05*** [0.000] 0.25*** [0.006] 1.14*** [0.000] Yes No Yes -2.87*** [0.000] 13314  18.5%  5.8%  22.6%  47%  29%  39%  83  Panel B: Using Ln(Volume) as the Proxy for Hedging Cost The sample consists of 10,123 firm-year observations from 1996 to 2005. All of the observations have their options traded on U.S. option exchanges. Pay-performance sensitivity (PPS) is calculated as the dollar value change of the stock and options held by a CEO for per $1,000 shareholders return. Portfolio equity incentive (PEI) is the sensitivity of the total value of stock and options held by a CEO to 1% change in stock price, and PEI is measured in $thousand. Volume is the average number of daily option contracts traded. Ln() denotes the natural logarithm transform. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Ln(Volume) Firmsize Firmsize2 CDF of Variance M/B ROE Leverage Age Cash Stockreturn Year Dummy Industry Dummy Firm Fixed Effects Intercept N Adjusted-R2/ Pseudo R2  (1) PPS OLS  (2) PPS Median  (3) PPS Fixed Effect  (4) Ln(PEI) OLS  (5) Ln(PEI) Median  (6) Ln(PEI) Fixed Effect  3.57*** [0.004] -30.71*** [0.000] 1.03** [0.026] 42.36*** [0.000] 2.21*** [0.000] 14.21*** [0.003] -35.52*** [0.000] 1.76*** [0.000] 19.08 [0.12] 47.62*** [0.000] Yes Yes No 96.00*** [0.006] 9837  1.14*** [0.000] -8.21*** [0.000] 0.23*** [0.000] 5.36*** [0.000] 0.57*** [0.000] 4.05*** [0.000] -3.13*** [0.000] 0.35*** [0.000] 6.13*** [0.000] 9.78*** [0.000] Yes Yes No 34.19*** [0.000] 9837  6.77*** [0.000] -63.76*** [0.000] 2.69*** [0.000] 4.41 [0.41] 2.38*** [0.000] 15.71*** [0.000] -23.17*** [0.000] 1.29*** [0.000] 1.97 [0.75] 41.59*** [0.000] Yes No Yes 256.29*** [0.000] 9837  0.11*** [0.000] 1.15*** [0.000] -0.05*** [0.000] 0.58*** [0.001] 0.06*** [0.000] 0.48*** [0.000] -0.47** [0.012] 0.04*** [0.000] 0.44** [0.034] 1.21*** [0.000] Yes Yes No -3.61*** [0.000] 9837  0.12*** [0.000] 1.01*** [0.000] -0.04*** [0.000] 0.45*** [0.000] 0.07*** [0.000] 0.48*** [0.000] -0.32*** [0.001] 0.05*** [0.000] 0.75*** [0.000] 1.24*** [0.000] Yes Yes No -3.35*** [0.000] 9837  0.13*** [0.000] 0.68*** [0.000] -0.02*** [0.001] -0.25*** [0.007] 0.07*** [0.000] 0.34*** [0.000] -0.67*** [0.001] 0.05*** [0.000] 0.35*** [0.001] 1.09*** [0.000] Yes No Yes -2.64*** [0.000] 9837  19.5%  6.3%  25.3%  43.5%  27%  41.5%  84  Table 3.3. Managerial Hedging Cost and Sensitivity of CEO Wealth to Stock Volatility The sample consists of 13,691 firm-year observations from 1996 to 2005, 10,123 of which have their options traded on U.S. option exchanges. The dependent variable is Ln(Vega) and Vega (in $thousand) is the dollar value change of the stock and options held by a CEO for 0.01 change in standard deviation of stock return. Option equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Volume is the average number of daily option contracts traded. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (1) OLS  Option  0.17*** [0.005]  (2) Median  0.09*** [0.000]  (3) Fixed Effect  Firmsize2 CDF of Variance M/B ROE Leverage Age Cash Stockreturn Year Dummy Industry Dummy Firm Fixed Effects Intercept N Adjusted-R2/ Pseudo R2  0.52*** [0.000] 0.005 [0.618] 0.28** [0.04] -0.04*** [0.000] 0.07 [0.336] 0.97*** [0.000] -0.02*** [0.000] -0.02 [0.941] 0.34*** [0.000] Yes Yes No -0.32 [0.554] 13314 37%  0.71*** [0.000] -0.006* [0.095] 0.29*** [0.000] -0.02*** [0.000] -0.05 [0.310] 0.85*** [0.000] -0.01*** [0.000] 0.42*** [0.000] 0.36*** [0.000] Yes Yes No -1.56*** [0.000] 13314  (5) Median  (6) Fixed Effect  0.12*** [0.000] 0.44*** [0.008] 0.001 [0.912] -0.43** [0.011] -0.04*** [0.000] 0.11 [0.211] 0.94*** [0.000] -0.02*** [0.000] 0.06 [0.801] 0.28*** [0.000] Yes Yes No 0.24 [0.732] 9919  0.11*** [0.000] 0.65*** [0.000] -0.01** [0.041] -0.23*** [0.001] -0.03*** [0.000] 0.02 [0.814] 0.79*** [0.000] -0.01*** [0.000] 0.38*** [0.000] 0.31*** [0.000] Yes Yes No -1.08*** [0.001] 9919  0.07*** [0.000] 0.08 [0.359] 0.02*** [0.000] -0.16* [0.08] -0.03*** [0.000] 0.04 [0.414] 0.39*** [0.000] -0.01*** [0.000] 0.18* [0.077] 0.25*** [0.000] Yes No Yes 1.68*** [0.000] 9919  0.09*** [0.001]  Ln(Volume) Firmsize  (4) OLS  0.26*** [0.000] 0.016*** [0.000] 0.09 [0.18] -0.03*** [0.000] 0.03 [0.480] 0.33*** [0.000] -0.01*** [0.000] 0.21** [0.026] 0.29*** [0.000] Yes No Yes 0.83*** [0.002] 13314  29%  22%  85  35%  27%  22%  Table 3.4. Simultaneous Equations (3SLS): Managerial Hedging Cost, Pay-performance Sensitivity, and Sensitivity of CEO Wealth to Stock Volatility The sample consists of 13,691 firm-year observations from 1996 to 2005, 10,123 of which have options traded on U.S. option exchanges. The jointly determined variables are Ln(Vega) and Ln(PEI). The variable Option equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Volume is the average number of daily option contracts traded. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (1) Ln(Vega) Ln(PEI)  -0.03 [0.336]  0.16*** [0.000]  Firmsize2 Variance M/B ROE Leverage  1.11*** [0.000] -0.022*** [0.000] 0.81*** [0.000] 0.018*** [0.001] 0.35*** [0.000] 0.42*** [0.000]  Age Cash Stockreturn Ln(Cash Compensation) Year&Industry Dummy Intercept N R2  0.41*** [0.000] 0.92*** [0.000] 0.56*** [0.000] Yes -5.24*** [0.000] 13314 19%  (4) Ln(PEI)  -0.14*** [0.000]  0.06** [0.039]  Ln(Volume) Firmsize  (3) Ln(Vega) -0.58*** [0.000]  -0.61*** [0.000]  Ln(Vega)  Option  (2) Ln(PEI)  1.16*** [0.000] -0.038*** [0.000] 0.76*** [0.001] 0.061*** [0.000] 0.54*** [0.000] -0.31*** [0.001] 0.04*** [0.000] 0.55*** [0.000] 1.26*** [0.000]  Yes -3.84*** [0.000] 13314 46%  86  0.16*** [0.000] 1.03*** [0.000] -0.028*** [0.000] 0.14 [0.18] 0.012** [0.044] 0.32*** [0.000] 0.41*** [0.000]  0.41*** [0.001] 0.84*** [0.000] 0.47*** [0.000] Yes -4.18*** [0.000] 9837 18%  0.12*** [0.000] 1.21*** [0.000] -0.046*** [0.000] 0.53*** [0.000] 0.057*** [0.000] 0.50*** [0.000] -0.32*** [0.001] 0.04*** [0.000] 0.44*** [0.000] 1.24*** [0.000]  Yes -3.51*** [0.000] 9837 43%  Table 3.5. Regression Analysis on Capital Structure The sample consists of 59,381 firm-year observations from 1996 to 2005, 7,638 of which have their options traded on U.S. option exchanges. The dependent variable is the book leverage ratio (in percentage), defined as the book value of long-term debt/total assets. Option equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Volume is the average number of daily option contracts traded. Tangibility is computed as the ratio of the firm’s fixed assets over total assets. R&D is the ratio of the firm’s R&D expenses to sales. Advertising is the ratio of the firm’s advertising expenses to sales. Ln() denotes the natural logarithm transform. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values from robust standard errors clustered at the firm level are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (1)  Option  (2)  2.19*** [0.000]  (3)  0.83** [0.023] 0.43*** [0.000]  Ln(Volume)  -2.92*** [0.008]  Democracy  Democracy×Ln(Volume)  Stockreturn CDF of Variance M/B ROE Tangibility R&D Advertising Year&Industry Dummies Intercept N Adjusted-R2  0.56*** [0.000] -13.56*** [0.000]  1.23* [0.1]  Democracy×Option  Firmsize  (4)  0.51*** [0.002] -0.93*** [0.001] 1.65** [0.035] -0.27*** [0.000] -0.21 [0.18] 24.75*** [0.000] 0.02 [0.84] -3.73 [0.26] Yes 11.73*** [0.000] 59371 18.3%  -0.09 [0.65] -0.79*** [0.000] -0.37 [0.71] -0.26*** [0.000] 0.07 [0.85] 22.44*** [0.000] 0.65*** [0.000] 5.41 [0.41] Yes 22.97*** [0.000] 17637 20%  87  0.34 [0.78] -1.47*** [0.001] 4.28*** [0.000] -0.24*** [0.001] 0.34 [0.51] 17.34*** [0.000] 3.77*** [0.000] -10.89 [0.12] Yes 21.41*** [0.000] 12124 20.6%  2.31*** [0.000] -1.04*** [0.000] -0.65 [0.18] -1.79 [0.19] -0.29*** [0.000] 0.27 [0.62] 17.72*** [0.000] 4.75*** [0.000] -5.69 [0.47] Yes 32.99*** [0.000] 9214 22%  Table 3.6. Regression Analysis on Executive Option Holding The sample consists of 13,691 firm-year observations from 1996 to 2005. In the sample, 10,123 firm-year observations have their options traded on U.S. option exchanges. Opt1 is the dollar value of the exercisable in-the-money options held by the CEO. Opt2 is calculated as Opt1 over the CEO’s total annual income. Opt3 is defined as the number of shares underlying the CEO’s vested in-the-money option as the percentage of the firm’s stock shares owned by the CEO. Option is a dummy variable, which equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Volume is the average number of daily option contracts traded. Ln() denotes the natural logarithm transform. Ownership is the percentage of the firm’s shares owned by the CEO. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values from robust standard errors clustered at the firm level are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Option  (1) Ln(1+Opt1)  (2) Ln(1+Opt2)  (3) Ln(1+Opt3)  0.41*** [0.000]  0.22*** [0.000]  0.09** [0.033]  CDF of Variance Ownership M/B Age StockReturn Year&Industry Dummies Intercept N Adjusted-R2  (5) Ln(1+Opt2)  (6) Ln(1+Opt3)  0.028 [0.126] 0.13*** [0.005] -0.32 [0.139] -4.59*** [0.000] 0.085*** [0.000] 0.007 [0.225] 1.69*** [0.000]  0.092*** [0.000] -0.088* [0.060] -0.38* [0.064] -17.84*** [0.000] -0.013 [0.201] -0.033*** [0.000] -0.12** [0.016]  0.61*** [0.000] 0.19 [0.41] -7.69*** [0.000] 0.17*** [0.000] 0.004 [0.318] 2.52*** [0.000]  0.21*** [0.000] -0.09 [0.59] -3.33*** [0.000] 0.14*** [0.000] 0.005* [0.087] 1.66*** [0.000]  0.05*** [0.000] 0.19*** [0.000] -19.35*** [0.000] 0.001 [0.925] -0.023*** [0.000] -0.057 [0.175]  0.098*** [0.000] 0.46*** [0.000] -0.45 [0.162] -9.32*** [0.000] 0.11*** [0.000] 0.008 [0.343] 2.61*** [0.000]  Yes  Yes  Yes  Yes  Yes  Yes  -0.38 [0.595] 13689 30.6%  0.45 [0.343] 13689 30.6%  4.56*** [0.000] 13682 28%  1.94*** [0.004] 10121 27.6%  2.11*** [0.000] 10121 21.8%  6.27*** [0.000] 10114 29.1%  Ln(Volume) FirmSize  (4) Ln(1+Opt1)  88  Table 3.7. Regression Analysis on Corporate Dividend Policy The sample consists of 17,036 firm-year observations from 1997 to 2006, 11,247 of which have traded options on U.S. option exchanges. The dependent variable is the firm’s dividend normalized by its stock capitalization. Option equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Volume is the average number of daily option contracts traded. Optionpay is the Black-Scholes value of the CEO’s annual option grant normalized by her total compensation. Optionwealth is the Black-Scholes value of the CEO’s total option holding. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values from robust standard errors clustered at the firm level are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (1) (2) (3) (4) Optionpay  -1.22*** [0.000] -0.053*** [0.000]  Ln(Optionwealth)  Option×Optionpay  -0.76*** [0.000] -0.041*** [0.000]  0.71*** [0.000] 0.02*** [0.005]  Option×Ln(Optionwealth)  0.05*** [0.004]  Ln(Volume)×Optionpay  0.001 [0.487]  Ln(Volume)×Ln(Optionwealth) Option  -0.54*** [0.000]  -0.55*** [0.000]  Ln(Volume) Firmsize CDF of Variance M/B ROE Leverage Cash Year&Industry Dummies Intercept N Adjusted-R2  0.13*** [0.000] -3.02*** [0.000] -0.05*** [0.000] 0.02 [0.429] 0.23*** [0.000] -0.43*** [0.000] Yes 0.59*** [0.000] 16991 35.9%  89  0.15*** [0.000] -3.12*** [0.000] -0.04*** [0.000] 0.09*** [0.002] 0.22*** [0.000] -0.53*** [0.000] Yes 0.62*** [0.000] 17036 36.2%  -0.20*** [0.000] 0.33*** [0.000] -2.59*** [0.000] -0.05*** [0.000] 0.06* [0.066] 0.49*** [0.000] -0.38*** [0.000] Yes -0.57*** [0.002] 11247 36.4%  -0.21*** [0.000] 0.37*** [0.000] -2.63*** [0.000] -0.04*** [0.000] 0.12*** [0.001] 0.48*** [0.000] -0.42*** [0.000] Yes -0.54*** [0.009] 11246 37.3%  Table 3.8. Regression Analysis on Corporate Diversification The sample consists of 52,472 firm-year observations from 1996 to 2005. I obtain stock price data in CRSP, accounting data in Compustat, and option trading data in OptionMetrics. In the sample, 16,369 firm-year observations have their options traded on U.S. option exchanges. The dependent variables are the Herfindahl Index and the natural logarithm of number of business segments. The Herfindahl Index is calculated as the sum of the square of segment sales divided by the square of firm sales. Option is a dummy variable, which equals one if the firm’s option is traded on U.S. option exchanges, and zero otherwise. Volume is the average number of daily option contracts traded. Ln() denotes the natural logarithm transform. Industry dummies are constructed based on Fama and French’s (1997) 48 industries. Corresponding p-values from robust standard errors clustered at the firm level are reported in brackets. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (1) Herfindahl Index  Option  0.017*** [0.001]  Ln(Volume) Firmsize Stockreturn/100 CDF of Variance M/B/100 ROE/100 Cash Leverage Year&Industry Dummies Intercept N Adjusted-R2  (2) Herfindahl Index  -0.027*** [0.000] 0.27*** [0.000] 0.08*** [0.000] 0.41*** [0.000] 0.91*** [0.001] 0.13*** [0.000] -0.057*** [0.000] Yes 0.97*** [0.000] 51532 20%  (3) Ln(segments)  (4) Ln(segments)  -0.041*** [0.002] 0.004* [0.066] -0.028*** [0.000] 0.20** [0.042] 0.13*** [0.000] 0.42*** [0.000] 1.16*** [0.002] 0.15*** [0.000] -0.043** [0.029] Yes 0.98*** [0.000] 16220 26%  90  0.073*** [0.000] -0.61*** [0.000] -0.21*** [0.000] -1.12*** [0.000] -2.59*** [0.000] -0.35*** [0.000] 0.147*** [0.000] Yes 0.072 [0.152] 52472 24%  0.003 [0.644] 0.058*** [0.000] -0.56** [0.023] -0.39*** [0.000] -1.13*** [0.000] -3.48*** [0.000] -0.41*** [0.000] 0.094* [0.058] Yes 0.012 [0.925] 16369 31%  Table 3.9. Beta and Incentive Pay Panel A: Firm Beta The sample and regression specifications are the same as those in Panel A of Table 3.2, except that I replace Option with the firm beta. To estimate the beta of each firm-year observation, I use the preceding five years of monthly return and run an OLS regression of the firm’s returns on the returns of the CRSP value-weighted index. The coefficients of controls (unreported) are similar to those in Panel A of Table 3.2. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Beta  (1)  (2)  (3)  (4)  (5)  (6)  PPS OLS  PPS Median  PPS Fixed Effect  Ln(PEI) OLS  Ln(PEI) Median  Ln(PEI) Fixed Effect  -9.03*** [0.001]  -1.35*** [0.000]  -11.71*** [0.000]  -0.087* [0.052]  -0.13*** [0.000]  -0.14*** [0.000]  Panel B: Industry Beta The sample and regression specifications are the same with those in Panel A of Table 3.2, except that I replace Option with the firm’s industry beta. To compute the industry beta in each sample year, I run an OLS regression of the returns of value-weighted industry portfolio on the returns of the CRSP value-weighted index using the preceding five years of monthly return. I then assign the industry beta to the firm-year observations within the same industry. The industry classification is based on Fama and French’s (1997) 48 industries. The coefficients of controls (unreported) are similar to those in Panel A of Table 3.2. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Industry Beta  (1)  (2)  (3)  (4)  (5)  (6)  PPS OLS  PPS Median  PPS Fixed Effect  Ln(PEI) OLS  Ln(PEI) Median  Ln(PEI) Fixed Effect  -10.44*** [0.003]  -2.45*** [0.000]  -12.11*** [0.000]  -0.19*** [0.006]  -0.22*** [0.000]  -0.15*** [0.000]  91  Table 3.10. Financial Industry and Incentive Pay The sample and regression specifications are the same with those in Panel A of Table 3.2, except that I replace Option with the Financial Industry dummy (industry dummies and firm fixed effects are also excluded). The Financial Industry dummy takes the value of one if the firm is in banking, insurance, or trading industries, and zero otherwise. The coefficients of controls (unreported) are similar to those in Panel A of Table 3.2. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Financial Industry  (1)  (2)  (3)  (4)  PPS OLS  PPS Median  Ln(PEI) OLS  Ln(PEI) Median  1.11 [0.76]  2.91*** [0.000]  0.41*** [0.000]  0.49*** [0.000]  92  Table 3.11. Direct Executive Hedging Transactions and Incentive Pay Panel A: ExecuComp Population Sample The sample and regression specifications are the same with those in Panel A of Table 3.2, except that I replace Option with the Hedged dummy. The Hedged dummy takes the value of one if CEO-year observation is associated with purchases of equity swaps, forward sales or put options identified in the Thomson Reuters database (75 observations), and zero otherwise. The coefficients of controls (unreported) are similar to those in Panel A of Table 3.2. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Hedged  (1)  (2)  (3)  (4)  PPS OLS  PPS Median  Ln(PEI) OLS  Ln(PEI) Median  41.35*** [0.004]  27.26*** [0.000]  1.01*** [0.000]  1.12*** [0.000]  Panel B: Matched-firm Sample For each of the 75 firm-year observations specified in Panel A, I form a control firm in the same Fama and French (1997) 48 industries with the closest market value of equity in the preceding year. The Hedged dummy takes the value of one if CEO-year observation is associated with purchases of equity swaps, forward sales or put options identified in the Thomson Reuters database, and zero otherwise. The coefficients of controls (unreported) are qualitatively similar to those in Panel A of Table 3.2. Corresponding p-values are reported in brackets. The p-values for OLS regressions are based on robust standard errors clustered at the firm level. The p-values for median regressions are according to bootstrapped standard errors based on 20 replications. The notation ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Hedged  (1)  (2)  (3)  (4)  PPS OLS  PPS Median  Ln(PEI) OLS  Ln(PEI) Median  46.07** [0.012]  28.64*** [0.000]  0.95*** [0.000]  1.25*** [0.000]  93  References Acharya, V. V., and A. Bisin, 2005, Managerial hedging, equity ownership, and firm value, London Business School working paper. Aggarwal, R., and A. Samwick, 1999, The other side of the trade-off: The impact of risk on executive compensation, Journal of Political Economy 10, 65-105. Amihud, Y., and B. Lev, 1981, Risk reduction as a managerial motive for conglomerate mergers, Bell Journal of Economics 12, 605-617. Antle, R., and A. Smith, 1986, An empirical examination of relative performance evaluation of corporate executives, Journal of Accounting Research 24, 1-39. Baker, P. G., and J. B. Hall, 2004, CEO incentives and firm size, Journal of Labor Economics 22, 767-798. Bebchuk, L.A., J. M. Fried, and D. I. Walker, 2002, Managerial power and rent extraction in the design of executive compensation, University of Chicago Law Review 69, 751-846. Bettis, J. C., J. M. Bizjak,, and M. L. Lemmon, 2001, Managerial ownership, incentive contracting, and the use of zero-cost collars and equity swaps by corporate insiders, Journal of Financial and Quantitative Analysis 36, 345-370. Bettis, J. C., J. M. Bizjak, and M. L. Lemmon, 2005, Exercise behavior, valuation, and the incentive effects of employee stock options, Journal of Financial Economics 76, 445-470. Bisin, A., P. Gottardi, and A. A. Rampini, 2006, Managerial hedging and portfolio monitoring, New York University working paper. Breen, W. J., L. S. Hodrick, and R. A. Korajczyk, 2002, Predicting equity liquidity, Management Science 48, 470-483. Carpenter, J.N., 1998, The exercise and valuation of executive stock options, Journal of Financial Economics 48, 127-158. Coles, J. L., N. D. Daniel, and L. Naveen, 2006, Managerial incentives and risk-taking, Journal of Financial Economics 79, 431-468. Core, J., and W. Guay, 1999, The use of equity grants to manage optimal equity incentive levels, Journal of Accounting and Economics 28, 151-184. 94  Core, J., and W. Guay, 2002, Estimating the value of employee stock option portfolios and their sensitivities to price and volatility, Journal of Accounting Research 40, 613-630. Economist, 1999, Share options: Executive relief, April 3, 64. Fama, E., and K. French, 1997, Industry costs of capital, Journal of Financial Economics 43, 153-193. Fenn, G. W., and N. Liang, 2001, Corporate payout policy and managerial stock incentives, Journal of Financial Economics 60, 45-72. Garvey, G., 1997, Marketable incentive contracts and capital structure relevance, Journal of Finance 52, 353-378. Garvey, G., and T. Milbourn, 2003, Incentive compensation when executives can hedge the market: Evidence of relative performance evaluation in the cross section, Journal of Finance 58, 1557-1582. Gibbons, R., and K. J. Murphy, 1992, Optimal incentive contracts in the presence of career concerns: Theory and evidence, Journal of Political Economy 100, 468–505. Gompers, P., J. Ishii, and A. Metrick, 2003, Corporate governance and equity prices, Quarterly Journal of Economics 118, 107-155. Guay, W. R., 1999, The sensitivity of CEO wealth to equity risk: An analysis of the magnitude and determinant, Journal of Financial Economics 53, 43-71. Guner, B., G. Tate, and U. Malmendier, 2008, Financial expertise of directors, Journal of Financial Economics 88, 323-354. Hall, B. J., and J. Liebman, 1998, Are CEOs really paid like bureaucrats?, The Quarterly Journal of Economics 103, 653-691. Hall, B. J., and K. J. Murphy, 2002, Stock options for undiversified executives, Journal of Accounting and Economics 33, 3–42. Harris, M., and A. Raviv, 1991, The theory of capital structure, Journal of Finance 46, 297-356. Hemmer, T., S. Matsunaga, and T. Shevlin, 1996, The influence of risk diversification on the early exercise of employee stock options by executive officers, Journal of Accounting and Economics 21, 45-68. Holmstrom, B., and P. Milgrom, 1987, Aggregation and linearity in the provision of 95  intertemporal incentives, Econometrica 55, 303-328. Holmstrom, B and P. Milgrom, 1991, Multi-Task principal agent analysis: Linear contracts, asset ownership and job design, Journal of Law, Economics, and Organization 7, 24-52. Huddart, S., and M. Lang, 1996, Employee stock option exercises: An empirical analysis, Journal of Accounting and Economics 21, 5-43. Jagolinzer, A. D., S. Matsunaga, and E. Yeung, 2007, An analysis of insiders' use of prepaid variable forward transactions, Journal of Accounting Research 45, 1055-1079. Jensen, M., 1986, Agency cost of free cash-flow, corporate finance, and takeovers, American Economic Review 76, 323-329. Jensen, M., and W. Meckling, 1976, Theory of the firm: managerial behavior, agency costs, and ownership structure, Journal of Financial Economics 3, 305-360. Jensen, M., and K. Murphy, 1990, Performance pay and top management incentives, Journal of Political Economy 98, 225-264. Jin, L., 2002, CEO compensation, diversification, and incentives, Journal of Financial Economics 66, 29-63. John, T. A., and K. John, 1993, Top management compensation and capital structure, Journal of Finance 48, 949-974. Klein, A., 1998, Firm performance and board committee structure, Journal of Law and Economics 41, 275–303. Korajczyk, R. A., and R. Sadka, 2004, Are momentum profits robust to trading costs?, Journal of Finance 59, 1039-1082. Lambert, R.A., W.N. Lanen, and D.F. Larcker, 1989, Executive stock option plans and corporate dividend policy. Journal of Financial and Quantitative Analysis 24, 409–425. Lavelle, L., 2001, Undermining pay performance, Business Week 15, 70-71. Malmendier, U., and G. Tate, 2005, CEO overconfidence and corporate investment, Journal of Finance 60, 2661-2700. Manski, C., and S. Lerman, 1977, The estimation of choice probabilities from choice based samples, Econometrica 45, 1977-1988. May, D., 1995, Do managerial motives influence firm risk reduction strategies?, Journal of 96  Finance 50, 1291–1308. Mayhew, S., and V. Mihov, 2004, How do exchanges select stocks for option listing?, Journal of Finance 59, 447-471. Merton, R., 1973, Theory of rational option pricing, Bell Journal of Economics and Management Science 4, 141-183. Milbourn, T. T., 2003, CEO reputation and stock-based compensation, Journal of Financial Economics 68, 233-262. Muelbroek, L. K., 1992, An empirical analysis of illegal insider trading, Journal of Finance 7, 1661-1701. Ofek, E., and D. Yermack, 2000, Taking stock: Equity-based compensation and the evolution of managerial ownership, Journal of Finance 55, 1367-1384. Ortiz-Molina, H., 2007, Executive compensation and capital structure: The effect of convertible debt and straight debt, Journal of Accounting and Economics forthcoming. Ozerturk, S., 2006, Financial innovations and managerial incentive contracting, Canadian Journal of Economics 39, 434-454. Prendergast, C., 2002, The tenuous trade-off between risk and incentives, Journal of Political Economy 110, 1071-1102. Puri, S., 1997, New tools for the options crowd, Fortune, November 10. Schizer, D. M., 2000, Executives and hedging: The fragile legal foundation of incentive compatibility, Columbia Law Review 100, 440-504. Simon, R., 2000, Hedging a single stock as ups, downs, Wall Street Journal, February 2. Smith, C., and R. Stulz, 1985, The determinants of firm’s hedging policies, Journal of Financial and Quantitative Analysis 20, 391-405. Stulz, R., 1984, Optimal hedging policies, Journal of Financial and Quantitative Analysis 19, 127-140. Stulz, R., 1990, Managerial discretion and optimal financing policies, Journal of Financial Economics 7, 117-161. Yermack, D., 1995, Do corporations award CEO stock options effectively? Journal of Financial Economics 39, 237-269. 97  Chapter 4 Incentive Effects of Extreme CEO Pay Cuts24 4.1. Introduction Executive compensation has received enormous attention from both academics and the popular press. However, most of that attention has focused on the overall rise in pay, the increasing use of equity-based compensation, and the debate over whether pay is appropriately sensitive to performance (see for example, Jensen and Murphy (1990), and Bebchuk and Fried (2003)). In this paper, we study one aspect of compensation in corporate governance that has been largely overlooked—large, discrete pay cuts. Our study fits within the general question of how do boards alter compensation to motivate CEOs (see Core and Guay (1999), for example)? While some prior studies such as Acharya, John, and Sundaram (2000) have examined the practice of repricing executive stock options after a performance decline in order to preserve incentives, we examine what is effectively the opposite practice—sharply reducing CEO pay following poor performance with the implicit or explicit promise of restoring it if performance improves. Our study is increasingly relevant given the US government’s effective broad-based adoption of this practice in the financial industry bailout. For ExecuComp firms over our sample period of 1994-2005, we identify over 1,000 instances of extreme pay cuts where a CEO’s pay is reduced by at least 25% from the prior year, representing roughly 10% of the firm-year observations. The average (median) pay cut in our sample is 46% (42%) of the CEO’s pay in the prior year. We ensure that these cuts are not mechanical reversals of a prior pay spike. Further, using a model of normal compensation, we show that only about 60% of our sample CEOs have abnormally high 24  A version of this chapter will be submitted for publication. Gao, H., Harford, J., and Li, K., Incentive Effects of Extreme CEO Pay Cuts. 98  compensation in the year prior to the pay cut. The reduction in total pay is mainly due to a decrease in the units of stock and options grants leading to a major reduction in the value of equity-based compensation. In our sample of extreme pay cuts, the median CEO experiences a 75% reduction in his equity-based pay but only a 21% reduction in his salary and bonus. We find, unsurprisingly, that poor firm performance predicts a pay cut. The CEO of a firm with an industry-adjusted return of below −15% has a 13% probability of seeing his pay cut by 25% or more. However, we also find that the likelihood of receiving a sharp pay cut following poor performance is higher in firms with stronger governance mechanisms and limited opportunities to find replacing CEOs. While a pay cut can be thought of as a reprieve from dismissal in the year of the pay cut, it is not a long-term substitute for dismissal. About one quarter of our sample CEOs is dismissed in the year following the pay cut. However, those CEOs who do engineer a turnaround see their pay restored to normal levels. The performance turnaround is accompanied and aided by abnormal reductions in capital expenditures and R&D expenses, and a sharp reduction in leverage through active debt retirement. We implement a number of additional tests to understand the incentive mechanism underlying this extreme form of pay cut documented in the paper. We first provide evidence that the pay cut is not limited to the CEOs but also applies to other members of the top management team. Second, we single-out the cluster of pay cuts during the economic and stock market downturn of 2000-2001. Reflecting some degree of relative performance evaluation (Holmström (1979)), boards are less likely to cut pay and to dismiss CEOs for poor performance during a market downturn. However, those that do cut see greater performance improvement than those cutting outside the downturn period do. Notably, CEOs whose pay is cut during the downturn experience less pay recovery, consistent with 99  the outside employment opportunity theory of CEO pay (Oyer (2004)). Third, we investigate whether the level of CEO (excess) pay prior to the pay cut is important in the pay-cutting decision. While CEOs with excess pay are more likely to have their pay cut and tend to have smaller pay recovery later on, the level of pay before the cut has little influence on retention or performance change following the cut. Fourth, to explore the possibility that the performance improvement following a pay cut is due to performance mean reversion, we form a control sample for our pay cut sample based on year, industry, and stock return prior to the pay cut. Relative to these control companies, pay-cutting firms still experience significant improvement in stock returns after the cuts. Finally, we compute turnover statistics for pay-cut CEOs up to three years after the pay cuts. Compared to the control firm CEOs or to any CEOs covered by ExecuComp, the pay-cut CEOs experience much lower retention in future years. In summary, we conclude that boards use extreme pay cuts to motivate poorly performing managers to improve performance. In contrast to the way repriced options are used to maintain incentives, these boards use the pay cut to penalize poor performance and the (likely explicit) promise of restored pay to provide incentives to improve. The approach is generally effective, with firm performance strengthening and the successful CEO remaining in his post with restored pay following the pay cut. Our paper contributes to the literature on executive compensation by providing the first systematic examination (to our knowledge) of CEO pay cuts. We demonstrate that boards’ use of sharp cuts to CEO pay provides effective ex ante incentives for them to exert effort to avoid poor performance and ex post incentives to improve poor performance if it occurs. Our evidence is generally supportive of the optimal contracting view of current compensation practice. The plan of the paper is as follows. We review the literature and develop our 100  hypotheses in the next section. We describe our sample and variable construction in Section 4.3. We explore the causes and consequences of extreme pay cuts in Section 4.4. Additional investigation is implemented in Section 4.5, and we conclude in Section 4.6.  4.2. Prior Literature and Hypothesis Development 4.2.1. Literature Review The principal-agent conflict is one of the most well-known conflicts in corporate finance. Since the seminal work by Jensen and Meckling (1976) that identifies the agency costs associated with the separation of ownership and control in modern corporations, many papers have explored the incentive mechanisms that overcome this conflict. It is well recognized that an efficient compensation system should reward executives for their effort and resulting good firm performance and punish them if they fail to deliver. Jensen and Murphy (1990) provide the first systematic evidence on executive compensation in the US. They show that senior managers on average experience relatively little reduction in their personal wealth when their firms are unprofitable: CEO wealth typically decreases by only $3.25 (per $1,000 decrease in shareholder wealth) for a sample of 2,213 CEOs listed in Forbes’ Executive Compensation Surveys from 1974-1986. Hall and Liebman (1998) find that by the end of the 1990s, the pay-for-performance link for CEOs jumps almost ten folds since 1980. Here, rather than studying the general pay-for-performance link, we focus on discrete drops in pay as a response to poor performance. We are interested in how the threat of such a drop in pay induces effort and how, conditional on poor performance, the drop in pay creates incentives to turn the performance around. As such, although our sample firms are generally not in distress, our paper is most closely related to studies of CEO compensation during financial difficulty and restructuring. 101  Using a sample of 77 financially distressed firms between 1981-1987, Gilson and Vetsuypens (1993) show that about one third of the CEOs in their sample are replaced in a given year around default, and those who remain experience significant salary and bonus reductions, and are often granted new stock options in an attempt to closely align the interests of CEOs and shareholders. Using a sample of 263 Swedish bankruptcy cases, Eckbo and Thorburn (2003) find that the median income loss of CEOs of bankrupt firms is −47% as compared to the contemporaneous income change of CEOs of nonbankrupt industry rival firms of similar size. Using more recent data and the entire population of ExecuComp CEOs, Bebchuk and Grinstein (2007) show that there is an asymmetry in CEO pay responses to size increases and decreases: Firm expansion is positively correlated with subsequent CEO pay, while firm contraction is not correlated with subsequent CEO pay. In the literature focusing on CEO incentives during downturns, Dial and Murphy (1995) document significant increases in equity-based compensation at the defense contractor General Dynamics when the entire industry was in decline following the end of the Cold War. Mehran, Nogler, and Schwartz (1998) study the effects of managerial equity ownership and compensation on voluntary liquidation decisions and find that at least 41% of CEOs who close down their firms are personally better off. They conclude that current compensation plans motivate CEOs not only to expand but also to downsize for the purpose of value creation. Our work is also related to the literature on option repricing. Repricing refers to the practice of canceling out-of-the-money options and reissuing options with a lower strike price. Prior work has shown that repricings are economically significant compensation events: The new strike prices are often 30%-40% lower than the old strike prices (see for example, Chance, Kumar, and Todd (2000)). Using a sample of ExecuComp firms from 1992-1995, Brenner, Sundaram, and Yermack (2000) find an incidence of repricing of less 102  than 1.3% per firm year, and there is a strong negative relation between repricing and firm performance even after correcting for industry performance. Using a sample of 213 instances of repricing from 1992-1997, Chidambaran and Prabhala (2003) show that negative shocks to firm performance lead to repricing and the performance is not reversed in the next two years. Following repricing, CEOs still experience high turnover rates. Overall, there is some debate in the literature about whether repricings are effective governance mechanism. In our paper, we examine the efficacy of significantly reducing CEO pay following poor performance. Ostensibly, the two practices can be aimed at achieving the same goal—giving managers incentives to improve performance.  4.2.2. Our Hypotheses Our general hypothesis is that the threat of sharp pay cuts is one way boards provide managers with ex-ante incentives to maintain strong performance. Here we are assuming that in response to poor performance, boards evaluate whether the performance problem is due to CEO effort or skill. If the board concludes that CEO skill is low relative to the average skill in the CEO labor pool, then it will rationally dismiss the CEO and draw a new CEO from the labor pool. If the board puts enough weight on an effort problem, then it will rationally retain the CEO with potentially different inducements to effort. In this way, the cuts are a form of the ex-post settling-up incentives discussed in Fama (1980): Once the pay cuts are enacted, a promise of restored pay (and retaining his job) provides incentives to the CEO to improve performance. Moreover, a pay cut can shrink the manager’s outside opportunity and press him to work hard to get more pay. In summary, the incentive effects of pay cuts consist of two parts: an ex ante incentive for managers to avoid poor performance and an ex post incentive for them to exert more effort to retain their positions and get the pay back. 103  While most CEOs observably work hard, the kind of effort change we envision is a redirection of existing effort toward tasks that the CEO may have preferred to avoid, but which might be necessary to improve performance. For example, a CEO may not want to fire a top lieutenant, divest a division, or kill a pet project, but the board uses the pay cut and promised restoration of pay to increase his incentives to perform these tasks that otherwise provide him with disutility. Boards will be more likely to employ a pay cut over dismissal when they perceive there to be a lower expected replacement value from the CEO labor pool. For example, there are fewer potential substitutes with the experience necessary to run a very large company relative to a small company. It is worth noting that we do not assume that the CEO’s original contract was suboptimal. Rather, we view the flexibility of the board to sharply alter pay as part of the optimal contract with the CEO. This flexibility comes from discretion over the size of bonus, and stock and options grants. Gillan, Hartzell, and Parrino (2009) find that fewer than half of the S&P 500 CEOs have an explicit employment contract that would limit the board’s flexibility.25 Several specific hypotheses follow from the general hypothesis above in which pay cuts are viewed as part of the optimal contracting environment. H1: Poor performance increases the likelihood of a sharp pay cut and the link is stronger in firms with better governance. H2: Larger firms will be more likely to react to poor performance with a pay cut because the dismissal option is not as viable for them.  25 Equity-based compensation itself may already have the feature that the CEO’s wealth declines with declining stock price. A pay cut may differ from equity-based compensation in two ways. First, a pay cut is an active action taken by the board while the change of value of the CEO’s existing portfolio is largely out of control of the board. Second, a pay cut can send the CEO a clear message that the board is unhappy with the firm performance and is willing to take actions. 104  H3: Relative performance evaluation will cause the likelihood of a pay cut to be insensitive to industry performance.  In order for pay cuts to be a part of the optimal compensation scheme, on average they should be effective in producing improved performance. Thus, we hypothesize the following: H4: Following pay cuts, managers will take actions that effect a turnaround in performance. For a given level of performance, pay cuts will make such actions more likely and hence pay cuts will lead to greater performance improvements.  Finally, for the pay cut to have the appropriate ex ante incentives, improved performance must be met with restored pay and continued poor performance should result in dismissal such that the pay cut is only a short-term substitute for performance-driven dismissal. H5: CEOs who are successful in improving performance will have their pre-pay-cut pay restored.  4.3. Sample Formation and Variable Construction To construct our sample, we employ the ExecuComp database for the period of 1994-2005.26 We first identify CEOs who experience an extreme form of pay cut of at least 25% in total compensation (2,633 firm-year observations). CEO pay will fluctuate over time if stock and options grants, the largest component in CEO compensation, are not granted every year. Suppose a CEO is granted stock and options awards once every two years, we will mechanically observe “compensation reductions” every second year. To address this issue, we further require that the increase in CEO pay in the year prior to the pay cut is no more than 25% (1,572 firm-year observations). This additional filter helps  26 The ExecuComp database starts in 1992, but our sample identification scheme requires information on CEO pay for at least two years before a pay cut. As a result, our final sample period is from 1994-2005. 105  ensure that pay cuts identified in our sample are not due to the normal fluctuation in pay.27 As we discuss later, Figure 4.1 shows that our sample of CEOs do not experience a spike in pay in the year prior to the cut. In summary, a CEO experiences a major pay cut and thus is included in our sample if (1) the same CEO keeps his position from year -2 to the pay cut year 0; (2) his total pay in year 0 is no more than 75% of his pay in year -1; and (3) his total pay in year -1 is no more than 125% of his pay in year -2. Our final sample consists of 1,061 instances of pay cuts.28 Table 4.1 presents an overview of our pay cut sample. Panel A reveals that the frequency of pay cuts has increased over time. It also reveals that the majority of the pay cuts took place during the early 2000s as the economy entered into a recession and the stock market fell considerably from its peak. Panel B shows that the vast majority (85%) of our sample firms reduce their CEOs’ pay by between 25% and 65% of their pay in the prior year. Unreported in the table, the average (median) size of pay cut is 46% (42%) of the CEO’s pay in year -1. In untabulated analysis, we find that three Fama-French industries (Fama and French (1997)): Computers, electronic equipment, and measuring and control equipment, have a noticeably higher representation in our pay cut sample (17.3%) than in the overall ExecuComp population (11.4%). Given that our pay-cutting sample is formed from ExecuComp and the ExecuComp firms are not random draws of the US listed  27  We also formulate our pay-cutting sample by using a longer pay window so that a pay cut is deemed to occur if the  CEO’s total pay in the pay cut year is no more than 75% of his average pay over the previous two years. We end up with 2,891 pay cuts in the sample. Further investigation suggests that most of our key results remain unchanged (results available upon request). 28  In our sample, there are 662 firms experiencing only one CEO pay cut, 169 firms with two pay cuts, 19 firms with  three pay cuts, and one firm with four pay cuts (Alcoa made a pay cut to its CEO Paul O’Neill in 1997, and made three consecutive pay cuts to its CEO Alain Belda between 2003-2005). Among the 169 firms that make two pay cuts (338 pay cuts), 45 firms make pay cuts to different CEOs and 124 firms make pay cuts to the same CEO. Within the latter set, 59 firms make pay cuts in consecutive years, 13 firms make pay cuts within three years (e.g., year t for the first pay cut and year t+2 for the second one), 24 firms within four years, and 28 firms make pay cuts with a window longer than five years. 106  companies (as they tend to be larger than the overall Compustat population), we will employ the ExecuComp firms to establish performance benchmarks in our later analyses. Table 4.2 reports CEO pay and firm characteristics from the year before (year -1) to the year after the pay cut (year +1). All dollar values are in 2006 dollars. To mitigate the effect of outliers, we winsorize all continuous variables at the one percent level in both tails of the distribution. We define an executive’s total compensation (Totalpay) in a given year as the sum of the executive’s salary, bonuses, long-term incentive plans, the grant-date value of restricted stock awards, and the Black-Scholes value of granted options. Panel A shows that the median CEO receives Totalpay of $2.9 million in year -1 and experiences a drastic pay reduction to $1.5 million in the pay cut year. Afterwards, the median Totalpay increases to $2.5 million in year +1. We also decompose Totalpay into Cashpay and Equitypay, where Equitypay is the value of restricted stock granted and the Black-Scholes value of stock options granted, and Cashpay accounts for the remainder. The majority of the compensation reduction is due to Equitypay, as Cashpay is stable over time. Specifically, from year -1 to the pay cut year the median Equitypay decreases from $1.7 million to $429,000. In contrast, the corresponding change in Cashpay is from $1.1 million to $853,000. Put differently, the median CEO experiences a 75% cut in Equitypay, but only a 21% cut in Cashpay. A CEO at a firm that has a policy of granting a constant number of options would experience fluctuations in Equitypay simply due to increases and decreases in the stock price. To rule out the possibility that the pattern in Equitypay is being caused by fluctuations in the firm’s stock performance, we examine the units of stock and options granted. Stockshare is defined as the number of shares of the CEO’s annual stock grant, 107  normalized by the total number of shares outstanding. Optionshare is the number of shares underlying the CEO’s annual options grants, normalized by the total number of shares outstanding. The last three rows in each sub-panel of Panel A indicate that the number of CEO stock and options grants displays exactly the same pattern as the value of Equitypay and Totalpay. The mean Stockshare is reduced sharply from 0.03% in year -1 to 0.01% in the pay cut year (the corresponding median value is zero). In the meantime, the median Optionshare drops from 0.19% to 0.03%. In the year before the pay cut (Table 4.2, first sub-panel of Panel B), the median sample firm has poor stock performance with the raw stock return of −4.5% and industry-adjusted return of −19%, while its accounting performance looks normal with ROA of 12%, and industry-adjusted ROA of 0%. Noting that the 95th percentile of industry-adjusted return is 52.3%, we are concerned about the noise in our performance measure and for some of our analyses later on, we will also employ a subsample of pay-cutting firms that, by construction, underperform their industry peers. The fact that pay-cutting firms have poor stock returns, but relatively good accounting performance, implies that the stock market attributes low potential for growth to the firm. The median sample firm is quite large with annual sales of $1.1 billion and the market value of total assets of $2.4 billion. The median M/B ratio is 2.1, and the median volatility is 0.42. Turning to governance measures, Democracy takes the value of one if the value of the G-index is less than the median of ExecuComp companies (9), and zero otherwise. 43% of the sample firms have good governance as measured by Democracy, and the median Institutional Ownership is 64%. The median firm has sizeable capital expenditures of 4.7% of total assets, which is about one percentage point higher than the industry median, while the median R&D is close to zero. The median Book Leverage is 22%; the median Market Leverage is 14%. The median CEO age in the year before the pay cut is 55. About 5% of 108  the sample CEOs are at the retirement age (between 63 and 65 years old). Finally, about three quarters of the pay cut CEOs remain on their posts after the pay cut (last sub-panel of Panel B). Figure 4.1 presents the trend in CEO pay from three years before (year -3) to three years after the pay cut (year +3). In Panel A, we present the raw levels of our three pay measures. We show that CEO Totalpay increases from year -3 to -2 followed by a drop in year -1 and a further bigger drop in the pay cut year, and then the compensation level goes back up in year +1 and stays stable subsequently.29 It is clear that Totalpay and Equitypay move in tandem, while Cashpay is stable over the entire seven-year period we cover.30 The plot in Panel A could be misleading if firm, industry, and CEO characteristics justify the high level of pay before the pay cut. As a result, we also compute a measure of abnormal pay, which is the difference between the actual level of pay and the “normal” level of pay implied by firm, industry, and executive characteristics. Using the CEO population in ExecuComp, we estimate our benchmark model for expected compensation following prior research in this area (for example, Core, Holthausen, and Larcker (1999), and Murphy (1999)): Ln( Pay )it = a0 + a1Stockreturnit + a2 ROAit + a3 Firmsizeit + a4Volatilityit + a5CEOtenureit + Year Fixed Effects + Industry Fixed Effects + ε it ,  (4.1)  where i indexes firm and t indexes year. The estimated residual, actual Ln(Pay) − predicted Ln(Pay), is our measure of abnormal pay. The industry classification is from Fama and French (1997). Figure 4.1, Panel B presents the trend of abnormal CEO pay. Panel B shows a very similar pattern to that of Panel A: Abnormal CEO pay 29  While it appears that one could say that the pay cut event started in year -1, this is an artifact of our requirement that pay not increase substantially in year -1 to avoid mechanical reversals. If we drop that requirement, the graph does not show a decrease in year -1. 30 As a robustness check, we also plot CEO pay based on a subsample of 297 CEOs with the same CEO over the entire seven-year period. The pattern in pay remains the same. 109  increases from year -3 to -2 followed by a continued decline between year -2 and the pay cut year, a reversal in year +1, and then stays stable afterwards. During the three years prior to the pay cut, the average CEO is receiving sizeable abnormal pay. However, in the period after the pay cut, his pay is pulled back to normal (the level of abnormal pay is down to zero). As before, the change in abnormal Equitypay contributes the most to the fluctuation of abnormal Totalpay. Panel C presents the number of stock and options grants for the three years before and three years after the pay cut. It is clear that the change in pay is not driven by the ups and downs in the stock price, but primarily driven by the change in the units of stock and options granted. Panel C also suggests an important difference between a pay cut and a decline in pay. The former is expected to an overt penalty imposed by the board to the managers in response to poor firm performance; the latter is a passive outcome of fluctuating stock price. Panel C indicates that a pay cut, largely driven by a reduction in units of stock and option grant, is different from a simple decline in pay. Figure 4.2 presents firm performance over the seven-year period surrounding the pay cut. Panel A presents the stock performance measures, while Panel B presents the operating performance measures. We show that stock return plunges in the year before and the year of the pay cut. The raw stock return is about −5% and −8% in year -1 and the pay cut year, respectively. After adjusting for industry median stock returns, the underperformance right before the pay cut is even more striking. The industry-adjusted stock return is around −19% in year -1 and −22% in the pay cut year, suggesting that the poor stock price performance is more likely to be firm-specific instead of due to negative industry-wide shocks. After the pay cut, stock return rises back to about 10% per year and stays the same in subsequent years. Panel A shows that the pick-up in stock return for pay-cutting companies still does not match the performance of their industry peers. The 110  temporal pattern in accounting performance measures mirrors that of the stock market performance, but at a much smaller magnitude.  4.4. Causes and Consequences of CEO Pay Cuts 4.4.1. Why Cut Pay? In this section, we examine why the board decides to cut its CEO’s pay. Our predictions are that pay cuts follow abnormally poor performance and that for a given level of performance pay cuts are more likely when governance mechanisms are stronger and/or when it is relatively easy to replace a CEO. We estimate the following logit regression:  Pr( Paycutit ) = F ( Abretit −1 , Abroait −1 , Corporategovernanceit −1 , Abretit −1 × Corporategovernanceit −1 , Abroait −1 × Corporategovernanceit −1 , Largefirmit −1 , Abretit −1 × Largefirmit −1 , Abroait −1 × Largefirmit −1 , Indretit −1 ,  (4.2)  M / Bit −1 ,Volatilityit −1 , Ln(Totalpay)it −1 , Year Fixed Effects, Firm Fixed Effects). The dependent variable, Paycutit , takes the value of one if a CEO pay cut occurs in firm i and in year t, and zero otherwise. The function F denotes the logit cumulative distribution function. The independent variables include the industry-adjusted stock return, industry-adjusted ROA, different measures of corporate governance alone and interacted with performance, an indicator variable for firms in the highest quartile of sales and interacted with performance, industry median stock return, market-to-book ratio, stock return volatility, and CEO total pay. We also include year fixed effects to account for the time trend, and firm fixed effects to control for the unobserved firm heterogeneity. All the independent variables are measured in year -1. The sample for estimation consists of firms that have CEO compensation data in the ExecuComp database and whose CEOs have kept their job for at least three years. Table 4.3 presents the results. Table 4.3, Panel A Column (1) shows that poor stock performance strongly predicts CEO pay cuts. The coefficient on Abret is significantly negative across all model 111  specifications. In the population of ExecuComp CEOs, 4,629 firm-year observations have Abret less than −15%, and 595 of them experience extreme pay cuts. Put differently, the CEO of a firm with an abnormal return below −15% has a 13% (= 595/4,629) probability of seeing his pay cut by 25% or more. Although extant literature such as Garvey and Milbourn (2006) has found asymmetry in pay for performance, we show that there is some probability of a large reduction in pay below a certain level of performance. This nonlinear relation mitigates, but does not eliminate, the overall finding of lower sensitivity of pay to performance for poor performance. We do not find that the board’s pay-cutting decision is driven by firm operating performance.31 This is not surprising if pay cuts are due to pressure from shareholders, and shareholders rationally focus on stock returns. Similarly, Gilson and Vetsuypens (1993) find that earnings performance does not explain cross-sectional variation in CEO pay when firms are financially distressed. Further, the coefficient on Largefirm the indicator variable identifying the largest firms is positive and significant, suggesting that controlling for performance, CEOs of large firms are more likely to get a pay cut. The coefficient on Indret is negative and significant, suggesting that firms in poorly performing industries are more likely to cut their CEOs’ pay, similar in spirit to the findings in Jenter and Kanaan (2008) on CEO turnover. There is overwhelming evidence that CEO compensation increases when the sector performs well (usually referred to as pay-for-luck). This fact is often interpreted as evidence in support of the managerial power hypothesis, which argues that powerful CEOs have exerted undue influence on the pay process in their favor (Bebchuk and Fried (2003)). Our evidence suggests that the above view is at least incomplete, as CEOs also suffer major pay cuts when their sector performs poorly. That is, we have provided evidence on “pay cut for bad 31  Using other measures of accounting performance, such as net income normalized by total assets, gives the same result. 112  luck,” complementing the literature on pay for luck. Similar to our conclusions for asymmetric pay for performance, we conclude that the probability of an extreme, non-linear reaction to poor overall performance mitigates the overall asymmetry in pay for luck found in Garvey and Milbourn (2006) and Rajgopal, Shevlin, and Zamora (2006). CEOs with high compensation in the prior year are also more likely to be subject to pay cuts: The coefficient on Ln(Totalpay) is significantly positive at the 1% level. This result, while unsurprising, indicates that boards are more likely to employ a pay cut when pay is high to begin with. In the face of poor performance, activist shareholders are more likely to target overpaid CEOs for pay cuts, as evidenced by the public call for massive pay cuts to CEOs of financial companies during the most recent crisis. We note that our sample selection procedure ensures that these pay cuts are not simply reversals of pay spikes. Further, these CEOs are not all being overpaid relative to their peers prior to the pay cut. Slightly less than half (40%) of our sample CEOs have pay at or below the level predicted by a model of normal pay in the year prior to the pay cut. We predict that companies with stronger governance will be more likely to use pay cuts in response to poor performance. There is extensive literature that documents positive association between effective corporate governance mechanisms and shareholder value. Gompers, Ishii, and Metrick (2003) construct the “G-index” to measure governance from the perspective of firm-level anti-takeover protection. They show that better governed firms (which they call firms with “democracy”) have higher firm value and better performance.32 Davila and Penalva (2006) show that firms with strong governance structures as measured by the G-index design compensation contracts that emphasize stock performance over accounting performance.  32 Using Gompers et al.’s (2003) definition of Democracy (with a cut off value of 5) leads to similar results. However, the reduced variation in democracy (fewer firms qualify) reduces the significance. 113  We also use an alternative measure of governance: institutional ownership. Hartzell and Starks (2003) show that the presence of monitoring institutional investors is positively associated with higher CEO pay-for-performance sensitivity and negatively associated with the level of CEO pay. Examining acquisition decisions, Chen, Harford, and Li (2007) find that the presence of long-term institutions increases monitoring. Table 4.3, Panel A Columns (2) through (4) show that better governed firms are more likely to use pay cuts in response to poor performance. Specifically, the coefficient on the interaction term, Abret×Democracy, is negative and significant at the 10% level. We continue to find that pay cuts are based on stock performance rather than operating performance; when we interact Abroa with Democracy in Columns (3) and (4), we find that its coefficient is positive but insignificant. In Panel B, we replace Democracy with Institutional Ownership as an alternative measure for corporate governance, and obtain very similar results. The coefficient on the interaction term, Abret×Institutional Ownership, is negative and significant at the 10% level. Consistent with Hartzell and Starks’ (2003) findings on the monitoring role of institutional ownership, we show that the sensitivity of a pay cut to poor performance is increasing in the presence of greater institutional ownership.33 We also predict that pay cuts will be used when it is harder to replace the CEO. Our simple proxy for these situations is when the firm is one of the largest firms in the economy. In Column (2), the coefficient on the interaction of the firm’s abnormal return and the large firm indicator variable, Abret×Largefirm, is negative and significant at the 5% level, suggesting that large firms are twice as likely to use a pay cut in response to poor performance as are smaller firms. In summary, Table 4.3 shows that a pay cut is more likely to occur when the firm 33  Using institutional ownership by the top five institutional shareholders provides similar results. 114  performs poorly, and the use of a major pay cut to discipline poorly performing CEOs is more likely to happen in firms with strong governance structures and is less likely to happen if it is difficult to find a replacement CEO. These findings are consistent with our first two hypotheses (H1 and H2). On the other hand, we also show that when the firm’s industry performs poorly, and when the CEO was receiving high pay before the cut, the pay cut is more likely to take place. The former finding does not support our third hypothesis (H3) on the use of relative performance evaluation in pay cuts.  4.4.2. CEO Retention after a Pay Cut What happens to the career path of a CEO experiencing a major cut in pay? The pay cut could be a substitute for firing the CEO, or it could be a first step towards the eventual dismissal. By imposing a pay cut, a board signals its willingness and ability to take action against a poor-performing CEO. Thus, while it is clearly a substitute in the year of the pay cut, it likely does not eliminate the threat of dismissal should performance fail to improve. In this section, we examine whether the incumbent CEO keeps his job in the year following the pay cut. In particular, we estimate the following logit model: Pr( Retentionit +1 ) = F ( Paycutit , Abretit , Abroait , Paycutit × Abretit , Paycutit × Abroait , Corporategovernanceit , Paycutit × Corporategovernanceit , Largefirmit , Paycutit × Largefirmit , Indretit , M / Bit ,Volatilityit , Bookleverageit , Retirementit ,  (4.3)  Year Fixed Effects, Firm Fixed Effects). The dependent variable, Retention, equals one if the incumbent CEO remains in office in year +1, and zero otherwise. 775 CEOs out of our sample of 1,061 instances of pay cuts remain on their post in year +1.34 The key independent variable, Paycut, takes the value of one if the CEO receives a pay cut in year t, and zero otherwise. We also control for firm and CEO characteristics, and year as well as firm fixed effects. In particular, we  34  Four pay-cut CEOs join other ExecuComp firms as their CEOs in year +1 and are not counted as CEO retention in our analysis. 115  account for voluntary turnover by including an indicator variable, Retirement, which takes the value of one if the CEO is between 63 to 65 years old, and zero otherwise. The sample for estimation consists of firms that have CEO compensation data in the ExecuComp database and whose CEOs have kept their job for at least three years. Table 4.4 presents the results. In Table 4.4 Column (1), the coefficient on Paycut is −0.45 and is significant at the 1% level. This result indicates that CEOs who have received a pay cut are less likely to remain in office in the subsequent year. In a different setting, Chidambaran and Prabhala (2003) show that repricing companies have abnormally high CEO attrition rates. The coefficients on Abret and Abroa are significant and positive, suggesting that CEOs with good (industry-adjusted) performance are more likely to keep their jobs. The significant result on Abroa also suggests that when boards make the CEO retention decisions, they consider additional measures of performance other than just stock returns. The coefficient on Indret is negative but insignificant, indicating that shocks to industry have no significant impact on CEO retention decisions. This result is consistent with the theory on relative performance evaluation that CEO retention decisions should not be based on factors that are out of the CEOs’ control. Further, consistent with DeFond and Park (1999) we find that CEOs in firms with volatile stock returns are less likely to be retained. DeFond and Park (1999) suggest that firms with volatile performance are more likely to experience the severe negative shocks that lead to CEO turnover. As expected, CEO turnover is higher at retirement age. In Column (2), we include the interaction terms Paycut×Abret and Paycut×Abroa. None of the coefficients on these interaction terms is significantly different from zero, suggesting that a pay cut does not change the subsequent sensitivity of retention to firm performance. That is, while the board decides that following the initial performance decline, 116  it will cut the CEO’s pay severely rather than fire him, it is only a one-year reprieve; in the following year, for a given level of performance, he is just as likely to be fired as another CEO. Combined with the negative coefficient on Paycut, a CEO is less likely to be retained following a pay cut. In Columns (3) and (4), we investigate the role of corporate governance in CEO retention decisions by successively including Democracy and Institutional Ownership and their interactions with Paycut. The coefficient on Institutional Ownership is positive and significant, showing that, controlling for performance, CEOs of firms with greater institutional ownership are more likely to be retained. This effect is canceled if the CEO’s pay has been cut. In contrast, Democracy has no discernable effect on CEO retention post pay cuts. In Column (5), we investigate whether the potential difficulty of replacing the CEO has any impact on retention decisions by including Largefirm and its interaction with Paycut. The coefficient on the interaction term, Paycut×Largefirm, is positive and significant at the 10% level, suggesting that CEOs of large firms are more likely to be retained after pay cuts. In Column (6), we introduce a new indicator variable Underperforming Paycut that takes the value of one if the firm makes a pay cut and its stock return underperforms its industry median in year -1, and zero otherwise. The Underperforming Paycut variable captures the subsample of pay-cutting firms that experience poor stock performance relative to their industry peers before the pay cut (687 of them from our pay cut sample). The results are consistent with the specifications in Columns (1)-(3) and (5) using the full sample. In Column (7), we add Institutional Ownership and interaction with Paycut. We show that  institutional ownership strengthen the link between underperformance and  CEO turnover: the coefficient on the interaction term, Paycut × Institutional Ownership, is 117  negative and significant at the 10% level. Overall, Table 4.4 reveals that a pay cut appears to be the first step towards CEO dismissal. While the pay cut is a substitute for dismissal in that year, the CEO is not yet secure. All else equal, the board is more likely to dismiss the CEO in the following year. This suggests that the initial pay cut demonstrates an activist disposition on the part of the board.  4.4.3. Firm Performance after a Pay Cut In this section, we examine firm performance after the pay cut. Table 4.2 and Figure 4.2 reveal that, on average, there is a significant rebound in stock returns in the year following the pay cut. In particular, from the pay cut year to year +1, the median stock return increases from −7.8% to 11.1%. This contributes to an increase in the industry-adjusted stock return from −21.9% to −6.2%. Accounting performance improves as well during the same period. Figure 4.2 shows that after the improvement, firm performance remains stable out to at least year +3 relative to the pay cut year. To further investigate the effect of a pay cut on subsequent firm performance, we estimate the following regression to explain the change in performance in the year following the pay cut: ΔPerformanceit +1 = β 0 + β1 Paycutit + β 2 Firmsizeit + β3 M / Bit + β 4Volatilityit + β5 Bookleverageit + β 6Turnoverit +1 + Year Fixed Effects + Firm Fixed Effects + ε it .  (4.4)  The dependent variable is the change in firm performance from the pay cut year to year +1.35 Our variable of interest is the indicator variable Paycut. We add one new control variable, Turnover, which takes the value of one if the incumbent CEO leaves office in year +1, and zero otherwise. Table 4.5 presents the results. 35  It is worth noting that our performance measures in the year after the pay cut are not driven by bankruptcy filings or acquisitions as only one percent of our sample are lost from Compustat (and less than ten percent of our sample are lost from ExecuComp). 118  In Column (1) where the dependent variable, △Abret, is the change in the industry-adjusted stock return from the pay cut year to year +1, the coefficient on Paycut is 0.14 and is significant at the 1% level. The effect is also economically significant: For a pay-cutting company, △Abret increases by 14 percentage points compared to the population median of zero. The coefficient on Turnover is positive and significant, indicating that CEO turnover is also associated with improvement in firm performance. In Column (2), instead of △Abret, we use △Abroa, the change in the industry-adjusted ROA from the pay cut year to year +1, as the dependent variable. The coefficient on Paycut is 0.013 and is significant at the 1% level, indicating a pay-cutting company experiences an increase in △Abroa by 1.3 percentage points relative to the population median of zero. We again analyze the subsample of pay-cutting firms that are underperforming their industry peers at the time of the pay cut. Specifically, we estimate the change in performance regressions replacing the Paycut indicator variable with Underperforming Paycut indicator variable in the last two columns of Table 4.5. Similar to results reported in Columns (1) and (2), the coefficients on Underperforming Paycut are positive and significant. The pay-cutting firms that underperform their industry peers show a significant performance improvement after the pay cut. In summary, firms that cut their CEOs’ pay on average experience significant performance improvement in the year following the event, consistent with our fourth hypothesis (H4).  4.4.4. Corporate Policies after a Pay Cut In this section, we explore the actions taken by the CEO to effect the performance change. Specifically, we study capital expenditures, R&D expenses, and capital structure 119  decisions after a CEO pay cut.36 Figure 4.3 presents the trend in corporate investment and financing in the seven-year period surrounding the pay cut. Panel A shows that capital expenditures stay constant at around 5% of total assets from year -3 to -1, decrease to 4% in the pay cut year, and further decrease to about 3% in year +3. The industry-adjusted measure gives a similar story, showing that capital expenditures are slightly above the industry norm until the pay cut, at which point they drop down to the industry median level. Panel B shows a small decline in R&D expenses around a pay cut. The average R&D is around 3.8% of total assets before the pay cut, and it shrinks to about 3.6% afterwards.37 The industry-adjusted R&D presents the same pattern; it stays roughly at 1.5% of total assets in the year of pay cut and decreases to about 1.4% of total assets from year +1 to +3.38 The funds freed-up by the declines in capital expenditures and R&D go toward debt reduction. Panel C shows that leverage ratios increase before the pay cut and decrease afterwards. For example, Book Leverage is around 20% in year -3, and rises to 23% in the pay cut year, followed by a gradual decline reaching 20% in year +3. A similar pattern can be found by examining industry-adjusted leverage ratios. Industry-adjusted Book Leverage is close to zero from year -3 to -1, rises to about 2.5% in the pay cut year, and gradually declines to 2% in year +3. Market Leverage shows a similar trend around the pay cut. We further explore what proportion of the decrease in leverage is due to debt reduction versus equity issuance. Panel D shows that the median value of gross debt issuance relative to total assets drops from about 1.6% in year -1 to about 1.0% in year +1; 36 In unreported analysis, we find that the number of employees generally increases after the pay cut. This evidence supports our overall message in the paper that we are not simply observing firms in distress taking distress-mitigation actions and are instead truly capturing managers taking discretionary actions to increase performance after being “told” to shape-up. 37 The median value of R&D is always zero from year -3 to +3 around the pay cut. 38 Most of the industry median R&D is zero; therefore we compute the industry average R&D as the benchmark. 120  in the meantime the median gross debt retirement relative to total assets increases from 1.7% to 2.3%. Over the seven-year period surrounding the pay cut, the amount of gross equity issuance and the amount of equity repurchase relative to total assets are quite stable at 0.6% and 0.2%, respectively. Panel E presents similar decomposition using industry-adjusted measures. Again, we see that debt retirement is rising above the industry level, while both equity issue and retirement are converging to the industry norm after the pay cut. Industry-adjusted debt issuance is always at zero, suggesting that the median pay-cutting firm is issuing as much debt as the industry median. In summary, we conclude that the decrease in leverage is mainly driven by the reduction in debt. We formally examine the changes in corporate policy associated with pay cuts by estimating the following regression: ΔCoporatepolicyit +1 = γ 0 + γ 1 Paycutit + γ 2 Firmsizeit + γ 3 M / Bit + γ 4Volatilityit +γ 5 Stockreturnit + γ 6Turnoverit +1 + Year Fixed Effects + Firm Fixed Effects + ε it .  (4.5)  The results on capital expenditures surrounding a pay cut are reported in Table 4.6. We examine the change in capital expenditures subsequent to a pay cut in Column (1). The coefficient on Paycut is −0.30 and is significant at the 1% level. Given that the median growth rate in capital expenditures is actually −0.04%, a slight decline, we show that pay-cutting firms further reduce capital expenditures by 0.30 percentage points. Column (2) confirms the result for the underperforming subsample. We conclude that after a pay cut, firms tend to invest less. The findings in Titman, Wei, and Xie (2004) on poor returns following capital expenditure increases, as well as the turnaround in performance documented in our sample, suggest that the decision to curtail investment is probably value enhancing. In Table 4.7, we conduct regression analysis on the effect of pay cuts on R&D expenses. The dependent variable is the change in R&D from the pay cut year to year +1. 121  The coefficients on both Paycut and Underperforming Paycut are −0.15 and are significant at the 1% level, indicating that the company that makes a pay cut experiences less R&D growth in the subsequent year. Given that the sample average growth rate in R&D is 0%, pay-cutting firms then reduce R&D by 0.15 percentage points. Table 4.8 reports the regression results using capital structure measures as the dependent variables. The dependent variable in Column (1) is the change in Book Leverage from the pay cut year to year +1. The coefficient on Paycut is −1.07 and is significant at the 1% level. Thus, relative to non-pay-cutting peers, firms cutting their CEOs’ pay, have a subsequent negative growth in leverage of 1.1 percentage points on average. Using the change in Market Leverage as the dependent variable in Column (2) yields a similar result. We further confirm the capital structure results on the underperforming subsample. Consistent with Baker and Wurgler (2002) who show that higher stock returns are positively associated with the greater use of equity, we find significantly negative coefficients on Stockreturn when different leverage measures are used as dependent variables. Dierkens (1991) argues that higher stock volatility, a proxy for greater asymmetric information, should be positively associated with the use of debt. However, the empirical evidence about the relation between stock return volatility and financing decisions is quite mixed (Dierkens (1991), Denis (1994), and Jung, Kim, and Stulz (1996)). We find that firms with high return volatility tend to use less debt. In summary, after a pay cut the CEO changes his firm’s capital structure by reducing debt.  4.4.5. CEO Pay Recovery after a Pay Cut Finally, we examine the effect of performance improvement and changes in corporate policies on CEO pay going forward. We hypothesize that extreme pay cuts can be effective in providing incentives if there is an implicit or explicit promise of restored pay 122  should the CEO improve performance (H5). We investigate the CEO’s pay in the year following the pay cut by estimating the following regression model: ΔPayit +1 = δ 0 + δ1 Abretit + δ 2 Abroait + δ3 Indretit + δ 4 Firmsizeit + δ5 M / Bit +δ 6Volatilityit + δ 7CEOageit + Year Fixed Effects + Firm Fixed Effects + ε it .  (4.6)  The sample for estimation consists of 689 instances of pay cuts where the same CEO remains on his post in the year after the pay cut.39 The dependent variable is △Pay, the percentage change in CEO total pay from the pay cut year to year +1. The coefficient on Abret is 0.33 and is statistically significant at the 10% level. In terms of economic significance, a one-standard-deviation increase in Abret is associated with an increase in  △Pay by 0.15 (= 0.445×0.33), relative to the sample median of 0.30. Column (2) presents the similar regression result based on the subsample of pay cuts where the pay-cutting firms have poorer stock performance than their industry peers prior to the pay cut. The coefficient on Abret is 0.29 and is significant at the 10% level. In both specifications, the coefficient on either Abroa or Indret is not significantly different from zero. This result suggests that the recovery in CEO pay is mainly driven by firm-specific stock performance over operating performance or industry-wide factors. In summary, CEO pay increases in tandem with improved firm performance after the pay cut, consistent with our final hypothesis (H5). In untabulated results, we examine the effect of CEO turnover on pay changes. We find that if the board turns to a new CEO, his pay is much higher, consistent with a premium for being brought in for a turnaround and with the fact that the original poor performance is not his fault.  39  The requirement for our sample firms with the same CEOs that they also have data for all control variables used in Equation (6) leads to the final sample size that is different from 775 reported in Table 4.4 which only require information from ExecuComp in year +1. 123  4.5. Additional Investigation 4.5.1. Pay Change for Top Management Team So far, our analysis focuses on CEO pay cuts. Chidambaran and Prabhala (2003) show that in their sample of option repricers, a significant 40% of repricings does not include CEOs. They interpret their finding as companies use repricing to deal with within-management incentive distortions created by negative return shocks. To further our understanding of how the board uses changes in compensation to manage the executives’ incentives, we examine how widespread the pay cut is within the senior management ranks. The ExecuComp database includes compensation information for up to the five highest paid managers. After excluding the CEO, we compute the median pay for members of the senior management team in each year from three years before to three years after the pay cut. Figure 4.4 presents the trend in top management pay over the seven-year period surrounding the CEO pay cut. Both Totalpay and Equitypay drop in the pay cut year. For example, the median Totalpay for top executives is around $1.2 million between year -3 to year -1. It goes down to $900,000 in the pay cut year, goes back up to about $1.2 million in year +1, and stays constant up to year +3. Similar to our findings on the cash pay of CEOs, the cash pay of top management team stays relatively stable over the entire seven-year period. We conclude that when the CEO receives a pay cut, the whole senior management team is subject to similar cut in their pay as well.  4.5.2. Controlling for Business Cycle Table 4.1 Panel A revealed a clustering of pay cuts during the market decline and recession in 2000 and 2001. Our base specifications control for industry performance and year fixed effects, but we are also interested in specifically isolating the effect of a 124  market-wide downturn on pay cuts. The clustering of pay cuts during a downturn suggests a general lack of relative performance evaluation with respect to economy-wide performance. Also, one could expect the actions taken by a CEO responding to a pay cut during a downturn to be different from those taken outside a downturn. Finally, we will be able to determine whether our results are driven by the downturn subsample or are more general. We add an indicator variable, Downturn, which is set equal to one if the firm-year observations are in 2000 or 2001, and zero otherwise, to our specifications on its own and interacted with the appropriate variables. We use this variable to investigate whether the overall market downturn during our sample period influences the CEO pay cut and retention decisions, subsequent performance changes, and CEO pay recovery (results available upon request). We find that during the period of overall market downturn, the board is less likely to cut its CEO’s pay in response to poor performance. Thus, the CEO benefits from some form of relative performance evaluation. Nonetheless, we note our finding in Table 4.3 that poor industry performance still is significantly and positively associated with subsequent pay cuts. Further, the board is less likely to dismiss its CEO in the period of overall stock market downturn. However, we also find that the negative association between pay cut and CEO retention is not moderated by the overall stock market performance. Thus, once the board has enacted a pay cut, it is just as likely to follow-up with a dismissal inside a downturn as it is outside a downturn. As for CEO actions following a pay cut in a downturn, we find that the CEO reduces leverage, but not investment, more drastically after a pay cut in the downturn period than after a pay cut outside that period. At the same time, we find that CEOs going through pay cuts during market downturns experience more negative growth in pay. This 125  result is consistent with the outside opportunities theory of CEO pay, that is, when the opportunity cost of CEO pay worsens during downturns, the extent of their pay recovery suffers.  4.5.3. Controlling for Initial Abnormal Pay In our sample of 1,061 instances of pay cuts, 602 (about 60%) of these CEOs receive higher compensation in the year prior to the pay cut than their normal level of pay based on Equation (4.1). We are interested in whether a pay cut from abnormally high pay has differential effects on CEOs than a pay cut relative to normal pay. There are a number of reasons why a difference would exist. CEOs earning abnormal pay may have perceived themselves to be strong relative to their boards and the pay cut would indicate a substantial shift in bargaining power. Further, higher compensation levels should be associated with greater sensitivity of pay to performance, so for a given level of poor performance, highly paid CEOs should be more likely to have their pay cut. Finally, pay cuts relative to normal pay are inherently more punitive than pay cuts relative to abnormally high pay. We construct an indicator variable Highpay, which takes the value of one if the CEO receives higher-than-predicted pay in the year prior to the pay cut, and zero otherwise. Including this indicator variable and interacting it with appropriate other variables, we find that CEOs receiving abnormally high pay are more likely subject to a pay cut, and companies are more likely to cut their CEOs’ pay in response to poor performance when their CEOs are over paid in the first place. Our results are consistent with the standard contract theory which suggests that high compensation level should be associated with high sensitivity to performance (Holmström and Milgrom (1987)). We also examine the relation between Highpay and CEO retention decisions by adding Highpay and its interaction with Paycut in Equation (4.3). We find that a CEO who receives abnormally high compensation prior to the pay cut is as likely to be dismissed as other CEOs. 126  While a highly paid CEO does not curtail leverage or investment differently from other CEOs following a pay cut, his firm experiences a smaller improvement in stock performance and his pay recovers less. Overall, it appears that a pay cut relative to a normal level of pay imposes more pressure on the CEO than a pay cut on an abnormally high level of pay.  4.5.4. Results from Performance-based Control Firms Section 4.3 has shown significant performance improvement after a pay cut. However, this result could be due to mean reversion of industry and firm-specific factors. To examine this possibility, we apply the performance-based control group matching method used by Huson, Malatesta, and Parrino (2004). Each pay-cutting firm is matched to comparison firms in the same Fama and French (1997) 48 industries whose stock returns in year -1 are within [90%, 110%] of the sample firm’s stock return. Control-adjusted stock return (ROA) of each sample firm is computed as the difference between its stock return (ROA) and the median stock return (ROA) of its control group. As suggested by Huson et al. (2004), this method can isolate the component of performance change attributable to pay cuts from that due to mean reversion in performance time series. Consistent with Figure 4.2 Panel A, Panel A of Figure 4.5 highlights a clear “V” shape of the control-adjusted stock return. The median control-adjusted return is almost zero from year -3 to -1, it drops to around −14% in the year of the pay cut, and bounces back to above zero during the three years afterwards. The change in control-adjusted stock return from the pay cut year to year +1 is 20%on average and 16% in median; both are significantly different from zero at the 1% level. Figure 4.5 Panel B shows that the time series of control-adjusted ROA is relatively flat. Although the sample firm’s control-adjusted ROA deteriorates from year -3 to year +1 127  and it begins to improve afterwards, the magnitudes of the changes are small. For example, the medians of changes in control-adjusted ROA from the pay cut year to year +1 and from year +1 to year +2 are −0.2% and 0.2%, respectively; neither is significantly different from zero at the 10% level. This finding is consistent with our earlier comment that pay cuts are more likely driven by stock market performance instead of operating performance and as a result, we expect the incentive effect of pay cuts to show up in the former not the latter. Overall, Figure 4.5 indicates that pay cuts are indeed associated with performance turnaround, especially in terms of stock market performance. This turnaround is not simply driven by mean reversion in performance.40  4.5.5. Pay Cut versus CEO Turnover Further, we examine the likelihood that pay-cut CEOs retain their posts up to three years after the event. To avoid double-counting in the cases of multiple pay cuts occurring within the same firm, we only consider the first pay cut of a CEO in our sample and track the CEO’s turnover subsequently. Of the 851 CEOs receiving their first pay cuts during our sample period, 72% of them (619 instances) remain CEOs of the same company in year +1, 54% of them (460 instances) remain up to year +2, and 42% of them (356 instances) remain up to year +3. Among the CEOs of the control sample as constructed in the above section, the corresponding turnover numbers are 81%, 64%, and 47%, respectively. The corresponding  40  In unreported analysis, we separate the control sample into the turnover subsample where the CEO leaves in year 0 and  the “do nothing” subsample. The performance plot of the turnover sample is characterized by prolonged underperformance (relative to our pay cut sample) prior to the turnover. The pay cut sample performs worse during year 0, probably because the market reacts positively to the turnover in the worse-performing firms, and then quickly rebounds to similar performance. This suggests to us that the pay cut is, on average, an effective choice for those firms, and that firms typically only choose turnover over pay cut when performance has been bad for several years. That is sensible in that they would have then decided that the poor performance is more an ability than effort issue. 128  turnover numbers for the CEO population in ExecuComp during 1994-2005 are 82%, 64%, and 48%, respectively. These numbers show that CEOs are less likely to be retained in the year following a sharp pay cut, but after that, their incremental retention likelihood is normal (the difference in retention rates does not widen in years +2 or +3. This is consistent with our regression finding that pay-cut CEOs are more likely than otherwise similar CEOs to be dismissed in the year immediately following the pay cut. The results here suggest that if they survive past that year, they have sufficiently improved performance to avoid dismissal.  4.5.6. Pay Cut versus Option Repricing As we have noted, a pay cut and option repricing are essentially the opposite responses to poor performance, so it is interesting to compare their efficacy in achieving improved performance. While our results indicate a strong performance turnaround following a pay cut, the existing literature fails to find evidence of improvement in firm performance after an option repricing. For example, Chidambaran and Prabhala (2003) find that industry-adjusted operating performance of repricers is −0.5% two years preceding repricing. In the repricing year, it suddenly falls to –6.7% and remains at about –5% for the next two years. Similar results can be found by using other measures of performance. They claim that “Repricers never regain their historical profitability levels or growth rates” (page 163). Moreover, the frequency of option repricings is much smaller than that of pay cuts. Based on ExecuComp, Chidambaran and Prabhala (2003) only identify 213 repricing cases in the 1992-1997 period (about 36 events per year). In another study, Callaghan, Saly, and Subramnian (2004) identify 236 repricing events during the same period. Compared to our sample of 1,061 instances of pay cuts from 1994 to 2005 (about 88 pay cuts per year), option repricing is clearly less frequent and has almost disappeared since the accounting rule change in 1998, while the phenomenon we study has become more economically important. 129  4.6. Conclusion Creating incentives for managers to exert effort to perform well and to improve poor performance is a complex process. In this paper we study changes to CEOs’ compensation packages that have the potential to create ex ante incentives to exert effort to avoid poor performance as well as ex post incentives to improve poor performance if it is experienced. Specifically, we examine the causes and consequences of sharp pay cuts, an action that has been mostly overlooked in the attention given to overall rising pay levels. We find that poor performance significantly increases a CEO’s chance of having his pay cut sharply and that firm governance is important in establishing this link. The CEO can restore his pay level by reversing the poor performance. Pay cuts are only a short-term substitute for dismissal—a CEO continuing poor performance following a pay cut is just as likely to be dismissed as a non-pay-cutting CEO with similar performance. On average, CEOs respond to the pay cut by curtailing capital expenditures and R&D and allocating funds to reduce leverage. For most firms, performance improves and the CEO’s pay is restored. Compared to option repricing, pay cuts appear to be more effective in improving firm performance.  130  Table 4.1. Sample Distribution The sample consists of 1,061 instances of CEO pay cuts from 1994 to 2005. A CEO experiences a major pay cut and thus is included in our sample if (1) the same CEO keeps his position from year -2 to the pay cut year; (2) his total pay is no more than 75% of his pay in year -1; and (3) his total pay in year -1 is no more than 125% of his pay in year -2. In Panel B, pay cut is defined as one minus Pay (t)/Pay(t-1) where Pay(t) and Pay(t-1) refer to the CEO’s total pay in the pay cut year t and year t-1, respectively.  Panel A: Distribution of CEO Pay Cut by Year Year  Frequency (1)  Percent (2)  Number of firms in ExecuComp (3)  Percentage of firms in ExecuComp (4)  1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Total  13 53 52 60 78 69 100 120 151 165 90 110 1061  1.23% 5.00% 4.90% 5.66% 7.35% 6.50% 9.43% 11.31% 14.23% 15.55% 8.48% 10.37% 100%  1540 1591 1636 1661 1721 1793 1779 1651 1656 1678 1681 1681 20068  7.67% 7.93% 8.15% 8.28% 8.58% 8.93% 8.86% 8.23% 8.25% 8.36% 8.38% 8.38%  131  (1)/(3)  0.84% 3.3% 3.2% 3.6% 4.5% 3.8% 5.6% 7.3% 9.1% 9.8% 5.4% 6.5% 5.3%  Panel B: Distribution of CEO Pay Cut by Size Pay Cut  Frequency  Percent  Cumulative Percent  [25%, 35%) [35%, 45%) [45%, 55%) [55%, 65%) [65%, 75%) [75%, 85%) [85%, 95%) [95%, 100%] Total  358 240 181 121 75 51 26 9 1061  33.74% 22.62% 17.06% 11.40% 7.07% 4.81% 2.45% 0.85% 100%  33.74% 56.36% 73.42% 84.83% 91.89% 96.70% 99.15% 100%  132  Table 4.2. Sample Descriptive Statistics This table reports CEO pay and firm characteristics from one year before the pay cut to one year after. The sample consists of 1,061 instances of CEO pay cuts from 1994 to 2005. Totalpay is the sum of the CEO’s salary, bonuses, long-term incentive plans, the grant-date value of restricted stock awards, and the Black-Scholes value of granted options. Cashpay is the sum of the CEO’s salary, bonus, payouts from long-term incentive plans, and all other cash-based compensation. Equitypay is the value of restricted stock and the Black-Scholes value of stock options. Stockshare is the number of shares in the CEO’s annual stock grants as a percentage of the total number of shares outstanding. Optionshare is the number of shares underlying the CEO’s annual options grants as a percentage of the total number of shares outstanding. Incentiveshare is the sum of Stockshare and Optionshare. ROA is operating income before depreciation over total assets. Abret is the difference between firm stock return and industry median stock return. Abroa is the difference between firm ROA and industry median ROA. The industry classification follows Fama and French (1997) 48 industries. MV Equity is the product of total number of shares outstanding and the fiscal year end closing price. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. Democracy takes the value of one if the value of the G-index as constructed by Gompers et al. (2003), is less than the median of ExecuComp firms (9), and zero otherwise. Institutional Ownership is the number of shares owned by institutional investors as a percentage of the total number of shares outstanding. Capex is capital expenditures over total assets. R&D is research and development expenses over total assets. Book Leverage is the ratio of long-term debt and current debt over total assets. Market Leverage is the ratio of long-term debt and current debt over the market value of total assets. Retirement takes the value of one if the CEO is between 63 and 65 years old, and zero otherwise. Retention takes the value of one if the CEO stays on his job in year +1, and zero otherwise. All dollar values are in 2006 dollars. All continuous variables are winsorized at the 1st and 99th percentiles. Panel A: CEO Pay Mean  Std  5th Pct  Median  95th Pct  5917 1505 4350 0.03% 0.35% 0.35%  8265 1499 7452 0.29% 0.63% 0.46%  736 336 13 0 0 0  2925 1085 1735 0 0.19% 0.22%  21797 4159 18124 0.13% 1.09% 1.13%  2725 1200 1508 0.01% 0.13% 0.14%  3248 1096 2603 0.06% 0.29% 0.24%  394 290 2 0 0 0  1474 853 429 0 0.03% 0.04%  10139 3344 7234 0.05% 0.57% 0.58%  4811 1402 3348 0.03% 0.28% 0.31%  6477 1346 5591 0.11% 0.49% 0.53%  486 310 6 0 0 0  2463 986 1375 0 0.12% 0.14%  17190 3666 13291 0.17% 1.18% 1.29%  Year -1 Totalpay ($K) Cashpay ($K) Equitypay ($K) Stockshare Optionshare Incentiveshare Pay Cut Year Totalpay ($K) Cashpay ($K) Equitypay ($K) Stockshare Optionshare Incentiveshare Year +1 Totalpay ($K) Cashpay ($K) Equitypay ($K) Stockshare Optionshare Incentiveshare  133  Panel B: Firm Characteristics Mean  Std  5th Pct  Median  95th Pct  -1.77% 11.68% -18.51% -0.56% 4047 8524 12491 5587 2.85 0.49 0.43 61% 6.21% 3.8% 23% 17% 55 0.05  44.6% 11.71% 41.08% 10.71% 8297 26538 32450 12854 2.88 0.24 0.49 19% 5.62% 6.8% 19% 16% 7.47 0.23  -67.99% -8.07% -82.43% -18.48% 103 144 255 130 0.53 0.22 0 25% 0.67% 0 0 0 43 0  -4.51% 12.46% -19.05% 0% 1135 1284 2429 1343 2.08 0.42 0 64% 4.71% 0 22% 14% 55 0  78.24% 28.65% 52.33% 14.37% 18211 38797 57325 27822 8.07 0.97 1 89% 18% 18.3% 57% 50% 68 1  -1.37% 9.64% -19.17% -2.29% 4079 9134 12599 5140 2.47 0.51 0.43 61% 5.51% 3.8% 24% 18% 56 0.06  48.61% 12.2% 44.54% 11.46% 8249 28960 34383 11476 2.51 0.25 0.49 20% 5% 6.8% 19% 17% 7.57 0.24  -71.31% -11.72% -82.02% -23.02% 86 139 206 85 0.37 0.21 0 23% 0.64% 0 0 0 44 0  -7.75% 10.61% -21.89% -1.05% 1118 1347 2267 1179 1.85 0.44 0 63% 4.22% 0 23% 15% 56 0  83.01% 26.66% 54.58% 13.11% 19875 39925 56982 24054 7.06 0.99 1 91% 16% 18% 59% 54% 69 1  Year -1 Stockreturn ROA Abret Abroa Sales ($M) BV Total Assets ($M) MV Total Assets ($M) MV Equity ($M) M/B Volatility Democracy Institutional Ownership Capital Expenditures R&D Book Leverage Market Leverage CEO age Retirement Pay Cut Year Stockreturn ROA Abret Abroa Sales ($M) BV Total Assets ($M) MV Total Assets ($M) MV Equity ($M) M/B Volatility Democracy Institutional Ownership Capital Expenditures R&D Book Leverage Market Leverage CEO age Retirement  134  Year +1 Stockreturn ROA Abret Abroa Sales ($M) BV Total Assets ($M) MV Total Assets ($M) MV Equity ($M) M/B Volatility Democracy Institutional Ownership Capital Expenditures R&D Book Leverage Market Leverage CEO age Retirement Retention  20.39% 9.83% 1.21% -2.25% 4301 10053 13923 5884 2.37 0.49 0.42 64% 4.72% 3.7% 23% 18% 55.62 0.06 0.73  67.22% 10.82% 63.34% 10.29% 8413 33401 39202 13494 2.09 0.25 0.49 21% 4.33% 6.7% 19% 16% 7.51 0.24 0.44  135  -62.86% -8.24% -74.11% -19.68% 86 147 221 77 0.36 0.20 0 25% 0.51% 0 0 0 44 0 0  11.06% 10.47% -6.21% -0.96% 1268 1498 2439 1398 1.90 0.44 0 65% 3.38% 0 22% 15% 56 0 1  122.40% 24.74% 98.00% 11.94% 21896 37987 58845 27751 6.26 0.99 1 93% 13.87% 17.6% 57% 50% 68 1 1  Table 4.3. What Predicts a CEO Pay Cut? We run a logit regression using the ExecuComp firms whose CEOs stay in office for at least three years during the period from 1994 to 2005. The dependent variable, Paycut, takes the value of one if the firm makes a CEO pay cut in year t, and zero otherwise. There are 1,061 instances of pay cuts in the estimation sample. Abret is the difference between firm stock return and industry median stock return. Abroa is the difference between firm ROA and industry median ROA. The industry classification follows Fama and French (1997) 48 industries. Democracy takes the value of one if the value of the G-index as constructed by Gompers et al. (2003), is less than the median of ExecuComp firms (9), and zero otherwise. Institutional Ownership is the number of shares owned by institutional investors as a percentage of the total number of shares outstanding. Largefirm takes the value of one if sales is in the top quartile, and zero otherwise. Indret is the industry median annual stock return. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. Ln(Totalpay) is the natural logarithm of CEO’s total annual compensation. Corresponding p-values are reported in brackets. The superscripts ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Panel A: Using Democracy to Measure Corporate Governance Abret Abroa  (1)  (2)  (3)  (4)  -0.86*** [0.000] -0.35 [0.55]  -0.65*** [0.000] 0.12 [0.87] 0.06 [0.79] -0.38* [0.09]  -0.95*** [0.000] 0.16 [0.86] 0.13 [0.59]  -0.64*** [0.000] -0.16 [0.87] 0.12 [0.63] -0.41* [0.09] 0.97 [0.48] 0.31 [0.31] -0.56* [0.06] -1.36 [0.54] -1.69*** [0.000] -0.014 [0.59] 0.12 [0.87] 0.65*** [0.000] Yes Yes 9013 9.3%  Democracy Abret × Democracy Abroa × Democracy Largefirm  0.54** [0.02]  Abret × Largefirm  0.41* [0.1] -0.59** [0.05]  Abroa × Largefirm Indret M/B Volatility Ln(Totalpay) Year Fixed Effects Firm Fixed Effects No. of Obs. Pseudo-R2  -1.52*** [0.000] -0.007 [0.74] 0.41 [0.48] 0.59*** [0.000] Yes Yes 11189 8.5%  -1.68*** [0.000] -0.015 [0.54] 0.11 [0.88] 0.64*** [0.000] Yes Yes 9013 9.3% 136  0.59 [0.67] 0.33 [0.27]  -1.87 [0.39] -1.66*** [0.000] -0.015 [0.54] 0.17 [0.82] 0.64*** [0.000] Yes Yes 9013 9.1%  Panel B: Using Institutional Ownership to Measure Corporate Governance Abret Abroa  (1)  (2)  (3)  (4)  -0.86*** [0.000] -0.35 [0.55]  -0.28 [0.33] -0.42 [0.47] 0.18 [0.66] -0.84* [0.07]  -0.86*** [0.000] 0.82 [0.56] 0.07 [0.89]  -0.29 [0.31] 0.54 [0.69] -0.006 [0.99] -0.81* [0.08] -1.78 [0.45] 0.49* [0.08] -0.53* [0.06] 0.04 [0.98] -1.56*** [0.000] -0.008 [0.68] 0.34 [0.55] 0.58*** [0.000] Yes Yes 11189 8.8%  Institutional Ownership Abret × Institutional Ownership Abroa × Institutional Ownership Largefirm  0.54** [0.02]  Abret × Largefirm  0.49** [0.04] -0.52* [0.06]  Abroa × Largefirm  -2.16 [0.36] 0.49* [0.08]  Indret  -1.52*** [0.000]  -1.56*** [0.000]  -0.62 [0.77] -1.53*** [0.000]  M/B  -0.007 [0.74] 0.41 [0.48] 0.59*** [0.000] Yes Yes 11189 8.5%  -0.009 [0.67] 0.34 [0.55] 0.58*** [0.000] Yes Yes 11189 8.7%  -0.006 [0.75] 0.41 [0.48] 0.58*** [0.000] Yes Yes 11189 8.5%  Volatility Ln(Totalpay) Year Fixed Effects Firm Fixed Effects No. of Obs. Pseudo-R2  137  Table 4.4. CEO Retention After a Pay Cut We run a logit regression using the ExecuComp firms whose CEOs stay in office for at least three years during the period from 1994 to 2005. The dependent variable, Retention, takes the value of one if the CEO remains in position in year +1, and zero otherwise. Paycut takes the value of one if the firm makes a pay cut in year t, and zero otherwise. Underperforming Paycut takes the value of one if the firm makes a pay cut and its stock return is less than the ExecuComp industry median return in year -1, and zero otherwise. Abret is the difference between firm stock return and industry median stock return. Abroa is the difference between firm ROA and industry median ROA. The industry classification follows Fama and French (1997) 48 industries. Democracy takes the value of one if the value of the G-index as constructed by Gompers et al. (2003), is less than the median of ExecuComp firms (9), and zero otherwise. Institutional Ownership is the number of shares owned by institutional investors as a percentage of the total number of shares outstanding. Largefirm takes the value of one if sales is in the top quartile, and zero otherwise. Indret is the industry median annual stock return. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. Book Leverage is the ratio of long-term debt and current debt over total assets. Retirement takes the value of one if the CEO is between 63 and 65 years old, and zero otherwise. Corresponding p-values are reported in brackets. The superscripts ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Paycut Underperforming Paycut Abret Abroa  (1)  (2)  (3)  (4)  (5)  -0.45*** [0.000]  -0.46*** [0.001]  -0.56*** [0.000]  -0.38** [0.017]  0.078 [0.82]  0.25*** [0.002] 2.46*** [0.000]  Paycut × Abret Paycut × Abroa  0.25*** [0.004] 2.47*** [0.000] -0.004 [0.98] -0.055 [0.94]  Democracy  0.22** [0.019] 2.20*** [0.000]  0.23*** [0.006] 2.39*** [0.000]  (7)  -0.52*** [0.000] 0.26*** [0.001] 2.41*** [0.000]  0.09 [0.81] 0.24*** [0.003] 2.33*** [0.000]  -0.21 [0.20] -0.15 [0.49]  Paycut× Democracy Institutional Ownership Paycut×Institutional Ownership Underperforming Paycut× Institutional Ownership Largefirm  0.25*** [0.002] 2.43*** [0.000]  (6)  1.06*** [0.006]  1.02*** [0.008] -0.84* [0.10]  -0.97* [0.09] -0.23 [0.19]  -0.23 [0.19]  -0.26 [0.14]  138  -0.19 [0.33]  -0.22 [0.23]  -0.23 [0.19]  -0.22 [0.23]  (1)  (2)  (3)  (4)  (5)  (6)  (7)  -0.21 [0.33] -0.025 [0.16] -1.41*** [0.006] 0.79 [0.15] -1.24*** [0.000] Yes Yes 11167 9.2%  0.38* [0.08] -0.24 [0.23] -0.008 [0.63] -1.12** [0.012] 0.03 [0.95] -1.23*** [0.000] Yes Yes 11167 9.1%  -0.19 [0.33] -0.006 [0.72] -1.32*** [0.003] -0.06 [0.91] -1.22*** [0.000] Yes Yes 11167 9%  -0.23 [0.25] -0.009 [0.59] -1.16*** [0.01] 0.02 [0.96] -1.22*** [0.000] Yes Yes 11167 9.2%  Paycut × Largefirm Indret M/B Volatility Book Leverage Retirement Year Fixed Effects Firm Fixed Effects No. of Obs. Pseudo-R2  -0.21 [0.29] -0.004 [0.79] -1.29*** [0.003] -0.06 [0.91] -1.22*** [0.000] Yes Yes 11167 9%  -0.21 [0.29] -0.004 [0.79] -1.29*** [0.003] -0.06 [0.91] -1.22*** [0.000] Yes Yes 11167 9%  -0.21 [0.29] -0.004 [0.81] -1.31*** [0.003] -0.07 [0.89] -1.22*** [0.000] Yes Yes 9127 8.8%  139  Table 4.5. Firm Performance After a CEO Pay Cut This regression is based on the ExecuComp firms whose CEOs stay in office for at least three years during the period from 1994 to 2005. Abret is the difference between firm stock return and industry median stock return. Abroa is the difference between firm ROA and industry median ROA. The industry classification follows Fama and French (1997) 48 industries. △Abret is defined as Abret(t+1) − Abret(t). △Abroa is defined as Abroa(t+1) − Abroa(t). Paycut takes the value of one if the firm makes a pay cut in year t, and zero otherwise. Underperforming Paycut takes the value of one if the firm makes a pay cut and its stock return is less than the ExecuComp industry median return in year -1, and zero otherwise. Firmsize is the natural logarithm of sales. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. Book Leverage is the ratio of long-term debt and current debt over total assets. Turnover takes the value of one if the CEO is replaced between the pay cut year and year +1, and zero otherwise. Corresponding p-values are reported in brackets. The superscripts ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Paycut  (1) △Abret  (2) △Abroa  0.14*** [0.000]  0.013*** [0.000]  Underperforming Paycut Firmsize M/B Volatility Book Leverage Turnover Constant Year Fixed Effects Firm Fixed Effects No. of Obs. Adjusted-R2  -0.15*** [0.000] -0.11*** [0.000] -0.001 [0.99] 1.43*** [0.000] 0.058*** [0.002] 2.64*** [0.000] Yes Yes 10466 19%  -0.024*** [0.000] 0.001** [0.017] 0.027** [0.023] 0.21*** [0.000] 0.003 [0.15] 0.37*** [0.000] Yes Yes 10484 6%  140  (3) △Abret  (4) △Abroa  0.11*** [0.000] -0.15*** [0.000] -0.11*** [0.000] 0.002 [0.98] 1.44*** [0.000] 0.06*** [0.001] 2.65*** [0.000] Yes Yes 10466 19.8%  0.018*** [0.000] -0.023*** [0.000] 0.001** [0.015] 0.027** [0.022] 0.21*** [0.000] 0.003 [0.17] 0.36*** [0.000] Yes Yes 10484 6.1%  Table 4.6. Capital Expenditures After a CEO Pay Cut This regression is based on the ExecuComp firms whose CEOs stay in office for at least three years during the period from 1994 to 2005. Capex is capital expenditures over total assets. △Capex is defined as Capex (t+1) − Capex (t), measured in percentage points. Paycut takes the value of one if the firm makes a pay cut in year t, and zero otherwise. Underperforming Paycut takes the value of one if the firm makes a pay cut and its stock return is less than the ExecuComp industry median return in year -1, and zero otherwise. The industry classification follows Fama and French (1997) 48 industries. Firmsize is the natural logarithm of sales. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. Turnover takes the value of one if the CEO is replaced between the pay cut year and year +1, and zero otherwise. Corresponding p-values are reported in brackets. The superscripts ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (2) △Capex  (1) △Capex Paycut  -0.30*** [0.01] -0.35*** [0.004] 0.20* [0.056] -0.013 [0.41] 0.54 [0.24] 0.84*** [0.000] -0.006 [0.95] -4.22* [0.07] Yes Yes 9580 4%  Underperforming Paycut Firmsize M/B Volatility Stockreturn Turnover Constant Year Fixed Effects Firm Fixed Effects No. of Obs. Adjusted-R2  0.17 [0.12] -0.02 [0.15] 0.63 [0.19] 0.84*** [0.000] -0.025 [0.79] -3.61 [0.13] Yes Yes 9580 3.8%  141  Table 4.7. R&D Expenses After a CEO Pay Cut This regression is based on the ExecuComp firms whose CEOs stay in office for at least three years during the period from 1994 to 2005. R&D is research and development expenses over total assets. △R&D is defined as R&D (t+1) − R&D (t), measured in percentage points. Paycut takes the value of one if the firm makes a pay cut in year t, and zero otherwise. Underperforming Paycut takes the value of one if the firm makes a pay cut and its stock return is less than the ExecuComp industry median return in year -1, and zero otherwise. The industry classification follows Fama and French (1997) 48 industries. Firmsize is the natural logarithm of sales. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. Turnover takes the value of one if the CEO is replaced between the pay cut year and year +1, and zero otherwise. Corresponding p-values are reported in brackets. The superscripts ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (2) △R&D  (1) △R&D Paycut  -0.15*** [0.01] -0.15*** [0.008] 0.088* [0.08] -0.044*** [0.000] -0.029 [0.89] -0.083** [0.03] -0.003 [0.94] -1.71 [0.124] Yes Yes 10523 1%  Underperforming Paycut Firmsize M/B Volatility Stockreturn Turnover Constant Year Fixed Effects Firm Fixed Effects No. of Obs. Adjusted-R2  0.079 [0.12] -0.042*** [0.000] 0.027 [0.90] -0.083** [0.03] -0.002 [0.97] -1.48 [0.18] Yes Yes 10523 1%  142  Table 4.8. Capital Structure After a CEO Pay Cut This regression is based on the ExecuComp firms whose CEOs stay in office for at least three years during the period from 1994 to 2005. Book Leverage is the ratio of long-term debt and current debt over total assets. Market Leverage is the ratio of long-term debt and current debt over the market value of total assets. △Book (Market) Leverage is defined as Book (Market) Leverage (t+1) − Book (Market) Leverage (t), measured in percentage points. Paycut takes the value of one if the firm makes a pay cut in year t, and zero otherwise. Underperforming Paycut takes the value of one if the firm makes a pay cut and its stock return is less than the ExecuComp industry median return in year -1, and zero otherwise. The industry classification follows Fama and French (1997) 48 industries. Firmsize is the natural logarithm of sales. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. Turnover takes the value of one if the CEO is replaced between the pay cut year and year +1, and zero otherwise. Corresponding p-values are reported in brackets. The superscripts ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively.  Paycut  (1) △Book Leverage  (2) △Market Leverage  -1.07*** [0.000]  -0.86*** [0.000]  Underperforming Paycut Firmsize M/B Volatility Stockreturn Turnover Constant Year Fixed Effects Firm Fixed Effects No. of Obs. Adjusted-R2  -0.64*** [0.01] -0.14*** [0.000] -6.56*** [0.000] -0.63*** [0.000] 0.096 [0.67] 17.36*** [0.000] Yes Yes 10289 2.7%  0.41* [0.07] 0.12*** [0.000] -9.23*** [0.000] -0.32** [0.05] -0.14 [0.48] -5.24 [0.28] Yes Yes 10289 5.2%  143  (3) △Book Leverage  (4) △Market Leverage  -1.06*** [0.000] -0.63*** [0.01] -0.13*** [0.001] -5.54*** [0.000] -0.64*** [0.001] 0.11 [0.62] 16.64*** [0.002] Yes Yes 10289 2.5%  -0.87*** [0.000] 0.35 [0.11] 0.12*** [0.000] -8.11*** [0.000] -0.33** [0.044] -0.11 [0.57] -4.52 [0.34] Yes Yes 10289 5.1%  Table 4.9. CEO Pay Recovery This table examines the recovery in CEO pay after the pay cut. We limit our sample to CEOs without experiencing turnover after the pay cut. There are 689 CEOs in the estimation sample. The regression is based on a subsample of pay-cutting firms with the same CEOs from year t to t+1. Totalpay is the sum of the CEO’s salary, bonuses, long-term incentive plans, the grant-date value of restricted stock awards, and the Black-Scholes value of granted options. The dependent variable ΔPay is defined as Ln(Totalpayt+1) − Ln(Totalpayt), measuring the change in the CEO’s total pay from year t to t+1. Abret is the difference between firm stock return and industry median stock return. Abroa is the difference between firm ROA and industry median ROA. Indret is the industry median annual stock return. The industry classification follows Fama and French (1997) 48 industries. Firmsize is the natural logarithm of sales. M/B is the ratio of market value of equity over book value of equity. Volatility is the standard deviation of stock returns based on monthly returns over the past 60 months. CEO age is the age of the CEO. Column (2) is based on the subsample where the pay-cutting firm’s stock return is less than the ExecuComp industry median in year -1. Corresponding p-values are reported in brackets.The superscripts ***, ** and * denote statistical significance at the 1%, 5% and 10% level, respectively. (1)  (2)  ΔPay Abret Abroa Indret Firmsize M/B Volatility CEO age Constant Year Fixed Effects Firm Fixed Effects No. of Obs. Adjusted-R2  ΔPay  0.33* [0.09] 0.27 [0.69] 0.79 [0.12] -0.17 [0.38] -0.01 [0.84] -1.56 [0.21] -0.01 [0.89] 5.43 [0.23] Yes Yes 689 21%  0.29* [0.1] 1.28 [0.22] 0.55 [0.42] -0.11 [0.69] -0.03 [0.61] -0.42 [0.76] -0.02 [0.73] 4.31 [0.53] Yes Yes 509 49%  144  Figure 4.1. CEO Pay Around a Pay Cut Panel A: CEO Pay This figure presents the trend in CEO pay over the seven-year period surrounding a pay cut. Totalpay is the sum of the CEO’s salary, bonuses, long-term incentive plans, the grant-date value of restricted stock awards, and the Black-Scholes value of granted options. Cashpay is the sum of the CEO’s salary, bonus, payouts from long-term incentive plans, and all other cash-based compensation. Equitypay is the value of restricted stock and the Black-Scholes value of stock options. Year 0 is the year when the pay cut occurs. The median value of CEO pay measured in thousands of 2006-constant dollars is presented.  4500 4000 3500 3000 2500  Totalpay  2000  Cashpay  1500  Equitypay  1000 500 0 -3  -2  -1  0  1  2  3  Panel B: Abnormal CEO Pay This figure presents the trend in abnormal CEO pay over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. Abnormal CEO pay is the difference between CEO pay and the predicted CEO pay based on Equation (4.1).  1.2 1 0.8 0.6 Totalpay  0.4  Cashpay  0.2  Equitypay  0 -0.2  -3  -2  -1  0  1  -0.4 -0.6  145  2  3  Panel C: Number of Stock and Options Granted This figure presents the trend in the number of stock and options granted to the CEO over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. Stockshare is the number of shares in the CEO’s annual stock grants as a percentage of the total number of shares outstanding. Optionshare is the number of shares underlying the CEO’s annual options grants as a percentage of the total number of shares outstanding. Incentiveshare is the sum of Stockshare and Optionshare. The median values are presented.  0.35% 0.30% 0.25% 0.20% Optionshare  0.15%  Incentiveshare  0.10% 0.05% 0.00% -3  -2  -1  0  1  146  2  3  Figure 4.2. Firm Performance Around a Pay Cut Panel A: Stock Performance This figure presents the trend in stock performance over the seven-year period surrounding a pay cut. Abret is the industry-adjusted stock return, based on Fama and French (1997) 48 industries. Year 0 is the year when the pay cut occurs. The median value of various performance measures is presented.  15.00% 10.00% 5.00% 0.00% -5.00%  -3  -2  -1  0  1  2  Stockreturn  3  Abret  -10.00% -15.00% -20.00% -25.00%  Panel B: Operating Performance This figure presents the trend in operating performance over the seven-year period surrounding a pay cut. Abroa is the industry-adjusted ROA, based on Fama and French (1997) 48 industries. Year 0 is the year when the pay cut occurs. The median value of various performance measures is presented.  16.00% 14.00% 12.00% 10.00% 8.00%  ROA  6.00%  Abroa  4.00% 2.00% 0.00% -2.00%  -3  -2  -1  0  1  147  2  3  Figure 4.3. Capital Expenditures, R&D, and Capital Structure Around a Pay Cut Panel A: Capital Expenditures This figure presents the trend in capital expenditures over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. Capex is capital expenditures over total assets. Abcapex is the industry-adjusted Capex, based on Fama and French (1997) 48 industries. The median value is presented.  6.0% 5.0% 4.0% 3.0%  Capex  2.0%  Abcapex  1.0% 0.0% -1.0%  -3  -2  -1  0  1  2  3  Panel B: R&D Expenses This figure presents the trend in R&D expenses over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. R&D is research and development expenses over total assets. AbR&D is the industry-adjusted R&D, based on Fama and French (1997) 48 industries. The mean value is presented (The corresponding median value is always zero).  4.50% 4.00% 3.50% 3.00% 2.50% R&D  2.00%  AbR&D  1.50% 1.00% 0.50% 0.00% -3  -2  -1  0  1  148  2  3  Panel C: Capital Structure This figure presents the trend in capital structure over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. Book Leverage is the ratio of long-term debt and current debt over total assets. Market Leverage is the ratio of long-term debt and current debt over the market value of total assets. Abbooklev and Abmktlev are the industry-adjusted Book Leverage and Market Leverage, respectively, based on Fama and French (1997) 48 industries. The median value is presented.  25.0% 20.0% 15.0%  Book Leverage Market Leverage  10.0%  Abbooklev Abmktlev  5.0% 0.0% -3  -2  -1  0  1  2  3  -5.0%  Panel D: Decomposing Leverage This figure presents the trend in debt issuance, debt retirement, equity issuance, and equity retirement over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. Debt_issue is the firm’s annual debt issuance over total assets. Debt_retire is the firm’s annual debt retirement over total assets. Equity_issue is the firm’s equity issuance over total assets. Equity_retire is the firm’s equity retirement over total assets. The median value of various measures is presented.  2.50% 2.00% Debt_issue  1.50%  Debt_retire 1.00%  Equity_issue Equity_retire  0.50% 0.00% -3  -2  -1  0  1  149  2  3  Panel E: Decomposing Industry-adjusted Leverage This figure presents the trend in industry-adjusted debt issuance, debt retirement, equity issuance, and equity retirement over the seven-year period surrounding a pay cut. The industry classification follows Fama and French (1997) 48 industries. Year 0 is the year when the pay cut occurs. The median value of various measures is presented. The median value of industry-adjusted Debt_issue (Abdebt_issue) is always zero from year -3 to year +3.  0.15% 0.10% 0.05% Abdebt_issue  0.00% -0.05%  -3  -2  -1  0  1  2  3  Abdebt_retire Abequity_issue  -0.10%  Abequity_retire  -0.15% -0.20% -0.25%  150  Figure 4.4. Top Executive (Excluding CEO) Pay Around a CEO Pay Cut This figure presents the trend in top executive (excluding the CEO) pay over the seven-year period surrounding a pay cut. Totalpay is the sum of salary, bonuses, long-term incentive plans, the grant-date value of restricted stock awards, and the Black-Scholes value of granted options. Cashpay is the sum of salary, bonus, payouts from long-term incentive plans, and all other cash-based compensation. Equitypay is the value of restricted stock and the Black-Scholes value of stock options. Year 0 is the year when the pay cut occurs. The median value of top executive pay measured in thousands of 2006-constant dollars is presented.  1400 1200 1000 800  Totalpay  600  Cashpay Equitypay  400 200 0 -3  -2  -1  0  1  151  2  3  Figure 4.5. Control-adjusted Performance Measures Panel A: Control-adjusted Stock Return This figure presents the trend in control-adjusted stock return over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. Following Huson, Malatesta, and Parrino (2004), each pay-cutting firm is matched to control firms in the same Fama and French (1997) 48 industries whose stock returns in year -1 are within [90%, 110%] of the sample firm’s stock return. Control-adjusted stock return of each sample firm is computed as the difference between its stock return and the median stock return of its control group. The median value is presented. 15.00% 10.00% 5.00% 0.00% -5.00%  -3  -2  -1  0  1  2  3  -10.00% -15.00% -20.00% -25.00%  Panel B: Control-adjusted Operating Performance This figure presents the trend in control-adjusted operating performance (ROA) over the seven-year period surrounding a pay cut. Year 0 is the year when the pay cut occurs. Following Huson, Malatesta, and Parrino (2004), each pay-cutting firm is matched to control firms in the same Fama and French (1997) 48 industries whose stock returns in year -1 are within [90%, 110%] of the sample firm’s stock return. Control-adjusted ROA of each sample firm is computed as the difference between its ROA and the median ROA of its control group. The median value is presented. 16.00% 14.00% 12.00% 10.00% 8.00% 6.00% 4.00% 2.00% 0.00% -2.00%  -3  -2  -1  0  152  1  2  3  References Baker, M., and J. Wurgler, 2002, Marketing timing and capital structure, Journal of Finance 57, 1-32. Bebchuk, L.A., and J. Fried, 2003, Executive compensation as an agency problem, Journal of Economic Perspectives 17, 71-92. Bebchuk, L.A., and Y. Grinstein, 2005, The growth of executive pay, Oxford Review of Economic Policy 21, 283-303. Bebchuk, L.A., and Y. Grinstein, 2007, Firm expansion and CEO pay, Harvard Law School and Cornell University working paper. Brenner, M., R.K. Sundaram, and D. Yermack, 2000, Altering the terms of executive stock options, Journal of Financial Economics 57, 103-128. Callaghan, S.R., P.J. Saly, and C. Subramaniam, 2004, The timing of option repricing, Journal of Finance 59, 1651-1676. Chance, D.M., R. Kumar, and R.B. Todd, 2000, The ‘repricing’ of executive stock options, Journal of Financial Economics 57, 129-154. Chen, X., J. Harford, and K. Li, 2007, Monitoring: Which institutions matter?, Journal of Financial Economics 86, 279-305. Chidambaran, N.K., and N.R. Prabhala, 2003, Executive stock option repricing, internal governance mechanisms, and management turnover, Journal of Financial Economics 69, 153-189. Core, J., and W. Guay, 1999, The use of equity grants to manage optimal equity incentive levels, Journal of Accounting and Economics 28, 151-184. Core, J., Holthausen, R., Larcker, D., 1999. Corporate governance, chief executive officer compensation, and firm performance, Journal of Financial Economics 51, 371-406. Davila, A., and F. Penalva, 2007, Governance structure and the weighting of performance measures in CEO compensation, Review of Accounting Studies 11, 463-493. DeFond, M. L., and C. W. Park, 1999, The effect of competition on CEO turnover, Journal of Accounting and Economics 27, 35-56. Denis, D. J., 1994, Investment opportunities and the market reaction to equity offerings. Journal of Financial and Quantitative Analysis 29, 159-77. 153  Dial, J., and K.J. Murphy, 1995, Incentive, downsizing, and value creation at General Dynamics, Journal of Financial Economics 37, 261-314. Dierkens, N., 1991, Information asymmetry and equity issues. Journal of Financial and Quantitative Analysis 26, 181-200. Easton, P., P. Shroff, and G. Taylor, 2000, Permanent and transitory earnings, accounting record lag, and the earning coefficients, Review of Accounting Studies 5, 281-300. Eckbo, B. E., and K. S. Thorburn, 2003, Control benefits and CEO discipline in automatic bankruptcy auctions, Journal of Financial Economics 69, 227-258. Fama, E., 1980, Agency problems and the theory of the firm, Journal of Political Economy 88, 288-307. Fama, E., and K. French, 1992, The cross-section of expected stock returns, Journal of Finance 47, 427-465. Fama, E., and K. French, 1995, Size and book-to-market factors in earnings and returns, Journal of Finance 50, 131-155. Fama, E., and K. French, 1997, Industry costs of capital, Journal of Financial Economics 43, 153-193. Garvey, G., and T. Milbourn, 2006, Asymmetric benchmarking in compensation: Executives are rewarded for good luck but not penalized for bad, Journal of Financial Economics 82, 197-226. Gillan, S.L., J.C. Hartzell, and R. Parrino, 2009, Explicit vs. implicit contracts: Evidence from CEO employment agreements, Journal of Finance forthcoming. Gilson, S. C. and M. R. Vetsuypens, 1993, CEO compensation in financially distressed firms: An empirical analysis, Journal of Finance 48, 425-458. Gompers, P., J. Ishii, and A. Metrick, 2003, Corporate governance and equity prices, Quarterly Journal of Economics 118, 107-155. Hall, B.J., and J.B. Liebman, 1998, Are CEOs really paid like bureaucrat? Quarterly Journal of Economics 103, 653-691. Hartzell, J. C., and L. T. Starks, 2003, Institutional investors and executive compensation, Journal of Finance 58, 2351-2374. Holmström, B., 1979, Moral hazard and observability, Bell Journal of Economics 10, 74-91. 154  Holmström, B., and P. Milgrom, 1987, Aggregation and linearity in the provision of intertemporal incentives, Econometrica 55, 303-328. Huson, M. R., P. H. Malatesta, and R. Parrino, 2004, Managerial succession and firm performance, Journal of Financial Economics 74, 237-275. Jegadeesh, N. and S. Titman, 1993, Returns to buying winners and selling losers: Implication for stock market efficiency, Journal of Finance 48, 65-91. Jensen, M.C., and W. Meckling, 1976, Theory of the firm: Managerial behavior, agency costs and ownership structure, Journal of Financial Economics 3, 305-360. Jensen, M.C., and K.J. Murphy, 1990, Performance pay and top-management incentives, Journal of Political Economy 98, 225-264. Jenter, D., and F. Kanaan, 2008, CEO turnover and relative performance evaluation, Journal of Finance forthcoming. Lewellen, J., 2002, Momentum and autocorrelation in stock returns, Review of Financial Studies 15, 533-563. Jung, K., Y. Kim, and R. M. Stulz, 1996, Timing, investment opportunities, managerial discretion, and the security issue decision, Journal of Financial Economics 42, 159-185. Mehran, H., G.E. Nogler, and K.B. Schwartz, 1998, CEO incentive plans and corporate liquidity policy, Journal of Financial Economics 50, 319-349. Murphy, K. J., 1986, Incentives, learning, and compensation: A theoretical and empirical investigation of managerial labor contracts, Rand Journal of Economics 17, 59-76. Murphy, K., 1999, Executive compensation. In: Ashenfelter, O., Card, D. (Eds.), Handbook of Labor Economics 3, Elsevier, Amsterdam, 2485-2563. Oyer, P., 2004, Why do firms use incentives that have no incentive effects? Journal of Finance 59, 1619-1649. Rajgopal, S., T. Shevlin, and T. Zamora, 2006, CEO’s outside employment opportunities and the lack of relative performance evaluation in compensation contracts, Journal of Finance 61, 1813-1844. Titman, S., K.C.J. Wei, and F. Xie, 2004, Capital investments and stock returns, Journal of Financial and Quantitative Analysis 39, 677-700.  155  Chapter 5 Conclusion This thesis undertakes empirical studies on various corporate decisions, including acquisitions, executive compensation, and governance structure. The first essay provides insights to understand the managers’ motivation in making acquisitions from the perspective of managerial horizon. It identifies managerial horizon as an important factor in the M&A decision process, and enhances our understanding of the causes and effects of takeovers. Further, this article suggests that career concern and compensation scheme are two determinants of managerial horizon. This paper makes a number of contributions to the existing literature. First of all, my study enhances the literature on mergers and acquisitions by identifying managerial horizon as an important factor in the M&A decision process, and also provides new insights into understanding the causes and effects of takeovers. Further, this article suggests that career concern and compensation scheme are two determinants of managerial horizon. To the best of my knowledge, this is the first paper providing an empirical measure of managerial horizon. Moreover, my paper adds to the literature on corporate behaviors from the perspective of managerial characteristics (e.g., Bertrand and Schoar (2003) and Malmendier and Tate (2008)). Finally, this study also contributes to the rapidly growing field of behavioral corporate finance, which sees corporate policies as a response to market mispricing (see Baker et al. (2006) for a comprehensive survey). One limitation to my results is the endogeneity issue. A CEO’s horizon could be endogenously determined by some unobservable factors that drives acquisition decisions. 156  Although it is difficult to solve this problem completely, I am able to alleviate some endogeneity concerns by adding a number of additional controls in the regression. I show that my results are not driven by time trend effects, industry effects, firm effects (where possible) or tangible firm and CEO characteristics. The second essay studies the managerial incentive and corporate governance mechanism when managers can hedge their stock and option award. It identifies executive hedging cost as an important determinant to executive compensation structure. To my best knowledge, this is the first paper that empirically examines how management compensation schemes depend on managers’ opportunities to hedge. In addition, my research also furthers our understanding of capital structure as a substitute mechanism for compensation contracts in resolving managerial agency problems. Moreover, this article provides insight into managers’ personal financial decision-making. The exercise behavior of executive stock options is quite an important topic for compensation research, because it is crucial to the valuation of employee stock options (Huddart and Lang (1996)). My paper contributes to this literature by documenting the important impact of managerial hedging on the managers’ option-exercising decisions. Beyond the implications for CEO pay, this study also improves our understanding of managerial incentive in making corporate policies. Stock-based compensation is an often-cited factor that influences corporate decisions. This influence clearly depends on whether or not managers can hedge. Therefore, this research sheds more light on dividend policy and corporate diversification through the lens of executive hedging. Lastly, this paper documents indirect evidence that executives tend to use public 157  derivative markets to unwind their incentive portfolios and that shareholders take those publicly-tradable derivatives into consideration when designing compensation contracts. There are a few extensions that I can explore for future research on executive hedging. First of all, it would be interesting to investigate how managerial hedging influences mergers and acquisitions decisions. Datta et al. (2001) show that equity-based compensation induces managers to make better acquisition decisions. Harford and Li (2007) document that CEO pay tend to rise after an acquisition. Presumably, the relation between CEO personal wealth and acquisition decisions depends on whether or not managers can hedge their personal portfolios. Moreover, while my study focuses on the possibility that managers hedge through public option market, another way for managers to hedge is to trade their competitors’ stocks. It would be interesting to investigate this issue in a future research project. The third essay examines the incentive effects from a sharp cut in CEO compensation, which is largely overlooked by the current literature. It shows that cutting CEO pay sharply is not uncommon. CEO pay cuts usually follow poor stock performance, and this relation is stronger for better-governed companies. On average, the CEOs can improve their firm performance and have their pay restored after the cuts. These results indicate that the dynamic change in CEO compensation is consistent with optimal contracting view: Pay cuts provide managers ex ante incentives to avoid bad performance and ex post incentive to turnaround the firm if poor performance happens. This study is particularly relevant as the US government is currently taking aggressive actions in limiting executive compensation level. 158  This article shows that CEOs do suffer personal wealth loss via pay cuts when their firms perform poorly, complementing the overwhelming literature on overall CEO pay rise (Bebchuk and Fried (2003)). It also contributes to the literature of option repricing by demonstrating that pay cuts, relative to repricing, seem more widely used by boards and more effective to incentive managers to improve firm performance.  159  References Baker, M., R. Ruback, and J. Wurgler, 2006, Behavioral corporate finance: A survey, forthcoming in B.Espen Eckbo (ed.), Handbook of Corporate Finance: Empirical Corporate Finance (Handbooks in Finance Series, Elsevier/North Holland). Bebchuk, L.A., and J. Fried, 2003, Executive compensation as an agency problem, Journal of Economic Perspectives 17, 71-92. Bertrand, M. and A. Schoar, 2003, Managing with style: The effect of managers on firm policies, Quarterly Journal of Economics 118, 1169-1208. Datta, S., M. Iskandar-Datta, and K. Raman, 2001, Executive compensation and corporate acquisition decisions, Journal of Finance 56, 2299-2336. Harford, J., and K. Li, 2007, Decoupling CEO wealth and firm performance: The case of acquiring CEOs, Journal of Finance 62, 917-949. Malmendier, U., and G. Tate, 2008, Who makes acquisitions? CEO overconfidence and the market’s reaction, Journal of Financial Economics 89, 20-43. Huddart, S., and M. Lang, 1996, Employee stock option exercises: An empirical analysis, Journal of Accounting and Economics 21, 5-43.  160  Appendix:  Examples of CEO pay cuts in our sample  1. Edward W. Barnholt, CEO of Agilent Technologies Inc In 2002, Agilent’s sales were down 28%, the stock was off 35%, and the firm posted a $1 billion loss. So when it came time for the board to decide Barnholt’s pay, the board decided to cut his base salary by 10%, to $925,000, and give no bonus or restricted-stock grant for the second consecutive year. Says Barnholt: “I don't expect anything different. If the company doesn’t perform, I shouldn’t be getting any rewards.” Source: Business Week. www.businessweek.com/magazine/content/03_16/b3829002.htm 2. Richard M. Rodstein, CEO of K2 INC In determining the CEO’s incentive compensation award for 2001, the Committee considered K2’s performance for the year in meeting earnings targets, stock price performance, improvement in margins, returns on investment and meeting cash flow objectives, implementation of cost reduction programs, and augmenting K2’s long-term strategic plan for sustainable growth. The Committee noted that while K2’s stock price decreased 10% for the year, K2’s peer group index decreased 16% in the same period. The Committee also noted that despite a significant decline in the sales of inline skates and a collapse of the scooter market, K2’s remaining businesses reported significant improvements in operating earnings in 2001 due in part to sales of new products, to the transfer of certain production to China and an aggressive cost reduction program. The Committee noted the successful transfer of production of alpine skis to China and the implementation of significant cost reduction measures that should benefit future years. Finally, the committee considered the significant cash flow and debt reduction of K2 during the period despite the substantial decline in sales. After consideration of the above factors, the committee elected not to grant any award to the Chief Executive Officer for the year 2001 compared to an award of $285,000 in the prior year. The 2001 total compensation for the CEO represents a 47% shortfall from the 50th percentile for total compensation of the marketplace for similar positions, according to survey data. Source: Def 14A 2002 for K2 INC www.sec.gov/Archives/edgar/data/6720/000091205702012792/a2072243zdef14a.htm 3. Philip J. Purcell, CEO of Morgan Stanley Morgan Stanley cut the compensation of its chairman and chief executive, Philip J. Purcell, by 26 percent in 2002. The cut in pay follows a 17 percent decline in stock price and a 15 percent decline in net income at Morgan Stanley. The company paid Mr. Purcell $11 million in 2002, down from $15 million in 2001. Moreover, the aggregate compensation paid to the five most highly compensated officers for 2002 also decreased approximately 26% from 2001. Source: Def 14A 2002 for Morgan Stanley www.sec.gov/Archives/edgar/data/895421/000095013003001281/ddef14a.htm#tx814_16  161  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0067252/manifest

Comment

Related Items