CEO Pay Gap and Corporate Debt Structure

Transcription

1 CEO Pay Gap and Corporate Debt Structure Di Huang School of Business University of Connecticut Chinmoy Ghosh School of Business University of Connecticut Hieu Phan The Eli Broad College of Business Michigan State University

2 CEO Pay Gap and Corporate Debt Structure Abstract Prior studies document that CEO pay gap, measured as the difference between a CEO s total pay and the median total pay of the next group of senior executives, is positively related to firm value and performance. In this study, we examine the relation between CEO pay gap and the firm's debt structure. We first establish a negative association between CEO pay gap and the firm's default risk, suggesting CEO pay gap is beneficial to the creditors. Consistent with this relation, we subsequently present robust evidence that CEO pay gap is positively related to debt maturity but negatively related to cost of debt and number of debt covenants, indicating creditors take into consideration the effect of CEO pay gap when designing the debt contracts. Our study further highlights the importance of CEO pay gap in the compensation and debt structure literature. I. Introduction Conventional literature has largely focused on the bilateral relations between shareholders and managers or between shareholders and bondholders. Recently, some studies examine the simultaneous interactions among shareholders, bondholders and managers (e.g., Chava, Kumar and Warga (2009); Brockman, Martin and Unlu (2010)). These studies show that corporate bondholders are fully aware of the alignment of shareholders and managers interests and of the impacts such interest alignment potentially has on the interest of bondholders. They also report on how bondholders design the debt contracts to protect their own interests. For example, Chava et al. (2009) show that bondholders use various types of bond covenants in response to managerial entrenchment and the risk of managerial fraud. Brockman et al. (2010)

3 find that bondholders use short-term debt to mitigate managerial risk-taking incentives arising from CEO compensation arrangements. The interactions among shareholders, bondholders and managers tend to be more multifaceted than suggested by the research mentioned above. One notable shareholder-manager interest-alignment mechanism that receives increasing attention from a growing stream of recent research is the promotional incentives proxied by pay gap between CEO and other top executives. One stream of literature suggests that the CEO pay gap represents the reward for an intra-firm promotional tournament for the senior executives. Higher CEO pay gap represents a bigger tournament prize and motivates managers to take risk-increasing activities in order to maximize the outcomes and thus increase their chance of winning (Kale, Reis, and Venkateswaran (2009) and Kini and Williams (2012)). This tournament paradigm is challenged by the productivity-based paradigm (Masulis and Zhang (2013)), which hypothesizes that CEO pay gap represents a CEO productivity relative to that of other senior executives or, in other words, the marginal contribution of the CEO to the firm. A more productive CEO contributes positively to the firm's operating performance and overall value of the firm, thus receiving higher compensation relative to other top executives. Although these two streams of literature diverge on what CEO pay gap represents, they predict similar relation between CEO pay gap and firm value: higher CEO pay gap is associated with higher firm value, at least for certain types of firms (Kale, Reis, and Venkateswaran (2009), Zhang (2013), and Masulis and Zhang (2013)). In this research, we build on the positive relation between CEO pay gap and firm value and further examine how CEO pay gap impacts the firm's debt structure. We argue that higher CEO pay gap should motivate top executives including both CEO and second-tier executives to act more cautiously when increasing firm value for their job security. Therefore, higher CEO pay

4 gap should be associated with lower likelihood of bankruptcy. To examine this, we adopt a measure of default likelihood proxied by distance to default. Distance to default is the z-score estimated based on Merton (1974) model, in which the equity of the firm is considered as a call option on the underlying value of the firm and the strike price equals the value of the firm's debt. The larger the distance to default, the lower likelihood of default. Consistent with our expectation, we find that CEO pay gap is positively associated with distance to default. That is, greater CEO pay gap is related to lower likelihood of default. Then the next question is whether the relation between CEO pay gap and default likelihood is reflected in the firm's debt structure. Prior literature has documented several arrangements by the bondholders to address the agency issues arising from shareholder-bondholder conflicts, such as shortening debt maturity, increasing cost of debt, and increasing the number of debt covenants (Chava et al. (2009); Brockman et al. (2010)). Since CEO pay gap aligns the interest between shareholders and bondholders by increasing the value of overall firm and lowering the bankruptcy likelihood, we expect positive reaction from bondholders to CEO pay gap. We start by testing the impact of CEO pay gap on debt maturity. We employ two alternative variables in this analysis i) the proportion of short-term debt out of total debt on balance sheet obtained from Compustat; and ii) years to maturity of newly issued debt retrieved from SDC Platinum. We find a significantly positive relation between CEO pay gap and debt maturity using either sample for analysis, indicating bondholders lengthen debt maturity in response to larger CEO pay gap. We then use the new debt issues dataset to examine the relations between CEO pay gap and the cost of debt. We measure the cost of debt as the yield spread between the yield to maturity of newly issued debt and the corresponding Treasury benchmark yield obtained from SDC Platinum. We find that CEO pay gap is negatively related to cost of debt, indicating that

5 bondholders demand a lower risk premium when CEO pay gap of the firm is higher. This finding corroborates with our expectation that bondholders react positively to CEO pay gap. Finally, we analyze the effect of CEO pay gap on debt covenants using loan contracts retrieved from the Dealscan database. We find that CEO pay gap is negatively related to the number of debt covenants, suggesting that bondholders are less restrictive when lending to a firm with a larger CEO pay gap. Again, the positive response of bondholders towards CEO pay gap is consistent with our expectation. To ensure the robustness of our results, we control for other variables that potentially explain the risk-taking incentives of the CEOs such as CEO delta, defined as the sensitivity of CEO compensation portfolio to changes in stock price, and CEO vega, defined as the sensitivity of CEO compensation portfolio to stock return volatilities, throughout our analysis. Furthermore, since the relation between CEO compensation and debt structure could be endogenous due to a possible joint determination of executive compensation structure and debt structure or reverse causality, we address these endogeneity concerns by using several estimation methods which include: (i) ordinary least square (OLS) regressions using lagged independent variables; (ii) instrumental-variable regressions in which CEO pay gap, CEO delta and CEO vega are instrumented; and (iii) simultaneous equation model in which the cost of debt and debt maturity are simultaneously determined. At last, we exploit the significant jump in CEO pay gap following the adoption of FAS 123R in FAS 123R changes the accounting treatment of stock options and as a result, firms dramatically reduce the use of stock options and substitute options with other pay-performance alignment compensation forms such as bonuses, restricted stock and long-term incentive awards (Hayes, Lemmon and Qiu(2012)). This substitution leads to a notable increase in CEO pay gap after 2005 as shown in Figure 1. We find that after 2005

6 the effect of CEO pay gap on debt structure also becomes more pronounced, further confirming the link between CEO pay gap and debt structure. Our research makes several contributions to the current literature. First, we document a negative association between CEO pay gap and the likelihood of default. Our finding indicates regardless of the possible contributors to CEO pay gap, it provides incentives for top executives to act cautiously to ensure the continuity of the firm. Consequently, CEO pay gap serves as a mechanism that aligns the interests of both shareholders and bondholders. Second, we systematically examine how this interest alignment mechanism is reflected in various aspects of debt structure. Specifically, we find that CEO pay gap is associated with longer debt maturity, lower cost of debt and fewer debt covenants. We provide comprehensive evidence that bondholders take into account the effects that CEO pay gap may have on their own interests in rationally designing the debt contract. This finding indicates that the bondholders' responses are not limited to the well-documented shareholder-management interest-alignment incentives through CEO delta and CEO vega; indeed, our evidence shows that the less documented effects of the pay gap also receive significant attention of the bondholders. Our research is built on two streams of literature. One stream of literature examines the pay gap between CEO and other senior executives. Kale et al. (2009) consider CEO pay gap as a proxy for tournament incentives and find that pay gap is positively related to firm value and operating performance. Also based on the tournament paradigm, Kini and Williams (2012) report a positive relation between CEO pay gap and corporate risk-taking behavior. They argue that the positive relation is attributable to the greater risk-taking incentives of the senior managers stemming from a larger tournament prize proxied by the pay gap in order to increase their chance of promotion. However, the tournament explanation of CEO pay gap is challenged by Masulis

7 and Zhang (2013), who examine the tournament paradigm alongside the productivity-based paradigm and find evidence in support of the latter paradigm. That is, CEO pay gap represents the CEO's relative contribution to the firm. Zhang (2013) examine the effect of CEO pay gap on firm value for different types of firm and find that CEO pay gap increases firm value for complex, capital intensive firms and firms in symmetric industries. Therefore, even though tournament paradigm and productivity-based paradigm differ with regard to what CEO pay gap stands for, they agree that CEO pay gap tends to increase firm value especially for certain types of firms. Another stream of literature related to our research focuses on bondholders' reaction to the interest-alignment arrangements between shareholders and managers. Brockman et al. (2010) investigate bondholders' responses to the differing effects of CEO delta and CEO vega. Due to the convex payoff structure of the options, the value of CEO stock option-based compensation increases with stock return volatility. Thus, large CEO vega encourages managerial risk-taking. On the other hand, CEO delta, representing the direct link between option pay-off and stock price, discourages risk-averse managers from taking on too much risk. As a result, they find CEO vega is negatively associated with debt maturity and positively associated with cost of debt whereas CEO delta is positively related to debt maturity and negatively related to the cost of debt. Chava et al. (2009) analyze how the bondholders use the debt covenants in response to the managerial entrenchment and fraud and report that managerial entrenchment and fraud have differing effects on the use of debt covenants: entrenchment is positively related to the use of investment covenants to restrain entrenched managers from "empire building" but negatively related to the use of dividend payout covenants due to the tendency of the entrenched managers to conserve cash. Begley and Feltham (1999) investigate the influence of managerial

8 compensation-based incentives on the use of debt covenants in restricting dividends and additional borrowing. They find evidence that the use of debt covenants is significantly and positively related to the ratio of a CEO's stock value to her total compensation, suggesting that creditors are concerned with managerial opportunism when a CEO's interest is aligned with those of the shareholders. The remainder of this paper is organized as follows: In section II, we review related research literature and develop testable predictions. Section III describes the sample selection and data. Section IV reports the empirical findings and discussions. Section V concludes the paper. II. Empirical Predictions Previous literature has established there is a positive relation between CEO pay gap and firm value especially for certain types of firms (Kale, Reis, and Venkateswaran (2009) and Zhang (2013)). In this study, we extend their analysis to examine CEO pay gap from bondholders' perspective. We start our analysis with the relation between CEO pay gap and the firm's bankruptcy likelihood. We argue that since CEO pay gap contributes positively to firm value, it should be associated with lower bankruptcy likelihood. In addition, larger CEO pay gap should cause top executives including both CEO and second-tier top executives to be more concerned about the continuity of the firm, lowering a firm's bankruptcy risk. To examine this relation, we adopt the distance to default as a measure of default likelihood. Consequently, we hypothesize higher CEO pay gap and distance to default are positively related. We next examine the relation between CEO pay gap and a firm's debt structure including debt maturity, cost of debt and total number of debt covenants. We argue that because higher CEO pay gap is associated with higher firm value and lower bankruptcy risk, it is beneficial to

9 both shareholders and bondholders. As a result, we expect bondholders to respond positively to the pay gap. To be more specific, we predict a positive relation between CEO pay gap and debt maturity and negative relations between CEO pay gap and the cost of debt and the number of debt covenants. In the following sections, we empirically examine if our predications are supported by the data. III. Sample Selection and Data Description A. Sample Selection We obtain compensation data of CEO and other senior executives from Execucomp. We use the compensation data to calculate CEO pay gap, CEO delta, and CEO vega. We include a description of the CEO delta and CEO vega estimation in Appendix B. We construct distance to default following the steps detailed in Bharath and Shumway (2008). Our debt-related data comes from a number of sources: short-term debt proportion (ST3) constructed as the proportion of short-term debt maturing within three years out of total debt is based on Compustat data; Maturity, defined as years to maturity of newly-issued debts, and Yield Spread, defined as the difference between yield to maturity of newly-issued debt and the corresponding Treasury benchmark yield are obtained from SDC Platinum; and debt covenants are obtained from Dealscan. To obtain the proportion of short-term debt, we start with 116,355 firm-year observations in Compustat Fundamental Annual dataset with non-missing values of debt to calculate ST3 over the period from 1993 to We calculate CEO delta, CEO vega, pay gap and other compensation-based variables using Execucomp from 1992 to 2010 and merge them with the one-year lead ST3 dataset. After merging Compustat data with executive compensation data and removing missing data, our sample size decreases to 30,204 firm-year observations. Following

10 Brockman et al. (2010), we limit our sample to firms with four-digit SIC codes between 2000 and 5999 and as the result, further eliminate 10,359 firm-year observations. We construct the variables following the procedures described in Appendix A and require their values to be nonmissing. Our final sample has 11,874 firm-year observations based on 1,523 unique firms. To minimize the effect of outliers, we winsorize ST3, CEO pay gap, CEO delta, and CEO vega at their 1% and 99% values. We construct the bond yield sample by retrieving new bond issues data from SDC Platinum and merge it with stock return data from CRSP, accounting data from Compustat, and compensation data from Execucomp. Requiring the SIC codes of the issuers to be between 2000 and 5999 causes us to lose a substantial amount of observations since issuers with SIC code 6111 (federal and federally-sponsored credit agencies) is a major player in the market 12. To ensure the robustness of our findings, we conduct analysis both on the basis of single issue and of each firm-year. For firm-year analysis, we calculate the issue size weighted average maturity and spread with the issue size defined as total proceeds from the issue. B. Sample Distribution and Summary Statistics We present the time series distribution of medians of CEO pay gap, CEO delta and CEO vega in Table 1. The sample covers 31,045 Execucomp firm-year observations of 3,278 firms from 1993 to To uncover the real growth pattern of CEO pay gap, CEO delta and CEO vega, we adjust the three variables for inflation using the Consumer Price Index (CPI). CEO pay gap exhibits an upward trend from 1993 to 2005, a sudden jump from 2005 to 2006, a notable drop from 2007 to 2008, and a gradual recovery from 2009 to Given the 1 22,495 of 32,903 issues in the original sample are from issuers with sic code Including financial firms into our analysis doesn't alter our findings.

11 positive association between CEO pay gap and firm value documented by Kale, Reis, and Venkateswaran (2009) and Zhang (2013), the general ascending trend from 1993 to 2005 possibly reflects recognition by the board of directors of the importance of CEO pay gap. The significant jump in CEO pay gap from 2005 to 2006 coincides with the adoption of FAS 123R, which changes the accounting treatment of stock options. Hayes, Lemmon and Qiu (2012) document a decline in usage of options following the implementation of FAS 123R in 2005 and an increase in reliance on bonuses, restricted stock and long-term incentive awards. The jump in CEO pay gap possibly indicates that board of directors substitute options with other forms of compensation that increase the pay gap between CEO and other top executives. On the other hand, CEO delta and vega experience noticeable drop from 2005 to 2006, corroborating the findings by Hayes, Lemmon and Qiu (2012). From 2007 to 2008, The effect of the financial crisis can be seen in all three variables. CEO pay gap, CEO delta and vega all experience a decline in After the financial crisis, the three variables recover gradually and reach the levels of 2007 at the end of our sample period in In Table 2, we present the summary statistics of the variables. Panel A reports the distribution of compensation variables, Panel B of firm characteristics variables and Panel C of bond issues variables. In Panel A, the sample size of 31,045 observations represents the full sample extracted from Execucomp. CEO pay gap has a mean value of $2, thousands and a median value of $ thousands. The values are in line with those reported by Kale et al. (2009) and Kini and Williams (2012). CEO delta has a mean value of $ thousands and a median value of $ thousands. CEO vega has a mean value of $73.40 thousands and a median value of $25.22 thousands. All three variables are right-skewed. To address the right

12 skewness of these compensation variables, we use their natural logarithm transformation in all regressions. Variables reported in Panel B are firm characteristics variables used in the regressions. We constructed these variables from Compustat and CRSP. The discrepancy in number of observations for these variables is due to the variability in sample size in different sets of regressions. Regarding dependent variables, distance to default, the z-score of the normal cumulative density function that predicts the probability of default, has a mean value of 7.36 and a median value of 6.57, indicating on average the default probability is quite low. ST3, the proportion of short-term debt, has a mean value of 0.40 and a median value of 0.32, similar to the levels reported by Brockman et al. (2010). In terms of other important control variables,, the mean (median) value of size,defined as market value of total assets, is $11,414 million ($2,626 million). Following prior literature, we use its log transformation throughout our analysis. The leverage ratio has mean and median values of 0.16 and 0.13 respectively. CEO stock ownership exhibits a mean value of 0.02 and a median value of The mean and median values of market to book ratio are 1.84 and 1.48 respectively. Panel C reports summary statistics of debt issues variables. On average, the new debt issues have years to maturity, 1.72 debt covenants, 1.88% yield spread over treasury yield and the total proceed from new issue is $401 million. Since the mean value of years to maturity at years is greater than its median value at years, evidence that it is right-skewed, we use its natural log transformation in our empirical analysis. IV. Empirical Results and Discussions A. CEO Pay Gap and Proportion Distance to Default

13 To examine the relation between CEO pay gap and the firm's debt structure, we need to first establish how CEO pay gap affects the firm's bankruptcy risk. The findings by Kale, Reis, and Venkateswaran (2009) and Zhang (2013) suggest CEO pay gap is positively related to firm value. We further argue that since CEO pay gap represents either the ultimate reward of winning internal promotion for the second-tier executives or the compensation to CEO for her productivity level, it provides incentives for top executives to ensure the continuity of the firm, reducing the firm's bankruptcy risk. Thus, we predict higher level of CEO pay gap is associated with lower level of bankruptcy risk. In this study, we measure bankruptcy risk by distance to default. Distance to default is constructed according to the model suggested by Merton(1974), in which the equity of the firm represents a call option on the underlying value of the firm and the strike price of the call option equals to the face value of the firm's debt. Distance to default is the z-score of the normal distribution that determines the value of the call option. Higher z-score indicates lower bankruptcy likelihood. We calculate distance to default based on steps detailed in Bharath and Shumway (2008). Table 3 reports OLS regressions with distance to default as dependent variable on CEO pay gap. In column (1) and (3), the only explanatory variable we include is CEO pay gap. In column 3, we control for firm fixed effects. Our objective is to ensure that the identified relation between CEO pay gap and distance to default is not driven by the presence of control variables that are potentially endogenously determined. Coefficient estimates of CEO pay gap in column and column 3 indicate that, consistent with our expectation, the CEO pay gap is positively related to distance to default and the effect is statistically significant at 1% level, reflecting the firm's bankruptcy risk decreases with CEO pay gap. In column (2) and column (4), we further

14 include control variables size, leverage and number of business segments and in column 4, we include firm fixed effects. We find that CEO pay gap continues to have a significantly positive effect on distance to default in both sets of regressions. In addition, the results show that distance to default increases with size and decreases with leverage, consistent with the findings by Sundaram and Yermack (2007). Overall, our findings reported by Table 3 support our expectation that the firm's default risk decreases with CEO pay gap. In the following analysis, we will examine if this relation is reflected in the firm's debt structure. Or in another word, if the bondholders take into consideration the effect of CEO pay gap when designing the firm's debt contracts. The debt features we examine in this study include debt maturity, cost of debt and debt covenants. B. CEO Pay Gap and Proportion of Short-term Debt Regression (Balance Sheet Data) Financial economic literature suggests that bondholders can use debt maturity, particularly short-term debt, to mitigate the risk arising from conflict of interest between bondholders and shareholders (Leland and Toft (1996), Rajan and Winton (1995); Brockman et al. (2010)). From the bondholders' perspective, the advantage of short-term debt comes from its contracting flexibility. Specifically, by lending short term and subjecting the borrowing firms to the risk of failing to rollover short-term debt, bondholders can steer managers away from riskincreasing activities induced by executive option-based compensation. For example, Brockman et al. (2010) show a positive relation between short-term debt and CEO vega because CEO vega encourages managerial risk-taking; however, the proportion of short-term debt is lower when CEO delta is large because delta discourages managerial risk-taking. Regarding CEO pay gap, prior researchers have documented a positive association between CEO pay gap and firm value which benefits both shareholders and bondholders.

15 Furthermore, the empirical evidence we present in the last section suggests that higher CEO pay gap is associated with lower risk of bankruptcy, again a positive perspective for bondholders. Consequently, we expect a negative relation between CEO pay gap and the proportion of shortterm debt. To exam the effect of CEO pay gap on proportion of short-term debt, we estimate the following multivariate regression model: ST3 i,t = α 0 + α 1 Log(CEO Pay Gap) i,t-1 + α 2 Log(CEO Delta) i,t-1 + α 3 Log(CEO Vega) i,t-1 + α 4 Log(Size) i,t-1 + α 5 Log(Size) 2 i,t-1 + α 6 Leverage i,t-1 + α 7 Asset Maturity i,t-1 + α 8 Ownership i,t-1 + α 9 Market/Book i,t-1 + α 10 Term Structure i,t + α 11 Abnormal Earnings i,t-1 + α 12 Return Volatility i,t-1 + α 13 Rate(dummy) i,t-1 + α 14 Altman Z-Score(dummy) i,t-1 + α 15 Regulated Industry (dummy) i,t-1 + ε i,t (1) We report our findings of the effects of CEO pay gap on the proportion of short-term debt (ST3) in Table 4. Following Brockman et al. (2010), we estimate the debt maturity regression with industry-year clustered standard errors where industries are based on the two-digit SIC codes. The first column in Table 4 replicates the debt maturity regression of Brockman et al. (2010) that examines the effects of CEO delta and CEO vega while controlling for other variables but excluding the CEO pay gap. The coefficient estimate of CEO delta (CEO vega) is significantly negative (positive), which is essentially similar to that reported by Brockman et al. Since CEO vega (CEO delta) encourages (discourages) managerial risk-taking, the findings further support the argument that bondholders use short-term debt to mitigate managerial risktaking incentives.

16 We examine the effect of CEO pay gap on short-term debt while controlling for CEO delta and CEO vega and other variables in column (2). The coefficient estimate of CEO pay gap is negative ( ) and highly significant, indicating that a larger CEO pay gap is associated with more long-term debt. This finding is consistent with our expectation. In addition, the coefficient on CEO delta (CEO vega) remains negative (positive) and statistically significant. The signs and significance of the coefficients on other control variables are also consistent with those documented in the literature. For instance, the coefficient estimates of Size Square and Ownership are positive whereas the coefficient estimates of Size, Leverage, Asset Maturity (book value-weighted average of maturities of property, plant and equipment and current asset), Rate Dummy (an indicator that takes a value of one if a firm has S&P credit rating in a given year, and zero otherwise) and Altman Z-Score are negative and statistically significant. We next consider the possibility that the relation between CEO compensation variables (CEO pay gap, CEO delta, and CEO vega) and firm' use of short-term debt are endogenous due to either a possible joint determination of corporate debt structure and executive compensation or reverse causality. Column (3) in Table 4 presents results of the instrumental variable (IV) regressions in which CEO pay gap, CEO delta and CEO vega are instrumented. We require the instruments to satisfy all the relevance and validity tests. In this set of regression, we identify log(median Industry CEO Pay Gap), log(median Industry CEO Delta), log(median Industry CEO Vega), Number of VPs and Inside Promotion as the relevant and valid instruments because the coefficients of these instruments are significant in the first-stage regressions. In addition, the Anderson-Rubin F-test for joint significance rejects the null hypothesis and indicates that the endogenous variables are jointly significant. Since we have more instruments than the endogenous variables, we present the statistics of the Hansen J-statistic of over-identification test.

17 The test statistic is so we cannot reject the null hypothesis that the instruments are valid. Finally, the Difference-in-Sargan C-statistic is and statistically significant at 1% level, rejecting the null hypothesis that CEO pay gap, CEO delta, and CEO vega are jointly exogenous and further implying that the endogeneity correction is necessary. The coefficients on the predicted CEO pay gap and CEO delta are and , respectively, and both are statistically significant at the 1% level. The coefficient on the predicted CEO vega is and significant at the 5% level. This evidence suggests that our results are robust to the correction for potential endogeneity bias. Furthermore, the negative and significant relation between the predicted CEO pay gap and the use of short-term debt indicates that creditors consider CEO pay gap beneficial to their interests, consistent with our finding that higher CEO pay gap is associated with lower bankruptcy risk. C. CEO Pay Gap and Maturity of New Debt Issues Using the maturity structure of outstanding debt reported on firms balance sheets in our earlier debt maturity analysis provides important advantages since we can track the impacts of CEO pay gap on corporate debt structure on both the cross-sectional and time series bases. However, because firms do not issue debts frequently, corporate debt maturity is likely to be stale and follow a decreasing trend whereas CEO pay gap could change dynamically and any documented effects of CEO pay gap on the debt structure could be artificial. Therefore, in this section, we analyze the effect of CEO pay gap on the maturity of new debt issues obtained from SDC Platinum instead of outstanding debt reported on the balance sheet. It is noteworthy that as firm value increases in CEO pay gap while default risk decreases in it, prospective bondholders should extend more favorable terms to firms with larger CEO pay

18 gap by lengthening the debt maturity. We use the following empirical model to examine the relation between CEO pay gap and years to maturity of newly issued debt: Log(Maturity) i,t = α 0 + α 1 Log(CEO Pay Gap) i,t-1 + α 2 Log(CEO Delta) i,t-1 + α 3 Log(CEO Vega) i,t-1 + α 4 Log(Size) i,t-1 + α 5 Log(Size) 2 i,t-1 + α 6 Leverage i,t-1 + α 7 Asset Maturity i,t-1 + α 8 Ownership i,t-1 + α 9 Market/Book i,t-1 + α 10 Abnormal Earnings i,t-1 + α 11 Return Volatility i,t-1 + α 12 Average Return i,t-1 + α 13 Interest Coverage i,t-1 + α 14 Term Structure i,t + α 15 Altman Z- Score(dummy) i,t-1 + α 16 Regulated Industry (dummy) i,t-1 + ε i,t (2) Table 5 presents the results of our analysis of the relation between the maturity of new debt issues and CEO pay gap. We use the natural log transformation of years to maturity as dependent variable. Column (1) reports the regression results that includes only CEO pay gap as the main test variable while column (2) further control for CEO delta and CEO vega. In both regressions, CEO pay gap is positively related to debt maturity, indicating that lenders offer longer debt maturity in response to the borrower s larger CEO pay gap. The positive reaction from the bondholders towards the borrower s CEO pay gap is consistent with our earlier evidence and further confirms our expectation. We also run IV regression and report the results of the outcome regression in column (3). The instruments that pass the relevance and validity tests in this set of regressions are industry-median CEO pay gap, industry-median CEO delta, industry-median CEO vega and CEO tenure. Again, the predicted CEO pay gap is positive and statistically significant, further supporting the CEO productivity hypothesis. In terms of the control variables, consistent with prior literature, our results indicate that firms with high leverage and high growth opportunity in terms of market to book ratio tend to raise debts with shorter maturity.

19 Taken together, the results in Table 4 and Table 5 consistently show that debt maturity increases in CEO pay gap. This finding remains qualitatively unchanged when we additionally control for other risk-taking incentives induced by CEO delta and CEO vega. The positive relation between CEO pay gap and debt maturity provides first line evidence that bondholders react positively to CEO pay gap, consistent with our expectations. D. CEO Pay Gap and Cost of Debt In this section, we examine the relation between CEO pay gap and cost of debt for new debt issues. Prior literature shows that bondholders are aware of the potential effects of CEO compensation arrangements on their own interest and use cost of debt as a mechanism to constrain managerial risk-taking behavior and to compensate for their lending risk (Brockman et al. (2010)). Specifically, bondholders increase the cost of debt in response to managerial riskseeking incentives but lower the cost of debt when the compensation incentives are beneficial to their interests. Therefore, based on the positive relation between CEO pay gap and firm value and distance to default, we should observe a negative relation between CEO pay gap and costs of debt. The multivariate regression model is specified as follows: Spread i,t = α 0 + α 1 Log(CEO Pay Gap) i,t-1 + α 2 Log(CEO Delta) i,t-1 + α 3 Log(CEO Vega) i,t-1 + α 4 Return Volatility i,t-1 + α 5 Average Return i,t-1 + α 6 Return on Sales i,t-1 + α 7 Leverage i,t-1 + α 8 Interest Coverage i,t-1 + α 9 Log(Total Proceeds) i,t + α 10 Treasury Benchmark Yield i,t + α 11 Yield Curve Slope i,t + α 12 Regulated Industry (dummy) i,t-1 + ε i,t (3) Table 6 presents the results of the cost of debt regressions. Again, the sample we use in this analysis is from Global New Issues of SDC Platinum. The dependent variable is yield spread which is defined as the difference between yield to maturity of new debt issues and the Treasury benchmark yields which have similar maturity. Column (1) includes only CEO pay gap as the

20 test variable and other control variables. The coefficient on CEO pay gap is negative ( ) and statistically significant at 1% level, indicating that bondholders lower the cost of debt when CEO pay gap is large. Column (2) presents the regression results that additionally control for CEO delta and CEO vega. The magnitude of the coefficient estimate on CEO pay gap increases to and remains significant at the 1% level, indicating that our results are robust to the control of CEO equity-based compensation. Furthermore, similar to Brockman et al. (2010), we find that cost of debt is negatively (positively) related to CEO delta (CEO vega). To address a concern about a potential endogenous relationship between executive compensation and cost of debt, we run IV regression with CEO pay gap, CEO delta, and CEO vega being instrumented by CEO tenure, number of VPs, inside promotion, and succession plan, and report the results of the outcome regression in column (3). The coefficient on CEO vega remains statistically significant while the coefficient on CEO delta becomes statistically insignificant. Crucially, the coefficient on CEO pay gap remains negative and statistically significant at 5% level, suggesting that our results are robust to endogeneity correction and further supporting the our previous findings. Finally, the test statistics of Anderson-Rubin Wald test, Hansen J test and Endogeneity test indicate that our selected instruments are relevant and valid. Regarding control variables, we find that firms with high return volatility, highly leveraged and raising large amount of debt tend to have higher cost of debt. Conversely, firms with high average returns and high return on sales can issue debt more cheaply. E. CEO Pay Gap,Debt Maturity and Cost of Debt at Firm Level In the last two sections, we conduct our analysis using the original sample of single issues. In this section, we consolidate the sample into a firm-year format. We compute proceedsweighted average years to maturity and yield spread for firms with multiple issues in a year.

21 Panel A of Table 7 presents regression results using proceeds-weighted average maturity for each firm-year as dependent variable while Panel B reports the results using proceeds-weighted average yield spread for each firm-year as dependent variable. In column 1 of Panel A, we only include log (CEO pay gap) as the test variable. In column 2, we further include log (CEO delta) and log (CEO vega). In both columns, the coefficient estimates of CEO pay gap is positively significant, indicating larger CEO pay gap is associated with longer debt maturity at firm level as well. In panel B, we follow the same order of reporting: column 1 reports the regression results including only log (CEO pay gap) as the main test variable while column 2 reports the results include all three compensation variables. In both columns, CEO pay gap is significantly and negatively related to proceeds-weighted yield spread, consistent with our findings based on single issues. F. CEO Pay Gap and the Joint Effect of Debt Maturity and Cost of Debt To account for a possible joint determination of debt maturity and cost of debt, we follow Brockman et al. (2010) in estimating a system of simultaneous equations with debt maturity and cost of debt as endogenous variables. Table 7 reports the simultaneous equation results using the SDC new debt issues sample. The first two columns (model 1) present the results based on single issues while columns 3 and 4 (model 2) reports the results based on the consolidated sample of firm-year observations. The coefficient estimate of CEO pay gap on yield spread is ( ) and significant at the 1% level in model 1 (model 2), indicating that bondholders lower the cost of debt when the borrowing firms have larger CEO pay gap. The coefficient estimate of CEO pay gap on debt maturity is (0.0282) and significant at the 5% (10%) level in model 1 (model 2), suggesting a positive relation between CEO pay gap and debt maturity. Taken together, the

22 coefficient estimates of CEO pay gap on cost of debt and debt maturity in the simultaneous equations are consistent with our previous findings that bondholders lengthen debt maturity and lower cost of debt in response to a greater CEO pay gap. The positive reactions of bondholders to CEO pay gap are consistent with our previous findings that the firm's bankruptcy risk decreases in CEO pay gap. G. CEO Pay Gap and Debt Covenants Besides the maturity structure of debt and cost of debt, another mechanism through which bondholders can deploy to protect themselves from the agency cost of debt is debt covenants. Previous studies show that bondholders use debt covenants to mitigate the agency conflicts between shareholders and bondholders. For instance, Begley and Feltham (1999) find that bondholders are more likely to use debt covenants restricting dividends and additional borrowing if there is a larger threat of CEO opportunism on behalf of shareholders as motivated by CEO stock ownership. Billett, King and Mauee (2007) argue that short-term debt and debt covenants are substitutes in mitigating shareholder-bondholder conflicts. Chava et al. (2009) document that bondholders use debt covenants in response to managerial entrenchment and the risk of managerial fraud. In this section, we examine the impact of CEO pay gap on the use of debt covenants. Table 8 reports the results for the debt covenants analysis estimated based on the following model: Log(Sum of Debt Covenants) i,t = α 0 + α 1 Log(CEO Pay Gap) i,t-1 + α 2 Log(CEO Delta) i,t-1 + α 3 Log(CEO Vega) i,t-1 + α 4 Log(Maturity) i,t + α 5 Leverage i,t-1 + α 6 Asset Maturity i,t-1 + α 7 Market/Book i,t-1 + α 8 Return Volatility i,t-1 + α 9 Ownership i,t-1 + α 10 Abnormal Earnings i,t-1 + α 11 Altman Z-Score(dummy) i,t-1 + α 12 Regulated Industry (dummy) i,t-1 + ε i,t (4)

23 Column (1) presents the regression result that includes only CEO pay gap as the test variable and other controls. The coefficient estimate of CEO pay gap is and statistically significant at the 1% level, indicating that bondholders impose fewer debt covenants when dealing with firms with larger CEO pay gap. This result is consistent with our previous findings. Our finding is essentially similar when we additionally control for CEO delta and CEO vega in column (2). Finally, we use the IV regression approach to account for the possibility that CEO pay gap, CEO delta and CEO vega are endogenously related to the number of debt covenants. Column (3) reports the results of the outcome regression. The set of instruments we use for CEO pay gap, CEO delta and CEO vega include industry-median CEO pay gap, industry median CEO delta, industry median CEO vega and Inside promotion dummy. These instruments pass all our relevance and validity tests. The coefficient estimate of predicted CEO pay gap remains negative and statistically significant, indicating that the number of debt covenants decreases in CEO pay gap even after we correct for potential endogeneity bias. Consistent with Billet et al. (2007), we find that the number of debt covenants increases in leverage. Furthermore, we find number of debt covenants increases in return volatility and decreases in market to book ratio and abnormal earnings. H. CEO Pay Gap Before and After 2005 The implementation of FAS 123R in 2005 changes the accounting treatment of stock options. Consequently, there is a significant decline in stock options awarded to CEOs and instead options have been substituted by other forms of compensation that increase CEO pay gap (Hayes, Lemmon and Qiu (2012)). This change is reflected in Figure 1 which depicts the time series distribution of de-trended pay gap. There is a notable jump in CEO pay gap after More specifically, as shown in Table 1, CEO pay gap increases from thousands to

24 thousands from 2005 to In this section, we formally examine whether there is a significant change in CEO pay gap after 2005 and present the results in panel A of Table 10 based on the following model: Log(CEO Pay Gap) i,t = α 0 + α 1 Post_ α 2 Trend+ α 3 Post_2005 * Trend + α 4 Log(Size) i,t-1 + α 5 Number of Segments i,t-1 + α 6 Tobin's Q i,t-1 + α 7 CEO Above 62 (dummy) i,t-1 + α 8 Log(CEO Tenure) i,t-1 + ε i,t (5) Column (1) and (2), we only include post_2005, trend, and the interaction term of the two variables in the regression. In column (3) and (4), we regress on the full set of control variables. In column (2) and column (4), we further control for firm fixed effects. The coefficient estimates of variable post_2005 across the 4 columns are qualitatively the same and stay significantly positive at 1% level, consistent with evidence shown by Figure 1 and Table 1 that on average, CEO pay gap increases notably after We next examine whether the effect of CEO pay gap on debt structure changes after the increase in CEO pay gap. We choose balance sheet variable proportion of short-term debt (ST3) as the dependent variable in this set of regressions given it is relatively stable over time. Due to the decreased importance of CEO delta and CEO vega after 2005, we expect that the effect of CEO pay gap on debt structure becomes more pronounced after We test the change in effect of CEO pay gap on debt structure using the interaction term between CEO pay gap and variable post_2005. We find that the coefficient estimate of this interaction variable significantly negative at 1% level with or without CEO delta and CEO vega, consistent with the idea that CEO pay gap becomes more important in post-fas 123R era.

25 V. Conclusion In this study, we build on prior literature which suggests CEO pay gap and firm value are positively related (Kale, Reis, and Venkateswaran (2009) and Zhang (2013)) and further examine the impact of CEO pay gap on debt structure. We start our analysis by examining how CEO pay gap affects firm's default risk as measured by distance to default and find that default risk decreases in CEO pay gap. Previous literature shows that bondholders can incorporate various features into debt contracting to constrain managerial risk-taking preferences and to protect their own interests. We therefore examine the effect of CEO pay gap on various features of debt contracts. We find robust evidence that CEO pay gap is positively associated with debt maturity and negatively associated with cost of debt and number of debt covenants. Our finding is robust to the control of CEO equity-based compensation and corrections for potential endogeneity concerns. The finding of our research has important implications for executive compensation design and debt contracting.

26 References Altman, Edward I., 1977, The Z-score Bankruptcy model: past, present, and future, Financial Crises, New York 1977, Barclay, Michael J., and Clifford W. Smith, 1995, The maturity structure of corporate debt, the Journal of Finance 50, Begley, Joy, and Gerald A. Feltham, 1999, An empirical examination of the relation between debt contracts and management incentives, Journal of Accounting and Economics 27, Bharath, Sreedhar T., and Tyler Shumway, 2008, Forecasting default with the Merton distance to default model, Review of Financial Studies 21, Billett, Matthew T., TAO HSIEN D. KING, and David C. Mauer, 2007, Growth opportunities and the choice of leverage, debt maturity, and covenants, The Journal of Finance 62, Billett, Matthew T., David C. Mauer, and Yilei Zhang, 2010, Stockholder and bondholder wealth effects of CEO incentive grants, Financial Management 39, Black, Fischer, and Myron Scholes, 1973, The pricing of options and corporate liabilities, The journal of political economy, Carpenter, Jennifer N., 2000, Does option compensation increase managerial risk appetite? The journal of finance 55, Chava, Sudheer, Praveen Kumar, and Arthur Warga, 2010, Managerial agency and bond covenants, Review of Financial Studies 23,

27 Chava, Sudheer, and Amiyatosh Purnanandam, 2010, Is default risk negatively related to stock returns? Review of Financial Studies, hhp107. Chava, Sudheer, and Michael R. Roberts, 2008, How does financing impact investment? The role of debt covenants, The Journal of Finance 63, Coles, Jeffrey L., Naveen D. Daniel, and Lalitha Naveen, 2006, Managerial incentives and risktaking, Journal of Financial Economics 79, Core, John, and Wayne Guay, 2002, Estimating the value of employee stock option portfolios and their sensitivities to price and volatility, Journal of Accounting Research 40, Daniel, Naveen, J. S. Martin, and Lalitha Naveen, 2004, The hidden cost of managerial incentives: Evidence from the bond and stock markets, Available at SSRN Datta, Sudip, MAI ISKANDAR DATTA, and Kartik Raman, 2005, Managerial stock ownership and the maturity structure of corporate debt, the Journal of Finance 60, Frank, M., and V. Goyal, 2007, Corporate leverage adjustment: How much do managers really matter. Hayes, Rachel M., Michael Lemmon, and Mingming Qiu, 2012, Stock options and managerial incentives for risk taking: Evidence from FAS 123R, Journal of Financial Economics 105, Kale, Jayant R., Ebru Reis, and Anand Venkateswaran, 2009, Rank Order Tournaments and Incentive Alignment: The Effect on Firm Performance, The Journal of Finance 64,

28 Kini, Omesh, and Ryan Williams, 2012, Tournament incentives, firm risk, and corporate policies, Journal of Financial Economics 103, Knopf, John D., Jouahn Nam, and John H. Thornton Jr, 2002, The volatility and price sensitivities of managerial stock option portfolios and corporate hedging, The Journal of Finance 57, Masulis, Ronald W., and Shage Zhang, 2012, Compensation Gaps among Top Executives: Evidence of Tournament Incentives or Productivity Differentials? Available at SSRN Merton, Robert C., 1974, On the pricing of corporate debt: The risk structure of interest rates*, The Journal of Finance 29, Merton, Robert C., 1973, Theory of rational option pricing, The Bell Journal of Economics and Management Science, Shaw, Kenneth W., 2012, CEO incentives and the cost of debt, Review of Quantitative Finance and Accounting 38, Sundaram, Rangarajan K., and David L. Yermack, 2007, Pay me later: Inside debt and its role in managerial compensation, The Journal of Finance 62, Zhang, Shage, 2013, Pay Gap among Executives and Firm Value, Available at SSRN