Title: Voluntary Clawback Provisions and Executive Risk-taking Behavior. The Catholic University of Eastern Africa, Nairobi, Kenya.

Transcription

1 Title: Voluntary Clawback Provisions and Executive Risk-taking Behavior Author 1: Affiliation: Henry Kimani Mburu The Catholic University of Eastern Africa, Nairobi, Kenya Tel: Author 2: Alex P. Tang * Affiliation: Morgan State University, Baltimore, US Tel: alex.tang@morgan.edu Abstract Clawback Provisions are corporate governance mechanisms intended to reduce executive risktaking and deter opportunistic behavior. Though voluntary clawback provisions have been associated with several positive consequences in extant literature, there is little empirical evidence of their effect on executive risk-taking and deterrence of opportunistic behavior. On the contrary, recent clawback research shows that clawback firms engage more in real activities earnings management suggesting increasing risk-taking and opportunistic behavior in clawback firms. Additionally, there is a need to better understand the consequences of voluntary clawback adoption as the debate on mandatory clawback provisions for all publicly traded companies in the U.S. rages on. In this study, we focus on voluntary restatement triggered clawback provisions, and examine whether clawback adoption leads to changes in executives risk-taking behavior. We use a difference-in-difference research design on a propensity score-matched sample of 418 firms for the period Our findings are consistent with more risk-averse executives in clawback firms compared to those in non-clawback firms. Furthermore, we find empirical evidence suggesting that CFOs are more risk averse than CEOs. Keywords: Executive Risk-taking, voluntary clawback, firm-initiated clawback, information environment, Vega * Presenter

2 INTRODUCTION In this study, we extend the growing research on firm-initiated restatement triggered clawback provisions in executive compensation. Specifically, we examine the impact of firminitiated, or voluntary restatement triggered clawback adoption on the risk-taking behavior of executives. We focus on the executive risk-taking behavior and use both direct and indirect measures of executive risk-taking. The direct measures are executive Vega and proportion of unexercised exercisable in-the-money stock options and the indirect measures are the proportion of bonus to total compensation, the executive total compensation, and engagement in real activities earnings management. Executive compensation and its accompanying effects on managerial risk-taking behavior has been an issue of great concern to researchers (Coles, Daniel, and Naveen 2006; Cohen, Dey, and Lys 2013). This is because, major scandals and market failures have partly been blamed on problems associated with corporate governance (Agrawal and Chandha 2005). The assumption is that existing governance mechanisms has failed to adequately deal with the attendant principal-agent relationship problems. Regulators have responded to these problems by enacting laws to ensure good governance and disclosure. The Sarbanes-Oxley Act of 2002 (SOX) has introduced many regulations to enhance corporate governance. Additionally, the SOX has introduced regulations in relation to executive compensation under Section 304. These regulations are commonly referred to as clawback provisions. Also, the Dodd-Frank Wall Street Reform and Consumer Protection Act (simply referred to as the Dodd-Frank Act) has made various recommendations that are meant to enhance corporate governance through clawback provisions. The Securities Exchange Commission (SEC) is required to provide guidelines on the implementation of clawback provisions in listed companies in the U.S. Extant literature describes clawback provisions as corporate governance mechanisms that are intended to deter executive risk-taking (Dehaan, Hodge, and Shelvin 2013), reduce managerial opportunistics behavior (Iskandar-Datta and Jia 2013), limit risk of manipulation, penalize bad behavior and prevent windfall earnings (Earlie and Wilkerson 2012). Furthermore, researchers use two hypotheses to explain voluntary clawback adoption, the causal hypothesis 2

3 (Chan, Chen, Chen, and Yu 2012) and the signaling hypothesis (Chan et al. 2012; Dehaan et al. 2013; Iskandar-Datta and Jia 2013). On one hand, the causal hypothesis argues that clawback adoption would lead to changes in executive behavior resulting in certain positive outcomes as mentioned hereunder. On the other hand, the signaling hypothesis argues that clawback adoption may be used either to signal the market about the high quality corporate governance, or to manage the impression of management to outsiders. Neither of these two reasons would be expected to lead to behavior change of the executives. Prior studies have documented that clawback adoption is associated with positive outcomes including: higher earnings quality, reduction in restatements (Dehaan et al. 2013), perception of lower audit risk (Chan et al. 2012), and better quality corporate governance (Iskandar-Datta and Jia 2013). Other studies also find that managers of clawback-adopting firms use less accruals earnings management, but more real activities earnings management (Chan, Chen, and Chen 2013; Yu 2013; Chan, Chen, and Yu 2014). However, to the best of our knowledge, the impact of voluntary clawback adoption on executive risk-taking behavior has not been directly examined. There is conflicting evidence on the issue of the relation between clawback adoption and executive risk-taking. On one hand, adoption of clawback provisions is envisaged to reduce executive risk-taking (Dehaan et al. 2013; Chan et al. 2014). On the other hand, if adoption of clawback provisions were merely for impression management according to the signaling hypothesis (Chan et al. 2012), it would be unreasonable to expect any change in executive risktaking. In addition, recent research indicates that executives in clawback firms engage more in earnings management compared to those in non-clawback firms. This finding is against the expectation of executives being less risk-taking. We, therefore, seek to reconcile these seemingly contradicting research findings. Our study extends this research by investigating whether firm-initiated restatement triggered clawback adoption is associated with changes in executive risk-taking behavior. In addition, we also examine whether clawback adoption has an impact on risk-taking behavior of executives. Assuming the causal hypothesis of clawback adoption, change in the executive risk- 3

4 taking behavior would be a reasonable expectation. Specifically, we conjecture that in the presence of clawback provisions, executives would be risk-averse. Our argument is motivated by the Prospect Theory (Kahneman and Tversky 1979; Tversky and Kahneman 1986) and the Behavioral Agency Theory (Wiseman and Gomez-Mejia 1998). These theories explain the agent s behavior when governance mechanisms shift risk to the agent as in the case of restatement triggered clawback provisions. We use two direct measures of executive risk-taking: 1) Vega (Coles et al. 2006; Low 2009), and 2) the proportion of unexercised in-the-money stock options (Huddart 1994; Hall and Murphy 2002; Brisley 2006). In addition, we use indirect measures of executive risk-taking: executive total compensation and the proportion of executive bonuses to total compensation. Our study builds on prior studies that find improvement in information environment and quality of corporate governance after clawback adoption (Dehaan et al. 2013; Iskandar-Datta and Jia 2013), decline in restatements (Dehaan et al. 2013), and reduced perception of audit risk (Chan et al. 2012). However, our study differs from these studies in that we focus on executive risk-taking behavior which has not been examined in prior studies. Furthermore, we use Vega and unexercised exercisable stock option which are direct measures of executive risk-taking. Our study contributes to the extant literature on voluntary clawback adoption in several ways. First, it enhances our understanding of the consequences of voluntary clawback adoption. We use the difference-in-difference research design and are therefore able to test for causal relationships between clawback adoption and executive risk-taking. This provides additional evidence of the consequences of voluntary clawback adoption. Furthermore, our study contributes to the existing literature by directly testing the changes in executive risk-taking. This is important because as Cohen et al. (2013) indicate, there is a need to assess to what extent corporate governance regulations impact operation and investment strategy through interaction with the executives incentives. In addition, although Dehaan et al. (2013) indicate that clawback provisions are meant to discourage executive risktaking, to date, there is little direct empirical evidence to support this claim. We have organized our paper as follows: in the following section we review extant literature and develop our hypotheses, then we explain our research design and methodology, 4

5 present our empirical results, and discuss our findings. The conclusions and suggestions for further research are presented in the last section. LITERATURE REVIEW AND HYPOTHESES DEVELOPMENT Clawback Provisions Clawback provisions, also referred to as recoupment clauses or simply clawbacks, have been used in executive compensation for a long time (Langevoort 2007). Clawback provisions are the rights of a firm to recoup from an executive of the firm benefits already earned and paid following a predetermined triggering event (Earlie and Wilkerson 2012). Such triggering events include: misconduct, breach of restrictive covenants such as non-competition and nonsolicitation agreements, influence on employees, termination of employment due to unethical behavior, and financial statement restatement. The U.S. Securities and Exchange Commission (SEC, 2003) describes the adoption of clawback provisions through enactment of the Sarbanes-Oxley Act (SOX) Section 304 as an action to improve the tone at the top. According to Robert Khuzami, Director of the SEC s Division of Enforcement, clawback provisions are: An important incentive for senior executives to be vigilant in preventing misconduct and ensuring that companies comply with financial reporting requirements. 1 Earlie and Wilkerson (2012) identify three main purposes of clawback provisions: limiting the risk of manipulation, penalizing bad behavior, and preventing windfall earnings. 2 In addition, Earlie and Wilkerson (2012) indicate that, broadly speaking, clawback provisions permit recovery of compensation regardless of misconduct on the part of the executive. Dehaan et al. (2013) add that clawback provisions are corporate governance mechanisms meant to deter executives from risk-taking behavior that would result in restatements, and at the same time penalize the executives if they engage in such behavior. Furthermore, Iskandar-Datta and Jia (2013) add that the key benefits of clawback adoption are reduction of managerial opportunistic behavior and assurance of high quality financial reporting. 1 The Director was commenting on the clawback enforcement case of Heazer Homes USA. Press Release available at 2 Emphasizing this point, Robert Khuzami, director of the SEC s Division of Enforcement said, "The personal compensation received by CEOs... can be clawed back. The costs of such misconduct need not be borne by shareholders alone." In addition, a quote from filing DEF 14A of Mineral Technologies Inc. dated 04/03/2013,... ensure that our executives do not retain undeserved windfalls... illustrates this point too. 5

6 Clawback provisions have thus created a growing interest in both research and practitioner publications in accounting, law, and corporate governance in recent years. A stream of accounting research has been evolving to better understand both determinants (Brown, Davis- Friday, and Guler 2011; Addy, Chu, and Yoder 2014; Brown, Davis-Friday, and Guler 2014) and consequences (Chan et al. 2012; Chan et al. 2013; Dehaan et al. 2013; Iskandar-Datta and Jia 2013; Mariola and Ryan 2013; Chan et al. 2014) of voluntary or firm-initiated clawback provisions. The majority of these studies have documented positive consequences of clawback adoption. These include higher financial reporting quality and reduced analysts forecast dispersion (Dehaan et al. 2013), positive stock-valuation effects (Iskandar-Datta and Jia 2013), decline in restatements, and a perception of lower audit risk (Chan et al. 2012). Furthermore, firms whose management has a monitoring orientation are more likely to adopt clawback provisions compared to those whose management has an entrenchment orientation (Addy et al. 2014). Adoption of clawback provisions also results in an oversight impression and could reduce the cost of recovery of paid compensation from errant executives (Addy et al. 2014). Beck (2012) finds that, conditional on the quality of corporate governance, firms which adopt clawback provisions are associated with lower income-increasing accrual earnings management. She further argues that clawback adoption may be viewed as a strong governance mechanism. Firm-initiated clawback provisions have many different triggers, including but not limited to: restatements, breach of non-solicitation and non-competition covenants, termination for cause, resignation, change of control, intentional misconduct, fraud, negligence, and breach of fiduciary duty. In this study, we focus only on restatement triggered clawback provisions, which are the subject of the SOX Section 304 and Section 954 of the Dodd-Frank Act. Firm-initiated Clawback Adoption and executive risk-taking Adoption of firm-initiated clawback provisions is explained by two theories in the extant literature. One, the causal hypothesis which envisages changes in executive behavior upon adoption and two, the signaling hypothesis that suggests that management may not be committed to the implementation of clawback provisions. A behavioral approach to clawback adoption may help to extend the causal hypothesis and directly link it to executive risk-taking behavior. We, therefore, use two behavioral theories to motivate this link. First, the Prospect Theory 6

7 (Kahneman and Tversky 1979) identifies two phases in the process of decision making under uncertainty (Tversky and Kahneman 1986) whose outcomes are expressed as positive (gains) or negative (losses) based on a neutral reference outcome assigned the value of zero. Thus, according to the Prospect Theory, agents behave differently depending on whether their prospects are gains, when they are said to be in a loss domain, or losses, when they are said to be in a gain domain. Accordingly, agents risk-taking behavior does not remain constant with changing prospects as would be predicted by the expected utility theory. Agents are risk-averse in the gain domain and risk-seeking in the loss domain according to the Prospect Theory. Second, according to the Behavioral Agency Theory (Wiseman and Gomez-Mejia 1998), executives will avoid risky activities that potentially threaten their endowed wealth and will make a choice of activities depending on the contract framing. Wiseman and Gomez-Mejia (1998) explain that contracts are positively framed when the available options generally promise desirable outcomes and are negatively framed when the options generally promise undesirable outcomes. Executives can, therefore, exhibit both risk-seeking and risk-averse behavior depending on whether the contracts are positively or negatively framed. Restatement triggered clawback provisions introduce negative framing in executive compensation contracts, placing managers in a gain domain according to the Prospect Theory (Tversky and Kahneman 1986). Restatement triggered clawback provisions also increase the executives risk-bearing. Risk-bearing results when corporate governance mechanisms transfer risks to the agent and places part of the agent s income at risk (Wiseman and Gomez-Mejia 1998). Agents with high risk-bearing would be more risk-averse. Accordingly, from a behavioral perspective, executives in restatement triggered clawback adopting firms would be expected to change their risk-taking behavior. This is because restatement triggered clawback provisions effectively place the executives in a gain domain exposing them to the risk of losing their endowed wealth. Such executives would, therefore, behave in such a manner to protect their potentially endowed wealth and general welfare. This argument is consistent with Tversky and Kahneman (1986) who indicate that risk-averse agents are more sensitive to losing than to gaining wealth. Hence the adoption of restatement triggered clawback provisions would directly affect the executive risk-taking behavior consistent with the causal hypothesis of voluntary clawback adoption. 7

8 Prior studies have used Vega, a direct measure of executive risk-taking, (Coles et al. 2006; Low 2009) to examine changes in executive risk-taking. Vega is defined as the percentage change in the executive wealth given a unit percentage change in the standard deviation of the firm s daily returns. Executives with higher Vega are known to undertake higher risk policy decisions (Coles et al. 2006). We, therefore, conjecture that executives in clawback firms will have low Vega compared to those in non-clawback firms consistent with our argument that they should be more risk-averse. This leads us to our first hypothesis: H1: All else being equal, CEOs and CFOs in clawback firms are more likely to have lower Vega compared to those in non-clawback firms. Accounting research has also examined components of executive compensation in relation to executive risk-taking behavior. Results show that executives who have higher stock option components in their compensation have higher incentives to engage in earnings management. Cheng and Warfield (2005) and Bergstresser and Philippon (2006) find a positive and significant association between executive stock options and earnings management. In addition, finance research documents that rational risk-averse executives exercise more in-themoney stock options earlier than the risk-seeking executives (Huddart 1994; Hall and Murphy 2002; Brisley 2006). Hence, for the risk-averse executives who sign compensation contracts with clawback provisions, reasonable strategies of lowering risk would be to reduce the proportion of stock options and bonus components of their compensation and exercise more of the exercisable inthe-money stock options at the earliest opportunity. This meaningfully reduces the risk of losing the executives endowed wealth consistent with the Behavioral Agency Theory, in line with their risk-averseness. In addition, prior research finds that executive risk-taking behavior is affected by contract framing (Christ, Sedatole, and Towry 2012). Because executive compensation contracts in clawback firms are negatively framed, it is reasonable to expect changes in executive risk-taking in line with this research. These arguments lead to our second hypothesis: H2: All else being equal, CEOs and CFOs in clawback firms are more likely to have a lower proportion of unexercised exercisable in-the-money stock options and bonuses to total compensation compared to those in non-clawback firms. 8

9 RESEARCH DESIGN AND METHODOLOGY Data Sources We retrieve data from multiple public sources: fundamentals data from COMPUSTAT, analyst forecast data from Institutional Brokers Earnings Services (I/B/E/S), auditor and internal control material weaknesses data from Audit Analytics, and market data from CRSP. We obtain executive compensation data from ExecuComp and clawback data from CapitalIQ and the SEC filings. Data on board characteristics, stock holding, and executive tenure are all from the SEC filings. 3 Sample Construction We start the sample construction with clawback data for the period 2007 to In 2006, the SEC amended regulation S-K requiring clawback provision adoption to be disclosed on form DEF 14A. This makes the year 2007 to be a reasonable starting point. We start with 5,551 firms screened from CapitalIQ using the term clawback. We select firms in the initial sample using the following preliminary criteria: adoption of clawback provisions, non-financial and nonutility industry firms, data availability in COMPUSTAT, CRSP, I/B/E/S, and ExecuComp. This process leaves us with 713 firms that had adopted some form of clawback provisions in the period 2007 to We drop firms that have non-restatement triggered clawback provisions. This leaves us with 457 firms. We summarize our selection procedures in Table 1. Insert Table 1 here We then construct a one-to-one matched sample, of clawback and non-clawback firms, based on the probability of clawback adoption (propensity scores). We compute our propensity scores using a modified Dehaan et al. (2013) model 4. Our final sample has 418 firms, comprising 209 clawback firms and 209 non-clawback firms for the period 2010 to 2013 comprising 1,530 firm-year observations. 3 We confirm all clawback data from CapitalIQ screening with the SEC DEF 14A filings. In addition, we also use the following SEC filings: DEFA14A, DEFM14A, DEF 14C, 10-K, and 20-F for some firms. 4 Results from this estimate (not tabulated) have N = 30,114, pseudo-r 2 of and a maximum rescaled pseudo- R 2 of All variables used in the model are defined in Appendix A. 9

10 Research Design We first conduct univariate analysis for preliminary examination of each hypothesis. We conduct t-test for differences in means, Wilcoxon test for differences in medians, and the Pearson s correlation analysis. We then estimate the multivariate OLS regression models to test our two hypotheses. To check for multicollinearity concerns, we measure the variance inflation factor (VIF) in all the multivariate models in addition to assessing the size and significance of Pearson s correlation coefficients between independent variables. Because the distribution by stock exchange and S&P index show clustering of clawback and non-clawback firms in the final sample, we cluster standard errors by stock exchange and S&P index. Furthermore, there is a large body of research that documents that though both CEOs and CFOs show opportunistic behavior, their incentive levels differ (Chava and Purnanandam 2010; Jiang, Petroni, and Wang 2010; Friedman 2014). We, therefore, contend that changes in risk-taking behavior may also differ between CEOs and CFOs. To examine this prediction, we conduct additional analysis and estimate all the models using the seemingly unrelated regression (SUR) technique to compare the differences in the coefficients of the variables CLAWBACK and AFTER*CLAW. We now explain the models for each hypothesis. Executive risk-taking and Clawback Adoption To examine the relationship between executive risk-taking and clawback adoption in hypotheses H1 and H2, we use two direct measures of executive risk-taking: a) Executive Vega and b) Unexercised exercisable in-the-money stock options. We also use two indirect measures of executive risk-taking: a) executive total compensation and b) proportion of executive bonuses to total compensation. We estimate the following models to examine these relationships: VEGA i = α + β 1 CLAWBACK i + β 2 AFTER i + β 3 AFTER*CLAW i + β 4 MKV i + β 5 LNANFOL i + β 6 RDV i + β 7 LEV i + β 8 PERF i + β 9 CAPEX i + β 10 DELTA i + β 11 COMP i + β 12 TENURE i + ε i ---- (Model 1) VEGA is the change in executive total compensation scaled by the percentage change in the standard deviation of daily returns for one year period. CLAWBACK is equal to 1 if the firm has adopted clawback and zero otherwise, AFTER is equal to 1 for all years after clawback adoption and zero otherwise; MKV is log of market value of equity, PERF is net income before 10

11 extraordinary items scaled by lagged total assets, LEV is long-term debt scaled by lagged total assets; LNANFOL is the natural log of the number of analysts following the firm; RDV is ( R&D + advertising expenses) lagged by total operating expenses; CAPEX is the capital expenditure scaled by total assets; DELTA is measured as the log of change in executive compensation scaled by percentage change in the stock price; COMP is the log of the executive total compensation and TENURE is the executive tenure. PAY i = α + β 1 CLAWBACK i + β 2 AFTER i + β 3 AFTER*CLAW i + β 4 MKV i + β 5 PERF i + β 6 LEV i + β 7 MTB i + β 8 RTENURE i + β 9 ANFOL i + β 10 DELTA i + µ j IND + δt (Model 2) n YR t + ε i PAY is measured in three different ways: (1) Total executive compensation, COMP, (2) Proportion of executive bonus to the total compensation, PBONUS, and (3) Proportion of unexercised exercisable in-the-money stock options. MTB is the ratio of market to book; RTENURE is executive tenure/average tenure of all independent directors during the year; DELTA is measured as the log of change in executive compensation scaled by percentage change in the stock price; FIXED IND and YR are the industry and year dummies respectively. All other variables are defined as in Model 1. Univariate analysis EMPIRICAL RESULTS AND DISCUSSION We show the results from the test of differences in means and medians in Table 2. We also incorporate the difference-in-difference (DiD) results in Table 2. Results show significant differences between clawback firms and non-clawback firms in executive compensation measures, real activities earnings management measures, and executive Vega. For example, t- values for differences in total compensation are for the CEO and for the CFO while t-values for differences in proportion of bonuses are for the CEO and for the CFO. In addition, we find significant differences in executive total compensation, proportion of executive bonuses to total compensation, executive Vega before and after clawback adoption (Table 2 column DiD). Moreover, the DiD results provide preliminary evidence that differences in real activities earnings management, executive compensation structure, and executive Vega may directly attributed to clawback adoption. 11

12 Insert Table 2 here Pearson s Correlation Analysis Correlation analysis results do not show significant correlation between the measure of executive of risk-taking, Vega, and CLAWBACK. This finding suggests that risk-taking has no association with clawback, which is contrary to our prediction. However, there is a positive and significant correlation between AFTER*CLAW and the CEO Vega (p =.0384). This suggests that CEO risk-taking changes in the period after clawback adoption which supports our contention for this study. The proportion of the CEOs unexercised exercisable in-the-money options is negatively and significantly correlated to the CLAWBACK variable. The correlation with the AFTER*CLAW interaction term is also negative but not significant. For the CFOs, the proportion of the unexercised exercisable in-the-money stock options to total compensation is negative but not significant. However, it is positively correlated with the interaction term AFTER*CLAW. The negative correlation (r = -.07 (p =.0286) for the CEO and r = -.02 (p =.5229 for the CFO) between the proportions of the unexercised exercisable stock options to the total compensation and CLAWBACK provides preliminary evidence of the risk-averseness of the executives as documented in the finance literature (Huddart 1994; Hall and Murphy 2002; Brisley 2006). The proportion of executive bonuses and stock options to the total compensation is lower in clawback firms compared to non-clawback firms. In addition, there is high positive correlation between CEO and CFO proportion of bonuses (r =.77, p <.0001) and the proportion of unexercised exercisable in-the-money options (r =.20, p <.0001). These findings support the view that it is more desirable to consider both the CEO and CFO together in the study. This is because the CEO and CFO incentives and risk-taking are associated (Jiang et al. 2010; Feng, Ge, Luo, and Shevlin 2011; Friedman 2014; Lefebvre and Vieider 2014). Multivariate Analysis Clawback Adoption and executive risk-taking We estimate models 1 and 2 to test hypotheses H1 and H2 respectively and the results are presented in Tables 3 and 4. We, first, examine the association between the executive Vega and clawback adoption to test H1. Additionally, we also examine indirect measures of executive 12

13 risk-taking as predicted by the Prospect Theory and the Behavioral Agency theory. The number of observations reduces significantly after lagging and merging ExeCuComp compensation data with CRSP market data. We show the results in Table 3. Insert Table 3 here Consistent with our prediction, we find a negative and significant association between Vega and the variable CLAWBACK for both CEO (p =.0359) and CFO (p =.0210) models. This result suggests the existence of clawback provision influences the risk-taking behavior of CEOs and CFOs. In both CEO and CFO models, the coefficients for the AFTER*CLAW variable are positive but not significant (p-values are.8623 and.3597 respectively). According to Coles et al. (2006), executives with a lower Vega have a decreasing likelihood of implementing riskier policy choices. We then test hypothesis 2 by estimating model 2 and the results are presented in Table 4 and Table 5. Insert Table 4 here Table 4 shows the results for the CEO and the CFO proportion of unexercised exercisable stock options to total compensation. The coefficient β 1 on the CLAWBACK variable is not significant in either the CEO or the CFO model. However, the coefficient β 3 on the AFTER*CLAW variable is negative and significant at the ten percent level for the CEO model and is not significant for the CFO model. Results for the proportion of unexercised exercisable in-the-money stock options, show that the CEOs have significantly lower unexercised exercisable in-the-money stock options in the post-clawback adoption period compared to the period before. The coefficient β 3 is negative and significant at the 10 percent level in the CEO model while it is positive but not significant in the CFO model. The coefficient β 1 is not significant in either the CEO or CFO model. These results provide partial support for hypothesis H2 to the extent that it is significant for the CEO model. In addition, they are consistent with pertinent literature in finance that documents that risk-averse executives exercise their in-themoney stock options earlier than risk-seeking executives (Huddart 1994; Hall and Murphy 2002; Brisley 2006). We show the results for the two indirect measures, proportion of executive bonus, PBONUS, and executive total compensation, COMP in Table 5. These results show that 13

14 executive total compensation including that of the CEOs and CFOs is significantly higher in clawback firms compared to that in non-clawback firms. In addition, the coefficient β 3 on the interaction term, AFTER*CLAW is not significant in either the CEO or the CFO total compensation model. Overall, these results support prior research that documents higher executive total compensation in clawback firms compared to that in non-clawback firms. This is in line with increasing risk-bearing of executives in clawback firms as predicted by The Behavioral Agency Theory. Insert Table 5 here Multivariate analysis for the proportion of bonuses shows that the coefficient β 1 on the variable CLAWBACK is negative, as expected, but not significant for both the CEO and the CFO models. The coefficient β 3 on the variable AFTER*CLAW is negative and significant at the five percent level for the CFO model but not significant for the CEO model. This suggests that in the post-clawback adoption period, the proportion of executive bonuses to the total compensation decrease. This result is consistent with our prediction for hypothesis H2. Taken together, these results provide further support for hypothesis H2 and for our contention that executives in clawback firms would be more risk-averse compared to those in non-clawback firms. These findings support the thesis of our study that executives of firms that adopt firm-initiated restatement triggered clawback provisions are more risk-averse compared to the executives in non-clawback firms. Moreover, the finding that the total compensation for both CEOs and CFOs is significantly higher in clawback firms than in non-clawback firms is consistent with higher risk-bearing executives in clawback adopting firms as predicted by the Behavioral Agency Theory (Wiseman and Gomez-Mejia 1998). These findings also support prior research findings by Sawers, Wright, and Zamora (2011) that those executives whose earnings are at risk are more likely to be less risk-taking. Difference-in-difference results do not, however, support a causal relationship between total compensation and clawback adoption because the coefficient β 3 is not significant. Overall, we find adequate support for hypothesis H1 and H2. While the executive total compensation increases in clawback firms, the proportion of bonuses significantly decreases. In addition, executives in clawback firms exercise significantly more of the in-the-money exercisable options as a proportion of their total compensation, a further indication of their risk- 14

15 averseness. Together, these findings suggest changes in risk-taking behavior of the executives in clawback firms consistent with risk-averseness. Additional Analysis Earnings management is a risk-taking behavior. We, therefore, conjecture less engagement in real activities management is clawback firms all else being constant to avoid any possibility of triggering clawback provisions (Ettredge, Scholz, Smith, and Sun 2010). Our analysis (not tabulated) shows that clawback firms engage less in manipulation of discretionary expenses but more in production costs and cash flow from operations compared to non-clawback firms. These findings are both consistent with our other findings and support prior evidence of increasing earnings quality in clawback firms. Prior research has also documented differences in CEO and CFO incentives to engage in opportunistic behavior and influence corporate policies (Chava and Purnanandam 2010; Feng et al. 2011; Kim, Li, and Zhang 2011), and also to influence audit scope (Kannan, Skantz, and Higgs 2014). Other research also documents that the CFOs may be under undue pressure from powerful CEOs to manipulate earnings with implications on compensation, reporting, firm value, and the information environment (Friedman 2014). Overall, the literature suggests that any changes in executive-risk taking would differ between CEOs and CFOs. Therefore, we use the seemingly unrelated regression (SUR) technique to further examine and test for differences in the coefficients of the variables CLAWBACK and AFTER*CLAW in the real activities earnings management, pay, and Vega models. The results are presented in Table 6. Insert Table 6 here Panel A of Table 6 shows that executives use real activities management tools to different extents. Abnormal production costs, abnormal cash flow from operations, and abnormal discretionary expenses may be used to different extents to meet different objectives. For the real activities earnings management models, the SUR results show significant differences in the coefficients of the variable AFTER*CLAW, while the coefficients on CLAWBACK are not significantly different. This result suggests that differences in the way different real activities earnings management tools are used differs significantly after clawback adoption which is in consistent with changing executive risk-taking behavior. 15

16 The SUR results for the pay models show significant differences in the coefficients on both variables CLAWBACK and AFTER*CLAW except for coefficient β 1 which is not significant in the unexercised exercisable in-the-money stock options models. These results suggest that there are significant differences between the CEO and CFO both between clawback and non-clawback firms and in the periods before and after clawback adoption. Furthermore, these results suggest that the changes in the risk-taking behavior differ between the executives. This result is consistent with prior research in finance mentioned earlier and suggests that the risk-taking behavior of the CEOs is different from that of the CFOs. In addition, the results show that there is a significant difference between the CEO Vega and CFO Vega comparing clawback and non-clawback firms, which further support the above findings. To further examine the differences between the CEO and CFO risk-taking, we conduct a 2 x 1 analysis of variance (ANOVA) for the executive Vega. The ANOVA results are shown in Panel B of Table 6. Results show that there is a significant difference at one percent level between the CEO risk-taking and CFO risk-taking. The Brown-Forsythe test for equality of means is also significant at one percent level. The sample mean CEO Vega is significantly higher than that of the CFO. This result suggests that the CFOs are more risk-averse compared to the CEOs. This is in support of prior research findings that CFOs are more involved in risktaking behavior (Jiang et al. 2010). We attribute this to the possible coercion from powerful CEOs (Friedman 2014). CONCLUSIONS In this study, we examine the effect of clawback adoption on executive risk-taking behavior. We find empirical support for our two hypotheses, H1 and H2. Results show that executives in clawback firms have lower proportions of bonuses to total compensation and exercise more of the in-the-money stock options compared to those in non-clawback firms. In addition, we find that executives in clawback firms have lower Vega. 5 These findings lead to our conclusion that voluntary clawback adoption is associated with changing executive risk-taking behavior. Specifically, executives are more risk-averse in firms that voluntarily adopt 5 An alternative interpretation of this finding may be the positive association between R&D expenditure and the CEO compensation documented by Cheng (2004). However, other findings in support of risk-averse executives seem to be the more compelling view. 16

17 restatement triggered clawback provisions. In addition, we find differences in the degree of riskaversion between CEOs and CFOs. This is consistent with prior findings in existing literature. These results are, therefore, consistent with the risk-averse behavior of executives in clawback firms as anticipated by the Prospect and Behavioral Agency theories. Overall, our results show that executives in clawback firms are more risk-averse than executives in non-clawback firms. They are, nevertheless, more keen to meet short-term targets even if it is at the expense of long-term firm-value as documented in existing literature (Graham, Harvey, and Rajgopal 2005). This reduces their own risks but not necessarily the risks associated with long-term firm-value. 17

18 References Addy, N., X. Chu, and T. Yoder Voluntary adoption of clawback provisons, corporate governance, and interlock effects. Journal of Accounting and Public Policy 33(2): Agrawal, A., and S. Chandha Corporate governance and accounting scandals. Journal of Law and Economics 48(2): Beck, A. K Clawback provisions: How sharp are the claws? An analysis of the deterrence effectiveness of voluntary clawback provisions: Available at: Bergstresser, D., and T. Philippon CEO incentives and earnings management. Journal of Financial Economics 80(3): Brisley, N Executive stock options: Early exercise provisions and risk-taking executives. The Journal of Finance 61(5): Brown, A. B., P. Y. Davis-Friday, and L. Guler Economic determinants of the voluntary adoption of clawback provisions in executive compensation contracts. NY: Available at 17Aug2011.pdf. Brown, A. B., P. Y. Davis-Friday, and L. Guler M&A performance and the voluntary adoption of clawback provisions in executive compensation contracts. NY: Available at Chan, L. H., K. C. Chen, T.-Y. Chen, and Y. Yu The effects of firm-initiated clawback provisions on earnings quality and auditor behavior. Journal of Accounting and Economics 54(2/3): Chan, L. H., K. C. W. Chen, and T.-Y. Chen The effects of firm-initiated clawback provisions on bank loan contracting. Journal of Financial Economics 110(3): Chan, L. H., K. C. W. Chen, and Y. Yu Substitution between real and accruals-based earnings management after voluntary adoption of compensation clawback provisions. The Accounting Review (forthcoming). 18

19 Chava, S., and A. Purnanandam CEOs versus CFOs: Incentives and corporate policies. Journal of Financial Economics 92: Cheng, Q., and T. D. Warfield Equity incentives and earnings management. The Accounting Review 80(2): Cheng, S R&D expenditures and CEO compensation. The Accounting Review 79(2): Christ, M. H., K. L. Sedatole, and K. L. Towry Sticks and carrots: The effect of contract frame on effort in incomplete contracts. The Accounting Review 87(6): Cohen, D. A., A. Dey, and T. Z. Lys Corporate governance reform and executive incentives. Implications for investments and risk taking. Contemporary Accounting Research 30(4): Cohen, D. A., and P. Zarowin Accrual-based and real earnings management activities around seasoned equity offerings. Journal of Accounting and Economics 50(1): Coles, J. L., N. D. Daniel, and L. Naveen Managerial incentives and risk-taking. Journal of Financial Economics 79(2): Dehaan, E., F. Hodge, and T. Shelvin Does voluntary adoption of a clawback provision improve financial reporting quality. Contemporary Accounting Research 30(3): Earlie, J. E., and A. Wilkerson Dodd-Frank clawback: hot issue for Journal of Investment Compliance 13(1): Ettredge, M., S. Scholz, K. R. Smith, and L. Sun How do restatements begin? Evidence of earnings management preceding restated financial reports. Journal of Business Finance & Accounting 37(3/4): Feng, M., W. Ge, S. Luo, and T. Shevlin Why do CFOs become involved in material accounting manipulations? Journal of Accounting and Economics 51(1/2): Friedman, H. L Implications of power: when the CEO can pressure the CFO to bias reports. Journal of Accounting and Economics 58(1): Graham, J. R., C. R. Harvey, and S. Rajgopal The economic implications of corporate financial reporting. Journal of Accounting and Economics 40(1/3):

20 Hall, B. J., and K. J. Murphy Stock options for undiversified executives. Journal of Accounting and Economics 33(1): Huddart, S Employee stock options. Journal of Accounting and Economics 18(2): Iskandar-Datta, M. Y., and Jia Valuation consequences of clawback provisions. The Accounting Review 88(1): Jiang, J., K. R. Petroni, and I. Y. Wang CFOs and CEOs: Who have the most influence on earnings management? Journal of Financial Economics 96: Kahneman, D., and A. Tversky Prospect theory: An analysis of decision making under risk. Econometrica 47(2): Kannan, Y. H., T. R. Skantz, and J. L. Higgs The impact of CEO and CFO equity incentives on audit scope and perceived risks as revealed through audit fees. Auditing: A Journal of Practice & Theory 33(2): Kim, J.-B., Y. Li, and L. Zhang CFOs versus CEOs: Equity incentives and crashes. Journal of Financial Economics 101(3): Langevoort, D. C On leaving corporate executives "naked, homeless and without wheels": Corporate fraud, equitable remedies, and the debate over entity versus individual liability. Wake Forest Law Review 42(3): Lefebvre, M., and F. M.Vieider Risk taking of executives under different incentive contracts: experimental evidence. Journal of Economic Behavior & Organization 97: Low, A Managerial risk-taking behavior and equity-based compensation. Journal of Financial Economics 92(3): Mariola, E., and H. Ryan Clawback provisions, firm performance, and risk shifting. Journal of Business and Accounting 6(1): Sawers, K., A. Wright, and V. Zamora Does greater risk-bearing in stock option compensation reduce the influence of problem framing on managerial risk-taking behavior?. Behavioral Research in Accounting 23(1): Securities and Exchange Commission (SEC) News. [Online] Available at: 20

21 Tversky, A., and D. Kahneman Rational choice and the framing of decisions. The Journal of Business 59(4): s252-s278. Wiseman, R. M., and L. R. Gomez-Mejia A behavioral agency model of managerial risktaking. Academy of Management Review 23(1): Yu, L.-H The effects of firm-initiated clawback provisions on earnings management. :UMI Dissertation publishing (UMI Number: ). 21

22 Variable definitions APPENDIX Variable ACFO APROD ADISX CZ1 (CZ2) AFTER CLAWBACK AFTER*CLAW ABSDA ANFOL LNANFOL MKV PERF LEV Definition Abnormal cash flow from operations, the difference between actual and normal operational cash flows Abnormal production costs, the difference between actual and normal production costs Abnormal discretionary expenses, the difference between actual and normal discretionary expenses Combined measure of real activities earnings management defined as CZ1 = (-1) x ADISX + APROD (CZ2 = (-1) x ACFO + (-1) x ADISX) consistent with Cohen and Zarowin (2010) A dummy variable equal to 1 for all fiscal years after clawback adoption and 0 otherwise. Equal to 1 if firm has adopted clawback provisions at the end of fiscal year and 0 otherwise An interaction term between the variables AFTER and CLAWBACK Absolute abnormal discretionary accruals computed using the modified Jones model Number of analysts following the firm, measured as the number of analysts forecasting earnings in I/B/E/S annual estimates file The natural log of the number of analysts (ANFOL) following the firm The natural logarithm of market value of equity Measure of performance given by net income before extraordinary items scaled by lagged total assets Measures leverage and is given by long term debt scaled by 22

23 MTB VEGA(DELTA) LCOMP PBONUS PoptUEx PoptFV IND YR RDV RETVAR CAPEX TENURE lagged total assets Measure of the firm s growth opportunities and is given by market value of equity divided by book value of equity Change in executive total compensation scaled by the percentage change in the standard deviation of daily returns (by the percentage change in stock price) for the year. Natural logarithm of the executive total compensation The proportion of bonus to total compensation of the executive The proportion of the executive s unexercised, exercisable inthe-money stock options to the total compensation The proportion of the executive s fair value of granted stock options to the total compensation Fama & French 48 industry classification dummy Year dummy Controls for expenditure on innovative activities; measured as (R&D + advertising expenses)/total operating expenses Standard deviation of the firm s daily returns for one year The capital expenditure scaled by total assets The executive tenure in years Table 1: Sample selection procedure Description Total Original Capital IQ screened list ,551 Drop private firms (28) (53) (95) (115) (133) (124) (54) (602) Public firms ,949 Drop duplicates in prior year lists (0) (23) (52) (80) (165) (286) (381) (987) Drop utility (SIC ) and financial (SIC ) firms (32) (58) (194) (254) (344) (399) (472) (1,753) Drops firms with missing data (18) (30) (61) (106) (211) (275) (346) (1,047) Firms with data in databases ,162 23

24 Drop non-clawback firms (8) (19) (34) (46) (115) (110) (117) (449) Firms with clawback provisions Drop non-restatement triggered (11) (12) (19) (28) (45) (69) (72) (256) clawback provisions Firms with restatement triggered clawback provisions Matched study firms a Matched control firms Final study sample 418 a We used observations for the period to compute propensity scores. Consistent with prior research (Dehaan et al. 2013), these observations should be excluded from model estimations for the study. 24

25 Table 2: Test of differences in the dependent variables Variable Clawback (1) Non-clawback (0) Differences (1 0) DiD a Mean Median Mean Median t-value z t-value z APROD * 2.43 *** ACFO ** * ADISX ** *** CZ * 2.23 ** 1.68 * 2.92 *** CZ ** ANFOL *** *** 8.34 *** 8.83 *** CEOCOMP *** *** 7.71 *** 7.15 *** CFOCOMP *** *** 6.73 *** 6.78 *** CEOCOMP CFOCOMP *** CEOPBONUS *** *** *** ** CFOPBONUS ** *** *** ** POptUeX_CEO *** POptUeX_CFO ** 1.66 * 3.62 *** POptFV_CEO *** 4.38 *** ** POptFV_CFO ** 3.15 *** ** CEOVEGA ** 2.12 ** 2.24 ** CFOVEGA ** a DiD refers to the difference-in-difference test statistics being means and medians after clawback adoption minus means and medians before clawback adoption CLAWBACK = 1 if the firm has adopted clawback and zero otherwise, AFTER*CLAW is an interaction between AFTER (AFTER = 1 for all years after clawback adoption and zero otherwise) and CLAWBACK; APROD is the abnormal production costs, ACFO is the abnormal cash flows from operations, ADISX is the abnormal discretionary expenses, CZ1 = APROD + - ADISX, and CZ2 = -ACFO + -ADISX; ACCURACY1 is the absolute difference between actual 25