The Impact of Organized Labor on CEO Inside Debt: Strategic Substitution or Incentive Alignment?

Transcription

1 The Impact of Organized Labor on CEO Inside Debt: Strategic Substitution or Incentive Alignment? Tao-Hsien Dolly King Department of Finance University of North Carolina at Charlotte Charlotte, NC Takeshi Nishikawa Business School University of Colorado Denver Denver, CO Andrew K. Prevost Grossman School of Business University of Vermont Burlington, VT Abstract We examine the impact of organized labor on the debt-like components of CEO compensation. We show that unionization intensity is significantly and robustly correlated with the proportion of debt-like compensation to equity incentives. Consistent with the strategic substitution view, this result is driven by the deferred component of debt-like compensation, suggesting managers substitute current- for deferred compensation in order to improve their bargaining position over labor. We find the speed of adjustment to the optimal relative debt ratio is slower when firms face strong unions, highlighting a potential consequence of strategic substitution. Finally, we provide evidence that a mechanism of strategic substitution is the interplay between cash bonus and deferred compensation, providing a new insight on the interrelationship of compensation components in the presence of strong unions.

2 1. Introduction The structure of executive compensation represents an important contracting mechanism when viewing the firm as a nexus of contracts. Motivated by the wealth-maximizing objective of financial stakeholders, a well-developed literature examines if incentive compensation effectively aligns shareholder and managerial interests (e.g., Murphy, 1985; Morck, Schleifer and Vishny, 1988; Guay, 1999; Coles, Daniel and Naveen, 2006), and if financial stakeholders such as institutional investors influence equity incentives (e.g. Yermack, 1995; Core, Holthausen and Larcker, 1999; Hartzell and Starks, 2003). Labor unions are unique yet important nonfinancial stakeholders whose objective is to maximize the wealth of their constituent members. In their role as collective bargaining agents, extent research shows that organized labor alters the contracting environment. Consistent with the view of some unions that executives capture an excessive share of firms returns (Banning and Chiles, 2007), anecdotal evidence and prior research find a negative association between union presence and compensation. For example, Banning and Chiles (2007) document a negative relation between unionization rate and total CEO compensation, while Gomez and Tzioumis (2011) find that CEOs of unionized firms have lower performance sensitivities than those of non-unionized firms. In recent work, Huang, Jiang, Lie and Que (2015) show highly unionized firms strategically reduce executive pay before union contract negotiations to potentially obtain concessions from unions, suggesting that there may be a shift in bargaining power between management and strong unions. Because top executives have considerable influence over their own pay arrangements (Bebchuk and Fried, 2004), it is plausible that managers make strategic choices regarding the structure and level of their compensation when faced with strong labor union presence. However, this issue has been relatively unexplored in the extant literature. 1

3 A growing line of research explores the prevalence and impact of debt-like compensation in executive pay. 1 Recent research shows that executive compensation typically includes a substantial amount of pension and deferred compensation along with cash and equity incentives (e.g., Bebchuk and Jackson, 2005; Wei and Yermack, 2011). For example, Bebchuk and Jackson (2005) report that in 2003 the average actuarial value of pension holdings was $19.6 million for the CEOs of S&P 500 firms with ages between 63 and 67, indicating that debt-like compensation is a prominent component of CEO compensation packages. In this paper, we explore the role played by union strength in the determination, and consequences, of debt-like compensation. In particular, we explore the following research questions: Does union strength induce managers to strategically substitute cash and/or incentive compensation for debt-like (i.e. deferred and pension) compensation? Does labor presence impact the amount and the optimal form of compensation policy among the component parts? Prior research establishes that managers of firms in more unionized industries make strategic financial decisions in order to gain bargaining advantages over unions; for example, Klasa, Maxwell and Ortiz-Molina (2009) show that firms strategically choose lower cash holdings as a means of sheltering corporate income and improving their bargaining position over labor. In a similar vein, there are a number of reasons to believe that debt-like compensation strategically paid to managers alters the balance of power with labor unions. Prior work (e.g. Cassell, Huang, Sanchez and Stuart, 2012) suggests debt-like compensation promotes greater alignment with creditor interests that coincide with the interests of labor. As with wages paid to unionized workers, debt-like compensation has a fixed payoff that is contingent on the firm s long-term survival; therefore, labor unions could prefer the use of the debt-like compensation in CEO pay relative to other forms of compensation that reward risk taking. In addition, Jensen and Murphy (1990) argue that there is an implicit connection between 1 The extant literature generally refers to debt-like compensation as inside debt. We use debt-like compensation and inside debt interchangeably. 2

4 executive- and worker compensation in unionized firms due to the public disclosure of executive compensation which forms the basis for union demands. Consistent with this argument, the American Federation of Labor and Congress of Industrial Organizations (AFL-CIO) closely monitors executive compensation (Farber, Jung, Lee and Yi, 2012), and labor unions are prolific sponsors of shareholder proposals addressing CEO compensation (Ertimur, Ferri and Stubben, 2010; Del Guercio and Woidtke, 2012). Anecdotally, the AFL-CIO emphasizes pension benefits as one of the important benefits of union membership. Extending the argument of Jensen and Murphy (1990), it is plausible that labor unions may be more tolerant of executive debt-like compensation. Third, while pension and deferred compensation have long been components of CEO compensation, their disclosure was not required until the SEC regulatory changes in 2006 and therefore they are relatively new to the debate over the perceived excessive executive compensation. Combined with a complex actuarial-based method of valuation (Bebchuk and Fried, 2004), it is possible that unions do not fully understand the valuation of debt-like compensation. 2 Fourth, despite the disclosure requirements of debt-like compensation, the extent of information disclosure varies among firms, particularly for deferred compensation. For example, while deferred compensation must be disclosed in the Non-Qualified Deferred Compensation Table, firms are not required to report all components of deferred compensation in the Summary Compensation Tables, thereby providing an opportunity to obfuscate the total compensation (Fulmer, Ang and Cheng, 2014). Fifth, managers may strategically substitute current compensation for deferred compensation in the face of strong union presence. In addition to low valuation transparency, deferred compensation plans often feature withdrawal options prior to at retirement and executives are able to invest in the firm s equity (e.g. Anantharaman, Fang and Gong, 2 Francis and Ackerman (2015) discuss how the value of pension benefits for top executives in the U.S. increased significantly in 2014 due to arcane pension accounting. See 3

5 2014). These characteristics of deferred compensation, along with the influence of top executives over their own pay arrangements (Bebchuk and Fried, 2004), could be exploited by managers and board compensation committees to strategically shift executive compensation away from current compensation. Our study contributions to the literatures on union presence and managerial compensation policy in several important ways. Based on a broad sample of firms spanning fiscal years , we first show that union presence is positively associated with the ratio of CEO inside debt compensation to equity compensation and to the CEO-firm relative debt ratio. These links are robust to alternative model specifications, controls for potential endogeneity, and the alternative use of firmlevel unionization information. We next explore the significant association between unionization intensity and inside debt by decomposing debt-like compensation into pension and deferred components. Our evidence suggests that deferred compensation drives the union effect. Our evidence illustrates that labor strength has a positive and significant impact on the level of deferred compensation balance level and to the annual contribution to the deferred account scaled by annual total cash compensation. Additionally, labor strength is positively associated with the CEO s deferred balance scaled by his or her equity portfolio value and, finally, with the CEO-firm relative debt ratio using the CEO s deferred balance. These results collectively provide new evidence suggesting that managers strategically use deferred compensation in order to improve their bargaining position. In support of this premise, we find that firms headquartered in states where organized labor is strong, and led by CEOs who are closer to retirement, drive these results. In contrast, labor strength is generally unrelated to the level and proportion of pension benefits. Our analysis of optimal debt-like compensation policy shows that labor union presence and its effect on deferred compensation choices slows the speed of adjustment to the optimal CEO-firm relative debt ratio. Finally, we provide direct 4

6 evidence of the strategic substitution by managers by examining the interplay between the components of cash, incentive, and debt-like compensation. Our results consistently suggest that managers strategically substitute cash bonus compensation for deferred compensation, after controlling for the additional forms compensation. Finally, we employ a difference-in-differences approach to test the robustness of our cross-sectional results pertaining to the strategic substitution hypothesis using the passage of right-to-work (RTW) legislation and union contract renegotiations during our sample period as the context for a natural experiment. The difference-in-differences results provide corroborating evidence for the robustness of our cross-sectional findings. We structure the remainder of the paper as follows. Section 2 reviews the literature on inside debt and labor union strength. Section 3 describes our hypotheses. Section 4 reports the empirical results of our hypotheses. Finally, Section 5 summarizes our findings and offers concluding remarks. 2. Literature Review 2.1. Inside Debt and Corporate Policy Jensen and Meckling (1976) hypothesize that debt-like compensation may alleviate riskshifting conflicts between stockholders and bondholders that result from equity incentive compensation. Because inside and outside debt feature similar payoffs, debt-like compensation can be used to dissuade managers from taking excessive risk at the expense of bondholders. Consistent with this premise, recent research investigates the prevalence and effects of inside debt as a component of managerial compensation. Bebchuk and Jackson (2005) and Sundaram and Yermack (2007) demonstrate that pension compensation comprises a significant portion of CEO compensation. The Securities and Exchange Commission s 2007 reform that promotes greater transparency in reporting of pension and deferred compensation further allows researchers to examine in detail the components 5

7 of inside debt. For example, Wei and Yermack (2011) show more than two-thirds of the CEOs in the Execucomp database had non-zero inside debt with a mean value of $5.7 million in Edmans and Liu (2011) provide a theoretical motivation for the use of inside debt by extending the intuition of Jensen and Meckling (1976) to focus on manager s debt-equity ratio relative to that of the firms. In a firm facing agency costs of debt and equity, Edmans and Liu (2011) postulate that firms use managerial inside debt to mitigate shareholder-bondholder agency conflicts. For example, if the agency cost of debt is more significant than the agency cost of equity, then greater proportions of managerial inside debt-to-equity, relative to firm debt-to-equity, can reduce the agency cost of debt relative to equity by increasing the manager s relative exposure to debt-like payoffs. Edman and Liu s (2011) model predicts, among others, that managers with higher proportions of inside-to-firm debt ratios are more likely to choose conservative operating policies. Consistent with the theoretical predictions of Edmans and Liu (2011), Sundaram and Yermack (2007) find a negative relation between the level of managerial pension holdings and the probability of default, suggesting that managers with high inside debt behave more conservatively. Similarly, Wei and Yermack (2011) find that bond prices increased and stock prices fell when the heightened disclosure requirements for pension and deferred compensation in executive compensation packages came into effect in 2007, with pronounced price reactions for firms with higher amounts of CEO inside debt. Cassell, Huang, Sanchez and Stuart (2012) find that CEO risk-seeking behavior is influenced by the level of inside debt she holds: When CEO inside debt is high, future stock return volatility, R&D expenditure and financial leverage are lower, while the extent of diversification and asset liquidity are higher. Phan (2014) extends these findings by showing that firms with high CEO-firm relative debtequity ratios experience significantly greater likelihoods of diversifying acquisitions. In a similar vein, Liu, Mauer and Zhang (2014) find that cash holdings are significantly higher for firms in which CEOs 6

8 have a higher level of inside debt. Chava, Kumar and Warga (2010) take an alternative approach by examining the link between the structure of bond covenants and inside debt. They find a negative relation between the number of bond covenants and the level of CEO inside debt compensation. Finally, Anantharaman et al. (2014) present empirical evidence that bank loans are priced higher and have fewer covenants when the CEO-firm relative debt-to-equity ratio is higher Labor Strength and Corporate Policy As collective bargaining agents, unions seek to maximize the utility of their members. Their success at achieving this objective has been the subject of a well-developed body of research. For example, Lewis (1986) reports that there are substantial markups in the salaries of unionized workers while Abraham and Medoff (1984) show that senior employees at unionized firms have greater protection from layoffs than those of non-unionized firms resulting from collective bargaining agreements. Similarly, Card (1996) reports that earnings of workers in unionized firms are significantly higher than those in non-unionized firms. Unions are also motivated to curb what they perceive to be excessive executive compensation. Banning and Chiles (2007) show a negative relation between the unionization rate and total CEO compensation, while Gomez and Tzioumis (2011) further find that the negative correlation is primarily driven by option compensation. Gomez and Tzioumis (2011) propose direct and indirect channels linking union presence to executive compensation. Unions pressure firms by direct threats and through grassroots efforts to create public awareness about the perceived excessive salaries. Unions and affiliated funds are also prolific shareholder activists who use shareholder proposals to target the firms executive compensation policies, among others. 3 Alternatively, Gomez and Tzioumis (2011) propose an indirect channel where stock market participants penalize strong union presence which in turn reduces the value of equity incentive 3 Copland and O Keefe (2013) note that labor-affiliated institutional investors sponsored about one-third of all shareholder proposals submitted during

9 compensation (e.g., Banning and Chiles, 2007; Huang et al., 2015). In support of these arguments, Gomez and Tzioumis (2011) find that unionization intensity is negatively associated with executive compensation through the restraining effect union presence has on stock option compensation. Organized labor and bondholders share common objectives. In a contingent claims context, unions seek to maximize the utility of workers, resulting in a payoff function that resembles those of bondholders. Similar to a bondholder s payoff, worker wages resemble a long position in the firm s assets and a short call option position written on the same assets (Merton, 1973). In solvent states, the wages paid to workers is fixed; in insolvent states, promised payments are unfulfilled by the firm thereby creating a payoff with limited upside potential and potentially unlimited loss. Since shareholders capture the positive payoffs from risky investments while workers and bondholders bear the consequences of negative payoffs, unions share a preference for safe investments with corporate bondholders. While the interests of unions broadly coincide with those of bondholders, they contrast with those of shareholders. As labor costs are frequently one of firms largest expenses, managers have an incentive to improve their bargaining position against unions as a means of maximizing firm value. 4 A line of research highlights this tension between management and unions by showing that managers strategically attempt to counteract unions rent-seeking behavior via corporate decisions. In early work, DeAngelo and DeAngelo (1991) show that steel manufacturers significantly reduce dividends prior to their negotiations with unions. Bronars and Deere (1991) show that firms credibly reduce funds potentially available to unions by issuing debt, thereby sheltering income from union demands. More recently, Klasa, Maxwell and Ortiz-Molina (2009) argue that managers use cash as a mechanism 4 For example, early work documents a negative relation between unionization rates and firm value due to lower productivity and higher production costs (Clark, 1984; Ruback and Zimmerman, 1984; Addison and Hirsch, 1989; Vender and Gallaway, 2002). 8

10 to strengthen their bargaining position over organized labor. Klasa et al. (2009) provide evidence that firms in more unionized industries generally hold less cash, that managers adjust cash holdings downwards prior to negotiations with organized labor, and that higher cash holdings are associated with a greater probability of labor strikes. In a similar vein, Matsa (2010) shows that managers strategically use debt financing when corporate liquidity is high to improve their bargaining position with workers, while Huang et al. (2015) report managers and boards of directors strategically reduce CEO compensation prior to negotiations with labor as unions use the level of CEO compensation to gauge the firm s well-being. Additionally, there is evidence that managers disclose information strategically when faced with a strong union: Hilary (2006) and Bova (2013) find that firms partially disclose information when faced with strong union presence, resulting in greater information asymmetry with the market. Further, a related line of research shows that unions are able to influence corporate policies that result in greater alignment with their preference for lower risk. For example, Connolly, Hirsch and Hirschey (1986) find the elasticity of R&D-to-sales with respect to the unionization rate is negative. Using survey data, Hirsch and Link (1987) report that innovative activities are less prevalent in highly unionized firms. Chyz, Lueng, Li and Rui (2013) argue that unions prefer less tax risk because they ultimately bear the full cost of aggressive tax strategies that may result in additional taxes and Internal Revenue Service penalties. Their evidence demonstrates a significant negative association between union bargaining strength and corporate tax aggressiveness, supporting the view that unions prefer lower risk and that managers respond to this preference by adopting more conservative tax strategies. 9

11 3. Hypotheses Development Prior research demonstrates that unions incentive to maximize the utility of their members leads them to prefer conservative corporate policies, thereby aligning their interests with those of bondholders. In contrast, managers seek to maximize firm value. These competing objectives create tension during the collective bargaining process to which managers respond strategically. These responses are designed to produce greater uncertainty about the firm s long-term survivability, including limited disclosure of earnings, higher leverage, less cash holdings, and lower dividends. Because firm failure is costly to workers, these tactics increase the likelihood that managers gain concessions (e.g. lower wages) from unions. In addition, the extant literature shows that managers strategically choose less risky policies to better align with organized labor s preference for conservatism, which reduces the compensation demanded by unions in exchange for managerial risktaking. The level of executive compensation, and the gap between executive- and worker compensation, is an important consideration for unions in the bargaining process. 5 While the existing literature is limited, the consensus is that there is a significant negative association between unionization and executive compensation levels. More importantly, there is empirical evidence suggesting that firms strategically use CEO compensation in order to curb union demands, however the extant literature (e.g., Huang, et al. 2015) has yet only focused on components of current compensation (salary, bonus, and equity compensation). We postulate that unions have a direct impact on inside debt, based on the premise that labor unions are fixed income claimants who are concerned 5 Direct evidence for union interest in the gap between executive and worker compensation can be found in shareholder proposals sponsored by unions and union-affiliated funds. For example, General Electric faced such a shareholder proposal in its 2010 proxy to request a review of GE s top executive compensation policies. The proposal sought to promote such issues as a comparison of the total compensation package of GE s top executives and its lowest paid employees in the U.S. between 2000 and 2009, rationales for such gaps, the trend in this gap over time between the two groups, and a greater equity between the two groups as the goal. 10

12 about downside risk. The rapidly growing inside debt literature provides empirical support for the theoretical predictions of Jensen and Meckling (1974) and Edmans and Liu (2011) that inside debt produces better alignment between managerial and bondholder interests. Because unions have payoffs similar to those of bondholders, and because inside debt incentivizes choices that are associated with less risk, we conjecture that unions prefer debt-like compensation vs. equity compensation. Thus, we expect union presence is associated with a greater proportion of inside debt in executive compensation contracts. This leads to our first hypothesis: H1: The association between labor strength and inside debt compensation is positive. While the fixed income claimant rationale may be one reason for empirical evidence supporting Hypothesis H1, there is an alternate explanation. Consistent with the above-described literature demonstrating that managers disclose information selectively, manage earnings downwards, and otherwise attempt to shield income from union demands, managers may employ the components of their compensation strategically as an alternative bargaining tool in the face of strong unions. Consistent with the findings of Huang et al. (2015) showing that board of directors and managers strategically reduce executive compensation to improve their bargaining position with unions, we posit that managers may opt to strategically substitute their current cash and / or equity incentive income to deferred income, which mitigates the perceived executive-to-worker pay gap. Because deferred income is a component of debt-like compensation, the strategic substitution view may be a contributing driver for the positive association between the unionization rate and debt-like compensation. By examining the pension and deferred compensation components of debt-like compensation separately, we are able to test whether the fixed income claimant or strategic substitution argument best explains the positive effect of labor union on inside debt. While most research on inside debt does 11

13 not distinguish between pension and deferred compensation, there are important differences. Executives contractually receive retirement pension contracts (SERPs) which have defined postretirement benefits that resemble bond payoffs (Bebchuk and Fried, 2004). Unlike SERPs, executives make annual deferred compensation decisions. Deferred income can be invested in the firm s own stock, and deferred compensation plans typically offer some flexibility in repayment prior to retirement (Anantharaman et al., 2014). These characteristics result in a divergence in payoffs between deferred compensation and risky debt. If the fixed income claimant view is the underlying reason for a positive association between union intensity and debt-like compensation, then we expect to find a significant association between labor strength and the value of the CEO s pension relative to her current income and portfolio of equity incentives. Alternatively, if the strategic substitution view is the primary driver for a positive association between union presence and inside debt, then we expect that deferred compensation drives the association. We express these two alternative hypotheses as follows: H2a: Under the fixed income claimant view, managers use pension compensation to align with unions resulting in a positive association between pension compensation and union intensity. H2b: Under the strategic substitution view, managers use deferred compensation to reduce current compensation resulting in a positive association between deferred compensation and union intensity. To enhance our understanding of the relation between CEO compensation and union presence, we explore how union presence affects optimal contracting. In particular, the predictions of Hypothesis 2b suggest additional insights into the use of debt-like compensation in the presence of strong organized labor. First, we surmise that changes to deferred compensation, apart from those predicted by the economic determinants of debt-like deferred compensation, result in above-optimal relative debt ratios and friction in the adjustment towards the optimum. Campbell, Galpin and John (2015) 12

14 calculate predicted CEO-firm relative debt ratios based on optimal contracting variables specified by the theoretical framework of Edmans and Liu (2011) and the empirical findings of Sundaram and Yermack (2007). They regress changes in the relative debt ratio on the lagged residual relative debt ratio and additional control variables. Negative significance of the lagged residual supports the intuition that a relative debt ratio that is below (above) the theoretical optimum is associated with an increase (decrease) in the relative debt ratio towards the optimal target ratio. Based on our prior hypotheses, we expect union presence to play a role on the speed of adjustment towards the optimum. To the extent debt-like (i.e. deferred) compensation is used strategically as predicted by Hypothesis 2b, CEO choices to substitute current compensation into their deferred accounts result in non-optimal relative debt ratios that are not explained by the optimal contracting variables and represent a constraint on the movement towards the optimum. This leads to Hypothesis 3: H3: Strategic use of deferred compensation in the face of strong unions slows the speed of adjustment towards the optimal contracting-based CEO-firm relative debt ratio Finally, we examine the interplay, or trade-off, between the various components of executive compensation. Optimal contracting theory specifies underlying CEO- and firm-level characteristics to explain debt-, cash (salary and bonus), and equity incentive compensation. For example, Gerakos (2010) shows that underlying economic determinants explain most of the variation in CEO pension balances. However, Gerakos (2010) also reports limited evidence that powerful managers use pensions to extract rents, leading to over-optimal pension benefits. Similarly, Core, Holthausen and Larcker (1999) hypothesize that board and ownership variables enable managers to extract rents. They use the economic determinants of total compensation (salary, bonus, and equity incentives) to illustrate that board and ownership structure measure agency problems that are associated with out-of-equilibrium 13

15 compensation levels. Consistent with this approach, recent work by Huang et al. (2015) demonstrates that union presence results in suboptimal equity incentives that become more negative when the firm enters into contract negotiations and the CEO chooses less equity compensation in order to improve their bargaining position. We build on these findings, and add to extant work on the role nonfinancial stakeholders play in the determination of policy choices, by providing a comprehensive picture of how union presence affects the interplay among the various components of compensation. Boards make CEO compensation choices simultaneously and the presence of non-economic determinants may result in complementary or substitute effects between the various components. For example, Gerakos (2010) finds support for the notion that managerial power results in rent extraction by showing that positive excess total compensation significantly complements annual pension benefit accruals. Following the approach of prior research that measures (actual predicted) excess compensation based on underlying economic determinants, we examine how labor strength affects the interplay among the components of the CEO s optimal compensation structure. Similar to Gerakos (2010), we examine if the excess components of current compensation (base salary, bonus, and equity compensation) explain excess deferred compensation: a positive coefficient would suggest a complementary effect resulting from managerial power allowing rent extraction, while a negative coefficient would suggest a substitution effect. Following our earlier hypotheses, greater labor strength should alter these relationships, leading to the substitution of one or more types of current compensation into the deferred account: H4: Strong union presence leads to a negative relation between one or more excess current compensation components and excess deferred compensation. 14

16 4. Labor Strength and Debt-like Compensation 4.1. Data and Sample Selection Our sample period spans FYE 2006 to 2014, where the 2006 starting point is due to changes in SEC rules on August 29, 2006 that require the disclosure of the present value of benefits related to pension and other deferred compensation plans. Because the effects of union strength may be different in regulated firms, we exclude utilities and financial firms. Table 1 Panel A provides descriptive statistics for CEO debt-like compensation and cash compensation. The average annual executive contribution to his/her deferred compensation account is $196,755 while average total cash compensation is $1,059,860. Debt-like compensation is the sum of the aggregate non-tax-qualified deferred compensation balance and the aggregate actuarial present value of the accumulated pension benefit reported by Execucomp: the deferred compensation balance for the typical CEO firm-year has a mean (median) of $2,599,560 ($171,893) and the mean (median) present value of pension benefits is $3,061,160 ($0). Following the methodology described by Sundaram and Yermack (2007) and Daniel, Li and Naveen (2013), CEO equity compensation (or Inside equity) is the sum of stock grants, restricted stock, and the present value of stock option holdings. The Inside debt ratio is the CEO s personal debt-equity ratio and is based on total debt-like compensation divided by equity compensation. Mean (median) Inside debt ratio is (0.048). The Relative debt ratio is CEO Inside debt ratio divided by the firm s debt-equity ratio (e.g., Edmans and Liu, 2011; Anantharaman et al., 2014) and has a mean (median) of (0.194). 6 In Panel B we present firm-level characteristics. We base the strength of union bargaining power (Labor strength) on industry- and firm-level characteristics. The number of unionized workers 6 A small number of firms have very small values of debt-equity ratios that result in large values of relative debt. For example, in 2011 ITT Corporation had a firm debt-equity ratio of 0.003, a CEO inside debt ratio of 6.993, and a resulting relative debt ratio of 2,106. To reduce the likelihood of bias to our results caused by extreme outlying values, we exclude the top 1 percent of the relative debt distribution from our analysis. 15

17 covered by bargaining agreements at the firm level is voluntarily reported and often unavailable; therefore, we obtain unionization rates from the Union Membership and Coverage Database ( maintained by Barry Hirsch and David Macpherson. Unionization rates are reported at the Census Industry Classification (CIC) industry level. We convert the CIC codes to fourdigit SIC codes that serve as the unionization rate at the firm level. A variety of studies (e.g. Klasa, Maxwell and Ortiz-Molina, 2009; Chen, Kacperczyk and Ortiz-Molina (2011a, 2011b); Chen, Chen and Liao, 2011) provide evidence that industry CIC-level unionization rates effectively proxy for firmlevel rates. Following Hilary (2006) and Chen et al. (2011), Labor strength gauges the intensity of union presence relative to firm size. We calculate Labor strength as the number of unionized employees (unionization rate number of employees) divided by total assets. As shown in Table 1 Panel A, Labor strength has a mean (median) of (0.018). As a robustness check, we hand- collect firm-level unionization data from SEC filings when available (approximately 65 percent of our sample), following the method discussed by Huang et al. (2015). Huang et al. (2015) report a similar decrease in their sample size when they use firm-level unionization information. Summary statistics for our firm-level unionization data are very similar to those reported by Huang et al. (2015). Labor strength using union rates collected at the firm level has a mean (median) of (0.003). Additional firm-level explanatory variables for debt-like compensation are provided in Table 1 Panels B-C and are based on the specification of Sundaram and Yermack (2007) and related research. Founder firm is a binary variable equal to one if the CEO is also the company founder and Firm age measures the number of years since its listing date using the CRSP Header File. Stock price performance may relate to compensation choices: we measure stock performance and risk with Market-adjusted CAR (Stock return volatility), respectively. To control for the CEO s incentive to accept debt-like compensation as compensation in low-risk firms (Campbell, Galpin and Johnson, 16

18 2015), we include stock return performance and risk metrics. Market-adjusted CAR is the cumulative market-adjusted stock return for the 24 months leading to the FYE date using the CRSP valueweighted index, and Stock return volatility is the standard deviation of monthly returns over the 24- month period. Following Sundaram and Yermack (2007), Liquidity constraint is a binary variable equal to one if the operating cash flow is negative and zero otherwise. We define Sales growth as the five-year geometric growth in sales and we use Market-book ratio as a market-based alternative measure of future growth. Tax loss carry-forwards control for the tax benefits associated with deferral of income to future years, and the Herfindahl-Hirschman Index (HHI) controls for potentially higher rents exerted by unions in more concentrated industries (e.g., Salinger (1984), Karier (1985), and Klasa et al. (2009)). Net PPE controls for capital intensity; less free cash flows can be committed to fixed payments in capital-intensive industries, and capital intensity is associated with unionization rates. Finally, logged total assets controls for firm size. At the CEO level, we include variables measuring CEO age, CEO influence over the board as a chairperson (CEO duality) and the number of years the CEO has served the firm (CEO tenure). We summarize additional details about the sources and construction of the variables in the Appendix Empirical Results Labor Strength and Debt-like Compensation We follow Sundaram and Yermack (2007) by employing the Inside debt ratio (CEO debt-like compensation divided by equity incentive compensation) as the dependent variable. Following recent work (e.g. Anantharaman et al., 2014), we also employ Relative debt (Inside debt divided by the firm s debt-equity ratio) to examine if union strength alters the balance of compensation incentives provided by inside debt. Both measures are truncated at zero; therefore, we estimate the regressions using the Tobit methodology. Since both Inside- and Relative Debt are highly skewed as demonstrated in Table 17

19 1 Panel A, we normalize the distributions by logging (1 + each measure). To insure that skewness of the CEO inside debt measures does not induce bias in our Tobit-based results, we also estimate median regressions. Intersecting the various databases results in a total of 8,690 firm-year observations used in the cross-sectional models. We specify the regression model as follows: Inside(or Relative) debt α 5 α α Founder firm α Liquidity constraint α Net PPE α 16 6 α 0 α Labor strength α Log(1 Firm age) α Book leverage α Market adjusted CAR α Stock volatility 11 Sales growth α Log(CEO age) α CEO duality α Firm size α Fama - French 30 industry fixed effects α Year i 1 i Market - book ratio α Tax loss carry - i i 4 Log(1 CEO tenure) 9 forwards α fixed effects 14 HHI (1) As illustrated in Table 2, Labor strength is significantly associated with Inside debt (Relative debt) at the 1 percent levels, using the Tobit methodology (Models 1-2) and median regressions (Models 3-4). These results are consistent with the predictions of Hypothesis H1, which points to a positive relationship between union bargaining strength and the CEO s debt-like compensation. The signs and significance of the control variables are generally consistent with the findings of Sundaram and Yermack (2007) and related research: Older CEOs are associated with greater debt-like compensation. CEOs that are board chairs have a higher proportion of debt-like compensation, while company founders who are CEOs have lower debt-like compensation. Consistent with Sundaram and Yermack s (2007) findings, CEOs of older firms have a greater proportion of debt-like compensation. Among the financial control variables, stock price performance is negatively associated with Inside debt and consistent with the results of Campbell et al. (2016), stock return volatility is also negatively related. Growth opportunities (Market-book ratio) are negatively associated with inside debt. Similar to Sundaram and Yermack s (2007) findings, CEOs of firms with operating loss carry-forwards have less Inside debt. Finally, Firm size is positively associated with debt-like compensation for all Models in Table 2 as reported in related work. We re-estimate the model with quantile (median) regressions in Models 6-8. The results from these regressions are quantitatively and qualitatively similar to the Tobit model results. 18

20 A limitation of many related prior studies is the use of industry level unionization rates. To mitigate this problem and to account for variation in union coverages within an industry, we handcollect firm level unionization rates from financial statements and rerun our models in Models 5-8. Because firms are not required to disclose their union information, our sample size decreases by almost 3,000. Huang et al. (2015) report the similar decrease in their sample size when they use the firm-level unionization information. Similar to the Tobit regression results of Models 1-2, Labor strength in Models 5-6 is significant and positive, suggesting that industry-level unionization rates do not drive our results. These results are robust to the use of median regression methodology (Models 7-8), where Labor strength continues to be significantly related to CEO inside- and relative debt ratios at the 1 percent level. The remaining control variables are broadly similar to the industry-level union rate models Omitted Variables and Endogeneity While Table 2 Models 1-8 illustrate that the impact of Labor strength on inside- and relative debt is positive and significant, it is possible that an omitted variable not captured by the firm-level regressors and the industry fixed effects is driving the results. In particular, geographic regions not only have a significant bearing on the pattern of unionization (Hirsch, 1980), but may also be correlated with a firm s major financial characteristics including executive compensation. To address this possibility, we re-estimate Equation (1) including location variables to test if the Labor strength variable continues to have significant incremental explanatory power. We define the region in which the firm has the majority of its operations using six interstate megalopolises, or mega-regions. These regions are the Piedmont Atlantic Mega Region, Northeast Megalopolis, Cascadia, Great Lakes, California, and the Arizona Sun Corridor. We create a binary variable equal to one if Compustat data 19

21 item STATE is in a given region. 7 Since Labor strength is correlated with the region variables, the impact of region on debt-like compensation measures is confounded because its effects are already captured in Labor strength. Hence, we create a Labor strength residual variable representing the component of Labor strength that is unrelated to the region variables by regressing Labor strength on the six region variables and obtain the difference between actual and predicted Labor strength. In Table 3 Models 1-2, we report the Tobit coefficients using Inside debt (Relative debt) as the dependent variable. At least two of the five location variables are significant in each model, indicating that region impacts firm-level compensation policies as predicted. Consistent with the results of Table 2, Labor strength residual is positive and significant at the 1 percent levels (p=0.00) using Inside debt and Relative debt as the dependent variables, respectively. These results demonstrate that unionization intensity has a significant impact on deferred compensation independent of the effects related to firm location. 8 We next consider whether Labor strength and debt-like compensation are jointly determined. We re-estimate Equation (1) using an instrumental variable (IV) Tobit approach, specifying Labor strength as endogenous. The key requirements for instrument validity are relevance and exogeneity: we identify two instruments for Labor strength based on labor force characteristics and economies of scale related to the costs of organizing that are unlikely to be directly associated with compensation choices. First, Hirsch (1980) argues that labor mobility is likely to be greater, and demand for unionization lower, in areas with more rapid growth in the labor force. Using the Bureau of Labor Statistics Geographic Profile of Employment and Unemployment, we calculate one-year growth in 7 See for a description of states within each megalopolis. We also use the nine subdivisions of the four primary U.S. Census regions (Northeast, South, Midwest, and West) as alternative region measures; the results are qualitatively the same as those reported here and are available upon request. 8 Our firm-level Labor strength measure produces qualitatively similar results. For brevity, we do not report these results but are available upon request. 20

22 the labor force for each state. Second, Stephens and Wallerstein (1991) argue that there are scale economies to organizing: labor is more costly to organize when industrial concentration is less, leading to lower union density. We collect the number of publicly listed firms in each state using the LexisNexis database as a proxy for concentration. Using these two variables as instruments for Labor strength, we estimate the Inside- and Relative debt models using maximum likelihood estimation. Table 3 Models 3-4 report the IV Tobit coefficient estimates. In the final two rows of each model, we report the instruments first-stage coefficient estimates. As expected, both instruments are negatively related to Labor strength and statistically significant in both models. In the second stage, the Labor strength instrumental variables are significant at the 5 percent levels using Inside debt and Relative debt as the dependent variables, respectively. The Kleibergen-Paap LM Statistic tests the null hypothesis that the model is underidentified: In both models, the test statistics are significant at the 1 percent levels indicating that the models are not underidentified. Finally, the Hansen J Statistic tests the null hypothesis that the overidentification restrictions are valid. The J Statistic s p-values of 0.72 in Model 3 (0.18 in Model 4) suggests the instruments are appropriate and well identified in each model Fixed Income Claimant- and Strategic Substitution Views The previous sub-section establishes significant positive correlation between union bargaining strength and CEO debt-like incentives. To determine whether the fixed income claimant- or strategic substitution view is driving the results, we analyze the deferred - and pension components of the CEO s debt-like compensation. Because prior work shows that unionization is negatively related to equity incentives, and because equity incentives are part of the Inside debt and Relative debt measures, we use multiple metrics to insure that our results are not merely related to the effect of Labor strength on equity incentives. According to the strategic substitution view, executives substitute current- for 21

23 deferred compensation in the presence of strong labor, resulting in a higher deferred balance over time and Deferred contribution-to-tcc ratio in a given year. Alternatively, the fixed claimant view implies that corporate policies that align CEO interests with those of organized labor may result in a higher pension balance and pension contributions relative to cash compensation. The strategic substitution (pension claimant) view implies that CEOs hold a greater proportion of deferred compensation (pension benefits) relative to their portfolio of equity incentives. Inside deferred debt and Inside pension debt are based on the Inside debt measure decomposed into its deferred and pension components, respectively. Finally, we investigate if strategically substituting current- for deferred compensation, or alternatively responding to union preference for greater pension compensation, induces a tilt in the CEO-to-firm debt-equity ratio as measured by the Relative deferred- or Relative pension debt ratios, respectively. Inside deferred debt and Inside pension debt are based on the Relative debt ratio decomposed into its deferred and pension components, respectively. We sort the sample by labor strength quartiles within each year and calculating the mean and median value of each measure for each quartile. We calculate the difference in means and medians between the first and fourth labor strength quartiles and present the p-value for this difference in Column 5. Mean deferred- and pension balances increase as Labor strength increases. Mean (median) Deferred contribution-to-tcc increases significantly, however mean Pension contribution-to-tcc decreases from the lowest to the highest quartiles. The ratios of deferred- and pension compensation to equity compensation increase significantly over the four labor strength quartiles, as do the Relative deferred- and Relative pension debt ratios. While the results of Table 4 generally provide evidence that the deferred and pension-based components of debt-like compensation are increasing in Labor strength, they do not allow us to differentiate the fixed income claimant- and strategic substitution views. Because additional CEO- and 22

24 firm-level characteristics also affect these measures, we proceed to a multivariate framework where we estimate Tobit regressions of the logged CEO debt-like compensation measures on Labor strength and the additional explanatory variables described by Equation (1). These results are reported in Table 5. Following the format of Table 4, we examine multiple metrics. In Model 1 (Model 2), we use the logged deferred compensation (pension) balance as the dependent variable. In support of the strategic substitution view s intuition that CEOs are more likely to substitute current- for deferred compensation when union presence is stronger, greater Labor strength is significantly associated with higher deferred balance levels at the 1 percent level in Model 1; in contrast, Labor strength is statistically unrelated to the logged pension balance in Model 2. In Models 3-4, we use the logged ratios of deferred contribution-to-tcc and pension contribution-to-tcc. Model 3 demonstrates that Labor strength significantly predicts the proportion of annual income allocated to deferred compensation divided by cash compensation and, interestingly, is negatively related to the pension contribution in Model 4. In Models 5-6 we repeat these regressions using the logged (1+Inside deferred debt) and (1+Inside pension debt) measures. Similar to the results of Models 1-4, there are contrasting effects. Labor strength is significantly positively related to Inside deferred debt (p=0.00) and is marginally statistically related to Inside pension debt (p=0.10). Similarly, Model 7 shows the Labor strength coefficient estimate using is positive and strongly statistically significant at the 1 percent level using Relative deferred debt as the dependent variable. In contrast, the coefficient estimate using Relative pension debt (Model 8) is approximately one-third the magnitude and is statistically insignificant. Viewed collectively, Table 5 provides preliminary support for the strategic substitution view: Deferred compensation balances and CEO contributions to deferred compensation, relative to their total cash compensation, are greater when unionization intensity is higher. This association results in a greater 23