Content area
Purpose
The purpose of this paper is to investigate the relationship between CEOs' inside debt holdings (pension benefits and deferred compensation) and the operating leverage of the firms they manage, with the aim to examine whether CEO incentives play a role in corporate risk-taking.
Design/methodology/approach
The authors investigate the relation between CEO inside debt holdings (CIDH) (pension benefits and deferred compensation) and the operating leverage (DOL) of the firms they manage. Using a sample of 11,145 US firm-year observations over the period 2006–2017, the authors find a strong negative association between CIDH and DOL. Additional analyses reveal that the relationship between CIDH and DOL is more pronounced in firms with heightened agency issues, powerful CEOs and for CEOs with stronger professional networks. The results are robust to various sensitivity and endogeneity tests.
Findings
The authors find strong evidence confirming the expected negative association between CEO inside debt and DOL suggesting that firms with higher inside debt tend to maintain lower levels of operating leverage. These findings continue to hold with the alternative measure for the inside debt and operating leverage, and across a range of tests designed to rule out the possibility that the primary findings are in any way driven by potential endogeneity. In addition, the findings demonstrate that the presence of manager-shareholder agency conflicts can strengthen the inside debt–DOL relationship suggesting the strong role of inside debt in reducing firm risk.
Research limitations/implications
Findings in this paper have implications for design of compensation structures so that corporate boards can establish incentives as a tool for risk management. A limitation of this study is that it is focused on one market, i.e. US listed companies, so the findings may not be applicable on a global scale.
Originality/value
To the best of the authors’ knowledge, this is the first study that links firm-level management of operating leverage through design of CEO inside debt incentives (two obvious choices for risk-reduction at the CEOs’ disposal include reducing financial risk through reduction of firm leverage and reducing operating risk through reduction of operating leverage). While use of firm leverage as an instrument of choice has been explored in the past, use of operating leverage to achieve risk reduction when CEO possess high inside holding, has received very little attention.
1. Introduction
In this paper, we investigate whether changes to managerial incentives resulting from pension benefits and/or deferred compensation (frequently referred to as CEO inside debt) lead to changes in a firm’s cost structures through changes in operating leverage (OL) [1]. A causal link between risk-taking incentives and OL was first established by Aboody et al. (2018) who showed that CEOs substitute fixed costs with variable costs in response to lowered risk-taking incentives engendered by a regulatory change (passage of FAS 123 R) that required firms to expense options grants to executives and hence incentivized corporate boards to reduce options in CEO pay packages. The authors, however, caution against generalizing their results across other policy changes that may trigger changes in risk-incentives. For example, they aver that the nature and direction of the impact on a firm’s OL from a firm’s decision to increase option grants may not easily be inferred from their findings around decreases in option intensity. It, therefore, appears that the role of managerial incentives in altering their firms’ cost structures – in particular, OL – needs further investigation. In an effort to improve our understanding of how incentives impact OL, we undertake an examination of changes to risk-incentives resulting from an increase in CEO inside-debt holdings. We strongly believe that such an examination will not only add to our understanding of factors that impact cost structures, but it is also timely given the surge in debt-like compensation structures in recent times.
While equity-driven incentives have been extensively studied, the literature is only starting to take cognizance of incentives driven by CEO pay packages that are increasingly reflecting greater use of debt-like compensation components. Sundaram and Yermack (2007) report that CEO pension plans constitute around 10% of total compensation in their sample of large US companies. Anecdotally at least, there is evidence to suggest that the magnitude of such pay components can be significant. For example, pension, deferred compensation and other retirement benefits of Michael Terry Duke, ex-CEO of Walmart, amounted to $113m in 2013. Wei and Yermack (2011) report that 84% of CEOs in their sample have inside debt close to $10m on average. There is, in fact, growing evidence that CEO debt-like compensation very often exceeds equity-linked compensation in many companies (Bebchuk and Jackson, 2005; Sundaram and Yermack, 2007; Gerakos, 2010 and Wei and Yermack, 2011). It is, therefore, not surprising that recent compensation literature is increasingly focussing on the incentive effects of debt-like pay components that have payoffs closely resembling those of fixed income obligations.
The global financial crisis of 2009 brought to the fore the risks to personal wealth of CEOs as many whose firms did not survive the crisis incurred significant personal losses through loss of pension benefits and/or deferred compensation. During times of economic distress, inside debt holdings are particularly vulnerable as they generally represent unsecured and unfunded firm liabilities rendering them sensitive to default risk (Sundaram and Yermack, 2007; Edmans and Liu, 2011). Therefore, inside debt is likely to reduce CEO appetite for risk given that an average CEO’s personal portfolio is largely undiversified (Sundaram and Yermack, 2007; Cassell et al., 2012). Indeed, Wei and Yermack (2011) state:
To the extent that managers have large unfunded deferred compensation claims against their firms, outside investors expect them to manage their firms conservatively, implying lower-risk investment strategies that would tend to make inside debt safer (p. 3817).
While their toolkit offers several alternative mechanisms to achieve their risk-reduction objective, two obvious choices at the CEOs’ disposal include reducing financial risk (through reduction of leverage) and reducing operating risk (through reduction of OL by reducing the firm’s fixed cost structures). Although these two risks may not necessarily operate independent of each other, they both afford CEOs (particularly those with high inside debt) with opportunities to limit their exposure to firm failure. While use of firm leverage as an instrument of choice was the subject of Chen et al. (2019) (who reported an inverse relationship between firm leverage and CEO inside debt), use of OL to achieve risk reduction for CEOs with high inside holdings has received very little attention. Therefore, risk-reduction through changes in OL is the subject of our study.
Incentive conflicts between managers and outside debt providers result when managers engage in risk-increasing policy choices designed to benefit shareholders but may potentially be detrimental to the interests of debtholders (frequently referred to as agency costs of risk-shifting; see, for e.g. Jensen and Meckling, 1976). This is because, while payoffs to debtholders are capped during good times, they stand to potentially lose their entire investment upon firm failure. Debt holders would therefore prefer managers to pursue conservative investment/financing policies given their asymmetric payoff function with respect to the firm’s net assets. Inside debt holdings, on the other hand, can encourage CEOs to make investment/financing decisions that could potentially mitigate the agency costs of debt arising out of risk shifting, as CEOs with inside debt will likely engage in policies geared more towards risk-reduction. Consistent with this, Edmans and Liu (2011) argue that inside debt is more effective than cash compensation in alleviating agency conflicts arising out of debt financing.
Empirical evidence supports this argument since the literature on CEO inside debt is increasingly documenting a shift in policy prerogatives tilted more towards lowering firm risk. For example, Wei and Yermack (2011) report that announcements of large inside debt holdings of CEOs result in a wealth transfer from equity to debtholders and a decrease in the volatility of both debt and equity securities. Sundaram and Yermack (2007) find that a higher percentage of pension compensation leads to more conservative decision making by CEOs. In addition, CEO inside debt has been shown to be positively related to firm cash holdings (Liu et al., 2014) and negatively related to obtaining trade credit. Also, high inside debt has been shown to lower cost of debt as well as enabling firms to raise debt capital with a smaller number of restrictive debt covenants (Anantharaman et al., 2014; Chava et al., 2010), and that the debt of such firms is priced higher in the secondary bond markets (Wei and Yermack, 2011) Inside debt is also associated with lower incidence of tax sheltering (Chi et al., 2017), lower payout (Eisdorfer et al., 2015), higher firm liquidation values (Chen et al., 2010) and better acquisition performance (Phan, 2014; Bhabra et al., 2021). Extant evidence, therefore, suggests that higher CEO inside debt is associated with lower firm risk. It is noteworthy however, that past studies that have documented a negative association between CEO inside debt and riskiness of firm policy choices, have focussed largely on financial risk. One exception is Cassell et al. (2012) which finds that high CEO inside debt leads to lower levels of stock return volatility that is largely achieved through lower R&D investments, higher diversification and through maintenance of higher working capital.
We build on existing evidence regarding inside debt-driven disincentives for risk-taking by examining the impact of CIDH on investment choices that reduce the firm’s OL. In the absence of market imperfections (no bankruptcy or no deadweight costs of distress), lowering OL should not matter. When markets are imperfect, however, reducing the variability of a firm’s cash flows can be valuable to CEOs given their personal portfolio. Hence, managers with inside debt may engage in investment choices with potential to reduce OL since this could reduce volatility of cash flows and hence their risk exposure. The global financial crisis of 2008–2009 did highlight the risks to the personal portfolios of CEOs when the firms they manage suffer financial distress. Such risks would be of less concern in perfect capital markets. Secondary markets for assets, however, are not perfect, suggesting that firm failure has potential for real wealth loss to stakeholders and this raises the possibility that CEOs with large inside debt holdings may engage in policies aimed less towards shareholder wealth maximization and more towards limiting their personal exposure to downside risk. In this paper, we explore the use of changes in OL as one approach to risk-reduction that CEOs may use to shield their personal portfolios.
We use two of the most commonly used proxies for CEO inside debt holding, namely, CEO debt-to-equity scaled by a firm’s debt-to-equity (CEORELDE) and CEO debt-to-equity (CEODE) ratios (Cassell et al., 2012; Chi et al., 2017; Edmans and Liu, 2011; Wei and Yermack, 2011). CEO inside debt-holding is computed as the present values of deferred incentives and accumulated pension, and equity-holding as accumulated stock options and stock holdings (Chi et al., 2017). Consistent with our primary research question, we find strong evidence confirming the expected negative association between CEO inside debt and DOL. The coefficient of the inside debt proxy is significantly negative suggesting that CEOs with higher inside debt holdings tend to maintain lower levels of OL in the firms they manage. These findings withstand several robustness tests. For example, our results continue to hold with an alternative measure for inside debt proxy (CEODE), and an alternative measure for OL (Novy-Marx, 2011). Our results also hold across a range of tests [2] designed to rule out the possibility that our primary findings are in any way driven by potential endogeneity. In addition, our findings demonstrate that the inside debt–DOL relationship is stronger in firms with elevated levels of manager-shareholder agency conflicts.
Our study extends the literature examining the incentive effects of deferred compensation along several dimensions. This is timely since the literature hitherto has focussed primarily on the risk and value implications of CEO equity-based compensation. This literature shows that equity-based pay generally provides increased risk-taking incentives (see, for example, Guay, 1999; Coles et al., 2006) [3]. By focussing on a very different pay component, we find that compensating CEOs with deferred compensation can incentivize them to adopt a diametrically opposite approach to risk management relative to equity-based compensation. In that sense, our study stresses the divergent roles of the various compensation components that can be used to incentivize CEOs. Our study, therefore, sheds light on the design of pay packages that ensure a more equitable approach to stakeholder welfare rather than a singular focus on profits. Analysis surrounding the implementation of IRC 409 A and the associated findings clearly suggest that findings in this study can be used to design policies aimed at ensuring risk-reducing managerial actions. Such risk-reducing policy changes can have wide-ranging implications as firm failure impacts not only the CEO but also a number of other stakeholders such as employees, suppliers and customers. Our study also extends the literature examining the role of OL as a tool for risk management. Dugan et al. (1994), for example, examine the trade-off between operating and financial leverages that managers can engage in their drive to achieve a desired risk-exposure. Since OL can significantly impact a firm’s systematic risk (Lev, 1974; Gahlon, 1981; Garcia-Feijoo and Jorgensen, 2010), findings in our study suggest a possible incentive for CEOs to use changes to OL as a route to achieving the desired risk profile. Finally, consistent with Aboody et al. (2018), findings in this paper provide further evidence that managerial incentives play a central role in shaping/designing their firms’ cost structures.
2. Hypothesis development
Examination into factors that determine a firm’s cost structures, in particular OL, is important as adjustment of costs reflect the collective impact of a range of operational decisions. Hence, a clear understanding of cost drivers is crucial. Past studies have explored, for example, the impact of regulatory changes on cost structures (Holzhacker, et al., 2015), earnings driven incentives and their impact on operating expenses (Kama and Weiss, 2013), impact of product market competition on OL (Babar and Habib, 2022) and OL changes triggered by risk-reducing incentives engendered through reduced option intensity in pay packages (Aboody et al., 2018). While Aboody et al. (2018) come closest in spirit to our central thesis, the authors however, caution against generalizing their finding of a positive association between decrease in option grants and OL across other policy choices/regulatory events that may impact incentives differently.
To extend our understanding of the link between incentives and OL, we explore the potential role/incentive effects of the recent surge in inside-debt of executive pay packages. There is growing evidence that a significant portion of CEOs’ personal portfolios in recent times reflect features that resemble those of debt instruments (Sundaram and Yermack, 2007; Wei and Yermack, 2011). A nascent strand of CEO compensation literature is starting to explore the incentive effects of CEO debt-like compensation and provides preliminary evidence on the implications of CEO inside debt holdings. For example, Wei and Yermack (2011) find that disclosure of CEO inside debt holdings generally results in an increase in existing bond prices, a decrease in equity prices and decreased volatility for both types of securities. Sundaram and Yermack (2007) document a positive association between CEO inside debt holdings and time to default. Others report a negative association between inside debt holdings and the cost of debt (Wang et al., 2010), the use of restrictive covenants in debt contracts (Chava et al., 2010; Chen et al., 2010) and accounting conservatism (Chen et al., 2010; Wang et al., 2010). Still others find that inside debt holdings are associated with higher firm liquidation values (Chen et al., 2010) and lower credit default swap spreads (Bolton et al., 2010). Collectively, these studies suggest that CEOs in firms with inside debt are likely to pursue conservative policies that are more aligned with the interests of outside debt holders rather than with those of shareholders.
In a corporate form of organizational structure, managers engage in risk-reducing policies driven by the asymmetric nature of their claims on firm assets. This is because CEOs capture only a portion of firm profits as part of compensation for their efforts but stand exposed to the full extent of their inside debt holding upon firm failure. Not surprisingly, given that CEOs’ wealth is less than perfectly diversified, they generally tend to have a lower appetite for risk than what shareholders would desire (Jensen and Meckling, 1976). As seen during the recent financial crises, firm failures brought significant wealth losses to those CEOs with large pension and deferred compensation. Such losses would be less concerning in perfect capital markets since deadweight loss from financial distress or firm-failure would be irrelevant. Secondary markets for assets, however, are not perfect, suggesting that firm failure has potential for real wealth loss to stakeholders. Decision-making in the presence of inside debt will therefore hinge on risk choices that ensure firm-survival as the decision-makers (i.e. management and more importantly, the CEO) have significant personal welfare at stake. It is therefore not surprising that firms with more variable cash flows find tax shields less valuable and perceive the expected costs of financial distress higher, and hence, such firms tend to adopt low-debt capital structure policies (Mackie-Mason, 1990).
A question then is, what levers do CEOs have at their disposal to ensure that risk of firm failure is minimized? To gain a better understanding around how incentives around risk-taking influence managers’ decisions to alter cost structures, we explore changes to OL in response to increase in inside debt of CEOs. We choose OL as a risk measure since high (low) OL magnifies (dampens) earnings volatility when revenues rise (decrease). However, managers’ ability to alter OL is not without frictions. Anderson et al. (2003) show that adjustment frictions to volume-based costs are asymmetric across revenue increases and decreases. In other words, there is lower managerial engagement in cost-reduction in response to decreases in revenue while the opposite is true when revenues increase. Such asymmetry in volume-based costs adjustment leaves firms with high OL exposed to large decreases in earnings when revenues decrease (Shust and Weiss, 2014). It, therefore, appears that high OL can prove potentially costly/risky to managers. Others such as Lev (1974), Gahlon (1981) and Novy-Marx (2011) demonstrate a strong association between OL and a firm’s systematic risk, while Kallapur and Eldenburg (2005) find an inverse relationship between demand uncertainty and OL.
Therefore, as one important driver of cash flow variability is a firm’s degree of OL, and as financial distress is costly, reducing OL (and hence the variability of cash flows) would be valuable to CEOs whose personal portfolios are vulnerable to firm failure. Hence, the more variable a firm’s cash flows, the more valuable this reduction in cash flow variability would be to managers. In addition, such gains would vary cross-sectionally with the level of inside debt [4]. We therefore test the following hypothesis:
CEOs with high levels of inside debt holdings will choose to maintain low levels of OL in the firms they manage.
The assumption here is that CEOs not only have a desire to pursue risk-reducing policies but, more importantly, also have the ability to engage in such actions. Such an assumption may not be tenable given past observations on the asymmetric frictions in volume-based cost adjustment in response to revenue shocks (Anderson et al., 2003). However, given that benefits to shareholders accrue from risk-increasing policies (asset substitution where safe assets are replaced with risky assets or make financing decisions that increase firm leverage), boards may incentivize CEOs to increase firm risk. Corporate governance is one tool that shareholders use to ensure managers work to maximize shareholder value and hence strong corporate governance will ensure CEOs pursue shareholder-friendly policies. Conversely, and consistent with findings in Core et al. (1999), when corporate governance is weak, manager-shareholder agency conflicts will increase thereby permitting CEOs the flexibility to engage in policies that improve their personal welfare to the detriment of shareholders. In our context, when CEOs have large inside debt holdings, they will tend to pursue policies that reduce firm risk (and by extension risk to their personal portfolio) even if doing so can harm shareholder interests. We, therefore, test the following contention:
The inverse relationship between CEO inside debt holdings and OL in the firms they manage is more pronounced in poorly governed firms.
3. Sample construct and research methodology
3.1 Sample construct
In formulating our final sample, we obtained data from various sources. For example, we collected information on CEO compensation and CEO attributes from ExecuComp; board related data from Boardex; institutional ownership data from Thomson 13f filings; accounting and financial data from COMPUSTAT; and return data from CRSP. Our sample spans the period 2006–2017 where the starting year of the data is constrained by availability of inside debt holding data. We removed firm-year observations if data on any of our dependent/independent or control variables were missing. This process yielded a final sample of 11,145 firm-year observations from 1,722 unique US listed companies.
3.2 Main regression model
To examine the relationship between CEO inside debt and DOL, we estimate the following ordinary least square (OLS) regression model: where operating leverage, DOL, is the dependent variable (described in Section 3.3); and CEO inside debt proxies, CEORELDE/CEODE, are our primary variables of interest (described in Section 3.4). We also employ an array of controls motivated by extant literature including CEO attributes, board characteristics, monitoring traits, firm characteristics and year and industry dummies (Section 3.5). In our estimation of regression coefficients, we standardize errors after correcting for firm-level clustering. Definitions of all the variables are located in Appendix.
3.3 Dependent variable (operating leverage – DOL)
Following the approach in Chen et al. (2019), we compute operating leverage (DOL) as the ratio of selling, general and administrative expenses scaled by lagged total assets. The authors describe this as a simple measure that is not susceptible to many of the issues that affect the more sophisticated measures used in earlier studies (Mandelker and Rhee, 1984; Kahl et al., 2019). We find that that this measure of OL correlates positively with profitability (ρ = 0.23, p < 0.01 between DOL and ROA which is similar to that reported in Chen et al. (2019).
3.4 Main variables of interest (CEO inside debt – CEORELDE/CEODE)
We use two of the most commonly used proxies for CEO inside debt holding: CEO’s debt-to-equity scaled by a firm’s debt-to-equity (CEORELDE) and CEO debt-to-equity (CEODE) ratio (Cassell et al., 2012; Chi et al., 2017; Edmans and Liu, 2011; Wei and Yermack, 2011). Both these measures include debt and equity components of CEOs’ holdings. Following the methodology in Chi et al. (2017) and others, we define CEO inside debt-holding as the present value of deferred incentives and accumulated pension, and equity-holding as accumulated stock options and stock holdings. All of these data were obtained from the ExecuComp database.
Our first measure (CEORELDE) is calculated as the natural log of one plus CEO-to-firm debt-to-equity ratio. This measure is motivated by the theoretical framework in Jensen and Meckling (1976) and Edmans and Liu (2011). These authors suggest that the closer the debt-to-equity ratios of the personal portfolios of the CEOs are to the firms they manage, the better the probability of eliminating agency issues arising from CEO-bias towards stockholders at the expense of debtholders. In addition, there is evidence to show that CEOs become more risk-averse when their debt-to-equity ratios surpass those of their firms. We would, therefore, expect that an alignment of debt-to-equity ratios will ensure alignment of risk appetites as well. We therefore implement this as our first measure of inside debt. The second proxy we use is calculated as the natural log of one plus a CEO’s own debt-to-equity ratio (CEODE). Although this measure is much simpler in nature, it is still widely used in the literature.
3.5 Control variables
In our multivariate analyses, we include an array of control variables likely to affect a firm’s level of OL (also see, for e.g. Hoi et al., 2019; Chen et al., 2019). CEO attributes such as the age of the CEO (CEO Age) and the number of years the CEO has been in her/his current role (CEO Tenure) control for the fact that older CEOs and CEOs who have been in their jobs longer are more likely to accumulate larger amounts of inside debt and hence moderate their attitude towards OL risk. In addition, to control for potential agency conflicts, we include an indicator variable set equal to 1 if the CEO is also the chairman of the board (CEO Dual) and 0 otherwise, as well as a dummy variable that equals 1 if the CEO is female (CEO Female). Past literature such as Aboody et al. (2018) has demonstrated the association between CEO pay incentives and OL, and hence, we include the natural log of (1+ CEO total pay) as reported in ExecuComp (CEO Total Pay) as well as risk-incentives given to the CEO as defined in Coles et al. (2006) (CEO Vega). To control for agency conflicts emanating from board structure, we next include a suite of board characteristics such as, size of the board (Board Size), the percentage of independent (outside) directors on the board (Board Independence) and the percentage of female directors on the board (Board Diversity). Likewise, governance strength driven by external monitoring is captured/controlled by the percentage of total institutional ownership (Inst. Own. Total) and institutional ownership concentration (Inst. Own. HHI), both sourced from Thomson 13f filings. Firm-level OL will also be driven by factors that are peculiar to the firm. Therefore, we also include firm-specific controls such as the natural log of the market value of equity (Firm Size), the ratio of total debt to total assets (Leverage). These variables serve to control for the fact that as leverage increases, the substitution between operating and financial leverage will drive OL down (Dugan et al., 1994). As growth firms have greater proportion of intangible assets and consequently lower OL, we control for growth opportunities by including the market-to-book value of equity (M/B), and asset tangibility by the property, plant and equipment divided by total assets (Tangibility). A negative association between M/B ratio and OL is also predicted in Garcia-Feijoo and Jorgensen (2010). Firms with higher levels of liquid assets (cash and cash equivalents divided by the book value of total assets – Liquidity) are likely to be more financially constrained and such firms will strive to maintain lower levels of OL. We, therefore, include Liquidity as well as the Kaplan and Zingales (1997) “KZ” index (KZ Score) which proxies for financing constraints. We also include short-term-debt over total debt ratio (Debt Maturity) to capture a firm’s short term liquidity needs. Firms with established dividend policies will ensure that their earnings are less volatile and hence we include dividends paid scaled by total assets (Dividend Payout), and the standard deviation of monthly stock returns for the past twelve months (Return Volatility) (Gahlon, 1981; Novy-Marx, 2011, for the link between OL and stock returns). Finally, as CEO compensation structures can vary over time and across industries (Hoi et al., 2019), we control for year and industry (Fama-French 48) fixed effects. Appendix describes all the variables and their measurements.
4. Discussion of empirical analyses
4.1 Descriptive statistics
In Table 1, we provide a yearly (Panel A) and industry-wise (Panel B) distribution of operating leverage (DOL) and CEO incentives, namely, CEO debt holdings and CEO total pay in raw dollars. Results in Panel A show that the DOL of sample firms is fairly stable over our sample period. Also, the CEOs’ debt holdings are on average higher than their total pay except for the last two years of the sample suggesting why academics are increasingly focussing on the personal debt of CEOs. On the other hand, industry distribution is very steady with respect to both DOL and debt-to-total-pay ratio. This indicates that industry variation is not a significant concern (though we still control for possible industry effects in our regressions).
Table 2 contains descriptive statistics of the variables used in our main model. We find that the mean (median) of our OL variable, DOL, is 0.228 (0.184) with a standard deviation of 0.198. Our main independent variables CEORELDE and CEODE have mean values of 0.139 and 0.180, respectively. These numbers are comparable to those reported in previous studies (see e.g. Wei and Yermack, 2011; Chi et al., 2017). This table also shows that the average firm in our study is mid-sized [market value of $2.6bn, exp (7.865)], less leveraged (leverage = 21%) and has good growth potential (market-to-book = 3.4). Notably, about 51.5% of the observations are from firms where the CEO and the Chair of the board are the same person; 3.5% of all the CEOs in our sample are female; women occupy about 13.1% of the board seats; almost 85% of the directors are independent; and institutional ownership is almost 80%.
4.2 Correlations and variance inflation factors
Bivariate Pearson correlations between variables used in this study are presented in Table 3. As a primitive support to our stated hypothesis, we find that the correlation between DOL and CEORELDE is negative and statistically significant (ρ = −0.244, p < 0.01). Similarly, the correlation between DOL and CEODE is also negative and significant (ρ = −0.229, p < 0.01). Almost all the control variables have significant correlations with our dependent variable DOL, suggesting that we should include them in our regression analyses to better isolate the role of our primary independent variable.
To further mitigate concerns around potential multicollinearity, we also check the variance inflation factors (VIF) of the variables included in the analysis. In untabulated results, we find that the highest VIF is 3.18 for Firm Size, followed by 2.42 for CEO Total Pay. The rest of the VIFs are all below 1.82 with the average being 1.49. These VIFs indicate that multicollinearity is not a concern for our analysis.
4.3 Univariate results
Univariate tests of means are presented in Table 4. To study the incentive effects of inside debt intensity, we divide our sample into “High CEO inside debt holding” and “Low CEO inside debt holding”. Consistent with the theoretical underpinnings in Jensen and Meckling (1976) and Edmans and Liu (2011), if a CEO’s D/E is greater (less than or equal to) than that of the firm’s D/E, we classify that firm-year observation into the High CEO inside debt holding (Low CEO inside debt holding) group. We find that OL (DOL) is significantly lower for the High inside debt group (diff. between high and low groups = −0.040, t-stat for diff. = −9.26). This means that selling, general and administrative expenses for the High CEO inside debt firms are lower by 4% of lagged total assets on average, i.e. almost US$382m (mean lagged total assets of 9,545.62 × 0.04) or 7.4% of the market cap of an average High CEO inside debt firm. To sum up, the statistically significant lower DOL for high inside debt firms is also economically significant. These findings further substantiate our earlier findings on the correlations reported in Section 4.2.
4.4 Baseline regression results
Results from estimating our main OLS specification [equation (1)] are presented in Table 5. In all eight of the model specifications, the dependent variable is our proxy for OL, DOL. We use both the proxies for CEO inside debt holding in our analyses. We also use all the controls including year and industry dummies as described in Section 3.5. As discussed in our hypotheses and introduction sections, we expect to find a negative association between CEO inside debt holding and DOL.
In Column (1), we report the OLS regression estimates from our main model with CEORELDE as the proxy for inside debt and all the controls. Consistent with our hypothesis, we find a strong negative relationship between CEORELDE and DOL (coefficient = −0.0329; p < 0.05), suggesting that firms with higher CEO inside debt tend to have lower levels of OL. In Columns (2) to (4), we rerun the regression with various restricted specifications. Consistent with the findings in Column (1), we report that CEORELDE and DOL have a strong negative relationship in each of these instances. In Column (5), we estimate our main model but with CEODE as our primary independent variable. Consistent with our findings using CEORELDE, we find a negative and significant association between our inside debt proxy, CEODE and DOL (coefficient = −0.0192; p < 0.05). This association holds for restricted model specifications as well (Columns 6–8). These results are consistent with our findings from bivariate correlation analysis and our univariate results. We also note that the economic magnitude of our estimates is nontrivial. For example, coefficients in Columns (1) and (5) suggest that a one standard deviation increase in CEORELDE (CEODE) is related to a 3.5% (3%) decrease in the standard deviation of DOL of a firm. These results support the risk-aversion characteristics of inside debt since lowering OL is a way to reduce risk.
4.5 Robustness tests
4.5.1 Sensitivity to alternate measures for DOL.
So far, the dependent variable (degree of OL), in our model was computed as per the methodology described in Chen et al. (2019). However, an alternative method of computing DOL as described in Novy-Marx (2011) is also widely used in the literature. Therefore, in addition to our earlier proxy, we also use the Novy-Marx way of calculating DOL to ensure that our results are not sensitive to any specific measurement type. DOL using this method is measured as cost of goods sold (COGS) plus selling, general and administrative expenses (XGSA), divided by lagged total assets. Chen et al. (2019) note the following:
Different from Novy-Marx (2011), we do not include COGS in the numerator for two reasons. First, XGSA is much stickier than COGS. Compared with XSGA, COGS is much more responsive to fluctuations in sales and so is better characterized as variable costs. Second, the exclusion of COGS helps mitigate endogeneity concerns. COGS depends on the production of goods, and therefore the inclusion of COGS would make the firm’s operating leverage dependent on the amount produced (p. 373).
In addition to this measure, we have implemented the Kulchania (2016) and Aboody et al.’s (2018) measures [5]. Our results continue to hold. Results using this alternate measure as our dependent variable are presented in the Supplemental Appendix (SA henceforth) (see SA.1). These results are consistent with our main hypothesis – H1.
4.5.2 Sensitivity to alternate model specifications.
To control for cross-sectional and serial dependence (Petersen, 2009; Gow et al., 2010; Attig et al., 2014; Shen and Zhang, 2020) we use various alternate estimation procedures. Results are presented in SA.2. We report the Fama–Macbeth regressions (Columns 1–2), generalized linear model estimation “GLM” (Columns 3–4) and the Newey-West procedure to correct autocorrelation among the residuals (Columns 5–6). We continue to find a negative association between CEO inside debt and DOL with a strong statistical significance (e.g. p < 0.01 for all but one coefficient for the CEO inside debt holding; in Columns 6 the coefficient is significant at p < 0.05).
4.5.3 Sensitivity to alternate sample construct.
While it is not common in the OL literature to exclude firms from the regulated industries (such as financials and utilities), excluding such firms is quite common in the executive compensation literature. It is also true that our sample period covers the extreme years of Global Financial Crisis (2008–2009). To ensure that our results are not sensitive to any of these factors, we exclude financials (Columns 1–2), utilities (Columns 3–4) and GFC period (Columns 5–6). Our results continue to hold. These results are reported in SA.3.
4.6 Endogeneity tests
In results reported so far, we have found strong and robust evidence supporting our main hypothesis (H1) as we consistently find a negative association between CEO inside debt and OL. Regardless of our main findings, however, any empirical exploration like ours is susceptible to criticism around inherent endogeneity surrounding such relationships. There are two specific concerns: first, endogeneity could arise if unobservable factors influence both inside debt and DOL simultaneously, and the second concerns reverse causality. It could be argued that firms with a lower OL choose to provide their CEOs with higher debt-like incentives. We address these potential endogeneity concerns in different ways.
4.6.1 Difference-in-differences analysis using the implementation of IRC 409 A.
In the first of a string of tests to address, and hence assuage, endogeneity concerns, we use the tax-law change triggered by the final implementation of Internal Revenue Service provision 409 A as an exogenous shock in conjunction with the difference-in-difference (DiD) regression method in an effort to stablish a causal relation between CEO inside debt and DOL. Provision 409 A (final implementation effective on 2009/01/01), added to the Internal Revenue Code to regulate deferred compensation among other things, was ratified as a reaction to Enron executives fast-tracking their withdrawal of deferred compensation before the final demise of the firm.
Before IRC 409 A, corporate top officers could withdraw deferred incentives at any time and by any amount which potentially reduced their exposure to bankruptcy risk. IRC 409 A imposed limitations on the timing of such payouts. Notably, if any executive breached the clause, s/he will be subject to an immediate 20% additional tax and 1% interest. In essence, this provision makes early access to deferred incentives exorbitantly costly. Thus, it enhances bankruptcy risk exposure for top executives, which in turn improves the usefulness of the CEO’s deferred compensation. Accordingly, we expect the negative relationship between inside debt and DOL to be stronger during the post-implementation period.
In our exploration, following Shen and Zhang’s (2020) methodology, we identify 2009 as the event year. As argued by Shen and Zhang (2020), the final implementation date (2009/01/01) is of relevance rather than the date when the regulation was passed in 2004. Subsequently, [like Shen and Zhang] we identify 2006–2008 as our pre-event years and 2010–2012 as post-event years. The variable After is defined as an indicator variable that equals 1 if the firm-year observation belongs to the post-event years (2010–12), and 0 otherwise (i.e. pre-event years, 2006–2008). Furthermore, we require the sample firms to have at least one observation on either side of the event to qualify for the DiD test sample.
We calculate firm-specific change in CEORELDE (ΔCEORELDE) as the post-event period mean value of CEORELDE minus pre-event period mean value of CEORELDE. Next, we use ΔCEORELDE to rank the sample and then mark those that are above (below) median as treatment (control) firms. Subsequently, Treatment is a dichotomous variable that equals 1 (0) if ΔCEORELDE is above (below) median. Following extant literature, we use a propensity-score matching method to identify an observation from the control group for each observation in the treatment group (Brogaard et al., 2017; Shen and Zhang, 2020).
We use the propensity score matching technique where we use all controls in our main regression as explanatory variables for the logit model. Using the sample from the pre-event years, this logit model predicts the propensity score to match each observation from the treatment group (an observation with Treatment = 1) with an observation from the control group (Treatment = 0). To ensure a closer match, we require that the difference in probability is kept to a maximum of 0.01. Once this pre-event period PSM was done, only observations from the firms in the pre-event period were allowed in the post-event subsample. Our final sample for this DiD test includes 3,377 (3,410) firm-year observations over a six-year period for the CEORELDE (CEODE)-based analysis.
Univariate comparisons of firm characteristics between the control and treatment firms in the pre-event years are presented in Panel A of Table 6. The literature suggests that, for the DiD test, the treatment and matched control group observations should exhibit similar trends with respect to covariates (Lemmon and Roberts, 2010). In accordance with parallel trend assumption, in Panel A, we find that there is no significant difference between the treatment (HIGH ΔCEORELDE = 1) and control ((HIGH ΔCEORELDE = 0) observations with respect to the covariates [Columns (1) – (4) for CEORELDE-related results].
In Panel B of Table 6, we report results from the DiD regression. Our regression consists of three key independent variables: After (equals one if the observation is from 2010 to 2012), Treatment (equals one if ΔCEORELDE is above the median), and the interaction between the two (After × Treatment) along with controls. Our main variable of interest is the coefficient of the interaction term (After × Treatment), which captures the difference in the OL between firms with an increase and decrease in CIDH across the two periods surrounding the event year.
We report the regression estimates based on ΔCEORELDE (ΔCEODE) in Columns 1 (Columns 2). We find that the coefficient of the interaction variable is negative and statistically significant [see Column (1): coefficient = −0.0179; p < 0.05], indicating that firms with a larger increase in CEO inside debt holding have a significantly lower OL during the post-event years relative to firms with a larger reduction in inside debt holding. This result is in line with our expectation that the final implementation of provision 409 A improves the usefulness of CEO inside debt, which in turn results in lower OL in our case. Our results, as tabulated in Panels A and B in Table 7 show that the findings are qualitatively similar when we use CEODE as a proxy for CEO inside debt. Overall, the results in this Table not only provide evidence of causal relationship but also help allay concerns surrounding potential endogeneity.
4.6.2 Other endogeneity tests.
To further alleviate concerns related to endogeneity, we undertake two stage least squares (2SLS) analysis using instrumental variables (IVs) (SA.4); instrumental variable (IV) regression estimates using heteroskedasticity-based instruments (Lewbel, 2012) (SA.5); propensity score matching (PSM) (SA.6); entropy balancing (EB) approach (SA.7); various lagged specification that includes change regression (SA.8); and system GMM method (SA.9); change regression (SA.11); and firm-fixed effects specification (SA.12). The results and relevant discussions have been relegated to the Supplemental Appendix to conserve space.
It is to be noted that 2SLS-IV analysis address both omitted variable and reverse causality (Jha and Cox, 2015, p. 253, second column). PSM analysis addresses model misspecification and is better designed to assess a causal relation (Rosenbaum and Rubin, 1983). Entropy balancing is a newer technique used in accounting literature that tends to serve a purpose as PSM, but it does not suffer from sample size reduction (McMullin and Schonberger, 2020). Change regression analysis alleviates the concerns related to unobserved factors (Kang et al., 2018). Firm fixed effects specification is often used to alleviate omitted variables bias (Hossain and Masum, 2022), and hence, we use that as our final test of endogeneity.
4.7 The role of agency: negative association between CEO inside debt and operating leverage is more prominent for firms with more agency issues
Agency problems and incentive alignment are used interchangeably in the compensation literature (Bertrand and Mullainathan, 2001; Bebchuk et al., 2002; Bebchuk and Fried, 2003). In this section, we investigate if these issues moderate the inside debt–DOL relationship presented earlier. Extant literature on CEO inside debt holdings generally reports that deferred compensation helps assuage agency problems. Thus, we would expect the inside debt–DOL association to be stronger for firms with inherent agency (or incentive alignment) issues (H2).
First, we use a popular and widely used proxy for agency conflicts, i.e. the Entrenchment (or E-) index first introduced in Bebchuk et al. (2009) where a higher (lower) score of the E-index indicates higher (lower) governance. Results of this analysis are presented in Table 7 (Columns 1–2). In Columns (1) and (2), we classify agency as high (low) if the E-index score is higher (lower) than median. We report that the coefficient of CEORELDE is negative, significant and more pronounced only for poor governance sub-samples (coef. for high agency subsample = −0.0500, p < 0.01; and the coef. for a low agency subsample = 0.0114, p > 0.10). A Chi-test rejects the null hypothesis (p < 0.01) that coefficients for CEORELDE are the same for both subsamples.
Second, we use CEO power as a measurement of agency issues. Extant studies report that powerful CEOs have more decision-making authority and are more prone to overcompensation, which harms both shareholders and bondholders (Core et al., 1999). In other words, firms with powerful CEOs are prone to agency issues. We use CEO pay slice (Bebchuk et al., 2011) as a proxy for CEO power. A firm with above (below) median CEO pay slice is marked as one with high (low) agency issue. Results presented in Columns (3) and (4) are consistent with the previous findings in that CIDH-DOL association is more pronounced for firms with high agency issues (i.e. more powerful CEOs).
Third, we use CEO network size as a proxy for agency issues. Connections are often viewed as a source of CEO power (Engelberg et al., 2013). Thus, firms with CEOs with strong (weak) connections are more likely to have high (low) agency issues. Consistent with findings reported earlier in this section, we find that firms with Strong CEO Network (i.e. above median network size) are more prominent when it comes to the association between CIDH and DOL. These results are presented in Columns (5) and (6).
Results are qualitatively similar when we use CEODE as our proxy for CEO inside debt holding. These results are presented in SA.10. To sum up, results in this section consistently show that the negative association between CEO inside debt and OL is driven by firms with higher agency issues (i.e. poorer governance or powerful CEOs).
4.8 Does macro-level uncertainty play any role in influencing the negative association between CEO inside debt and operating leverage? [6]
Readers may argue that our main measure of OL is constructed using accounting information and business cycle may have profound implications for financial reporting. Therefore, it is important to investigate whether the documented negative association vary during economic certainty versus uncertainty.
We undertake two different tests to address this concern. First, we split our sample based on above (high uncertainty) and below (low uncertainty) average VIX years. The average value of VIX for our sample period (2006–2017) is 19.26. So, years 2007–2009 and 2011 were high uncertainty years and the rest were low uncertainty years. Results are presented in online appendix Table SA.13. Though the coefficient for CEORELDE is higher for High Uncertainty subsample, the differences in coefficients of CEORELDE between High and Low Uncertainty subsamples is not statistically significant (Chi-stat = 0.18; p = 0.67). The results are similar when we use CEODE as the proxy for CEO inside debt.
Second, we split the sample based on financial crisis (2008–9) vs non crisis years and reach similar conclusion as in the case of VIX splits mentioned. Results are presented in online appendix Table SA.14.
5. Conclusions
In this study, we have examined the risk-reducing incentives that result from the presence of deferred compensation components in CEO pay packages. Attempts to fast-track payout of deferred compensation by managers at Enron prior to its bankruptcy and the loss to personal wealth of CEOs following widespread firm failures resulting from the 2009 global financial crisis clearly suggest that CEOs with vulnerable personal portfolios will have strong incentives to reduce the risk of the firms they manage. While the choice of operating and capital structures constitute two obvious risk-reducing tools at management’s disposal, we limit our focus on inside debt-driven incentives for CEOs to engage in risk-reduction through changes in OL.
A growing body of literature associates high OL with firm risk, driven by the asymmetry in the cost of adjusting operating expenses in response to revenue shocks. Also, a large body of academic literature points to the value and risk implications of CEO inside debt. For example, there is evidence that announcements of large inside debt holdings of CEOs result in a wealth transfer from equity to debtholders, and a decrease in the volatility of both debt and equity securities. In addition, a higher percentage of pension compensation leads to more conservative decision-making by CEOs. Also, inside debt is positively related to firm cash holdings, a lower cost of debt financing along with a smaller number of restrictive debt covenants, a lower incidence of tax sheltering and a lower payout policy. Inside debt also attracts higher price for the debt of such firms in the secondary bond market. More importantly, extant evidence reports that higher CEO inside debt is associated with lower firm risk and higher firm liquidation values.
We tested our hypothesis of a negative association between CEO inside debt and the degree of OL using a large sample of listed U.S. companies over the period 2006 to 2017 consisting of 11,145 firm-year observations from 1,722 unique companies. We employ two commonly used proxies for CEO inside debt holding, namely, CEO debt-to-equity scaled by a firm’s debt-to-equity, and CEO debt-to-equity ratios. Debt-holding is computed as the present value of deferred incentives and accumulated pension, and equity-holding as accumulated stock options and stock holdings (see, for example, Chi et al., 2017). We find strong evidence confirming the expected negative association between CEO inside debt and DOL suggesting that firms with higher inside debt tend to maintain lower levels of OL. These findings continue to hold with an alternative measure for the inside debt proxy (an alternative measure for OL [Novy-Marx, 2011], and survive a range of tests designed to rule out potential endogeneity concerns. In addition, our findings demonstrate that the presence of manager-shareholder agency conflicts can strengthen the inside debt–DOL relationship suggesting the strong role of inside debt in reducing firm risk.
By examining the incentive effects of deferred compensation, a very significant contribution of this study centers on the design of compensation packages that serve the interests of a broader group of stakeholders instead of the prevailing emphasis on shareholder wealth maximization. As the primary focus of the executive compensation literature is on the risk and value implications of CEO equity-based compensation, a study of the incentive effects of an equally important and growing component of CEO pay packages, namely, inside debt, will add significantly to this burgeoning literature. A focus on a pay component with a completely opposite impact on firm risk relative to equity-based compensation underlines the importance of this analysis. We are thereby contributing to the ongoing debate on the design of CEO pay packages that serve the interests of a broader group of stakeholders rather than focussing purely on shareholder wealth maximization. Indeed, our study shines a spotlight on the fact that the failure of a firm affects not just a CEO, but many other interested parties.
Notes
1.Pension and deferred compensation are fixed claims on firm assets held by corporate insiders. Such claims enjoy priority over the claims of shareholders and hence resemble the payout pattern of outside debt holders. See, for e.g., Sundaram and Yermack (2007) and Wei and Yermack (2011) for further details.
2.These tests include difference-in-difference analysis surrounding implementation of IRC409A, instrumental variable approach, propensity score matching approach, lagged model specification approach and an entropy balancing approach.
3.Equity-based compensation does not uniformly provide risk-increasing incentives, however. For example, studies show that deep in-the-money options could lead to higher levels of CEO risk aversion (see, e.g., Lambert et al., 1991; Lewellen, 2006).
4.This is akin to cross-sectional variation in risk-reduction benefit across firms with varying levels of tangible and intangible assets (growth opportunities, brand name, consumer loyalty, etc.) as firms with mostly intangible assets have less debt and would therefore value reductions in cash flow variability more highly (see, for e.g., Smith and Stulz, 1985).
5.We thank an anonymous reviewer for suggesting these measures. Kulchania (2016) first computes ex ante expectations of operating costs and sales that are based on the geometric growth rate over the prior two years. Differences from expectations (innovation) are then computed. The cost structure is then obtained as the coefficient for the computed innovations in the growth rate of sales by regressing innovations of growth in operating costs on the innovations of growth in sales. For additional details on this methodology, please refer to Appendix A in Kulchania (2016).
6.We thank an anonymous reviewer for suggesting these tests.
The authors thank Harjeet Bhabra, Hasibul Chowdhury, Mostafa Hasan, Xiaobing Ma, and Jinghua Nie for their comments on the earlier version of the paper. The authors also acknowledge the significant contributions of two anonymous referees whose insightful comments and suggestions significantly helped in bringing the document to its final form. Hossain thanks Memorial University of Newfoundland and the Social Sciences and Humanities Research Council of Canada (SSHRC, Grant #430-2020-00275) for providing financial support.
Sample distribution
| Year | N | DOL | CEO DEBT | CEO TOTAL PAY | Debt to total pay |
|---|---|---|---|---|---|
| Panel A: Year-by-year sample distribution | |||||
| 2006 | 825 | 0.239 | 4,326 | 3,753 | 1.15 |
| 2007 | 880 | 0.243 | 5,015 | 4,177 | 1.20 |
| 2008 | 881 | 0.236 | 4,453 | 4,069 | 1.09 |
| 2009 | 937 | 0.226 | 4,879 | 3,636 | 1.34 |
| 2010 | 1,018 | 0.241 | 4,958 | 4,203 | 1.18 |
| 2011 | 984 | 0.240 | 5,084 | 4,541 | 1.12 |
| 2012 | 975 | 0.229 | 5,102 | 4,558 | 1.12 |
| 2013 | 977 | 0.227 | 5,315 | 4,909 | 1.08 |
| 2014 | 960 | 0.219 | 5,636 | 5,184 | 1.09 |
| 2015 | 927 | 0.219 | 5,497 | 5,385 | 1.02 |
| 2016 | 929 | 0.217 | 4,953 | 5,528 | 0.90 |
| 2017 | 852 | 0.198 | 4,086 | 6,008 | 0.68 |
| Panel B: Industry-by-industry sample distribution | |||||
| Industry description | |||||
| Consumer nondurables | 750 | 0.228 | 4,980 | 4,666 | 1.07 |
| Consumer durables | 367 | 0.227 | 4,975 | 4,714 | 1.06 |
| Manufacturing | 1634 | 0.228 | 4,956 | 4,636 | 1.07 |
| Energy | 460 | 0.228 | 4,906 | 4,645 | 1.06 |
| Chemicals | 484 | 0.228 | 4,964 | 4,683 | 1.06 |
| Business equipment | 2303 | 0.229 | 4,967 | 4,634 | 1.07 |
| Telecom | 228 | 0.225 | 5,007 | 4,820 | 1.04 |
| Utilities | 645 | 0.228 | 4,940 | 4,633 | 1.07 |
| Shops | 1366 | 0.229 | 4,987 | 4,635 | 1.08 |
| Health care | 985 | 0.228 | 4,941 | 4,656 | 1.06 |
| Finance | 353 | 0.225 | 4,946 | 4,818 | 1.03 |
| Other | 1570 | 0.227 | 4,952 | 4,706 | 1.05 |
Note:This table reports the sample distribution by year (Panel A) and by industry (Panel B)
Source: Tables by authors
Descriptive statistics
| Variables | N | Mean | SD | 25th pct. | Median | 75th pct. |
|---|---|---|---|---|---|---|
| DOL | 11,145 | 0.228 | 0.198 | 0.079 | 0.184 | 0.329 |
| CEORELDE | 11,145 | 0.139 | 0.212 | 0.000 | 0.028 | 0.206 |
| CEODE | 11,145 | 0.180 | 0.309 | 0.000 | 0.028 | 0.229 |
| CEO Age | 11,145 | 63.722 | 7.319 | 59.000 | 63.000 | 69.000 |
| CEO Tenure | 11,145 | 7.717 | 7.035 | 2.000 | 6.000 | 11.000 |
| CEO Dual | 11,145 | 0.514 | 0.500 | 0.000 | 1.000 | 1.000 |
| CEO Female | 11,145 | 0.035 | 0.183 | 0.000 | 0.000 | 0.000 |
| CEO Total Pay | 11,145 | 8.035 | 0.944 | 7.369 | 8.060 | 8.731 |
| CEO Pay Vega | 11,145 | 3.285 | 2.167 | 1.379 | 3.714 | 5.007 |
| Board Size | 11,145 | 2.204 | 0.239 | 2.079 | 2.197 | 2.398 |
| Board Independence | 11,145 | 0.849 | 0.078 | 0.818 | 0.875 | 0.900 |
| Board Diversity | 11,145 | 0.131 | 0.103 | 0.000 | 0.125 | 0.200 |
| Inst. Own. Total | 11,145 | 0.790 | 0.153 | 0.708 | 0.826 | 0.906 |
| Inst. Own. HHI | 11,145 | 0.050 | 0.024 | 0.034 | 0.044 | 0.057 |
| Firm Size | 11,145 | 7.865 | 1.636 | 6.651 | 7.737 | 8.983 |
| Leverage | 11,145 | 0.210 | 0.167 | 0.056 | 0.200 | 0.323 |
| M/B | 11,145 | 3.396 | 3.459 | 1.579 | 2.390 | 3.853 |
| Dividend Payout | 11,145 | 0.015 | 0.023 | 0.000 | 0.007 | 0.021 |
| Tangibility | 11,145 | 0.265 | 0.231 | 0.085 | 0.182 | 0.385 |
| Return Volatility | 11,145 | 0.121 | 0.070 | 0.073 | 0.120 | 0.169 |
| KZ Score | 11,145 | 0.459 | 0.967 | 0.042 | 0.524 | 1.012 |
| Debt Maturity | 11,145 | 0.306 | 0.327 | 0.014 | 0.202 | 0.469 |
Notes:This table presents descriptive statistics of the variables used in this study. We winsorize the continuous variables at the 1% and 99% levels. Variable definitions are provided in Appendix
Source: Tables by authors
Correlation matrix
| No. | Variables | [1] | [2] | [3] | [4] | [5] | [6] | [7] | [8] | [9] | [10] | [11] | [12] | [13] | [14] | [15] | [16] | [17] | [18] | [19] | [20] | [21] | [22] |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| [1] | DOL | 1 | |||||||||||||||||||||
| [2] | CEORELDE | −0.244** | 1 | ||||||||||||||||||||
| [3] | CEODE | −0.229** | 0.990** | 1 | |||||||||||||||||||
| [4] | CEO Age | −0.052** | 0.079** | 0.073** | 1 | ||||||||||||||||||
| [5] | CEO Tenure | 0.082** | −0.104** | −0.095** | 0.389** | 1 | |||||||||||||||||
| [6] | CEO Dual | −0.068** | 0.150** | 0.138** | 0.293** | 0.326** | 1 | ||||||||||||||||
| [7] | CEO Female | 0.002 | 0.026** | 0.022* | −0.067** | −0.067** | −0.009 | 1 | |||||||||||||||
| [8] | CEO Total Pay | −0.159** | 0.177** | 0.155** | −0.037** | −0.105** | 0.142** | 0.017 | 1 | ||||||||||||||
| [9] | CEO Pay Vega | 0.009 | 0.037** | 0.020* | 0.056** | −0.049** | 0.112** | −0.038** | 0.392** | 1 | |||||||||||||
| [10] | Board Size | −0.191** | 0.304** | 0.274** | 0.025** | −0.192** | 0.091** | 0.016 | 0.439** | 0.213** | 1 | ||||||||||||
| [11] | Board Independence | −0.171** | 0.221** | 0.200** | −0.123** | −0.275** | 0.043** | 0.030** | 0.289** | 0.160** | 0.373** | 1 | |||||||||||
| [12] | Board Diversity | −0.022* | 0.209** | 0.192** | −0.086** | −0.154** | 0.097** | 0.255** | 0.295** | 0.099** | 0.355** | 0.260** | 1 | ||||||||||
| [13] | Inst. Own. Total | 0.011 | −0.072** | −0.077** | −0.117** | −0.045** | −0.055** | −0.022* | 0.187** | 0.134** | −0.049** | 0.114** | 0.024* | 1 | |||||||||
| [14] | Inst. Own. HHI | 0.102** | −0.081** | −0.064** | −0.011 | 0.056** | −0.085** | 0.036** | −0.397** | −0.252** | −0.233** | −0.146** | −0.151** | −0.348** | 1 | ||||||||
| [15] | Firm Size | −0.237** | 0.210** | 0.188** | −0.007 | −0.105** | 0.167** | 0.005 | 0.723** | 0.361** | 0.549** | 0.250** | 0.325** | 0.062** | −0.486** | 1 | |||||||
| [16] | Leverage | −0.338** | 0.199** | 0.182** | −0.020* | −0.083** | 0.062** | −0.002 | 0.273** | 0.061** | 0.285** | 0.175** | 0.183** | 0.012 | −0.052** | 0.224** | 1 | ||||||
| [17] | M/B | 0.161** | −0.039** | −0.039** | −0.077** | −0.010 | 0.015 | 0.019* | 0.171** | 0.073** | 0.064** | 0.034** | 0.124** | 0.011 | −0.099** | 0.287** | 0.189** | 1 | |||||
| [18] | Dividend Payout | 0.022* | 0.113** | 0.105** | 0.040** | −0.022* | 0.059** | 0.048** | 0.094** | −0.014 | 0.127** | −0.003 | 0.124** | −0.169** | −0.043** | 0.237** | 0.015 | 0.290** | 1 | ||||
| [19] | Tangibility | −0.343** | 0.215** | 0.205** | 0.094** | −0.035** | 0.081** | 0.033** | 0.020* | −0.097** | 0.135** | 0.054** | 0.058** | −0.127** | 0.004 | 0.097** | 0.279** | −0.093** | 0.048** | 1 | |||
| [20] | Return Volatility | 0.006 | 0.011 | 0.010 | −0.002 | −0.009 | −0.004 | 0.009 | 0.010 | 0.014 | 0.014 | 0.009 | 0.025** | 0.011 | −0.005 | 0.019* | −0.002 | 0.012 | 0.015 | −0.003 | 1 | ||
| [21] | KZ Score | −0.115** | −0.019 | −0.019* | −0.072** | −0.011 | −0.010 | −0.041** | 0.116** | 0.053** | 0.046** | 0.073** | 0.021* | 0.159** | −0.037** | 0.023* | 0.515** | 0.128** | −0.727** | 0.091** | −0.013 | 1 | |
| [22] | Debt Maturity | 0.067** | −0.021* | −0.022* | 0.009 | 0.003 | 0.003 | −0.011 | −0.076** | −0.009 | 0.002 | −0.018 | −0.006 | −0.074** | 0.065** | −0.081** | −0.027** | −0.033** | −0.055** | −0.079** | −0.009 | 0.027** | 1 |
Notes:This table presents correlations between variables used in this study. * and ** are significant at p < 0.05 and p < 0.01 levels, respectively. Variable definitions are provided in AppendixSource: Tables by authors
Univariate mean-test results
| Variables | CEO D/E > Firm D/E | CEO D/E < Firm D/E | Differences of Mean | t-stat for mean diff. |
|---|---|---|---|---|
| (1) | (2) | (3) = (1) – (2) | (4) | |
| DOL | 0.198 | 0.238 | −0.040 | −9.26*** |
| CEO Age | 64.586 | 63.428 | 1.157 | 7.28*** |
| CEO Tenure | 6.617 | 8.091 | −1.474 | −9.67*** |
| CEO Dual | 0.638 | 0.472 | 0.166 | 15.38*** |
| CEO Female | 0.042 | 0.032 | 0.010 | 2.53*** |
| CEO Total Pay | 8.330 | 7.934 | 0.396 | 19.59*** |
| CEO Pay Vega | 3.657 | 3.158 | 0.499 | 10.62*** |
| Board Size | 2.300 | 2.171 | 0.129 | 25.41*** |
| Board Independence | 0.873 | 0.841 | 0.031 | 18.90*** |
| Board Diversity | 0.162 | 0.120 | 0.042 | 18.98*** |
| Inst. Own. Total | 0.783 | 0.793 | −0.010 | −2.95*** |
| Inst. Own. HHI | 0.044 | 0.051 | −0.007 | −13.34*** |
| Firm Size | 8.553 | 7.631 | 0.922 | 26.72*** |
| Leverage | 0.209 | 0.211 | −0.002 | −0.55 |
| M/B | 3.673 | 3.302 | 0.371 | 4.93*** |
| Dividend Payout | 0.022 | 0.013 | 0.009 | 18.04*** |
| Tangibility | 0.300 | 0.252 | 0.048 | 9.62*** |
| Return Volatility | 0.122 | 0.121 | 0.001 | 0.59 |
| KZ Score | 0.245 | 0.531 | −0.286 | −13.72*** |
| Debt Maturity | 0.349 | 0.291 | 0.058 | 8.11*** |
Notes:This table presents univariate mean difference test results of the variables included in the main model. We classify the sample into two groups: we mark as high (low) CIDH groups those where CEO’s debt-to-equity ratio is greater (lower) than the firm’s debt-to-equity. ***, ** and * represent significance at 1, 5 and 10% levels, respectively. Variable definitions are provided in Appendix
Source: Tables by authors
Baseline results
| Variables | Dependent variable = DOL | |||||
|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | |
| CEORELDE | −0.0329** (−2.12) | −0.0371** (−2.36) | −0.0258* (−1.75) | |||
| CEODE | −0.0192** (−1.97) | −0.0222** (−2.24) | −0.0156* (−1.65) | |||
| CEO Age | −0.0021*** (−4.01) | −0.0021*** (−4.04) | ||||
| CEO Tenure | 0.0018*** (3.17) | 0.0018*** (3.20) | ||||
| CEO Dual | 0.0016 (0.25) | 0.0014 (0.22) | ||||
| CEO Female | −0.0085 (−0.57) | −0.0085 (−0.57) | ||||
| CEO Total Pay | 0.0095** (2.16) | 0.0094** (2.14) | ||||
| CEO Pay Vega | 0.0047*** (3.48) | 0.0046*** (3.47) | ||||
| Board Size | 0.0448** (2.47) | 0.0401** (2.15) | 0.0439** (2.43) | 0.0391** (2.11) | ||
| Board Independence | −0.0226 (−0.42) | −0.0142 (−0.26) | −0.0238 (−0.44) | −0.0156 (−0.29) | ||
| Board Diversity | 0.0976** (2.57) | 0.0943*** (2.60) | 0.0963** (2.53) | 0.0928** (2.55) | ||
| Inst. Own. Total | −0.0537** (−2.25) | −0.0405* (−1.66) | −0.0538** (−2.26) | −0.0407* (−1.67) | ||
| Inst. Own. HHI | −0.0139 (−0.10) | −0.0257 (−0.18) | −0.0131 (−0.09) | −0.0244 (−0.17) | ||
| Firm Size | −0.0363*** (−10.32) | −0.0303*** (−10.11) | −0.0259*** (−11.25) | −0.0363*** (−10.32) | −0.0304*** (−10.11) | −0.0261*** (−11.34) |
| Leverage | −0.3877*** (−11.26) | −0.3817*** (−11.22) | −0.3645*** (−10.84) | −0.3885*** (−11.28) | −0.3827*** (−11.24) | −0.3656*** (−10.88) |
| M/B | 0.0104*** (9.92) | 0.0103*** (9.74) | 0.0103*** (9.81) | 0.0104*** (9.92) | 0.0103*** (9.73) | 0.0103*** (9.81) |
| Dividend Payout | 1.3319*** (4.03) | 1.2918*** (3.93) | 1.2320*** (3.69) | 1.3303*** (4.02) | 1.2908*** (3.93) | 1.2336*** (3.69) |
| Tangibility | −0.0992*** (−4.06) | −0.1051*** (−4.26) | −0.1037*** (−4.15) | −0.0996*** (−4.07) | −0.1055*** (−4.27) | −0.1040*** (−4.16) |
| Return Volatility | 0.0093 (0.52) | 0.0085 (0.46) | 0.0105 (0.57) | 0.0093 (0.51) | 0.0085 (0.46) | 0.0105 (0.57) |
| KZ Score | 0.0365*** (4.38) | 0.0377*** (4.54) | 0.0343*** (4.11) | 0.0366*** (4.38) | 0.0378*** (4.55) | 0.0344*** (4.12) |
| Debt Maturity | 0.0095 (1.25) | 0.0088 (1.14) | 0.0119 (1.53) | 0.0095 (1.25) | 0.0088 (1.14) | 0.0119 (1.53) |
| Constant | 0.5306*** (8.62) | 0.4485*** (8.88) | 0.4654*** (23.50) | 0.5346*** (8.72) | 0.4517*** (9.02) | 0.4658*** (23.52) |
| Observations | 11,145 | 11,145 | 11,145 | 11,145 | 11,145 | 11,145 |
| R2 | 0.533 | 0.525 | 0.521 | 0.533 | 0.525 | 0.521 |
Notes:This table presents the baseline results based on our main model [Equation (1)]. The sample contains 11,145 firm-year observations from 1,722 unique firms during the period 2006 to 2017. Our starting year is restricted by the fact that CEO debt data is only available from fiscal year 2006. We remove firm-year observations with missing dependent, independent and control variables. We winsorize the continuous variables at the 1% and 99% levels. Variable definitions are provided in Appendix. Our main dependent variable is DOL which is selling, general and administrative expenses scaled by lagged total assets. Our main variables of interest are CEO inside debt measures proxied by CEORELDE [LN (1+ CEO D/E divided by Firm’s D/E)] and CEODE [LN (1+CEO D/E)]. We control for various CEO, board, monitoring, and firm attributes. We use Fama-French industry (FF-48) and year dummies in all our regressions. Robust t-statistics using standard errors clustered at firm-level are reported in parentheses. Significance at the 10%, 5% and 1% level is indicated by *, ** and ***, respectively. In Columns (1) and (5), we present results with our full model, while other columns represent various restricted versions of our main model
Source: Tables by authors
Difference-in-differences analysis
| Variables | Treatment group |
Control group |
Diff. |
t-stat for diff. | Treatment |
Control |
Diff. |
t-stat for diff. |
|---|---|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |
| Panel A: Pre event period (2006–2008) covariate mean comparison of treatment and control groups | ||||||||
| CEO Age | 67.230 | 67.826 | −0.596 | −1.60 | 67.126 | 67.791 | −0.665 | −1.59 |
| CEO Tenure | 7.506 | 7.910 | −0.404 | −1.02 | 7.528 | 7.918 | −0.390 | −0.98 |
| CEO Dual | 0.478 | 0.476 | 0.001 | 0.05 | 0.474 | 0.477 | −0.003 | −0.11 |
| CEO Female | 0.016 | 0.022 | −0.006 | −0.79 | 0.016 | 0.020 | −0.004 | −0.61 |
| CEO Total Pay | 7.787 | 7.766 | 0.021 | 0.41 | 7.783 | 7.758 | 0.025 | 0.51 |
| CEO Pay Vega | 3.624 | 3.515 | 0.109 | 1.11 | 3.622 | 3.555 | 0.067 | 0.69 |
| Board Size | 2.200 | 2.213 | −0.014 | −1.14 | 2.206 | 2.212 | −0.006 | −0.48 |
| Board Independence | 0.838 | 0.838 | 0.000 | 0.08 | 0.838 | 0.837 | 0.001 | 0.12 |
| Board Diversity | 0.103 | 0.105 | −0.001 | −0.27 | 0.101 | 0.104 | −0.003 | −0.59 |
| Inst. Own. Total | 0.778 | 0.771 | 0.007 | 0.84 | 0.779 | 0.770 | 0.009 | 1.03 |
| Inst. Own. HHI | 0.050 | 0.052 | −0.002 | −1.13 | 0.051 | 0.052 | −0.001 | −0.93 |
| Firm Size | 7.527 | 7.497 | 0.029 | 0.36 | 7.511 | 7.494 | 0.016 | 0.20 |
| Leverage | 0.201 | 0.201 | 0.000 | −0.03 | 0.197 | 0.200 | −0.002 | −0.26 |
| M/B | 2.775 | 2.642 | 0.133 | 1.01 | 2.853 | 2.648 | 0.205 | 1.47 |
| Dividend Payout | 0.013 | 0.013 | 0.000 | 0.08 | 0.013 | 0.013 | −0.001 | −0.51 |
| Tangibility | 0.285 | 0.297 | −0.012 | −0.97 | 0.286 | 0.299 | −0.013 | −1.06 |
| Return Volatility | 0.120 | 0.123 | −0.003 | −0.80 | 0.120 | 0.123 | −0.003 | −0.72 |
| KZ Score | 0.459 | 0.452 | 0.007 | 0.14 | 0.491 | 0.453 | 0.038 | 0.78 |
| Debt Maturity | 0.310 | 0.313 | −0.003 | −0.15 | 0.320 | 0.308 | 0.012 | 0.67 |
| Dependent variable = DOL | ||
| Variables | Treatment is HIGH ΔCEORELDE = 1 | Treatment is HIGH ΔCEODE = 1 |
| (1) | (2) | |
| Panel B: Regression analysis | ||
| After × Treatment | −0.0179** (−2.10) | −0.0024** (−2.13) |
| Treatment | −0.0209** (−2.16) | −0.0101* (−1.91) |
| After | −0.0148 (−1.15) | −0.0115 (−0.80) |
| Observations | 1,392 | 1,416 |
| R2 | 0.0022 | 0.0026 |
Notes:This table presents difference-in-differences (DiD) analyses that exploits the final enforcement of Section 409 A of the Internal Revenue Code (IRC) as an exogenous shock to CIDH. Panel A reports the comparison between treatment and matched firms in pre-event period (2006–2008). Panel B reports the DiD regression estimates. In Panel B, After is a dummy variable that takes a value of one if the firm-year observation belongs to the post-event period (2010–2012), and zero if it is from the pre-event period (2006–2008); Treatment is a dummy variable that takes the value of one if a firm experiences top-tercile changes (i.e. greater increases) in CEORELDE (CEODE), and zero if the firms experience bottom-tercile changes (i.e. lowest increase) in CEORELDE (CEODE). We use the predicted propensity score from the logit regression to the match treatment observation (Treatment = 1) with a matched observation (Treatment = 0) using the closest propensity score. We report the regression estimations based on ΔCEORELDE (ΔCEODE) in Columns 1 and 2, respectively. Robust t-statistics using standard errors clustered at firm-level are reported in parentheses. Significance at the 10, 5 and 1% level is indicated by *, ** and ***, respectively. Variable definitions are provided in Appendix
Source: Tables by authors
Negative association between CEO inside debt and operating leverage is more pronounced for firms with agency issues
| Dependent variable = DOL | ||||||
|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | |
| High agency | Low agency | High agency | Low agency | High agency | Low agency | |
| Variables | (High entrenchment) | (Low entrenchment) | (High CEO pay slice) | (Low CEO pay slice) | (Strong CEO network) | (Weak CEO network) |
| CEORELDE | −0.0500*** (−3.18) | −0.0114 (−0.49) | −0.0531*** (−3.63) | −0.0085 (−0.34) | −0.0519*** (−3.31) | −0.0149 (−0.56) |
| CEO Age | −0.0015* (−1.93) | −0.0016** (−2.43) | −0.0017*** (−2.78) | −0.0023*** (−3.44) | −0.0022*** (−2.91) | −0.0017** (−2.35) |
| CEO Tenure | 0.0006 (0.79) | 0.0020*** (2.98) | 0.0019*** (2.96) | 0.0015** (2.01) | 0.0020** (2.28) | 0.0015** (2.12) |
| CEO Dual | 0.0090 (0.95) | 0.0001 (0.01) | −0.0096 (−1.40) | 0.0131 (1.51) | −0.0042 (−0.46) | 0.0065 (0.76) |
| CEO Female | −0.0095 (−0.33) | 0.0074 (0.46) | 0.0057 (0.37) | −0.0282 (−1.42) | −0.0254 (−1.62) | 0.0058 (0.25) |
| CEO Total Pay | 0.0106 (1.53) | 0.0091 (1.57) | 0.0137** (2.06) | 0.0172** (2.53) | 0.0052 (0.94) | 0.0100* (1.68) |
| CEO Pay Vega | 0.0049** (2.34) | 0.0029* (1.73) | 0.0052*** (3.65) | 0.0050*** (2.75) | 0.0056*** (3.51) | 0.0033 (1.56) |
| Board Size | 0.0721*** (2.76) | 0.0247 (1.05) | 0.0543*** (2.76) | 0.0359 (1.48) | 0.0457* (1.91) | 0.0458* (1.79) |
| Board Independence | −0.0049 (−0.06) | −0.0039 (−0.06) | −0.0099 (−0.14) | −0.0161 (−0.28) | −0.0679 (−0.95) | −0.0191 (−0.26) |
| Board Diversity | 0.0893 (1.59) | 0.0740 (1.64) | 0.1078** (2.43) | 0.0963** (1.97) | 0.0933* (1.73) | 0.0908* (1.93) |
| Inst. Own. Total | −0.0047 (−0.14) | −0.0721** (−2.08) | −0.0631** (−2.34) | −0.0484* (−1.65) | −0.1047*** (−3.28) | −0.0215 (−0.63) |
| Inst. Own. HHI | −0.0425 (−0.18) | −0.0350 (−0.18) | −0.0378 (−0.20) | −0.0433 (−0.26) | 0.1646 (0.77) | −0.1354 (−0.79) |
| Firm Size | −0.0428*** (−8.10) | −0.0329*** (−7.75) | −0.0384*** (−8.54) | −0.0390*** (−8.36) | −0.0362*** (−8.23) | −0.0393*** (−7.75) |
| Leverage | −0.4463*** (−7.85) | −0.3230*** (−7.67) | −0.3912*** (−9.00) | −0.3892*** (−8.98) | −0.4414*** (−9.83) | −0.3479*** (−7.35) |
| M/B | 0.0120*** (6.76) | 0.0103*** (7.26) | 0.0095*** (7.64) | 0.0115*** (7.24) | 0.0095*** (7.06) | 0.0111*** (7.04) |
| Dividend Payout | 2.2371*** (4.50) | 0.8723* (1.89) | 1.7070*** (3.21) | 1.1424*** (3.09) | 1.8737*** (3.80) | 1.1098*** (2.82) |
| Tangibility | −0.1628*** (−5.57) | −0.0284 (−0.96) | −0.1081*** (−4.12) | −0.0816** (−2.51) | −0.0778*** (−2.64) | −0.1123*** (−3.29) |
| Return Volatility | −0.0044 (−0.17) | 0.0139 (0.55) | 0.0221 (0.98) | −0.0098 (−0.35) | 0.0297 (1.31) | −0.0024 (−0.09) |
| KZ Score | 0.0557*** (3.95) | 0.0196* (1.79) | 0.0450*** (3.69) | 0.0322*** (3.25) | 0.0482*** (4.09) | 0.0298*** (2.67) |
| Debt Maturity | −0.0035 (−0.33) | 0.0154* (1.66) | 0.0161* (1.70) | 0.0050 (0.55) | 0.0121 (1.18) | 0.0099 (1.00) |
| Constant | 0.4251*** (4.34) | 0.5031*** (6.62) | 0.4533*** (5.37) | 0.5153*** (6.86) | 0.6338*** (7.53) | 0.5032*** (6.00) |
| Observations | 4,135 | 5,256 | 5,573 | 5,572 | 5,559 | 5,586 |
| R2 | 0.562 | 0.585 | 0.564 | 0.521 | 0.554 | 0.541 |
| Chi-stat | 8.75 | 10.75 | 7.43 | |||
| p-value | 0.00 | 0.00 | 0.01 | |||
Notes:This table reports the results from our cross-sectional analyses. For all the subsample tests, the dependent variable is DOL and the primary independent variable is CEORELDE. We use Fama-French industry (FF-48) and year dummies in all our regressions. Robust t-statistics using standard errors clustered at firm-level are reported in parentheses. Significance at the 10, 5 and 1% level is indicated by *, ** and ***, respectively. Chi-square test statistics and p-value are provided where the null hypothesis is coefficients for CEORELDE for both subsamples are the same. We use three proxies for agency issues, namely, Entrenchment (i.e. Bebchuk et al., 2009’s E-index); CEO pay slice (Bebchuk et al., 2011); and CEO network size (Engelberg et al., 2013). For each of these measures, a higher score means more agency issues. We split our sample along the median for each of these proxies for agency issues. We replicate this table using our alternate measure of CEO inside debt (i.e. CEODE). Those results are relegated to Supplemental Appendix (see SA.10)
Source: Tables by authors
Summary of online appendix tables
| Table No. | Type | Description | Includes discussion of results? | Reference in text |
|---|---|---|---|---|
| SA.1 | Robustness test | This table reports results using Novy-Marx (2011) measure of operating leverage | no | 4.5.1 |
| SA.2 | Robustness test | This table reports some alternate model specifications | no | 4.5.2 |
| SA.3 | Robustness test | This table reports some alternate sample construct | no | 4.5.3 |
| SA.4 | Endogeneity test | This table presents two-stage least squares (2SLS) regression results | yes | 4.6.2 |
| SA.5 | Endogeneity test | This table reports instrumental variable (IV) regression estimates using heteroskedasticity-based instruments (Lewbel, 2012) | yes | 4.6.2 |
| SA.6 | Endogeneity test | This table reports results using propensity score matched (PSM) samples | yes | 4.6.2 |
| SA.7 | Endogeneity test | This table reports our results from the entropy balancing approach | yes | 4.6.2 |
| SA.8 | Endogeneity test | This table presents results from various lagged model specifications | yes | 4.6.2 |
| SA.9 | Endogeneity test | This table reports the two-step system GMM results for the association between CEORELDE (CEODE) and DOL | yes | 4.6.2 |
| SA.10 | Sensitivity test | This table reports the cross-sectional tests using our second measure of CEO inside that, i.e. CEODE | no | 4.7 |
| SA.11 | Endogeneity test | This table reports the change regression analysis | no | 4.6.2 |
| SA.12 | Endogeneity test | This table reports firm-fixed-effects and high-definition-fixed-effects regression analysis | no | 4.6.2 |
| SA.13 | Cross-sectional test | This table reports cross-sectional tests using below vs above average VIX years | no | 4.8 |
| SA.14 | Cross-sectional test | This table reports cross-sectional tests using financial crisis vs non-financial crisis years | no | 4.8 |
Source: tables by authors
Variable definitions
| Variables | Description |
|---|---|
| Main variables | |
| DOL | As in Chen et al. (2019), selling, general and administrative expenses scaled by total assets in year (t-1) |
| CEORELDE | Natural log [1+(CEO D/E)/(Firm D/E)] |
| CEODE | Natural log [1+CEO D/E] |
| CEO Age | Age of the CEO in year t |
| CEO Tenure | Number of years the CEO has been in his/her current role as of year t |
| CEO Dual | An indicator variable that equals to one if the CEO and chairperson of the board is the same person |
| CEO Female | An indicator variable that equals to one if the CEO is female |
| CEO Total Pay | Natural log of one plus total pay for the year t for the CEO as reported in ExecuComp |
| CEO Pay Vega | Vega is defined by Coles et al., 2013 as “Dollar change in wealth associated with a 0.01 change in the standard deviation of the firm's returns (in $000 s)”. Our measure is calculated as natural log of one plus the Vega as in Coles et al. (2006 and 2013) |
| Board Size | Natural log of the number of directors on the board |
| Board Independence | Percent of independent directors on the board |
| Board Diversity | Percent of female directors on the board |
| Inst. Own. Total | Percentage of total institutional ownership |
| Inst. Own. HHI | Concentration (HHI) of institutional ownership as reported in Thomson 13f |
| Firm Size | Natural log of market value of equity |
| Leverage | The ratio of total debt (DLC+DLTT) to total assets |
| M/B | Market value of equity to book value of equity |
| Liquidity | Cash and cash equivalents scaled by total assets |
| Dividend Payout | Dividend paid scaled by total assets |
| R&D | R&D expense scaled by total assets |
| Tangibility | Property, plant and equipment divided by total assets |
| Return Volatility | Standard deviation of monthly returns for the past twelve months |
| KZ Score | Kaplan and Zingales (1997) score as calculated in Jha and Cox (2015).It is the Kaplan and Zingales (1997) index of financial constraint. Specifically, it is calculated as follows: −1.002 * (CF/L.at) − 39.368 * (div/L.at) − 1.315 * (C/L.at) + 3.139 * lev + 0.283 * Qraj where Qraj = ((prcc_f*csho) + at − (ceq + txdb))/at. Source: COMPUSTAT |
| Debt Maturity | [(debt maturing in current liabilities, DLC + debt maturing in two years, DD2 + debt maturing in three years, DD3)/(debt in current liabilities, DLC + long-term debt, DLTT)] |
| Other variables (in alphabetical order) | |
| CEO Network | This data is collected from Boardex and includes the total number of connections that the CEO has. It is basically his/her networking strength |
| CEO Pay Slice | It is calculated as the ratio of CEO-pay to top-5 executives’' pay where top-executives include CEO himself/herself. We use CEO pay slice (Bebchuk et al., 2011) as a proxy for CEO power. A firm with above (below) median CEO pay slice is marked as one with high (low) agency issue |
| DOLNOVY-MARX | Operating leverage measure as used in Novy-Marx (2011) |
| DOL_KULCHANIA | Operating leverage measure as used in Kulchania (2016) |
| DOL_ABOODYETAL | Operating leverage measure as used in Aboody et al. (2018) |
| Entrenchment | We use the Entrenchment Index (popularly known as the E-index) as proposed by Bebchuk et al. (2009). A higher (lower) score of the E-index indicates higher (lower) governance |
| State Income Tax | Marginal income tax rate at the state level |
| State Mortgage Subsidy | This is the state-level mortgage subsidy rate |
© Emerald Publishing Limited.