Firm cash holdings and CEO inside debt

Journal of Banking & Finance 42 (2014) 83–100

Contents lists available at ScienceDirect

Journal of Banking & Finance journal homepage: www.elsevier.com/locate/jbf

Firm cash holdings and CEO inside debt Yixin Liu a,1, David C. Mauer b,⇑, Yilei Zhang c,2 a

University of New Hampshire, Peter T. Paul College of Business and Economics, 10 Garrison Avenue, Durham, NH 03824, United States Department of Finance, Tippie College of Business, University of Iowa, 108 John Pappajohn Business Bldg., S320, Iowa City, IA 52242-1994, United States c Department of Finance, College of Business and Public Administration, University of North Dakota, Grand Forks, ND 58202, United States b

a r t i c l e

i n f o

Article history: Received 13 February 2013 Accepted 18 January 2014 Available online 30 January 2014 JEL classification: G30 G32 G34

a b s t r a c t We examine the effect of CEO pensions and deferred compensation (inside debt) on firm cash holdings and the value of cash. We document a positive relation between CEO inside debt and firm cash holdings. This positive effect is magnified by firm leverage and mitigated by the presence of financial constraints. We further find that the marginal value of cash to shareholders declines as CEO inside debt increases. Our evidence supports the view that inside debt tilts managerial incentives toward bondholders and helps balance the competing interests of stockholders and bondholders. The evidence also suggests, however, that inside debt can harm shareholder value by encouraging excess cash holdings. Ó 2014 Elsevier B.V. All rights reserved.

Keywords: Cash holdings Inside debt Pension Deferred compensation

1. Introduction Executive compensation packages typically include salaries, bonuses, stock and options, pensions, and deferred compensation. The stock and option components are widely believed to align a manager’s interests with those of stockholders. Lesser known are the incentive effects of debt-like components such as pensions and deferred compensation. These compensation components, collectively referred to as inside debt, are unsecured and typically underfunded obligations that resemble debt-like claims against the company and thereby help align managers’ incentives with those of the firm’s bondholders. Recent studies conclude that inside debt represents a significant component of CEOs’ compensation and that it has become increasingly popular to compensate CEOs with inside debt.3 ⇑ Corresponding author. Tel.: +1 319 335 0944. E-mail addresses: [email protected] (Y. Liu), [email protected] (D.C. Mauer), [email protected] (Y. Zhang). 1 Tel.: +1 (603) 862 3357. 2 Tel.: +1 (701) 777 3407. 3 Bebchuk and Jackson (2005) and Sundaram and Yermack (2007) document the contribution of pensions to CEO compensation and recent studies by Cen (2010), Lee and Tang (2011), Cassell et al. (2012), and Anantharaman et al. (2013) document the increasing contribution of pensions and deferred compensation for the ExecuComp universe of firms starting with the period from 2006 when the SEC’s executive compensation disclosure requirements were expanded to include pensions and deferred compensation. http://dx.doi.org/10.1016/j.jbankfin.2014.01.031 0378-4266/Ó 2014 Elsevier B.V. All rights reserved.

In an attempt to understand how inside debt affects corporate decision-making, in this paper we study the effect of inside debt on corporate cash policy. As shown by Edmans and Liu (2011), when CEOs hold debt-like claims they are more likely to behave like bondholders. The closer alignment of CEOs with bondholders as induced by inside debt has important implications for stockholder–bondholder conflicts. Corporate cash policy seems to be an ideal policy in which to explore the links between debt compensation incentives and stockholder–bondholder conflicts. On the one hand, firm cash policy choices have been shown to be significant policy choices for both stockholders and bondholders because of the potentially high impact of firm cash balances on the riskiness of their claims.4 On the other hand, corporate cash policy, to a large extent, is at the discretion of managers with little scrutiny from outside investors. As such, cash policy can serve as an interesting and useful backdrop with which to study how managerial debtlike compensation incentives affect firm stakeholders. Using compensation data from ExecuComp, we compute CEO inside debt as the sum of pension value and deferred compensation. Pension value is the aggregate present value of the CEO’s 4 For example, see Acharya et al. (2007) and Acharya et al. (2012). Other papers that examine the determinants and value of corporate cash policy include Kim et al. (1998), Opler et al. (1999), Faulkender and Wang (2006), Pinkowitz et al. (2006), Dittmar and Mahrt-Smith (2007), Harford et al. (2008), Bates et al. (2009), Denis and Sibilkov (2010), Liu and Mauer (2011), and Dittmar and Duchin (2012).

84

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

accumulated benefits under the firm’s pension plan at the end of a fiscal year and deferred compensation is the aggregate balance in non-tax-qualified deferred compensation plans at the end of a fiscal year. Excluding financial and utility firms, we then match this data with Compustat and CRSP data, which results in a sample of 6009 firm-year observations over the period of 2006–2011.5 We test several hypotheses about the relation between firm cash policy and CEO inside debt. We find that our evidence is more consistent with CEO risk aversion than a spending hypothesis originally put forth by Harford et al. (2008) to explain their finding that cash balances are smaller in firms with weaker corporate governance structures. In particular, we find that inside debt and its components are a greater share of total CEO wealth in firms with weaker governance structures. We document a positive relation between cash holdings and both the proportion of the CEO’s wealth represented by inside debt and the relative CEO-firm debt-to-equity ratio. The effect of inside debt on cash balances is economically significantly. In particular, regression estimates indicate that a one standard deviation increase in inside debt increases cash balances by 3.7–6.2% for the mean firm in our samples. This positive influence of inside debt on cash balances is driven by both pension and deferred compensation components, although deferred compensation tends to be more influential. Interestingly, we also find a nonlinear relation between cash balances and inside debt by firm leverage level. As firm leverage increases the magnitude of the positive relation between cash and inside debt increases but eventually turns negative for firms with relatively high levels of leverage. Our finding that CEO debt-like compensation is associated with higher firm cash balances is consistent with the view that CEOs have incentives to reduce risk as they become more aligned with bondholders. In their study of CEO equity-compensation, Liu and Mauer (2011) document that greater equity incentives as measured by the sensitivity of equity compensation to volatility (vega) are also associated with higher corporate cash holdings. Given that both CEO debt- and equity-compensation incentives increase cash holdings, it is interesting to ask whether the effect of inside debt is independent of the vega effect. We find that the inside debt has a positive effect on cash holdings regardless of whether we control for vega. This suggests that inside debt and vega influence cash holdings through different channels. If inside debt encourages greater risk aversion, and therefore induces CEOs to hold excess cash, CEOs may face constraints in their ability to do so. One constraining factor could be the financial constraint status at the firm level. Unlike CEOs at financially unconstrained firms that can raise capital relatively easily, CEOs in financially constrained firms may face difficulty accumulating cash reserves as their inside debt increases, simply because capital is limited. This argument implies a mitigating effect of the firm’s financial constraint status on the relation between inside debt and cash holdings. To investigate, we interact several measures of financial constraint with CEO inside debt in our cash holding regressions. We generally find significantly negative coefficients on the interaction terms, lending support to the mitigating role of financial constraints on the cash-inside debt relation. Our cash results suggest but are insufficient to show that although the cash policy of managers may better align them with bondholders, the excess cash attributable to inside debt may harm shareholder wealth and thereby portend greater stockholder– bondholder conflicts. Using the methodology in Faulkender and Wang (2006), we examine the marginal value of cash as a function of CEO inside debt holdings. We find the marginal value of cash to

5 The sample starts in 2006 because it is the first year the SEC required firms to report executive pension benefits and deferred compensation plans.

shareholders is decreasing in CEO inside debt. For example, we find that the value of an additional dollar of cash for the mean firm with below median inside debt is $1.89, while the value of an additional dollar of cash for the mean firm with above median inside debt is $1.01, which is a decrease of 47%. This evidence strongly suggests that stockholders internalize the negative effect of inside debt on managerial incentives and discount the value of cash when inside debt is high. Our paper makes several contributions to the literature. First, we contribute to the emerging literature on how managerial debt-like compensation influences the policy choices of firms. For example, Lee and Tang (2011) and Cassell et al. (2012) find that firm leverage, R&D expenditures, and stock return volatility are decreasing in inside debt, while firm diversification is increasing in inside debt. We are the first paper, however, to examine the influence of inside debt on firm cash policy.6 Second, to our knowledge we are the first paper that examines the joint influence of CEO equity compensation incentives and debt compensation incentives on firm policy decisions. We find that the effect of inside debt on cash policy is independent of the effect of equity compensation on cash policy, with both having a positive effect on cash policy.7 Third, we are the first to document that the marginal value of cash to equityholders is decreasing in inside debt, which suggests that inside debt may exacerbate stockholder–bondholder conflicts by inducing firms to hold excessive amounts of cash. The finding that inside debt increases as shareholder rights decrease and that cash holdings are increasing in inside debt supports this view. The remainder of the paper is organized as follows. We develop our hypotheses in Section 2 and provide a brief review of the cash literature and discussion of CEO inside debt. Section 3 describes our data and the variables that we use in our empirical analysis. Section 4 reports our results, and Section 5 concludes. 2. CEO inside debt and corporate cash balances We start with a brief discussion of the literature on the determinants of cash balances. We then discuss the concept of inside debt and the two types of CEO inside debt. This is followed by a discussion of hypotheses on how CEO inside debt will influence corporate cash balances.

2.1. Determinants of corporate cash balances There is a large literature examining the determinants of corporate cash holdings. The early studies by Kim et al. (1998) and Opler et al. (1999) find strong evidence that cash balances are built to hedge external financing frictions. In particular, these studies find that smaller firms with strong growth opportunities, riskier cash flows, and higher information asymmetry hold more cash, while larger firms with ready access to external sources of finance hold less cash. Denis and Sibilkov (2010) extend this literature by examining the relation between cash holdings, investment, and financial constraints. They find that high cash balances are used by financially constrained firms to finance positive net present value investments and not simply to pursue empire building. Along other dimensions, Harford et al. (2008) find that poorly governed firms hold less cash – the spending hypothesis – and Dittmar and Duchin (2012) find that managerial conservatism can help explain large 6

Cassell et al. (2012) find that working capital is increasing in inside debt. Liu and Mauer (2011) find that CEO equity compensation incentives as measured by vega have a positive effect on cash holdings and attribute their finding to bondholders anticipating greater risk-taking in high vega firms and thereby requiring greater liquidity. Our results suggest that the positive relation between cash holdings and CEO inside debt is attributable to the separate effect of CEO risk aversion. 7

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

cash balances in firms with ready access to external finance. Finally, Bates et al. (2009) find that the secular growth in cash held by U.S. firms over the last three decades is partly driven by an increase in cash flow risk which is linked to an increase in idiosyncratic risk. The literature has paid much less attention to the role which managerial compensation plays in determining cash balances. Liu and Mauer (2011) find that firms with CEOs that have compensation which is sensitive to equity volatility (i.e., high vega compensation like stock options) hold more cash. They explain this finding by noting that creditors will demand higher cash balances in firms when CEOs have greater incentives to take risk. The analysis in this paper focuses on the influence of CEO debt-like compensation (e.g., pensions and deferred compensation) on corporate cash balances. We develop our testable hypotheses below. 2.2. CEO inside debt CEO pensions and deferred compensation are collectively referred to as inside debt. Pensions for CEOs invariably involve much larger payments than regular employee pensions and are classified as non-qualified compensation plans. In these plans, contributions are not taxed until they are received – retirement or separation from the firm – and the firm does not get an income tax deduction until the contributions are distributed and the employee pays tax. Deferred compensation plans for CEOs have the same characteristics, except the CEO elects to defer part of his/her compensation to a later date. To encourage executives to defer part of their salary, firms usually specify a rate at which they will match salary deferrals. Importantly, both CEO pensions and deferred compensation plans are unfunded and unsecured to preserve their tax-deferral benefits and to be exempt from the rules applicable to ordinary tax-qualified plans under ERISA (Employee Retirement Income Security Act). In particular, ‘‘unfunded’’ means that although pension and deferred compensation accounts accumulate and grow in accordance with the stated asset allocation, the firm never actually allocates money to the pension or the deferred compensation plan of the employee.8 The firm is obligated only to make payments at retirement. Furthermore, since these compensation plans are not secured; CEOs and other top executives as beneficiaries of these plans are classified by bankruptcy courts as just another unsecured creditor of the firm. 2.3. Hypotheses Following Jensen and Meckling (1976) and Edmans and Liu (2011), much of the empirical literature on inside debt takes an optimal contracting perspective focusing on the relation between the CEO’s debt-to-equity ratio and the firm’s debt-to-equity ratio. For example, Wei and Yermack (2011) examine how the CEO-firm relative debt-to-equity ratio influences stockholder and bondholder reactions to the disclosure of CEO inside debt, while Anantharaman et al. (2013) examine how the relative leverage ratio influences the cost of debt and covenants in bank loans (i.e., debt contracting). Our objective is to examine how CEO inside debt influences the choice of cash balances, and not necessarily whether inside debt is part of an optimal contracting solution. We adopt the perspective that the CEO’s compensation (e.g., cash, equity, and in8 Note that because deferred compensation plans are unfunded, compensation experts often claim that deferred executive salaries can be used as a source of financing for the firm. This might even be true for pensions of CEOs since the literature shows that there is a substitution effect between salary and pension benefits. For example, in a sample of Fortune 500 companies, Gerakos (2010) finds that an additional dollar of pension benefits decreases salary by 48 cents.

85

side debt) is optimal subject to frictions such as relative bargaining power between managers and shareholders (i.e., corporate governance). In much of our analysis we therefore focus on how the choice of cash balances is influenced by the ratio of CEO inside debt (and its pension and deferred compensation components) to the sum of inside debt and total equity compensation, given the (constrained) optimal compensation choice.9 There are several testable hypotheses about the relation between corporate cash holdings and CEO inside debt.

2.3.1. Risk aversion hypothesis Given that with pension and deferred compensation the CEO forgoes current salary for future benefits and that the present value of these future benefits are unsecured, inside debt exposes the CEO to bankruptcy risk just as if the CEO held a piece of the firm’s risky unsecured debt. Since cash balances reduce firm asset volatility and are the firm’s most liquid assets, the risk aversion hypothesis predicts that cash balances are increasing in the ratio of inside debt to total compensation. Since the risk aversion hypothesis focuses on the view that inside debt is essentially unsecured corporate debt, we would expect that the positive relation between firm cash balances and CEO inside debt gets stronger as the firm’s leverage increases. Thus, if the risk aversion hypothesis is correct, we would expect little relation between cash balances and inside debt for zero leverage firms10 and an increasingly stronger positive relation between cash balances and inside debt as firm leverage increases. Interestingly, we might also expect that the relation turns negative for firms with very high leverage, since the debt service burden will likely prevent CEOs from building large cash balances.

2.3.2. Spending hypothesis In contrast to the risk aversion hypothesis, the spending hypothesis predicts a negative relation between cash balances and inside debt. The logic follows from the negative relation between cash balances and firms with weak corporate governance and the positive relation between CEO inside debt and weak corporate governance. In particular, Harford et al. (2008) find that firms with weak corporate governance as measured by, for example, the Gompers et al. (2003) G-Index or CEO-chair duality, have significantly smaller cash balances than firms with good governance structures. Harford et al. sum up their findings as follows: ‘‘The conclusion we draw from our study is that lower cash reserves in poorly governed firms are the result of a decision by managers to spend the cash flow and any accumulated cash quickly, rather than allowing it to accumulate even though such accumulation might provide future flexibility’’ (p. 537). Similarly, Cen (2010) and Lee and Tang (2011) and this study (see Table 3) document a positive relation between CEO inside debt and weak corporate governance. The upshot is that the spending hypothesis predicts that cash balances will be decreasing in the ratio of inside debt to total compensation. It is important to note that a positive relation between inside debt and cash balances could be explained by powerful CEOs in poorly governed firms. In particular, a powerful CEO (e.g., a CEO who is also the chairman of the board of directors) might push for more debt-like compensation when the firm has high cash balances and thereby low financial distress risk.11 Although this is certainly plausible, as noted below we find that the relation between CEO inside debt and cash is not (significantly) influenced by variables predicting CEO power and overall firm corporate governance. 9 We show, however, that our results are robust to using the CEO-firm relative debt-to-equity ratio. 10 Except perhaps because of a concern about economic distress (as opposed to financial distress) caused by, for example, product market competition. 11 We thank an anonymous referee for bringing this to our attention.

86

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

2.3.3. Financial contracting hypothesis Sundaram and Yermack (2007) find that CEO pensions are increasing in firm leverage which is consistent with the financial contracting perspective that inside debt offsets the incentive of equity-compensated managers to favor policies that benefit equityholders at the expense of bondholders. Given a negative relation between cash balances and leverage (see, e.g., Kim et al. (1998)) and in particular, the view that cash is negative debt,12 the financial contracting hypothesis predicts a negative relation between cash balances and inside debt. It is important to note that financial constraints should influence the predicted relations between cash balances and inside debt. Financially constrained firms tend to be smaller with a larger portion of their investment opportunity sets consisting of growth options and low or no payouts. The literature documents that financially constrained firms tend to build cash balances for hedging and/or precautionary savings motives (see, e.g., Kim et al. (1998) and Opler et al. (1999)), since external financing may be costly or not available. Given that inside debt (e.g., pensions) provides future cash commitments, Sundaram and Yermack (2007) argue that constrained firms are more likely to eschew debt-like compensation in favor of stock and options which do not directly affect the firm’s cash flows. As such, firms facing significant financial constraints are likely to have higher cash balances and lower inside debt. The upshot is that financial constraints are predicted to induce a negative relation between cash balances and inside debt, which would bias our results in favor of the spending and optimal contracting hypotheses and against the risk-aversion hypothesis.

3. Data and descriptive statistics We use the ExecuComp database to construct a sample of US firms from 2006 to 2011. The ExecuComp database provides yearly data on executive compensation such as salary, bonus, stock options, restricted stock grants, and accumulated stock and option holdings for the top executives of firms in the Standard and Poor’s (S&P) 500, S&P Midcap 400, and S&P Smallcap 600. We start from 2006 because it is the first year firms were required by the SEC to disclose and describe their top executives’ deferred compensation plans, pension benefits and other post-employment payments which are reported in the ExecuComp database. We focus our analysis on the CEO’s pension benefits and deferred compensation, where pension benefits are the aggregate actuarial present value of the CEO’s accumulated benefits under the company’s pension plan and deferred compensation is the CEO’s aggregate balance in non-tax-qualified deferred compensation plans. We exclude firms in the financial service industries in which liquidity is hard to assess (Standard Industrial Classification (SIC) codes 6000–6999) and in the utility sector due to their special regulatory status (SIC codes 4900–4999). Firm-specific accounting variables are obtained from Compustat, and stock returns are obtained from CRSP. Following Faulkender and Wang (2006), we eliminate firm-years for which net assets are negative, the market value of equity is negative, or dividends are negative. Finally, we require our sample firm-years to have non-missing ExecuComp, Compustat, and CRSP data to compute the variables used in our analyses. Our final sample consists of 6009 firm-years. Table 1 reports descriptive statistics for variables used in our cash holdings, inside debt, and cash value regressions. We define and discuss the variable choices below.13

2.4. CEO inside debt and the value of cash 3.1. Cash The risk aversion hypothesis predicts that CEO inside debt has a negative effect on the marginal value of an additional dollar of cash from shareholders’ perspective because CEOs with significant amounts of inside debt accumulate cash to lower risk. This follows from the contingent-claims perspective that equity is an option on the value of the firm’s assets whose value therefore decreases as firm value volatility decreases. The spending hypothesis, however, has an indeterminate effect on the market value of cash. On the one hand, it can be argued that since inside debt coincides with weak shareholder rights and since managers of such firms suboptimally spend cash, the effect of inside debt on the marginal value of an additional dollar of cash from shareholders’ perspective could be positive. This presumes, however, that the additional cash will add to cash reserves and be used to invest in positive net present value projects. If this is not the case and/or if equityholders perceive that the additional cash will be spent on wasteful projects, then the marginal value of an additional dollar of cash will be negative. The latter implication means that we must depend on the cash-inside debt relation to distinguish between the risk aversion and spending hypotheses, because the value of cash could be negative under both hypotheses. In contrast to the spending hypothesis, the financial contracting perspective largely predicts a mechanical negative relation between cash balances and inside debt, and does not by itself have any clear predictions about the effect of inside debt on the marginal value of cash to equityholders. We examine the relation between the value of cash and CEO inside debt by augmenting the methodology of Faulkender and Wang (2006) to estimate the influence of CEO inside debt on the value of an additional dollar of cash to equityholders. 12 See, e.g., Acharya et al. (2007) for a review of the evidence on the negative relation between cash and leverage and the conditions under which such a view is theoretically valid.

Following Opler et al. (1999), we measure corporate cash holdings as the ratio of cash and marketable securities to net assets, where net assets are total assets minus cash and marketable securities.14 3.2. Debt compensation incentives For a firm-year, inside debt is the sum of CEO pension and deferred compensation. The CEO’s pension is the aggregate actuarial present value of the executive’s accumulated benefit under the company’s pension plans at the end of the fiscal year. The CEO’s deferred compensation is the aggregate balance in non-tax-qualified deferred compensation plans at the end of the fiscal year. We scale inside debt and its components by the sum of CEO inside debt and equity value, where equity value is the sum of the value of the CEO’s common stock holdings in the firm plus the dividendadjusted Black–Scholes value of option holdings, all measured at the end of the fiscal year end. Note that stock and option holdings include current year grants and all accumulated stock and option holdings (i.e., total CEO equity wealth in the firm). Finally, the CEO-firm relative debt-to-equity ratio is the CEO’s debt-to-equity ratio divided by the firm’s debt-to-equity ratio, where CEO debt is the value of inside debt, CEO equity is the total value of CEO equity wealth in the firm, the firm’s debt is the sum of long-term debt plus debt in current liabilities, and the firm’s equity is the market value of equity. We cannot compute the CEO-firm relative 13 We discuss the computation of the cash value variables later when we present the cash value regressions. 14 Our regression results are robust to alternative measures of cash holdings (e.g., the ratio of cash plus marketable securities to total assets). These results are available upon request.

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

debt-to-equity ratio for firm-years when the firm has no long-term debt or debt in current liabilities. Thus we have 5218 (out of 6009) valid firm-year observations for the CEO-firm debt-to-equity ratio. 3.3. Equity compensation incentives We compute the vega of a CEO’s equity compensation as the dollar change in the value of the CEO’s option grants and any option holdings for a 0.01 change in the annualized standard deviation of stock returns, scaled by CEO total current compensation, where total current compensation in a year includes bonus, restricted stock and option grants, long-term incentive payouts, and any other compensation. The vega computation follows the methods in Core and Guay (2002), who use the dividend-adjusted version of the Black and Scholes model to compute the value of executive stock options. We also follow Coles et al. (2006) in assuming that the vega of any stockholdings, including restricted stock, is zero. 3.4. Instruments We use several different instruments for inside debt in our regression models. The instruments include firm age, CEO age, and CEO tenure. A firm’s age in a given sample year is the number of years since the first year that the firm is reported in Compustat; CEO age is the age of the CEO as reported in the ExecuComp database; and CEO tenure is the number of years that the current CEO has served in that capacity as reported in ExecuComp. These instruments for CEO compensation are used by, for example, Coles et al. (2006), Brockman et al. (2010), and Liu and Mauer (2011). In their study of CEO pensions, Sundaram and Yermack (2007) find that the value of the CEO’s pension is positively related to firm age, CEO age, and CEO tenure. Indeed, as seen in Table 2, we also find strong correlations between all three instruments and our measure of inside debt and its pension and deferred compensation components. Thus, on a first pass at least, firm age, CEO age, and CEO tenure do not appear to be weak instruments. Another potential concern, however, is whether these instruments are correlated with cash in a direct way beyond their indirect association with inside debt which would render them invalid.15 We address these concerns below with formal tests of weak instruments and instrument validity (i.e., exogeneity). 3.5. Governance variables We use several measures of corporate governance. Following the cash papers of Dittmar and Mahrt-Smith (2007) and Harford et al. (2008), we use the G-Index of Gompers et al. (2003). The G-index is the number of antitakeover provisions in the firm’s charter as reported by the Investor Responsibility Research Center (IRRC) and varies from zero (perfect shareholder rights) to 24 (perfect managerial entrenchment).16 Following Masulis et al. (2007), we also separate the G-index into Democracy and Dictator dummy variables based on whether the index is below or above a critical value, 15 It is difficult to argue that older firms need to hold more cash than younger firms, especially since older well established firms are much less likely to face financing constraints. From an economic perspective, it therefore seems quite plausibly that firm age is related to firm (excess) cash holdings only through its association with CEO debt-like compensation. Similarly, older CEOs and CEOs with longer tenures would not be expected to be concerned about building excess cash balances unless higher cash balances are one way in which they can mitigate the risk of losing their pensions and deferred compensation. Overall, it seems economically reasonable that firm age, CEO age, and CEO tenure are valid instruments (i.e., uncorrelated with the error term in our cash regressions). 16 IRRC’s most recent data is 2006 and so we use a time-invariant G-index based on 2006 data in our analysis of the determinants of inside debt (see Table 3).

87

respectively. We create another dummy variable which is set equal to one if the CEO is also the chairman of the board of directors, and zero otherwise. Finally, we use the median dollar value of director stock ownership in the firm. Bhagat and Bolton (2013) argue that dollar ownership of board members is more reliable and less prone to bias and measurement error than other corporate governance measures.17 3.6. Financial constraint variables We use a number of variables to measure financial constraint status. In addition to a number of variables based on firm size, growth opportunities, and payouts, we create a dummy variable for whether the firm has a bond rating and we use the Whited and Wu (WW) index. The index developed by Whited and Wu (2006) is rigorously grounded in theory and is shown by them and Hennessy and Whited (2007) to outperform competing indices (e.g., the Kaplan and Zingales (1997) index) in identifying financially constrained firms.18 For a firm-year, the Whited and Wu index is computed as

WWindexi;t ¼ 0:091CF i;t  0:062DIVPOSi;t þ 0:021TLTDi;t  0:044LNTAi;t  0:035SGi;t þ 0:102ISGi;t

ð1Þ

where for firm i in year t, CFi,t is the ratio of cash flow to book assets, DIVPOSi,t is an indicator that equals one if the firm pays dividends, and zero otherwise, TLTDi,t is the ratio of long-term debt to total assets, LNTAi,t is the natural log of total assets, SGi,t is own-firm sales growth computed as Sales(t)/Sales(t  1), and ISGi,t is the firm’s three-digit industry sales growth. The index is design so that higher values for the index indicate greater financial constraint. 3.7. Control variables We follow Bates et al. (2009) and use their cash holding regression variables as controls in our cash holdings regressions and we follow Sundaram and Yermack (2007) and use their pension regression variables as controls in our inside debt regressions.19 There are many common variables across the two papers, so for brevity the following list of controls will not separate the variables into those used in the cash holdings regressions and those used in the inside debt regressions. Firm size is measured by the logarithm of net assets. Market-tobook is computed as the book value of net assets minus the book value of equity plus the market value of equity, all divided by the book value of net assets. Cash flow/net assets is earnings after interest, dividends and taxes but before depreciation divided by the book value of net assets. NWC/net assets is the net working capital-to-net assets ratio. CAPEX/net assets is the ratio of capital expenditures to the book value of net assets. Leverage is the sum of long-term debt and debt in current liabilities divided by the book value of net assets. Industry sigma is the mean of the standard deviations of cash flow/net assets over 10 years for firms in the same industry, as defined by two-digit SIC codes.20 Dividend is a dummy variable equal to one in years in which a firm pays a common dividend, and zero otherwise. R&D/sales is the ratio of research and development expense to sales. This ratio is set equal to 17 We also use board size, board independence, and institutional ownership but the relation of these governance measures to CEO inside debt is much weaker. These results are available upon request. 18 The financial constraint index developed by Kaplan and Zingales (1997) is also inappropriate for our empirical analysis because one of the five firm characteristics used to construct the index is firm cash holdings. 19 See Table 3 in Bates et al. (2009) and Table 6 in Sundaram and Yermack (2007). 20 There are always at least three firms to compute industry sigma for the firms in our sample.

88

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

Table 1 Descriptive statistics and correlations of firm characteristics and CEO inside debt and its components. Variable

Mean

1st Quartile

Median

3rd Quartile

Std. dev.

N

Cash

0.263

0.039

0.113

0.294

0.421

6009

0.030 0.000 0.003 0.195

0.181 0.070 0.056 1.154

0.168 0.132 0.087 2.060

6009 6009 6009 5218 5218

CEO inside debt, inside debt components, and CEO-firm relative debt-equity ratio Inside debt 0.116 0.000 Pension 0.067 0.000 Deferred compensation 0.047 0.000 Relative debt-equity ratio 1.128 0.000 Relative debt-equity ratio > 1 0.273 Instruments Firm age CEO age CEO tenure

22.040 55.120 7.638

13.000 50.000 2.916

20.000 55.000 5.671

37.000 60.000 10.000

11.460 6.776 6.703

6009 5935 5908

Governance variables G-Index Democracy (G-Index 6 5) Dictator (G-Index P 14) Median director stock ($1000) CEO-Chair duality

9.246 0.064 0.041 1530.000 0.261

8.000

9.000

11.000

2.448

296.563

827.764

1870.000

2150.000

4281 4281 4281 3962 6009

Financial constraint variables NoRating WW Index

0.497 0.208

0.326

0.257

0.189

0.336

0.032 7.320 2.008 0.101 0.071 0.058 0.249 0.132 0.456 0.046 0.036 0.103 0.520

0.009 6.189 1.211 0.063 0.022 0.023 0.068 0.052

0.021 7.279 1.618 0.101 0.077 0.040 0.224 0.107

0.041 8.395 2.378 0.149 0.186 0.073 0.361 0.154

0.035 1.637 1.145 0.130 0.188 0.056 0.219 0.113

0.000 0.000

0.003 0.002

0.052 0.032

0.085 0.076

0.064

0.213

0.017

0.207

0.598

6009

0.008 0.151 0.025 0.008 0.000 0.000 0.000 0.151 0.005

0.022 0.038 0.018 0.052 0.000 0.002 0.000 0.029 0.052

0.003 0.092 0.006 0.011 0.000 0.000 0.000 0.121 0.010

0.034 0.194 0.028 0.073 0.001 0.002 0.001 0.226 0.013

0.094 0.179 0.225 0.304 0.011 0.011 0.008 0.142 0.125

6009 6009 6009 6009 6009 6009 6009 6009 6009

Control variables Vega Log firm size Market-to-book Cash flow/net assets NWC/net assets CAPEX/net assets Leverage Industry sigma Dividend R&D/sales Acquisition activity Liquidity Tax status Cash value regression variables rit  RBit DCt Ct1 DE t DNAt DRDt DIt DD t Lt NFt

6009 6009 6009 6009 6009 6009 6009 6009 6009 6009 6009 6009 6009 6009 6009

The sample includes all firm-years in the ExecuComp database from 2006 to 2011 where data is available to compute CEO inside debt and compensation incentives and where accounting data is available on Compustat. The sample starts in 2006 because it is the first year the SEC required firms to report executive deferred compensation plans, pension benefits, and other post-employment payments. The sample excludes financial and utility firms. Cash is the ratio of cash plus marketable securities to net assets, where net assets are total assets minus cash and marketable securities. Inside debt is the sum of CEO pension and deferred compensation scaled by total compensation, where pension is the aggregate actuarial present value of the CEO’s accumulated benefits under the company’s pension plan at the end of the fiscal year, deferred compensation is the aggregate balance in non-tax-qualified deferred compensation plans at the end of the fiscal year, and total compensation is the sum of CEO inside debt and equity value. Equity value is the sum of the value of the CEO’s common stock holdings in the firm plus the dividend-adjusted Black–Scholes value of option holdings, all measured at the fiscal year end. Note that stock and option holdings include current year grants and all accumulated stock and option holdings (i.e., total CEO equity wealth in the firm). The variables Pension and Deferred compensation are the components of inside debt scaled by the same measure of total compensation (defined above) used to scale inside debt. The relative debt-equity ratio is the CEO’s debt-to-equity ratio divided by the firm’s debt-to-equity ratio. CEO debt is the sum of CEO pension and deferred compensation and CEO equity value is the sum of the value of the CEO’s common stock holdings in the firm plus the Black–Scholes value of option holdings, all measured at the fiscal year end. The firm’s debt-to-equity ratio is total debt (long-term debt plus debt in current liabilities) divided by the market value of equity. Relative debt-to-equity ratio > 1 is a dummy variable which is equal to one when the relative ratio is greater than one, and zero otherwise. A firm’s age in a given sample year is the number of years since the firm is first reported in the Compustat database. CEO age is the age of the CEO as reported in the ExecuComp database. CEO tenure is the number of years that the current CEO has served in that capacity as reported in the ExecuComp database. The G-Index is the number of antitakeover provisions in the firm’s charter as reported by the Investor Responsibility Research Center (IRRC) in 2006 and varies from zero to 24. Democracy is a dummy variable which is equal to one if the G-Index is less than or equal to 5, and zero otherwise. Dictator is a dummy variable which is equal to one if the G-Index is greater than or equal to 14, and zero otherwise. Median director stock is the median dollar value of the stock in the firm owned by members of the board of directors. CEO-Chair duality is a dummy variable equal to one if the CEO is also the chairman of the board of directors, and zero otherwise. No rating is a dummy variable which is equal to one if the firm does not have a credit rating, and zero otherwise. WW Index is the Whited and Wu (2006) index which for a firm-year is computed as WW Index = 0.091  CF  0.062  DIVPOS + 0.021  TLTD  0.044  LNTA  0.035  SG + 0.102  ISG, where CF is the ratio of cash flow to book assets, DIVPOS is an indicator that equals one if the firm pays dividends, and zero otherwise, TLTD is the ratio of long-term debt to total assets, LNTA is the natural log of total assets, SG is own-firm sales growth computed as Sales(t)/Sales(t  1), and ISG is the firm’s three-digit industry sales growth. Vega is the change in the value of the CEO’s option grant in a year and any accumulated option holdings for a 0.01 change in the annualized standard deviation of stock returns, scaled by total current compensation. Following Liu and Mauer (2011), Vega is computed using the dividend-adjusted Black–Scholes model and total current compensation in a year includes salary, bonus, restricted stock and option grants, long-term incentive payouts, and any other compensation. Log firm size is the natural logarithm of net assets. Market-to-book ratio is computed as the book value of net assets minus the book value of equity plus the market value of equity, all divided by the book value of net assets. Cash flow/net assets is the ratio of earnings after interest, dividends and taxes but before depreciation divided by the book value of net assets. NWC/net assets is the net working capital-to-net assets ratio. CAPEX/net assets is the ratio of capital expenditures to the book value of net assets. Leverage is the sum of long-term debt and debt in current liabilities divided by the

89

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

book value of net assets. Industry sigma is the mean of the standard deviations of cash flow/net assets over 10 years for firms in the same industry, as defined by two-digit SIC codes. Dividend is a dummy variable equal to one in years in which a firm pays a common dividend, and zero otherwise. R&D/sales is the ratio of research and development expense to sales. This ratio is set equal to zero when research and development expense is missing. Acquisition activity is measured by the ratio of expenditures on acquisitions to the book value of net assets. Liquidity is a dummy variable which is equal to one if the firm has negative operating cash flow (i.e., earnings after interest, dividends, and taxes but before depreciation), and zero otherwise. Tax status is a dummy variable equal to one if a firm has net operating loss carry-forwards, and zero otherwise. The cash value regression variables are defined as follows. Excess stock return, r it  RBit , is the stock return of firm i during fiscal year t, rit, minus stock i’s benchmark return in year t, RBit , where the benchmark return is the return of the Fama and French size and book-to-market portfolio to which stock i belongs at the beginning of fiscal year t. The following variables are scaled by the lagged market value of equity, Mt1, and the one-year change is computed over t  1 to t. DCt is the one-year change in cash plus marketable securities, Ct1 is cash plus marketable securities at the end of fiscal year t  1. DEt is the one-year change in earnings before extraordinary items plus interest, deferred tax credits, and investment tax credits. DNAt is the one-year change in total assets minus cash holdings. DRDt is the one-year change in research and development expense (which is set equal to zero if research and development is missing in either or both years). DIt is the one-year change in interest expense. DDt is the one-year change in common dividends. Lt is the ratio of long-term debt plus debt in current liabilities to the market value of assets at time t, where the market value of assets is computed as the book value of assets plus the difference between the market and book values of equity. DNFt is the one-year change in total equity issuances minus repurchases plus debt issuances minus debt redemption. All continuous variables are winsorized at the 1st and 99th percentiles. All dollar values are CPI-adjusted to 2011 constant dollars.

Table 2 Correlations between cash, CEO inside debt, compensation incentives, and firm characteristics.

Panel A. Cash holding variables Cash Inside debt Pension Deferred compensation Relative debt-equity ratio Relative debt-equity ratio > 1 Firm age CEO age CEO tenure G-Index Democracy (G-Index 6 5) Dictator (G-Index P 14) Median director stock CEO-Chair duality SmallSales SmallNA HighMB LowPayout NoRating WW Index Vega Log firm size Market-to-book Cash flow/net assets NWC/net assets CAPEX/net assets Leverage Industry sigma Dividend R&D/sales Acquisition activity Liquidity Tax status r it  RBit

Cash

Inside debt

Pension

Deferred compensation

Relative D–E ratio

Relative D–E ratio > 1

1.000 0.207*** 0.177*** 0.132*** 0.004 0.063*** 0.185*** 0.055*** 0.045*** 0.142*** 0.071*** 0.046*** 0.116*** 0.098*** 0.279*** 0.309*** 0.281*** 0.229*** 0.310*** 0.187*** 0.060*** 0.451*** 0.504*** 0.067*** 0.218*** 0.021* 0.047*** 0.213*** 0.229*** 0.562*** 0.043*** 0.211*** 0.012

1.000 0.829*** 0.608*** 0.563*** 0.632*** 0.367*** 0.184*** 0.052*** 0.211*** 0.091*** 0.072*** 0.147*** 0.112*** 0.255*** 0.256*** 0.184*** 0.316*** 0.284*** 0.107*** 0.077*** 0.304*** 0.230*** 0.054*** 0.059*** 0.103*** 0.098*** 0.084*** 0.316*** 0.192*** 0.055*** 0.056*** 0.008

1.000 0.071*** 0.441*** 0.508*** 0.328*** 0.164*** 0.063*** 0.182*** 0.086*** 0.054*** 0.144*** 0.109*** 0.227*** 0.234*** 0.180*** 0.274*** 0.256*** 0.093*** 0.054*** 0.270*** 0.197*** 0.051*** 0.032** 0.104*** 0.092*** 0.064*** 0.274*** 0.159*** 0.049*** 0.061*** 0.015

1.000 0.385*** 0.422*** 0.204*** 0.098*** 0.005 0.118*** 0.048*** 0.059*** 0.056*** 0.046*** 0.149*** 0.142*** 0.070*** 0.193*** 0.158*** 0.064*** 0.054*** 0.183*** 0.135*** 0.018 0.060*** 0.040*** 0.047*** 0.057*** 0.193*** 0.123*** 0.029** 0.027** 0.038***

1.000 0.751*** 0.220*** 0.138*** 0.031** 0.117*** 0.028* 0.060*** 0.002 0.048*** 0.101*** 0.045*** 0.117*** 0.238*** 0.018*** 0.040*** 0.015 0.093*** 0.105*** 0.097*** 0.133*** 0.043*** 0.308*** 0.010 0.238*** 0.045*** 0.074*** 0.053*** 0.040***

1.000 0.271*** 0.142*** 0.043*** 0.154*** 0.055*** 0.082*** 0.020 0.070*** 0.164*** 0.115*** 0.102*** 0.298*** 0.088*** 0.059*** 0.016 0.161*** 0.061*** 0.079*** 0.131*** 0.057*** 0.247*** 0.018 0.298*** 0.074*** 0.076*** 0.086*** 0.028***

DCt

Ct1

DE t

DNAt

DRDt

DIt

DDt

Lt

NFt

1.000 0.178*** 0.150*** 0.153*** 0.043*** 0.019 0.033** 0.008 0.202***

1.000 0.246*** 0.185*** 0.208*** 0.015 0.057*** 0.017 0.064***

1.000 0.129*** 0.162*** 0.057*** 0.135*** 0.005 0.092***

1.000 0.250*** 0.296*** 0.168*** 0.161*** 0.410***

1.000 0.031** 0.049*** 0.047*** 0.080***

1.000 0.003 0.097*** 0.356***

1.000 0.109*** 0.036***

1.000 0.102***

1.000

Panel B. Cash value variables 1.000 r it  RBit

DCt Ct1 DE t DNAt DRDt DIt DD t Lt NFt

0.254*** 0.189*** 0.393*** 0.108*** 0.080*** 0.071*** 0.046*** 0.095*** 0.051***

All variables are defined in Table 1. *** Significance at 1% level. ** Significance at 5% level. * Significance at 10% level.

zero when research and development expense is missing. Acquisition activity is measured by the ratio of expenditures on acquisitions to the book value of net assets. Liquidity is a dummy variable which

is equal to one if the firm has negative operating cash flow (i.e., earnings after interest, dividend, and taxes but before depreciation), and zero otherwise. Tax status is a dummy variable equal to one if a firm

90

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

Table 3 Regressions of Inside Debt on Governance and Controls. Independent variables

G-Index

Dependent variable Inside debt (1)

Pension (2)

Def. comp. (3)

0.006*** (5.16)

0.004*** (3.81)

0.003*** (3.54)

CEO tenure CEO age Log firm size Leverage R&D/sales Liquidity Tax status Firm age Industry dummies Year dummies Observations Adj. R2

Def. comp. (6)

Inside debt (7)

Pension (8)

Def. comp. (9)

0.019*** (2.88) 0.011 (0.94) 0.020*** (8.27) 0.019*** (2.77) 0.002*** (5.53) 0.003*** (7.99) 0.013*** (6.93)

0.006 (0.95) 0.014* (1.91) 0.005*** (2.88) 0.001 (0.26) 0.000 (0.30) 0.001* (1.93) 0.003** (2.16)

0.020*** (2.97) 0.012* (1.65) 0.026*** (8.51) 0.019** (2.28) 0.002*** (3.65) 0.004*** (7.73) 0.015*** (5.97)

0.019*** (3.40) 0.003 (0.40) 0.020*** (8.36) 0.020*** (2.82) 0.002*** (5.09) 0.003*** (7.85) 0.013*** (6.79)

0.001 (0.30) 0.009** (2.13) 0.005*** (2.81) 0.001 (0.29) 0.000 (0.41) 0.001* (1.88) 0.003** (2.03)

0.004 (0.34) 0.123*** (3.95) 0.002 (0.22) 0.014*** (2.93) 0.002*** (8.94) Yes Yes 3094 0.27

0.011 (1.44) 0.120*** (4.49) 0.016** (2.02) 0.004 (1.05) 0.001*** (7.43) Yes Yes 3094 0.13

0.012 (0.94) 0.247*** (5.99) 0.017 (1.31) 0.012** (2.00) 0.004*** (11.41) Yes Yes 3094 0.30

0.003 (0.28) 0.121*** (3.89) 0.002 (0.21) 0.015*** (3.07) 0.002*** (8.82) Yes Yes 3094 0.27

0.010 (1.40) 0.120*** (4.48) 0.016** (2.11) 0.003 (0.96) 0.001*** (7.21) Yes Yes 3094 0.13

0.027*** (8.62) 0.018** (2.18) 0.002*** (3.52) 0.004*** (7.67) 0.015*** (6.16)

0.021*** (8.38) 0.019*** (2.74) 0.002*** (5.09) 0.003*** (7.85) 0.013*** (6.95)

0.006*** (2.97) 0.002 (0.35) 0.000 (0.62) 0.001* (1.80) 0.003** (2.12)

0.011 (0.89) 0.244*** (5.94) 0.017 (1.33) 0.011** (1.98) 0.004*** (11.07) Yes Yes 3094 0.30

0.003 (0.31) 0.120*** (3.88) 0.002 (0.15) 0.014*** (3.01) 0.002*** (8.52) Yes Yes 3094 0.27

0.009 (1.26) 0.117*** (4.39) 0.016** (2.05) 0.003 (0.91) 0.001*** (6.88) Yes Yes 3094 0.13

0.013 (1.05) 0.251*** (6.06) 0.016 (1.25) 0.010* (1.80) 0.004*** (11.77) Yes Yes 3094 0.30

Dictator

CEO-Chair duality

Pension (5)

0.023** (2.11) 0.023* (1.88) 0.026*** (8.43) 0.019** (2.27) 0.002*** (4.06) 0.004*** (7.86) 0.015*** (6.14)

Democracy

Log median value of director stockholding

Inside debt (4)

The table reports pooled time-series cross-sectional OLS regressions of inside debt and its components on governance variables and controls. The dependent variables are CEO inside debt (i.e., pension plus deferred compensation), pension, and deferred compensation scaled by the sum of CEO inside debt and the value of the CEO’s equity and option holdings, respectively. The governance variables common to all models include the logarithm of the median dollar value of director stock ownership in the firm and a dummy variable equal to one when the CEO is also the chair of the board of directors and zero otherwise. In addition, Models (1)–(3) include the G-Index, Models (4)–(6) separate the G-Index into Democracy (dummy equal to one if G-Index 6 5, and zero otherwise) and Dictator (dummy variable equal to one if G-Index P 14, and zero otherwise), and Models (7)–(9) redefine Democracy and Dictator according to sample quartiles of the G-Index (i.e. Democracy equals one if G-Index 6 25th and Dictator equals one if G-Index P 75th percentile). All dependent and independent variables are defined in Table 1. Industry dummies are based on Fama–French 48 industry classifications. We report t-statistics in parentheses below parameter estimates. The t-statistics are computed using robust standard errors. *** Significance at 1% level. ** Significance at 5% level. * Significance at 10% level.

has net operating loss carry-forwards, and zero otherwise. Finally, all continuous variables are winsorized at the 1st and 99th percentiles and all dollar values are CPI-adjusted to 2011 constant dollars. As seen in Table 1, average cash balances over the period from 2006 to 2011 are 26% of net assets, although the median firm’s cash reserves are smaller at 11% of net assets. CEO debt compensation appears to be relatively small. The average (median) ratio of CEO inside debt to the sum of CEO inside debt and total CEO equity wealth is 12% (3%), with the average (median) ratios for pensions and deferred compensation 7% (0%) and 5% (0.3%), respectively. The average CEO-firm relative debt-to-equity ratio is 1.1 with 27% of the firm-year observations above one. Contrary to the intuition in Jensen and Meckling (1976) that the relative debt-to-equity ratio should be one to balance managerial incentives between equityholders and debtholders, Edmans and Liu (2011) argue that the optimal debt ratio for the CEO should be less than the firm’s, because relatively more equity compensation is needed to induce CEO effort. Consistent with the Edmans and Liu logic, the majority of firm-year observations have a relative debt-to-equity ratio below one and the median is only 0.2. Table 2 reports Pearson correlation coefficients between the variables. Interestingly, the correlations between cash, inside debt, pension, and deferred compensation are negative. Also note that the correlations between inside debt, pension, and deferred compensation and the bad governance variables G-Index, Dictator,

and CEO-Chair duality are positive which suggests that powerful entrenched CEOs may influence their compensation toward inside debt. These relations are consistent with the spending hypothesis which predicts that CEOs with relatively high debt-like compensation in poorly governed firms tend to spend cash reserves. Of course, this may be a premature conclusion since cash, inside debt, and the governance variables are highly correlated with many other firm characteristics. For example, note that CEOs of larger companies have high relative amounts of inside debt and these companies also tend to have poor governance characteristics and lower cash balances.21 Two additional correlations are noteworthy. First, there is a positive correlation between cash and vega. Liu and Mauer (2011) find a similar result and argue that bondholders anticipate that CEOs with high vega compensation are incentivized to pursue riskier policies and therefore require larger cash balances. Second, as mentioned earlier, note that the instruments Firm age, CEO age, and CEO tenure are all highly correlated with inside debt and its components. Note too, however, that all three are correlated with cash balances (i.e., the variable Cash). Of course, this is expected.

21 The correlations between log firm size and the G-Index, Dictator, and CEO-Chair duality are 0.15, 0.05, and 0.14, respectively, and all are statistically significant. To preserve space, these correlations are not reported in Table 2 and are available upon request.

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

The key is that the instruments should be correlated with cash balances only through their association with inside debt. We test this below. 4. Results We first report regressions that examine the determinants of CEO inside debt and its pension and deferred compensation components. We then report regressions of corporate cash holdings on CEO inside debt and equity incentives, and examine the effects of leverage groups and financial constraints on the cash-inside debt relation. Lastly, we examine the impact of inside debt on the marginal value of cash. 4.1. The determinants of inside debt The value of a CEO’s pension and deferred compensation depends on the firm’s ability to make future payments and should the firm get into financial distress, the CEO’s claims to his/her pension and deferred compensation will be like that of any other unsecured creditor of the firm. These features of inside debt serve to align managers’ interests with those of bondholders, and as originally argued by Jensen and Meckling (1976) and developed by Edmans and Liu (2011), may help mitigate stockholder–bondholder conflicts. This line of reasoning suggests that the use of inside debt to compensate CEOs should be correlated with the likelihood of stockholder–bondholder conflicts. Thus from an optimal contracting perspective, firms may mitigate stockholder–bondholder conflicts by tilting CEO compensation toward inside debt and dialing down the risk-seeking behavior induced by equity-based compensation such as stock options. Furthermore, if CEOs have a preference for debt-like compensation, then this would encourage more powerful CEOs in companies with weak shareholder rights to influence board of director compensation committees to award pension benefits and establish deferred compensation plans.22 As noted above, we use a variety of proxies for CEO power and corporate governance including the G-index, Dictator and Democracy variables constructed from the G-index, whether the CEO is also the chairman of the board of directors (CEO-Chair duality), and the median dollar value of director shareholding in the firm. We predict that a larger G-index (i.e., more anti-takeover amendments), a dictator (i.e., very high G-index), and a CEO who is also the board chair will have larger relative inside debt. In contrast, more director skin in the game (i.e., larger median director shareholdings) and a more democratic environment (i.e., few anti-takeover amendments) should be associated with lower inside debt.23 Table 3 examines the determinants of CEO inside debt and its pension and deferred compensation components. The dependent variables in Models {(1), (4), (7)}, {(2), (5), (8)}, and {(3), (6), (9)}, respectively, are CEO inside debt (i.e., pension plus deferred compensation), pension, and deferred compensation scaled by the sum of CEO inside debt and the value of the CEO’s equity and option holdings. The common variables in all regression models are based on those used by Sundaram and Yermack (2007) and Cen (2010). These include CEO tenure, CEO age, Log firm size, Leverage, 22 Of course, it is not clear whether CEOs will have a preference for debt-like compensation over equity-like compensation (i.e., stock options and restricted stock); especially given the well-established substitution between salary and pension benefits (see, e.g., Gerakos (2010) and the references cited therein) or if the firm’s match to deferred compensation is low or none at all. A more powerful CEO, however, may be able to influence his/her compensation and mitigate undesirable aspects of pensions and deferred compensation. 23 We also use board size, fraction of independent board members, and various measures of institutional share ownership as proxies for CEO power and corporate governance. We find no reliable relation, however, between CEO inside debt and any of these variables. These results are available upon request.

91

R&D/sales, Liquidity, Tax status, Firm age, industry dummies based on Fama–French 48 industry classifications, and year dummies. In all regressions we also include log median value of director stockholdings and CEO-Chair duality. Models (1)–(3) include the G-Index, Models (4)–(6) separate the G-Index into Democracy (dummy equal to one if G-Index 6 5, and zero otherwise) and Dictator (dummy variable equal to one if G-Index P 14, and zero otherwise), and Models (7)–(9) redefine Democracy and Dictator according to sample quartiles of the G-Index (i.e. Democracy equals one if G-Index 6 25th and Dictator equals one if G-Index P 75th percentile). Finally, the t-statistics reported in parentheses below parameter estimates in Table 3 and all subsequent regression tables in the paper are computed using robust standard errors. As expected, the coefficients on G-Index are significantly positive and the coefficients on Log median director ownership are significantly negative. For example, in Model (1) a one-standard deviation increase in G-Index increases CEO inside debt for the mean firm by 13% and a one standard deviation increase in the median director ownership decreases CEO inside debt for the mean firm by 26%. Interestingly, the most obvious measure of CEO power to influence inside debt compensation, CEO-Chair duality, is significantly positive only in the inside debt and pension regressions. CEO-Chair duality has no influence on deferred compensation. This might be explained by the fact that a CEO must actively defer his/her compensation to participate in a deferred compensation plan. Note also that separation of G-Index into democracy and dictator groups (i.e., the Democracy and Dictator dummy variables in Models (4)–(9)) shows that CEOs in democracies (dictatorships) have significantly less (more) inside debt than do CEOs in firms not classified as democracies or dictatorships. Overall, the regressions in Table 3 provide evidence that CEO power and corporate governance have a significant influence on CEO inside debt. With a couple of exceptions, the results for the other variables are similar to those reported in Sundaram and Yermack (2007) and Cen (2010). In particular, we see in Table 3 that older CEOs in larger firms with few growth opportunities (as measured by R&D to sales) have relatively more inside debt and pension and deferred compensation components. Older firms also compensate their CEOs with more inside debt. Where we differ from Sundaram and Yermack (2007) and Cen (2010) is that we do not find that firm leverage is related to inside debt and we find that CEO tenure has a negative rather than a positive influence on inside debt.24 The reason for the latter result is that we scale inside debt by the sum of CEO inside debt and CEO equity wealth in the firm. As CEO tenure increases, equity wealth tends to dominate the ratio which generates the negative relation between our measures of CEO inside debt and CEO tenure.25

24 Sundaram and Yermack (2007) find a positive relation between CEO pension and firm leverage and Cen (2010) finds that CEO pension is first increasing and then decreasing in firm leverage. Cen finds no relation between CEO deferred compensation and firm leverage. Two differences between the sample in Sundaram and Yermack and our paper that might explain the difference in result are that their sample is composed only of large Fortune 500 companies and their data runs from 1996 to 2002. In contrast, we use the ExecuComp database which is a stratified sample of large-, medium-, and small-sized companies and our sample runs from 2006 to 2011. Similar to our paper, Cen uses the ExecuComp database but his sample runs from 2006 to 2008. When we estimate our inside debt regressions over that shorter time frame we find Cen’s result but the relation between inside debt and leverage is still fairly weak. 25 Sundaram and Yermack (2007) find a consistent positive relation between CEO pension and CEO tenure regardless of whether CEO pension is not scaled, scaled by salary plus bonus, or scaled by the value of stock plus options. In contrast, Cen (2010) finds no relation between CEO inside debt (pension plus deferred compensation) when inside debt is scaled by the value of the CEO’s stock plus option holdings.

92

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

4.2. Cash holdings and CEO debt compensation incentives Table 4 reports regressions of cash holdings on CEO inside debt and a common set of control variables which follow those used in the cash regressions of Bates et al. (2009). Each regression in the table also includes the vega of the CEO’s compensation to control for compensation incentives other than inside debt. Models (1)–(3) are pooled time-series cross-sectional regressions of cash holdings in year t on year t inside debt (i.e., pension plus deferred compensation), pension, and deferred compensation, Models (4)–(6) are corresponding lagged pooled regressions of cash holdings in year t on year t  1 inside debt variables, Models (7)–(9) are corresponding firm fixed effects regressions of cash holdings in year t on year t  1 inside debt variables, and Models (10)–(12) are twostage least squares (2SLS) regressions of cash holdings in year t on year t predicted inside debt variables. The predicted values of CEO inside debt, pension, and deferred compensation are computed from first stage regressions of these variables on all of the independent variables plus the instruments firm age, CEO age, and CEO tenure. The control variables in Models (1)–(3) and Models (10)–(12) are measured in year t and the control variables in Models (4)–(6) and Models (7)–(9) are measured in year t  1. The lagged specifications in Models (4)–(9) are intended to correct for possible simultaneity bias between cash holdings and the inside debt variables, while the firm fixed effects specifications in Models (7)–(9) are intended to correct for possible omitted variables bias provided that firms differ significantly in any omitted variables. The 2SLS regressions in Models (10)–(12) address concerns about endogeneity of the inside debt variables. Note that for the 2SLS regressions we report tests for whether the instruments – firm age, CEO age, and CEO tenure – are exogenous (i.e., uncorrelated with the error term in the second stage regression) and whether the instruments are relevant (i.e., correlated with the endogenous regressors inside debt, pension, and deferred compensation). Since we have three instruments and one endogenous regressor in each 2SLS regression, we use the Sargan test for overidentifying restrictions to assess whether the instruments are uncorrelated with the second-stage error. If the test statistic – which is distributed chi-square – exceeds the critical value we reject the null hypothesis that the instruments are uncorrelated with the structural error and conclude that at least some of the instruments are not exogenous. As reported for Models (10)–(12), none of the p-values for the Sargan test warrant rejection of the null hypothesis that the instruments are exogenous. We use the Cragg–Donald statistic to assess whether the instruments are weak. When there is one endogenous regressor as in our 2SLS models, this statistic has an F distribution under the null hypothesis that the instruments have no explanatory power in the first stage regression. As reported in Table 4, the significant F-statistics for Models (10)–(12) easily reject the null hypothesis of weak instruments. Finally, note in Models (4)–(9) that by lagging right-hand-side variables we lose more than 20% of our firm-year observations given the short time-series of reported CEO inside debt. Also note that the firm fixed effects specifications in Models (7)–(9) are implemented by including firm-specific intercepts. Since this specification is equivalently (and more efficiently) estimated by subtracting the time-series mean for each variable from each observation, it is clear that although the fixed effects estimator preserves time-series dispersion in the sample it sweeps out much of the cross-sectional variation.26 For these reasons and like the 2SLS specification, we view the lagged and firm fixed effects specifi26 For example, see Barclay and Smith (1995), Stohs and Mauer (1996), Barclay et al., 2003, and Billett et al., 2007 for discussions of this limitation of the firm fixed effects estimator.

cations as providing robustness tests of the pooled time-series crosssectional specification which we use in subsequent analysis. Focusing first on overall inside debt, we see in Models (1), (4), (7), and (10) that corporate cash holdings are significantly increasing in CEO inside debt. The effect is economically significant. For example, in Model (1) a one-standard deviation increase in inside debt increases cash holdings by 5.1% for the mean firm in the sample. Partitioning inside debt into pension (Models (2), (5), (8), and (11)) and deferred compensation (Models (3), (6), (9), and (12)), we see that the positive relation between cash and inside debt is present for both debt-like compensation components. Thus, for example, in Models (2) and (3) a one standard deviation increase in pension and deferred compensation increases cash balances for the mean firm by 2.8% and 3.5%, respectively. Overall, the positive relation between cash and CEO inside debt is consistent with the managerial risk-aversion hypothesis and inconsistent with the spending and financial contracting hypotheses. The signs of the coefficients on the other variables in the regressions are generally consistent with the results in the literature. We find, for example, that larger firms (Log size) with more investment activity (CAPEX/net assets and Acquisition activity) have smaller cash holdings while firms with larger growth opportunities (Market-to-book ratio and R&D/sales) have larger cash holdings. These findings are consistent with results in for example, Kim et al. (1998), Opler et al. (1999), Bates et al. (2009), and Liu and Mauer (2011). However, we do not find any evidence that cash holdings are related to firm cash flow (Cash flow/net assets) which appears to be inconsistent with Almeida et al. (2004) who find that the change in cash holdings is positively related to cash flow in financially constrained firms. Our finding of no relation between cash and cash flow in the regressions of Table 4 is because these regressions do not condition on financial constraints. When we group our sample into financially constrained and unconstrained subsamples, we find a strong positive relation between cash holdings and cash flow for financially constrained firms.27 We also estimate cash regressions that interact CEO inside debt and its pension and deferred compensation components with corporate governance variables. The purpose of these regressions is to see whether CEO power – as for example reflected in whether the CEO is also the chair of the board of directors – and corporate governance in general influence the relation between cash holdings and CEO inside debt. The coefficients on these interaction variables are never significantly different from zero. Finally, we estimate regressions before and after the financial crisis of 2007–2008 to see whether the crisis has any influence on the results. Although we lose power when the sample is broken up into subsamples, we find no noteworthy differences in results when the sample is grouped into sub-periods 2006–2008 and 2009–2011 or 2006– 2007, 2008–2009, and 2010–2011.28 Table 5 repeats the cash-inside debt regressions using the CEOfirm relative debt-to-equity ratio instead of the ratio of CEO inside debt scaled by CEO inside debt plus total CEO equity wealth in the firm. Recall that the CEO-firm relative debt-to-equity ratio is the ratio of CEO inside debt to total CEO equity wealth in the firm divided by the firm’s debt-to-equity ratio.29 Model (1) uses the relative debt-to-equity ratio, Model (2) uses the log of the ratio, and Model (3) uses a dummy variable which is equal to one when the relative debt-to-equity ratio is greater than one and zero otherwise. As seen in the table, all of the coefficient estimates are significantly

27

These results are available upon request. The interaction regressions and the sub-period regressions are available upon request. 29 Note that the regression sample size in Table 5 is smaller than the 6,009 firm-year observations in the sample because we exclude zero leverage firms when computing the CEO-firm relative debt-to-equity ratio. 28

Table 4 Regressions of Cash Holdings on Inside Debt and its Components. Independent variables

OLS (1)

Inside debt

Lagged OLS (2)

0.080*** (3.90)

Market-to-book ratio Cash flow/net assets

CAPEX/net assets Leverage Industry sigma Dividend R&D/sales Acquisition activity Industry dummies Year dummies Instrument Exogeneity: Sargan test p-value Instrument Relevance: Cragg–Donald statistic Observations Adj R2

(6)

0.027 (0.22) 0.081*** (19.04) 0.106*** (15.49) 0.006 (0.07) 0.426*** (8.93) 0.118 (1.01) 0.050 (1.50) 0.084 (1.10) 0.005 (0.63) 1.502*** (11.92) 0.458*** (7.93) Yes Yes

(7)

0.246* (1.71) 0.067*** (14.08) 0.086*** (11.34) 0.090 (1.03) 0.317*** (5.66) 0.254** (1.99) 0.035 (0.95) 0.040 (0.51) 0.020** (2.12) 1.569*** (9.98) 0.401*** (6.19) Yes Yes

2SLS (9)

0.097*** (3.03)

0.233 (1.62) 0.067*** (14.02) 0.086*** (11.28) 0.092 (1.05) 0.315*** (5.63) 0.257** (2.01) 0.034 (0.92) 0.038 (0.48) 0.018* (1.88) 1.565*** (9.97) 0.405*** (6.24) Yes Yes

(10)

(11)

0.080* (1.86) 0.238* (1.66) 0.067*** (14.02) 0.086*** (11.34) 0.092 (1.05) 0.317*** (5.68) 0.261** (2.04) 0.034 (0.91) 0.041 (0.52) 0.018** (1.96) 1.566*** (9.98) 0.406*** (6.27) Yes Yes

0.051 (0.37) 0.124*** (4.81) 0.009 (0.89) 0.014 (0.14) 0.095 (1.21) 0.422** (2.53) 0.092 (1.29) 0.012 (0.15) 0.015 (0.84) 0.284 (0.71) 0.129** (2.33) – –

0.045 (0.32) 0.124*** (4.82) 0.007 (0.71) 0.013 (0.13) 0.093 (1.19) 0.410** (2.46) 0.096 (1.35) 0.012 (0.16) 0.017 (0.94) 0.283 (0.70) 0.132** (2.39) – –

0.812*** (4.94) 0.176** (2.28) 0.053 (0.38) 0.125*** (4.84) 0.008 (0.83) 0.014 (0.14) 0.092 (1.18) 0.426** (2.55) 0.094 (1.32) 0.009 (0.12) 0.016 (0.88) 0.285 (0.71) 0.132** (2.40) – –

0.187 (1.37) 0.092*** (17.05) 0.114*** (16.05) 0.010 (0.12) 0.418*** (8.41) 0.010 (0.09) 0.046 (1.32) 0.093 (1.09) 0.038*** (3.65) 1.533*** (11.77) 0.390*** (6.27) Yes Yes

0.139 (1.04) 0.091*** (16.98) 0.113*** (15.87) 0.008 (0.09) 0.401*** (8.09) 0.022 (0.18) 0.047 (1.35) 0.077 (0.89) 0.031*** (3.09) 1.520*** (11.66) 0.390*** (6.26) Yes Yes

2.179*** (4.89) 0.277* (1.77) 0.096*** (15.63) 0.115*** (15.74) 0.008 (0.09) 0.462*** (8.71) 0.014 (0.11) 0.046 (1.27) 0.137 (1.48) 0.058*** (4.10) 1.552*** (11.66) 0.395*** (5.83) Yes Yes

0.324

0.154

0.863

***

6009 0.54

6009 0.54

6009 0.54

4615 0.49

4615 0.49

4615 0.49

4615 0.82

(12)

0.590*** (5.18) 0.053* (1.95)

0.032 (1.49) 0.106*** (2.71) 0.031 (0.25) 0.081*** (19.03) 0.105*** (15.53) 0.006 (0.08) 0.429*** (8.99) 0.122 (1.05) 0.051 (1.51) 0.087 (1.14) 0.006 (0.69) 1.503*** (11.93) 0.460*** (7.98) Yes Yes

(8)

4615 0.82

4615 0.82

132.62 5839 0.50

101.30 5839 0.49

***

30.18*** 5839 0.36

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

NWC/net assets

0.045 (0.36) 0.082*** (19.10) 0.106*** (15.58) 0.004 (0.04) 0.428*** (8.97) 0.112 (0.97) 0.049 (1.47) 0.086 (1.12) 0.008 (0.96) 1.507*** (11.95) 0.453*** (7.86) Yes Yes

Firm Fixed Effects (5)

0.058** (2.56)

Deferred compensation

Log firm size

(4)

0.056*** (2.94)

Pension

Vega

(3)

The dependent variable is the ratio of cash plus marketable securities to net assets, where net assets is the book value of total assets minus cash plus marketable securities. All independent variables are defined in Table 1. Models (1)–(3) are pooled time-series cross-sectional OLS regressions of cash holdings in year t on year t inside debt (i.e., pension plus deferred compensation), pension, and deferred compensation, respectively. Models (4)–(6) are pooled time-series cross-sectional OLS regressions of cash holdings in year t on year t  1 inside debt, pension, and deferred compensation, respectively. Models (7)–(9) are firm fixed effects regressions of cash holdings in year t on year t  1 inside debt, pension, and deferred compensation, respectively. Models (10)–(12) are regressions of cash holdings in year t on year t predicted CEO inside debt, predicted pension, and predicted deferred compensation, respectively. The predicted values of CEO inside debt, pension, and deferred compensation are computed from first stage regressions of these variables on all of the independent variables plus the instruments firm age, CEO age, and CEO tenure. CEO inside debt, pension, and deferred compensation are scaled by the sum of CEO inside debt and the value of the CEO’s equity and option holdings. The control variables in Models (1)–(3) and Models (10)–(12) are measured in year t and the control variables in Models (4)–(9) are measured in year t  1. Industry dummies are based on Fama–French 48 industry classifications. For the 2SLS regressions, we report tests for whether the instruments are exogenous (i.e., uncorrelated with the error term in the second stage regression) and whether the instruments are relevant (i.e., correlated with the endogenous regressors inside debt, pension, and deferred compensation). Since we have three instruments and one endogenous regressor in each 2SLS regression model, we use the Sargan test for overidentifying restrictions to assess whether the instruments are uncorrelated with the second-stage error. If the test statistic – which is distributed chi-square – exceeds the critical value we reject the null hypothesis that the instruments are uncorrelated with the structural error and conclude that at least some of the instruments are not exogenous. We use the Cragg–Donald statistic to assess whether the instruments are weak. When there is one endogenous regressor as in our 2SLS models, this statistic has an F distribution under the null hypothesis that the instruments have no explanatory power in the first stage regression. We report t-statistics in parentheses below parameter estimates. The t-statistics are computed using robust standard errors. *** Significance at 1% level. ** Significance at 5% level. * Significance at 10% level.

93

94

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

Table 5 Regressions of cash holdings on CEO-firm relative debt ratio. Independent variables Relative debt-equity ratio

(1)

(2)

0.009 (4.79)

0.025*** (5.01)

Log relative debt-equity ratio Relative debt-equity ratio > 1 Vega Log firm size Market-to-book ratio Cash flow/net assets NWC/net assets CAPEX/net assets Leverage Industry sigma Dividend R&D/sales Acquisition activity Industry dummies Year dummies Observations Adj R2

(3)

***

0.033 (0.32) 0.060*** (15.42) 0.089*** (11.88) 0.006 (0.07) 0.403*** (8.22) 0.128 (1.22) 0.184*** (5.06) 0.026 (0.46) 0.009 (1.18) 1.406*** (10.66) 0.389*** (7.31) Yes Yes 5218 0.52

0.033 (0.32) 0.060*** (15.46) 0.088*** (11.80) 0.008 (0.09) 0.404*** (8.26) 0.123 (1.16) 0.189*** (5.18) 0.025 (0.44) 0.010 (1.37) 1.408*** (10.68) 0.385*** (7.24) Yes Yes 5218 0.52

0.025*** (3.40) 0.019 (0.18) 0.060*** (15.31) 0.090*** (12.15) 0.005 (0.05) 0.401*** (8.17) 0.128 (1.21) 0.171*** (4.79) 0.025 (0.44) 0.007 (0.93) 1.405*** (10.63) 0.395*** (7.39) Yes Yes 5218 0.51

The table reports pooled time-series cross-sectional OLS regressions of cash holdings on the CEO-firm relative debt-to-equity ratio and controls. The dependent variable is the ratio of cash plus marketable securities to net assets, where net assets is the book value of total assets minus cash plus marketable securities. The relative debt-equity ratio is the CEO’s debt-to-equity ratio divided by the firm’s debt-to-equity ratio. The computation of the ratio is described in Table 1. All control variables are defined in Table 1 and are the same as those used in Table 4, including industry and year dummies. Model (1) uses the CEO-firm relative debt-to-equity ratio, Model (2) uses the logarithm of one plus the CEO-firm relative debt-to-equity ratio, and Model (3) uses a dummy variable which is equal to one when the CEOfirm relative debt-to-equity ratio is greater than one and zero otherwise. The regression samples exclude all-equity firm-year observations because the CEO-firm relative debt-to-equity ratio cannot be computed. We report t-statistics in parentheses below parameter estimates. The t-statistics are computed using robust standard errors.  Significance at 5% level.  Significance at 10% level. *** Significance at 1% level.

positive. For example, using Model (1) a one standard deviation increase in the relative debt-to-equity ratio increases cash holdings by 6.5% for the mean firm. The economic significance of the dummy variable specification in Model (3) is even greater; a relative debt-toequity ratio above one increases cash balances by 10% for the mean firm. Overall, the results in Tables 4 and 5 are strongly supportive of the risk aversion hypothesis. 4.3. Cash holdings, CEO debt compensation incentives, and leverage groups An implication of the risk-aversion hypothesis is that the positive relation between cash and inside debt should strengthen as firm leverage increases. In particular, we would expect that CEOs with a significant amount of inside debt will increase cash holdings as firm leverage increases to hedge the risk of financial distress and thereby preserve the value of their pension and/or deferred compensation. Table 6 examines the influence of leverage on the cash-inside debt relation by grouping sample firms into zero

leverage and positive leverage groups and interacting inside debt with a zero leverage dummy variable (Models (1)), and by grouping sample firms into zero leverage, low leverage, middle leverage and high leverage groups and interacting inside debt with a zero leverage, low leverage, and middle leverage dummy variable (Model (2)).30 The left-out or baseline group in Model (1) is positive leverage firms and the left-out or baseline group in Model (2) is high leverage firms. The regressions in Panel A of Table 6 show that the influence of leverage on the cash-inside debt relation is nonlinear.31 First note in Model (1) that there is a positive relation between cash holdings and inside debt for both positive leverage and zero leverage firms (i.e., the coefficient on inside debt is positive and the coefficient on the interaction of inside debt and the zero leverage dummy is not different from zero). As noted earlier (see Footnote 8), the positive relation for zero leverage firms may reflect a desire to hedge economic distress as opposed to financial distress. Next note in Model (2) that the coefficient on inside debt is negative and the coefficients on the interactions of inside debt with the leverage group dummies are monotonically increasing. Since the high leverage group is the left-out or baseline group, the results in Model (2) indicate that the influence of leverage on the cash-inside debt relation is nonlinear; positive and increasingly so as leverage increases but eventually turning negative at high leverage. The influence of leverage on the cash-inside debt relation can be clearly seen in Panel B which reports the percentage change in cash for a one standard deviation change in inside debt by leverage group. As seen there, in Model (2) there is a 5.8% increase in cash for zero leverage firms, a 7.7% increase in cash for low leverage firms, a 10.8% increase in cash for middle leverage firms and a 2.1% decrease in cash for high leverage firms. The negative effect of inside debt on cash holdings for high leverage firms has two possible explanations. On the one hand, it could simply reflect the inability of firms with very high leverage to build cash reserves.32 Alternatively, since the CEO’s inside debt holdings are treated as unsecured claims by bankruptcy courts and therefore the CEO is unlikely to recover much of her accumulated pension and/or deferred compensation should the firm enter bankruptcy, when firm leverage is high the CEO has an incentive to pursue risky policies which is consistent with a reduction in cash holdings.33 Overall, regardless of which story explains the negative relation between cash holdings and inside debt at high leverage, the results in Table 6 are generally consistent with the risk-aversion hypothesis in that leverage clearly influences the relation between cash holdings and CEO inside debt. 4.4. Cash holdings, CEO debt compensation incentives, and financial constraints If inside debt makes CEOs more risk averse, and therefore induces them to hold excess cash, they may face constraints in 30 There are 791 zero leverage firm-year observations in the sample, or a little more than 13% of the total sample of 6009 firm-year observations. 31 Although we report results only for overall inside debt (i.e., the sum of pension and deferred compensation), the results are similar for pension and deferred compensation. 32 Although he does not examine the relation between cash policy and inside debt for firms in different leverage groups, Cen (2010) finds a nonlinear relation between firm leverage and CEO inside debt. In particular, he finds an inverted U-shaped relation in that middle leverage firms have higher inside debt than do low- and highleverage firms. He attributes this relation to firm financial distress risk and CEO risk aversion. 33 This change in unsecured creditors’ appetite for risk at high levels of leverage was first demonstrated by Smith (1979) in the case of junior and senior pure discount debt and later examined by Leland (1994) in the case of one homogenous class of coupon debt and Hackbarth and Mauer (2012) in the case of two classes of coupon debt with different priorities in bankruptcy.

95

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100 Table 6 Regressions of cash holdings on inside debt and firm leverage groups. Independent variables Panel A. Regressions of cash holdings on leverage groups Inside debt Inside debt  zero leverage

(1)

(2)

0.085*** (4.91) 0.013 (0.09)

0.032 (1.11) 0.123 (0.84) 0.153*** (3.44) 0.201*** (5.65) 0.053** (2.44) 0.061*** (4.42) 0.068*** (5.97) 0.016 (0.13) 0.074*** (17.76) 0.101*** (14.54) 0.009 (0.11) 0.446*** (9.46) 0.150 (1.31) 0.085 (1.12) 0.009 (1.04) 1.454*** (11.54) 0.440*** (7.71) Yes Yes 6009 0.54

Inside debt  low leverage Inside debt  middle leverage Zero leverage

0.097*** (5.23)

Low leverage Middle leverage Vega Log firm size Market-to-book Cash flow/net assets NWC/net assets CAPEX/net assets Industry sigma Dividend R&D/sales Acquisition activity Industry dummies Year dummies Observations Adj R2 Panel B. Percentage change in cash for a one-standard deviation change in inside debt Zero leverage firms Positive leverage firms Low leverage firms Middle leverage firms High leverage firms

0.006 (0.05) 0.073*** (17.78) 0.101*** (14.62) 0.028 (0.36) 0.449*** (9.44) 0.147 (1.29) 0.074 (0.98) 0.011 (1.37) 1.454*** (11.50) 0.428*** (7.58) Yes Yes 6009 0.54 6.22 5.41

5.79 7.71 10.77 2.06

In Panel A, the dependent variable is the ratio of cash plus marketable securities to net assets, where net assets is the book value of total assets minus cash plus marketable securities. All control variables are defined in Table 1 and are the same as those used in Table 4, including industry and year dummies. Model (1) regresses cash holdings on inside debt (pension plus deferred compensation) and inside debt interacted with zero leverage firms with the baseline (excluded) group positive leverage firms. Models (2) regresses cash holdings on inside debt and inside debt interacted with zero leverage firms, low leverage (first positive leverage tercile) firms, and middle leverage (second positive leverage tercile) firms with the baseline (excluded) group high leverage (third positive leverage tercile) firms. CEO inside debt is scaled by the sum of CEO inside debt and the value of the CEO’s equity and option holdings. The models exclude leverage as a separate control variable. Panel B reports the percentage change in cash for a onestandard deviation increase in inside debt for the leverage groups in regression Models (1) and (2). The t-statistics in parentheses below parameter estimates are computed using robust standard errors. ⁄ Significance at 10% level. *** Significance at 1% level. ** Significance at 5% level.

their ability to do so. One constraining factor would be the financial constraint status of the firm. Unlike CEOs at financially unconstrained firms that can raise capital relatively easily, CEOs in financially constrained firms may face difficulty accumulating excess cash as their inside debt goes up, simply because capital is limited. This argument implies that the positive relation between cash and inside debt should be attenuated by financial constraints. To determine whether financial constraints influence the relation between inside debt and cash holdings, we interact inside debt with variables that proxy for the degree to which a firm is financially constrained. Negative coefficients on these interaction terms suggest that the positive effect of inside debt on cash holdings is smaller for financially constrained firms than for firms that are not financially constrained.

We construct a number of financial constraint proxies suggested by the literature. Gertler and Gilchrist (1994) use firm size as a measure of financial constraint based on the argument that small firms are less likely than large firms to have access to external funds. The importance of firm size as a proxy for financial constraint is emphasized by Hennessy and Whited (2007) who show that firm size generally outperforms any known proxies for financial constraint including many of the constraint indices (e.g., the Kaplan and Zingales (1997) index).34 Lamont et al. (2001) use the market-to-book ratio as a proxy for financial constraint based on 34 As noted above, we use the Whited and Wu index instead of the Kaplan and Zingales index because the Whited and Wu index has been shown by for example, Whited and Wu (2006) and Hennessy and Whited (2007), to be a superior indexbased measure of financial constraint.

96

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

the idea that firms with large amounts of growth options and few assets in place generate less internal funds and have a higher cost of external funds because of information asymmetry over their investment opportunities. We also use low payout as a constraint proxy based on the arguments in Fazzari et al. (1988) that firms with low or no payouts are preserving cash for investment because they have limited access to external finance. Finally, as discussed above we construct a credit rating dummy variable and we use the Whited and Wu index (2006). Table 7 reports cash regressions where we interact inside debt with these various constraint proxies. Models (1) and (2) interact inside debt with the firm size dummy variables SmallSales and SmallNA, respectively. SmallSales (SmallNA) is equal to one if a firm’s sales (net assets) in a sample year is below the median firm’s sales (net assets) in that year, and zero otherwise. Model (3) interacts inside debt with HighMB which is a dummy variable equal to one if the firm’s market-to-book asset ratio in a sample year is above the median firm’s market-to-book asset ratio in that year, and zero otherwise. Model (4) interacts inside debt with LowPayout which is a dummy variable equal to one if the firm’s payout ratio in a sample year is below the median firm’s payout in that year, and zero otherwise. Note that the payout ratio in a given year is computed as the ratio of common dividends plus share repurchases to earnings before extraordinary items plus interest, deferred tax credits, and investment tax credits. Finally Models (5) and (6), respectively, interact inside debt with a dummy variable equal to one if the firm does not have a credit rating in a given sample year (NoRating) and by the Whited and Wu index of financial constraint (WW Index).35 Consistent with the predicted effect of financial constraints, the coefficients on inside debt interacted with the financial constraint proxies are all negative, and except for the coefficients on the interactions with LowPayout and NoRating, statistically significant. The magnitudes of the coefficients on the interaction terms are large enough to completely offset the positive effect of inside debt on cash. For example, using Model (2) the effect of a one standard deviation increase in inside debt for sample firms with below median net assets in a given year is computed as 0.050  0.168 + 0.103  0.168  1 = 0.009, or a decrease in cash balances for the average firm of 3% [(0.009/0.263)  100]. Similar effects can be computed for the other models in Table 7. This indicates that financial constraints attenuate and in some cases completely offset the incentive of CEOs with inside debt to build excess cash reserves.

4.5. Value of cash and CEO debt compensation incentives To examine how stockholders perceive the incentive alignment of managers with bondholders attributable to inside debt, we use the methodology developed in Faulkender and Wang (2006), which focuses on the value of cash to stockholders in regressions in which excess equity returns are regressed on the change in cash and other corporate policy variables.36 We estimate the Faulkender and Wang regression augmented to include CEO debt compensation 35 Except where noted below, all control variables in Table 7 are defined in Table 1 and are the same as those used in Table 4. In Models 1 and 2 we exclude log firm size as a control variable because we use two proxies for size to measure financial constraint, and in Model 3 we exclude market to book as a control variable because we use market to book to proxy for financial constraint. 36 Note that it would not be appropriate to use a total firm value approach to estimate the value of cash as in Pinkowitz and Williamson (2004) and Pinkowitz et al., 2006, because the risk-aversion and spending hypotheses give different predictions for how equityholders and debtholders view the marginal value of cash. A total firm value approach yields a net value of cash, which combines the assessments of equityholders and debtholders.

incentives (i.e., inside debt and its components). For example, in the case of inside debt, we run the following regression:

DC i;t DEi;t DNAi;t DRDi;t þ c2 þ c3 þ c4 M i;t1 M i;t1 Mi;t1 Mi;t1 DIi;t DDi;t C i;t1 NF i;t þ c5 þ c6 þ c7 þ c8 Li;t þ c9 M i;t1 Mi;t1 M i;t1 M i;t1 C i;t1 DC i;t DC i;t þ c10  þ c11 Li;t  þ c12 ðInside debtÞi;t M i;t1 M i;t1 M i;t1 DC i;t þ c13 ðInside debtÞi;t  þ ei;t ; ð2Þ Mi;t1

ri;t  RBi;t ¼ c0 þ c1

where the dependent variable is the difference between firm i’s stock return over year t  1 to year t (ri,t) (computed using monthly returns from CRSP) and the Fama and French (1993) size and bookto-market matched portfolio return from year t  1 to year t   RBi;t .37 For the right-hand-side variables, DXi,t indicates a change in variable X for firm i over year t  1 to year t, where the scaling variable, Mi,t1, is firm i’s market value of equity at time t  1. The righthand-side variables include cash and marketable securities (Ci,t), earnings before extraordinary items (Ei,t), net assets (NAi,t), research and development expense (RDi,t) (set equal to zero if missing), interest expense (Ii,t), common dividends (Di,t), long-term debt plus debt in current liabilities divided by the market value of assets at time t (Li,t), and net new finance (NFi,t).38 Table 1 and Panel B of Table 2 provide descriptive statistics and correlations for these variables. The coefficient on Inside debt (c12) measures the direct effect of Inside debt on excess equity returns, and the coefficient on the interaction of Inside debt with the change in cash (c13) measures the effect of CEO debt compensation incentives on the value of an additional dollar of cash. Since our goal is to examine how the marginal value of cash changes with debt compensation incentives, we focus on the sign of c13. As discussed in Section 2, the risk aversion hypothesis predicts that CEO inside debt has a negative effect on the marginal value of cash (c13 < 0) since CEOs build excess cash reserves for their own self-interest (i.e., to protect their future debt-like compensation payments). The spending hypothesis allows for either a positive or negative effect of CEO inside debt on the value of cash to shareholders. On the one hand, since inside debt coincides with weak shareholder rights as shown in Table 3 and since managers in these firms according to the spending hypothesis suboptimally spend cash, the effect of inside debt on the marginal value of an additional dollar of cash could be positive (c13 > 0). But this assumes that the additional cash will be used for positive net present value projects. Thus, on the other hand, if shareholders perceive that the firm will waste additional cash on negative net present value projects the marginal value of an additional dollar of cash could be negative (c13 < 0). Lastly, recall from the discussion in Section 2 that the financial contracting hypothesis does not have a clear prediction for the effect of inside debt on the marginal value of cash to equityholders. Overall, the risk aversion, spending, and financial contracting hypotheses cannot be distinguished by assessing only the value of cash to shareholders, unless CEO inside debt has a positive effect on the marginal value of cash which would be consistent with the spending hypothesis. Nevertheless, since the 37 Specifically, for each year, we group every firm in our sample into one of 25 size and book-to-market portfolios based on the intersection between size and book-tomarket independent sorts. Thus stock i’s benchmark return in year t is the return to which stock i belongs at the beginning of fiscal year t. Returns on these 25 portfolios are from Kenneth R. French’s website http://mba.tuck.dartmouth.edu/pages/faculty/ ken.french/data_library.html. We thank Professor French for graciously providing these data. 38 Net new finance, NFi,t, is computed as sales of common and preferred stock net of stock repurchases, plus issuance of long-term debt net of long-term debt reduction.

97

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100 Table 7 Regressions of cash holdings on inside debt and financial constraints. Independent variables Inside debt Inside debt  SmallSales

(1) 0.029 (1.52) 0.079* (1.89)

(2)

(3) ***

0.050 (2.81)

(4) ***

0.101 (5.06)

(5) ***

0.092 (3.45)

(6) ***

0.086 (4.80)

0.103** (2.45)

Inside debt  SmallNA

0.164*** (3.36)

Inside debt  HighMB Inside debt  LowPayout

0.031 (0.81)

Inside debt  NoRating

0.033 (0.73) 0.258*** (3.51)

Inside debt  WW Index SmallSales

0.113*** (10.33) 0.139*** (12.97)

SmallNA

0.089*** (7.70)

HighMB LowPayout

0.012 (1.11) 0.034*** (2.80)

NoRating WW Index Independent variables Vega

(1) 0.460*** (3.56)

(2) 0.378*** (2.93)

0.132*** (18.31) 0.159* (1.87) 0.312*** (6.23) 0.070 (0.58) 0.016 (0.46) 0.073 (0.90) 0.053*** (6.42) 1.607*** (11.73) 0.520*** (8.30) Yes Yes 6009 0.49

0.127*** (17.58) 0.159* (1.90) 0.331*** (6.62) 0.096 (0.81) 0.001 (0.03) 0.076 (0.94) 0.052*** (6.41) 1.599*** (11.76) 0.496*** (8.05) Yes Yes 6009 0.50

Log firm size Market-to-book Cash flow/net assets NWC/net assets CAPEX/net assets Leverage Industry sigma Dividend R&D/sales Acquisition activity Year dummies Industry dummies Observations Adj R2

0.017 (0.60)

(3) 0.249* (1.94) 0.099*** (21.39)

0.224*** (2.66) 0.528*** (10.59) 0.069 (0.56) 0.062* (1.77) 0.165** (2.05) 0.022** (2.49) 1.804*** (13.91) 0.517*** (8.44) Yes Yes 6009 0.49

(4) 0.044 (0.35) 0.082*** (19.08) 0.106*** (15.58) 0.004 (0.04) 0.427*** (8.94) 0.114 (0.98) 0.050 (1.49) 0.086 (1.12)

1.504*** (11.93) 0.454*** (7.86) Yes Yes 6009 0.54

(5) 0.058 (0.47) 0.089*** (17.57) 0.106*** (15.51) 0.002 (0.02) 0.424*** (8.89) 0.120 (1.04) 0.032 (0.89) 0.086 (1.13) 0.007 (0.85) 1.517*** (11.99) 0.449*** (7.78) Yes Yes 6009 0.54

0.035 (1.40) (6) 0.043 (0.34) 0.081*** (18.60) 0.106*** (15.56) 0.002 (0.03) 0.427*** (8.95) 0.117 (1.01) 0.048 (1.45) 0.081 (1.07) 0.006 (0.76) 1.512*** (11.99) 0.449*** (7.82) Yes Yes 6009 0.54

The dependent variable is the ratio of cash plus marketable securities to net assets, where net assets is the book value of total assets minus cash plus marketable securities. Inside debt (i.e., pension plus deferred compensation) is scaled by the sum of CEO inside debt and the value of the CEO’s equity and option holdings. SmallSales is a dummy variable equal to one if a firm’s sales in a sample year is below the median firm’s sales in that year, and zero otherwise. SmallNA is a dummy variable equal to one if a firm’s net assets (total assets minus cash plus marketable securities) in a sample year is below the median firm’s net assets in that year, and zero otherwise. HighMB is a dummy variable equal to one if the firm’s market to book asset ratio in a sample year is above the median firm’s market-to-book asset ratio in that year, and zero otherwise. LowPayout is a dummy variable equal to one if the firm’s payout ratio in a sample year is below the median firm’s payout in that year, and zero otherwise. The payout ratio in a given year is computed as the ratio of common dividends plus share repurchases to earnings before extraordinary items plus interest, deferred tax credits, and investment tax credits. NoRating is a dummy variable equal to one if a firm does not have a credit rating in a sample year, and zero otherwise. WW Index is the Whited and Wu (2006) index which for a firm-year is computed as WW Index = 0.091  CF  0.062  DIVPOS + 0.021  TLTD  0.044  LNTA  0.035  SG + 0.102  ISG, where CF is the ratio of cash flow to book assets, DIVPOS is an indicator that equals one if the firm pays dividends, and zero otherwise, TLTD is the ratio of long-term debt to total assets, LNTA is the natural log of total assets, SG is own-firm sales growth computed as Sales(t)/Sales(t  1), and ISG is the firm’s three-digit industry sales growth. Except where noted below, all control variables are defined in Table 1 and are the same as those used in Table 4, including industry and year dummies. In Models (1) and (2) we exclude firm size as a control because we use two proxies for size to measure financial constraints, in Model (3) we exclude market to book as a control because we use market to book to proxy for financial constraint, and in Model (4) we exclude the dummy variable Dividend as a control because we use payouts to proxy for financial constraints. The t-statistics in parentheses below parameter estimates are computed using robust standard errors. *** Significance at 1% level. ** Significance at 5% level. * Significance at 10% level.

98

Table 8 The Impact of CEO Inside Debt on the Value of Cash. Independent variables

Faulkender and Wang base model (1)

Above versus below median inside debt measures

Top tercile versus bottom tercile inside debt measures

Continuous inside debt measures

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

DCt

1.598*** (6.13) 0.100 (0.11) 0.783 (1.32)

1.846*** (5.93) 0.386 (0.41) 0.740 (1.25) 0.887** (2.22)

1.736*** (5.76) 0.280 (0.30) 0.746 (1.26)

1.803*** (5.92) 0.342 (0.37) 0.760 (1.28)

1.831*** (5.14) 0.437 (0.39) 0.638 (0.84) 1.304*** (2.61)

1.744*** (5.72) 0.256 (0.27) 0.704 (1.18)

1.858*** (5.36) 0.515 (0.47) 0.603 (0.88)

1.835*** (6.13) 0.392 (0.42) 0.734 (1.24) 4.743** (2.47)

1.723*** (5.99) 0.266 (0.28) 0.760 (1.28)

1.735*** (6.06) 0.267 (0.29) 0.764 (1.29)

Ct1  DCt Lt  DCt Inside debt  DCt

0.845* (1.78)

Pension  DCt

0.794** (2.04)

Def. comp.  DCt Inside debt

0.001 (0.09)

DRDt DIt DDt Ct1 Lt NFt Intercept Observations Adj R2

0.841*** (7.15) 0.025 (0.43) 0.586 (0.45) 2.023 (1.18) 0.286 (0.22) 0.407*** (3.23) 0.500*** (5.68) 0.217 (1.55) 0.016 (0.66) 6009 0.23

0.024* (1.75) 0.009 (0.78) 0.841*** (7.15) 0.025 (0.45) 0.518 (0.40) 2.003 (1.17) 0.332 (0.25) 0.394*** (3.16) 0.497*** (5.72) 0.217 (1.55) 0.024 (1.03) 6009 0.23

0.906*** (6.62) 0.026 (0.37) 1.698 (1.09) 2.441 (1.12) 1.288 (0.82) 0.444*** (2.83) 0.432*** (3.80) 0.217 (1.22) 0.000 (0.01) 4223 0.23

0.843*** (7.10) 0.017 (0.29) 0.634 (0.48) 1.916 (1.10) 0.352 (0.27) 0.419*** (3.25) 0.494*** (5.48) 0.217 (1.51) 0.014 (0.55) 5761 0.23

0.115** (2.52) 0.000 (0.03) 0.901*** (6.92) 0.006 (0.09) 0.729 (0.51) 2.199 (1.12) 0.834 (0.52) 0.409*** (2.89) 0.447*** (4.37) 0.172 (1.03) 0.013 (0.47) 4827 0.23

0.835*** (7.12) 0.017 (0.29) 0.533 (0.41) 2.051 (1.20) 0.223 (0.17) 0.378*** (3.02) 0.458*** (5.14) 0.219 (1.57) 0.042* (1.83) 6009 0.23

0.838*** (7.14) 0.023 (0.41) 0.540 (0.41) 2.081 (1.21) 0.272 (0.21) 0.386*** (3.07) 0.470*** (5.27) 0.221 (1.58) 0.034 (1.47) 6009 0.23

0.169*** (3.08) 0.840*** (7.12) 0.021 (0.37) 0.513 (0.39) 1.988 (1.16) 0.276 (0.21) 0.381*** (3.05) 0.485*** (5.62) 0.217 (1.56) 0.037 (1.61) 6009 0.23

The dependent variable is excess stock returns, r it  RBit , defined as the stock return of firm i during fiscal year t, rit, minus stock i’s benchmark return in year t, RBit , where the benchmark return is the return of the Fama and French size and book-to-market portfolio to which stock i belongs at the beginning of fiscal year t. All variables except Inside debt, Pension, Deferred compensation, and Leverage are scaled by the lagged market value of equity, Mt1. Model (1) is the benchmark Faulkender and Wang (2006) specification. Models (2)–(4) set inside debt, pension, and deferred compensation equal to one if the respective continuous measures are above their yearly sample medians, and zero otherwise. Models (5)–(7) set inside debt, pension, and deferred compensation equal to one if the respective continuous measures are in the top terciles of their yearly sample distributions, and zero otherwise. In all cases, the benchmark group is the bottom tercile, so all sample observations in the middle tercile are eliminated from these regressions. Models (8)–(10) use the continuous variables inside debt, pension, and deferred compensation, respectively. CEO inside debt (i.e., pension plus deferred compensation), pension, and deferred compensation are scaled by the sum of CEO inside debt and the value of the CEO’s equity and option holdings. In the regressions, Ct is cash plus marketable securities, Et is earnings before extraordinary items plus interest, deferred tax credits, and investment tax credits, NAt is total assets minus cash holdings, RDt is research and development expense (which is set equal to zero if missing), It is interest expense, Dt is common dividends, Lt is the ratio of long-term debt plus debt in current liabilities to the market value of assets at time t, and NFt is total equity issuances minus repurchases plus debt issuances minus debt redemption. DXt is notation for the one-year change, Xt  Xt1, where t (t  1) denotes end of fiscal year t (t  1). The t-statistics in parentheses below parameter estimates are computed using robust standard errors. *** Significance at 1% level. ** Significance at 5% level. * Significance at 10% level.

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

DNAt

0.839*** (7.15) 0.025 (0.45) 0.531 (0.41) 2.054 (1.20) 0.304 (0.23) 0.394*** (3.14) 0.489*** (5.45) 0.223 (1.59) 0.027 (1.16) 6009 0.23

8.717* (1.96) 0.141*** (3.97)

0.007 (0.42)

Def. comp. 0.841*** (7.14) 0.025 (0.44) 0.527 (0.40) 2.020 (1.18) 0.300 (0.23) 0.390*** (3.11) 0.494*** (5.69) 0.217 (1.55) 0.028 (1.21) 6009 0.23

6.174** (2.02) 1.047** (2.25)

0.031** (2.42)

Pension

DEt

0.967* (1.88)

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

relation between cash holdings and CEO inside debt is robustly positive, a negative effect of CEO inside debt on the marginal value of cash would support the risk aversion hypothesis. Table 8 reports a baseline Faulkender and Wang (2006) regression specification without inside debt, Model (1), and a number of specifications like Eq. (2) augmented with inside debt variables interacted with the change in cash, Models (2)–(10). For the augmented specifications, Models (2)–(4) set inside debt, pension, and deferred compensation equal to one if the respective continuous measures are above their yearly sample medians, and zero otherwise; Models (5)–(7) set inside debt, pension, and deferred compensation equal to one if the respective continuous measures are in the top terciles of their yearly sample distributions, and zero otherwise; and Models (8)–(10) use the continuous variables inside debt, pension, and deferred compensation, respectively. Note that all continuous measures of CEO inside debt are scaled by the sum of CEO inside debt and the value of the CEO’s equity and option holdings. Finally, note in Models (5)–(7) that the benchmark group is the bottom tercile, so all sample observation in the middle tercile are eliminated from these regressions.39 From the baseline specification in Model (1), we see that the value of an additional dollar of cash for a firm with zero cash balances and zero leverage is $1.60. The value of an additional dollar of cash for the mean firm with cash holdings of 15.1% of the market capitalization of equity at the beginning of the fiscal year and a market leverage ratio of 15.1% is 1.598  0.100  0.151  0.783  0.151 = $1.46.40 Observe in Models (2)–(10), however, that when the baseline regression specification is augmented to include inside debt and its pension and deferred compensation components all of the coefficients on the interactions of these variables and the change in cash are significantly negative. This indicates that CEO inside debt decreases the marginal value of cash. For example, in Model (2) we find that the value of an additional dollar of cash for the mean firm with below median inside debt is 1.846  0.386  0.151  0.740  0.151 = $1.68, and the value of an additional dollar of cash for the mean firm with above median inside debt is 1.846  0.386  0.151  0.740  0.151  0.887  1 = $0.79.41 Thus moving from below to above median inside debt reduces the value of cash for the mean firm by 53% and the net present value of an additional dollar of cash is now negative (i.e., $0.79  1 = $0.21). Marginal value of cash calculations for pension and deferred compensation components of inside debt using Models (3) and (4) yield similar results, as do the tercile inside debt regressions in Models (5)–(7) and the continuous inside debt regressions in Models (8)–(10). Overall, our evidence on the marginal value of cash in conjunction with the positive relation between cash balances and CEO inside debt supports the risk aversion hypothesis. In unreported regressions, we estimate the marginal value of cash by (i) interacting debt compensation incentives times the change in cash by measures of corporate governance and (ii) estimating the relation

39 Note in Table 8 that less than one-third of the 6,009 firm-year observations are dropped when we delete the middle tercile in Models (5)-(7). The reason is that there are firm-year observations with zero CEO pension and/or deferred compensation and these observations by default are grouped in the bottom tercile. 40 The sample means for Ct-1 and Lt are reported in Table 1. The fact that both are 0.151 is purely coincidental. 41 Although these point estimates are far apart, it is useful to construct confidence intervals around them to see if the bounds overlap. Using the STATA command lincom we compute a 95% confidence intervals for the marginal value of cash for the below and above median inside debt groups of 1.68 ± 1.96  0.26 and 0.79 ± 1.96  0.28, respectively. Since the lower bound of the first estimate (1.17) overlaps the upper bound of the second estimate (1.34) we cannot be statistically certain that the marginal value of cash is different in the two groups. Similar calculations for the marginal value of cash in the other models reported in Table 8 also yield overlapping estimates. These calculations are available upon request. We thank an anonymous referee for suggesting that we compute confidence intervals for our estimates of the marginal value of cash using the STATA command lincom.

99

between debt compensation incentives and the marginal value of cash by grouping firms into samples with good and bad corporate governance.42 We find no evidence that corporate governance influences the relation between the marginal value of cash and debt compensation incentives. Accordingly, we find little evidence in support of the spending hypothesis.

5. Conclusions The impact of CEO pensions and deferred compensation on corporate policy decisions is a relatively new area of compensation research despite the fact that Jensen and Meckling (1976) first speculated over thirty years ago that debt-like compensation could be used to mitigate the excessive risk-taking incentives of equity compensation and the resulting agency conflicts between stockholders and bondholders. The objective of this paper is to rigorously examine the impact of CEO debt-like compensation on firm cash policy. We choose to study cash policy for a number of reasons. First, extensive research in recent years has explored the determinants of cash balances with the result that we are now fairly comfortable with our understanding of the factors that drive cash. Second, methods have been developed in recent years to examine the value of cash and the impact on the value of cash of corporate policy decisions. Third, recent research by, for example, Liu and Mauer (2011) establishes that cash balances are increasing in CEO equity compensation incentives and the marginal value of cash to equityholders is decreasing in equity compensation incentives. The interesting and important question is what role CEO debt-like compensation plays in determining cash policy and the value of cash. Finally, it is now well established that firms appear to be building excessive cash balances and this phenomenon is hard to explain with firm characteristics alone (see, e.g., Bates et al. (2009) and Dittmar and Duchin (2012)). We first examine the determinants of CEO debt compensation incentives (i.e., inside debt and its pension and deferred compensation components). Importantly, we find that CEO debt-like compensation is higher in poorly governed firms, which suggests that CEO debt-like compensation likely causes conflicts between stockholders and bondholders and thereby influences firm cash balances in unexpected ways. We test primarily two hypotheses. The risk aversion hypothesis argues that CEOs with more debt-like compensation will choose to carry higher cash balances to preserve the value of their debt-like compensation. The spending hypothesis, in contrast, predicts that CEOs with high debt-like compensation tend to spend cash. This hypothesis is based on the finding that CEO debt-like compensation is higher in poorly governed firms and the finding in Harford et al. (2008) that CEOs in poorly governed firms spend cash rather than build cash reserves. Overall, our results support the risk aversion hypothesis in that cash balances are higher in firms when CEOs have larger amounts of debt-like compensation. We find that the positive effect of CEO debt-like compensation incentives on cash balances is independent of the positive effect of CEO equity compensation incentives on cash balances. This suggests that the factors that drive both relations are separate. We further find that CEOs with significant amounts of inside debt increase cash balances as firm leverage increases, which suggests that cash is used to hedge the risk of financial distress and the likely loss of debt-like compensation in bankruptcy. Finally, we find that the positive relation between cash and debt-like compensation can be offset by financial constraints which limit the ability of the CEO to build excess cash reserves. Our final set of tests examines the influence of CEO debt-like compensation on the marginal value of cash to equityholders. 42

These results are available upon request.

100

Y. Liu et al. / Journal of Banking & Finance 42 (2014) 83–100

We find that debt-like compensation has a strong negative effect on the marginal value of cash to shareholders. In conjunction with the positive effect of CEO inside debt on cash balances, the negative effect of CEO inside debt on the marginal value of cash supports the risk aversion hypothesis. Acknowledgments We thank Matt Billett, Audra Boone, Jon Garfinkel, Kristine Hankins, Shane Johnson, Dolly King, Wendy Liu, Yiming Qian, Nan Yang, and two anonymous referees for many helpful comments and suggestions. Yixin Liu would like to thank David Greenlaw and the Peter T. Paul Financial Policy Center for support. References Acharya, V., Almeida, H., Campello, M., 2007. Is cash negative debt? Corporate financial policies. Journal of Financial Intermediation 16, 515–554. Acharya, V., Davydenko, S.A., Strebulaev, I., 2012. Cash holdings and credit risk. Review of Financial Studies 25, 3572–3609. Almeida, H., Campello, M., Weisbach, M.S., 2004. The cash flow sensitivity of cash. Journal of Finance 59, 1777–1804. Anantharaman, D., Fang, V.W., Gong, G., 2013. Inside debt and the design of corporate debt contracts. Management Science (in press). Barclay, M.J., Smith, C.W., 1995. The maturity structure of corporate debt. Journal of Finance 50, 609–631. Barclay, M.J., Marx, L.M., Smith, C.W., 2003. The joint determination of leverage and maturity. Journal of Corporate Finance 9, 149–167. Bates, T.W., Kahle, K.M., Stulz, R., 2009. Why do U.S. firms hold so much more cash than they used to? Journal of Finance 64, 1985–2021. Bebchuk, L., Jackson, R., 2005. Executive pensions. Journal of Corporate Law 30, 823–855. Bhagat, S., Bolton, B., 2013. Director ownership, governance, and performance. Journal of Financial and Quantitative Analysis 48, 105–135. Billett, M.T., King, T.-H.D., Mauer, D.C., 2007. Growth opportunities and the choice of leverage, debt maturity, and covenants. Journal of Finance 62, 697–730. Brockman, P., Martin, X., Unlu, E., 2010. Executive compensation and the maturity structure of corporate debt. Journal of Finance 65, 1123–1161. Cassell, C.A., Huang, S.X., Sanchez, J.M., Stuart, M.D., 2012. Seeking safety: the relation between CEO inside debt holdings and the riskiness of firm investment and financial policies. Journal of Financial Economics 103, 588–610. Cen, W., 2010. The Determinants of CEO Inside Debt and Its Components. SSRN Working Paper. Coles, J.L., Daniel, N.D., Naveen, L., 2006. Managerial incentives and risk-taking. Journal of Financial Economics 79, 431–468. Core, J., Guay, W.R., 2002. Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. Journal of Accounting Research 40, 613–630. Denis, D., Sibilkov, V., 2010. Financial constraints, investment, and the value of cash holdings. Review of Financial Studies 23, 247–269. Dittmar, A., Duchin, R., 2012. The Concentration of Cash: Cash Policies of the Richest Firms. SSRN Working Paper.

Dittmar, A., Mahrt-Smith, J., 2007. Corporate governance and the value of cash holdings. Journal of Financial Economics 83, 599–634. Edmans, A., Liu, Q., 2011. Inside debt. Review of Finance 15, 75–102. Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3–56. Faulkender, M., Wang, R., 2006. Corporate financial policy and the value of cash. Journal of Finance 61, 1957–1990. Fazzari, S., Hubbard, R.G., Peterson, B., 1988. Financing constraints and corporate investment. Brookings Papers on Economic Activity 1, 144–195. Gerakos, J., 2010. Chief executive officers and the pay-pension tradeoff. Journal of Pension Economics and Finance 9, 303–319. Gertler, M., Gilchrist, S., 1994. Monetary policy, business cycles and the behavior of small manufacturing firms. Quarterly Journal of Economics 109, 309–340. Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118, 107–155. Hackbarth, D., Mauer, D.C., 2012. Optimal priority structure, capital structure, and investment. Review of Financial Studies 25, 747–796. Harford, J., Mansi, S.A., Maxwell, W.F., 2008. Corporate governance and firm cash holdings in the US. Journal of Financial Economics 87, 535–555. Hennessy, C.A., Whited, T.M., 2007. How costly is external financing? Structural estimation. Journal of Finance 62, 1705–1745. Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: managerial behavior, agency costs, and ownership structure. Journal of Financial Economics 3, 305–360. Kaplan, S., Zingales, L., 1997. Do financing constraints explain why investment is correlated with cash flow? Quarterly Journal of Economics 112, 169–216. Kim, C., Mauer, D.C., Sherman, A.E., 1998. The determinants of corporate liquidity: theory and evidence. Journal of Financial and Quantitative Analysis 33, 305– 334. Lamont, O.A., Polk, C., Saa-Requejo, J., 2001. Financial constraints and stock returns. Review of Financial Studies 14, 529–554. Lee, G., Tang, H., 2011. CEO Pension and Deferred Compensation. Working Paper. Seton Hall University. Leland, H.E., 1994. Corporate debt value, bond covenants, and optimal capital structure. Journal of Finance 49, 1213–1252. Liu, Y., Mauer, D.C., 2011. Corporate cash holdings and CEO compensation incentives. Journal of Financial Economics 102, 183–198. Masulis, R.W., Wang, C., Xie, F., 2007. Corporate governance and acquiring returns. Journal of Finance 62, 1851–1889. Opler, T., Pinkowitz, L., Stulz, R., Williamson, R., 1999. The determinants and implications of cash holdings. Journal of Financial Economics 52, 3–46. Pinkowitz, L., Williamson, R., 2004. What is a Dollar Worth? The Market Value of Cash Holdings. Working Paper. Georgetown University. Pinkowitz, L., Stulz, R., Williamson, R., 2006. Does the contribution of corporate cash holdings and dividends to firm value depend on governance? A cross-country analysis. Journal of Finance 61, 2725–2751. Smith, C.W., 1979. Applications of option pricing analysis. In: Bicksler, J. (Ed.), Handbook of Financial Economics. North Holland, Amsterdam, pp. 79–121. Stohs, M.H., Mauer, D.C., 1996. The determinants of corporate debt maturity structure. Journal of Business 69, 279–312. Sundaram, R., Yermack, D., 2007. Pay me later: inside debt and its role in managerial compensation. Journal of Finance 62, 1551–1588. Wei, C., Yermack, D., 2011. Investor reactions to CEOs’ inside debt incentives. Review of Financial Studies 24, 3813–3840. Whited, T., Wu, G., 2006. Financial constraints risk. Review of Financial Studies 19, 531–559.