DEBT COVENANTS AND RISK TAKING
Huijing Fu
Texas Christian University
Jieying Zhang*
University of Southern California
January 2011
*Corresponding author. Tel: (213) 7408705
Email:
Acknowledgments: We thank Kevin Murphy, Oguzhan Ozbas, K.R. Subramanyam, workshop participants at USC Finance brownbag for helpful comments and suggestions.
Debt Covenants and Risk Taking
ABSTRACT
This paper investigates whether debt covenants protect lenders from borrowers’ risk taking. Using a large sample of private credit agreements, we find that firms with more restrictive covenants spend less on R&D, diversify more, and their CEOs have less risk-taking incentives from option compensation. We use a change specification, a simultaneous regression, and a bond market reaction test to address the endogeneity. In particular, we find a positive bond market reaction to additional covenants, suggesting that covenant restrictions do not simply capture expected lower risk taking. Additional analyses show that (1) covenant violations trigger a further reduction in risk-taking activities; (2) the effect of covenants in mitigating risk taking is not driven by short maturity, and (3) the results from private debt covenants can be replicated with public debt covenants. Overall, our findings provide important evidence that debt covenants are effective in mitigating risk taking, an agency problem that cannot be explicitly contracted on.
Keywords: debt covenants, risk taking, risk-return tradeoff, covenant violation
1. Introduction
Agency theory suggests that debt covenants exist to mitigate debtholder-stockholder conflict by restricting wealth expropriation from borrowers (Jensen and Meckling, 1976; Myers, 1977; Smith and Warner, 1979). One such action involves borrowers undertaking suboptimal risky projects after debt is in place.[1] However, there is no empirical evidence on whether debt covenants are effective in restricting borrowers’ risk taking and thereby protecting lenders from wealth expropriation.[2] In this paper, we fill this void by investigating whether stronger debt covenants lead to lower risk taking.
It is important to study the effectiveness of debt covenants because debt covenants are widely used to reduce the agency cost of debt (Tirole, 2006). While theoretically covenants are designed to protect lenders from wealth expropriation including risk taking, incomplete contracting suggests that covenants cannot explicitly specify the exact projects that a firm may or may not undertake. Therefore it is unclear whether covenants are effective in mitigating risk taking through a combination of indirect restrictions (McDaniel, 1986).
Our primary hypothesis is that when debt covenants are more restrictive, borrowers engage in less risk taking. We expect this negative relation for two reasons. First, restrictive covenants limit borrower’s abilities to engage in risk taking (Smith and Warner, 1979). For example, asset sweep covenant limits borrowers’ ability to replace less risky assets with more risky ones; debt sweep and equity sweep covenant limit borrowers’ ability to finance risky projects from raising additional capital; and secured debt covenants limit borrowers’ ability to substitute the pledged assets for risky ones. Second, restrictive covenants reduce borrowers’ incentives to take risk through the threat of covenant violations. Risk taking increases the odds of covenant violations. For example, risky projects produce more volatile financial performance and financial covenants are more likely to be violated. As a result, borrowers would be less willing to take excess risk to avoid the anticipated cost associated with covenant violations.[3]
However, two factors work against our hypothesis. First, debt contracts are incomplete and covenants can only restrict risk taking indirectly. Because none of the covenants mentioned above can specify which projects the borrower may or may not undertake, these covenants can only restrict risk taking indirectly through reducing borrowers’ incentives and abilities to take additional risk. Second, the threat of covenant violation may not be considered substantial to prevent risk taking. In that case, borrowers would only cut down risk taking after a covenant is violated and the contract is renegotiated, not before. Therefore, it is an open empirical question whether restrictive covenants are associated with lower risk taking in the absence of a control transfer.
We measure the restrictiveness of debt covenants following Bradley and Roberts (2004). Specifically, we define covenant intensity of a private debt contract as the sum of six dummy variables representing the existence of a debt sweep covenant, an equity sweep covenant, an asset sweep covenant, a dividend covenant, at least two financial covenants, and a secure covenant. This measure ranges from zero (the least restrictive) to six (the most restrictive). We measure risk taking using two sets of variables. The first set is risk-taking activities, captured by R&D expenditure, number of segments, and segment Herfindahl index.[4] The second set is risk-taking incentives, captured by the sensitivity of CEO option portfolio to stock return volatility (vega). We use both vega from newly granted options and total vega, which is the sum of vega from newly granted options, vega from unexercisable options, and vega from exercisable options.
Our sample is the intersection of Dealscan and Compustat database. Essentially we include all Dealscan loan contracts to the U.S. public corporations for which the risk taking variables are available. Our risk taking activity sample (hereafter, the full sample) includes 37,456 firm-year observations from 1987 to 2008, representing 6,102 firms. The data requirement for vega reduces the risk-taking-incentive sample (hereafter, the reduced sample) to 13,086 firm-year observations (2,014 firms).
We first document a negative relation between risk taking and covenant intensity. Univariate analysis indicates that all five risk taking measures, i.e., R&D, number of segments, segment Herfindahl, vega from newly granted options, and total vega, monotonically decrease with the covenant intensity. Multivariate regression analysis finds the same pattern, that is, borrowers’ risk-taking activities and incentives decrease with covenant intensity, after controlling for leverage, size, ROA, Intangibles, sales growth, M/B, stock return, return volatility, cash balance, cash surplus, and dividend cut. Nonparametric analysis also yields consistent results that high covenant intensity firms have significantly lower risk taking than low covenant intensity firms that have the closest covenant propensity score. Collectively, these results suggest that covenants are effective in reducing risk taking.
We then address the potential endogeneity between debt covenants and risk taking. We first employ a change specification to examine whether an increase in covenant intensity leads to a decrease in risk taking, using the borrower itself as the control. We find consistent results that borrowers take less risk when there is an increase in covenant intensity, for both a short time series that includes only the year before and the year of loan initiation, and a longer time series that spans all the years before and after loan initiation. Then we run a two-stage-least-square regression where the instrumental variable is the average covenant intensity of the lending bank(s). We find strong and consistent results that borrowers’ risk-taking activities and incentives decrease with covenant intensity in the second stage. Finally and most importantly, we exploit the bond market reaction to additional covenants to gauge whether additional covenants simply capture expected reduction in risk taking. If only borrowers that expect lower risk taking in the future agree on restrictive covenants, we expect the bond market not to react positively to additional covenants because the lower risk taking is anticipated. Using daily bond prices from TRACE, we find a positive bond market reaction to the additional covenants introduced by new bank loans, suggesting that additional covenant restrictions do not simply capture expected lower risk taking. Taken together these three tests mitigate the concern of the endogeneity between covenants and risk taking.
We also conduct four additional analyses. First, we address the concern that covenant intensity is a surrogate for leverage; that is, borrowers with high leverage have lower risk taking through other mechanisms, such as short maturity.[5] After sorting the sample into four groups by leverage and covenant intensity, we find that firms with low leverage but high covenant intensity have lower risk taking than firms with high leverage but low covenant intensity. This indicates that leverage does not drive the negative relation between covenant and risk taking. In addition, covenant intensity has a consistently negative relation with risk taking after controlling for leverage, yet leverage does not have a consistently negative association with risk taking, suggesting that leverage has a more complex relation with risk taking.
Second, we examine whether covenant violations further limit risk taking. While it is important that the existence of covenants mitigates risk taking, covenant violations are important control transfer mechanisms that allow lenders further influence on firm decisions. We find a decrease in both risk-taking activities and incentives after covenant violations in the univariate analysis. However, in the multivariate analysis, we find that risk-taking activities drop significantly in the four years after covenant violations, but not risk-taking incentives. Overall, our evidence suggests that covenant violations have a further impact on curbing risk taking.
Third, we explore the role of short maturity in curbing risk taking. Unlike state contingent covenants, short maturity allows frequent renegotiation and thereby could also curb agency problems including risk taking. While Brockman, Martin, and Unlu (2009) find that short term debt constrains managerial risk preference, we find that covenants continue to reduce risk taking after controlling for short maturity. In fact, in most of the regressions, covenants dominate short maturity in reducing risk taking.
Lastly, we test the role of bond covenants in mitigating risk taking. We use loan covenants in our primary analyses because loan covenants are more binding due to lower renegotiation costs between banks and borrowers. Nevertheless, we provide evidence that bond covenants are also effective in reducing risk taking.
Our study contributes to the literature in several ways. First, we are the first study to directly examine whether debt covenants are effective in indirectly restricting borrowers’ risk taking. Chava and Roberts (2008) and Roberts and Sufi (2009) study how covenant violations restrict borrower’s subsequent investments and debt financing and covenants serve as “tripwires” in both cases. These papers do not address the question of whether restrictive covenants cause borrowers to reduce risk taking absent of a control transfer. The primary focus of this paper is to investigate whether stronger covenants restrict risk taking.
Second, we add to the literature on how debtholder-stockholder conflict impacts operating and investing decisions. In a closely related study, Nini, Smith, and Sufi (2009) document that capital expenditure restrictions in private debt contracts cause a reduction in firm investment. Our paper complements Nini et al. (2009) by investigating whether debtholder-stockolder conflict impacts operating decisions.[6] Instead of examining a specific covenant (i.e. capital expenditure restriction) and a specific wealth transfer action (i.e., over-investment), we study the overall covenant restriction (six major covenants collectively) and a broad set of risk taking behavior, including R&D, focus, and CEO risk-taking incentives. Overall, our findings highlight the importance of debtholder-shareholder conflict in impacting operating decisions and the importance of debt governance in mitigating risk taking.
The remainder of the paper is organized as follows. The next section summarizes related literature and develops the hypotheses. Section 3 describes our data, empirical proxies and research design. Section 4 presents the empirical results of our analyses and Section 5 concludes the study.
2. Literature and hypotheses
The early empirical literature on debt covenants examines the ex ante efficiency in the trade-off between the benefit and cost of including restrictive covenants. If the benefit of reducing the cost of debt outweighs the cost of loosing future flexibility, then the borrower would agree on restrictive covenants. Consistent with the theoretical predictions, firms are more likely to include restrictive covenants in the debt contracts if the leverage is high (Malitz, 1986), or the shareholder-bondholder conflicts are high (Begley and Feltham, 1999), or managerial entrenchment and the likelihood of committing fraud is high (Chava, Kumar, and Warga, 2009). On the benefit of covenants, Billett, King, and Mauer (2007) document that the negative relation between leverage and growth opportunities is attenuated by covenant protection. Their paper shows that covenants can mitigate the agency cost of debt for high growth firm through an increase in debt financing. However, all these papers study the ex ante contract design and are silent about whether and how covenants protect lenders ex post after debt is in place. [7]
Recently there is an upsurge of studies that investigate how covenants violations affect firm investing and financing. Chava and Roberts (2008) document that capital investment declines sharply following a financial covenant violation. Roberts and Sufi (2009) show that net debt issuing activity experiences a sharp and persistent decline following covenant violations. In both papers covenants serve as a trigger that transfers control rights to lenders, and lenders influence firm investing and financing using their right to accelerate the debt, increase interest rates, etc. While the tripwire function of covenants is undoubtedly important, these papers do not address the question whether covenants are effective in protecting lenders via the threat of a control transfer without an actual control transfer.
We are aware of only one recent paper that investigates the impact of covenants on firm behavior in the absence of a control transfer. Nini, Smith, and Sufi (2009) find that capital expenditure restrictions in private debt contracts are effective in reducing borrowers’ investment.[8] Our paper complements Nini et al. (2009) by studying whether six major covenant restrictions collectively (as opposed to the capital expenditure restriction) mitigate borrowers’ risk taking (as opposed to borrower’s over-investment). While the effect of capital expenditure restriction on over-investment is direct and less obscure, we focus on risk taking because incomplete contracting makes it unclear whether covenants are effective in limiting risk taking indirectly and there is no empirical evidence on this issue.
We expect covenants to mitigate risk taking because theory suggests that covenants exist to protect debtholders from agency problems and risk taking is one of the agency problems that transfer wealth from debtholders to shareholders. Debt covenants can mitigate risk taking in two ways. First, restrictive covenants limit borrower’s abilities to engage in risk taking (Smith and Warner, 1979). For example, asset sweep covenant restrict borrowers from disposing existing assets and thereby limiting borrowers’ ability to dispose less risky assets and replace them with more risky ones. Debt sweep and equity sweep covenant prevents borrowers from raising additional capital to finance risky projects. Secured debt covenants reduce borrowers’ abilities to substitute the pledged assets for risky ones because debtholders hold the title to these assets.