How Effectively Can Debt Covenants Alleviate Financial Agency Problems? Andrea Gamba Alexander J. Triantis∗ February 15, 2013 Abstract We examine the effectiveness of debt covenants in alleviating financial agency problems. Distortions in both investment and financing policies are captured in a structural dynamic model where both policies are endogenously determined by shareholders. The combined and compounding effect of these distortions is shown to be large. We impose covenants that restrict the level of debt, or control the use of proceeds from asset sales or debt issuance, and analyze how, and how much, theymitigatefinancialagencycosts. Weinvestigatethedirectandindirectimpact of covenants on financing and investment policies, including at the point where covenants are violated, providing alternative interpretations of recent empirical evidence. We conclude that the presence and enforcement of debt covenants sig- nificantly alters dynamic financing and investment policies, not only at the point of covenant violations, and thus should be an important element of structural models. ∗Gamba is at the Warwick Business School, University of Warwick ([email protected]). Triantis is at the Smith School of Business, University of Maryland ([email protected]). TriantisgratefullyacknowledgesfinancialsupportfromtheCenterforFinancialPolicyattheRobertH. Smith School of Business. We thank participants at seminars at Warwick Business School and Arizona State University for their helpful comments. 1 Electronic copy available at: http://ssrn.com/abstract=2101645 Introduction It has long been argued in the corporate finance literature that the use of debt financing introduces financial agency problems in the form of distortions to corporate investment and financing policies. The evolution from static to dynamic corporate finance models has expanded our perspectives on the nature and magnitude of these distortions, and the extent to which different debt contract designs can mitigate these problems. These have included a focus on seniority, maturity, and call and reset features.1 However, the role of debt covenants in resolving financial agency problems has not yet been carefully explored in a dynamic setting, despite much recent interest in the empirical literature on the role and impact of covenants. In this paper, we introduce different debt covenants into a dynamic structural model of the firm to analyze how, and how well, they mitigate investment and financing distortions. Given the prevalence of debt covenants, particularly in private debt contracts, they are generally viewed as value enhancing design features. The underlying concept is that they allow a state-contingent transfer of control from shareholders to bondholders which can mitigate financial agency problems. Smith and Warner (1979) provide the first detailed evidence on the types of covenants used, and tied these covenants to agency problems they were designed to mitigate. More recently, Bradley and Roberts (2004) and Billett, King, and Mauer (2007) provide additional empirical evidence on the types of covenants embedded into private and public debt contracts, respectively. Bradley and Roberts (2004) find that covenants have a direct effect on decreasing yields. Billett, King, and Mauer (2007) explore the relationship of different covenants to the leverage ratio, debt maturity structure, and the prevalence of growth opportunities. Another strand of empirical literature on covenants focuses specifically on the states where covenants are violated. Beneish and Press (1993), Beneish and Press (1995), Dichev and Skinner (2002), and Sweeney (1994) document that covenants are written with triggers that are quite tight relative to the firm’s financial condition at the time of debt issuance, and as a result the frequencies of covenant violations are quite high. They also explore firms’ accounting choices as they come close to triggering covenant 1HackbarthandMauer(2012)exploreoptimaldebtpriorityinadynamicsetting. RajanandWinton (1995) discuss the role of short-term debt, but also the impact of long-term loans subject to covenants. Bhanot and Mello (2006) explore shareholder incentives when debt includes rating triggers that affect debt contract structure as a firm’s credit risk increases. 2 Electronic copy available at: http://ssrn.com/abstract=2101645 violations. More recently, Chava and Roberts (2008), Roberts and Sufi (2009), and Nini, Smith, andSufi(2009)examinefirms’investmentandfinancingbehavioronceafinancial covenant has been violated, finding a decrease in both investment rates and debt levels, and concluding that lenders use the opportunity of a technical default trigger to pressure firms to alter their policies and reduce credit risk. We examine the investment and financing policies that shareholders would follow absent any covenant restrictions, and the resulting magnitude of financial agency costs. While several recent papers have focused on measuring agency costs using dynamic models, their models have imposed either a fixed level of debt (Childs, Mauer, and Ott (2005), Mello and Parsons (1992), Moyen (2007)), a specified maturity schedule for debt (Leland (1998), Parrino and Weisbach (1999)), or restrictions on the issuance of new debt (Titman and Tsyplakov (2007)). While a flexible financing policy could potentially help to mitigate investment distortions, we show how shareholders will instead game the financing policy, taking advantage of legacy debt holders. This financing distortion further exacerbates underinvestment problems, and the resulting compounding effect leads to significant value loss, far greater in magnitude than documented in the existing literature. Using this baseline, we measure how effective debt covenants can be in moving share- holders’ investment and financing policies closer towards first–best policies, and the ex- tent to which agency costs can thus be mitigated. While a wide range of covenants are used in practice, we focus our attention on three representative covenants that highlight some key characteristics of covenants. The first is an asset sweep covenant that requires shareholders to use proceeds from asset sales to pay down debt, and thus discourages asset sales designed merely to fund shareholder payouts.2 The second is a debt sweep covenant that specifies that proceeds from new debt issuance be used to pay down ex- isting debt, and thus targets opportunistic leverage increases that expropriate current bondholders’ wealth.3 The third is a financial accounting covenant which is violated if the firm’s Debt/EBITDA ratio exceeds a specified threshold. Shareholders can remedy 2Sweep provisions are highlighted by Bradley and Roberts (2004). Debt encumbered with an asset sweep covenant is similar to secured debt. Morellec (2001) explores the joint effects of asset liquidity and pledging assets as collateral on the debt capacity of a firm. 3Itisalsocommontofindcovenantsthatrestrictdividendpayments. Wedonotmodelcashretention in our model, so the drivers of excessive dividend payments in our context would be either liquidated capital or new debt issuance, and the two sweep covenants address these potential agency problems directly. 3 this technical default by making investment and/or financing decisions that will limit the firm’s Debt/EBITDA ratio under similar profitability scenarios. We find that these covenants are effective in different ways, and to varying degrees, in reducing investment and financing distortions and mitigating the value loss associ- ated with financial agency problems. Due to the compounding effects of investment and financing distortions, the covenants that reduce the likelihood of new debt issues or increase the incentive to reduce debt are also quite effective in mitigating investment distortions despite not targeting investment directly. Similarly, the covenant that ad- dresses asset liquidations also indirectly mitigates financing distortions. Furthermore, we show that covenants alter investment and financing policies across many states, and not simply at the points where the covenants are binding or violated. These policy modi- fications result in significant value creation, particularly when the propensity to increase leverage is controlled, and in low states of profitability due to either macroeconomic or firm-specific factors. Our state–contingent framework is particularly well-suited to examine the impact of a debt covenant violation on investment and financing policies, and to provide insights on the recent empirical literature on this topic. A key advantage of our model is that we can conduct parallel simulations where debt is, or is not, subject to a financial covenant. We can thus examine in a controlled manner the investment and financing policies at the specific states where the covenant would get violated if it existed, and see whether the existence of the covenant indeed makes a difference. From this controlled experiment, we find that while the investment rate drops fol- lowing a covenant violation, once other factors related to these low states are controlled for in a multivariate regression, the violation itself actually results in higher investment. We put these results in the context of conflicting empirical conclusions in Chava and Roberts (2008), Roberts and Sufi (2009), and Nini, Smith, and Sufi (2009). In contrast, the empirically documented drop in debt levels at the time of a financial covenant vio- lation appears to be an unambiguous consequence of the violation and its remedy. Our simulations also produce a significant drop in net payout once the covenant is violated. 4 Our model is closest in structure to Brennan and Schwartz (1984) and Titman and Tsyplakov (2007).4 In their path-breaking article, Brennan and Schwartz (1984) laid the foundations for models that incorporate both dynamic investment and dynamic financing. Their model also includes a financial accounting covenant. However, the main purpose of their paper is to show how a contingent-claims framework can be applied to value a firm allowing for dynamic corporate policies. Their results regarding the effect of the financial covenant are thus very limited. More recently, Titman and Tsyplakov (2007) present a fully dynamic model of en- dogenous investment and leverage decisions that incorporates the effects of taxes, finan- cial distress costs, and adjustment costs for both financing and investment. While our model shares many similar structural characteristics with theirs, they do not explicitly incorporate debt covenants. They do impose a debt restructuring condition similar to that in Fischer, Heinkel, and Zechner (1989) which requires all debt to be repurchased andnewdebttobeissuedwhenthefirmchangesitsdebtlevel. Ourdebtsweepcovenant shares some commonality with this condition, but is less restrictive, and its effects are explicitly explored. In a simpler dynamic setting that leads to closed form solutions, Leland (1994) ex- amines the effect of a positive net-worth covenant on the propensity of shareholders to increase the riskiness of assets, and on the firm’s capital structure. Consistent with Leland (1994) and the related dynamic corporate finance literature, we do not seek to deriveoptimalsecuritydesigns, butrathertoexaminetheeffectofcommondebtfeatures on dynamic investment and financing strategies and the resulting firm value. A differ- ent strain of the corporate finance literature studies the optimal allocation of control, and derives the design of covenants endogenously (within the class of debt contracts). For example, Garleanu and Zwiebel (2009) examine entrepreneur-lender conflicts under asymmetric information, while Mao (2011) models managerial-shareholder conflicts and derives debt covenants that address the moral hazard problem.5 4Other papers such as Childs, Mauer, and Ott (2005)) examine dynamic investment in the presence of debt, but allow only for a single growth option that can be exercised once rather than allowing for switching between various capital levels, and they don’t incorporate dynamic debt policies. 5Earlier related papers on debt covenants include Berlin and Mester (1992), Rajan and Winton (1995), and Sridhar and Magee (1997). In an interesting departure to the conventional literature on debt covenants, Murfin (2012) examines the issue of covenant strictness from the supply side rather than from the firm’s perspective. He finds evidence that defaults in a bank’s loan portfolio tends to 5 The next section presents the details of our model. Section 2 provides results re- garding the measurement of agency costs, the impact of covenants on investment and financing policies in general and at covenant violation points, and the overall effective- ness of covenants in reducing agency costs. Section 3 concludes the paper. 1 The Model 1.1 Economic and Financial Setting We model investment and financing decisions in an infinite-horizon discrete-time dy- namic and stochastic framework. The control variables are the book value of assets in place, k, and the face value of outstanding debt, b. We denote macroeconomic risk by x, and assume that it is an autoregressive process6 x(cid:48) = ρ x+σ ε(cid:48), ε(cid:48) ∼ N(0,1), |ρ | < 1 (1) x x x x x where ρ is the autocorrelation, σ is the conditional standard deviation, and the terms x x ε are serially i.i.d.. x We model a finite set of heterogeneous firms in this economy, each driven by an independent company–specific risk, denoted with z , and assume it also follows an au- j toregressive process z(cid:48) = ρ z +σ ε(cid:48), ε(cid:48) ∼ N(0,1), |ρ | < 1 (2) j z j z j j z where the parameters have similar meanings as above, and ε are serially i.i.d.. The j two contemporaneous random terms ε and ε are independent, and the random terms x j ε are cross–sectionally independent across the firms. In the rest of this section we will j describe the behavior of an individual firm, so we will drop the index j, for convenience. lead to increased strictness of covenants on subsequent portfolio loans, likely due to learning about the lender’s screening ability. 6We describe the evolution of a system at the steady–state. Hence, only the current period, t, and the next one, t+1 are relevant. We use a prime to denote the value of a variable in period t+1, the same variable without a prime denoting its value in period t. 6 The EBITDA (operating cash flow before taxes) is π(x,z,k) = ex+zkα −ψ, (3) where k > 0 is the book value of assets, ψ > 0 is a fixed cost that includes all expenses, and α < 1 models decreasing returns to scale of the asset. We assume that capital is homogeneous and that it depreciates both economically and for accounting purposes at a constant rate δ > 0. The company’s debt consists of risky consol bonds of equal priority with face value b ≥ 0 and a coupon rate equal to the risk free rate r.7 We assume that the firm can not hold a cash balance in order to simplify the model.8 The Earnings Before Taxes (EBT) is equal to EBITDA minus depreciation and in- terest: y = y(x,z,k,b) = π −δk −rb. (4) We model corporate taxes on earnings net of personal taxes. The net corporate tax function is g(y) = τy, where τ is the marginal net tax rate on earnings. At any date, given the current state (x,z,k,b), the firm can decide to invest or disinvest to get the capital stock for next period, k(cid:48).9 If there is positive investment, then the capital expenditure is k(cid:48) −(1−δ)k and investment can be financed either by using current cash flows or by issuing debt or equity. On the contrary, if the firm decides to disinvest, the cash inflow is (cid:96)((1−δ)k −k(cid:48)), with (cid:96) ≤ 1 denoting a liquidation price. Thefirmmayalsodecideatanytimetoincreaseorreduceitsdebttoanewlevelb(cid:48) for thenextperiod. Anyvariationofdebtfrombtob(cid:48) entailsadirectcost, q(b(cid:48),b) = η|b(cid:48)−b|, where η ≥ 0. Additional debt is issued at market value, and old and new debt have equal seniority. We later explore the possibility that covenants impose restrictions on 7Several structural corporate finance models explicitly model debt maturity using an exogenously specified debt repayment rate (e.g., Leland (1998), Dangl and Zechner (2004), and Titman and Tsy- plakov (2007)). As we discuss below, debt is dynamically adjusted in our model, and thus while debt hasinfinitematurityonanex-antebasis,theeffectiveex-postmaturityisendogenouslydeterminedand state-dependent, reflecting financing and investment decisions, as well as the impact of any covenants. 8Gamba and Triantis (2008) and Riddick and Whited (2009) address the decision to raise external capital and simultaneously hold a cash balance. 9We assume that investment instantaneously increases the productive capacity of the firm. In con- trast, Tsyplakov (2008) examines the effect of the time-to-build characteristic of capital in certain industries, and shows that this can affect dynamic capital structure choices. 7 new debt issuance. We assume that the debt level can be reduced by repurchasing a portion of the debt at its par value, rather than its market value. Mao and Tserlukevich (2012) show that in the absence of frictions, lenders will sell their debt back to the firm only at par value. Even with frictions, debt will be repurchased at a premium to market value, though the size of the premium depends on the nature of the frictions. Mao and Tserlukevich (2012) also detail institutional considerations, including securities and tax regulations, that make it more likely that debt is repurchased at or close to par value. For notational convenience, where appropriate, we will denote the states θ = (x, z), as specific combinations of the macroeconomic and firm-specific states. At any date, given specific capital (k) and debt (b) levels, the state θ and the resulting cash flow will be observed, and new levels of capital (k(cid:48)) and debt (b(cid:48)) are chosen. While the state is observable to shareholders (and, as will be discussed below, to debtholders to price their claims), we assume that the debt contract is incomplete in that a state- contingent contract that imposes specific investment and financing decisions in all states is infeasible. When we later introduce covenants, we will assume that specific events are contractible. Investment and financing decisions can be made with no restriction as long as the firm is not in financial distress or in a state of default (or as described later, restricted by, or in violation of, a covenant). We assume that financial distress takes place when the after-tax operating cash flow is lower than the coupon payment, rb > π −g(y). In this case, the firm sells the minimum amount of capital, at a discount s, to make the promised payment to debt holders. In contrast, as we shall soon define, default occurs at states which shareholders select to maximize the value of their equity given the option created by limited liability. Based on the assumptions above, we can determine the residual cash flow to share- holders in state (θ,k,b) when the firm is solvent, given a decision (k(cid:48),b(cid:48)). For notational convenience, we define the function χ(ξ,(cid:96)) as χ(ξ,(cid:96)) = ξ if the capital stock change ξ = k(cid:48) −k ≥ 0, and χ(ξ,(cid:96)) = ξ(cid:96) if ξ < 0, where (cid:96) ≤ 1 is the liquidation price of capital. D(θ,k(cid:48),b(cid:48)) is the market price of newly issued debt, assuming the firm is solvent, based 8 on the new book value of assets, k(cid:48), and the new book value of debt, b(cid:48). The residual cash flow to shareholders, where I(·) is the indicator function for the event b(cid:48) < b, is e(θ,k,b,k(cid:48),b(cid:48)) = max{π −g(y)−rb,0} (cid:101) D(θ,k(cid:48),b(cid:48)) +I(b(cid:48))(b(cid:48) −b)+(1−I(b(cid:48))) (b(cid:48) −b)−q(b(cid:48),b) b(cid:48) (cid:18) (cid:26) (cid:27) (cid:19) rb+g(y)−π −χ k(cid:48) −k(1−δ)+max ,0 ,(cid:96) . (5) s The first line of (5) is the after-tax operating profit; the second line is the cash flow from altering the debt level (first term is for a decrease in debt, and the second term is for an increase); and the last line is the cash flow due to investment or disinvestment, where the latter may be subject to a liquidation and/or distress sale discount. If the optimal residual cash flow for shareholders is negative, then funds are raised by issuing new equity, and a proportional flotation cost λ (0 < λ < 1) is incurred. Thus, when the firm is solvent, the net cash flow to equity holders is e(θ,k,b,k(cid:48),b(cid:48)) = e−λmax{−e,0} (6) (cid:101) (cid:101) 1.2 Optimization We now set up the valuation of the firm (and its corporate securities) under different optimization conditions. In Section B.1, we model the cases where the investment and financing decisions are made to maximize either equity value or total firm value.10 In Section B.2, we introduce debt covenant restrictions into the second–best optimization problem. 10Managerialagencyproblemswouldaddaninterestinglayerofcomplexitytotheproblemweexplore. Covenants could conceivably either mitigate or exacerbate managerial agency problems, depending on thenatureofthoseproblemsandtheothermechanisms, suchascompensationdesign, thatareinplace to resolve the conflict between managers and shareholders. We leave these interesting issues to further research given the complexities they would introduce into our framework. 9 1.2.1 Equity and Firm Value Maximization Following Berk, Green, and Naik (1999), Zhang (2005), and others, we assume that the stochastic discount factor depends on macroeconomic conditions captured by x. We use the convenient functional form introduced by Jones and Tuzel (2012), where the one–period discount factor given the transition to state x(cid:48) from the current state x is M(θ,θ(cid:48)) = βe−γ(x)ε(cid:48)x−12(γ(x))2σx2, (7) where the state–dependent coefficient of risk–aversion is defined as γ(x) = exp(γ +γ x), 0 1 with 0 < β < 1, γ > 0 and γ < 0.11 0 1 We first examine equity value maximization. E(θ,k,b), the value of equity at state (θ,k,b), is the fixed point of the Bellman operator (cid:26) (cid:27) E(θ,k,b) = max max{e(θ,k,b,k(cid:48),b(cid:48))+E [M(θ,θ(cid:48))E(θ(cid:48),k(cid:48),b(cid:48))]},0 , (8) θ (k(cid:48),b(cid:48)) where the expectation is conditional on the current state θ. The optimal policy at (θ,k,b) is (k∗,b∗) = argmax{e(θ,k,b,k(cid:48),b(cid:48))+E [M(θ,θ(cid:48))E(θ(cid:48),k(cid:48),b(cid:48))]}, θ (k(cid:48),b(cid:48)) as long as the resulting equity value from this policy is not negative. Otherwise, the shareholders default on servicing debt and surrender the firm to debt holders. When default occurs, a bankruptcy cost is incurred, which is a proportion ζ (0 ≤ ζ < 1) of the debt holder payoff b(1+r). We define ∆ = ∆(θ,k,b) to be the indicator function of the default event in the cur- rent state. When default occurs, the former bondholders become the new shareholders, taking over an unlevered firm with depreciated assets, k(1 − δ), so that the policy is (k(1−δ),0). Hence, the optimal policy for shareholders at (θ,k,b) is ϕ(θ,k,b) = (k∗,b∗)·(1−∆)+(k(1−δ),0)·∆. (9) 11With this functional form of the stochastic discount factor, the return of a risk–free one–period zero coupon bond, 1/E [M ] = 1/β, is independent of the state. This is an important feature when t t+1 modelling a firm that can issue long term debt, whose par value is state dependent. 10