Collateral and Debt Capacity in the Optimal Capital Structure

Transcription

1 IRES IRES Working Paper Series Collateral and Debt Capacity in the Optimal Capital Strctre Erasmo Giambona Antonio Mello Timothy Riddiogh May 2011

2 Collateral and Debt Capacity in the Optimal Capital Strctre Erasmo Giambona Antonio Mello Timothy Riddiogh University of Amsterdam University of Wisconsin University of Wisconsin Febrary 2009 Abstract This paper considers how collateral is sed to finance a going concern. We focs on firms that offer collateral with significant debt capacity, and a setting where debt is nconditionally an optimal contract. A theory is developed in which firms with better qality collateral se that collateral to finance new investment opportnities with nsecred debt, while firms endowed with lower qality collateral se secred debt. Better qality firms mst also isse eqity to separate themselves from lesser qality firms, implying lower leverage and greater ncommitted cash flow from operations with which to meet debt service reqirements. Empirical evidence from Real Estate Investment Trsts (REITs) spport predictions of the theory, where we find that firms with less reliance on secred debt and lower total leverage invest more, are valed more highly and financially otperform firms with higher levels of secred debt and total leverage. Or findings provide an alternative perspective on risk-shifting and the vale of free cash flow, and demonstrate limits to the benefits of leverage as it depends on debt capacity.

3 Collateral and Debt Capacity in the Optimal Capital Strctre I. Introdction The role of collateral in lending is important de to its effects on the real economy, as propagated throgh the credit channel. Bernanke and Gertler (1989) and Kiyotaki and Moore (1997) show that the vale of collateralized assets impact firms debt capacity, in trn affecting investment and otpt. Becase lenders have limited ability to enforce repayment, pledging collateral is indispensable to ameliorate credit frictions. The importance of collateral is docmented in Berger and Udell (1990), and Harhoff and Korting (1998), who find that nearly 70% of commercial and indstrial loans in the U.S., the U. K., and Germany are secred. Yet, for all of the attention paid to the topic there remain nmeros nanswered qestions abot how collateral is sed by firms to invest and isse claims to finance their investments, and more generally how collateral affects debt capacity, which in trn helps govern the optimal capital strctre of firms as going concerns. 1 Perhaps the most important form of collateral is real estate. A recent srvey of the World Bank reports that in 58 emerging contries, real estate represents an average of 50% of firm s collateral. 2 Similarly, Gan finds that 70% of secred loans in Japan were backed by land. The role of real estate assets in collateralizing financial claims makes Real Estate Investment Trsts (REITs) an interesting and niqe object of analysis. REITs are tax exempt pblicly traded companies that typically hold ownership positions in commercial real estate. In retrn for taxexemption, REITs are reqired to distribte at least 90% of the income generated. For these 1 Gan (2007) considers many of these isses in her analysis of bank lending in the face of the collapse of the Japanese land market. 2 See World Bank Investment Climate Srvey at 1

4 firms, drability and redeployability of assets imply low marginal expected deadweight costs of financial distress as a fnction of leverage. Ths, even when a small wedge between the cost of issing eqity verss debt exists, firms will be tempted to exploit this wedge by tilizing significant amonts of debt (Almeida and Campello (2007)). Bt, at the same time, the provisions that REITs do not pay taxes at the firm level and are reqired to distribte a large percentage of operating cash flow as dividends remove two key incentives to employ debt relative to eqity. Moreover, traditional pecking-order (Donaldson (1961) and Myers and Majlf (1984)), and principal-agent,-free cash flow (Easterbrook (1984) and Jensen (1986)) rationales for debt issance are less compelling for these firms, as significant constraints on the retention of cash redces managerial discretion to redeploy retained earnings and free cash flow to fnd investment. Table 1 docments leverage for a sample of REITs from 1991 to 2007 as they compare to non-financial C-Corporations (the Compstat Sample ), where we also classify debt based on whether it is secred or nsecred. This table shows that REITs employ significantly greater leverage and finance themselves with significantly more secred debt than the comparison grop, sggesting that debt capacity of real estate assets create powerfl incentives for REITs to lever themselves p, and that they do so with a lot of secred debt. Table 1 Here While REITs employ a lot of leverage, the existence of debt capacity does not necessarily imply that a capital strctre with a lot of debt, particlarly secred debt, is an optimal capital strctre given the noted offsetting effects of no-taxes and an inability to retain cash flow. To begin to get at this isse, in Table 2 we calclate average q vales of REITs. Firms are classified as being above or below the sample median in three different ways: 1) by the ratio of secred 2

5 debt to total debt; 2) by the ratio of total debt to total assets; and 3) by the cross term of the two previos measres, the ratio of secred debt to total assets. Note that total assets sed to calclate total leverage and secred leverage are reported on a market vale as well as a book vale basis. Table 2 Here Reslts show that firms that employ greater leverage and se a greater proportion of secred debt in their capital strctre have lower q vales. This simple relation sggest that, at a minimm, there are limits to collateral as a mechanism to debt finance, particlarly with secred debt. Bt what are the factors that wold lead to an otcome where, say 40 percent total leverage with limited secred debt is an optimal capital strctre for firms whose assets provide terrific collateral vale and received wisdom which sggests that debt, and often secred debt, is an optimal contract? The verdict is still nclear. For example, Boot, Thakor and Udell (1991) indicate that firms with good fndamentals borrow more and se more collateral to signal their qality. On the other hand, Rajan and Winton (1995) show that banks lend less and reqire more collateral from low qality firms. While real estate makes for excellent collateral in general, there can be significant differences in the prodctivity of the collateral in terms of its intrinsic qalities and how it is managed. To appreciate the robstness of the reslts noted in Table 2, in Table 3 we report financial performance reslts of REITs dring the time period a period in which real estate asset prices generally increased and the time period which is the time frame covering the beginning and end of the most intensive period of the financial market crisis. The table shows that firms classified as being below the median in terms of the ratio of secred debt to total assets generated higher total retrns for their shareholders than above-median firms 3

6 in both time periods; i.e., firms that tilized less secred debt and total leverage generated higher retrns for their shareholders in good times as well as bad. Table 3 Here Table 3 also shows that firms that had below-median amonts of secred debt to total assets in their capital strctres incrred total lower interest expenses as a percentage of total debt otstanding, implying that secred debt is more expensive than nsecred debt. With this collection of stylized empirical facts in mind REITs employ significantly more leverage and secred debt relative to indstrial firms, bt also that REITs which generate higher q vales have better stock market performance, lower debt financing costs, and employ relatively less leverage as well as have lower levels of secred debt we consider theory and evidence on the optimal capital strctre of going concerns that own collateralizable assets with significant debt capacity. We develop a theoretical model that is in the spirit of Bester (1985), and that also borrows from the seminal work of Stlz and Johnson (1985) and Myers and Rajan (1998). In or model, firms are characterized as going concerns with collateralizable assets in place. The assets in place have prodctive qalities that vary across firms and that are nobservable to otsiders. Debt has been sed to finance ownership of the assets in place. Moreover, debt is nconditionally an optimal contract becase there are fixed costs associated with issing eqity. A new constant-qality positive NPV project is available for investment, and firms are interested in financing the investment. The firm is nable to credibly commit any of its existing liqid resorces to finance the new project, bt its real assets in place are available as additional collateral. 3 If real assets in place are offered as collateral to finance the new project, we say the new financing is nsecred (that is, all of the available assets of the firm collateralize all of the 3 Note that in the case of REITs the reqirement to pay ot a large proportion of the operating cash flows drastically limits the amont of inside eqity and holdings of liqid assets. 4

7 debt of the firm, old as well as new). Otherwise, if only the new investment is available to collateralize the new debt, we say that the debt is secred. 4 The critical qestion we consider is how firms, which are endowed with assets in place that have an nobservable component that varies in qality, shold optimally finance their investment and how creditors shold optimally screen firms in order to ascertain their tre type. We show that, conditional on their identity being known to otside financiers, higher qality firms prefer to finance themselves with nsecred debt, and lower qality firms prefer to finance themselves with secred debt. The basic intition is that of Bester (1985), with an important twist: better qality firms, which possess more valable assets in place, can improve issance proceeds by pledging those assets as collateral to finance the new investment. The opposite effect occrs with lower qality firms, so they finance investment with secred debt. These relations are robst to any participation constraints that might be imposed by existing debt holders to limit wealth transfers associated with financing new investment. The problem of corse is that lower qality firms wold like to represent to otsiders that they are higher qality firms, in order to isse nsecred debt and ths increase issance proceeds. Withot some method to differentiate firms,, lenders will otherwise decide to pool all borrowers. Competition between lenders will reslt in screening, which is done by reqiring firms that wish to finance themselves with nsecred debt to also isse eqity. This leverage decreasing action, while costly to higher qality firms, is more costly for lower qality firms than it is for higher qality firms. As a reslt, screening is sch that higher qality firms isse eqity 4 We are interested in the cross-collateralization properties of nsecred debt relative to secred debt. Conseqently we limit recovery rights to secred debt to the vale of a specified asset. There are other ways to characterize secred and nsecred debt. For example, Stlz and Johnson (1985) focs on priority of secred debt withot liability being limited to the vale of any specific asset. Boot, Thakor and Udell (1991) also focs on priority and simply assme that nsecred debt generates a zero payoff in defalt while secred debt generates a positive payoff. 5

8 in sfficient qantity to make it too expensive for lower qality firms to mimic them, and a separating eqilibrim reslts. Eqilibrim implies that, consistent with the stylized empirical facts otlined earlier, higher qality firms will, throgh their financing decisions, reveal themselves with higher q vales, ths generating greater retrns to shareholders over time. They will also have lower incremental costs of debt finance, and will se less leverage with a greater proportion of nsecred debt to finance new investment opportnities. We expect these effects to be persistent and self-reinforcing. Less leverage by higher qality firms is a credible commitment to insring creditors against the risk of loan defalt. To the extent that this can relax financial constraints of firms, higher qality firms will also invest more. Formal tests of model implications are ndertaken sing an nbalanced panel of annal eqity REIT data over the years 1991 to In the first of two basic empirical specifications, we match firms based on growth opportnities, size, profitability, earnings volatility and age, and assess the effects of financing decisions on firm performance. Based on or smmary measre of the joint financing type-leverage decision the ratio of secred debt to total assets we categorize firms as being above and below the median vale in the sample. Performance measres are calclated at 12, 24 and 36 months into the ftre after the match date. For each and every performance measre considered which incldes investment, retrn on assets, and vale mltiple we find that below-median firms otperform above-median firms. We also docment that below-median firms isse greater amonts of eqity in the ftre as a percentage of investment, indicating that financing differences that distingish firms persist into the ftre and therefore are self-reinforcing. 6

9 Or second approach is to estimate the relation between q and financing decisions sing a GMM estimation methodology. Consistent with predictions of the theory, we find a robst casal relation going from the joint financing choice-leverage decision to changes in q, where an increase in the secred debt-to-total asset ratio is negatively related to the change in q. At the same time, q is fond to case and have a negative effect on the secred debt-to-total asset ratio, sggesting feed-back and ths persistence in the categorization of firms as higher and lower qality. Reslts based on the smmary financial measre are robst in the sense that q is negatively related to the component parts of the joint financing decision, those being secred debt-to-total debt and total debt-to-total assets. We believe that stdying REITs, which are going concerns that possess collateralizable assets with high debt capacity, is sefl for stdying broader isses in corporate finance and macroeconomics. These firms do not pay taxes and are exogenosly cash flow constrained, which provides a controlled environment for isolating effects associated with collateral, information and agency as they relate to isses of optimal capital strctre and the role of collateral in the economy. Or reslts sggest that a nmber of the firms we analyze were over-levered and sed secred debt in excess, and that they did so ot of weakness. We conclde that, jst becase assets are collateralizable with significant debt capacity, flly exploiting that debt capacity is sally not an optimal strategy. Rather, like everything else in the world, delicate and oftentimes sbtle tradeoffs need to be considered. II. Theory and Empirical Implications II.A. Model 7

10 Or model contains elements of three seminal papers that consider collateral and debt financing: Bester (1985), Stlz and Johnson (1985) and Myers and Rajan (1998). The basic model strctre and mch of the economic intition follows from Bester (1985), which is a screening-signaling model in which collateral is sed to indce separation between projects of varying qality. There are, however, some important modifications to the Bester model to properly characterize financing decisions associated with a going concern as opposed one-off project financing. As in Stlz and Johnson (1985), we distingish between secred and nsecred debt, and consider incremental financing decisions when a firm is presented with a new investment opportnity. Major differences between their paper and or approach are: i) distingishing between the collateralizability of existing assets as they relate to secred verss nsecred claims; ii) the consideration of eqity issance in addition to the possibility of debt issance; and iii) or focs on asymmetric information as being instrmental in motivating financial decision making. Like in Myers and Rajan (1998) we consider the commitment vale of liqid verss illiqid assets, positing that cash on the balance sheet has no commitment vale. Given that REITs are reqired to pay a large percentage of their operating income as dividends, liqid assets represent a small fraction of REITs balance sheets. However, we consider the isse of commitment as it occrs vis-à-vis eqity issance, by arging that firms can affect commitment vale of operating cash flow by redcing debt service reqirements. Ths, cash from an eqity issance, in effect, sbstittes for debt that wold otherwise increase debt service reqirements and hence the credit risk of the firm. The commitment to isse eqity is a separating and 8

11 incentive compatible screening devise that is reqired by lenders, in that only higher qality firms will optimally choose to isse eqity. The formal model is as follows. Consider two types of firms (real estate firms) that are characterized by their assets in place. These assets are a project or a collection of projects of different measrable qality, w j, where j=g,b indicates good or bad. The assets in place collectively generate a binary payoff in one period, where the otcome is either high, H, or low, L. For simplicity, consider the otcomes to be eqally likely. This assmption does not affect the reslts in any way, and makes the analysis cleaner and easier to follow. The expected payoff of the good project is higher than the expected payoff of the bad project; i.e., E[w G ]>E[w B ]. However, a low otcome of a good project pays off less than a good otcome of a bad project: i.e., w w L G H B. This implies that the bad project is not necessarily riskier than the good project. Alternatively, we cold have specified the qality of the projects sing different probabilities of a low verss high otcome. The assets in place, represented by project w, were lanched sometime in the past, say time t 1. At the present time, t 0, otside financiers cannot distingish between firm type in terms of whether the firm is endowed with w G or w B. Assme the firm previosly issed debt with a crrent vale eqal to B w to help fnd project w. Becase asset qality is indistingishable ex ante, B w is a price that reflects ignorance, or a pooling, with respect to the tre w. Said differently, based on fll information, a G-qality firm has lower crrent leverage and greater debt capacity than a B-qality firm at time t 0. At time t 0+ the firm has the opportnity of lanching another project,. Payoffs to the project do not depend on firm type, and will be realized in one period. These payoffs are binary otcomes, either high or low. The high payoff, H, is eqal to the expectation of H wb and H w G, 9

12 and the low payoff, L, eqal to the expectation of L wb and L w G. Qality and size of the new project are ths exogenos, so they cannot be sed to reveal information to otsiders. The new project is, frthermore, average as it compares to the assets in place of the two existing firms. More specifically, these payoffs imply that the new project has a lower expected payoff for the G-qality firm than the expected payoff to assets in place, whereas the opposite is tre for the B- qality firm. To simplify the analysis, assme that the payoffs to the projects are independent; that is, w =0. This assmption is also not necessary to generate or reslts. Unconditionally, the new project is optimally financed with debt becase eqity issance occrs at greater relative cost. For the moment assme that prospective creditors for, who operate in a competitive loan market, are nable to identify the qality of the firm of a given type w. The issance of new debt poses a problem to the creditors of project. Shold the debt offered to finance project be secred or nsecred? How can creditors of project screen borrowers in an attempt to identify their tre qality? What does this do to the capital strctre of project, in terms of the price and qantity of debt offered? To begin to answer these qestions, consider payoffs to the bondholders as they depend on the type of debt issed to finance project. For simplicity, we will assme that the project can be flly debt financed in the secred debt market. We will also assme that there is no commitment vale to any liqidity held by the firm that exists prior to investment in (Myers and Rajan (1998)). Conseqently, if the new debt is secred, bondholders of w are naffected by financing and investment in. This follows becase, in a H w j state, the debt is paid in fll in the amont of D w, where D w is the face vale of the debt in place. In contrast, in a L w j state, defalt L occrs and w D is recovered by the creditor in lie of fll contractal debt repayment. j w 10

13 Similarly, bondholders of receive D in a H state; otherwise, they receive L Dw. As stated earlier, payoffs to secred debt on are independent of firm type. We assme a zero discont rate and that all agents are risk netral. By fixing the payment dates at loan matrity, we see that B w and B are jst the expected vale of payoffs to bondholders of projects w and in the next period, t 1, when the projects payoffs are realized and the firm is assmed to be liqidated. Now consider the case of issing nsecred debt to finance. Issing nsecred debt to finance project implies that either both projects pay off as promised or both projects defalt. Defalt is sch that payoffs to w and bondholders depend on the pooled payoffs across projects and a pro rata priority allocation rle that applies in the case of defalt. Conseqently, in contrast to others that characterize nsecred debt as having little or no recovery vale in defalt, we do so in the context of cross-collateralizing all of the assets to secre all of the debt of the firm. To calclate proceeds from an nsecred debt issance, we will need to consider whether bondholders are facing a G-qality firm or a B-qality firm. Sppose first that creditors of project nknowingly face a B-qality firm with asset qality w B. In this case defalt is only avoided if both projects w and have a high state otcome. Ths, if either w or experience a low state otcome, there will not be enogh proceeds to repay the old and new creditors in fll. As a reslt of defalt, creditors will receive their fair share of the defalt vale of the firm across the two projects, where the fair share allocated to creditors follows a pro rata distribtion rle of =D /(D +D w ). In order to simplify and maintain symmetry in the analysis, we will assme that D =D w, implying that =1/2. Reslts can begeneralized to an arbitrary D and D w. 11

14 Alternatively, sppose now that creditors of project nknowingly face a G-qality firm with asset qality w G. In this case defalt occrs only when L and w are jointly realized, implying that there are sfficient proceeds from L + w as well as H + w to fnd the joint promised debt payoffs of D +D w. Ths, from a financing perspective, the fndamental distinction between the G-qality firm and the B-qality firm is defalt risk associated with nsecred debt issance, in which the B-qality firm is a greater credit risk than the G-qality firm. This credit risk manifests itself in the probability of defalt as well as in the severity of loss associated with defalt. Payoffs to creditors conditional on an nsecred debt issance can now be stated, and are smmarized in Table 4. H G L G L G Table 4 Payoffs to Project Unsecred Creditors as a Fnction of State Otcome and Firm Type Firm Type ( H, State Otcome Combination w ) ( H, w ) ( L, w ) ( L, H j L j H j L w j ) j=b D ( H + w L B ) ( L + w H B ) ( L + w L B ) j=g D D D ( L + w L G ) Conditional on creditors knowing firm type, we are now in a position to consider the firm s preferred choice of debt to finance investment in. Withot any eqity commitment vale, existing eqityholders are nable to affect management s financing decisions. This implies that management will conditiontheir financing choice decision on that which maximizes proceeds from secrity issance. As a reslt, financing choice reqires a comparison of expected payoffs to creditors conditional on the se of secred verss nsecred debt. The following proposition states the reslt. 12

15 Proposition 1: Assme that a firm, after being presented a profitable investment project,, is motivated to ndertake the investment based on maximizing secrity issance proceeds (minimizing its crrent cost of capital). Conditional on firm type being known to the lender, the G-qality firm will prefer to finance project with nsecred debt. A moderate B-qality firm will also prefer to finance the project with nsecred debt while a low B-qality firm will prefer to finance project with secred debt. Debt issance proceeds are always greater for the G-qality firm. Proof: See Appendix A. The intition for this reslt is straightforward. G-qality firms can obtain greater proceeds, or eqivalently a lower cost of debt capital, by offering its existing stock of high qality projects as collateral for the new nsecred loan. Doing so simltaneosly decreases the risk of defalt on the new issance and improves recovery conditional on defalt. This implies that the G-qality firm has higher ex ante debt capacity than the B-qality firm. In contrast, the existing stock of projects for the B-qality firm is of lower qality as compared to the new project. Pledging the existing stock as collateral for an nsecred loan is, conseqently, less advantageos as compared to the G-qality firm. For moderate B-qality firms, nsecred debt issance proceeds exceed secred debt issance proceeds even the thogh the risk of defalt with nsecred debt exceeds that associated with secred debt. This follows becase recovery proceeds conditional on defalt are high enogh in the joint High-Low defalt states to offset the increased risk of defalt. Low B-qality firms i.e., those firms with relatively lower recoveries in defalt states do not generate high enogh proceeds in defalt to offset the increased risk of defalt. Conseqently, the low B- qality firm can generate higher proceeds throgh a secred debt issance. Being able to ndertake an nsecred debt issance as described may depend on satisfying a participation constraint, stated as a bond covenant, that garantees existing creditors 13

16 they are no worse off as a reslt of a new investment-financing transaction. The following corollary states the reslt as it applies to firm preferences for nsecred debt. Corollary 1: Firms that prefer to isse nsecred debt will satisfy a participation constraint, shold it exist, that garantees that w creditors are no worse off as a reslt of the financing decision. Proof: See Appendix A. This corollary is really abot the risk associated with asset sbstittion. Very low qality firms wold like to engage in asset sbstittion, bt their revealed identity leads to secred debt issance that eliminates asset sbstittion as a concern to creditors. Higher qality firms, in contrast, do not water down their claim with addition of new collateral. They can therefore isse nsecred debt at an attractive price. Note that asset sbstittion in this case is not as traditionally characterized, with firms ndertaking a new high-risk negative NPV project, bt is instead abot bringing older assets to the table as potential collateral for financing a new project. Reslts stated in the above proposition and corollary assme that creditors can identify firm type prior to debt issance. Creditors cannot, in fact, differentiate between firms at time t 0. They do know that, conditional on firm identity being known at that time of issance, a G-qality firm wold prefer to finance investment in with nsecred debt, and that it will be given high issance proceeds relative to a B-qality firm. Bt they also know that a B-qality firm wold like to mimic the G-qality firm to obtain the same terms as a G-qality firm. Conseqently, withot some way to screen firms, nsecred debt to finance project will not be offered to any firm at terms available to the known G-qality firm. This will case financiers and the G-qality firm to explore mechanisms capable of revealing type. 14

17 Creditors of can attempt to screen borrowers by way of reqiring eqity issance coincident with nsecred debt issance. This imposes a threshold cost on the issing firm de to the costs of eqity issance. The qestion is whether firms are willing to isse eqity, and, if so, how mch and nder what conditions. Define eqity as a state contingent claim a credible and enforceable garantee made by the firm that pays off to the nsecred lender in some or all of the defalt states. This garantee is relatively low cost to the G-qality firm, since it has a lower probability of defalt (and eqity loses everything) than the B-qality firm. In contrast, eqity is clearly more costly to the B- qality firm that might wish to mimic the G-qality firm. The nsecred lender knows that the G-qality firm, if given the opportnity, will have an incentive to isse eqity in an attempt to separate itself from the B-qality firm. As a first step in this analysis, we will need to specify issance proceeds associated with nsecred debt in a pooling eqilibrim. This is becase B-qality firms can increase issance proceeds above their fll-information vale by mimicking the G-qality firm. Sccessfl mimicry will, in trn, case the nsecred creditors to pool. For the analysis that follows, we will benignly assme that G- and B-qality firms exist in eqal proportion to one another and that H + L is less than D +D w. Neither assmption is restrictive, bt rather is made to simplify the analysis. With this, we state nsecred debt issance proceeds nder pooling in the following lemma. Lemma 1: Unsecred debt issance proceeds in a pooling eqilibrim are P B Uns =(D + H +2 L )/4. These proceeds exceed issance proceeds available to the B-qality firm when its type is known, bt are less than proceeds available to a known G-qality firm. Proof: See Appendix A. 15

18 This lemma establishes that a B-qality firm wold like to try to mimic a G-qality firm, and that a G-qality firm wold like to try to separate itself from a B-qality firm. The feasibility of separation comes down to the fixed cost of eqity issance as it applies to the G-qality firm and the relative cost of issing a state-contingent eqity claim for the B-qality firm. Denote nsecred debt proceeds associated with a known G-qality firm as B G Uns and let E be the fixed cost of eqity issance. The following lemma establishes conditions nder which an nsecred debt issance pooling eqilibrim is realized. Lemma 2: If E < B G Uns B P Uns, nsecred creditors will consider screening firms by reqiring coincident eqity issance. Otherwise, if E B G Uns B P Uns nsecred creditors pool by offering P nsecred debt with proceeds B Uns with no eqity issance reqirement. Proof: See Appendix A. The pooling otcome is of corse less interesting to s, as we do not believe it describes how credit markets actally fnction, so going forward we assme that E < B G Uns B P Uns so that separation throgh eqity issance is a feasible screening mechanism. Now, sppose that eqity as a state-contingent claim has the defining characteristic of paying off to nsecred creditors in joint High-Low states. This implies that issing eqity creates no risk to the G-qality firm, since there is sfficient collateral vale in those states to flly repay nsecred creditors. In contrast, issing sch a claim does create risk for the B- qality firm, since existing collateral is insfficient to fnd fll debt repayment in those states. More specifically, for the B-qality firm there is a 0.5 cost of issing eqity for each nit of insrance provided to nsecred creditors. B-qality firms are willing to internalize this cost, in addition to the fixed cost of eqity issance, as long as the benefits to pooling exceed the associated costs. That is, the moderate B-qality firm will isse eqity as long as 16

19 P BUns B B Uns E.5E, where B B Uns is proceeds from an nsecred debt issance by a moderate B- qality firm and E denotes the face vale of eqity that pays off E nits to nsecred creditors in a joint High-Low state. In comparison, the low B-qality firm will isse eqity as long as P BUns B E.5E, where we recall that B denotes issance proceeds from secred debt. To make things interesting assme that the left-hand side of each of the previos two ineqalities is positive, implying that a non-trivial amont of eqity mst be issed by G-qality firms to indce separation. We are finally in a position to state the major reslt in this section. Proposition 2: Assme that fixed eqity issance costs are sfficiently low so that the G- and B- qality firms have incentives to isse eqity. For the G-qality firm to separate itself from the moderate B-qality firm, it finances investment in with nsecred debt and isses eqity shares P that exceed 2[ BUns B B Uns E ]. For the G-qality firm to separate itself from the low B-qality firm, it isses nsecred debt and more than 2[ B B E ] eqity shares. Proof: See Appendix A. P Uns Ths, when eqity issance costs are sfficiently low, nsecred creditors can screen firms by reqiring eqity issance in excess of the limits prescribed in Proposition 2. The G- qality firm will isse the eqity. If eqity issance costs were variable instead of fixed (reqiring only cosmetic changes in the model), the G-qality firm wold isse no more shares than necessary, since eqity issance at the margin wold be costly. In any case, the B-qality firm will not isse the reqired amont of eqity, since the increasing marginal costs to eqity issance are sch that the lower-risk firm is better off issing nsecred debt and no eqity while the higher-risk firm is better off issing secred debt and no eqity. The eqilibrim otcome is flly revealing and incentive compatible. Moreover, as shown in corollary 1, participation constraints for w creditors are satisfied (shold they exist) 17

20 when nsecred debt issance is optimal. This means that secrity issance choice is incentive compatible as it relates to satisfying asset sbstittion concerns of existing debt holders. 5 To the extent that issing eqity is a screening mechanism bt also introdces monitoring, one can characterize management of better qality firms as willing to commit themselves to increased market scrtiny in order separate themselves from lower qality firms and hence increase their market valation. Since G-qality firms will defalt less both becase of their inherent type and becase of the eqity cshion arising from screening, the cost of collateralization from sboptimal maintenance near distress shold be insignificant. Althogh not present in the model, the event of liqidation mst also determine the choice of debt. If the costs of liqidation increase with the nmber of different debt contracts and of debt holders, then only higher qality firms shold be able to isse nsecred debt, which is harder to disentangle and renegotiate than secred debt. II.B. Empirical Implications The model generates a rich set of empirical implications that can be tested with data. First, a separating eqilibrim implies that higher qality firms have latent vale and debt capacity that is revealed as a reslt of a joint issance choice-leverage financing decision. Creditors are able to identify higher qality firms type, which alleviates asset sbstittion concerns of existing debtholders. Latent vale, as smmarized by q, is cased by and is negatively associated with incremental increases in secred debt otstanding and overall leverage. 5 In a similar vein, it is straightforward to consider providing inside eqity holders control rights, in the sense of having to satisfy a stats qo participation constraint for existing eqity holders as related to financing with nsecred debt. Doing so may impose a cost on the w creditors, however, who might be forced to srrender some of their gains associated with investment and financing of. 18

21 Firms with liqid assets on their balance sheet have a commitment problem as it relates to facilitating new investment, as cash can easily be distribted to shareholders or otherwise diverted as a sorce of defalt insrance to creditors. Higher qality firms separate themselves from lower qality firms by maintaining lower total leverage, and ths higher levels of operating cash flow with which to service debt, to decrease credit risk. Ths, higher qality firms are predicted to have higher debt service coverage ratios. And to the extent that lower leverage and access to eqity markets relaxes financial constraints, we wold expect to observe higher rates of investment in the ftre by higher qality firms. 6 Finally, we wold also expect to see a high incidence of eqity issances per nit of investment for the higher qality firms. III. Data, Specification and Estimation Reslts III.A. Data Or sample consists of REITs identified from SNL Datasorce for the years from 1991 to These firms tend to concentrate their asset holdings by a particlar property type, and are classified as sch by SNL. The sample incldes residential, retail, office, and indstrial REITs, which constitte the for major property type classifications. 7 We inclde firms that report fllyear earnings in a particlar year within the sample period, and exclde firm-years for which secred debt exceeds the amont of total debt otstanding, which we infer as a reporting mistake. Table 4 reports smmary statistics for the variables sed in the estimation. As noted earlier, REITs are on average seen to operate at relatively high leverage ratios and to finance themselves with significant amonts of secred debt. In particlar, the average REIT has abot 6 In or model we assmed that proceeds from secred debt issance were sfficient to fnd investment. In reality, absent additional garantees or collateral, there are down payment reqirements with secred debt, implying that cash constraints may redce investment. 7 In percentages, or sample incldes 25% of residential REITs, 42% of retail REITs, 22% of office REITs, and 11% of indstrial REITs. 19

22 50 percent debt in its capital strctre as a percentage of market asset vale, of which jst less than two-thirds of the debt is secred. Frther note the significant variation in these measres across firms as seen most clearly by the 25 th and 75 th percentile statistics. Table 4 Here Other profitability measres and control variables are as follows. EarningsChanges is the ratio of next year earnings mins this year earnings to this year common book eqity; Size is book vale of total assets (measred in billions of 2006 dollars sing the Prodcer Price Index (PPI) pblished by the U.S. Department of Labor as the deflator); Profitability is the ratio of earnings before interest, taxes, depreciation, and amortization to the book vale of total assets; EarningsVolatility is the ratio of the standard deviation of earnings before interest, taxes, depreciation and amortization sing p to five years of consective observations to the average book vale of total assets estimated over the same time horizon; Age is the nmber of years since the IPO year measred as of the end of the sample period in As implied by or theory, we are particlarly crios abot how Tobin s q varies as a fnction of changes in capital strctre. Following the literatre, q is the ratio of market vale of total assets to book vale of total assets. As discssed by Gentry and Mayer (2008), Hartzell, Sn and Titman (2006) and Riddiogh and W (2009), REIT q vales are believed to provide more accrate and less noisy measres of marginal q than can generally be obtained with Compstat indstrial data (see also Erickson and Whited (2000)). We wold like to identify a single variable that jointly measres (1) the secrednsecred debt choice and (2) total leverage decision. The measre we employ is SecredMarket (SecredBook)Leverage, which is the ratio of secred pls mezzanine debt to the market (book) vale of total assets. This measre is a composite of SecredDebt, the ratio of secred pls 20

23 mezzanine debt to total liabilities pls mezzanine debt, and Market (Book)Leverage, the ratio of total liabilities pls mezzanine debt to the market (book) vale of total assets. Ths, in the context of or theory, this measre is hypothesized to be strictly decreasing in q as it reveals the qality characteristics of the firm. Table 5 displays pair-wise correlation coefficient estimates between and among secred debt and leverage ratio measres, as well as how these measres correlate with q. All debt ratio measres correlate negatively with q, with the exception of total book leverage. Or model sggests that the relevant measres are based on market vales, bt we are cognizant of concerns regarding co-movement in q and market leverage de to se of contemporaneos stock price in both measres. Observe the negative and significant correlations between secred leverage (market and book) and q, where secred leverage is or measre of the joint secred-nsecred debt choice and leverage decision. Also note the significant positive correlations between the secred debt-to-total debt ratio and total leverage, implying that firms which employ secred debt relative to nsecred debt also employ greater overall leverage. Table 5 Here Table 6 shows the components of secred and nsecred debt given the available data. Secred debt is composed of first mortgages, mezzanine debt (which are jnior mortgages that are issed together with a first mortgage) and secred bank lines of credit. We can see that first mortgages are by far the largest category of secred debt, bt also that mezzanine debt contribtes significantly to the total capital strctre of REITs. There is also significant variation within each secred debt category where, for example, many firms do not finance with mezzanine debt at all. Note that secred bank lines of credit are only abot 2 percent of the debt capital strctre, and that less than 25 percent of all firms tilize secred bank lines at all. 21

24 Table 6 Here Exclding the catch-all category of sbordinated debt and other liabilities, nsecred debt has two components: corporate-level debt and bank lines of credit. Corporate-level nsecred debt is seen to comprise abot three-forths of the total nsecred debt. It is interesting to note that abot 38 percent of REITs se secred mortgage debt to finance themselves bt do not se corporate-level nsecred debt, while less than 3 percent of all REITs that se corporate-level nsecred debt do not se secred debt to finance themselves. 8 III.C. Specification We will take a two-pronged approach to formally testing or theory of optimal capital strctre. First, we employ the matching methodology of Abadia and Imbens (2002) to assess the conseqences of financing choices on firm performance otcomes. At a given point in time, firms are classified based on whether they are above or below the median vale of or smmary measre of the joint secrity choice-leverage decision, defined as the ratio of secred debt to total assets. Firms below the median are hypothesized to be higher qality firms based on their revealed capital strctre decisions as related to or theory. Firms in each grop are matched based on observables, referred to as matching variables. An eqally weighted sqared distance measre is calclated based on matching variables, with minimm variance firms in the below-median grop matched to the relevant firm in the abovemedian grop. Performance of matched firms is tracked over time to assess model predictions. 8 Stlz and Johnson (1985) claim that secred debt provides protection against asset sbstittion when monitoring covenants is expensive, and also solves the nderinvestment problem. However, this implies that eqity holders are able to divert all valed added created by new projects from existing nsecred creditors, which wold not be possible in a firm with covenants that forbid the payment of dividends and where nsecred bond holders sccessflly engage in monitoring. Then, secring the debt of ftre investments wold not prevent existing bondholders sharing the benefits of the vale created by new investments. 22

25 We believe this methodology is apropriate, since or theory sggests that latent vale exists with better qality firms that is sbseqently revealed as a reslt of (persistent) joint secrity issance and leverage choices made by firms. One possible concern with the reslts from the Abadia-Imbens matching methodology is that debt type might be endogenos to ftre otcome. To mitigate this concern, we employ a GMM approach to estimate how the joint debt secrity issance-leverage decision affects firm vale as measred by Tobin s q. A GMM approach is attractive within or framework as a complement to the Abadia-Imbens matching methodology, since it handles simltaneity and estimation is cross-sectional with a steady-state interpretation. Tobin s q as a left-hand side variable has been sed by McConnell and Servaes (1995) and Agrawal and Knoeber (1996) to identify the determinants of optimal capital strctre. Bt we also note that capital strctre stdies starting from Rajan and Zingales (1995) consistently se q to explain leverage, which is why the GMM approach is sefl as it addresses endogeneity concerns. III.C. Estimation Reslts We will first report the matching estimation reslts. For each firm in the treated grop, we identify three REITs as a match in the control grop. Matching criteria are based on five matching variables: q, Size, Profitability, Earnings Volatility and Age. Table 7 displays smmary statistics for the five matching variables as they apply to the treatment and control samples, respectively. The key finding in Table 7 is that treatment and control firms do not differ in any significant way based on the matching variables. Table 7 Here 23

26 Tables 8 throgh 11 display reslts on a and 36-month forward looking basis as they apply to investment, profitability, firm valation and eqity issance, respectively. The reslts are clear and distinct. Firms with secred debt-to-total asset ratios that are below the median invest more, generate higher retrns, are valed more highly, and se a high proportion of eqity to finance investment. The initial and latter reslts sggest persistence and selfreinforcement in the gropings, implying that once a firm makes capital strctre choices that reveal type they tend to get stck in that category. Tables 8, 9, 10, 11 Here The differences are statistically and economically significant. For example, firms with lower-than-median secred debt-to-total asset ratios invest approximately for percent more per year and earn approximately two percent more per year on assets on a look-forward basis. Firm vale relative to initial total book assets for firms with lower-than-median secred debt-to-total asset ratios increase at a rate of approximately 3 percent fast than treatment grop, althogh the statistical significance is not qite as strong with this measre than the other two measres. We now offer reslts that are based on GMM estimation. In a GMM framework endogenos variables are estimated as part of a system in the cross section, in which all variables are first-differenced, implying that simltaneity as well as timing concerns with respect to q and financing decisions are addressed. Specifically, we regress changes in q and changes in the secred debt-to-total asset ratio (market as well as book) on each other, together with a set of appropriate control variables. Note that the earnings change variable is forward-looking, which goes to the isse of disentangling changes in q that are related to fndamentals verss nobservables. Estimation reslts are 24

27 displayed in Panels A and B in Table 12, where for comparison we report OLS estimation reslts in levels. Table 12 Here As seen in Panel A of the table, firms that make leverage-increasing secred debt financing choices generate lower q vales. The reslts are economically significant, in which a 10 percent increase in the secred debt ratio reslts in almost a 4 percent decrease in q as measred in market vale and almost a 3 percent decrease in q as measred by book vale. OLS estimation reslts are consistent with GMM estimation reslts, where the OLS reslts in levels sggest strong persistence in financing-indced firm valation otcomes. Panel B shows that increases in q simltaneosly case a decrease in the total secred leverage ratio, a reslt that is statistically significant in the market vale GMM regression. Together with reslts from Panel A, we infer a feedback that is consistent with implications of or model and the matching estimation reslts previosly reported, in which higher secred leverage cases lower q-vales that in trn case increases in the secred debt ratio. In other words, firms reveal themselves throgh their financing decisions as higher or lower qality firms, an effect that is intertemporally self-reinforcing. In Table 13 we report GMM estimation reslts where we now consider separately the component parts of the secred debt-to-total asset ratio: the ratios of secred debt to total debt and total debt to total assets. Prior to considering the reslts, recall the significant positive correlation that exists between the two component leverage measres. Strong positive correlation, in addition to being consistent with or theory, has the effect of biasing the relevant regression coefficients towards zero. Table 13 Here 25

28 Reslts show that, in the case of sing market leverage, total market leverage has a significantly negative effect on q, and, simltaneosly, q has a significantly negative effect on market leverage. The secred debt-to-total debt ratio is insignificant. In comparison, in the case of sing book leverage, secred debt-to-total debt has a significantly negative effect on q, and, simltaneosly q has a significantly negative effect on the secred debt-to-total debt ratio. We interpret these reslts as consistent and complementary with or matching methodology reslts as well as previos GMM specification, and spportive of the theory otlined earlier in the paper. As a final exercise, in Table 14 we compare allocation of income from operations to debt and eqityholders. Firms with below-median levels of secred debt-to-total assets are slightly more profitable, in the sense that they generate greater cash from operations as a percentage of assets in place. They are seen to commit to lower debt service obligations, and ths generate higher pre-dividend cash flow relative to above-median firms. These reslts show strong persistence over time. Bt then observe that the below-median firms pay this cash ot in the form of dividends to reslt in no difference in incremental cash directed to the balance sheet. Table 14 Here This is consistent with or model strctre that follows from Myers and Rajan s (1998) argment that cash has little commitment vale, and moreover that there is a significant opportnity cost to retaining cash on the balance sheet. Instead, higher qality firms redce debt service obligations relative to available cash flow from operations, which is the most credibly commitable cash flow available to mitigate defalt risk. Donaldson (1961) and Myers and Majlf (1984) pecking order of financing choices is consistent with this finding, in that firms with excess cash tend to pay down debt faster. 26