On the impact of financial distress on capital structure: The role of leverage dynamics

On the impact of financial distress on capital structure: The role of leverage dynamics Evangelos C. Charalambakis Susanne K. Espenlaub Ian Garrett Corresponding author. Manchester Business School, University of Manchester, UK, email: evangelos.charalambakis@mbs.ac.uk Manchester Business School, University of Manchester, UK, email: susanne.espenlaub@mbs.ac.uk Manchester Business School, University of Manchester, email: ian.garrett@mbs.ac.uk. We would like to thank Michael Brennan, John Graham, participants at the 2008 FMA Conference in Dallas and seminar participants at Manchester Business School for helpful comments. 1

On the impact of financial distress on capital structure: The role of leverage dynamics Abstract This paper uses a research design that addresses the endogenous effect of financial distress on debt ratios. Using a probability of financial distress derived from a hazard model as a measure of financial distress, we find that leverage dynamics are crucial in unraveling the true effect of financial distress on leverage. Our findings offer an explanation for the prior conflicting evidence on the association between leverage and financial distress and shed new light on the role of leverage dynamics in understanding capital structure decisions. We then show that firms balance the tax benefit of debt and financial distress costs when making financing decisions in a dynamic process. 2

1 Introduction While the theoretical underpinnings of capital structure suggest a negative association between financial distress costs and leverage, quantifying the impact of financial distress costs on debt ratios is difficult. Early empirical studies of capital structure (e.g., Kim and Sorensen (1986) and Titman and Wessels (1988)) use a firm s operating risk, measured as either the coefficient of variation or the standard deviation of earnings before interest and taxes (EBIT), to proxy for financial distress costs. These studies find no evidence of a negative relationship between financial distress costs and leverage. Several other studies that investigate the relationship between leverage and financial distress costs do so incorporating firm size as an inverse proxy for expected financial distress costs in their empirical specification (see, for example, Shyam- Sunder and Myers (1999), Fama and French (2002), and Flannery and Rangan (2006)). Even if firm size is a suitable proxy for financial distress costs, it is subject to criticism as it is likely to capture other firm characteristics, such as information asymmetry, access to public debt markets and the extent to which firms assets are diversified. 1 Recent studies have used measures relating to the likelihood of financial distress, most notably Altman s Z-score, or a modified version of it (see, for example, Graham (2000) and Byoun (2008)). Studies using Z-score typically find that a higher likelihood of financial distress, i.e., lower Z-score, leads to higher debt ratios, a finding that seems puzzling. There is an additional concern that financial distress costs are also endogenously related to debt ratios. Increasing leverage increases the probability of financial distress while an increase in the probability of financial distress should bring about decreases in the amount of debt a firm has in its capital structure. This issue of endogeneity has, to our knowledge, not been considered in the existing literature. Taken together prior studies that investigate the impact of financial distress on debt ratios have used either relatively poor proxies for financial distress, or misspecified models ignoring that financial distress is endogenous. In this paper we properly address the endogenous association between financial distress costs and debt ratios and provide new insights on how leverage dynamics affect corporate 1 Frank and Goyal (2003) argue that small firms are more likely to face information asymmetries as these firms are more likely to be younger and therefore less well known. Fama and French (2002) argue that larger firms have easier access to public debt markets as they find it more economical to produce the information required for public securities. Titman and Wessels (1988) argue that larger firms tend to be more diversified. 3

financing decisions. As financial distress costs cannot easily be observed, we assume that firms with a higher probability of financial distress are more likely to face higher financial distress costs. We use a probability-based estimate of financial distress rather than the Z- score that previous studies tend to use to test the association between leverage and financial distress costs. Specifically, we follow Shumway (2001) and estimate the probability of financial distress using a hazard model with time-varying covariates. We use predetermined equity market-driven variables to estimate the probability of financial distress, so that the estimated probability of financial distress at time t is based on variables dated t 1. An advantage of using the estimated probability of financial distress is that, unlike Z-score, the estimated probability of financial distress is based on a hazard model which accounts for how long the firm has survived before moving into financial distress and treats the estimation of the probability of financial distress as a dynamic multi-period problem that uses all available firmyear observations to estimate the probability. The Z-score is derived from a static bankruptcy prediction model. Unlike hazard models, static bankruptcy prediction models (e.g. see, Altman (1968) and Ohlson (1980)) are estimated only with each firm s last observation. As a result, static models produce inconsistent and biased estimates. 2 We perform our tests in a) a static framework which assumes that there are no adjustment costs when firms adjust toward their target capital structures and b) a dynamic framework in which leverage dynamics enter in the model, allowing for costly adjustment. We formulate our dynamic model using the first-differenced Arellano and Bond (1991) estimator and the twostep GMM-system estimator. These two econometric techniques are appropriate to estimate our dynamic model because they enable us to treat leverage dynamics and the probability of financial distress as endogenous variables. To test the effect of taxes on debt ratios we take into consideration the endogeneity of corporate tax status documented by Graham et al. (1998). Following Graham et al. (1998) we overcome the endogeneity problem by using a measure of corporate tax rate that is based on earning before interest deductions (before-financing). The results show that when leverage dynamics are excluded from the model, there is a positive relationship between the probability of financial distress and leverage. However, once we allow for leverage dynamics, the relationship between leverage and the probability 2 Shumway (2001) discusses this problem in detail. 4

of financial distress becomes negative and statistically significant, which reconciles with the prediction of the tradeoff theory. We also find, regardless of whether dynamics are included in the model, a positive relationship between corporate tax and leverage once we consider the endogeneity of corporate tax status, consistent with the finding of Graham et al. (1998). The only exception to this is for the effect of taxes on market leverage using the two-step Arellano- Bond GMM estimator in our dynamic empirical model. Based on this GMM estimator, we document that there is no association between market leverage and corporate taxes. Our findings have important implications for capital structure. The assumption that firms are optimally levered can lead to a misleading effect of financial distress on debt ratios. We show how important leverage dynamics are in corporate financing decisions. It is leverage dynamics in conjunction with the endogenous association between financial distress and debt ratios that generates a negative relationship between financial distress and debt ratios. Importantly, and unlike other studies, only when using a measure of the probability of financial distress and allowing for leverage dynamics do we find a negative relationship between financial distress and debt ratios. We then provide evidence that firms tradeoff between the tax advantage to debt and financial distress costs when we consider the role of corporate refinancing, i.e., leverage dynamics. This paper suggests that we need to focus on leverage dynamics to obtain a more comprehensive depiction of capital structure behavior. The rest of the paper is organized as follows. Section 2 motivates our empirical approach and describes the data. Section 3 presents and interprets the results while section 4 offers concluding remarks. 2 Model Specification 2.1 Target adjustment model Shyam-Sunder and Myers (1999), Flannery and Rangan (2006) and Huang and Ritter (2009), among others, have used a target adjustment model to investigate corporate financing behavior. Following previous studies we use a partial adjustment formulation which is described by 5

the following equation: Lev i,t = (1 λ)lev i,t 1 + λlev i,t + ϵ i,t (1) where Lev i,t is actual leverage for firm i in year t, Lev i,t 1 is actual leverage for firm i in year t 1, Levi,t is target leverage for firm i in year t, λ is the speed of adjustment towards target leverage, and ϵ i,t is an error term. The target is assumed to depend upon a vector of variables such that Levi,t = β x i,t 1. Substituting in (1) gives the basis of our empirical model: Lev i,t = (1 λ)lev i,t 1 + λ(β x i,t 1 ) + ϵ i,t (2) For (2) to be operational, we need to specify the variables that determine target leverage. According to the tradeoff theory, target leverage is determined by trading off the tax benefit of debt against financial distress costs. In addition to tax and distress variables, the other variables we use in specifying the target are well-established in the literature (see, for example, Rajan and Zingales (1995) and Frank and Goyal (2009).) In particular, we specify the following firm-specific factors: 1. Probability of Financial Distress (PROBFD) Following Shumway (2001) we estimate the probability that a firm will enter financial distress using market-driven predictors. 3 In particular, the probability of financial distress is the fitted value from the multi-period logistic regression P i,t = 1 1 + e ( α+β x i,t 1 ) (3) where P i,t is the probability that firm i will enter either bankruptcy or liquidation at time t and β x i,t 1 = β 1 REL SIZE i,t 1 + β 2 EXRET i,t 1 + β 3 σ i,t 1.The dependent variable is a dummy equalling zero if the firm has not filed for bankruptcy or entered liquidation. If the firm has entered liquidation or bankruptcy, then the dependent 3 Shumway (2001) also estimates the probability of bankruptcy using two additional accounting ratios, i.e, profitability and leverage. However, as book leverage and profitability are both incorporated in the dynamic panel data model we estimate, we choose not to include profitability and leverage in the estimation of the probability of financial distress as this could bias our main empirical specification described by (2) and our results. 6

variable equals one only for its last firm-year observation. REL SIZE is a firm s market capitalization expressed relative to the total market capitalization of NYSE and AMEX firms, EXRET is a firm s past return in excess of the market and σ i is firm i s stock return volatility. I calculate each firm s σ for a given year by regressing each stock s monthly returns in year t 1 on the value-weighted NYSE/AMEX index return in year t 1. The σ is the standard deviation of the residual of this regression. We expect there to be a negative relationship between the probability of financial distress and leverage. 2. Corporate Tax Rate Before Financing (CTRBF) CTRBF is a measure of the firm s corporate tax rate, which reflects before-financing decisions to address the endogeneity of corporate tax status associated with the debt ratio. It is calculated as income tax expense plus (interest expense the top statutory tax rate), divided by earnings before interest and tax. 4 Since we add back a proxy for the interest tax shield, i.e., interest expense the top statutory tax rate, to the income tax expense in the numerator and we use before-financing taxable income in the denominator CTRBF is exogenous to debt ratios. 5 Based on the prediction of the trade-off theory, we expect there to be a positive relationship between CTRBF and leverage. 3. Firm Size (SIZE) We define this as the natural logarithm of sales. Large firms are more profitable and hence have more tax benefits of debt. As large firms have more stable profit streams, they are less likely to go bankrupt. We therefore expect to see a positive relationship between firm size and leverage. 4. Tangibility (TANG) This is defined as fixed assets divided by total assets. If a firm has a large amount of fixed (tangible) assets then these assets can serve as collateral to debtholders. If debt is collateralized then the risk of the lender suffering agency costs of debt diminishes and 4 Following Sharpe and Nguyen (1995) and Graham et al. (1998), CTRBF is set to zero if the numerator is negative, and is set to one if the numerator is positive and the denominator is negative. 5 Interest expense divided by EBIT (interest coverage ratio) is inevitably a component of CTRBF. This could bias our measure of corporate tax rate. However, when we performed a correlation test between CTRBF and the interest coverage ratio, the correlation coefficient is low (0.04). 7

the firm s debt capacity increases. We therefore expect to see a positive relationship between tangibility and leverage. 5. Profitability (PROF) This is defined as earnings before interest, tax, depreciation and amortization (EBITDA) divided by total assets. More profitable firms are more likely to have accumulated retained earnings and thus have less incentive to issue debt. 6. Market to book (MTB) This is defined as the market value of assets divided by book value of assets. Market to book proxies for growth opportunities. Due to the agency costs of debt firms issue less debt to protect their investment opportunities; see Myers (1977). 7. Industry Leverage (IND LEV) This is defined as the industry median book leverage, based on four-digit SIC codes. This factor accounts for industry effects on leverage. MacKay and Phillips (2005) and Frank and Goyal (2009) find strong industry effects in the cross section of firms leverage. With regard to the definition of Lev, we use a book measure of leverage and a marketbased measure to assess the robustness of our results. Book leverage is defined as book value of debt divided by book value of debt plus stockholders equity. Market leverage is measured as book value of debt divided by book value of debt plus market value of equity. We provide more complete information about the definition of our variables in Appendix A. 2.2 Estimation of the target adjustment model Allowing for a lagged dependent variable to appear on the right hand side in (2) creates a dynamic panel data model. The error term ϵ i,t in (2) consists of two components: an unobserved firm-specific component η i and the residual component u i,t. An OLS-estimated coefficient on Lev i,t 1 would be upward biased as η i is correlated with u i,t. Including fixed effects in (5) to control for unobserved heterogeneity will also induce a bias in the coefficient on Lev i,t 1. This is because firm-specific effects are correlated with the lagged dependent variable; see, for example, Nickell (1981) and Baltagi (2008). Expressing all variables as 8

deviations from their firm-specific time-series means (time-demeaned variables) removes the time-invariant firm-specific effect. However, this simultaneously creates a correlation between the time-demeaned lagged dependent variable and the time-demeaned error term, introducing a bias in the dynamic panel data model. To obtain unbiased coefficient estimates for a dynamic panel data model similar to that described in (2) the econometric literature suggests various techniques. Anderson and Hsiao (1981) first-difference the dynamic model to eliminate the fixed effects and then use the second lag of the dependent variable as an instrument for the first-differenced lagged dependent variable. However, this technique cannot be applied to our case as it restricts the remaining independent variables to be strictly exogenous. Therefore, we cannot address the endogeneity of the probability of financial distress. Arellano and Bond (1991) first difference the dynamic panel data model applying a generalized method of moments (GMM) framework to develop valid instruments. In particular they use further lags of the endogenous variables as instruments for those independent variables that are endogenous, provided that the residuals have no second-order autocorrelation. Arellano and Bover (1995) and Blundell and Bond (1998) augment the Arellano-Bond estimator by using the lagged first differences of the exogenous independent variables in a non-transformed (level) equation. They build a system of two equations; the level equation as well as the first differenced one. This technique is widely known as GMM-system. Both the Arellano- Bond estimator and the GMM-system estimator are appropriate for our empirical model because they consider the probability of financial distress to be an endogenous variable. Flannery and Rangan (2006) use the fixed effects instrumental variables (IV) approach by instrumenting lagged book (market) leverage with lagged market (book) leverage for the dynamic book and market leverage regressions, respectively. Huang and Ritter (2009) use the long differencing technique to estimate a partial adjustment model of capital structure. However, both of these estimation methods are not suitable for the empirical specification described in (2) as they assume that apart from the lagged dependent variable all the other independent variables are strictly exogenous. We first use the two-step Arellano and Bond GMM first-difference estimator for our dy- 9

namic panel data model. We therefore estimate a first-differenced version of (2), Lev i,t = λ(β x i,t 1 ) λ Lev i,t 1 + ϵ i,t ϵ i,t 1 (4) We also estimate (2) using the two-step GMM-system technique. Our dynamic panel data model is estimated in both levels and first-differences. Based on the GMM-system method level equations are simultaneously estimated using lagged first differenced regressors as instruments. The GMM-system estimator increases the number of instruments and imposes additional moment restrictions, which can dramatically improve efficiency. We use the approach of Windmeijer (2005) to correct for the finite sample bias associated with the two-step first-differenced GMM and the two-step GMM-system estimators. 2.3 Data Our sample initially comprises active and inactive non-financial (SIC codes 6000 6999 are excluded) and non-utility (SIC codes 4900 4949 are excluded) firms traded on NYSE, AMEX and NASDAQ over the period 1963 2006. The accounting and market data are obtained from the CRSP/Compustat Merged Database. We obtain data on the top statutory tax rates from the Office of Tax Policy Research at the University of Michigan. We exclude firms with missing data for any variable. We restrict the sample to firms with available data for at least five consecutive years over the period 1963 2006 because of the use of the Arellano-Bond firstdifferenced GMM estimator and the GMM-system estimator. To estimate the probability of financial distress we need to identify which of the inactive firms were financially distressed. All inactive listed firms that entered any type of bankruptcy or liquidation are considered financially distressed. We identify from our sample 195 financially distressed firms between 1963 2006. The final sample consists of 6,901 firms with 98,583 firm-year observations. All the variables are winsorized at the upper and lower 0.5 tails except the corporate tax rate before-financing, market leverage, size and probability of financial distress. 6 Following Sharpe and Nguyen (1995) and Graham et al. (1998), we censor CTRBF to be bounded between zero and one. Table I presents some descriptive statistics for the variables. Prof- 6 We did not winsorize market leverage and size because descriptive statistics indicate that they are normally distributed, although the results remain unaltered if we also winsorize these two variables. 10

itability and the probability of financial distress (PROF and PROBFD, respectively) are the most volatile variables. 3 Results 3.1 A Static Model As a benchmark, we begin by estimating a static version of the model by setting λ = 1 in (2). The static model we estimate is Lev i,t = β 0 + β 1 P ROBF D i,t 1 + β 2 CT RBF i,t 1 + β 3 SIZE i,t 1 +β 4 T ANG i,t 1 + β 5 P ROF i,t 1 + β 6 MT B i,t 1 (5) +β 7 IND LEV i,t 1 + ϵ i,t 1 We estimate the static model using Tobit and fixed effects regression models. 7 In particular we use a double-censored Tobit estimator as the dependent variable is restricted to the range zero to one. We also use a Fixed effects estimator to control for unobserved sources of firm heterogeneity that are relatively constant over time. 8 Finally we perform regressions with clustered standard errors to account for cross-sectional and time-series dependence. Table II shows the regression results for book leverage. Size, tangibility, profitability, growth opportunities and median industry leverage are all significant and have the expected signs. The results are also robust to the method of estimation, so we will discuss the results as a whole. The results indicate that the coefficient on PROBFD is significantly positive, suggesting that as the probability of financial distress increases, so does leverage. This seems counter-intuitive and is inconsistent with the prediction of the tradeoff theory. On the other hand, we find a positive and significant association between corporate tax rate before-financing (CTRBF) and book leverage, consistent with the findings of Graham et al. (1998) and with the trade-off theory. To examine whether the results reported in Table II are due to the use of book leverage, Table III reports regression results for (5) with market leverage replacing book leverage as the dependent variable. As for book leverage, there is a significantly positive relationship 7 For the fixed effects regressions, ϵ i,t = η i + υ i,t in (5) where η i are the fixed effects. 8 We also estimated random-effects regressions. However, a Hausman specification test suggests that the fixed effects specification is most appropriate in estimating the static model. 11

between PROBFD and market leverage. We also document a positive association between CTRBF and market leverage, as we would expect. The effect of size, tangibility, profitability, growth opportunities and median industry leverage is the same as with the book leverage regressions. In the next section we explore whether the static model is misspecified leading to a positive bias on the coefficient of the probability of financial distress. 3.2 Enter Leverage Dynamics Several studies have documented that leverage dynamics are important in explaining capital structure empirically (see, for example, Leary and Roberts, 2005, Flannery and Rangan, 2006, and Byoun, 2008). To examine the effects of leverage dynamics on the results of the previous section, we relax the restriction that λ = 1 in (2). The model we estimate is Lev i,t = β 0 + β 1 Lev i,t 1 + β 2 CT RBF i,t 1 + β 3 P ROBF D i,t 1 + β 4 SIZE i,t 1 +β 5 T ANG i,t 1 + β 6 P ROF i,t 1 + β 7 MT B i,t 1 (6) +β 8 IND LEV i,t 1 + η i + υ i,t As mentioned earlier we estimate our dynamic panel data model using the two-step Arellano and Bond (A-B) first-differenced GMM estimator and the two-step GMM-system estimator. Tables IV and V report the results for the dynamic panel data model for book and market leverage, respectively. The estimated coefficient on lagged leverage is positive and statistically significant. The magnitude of this coefficient indicates that firms adjust rapidly toward their target capital structures. The GMM system estimator for book and market debt ratio produces lower speed of adjustment (36% and 27%, respectively) than that of the A-B GMM estimator for book and market debt ratio (45% and 34%, respectively). The most striking result, however, is the change in sign on the probability of financial distress, PROBFD using both the two-step GMM estimation methods. The presence of lagged leverage in the model now generates a significant negative relationship between financial distress and leverage. Using the two-step GMM-system estimator we find that the impact of CTRBF on leverage is positive and significant, in line with the prediction of the tradeoff theory. While we document a positive association between corporate tax rate and book leverage using the two-step A-B 12

GMM estimator, there is no association between the corporate tax rate and market leverage. In the dynamic model, irrespective of whether we use book leverage or market leverage, an increase in the probability of financial distress leads to a decrease in leverage, which is as we would expect. With respect to the GMM-system estimates, all the remaining firm-specific variables enter significantly and with the expected sign. While there is a negative sign on profitability and growth opportunities and a positive sign on industry leverage when we use the two-step A-B GMM estimator, there is no evidence on the effect of size and tangibility on leverage. We also report in Tables IV and V the AR(2) test statistic, that examines the null hypothesis of no second-order serial correlation in the error term, is statistically insignificant. Therefore, unlike Flannery and Rangan (2006), we show that there is no second-order correlation of υ it in the dynamic empirical specification using either the Arellano-Bond GMM estimator or the GMM-system estimator. Overall, our results shed new light on the role of leverage dynamics in corporate financing decisions. Our evidence shows that accounting for leverage dynamics as well as addressing properly the endogeneity of financial distress, allows us to uncover a negative relationship between distress and leverage, something which the existing literature has not been able to unravel. The positive effect of taxes on debt ratios and the negative effect of financial distress on debt ratios documented in Tables IV and V show that firms trade-off the tax benefit of debt against financial distress costs when determining their financing policy. 4 Summary and conclusions In this paper we thoroughly estimate the impact of financial distress costs on debt ratios addressing properly the endogenous association between financial distress and leverage. In contrast to previous studies, we use the estimated probability of financial distress derived from a hazard model based on equity market-driven variables as a measure of financial distress. To measure the tax impact on capital structure we use a before-financing measure of the corporate tax rate to overcome the endogenous association between corporate tax rates and debt ratios. We use a static and a dynamic model of capital structure to quantify the effect of financial distress costs on debt ratios investigating the role that leverage dynamics play. 13

The static model assumes that firms adjust immediately toward their target capital structures whereas the dynamic model includes leverage dynamics allowing for costly adjustment. We estimate the dynamic model using the two-step first-differenced GMM estimator and the twostep GMM-system estimator, which enable us to account for the endogenous effect of financial distress on debt ratios. When there are no leverage dynamics in the model, we find that an increase in the probability of financial distress increases leverage. This result is inconsistent with what theory of capital structure suggests. Using a dynamic model that accounts for leverage dynamics we find that leverage decreases with the probability of financial distress as predicted by the trade-off theory. We find a positive relationship between the before-financing corporate tax rate and leverage irrespective of whether we allow for leverage dynamics in the regression model. When we use the first-differenced GMM estimator for our dynamic empirical specification there is no association between corporate tax rates and market leverage. Our results show that only when accounting leverage dynamics and considering the endogeneity of financial distress do we provide evidence that firms with higher probability of financial distress issue less debt, in line with the tradeoff theory. This finding provides new evidence on how leverage dynamics affects corporate financing decisions. Leverage dynamics enables us to understand the endogenous association between financial distress and debt ratios and therefore to estimate properly the effect of the probability of financial distress on leverage. Having thoroughly examined the association between debt ratios and financial distress costs, we show that firms tradeoff between the tax benefit of debt and financial distress costs taking into account that firms rebalance their capital structures over time. This paper suggests that the role of leverage dynamics is important in order to understand how firms make their financing decisions. References Altman, E., 1968, Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy, Journal of Finance 23, 589 609. Anderson, T.W. and C. Hsiao, 1981, Estimation of dynamic models with error components, Journal of the American Statistical Association 76, 589 606. 14

Antoniou, A. Y.Guney, and K.Paudyal, 2008, The Determinants of Capital Structure: Capital Market Oriented versus Bank Oriented Institutions, Journal of Financeial and Quantitative Analysis 43, 59 92. Arellano, M. and S.R. Bond, 1991, Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations, Review of Economic Studies 58, 233 257. Arellano, M. and O. Bover, 1995, Another Look at Instrumental-Variable Estimation of Error- Components Models, Journal of Econometrics 68, 29 52. Baltagi, B.H., 2008, Econometric Analysis of Panel Data, New York, John Wiley and Sons. Blundell, R. and S. Bond, 1998, Initial conditions and moment restrictions in dynamic panel data models, Journal of Econometrics 87, 115 143. Booth, L., v. Aivazian, A. Demirguc-Kunt, and M. Maksimovic, 2001, Capital Structures in Developing Countries, Journal of Finance 56, 87 130. Bun, M. and J.F. Kiviet, 2006, The Effects of Dynamic Feedbacks on LS and MM Estimator Accuracy in Panel Data Models, Journal of Econometrics 132, 409 444. Byoun, S., 2008, How and When Do Firms Adjust Their Capital Structures toward Targets? Journal of Finance 63, 3069 3096. Fama, E., 1985, What s Different about Banks? Journal of Monetary Economics 15, 29 37. Fama, E. and K.R. French, 2002, Testing Trade-Off and Pecking Order Predictions about Dividends and Debt, Review of Financial Studies 15, 1 33. Flannery, M. and K. Rangan, 2006, Partial Adjustment towards Target Capital Structures, Journal of Financial Economics 79, 469 506. Frank, M.Z. and V.K. Goyal, 2003, Testing the pecking order theory of capital structure, Journal of Financial Economics 67, 217 248. Frank, M.Z. and V.K. Goyal, 2009, Capital Structure Decisions: Which Factors are Reliably Important? Financial Management 38, 1 37. 15

Graham, J.R., M. Lemmon, and J. Schallheim, 1998, Debt, Leases and the Endogeneity of Corporate Tax Status, Journal of Finance 53, 131 162. Graham, J.R., 2000, How Big Are the Tax Benefits of Debt? Journal of Finance 55, 1901 1941. Hahn, J., J. Hausman, and G. Kuersteiner, 2007, Long Difference Instrumental Variables Estimation for Dynamic Panel Models with Fixed Effects, Journal of Econometrics 140, 574 617. Huang, R. and J.R. Ritter, 2009, Testing Theories of Capital Structure and Estimating the Speed of Adjustment, Journal of Financial and Quantitative Analysis 44, 237 271. Kim, W. and E. Sorensen, 1986, Evidence on the Impact of the Agency Costs of Debt on Corporate Debt Policy, Journal of Financial and Quantitative Analysis 21, 131 144. Leary, M.T. and M.R. Roberts, 2005, Do Firms Rebalance Their Capital Structures? Journal of Finance 60, 2575 2619. MacKay, P. and G.M. Phillips, 2005, How Does Industry Affect Firm Financial Structure? Review of Financial Studies 18, 1433 1466. Myers, S.C., 1977, Determinants of Corporate Borrowing, Journal of Financial Economics 5, 147 175. Nickell, S., 1981, Biases in Dynamic Models with Fixed Effects, Econometrica 49, 1417 1426. Rajan, R.G. and L. Zingales, 1995, What Do We Know about Capital Structure: Some evidence from International Data Journal of Finance 50, 1421 1460. Ohlson, J.A., 1968, Financial Ratios and the Probabilistic Prediction of Bankruptcy, Journal of Accounting Research 18, 109 131. Sharpe, S.A. and H.H. Nguyen, 1995, Capital Market Imperfections and the Incentive to Lease, Journal of Financial Economics 39, 271 294. 16

Shumway, T., 2001, Forecasting Bankruptcy More Accurately: A Simple Hazard Model, Journal of Business 74, 101 124. Shyam-Sunder, L. and S.C. Myers, 1999, Testing Static Tradeoff against Pecking Order Models of Capital Structure, Journal of Financial Economics 51, 219 244. Titman, S. and R. Wessels, 1988, The Determinants of Capital Structure Choice, Journal of Finance 43, 1 21. Windmeijer, F., 2005, A Finite Sample Correction for the Variance of Linear Efficient Two- Step GMM Estimators, Journal of Econometrics 126, 25 51. 17

Table I Summary Statistics The table presents the summary statistics of the variables used in the study. We exclude firm-years in which the firm has missing data. The sample contains 6,901 firms and 98,583 firm-year observations from 1963 2006. We censor the corporate tax rate before-financing (CTRBF) to be bounded between zero and one. All the remaining variables except market leverage, size and the probability of financial distress are winsorized at the 0.5 th and 99.5 th percentiles. Book leverage is book value of debt divided by book value of debt plus book value of stockholders equity. Market leverage is book value of debt divided by book value of debt plus market value of equity. Lagged book leverage is the book leverage in year t-1. Lagged market leverage is market leverage in year t 1. The before-financing tax rate, CTRBF, is measured as total income tax plus interest expense multiplied by the top statutory tax rate, all divided by earnings before interest and tax (EBIT). PROBFD is the estimated probability of financial distress. SIZE is defined as the natural logarithm of net sales. TANG is the ratio of fixed assets to total assets. PROF is measured as earnings before tax,interest, depreciation and amortization divided by total assets. MTB is the market value of assets divided by the book value of assets. IND LEV is the median industry book leverage based on the SIC four-digit code. Variable Mean Median Std.dev Min Max Book Leverage t 1 0.3264 0.3006 0.2748 0.0000 1.9163 Market Leverage t 1 0.2648 0.2081 0.2398 0.0000 0.9985 PROBFD t 1 0.0036 0.0024 0.0040 0.0001 0.1127 CTRBF t 1 0.3832 0.3826 0.2801 0.0000 1.0000 SIZE t 1 5.2759 5.2912 2.1632-6.9078 12.5918 TANG t 1 0.3187 0.2733 0.2172 0.0016 0.9267 PROF t 1 0.0045 0.0430 0.1896-1.7743 0.2885 MTB t 1 1.7476 1.2683 1.5796 0.4883 15.5373 IND LEV t 1 0.2963 0.2998 0.1624 0.0011 0.7974 18

Table II Tobit Fixed Effects and Clustered Regressions, Book Leverage This table contains results from a static model of capital structure using censored Tobit, Fixed effects and clustered regressions. The dependent variable is book leverage which is book value of debt divided by book value of debt plus book value of stockholders equity. The sample consists of 98,583 observations from 1963 2006. PROBFD is the estimated probability of financial distress. The before-financing tax rate, CTRBF, is measured as total income tax plus interest expense multiplied by the top statutory tax rate, all divided by earnings before interest and tax (EBIT). SIZE is defined as the natural logarithm of net sales. TANG is the ratio of fixed assets to total assets. PROF is measured as earnings before tax,interest, depreciation and amortization divided by total assets. MTB is the market value of assets divided by the book value of assets. IND LEV is the median industry book leverage based on the SIC four-digit code. The regression is estimated using a Tobit model censoring at zero at the lower end and one at the upper end with robust standard errors, a Fixed effects(fe) and a clustered model. The estimated model is: Lev i,t = α+β 1 P ROBF D i,t 1 +β 2 CT RBF i,t 1 + β 3 SIZE i,t 1 +β 4 T ANG i,t 1 +β 5 P ROF i,t 1 +β 6 Market to book i,t 1 +β 7 IND LEV i,t 1 +ϵ it. ***, ** and * denote significance at the 1, 5 and 10 percent levels, respectively. Dependent Variable=Book leverage Censored Tobit FE Clustered Constant -0.0873-0.0040-0.0473 (-22.57) (-0.61) (-3.51) PROBFD t 1 6.4291 3.3649 6.9202 (26.85) (14.82) (6.82) CTRBF t 1 0.1233 0.0509 0.1134 (42.11) (18.94) (8.92) SIZE t 1 0.0256 0.0319 0.0235 (55.63) 32.57 (11.08) TANG t 1 0.1204 0.1100 0.0953 (31.44) (14.62) (8.22) PROF t 1-0.2402-0.2329-0.2812 (-50.13) (-47.79) (-11.14) MTB t 1-0.0180-0.0051-0.0107 (-32.35) (-8.66) (-5.45) IND LEV t 1 0.6290 0.3810 0.6011 (118.30) (52.92) (34.52) Number of observations 98,583 98,583 98,583 19

Table III Tobit Fixed Effects and Clustered Regressions, Market Leverage This table contains results from a static model of capital structure. The dependent variable is market leverage which is book value of debt divided by book value of debt plus market value of equity. The sample consists of 98,583 observations from 1963 2006. PROBFD is the estimated probability of financial distress. The before-financing tax rate, CTRBF, is measured as total income tax plus interest expense multiplied by the top statutory tax rate, all divided by earnings before interest and tax (EBIT). SIZE is defined as the natural logarithm of net sales. TANG is the ratio of fixed assets to total assets. PROF is measured as earnings before tax,interest, depreciation and amortization divided by total assets. MTB is the market value of assets divided by the book value of assets. IND LEV is the median industry book leverage based on the SIC four-digit code. The regression is estimated using a Tobit model censoring at zero at the lower end and one at the upper end with robust standard errors and a Fixed effects(fe) model and a clustered model. The estimated model is Lev i,t = α+β 1 P ROBF D i,t 1 +β 2 CT RBF i,t 1 +β 3 SIZE i,t 1 +β 4 T ANG i,t 1 +β 5 P ROF i,t 1 + β 6 Market to book i,t 1 + β 7 IND LEV i,t 1 + ϵ it. ***, ** and * denote significance at the 1, 5 and 10 percent levels respectively. Dependent Variable=Market leverage Censored Tobit FE Clustered Constant -0.0304-0.0142 0.0200 (-8.81) (-2.93) (1.85) PROBFD t 1 7.1987 3.1057 6.7723 (33.90) (18.43) (8.94) CTRBF t 1 0.1050 0.0415 0.0821 (40.21) (20.79) (10.52) SIZE t 1 0.0197 0.0287 0.0157 (48.13) (39.46) (11.28) TANG t 1 0.0898 0.1439 0.0690 (26.36) (25.77) (6.05) PROF t 1-0.1786-0.1261-0.1577 (-41.72) (-34.85) (-7.09) MTB t 1-0.0489-0.0209-0.0397 (-94.69) (-47.74) (-12.17) IND LEV t 1 0.5761 0.3215 0.5315 (121.84) (60.15) (28.33) Number of observations 98,583 98,583 98,583 20

Table IV Two-step GMM Estimation Results, Book Leverage This table presents the results from a dynamic model of capital structure. We use firm-year observations with available data for at least five consecutive years. The dependent variable is book leverage which is book value of debt divided by book value of debt plus book value of stockholders equity. Lagged book leverage is the book leverage in year t-1. PROBFD is the estimated probability of financial distress. The before-financing tax rate, CTRBF, is measured as total income tax plus interest expense multiplied by the top statutory tax rate, all divided by earnings before interest and tax (EBIT). SIZE is defined as the natural logarithm of net sales. TANG is the ratio of fixed assets to total assets. PROF is measured as earnings before tax,interest, depreciation and amortization divided by total assets. MTB is the market value of assets divided by the book value of assets. IND LEV is the median industry book leverage based on the SIC four-digit code. The dynamic panel data model we estimate is of the following form: Lev i,t = α + β 1 Lev i,t 1 + β 2 P ROBF D i,t 1 + β 3 CT RBF i,t 1 + β 4 SIZE i,t 1 + β 5 T ANG i,t 1 +β 6 P ROF i,t 1 +β 7 Market to book i,t 1 +β 8 IND LEV i,t 1 +η i +η t +υ i,t. We also include year dummies (not reported). The model is estimated using the two-step Arellano- Bond GMM estimator and the two-step GMM-system estimator, which treat PROBFD as an endogenous variable. For the two-step GMM estimation methods we apply Windmeijer s correction to the standard errors. ***, ** and * denote significance at the 1, 5 and 10 percent levels respectively. Dependent Variable=Book leverage A-B GMM Estimator GMM-system Estimator Book Leverage t 1 0.5509 0.6364 (17.73) (29.48) PROBFD t 1-4.6159-4.0340 (-6.87) (-6.59) CTRBF t 1 0.1110 0.1663 (5.14) (11.93) SIZE t 1 0.0053 0.0084 (0.59) (7.53) TANG t 1 0.0650 0.0685 (1.25) (7.70) PROF t 1-0.3341-0.2711 (-5.64) (-11.37) MTB t 1-0.0106-0.0215 (-2.89) (-13.86) IND LEV t 1 0.0441 0.1052 (2.04) (6.45) Number of observations 84,594 91,495 AR(1) -16.87-17.36 AR(2) 1.32 1.46 21

Table V Two-step GMM Estimation Results, Market Leverage This table presents the results from a dynamic model of capital structure. We use firmyear observations with available data for at least five consecutive years. The dependent variable is market leverage which is book value of debt divided by book value of debt plus market value of equity. PROBFD is the estimated probability of financial distress. The before-financing tax rate, CTRBF, is measured as total income tax plus interest expense multiplied by the top statutory tax rate, all divided by earnings before interest and tax (EBIT). SIZE is defined as the natural logarithm of net sales. TANG is the ratio of fixed assets to total assets. PROF is measured as earnings before tax,interest, depreciation and amortization divided by total assets. MTB is the market value of assets divided by the book value of assets. IND LEV is the median industry book leverage based on the SIC four-digit code. The The dynamic panel data model we estimate is of the following form: Lev i,t = α + β 1 Lev i,t 1 + β 2 CT RBF i,t 1 + β 3 P ROBF D i,t 1 + β 4 SIZE i,t 1 + β 5 T ANG i,t 1 + β 6 P ROF i,t 1 +β 7 Market to book i,t 1 +β 8 IND LEV i,t 1 +η i +η t +υ i,t. We also include year dummies (not reported). The model is estimated using the two-step Arellano-Bond GMM estimator and the two-step GMM-system estimator, which treat PROBFD as an endogenous variable. For the two-step GMM estimation methods we apply Windmeijer s correction to the standard errors. ***, ** and * denote significance at the 1, 5 and 10 percent levels respectively. Dependent Variable=Market leverage A-B GMM Estimator GMM-system Estimator Market Leverage t 1 0.6606 0.7264 (33.64) (64.19) PROBFD t 1-3.0821-3.5281 (-5.57) (-7.84) CTRBF t 1 0.0139 0.0661 (0.79) (7.81) SIZE t 1 0.0082 0.0019 (1.17) (2.96) TANG t 1 0.0428 0.0299 (1.10) (5.75) PROF t 1-0.0838-0.1906 (-2.39) (-16.56) MTB t 1-0.0304-0.0271 (-11.58) (-20.79) IND LEV t 1 0.0673 0.0988 (3.63) (7.31) Number of observations 84,594 91,495 AR(1) -38.96-42.09 AR(2) -1.25-1.47 22

Appendix A Table A1 Definition of Variables This appendix defines the variables used in the study. All numbers in parentheses refer to the Compustat code of each accounting item. Variable Name Variable definition Total Debt Debt in Current Liabilities (34) + Long term Debt (9) Book Leverage Total Debt Total Debt + Stockholders Equity (216) Total debt Debt in current liabilities (34) + Long term debt (9) Book Leverage Total debt Total debt + Stockholders equity (216) Market value of equity Stock Price (199) Shares outstanding (54) Market Leverage Total debt Total debt + Market value of equity (mcap) EBIT Pretax income (170) + Interest expense (15) CTRBF PROBFD (Income tax (16) + (Interest expense Top Statutory Tax Rate)) EBIT Estimated probability of financial distress from a hazard model Working Capital Size Tangibility Profitability Market to book Ind LEV Current assets(4) Current liabilities(5) Natural logarithm of net sales, where net sales are deflated by the GDP Property, plant and equipment (8) Book value of assets (6) Operating income before depreciation (13) Book value of assets Book value of assets Common equity (60) + Market value of equity Book value of assets the median industry book leverage, based on the SIC four-digit code 23