Debt Structure and Firm Leverage Adjustments

Transcription

1 Debt Structure and Firm Leverage Adjustments Kai Chen, Wenlian Gao, and Feifei Zhu This Version: September 2017 Authors contact information: Chen, SUNY Oneonta. Gao, Chinese Academy of Finance and Development, Central University of Finance and Economics. Zhu, Radford University.

2 Debt Structure and Firm Leverage Adjustments ABSTRACT This paper examines the impact of debt structure characteristics on firm leverage adjustments. Results show that debt maturity impedes leverage adjustments whereas debt concentration accelerates adjustments, suggesting that both the interest conflicts between debtholders and equity holders and those among different groups of debtholders have implications for the dynamics of leverage. Their effect is more notable in overlevered firms and in the period of financial crisis. Further evidence indicates that the negative impact of debt maturity is stronger in firms with more public debt and lower in firms with more bank loans. Keywords: Dynamics of Leverage, Debt Maturity, Debt Concentration, Bank Loan, Public Debt JEL Classification Number: G12, G14, G32

3 1 1 Introduction The Trade-off theory maintains that firms maximize the wealth of their shareholders when their capital structure reaches the optimal level via a trade-off of tax benefits against financial distress costs of debt. Consequently, firms take positive steps to offset deviations from their optimal debt ratio. However, the literature suggests slow and incomplete leverage rebalancing. To examine the determinants of leverage adjustment speed, the conventional literature has focused on market frictions, such as equity issuance costs (Altinkili and Hansen, 2000) and transactions costs (Fischer, Heinkel, and Zechner 1989; Korajczyk and Levy, 2003; Strebulaev, 2007; Shivdasani and Stefanescu, 2010). More recent studies have proposed that other factors including cash flow realizations (Faulkender, Flannery, Hankins, and Smith (2012)), institutional and macroeconomic factors (Oztekin and Flannery, 2012; Oztekin forthcoming), and agency conflicts (Morellec, Nikolov, and Schurhoff, 2012; Liao, Mukherjee, and Wang, 2015) also play a role in capital structure rebalancing. Our study compliments this literature by focusing on the characteristics of corporate debt structure, i.e., debt maturity and debt concentration, and investigating how they may shape the capital structure dynamics. Debt maturity has been argued to play a role in investment policies and value maximization as it captures the conflicts of interest between debtholders and equity holders (Myers,1977). The debtholder/equity holder conflicts not only influence the level of target debt ratio but also how debt ratios evolve over time (Titman and Tsyplakov, 2007; Dangl and Zechner, 2007). If a firm is primarily financed with long-term debt, then replacing debt with equity would always lead to a large wealth transfer to the remaining debtholders so that new debt is used to refinance maturing debt. In contrast, firms with short debt may find it optimal to replace maturing debt with equity, especially when their profitability drops (Hovakimian, Opler, and Titman, 2001; Dangl and Zechner, 2007). Thus, long debt maturities are expected to be an impediment on debt adjustment. On the other hand, debt concentration reflects the conflicts among different groups

4 2 of creditors that may also affect capital structure (Welch, 1997; Bris and Welch, 2005; Hackbarth and Mauer, 2012). Particularly, a concentrated debt structure helps minimize expected bankruptcy costs and economizes on information collection and monitoring costs (Paolo et al., 2013). It has been explicitly formalized that debt concentration lowers negotiation costs (Bolton and Scharfstein, 1996) and increases the speed of restructuring under Chapter 11 and lowers the likelihood of liquidation (Ivashina, Iverson, and Smith, 2011). Creditors, including both relational lenders and bond holders have incentives to monitor as long as they have a sufficiently large claim in the firm. 1 Consequently, debt concentration should have a positive effect on leverage adjustment. We obtain debt structure information from the database of Capital IQ, an affiliate of Standard and Poor s. It decomposes a firms debt into seven categories, i.e., commercial paper, drawn credit lines (also known as revolving credit facilities), term loans, senior and subordinated bonds and notes, and capital leases. We measure debt maturity as the weighted average maturity of all types of debt used by a firm. Following Colla, Ippolito, and Li (2013), we measure debt concentration using a normalized Herfindahl-Hirschman Index of a firm s debt type usage and a dummy variable which equals one if a firm obtains at least 90% of its debt from one debt type. Results show that debt maturity impedes capital structure adjustment speed whereas debt concentration expedites the adjustment speed. Their effect is more notable in overlevered firms and in the period of financial crisis. The results remain significant after controlling for the agency conflicts between firm management and shareholders. We further examine how the effect of debt maturity interacts with a firm s debt type, i.e., bank loan versus public debt. Relational lenders have access to non-public information at low cost (Nakamura, 1993) and provide more intense monitoring on firm management (Holstrom and Tirole, 1997; Repullo and Suarez, 1999). In contrast, information acquisition is more costly for arm s-length bondholders and it is less efficient for multiple 1 See Park (2000), Sufi (2007), Nini, Smith, and Sufi (2009), and Roberts and Sufi (2009) for bank monitoring and Jiang, Li and Wang (2012) and Li and Wang (2015) for bondholder monitoring.

5 3 individual bondholders to monitor a firm due to free-rider prolems. (Diamond, 1984; Rajan, 1992). Consistent with our projection, the results show that the negative effect of debt maturity on adjustment has been mitigated in firms with more bank loans and been exacerbated in firms with more public debt. The findings of our study have important implications for the capital structure dynamics literature. A series of recent papers have been devoted to examine the determinants of leverage dynamics (See Altinkili and Hansen (2000) for equity issuance costs, Fischer, Heinkel, and Zechner (1989), Korajczyk and Levy (2003), Strebulaev (2007) and Shivdasani and Stefanescu (2010) for transactions costs, Faulkender, Flannery, Hankins, and Smith (2012) for cash flow realizations, and Oztekin and Flannery (2012) and Oztekin (forthcoming) for institutional and macroeconomic factors) and our work extends this literature by showing the material effect of debt structure characteristics on leverage dynamics. This piece of evidence suggests that, in addition to the well documented agency conflicts between firm management and shareholders (e.g., Morellec, Nikolov, and Schurhoff, 2012; Liao, Mukherjee, and Wang, 2015), the interest conflicts between equity holders and debt holders and those among different groups of debt holders have important implications in shaping capital structure decisions. The remaining of the paper is structured as follows. Section 2 reviews the related literature and develops hypotheses. Section 3 describes the empirical approach and the data. Sections 4 examines the impacts of debt structure characteristics, i.e., debt maturity and debt concentration, on firm leverage adjustments. Section 5 discusses how these impacts vary with firms use of bank loan and public debt. The final section concludes the paper.

6 4 2 Hypothesis Development 2.1 Debt Maturity and Leverage Adjustments The role of maturity in firms investment policies and thus value maximization can be traced back to Myers (1977). He proposes a concept of debt overhang which refers that risky debt that matures in the future leads to underinvestment today. The insight is that part of the cash flows generated by investment goes to debt holders at maturity, and unfortunately the equity holders who make the investment decision will not internalize this benefit. Thus, the core of debt overhang is wealth transfer from equity holders to debt holders. He therefore suggests the solution of short-term debt to the debt overhang problem, because if all debt matures before the investment opportunity, the firm can make the investment decision as if an all-equity firm. Titman and Tsyplakov (2007) further argue that the conflicts of interest between debt holders and equity holders have been argued to have a first order effect not only on the level of target debt ratio but also on how debt ratios evolve over time. They find that debt holder/equity holder conflicts reduce the tendency of firms to move towards their target debt ratios. Along the same line, Dangl and Zechner (2007) concentrate on the role of debt maturity to mitigate the adverse effects of conflicts of interest between debtholders and equityholders on capital structure dynamics. For firms financed primarily with longterm debt, the wealth transfer is much larger. They find that equityholders never reduce leverage if the debt maturities are sufficiently long. In this case, new debt is used to refinance maturing debt as replacing maturing debt with equity would always lead to a sufficiently large wealth transfer to the remaining debtholders. On the other hand, they find that sufficiently short debt maturities make it optimal for equityholders to replace maturing debt with equity, especially when the firms profitability drops. Moreover, short debt maturities require the firm to refund a large fraction of its debt over any period of time and thus lead to more pronounced debt reductions. Thus, we expect

7 5 H1: The average maturity of firm debt has a negative effect on the speed of leverage adjustments. 2.2 Debt Concentration and Leverage Adjustments In addition to the conflicts between shareholders and debt holders, the literature suggests conflicts among different groups of debt holders may also affect capital structure (Welch (1997), Bris and Welch (2005), Hackbarth and Mauer (2012)). Especially, a high degree of debt concentration helps minimize expected bankruptcy costs and economizes on information collection and monitoring costs (Paolo et al., 2013). Bolton and Scharfstein (1996) formalize the idea that debt concentration lowers negotiation costs by demonstrating that borrowing from a single creditor maximizes liquidation value for firms with low-credit quality. They also find that debt concentration facilitates bargaining with an outside buyer for firms with easily redeployable assets, i.e., assets with a higher value to another firm in the industry or some other investor who can manage the assets. Furthermore, higher creditor concentration is associated with better restructuring outcomes, increasing the speed of restructuring under Chapter 11 and reducing the likelihood of liquidation (Ivashina, Iverson, and Smith, 2011). Prior literature also suggests the information collection and monitoring costs will be minimized if the firm has a concentrated debt structure. Creditors, including both relational lenders and bond holders have incentives to monitor as long as they have a sufficiently large claim in the firm. For relational lending, Park (2000) shows that the optimal debt contract delegates monitoring to a single senior lender and that seniority allows the monitoring senior lender to appropriate the full return from his monitoring activities. Employing data on syndicated loans and on the composition of lending syndicates, Sufi (2007) shows that the lead bank in a lending syndicate retains a larger share of the loan and forms a more concentrated syndicate when the borrowing firm requires more intense monitoring and due diligence. Nini, Smith, and Sufi (2009) document the impor-

8 6 tance of bank control by focusing on explicit contractual restrictions on firm investment policy contained in bank credit agreements. Roberts and Sufi (2009) examine covenant violations and find that creditors respond to the violation by reducing the credit facility, increasing the interest spread, or demanding additional collateral, which leads to a large and persistent decline in the flow of debt and a corresponding decline in leverage ratios. Bondholders influence corporate decisions especially when there is violation of covenants or when a firm enters Chapter 11. Jiang, Li and Wang (2012) document that the presence of hedge fund unsecured creditors increases the likelihood of a successful reorganization, which is usually associated with a both higher debt recovery and a more positive stock market response at the time of a bankruptcy filing. Li and Wang (2015) show that hedge fund and private equity fund of Chapter 11 firms strategically target firms to enforce power and exert influence by seeking board representation and improving corporate governance and the operating performance of emerged firms. Given the lower bankruptcy costs and monitoring costs associated with concentrated debt structure, we conjecture that firms with more concentrated debt structure tend to adjust faster towards their optimal leverage, especially for firms that are underlevered. H2: The degree of debt concentration has a positive effect on the speed of leverage adjustments. 2.3 Joint Effect with Debt Type The composition of corporate debt varies substantially across US firms. However, the majority corporate debt can be classified into two broad categories, bank loans and public debt. Given the distinct properties of these two debt types, the effect of debt structure on leverage dynamics could possibly be influenced by the use of each of them in a firm. As financial intermediaries, banks tend to build a closer relationship with entrepreneurs than dispersed investors. Banks have access to the firm s transaction accounts and can gather much of the required non-public information at low cost (Nakamura, 1993). They

9 7 are willing to spend resources to acquire and assess information about firms uncertain productive prospects. It has been widely documented that monitoring is more intense under bank finance (Holstrom and Tirole, 1997; Repullo and Suarez, 1999). Recently, banks also take a role of originators of asset-backed securities which requires screening of applicants projects. The costly information acquisition will make banks more cautious in initiating a new or terminating an old bank-entrepreneur relationship. In firms with more bank lans, the long-term debt maturity implies a long-term banking relationship. Under such a relationship, the bank knows well about the firm and the information acquisition and monitoring should be less costly for the bank. Banks might be willing to provide higher flexibility to the firm and, in return, it should be easier for the firm to adjust its use of bank loans. Thus, we expect H1a: The negative effect of debt maturity on leverage adjustments has been mitigated in firms with more bank loans. Compared to informed bank lenders, information acquisition is more costly for arm slength bondholders. They typically receive only public information. Producing the information required to issue public securities is expensive as prospective borrowers are required to submit certified financial statements and apply for SEC (Securities and Exchange Commission) registration (Blackwell & Kidwell, 1988). Also, it is less efficient for multiple individual bond investors to monitor a firm because each investor has the incentive to free-ride on someone elses monitoring effort (Diamond, 1984). It is hard to contact these dispersed holders and any renegotiation suffers from information and free-rider problems (Rajan, 1992). Thus we expect H1b: The negative effect of debt maturity on leverage adjustments has been exacerbated in firms with more public debt.

10 8 3 Empirical Approach and Data 3.1 Empirical Approach We follow Faulkender, et al. (2012) to estimate firms capital structure adjustments toward target. Conventionally, the capital structure adjustment speed is estimated using the following regression specification: Lev i,t Lev i,t 1 D i,t A i,t D i,t 1 A i,t 1 = λ(lev i,t Lev i,t 1 ) + ε i,t, (1) Where D i,t is the firm s outstanding debt at time t, A i,t is the book value of assets at time t, Lev i,t is contemporaneous leverage, Lev i,t 1 is lagged leverage, Levi,t is the estimated target leverage, given firm characteristics at t 1. λ captures the firm s speed of adjustment toward target, measuring the fraction of leverage deviation that is eliminated during period t. However, as argued by Faulkender et al. (2012), λ in equation (1) will change without any active capital structure adjustment when the firm posts its annual income to its equity account. They suggest that target adjustment models should focus on active adjustments, where the firms access capital markets in some way and pay the associated transaction costs. We follow their approach and revise (1) as below: Lev i,t Lev p i,t 1 = γ(lev i,t Lev p i,t 1 ) + ε i,t, (2) where Lev p i,t 1 D i,t 1 A i,t 1 + NI i,t and NI i,t is equal to net income during year t. Lev p i,t 1 is the leverage at t if the firm engages in no net capital market activities. Therefore, γ in equation (2) captures the firm s speed of active adjustment toward its target leverage ratio.

11 9 To construct an estimate of the target leverage ratio, L i,t, we follow the Faulkender et al. (2012) and use the restriction that Lev i,t = βx i,t 1 Lev i,t = γβx i,t 1 + (1 γ)l i,t 1 + ε i,t, (3) Where β is a coefficient vector to be estimated concurrently with γ and X i,t 1 is a vector of firm-specific characteristics, along with firm- and year- fixed effects. Our firmspecific characteristics are defined as: LnT A = ln(total assets deflated by the consumer price index to 2001 dollars), EBIT T A = (Income before extraordinary items + Interest expense + Income taxes)/total assets, M B = (Book liabilities plus market value of equity)/total assets, F A T A = Fixed assets/total assets, DEP T A = Fixed asset depreciation/total assets, R&D T A=Research and development expense/total assets (missing R&D expenses are set to zero), R&D D=1 if Research and development expense =0, else zero, and Ind med Lev = Median leverage for the firm s Fama and French (1997) industry. We estimate equation (3) using Blundell and Bond s (1998) system GMM estimation method to compute ˆL i,t. Then equation (2) is estimated using OLS with bootstrapped standard errors to account for the generated regressor (Pagan, 1984). Following Faulkender et al. (2012) and Oztekin and Flannery (2012), we take equation (2) as our baseline model and interact deviation from target leverage with debt duration, debt concentration, and other variables of interest to see how these variables affect leverage adjustment speeds.

12 Data We start with U.S. firms covered by Compustat from 2001 to The sample begins in 2001 because the coverage of Capital IQ is incomplete in year Following the literature, we exclude financial and utility firms (SIC codes and ). We further exclude firms with non-positive assets, firms with book leverage outside the unit interval, and firms with fewer than two consecutive years of data. We then merge the resulting Compustat sample with Capital IQ and exclude observations for which the difference between total debt as reported in Compustat and the sum of debt types as reported in Capital IQ exceeds 10% of total debt. Our final study sample consists of 37,634 firm-year observations. Firm-level debt structure variables are from Capital IQ that decomposes total debt into seven mutually exclusive debt types: commercial paper (CP), drawn credit lines (DC), term loans (TL), senior bonds and notes (SBN), subordinated bonds and notes (SUB), capital leases (CL), and other debt (Other) (See Colla, Ippolito, and Li (2013) for more details on the definition of each type of debt). Our debt maturity measure is defined as the weighted average duration of all different types of debt used by a firm. **How about debt with maturity less than 1 year?** Our debt concentration measures are defined as in Colla et al. (2013). The first one is a normalized Herfindahl-Hirschman Index (HHI) of debt type usage. It is computed as where HHI i,t = SS i,t 1/7 1 1/7 SS i,t = ( CP i,t T D i,t ) 2 + ( DC i,t T D i,t ) 2 + ( T L i,t T D i,t ) 2 + ( SBN i,t T D i,t ) 2 + ( SUB i,t T D i,t ) 2 + ( CL i,t T D i,t ) 2 + ( Other i,t T D i,t ) 2 Thus, SS i,t is the sum of the squared debt type ratios for firm i in year t and T D refers to total debt. HHI equals one if a firm employs exclusively one single debt type and zero if a firm simultaneously employs all seven debt types in equal proportion. Hence, higher

13 11 values of HHI indicate firms tend to use fewer debt types and have higher degree of debt concentration. The alternative measure of debt concentration is a dummy variable, Ex90. It equals one if a firm obtains at least 90% of its debt from one debt type, and zero otherwise. The percentage of bank loans used by a firm is defined as the sum of drawn credit lines and term loans over total debt while the percentage of public debt is defined as the sum of senior bonds and notes and subordinated bonds and notes over total debt. Table I presents summary statistics of all the variables. 4 Debt Structure Characteristics and Leverage Adjustments To compare with existing studies, we start with baseline estimation of equations (1) and (2) and the results are reported in Table II. The annual leverage adjustment speed and the active adjustment speed are and 0.308, respectively, which correspond closely to estimates documented previously in the literature (See, e.g., Faulkender, et al., 2012). Given the advantages of using Lev p i,t 1 as the firm s starting point in adjusting leverage, we continue with it throughout the rest of the paper To estimate the effect of debt structure characteristics on the capital structure adjustment speed, we generalize equation (2) by specifying that a firm s adjustment speed γ depends on a variable of interest (in our case, that variable is debt maturity or debt concentration) as follows (Faulkender, et al., 2012): Lev i,t Lev p i,t 1 = (γ 0 + γ 1 DS Char i,t )(Lev i,t Lev p i,t 1 ) + ε i,t, (4) Where γ 1 is the coefficient on the interaction between active deviation from target leverage and debt structure characteristics. The coefficient γ 1 captures the influence of debt structure characteristics on the speed of leverage adjustment. When γ 1 is positive, the debt structure variable speeds up the adjustment, and when it is negative, the bank

14 12 structure variable slows down the adjustment. Table III reports the estimation results of equation (4). Column 1 shows that M aturity has a significantly negative coefficient of , suggesting the average maturity of corporate debt is negatively associated with leverage adjustment speed. In columns 2 and 3, two concentration measures, HHI and Ex90, have positive coefficients of and 0.022, respectively. This indicates debt concentration is positively related to leverage adjustment. Columns 4 and 5 put maturity and concentration measures together and each variable still remains statistically significant. The evidence provides support to both H1 and H2. That is, debt maturity impedes leverage adjustment while concentration accelerates adjustment. The last two columns further include the interactions between maturity and concentration measures. Interestingly, both interaction terms are significant and bear a positive sign, indicating the effect of two aspects of debt structure are not independent from each other. Concentrated debt structure help mitigate the negative effect of maturity on leverage adjustment. As argued by Faulkender, et al. (2012), it is not appropriate to assume under- and over-levered firms adjust their leverage ratios at the same rate because they differ not only in adjustment costs but also in perceived benefits. To eliminate the symmetry between under- and over-levered firms, we split our sample into two groups and examine the effect of debt structure in under- and over-levered firms separately. We create a dummy variable Underlev, which equals 1 for firm-years with leverage below target leverage and 0 for those with leverage above target. Column 1 in Table IV shows that, consistent with prior literature, under-levered firms exhibit a slower adjustment speed compared to their over-levered counterparts. The coefficient of U nderlev is and it is statistically significant, suggesting the adjustment speed of under-levered firms is less than that of over-levered firms. Columns 2 through 4 add debt structure variables and their effect remain similar to those reported in Table III. Columns 5 throug 7 further include the interaction terms between U nderlev

15 13 and maturity and concentration measures. As it shows, the effect of debt structure on leverage adjustment is not symmetric in under- and over-levered firms. Specifically, the coefficient of Maturity Underlev is significantly positive whereas that of HHI Underlev and Ex90 U nderlev are significantly negative. This suggests, in over-levered firms, both the negative effect of debt maturity and the positive effect of debt concentration for leverage dynamics are stronger. 4.1 Robustness Checks Financial Crisis and Debt Structure Our study period covers the most recent financial crisis during which many US firms become financially distressed and credit market condition deteriorates. Banks experienced major difficulties to finance themselves in money markets. Liquidity drying-up and funding difficulties in the banking sector were soon spread to the corporate sector. As a result, the financing pattern of US firms has been changed accordingly. Table V examines whether the financial crisis impacts the relation between debt structure and leverage adjustment. We define a variable CrashY to denote the financial crisis years. It equals 1 for years and 0 otherwise. Results show that the effects of debt maturity and concentration are unchanged. Interestingly, CrashY carries a significantly positive coefficient in most of the models, indicating firms adjust faster towards their target leverage during the financial crisis. This is consistent with the stylized fact that the debt to equity ratio rose sharply for US firms in crisis years. The interaction term between CrashY and Maturity has a significantly negative coefficient of The long-term maturity acts as a more severe impediment on adjustment during financial crisis. On the other hand, the interaction between CrashY and HHI has a significant positive coefficient of The positive effect of debt concentration on leverage adjustment is higher in financial crisis years. More concentrated debt structure might help with firm s renegotiation with creditors as a significant portion of firms became

16 14 financially constrained during the crisis Agency costs and Debt Structure Morellec, Nikolov, and Schurhoff (2012) demonstrate that manager-shareholder conflicts have a first-order effect on capital structure decisions. When making financing decisions, self-interested managers trade off the tax benefits of debt against the total costs of debt, which include not only the costs of financial distress but also those associated with the disciplining effect of debt, that is, the loss of free cash flow at managers discretion for potential private benefits (Jensen, 1986). As a result, agency conflicts would lower the firm s target leverage and its propensity to refinance. This argument is supported by the empirical evidence of Liao, Mukherjee, and Wang (2015). To ensure the effect on leverage adjustment captured by our debt structure variables is not driven by firms agency costs, we include the corporate governance index developed by Gompers, Ishii, and Metrick (2003), Gindex, and a subindex by Bebchuk, Cohen, and Ferrell (2009), Index6, in equation (4). The estimation results are presented in table VI. Two interesting results emerge. First, incorporating governance indices into our model specification does not affect the significance of debt structure variables, suggesting the interest conflicts between debtholders and equity holders and those among debtholders play an independent role in capital structure adjustment. Second, the interaction terms between governance indices and debt maturity have a significant effect on leverage dynamics. Maturity Gindex and Maturity Index6 have a significant coefficient of and 0.062, respectively. Last, governance indices are negatively associated with leverage adjustment, which is in line with the finding of extant literature. 5 The Role of Debt Type As discussed in section 2, private credit and arms-length debt differ substantially in information acquisition costs and monitoring intensity. To investigate whether and how

17 15 the relative use of each debt type might affect the observed effect of debt maturity and concentration, we construct two variables, BLp and P Dp. The former is defined as the percentage of bank loans (i.e., the sum of drawn credit lines and term loans) out of total debt and the latter is as the percentage of public debt (i.e., the sum of senior bonds and notes and subordinated bonds and notes) out of total debt. The regression results are presented in Table VII. Results show that the use of bank loans versus public debt has material effect on leverage dynamics. The percentage of bank loan per se, BLp, has a significantly negative coefficient across columns 1 to 3. More interestingly, the coefficient of M aturity BLp is and it is significant, suggesting the use of bank loans help mitigate the negative effect of maturity on adjustment speed. On the other hand, the interaction between maturity and the use of public debt, Maturity P Dl has a negative coefficient of , though P Dp per se carry a positive sign. This implies that, in firms primarily financed with public debt, long-term debt maturity would make it more difficult to adjust towards optimal leverage ratio. Overall, the evidence in Table VII is consistent with our expectation and provides support to H1a and H1b. 6 Conclusion A large literature has been devoted to evaluate how firms are to attain their target leverage ratios and a series of factors have been documented as possible determinants of capital structure dynamics. We extend this literature by showing the characteristics of corporate debt structure have significant influence on adjustment speed. We find that firms with long-term debt maturity moves slowly towards their target leverage ratio while firms with concentrated debt structure moves quickly. This is consistent with our expectation as replacing debt with equity would always lead to a large wealth transfer to the remaining debtholders if a firm has long-term debt maturity while concentrated debt structure might lower the bankruptcy costs and monitoring costs. This

18 16 piece of evidence suggests that both the interest conflicts between debtholders and equity holders and those among different groups of debtholders have implications for the dynamics of leverage. The results are more pronounced in over-levered firms and in the period of financial crisis. Further evidence indicates that the negative impact of debt maturity is stronger in firms with more public debt and lower in firms with more bank loans. Compared to arm s-length creditors, relational lenders have lower information acquisition costs and monitoring costs, which makes the negotiation easier and thus provides more flexibility to firms.

19 17 References Altinkili, O. and R.S. Hansen, 2000, Are there economies of scale in underwriting fees? Evidence of rising external costs, Review of Financial Studies 13, Bebchuk, L., Cohen, A., Ferrell, A., 2009, What matters in corporate governance? 22, Bolton, P., and D. S. Scharfstein, 1996, Optimal debt structure and the number of creditors, Journal of Political Economy 104, Bris, A., and I. Welch, 2005, The optimal concentration of creditors, Journal of Finance 60, Childs, P. D., D. C. Mauer, and S. H. Ott, 2005, Interactions of corporate financing and investment decisions: The effect of agency conflicts, Journal of Financial Economics 79, Dangl, T., and J. Zechner, 2004, Debt maturity and the dynamics of leverage, Working paper. Diamond, D. W., 1984, Financial intermediation and delegated monitoring. The Review of Economic Studies, 51, Faulkender, M., Flannery, M.J., Hankins, K.W., Smith, J.M., Cash flows and leverage adjustments., Journal of Financial Economics 103, Fischer, E. O., R. Heinkel, and J. Zechner, 1989, Dynamic capital structure choice: Theory and tests, Journal of Finance 44, Gompers, P., Ishii, J, Metrick, A, 2003, Corporate governance and equity prices, Quarterly Journal of Economics 118, Hackbarth, D., and D. C. Mauer, 2012, Optimal priority structure, capital structure, and investment, Review of Financial Studies 25,

20 18 Hovakimian, A., T. Opler, and S. Titman, 2001, The debt-equity choice, Journal of Financial and Quantitative Analysis 36, 124. Jiang, W., K. Li, and W. Wang, 2012, Hedge funds and Chapter 11, Journal of Finance 67, Korajczyk, R., and A. Levy, 2003, Capital structure choice: Macroeconomic conditions and financial constraints, Journal of Financial Economics 68, Li, Kai, and Wei Wang, Creditor governance with loan-to-loan and loan-to-own, Working paper, University of British Columbia. Liao, L., Mukherjee, T., Wang, W., Corporate governance and capital structure dynamics: An empirical study, Journal of Financial Research, Morellec, E., B. Nikolov, and N. Schurhoff, 2012, Corporate governance and capital structure dynamics, Journal of Finance 67, Myers, S. C., 1977, The determinants of corporate borrowing, Journal of Financial Economics 5, Nakamura, L. I., 1993, Commercial bank information: information for the structure of banking. In L. J. White, & M. Klausner (Eds.), Structural change in banking. Homewood, IL: Business One Irwin. Nini, G., David Smith, and Amir Sufi, Creditor control rights and firm investment policy, Journal of Financial Economics 92, Oztekin, O., forthcoming, Capital structure decisions around the world: Which factors are reliably important? Journal of Financial and Quantitative Analysis. Oztekin, O., and M. Flannery, 2012, Institutional determinants of capital structure adjustment speeds, Journal of Financial Economics 103,

21 19 Paolo, C., F. Ippolito, and K. Li, 2013, Debt specialization, Journal of Finance, Park, C., 2000, Monitoring and the structure of debt contracts, Journal of Finance 55, Rajan, R. G., 1992, Insiders and outsiders: the choice between informed and arms-length debt. The Journal of Finance, 47, Roberts, Michael, and Amir Sufi, Control rights and capital structure: An empirical investigation, Journal of Finance 64, Strebulaev, I. A., 2007, Do tests of capital structure theory mean what they say? Journal of Finance 62, Sufi, A., 2007, Information asymmetry and financing arrangements: Evidence from syndicated loans, Journal of Finance 62, Titman, S., and S. Tsyplakov, 2007, A dynamic model of optimal capital structure, Review of Finance 11, Welch, I., 1997, Why is bank debt senior? A theory of asymmetry and claim priority based on influence costs, Review of Financial Studies 10,

22 20 Table I: Summary Statistics This table provides the mean, median, standard deviation, the 1st and 3rd quintile for all variables. In Panel A and B, the mean under-levered and mean over-levered values are also provided. is the target leverage ratio. depblev is the change in book leverage (Lev i,t Lev i,t 1 ). depactblev is the change in book leverage restricted to active adjustments only (Lev i,t Lev p i,t 1 ). IndepBlev is the book target less the book leverage from the previous year (Levi,t Lev i,t 1). IndepActBlev is the book target less the book leverage adjustment (Lev p i,t 1 ), which is capped at two to reduce the effect of income over-realization. Maturity is the weighted-average debt maturity. HHI is the debt specialization index. Ex90 is an indicator variable, equals one if a firm obtains at least 90% of its debt from one debt type, and zero otherwise. PDp is the percentage of public debt and BLp is the percentage of bank loans. Gindex is the governance index from Gompers, Ishii, Metrick (2003) and Index6 is the entrenchment index as in Bebchuk, Cohen and Ferrell (2009). Lev is the book leverage. EBIT T A is the EBIT over total assets. MB is the market-to-book ratio. DEP T A is depreciation over total assets. LnTA is the log of total assets in thousands. F A T A is the fixed assets over total assets. R&D T A is the research and development expenses over total assets and R&D D is a dummy variable, equals 1 if research and development expense =0, else zero. Ind med Lev is the median leverage for the firm s Fama and French (1997) industry. Sample period is from 2001 to Variable Mean Median StdD Q25 Q75 Overlev Underlev Panel A: Targets and deviations from target Levi,t depblev depactblev indepblev indepactblev Panel B: Debt maturity and structure variables Maturity Y HHI ex PDp BLP Gindex Index Panel C: Firm characteristics used in target leverage estimation lev EBIT TA MB DEP TA lnta FA TA R&D T A R&D D Ind med Lev

23 21 Table II: Baseline estimation of adjustment speed This table provides the baseline adjustment speed regression analysis for equation (1) and equation (2), where dependent variables are the change of book leverage and active change of book leverage, respectively. Lev i,t Lev i,t 1 D i,t A i,t D i,t 1 A i,t 1 = λ(lev i,t Lev i,t 1 ) + ε i,t, Lev i,t Lev p i,t 1 = γ(lev i,t Lev p i,t 1 ) + ε i,t, depblev is the change in book leverage (Lev i,t Lev i,t 1 ). depactblev is the change in book leverage restricted to active adjustments only (Lev i,t Lev p i,t 1 ). IndepBlev is the book target less the book leverage from the previous year (Levi,t Lev i,t 1). IndepActBlev is the book target less the book leverage adjustment (Levi,t Levp i,t 1 ). λ is estimated in the first regression and γ is estimated in the secon regression. P values are reported in the parenthesis. Nobs is the number of observations and adjusted R Sq is reported below it. Sample period is from 2001 to depblev depactblev indepblev (0.000) indepactblev (0.000) Nobs 39,621 39,621 R Sq 6.76% 22.22%

24 22 Table III: Adjustment speed, debt maturity and specialization structures This table provides regression analysis of equation (4) Levi,t Lev p i,t 1 = (γ0 + γ1ds Chari,t)(Lev i,t Lev p i,t 1 ) + ε i,t, where the debt structure variables in consideration are debt maturity and specialization. Their differential impact on adjustment speed is further examined for under-leveraged firms. Maturity is the weighted-average debt maturity. HHI is the debt specialization index. Ex90 is an indicator variable, equals one if a firm obtains at least 90% of its debt from one debt type, and zero otherwise. Underlev is a dummy variable, equals 1 if a firm s leverage is below its target level and zero otherwise. P values are reported in the parenthesis. Nobs is the number of observations and adjusted R Sq is reported below it. Sample period is from 2001 to Variable Maturity (0.000) (0.000) (0.000) HHI (0.000) (0.000) (0.000) Ex (0.000) (0.000) (0.000) Underlev (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Maturity * Underlev (0.000) HHI* Underlev (0.000) Ex90* Underlev (0.000) Constant (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Nobs 37,634 37,634 37,634 37,634 37,634 37,634 37,634 37,634 37,634 37,634 R Sq 23.55% 23.28% 23.26% 26.51% 27.29% 27.15% 26.97% 27.88% 29.27% 28.80%

25 23 Table IV: Interactions between debt maturity and specialization structures This table further tests equation (4) Lev i,t Lev p i,t 1 = (γ 0 + γ 1 DS Char i,t )(Lev i,t Lev p i,t 1 ) + ε i,t, where the interactions between debt maturity and specialization variables are examined. Maturity is the weighted-average debt maturity. HHI is the debt specialization index. Ex90 is an indicator variable, equals one if a firm obtains at least 90% of its debt from one debt type, and zero otherwise. P values are reported in the parenthesis. Nobs is the number of observations and adjusted R Sq is reported below it. Sample period is from 2001 to Variable Maturity (0.000) (0.000) (0.000) (0.000) HHI (0.001) (0.000) Ex (0.018) (0.000) Maturity * HHI (0.000) Maturity *Ex (0.000) Constant (0.000) (0.000) (0.000) (0.000) Nobs 37,634 37,634 37,634 37,634 R Sq 23.57% 23.55% 23.99% 23.85%

26 24 Table V: Financial crisis and debt structure This table examines the debt structure and adjustment speed association during financial crisis from 2007 to CrashY is a dummy variable, equals 1 for financial crisis years (2007 to 2010), and zero otherwise. Maturity is the weighted-average debt maturity. HHI is the debt specialization index. Ex90 is an indicator variable, equals one if a firm obtains at least 90% of its debt from one debt type, and zero otherwise. P values are reported in the parenthesis. Nobs is the number of observations and adjusted R Sq is reported below it. Sample period is from 2001 to CrashY (0.000) (0.000) (0.000) (0.000) (0.000) (0.205) (0.116) Maturity (0.000) (0.000) HHI (0.000) (0.014) Ex (0.000) (0.022) Maturity * CrashY (0.004) HHI * CrashY (0.011) Ex90* CrashY (0.293) Constant (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Nobs 37,634 37,634 37,634 37,634 37,634 37,634 37,634 R Sq 23.27% 23.58% 23.31% 23.29% 23.59% 23.32% 23.29%

27 25 Table VI: Agency costs and debt structure This table examines the debt structure and adjustment speed association given the agency problem between shareholders and debt holders. Gindex is the governance index using 24 anti-takeover attributes from Gompers, Ishii and Metrick (2003). Index6 is the entrenchment index as in Bebchuk, Cohen and Ferrell (2009). Maturity is the weighted-average debt maturity. HHI is the debt specialization index. Ex90 is an indicator variable, equals one if a firm obtains at least 90% of its debt from one debt type, and zero otherwise. P values are reported in the parenthesis. Nobs is the number of observations and adjusted R Sq is reported below it. Sample period is from 2001 to Variable Gindex (0.001) (0.001) (0.007) (0.007) (0.000) (0.008) (0.046) Index (0.55) (0.576) (0.476) (0.48) (0.003) (0.315) (0.897) Maturity (0.153) (0.116) (0.001) (0.000) HHI (0.008) (0.007) (0.015) (0.316) Ex (0.004) (0.004) (0.425) (0.014) Maturity * Gindex (0.001) Maturity * Index (0.001) HHI* Gindex (0.074) HHI * Index (0.427) Ex90 *Gindex (0.963) Ex90 * Index (0.341) Constant (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.424) (0.000) (0.000) (0.000) Nobs 13,367 13,367 13,367 13,367 13,367 13,367 13,367 13,367 13,367 13,367 13,367 13,367 13,367 13,367 R Sq 11.38% 11.31% 11.39% 11.32% 11.40% 11.35% 11.40% 11.36% 11.45% 11.39% 11.41% 11.35% 11.40% 11.36%

28 26 Table VII: Bank Loans, Public Debt and Debt Maturity Structure This table provides the debt maturity structure analysis given different levels of bank loan and public debt usage. BLp is the percentage of bank loans to a firm s total debt. PDp is the percentage of public debt of a firm s total debt. Maturity is the weighted-average debt maturity. P values are reported in the parenthesis. Nobs is the number of observations and adjusted R Sq is reported below it. Sample period is from 2001 to Variable BLp (0.000) (0.000) (0.000) PDp (0.000) (0.000) (0.000) Maturity (0.000) (0.000) (0.000) (0.000) Maturity*BLp (0.000) Maturity* PDp (0.000) Constant (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Nobs 37,634 37,634 37,634 37,634 37,634 37,634 R Sq 23.73% 24.15% 24.19% 24.92% 25.72% 25.89%