The effect of debt market imperfection on capital structure and investment: Evidence from the global financial crisis of 2008 in Japan∗ Hiromichi Iwaki† May 10, 2014

Abstract This paper investigates how debt market frictions affect real firm behaviors such as capital structures and investments differently based on whether the firm has access to the public debt market, taking debt structure differences into account. To this aim, using the natural experimental approach to examine the 2008 credit supply shock in Japan, I show that firms without access to the public debt market face decreased leverage and investment, accompanied by decreased bank debt, compared to firms with access. Considering that firms without access to the public debt market are more dependent on banks for their debt and are likely to have closer relationships with banks than those with access, it is intriguing that bank-dependent firms face reduced debt supplies from banks compared to other firms. Moreover, through investigation of the regression of investments where the interaction term with different debt structures is introduced, it is suggested that differences in debt structure or debt maturity between firms with access to the public debt and those without also play an important role in determining debt and investment and that bank-dependent firms face more underinvestment or uncertainty after the financial crisis of 2008 than firms with access to public debt market.

Keywords: Credit crisis, Bank loans, Public debt market, Capital structures, Debt structures, Debt source, Investment, Financing constraints JEL classification: E22, G01, G21, G32

∗

I am especially grateful to my Advisory Professor Takashi Misumi for guidance and support. We thank Hiroshi Kamae, Masaru Konishi, and Junichi Nakamura for helpful comments and continuous encouragements. We also thank the discussant, Zhaozhao He, at the 26th Asustralasian Finance and Banking Conference 2013 and the seminar participants at the 9th International Conference on Asian Financial Markets and Economic Development, the 13th Workshop at Hitotsubashi University in 2013, and the Kanto Area Study Group of Japan Society of Monetary Economics in 2014 for helpful comments. This paper was previously titled as “What do we know from empirical analysis of the credit crunch of 2008? Evidence from Japan”. All remaining errors are mine. † Graduate School of Commerce and Management, Hitotsubashi University, 2-1 Naka, Kunitachi, Tokyo 186-8601 Japan. E-mail address: [email protected]

1

1

Introduction

Firm capital structures involve different types of debt, such as bank loans or public bonds, though all are recognized as the same debt. Moreover, among publicly traded firms, some have access to the public debt market in which commercial paper and public bonds are issued, whereas other firms lack access and are therefore dependent on bank loans for debt financing. Although the former may use both bank loans and public debt, the latter are confined to using bank loans. In this paper, we study how firms make their decisions on debt and investment differently based on whether firms have access to the public debt market, taking into account the differences in detailed debt structure1 in the presence of the negative supply contraction that occurred after the Lehman shock on September 15, 2008. When there is friction in the debt capital market supply and some firms are restricted from substitute debt financing (while others can access the public debt market), that friction might have significant influence on firm behaviors such as leverage and investment decisions. In line with this prediction, Leary (2009), who studies the effect of debt source on capital structure through two events of bank loan credit expansion and contraction in the 1960s, shows evidence of less use of debt by firms without access to the public debt market. As for investment response to debt supply shock, Duchin, Ozbas, and Sensoy (2010) confirm that there are economically and statistically significant differences in investments among publicly traded firms due to financing constraints. — FIGURE 1 ABOUT HERE — The possible debt substitution during the credit crunches of 2008 and 1997-1998 can be seen in figure 12 . While banks’ lending attitude toward firms drops down sharply and straightly, there is an upsurge in public bond issuances. The capital market shrank after the Lehman Brothers bankruptcy because few institutional investors were taking the risk of investing, even in higher graded public bonds, as represented by the sudden increase 1

Raugh and Sufi (2010, p.4278) suggest that “recognition of debt heterogeneity might prove useful in examining the effect of financing on investment”. 2 Some recent studies point out firms’ substitution between loans and bonds (Becker and Ivashina, 2014; Adrian et al. 2012).

2

in bond yields shown in figure 2. The capital market soon regained its role because of the large demand from individual investors searching for attractive investment opportunities, constituting about 20% of total public bond issuances in the fiscal year of 2008. — FIGURE 2 ABOUT HERE — Institutions in Japan are generally viewed as being less influenced by the depressed subprime mortgage problems that increased distrust among financial institutions and led to the bankruptcy of Lehman Brothers because of less exposure to those products. However, a credit crunch and accompanying credit contraction did occur in Japan and had considerable effect on the real business sector, as seen in the number of bankruptcies in publicly traded firms. In figure 3, the number of bankruptcies in 2008 shows an abrupt surge, reaching the highest number ever. Furthermore, as shown in figures 4 and 5, the majority of public firms facing bankruptcy in 2008 belonged to industries highly dependent on debt, especially bank loans, or industries with higher leverage, such as real estate development, construction, finance (nonbank), and real estate fund management. Taken together, it could be said that firms in Japan also faced severely tightened credit due to the credit crunch during that period. — FIGURE 3 ABOUT HERE — — FIGURE 4 ABOUT HERE — — FIGURE 5 ABOUT HERE — Holmstrom and Tirole (1997) made the theoretical prediction that firms which do not have any access to the public debt market and therefore are dependent on banks for their debt are likely to face a contraction of the bank loan supply even if their financial health remains stable. This prediction implies that it is possible for a credit crunch to have asymmetric effects for firms depending on whether they have access to the public debt market. Consistent with Holmstrom and Tirole (1997), Chava and Purnanandam (2011) confirm, through the investigation of the Russian crisis of 1998, that firms dependent on 3

banks faced a higher decline in firm performance indicators such as stock return, investment, and profitability than firms with access to the public debt market.3 In the investigation, the researchers removed firms issuing “junk bonds” (below BBB grade) from their sample because of their research interest in unambiguous comparison between firms with access to the public debt market and firms without. This removal is valid because a junk bond issuance market exists in the U.S. debt capital market, and because of the limited number of investors investing in junk bonds it is not clear whether firms issuing such bonds maintain access to public debt markets. On the other hand, this elimination could potentially, and essentially, increase the differences between firms with investment-grade ratings (and therefore access to the public debt market) and firms without access. The Japanese debt capital market might be suitable for an investigation of the effect of different debt sources on firm behaviors because of its clearly separated debt capital market, as shown in figure 6. Firms with credit ratings above BBB can issue public bonds, and there is constant demand for those issuances. On the other hand, firms below BBB can never issue public bonds because of the lack of a junk bond issuance market. This obvious segmentation makes comparison among firms with similar characteristics more efficient for a natural experiment. — FIGURE 6 ABOUT HERE — When firms do not have access to the public debt market, they have little choice other than to be dependent on banks for debt financing, and thus their relationship with banks is more crucial than the relationships of firms that have access to alternative debt sources such as public bonds or commercial paper. Diamond (1984) and Haubrich (1989) point out the benefits of bank-firm relationships for both parties: Mitigating informational symmetries between firms and banks allows both to reap benefits from reduced costs of 3

There are some other studies on the impact of bank lending change on firm borrowers. Solvin, Suishka, and Polonchek (1993) show that firms dependent on bank are affected directly by their lending banks and moreover such firms are regarded as stakeholders of those banks. Camplello, Graham, and Harvey (2010) point out that, faced with the fears of future uncertainties on funding from banks, firms tend to cancel or delay the investment decisions. This evidence suggests that upon financial crisis, bankdependent firms are potentially restricted from ideal investment decisions as compared with firms with alternative debt sources other than bank debt.

4

both borrowing and lending. In this view, firms rated below BBB are likely to have less negative effect shock from bank lending compared with firms with access to public debt markets, which are less dependent on banks and less in need of deep relationships with banks. It is also interesting to examine the consistency of this view in the context of a credit crunch, when bank loan negative supply shock occurs. In light of this, we show the regression of bank loans outstanding through the credit crunch of 2008, with the result that firms more likely to be dependent on banks suffer more from decreased bank loan supply than firms with access to the public debt market. For the purpose of assessing how access to the public debt market affects firm behaviors such as leverage and investment, in this paper, we use the difference-in-differences approach and propensity score matching diagnosis as robustness checks of the main results. These approaches are seen in the recent studies of the effect of the credit crunch on firm behaviors (for example, Chava and Purnanandam, 2011; Lemmon and Roberts, 2010). Using these methods, we show evidence that access to the public debt market has significant influence on debt structure, especially debt maturity, even if measured by the ratio of short-term bank debt to total bank debt. In line with this view of debt maturity, Duchin, Ozbas, and Sensoy (2010) suggest that firms with a high short-termdebt-to-total-assets ratio face financial constraints and suffer more from credit supply shock. Almeida et al. (2012), who examined the causal effect of debt maturity on investment through the credit crunch of 2007, present findings that the more in need of long-term debt refinancing firms are, the more investment reductions those firms face. In this paper, we present additional evidence that debt maturity has a more significant effect on investment for firms without access to the public debt market than firms with access, as shown by the difference-in-differences method, in which an interaction term with debt maturity is additionally introduced. As a result, we find that in the course of the credit crunch of 2008, firms without access to the public debt market faced decreased leverage defined as the ratio of debt to total assets, increased short-term debt, and decreased long-term debt measured in

5

terms of amount compared with firms with access4 . Moreover, it is intriguing that bankdependent firms face a reduction in the ratio of short-term bank debt to total debt. Based on the result of the study on investment, consistent with previous research, firms without access to the public debt market invest less than firms with access. The investigation of investment taking debt structures into account shows that firms without access to the public debt market respond to an increase of both short-term debt and bank debt with less marginal investment than firms with access. This finding suggests that bankdependent firms face more underinvestment or uncertainty after a credit crunch than other firms. The evidence in this paper suggests that both access to the public debt market and differences in debt structure or debt maturity play important roles in determining investment. The remainder of this paper is structured as follows. In section 2, we describe the related literature. In section 3, the empirical method and approach using a robustness test are presented. In section 4, the sample, data, and data-generation methods are described. Section 5 shows the results from regression and robustness tests. Finally, we conclude in section 6.

2

Related literatures

Faulkender and Petersen (2006) focus on debt source differences and show that firms dependent on banks use less debt, with an economically significant magnitude, than firms with access to the public debt market, suggesting that firms could suffer from bank lending squeezes in different ways depending on whether they have alternative debt sources. This prediction is confirmed by Leary (2009). As to capital market frictions, which are inflexible for entry to and exit from the public debt market, Cantillo and Wright (2000) show evidence that once firms enter the public debt market, they continue to issue public debt, even if they drop below the threshold value for entry. This finding also implies that among firms with similar characteristics, 4

Uchino (2013) stresses that even firms with large public debt did not decrease investment relative to bank-dependent firms during the finanical ciris period of 2008 in Japan and the banking sector works efficiently.

6

some remain dependent on banks for their debt, but others continue to issue public debt and reap the benefit of having alternative debt sources. As for the literature on debt structure, there are several perspectives about the ways that debt maturity decisions are made. Myers (1977) presents a model in which firms with plenty of growth options choose more short-term debt than long-term debt because of the advantage of firm value accrued from higher investment option value. Diamond (1993) argues that the maturity choice made by firms varies depending on their credit quality. For firms with higher credit, short-term debt is preferred because they have confidence in their performance in the near future and expect the interest rate to be reduced. In the same manner, firms with somewhat lower credit quality favor long-term debt over short-term debt because they do not have enough confidence in their firms in the future. Firms with much lower credit quality have no choice other than short-term debt. Barclays and Smith (1995) show evidence supporting the prediction by Diamond (1993) in an empirical work in which they use the ratio of long-term debt to debt as a dependent variable; this is the ratio we use in regression in this paper as one of several measures of debt structure. The prediction that the less credit quality firms have, the more short-term debt they have, is consistent with results of this paper.

3 3.1

Methodology Difference-in-differences estimation

In order to measure the net effect of the difference in firm behavior responses such as leverage and investment through the credit crunch of 2008 in the light of a natural experiment, we employ the difference-in-differences approach. We take the following basic specification:

yit = α0 + α1 Tt + α2 Di + α3 Tt Di + Xit β + ϵit ,

7

(1)

where yit represents real outcomes such as leverage and investment; Tt is an indicator variable equal to one in the specific year: 2008, 2009, and years after 2010 respectively; Di denotes an indicator variable equal to one if the firm does not have access to the public debt market and thus is dependent on banks for its debt and restricted from issuing public debt, whether in the form of public bonds or commercial paper; and Xit is a vector representing control variables. Following the specification of Leary (2009), we divide the period after the year the Lehman shock occurred into three subsequent terms: 2008, 2009, and the remainder of the years. Doing so, we grasp the different time-varying response changes for the two types of firms.

3.2

Propensity score matching diagnosis

There are many differences in firm characteristics between firms with access to the public debt market and those without, as shown in table 1. There is also a concern that those differences have some influence on firm leverage or investment through some unobservable variable determining such corporate outcomes. To control for this potential endogeneity, Faulkender and Petersen (2006) introduced a method similar to the two-step investment variable approach, which we use for robustness in this paper. — TABLE 1 ABOUT HERE — As another method to tackle the possible endogeneity problems, we conduct a propensity score-matching diagnosis5 of the main result using difference-in-differences regression of leverage and investment. This method is employed by recent related studies6 . I use the following procedure. In the first stage, using probit estimation, we regress the indicator variable of whether firms have access to the public debt market on firm characteristics and the outcome before the credit crunch. At the second stage, we match7

5

Roberts and Whited (2012) present a detailed discussion of the propensity score-matching method. For example, Chava and Purnanandam (2011); Lemmon and Roberts (2010). Almeida et al. (2012) match not the form of propensity score but firm characteristics as covariates. 7 we follow the method suggested by Abadie and Imbens (2006) for the choice of distance metric. 6

8

probabilities as propensity scores, which are calculated from the first-stage probit estimation, between firms with credit ratings as a treatment group and firms without as a control group. For the purpose of controlling for the distance of propensity scores between each matched pair, we use the nearest neighborhood caliper-matching approach, in which we limit the distance to within one standard deviation. As Chava and Purnanandam (2011) noted, there is a trade-off between bias and efficiency: On the one hand, we require nearer matching so that the more similar firms in light of their characteristics are selected from both groups, but on the other hand, this reduces the size of the matching sample. A similar trade-off lies in the number of matches. Moreover, an appropriate and outright criterion based on which the number of matches is set appears to be lacking. This is why we show the results of several matching diagnoses, varying the number of matches for robustness. Theoretically, after matching based on the near propensity scores of firms with access to public debt and firms dependent on banks, the potential endogeneity is controlled and removed; however there still remains a concern that unobservable factors affect both groups of firms and lead to differences in outcomes such as leverage and investment. Thus, we report not only the average treatment effect for the treated (hereafter, ATT) but also the average treatment effect (hereafter, ATE), which is the same as the net effect calculated by the difference-in-differences method between the two samples. ATT and ATE are defined as follows:

1 ∑ af ter ATT = (y − yjaf ter ), n 1 i

(2)

1 ∑ af ter ATE = [(yi − yibef ore ) − (yjaf ter − yjbef ore )], n 1

(3)

n

n

where n is the number of matched sample observations; i and j represent firms among the treatment group with access to the public debt market and firms among the control group, respectively; y is outcomes such as leverage or investment; and bef ore and af ter 9

indicate whether the outcome is from the year before or the year after the collapse of Lehman Brothers, taking 2006 as bef ore and 2009 as af ter.

4

Data

4.1

Data source

I use data from Nikkei Needs Financial Quest for the financial information of each firm and from Rating and Investment Information, Inc. (R&I) for the information on the existence of a credit rating for each firm. For the purpose of investigating the effect of the credit crunch on publicly traded firms, which are regarded as providing relatively adequate information to outside investors compared with not-publicly-traded firms, we restrict the sample observations to those having information about stock price. We remove financial industry firms such as banks, securities, insurance companies, and leasing firms because of their structural differences. For the same reason, we also remove the electric power industry8 . The estimated period for the regression in the form of difference-in-differences is set to be balanced between the pre-credit crunch period (the fiscal years9 of 2004-2007) and the post-credit crunch period (the fiscal years of 2008-2011). During the pre-credit crunch period, the macro economy in Japan and other countries was stable and business conditions were moderate. Conversely, during the post-credit crunch period, business circumstances faced severe turmoil, as shown in figure 1.

4.2

Recognition of having access to public debt market

In this paper, public debt is defined as publicly traded bonds or commercial paper, those of which are recognized as having high liquidity. Because it is indispensable, when issuing public debt, to have a credit rating from a rating agency such as R&I or Standard and Poors (S&P), public debt issuance is generally accompanied by a credit rating. The 8

Exceptional public bond issuances by electric power companies have been made since long before 1996, when the deregulation of overall public bond issuance came into effect. 9 In Japan, a fiscal year is from April to March.

10

coincidence that firms with a credit rating have public debt outstanding was pointed out by Cantillo and Wright (2000); we therefore use the information on the existence of a credit rating and the experience of public debt issuance in classifying whether each firm year observation has access to the public debt market.10 For example, because R&I’s information on credit ratings before 2007 is not available publicly, we recognize each firm year observation for that period as having access to the public debt market if that firm had issued public debt since 1993. For consistency and to be conservative, in deciding whether a firm year observation is recognized as having access to the public debt market for the period from 2004 to 2006, we require a firm recognized as having access to have both public debt issuance since 1993 and credit ratings since 2007. In addition, we restrict the sample to firms with (without) consistent access to the public debt market; that is, any sample firms that do not have consistent experiences throughout the estimation period are not included. Moreover, we require whole sample observations to have successive information on financial data for the estimation period, to meet the criteria noted above, and to have debt outstanding. For the same reason as previous studies, such as those of Faulkender and Petersen (2006) and Chava and Purnanandam (2011), we exclude firms without debt because it is not certain whether they have access to the public debt market; they may have decided not to enter the public debt market even if they could.

4.3

Alternative measure of access to public debt market

As suggested by Faulkender and Petersen (2006), there is a possible endogeneity problem in using presence of a credit rating or past public debt issuance as an indicator variable in regressing the effect of leverage or investment on firm characteristics if unobservable variables related to some explanatory variable affect those outcomes. To circumvent that problem, as an alternative variable, we also use the predicted probability11 of having a credit rating, as introduced by Faulkender and Petersen (2006). 10

As for the recognition of having access to public debt, we follow the previous studies such as Faulkender and Peteresen (2006) and Chava and Purnanandam (2011). 11 Leary (2009) uses this methodology.

11

This methodology can be regarded as similar to the second-stage instrumental variable approach, as follows. In the first stage, we estimated the probit model, using the sample12 from 2007 to 2011, the period for which the credit rating information is available, of the existence of a credit rating on the explanatory variables as follows: tangibility; log of firm age; market-to-book ratio; an indicator as an instrument of whether the firm is listed on the Tokyo Stock Exchange’s first section at each fiscal year; and another indicator as an instrument of whether the firm meets a modified criterion implied by the criterion in use as of the year 1985 for firms to be entitled to issue convertible bonds. We chose the latter indicator variable13 as an instrument in the first-stage probit model estimation because firms meeting the modified criterion are likely to have easy access to the public debt market; for the same reason, following Faulkender and Petersen (2006), we chose the former instrumental variable because of the higher reputation accrued from being listed on the Tokyo Stock Exchange’s first section. Applying the coefficients from estimation of the probit model to the full sample, the predicted probability of public debt market access for the estimated years is generated. In the second stage, we use the predicted probability as an instrumental variable or alternative measure indicating the degree of access to the public debt market.

4.4

Sample description

As a result of the screening noted above, the whole sample consists of 13,855 firm year observations including 2,472 firm year observations recognized as having access to the public debt market and 11,383 others. Table 1 shows summary statistics for the sample. Differences in the averages of firm characteristics between two subsamples based on access to the public market are clearly observed for capital structures, debt structures, and investment level. The most remarkable differences in firm characteristics appear in firm size measured by assets or sales, as suggested by Faulkender and Petersen (2006). In 12

Here, the whole sample size used for the probit estimation is 13,518, which contains those observations excluded because of lacking the needed requirements for the difference-in-differences estimation done in this paper. 13 The correlation coefficient between the firms with credit ratings and the modified criterion for the period since 2007 is 0.52, a high enough value.

12

particular, firms dependent on banks for debt have, on average, about a 48% higher ratio of short-term debt to total debt than firms with access to the public debt market, which is in line with previous studies such as Leary (2009) and Barclays and Smith (1995). For almost all measures, firms with access to the public debt market are different from those without. This fact indicates that it is crucial to control for those differences when extracting the net effect from the credit crunch on firm without access to public debt market. In order to control for the principal differences, we run the regression in the form of difference-in-differences. In addition, as noted before, we test the main result through a propensity score matching diagnosis as an alternative approach to circumvent the potential endogeneity.

5 5.1

Results Impact on leverage and various types of debt

In the estimation of the credit crunch’s impact on debt, following the previous literature (Faulkender and Petersen, 2006; Rajan and Zingales, 1995), we basically select the control variables as follows: growth opportunity defined as market-to-book ratio (MB ), profitability (PROFITABILITY ), tangibility (TANGIBILITY ), log of book assets (LOG(ASSETS)), and log of firm age (LOG(1+AGE)). In addition, in Japan, some typical situations close to credit rationing might affect the bank lending market so that a firm’s decisions about debt could be distorted; we thus include another two control variables in the estimation: REDOPE2, an indicator which takes a value of one if the firm has two consecutive fiscal year records of operating deficit because Japanese banks almost automatically regard these firms negatively for lending; and UNDER MED ERATIO, another indicator which takes a value of one if the ratio of equity to assets is under the median value of the firm’s industry, based on the idea that the lower this ratio is, the less likely the firm is to borrow from banks because of comparison with other similar firms by banks. For the purpose of controlling for the macroeconomic factor, the confidence index (MACRO CI ), which is released monthly by the Cabinet Office, a government agency, is 13

introduced as an explanatory variable. Table 2 shows the estimation of difference-in-differences regression results of leverage, defined as the ratio of debt to total assets (in columns 1 and 2), and debt growth (in columns 3 and 4). In columns 1 and 3, the results using an indicator variable of having access to the public debt market are presented. In columns 2 and 4, instead of the indicator variable, a probability variable estimated from the probit model explained in the section 4.3 is used. It should be noted that in order to focus on how firms without access to the public debt market responded to the credit crunch, the variable NO ACCESS, referring to having no access to the public debt market, is interacted with dummy variables indicating the specific periods. — TABLE 2 ABOUT HERE — Regardless of whether the variable measuring access to the public debt market is defined as an indicator or a probability variable, the result in table 2 suggests that firms dependent on banks suffer a net decreasing effect of leverage compared with firms with access at 1.5% and 2.6%, respectively, in 2008, coefficients that are economically and statistically significant. For example, taking into account the average leverage ratio of firms without access to the public debt market, the magnitude of that net downward effect constitutes 6% or 11% of the average leverage of those firms. Similar effects are seen in the other post-Lehman shock period. The negative effect on debt usage from the credit crunch for firms dependent on banks is also confirmed by the result of regression of debt growth defined as the ratio of the annual difference in debt to debt in the previous year. — TABLE 3 ABOUT HERE — In table 3, the dependent variables are defined as the ratios of different types of debt (long-term debt, short-term debt, and bank debt, as well as total debt for comparison purpose) to total assets. This investigation allows us to observe the different impact of credit crunch on the amounts of different types of debt. Interestingly, although firms dependent on banks suffer more in terms of the amount of total debt, long-term debt, 14

and bank debt, they increase the amount of short-term debt. Even though firms without access to public debt market are generally dependent banks and they are regarded as keeping closer relationships with lending bank than firms with access, it is the counterintuitive evidence that the latter firms faced the reduction of bank debt. Taken together with the fact that the bank-dependent firms suffer the decreasing effect of other form of debt such as total debt, long-term debt, it is suggested that they have no other choice for debt financing than short-term debt, which is plausibly supplied by banks. In fact, as shown in column 5 in table 4, bank-dependent firms increase their short-term bank debt ratio to total bank debt more by 4.1% in 2008 than firms with access to public debt. In other words, this finding suggests that banks offer those bank-dependent firms relatively more short-term than long-term bank loans in comparison with firms with access to the public debt market. — TABLE 4 ABOUT HERE — In table 4, we further extend the investigation to a detailed analysis of bank debt, taking maturity differences into account. As Barclays and Smith (1995) suggest, in examining debt maturity structures, that the manner in which a specific debt is divided makes a difference. From this point of view, the dependent variables are defined as the ratio of long-term bank debt or short-term bank debt to total assets, total debt, and bank debt. The striking result is that when short-term bank debt is measured by the ratio to total assets, column 2 shows firms dependent on banks increased the ratio by 0.5% more than firms with access to public debt in 2008, but when measured by the ratio to total debt, column 4 shows that those firms decreased the ratio by 1% less than firms with access to public debt in 2008. In view of the debt structure, this evidence suggests that after the credit crunch, firms with access to the public debt market turned to banks for their debt, and that banks offered them relatively more not only long-term loans but also short-term loans than they did firms dependent on banks, which is consistent with the steep increase in outstanding bank loans to large enterprises around the credit crunch shown in figure 7. During the credit crunch, turmoil occurred in the debt capital market because of the upsurge in the yields on public bonds, as shown in figure 2. In other words, 15

during the credit crunch, even some firms with access to the public debt market, though perhaps briefly, faced difficulty in financing debt so that they had no other choice than to borrow from banks, even if it were in the form of short-term contracts. — FIGURE 7 ABOUT HERE — As a final test for the debt structure response during the credit crunch, in table 5, we show the regression result of the ratios of four types of debt to total debt. As already noted above, the possibility that firms having ex ante access to the public debt market increased their usage of bank loans more after the credit crunch than did other firms can be seen in column 1, with the net difference of 5.3%, an economically and statistically significant value. In column 2, the result from the regression of debt maturity shows that firms dependent on banks have shorter maturity than firms with access to the public debt market by 0.5% and 1.3% in 2008 and 2009, respectively. Columns 3 and 4 show the regression results of bonds and convertible bonds to total debt: Firms dependent on banks have a net increasing effect for those two outcomes relative to other firms, but the interpretation is hard to be made. For example, because of data limitations and inseparability for individual firms in terms of the fraction of public or private bonds of total bonds outstanding, the data on bonds used for regression presented in column 3 potentially include both public and private bonds. Particular to Japan, private bonds issued by listed firms are in most cases purchased directly by banks that have relationships with the issuer firms and are held by those banks until maturity, which is reasonably equivalent to a bank loan. — TABLE 5 ABOUT HERE — Through the detailed investigation of debt usage responses, it is implied that while firms with ex ante access to the public debt market benefit from bank lending relationships even if the debt capital market is in turmoil, mitigating the adverse effects of the credit crunch, ironically, firms dependent on banks face decreases of both absolute debt level and bank debt. Moreover, for the latter firms, the fraction of short-term debt in debt increases relatively as shown in column 5 in table 4, which is consistent with the argument 16

that banks try to mitigate risk in lending to such firms by acquiring firms’ control rights, as Rajan (1992) suggests.

5.2

Impact on investment

In this section, we investigate how access to the public debt market affects firms’ investment behavior. In addition to this, as Almeida et al. (2012) show that firms in more need of refinancing debt before the credit crisis faced decreases in investment relatively, the debt maturity difference might play a critical role in investment decisions. In light of this, we thus also investigate, in the next section, how the credit crunch affected the investment decisions of firms dependent on banks, taking into account the effect of the debt structure difference. Following Kashyap et al. (1994), because of their suggestion that sales information have a close relation to investment, we select explanatory variables for the regression of investment as follows: market-to-book ratio (MB ), log of sales (LOG(SALES)), and log of lagged sales. In addition, we add log of firm age (LOG(1+AGE)) to the regression, based on the idea that the stage at which firms stand are likely to affect investment decisions in a way that growth opportunity and sales information alone cannot capture. Table 6 shows the regression result of investment in response to the credit crunch. The first two columns present the result using the ratio of investment expenditure to lagged total assets as a dependent variable, and the last two columns present the result using the ratio of investment expenditure to lagged tangible assets. In columns 1 and 3, the dummy variable taking the value of one if the specific firm does not have access to the public debt market is used as an independent variable representing which condition of debt financing the specific firm faces, and in columns 2 and 4, instead of the dummy variable, the probability of access to the public debt market calculated from probit model estimation is introduced for robustness. — TABLE 6 ABOUT HERE — In every measure of access to the public debt market as an independent variable and of the definition of investment as a dependent variable, the result in table 6 confirms that 17

firms dependent on banks suffered more of a decrease in investment in 2008 and 2009 than did firms with access to the public debt market. Given the average ratio (0.04) of investment to lagged assets for those firms dependent on banks, column 1 suggests that they faced a net decreasing effect on investment level of 12.5% upon the credit crunch in 2008 compared with firms with access to the public debt market, which is a finding with economic and statistical significance. Similar effects are observed in the other columns. The result shown in table 6 is consistent with previous studies on the impact of the credit crunch on investments, such as the work of Chava and Purnanandam (2011).

5.3

Response of investment through debt

In this section, it is examined how the investments of firms dependent on banks responded to the credit crunch through interaction with debt structure differences. Table 7 shows the striking finding that, even after controlling for the effect from the change in debt structure, firms dependent on banks faced a relatively larger reduction in investments, as each type of their short-term debt ratio, shown in columns 1, 2, and 4, increased by the same magnitude as firms with access to the public debt market during the credit crunch of 2008 and 2009. In addition, the more remarkable finding is shown in column 3: Where the variable DEBT interacted is defined as the ratio of bank debt to assets, even if bank debt increases in amount, firms dependent on banks face a net reduction of investment in the course of the same increase of bank debt as firms with access to the public debt market. For example, if the ratio of bank debt to assets increased by 10%, then bank-dependent firm faced a decrease in investment 0.24% higher than firms with access to the public debt market after the credit crunch, which is equivalent to a downward effect on the investment ratio of 16.7% for the average bank-dependent firm. — TABLE 7 ABOUT HERE — Evidence shown in table 7 suggests that firms without access to the public debt market respond with less marginal investment to an increase of both short-term debt and bank debt than firms with access, a finding that implies the former firms are put 18

in situations of greater underinvestment or uncertainty than the latter firms. In other words, taking into account that upon the credit crunch, firms dependent on banks suffer more in terms of investment than firms with access to the public debt market, a nonnegligible segmentation or friction in the debt market does exist, as suggested by Leary (2009). This is why under which capital market condition firms are put is a decisive factor in determining investment level: Firms dependent on banks plausibly face financing constraints. These findings are in line with previous studies on financing constraints (Almeida and Campello, 2007; Almeida et al., 2004; Fazzari et al., 1988; Gilchrist and Himmelberg, 1995; Kashyap et al., 1993; Kashyap et al., 1994; Whited, 1992). In that firms in need of refinancing before the credit crunch are more likely to suffer more from the decrease in investment, the result in this paper supports the work of Almeida et al. (2012) and Duchin, Ozbas, and Sensoy (2010).

5.4

Matching test

As table 1 shows, there are large differences for various firm characteristics. In this sense, the method of propensity score matching can be regarded as an alternative measure because of its underlying idea of comparing observations with similar characteristics. The main findings of the test of leverage and investment responses after the credit crunch, as shown in tables 2 and 6, suggest that bank-dependent firms are more likely to suffer from credit supply shock than firms with access to public debt market in Japan, as in the United States. To ensure the robustness of these results, we run the propensitymatching diagnosis on the response of leverage and investment level for before and after the credit crunch of 2008. Panel A of table 8 shows the different leverage responses for firms with access to the public debt market (treatment group) and those without both before the credit crunch (in 2006) and after (in 2009). As noted in section 3.2, we report ATE (ATE (DID)) as well as ATT. Sample firms used for matching diagnosis are those with full sets of data used for the test both in 2006 and 2009. Explanatory variables (covariates) used for generating propensity scores through the probit model estimation of access to the 19

public debt market are as follows: log of firm age; market-to-book ratio; profitability; tangibility; log of assets; and leverage defined as the ratio of debt to total assets as of 2006, the year before the credit crunch. We include leverage as an explanatory variable to control for the ex ante leverage level. It is noticeable that the choice of the number of matched control firms for each treatment firm seems to lack a strict standard; thus, for robustness reasons, we report the results with one to 10 matches. Throughout the reported results in panel A, both measures of average difference in leverage response between treatment firms with access to the public debt market and control firms without confirm the main finding that leverage was affected by the debt supply shock that occurred after the credit crunch: Firms dependent on banks suffered greater debt decreases than firms with access to the public debt market that had characteristics similar to bank-dependent firms. In the economic sense, the magnitude of the ATE ranging from 0.011 to 0.024 is near the result from the regression of leverage shown in table 2. For the purpose of comparing with other cases of credit crunch and for relative investigation, we also report in panel B the result from the matching test on leverage throughout the credit crunch of 1997-1998 in Japan, which had complex causes such as the regulatory pressure on banks to meet capital requirements and to apply stricter standard in lending to firms as well as credit crises in both Asia and Russia. Although there seems to be similarity in the phenomenon during the two credit crunch periods, as shown in figure 1, the older credit crunch appears to have much less of a differential impact between bankdependent firms and those with alternative debt sources such as public debt than did the latest credit crunch. This comparison sheds light on how large the impact from the credit crunch following the collapse of the subprime mortgage market and the Lehman shock was, as implied in figure 3 by the difference in number of bankruptcies between the two crises. In table 9, using the matching method, we show the result of investigation of the different impacts on investment. In the same manner used in the matching diagnosis for leverage difference in table 8, matching pairs are constructed between treatment firms

20

with access to the public debt market and control firms without as of 2006, and then average differences in investment level within the treatment group and control group before the credit crunch (2006) and after (2009) are calculated. The explanatory variables (covariates) used to generate a propensity score through probit model estimation of access to the public debt market are as follows: log of firm age; market-to-book ratio; profitability; tangibility; log of assets; and investment defined as the ratio of investment expenditure to lagged total assets in 2006, the year before the credit crunch, to control for the ex ante investment level. For both ATT and ATE (DID), the net advantage in investment level for the treatment firms over the control firms is evident as a whole, except for the case of allowing for two matches, a result which is consistent with Chava and Purnanandam (2011). The magnitude of ATE (DID) is consistent with the result in table 6.

5.5

Caveats

There might be some caveats that a demand shock could affect our regression results. For example, Kahle and Stulz (2013) stress the significant role of the impact of the demand shock, rather than the bank lending supply shock, on firm behaviors such as capital expenditures and net debt issuances during the recent financial crisis.14 To overcome the concern that those factors described above, we introduce industry dummy variables for all regressions of the impacts of the credit crisis on debt, and individual firm dummy variables for all regressions of the impacts of the credit crisis on investment.15 As a result of controlling for these factors, our regression results remain unchanged. Another concern is related to the sample used in this paper. In order to meet the ideal requirement of difference-in-differences regression in light of a natural experiment, as we mention in section 4.2, we use firm-year observations with successive financial information and consistent status of whether or not access to public debt market over the estimation 14

On the other hand, in the study of the factors that drive the credit cycle, Mian and Sfui (2010) show that through the investigation of the subprime mortgage crisis, not the demand side but “an outward shift in the supply of credit from 2002 to 2006 was a primary driver of the macroeconomic cycle of 2002 to 2009”. 15 Notice that the demand shock common among all firm are controlled in the difference-in-differences specification used in this paper.

21

period. This methodology might potentially cause sample selection bias, so we rerun the regression using the alternative sample that includes firms that are dropped from the sample used in this paper because of the lack of successive financial information.16 Here again, the sample selection bias problem seems not to be plausible and the robustness of the major results in this paper hold.

6

Conclusions

The Lehman shock was called a “one in 100 years event.” During the credit crunch, the magnitude of the effect from the shock to the Japanese financial system was often portrayed as being relatively smaller than those in other developed countries such as the United States and European countries, but the evidence in this paper shows that the shock was extraordinary enough that even publicly traded firms faced negative impact from the shock. Moreover, there were asymmetric effects on capital structures, debt structures, and investments, depending on the debt source or whether firms had access to the public debt market. This point supports Faulkender and Petersen (2006) and Leary (2009). The investigation made by this paper through the credit crunch also sheds light on the fact that the marginal change in debt structures affected investment in different ways after the credit crunch depending on both whether firms had access to the public debt market and which debt structures firms had. The finding that debt structure may influence investment is consistent with the suggestion in Raugh and Sufi (2010). Taken together with these findings, it is suggested that the differences in debt structures as well as accessibility to the public debt market, through their interactive effects, play a significant role in investment and that bank-dependent firms face more underinvestment or uncertainty after the financial crisis of 2008 than firms with access to public debt market. Our results suggest that the segmented debt capital market imposes some relative disadvantage in debt financing and investment on firms without access to the public debt 16

The regressions are taken place for the estimations on leverage, debt growth, investment.

22

market, and that there remains room to expand the public debt market to at least slightly below the BBB rating in light of the alternative debt choice and its benefits. At the same time, the results also imply that firms dependent on banks might face somewhat more uncertainty in debt financing after a banking shock than other firms, and this might be related to capital structure, debt structure, and investment decisions, topics that remain for future research.

Appendix: Definitions  

AGE The period since the firm was founded. ASSETS ; assets The book value of assets. BANK DEBT ; bank debt Short-term bank loans matured within one year plus longterm bank loans matured over one year. DEBT ; debt Long-term debt plus short-term debt. INVEST ; Invest Capital expenditure reported on the annual financial statements. SALES Gross sales of the firm. Lt Debt matured over one year. St Debt matured within one year. MACRO CI Confidence index released monthly by the Cabinet Office, a government agency. MB The ratio of the market value to the book assets, where the market value of assets is defined as book assets minus book equity plus the market value of equity. LEH08 An indicator equal to one in 2008, and zero otherwise. LEH09 An indicator equal to one in 2009, and zero otherwise. 23

LEHAFTER An indicator equal to one after the credit crunch except for 2008 and 2009, and zero otherwise. PROFITABILITY Operating profit divided by sales. TANGIBILITY Net property plant and equipment scaled by the book value of assets. REDOPE2 An indicator equal to one if the firm records an operational deficit for two prior years, and zero otherwise. ST BANK Short-term bank debt matured within one year. LT BANK Long-term bank debt matured over one year. BOND Public bonds plus private bonds. CB Convertible bonds. BANK DEBT Short-term bank debt plus long-term bank debt. ST DEBT Short-term debt matured within one year. LT DEBT Long-term debt matured over one year. TANGIBLE ASSETS ; tangible assets Net property plant and equipment. UNDER MED ERATIO An indicator equal to one if the firm has a ratio of equity capital to total capital below the median value in the industry in which the firm operates, and zero otherwise.

References [1] Abadie, A., and Imbens, G., 2006. Large sample properties of matching estimators for average treatment effects. Econometrica 74, 235-267. [2] Adrian, T., Colla, P., Shin, H.S., 2012. Which financial frictions? Parsing evidence from the financial crisis of 2007-2009. NBER Working Paper No. 18335.

24

[3] Almeida, H., Campello, M., 2007. Financial Constraints, asset tangibility, and corporate investment. Review of Financial Studies 20, 1429-1460. [4] Almeida, H., Campello, M., Laranjeira, B., Weisbenner, S., 2012. Corporate debt maturity and the real effects of the 2007 credit crisis. Critical Financial Review 1, 3-58. [5] Almeida, H., Campello, M., Weisbach, M., 2004. The cash flow sensitivity of cash. Journal of Finance 59, 707-722. [6] Barclay, M.J., Smith, C.W., Jr., 1995. The maturity structure of corporate debt. Journal of Finance 50, 609-631. [7] Becker, B., Ivashina, V., 2014. Cyclicality of credit supply: Firm level evidence. Journal of Monetary Economics 62, 76-93. [8] Campello, M., Graham, J.R., Harvey, C.R., 2010. The real effects of financial constraints: evidence from a financial crisis. Journal of Financial Ecnomoics 97, 470-487. [9] Cantillo, M., Wright, J., 2000. How do firms choose their lenders? An empirical investigation. Review of Financial Studies 13, 155-189. [10] Chava, S., Purnanandam, A., 2011. The effect of banking crisis on bank-dependent borrowers. Journal of Financial Economics 99, 116-135. [11] Diamond, D.W., 1984. Financial intermediation and delegated monitoring. Review of Economics Studies 51, 393-414.. [12] Diamond, D.W., 1993. Seniority and maturity of debt contracts. Journal of Financial Economics 33, 341-368. [13] Duchin, R., Ozbas, O., Sensoy, B.A., 2010. Costly external finance, corporate investment, and the subprime mortgage credit crisis. Journal of Financial Economics 97, 418-435.

25

[14] Faulkender, M., Petersen, M., 2006. Does the source of capital affect capital structure? Review of Financial Studies 19, 45-79. [15] Fazzari, S., Hubbard, R.G., Petersen, B., 1988. Financing constraints and corporate investment. Brookings Papers On Economic Activity 1, 141-195. [16] Gilchrist, S., Himmelberg, C.P., 1995. Evidence on the role of cash flow for investment. Journal of Monetary Economics 36, 541-572. [17] Haubrich, J., 1989. Financial intermediation, delegated monitoring, and long-term relationships. Journal of Banking Finance 13, 9-20. [18] Holmstrom, B., Tirole, J., 1997. Financial intermediation, loanable funds and the real sector. Quarterly Journal of Economics 112, 663-692. [19] Kahle, K.M., Stulz, R.M., 2013. Access to capital, investment, and the financial crisis. Journal of Financial Economics 110, 280-299. [20] Kashyap, A., Lamont, O., Stein, J., 1994. Credit conditions and the cyclical behavior of inventories. Quarterly Journal of Economics 109, 565-592. [21] Kashyap, A., Stein, J., Wilcox, D., 1993. Monetary policy and credit conditions evidence from the composition of external finance. American Economic Review 83, 78-98. [22] Leary, M.T., 2009. Bank loan supply, lender choice, and corporate capital structure. Journal of Finance 64, 1143-1185. [23] Lemmon, M.L., Roverts, M. R., 2010. The response of corporate financing and investment to changes in the supply of credit. Journal of Financial and Quantitative Analysis 45, 555-587. [24] Mian, A., Sufi, A., 2010. The great recession: Lessons from Microeconomic Data. American Economic Review 100, 51-56.

26

[25] Myers, S., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5, 147-175. [26] Rauh, D.J., Sufi, A., 2010. Capital structure and debt structure. Review of Financial Studies 23, 4242-4280. [27] Roberts, M.R., Whited, T.M., forthcoming. Endogeneity in empirical corporate finance. Milton Harris, and Rene Stulz, eds., Handbook of the Economics of Finance Volume 2, Elsevier. [28] Slovin, M.J., Sushka, M.E., Polonchec, J.A., 1993. The value of bank durability: borrowers as bank stakeholders. Journal of Finance 49, 247-266. [29] Uchino, T., 2013. Bank dependence and financial constraints on investment: Evidence from the corporate bond market paralysis in Japan. Journal of The Japanese and International Economies 29, 74-97. [30] Whited, T.M., 1992. Debt, liquidity constraints, and corporate investment: evidence from panel data. Journal of Finance 47, 1425-1460.

27

Figure 1: Trends in publicly traded corporate bond issuances and views (Diffusion Index (D.I.)) of large enterprise in terms of banks’ lending attitudes from 1993 to 2012. Data for public bond issuances are from the Japan Securities Dealers Association, and data for D.I. are from the Short-Term Economic Survey of Enterprises (Tankan) that the Bank of Japan releases in every three months. The Tankan is a statistical survey based on the Statistics Law; it includes approximately 210,000 private enterprises (excluding financial institutions) and is constructed as follows: First, the responding firms respond to specific questions, indicating whether the situation is (1) favorable, (2) not so favorable, or (3) unfavorable. The percentage of enterprises responding (3) is then subtracted from the percentage of enterprises responding (1). The right vertical line represents the level of D.I., and the left vertical line represents the amount of public bonds issued per month.

28

Figure 2: Public bond yields for three different terms from 2006 to 2012: long-term, mid-term, and short-term. Data are from the Nikkei Bond Index.

Figure 3: Annual trends in the number of bankruptcies of publicly traded firms. Data are from Teikoku Databank from 1992 to 2011.

29

Figure 4: Number of bankruptcies among public firms based on industries to which firms belong. Data are from Teikoku Databank from 2007 to 2009. The figure shows the following four classifications of bankrupted firms for each year: (1) all industries; (2) real estate development, construction, real estate fund management, and finance (nonbank) industries; (3) real estate development, construction, and real estate fund management industries; and (4) real estate fund management industries.s.

30

Figure 5: Total amount of liabilities of bankrupted public firms based on industries to which firms belong. Data are from Teikoku Databank from 2007 to 2009. The figure shows the following four classifications of bankrupted firms for each year: (1) all industries; (2) real estate development, construction, real estate fund management, and finance (nonbank) industries; (3) real estate development, construction, and real estate fund management industries; and (4) real estate fund management industries.

31

Figure 6: Distribution of credit ratings as of August 31, 2012. Data are from R&I, a credit rating agency founded in Japan.

Figure 7: Total amount of outstanding bank loans to large enterprise and medium and small enterprise. Data are from the Financial and Economic Statistics Monthly by the Bank of Japan. Data described are adjusted so that both of the two datasets as of fourth quarter 1996 have the same value, 100, so that both of the two datasets show the relative amount to those as of fourth quarter 1996.

32

Table 1: Summary statistics Firm characteristics All Access 1.05 (0.51) 1.15 (0.33) 3.96 (0.20) 4.19 (0.39) 0.05 (0.12) 0.06 (0.06) 0.33 (0.18) 0.37 (0.19) 10.79 (1.56) 12.97 (1.28) 10.83 (1.57) 12.90 (1.28)

MB LOG(1+AGE) PROFITABILITY TANGIBILITY LOG(ASSETS) LOG(SALES)

DEBT/ASSETS BANK DEBT/ASSETS LT DEBT/ASSETS ST DEBT/ASSETS ST DEBT/DEBT LT BANK/ASSETS ST BANK/ ASSETS ST BANK/BANK DEBT LT BANK/DEBT ST BANK/ DEBT ST BANK/DEBT BOND/DEBT CB/DEBT

Capital structures and debt structures 0.24 (0.17) 0.28 (0.16) 0.21 (0.16) 0.21 (0.14) 0.11 (0.11) 0.17 (0.12) 0.13 (0.11) 0.11 (0.08) 0.59 (0.26) 0.42 (0.22) 0.09 (0.10) 0.12 (0.10) 0.12 (0.11) 0.09 (0.08) 0.62 (0.26) 0.49 (0.24) 0.33 (0.24) 0.35 (0.21) 0.56 (0.25) 0.35 (0.22) 0.90 (0.19) 0.73 (0.23) 0.08 (0.17) 0.21 (0.21) 0.02 (0.09) 0.04 (0.11)

INVEST /lagged ASSETS INVEST /lagged TANGIBLE ASSETS

0.05 0.20

Investment (0.05) 0.05 (1.07) 0.15

(0.03) (0.12)

No access 1.04 (0.54) 3.91 (0.51) 0.05 (0.12) 0.32 (0.18) 10.32 (1.17) 10.39 (1.25)

Difference 0.11∗∗∗ 0.28∗∗∗ 0.01∗∗∗ 0.05∗∗∗ 2.66∗∗∗ 2.51∗∗∗

0.23 0.21 0.10 0.13 0.63 0.08 0.19 0.65 0.33 0.61 0.93 0.06 0.01

(0.17) (0.16) (0.10) (0.11) (0.26) (0.10) (0.11) (0.25) (0.25) (0.27) (0.16) (0.14) (0.08)

0.05∗∗∗ 0.00 0.08∗∗∗ -0.02∗∗∗ -0.2∗∗∗ 0.03∗∗∗ -0.1∗∗∗ -0.16∗∗∗ 0.02∗∗∗ -0.25∗∗∗ -0.2∗∗∗ 0.16∗∗∗ 0.02∗∗∗

0.04 0.21

(0.05) (1.18)

0.01∗∗∗ -0.06∗∗∗

This table shows summary statistics for firm characteristics, capital structures, debt structures, and investments for all firms as well as subsample firm year observations divided based on whether a sample firm has access to the public debt market. The variables are generated based on data from Nikkei Needs Financial Quest. Sample firms are restricted to those that have the same status for access to the public debt market for the whole estimated period from 2004 to 2011. For detailed definitions of each variable, please see the Appendix. White’s heteroskedasticity-consistent standard errors, corrected for correlation across observations of a given firm, are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

33

Table 2: Impact of credit crunch on debt level and growth NO ACCESS: Constant NO ACCESS: LEH08 LEH09 LEHAFTER NO ACCESS*LEH08 NO ACCESS*LEH09 NO ACCESS*LEHAFTER MB PROFITABILITY TANGIBILITY LOG(ASSET BV) LOG(1+AGE) REDOPE2 UNDER MED ERATIO MACRO CI N R2

Leverage Dummy Probability 0.48∗∗∗ (0.088) 0.686∗∗∗ (0.076) −0.041∗∗∗ (0.002) −0.109∗∗∗ (0.007) 0.034∗∗∗ (0.002) 0.042∗∗∗ (0.003) −0.042 (0.026) −0.035 (0.025) −0.014 (0.012) −0.003 (0.012) −0.015∗∗∗ (0.002) −0.026∗∗∗ (0.003) −0.013∗∗∗ (0.001) −0.029∗∗∗ (0.003) −0.016∗∗∗ (0.002) −0.034∗∗∗ (0.004) 0.004 (0.007) 0.002 (0.007) −0.037 (0.023) −0.042∗ (0.023) 0.306∗∗∗ (0.005) 0.3∗∗∗ (0.005) −0.002∗∗∗ (0.000) −0.014∗∗∗ (0.001) −0.042∗∗∗ (0.002) −0.043∗∗∗ (0.003) 0.043 (0.125) −0.003 (0.131) 0.169∗∗∗ (0.003) 0.172∗∗∗ (0.003) −0.002∗∗ (0.001) −0.002∗∗∗ (0.001) 13,834 0.411

13,780 0.414

Debt growth Dummy Probability −0.428∗∗∗ (0.108) −0.526∗∗∗ (0.115) 0.048∗∗∗ (0.017) 0.102∗∗∗ (0.034) 0.106∗∗∗ (0.014) 0.194∗∗∗ (0.023) 0.035 (0.042) 0.062 (0.046) 0.023 (0.031) 0.05 (0.031) −0.031∗∗ (0.014) −0.138∗∗∗ (0.024) −0.016 (0.014) −0.046∗∗ (0.023) −0.034∗ (0.018) −0.065∗∗∗ (0.023) −0.01 (0.010) −0.009 (0.010) −0.18∗∗ (0.083) −0.178∗∗ (0.084) 0.035∗∗ (0.017) 0.034∗∗ (0.017) 0.023∗∗∗ (0.005) 0.027∗∗∗ (0.007) −0.033∗∗ (0.016) −0.034∗∗ (0.017) −1.535∗∗∗ (0.549) −1.518∗∗∗ (0.546) 0.024∗∗∗ (0.008) 0.023∗∗ (0.009) 0.002∗∗ (0.001) 0.003∗∗ (0.001) 13,834 0.017

13,780 0.017

This table shows the regression results for how the credit crunch affected leverage and debt growth in firms with access to the public debt market as compared with those without access throughout the period before the credit crunch (2004-2007) and after (2008-2011). The dependent variables are the ratio of total debt to total assets for columns 1 and 2, and the ratio of the difference for one year to debt one year prior for columns 3 and 4. Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole estimated period. NO ACCESS represents either a dummy variable (columns 1 and 3) that takes one if the firm has no access to the public debt market, and zero otherwise, or a probability variable (columns 2 and 4). The probability of access to the public debt market is calculated by applying the coefficients, which are generated by estimating the probit model of whether a firm has a credit rating with R&I from 2007 to 2011, to all sample observations. NO ACCESS, using the probability of access, is then constructed as one minus the generated probability of access. For detailed definitions of the other variables, please see the Appendix. White’s heteroskedasticity-consistent standard errors, corrected for correlation across observations of a given firm, are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

34

Table 3: Impact of credit crunch on the amount of each type of debt: long-term, shor-term, and bank debt Debt type: Constant NO ACCESS LEH08 LEH09 LEHAFTER NO ACCESS *LEH08 NO ACCESS *LEH09 NO ACCESS *LEHAFTER MB PROFITABILITY TANGIBILITY LOG(ASSETS) LOG(1+AGE) REDOPE2 UNDER MED ERATIO MACRO CI N R2

Total debt 0.48∗∗∗ (0.088) −0.041∗∗∗ (0.002) 0.034∗∗∗ (0.002) −0.042 (0.026) −0.014 (0.012) −0.015∗∗∗ (0.002) −0.013∗∗∗ (0.001) −0.016∗∗∗ (0.002) 0.004 (0.007) −0.037 (0.023) 0.306∗∗∗ (0.005) −0.002∗∗∗ (0.000) −0.042∗∗∗ (0.002) 0.043 (0.125) 0.169∗∗∗ (0.003) −0.002∗∗ (0.001) 13,834 0.411

Share of assets Lt debt St debt 0.169∗∗∗ (0.038) 0.311∗∗∗ (0.060) −0.037∗∗∗ (0.003) −0.004 (0.002) 0.022∗∗∗ (0.001) 0.012∗∗∗ (0.001) −0.008 (0.008) −0.034∗ (0.020) 0.004 (0.004) −0.018∗∗ (0.009) ∗∗∗ −0.019 (0.001) 0.004∗∗ (0.002) ∗∗∗ −0.024 (0.001) 0.011∗∗∗ (0.002) −0.02∗∗∗ (0.001) 0.004∗ (0.002) 0.003 (0.003) 0.001 (0.005) 0.025∗∗ (0.012) −0.062∗∗∗ (0.023) 0.229∗∗∗ (0.003) 0.077∗∗∗ (0.002) 0.008∗∗∗ (0.001) −0.01∗∗∗ (0.001) −0.03∗∗∗ (0.002) −0.013∗∗∗ (0.001) 0.048∗ (0.029) −0.005 (0.127) 0.076∗∗∗ (0.001) 0.093∗∗∗ (0.002) −0.001∗∗∗ (0.000) −0.001 (0.001) 13,834 0.361

13,834 0.253

Bank debt 0.355∗∗∗ (0.074) 0.008∗∗∗ (0.003) 0.037∗∗∗ (0.002) −0.012 (0.022) 0.004 (0.010) −0.015∗∗∗ (0.002) −0.017∗∗∗ (0.002) −0.019∗∗∗ (0.003) 0.004 (0.007) −0.074∗∗∗ (0.025) 0.29∗∗∗ (0.005) −0.005∗∗∗ (0.000) −0.034∗∗∗ (0.002) −0.053 (0.128) 0.157∗∗∗ (0.003) −0.001∗ (0.001) 13,834 0.391

This table shows the regression results for how the credit crunch affected specific debt levels in firms with access to the public debt market as compared with those without access throughout the period before the credit crunch (2004-2007) and after (2008-2011). The dependent variables are the ratios of total debt, long-term debt, short-term debt, and bank debt to total assets in columns 1 (reproduced from table 2 for comparison purposes), 2, 3, and 4, respectively. Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole estimated period. NO ACCESS represents either a dummy variable that takes one if the firm has no access to the public debt market, and zero otherwise. For detailed definitions of the other variables, please see the Appendix. White’s heteroskedasticityconsistent standard errors, corrected for correlation across observations of a given firm, are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

35

Table 4: Impact of credit crunch on long-term and short-term bank debt Debt type: Constant NO ACCESS LEH08 LEH09 LEHAFTER NO ACCESS*LEH08 NO ACCESS*LEH09 NO ACCESS*LEHAFTER MB PROFITABILITY TANGIBILITY LOG(ASSET BV) LOG(1+AGE) REDOPE2 UNDER MED ERATIO MACRO CI N R2

Lt bank 0.062∗∗ −0.004 0.025∗∗∗ 0.019∗∗∗ 0.018∗∗∗ −0.02∗∗∗ −0.024∗∗∗ −0.019∗∗∗ 0.003 −0.008 0.213∗∗∗ 0.004∗∗∗ −0.023∗∗∗ −0.02 0.067∗∗∗ 0∗∗ 13,834 0.321

Share of assets debt St bank (0.026) 0.293∗∗∗ (0.003) 0.012∗∗∗ (0.001) 0.012∗∗∗ (0.004) −0.031∗ (0.002) −0.014∗ (0.002) 0.005∗∗ (0.002) 0.007∗∗∗ (0.002) 0 (0.003) 0.001 (0.009) −0.067∗∗∗ (0.004) 0.076∗∗∗ (0.001) −0.01∗∗∗ (0.002) −0.011∗∗∗ (0.024) −0.034 (0.001) 0.089∗∗∗ (0.000) −0.001∗

debt (0.054) (0.003) (0.001) (0.018) (0.008) (0.002) (0.002) (0.002) (0.005) (0.024) (0.002) (0.001) (0.001) (0.124) (0.002) (0.001)

13,834 0.253

Lt bank 0.057 0.015∗ 0.043∗∗∗ 0.093∗∗∗ 0.073∗∗∗ −0.042∗∗∗ −0.04∗∗∗ −0.033∗∗∗ 0.006∗∗ 0.008 0.387∗∗∗ 0.013∗∗∗ −0.036∗∗∗ −0.103 0.044∗∗∗ 0.001∗∗∗ 13,834 0.110

Share debt (0.037) (0.009) (0.005) (0.007) (0.006) (0.007) (0.007) (0.007) (0.003) (0.017) (0.011) (0.002) (0.003) (0.068) (0.003) (0.000)

of debt St bank 0.483∗∗∗ 0.186∗∗∗ 0.019∗∗∗ 0.048∗∗∗ 0.023∗∗∗ −0.01∗∗∗ −0.016∗∗∗ −0.013∗ −0.015∗∗∗ −0.12∗∗∗ −0.338∗∗∗ −0.026∗∗∗ 0.055∗∗∗ −0.25∗ −0.043∗∗∗ 0.001∗∗∗ 13,834 0.196

debt (0.091) (0.009) (0.002) (0.017) (0.008) (0.003) (0.003) (0.007) (0.004) (0.030) (0.004) (0.003) (0.003) (0.138) (0.004) (0.001)

Share of bank debt St bank debt 0.785∗∗∗ (0.049) 0.073∗∗∗ (0.008) −0.036∗∗∗ (0.004) −0.053∗∗∗ (0.007) −0.049∗∗∗ (0.005) 0.041∗∗∗ (0.006) 0.037∗∗∗ (0.005) 0.027∗∗∗ (0.005) −0.012∗∗∗ (0.003) −0.056∗∗∗ (0.018) −0.406∗∗∗ (0.011) −0.023∗∗∗ (0.003) 0.05∗∗∗ (0.003) −0.007 (0.090) −0.047∗∗∗ (0.003) 0 (0.000) 13,784 0.161

This table shows the regression results for how the credit crunch affected long-term and short-term bank debt in firms with access to the public debt market as compared with those without access throughout the period before the credit crunch (2004-2007) and after (2008-2011). The dependent variables are the ratio of long-term and short-term bank debt to total assets (columns 1 and 2), total debt (columns 3 and 4), and bank debt (column 5). Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole estimated period. NO ACCESS represents either a dummy variable that takes one if the firm has no access to the public debt market, and zero otherwise. For detailed definitions of the other variables, please see the Appendix. White’s heteroskedasticity-consistent standard errors, corrected for correlation across observations of a given firm, are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

36

Table 5: Impact of credit crunch on the fraction of debt: bank debt, short-term debt, bond, and convertible bond Debt type: Constant NO ACCESS LEH08 LEH09 LEHAFTER NO ACCESS *LEH08 NO ACCESS *LEH09 NO ACCESS *LEHAFTER MB PROFITABILITY TANGIBILITY LOG(ASSETS) LOG(1+AGE) REDOPE2 UNDER MED ERATIO MACRO CI N R2

Bank debt 0.54∗∗∗ (0.061) 0.201∗∗∗ (0.009) 0.062∗∗∗ (0.007) 0.141∗∗∗ (0.019) 0.096∗∗∗ (0.011) −0.053∗∗∗ (0.008) −0.056∗∗∗ (0.008) −0.045∗∗∗ (0.009) −0.009∗∗ (0.004) −0.112∗∗∗ (0.035) 0.049∗∗∗ (0.010) −0.013∗∗∗ (0.001) 0.019∗∗∗ (0.002) −0.353∗∗ (0.157) 0.002 (0.003) 0.002∗∗∗ (0.001) 13,834 0.195

Share St debt 0.584∗∗∗ (0.093) 0.12∗∗∗ (0.008) 0.001 (0.004) 0.021 (0.020) −0.003 (0.009) 0.005∗ (0.002) 0.013∗∗∗ (0.002) 0.011∗∗∗ (0.002) −0.014∗∗∗ (0.004) −0.101∗∗∗ (0.023) −0.362∗∗∗ (0.006) −0.027∗∗∗ (0.003) 0.052∗∗∗ (0.002) −0.062 (0.088) −0.043∗∗∗ (0.004) 0.002∗∗ (0.001) 13,834 0.174

of debt Bond 0.481∗∗∗ (0.044) −0.167∗∗∗ (0.010) −0.052∗∗∗ (0.008) −0.108∗∗∗ (0.011) −0.056∗∗∗ (0.007) 0.047∗∗∗ (0.008) 0.032∗∗∗ (0.008) 0.015 (0.010) 0.007∗∗ (0.003) 0.07∗∗∗ (0.024) −0.019∗∗∗ (0.006) 0.003∗∗ (0.001) −0.018∗∗∗ (0.001) 0.264∗ (0.149) −0.002 (0.002) −0.002∗∗∗ (0.000) 13,834 0.145

CB −0.019 −0.012∗∗∗ −0.013∗∗∗ −0.027∗ −0.033∗∗∗ 0.009∗∗∗ 0.014∗∗∗ 0.022∗∗∗ 0.002 0.046∗∗∗ −0.026∗∗∗ 0.009∗∗∗ −0.003∗∗∗ 0.092∗∗ 0 0

(0.034) (0.001) (0.002) (0.014) (0.008) (0.002) (0.002) (0.003) (0.002) (0.018) (0.005) (0.001) (0.001) (0.040) (0.001) (0.000)

13,834 0.040

This table shows the regression results for how the credit crunch affected specific debt as a fraction of total debt in firms with access to the public debt market as compared with those without access throughout the period before the credit crunch (2004-2007) and after (2008-2011). The dependent variables are the ratio of bank debt, short-term debt, total bonds outstanding, and convertible bond to total debt in columns 1, 2, 3, and 4, respectively. Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole estimated period. NO ACCESS represents either a dummy variable that takes one if the firm has no access to the public debt market, and zero otherwise. For detailed definitions of the other variables, please see the Appendix. White’s heteroskedasticity-consistent standard errors, corrected for correlation across observations of a given firm, are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

37

Table 6: Impact of credit crunch on investment NO ACCESS: C NO ACCESS LEH08 LEH09 LEHAFTER NO ACCESS*LEH08 NO ACCESS*LEH09 NO ACCESS*LEHAFTER MB LOG(SALES) LOG( lagged SALES) LOG(1+AGE) N R2

Invest/lagged assets Dummy Probability 0.079∗∗∗ (0.013) 0.123∗∗∗ (0.016) −0.007∗∗∗ (0.002) −0.024∗∗∗ (0.004) 0.01∗∗∗ (0.002) 0.015∗∗∗ (0.003) −0.003 (0.003) 0.001 (0.004) −0.005∗ (0.003) −0.004 (0.004) −0.005∗∗∗ (0.001) −0.01∗∗∗ (0.003) −0.002∗ (0.001) −0.007∗∗∗ (0.003) −0.001 (0.001) −0.003 (0.003) 0.014∗∗∗ (0.001) 0.013∗∗∗ (0.001) 0.042∗∗∗ (0.009) 0.041∗∗∗ (0.009) −0.042∗∗∗ (0.010) −0.043∗∗∗ (0.009) −0.011∗∗∗ (0.002) −0.011∗∗∗ (0.001) 12,848 0.073

12,800 0.073

Invest/lagged Dummy 1.086∗∗∗ (0.258) −0.013 (0.030) 0.036 (0.026) −0.027 (0.032) 0.009 (0.015) −0.055∗∗∗ (0.011) −0.052∗∗∗ (0.011) −0.056∗∗∗ (0.016) 0.14∗∗∗ (0.017) −0.064 (0.140) 0.039 (0.139) −0.187∗∗∗ (0.033) 12,848 0.016

tangible assets Probability 1.251∗∗∗ (0.404) −0.07 (0.088) 0.084∗∗∗ (0.029) 0.013 (0.034) 0.042∗∗ (0.021) −0.113∗∗∗ (0.028) −0.099∗∗∗ (0.028) −0.097∗∗∗ (0.036) 0.133∗∗∗ (0.019) −0.072 (0.140) 0.038 (0.137) −0.187∗∗∗ (0.035) 12,800 0.015

This table shows the regression results for how the credit crunch affected investment by firms with access to the public debt market as compared with those without access throughout the period before the credit crunch (2004-2007) and after (2008-2011). The dependent variables are the ratio of investment expenditure to lagged total assets for columns 1 and 2, and to lagged tangible assets for columns 3 and 4. Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole estimated period. NO ACCESS represents either a dummy variable (columns 1 and 3) that takes one if the firm has no access to the public debt market, and zero otherwise, or a probability variable (columns 2 and 4). The probability of access to the public debt market is calculated by applying the coefficients, which are generated by estimating the probit model of whether a firm has a credit rating with R&I from 2007 to 2011 to all sample observations. NO ACCESS using the probability of access is then constructed as one minus the generated probability of access. For detailed definition of each variable, please see the Appendix. White’s heteroskedasticity-consistent standard errors, corrected for correlation across observations of a given firm, are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

38

Table 7: Impact of credit crunch on investment through debt structures DEBTTYPE:

St debt/debt Constant 0.091∗∗∗ (0.011) NO ACCESS 0.003 (0.002) LEH08 0.004 (0.002) LEH09 −0.011∗∗∗ (0.003) LEHAFTER −0.01∗∗∗ (0.002) DEBTTYPE −0.032∗∗∗ (0.003) NO ACCESS *DEBTTYPE −0.007∗∗ (0.003) LEH08 *DEBTTYPE 0.014∗∗∗ (0.002) LEH09 *DEBTTYPE 0.014∗∗∗ (0.002) LEHAFTER *DEBTTYPE 0.01∗∗∗ (0.002) NO ACCESS *LEH08 *DEBTTYPE −0.012∗∗∗ (0.002) NO ACCESS *LEH09 *DEBTTYPE −0.004∗∗∗ (0.002) NO ACCESS *LEHAFTER *DEBTTYPE −0.003 (0.002) MB 0.014∗∗∗ (0.001) LOG(SALES) 0.04∗∗∗ (0.009) LOG( lagged SALES) −0.04∗∗∗ (0.010) LOG(1+AGE) −0.01∗∗∗ (0.002) N R2

Invest/lagged assets St bank debt/bank debt Bank debt/assets 0.091∗∗∗ (0.010) 0.065∗∗∗ (0.013) 0.006∗∗∗ (0.002) −0.01∗∗∗ (0.003) 0.003 (0.002) 0.007∗∗∗ (0.001) −0.013∗∗∗ (0.003) 0.001 (0.002) −0.011∗∗∗ (0.002) −0.002∗ (0.001) −0.027∗∗∗ (0.004) 0.02∗∗ (0.008) −0.015∗∗∗ (0.004) 0.02∗∗ (0.008) 0.012∗∗∗ (0.002) 0.012∗ (0.007) ∗∗∗ 0.013 (0.002) −0.009 (0.007) 0.008∗∗∗ (0.003) −0.003 (0.009) −0.008∗∗∗ (0.001) −0.024∗∗∗ (0.006) −0.002 (0.001) −0.019∗∗∗ (0.005) 0 (0.003) −0.018∗∗ (0.008) 0.014∗∗∗ (0.001) 0.014∗∗∗ (0.001) 0.04∗∗∗ (0.009) 0.044∗∗∗ (0.009) −0.04∗∗∗ (0.009) −0.043∗∗∗ (0.010) −0.01∗∗∗ (0.002) −0.01∗∗∗ (0.002)

12,848 0.102

12,816 0.105

12,848 0.082

St debt/assets −0.031∗∗∗ (0.010) −0.01∗∗∗ (0.003) 0.007∗∗∗ (0.001) 0.001 (0.002) −0.002∗ (0.001) 0.02∗∗ (0.008) 0.02∗∗ (0.008) 0.012∗ (0.007) −0.009 (0.007) −0.003 (0.009) −0.024∗∗∗ (0.006) −0.019∗∗∗ (0.005) −0.018∗∗ (0.008) 0.014∗∗∗ (0.001) 0.044∗∗∗ (0.009) −0.043∗∗∗ (0.010) −0.01∗∗∗ (0.002) 12,848 0.075

This table shows the regression results for how firms dependent on banks are affected by the credit crunch through the effect interacted with the specific debt structures as compared with firms with access to public debt. The estimated period spans the period before the credit crunch (2004-2007) and the period after (2008-2011). The dependent variable is the ratio of investment expenditure to lagged total assets. Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole estimated period. NO ACCESS represents either a dummy variable that takes one if the firm has no access to the public debt market, and zero otherwise. For detailed definitions of the other variables, please see the Appendix. White’s heteroskedasticity-consistent standard errors, corrected for correlation across observations of a given firm, are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

39

Table 8: Propensity score matching diagnosis on the effect of credit crunch on leverage Panel A

Credit crunch effect on leverage: (2006 vs 2009) ATE (DID) Treatment.obs (sd.error) (sd.error) (droped.obs) 0.022∗∗∗ (0.007) 0.014∗∗∗ (0.003) 303 (0) 0.007 (0.007) 0.011∗∗∗ (0.004) 303 (0) ∗∗∗ 0.032 (0.008) 0.021∗∗∗ (0.004) 243 (60) 0.04∗∗∗ (0.008) 0.024∗∗∗ (0.004) 237 (66) 0.036∗∗∗ (0.008) 0.02∗∗∗ (0.004) 228 (75) 0.032∗∗∗ (0.007) 0.011∗∗∗ (0.004) 207 (96)

ATT M 1 2 4 6 8 10

Panel B

Control.obs (unique.obs) Total 303 (121) 1,732 606 (192) 1,732 972 (295) 1,732 1,422 (359) 1,732 1,824 (394) 1,732 2,070 (430) 1,732

Credit crunch effect on leverage: (1996 vs 1998) ATE (DID) Treatment.obs Control.obs (sd.error) (sd.error) (droped.obs) (unique.obs) Total −0.089∗∗∗ (0.009) −0.026∗∗∗ (0.003) 145 (0) 145 (74) 1,405 0.004 (0.010) 290 (117) −0.014∗∗∗ (0.003) 145 (0) 1,405 0.073∗∗∗ (0.010) −0.003 (0.003) 145 (0) 580 (179) 1,405 0.038∗∗∗ (0.010) 0.001 (0.002) 117 (28) 702 (223) 1,405 0.018∗ (0.009) 0.001 (0.003) 111 (34) 888 (266) 1,405 0.017∗ (0.009) 0.001 (0.003) 106 (39) 1,060 (304) 1,405 ATT

M 1 2 4 6 8 10

This table shows the results from propensity score-matching diagnosis of the difference in the effect of the credit crunch on leverage between the differences in leverage of treatment firms with access to the public debt market and the differences of control firms without. ATT denotes the average treatment effect for the treated and is computed as the average leverage of treatment firms less the average of control matched firms after the credit crunch, where the matching pairs are constructed using propensity scores generated from firm characteristics before the credit crunch. ATE (DID) denotes the average treatment effect and is computed as the average difference in leverage of treatment firms between after and before the credit crunch less the average difference in leverage of control matched firms between after and before the credit crunch, where the matching pairs are constructed using propensity scores generated from characteristics before the credit crunch. Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole period from 2004 to 2009 for the matching diagnosis of the credit crunch of 2008 (panel A), and from 1995 to 2000 for the Asian and Russian crises (panel B). In panel A (B), the diagnosis is made by comparing leverages of sample firms as of 2006 (1996), a fiscal year before the credit crisis event, with those of corresponding firms as of 2009 (1998), a fiscal year after the credit crisis. The propensity score is equal to the probability calculated by the estimation of the probit model where the dependent variable is an indicator of whether the firm has access to the public debt market, and the independent variables are growth opportunity, profitability, firm size, firm age, and leverage defined as the ratio of debt to total assets, all before the event. The nearest neighborhood caliper-matching approach is used in generating matched pairs, where we ensure that the treatment (access to the public debt market) firm’s propensity score is within one standard deviation of the control (dependent on banks) firm’s score, so that samples are dropped by that measure. M denotes the number by which a treatment firm is matched to control firms. Maching is done with replacement. Treatment.obs (Control.obs) is the number of treatment (control) group observations used in matching. The standard error (denoted as sd.error) is reported using the bootstrap method with 1,000 replications. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

40

Table 9: Propensity score matching diagnosis on the effect of credit crunch on investment Credit crunch effect on investment: (2006 vs 2009) ATE (DID) Treatment.obs Control.obs M (sd.error) (sd.error) (droped.obs) (unique.obs) Total 1 0.008∗∗∗ (0.002) 0.003∗∗∗ (0.002) 280 (0) 280 (120) 1,566 2 0.01∗∗∗ (0.002) 0.003 (0.002) 280 (0) 560 (187) 1,566 ∗∗∗ ∗∗∗ 4 0.009 (0.002) 0.006 (0.002) 238 (42) 952 (284) 1,566 6 0.006∗∗∗ (0.002) 0.007∗∗∗ (0.002) 228 (52) 1,368 (341) 1,566 8 0.003∗∗∗ (0.002) 0.008∗∗∗ (0.002) 197 (83) 1,576 (397) 1,566 10 0.004∗ (0.002) 0.008∗∗∗ (0.002) 170 (110) 1,700 (439) 1,566 This table shows the results from propensity score-matching diagnosis of the difference in the effect of the credit crunch on investments between the differences in investments of treatment firms with access to the public debt market and the differences of control firms without. ATT denotes the average treatment effect for the treated and is computed as the average investment of treatment firms less the average of control matched firms after the credit crunch, where the matching pairs are constructed using propensity scores generated from firm characteristics before the credit crunch. ATE (DID) denotes the average treatment effect and is computed as the average difference in investment of treatment firms between after and before the credit crunch less the average difference in investment of control matched firms between after and before the credit crunch, where the matching pairs are constructed using propensity scores generated from characteristics before the credit crunch. Data are from Nikkei Needs Financial Quest. Sample firms are required to meet the condition of no missing data and constant recognition of access to the public debt market for the whole period from 2004 to 2009 for the matching diagnosis of the credit crunch of 2008. The diagnosis is made by comparing investments of sample firms as of 2006, a fiscal year before the credit crisis event, with those of corresponding firms as of 2009, a fiscal year after the credit crisis. The propensity score is equal to the probability calculated by estimation of the probit model where the dependent variable is an indicator of whether the firm has access to the public debt market, and the independent variables are growth opportunity, profitability, firm size, firm age, and investment defined as the ratio of debt to lagged total assets, all before the event. The nearest neighborhood caliper-matching approach is used in generating matched pairs, where we ensure that the treatment (access to the public debt market) firm’s propensity score is within one standard deviation of the control (dependent on banks) firm’s score, so that samples are dropped by that measure. M denotes the number by which a treatment firm is matched to control firms. Maching is done with replacement. Treatment.obs (Control.obs) is the number of treatment (control) group observations used in matching. The standard error (denoted as sd.error) is reported using the bootstrap method with 1,000 replications. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. ATT

41