• No results found

However, multiple main bank relationships reduce the hold-up costs and lead firms to higher profitability

N/A
N/A
Protected

Academic year: 2023

Share "However, multiple main bank relationships reduce the hold-up costs and lead firms to higher profitability"

Copied!
26
0
0

Loading.... (view fulltext now)

Full text

(1)

Takato Hiraki(a), Akitoshi Ito(b), and Fumiaki Kuroki(c)

(a) Kwansei Gakuin University, Scholl of Business Administration (b) in transition

(c) NLI Research Institute

November 14, 2003 Updated August 18, 2004

Abstract

In this study, we investigate whether multiple main bank relationships reduce the so-called “hold-up costs” of bank financing (Rajan (1992)) in the context of main-bank system uniquely developed in Japan by examining the panel data of companies listed on the Tokyo Stock Exchange, the first and second sections, during the period from 1991 to 1998. Our empirical results show that main bank borrowing is negatively related to the profitability of the firm, which suggests the presence of significant hold-up costs.

However, multiple main bank relationships reduce the hold-up costs and lead firms to higher profitability. This hold-up cost reducing effect of multiple main bank relationships is larger for firms with the higher value of growth opportunities than those with the lower value of growth opportunities.

Correspondence to:

Takato Hiraki, School of Business Administration, Kwansei Gakuin University Nishinomiya, Hyogo-ken 992-8501, Japan

Phone: (+81)798-54-6973; Fax: (+81)798-54-0903; e-mail: [email protected]

JEL classification: G32; G34

Key words: hold-up costs, multiple bank relationships, main bank, and Japanese firms

(2)

Single versus Multiple Main Bank Relationships: Evidence from Japan

Abstract

In this study, we investigate whether multiple main bank relationships reduce the so-called “hold-up costs” of bank financing (Rajan (1992)) in the context of main-bank system uniquely developed in Japan by examining the panel data of companies listed on the Tokyo Stock Exchange, the first and second sections, during the period from 1991 to 1998. Our empirical results show that main bank borrowing is negatively related to the profitability of the firm, which suggests the presence of significant hold-up costs.

However, multiple main bank relationships reduce the hold-up costs and lead firms to higher profitability. This hold-up cost reducing effect of multiple main bank relationships is larger for firms with the higher value of growth opportunities than those with the lower value of growth opportunities.

Key words: hold-up costs, multiple bank relationships, main bank, and Japanese firms

(3)

The benefits and costs of bank financing for firms are the focus of recent studies. A number of authors argue that information production and monitoring by banks mitigate information asymmetries and reduce agency costs of debt (see for example Diamond (1984, 1991), Fama (1985), and Berlin and Loeys (1988)). Past empirical studies report evidence consistent with such benefits of bank financing (for example, James (1987), Lummer and McConnell (1989), and Peterson and Rajan (1994)). Bank financing induces certain costs, however. Rajan (1992) and Sharpe (1990) point out that obtaining a monopoly in information, a bank to monitor is able to extract rents from its client firms, which in turn leads them to distorted investment decisions. This logic is a central idea behind the so-called “hold-up problem” of bank financing emphasized by Rajan (1992).1

The hold-up problem due to bank financing can be mitigated by the use of public debts.

But this may not be a viable solution all the time because of the significant agency costs involved in public debt financing. Still taking advantage of bank financing, an alternative way of reducing the hold-up costs is the use of several banks as opposed to the use of a single bank. Rajan (1992) and von Thadden (1994) argue that multiple bank relationships can mitigate the hold-up problem by providing some room for competition among lender banks. In the meantime, multiple bank relationships can be costly in some other respects. First, duplicated monitoring is usually less efficient than monitoring by a single bank. Second, a free rider problem of multiple bank

1 Houston and James (1996) provide evidence consistent with significant hold-up costs for U.S.

firms that rely on borrowing from a single bank.

(4)

monitoring may result in rather weakened monitoring over the activities of the borrower.

Third, restructuring of debt claims with multiple banks is more difficult during the period of financial distress. Therefore, the overall effectiveness of multiple bank relationships in mitigating the hold-up problem is largely an empirical matter.

In this study, we investigate whether multiple main bank relationships reduce the hold-up costs of “main-bank” financing by using the panel data of Japanese companies listed on the Tokyo Stock Exchange (TSE), both first and second sections included.

Historically, a typical Japanese company has maintained a special tie with its main bank (Aoki, Patrick and Sheard (1994)). Following past studies (Kang and Shivdasani (1997), and Morck, Nakamura and Shivdasani (2000) among others), the main bank in this study is defined as the largest creditor to the firm among the creditor banks. This does not preclude the possibility that the number of largest creditors, i.e., main banks, can be more than one: two or more creditor banks can provide the firm with exactly the same and the largest amount of loans. It is very odd to find that most past empirical studies (see for example Hoshi, Kashyap and Sharfstein (1990), Kaplan and Minton (1994), Kang and Shivdasani (1995), and Morck and Nakamura (1999)) do not pay much attention to this (multiple main bank) possibility or they even implicitly assume that the firm has (to have) a single main bank.2,3 Our data set actually shows that about 9.4 % of the firms with bank loans have multiple main banks according to our definition.

Based on this definition, all firms in our sample are classified into either the group with

2 For example, both of the popular quarterly investment guidebooks published by Toyo-keizai and Nikkei disallow each firm to have multiple main banks by

discretionally choosing one out of multiple main bank candidates with the same largest amount of loans. (more proof required)

3 See Horiuchi (1991) for institutional details about multiple main banks.

(5)

the single main bank or that of the multiple main bank relationship without much concerned with the qualitative difference between the real main bank, if it exists, and the sub-main bank(s). We subsequently confirm that this treatment of main bank(s) is relevant by showing higher performance in profitability for the firms with multiple main banks than for those with a single main bank. We interpret this result as attributable to the mitigated hold-up problem among the firms in the former group.

The relationship between Japanese firms and their main banks provides us with a unique experimental setting given the Rajan’s (1992) study. With the strong tradition of indirect financing in Japan, most firms borrow from multiple banks including main bank(s).

Therefore, our main concern is not with the number of banking relationships but with the number of main bank relationships. The relationship between the main bank and its client firm is maintained based on a long-time implicit contract, thus quite stable. In other words, changing or adding a main bank represents a very unusual event for Japanese companies. The observed stability of the main bank relationships in our sample implies that the potential endogeneity problem associated with selecting single vs. multiple main bank relationships might not be an important issue in this study.4

Past empirical studies provide mixed evidence about the effectiveness of multiple banks in reducing the hold-up costs. Houston and James (1996) report evidence that multiple bank relationships reduce the hold-up costs for large U.S. firms. In contrast, Degryse

4 Among the firms listed on the TSE in fiscal year 1991 which had a single main bank, only 8.9 percent of the firms changed their largest lender bank over the four-year course through fiscal year 1995, implying that a little more than two out of the one hundred firms change their main bank(s) per year. Out of this rare change of main banks, the occurrence of changes between the single and the multiple main bank relationship is further reduced to almost nil.

(6)

and Ongena (2001) show that multiple bank relationships reduce the profitability of Norwegian firms as borrowers from bank(s). The results of Cole (1998) indicate that multiple sources of creditors are likely to reduce the probability of obtaining additional credit for small U.S. firms.

Turning to the case of Japanese bank-firm relationships, many authors show that the main bank plays an important monitoring function and the firm’s special tie with the main bank reduces information asymmetries (see Aoki, Patrick and Sheard (1991), Kaplan and Minton (1994), Kang and Shivdasani (1995), and Morck and Nakamura (1999) among others). However, more recent studies (Weinstein and Yafeh (1998), Pinkowitz and Williamson (2001), and Hiraki et al. (2003)) suggest that the hold-up costs are also significant for Japanese companies that maintain a tie with the main bank.

Weinstein and Yafeh (1998) find that Japanese companies with a main bank tie tend to exhibit low profitability and low growth. Hiraki et al. (2003) also report that main bank borrowing is negatively related to firm value. Finally, Pinkowitz and Williamson (2001) show that Japanese firms with a main bank tie tend to hold the excessive amount of cash.

Thus, whether multiple main bank relationships can reduce the hold-up costs of Japanese firms is of academic interest as well as of practical importance.

Our major findings are as follows: first, main bank borrowing is significantly and negatively related to the profitability of the firm, which suggests the presence of hold-up costs. Second, multiple main bank relationships reduce the hold-up costs and lead the firms with this feature to higher profitability. This mitigating effect of multiple main bank relationships on the hold-up problem is robust irrespective of the

(7)

firm’s access to the public debt market and the firm size. Third, the mitigating effect of multiple main bank relationships on the hold-up problem is larger for firms with the higher value of growth opportunities than those with the lower value of growth opportunities. Finally, the correction on possibly existing selection biases does not qualitatively change our main results.

The rest of this paper is organized as follows. The next section describes our data set and variables of interest. Section III reports our empirical results. Section IV concludes the study.

II. Data

The data set used in this study is drawn from various sources: financial statement and stock price data of individual companies are from Nikkei Needs Database; data on bank loans are from Toyo Keizai and Nikkei Needs Database. Our initial sample of characteristic variables on a fiscal year basis of all firms listed on the first and second sections of the Tokyo Stock Exchange (TSE) for the sample period from 1991 to 1998 are extracted from this data set.5

From the initial sample of all listed public firms on the TSE, we exclude firms in the financial services industry because the main bank relationship is irrelevant for this group of firms. We also exclude firms operating in the regulated industries

5 In Japan, a fiscal year typically starts in April and ends in March next year. For example, the fiscal year 1991 corresponds to the year from April 1, 1991 to March 31, 1992. We use this expression throughout the paper.

(8)

(telecommunication, electricity and gas utilities) because important variables such as return on assets, leverage and main bank borrowing are likely to be influenced by regulatory factors rather than by market forces. For the most part of subsequent analysis, we focus only on firms that have bank loans. While excluding firms without bank loans from each year’s sample can be subject to selection biases, correcting possible selection biases produce materially no difference in results.6 Therefore, we only document the results based on the sample of firms with bank loans with one typical result of the robustness checks to these biases.

For each firm in the sample, we formally define the main bank as the largest, in terms of credit providing, of all creditor banks. If the largest creditor bank is singularly identified, we classify the firm as having a single main bank. If the number of the largest creditor banks is more than one, the firm is classified as having multiple main banks, each of which provides an identical amount of loans with other co-main bank(s).

We capture the effect of multiple main bank relationships on the hold-up costs by regressing the profitability of the firm on the variable indicating the existence of multiple main bank relationships and several control variables of the same firm. The economic rationale for this treatment is simple: if the multiple main bank relationship reduces the hold-up costs, it then should increase the firm’s profitability or at least partially offset the negative impact of the hold-up costs.

6 First, we run the Probit regression that determines whether the firm obtains bank loans. Second, in a subsequent regression based on the sample of firms with bank loans, we include estimates of the inverse Mill ratio obtained from the Probit regression among independent variables to take into account selection biases. This is Heckman’s (1979) proposed procedure. See Greene (2000, p. 926) for the details of the procedure.

(9)

We use return on assets (denoted by ROA) as the dependent variable. The numerator of ROA corresponds to earnings before interest and tax (EBIT), and the denominator to the sum of the market value of equity and the book value of total liabilities. In order to capture the impact of multiple main bank relationships, we use a dummy variable (denoted by MULTIB) which is equal to one if the firm has multiple main banks and zero otherwise. We use the main bank borrowing ratio (MAIN_BORD) defined as the ratio of the total main bank loans to the total liabilities to measure the strength of firm’s tie with the main bank(s). We also include an interaction term between the multiple main bank dummy and the main bank borrowing ratio in the regression together with other remaining independent variables. The interaction term might capture the changing response of firm’s ROA to the main bank borrowing ratio depending on having or not having additional main bank(s).

We use the asset value, leverage, Tobin’s q, and a keiretsu dummy as control variables.

The total asset value of the firm, i.e., firm size, is defined as the market value of equity plus the book value of total liabilities (FVALUE). In the actual regression, however, we use the natural log transformed total asset value (denoted by LFVALUE) rather than the original total asset value measure (FVALUE). This variable may capture the economies of scale or market power of the firm, which might be positively related with the firm’s ROA. The leverage ratio of the firm (LEVERAGE) is measured by the total liabilities divided by the total assets (FVALUE). If the firm has high leverage, the firm may have difficulties with raising additional funds, which is likely to force the firm to give up some of the profitable projects. Tobin’s q (TOBINQ) is conventionally measured in this study as the ratio of the sum of the market value of equity and the book

(10)

value of total liabilities to the book value of the total assets. We include this variable to control the regression for a potential impact of the firm’s growth opportunities on its profitability. Finally, we include a keiretsu dummy variable (BIG6) which is equal to one if the firm belongs to one of the six major keiretsu groups and zero otherwise. Hoshi, Kashyap and Sharfstein (1991) show that firms belonging to major keiretsu groups are financially less constrained, which may lead to higher profitability. As already reviewed, other researchers more recently document the opposite result. Finally, we include yearly and industry dummy variables to control for the effects of business cycles and sector demand fluctuations, respectively, on the firm’s profitability.

Table 1 shows the descriptive statistics of the variables used in this study. Some of them are not used in the basis regression specifications but used only for the robustness check of the basic regression result. The overall number of firm years is 10,344 throughout the sample period from 1991 through 1998, all of which meet the condition of having bank loans. According to Table 1, about 90.6 % (9,374 firm years) of the sample firms on average have a single main bank (UNIB), and 9.4 % (970 firm years) of the sample firms have multiple main banks (MULTIB). Among the firms that have multiple main banks, the average number of main bank relationships maintained is 2.4 (not reported in Table 1) with more than 75 % having only two main banks. Table 1 also shows that firms with a single main bank tend to have lower ROA, lower (log of) firm value, higher leverage, lower main bank borrowing ratio, and higher probability of keiretsu membership than firms with multiple main banks.

(11)

III. Empirical results

We capture the effect of multiple main bank relationships on the hold-up costs by regressing the profitability of the firm on the variable(s) indicating multiple main bank relationships and several control variables. The most basic regression results are shown in Table 2 with two alternative regression specifications. Equation (1) includes a multiple main bank dummy variable (MULTIB) which is equal to one if the firm has multiple main banks and zero otherwise. Equation (2), in addition to the multiple main bank dummy, includes the interaction term between the multiple main bank dummy and the main bank borrowing ratio (MULTIB*MAIN_BORD).7 For subsequent regression analysis, White’s (1980) t statistics are reported in referring to the statistical significance of the regression coefficients estimated.

Table 2 shows the OLS regression results with and without the interaction term under equation (1) and equation (2), respectively. The result of equation (1) shows that main bank borrowing ratio (MAIN_BORD) is significantly and negatively related to ROA at the 1 % level of statistical significance. This is consistent with the empirical hypothesis that the hold-up costs of main bank relationships decrease the profitability of the firm depending on the strength of a tie between the main bank(s) and the firm.8 On the other hand, the coefficient estimate of the multiple main bank dummy is significantly positive at the 1 % level, which is also consistent with the hypothesis,

7 In the current and subsequent regression analysis, we do not report estimates of coefficients corresponding to yearly and industry dummies to save space. These omitted results are available upon request.

8 A similarly significant result on the hold-up costs is obtained from the regression using the ratio of loans made by the two largest creditor banks to the total liabilities of the firm with a single main bank in stead of the strict main bank loan ratio.

(12)

based on Rajan (1992), that having multiple main bank relationships reduces the hold-up costs. We then allow the sensitivity coefficient of the main bank borrowing ratio to differ between the two groups of firms with and without the multiple main bank relationship by including the interaction term (MULTIB*MAIN_BORD) in equation (2).

The difference in the sensitivity coefficient on the main bank borrowing ratio between the single and the multiple main bank relationship case, measured by the coefficient on (MULTIB*MAIN_BORD), is significantly positive at the 1 % level, although the total sensitivity for multiple main bank firms (coefficient on MAIN_BORD + coefficient on MULTIB*MAIN_BORD) is still negative and statistically significant. Notice that the multiple main bank dummy is not statistically significant any more in this regression.

The results in Table 2 thus suggest that multiple main bank relationships reduce the negative effect, in magnitude, of the hold-up costs on the firm’s ROA.

We check the robustness of our main results in Table 2 by dividing the sample firms in several ways. First, we divide the sample into the firms that have public debts (including straight bonds, convertible bonds, bonds with warrants, and commercial papers) on their liabilities side of the balance sheet and the firms that do not have debt securities issued. We use the presence of outstanding debt securities as a proxy for the firm’s ability to alternatively raise capital. The firm with this ability may have already reduced the hold-up costs because of alternative debt financing available in the public markets. The multiple main bank relationship in reducing the hold-up costs is relatively unimportant in the former but important in the latter group. However, it is said that the bond issuance in Japan is often captured by the firm’s main bank or its investment banking subsidiary as a lead manager in bond underwriting. Thus, the effect of multiple

(13)

main bank relationships may be important, under this conjecture, for the firms with the access to public debt markets, too. Second, we construct two equally divided sub-samples based on the firm’s size (FVALUE). Information asymmetries tend to be severer for smaller firms than for larger firms and therefore, the benefit of more concentrated (i.e., single main bank) financing may outweigh the cost of such financing for the smaller firms. Thus, the effectiveness in hold-up cost reducing of multiple main banks may be limited for the smaller firms. Third, we also equally sub-divide the sample by the value of growth opportunities faced by each firm in terms of Tobin’s q.

Firms that have the higher value of growth opportunities can be subject to severer hold-up costs (Rajan (1992)). Then, the hold-up cost reducing effect of multiple main bank relationships may be more important for such firms. Table 3 reports the results of these robustness checks.

The columns under equations (1) and (2) in Table 3 show the results for the group of firms with public debts on the liabilities side of their balance sheets. In this sub-sample, the multiple main bank dummy (MULTIB) in equation (1) and its interaction term with the main bank borrowing ratio (MULTIB*MAIN_BORD) in equation (2) are positive and still statistically significant at the 5 % and the 1 % level, respectively. On the other hand, the columns under equations (3) and (4) show the results for the group of firms that have no public debts outstanding. Among these firms, the multiple main bank dummy is positive and still significant at the 5 % level in equation (3), although the positive interaction effect is only significant at the 10 % level in equation (4). The overall results suggest that the hold-up cost reducing effect of multiple main bank relationships still exits for both two sub-samples even though the

(14)

effect is slightly weaker for the group of the firms that have no access to the public debt market. The result is consistent with our second alternative view presented earlier: the main bank extract rents even in client firm’s bond issuance by exploiting bond market imperfection and this cost is reduced by introducing additional main bank(s) similar to the other group without the access to the public debt markets.

The columns under equations (5) to (6) in Table 3 show the results for the sub-sample of larger firms. While the coefficient estimate of the multiple main bank dummy is not statistically significant for both specifications, the coefficient of the interaction term included in equation (6) is significant at the 5 % level. Next, the columns under equations (7) and (8) show the results for the sub-sample of smaller firms. While the positive coefficient estimate of the interaction term in equation (8) is only significant at the 10 % level, that of the multiple main bank dummy is significant at the 5 % level in equation (7). These results are somewhat inconsistent with our prediction and the international evidence. The hold-up problem mitigating effect of multiple main bank relationships for smaller firms is at least as significant as that for larger firms, and thus the result is robust regardless of firm-size ranges.

Our results on the Rajan’s (1992) conjecture are shown under the columns of equations (9) through (12) in Table 3. In the sub-sample consisting of the firms with more growth opportunities, the coefficients of the multiple main bank dummy in equation (9) and the interaction term in equation (10) are significantly positive at the 5 % and the 1 % level, respectively. In the sub-sample of the firms with fewer growth opportunities, on the other hand, neither the multiple main bank dummy in equations

(15)

(11) and (12) nor the interaction term in equation (12) is statistically significant at the conventional levels. Notice that the coefficients estimated on the main bank borrowing ratio (MAIN_BORD) for the firms with lower q is about 50 %, in size, of those for firms with higher q, which suggests that the hold-up costs for lower q firms are less important than for higher q firms.9 Our interpretation is that the firms with more growth opportunities can more effectively use the multiple main bank relationship as a means of reducing their otherwise very high hold-up costs. On the other hand, the multiple main bank relationship is not very effective for those with less growth opportunities (i.e., those with lower hold-up costs). Thus, the hold-up cost reducing effect of multiple main bank relationships is largely limited to the group of firms with more growth opportunities.

So far, our regression analysis has been based solely on the sample of the firms that have bank loans actually outstanding. Given the fact that we excluded firms with no bank loans from our sample, the coefficients estimated with this sample can be subject to sample selection biases. For example, some unknown factors that encourage the firm to borrow (or not to borrow at all) from the bank may also affect the profitability of the firm. In order to take into account this kind of possibly existing selection biases, we first estimate the Probit model that determines whether the firm borrows from the bank or not. Then, the implied inverse Mill ratio in the first-stage Probit regression is used as one of the independent variables in the second-stage ROA regression.10

9 The difference in the coefficients between two sub-samples is statistically significant at the 1 % level.

10 This procedure follows Heckman (1979).

(16)

For the Probit regression, as independent variables we include the log of firm value (LFVALUE), the leverage (LEVERAGE), Tobin’s q (TOBINQ), dummy variable for the membership in the six major keiretsu groups (BIG6), the log of the number of months since listing on the TSE (LAGE), and dummy variable (MBOND) which is equal to one if the firm has debt securities outstanding in their liabilities and zero otherwise. The rationale for each variable is as follows: the higher the firm value is, the larger the economies of scale in issuing debt securities, which implies the negative relationship between the firm value and the probability of having bank loans. The higher the leverage is, the higher the firm’s need for bank loans in addition to public debts and/or trade credits. The more important the value of firm’s growth opportunities is, the higher the agency cost of public debts is, which implies the positive relationship between Tobin’s q and the probability of having bank loans. Firms belonging to traditional, major keiretsu groups are more likely to borrow from keiretsu-affiliated banks. The older the firm is, the stronger the firm’s (transaction-based) tie with banks is, which leads to higher probability of having bank loans.11 Alternatively, younger firms may rely more on bank financing because of the higher degree of information asymmetries applied to them. Finally, firms that are able to have an access to the public debt markets tend to use bank loans less frequently. The result of the Probit regression is shown in Panel A of Table 4. All independent variables are statistically significant at the 1% level. In addition, the signage of all coefficients is consistent with the prediction made above.

In the next stage, we re-run the ROA regression by additionally including the inverse

11 The number of months since listing on the TSE is used as a proxy for the age of the firm. This is because we do not have precise data on the date of incorporation for many sample firms.

(17)

Mill ratio implied by the Probit regression in Panel A. The results of the ROA regression with the implied inverse Mill ratio are shown in Panel B of Table 4. While the coefficient estimate of the inverse Mill ratio (IMILL) is statistically significant at the 1 % level for both specifications, the overall results are qualitatively very similar to those previously documented in Table 2. Specifically, the coefficient of the multiple main bank dummy (MULTIB) is significantly positive in equation (1), and that of the interaction term between the multiple main bank dummy and the main bank borrowing ratio (MULTIB*MAIN_BORD) is also significantly positive at the 1 % level. These suggest that the results in Table 2 were not driven by sample selection biases. We have also examined all other regression results by including the estimated inverse Mill ratio in each equation and have found that the results are very similar in all cases.12

IV. Conclusion

In this study, we investigate whether the multiple main bank relationship is related to the reduction of the “hold-up costs” of bank financing (Rajan (1992)) by using the panel data of Japanese companies listed on the Tokyo Stock Exchange, the first and second sections included, during the period from 1991 to 1998. The relationship between firms and their main banks provides us with a unique experimental setting with the Japanese data. For most Japanese firms, the relationship with their main bank(s) is quite stable over time and, in the case of multiple main bank relationships, it is almost fixed throughout our sample. This implies that we are able to measure the effectiveness

12 The results are available upon request.

(18)

of multiple main bank relationships without being much concerned with the endogeneity issue associated with selecting main bank(s).

We capture the effect of multiple main bank relationships on the hold-up costs by regressing the profitability of the firm (ROA) on the variables indicating multiple main bank relationships and those to control the regression. Our empirical results are as follows: first, main bank borrowing is negatively related to the profitability of the firm, which suggests the presence of significant hold-up costs. Second, the multiple main bank relationship reduces the hold-up costs and lead firms with this feature to higher profitability. This hold-up problem mitigating effect of multiple main banks is robust irrespective of the firm’s access to the public debt markets and the firm’s size. Third, this mitigating effect of multiple main bank relationships is larger for firms with the higher value of growth opportunities than those with the lower value of growth opportunities. Finally, the correction on sample selection biases does not change the results qualitatively. Thus, the hold-up cost mitigating effect of multiple main bank relationship is robust except for the asymmetric effectiveness with respect to the value of firm’s growth opportunities.

The main bank features as both major creditor and shareholder do not equally apply: the main bank does not play the role as a principal of shareholders in general but do play the role as a major creditor. Rents extracted from their client firms are probably too high to justify as shown in the negative and significant sensitivity coefficient (from -0.06 to -0.07) to the main bank borrowing ratio (with an average of 0.098) when the average ROA is as low as 0.025. The average percentage hold-up cost over our sample is about

(19)

20 percent of the ROA, 0.006/(0.006 + 0.025) or 0.007/(0.007 + 0.025). which results in the sample average ROA at 0.025. Notice than the percent hold-up cost is the positive function of the main bank borrowing ratio for the given level of ROA. The corresponding hold-up cost with multiple main bank relationship is between 11 percent and 12 percent. Given this, it is likely that the multiple main bank relationships reflect one of the viable solutions that management may possess to reduce the hold-up costs. It is not known in this study why many firms (about 90 % of the sample) rather knowingly maintain the costly single main bank relationship over the less costly multiple main bank relationship. Probably these firms have been forced by the existing single main bank not to increase the number of largest creditor banks as the words of “hold-up costs” exactly indicate.

(20)

References

Aoki, M., H. Patrick and P. Sheard, 1994, “The Japanese main bank system: An introductory overview,” Chapter 1, in Aoki, M, and H. Patrick, eds., The Japanese Main Bank System - Its Relevance for Developing and Transforming Economies, Oxford University Press, New York , p3-50.

Berlin, M. and J. Loeys, 1988, “Bond covenants and delegated monitoring”, Journal of Finance 43, p397-412.

Cole, R.A., 1998, “The importance of relationships to the availability of credit”, Journal of Banking and Finance 22, p959-977.

Degryse, H., and S. Ongena, 2001, “Bank relationships and firm profitability”, Financial Management Spring 2001, p9-34.

Diamond, D., 1984, “Financial intermediation and delegated monitoring,” Review of Economic Studies 51 (1984), p393-414.

Diamond, D., 1991, “Monitoring and reputation: The choice between bank loans and directly placed debt, Journal of Political Economy 99, p689-721.

Fama, G., 1985, What’s different about banks?”, Journal of Monetary Economics 15, p29-39.

Greene, W., 2000, “Econometric Analysis”, forth edition, Prentice Hall, New Jersey.

Heckman, J., 1979, “Sample selection bias as a specification error”, Econometrica 47, p153-161.

Hiraki, T., H. Inoue, A. Ito, F. Kuroki, and H. Masuda, 2003, “Corporate governance and firm value in Japan: Evidence from 1985 to 1998”, Pacific-Basin Finance Journal, forthcoming.

Horiuchi, T., 1994, “The effect of firm status on banking relationships and loan syndication”, Chapter 8, in Aoki, M, and H. Patrick, eds., The Japanese

(21)

Main Bank System - Its Relevance for Developing and Transforming Economies, Oxford University Press, New York , p258-294.

Hoshi, T., A. Kashyap, and D. Sharfstein, 1990, "The role of banks in reducing the costs of financial distree in Japan," Journal of Financial Economics 27, p67-88.

Hoshi, T., A. Kashyap, and D. Sharfstein, "Corporate structure, liquidity, and investment: evidence from Japanese industrial groups," Quarterly Journal of Economics 106 (1991), p33-60.

Houston, J., and C. James, 1996, “Bank information monopolies and the mix of private and public debt claims”, Journal of Finance 51, p1863-1889.

James, C., 1987, “Some evidence on the uniqueness of bank loans, Journal of Financial Economics 19, p217-235.

Kang, JK., and A. Shivdasani, “Firm performance, corporate governance, and top executive turnover in Japan,” Journal of Financial Economics 38 (1995), p29-58.

Kang, JK. and A. Shivdasani, "Corporate restructuring during performance declines in Japan," Journal of Financial Economics 46 (1997), p29-65.

Kaplan, S. and B.A. Minton, "Appointments of outsiders to Japanese boards:

Determinants and implications for managers," Journal of Financial Economics 36 (1994), p225-58.

Lummer, S.L., and J.J. McConnell, 1989, “Further evidence on the bank lending process and the capital market response to bank loan agreements”, Journal of Financial Economics 25, p99-122.

Morck, R., and M. Nakamura, "Banks and corporate control in Japan," Journal of Finance 54 (1999), p319-39.

Morck, R., M. Nakamura, and A. Shivdasani, “Banks, ownership structure, and firm value in Japan,” Journal of Busienss 73 (2000), p539-567.

(22)

Peterson, M., and R. Rajan, 1994, “The benefits of lender relationships: Evidence from small business data”, Journal of Finance 49, p3-37.

Pinkowitz, L., and R. Williamson, 2001, “Bank power and cash holdings: Evidence from Japan,” Review of Financial Studies 14, p1059-1082.

Rajan, R., 1992, "Insiders and outsiders: The choice between relationship and arm's length debt," Journal of Finance 47, p1367-1400.

Sharpe, S.A., 1990, “Asymmetric information, bank lending, and implicit contracts: A stylized model of customer relationships”, Journal of Finance 45, p1069-1087.

von Thadden, EL, 1994, “The commitment of finance, duplicated monitoring and investment horizon”, ESF-CEPR Working Paper in Financial Markets 27.

Weinstein, D.E., and Y. Yafeh, "On the costs of a bank centered financial system:

Evidence from the changing main bank relations in Japan," Journal of Finance 53 (1998), p635-72.

White, H., 1980, “A heteroscedasticity-consistent covariance matrix estimator and a direct rest for heteroscedasticity”, Econometrica 48, p817-838.

(23)

Table 1.

Basic statistics of firm characteristic variables for firms that have bank loans from 1991 to 1998

# of Obs 10344 9374 970

Mean Std Median Mean Std Median Mean Std Median

ROA 0.0290 0.0327 0.0272 0.0284 ** 0.0328 0.0270 0.0345 0.0318 0.0295

FVALUE 277939 662420 79309 272873 641665 77758 326898 835716 99578

LFVALUE 11.4408 1.3673 11.2811 11.4183 ** 1.3744 11.2614 11.6580 1.2780 11.5087

LEVERAGE 0.5231 0.1892 0.5204 0.5265 ** 0.1891 0.5241 0.4902 0.1877 0.4740

AGE 385.92 161.64 413.00 388.44 ** 160.84 414.00 361.59 167.34 393.00

LAGE 5.7481 0.8736 6.0235 5.7572 ** 0.8704 6.0259 5.6602 0.9005 5.9738

TOBINQ 1.3069 0.4816 1.2169 1.3061 0.4760 1.2166 1.3144 0.5322 1.2198

UNIB 0.9062

MULTIB 0.0938

MAIN_BORD 0.0983 0.0871 0.0776 0.0958 ** 0.0842 0.0762 0.1226 0.1087 0.0936

BIG6 0.1068 0.1089 * 0.0866

MBOND 0.6204 0.6175 0.6485

* (**) The difference in means between firms with a single main bank and with multiple main banks is significant at 5% (1%).

Overall Single Main Bank Multiple Main Bank

ROA = return on assets; FVALUE = book value of total liabilities + market value of equity ; LFVALUE = log of FVALUE;

LEVERAGE = total liabilities divided by firm value (FVALUE); AGE = number of months since the firm's stock was listed on the TSE; LAGE = log of AGE; TOBINQ = Tobin's q measured by the ratio of the book value of total liabilities plus the market value of equity to the book value of total assets; UNIB = dummy variable equal to one if the firm has a single main bank and zero otherwise; MULTIB = dummy variable equal to one if the firm has multiple main banks and zero otherwise;

MAIN_BORD = main bank loans divided by total liabilities; BIG6 = dummy variable equal to one if the firm belongs to one of the six major keiretsu groups and zero otherwise; MBOND = dummy variable equal to one if the firm has corporate bonds, convertible bonds or bonds with warrants in its liabilities and zero otherwise.

(24)

Table 2.

Variable (1) (2)

C 0.0276 0.0286

(6.654)** (6.850)**

LFVALUE 0.0030 0.0029

(10.007)** (9.715)**

LEVERAGE -0.0565 -0.0563

(-21.756)** (-21.663)**

TOBINQ 0.0079 0.0080

(4.961)** (5.018)**

MULTIB 0.0027 -0.0012

(2.925)** (-0.855)

MAIN_BORD -0.0622 -0.0675

(-11.740)** (-11.383)**

MULITIB*MAIN_BORD 0.0327

(3.244)**

BIG6 -0.0097 -0.0097

(-11.029)** (-10.991)**

# of Obs 10344 10344

F-value 105.81 ** 103.74 **

Adjusted R-Square 0.2935 0.2944

The dependent variable is ROA. The definition of variables is as follows: ROA = return on assets; LFVALUE

= log of the book value of total liabilities + market value of equity; LEVERAGE = total liabilities divided by firm value (FVALUE); TOBINQ = Tobin's q measured by the ratio of the book value of total liabilities plus the market value of equity to the book value of total assets; MULTIB = dummy variable equal to one if the firm has multiple main banks and zero otherwise; MAIN_BORD = main bank loans divided by total liabilities; BIG6 = dummy variable equal to one if the firm belongs to one of the six major keiretsu groups and zero otherwise;

t statistics based on White (1980) standard error are shown in parentheses. * (**) Significant at 5% (1%).

OLS regression of ROA on firm characteristic variables and main bank relation for the pooled sample from 1991 to 1998

(25)

Table 3.

OLS regression of ROA on firm characteristic varaibles and main bank relation with various sample splits from 1991 to 1998

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Variable

C 0.0332 0.0345 0.0104 0.0110 0.0404 0.0410 0.0075 0.0085 0.0486 0.0502 0.0119 0.0122

(6.220)** (6.459)** (1.277) (1.347) (6.708)** (6.813)** (0.811) (0.915) (6.671)** (6.831)** (2.418)* (2.482)*

LFVALUE 0.0011 0.0010 0.0054 0.0054 0.0003 0.0002 0.0067 0.0066 0.0024 0.0022 0.0020 0.0020

(3.111)** (2.833)** (8.059)** (8.012)** (0.715) (0.577) (8.163)** (8.070)** (4.790)** (4.475)** (6.088)** (6.023)**

LEVERAGE -0.0528 -0.0528 -0.0525 -0.0524 -0.0546 -0.0543 -0.0562 -0.0562 -0.0707 -0.0704 -0.0414 -0.0413

(-12.950)** (-12.824)** (-13.831)** (-13.806)** (-12.932)** (-12.896)** (-17.983)** (-17.966)** (-14.585)** (-14.515)** (-16.447)** (-16.414)**

TOBINQ 0.0163 0.0162 0.0044 0.0045 0.0156 0.0157 0.0008 0.0009 0.0037 0.0038 0.0237 0.0237

(5.550)** (5.474)** (2.159)* (2.207)* (6.468)** (6.537)** (0.551) (0.608) (1.841) (1.919) (7.687)** (7.685)**

MULTIB 0.0023 -0.0013 0.0037 -0.0007 0.0013 -0.0014 0.0034 -0.0007 0.0033 -0.0030 0.0019 0.0008

(2.157)* (-0.886) (2.019)* (-0.233) (1.263) (-0.876) (2.203)* (-0.273) (2.254)* (-1.463) (1.724) (0.452)

MAIN_BORD -0.0664 -0.0732 -0.0576 -0.0615 -0.0466 -0.0527 -0.0627 -0.0668 -0.0846 -0.0925 -0.0434 -0.0450

(-8.450)** (-8.103)** (-7.913)** (-7.674)** (-6.941)** (-6.637)** (-9.167)** (-8.831)** (-9.114)** (-8.961)** (-8.626)** (-8.526)**

MULTIB*MAIN_BORD 0.0395 0.0265 0.0295 0.0281 0.0583 0.0085

(2.762)** (1.710) (2.326)* (1.906) (3.809)** (0.653)

BIG6 -0.0066 -0.0065 -0.0146 -0.0147 -0.0056 -0.0056 -0.0105 -0.0106 -0.0106 -0.0105 -0.0055 -0.0055

(-7.808)** (-7.713)** (-4.644)** (-4.661)** (-5.667)** (-5.656)** (-1.814) (-1.825) (-8.257)** (-8.131)** (-5.458)** (-5.471)**

# of Obs 6417 6417 3927 3927 5172 5172 5172 5172 5172 5172 5172 5172

F-value 82.93 ** 81.28 ** 33.93 ** 33.22 ** 80.10 ** 78.39 ** 40.99 ** 40.16 ** 57.91 ** 57.02 ** 35.48 ** 34.66 **

Adjusted R-Square 0.3436 0.3445 0.2559 0.2563 0.3854 0.3860 0.2407 0.2413 0.3109 0.3127 0.2147 0.2147

Tobin's q>median Tobin's q <=median

The dependent variable is ROA. The definition of variables is as follows: ROA = return on assets; LFVALUE = log of the book value of total liabilities + market value of equity; LEVERAGE = total liabilities divided by firm value (FVALUE); TOBINQ = Tobin's q measured by the ratio of the book value of total liabilities plus the market value of equity to the book value of total assets; MULTIB = dummy variable equal to one if the firm has multiple main banks and zero otherwise; MAIN_BORD = main bank loans divided by total liabilities; BIG6 = dummy variable equal to one if the firm belongs to one of the six major keiretsu groups and zero otherwise;

t statistics based on White (1980) standard error are shown in parentheses. * (**) Significant at 5% (1%).

Firms that have bonds Firms that do not have bond Firm value>median Firm value<=median

(26)

Table 4.

Probit Estimation and OLS regression: 1991 - 1998

Panel A. Probit Estimation Panel B. OLS regression with slection bias correction

(1) (2)

Variable Variable

C -0.6207 C 0.0150 0.0159

(-3.520) ** (3.413) ** (3.597) **

LFVALUE -0.1440 LFVALUE 0.0019 0.0019

(-9.808) ** (5.640) ** (5.371) **

LEVERAGE 4.6635 LEVERAGE -0.0262 -0.0257

(40.763) ** (-5.180) ** (-5.075) **

TOBINQ 0.5459 TOBINQ 0.0110 0.0112

(16.205) ** (6.737) ** (6.813) **

BIG6 0.3578 MULTIB 0.0029 -0.0012

(5.438) ** (3.205) ** (-0.837)

LAGE 0.1089 MAIN_BORD -0.0629 -0.0684

(6.760) ** (-11.953) ** (-11.647) **

MBOND -0.0982 MULTIB*MAIN_BORD 0.0346

(-2.697) ** (3.428) **

BIG6 -0.0068 -0.0068

(-6.890) ** (-6.819) **

IMILL 0.0250 0.0253

(7.011) ** (7.092) **

# of Obs 12355 # of Obs 10344 10344

Log Likelihood -4102.77 F-value 105.68 ** 103.72 **

R-Square 0.2641 Adjusted R-Square 0.2983 0.2993

Panel A shows the resutl of Probit estimation which predicts wether the firm has bank loans (=1) or not (=0).

Panel B OLS regression regresses ROA on firm chracteristic variables, main bank relation, and the inverse Mil ratio calculated in the Probit estimation in Panel A.

ROA = return on assets; FVALUE = book value of total liabilities + market value of equity ; LFVALUE = log of FVALUE; LEVERAGE = total liabilities divided by firm value (FVALUE); AGE = number of months since the firm's stock was listed on the TSE; LAGE = log of AGE; TOBINQ = Tobin's q measured by the ratio of the book value of total liabilities plus the market value of equity to the book value of total assets; MULTIB = dummy variable equal to one if the firm has multiple main banks and zero otherwise; MAIN_BORD = main bank loans divided by total liabilities; BIG6 = dummy variable equal to one if the firm belongs to one of the six major keiretsu groups and zero otherwise; MBOND = dummy variable equal to one if the firm has corporate bonds, convertible bonds or bonds with warrants in its liabilities and zero otherwise; IMILL = inverse Mill ratio calculated in the Probit estimation in Panel A to take into account selection bias.

t statistics based on White (1980) standard error are shown in parentheses. * (**) Significant at 5% (1%).

References

Related documents

• Overall Theme: 3D graphics visualization used to leverage the power of the human visual pattern recognition system • Geophysical field data visualized: – Micro-seismic event