Effects of Regional Bank Merger on Small Business Borrowing : Evidence from Japan

This study examines the impact of a regional bank merger in Japan on borrowing by small businesses, focusing on firms that borrow from the acquiring bank, the acquired bank, or both. First, we find that post-merger borrowing costs declined. This result suggests that small borrowers enjoy more favorable post-merger financing conditions because efficiencies from economies of scale lead to lower costs. Second, we find that post-merger borrowing costs decline for firms that borrow only from the acquiring or acquired bank, whereas they did not decline for firms that borrow from both. Third, we find that only small business loans to firms that borrow from both the acquiring and acquired banks decrease post-merger. This result suggests that small business lending might decline because of a merged bank’s loan portfolio and lending strategy.


Introduction
Since his 2012 re-election, Prime Minister Shinzo Abe has sought to revive Japan's sluggish economy through aggressive monetary easing by the Bank of Japan, massive fiscal stimulus, and structural reforms.To strength local economies, Abe's cabinet and the Financial Services Agency promote consolidations among local banks in a regional banking model resembling that of Europe and North America.Twenty years of mergers have consolidated Japan's city banks into three mega-banks, but the number of regional banks remains high despite mergers and acquisitions (M&A), causing -over banking‖ in some areas.The number of bankruptcies among small business enterprises (SBEs) has risen since March 2013 following the expiration of the Small and Medium-Sized Enterprises Financing Facilitation Act.The number of nonperforming loans held by regional banks also has risen, the competition rate among lenders intensifies, and Japan's greying population and declining birth rate complicate attracting borrowers.Consequently, Japan's regional banks face challenging conditions as Abe's government promotes their consolidation, and further widespread reorganization of regional banks is expected.
European and American regional banks have consolidated steadily since 1990, and researchers have widely examined the effects.Sapienza (2002) reached three conclusions regarding the effects of Italian bank mergers on business borrowers.First, if mergers occurred between banks operating in the same province (in-market mergers), post-merger borrowers from the consolidated bank paid significantly lower interest rates.Second, borrowers with multiple banking relationships had access to broader external sources of finance and were unaffected by merger activity.Third, newly merged banks reduced lending to SBEs.
Using proprietary data for US commercial banks, Erel (2011) found significant reductions in spreads for small loans following mergers.This implies that small borrowers pay lower interest rates after banks expanded through mergers.Unlike Sapienza (2002), Erel (2011) found that acquiring banks increase lending to SBEs after M&A.
Scant research has investigated M&A among Japanese banks.Shin, Fraser, and Kolari (2003) examined the 1999 Mizuho merger between the Industrial Bank of Japan, Dai-ichi Kangyo Bank, and Fuji Bank to investigate whether consolidation affected bank-firm relationships.Their event study methodology revealed that stock prices of all firms that had borrowed from the acquiring banks declined following the merger.However, declines were significantly less among firms that used one of the three as their main bank or had large credit exposure with one.Uchino and Uesugi (2012) suggested that Japan's largest bank merger-Bank of Tokyo-Mitsubishi and UFJ Bank in 2005-engendered higher borrowing costs partly through exogenous reduction in the number of firm-bank relationships and partly from changes in the acquired bank's organizational structure, including consolidation of branch networks.Ogura and Uchida (2013) found that small bank mergers impair bankers' acquisition of soft information about borrowers (Note 1).This suggests that mergers complicate managerial structures and hinder acquisition and maintenance of soft information among small banks.
To our knowledge, only the aforementioned three studies examine the impact of Japanese bank M&A on SBE borrowers.Our study differs from these studies in two respects.First, Shin et al. (2003) and Uchino and Uesugi (2012) analyzed how megabank mergers affect comparatively larger and well-established corporate borrowers.These borrowers likely are less dependent on indirect financing, whereas small unlisted firms-i.e., most regional businesses-lack easy access to capital markets and likely depend on bank borrowing.We employ a unique dataset of small, unlisted firms to more accurately observe the effects of a landmark regional bank consolidation.Our study offers the literature's first empirical insights into the effects of regional bank consolidation in Japan.
Second, Ogura and Uchida (2013) surveyed SME characteristics only once and only for 2005.Our analysis extends to periods following the merger, using time-series financial data.By considering differences between pre-and post-merger parameters, we directly measure its long-run effect.
Global studies on banking consolidation employ firm-bank matched panel data to analyze interest rates that consolidating banks charge corporate borrowers for credit lines.We could not employ such a dataset due to the paucity of interest rate data in Japan, a difficulty that may explain why few studies investigate the impact of Japanese bank M&A on borrowers.We could not isolate interest rates on individual loans from different borrowers, including the post-merger consolidated bank, and firms can borrow from non-merged banks in their market, making it difficult to gauge the effects of mergers on borrowers.Therefore, we sampled firms that borrowed only from the acquiring bank, the acquired bank, and both.We investigate the impact of the merger more precisely using this set of firms.In this respect, our study contributes to the literature of bank M&A.
This study is structured as follows.Section 2 discusses the theoretical background.Section 3 presents the dataset, model, and results of cross-sectional analysis.Section 4 summarizes and concludes the study.

Theoretical Background
Theoretical literature posits that bank M&A exerts positive and negative effects on corporate borrowers through loan rates and quantity of lending.Therefore, we investigate the positive effects of lower borrowing costs through efficiency gains from economies of scale and scope and the negative effects when banks exercise market power to charge corporate borrowers higher rates.M&A also might engender higher borrowing costs if merged banks persuade government to maintain their dominant market positions.Moreover, newly merged banks might be excessively large and complex, generating higher costs through losses of efficiency and soft information.
Hypothesis 1: Efficiency gains after a merger might produce favorable financing conditions for SBEs.Alternatively, lending conditions might worsen if mergers raise banks' costs or credit market concentration.
We investigated the relationship between borrowing costs and the number of banks from which each SBE in our dataset borrowed.Diamond (1984) showed that borrowing from a single bank is advantageous because it entails lower monitoring costs.These findings suggest that borrowing from a single bank might reduce monitoring costs and therefore borrowing costs.Alternatively, Peterson and Rajan (1994) showed that firms borrowing from multiple banks pay significantly higher rates in the US.This finding might indicate that the number of banks from which firms borrow is a proxy for their creditworthiness.This implies that banks might not lend additional funds to low-quality firms, forcing them to borrow elsewhere at higher rates.In sum, borrowing costs rise as the number of a firm's lenders increases.Contrary to Peterson and Rajan (1994), however, Harhoff and Körting (1998) did not draw that conclusion for German credit markets.
We formulate Hypothesis 2 to examine the foregoing issues in the Japanese regional credit markets: Hypothesis 2: Pre-merger firms that borrow from both the acquiring and acquired banks pay higher borrowing costs than firms that borrow only from the acquiring or the acquired bank.Moreover, as mergers potentially reduce the number of banks from which firms can borrow, borrowing costs might decline for firms that borrow from both banks.Third, we investigate whether M&A affects quantity of lending.This issue relates to the size and composition of banks' loan portfolios, their lending strategies, and other factors.Peek and Rosengren (1998) showed that mergers may promote small business lending if the acquiring bank is small or its portfolio has a larger share of small business loans than the acquired bank.Erel (2011) suggested that post-merger increases in numbers of SBE loans might result from changes in lending technologies, acquiring banks' diversification strategies, and technological improvements (e.g., credit scoring models).Mergers of regional banks may reduce competition.This, in turn, may reduce SBE lending in a manner similar to how competition may prompt SBE lending because competition has the tendency to influence banks to seek more profit opportunities.Berger et al. (1998) decomposed the effects of banking consolidation on shares of SBE lending into one static and three dynamic effects that curtail SBE lending.
We formulate Hypothesis 3 to examine conflicting findings in the literature:  Hypothesis 3: SBE lending rises after mergers.If acquiring banks are small or their portfolios hold a greater proportion of SBE loans than acquired banks, loans to firms that borrow from both the acquiring and acquired banks increase post-merger.
The next section uses the firm-level dataset to examine these three hypotheses.The list of variables and their definitions are indicated in Table 1.

Methodology
We focus on the Kyushu region, where a spate of regional bank consolidations has occurred.In addition, we particularly examine Japan's landmark regional bank merger: Shinwa Bank and Kyushu Bank in 2003.We select data from Tokyo Shoko Research, Ltd (TSR) for 150 business borrowers from Tosho Shinyo-Roku (the Kyushu region) that transacted with either or both banks during 2000-2007.Isolating these 150 firms enabled us to analyze the Shinwa-Kyushu merger clearly.This is a unique aspect of our study.Our data are obtained from the TSR-Van2 database that daily updates information for 1.9 million Japanese corporations, including business profiles, financial data, bankruptcy information, and marketing reports.We examined unlisted regional firms.
Table 2 reports descriptive statistics for variables.The upper row (lower row) of Table 3 shows pre-merger data (post-merger data).The mean value of ) log( it r declines from 1.141 pre-merger to 0.908 post-merger.The mean value of it it ASSETS LOANS / declines from 3.362 pre-merger to 3.017 post-merger.Empirical analysis determined whether the merger affected borrowing costs and quantity of lending. We divided the 2000-2007 period into pre-merger (2000)(2001)(2002)(2003) and post-merger (2004-2007) subsamples.The post-merger period may seem brief, but bank managers, consultants, researchers, and financial media regard three years as sufficient for a full transition to a post-merger equilibrium (Berger et al., 1998) (Note 2).The audit period for the analyzed firms varies, making it impossible to specify exact yearly data.Therefore, we use pooled ordinary least squares without considering year effects.First, we estimate the following regression equation for firm i at time t: ) log( it r is the logarithm of borrowing cost for firm i at time t and is calculated using interest and discount expenses divided by the sum of long-term loans, short-term loans, and notes discounted during the previous year. Within the framework of Equation (1), Hypothesis 1 yields testable predictions for all coefficients.The main independent variable ( i A AFTERM & ) equals 1 for 2004-2007 and 0 otherwise.Hypothesis 1 that post-merger financing might favor small borrowers due to efficiency gains implies a negative sign for 7  .The co-hypothesis that conditions worsen if mergers engender higher costs or credit market concentration implies a positive sign for .
To control for variations in borrowing costs, we include demand-side, supply-side, interest rate-specific, merger-related, and firm-specific variables., an indicator of a soundly managed and performing firm, is ordinary income divided by sales.The higher its value, the lower is the interest rate charged by the bank, implying a negative sign for 1 DEBTRATIO , a proxy for a firm's fundamental risk characteristics, is calculated as debt divided by total assets.A higher value indicates a riskier firm and a higher interest rate for lending.We expect the sign of to be positive.
Supply-side variable is calculated as (short-term and long-term) loans i divided by the sum of loans from both Shinwa (the acquiring bank) and Kyusyu (the acquired bank).As loans from the banks increase, rates decline.We expect the sign of 3  to be negative.
Three interest rate-specific variables control for economy-wide variations in underlying cost of capital.Among the three, The logarithm of a firm's age, ) log( it AGE is calculated as the number of years since a firm was established.
Young firms generally pay higher rates than established firms.However, lenders might offer younger firms a lower rate to encourage long-term relationships and future borrowing.In such a case, younger firms might pay lower rates than established firms.A negative value for 6  indicates an inverse relation between age and interest rate; a positive value suggests the opposite.
The regression includes creditworthiness dummies VALUEA and VALUEC .The Tosho Shinyo-Roku report rates firms A, B, and C, with A being the highest rating and C being the lowest.
VALUEA and VALUEC equal 1 if a firm's credit is A-rated or C-rated and 0 otherwise.Second, we examined Hypothesis 2 by estimating the effect of a reduction in the number of banks from which SBEs borrow.Our estimated model has these specifications: Alongside independent variables in Equation ( 1), we include it BOTH and the interaction variable BOTH equals 1 if the SBE had transacted with both Shinwa and Kyusyu and 0 otherwise.A positive value for  8 supports Hypothesis 2 that firms borrowing from the acquiring and acquired banks pay higher costs than firms that borrow from one or the other.A negative value for  9 suggests that post-merger costs for SBEs that borrowed from both Shinwa and Kyusyu declined.
To examine Hypothesis 3, we estimate the effect of mergers on quantity of lending.Our basic estimated model has these specifications: LOANS / is calculated as loans (short-term and long-term loans) divided by assets.1) and ( 2).According to Hypothesis 3, a positive coefficient for 6  implies that post-merger lending increases.A positive value for 8  suggests that loans to firms borrowing from both Shinwa and Kyusyu increased post-merger.

The main independent variables are
The logarithm of sales ( ) log( it SALES ) is a proxy for firm size.As it increases, lending to the firm generally increases.Therefore, the expected sign of 1  is positive.This regression includes VALUEA , and VALUEC employed in Equations ( 1) and ( 2).The higher the value of it it SALE PROFIT / , the more likely banks are to lend the firm money, implying a positive sign for 2  .As it r is the borrowing cost for firm i at time t, the expected sign for is negative.
We expect t i DEBTRATIO to negatively correlate with it it SALES LOANS / in Equation (3).However, after Japan's bubble economy burst, some banks lent to deeply indebted firms, particularly in real estate and construction, a practice known as forbearance lending.Although the term has no single definition, Sekine, Kobayashi, and Saita (2003) define it as refinancing a loan (or providing additional credit) even though the borrower likely cannot repay the original amount.Many theoretical models examine motives for forbearance lending, but we believe banks simply hope to prevent the borrower's bankruptcy.If the bank engages in Considering the positive and negative effects of , we control for nonlinearities of it DEBTRATIO in the lending relationship.Therefore, we estimate Equation (4) as We also included a vector of t X to control for prefectural economic conditions.This vector comprises the sum of loans by both Shinwa and Kyusyu divided by prefectural GDP and the sum of prefectural loans divided by prefectural GDP.We expect 6  to be positive.are negative and statistically significant in all models, indicating rates declined as borrowing increased.

Results in
In columns V and VI, estimated coefficients of ) log( it AGE are positive and significant, implying that older SBEs paid higher rates to banks.This finding contravenes our expectations but endorses those of Peterson and Rajan (1995), who surmise that rural banks seek continuing relationships from small businesses and accept lower returns on initial loans.We identify the effect of rate smoothing on long-term relationships.Findings may be specific to regional markets, which have fewer banks and therefore greater bargaining power that strengthens as borrowers are locked into relationships.Thus, surviving older firms must pay higher rates if they cannot borrow elsewhere.This would explain the positive coefficients of ) log( it AGE (Note 4).In columns IV and VI, estimated coefficients of are positive and statistically significant in all models.This finding implies that SBEs' borrowing costs rose as long-term loan ratios increased, a finding consist with the theory.Estimated coefficients of are not statistically significant in any model except in column IV.
Estimated coefficients of and are negative and statistically significant in all models.These findings imply lower borrowing costs for A-rated and C-rated than for B-rated firms.Column VI is as our preferred specification because it includes all variables.
We ran regressions adding it BOTH and the interaction variable 5).Estimated coefficients of it BOTH are positive and statistically significant in all models.These results support Hypothesis 2 that firms which borrow from two banks pay higher rates than firms borrowing from one because concentrated borrowing reduces monitoring and borrowing costs.Alternatively, the number of lenders might be a proxy for firm quality.If so, our findings for Japan echo those of Peterson and Rajan (1994)  are not statistically significant in any model.These results contravene Hypothesis 3 that small business loans increase post-merger.Estimated coefficients of it BOTH are positive and statistically significant in columns III and IV.These results indicate greater lending to firms that borrowed from both Shinwa and Kyusyu than those that borrowed from one or the other.However, estimated coefficients of are negative and statistically significant in columns III and IV.These results suggest that post-merger loans to SBEs borrowing from both Shinwa and Kyusyu decreased.Our findings for those firms endorse findings of Sapienza (2002) and Berger et al. (1998) but contradict those of Erel (2011) and Peek and Rosengren (1998).One interpretation of our results is that banks' portfolios and lending strategies call for less SBE lending.

Notes
Note 1.The definition of soft information is vague-it indicates information that cannot be directly verified by anyone other than the agent who collected it and is difficult to communicate even within an organization.Attributes such as an entrepreneur's zeal and competence and employee morale cannot be unambiguously documented in a report that a loan officer can show to his superiors.See Stein (2002) and Ogura and Uchida (2013).
Note 2. This squares with Rhoades (1998) and Houston, James, and Ryngaert (2001).In these studies, post-merger cost savings and revenue gains take about three years.
Note 3. Note that although we selected 150 borrowing firms and a sample period between 2000 and 2007, the number of observations is 261 because of some missing data.
Note 4. Similar arguments are presented in studies pertaining to the effects of the duration of the relationship between the lending bank and borrower on the loan rate.Degryse and Cayseele (2000) and Degryse and Ongena (2005) found that the lending rate increases in tandem with the duration of the relationship.In contrast, Berger and Udell (1995) report that the duration of the relationship has a negative effect on the lending rate.Moreover, there are studies that present conflicting findings.For a survey of these findings, see Degryse, Kim and Ongena (2009).
Note 5. Following Degryse and Cayseele (2000), we attempted to address the potential simultaneity problem between ) log( it r and independent variables.As shown in Tables 4 and 5, deleting one independent variable (for instance, it it SALES PROFITS / ) does not affect the estimates of other variables.These results suggest that simultaneity problems in our estimated Equations ( 1) and (2) are minimal.Note 6.In the same manner as Equations ( 1) and (2), we attempted to address the potential simultaneity problem between it it ASSETS LOANS / and independent variables.Deleting ) log( it r does not affect the estimates of other variables, which is not reported in this study.From these results, we believe that simultaneity problems in our estimated Equations ( 3) and (4) are minimal.

Copyrights
Copyright for this article is retained by the author(s), with first publication rights granted to the journal.
This is an open-access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

2 
is the core variable, representing the logarithm of loan rates for Shinwa and Kyusyu.Borrowing costs are linked to ) log( t RATE , suggesting that 4  has a positive sign. it LOANSRATIO is the ratio of long-term loans calculated as long-term loans divided by the sum of short-term and long-term loans.Borrowing 7  costs rise as increases, implying a positive sign for .
are the same as in Equations (

Table 1 .
List of variables and their definitions and long-term loans) divided by tangible assets Firms' variables LOG(SALES) Logarithm of sales (thousands of yen) PROFITS/SALES Ordinary income divided by sales DEBTRATIO Debt divided by total assets LOANSRATIO Long-term loan ratio (long-term loans divided by the sum of short-term and long-term loans) (LOANS)/LOANS_TWO Loans (short-term and long-term loans) divided by the sum of loans of both Shinwa Bank and Kyusyu Bank LOG(AGE) Logarithm of firm's age BOTH 1 if the firm transacted from both Shinwa Bank and Kyusyu Bank, 0 otherwise.VALUEA 1 if credit rating of the firm is A, 0 otherwise.VALUEC 1 if credit rating of the firm is C, 0 otherwise.Banks' and regional variables LOG(RATE) Logarithm of loan rates of Shinwa Bank and Kyushyu Bank LOANS_TWO/GDP The sum of loans of both Shinwa Bank and Kyusyu Bank divided by prefectural GDP LOANS_TOTAL/GDP The sum of prefectural loans divided by prefectural GDP Mergers' variables AFTERM&A 1 if the year after M&A (2004-2007), 0 otherwise.

Table 2 .
Descriptive statistics of variables

Table 3 .
Descriptive statistics of subsample variables Table 4 summarize the estimated coefficients for Equation (1) in our models (Note 3).Columns I and II differ in that column I includes only rate-specific variables.Column VI is our preferred specification because it includes all variables.
Erel (2011)002)esti A AFTERM &are negative and statistically significant in all models.These findings endorse those ofSapienza (2002)andErel (2011), suggesting post-merger borrowing costs declined because efficiency gains create favorable financing conditions for SBEs as per Hypothesis 1.

Table 5
t RATE as an interest rate-specific variable.Columns IV and V include both it AGE .
) are nearly the same as in Table4.Table6summarizes estimated coefficients for Equations (3) and (4) in our models (Note 6).Columns I-IV show regression results for Equation (3).Columns V-VIII show regression results for Equation (4).To control for prefectural economic conditions, column I includes the sum of loans from both Shinwa and Kyusyu divided by prefectural GDP, VALUEC

Table 5 .
Results for Equation (2) VALUEA ) are not statistically significant.Estimated coefficients of VALUEC are positive and statistically significant only in columns III, IV, and VIII.