Competition, specialization and bank–firm interaction: what happens in credit crunch periods?

Abstract

This article empirically investigates the relationship between interbank competition, bank orientation and credit availability for a sample of more than 30 000 loans granted by a large banking group operating in the Italian credit market. We test whether and how, during a credit crunch period, competition affects bank orientation and how relationship lending and interbank competition can mitigate the credit crunch problem, for financially distressed firms. Using a unique and large bank–firm level data set, the main results show that an increase in competition is associated with a stronger relationship in terms of the length of the bank–borrower interaction, whereas the distance bank branch-headquarter negatively affects it. Moreover, a strong lender–borrower relationship, in terms of length and exclusivity, is found positively significant in determining the change in the amount of credit granted. Nonlinearity and sector specialization effects are tested, too, and report interesting results, supporting the crucial role of relationship lending during a financial crisis.

Keywords:

• competition
• relationship lending
• sector specialization
• financial crisis

JEL Classification:

• G01
• G21

Notes

¹ Relationship lending is costly as well. Boot (2000) identifies two primary costs: (i) first, borrowers know that they can easily ask the bank for more credit in case of distress or renegotiate their contracts ex post, thus may have perverse incentives ex ante in not preventing a bad outcome from happening (soft-budget constraint problem); (ii) at the same time, banks have an information monopoly, thus borrowers could be informationally locked in the relationship (hold-up problem).

² More precisely, we have collected for 2008 and 2010, the sum of loans of bank j aggregated at the borrower level.

³ It was also provided a measure of this index based on the amount of loans granted. The correlation matrix between the two Herfindahl–Hirschman indices shows a high, positive and significant coefficient of 0.7447 (p-value = 0.000), confirming the possibility of using alternatively one of them without any consequence on the results.

⁴ According to an interpretation shared in the literature, the level of interbank competition can be analysed as follows: high competition (HHI < 10), moderate competition (10 < HHI < 18), low competition (HHI > 18) and monopoly (HHI = 1).

⁵ We do not have collateral information for all 31 012 observations. Following Bharath et al. (2011), we consider such loans as unsecured.

⁶ The original range of ratings, as provided by the banks, is lightly different from the one presented; we have decided to standardize the original ratings in a range 1–10 in order to preserve the privacy of banks providing data.

⁷ Data of this study cannot be matched with the firm’s balance sheet data; for example, since for confidentiality purposes the banking group altered the borrower identities before providing us the data.

⁸ Further, our results are consistent with Ono and Uesugi (2009) that find that borrowers who establish long-term relationships with their main banks are more likely to pledge collateral. By contrast, other studies argue that stronger bank–firm relationship is inversely related to the incidence of collateralization of loans (see e.g. Chakraborty and Hu, 2006; Brick and Palia, 2007).

⁹ The results of the regressions obtained from the second and the third sample splits are not reported in this article but just discussed. Regression results can be provided upon request.

¹⁰ The second condition is realized as follows: first, we order industry in our data set by number of downgrading in descending order. Industry positioned on the top has the highest ratio between the number of sample firms downgraded at 2010 and the total downgraded cases. Second, we consider firms that operate in the two industries positioned to the top of this list.

¹¹ We look at the ramo attività economica (RAE) codes associated to each firm, so a total of 189 different codes are identified.

¹² The first interaction term is D_LENGTH*NUMBER, where D_LENGTH is a dummy variable that assumes value 1 if the length is higher than 7. The second interaction term is D_NUMBER*LENGTH, where D_NUMBER is a dummy variable equal to 1 when the number of banks is greater than or equal to the median value of NUMBER, i.e. 4 (see Table 2).