During the last six years, there has been an important increase in top-published articles on empirical banking. Needless to say, the financial crisis has contributed to this increase; however, the most important change has come from the availability of very detailed data, mainly through credit registers. In a credit register (which is usually in the central bank of the country) you can find data at bank-borrower-time level; in other words, you can know, at each month/quarter, how much borrower X is borrowing from bank Y. This allows to better identify credit supply, which had been the main limitation of this literature in the ‘pre-credit-registers period’.
This literature is mainly interested in knowing how banks change their credit supply as a result of a shock. How could we empirically look at this question? To make it more precise, we can use the case that Iyer et al. (2014): what was the reaction (in terms of credit supply) of Portuguese banks to a shock in the international interbank markets in August 2007? An obvious answer would be to use a diff-in-diff approach: we see how bank credit changed after the shock depending on how exposed to the shock banks were. Without a credit register, only looking at the total credit given by the banks to the corporate sector, the main identification assumption of this approach would be that the pool of borrowers of the different banks react similarly to the shock. Put it differently, that the credit demand change after the shock is uncorrelated with how much banks are borrowing from abroad.
This assumption is not very realistic. Banks that borrow more from abroad may be lending more to firms that are also more exposed to international markets. If that were the case, the result could be driven by these firms demanding less credit as a result of the shock, even if more affected banks did not change their credit supply differently.
With the credit register information, we can do the same diff-in-diff analysis but in a finer way: we analyse the change in credit of a bank to a particular borrower after the shock. Given the structure of the data, we can use borrower fixed effects. This allows us to do the following: we can compare how credit given by two or more banks (differently affected by the shock) to the same borrower changes after the shock. With the borrower fixed effects we control for the average change in credit for each firm, which can be understood as controlling for credit demand. Now the identification assumption has changed: we are assuming that the credit demand of a particular firm borrowing from two or more banks changes similarly for each of these banks. In other words, if firm X is borrowing from bank Y and Z when the shock happens, we require that the change in credit demanded by firm X to bank Y is similar to the one demanded to bank Z. Since credit demand is usually driven by firm fundamentals, then this assumption seems more realistic.
Two notes to finish. First, by using this empirical approach we are restricting our attention to firms borrowing form at least two banks. Second, these papers usually go one step further and, after showing the reduction of credit supply by more affected banks, they look whether firms have been able to substitute credit by borrowing from other, less affected banks. At the end of the day, this is what really matters to understand if the reduction in credit supply can have real effects.
In the next post I will discuss the results of three papers that use the Spanish register to understand how banks react to monetary, economic, and regulatory shocks.
– Iyer, R., J.-L. Peydró, S. Lopes, and A. Schoar, 2014, “Interbank Liquidity Crunch and Firm Credit Crunch: Evidence form the 2007-2009 Crisis,” Review of Financial Studies 27(1): 347-72.
If interested, look at the very first paper using this empirical approach in banking:
– Khwaja, A. I., and A. Mian, 2008, “Tracing the Impact of Bank Liquidity Shocks: Evidence from an Emerging Market,” American Economic Review 98(4): 1413-42.