Identifying the Bank Lending Channel in Brazil through Data Frequency
In lieu of an abstract, here is a brief excerpt of the content:

Identifying the Bank Lending Channel in Brazil through Data Frequency

Monetary policy affects economic activity through different channels. One mechanism is the credit channel, that is, how monetary policy influences the real sector through its effect on the functioning of credit markets.1 There are two types of credit channels: the broad credit channel and the bank lending channel. The former is the channel through which monetary policy affects the balance sheet of lenders and borrowers in the economy. With regard to the latter, banks fund a significant part of their operations through deposits, as these are normally the cheapest source of funding. Because deposits and other sources of funding are less-than-perfect substitutes, monetary policy will shift the supply schedule of bank credit, insofar as it affects the amount of deposits in the banking system. This transmission mechanism is known as the bank lending channel.

Bernanke and Blinder first tried to identify the bank lending channel by looking at the relationship between monetary policy shocks and future amounts of loans.2 Interpretation of their empirical results is blurred by the fact that aggregate lending changes several months ahead of a monetary policy shock, because of both supply (bank lending channel) and demand factors (changes in investment and consumption decisions). In other words, one cannot disentangle demand and supply reactions to monetary policy with low frequency data (quarterly in the case of Bernanke and Blinder). Kashyap, Stein, and Wilcox also use quarterly data, but they explore the impact of monetary [End Page 47] policy on commercial paper, a substitute for bank loans.3 Contractions in monetary policy are associated with increases in future quantities of commercial paper, supporting the idea of a supply shock. However, identification remains unsatisfactory. Focusing the empirical analysis on quantities does not exclude the possibility that the demand for bank credit and the demand for commercial paper react differently to shocks in monetary policy.

Dissatisfaction with identification based on aggregate data led researchers to use bank-level data. In a seminal work, Kashyap and Stein use bank characteristics to identify the bank lending channel.4 They assume that smaller banks have more difficulty raising funds in money markets than larger banks. In this case, differences in the reactions of small and large banks to changes in monetary policy may be interpreted as evidence of the bank lending channel. Arena, Reinhart, and Vázquez also use this strategy.5

Kashyap and Stein and Arena, Reinhart, and Vázquez rely on theoretical arguments that bank characteristics are informative about the bank's ability to substitute away from deposits.6 Thus, they always test a joint hypothesis of the bank lending channel plus a better ability on the part of larger banks to substitute deposits. Furthermore, even if this theory is correct, banks with different characteristics serve different clients.7 Large banks tend to serve large corporations, while smaller banks tend to supply credit to small and medium-sized enterprises (SMEs). Large corporations have better access to capital markets than SMEs. Consequently, large corporations have a more elastic credit demand than SMEs, and large banks would lose market share to bond markets if they tighten credit in response to a shock in monetary policy.8 In this case, differences in bank market structure for SMEs and corporations explain the results in Kashyap and Stein without the bank lending channel being operative.9

We contribute to the empirical understanding of the bank lending channel by employing a sharper identification strategy. We use very high frequency bank-level data on loans to isolate supply shocks driven by monetary policy. [End Page 48] Our method bypasses both concerns with Kashyap and Stein's identification strategy. We have daily bank-level data on interest rate and quantity. The high frequency of the data allows us to isolate supply from demand shocks. The key identifying assumption is that supply reacts faster than demand to monetary shocks. Demand for credit depends on investment and consumption decisions that do not react immediately to changes in monetary policy (our estimation window is very short, at just a few...