One of the most important channels through which the Federal Reserve (the Fed, henceforth) can influence the economy is the so-called bank-lending channel. In the seminal papers by Bernanke & Blinder (1988, 1992), the original formulation of the bank-lending channel relies on the key assumption of binding reserve requirement constraints. Yet, more than a decade after the financial crisis the Fed’s monetary operational framework still differs from the pre-crisis system in several important aspects. One of the most salient differences is the still large amount of excess reserves in the financial sector, indicating thus, that the key assumption of the bank-lending channel does not hold anymore.
It is therefore interesting to analyze whether the effect of monetary policy on the lending behavior of individual banks has changed since the Great Financial Crisis. Or in other words, is the bank-lending channel still supported by post financial crisis U.S. data? This paper tests the bank-lending channel in both pre and post-financial-crisis times using micro-level data, which is collected from the Consolidated Report of Condition and Income (also known as “Call Reports”) between 1990Q1 and 2019Q3, and it encompasses over a million bank-quarters observations of all U.S. regulated financial institutions. I also test, from a bank-level perspective, the lending reaction to narrative and high-frequency identified monetary shocks. Finally, this paper provides new evidence on the influence that excess reserve holdings have on the relationship between monetary policy and banks’ lending decisions.
To test for the presence of the bank-lending channel, I use a standard two-step regression approach following the seminal contribution by Kashyap & Stein (2000), where the main assumption is that a change in monetary policy can be interpreted as a funding shock for banks that forces them to adjust their credit supply. The advantage of this method is that by using micro data one can exploit the cross-sectional variation in banks’ response to changes in monetary policy. On the other hand, studies that rely on aggregate data to test for an active lending channel suffer from an identification problem given the fact that monetary policy affects loan-demand and loan-supply simultaneously. However, the identification problem of the bank-lending channel can potentially be overcome by examining individual bank’s lending behavior instead of aggregate credit measures, as explained below.
My results in the sample prior to the crisis provide supporting evidence for an active bank- lending channel. To the extend of my knowledge, the most recent study using U.S. data that finds similar evidence for the bank-lending channel is the one by Cetorelli & Goldberg (2012). They implemented a similar two-step approach using data from 1980Q1 to 2005Q4, and find evidence for the lending channel not only for small banks, but also for large non-global banks. Several other studies have found evidence in favor of an active lending channel. For example, Campello (2002) analyzes the effects of small banks that belong to a financial conglomerate. Peek & Rosengren (2013) summarizes the pre-crisis evidence on the effectiveness of the bank lending channel, not only for the U.S., but also for other countries and regions including Japan, South Korea and Europe. Finally, Peek & Rosengren (2013) conjecture that after the large-scale asset purchases (LSAPs henceforth) by the Fed, it is unlikely for conventional monetary policy “to have an effect on bank lending through the usual mechanism associated with the lending channel”.
Moreover, I find that since the Great Financial Crisis, the bank-lending channel is no longer present in U.S. data. Instead, monetary tightening is now associated with growth in bank lending via a reduction in the sensitivity of a bank’s lending decision to its liquidity position. Recently, other studies have found similar counterintuitive results using bank-level data from Norway and Sweden. Cao & Dinger (2018) use 20 years of Norwegian data and find that the bank-lending channel is inactive. Similarly, Eggertsson, Juelsrud, Summers & Wold (2019) find that at the ZLB on the deposit rate the lending channel collapses. Salachas, Laopodis & Kouretas (2017) find similar results and conclude that “the financial crisis had a remarkable effect on bank lending behavior since the bank lending channel was distorted significantly.” However, while the present study concentrates on the entire U.S. regulated financial ioh nstitutions over 30 years
(1990Q1-2019Q3), Salachas et al. (2017)’s focus their analysis exclusively on large international banks, and their sample only covers the years between 2001 and 2013.
Overall, my findings suggest that unconventional policies implemented at the onset of the financial crisis unintentionally reshaped the transmission mechanism of conventional monetary policy. Additionally, I show that banks with abundant excess reserve holdings, i.e., banks with an excess reserves to total assets ratio larger than 1%, seem to be driving, at least partially, the counterintuitive relationship between changes in monetary policy and bank’s credit supply.
The results are robust to different balance-sheet liquidity indicators, lending measures, defini- tion of size categories, econometric specifications and to several monetary indicators, including shadow rates, narrative and high-frequency identified monetary policy surprises. Moreover, the results are in line with the theoretical predictions in Salgado-Moreno (2021), where the monetary transmission mechanism is analyzed in a regime-switching two-agent New Keynesian model with an interbank market. While the bank-lending channel works as predicted by Bernanke & Blinder
(1988) under a permanently-binding reserve-requirement regime, under the new operational system, contractionary monetary policy increases bank’s credit supply to the real economy stimulating output and aggregate demand.
The remainder of this paper is organized as follows. Section 2 presents the theoretical background of the lending channel, and section 3 describes the data set. Afterwards, in section 4, the econometric specifications are shown, and section 5 describes the main results. An extensive robustness analysis is given in section 6. Finally, section 7 concludes.