Anxious periods and bank lending

In a recent paper, Geanakoplos and Fostel (2008) suggest that financial markets operate under three conditions: the normal economy, when the liquidity wedge is small and leverage is high; the anxious economy, when the liquidity wedge is big, leverage is curtailed and the general public is anxiously selling risky assets to more confident natural buyers; and, finally, the crisis or panicked economy, when many formerly leveraged natural buyers are forced to liquidate or sell off their positions to a reluctant public, often going bankrupt in the process. How do banks perceive these three states of the economy and form their lending strategy? In this paper, we try to shed some light in the strategic behavior of banks concerning their lending behavior, by placing particular emphasis in the second state of affairs, i.e. at times where the economy is anxious. We use quarterly data on US banks over the period 1985-2010 and examine the response of bank lending to changes in (i) consumer confidence, (ii) the federal funds rate and (iii) CEO confidence. In line with expectations, initial results show that during the full sample banks expand their lending when consumer and CEO confidence rise and curtail their lending when monetary policy is contractionary. The findings become more unexpected when distinguishing between periods of anxiety and periods of distress. During periods of anxiety banks continue to expand their lending irrespective of whether anxiety is defined by consumer confidence, changes in monetary policy or CEO confidence. Banks seem to curtail their lending (and thus their risky assets) only in bad times and at the edge between anxious and bad times. These results provide new implications for the role of banks in exacerbating financial crises. The equation to be estimated is of the following dynamic form (examines the response of bank lending to a change in the state of the economy): Where: dl reflects annual loan growth for bank i (i.e. over t-4). ds is the change in the state of the economy, captured by the change in consumer confidence or the change in interest rates. b is a set of bank characteristics that affect lending, notably capitalization, size, liquidity, provisions, non-performing loans and bank efficiency. dm is a set of variables characterizing other prevailing macroeconomic and regulatory conditions. The panel is unbalanced and quite large both in terms of the cross sectional and the time dimension. For panel datasets of this size all estimators for dynamic panels (OLS, ML, GMM) tend to converge. Alvarez and Arellano (Econometrica 2003) show that show that for T < N (as in our case), GMM bias is always smaller than OLS bias and LIML bias is smaller than the other two. Thus, we favor the LIML estimator (sensitivity analysis confirms the theoretical priors of Alvarez and Arellano). 1. During anxious times, a positive change in consumer confidence, the federal funds rate and CEO confidence increases lending. 2. This shows that banks actually increase their lending when the economy is at the anxious state. 3. Certainly, this implies an increase in the risk-taking of banks and may lead to a worse equilibrium risk of failure if the state of anxiety actually develops to a recession. 4. In contrast, banks curtail their lending during bad times, as evident from the negative coefficients on the consumer confidence, federal funds rate and CEO confidence variables.