In this paper we show that periodic credit crunches, swings between high and low credit allocations, are an inherent part of banking due to the way banks compete for borrowers. The amount of information that banks produce about potential borrowers, and the amount of credit banks are willing to extend, varies through time due to strategic interaction between competing banks. Credit cycles can occur without any change in the macroeconomic environment. We investigate this amplification mechanism and provide empirical evidence that bank credit cycles are an important autonomous part of business cycle dynamics. Extensive empirical tests of the model are presented, based on parameterizing the public information that is the basis for banks’ beliefs about rivals’ strategies. These information measures concern rival banks’ relative performance, encapsulated in a Performance Difference Index (PDI). The empirical behavior of U.S. bank credit card lending, commercial and industrial lending, and bank profitability, are consistent with the model. Bank credit cycles are a systematic risk. We find that, consistent with this, the PDI is a priced factor in an asset pricing model of bank stock returns. Most importantly, the PDI is a priced factor for non-financial firms as well, and increasingly so as firm size declines.
Changes in bank credit allocation, sometimes called "credit crunches," appear to be an important part of macroeconomic dynamics. Bank lending is procyclical. Rather than change the price of loans, the interest rate, banks sometimes ration credit. A dramatic example in the U.S. is the period shortly after the Basel Accord was agreed in 1988, during which time the share of U.S. total bank assets composed of commercial and industrial loans fell from about 22.5 percent in 1989 to less than 16 percent in 1994. At the same time, the share of assets invested in government securities increased from just over 15 percent to almost 25 percent. More generally, it has been noted that banks vary their lending standards or credit standards. Bank “lending standards” or “credit standards” are the criteria by which banks determine and rank loan applicants’ risks of loss due to default, and according to which a bank then makes its lending decisions. While not observable, there is a variety of evidence showing that while lending rates are sticky, banks do, in fact, change their lending standards. The most direct evidence comes from the Federal Reserve System’s Senior Loan Officer Opinion Survey on Bank Lending Practices. Banks are asked whether their "credit standards" for approving loans (excluding merger and acquisition-related loans) have “tightened considerably, tightened somewhat, remained basically unchanged, eased somewhat, or eased considerably.” Lown and Morgan (2001) examine this survey evidence and note that, except for 1982, every recession was preceded by a sharp spike in the percentage of banks reporting a tightening of lending standards. Other evidence that bank lending standards change is econometric. Asea and Blomberg (1998) examined a large panel data set of bank loan terms over the period 1977 to 1993 and “demonstrate that banks change their lending standards - from tightness-to laxity-systematically over the cycle” (p. 89), and they conclude that cycles in bank lending standards are important in explaining aggregate economic activity.
In a macroeconomic context changes in the Fed Lending Standards Index (the percentage of respondents reporting tightening) Granger-causes changes in output, loans, and the federal funds rate, but the macroeconomic variables are not successful in explaining variation in the lending standards index. The Lending Standards Index is exogenous with respect to the other variables in the Vector Autoregression system. See Lown and Morgan (2001, 2002) and Lown, Morgan and Rohatgi (2000). The analysis in this paper is aimed at explaining the forces that cause lending standards to change and, in particular, to explaining how this can happen independently of macroeconomic variables.
When competing with each other to lend, banks produce information about potential borrowers in an environment where they do not know how much information is being produced by rival bank lenders. We study a model of bank competition in which banks collude to set high loan rates (hence loan rates are sticky), and they implicitly agree not to (over-) invest in costly information production about prospective borrowers. A bank can strategically produce more information than its rivals and then select the better borrowers, leaving unknowing rivals with adversely selected loan portfolios. Unlike standard models of imperfect competition, following Green and Porter (1984), there are no price wars among banks since banks do not change their loan rates. However, as in Green and Porter (1984), intertemporal incentives to maintain the collusive arrangement requires periods of "punishment." Here these correspond to credit crunches. In a credit crunch all banks increase their costly information production intensity, that is, they raise their "lending standards," and stop making loans to some borrowers who previously received loans. These swings in credit availability are caused by banks’ changing beliefs, based on public information about rivals, about the viability of the collusive arrangement.
Empirically testing models of repeated strategic interaction of firms has focused on price wars. See Reiss and Wolak (2003) and Bresnahan (1989) for surveys of the literature. However, our model predicts there are "information production wars." Since information production is unobservable, we can not follow the usual empirical strategy. We propose a new method for testing the model, which we believe to be of independent interest. Our approach tests a general implication of any equilibrium of the model with imperfect competition by identifying the relevant public information and its relation with "information production wars"—credit crunches. In theory, to detect deviations by rivals, banks must look at two sources of public information: the number of loans made in a period by each rival and the default performance of each rivals’ loan portfolio. We argue that the relative performance of other banks is the public information relevant for each bank’s decisions about the choice of the level of information production. More importantly, the use of relative bank performance empirically distinguishes our theory from a general learning story, which would predict past bank performance matters for bank credit decisions.
Broadly, the empirical analysis is in three parts. First, we examine a narrow category of loans, U.S. credit card lending, where there are a small number of banks that appear to dominate the market. Since it is not clear which banks are rivals, we first analyze this lending market by examining banks pairwise. If the PDI increases, banks should reduce their lending and increase their information production resulting in fewer loan losses in the next quarter. We also examine bank profitability, using stock returns. Second, we analyze macroeconomic time series, including the Lending Standard Survey Index. We form an aggregate bank Performance Difference Index (PDI) based on the absolute value of the differences on all commercial and industrial loans of the largest 200 banks. If beliefs are, in fact, based on this information, then we should be able to explain (in the sense of Granger causality) the time series of Fed’s Lending Standard Survey responses (the percentage of banks reporting "tightening" their standards) in Lown and Morgan (2001). Thirdly, if credit crunches are endogenous, and a systematic risk, then they should be a priced factor in an asset pricing model of stock returns. Therefore, our final test is to ask whether the parameterization of banks’ relevant histories is a priced risk factor in a four factor Fama-French asset pricing setting. We look at banks and nonfinancial firms by size, as credit crunches have larger effects on smaller firms. We find all the evidence to be consistent with the theory.
Other relevant work includes Rajan (1994). He argues that fluctuations in credit availability by banks are driven by bank managers’ concerns for their reputations (due to bank managers having short horizons), and that consequently bank managers are influenced by the credit policies of other banks. Managers’ reputations suffer if they fail to expand credit while other banks are doing so, implying that expansions lead to significant increases in losses on loans subsequently. We test Rajan’s idea in the empirical section. Two related theoretical models are provided by Dell’Ariccia and Marquez (2004) and Ruckes (2003). These papers show a link between lending standards and information asymmetry among banks, driven by exogenous changes in the macroeconomy. As distinct from these models, the fluctuation of banks’ lending behavior in our paper is purely driven by the strategic interactions between banks instead of an exogenously changing economic environment. We proceed in Section 2 to describe the stage game for bank lending competition, and we study the existence of stage Nash equilibrium and the model’s implications for lending standards. The stage game is a prelude to considering the infinitely repeated game, the subject of Section 3. In Section 4, we carry out empirical tests. Section 5 concludes the paper.
Download
PDF Ebook Bank Credit Cycles
