Why Does Fast Loan Growth Predict Poor Performance for Banks?

In our article, Why Does Fast Loan Growth Predict Poor Performance for Banks?, which was recently accepted for publication in the Review of Financial Studies, we show that the common stock of U.S. banks with loan growth in the top quartile of banks significantly underperforms the common stock of banks with loan growth in the bottom quartile.

Many recent papers find that credit booms generally end poorly and are followed by poor economic performance. A number of theories have been advanced to explain this phenomenon. Most of the empirical analyses examining these theories have focused on country-level evidence. We instead investigate bank-level credit growth and subsequent returns within a single country and ask what the results imply for these theories. We analyze a panel of U.S. publicly listed banks between 1972 and 2014. Such a long time series has the benefit of enabling us to show that the phenomenon we document is persistent and is not the result of changes in bank regulations or in bank governance.

We find that banks that grow quickly make loans that perform worse than the loans of other banks and that investors and equity analysts do not anticipate the poorer performance. Our evidence is consistent with theories of credit booms that rely on expectation formation mechanisms. According to these theories, banks and investors fail to account fully and in an unbiased way for the risks of loans that banks make during the period of accelerated growth of their loan book.

Macroeconomic rational expectations approaches to explaining why credit booms are followed by poor economic performance rely on shocks to lending opportunities. A positive shock leads banks to lend more as they have better opportunities to lend. Adverse economic shocks then decrease the quality of the loans, are accompanied by poor economic performance, and lead banks to become more fragile and lend less. Alternatively, a credit boom could occur because of expectations that fail to take risks correctly into account, so that lenders and market participants become too optimistic about the risks of new lending opportunities. When the ignored risks are revealed or when the factors that led to over-optimistic expectations are no longer present, investors and bankers reassess the quality of the loans. At that time, reserves are increased, bank stock prices underperform, banks reduce their lending, and analysts are surprised by bank earnings.

Both rational expectations and biased expectations theories imply that loans grow quickly before eventually leading to unexpected bank losses. We show that this result holds for our sample of individual banks. But we also provide evidence that the pattern of loan growth and poor subsequent performance for individual banks is not tied to the performance of the economy as a whole or to regional economic performance. The latter results are hard to reconcile with a purely macroeconomic rational expectations theory of credit booms. The fact that poor bank performance following bank-level credit booms is predictable but that the market and analysts fail to anticipate it is also hard to reconcile with the rational expectations theory.

The expectation mechanism in biased expectation models starts from a kernel of truth, but extrapolates excessively from it. Hence, at the bank level, we would expect bank loan growth to start from a positive shock, but then the market and the bank would favor loan growth beyond what is supported by the positive shock. The fact that many banks are, at different times, banks with high lending growth and banks with low lending growth in our sample is consistent with that type of expectation mechanism and inconsistent with lasting differences across banks due to business models, ownership, or incentives.

We first examine whether high bank loan growth predicts poor future bank stock returns. If banks grow quickly because they make risky loans, they will experience higher loan losses following a period of high growth. If the banks and their investors properly understood that the high growth was the result of riskier loans, the stock price should correctly reflect the expectation of higher loan losses, so that high loan growth should not predict lower performance. By analyzing a panel of banks, we focus on variation in growth across banks and eliminate the effects of economic conditions because they are common across banks. Using a wide variety of econometric approaches, we show that high loan growth banks significantly underperform low loan growth banks in subsequent years.

We then show that banks that grow faster make poorer loans. We find that the fast-growing banks have a much higher ROA than the banks in the quartile with lowest growth in the year in which we measure growth (the formation year). However, by year three after formation, the order is reversed and the banks in the fastest growing quartile have a significantly lower ROA than the banks in the lowest growth quartile. A similar pattern holds for loan loss provision levels, which are lower for high growth banks in the formation year, but are higher by year three after formation.

Our evidence suggests that banks that grow quickly through loan growth do not appear to believe that they are making poorer loans than the banks that grow slowly. If they thought they were making riskier loans and provisioned properly, they would have greater loan loss reserves in the formation year than the banks that grow slowly, which is not the case. We also find evidence that analysts are surprised by the poorer performance of the high-growth banks after formation, in that their forecasts are too optimistic for high-growth banks relative to low-growth banks.

Finally, our results are not driven by merger activity. We show that high organic growth leads to lower stock returns even after controlling for growth through mergers, and the decrease in accounting performance is primarily driven by organic loan growth.

Overall, our evidence indicates that banks do not fully appreciate the risk of the loans they are making when they grow quickly. Such an outcome is in line with theories that rely on biased expectations or neglected risks.

The complete article is available for download here.