The information banks produce drives their lending decisions and macroeconomic outcomes, but this information is inherently difficult to analyse because it is private. We construct a novel measure of bank information quality from confidential regulatory data that include banks’ private risk assessments for US corporate loans. Information quality improves as local economic conditions deteriorate, particularly for new loans, large loans, and loans with higher expected losses. Information quality also declines during periods of rapid local house price appreciation. Our results provide empirical support for theories of countercyclical information production in credit markets.

Short sellers care about (i) how overvalued an asset is and (ii) when the overvaluation will be corrected. Hence, short selling costs should be higher over horizons when negative information is more likely to arrive. This article presents a model formalizing this intuition and tests the model using the put–call parity condition. Forward shorting costs predict future costs and stock returns, consistent with an expectations hypothesis in the shorting market. Additionally, an upward sloping term structure around earnings announcements increases the probability of a negative earnings surprise, evidence that short selling costs are higher over horizons when negative information is more likely to arrive. My findings suggest that the term structure of short selling costs conveys how long overpricings are expected to persist.