In this paper, we examine the predictability of the cross-section of bank stock returns by taking advantage of the unique set of industry characteristics that prevail in the financial services sector. We examine predictability in the cross-section of bank stock returns using information contained in individual bank fundamental variables such as income from derivative usage, previous loan commitments, loan-loss reserves, earnings, and leverage. We find that variables related to non-interest income, loan-loss reserves, earnings, leverage, and standby letters of credit are all univariately important in forecasting the cross-section of bank stock returns. Surprisingly, neither book-to-market nor firm size is important in our sample. We examine whether this cross-sectional predictability is due to increased risk, or another explanation, such as investor under or overreaction. Our results suggest that this predictability is not due to increased risk, but rather is consistent with investor underreaction to changes in banks_ fundamental variables. Furthermore, out-of-sample testing demonstrates this underreaction appears to be exploitable using simple cross-sectional trading strategies.