We test a new cross-sectional relation between expected stock return and idiosyncratic risk implied by the theory of costly arbitrage. If arbitrageurs find it more difficult to correct the mispricing of stocks with high idiosyncratic risk, there should be a positive (negative) relation between expected return and idiosyncratic risk for undervalued (overvalued) stocks. We combine several well-known anomalies to measure stock mispricing and proxy stock idiosyncratic risk using an exponential GARCH model for stock returns. We confirm that average stock returns monotonically increase (decrease) with idiosyncratic risk for undervalued (overvalued) stocks. Overall, our results support the importance of idiosyncratic risk as an arbitrage cost.

Building on the growing literature on interfirm links and limited attention, we find evidence of return predictability across alliance partners. A long–short portfolio sorted on lagged returns of strategic alliance partners provides a return of 89 basis points per month that is robust to a number of specifications. Investor inattention and limits to arbitrage may be the source of the underreaction of a firm’s returns to that of its partners.