We offer new evidence on the risk versus mispricing explanations for the R&D anomaly. Return covariance with a characteristic-based factor captures the cross-sectional return variation on R&D portfolios not explained by asset pricing models. This is consistent with both covariance risk and mispricing. Under the framework of the ICAPM, we find little economic justification that an R&D factor is a proxy for innovations to a state variable. The characteristic subsumes the factor loading in direct tests, providing support to the mispricing hypothesis. Investigating the mispricing explanation further, we reject the assertion that the R&D anomaly arises from the correction of stocks mispriced by investor sentiment. A natural experiment exploiting the pilot program under Regulation SHO shows no evidence that the anomaly persists due to limits to arbitrage in the form of short sale constraints.