BW also present an interesting analysis of the long-run returns of an arbitrage long/short portfolio that invests long in dividend payers and short in nonpayers. They find that when the dividend premium is high, i.e., payers are more highly valued than nonpayers, the future raw returns of this portfolio are low. In these tests, BW do not control for risk. We reconsider the one-to-three-year-ahead returns adjusted for risk. The raw 1% per month spread drops to a statistically and economically insignificant 0.1% per month, or about 3.5% over three years once we adjust for risk. In any case, our view is that the portfolio results are at best indirect evidence of gains that could potentially exist if firms catered. These tests are thus secondary to more direct tests that examine the long-term returns of firms that do cater. In these direct tests, catering suggests that long term returns after increases or initiations should be negative, an implication that is overwhelmingly rejected by the data.