Roll's critique of tests of the CAPM shows that if the index portfolio is ex post efficient, it is mathematically impossible for abnormal returns, as measured by the empirical market line, to be statistically different from zero. Yet the Ibbotson study on new issues uses the cross-section empirical market line and finds significant abnormal returns in the month of issue and none in the following months. Given Roll's critique, this should have been impossible. How can the empirical results be reconciled with the theory?