We apply Heckman's (1979) correction method to deal with the sample selection problem in the literature on size effect. We show that the anomaly is a statistical artifice of failing to control for the ex post survivorship bias, and small firms do not generate higher returns once the bias is corrected for. Moreover, large firms have higher returns than small firms once the ex ante default risk premium is taken into account. The evidence is robust when the beta measurement, economic cycle, extreme returns, transaction costs, and seasonality are controlled for and Jensen alpha is employed.