Labour market discrimination against women and parental discrimination against daughters are two of the most commonly cited explanations of the gender gap in education in developing countries. This study empirically tests the labour market explanation for India using household survey data collected in urban Uttar Pradesh in 1995. It estimates workforce participation functions and selectivity-corrected earnings fluctuations, and calculates the rates of return to education for the two sexes. Using the Blinder-Oaxaca method, the gross gender difference in earnings is decomposed into the part that is explained by men and women's differential characteristics and the part that is due to labour market discrimination. The results reveal that there is substantial omitted family background bias in the estimates of returns and that, contrary to received wisdom, the rates of returns to education rise by education level. The analysis suggests that, as well as overall labour girls face significantly lower economic rates of returns to education than boys.