Using a national level sample survey on labour market in India, we analyze the role of education-occupation (mis-)match (EOM) in explaining within-group dispersion in returns to education. Applying a double sample selection bias correction and Mincerian quantile wage regression estimation, we document that accounting for EOM reduces the dispersion in returns to education. Further, overeducated workers face a wage penalty of 7 percent and undereducated workers get a wage reward of 3 percent as compared to adequately matched workers. This study argues to focus on EOM to increase both pecuniary and social benefits of education in terms of productivity gains and wages as well as to reduce wage inequality.