This paper uses quantile regression techniques to analyze heterogeneous patterns of return to education across the conditional wage distribution in four transition countries. We correct for sample selection bias using a procedure suggested by Buchinsky (2001), which is based on a Newey (1991, 2009) power series expansion. We also examine the empirical implications of allowing for the endogeneity of schooling, using the control function approach proposed by Lee (2007). Using household data from Bulgaria, Russia, Kazakhstan and Serbia in 2003, we show that the return to education is heterogeneous across the earnings distribution. It is also found that accounting for the endogeneity of schooling leads to a higher rate of return to education.