This paper develops a two-step approach for identifying changes in socioeconomic effects across educational transitions. The approach corrects for the two sources of heterogeneity known to bias observed trends in family background effects across transitions: selective attrition and heteroskedastic latent errors. The method is compatible with viewing school transitions as socially stratified educational decisions about leaving or staying in the educational system. Using the method in a reanalysis of the 1973 OCG data on white men born between 1907 and 1951 in the US fails to replicate the waning coefficients pattern for father's social status and mother's years of schooling. The effects of these variables appear to increase across transitions.