This is my final installment on estimating the ATT using matching versus saturated regression models. Imbens and Rubin (2015) note, with a proof, that exogeneity implies three things:
Continue reading this post for free, courtesy of scott cunningham.