?
Double machine learning estimates of marriage premium in Russia
This study compares classical and double machine learning (DML) estimates of the marriage wage premium in Russia. Using data from the RLMS for 2023–2024, the analysis employs a rich set of controls for non-cognitive abilities to mitigate confounding, thereby extending the control variables used in prior literature. Without correcting for endogeneity, both classical and DML estimates are positive for men and negative for women, and they are similar in magnitude. However, after addressing endogeneity, all estimators provide evidence that the marriage premium is absent — a result consistent with selection theory. These findings suggest that classical estimates are relatively robust. Additionally, we find statistical evidence that spousal conflicts may impose a wage penalty for both genders.