This paper is devoted to the analysis of influence of number and age of children on salaries of working women. Unlike most of the research concerning wages, the approach applied allows considering both an unobserved heterogeneity of Russian women and the sample selection bias. The mentioned approach is based on two-stage Heckman procedure within the framework of fixed effects model. Russian Longitudinal Monitoring survey (RLMS–HSE) data of 18–22 rounds is used. On the first step tobit-model for each round is estimated and residuals are derived. On the second step the model for salary is estimated, taking into account residuals from the first step. Various reference variables are used in the models: education level, qualification level and social-demographic characteristics of women. As a result, the revealed influence of children on salaries of working women is considerably higher in models with consideration of sample selection bias.