This paper develops a structural empirical model of contraception and participation choice under imperfect control of fertility, learning and unobserved heterogeneity to identify, estimate and give a behavioral content to the effect of the first born child on female labor supply. Family planning failures are exploited as sources of identification. The data are drawn from the 1995 US National Survey of Family Growth (NSFG) which no one has used before for this purpose and which contains full retrospective information on participation, contraception and children. The model is estimated combining the Nested Pseudo Likelihood Estimation and the Expected-Maximization algorithm. Key factors driving the importance of the effect are education, labor market experience, child's age and preferences for leisure and children. From a policy perspective, this heterogeneity is important in designing maternity leave and child care policies. Additionally, twins and gender composition of children are used as sources of variation in the model to estimate the effect of the second and the third born child. The Average Treatment Effect (ATE) of the first is -12.4%, of the second is -5.6% and of third born child is -4.9%. Finally, based on the dynamic model a weighting procedure is proposed to understand the Local Average Treatment Effects (LATE) found in the nonstructural literature.