Endogénéité d’une variable explicative dichotomique dans le cadre d’un modèle probit bivarié : Une application au lien entre fécondité et activité féminine

The endogeneity of regressors, due to a correlation with the error term, can involve significant biases in the estimation of behaviors. When both the explained variable and the regressors are continuous, several methods to detect the endogeneity and correct its effects are avaialbe. However, when the two variables are dichotomic, the resolution of the problems due to the endogeneity is more difficult and seldom carried out. A test of endogeneity, which can be computed very simply with the usual software, is presented in the paper. Moreover, a method of estimation based on numerical techniques of integration is developed. The two procedures are then applied to detect the endogeneity of fertility in a model of female activity when the husband is not self-employed. The two techniques conclude in a very similar way to an absence of endogeneity.