We study the impacts of family income and parental education on the probability of children's schooling delay using the 1996 PNAD. With the adoption of some hypotheses about the links between generations of children, parents and grandparents, we control for the existence of non observable factors that simultaneously affect the income formation of parents and decisions concerning children schooling (simultaneity bias) or that are transmitted from one generation to another (hereditary bias). We work with three instruments: number of schools by the time parents were children; family factor changes across generations (parents/grandparents); and the 1971 change of the educational system.
