The dissertation is a collection of 3 articles. The first estimates two types of models to investigate the role played by family background and individual characteristics in keeping children at school up to the end of secondary education. The first is a censored ordered probit model. It accounts for right-censoring of the schooling level of children enrolled at school at the time of survey. The second is a sequential model. It accounts for the “educational selectivity” and for the fact that, some family background characteristics or individual characteristics may affect the likelihood to reach one level of education but not others. Our striking findings arise from the results of the sequential model: (1) household’s head education reveals some threshold effect, university education is needed to enhance the probability to complete secondary education; (2) parental wealth has no effect on the probability to start primary education. However, it increases the probability to complete secondary education. Chapter 2, co-written with Bertrand Verheyden, examines the determinants of schooling in developing countries with a special emphasis on birth order effects. We present a theoretical model accounting for the dynamics of birth order and its interaction with credit constraints. We show that since elder children are the only source of additional income when constraints get tighter in poor families, they work more than their younger siblings and end up with lower levels of education. This discrimination is not at work among wealthier families. Controlling for household fixed effects, gender and age, our results confirm that earlier-born children’s education levels are relatively lower in poor households, and not in richer ones. These results are robust to various measures of birth order and household wealth. Chapter 3, examines whether gender gap in current enrolment rates prevails both among “regular” and “irregular” schooling children. Children who started primary school at age 6 and advanced one grade each year succeeded the first national standardized exam (CEP) at the age of 12. Regular children are those who, by the age of 12 have succeeded the CEP exam. Irregular children are the rest. We investigate the gender gap in schooling empirically. Our empirical framework allows the gender effect on the likelihood to be enrolled to be different for regular and irregular children. It also accounts for selection into the two groups. Results show no male-female difference among regular children. Among irregular children however, females are more likely to stay out of schools. We therefore suggest that, independently of the source of the gender gap, it seems to be at work mostly among irregular children.