Abstract: | We analyse the school participation decision of children between 13 and 18 years in Indonesia using the 1992 SUSENAS household survey. Our focus is on which household factors determine enrolment and delayed enrolment. We use the multinomial fixed effect model where the cluster-specific fixed effects correct for any regional-specific factors that may influence the demand for education. The model is estimated by conditional maximum likelihood. We find that parent's education has a positive effect on enrolment, where the effects are stronger for boys than for girls. On the other hand, literacy of parents has stronger effects on the girls education than on the boys'. The model without fixed effects is rejected against the model with fixed effects. Hence, omitting regional variation in the model would have led to biased estimates. |