Evidence from Vietnam
Pham Minh Nguyet
Nguyen Viet Cuong, National Economics University, Hanoi, Vietnam
Daniel Westbrook, Georgetown University School of Foreign Service in Qatar, Doha, State of Qatar[1]*
Measuring the impacts of key demographic and economic variables on educational outcomes at the individual level occupies a very large literature. One of the major challenges to estimating the impacts of interest is endogeneity due to unobservables that affect outcomes and their covariates. For example, estimating the impact of sibship size on childrens’ educational outcomes should take account of the fact that educational outcomes and the number of children in the household may both be affected by unobservable attitudes of the parents. An additional challenge arises when the educational outcome and the endogenous regressor of interest are discrete variables: this invalidates the conventional linear instrumental variables approach, though the conventional approach may provide a good approximation of the average treatment effect. We approach the estimation problem by jointly modeling performance and sibship size. First, we define dichotomous measures of educational performance and sibship size: for example, we record performance as excellent or not and sibship size as more than two or not. Then we estimate bivariate probit models of performance and sibship size, controlling for a variety of individual, household, and community characteristics. The control variables in the sibship size equation include instruments that provide exogenous variation in sibship size. Our data source is the Vietnam Household Living Standards Survey of 2006, which includes a special module on education. The module includes information on childrens’ performance, repetition of grades, enrollment in extra classes, and engagement of tutors.
JEL Classification: I2, J1, J13
Keywords: sibship size, birth order, education, household survey, Vietnam.