DP5514 Birth Order and Sibship Sex Composition as Instruments in the Study of Education and Earnings
|Author(s):||Robert J. Gary-Bobo, Natalie Picard, Ana Prieto|
|Publication Date:||February 2006|
|Keyword(s):||birth order, earnings, education, family, siblings, sibship sex composition|
|JEL(s):||I2, J12, J16, J24|
|Programme Areas:||Public Economics|
|Link to this Page:||cepr.org/active/publications/discussion_papers/dp.php?dpno=5514|
This paper presents an empirical study of birth-order and sibship sex-composition effects on educational achievement, and uses these variables as instruments to estimate returns to education, with the help of a rich set of individual data. Our sample includes more than 12,000 men and 10,000 women, who all left school in 1992, in France. The wages and educational achievements of individuals, as well as many aspects of family background, including birth order, number of sisters and brothers, are observed. An Ordered Probit model explains educational achievements. Sibship sex composition is shown to have an impact. Brothers and sisters have significant, non-negligible and different effects on educational achievement. A higher number of siblings has a negative effect in general, holding birth order constant, except when parents belong to the highest occupational groups; in other words, it is good to have many brothers and sisters if one's parents are well-to-do (the 'rich daddy effect'). On average, girls suffer significantly more from an additional brother than boys. Birth-order effects are both significant and substantial, even when many controls are included in the regressions. A high rank among siblings is detrimental for educational attainment (all other things equal), except in the case of fatherless children. Finally, a two stage method is used to estimate log-wage equations, taking care of education endogeneity, using birth order and the number of siblings as instruments. The OLS estimates of returns to education are biased downwards, when females are considered, but do not seem to be biased in the male sub-sample, given that many controls have been added in the wage equation.