Explaining returns to education in Egypt : effect of education on labor force participation and age of marriage -

Abstract

Extensive literature has studied the effect of education on wages and other labor market outcomes. However, it is hard to isolate the effect of education due to endogeneity, selection bias and omitted variable bias. Methods that employ institutional changes as sources of exogenous variation in years of schooling have been used in order to identify causal effects. For example, a recent paper by Assaad et al. (2016) uses Law No.233 in Egypt, that reduced the number of years of primary schooling from 6 to 5 years in year 1988-89, thus decreasing compulsory years of education from 9 to 8 years. Their findings show that the intervention significantly affected the years of education received for males but not for females. Through 2SLS regressions (within an IV framework), they then estimate returns to education to be between 2 and 5.7percent for men, which is lower than returns to education found in other countries. Our paper aims to look more closely at the effect of this intervention on other outcomes, such as labor force participation and age of marriage, that in turn impact wages. Results of Linear Probability Models show that receiving the treatment is correlated with a significant increase in the probability of leaving the labor force for females. However, the treatment has no significant effect on males' labor force participation when we control for age polynomials. In addition, the treatment is found to have a significant effect on the age of marriage of females, and an insignificant effect on the age of marriage of males when we control for age polynomials. These findings show that the treatment could have affected wages of females through links other than its main effect on years of schooling, which may explain the insignificance of the first stage results that Assaad et al(2016) found. Furthermore, given that when we control for our treatment variable along with cohort dummy variables instead of age polynomials, the coefficient of the treatment variable is significant for males, then the validity of this trea