Estimating the effect of education on earnings in Ghana: models for twins

Abstract

This thesis estimates the economic returns to an additional year of schooling in Ghana using data from a twins' survey. Consequently, the relative imp011ance of the roles of genetics and family background in determining earnings and returns to schooling in Ghana is examined. A number of models and estimation methods were utilized to illustrate the sensitivity of different estimators to model specification. The results indicate that the economic return to schooling m Ghana usmg Mincer's Human Capital model is about l 0%. Estimates of the economic returns to schooling using fixed effects and selection effects regression models and incorporating an instrumental variables approach to correct for measurement error in self-reported schooling levels was also assessed. The measurement error corrected return to schooling for monozygotic twins was larger than the standard ordinary least squares return to schooling estimate indicating a downward bias in the ordinary least squares return to schooling. Finally, the Restricted Maximum Likelihood approach was adopted to identify unobservable differences in the returns to schooling for twins'. The analysis revealed significant unobservable differences (p<0.05) in the REML returns to schooling for dizygotic twins' whiles, unobservable differences in the REML returns to schooling for monozygotic twins was not significantly different from zero. The estimated "pure" rate of return to education in Ghana could therefore be used as an indicator for considering policies related to education.