Returns to compulsory schooling in Britain : evidence from a Bayesian fuzzy regression discontinuity analysis

In this paper we reevaluate the returns to education based on the increase in the compulsory schooling age from 14 to 15 in the UK in 1947. We provide a Bayesian fuzzy regression discontinuity approach to infer the effect on earnings for a subset of subjects who turned 14 in a narrow window around the policy change and whose schooling was affected by the policy change. Our approach and our results are quite different from previous work that has focused on large sets of cohorts and 2SLS based approaches and has reported positive earnings and wage effects of 5% and above. Our empirical analysis, using data from the UK General Household Surveys, yields considerably lower earnings and wage effects for the additional year of compulsory schooling than previous work. These findings are consistent with the implementation of the policy change that affected students at the lower end of the schooling distribution and did not lead students to acquire additional qualifications. The results add further evidence to a number of recent studies that have found no effect from this policy change on socio-economic outcomes correlated with earnings. -- Bayesian inference ; causal effects ; imperfect compliance ; natural experiment ; principal stratification ; regression discontinuity ; returns to schooling