Analyzing Zero Returns to Education in Germany: Heterogeneous Effects and Skill Formation

We analyze the causal effect of education on old-age cognitive abilities using German Socio-Economic Panel (SOEP) data and regional variation in mandatory years of schooling and the supply of schools. Our outcome variable is the score an individual reaches in an ultra-short intelligence test. We explain this score, using instrumented education. Instrumental variable estimation is necessary since on the one hand, schooling is highly affected by cognitive skills (reverse causality), and, on the other hand, both are influenced by third factors like income and health (possibly omitted variable bias). Using variations in education policy in the German federal states we are able to instrument education for three groups of students which cover all levels of educational achievement. Unlike previous studies this allows us to estimate three different local average treatment effects. The estimated effects range from 14% to 53% of a standard deviation. We find the strongest effect for students with intermediate education.