摘要：This paper provides a novel test of a key prediction of human capital theory that educational investment decisions depend on the length of the pay-off period. We obtain causal estimates by leveraging a unique reform of the German public pension system that, across a sharp date-of-birth cutoff, increased the early retirement age by three years. Using RDD, DiD, and IV estimation strategies on census and household-panel data, we show that this reform causally increased educational investment in the form of on-the-job training. In contrast, non-job related training before retirement was not affected. We explore heterogeneity and additional outcomes.
