This paper presents a life-cycle model with human capital investment during working life through training and provides a novel empirical test of human capital theory. We exploit a sizable pension reform across adjacent cohorts in a regression discontinuity setting and find that an increase in working life increases training. We discuss and test further predictions regarding the relation between initial schooling, training, and the reform effect, showing that only individuals with a college degree increase human capital investment. Our results speak to a large class of human capital models as well as policies extending or shortening working life.