Abstract

The ongoing transformation of the labour market driven by demographic and technological changes, has a significant impact on both labour supply and skill demand, which may eventually result in a skills mismatch. One way to adjust to this is to continuously invest in human capital throughout one’s working life, a concept that is promoted by policymakers across all over Europe, including in Germany. However, before designing effective programs, it is important to understand how on-the-job training investment decisions are formed. To address the lack of exogenous variation in human capital measures in observational data, we have designed a conjoint survey experiment that introduces random variation in the characteristics of potential training candidates and training content. Decision-makers in German establishments then evaluate these randomly generated candidate profiles. The conjoint survey experiment is anchored in the IAB Linked-Personnel-Panel. We randomly vary the gender, age, working time arrangements and caring responsibilities of the prospective training candidates. We further vary the primary aim of the training and the primary initiator of the training. We hypothesize that the latter four attributes (working time arrangements, caring responsibilities, skill content and primary initiator) are particularly important in influencing firms’ training choices and also contribute to explaining gender differences in training participation.