This paper presents the first panel data evidence on the productivity effects of training in Germany. It uses a large and representative establishment panel data set for all profit-oriented sectors of the economy. Increasing the share of employees participating in training in the first half of a year has a positive and significant effect on firm productivity in the same and the subsequent year. The impact in the third year is positive but insignificant. While formal internal and external training courses increase productivity in the same year and the years after, their impact decreases over time. The positive productivity impact of quality circles increases over time, while training on the job has a persistent negative productivity effect.

When we control for selectivity and unobserved heterogeneity, the measured productivity effects further increase, suggesting that firms with an inefficient production structure deliberately use training in order to boost productivity. In a fixed effects panel regression, simultaneously taking endogeneity of training into account, formal internal and external courses, self-induced learning and quality circles have a positive impact on the fixed productivity effects, while training on the job, seminars and talks, and job rotation do not have an impact on structural productivity differences. When training intensity increases by 1%, structural productivity increases by 0.3%. Finally it is shown that omitted variable bias plays a crucial role in the estimation of productivity effects of training.

Keywords: Training, Firm Productivity, Panel Estimation