This paper explores the theory and evidence behind why firms invest in training their workers. The authors propose a model where the current employer has superior information about the worker's ability compared to other firms, which gives the employer ex post monopsony power over the worker. This power encourages the firm to provide training. The model can lead to multiple equilibria, with either high quit rates and low training levels or low quit rates and high training levels. The authors use microdata from Germany to test their model, finding that it better explains the data than other explanations of firm-sponsored training. They also discuss the implications of their model for labor market dynamics and the efficiency of training levels. The paper concludes by comparing their findings to the literature on training and wage determination in imperfect labor markets.This paper explores the theory and evidence behind why firms invest in training their workers. The authors propose a model where the current employer has superior information about the worker's ability compared to other firms, which gives the employer ex post monopsony power over the worker. This power encourages the firm to provide training. The model can lead to multiple equilibria, with either high quit rates and low training levels or low quit rates and high training levels. The authors use microdata from Germany to test their model, finding that it better explains the data than other explanations of firm-sponsored training. They also discuss the implications of their model for labor market dynamics and the efficiency of training levels. The paper concludes by comparing their findings to the literature on training and wage determination in imperfect labor markets.