Little is known about the payoffs to apprenticeship training in the German speaking countries for the participants. OLS estimates suggest that the returns are similar to those of other types of schooling. However, there is a lot of heterogeneity in the types of apprenticeships offered, and institutional descriptions suggest that there might be an important element of selection in who obtains an apprenticeship, and what type. In order to overcome the resulting ability bias we estimate returns to apprenticeship training for apprentices in failed firms in Austria. When a firm fails, current apprentices cannot complete their training in this firm. Because apprentices will be at different stages in their apprenticeship at that time, the failure of a firm will manipulate the length of the apprenticeship period completed for some apprentices. The time to the firm failure therefore serves as an instrument for the length of the apprenticeship completed both at the original firm and at other firms. We find instrumental variables returns which are similar or larger than the OLS returns in our sample, indicating relatively little selection. (Contains 5 tables, 1 figure and 6 footnotes.) [Additional funding for this paper was provided by the Austrian Science Fund.]