3 Summary Using data from the 2010 Continuing Vocational Training Survey for Austria, I estimate the probability of formal employer-provided training and analyze the returns to such training. The returns are measured by labor productivity obtained from the Austrian Structural Business Statistics. To address potential biases from self-selection into training, I employ an instrumen- tal variable approach where I use the training-related stipulations in the collective bargaining agreements as an instrument. I estimate, in line with earlier research, that firms differ in their likelihood of provid- ing formal training for their employees. For example, firms with a large share of female blue- collar workers are less likely to formally train their workers than firms with a large share of male blue-collar workers. I estimate an elasticity of labor productivity with respect to formal training of about 16% with standard linear probability models; using an instrumental variable approach results in larger estimates of between 17 to 37%, however, because the instrument is not very strong, these later estimates are estimated with large standard errors which include the estimates from the more na?ve linear probability models.