Human capital theory predicts that training should reduce starting wages, yet this relationship remains empirically undocumented. Estimation of how training affects wages must control heterogeneity bias. I do this by estimating first-difference starting wage regressions on a sample of workers...