Abstract: | This article adds the empirically verified, but analytically neglected, fact of delayed worker training to a model of general training. By delaying training, firms can gain partial information on worker‐firm matches, generating wage compression. Analytical results differ from those of other training models. In the noncooperative regime, all firm‐paid training is delayed, all worker‐paid training is immediate. With no externalities to other firms, the cooperative result is Pareto‐optimal. In a continuous‐time extension, firms choose a probationary period after which workers are trained. With longer probations, fewer bad‐match workers are trained and wage compression rises. Government training mandates are also studied. |