We improve the precision of the implicit contract model test proposed by Beaudry and DiNardo (1991). Our dataset allows us to define the exact industry and plant of a particular employment relationship, link local labor market characteristics to individual-level wages, and control for composition effects. We find evidence in favor of the spot-market model of wage setting in the whole sample, but there is significant variation across industries and education levels. In particular, the spot market matters most for low-skill workers, while the implicit contract model with one-sided limited commitment applies better to high-skill workers.