We study the matching patterns between heterogeneous principals and agents in a principal-agent model. The resulting equilibrium relationship between risk and incentives could be negative, positive or U-shaped. These results may provide an explanation for the absence of systematic empirical...