"We study the performance effects of payment schemes for freelancers offering services on an online platform in an RCT. Under the initial scheme, the firm pays workers a pure sales commission. The intervention reduces the commission rate and adds a fixed payment per processed order to insure workers against earnings risk. Our experiment tests predictions from a formal model on labor supply and performance for individuals with different degrees of risk aversion and intrinsic motivation for the task. The treatment did not affect labor supply and even though the commission rate was reduced by 50% we find no sizeable loss in sales per order. However, there is strong evidence for heterogeneous treatment effects. The treatment reduced performance for less intrinsically motivated workers. For more intrinsically motivated workers, however, we observe the opposite pattern as performance increased even though commission rates were reduced."

Support: Digital