Piece-rates and tournaments: Implications for learning in a cognitively challenging task

We compare the impact of piece-rate and tournament payment schemes on learning in a cognitively challenging task. In each one of multiple rounds, subjects are shown two cue values, Cue A and Cue B, and asked to predict the value of a third variable X, which is a noisy function of the two cue values. The subjects’ aim is to predict the value of X as accurately as possible. Our metric of performance is the absolute error, i.e., the absolute distance between the actual and predicted values of X. We implement four treatments which are based on two different payment schemes: (1) piece rates, where subjects are paid linearly on the basis of their own absolute errors and (2) a two-person winner-take-all-tournament, where subjects are paired and the one with a smaller absolute error earns a fixed payoff, while the other earns nothing. We find that it is only in the tournament payment scheme, and particularly in a more complex version of the task, that subjects show significant evidence of learning over time, in that their predictions get closer to the actual value of X. This learning process is driven by the all-or-nothing nature of the payoff structure in tournaments.