Rational expectations of own performance: An experimental study

Abstract

According to rational choice theory, people will choose their careers, level of work effort or investment based in part on their expectations of success. But are people’s expectations of their likelihood of success accurate? Evidence accumulated by psychologists suggest that on average people underestimate their risks and overestimate their abilities relative to others. We test for such optimistic bias in experiments where people must predict their relative or absolute success in incentivebased verbal and maximization contests. We ask subjects to provide initial and revised point estimates of their success rates, either hypothetically or with a scoring rule that rewards forecast accuracy. We then passively measure the quantity and quality of subjects’ efforts and subsequent outcomes. Bias in forecasts is evaluated at the aggregate level as done by psychologists, but also at the individual level using realized outcomes. We find limited evidence of excess optimism only in relative maximization contests when encountered first, and excess pessimism or accuracy elsewhere. Experience across contests and updating does not always increase the accuracy of self-assessed forecasts, and even when accuracy improves biases are rarely eliminated entirely. Methodologically, we find no evidence that a modest quadratic scoring rule introduces moral hazard for own-outcome forecasts, but neither does it increase forecast accuracy or lower variance.