Less than three percent of the highest-paid U.S. executives are women. Why? In Performance in Competitive Environments: Gender Differences, a new paper in the Aug. 2003 QJE, the authors suggest an intriguing answer.
The authors compare male and female performance at solving mazes across different incentive systems. In a simple piece-rate system men perform slightly but not markedly better than women, on average the men solved 11.23 mazes in 15 minutes compared to 9.73 for the women, a difference of 1.5. But in a tournament, in which only the highest-paid performer wins, the men significantly improve their performance and the women hardly improve at all. As a result, the gender-gap in performance rises (men complete 15 mazes, the women only 10.8 for a difference of 4.2, stat. significant at p=0.034).
Now here is where it gets really interesting. One might think that this shows that women are less competitive than men. To test this the authors run single-sex tournaments. Surprisingly, in the single-sex tournaments the women’s performance improves considerably relative to both their performance in the piece rate system and to their performance in the mixed tournament. Women do like to compete just not against men! Men’s performance stays about the same as in the mixed tournament. As a result, when comparing the peformance of the all-male groups versus the all-female group, the gender gap shrinks considerably. Results are summarized in the figure below.
What could account for these differences? Tournament theory suggests one answer. In a tournament only the best player wins; so if some of the players are known to be better than the others, this reduces the incentives to compete. Why expend effort if the other player will amost surely win anyway? The men are slightly better at the task than the women and this effect is magnified by the numbers – there are 6 players (3 men, 3 women) so the women have to contend with 3 people who on average have slightly higher maze-solving ability.
If this explanation were the case, however, then we would expect men and women of the same ability to perform similarly but in fact women compete less aggresively than men of the same ability. This suggests another possibility. Relative to women, men may be more (over?) confident. As a result, they think they have a greater chance of winning the tournament and therefore they compete more vigorously. When given the option of choosing what level of maze to solve (with increasing rewards for more difficult mazes) the men do systematically chose more difficult mazes than the women.
What do we make of all this? First, we have an additional explanation for wage differences between men and women, especially at the highest levels where competition for promotion is a tournament. Second, we have added support for single-sex education and perhaps even single-sex firms (Astute readers will recall what happened to the women on The Apprentice before and after the groups were mixed). See also the related first volume in this series.
The authors focus on a third potential implication – the benefits of making women feel more confident (e.g. in reducing drop out rates in science and engineering). The latter, conclusion, however, doesn’t take into account the costs of effort. If men are over-confident about their abilities then they put too much effort into tournaments. Increasing women’s confidence would only make them (and the men) worse off. Other than restaurant customers, would anyone be better off if more people thought they could become a Hollywood star?