Despite a considerable decline of in the gender gap in the labor market over the last decades, gender differences at the top have proven to be highly persistent. For example, in their 2017 Journal of Economic Literature article, Blau and Kahn report that in Fortune 500 companies women make up only 14 percent of executive officers, 4 percent of CEOs, and 17 percent of board members. According to the most recent Global Gender Gap Report of the World Economic Forum worldwide only 18 percent of firms are led by a woman.

A large literature based on lab experiments suggests that these gender gaps can be partially explained by gender differences in the willingness to compete. In order to get in lucrative job, one needs to be willing to compete for that job. In experiments, women are much more reluctant to compete than (equally skilled) men are. Most of the studies in this literature base their experimental design on a seminal paper by Niederle and Vesterlund.

These experiments typically work as follows. Participants determine how they are compensated for their performance in a series of math problems where they have to add up sets of two-digit numbers. They can choose between a non-competitive piece rate and a competitive tournament incentive scheme. If they opt for the piece rate, they earn €1 per correct answer; if they opt for the tournament, they compete against three others, with the highest performer receiving €4 per correct answer and the others receiving nothing. In most studies, male and female subjects perform equally well on the math problems, but men are much more likely to choose the competitive option.

Compared to the financial interests that are at stake in career competitions, the money amounts used in willingness-to-compete experiments are typically very small. Notwithstanding the copious experimental evidence for the existence of a gender difference in willingness to compete, these small stakes raise the question of whether the observed behavior can be generalized to consequential real-world decisions.

To study gender differences in willingness to compete for high stakes, we use the Dutch version of the game show Deal or No Deal. This show has been on the air since 2002. In the final part of every episode, the winner of an elimination competition gets to play a game of chance with a maximum prize of 5 million euro and an expected payoff of approximately half a million euro.

The finalist is selected from an audience of 500 through a series of elimination games built around quiz questions. Similar to the tasks that are typically used in willingness-to-compete experiments, these quiz questions are often numerical or arithmetic in nature. In every episode, several contestants are offered the option to opt out of the competition in exchange for a substantial, but comparatively small prize.

These high-stakes choice situations almost appear to be designed by an economist as an experiment to analyze competitiveness differences with substantial incentives. In the paper, we focus on two different games that differ in strategic complexity.

Game one

In the first, contestants who most accurately answered an estimation question have to choose between competing in the next elimination game and opting out for a prize. This decision resembles the tournament-entry decision typically made by subjects in lab experiments. Mirroring the laboratory result, we find that women are more than twice as likely as men to opt out of the competition.

The contestants opt into the competition in a sequential manner. The first person who gets to make her decision does not know yet what the gender of her competitors in the next game will be, but those who make their choice later do know the gender of at least some of their competitors. Next, we investigate whether men and women condition their opt-out decision on the gender composition of the existing set of competitors. First, we simply distinguish between situations where most of the occupied seats in the next game are taken by men (> 50% male) and situations where half or more of the seats are taken by women (≤ 50% male) at the time the contestant makes her opt-out decision.

Interestingly, the higher opt-out rate of women derives entirely from situations where the majority of their opponents in the next stage would be male. Women seem to have a particular dislike of competition against men, rather than a general dislike of competition.

The binary division of whether more or less than half of already selected opponents are male is rather crude, and also neglects that the ultimate composition depends on the gender of the opponents who take the remaining vacant seats. We therefore also investigate the effect of gender composition by looking at the expected proportion of males among the total of four competitors in the next elimination game. Again, we observe that women dislike competition against men. In contrast, men do not prefer to compete against women; if anything they prefer to compete against other men, but this is not statistically significant.

The results for this first game thus confirm that the gender difference in willingness to enter competitions—robustly found in lab and classroom experiments with low stakes—is also present in an environment with very high monetary stakes: 13 percent of men and 28 percent of women choose to opt out of the competition in return for a relatively modest prize. Interestingly, the comparatively high opt-out rate of women derives entirely from situations where the majority of their opponents are (or are expected to be) male, which suggests that women have a particular dislike of competition against men, rather than a higher general aversion to competition.

Game two

The second game that we study is a head-to-head confrontation at the end of the elimination competition. One quiz question determines the last elimination, unless one of the two remaining contestants voluntarily accepts an opt-out prize.

This game is more complex than the former because the optimal choice depends on the anticipated behavior of the opponent. In the previous game, people who want to compete do not benefit from others’ reluctance to compete (contestants who opt out are replaced by another, randomly selected contestant). In this second game, however, those who choose to compete automatically win the competition if the other opts out. Hence, if you believe that your opponent will budge, you should not. Such strategic interaction plays an important role in many real-life competition choices.

In this strategic game, we again do not find that women are less competitive overall. Women compete as much against women as men do against men. Women are, however, much more likely to be the one who opts out when they are facing a man.

Further analyses reveal that this difference arises partly because women again avoid competing against men, and partly because men—apparently anticipating this lower competitiveness of their female opponents—compete more against women in this strategic game.

In a third game, which we analyze in the Appendix, and which has a similar strategic nature as the second game shows the exact same thing: in strategic interactions women avoid competing against men and men compete more against women.

Conclusion

Combined, the analyses of the different games suggest that women have a particular dislike of competing against men, and that men are aware of this preference and exploit it in a strategic setting. There is little reason to believe that these results are driven by gender differences in the ability to answer the questions of the various games: men and women perform roughly equally well in the actual tasks. These findings show that the gender difference in willingness to compete also occurs in a setting with exceptionally high stakes, and underline the importance of the gender of competitors, a factor that is mostly ignored in the literature.

Our results are particularly relevant for understanding the persistent gender gap at the higher rungs of the career ladder. Typically, the competitor pool gradually becomes more male-dominated the closer one gets to the top. If women are more likely to give up when faced with a male competitor and men compete harder when faced with a female competitor, male dominance in professional environments could become self-perpetuating.

In short, the Dutch version of the TV game show Deal or no Deal allowed Thomas Buser, Martijn van den Assem, and myself to study gender differences in willingness to compete for high stakes. The paper is available on SSRN .