When do we observe a gender gap in competition entry? A meta-analysis of the experimental literature. Eva Markowsky, Miriam Beblo. Journal of Economic Behavior & Organization, Volume 198, June 2022, Pages 139-163. https://doi.org/10.1016/j.jebo.2022.03.030
Abstract: This paper systematizes the experimental evidence on gender differences in competition preferences with a meta-analysis of 110 studies and 409 effect sizes on observed or residual gender gaps in experimental tournament entry. Our meta-summary confirms that, across all studies, men choose a tournament scheme 13 percentage points more often than women, which is only about a third of the gap found in Niederle and Vesterlund's (2007) seminal paper. Our meta-regression analysis reveals that larger gender differences are indeed prevalent in studies that most closely apply the Niederle-Vesterlund design, i.e., differences are largest in lab experiments with student subject pools and when math tasks are involved, but almost negligible for other age groups, verbal tasks, and in field-like environments. Experimental interventions such as information treatments or affirmative action measures prove very effective in reducing or even eliminating the gender gap. Although some measures of risk preferences and confidence are systematically related to the estimated residual gender gap in tournament entry, they do not eradicate competitiveness as a distinct trait. Finally, higher gender equality at the country level seems to go along with larger differences in women's and men's competition preferences.
Keywords: CompetitivenessGenderExperimentsMeta-analysis
JEL: J16 (Economics of Gender)D91 (Role and effects of psychologicalemotionalsocialand cognitive factors on decision making)C9 (Design of experiments)
1. Introduction
When Jane Fraser became CEO of Citigroup in February 2021, the number of female CEOs in the Fortune 500 increased to 37, resulting in a women's share of all top-level executives in the largest corporations in the United States of just over 7% (Ghosh 2021).1 This strikingly low number illustrates a common observation: Despite recent progress towards gender equality in the work place, women are still significantly underrepresented in leadership positions. The imbalance is not limited to private corporations but noticeable in the public sector as well, although to a somewhat lesser extent (DeHart-Davis et al. 2020; Cotroneo et al. 2021). In academia, similar patterns exist where women's representation diminishes throughout academic careers (e.g., European Commission 2019). To this day, in every sector of the labor market, climbing the career ladder to the very top seems to be much less likely for women than it is for men.
The idea that this gender imbalance may be driven by the work environment in high profile jobs, with high levels of competitive pressure, has garnered significant attention (e.g., Sandberg 2013, Bertrand 2011). A rather new branch of experimental literature in economics seems to confirm that women perform less well in highly competitive environments and that they are more inclined to avoid competition than men, who, in turn, tend to compete too much (Niederle 2017). We contribute to this field of research by systematizing the experimental evidence on gender differences in willingness to compete in a quantitative meta-analysis that assesses the size of the gap as well as its moderators.
The concept of competitiveness is commonly thought of as a trait comprising observable and unobservable latent components – among them risk and feedback aversion, (over)confidence, ability to perform under pressure and the willingness to enter a competition. Shurchkov & Eckel (2018: 488) argue that the latter represents a “revealed ‘preference for competition’”, making it an obvious choice for experimental investigations of competitiveness. Niederle & Vesterlund (2007) were the first to investigate these gender differences systematically in a laboratory experiment. In their seminal paper, university students perform a mathematical real-effort task twice, once under a piece-rate compensation and once in a winner-takes-all tournament in groups of four, where the winner is paid four times the piece rate per correctly solved problem and the other group members receive no payment. After these two performances, the authors let the subjects choose between piece rate and tournament compensation for the third round, to measure their competitive preferences, and find that women choose the tournament substantially less often than men. The gender gap persists even after controlling for risk aversion, confidence, and performance in a regression framework. In the following years, numerous experimental studies built on the Niederle-Vesterlund (NV) design and tested competitive preferences of women and men with different subject pools, different tasks, under different rules, and in different experimental settings. In doing so, researchers created many of what Hamermesh (2007) classifies as “scientific replications” – tests of an initial finding with different data and methods.
Reviews of this body of literature agree that, as a whole, these ”replication” efforts confirm women to be less willing to enter competitions than men, while emphasizing that the magnitude of the gap depends on the context and other, potentially unobserved, covariates. Among the two most recent and comprehensive, Niederle (2015) underlines the importance of beliefs, risk attitudes, other-regarding preferences, and the experimental task as possible confounding factors in measurements of the gender gap in competition entry. She names a number of other factors that may contribute to the gender gap: hormones, age, socioeconomic status, Big Five personality traits, priming, and culture. Finally, Niederle sees affirmative action and same-sex competition as possible interventions that may induce women to compete at similar levels as men. The review by Shurchkov & Eckel (2018: 10–13) adds stereotyping (though related to beliefs) and subject-pool differences to the list of factors potentially explaining the gender gap in competitiveness and contrast these with moderators that presumably reduce the gap: same-sex tournaments, competition in teams, magnitude of the prize, affirmative action, information/feedback/advice, “priming with ‘professionalism’”, or less time pressure.
In this paper, we complement the qualitative reviews with a systematic quantitative assessment of existing experimental studies on the gender gap in willingness to compete. Given the large number of studies on the subject, a meta-analysis can provide valuable insights where qualitative reviews, however thoroughly conducted, have their limitations. A meta-analysis can pin down an exact effect size and quantify the relative importance of moderators and interventions. While the results of our meta-analysis do not contest the central findings of the existing reviews, we believe that a quantitative analysis helps painting a fuller picture. In particular, we show under which conditions a gender gap in tournament entry emerges and we contribute to the current debate on distinct preference traits and the influence of overall gender equality on the gender gap in willingness to compete. Our analysis informs future research and policy makers aiming to design environments where women feel safe to compete.
The contribution of our paper is hence threefold: First, we summarize all experimental studies on the subject and present standardized effect sizes and their statistical qualities. Our meta-analysis comprises 110 experimental studies on gender differences in the willingness to compete and thereby twice as many as the most comprehensive qualitative review by Shurchkov & Eckel (2018) which reports on 58 papers, including investigations of competitive performance. We systematize the circumstances under which gender gaps in competition entry emerge in these studies by departing from very close NV replications and differentiating between study moderators and intervention moderators when enlarging the sample by more diverting variants of the original experimental design. We investigate the effectiveness of different intervention types aimed at higher competition rates of women. Our results confirm the great importance of the subject-pool and the nature of the experimental task in shaping the gender gap in tournament entry. They also highlight feedback, information, and affirmative action as the most efficient tools for reducing it.
Secondly, we exploit the meta-perspective to study how the measured competition gender gap changes when related traits are considered in a regression framework. This part of our analysis complements the recent debate about competitiveness as a distinct trait. Gillen et al. (2019) point out that experimental elicitations of the gender gap might be subject to bias resulting from erroneous measurement of the related factors risk preferences and overconfidence. The argument is that once measurement error in risk taking and confidence is minimized, both factors explain the majority of the gender gap in competition entry, leaving little room for a separate trait competitiveness.2 The authors show that one feasible way of reducing measurement error is to elicit and include multiple measures of risk and confidence. We build on this discussion and exploit heterogeneity across studies when determining the residual gender gap in tournament entry by different measures and controls for risk and confidence. Our analysis shows that controlling for risk can indeed lower the gender gap in competition entry. However, across the studies in our data set, we do not find evidence that including measures of risk (or confidence) reduces the residual gap systematically towards zero or renders it insignificant.
Thirdly, we contribute to the discussion on the influence of the cultural environment on gendered preferences. Our meta-data set of experimental competition studies conducted in 30 different countries and over a decade enables us to compare gender gaps in tournament entry relative to these countries’ respective state of gender equality. By complementing our data set with an indicator of gender equality, we show that competition gender gaps are larger in countries with higher levels of equal opportunity and parity between women and men.
The rest of the paper is structured as follows: Section 2 introduces our data set of experimental studies as well as the relevant effect sizes, i.e., measures of gender gaps in competition entry, and the different types of explanatory factors (moderators) included in the meta-analysis. Section 3 uses meta-analytical tools to quantify the reported competition gap, including formal tests for the presence of a potential publication bias in this body of literature. Section 4 applies meta-regression analysis to explain the heterogeneity in the literature and to answer the questions on effective interventions, correlations with risk preferences and overconfidence, and the cultural environment raised above. In Section 5 we conclude.