Despite the gains made by women in the job market in recent decades, the access of women to the upper levels of the business hierarchy remains limited. A vast literature seeks to explain the barrier to female advancement widely known as the “glass ceiling,” which is regarded as “an egregious denial of social justice,” at least by the U.S. Department of Labor. But the two primary strands of this literature — one using experimental methods and the other using data gleaned from sports events — offer very different conclusions. In this essay, we examine what these studies have added to our understanding of the glass ceiling and point out the limits of the methods that they employ.
When it comes to athletes, it is generally accepted that slight variations in performances can vastly alter rewards. At the Masters Golf Tournament in 2013, for example, Adam Scott and Ángel Cabrera ended the regulation 72 holes tied for first place. After taking two extra holes to finally best Cabrera, Scott won US$1.44 million, while Cabrera received only US$864,000 for his second-place efforts, a difference of over half a million dollars. Overall, the performance of these two golfers was barely distinguishable, but the difference in their respective rewards was massive.
Winning contests also increases pay in much of the business world. Standard economic theory posits that linking worker pay to performance provides employees with an incentive to put out the optimal amount of effort. The most direct way of doing this is by offering a fixed payment per unit of output. A common application of the piece-rate payment system is a commission. A car salesperson who sells one more vehicle than a coworker, for example, receives slightly higher pay, while one who sells twice as many cars than other members of the sales team receives much higher pay. Unfortunately, firms cannot always determine employee contributions to output as clearly as an auto dealership. Even when management can observe productivity, monitoring worker performance can be costly. Firms also cannot always account for all the factors that influence productivity. As a result, managers often do not worry about absolute employee performance. Instead, they rate a worker’s output by comparing it to the productivity of his or her peers.
Simply put, determining the top-performing employee in a relative sense is easier than figuring out exactly how much every worker contributes to the firm. The workplace thus becomes a contest between workers — a contest in which the winner receives a relatively high reward and the loser receives a relatively low reward.
To generate the desired level of effort by the workplace contestants, a firm can adjust the difference in pay between the winner and loser in its rank-order tournament (ROT). Larger differentials increase the worker incentive to deliver optimal effort on the job. ROTs have been used to explain a number of observations about firms. For example, large firms tend to offer larger raises associated with promotion. In part, this is because the larger number of people competing for the promotion makes it more difficult for firms to identify the most productive worker. Since a weak link between effort and reward reduces the incentive that any given reward provides, some firms must offer a larger reward to maintain work intensity.
THE IMPLICATION OF EXPERIMENTAL STUDIES
While the theory behind ROTs is clear, measuring how people respond to the incentives presented by rank-order tournaments is not easy. Economists and other social scientists have attempted to increase our knowledge of how incentives influence performance via experiments. They have conducted experiments under “laboratory conditions” outside the messiness of the workplace, using randomly selected groups asked to perform specific tasks in carefully controlled settings. By modifying the setting for some individuals, these studies can compare the behaviour of experimental and control groups.
Recent experimental research (of which we present a sampling) has shown that women and men respond differently to tournament settings. In particular, the studies find that men respond positively to tournament environments while the performance of women has a tendency to falter. Gneezy, Niederle and Rustichini (2003), in a study conducted at Israel’s Technion engineering university, asked students to solve mazes. No difference in the performance of men and women was found when a fixed reward per puzzle solved was offered. But when the reward system was changed to a winner-take-all setting that only rewarded the person solving the most puzzles, the performance of men improved while the output of women stagnated. The male–female differences were largest in mixed-gender groups. A related study, by Gneezy and Rustichini (2004), showed that theses gender differences are robust to changes in both the sample and the nature of the activity. Israeli fourth graders were timed when asked to run both individual sprints and head-to-head competitions. Boys in this study ran faster in competitive settings, while the girls slowed down. The results were robust to the gender make-up of the competitive race and to the relative abilities of the runners.
Using a sample of British college students, Gill and Prowse (2010) shed more light on the competitive nature of men and women by adding a random element of luck while conducting contest experiments with multiple rounds. This study found that women reduced their effort to win future contests after experiencing a failure caused by misfortune. Men in this study reduced effort only after failing to win large prizes, while the female response to bad luck was independent of prize value. The difference in the response to misfortune was found to account for about half of the gender performance gap observed in the study.
With such a large difference in the responses of men and women to competitive settings, one might conclude that women would seek to avoid them. Using a sample of English high school students, Booth and Nolen (2012) showed just that. Given the choice between a piece-rate reward system and a tournament system, girls were more likely than boys to choose the piece-rate setting. Interestingly, girls who attended single-sex high schools were more likely to choose the riskier format.
There are three key findings from the experimental literature reviewed above:
- Women respond less positively than men to the incentives associated with competitive settings
- Women are more easily discouraged than men by random adversities
- Women are more likely than men to avoid competitive settings
As professional advancement becomes more reliant on contests, these conclusions indicate that women will continue to be at an increasing disadvantage as they move up the corporate ladder since the conclusions suggest that ending overt discrimination would not break the glass ceiling. But while the above experiments add greatly to our understanding of the preferences of men and women, they are limited by the two features that give them credibility.
In abstracting from the real world, experiments create a different environment from what people actually face in the workplace. Boys or men might prefer to enter a contest when all they risk is a few dollars, but they might reach a different conclusion when retirement savings or a child’s education fund is at stake. Furthermore, to generalize from experiments, one must be sure that the subjects of the experiment were randomly selected. If not, there is no reason to presume that a similar experiment conducted elsewhere would yield similar results. This limits the applicability of the experiments noted above because workers are not randomly chosen. Women and men who compete for executive promotions differ from women and men who are content to remain in lower-level positions.
Simply put, people choose specific occupations, making the random selection of participants in the experiments an artificial world that does not necessarily reflect reality. And since workers tend to sort themselves into settings for which they are particularly well suited, experimental findings must be tested against “real-world” data before drawing hard lessons from them. This, of course, presents its own set of problems. After all, ROTs supposedly arise precisely because employers find it so hard to measure workers’ performance.
THE IMPLICATION OF SPORTS RESEARCH
There is no question that sports data can provide economists with far more detailed information on compensation and individual performance than most other settings. Keep in mind that it is very hard to determine how much a mid-level accountant contributes to the output of a major accounting firm, even if pay records are available. But performance and pay measures abound for every major-league shortstop and shooting guard. For this reason, researchers turn to sports data to test hypotheses about how the labour market operates.
ROTs have become a particularly popular subject for study because sports such as tennis and golf explicitly follow tournament structures and reward participants on their rank finish rather than on their absolute level of performance. The literature began with Ehrenberg and Bognanno (1990a) and Ehrenberg and Bognanno (1990b), two seminal studies that examined how golfers respond to incentives on the PGA Tour. They found that PGA players do better in the final round of a tournament when they are closer to the lead and improving one’s ranking yields a greater reward.
Since the publication of these studies, ROT theory has been applied to a wide variety of sports, including NASCAR (von Allmen, 2001). Only recently, however, have economists turned their attention to whether male and female athletes respond differently to tournament incentives. And the findings of this literature generally — but not uniformly — contradict experimental results.
Studies involving professional tennis, for example, show that favourites in elimination tournaments win more frequently when the stakes are higher, regardless of whether they are men or women. Preliminary evidence also suggests that women on the Women’s Tennis Association tennis tour do not respond more negatively to setbacks than men, as they are no more likely to lose in straight sets than men are.
In a highly creative study, Paserman (2010) determined the crucial points in a series of tennis matches and used them to test whether women and men respond differently to pressure. Both women and men were found to play more conservatively in these situations. Women made more unforced errors, but the greater number of errors made did not appear to affect ultimate outcomes. Studies of golf tours, meanwhile, have found that women respond to prize structures and incentives in roughly the same way as men. In fact, stroke totals for final rounds of golf tournaments suggest that women respond more positively to pressure than men when competing in contests.
In studies of non-traditional sports, female skiers have been found to respond to the differences in rewards as they move up the leaderboard. Unfortunately, no comparable study has been done of men. In figure skating, women appear more responsive than men to incentives, as their scores improve more when close to the lead and when competing in high-profile events.
There is also evidence that women resemble men in terms of effort as well as performance. Studies of long-distance runners, for example, indicate that women train as hard as men.
Track-related studies have also found indirect evidence that suggests that top female sprinters tend to sort themselves among tournaments to avoid competing directly with each other. But when it comes to sports data, this is the only finding that really supports conclusions drawn from experimental studies.
While the findings of experimental studies may appear unduly gloomy, the findings based on sports data might be overly optimistic, since athletes are much more narrowly sorted than workers in the labour force. Virtually all sporting events are segregated by sex, with women competing only against other women. Furthermore, these women are in a narrow age band, ranging from the mid-teens to mid-thirties. As a result, they face less pressure from marriage and children than women in the wider workforce. And since studies using sports data are silent on the issue of how older women respond to incentives, the research might overstate positive responses to incentives.
The research cited in this essay points in opposite directions. According to the experimental literature, the existence of the glass ceiling stems from female preferences. That suggests that anti-discrimination public policy will have limited impact on women’s progress up the corporate ladder unless these preferences change, which would require women to somehow view themselves and their families, not to mention the world around them, differently. Indeed, if the conclusions of this body of literature are valid, nothing short of a complete overhaul of corporate reward systems will shatter the glass ceiling.
The literature based on sports data tells a different story. Our ability to draw workplace conclusions from studies involving female athletes obviously depends on how closely we can link women who choose these two spheres. Nevertheless, the sports-related literature finds that women and men behave similarly, which suggests that the glass ceiling is an artificial barrier that can be broken. And so, if women can compete with their male colleagues as well as they do with female competitors in sporting events, and if they can maintain this effort in the face of family demands until they retire, then we will see many more female executives once artificial barriers in the workplace are lifted.