![[Adobe Stock]](https://sp-ao.shortpixel.ai/client/to_webp,q_glossy,ret_img,w_750,h_375/https://www.psypost.org/wp-content/uploads/2024/03/Businessman-Using-Business-Computer-In-Office-750x375.jpg)
Even when men and women earn the same average score on a job test, differences in how widely their individual scores spread out can tilt hiring toward men whenever employers screen out most applicants. A new computer simulation study published in the Journal of Applied Psychology suggests that some common multi-step screening setups can also leave employers with a workforce that is both mostly male and less productive than it could be. The findings point to practical adjustments in how organizations combine job tests and set screening cutoffs to improve both gender balance and expected job performance.
At the center of the research is a statistical pattern known as greater male variability. Whenever a group takes a test or is measured on a trait, the scores have two basic properties: the average (called the mean, or the “first moment” in statistics) and the variance (the “second moment,” which measures how widely individual scores spread out from that average).
Greater male variability, often abbreviated as GMV, occurs when men’s scores spread out more widely than women’s, clustering more heavily at both the extreme high and extreme low ends while women’s scores cluster closer to the middle. Under this pattern, men are overrepresented among both top and bottom scorers even when men and women share the exact same average.
Many employers struggle with gender imbalances despite diversity initiatives. Women make up roughly a third of the workforce at companies such as Apple and Google, and fewer than one in five software developers in the United States. Efforts to explain hiring gaps usually focus on evaluator bias, stereotypes, or group differences in average performance.
Don’t miss the next big psychology finding
Every morning, PsyPost sends a short briefing on the most interesting new peer-reviewed research in psychology and neuroscience.
Yet even where average gender gaps on academic and cognitive tests have narrowed, wider male score spread has been reported in areas such as quantitative and mechanical reasoning, academic tests, personality traits like conscientiousness (being organized, dependable, and hardworking), and manager or peer ratings of social competence. Still, the size of these variance differences varies across traits and studies. A 2024 meta-analysis of creative ability, for example, found that differences between men and women in the spread of scores were small and inconsistent. Scholars have also raised concerns that research on greater male variability could be misused to justify gender discrimination.
Where wider male variability does appear, it can occur even on traits where women score higher on average. For instance, a 2021 meta-analysis in Psychological Science by three of the new study’s authors found greater male variability in cooperative workplace behaviors despite higher female averages.
“That was my first work involving greater male variability, but once I started thinking about its implications, I realized that it had huge implications for personnel selection processes,” study lead author Jose M. Cortina, a distinguished professor at Ohio State University’s Fisher College of Business, told PsyPost. “And that is exactly what our simulations showed.”
Because hiring research emphasizes group averages, organizations often assume that assessments with equal averages give men and women equal hiring odds. Cortina, who was at Virginia Commonwealth University when the study was conducted, and his colleagues modeled how score variance interacts with screening rules to shape workforce gender composition and productivity.
The team used Monte Carlo simulation, a computer-based technique that generates virtual datasets to test how variables interact under controlled conditions. The study’s design and analysis were not preregistered, though the authors posted their simulation code publicly online. Each simulated pool included 20,000 virtual job applicants (10,000 women and 10,000 men), with every scenario repeated 100 times.
In the main simulations, men and women had identical average scores, bell-shaped distributions, and job performance variability, isolating the effect of test score spread. Setting women’s variance to 1.0 across all measures, the researchers set men’s test score variance either 5% higher (a small gap seen in verbal reasoning or self-reported extraversion) or 40% higher (a large gap seen in quantitative reasoning, mechanical comprehension, or manager ratings of emotional competence).
Across 112 scenarios (11,200 simulated datasets), the authors compared two common hiring systems. In a multiple-hurdle system, candidates face sequential cutoffs: everyone takes a first test, top scorers advance to a second test, and a share of those finalists get hired. In a compensatory system, all applicants complete both tests and receive one combined score weighted by how well each test predicts job performance, letting strength on one measure offset weakness on another.
One assessment was modeled as a strong predictor of job performance and the other as a moderate predictor, with test overlap ranging from zero to substantial. Hurdle pass rates ranged from a strict 10% cutoff (selecting a small shortlist) to a lenient 90% pass rate (screening out only unqualified applicants), while each compensatory model hired the exact same number of workers as its hurdle counterpart.
Whenever a hiring system paired a strict 10% cutoff with high male variability, men ended up substantially overrepresented among hires across both sequential-hurdle and combined-score systems. Even when a lenient first screen eliminated more low-scoring men and left women as 51% of advancing candidates, following it with a strict cutoff on a high-variability test reversed that edge and produced a workforce that was roughly 60% male.
When two hurdles each accepted only the top 10% on high-variability tests, women accounted for about 31% of hires in the hurdle system and only about 18% in the combined-score system, showing that combined scoring does not fix gender balance on its own. Relaxing ultra-strict cutoffs modestly did far more to boost female representation than loosening already lenient screens.
Even a 5% difference in male variance tilted selective hiring toward men, leaving women at 48.7% of hires when a strict cutoff preceded a lenient one. Assuming equal numbers of male and female applicants, the authors calculate that if all of the roughly 66 million annual U.S. hires went through such a system, close to 1 million fewer women would be hired each year. Women outnumbered men slightly, reaching at most 52% of hires, only when employers accepted nearly all applicants.
For Cortina, the central takeaway for the public is straightforward: “That GMV is real, that some tools used for selection have more of it than others, and that if such tools are used to screen out large proportions of applicants, there will be gender disparities in the workforce.”
Sequential hurdles also reduced expected productivity. Across every scenario, multiple-hurdle systems produced lower predicted job performance than combined-score systems because rigid early cutoffs eliminate candidates whose strengths on the other test would have made them strong overall performers, though the gap was sometimes slight.
Both diversity and performance suffered worst when a lenient screen on the stronger predictor preceded a strict cutoff on a weaker, high-variability predictor. In one such scenario, the two-hurdle system produced a 60% male workforce, whereas a combined-score system yielded a nearly balanced group and stronger performers, with 64% of its hires exceeding the hurdle group’s average. For a job paying $100,000 annually, the authors estimated that this hurdle setup would cost an employer about $14,400 per hired worker each year in lost productivity.
The authors stress that these workforces are not less productive because they are mostly male. The shortfall comes from how the hurdle system weighs the two tests, while the male tilt comes from a strict cutoff on a high-variability test.
Cortina observed that this dual penalty on both diversity and productivity stood out as an unexpected finding. “A multiple hurdle selection system that applies a stringent cutoff to a hurdle that has GMV and is only modestly correlated with job performance will produce a workforce that is not only mostly male but also inferior with regard to expected job performance,” he explained.
Combined-score systems did not always help gender balance, however. When the high-variability test was also the stronger performance predictor, the compensatory model weighted it heavily, boosting productivity while hiring an even smaller share of women than a hurdle system that underweighted it.
Follow-up simulations tested altered assumptions. Lowering male variance to a moderate 20% gap softened gender disparities without erasing them, while adding a third hurdle compounded the effects: following a lenient screen with two strict, high-variability hurdles reduced women to roughly 32% of hires, while 73% of workers chosen via combined scores outperformed the hurdle group’s average.
When men were given a moderate advantage in average test scores, male hiring dominance widened further; giving women that same average advantage left women in the majority, though strict, high-variability tests still chipped away noticeably at their lead. Having men make up 75% of applicants worsened women’s proportional hiring rate under selective high-variability hurdles, whereas removing the lowest-scoring 20% of candidates beforehand left the main patterns largely unchanged.
Cortina cautioned against misreading these statistical patterns as claims about superior male ability. “The misinterpretation is, Oh, so you are saying that men are better??!! We are not,” he noted. “We are saying that there is more variance in male scores on many variables, even on variables for which the female mean is higher, and those variance differences have consequences.”
In the authors’ reading, greater male variability may shape careers beyond hiring. Because promotions act like sequential hurdles with tight cutoffs at higher ranks, the researchers suggest that high-variability promotion criteria could progressively shrink female representation toward top management. They also propose that having more men at the extreme upper end of assessment distributions may yield more competing job offers and higher starting pay for men, feeding into gender wage gaps.
“Any organization that is interested in reducing gender disparities in its workforce, and particularly at the top, should try to understand the paper,” Cortina said. “The problems described in the paper can’t be solved by people who don’t know that the problem is there in the first place.”
To reduce these disparities without losing productivity, the researchers recommend avoiding strict cutoffs on high-variability tests and substituting equally predictive tools with minimal variance differences, such as structured job interviews. Organizations can also expand targeted recruitment; in one moderate-variability scenario, the authors calculated that achieving hiring parity would require attracting roughly 12 female applicants for every 10 male applicants.
Cortina also built two online resources on his personal site so employers and readers can explore these statistical patterns firsthand. “One just illustrates GMV effects,” he added. “One is a calculator that helps organizations estimate the number of qualified women that they have to recruit in order to cancel out the effects of GMV on the gender makeup of the workforce.”
There are a few things to keep in mind when interpreting these findings. Most notably, the results come entirely from mathematical computer simulations rather than hiring records from actual employers. The authors also note that there is too little evidence to say how often the specific combinations they modeled occur in real hiring, and that any single combination is likely to be rare. In addition, the models assumed symmetrical bell-curve score distributions, whereas real applicant pools may be skewed and real job seekers may adapt strategically, such as declining to apply when a process seems stacked against them.
Researchers have also measured exact male-to-female score variance for only a subset of workplace tests, meaning estimation errors could shift how strongly variability operates in practice, though the authors report that the overall pattern held when they used alternative estimates. Future empirical and legal studies will need to evaluate the financial costs of expanded recruitment and examine how adjusting test weights or cutoffs interacts with antidiscrimination laws such as Title VII in the United States.
Asked about his long-term goals and next steps for this line of work, Cortina pointed to two priorities: “Raising awareness. Also, GMV effects in leadership-related phenomena.”
The study, “The Man of the (Second) Moment: How Greater Male Variability Affects the Gender Composition and Productivity of the Workforce,” was authored by Jose M. Cortina, Imran Kadolkar, Stefan Volk, and Christian Thöni.
