Beyond the Glass Ceiling: How Behavioral Science is Closing the STEM Gender Gap

By Lucía Rosich, Guillermina Suárez, Irina Sánchez, and Irene González

The global disparity in Science, Technology, Engineering, and Mathematics (STEM) is more than a statistical imbalance; it is a profound societal challenge that stifles innovation and limits economic mobility. While the digital age demands a diverse workforce capable of navigating complex technological landscapes, the reality remains that women are significantly underrepresented in these high-growth sectors.

In the United States, despite women comprising roughly half of the working-age population, they accounted for only 35% of the STEM workforce and a mere 16% of engineers as of 2021. This underrepresentation is not a singular event but a cumulative process. Research conducted by Speer (2023) indicates that the gap is distributed throughout the academic and professional pipeline: 35% is attributed to unequal preparation in STEM subjects prior to university, 26% to a disparity in initial enrollment in STEM majors, and 41% to the uneven transition into STEM careers post-graduation.

This issue is not confined to the Global North. In Uruguay, the Information and Computer Technology (ICT) sector faces a critical paradox: while the industry requires 24% of the national labor force and offers some of the country’s most lucrative wages, it suffers from a staggering 58% gender gap. To address this, organizations like Ceibal have launched initiatives such as "Jóvenes a Programar" (JaP), a program designed to bridge the skills gap for young adults. Yet, even within these inclusive programs, the gender gap persists—rooted in the very way we test and select talent.

The Psychological Barriers of High-Stakes Testing

The underrepresentation of women in STEM is frequently exacerbated by their performance on high-stakes entrance exams. Academic literature has consistently pointed to a "confidence gap" and heightened test anxiety as primary drivers. Ayuso et al. (2020) observed that differences in the perception of self-efficacy and test anxiety in mathematics begin at the primary school level and widen as students age.

Further research by Cotner et al. (2020) suggests that female students express significantly higher levels of test anxiety, which directly correlates to lower performance scores. Arias et al. (2023) reinforce this by highlighting that differences in risk aversion, self-confidence, a preference for collaboration over competition, and the intense pressure of timed environments create an uneven playing field for women.

For the JaP program, this created an unintended barrier. Because the program is in such high demand, applicants must pass a self-administered online entrance exam. Data showed that women were consistently achieving lower scores and leaving a higher proportion of questions unanswered compared to their male counterparts—even though the test included no penalties for incorrect answers. This phenomenon suggests that psychological, rather than intellectual, factors were dictating the outcomes of the selection process.

Chronology of a Behavioral Intervention

Recognizing that the status quo was failing to provide equal opportunity, Ceibal’s Behavioral Insights Lab initiated a project to re-engineer the testing environment. The objective was clear: mitigate the negative impact of test anxiety to ensure that the selection process measured potential, not just composure under pressure.

Phase 1: The Design of Interventions

The team developed two distinct interventions integrated directly into the online testing portal:

1. The Sanitized Intervention: This version removed potential sources of anxiety. It stripped away specific topic titles from question blocks that might trigger "stereotype threat," included a clear progress bar to provide transparency, and explicitly stated that incorrect answers carried no penalties.
2. The Stress Reappraisal Intervention: This version included all "sanitized" features but added a psychological layer. Before the exam, participants read a brief passage explaining how the physiological symptoms of stress (increased heart rate, alertness) are actually a tool to enhance focus. They were then asked to write a few sentences interpreting their own stress in a positive light. A secondary prompt was included halfway through the two-hour exam to encourage mindfulness and cognitive reframing.

Phase 2: The Randomized Controlled Trial (RCT)

To ensure empirical validity, the Lab conducted an RCT with 6,094 young adults. Participants were randomly assigned to one of seven test versions: four were "sanitized," one included the full "stress reappraisal" package, and two served as the traditional control group.

Supporting Data: Evidence of Impact

The findings were both conclusive and encouraging. Both the "sanitized" and "stress reappraisal" interventions led to a marked reduction in the gender gap regarding skipped questions and overall performance (the proportion of correct answers among attempted questions).

The "stress reappraisal" intervention proved most effective, nearly eliminating the gender performance gap entirely. Furthermore, the rate of skipped questions—a clear proxy for risk aversion and anxiety—was reduced by more than 50% compared to the control group. These results demonstrate that when the "noise" of anxiety is reduced, the performance gap between genders narrows significantly.

Official Perspectives: The Role of Behavioral Science

Irene González, who has led the Data and Behavioral Sciences area at Ceibal for the past two years, notes that these results represent a paradigm shift in how educational institutions view entrance requirements. "We have spent years looking at curriculum and access, but we must also look at the ‘architecture’ of the test itself," González explains. "If the way we test excludes qualified candidates due to psychological barriers, we are not just failing those students; we are failing the industry that needs their skills."

Irina Sánchez, head of the Behavioral Insights Lab, emphasizes that these interventions do not "lower the bar." Rather, they ensure that the bar is accessible to everyone. "By providing the tools to manage stress—such as the reappraisal exercises—we aren’t giving women an advantage; we are removing an artificial disadvantage caused by the high-pressure environment of the test," Sánchez says.

Implications for Global Education Policy

The implications of the Ceibal study extend far beyond a single programming course in Uruguay. The findings suggest that the global STEM gender gap is, in part, a failure of assessment design. If high-stakes exams are inherently biased toward those who thrive in high-pressure, competitive environments, they will continue to filter out talented individuals who possess the technical aptitude but suffer from the systemic pressures of stereotype threat or test anxiety.

Scaling the Solution

To move forward, educational institutions and corporate HR departments should consider the following strategies:

• Low-Stakes Testing: Implementing smaller, frequent assessment instances rather than one singular high-stakes "gatekeeper" exam.
• Default Transparency: Clearly communicating that incorrect answers will not be penalized, and providing visual cues like progress bars to reduce uncertainty.
• Psychological Framing: Integrating short, evidence-based writing exercises that allow students to reappraise their stress, which has been shown to benefit all students, not just those experiencing high anxiety.

Conclusion: A More Inclusive Future

The success of the interventions implemented by Ceibal’s Behavioral Insights Lab underscores a critical truth: equity is a design choice. By addressing the psychological barriers that disproportionately affect women, we can foster a more inclusive and equitable educational environment.

This research confirms that the path to closing the STEM gender gap involves more than just advocacy; it requires the meticulous application of behavioral science to the tools and structures of our education system. As we continue to refine these interventions, the potential to empower women in STEM fields becomes more tangible. Every student who enters the ICT workforce because they were given a fair chance to demonstrate their potential represents a victory for innovation, diversity, and economic progress.

The evidence is clear: when we remove the barriers created by anxiety, we don’t just see better test scores—we see a more capable, diverse, and representative future for the global STEM workforce. The work of Ceibal serves as a blueprint for organizations worldwide to begin dismantling the invisible walls that have kept women out of the future of technology for too long.