The gender gap in risk-taking disappears when economists account for guessing

A large study of Chinese schoolchildren shows that girls are not inherently more risk-averse than boys. By accounting for cognitive errors and guessing, researchers found that gender differences in risk-taking actually depend on the size of the potential reward. The findings were published in the NBER Working Paper Series.

Traditional economic models assume that people make decisions with perfect rationality. Under these classical models, a person’s choices directly reflect their true internal preferences. Behavioral economists now recognize that people often make cognitive mistakes or rely on mental shortcuts.

These inconsistencies create a measurement problem for researchers. When a subject makes a choice, it is difficult to tell if that choice reveals their true preference or simply a lapse in attention. This challenge is especially prominent when testing children, whose decision-making skills and attention spans are still developing.

Economists Cesar Chavez Padilla and James J. Heckman from the University of Chicago, along with Shuaizhang Feng from Jinan University and Zhe Yang from Liaoning University, conducted the study to isolate true preferences. They wanted to separate a child’s actual risk tolerance from random guessing and mental fatigue.

The researchers analyzed data from a large study of 2,204 fifth graders in China. The children were between 10 and 11 years old. They completed a series of interactive games on tablet computers to measure their willingness to take risks.

These games used a format called multiple price lists. In these tasks, students faced a sequence of choices between a guaranteed number of tokens and a gamble, such as a coin flip that could yield a much higher or lower amount. As the child moved down the list, the gamble became increasingly attractive, and the point where they switched from the safe option to the gamble revealed their risk tolerance.

A perfectly rational person would switch from the safe option to the gamble exactly once. However, children often switched back and forth multiple times within a single list. Rather than throwing out these inconsistent responses, the researchers used them to measure the quality of the students’ decision-making processes.

To understand the psychology behind the choices, the researchers measured the children’s intelligence using standard visual puzzles. They also collected personality assessments based on the Big Five traits. These traits include conscientiousness, agreeableness, extraversion, openness to experience, and emotional stability.

The researchers relied on teacher evaluations rather than student self-reports to measure these personality traits. Teachers observe students over long periods in many different contexts. This avoids the bias that often occurs when children judge their own behavior.

The research team built a mathematical model to separate the students’ true risk preferences from two types of errors. The first error type is deliberation noise, which occurs when a person tries to evaluate a choice but makes a mistake in their reasoning. The second error type is simply guessing, where a person stops trying to solve the problem and picks an option at random.

The researchers also separated the errors into temporary lapses and persistent individual traits. Some mistakes are completely random, like a momentary distraction. Other mistakes stem from persistent cognitive habits, meaning some children are consistently noisier in their thinking across all the games they play.

The model also abandoned a rigid assumption common in previous economic research. Standard models use an equation that forces a person’s risk aversion to remain constant regardless of the prize size. Instead, the researchers used a flexible framework called an Expo-Power utility function.

This mathematical approach allows a person’s willingness to take a risk to change as the potential rewards grow smaller or larger. For 10-year-olds, the difference between a one-token prize and a 200-token prize is substantial. A flexible model can capture how the size of the stakes alters a child’s psychological approach to the gamble.

When the researchers ran a basic analysis that ignored cognitive noise and guessing, they replicated a common finding in behavioral economics. This basic model suggested that girls were consistently more hesitant to take risks than boys.

Once the researchers applied their full model, which accounted for guessing, decision noise, and varying reward sizes, this gender gap vanished. The flexible model revealed that risk aversion actually depends heavily on the stakes involved. When playing for small amounts of tokens, boys were more risk-averse than girls.

As the potential rewards increased, the trend reversed entirely. For the largest gambles in the games, girls became more risk-averse than boys. This indicates that measuring risk tolerance as a single, fixed number can produce highly misleading conclusions about gender differences.

The study also showed that personality and intelligence heavily influence how consistently children make choices. Higher intelligence scores correlated with less random guessing for both genders, and with lower deliberation noise for boys. Overall, about 30 percent of all the choices the children made were categorized as mistakes rather than reflections of their true preferences.

The researchers broke down these mistakes into their exact sources. Transitory random noise accounted for about 12 percent of the errors. Persistent individual noise accounted for about nine percent, and outright guessing made up the remaining nine percent.

Beyond intelligence, personality traits functioned differently depending on the child’s gender. For boys, traits like conscientiousness directly affected their baseline risk preferences. Conscientious boys were more risk-averse, and they also displayed much less noise in their deliberation processes.

For girls, personality traits primarily influenced the consistency of their decision-making rather than altering their actual tolerance for risk. Traits determined how often girls guessed and how noisy their thinking was. None of the measured personality traits had a statistically distinguishable effect on the girls’ core risk preferences.

The researchers also looked at the students’ socioeconomic backgrounds, including factors like parental education and household stability. They found that a child’s environment had almost no direct impact on their risk preferences or their tendency to make decision errors. Socioeconomic factors only shaped choices indirectly by influencing the development of the children’s intelligence and personality traits.

A limitation of this research is its focus on a single point in time, which prevents the authors from observing long-term developmental changes. While traits influence choices, it is also possible that economic preferences and environments shape personality traits over time. Future studies could track these children as they grow to see how these characteristics evolve together.

Another challenge is defining a baseline level of wealth for children, which is mathematically necessary for calculating how they perceive financial risks. Because young students do not manage household finances, the concept of wealth is abstract in this context. The authors ran alternative analyses with different wealth assumptions and found that their primary conclusions remained stable.

The study, “Relating Cognitive Skills and Personality Traits to Economic Preferences: Decision-Making under Cognitive Shocks and Guessing,” was authored by Cesar Chavez Padilla, Shuaizhang Feng, James J. Heckman, and Zhe Yang.