In many countries around the world, girls still perform more poorly on mathematics tests than boys.

Some researchers have argued that this gender gap is biological in nature: as then-Harvard University President Larry Summers famously put it in 2005, "there are issues of intrinsic aptitude" underlying women's underrepresentation in math-heavy fields such as science and engineering.

But in recent years, most economists have come to accept a different explanation.

"The alternative explanation is that there are cultural views that affect the way boys and girls are being raised," said Paola Sapienza, a professor of finance at Kellogg. Lending credence to this cultural explanation, Sapienza and her colleagues previously found that the gender gap differs widely around the world, with girls performing better relative to boys in societies where women have more power and equality.

Still, just why these social factors have an impact on performance has remained unclear. One possibility, Sapienza explains, is a lack of representation in math-based careers. "It could be that there are not enough examples for little girls to look up to, so they don't get inspired." Alternatively, perhaps a biased educational system pushes girls into other disciplines. Or perhaps the bias starts at home.

"Home is an obvious place to look," says Sapienza. "But nobody has done a systematic study looking at whether parents' beliefs and preferences regarding the role women should have in society explain kids' achievement gaps."

Until now. A comprehensive new study presents evidence that parents are indeed transmitting biases about gender roles to their daughters, including the notion that math is a traditionally male pursuit, causing a corresponding hit to the girls' math performance in school.

A team that included two Northwestern researchers-Sapienza and David Figlio, dean of the School of Education and Social Policy-conducted two analyses to examine whether a family's attitude toward boys and girls could impact their daughters' performance in math. The research team also included Gaia Dossi of the London School of Economics and Paola Giuliano of UCLA's Anderson School.

First the researchers used data on a family's observed fertility choices to determine which families were likely biased toward having boys. They found that girls raised in these "boy-biased" homes score lower on math tests between 3rd and 10th grade compared to girls who are not subject to such biases at home. That difference is equivalent to one quarter of the math performance gap between children whose mothers finished high school and those whose mothers dropped out.

A second analysis similarly finds a strong relationship between maternal attitudes toward gender roles and girls' performance.

"Because we can't perfectly measure what families really think about their daughters, we wanted to investigate this question in multiple ways," says Figlio. "But regardless of whether we looked at stated attitudes regarding gender roles or apparent preferences for sons, we found that daughters performed worse when their families were apparently biased toward boys."

A preference for boys

To test their hypothesis that parents might be transmitting their own gender biases to their children, the researchers first needed to know which families are likely to hold traditional views about gender roles. So they found a clever proxy for gender bias: simply put, families that prefer boys tend to keep having children until they get a boy-and then stop. Would girls from these families score more poorly in math?

The researchers used data from the Florida Departments of Health and Education that merged nearly 1.6 million birth records from children born between 1994 and 2002 with public school records for those children. This created a dataset that correlated individual students' performance on standardized math tests with information about the child's family, including data pertaining to family composition and socioeconomic status.

"This is the first large-scale study where you see how parental beliefs may affect children's outcome."-Paola Sapienza

The researchers first confirmed an overall "boy bias" among the Floridian households in their sample, finding that families of firstborn girls had more children than families who produced boys first. Then they divided their sample into families that showed a fertility pattern consistent with a boy bias-namely, families with only girls except for their lastborn-and those that did not.

Controlling for a range of variables also known to impact performance-including age, race, low-income status, and whether the child had special education needs-they found that girls from boy-biased families indeed scored lower than girls raised in other families.

The difference was larger among wealthier, more educated families, an effect that was of borderline significance.

"This is a very interesting result," says Sapienza. "Affluent parents have more resources to spend on their children. If they're biased, they may use those resources in a biased fashion." These parents might be paying for their sons to receive math tutoring or go to engineering camp, for instance, while signing their daughters up for dance lessons. The children of less affluent parents, on the other hand, may not benefit from any extracurricular opportunities, and so their biases are less likely to impact their children.

When the researchers ran their analysis in the opposite direction-to see if boys raised in analogously "girl-biased" families would suffer a similar dent in their math performance-they observed no such effect.

A transmission of attitudes about gender

The first analysis was remarkable for its huge sample size. But a potential downside of using data about family composition is that they are noisy: boy bias isn't the only reason that a family might choose to stop having children after producing a boy. Such a family structure might also be the result of random chance. So the researchers wanted to investigate possible gender bias from another angle as well.

The most direct way to measure this bias would be to literally ask girls' parents their opinions. As it turns out, such data do exist: the 1979 National Longitudinal Survey of Youth (NLSY) collected information about thousands of youth as they aged into adulthood, including data from standardized IQ and achievement tests and from surveys about attitudes toward gender roles. The NLSY also included some test and survey data from the participants' children.

The researchers focused on 4,934 women from the NLSY who had at least one child and analyzed how much they agreed with three descriptions of traditional gender roles, including "It is much better for everyone concerned if the man is the achiever outside the home and the woman takes care of the home and family," and "Women are much happier if they stay at home and take care of their children."

The more mothers favored traditional gender roles, the worse their school-aged daughters performed on a math test. Daughters of mothers whose preference for traditional gender roles was one standard deviation above the mean had scores 3% lower. Boys' math performance, meanwhile, was unaffected by their mothers' gender-role attitudes.

"Even with all the love you can imagine, we may bring home certain types of biases."-Paola Sapienza

The researchers also examined how much the women's sons and daughters agreed with statements like, "Girls and boys should be treated the same in school," and "A girl should not let a boy know she is smarter than he is." The researchers confirmed a strong relationship between the views of mothers about traditional gender roles and the views of their children-a relationship that only strengthened as children entered adolescence.

"Mothers' views are correlated with both the views of the boys and the views of the girls," says Sapienza. This is consistent with other studies finding that adults-of both genders-are affected by having had a working mother, with adult daughters being more likely to work outside of the home and hold supervisory positions, and adult sons spending more time caring for family members and being more likely to marry a woman who prefers to work outside the home.

"This is the first large-scale study where you see how parental beliefs may affect children's outcome," says Sapienza.

A reminder we all have biases

In the end, the gender gap in math achievement-and its potential downstream effects on women's presence in STEM career fields such as computer science and engineering-is unlikely to have a single source. The researchers have identified parental attitudes as one explanation for this gap.

"If you think about all of the potential explanations in terms of social norms that may induce girls to think that they have less chance to succeed in traditionally male-dominated disciplines," says Sapienza, "it's not surprising that kids absorb that much from 18 years in a home."

But, she cautions, parental transmission could be a particularly complicated explanation to address.

"There is not a clear policy recommendation here. The only one is a reminder to all of us that we all have biases," she says. "Even with all the love you can imagine, we may bring home certain types of biases."

This article was previously published in Kellogg Insight. It was republished with permission of the Kellogg School of Management.
tag