Children learn from their interactions with external stimuli. They interact first with their parents and families, then later with their schools, the print and visual media they encounter, and many other sources. From these interactions, they learn a wide range of things: about societal norms and rules, about cause and effect, and about people both like and unlike themselves, just to name a few.
This process is important because it can shape both children’s futures, and broader societal structures. If a child learns that one path for their future has great promise – for example, learning to play a musical instrument or sport – it may lead them to spend more time on it. As they practice and improve, their actions create the reality behind the message.
Similarly, telling a child they are bad at a subject – for example, the study of math – may cause a lack of enthusiasm for it, less study effort on that subject, and worse performance, generating a self-fulfilling vicious cycle. Exposure to beliefs about the roles that various identity groups can or cannot inhabit also has the potential to entrench historical patterns of group-based exclusion.
Our study uses this framework to better understand a key problem: why do so many people, in so many societies, believe that men are inherently better than women in the study of mathematics? This belief is surprisingly common across the world, from countries in North and South America to many in Europe, Africa, and Asia, despite the fact that women now consistently perform as well or, sometimes, even better in math than men in many academic settings.
In this paper, we test a key hypothesis from the economics and psychology of belief formation: that the persistence of this belief, despite its frequent disconnect from reality, is partly due to the transmission of beliefs from one generation to the next.
Many beliefs form in childhood. If children learn primarily from adults, who themselves learned from the previous generation, then beliefs that were true in the past but are wrong today can nonetheless be passed on from one generation to the next. What’s worse is that, if these beliefs affect reality in a way that reinforces their key messages, they may be even harder to reverse.
This is a tricky hypothesis to test. Normally, scientists would run an experiment, randomly assigning some children to parents who hold this belief and others to parents who don’t. Randomly assigning children to parents with different beliefs, of course, isn’t feasible; even if it were feasible, it would be unethical because exposure to these beliefs has been shown to cause harm.
In our paper, we solve this problem with a quasi-experimental research design which takes advantage of naturally occurring, random variation to approximate just such an experiment. We use the fact that, in China, middle school students are randomly assigned to classrooms. In Chinese middle schools, we also see that girls outperform boys in all subjects, including mathematics. Surprisingly, and in spite of this, nearly 50 percent of parents, and an even greater proportion of children, still believe that boys are inherently better than girls at learning math.
Our data come from a nationally representative survey of students, their parents, and their teachers in Chinese middle schools. For each child in our data, we know whether or not their parent agrees with the statement that “boys’ natural ability in learning math is superior to that of girls.”
We study what happens when, by chance, a child gets randomly assigned to a classroom with a larger or smaller proportion of peers whose parents hold this belief. Specifically, we estimate whether that child is more likely to hold this belief after being exposed to these peers and, separately, how the child then performs on their school-administered math tests.
We find that children’s peers are a key site for the intergenerational transmission of this belief. Children who are randomly assigned to classrooms with more peers whose parents hold the belief are significantly more likely to hold the belief themselves. This is true for both boys and girls. We then show that this transmission is greater from peers of the same gender, following the notion of “homophily.”
Homophily is the idea that people associate with and learn more from similar others. In our context, this means that, for girls, the impact of being assigned to a classroom with more girl peers whose parents hold this belief is far greater than the impact of being assigned to a classroom with the same number of boy peers whose parents hold the belief. We see symmetric effects for boys, whose beliefs are far more influenced by the proportion of boy peers whose parents hold the belief than by the proportion of girl peers whose parents do. This pattern, we note, is exactly what we would expect if children are learning about the world and their place in it at least partially from their friends.
This exposure affects children’s academic performance in math differentially by gender: boys who are assigned to a classroom with a greater number of peers whose parents hold this belief score better in midterm math tests, and girls assigned to a classroom with the same composition of peers score worse on these same tests.
In other words, random levels of exposure to this belief generate impacts on math learning which reinforce the message behind the belief, causing boys to perform better than girls. Here, too, we see evidence of homophily: we find greater effects on math performance – which are positive for boys and negative for girls – from peers of the same gender than from peers of the opposite gender.
Our findings showcase the importance of the informational environment in which children grow up. We show that exposure to this very common message about gender-specific ability not only transmits across generations via children’s peers, but also has real impacts on children’s academic performance. This has clear implications for other external stimuli – for example, teachers, the content of school curricula, and the media – which also help shape children’s view of the world and their place in it.
Our findings also have important implications for societal beliefs beyond gender; we suspect these causal pathways would be similar for any situation where people are taught about their value in society based on identity group membership – e.g., ethnicity or socio-economic status. If exposure to such messages in other contexts generates differences in beliefs, actions, and outcomes, it could also perpetuate other inter-group inequalities and disadvantages.
Sadly, beliefs about innate ability differences by gender, race, ethnicity, socioeconomic status, and other identities are extremely common across many societies. In light of this, it is crucial that our education systems counter such messages and ensure that all children learn that they can flourish in a wide range of possible futures.