Cognitive disparities develop early and are very important
Lead, air pollution, structural racism, and the achievement gap
When people talk about the "achievement gap", they are usually talking about the difference in average educational attainment between Black and white students. When I started writing this post, that was what I intended to focus on. There is such a gap: it is pronounced, it is persistent, and its origins, though much-theorized, are ultimately unclear.
There are many ways of diagnosing achievement gaps, but among the simplest is by using tests. Standardized tests are not a perfect proxy for either academic or cognitive ability, but they are remarkably well-correlated with both. And these tests reveal a very reliable pattern: cognitive skills gaps appear as early as kindergarten.
Moreover, as the data make clear, there are other such gaps. The distance between the average cognitive test scores of Black and white kindergartners is about a third of a standard deviation. But this gap shrinks (though it does not disappear) once you account for socioeconomic status. In fact, cognitive skills among kindergartners vary far more much more by socioeconomic status than they do by race.
In this post, I'll make the following arguments based on the data:
Group differences in cognitive ability appear by kindergarten
Cognitive test scores at kindergarten predict scores in junior high school
Cognitive test scores for junior high students are highly predictive of important life outcomes.
So that makes this a very consequential puzzle. Something is causing the brains of poor children to develop differently than the brains of rich or middle-class ones. And that something, whatever it is, is conspiring to keep them poor.
Importantly, as I'll describe, this does not "explain away" racial disparities by recasting them as socioeconomic disparities. There are two reasons for this. First, because there are compelling reasons to believe that interpersonal racism still affects economic outcomes among adults. And second, because many Black families are themselves poor as a result of historical racism. Even if there were no racism today, the environmental factors acting on young brains are acting to perpetuate the effects of these historical practices. This is, in essence, what people are talking about when they talk about "structural racism."
The data
In this post, I'll use two datasets:
The Early Childhood Longitudinal Studies, which followed a representative cohort of about 18,000 students from the kindergarten class of 1998-1999 through eighth grade, keeping track of data about their families, their school achievement, and their environments more generally.
The 1997 National Longitudinal Survey, which followed a representative sample of about 9,000 people born between 1980 and 1984, starting in 1997, tracking not just their performance in school and facts about their families but also their higher education attainment, job market outcomes, and other facts about their adult lives.
My code, as always, is public.
Fact 1: Cognitive skills gaps appear early
I want to start with the basic facts about cognitive gaps, presented in two different ways. The students included in the ELS dataset were administered a cognitive test during their kindergarten year. The data below are based on students' combined score on the two main subsections of this test: a reading section and a mathematics section.
On the top, we have the difference between mean scores for Black and white students; it is this kind of difference in averages that is most commonly cited when discussing achievement gaps.
On the bottom, we have a picture of two overlapping distributions: one for Black students and one for white ones. This visualization demonstrates that the difference in means isn't simply a matter of unusually high or low scores. There are more Black students at the low end of the distribution and fewer at the high end as well.
Here's the same statistic, but visualized by socioeconomic status. As you can see clearly in the bottom panel, there is considerably more variation by socioeconomic status than there is by race.
Fact 2: Early-life gaps predict later-life gaps
The other important fact here is that early-life cognitive skills gaps don't currently tend to disappear as time passes. At least for now, gaps on cognitive tests at the kindergarten level strongly predict cognitive scores in eighth grade.
The kids in the ELS cohort were tested again in eighth grade; the plot below shows the relationship between their kindergarten scores and their eighth-grade ones.
Fact 3: Cognitive skills in high school predict life outcomes
For a longer-term view, we can turn to the National Longitudinal Studies, which follow kids into adulthood. The kids in the NLS97 cohort took the Princeton Individual Achievement Test in eighth grade. The plots below show the relationship between eight-grade PIAT scores and significant life outcomes: people's likelihood of having been arrested, likelihood of having attended college, and average income.
Three explanations that don't explain away these gaps
In reading these facts, many people will have considered some possible explanations. I think all of these explanations are at least theoretically possible, but I don't think they're plausible. Moreover, even if they do explain part of the observed disparities, they don't explain them away. None of them provides a comprehensive explanation for the observed data.
Racism of test proctors
This seems like a possible explanation for the observed gap in cognitive scores between Black and white children. Though the degree to which unconscious bias affects real-world outcomes seems to have been overstated, it definitely exists. It's reasonable to imagine that psychometricians who administer tests to kindergarteners could allow this kind of bias to leak into their work. It seems possible that this is responsible for at least part of the cognitive score gap between Black and white kindergarteners.
But this explanation doesn't work since we see achievement gaps by income as well as race: since most poor people in the US are still white, it's hard to account for these disparities on this basis alone.
Group differences in IQ
The first reason to discount this explanation is that there's generally no good support for it in the academic literature on IQ. Though IQ is heritable, there seems to be no good reason to assume genetic differences in intelligence at the population level: the populations we're talking about here (African-Americans and white Americans) are so heterogeneous that polygenic traits are unlikely to differ meaningfully across groups. Moreover, the IQ gap between Black and white Americans has decreased substantially in recent years, lending compelling support to the idea that environmental factors are to blame: this is not long enough for group-level genetic change to take hold.
Moreover, this also doesn't work for the same reason as above: we see variation in cognitive scores by socioeconomic status, but the groups at each SES level are extremely diverse. Hypothetical differences in IQ by race, ethnicity, or national origin don't make sense as an explanation here, since these characteristics are spread across different SES levels.
Individual-level genetic differences in IQ
It's true that IQ has a genetic component. Given this, some people — many of them on the Internet, and extremely vocal — might argue that there is a simple story here in which parents with low cognitive skills are (1) likelier to be poor and (2) likelier to have children with lower cognitive skills.
I won't say too much about this theory other than that the data doesn't bear it out: when looking only at parents with a graduate degree, there is still a clear relationship between SES and kindergarten-level cognitive scores.
I'll admit that (a) graduate-level education attainment is not the same as IQ and (b) that the overall disparity does appear attenuated in this visualization. But people with graduate degrees do have higher IQs on average and, as I say above, my point here is that this explanation does not account for the magnitude of the observed disparities.
Is this correlation or causation?
I think there are compelling reasons to think that what's happening here is causal. It doesn't take long to articulate why: we can be fairly confident that socioeconomic status itself has a causal impact on cognitive skills because children from low-SES households adopted into high-SES households experience an IQ gain of between 12 and 18 points - that is to say, between 0.8 and 1.2 standard deviations on a typical IQ test — roughly the size of the Black/white achievement gap. That is to say, adoption into a high-SES household raises cognitive test scores.
Structural racism in practice
Let's return now to the original question I started with. What is the cause of the achievement gap between Black and white students? I've argued that differences in cognitive scores at the kindergarten level are largely predicted by socioeconomic status, and that these scores help to produce socioeconomic status when those children become adults. Seen in this light, the achievement gap comes into focus as the means by which structural racism is perpetuated.
The argument here is simple: a collection of historical practices — redlining and exclusionary zoning chief among them — concentrated Black Americans in neighborhoods that were disproportionately exposed to environmental influences that irretrievably injure developing brains. This concentration is so profound that, even if institutional racism had simply stopped with the end of redlining (it didn't), it would still take generations for families to accumulate enough wealth to filter meaningfully out of these neighborhoods and away from their pernicious on children's cognition.
Moreover, Black Americans are disproportionately likely to live in urban counties. This is a legacy of the Great Migration, itself a direct consequence of the reign of terror and violence that afflicted Black people living in the Jim Crow South. As I will argue below, some of the main drivers of cognitive skills gaps are principally urban phenomena, which means that we should expect them to disproportionately affect Black Americans.
So — how, exactly, does low socioeconomic status translate into lower cognitive scores?
Possible culprits
I think this leaves only one good set of explanations: there is something happening roughly between conception and kindergarten that is causing cognitive disparities that persist through childhood and predict important differences in adulthood.
Research over the past few years has provided some insight into the potential culprits; I list the top candidates below. Note that I've left out factors for which I don't think there is good evidence — parental environment (like the famous but spurious "word gap"), preschool attendance, and maternal stress in utero. I don't think it's the case that these factors don't matter at all; I do think it’s the case that the factors I list below are much more important
Lead
Lead lowers children’s IQ. This has been known since the 1970s, when accumulated evidence on the topic led directly to the ban on leaded gasoline. Since then, the evidence has stacked up, with evidence from Australia to Chile suggesting that an increase of one microgram of lead per cubic meter of blood is associated with test scores reduced by 0.1-0.2 standard deviations.
Lead exposure is one of the effects of poverty. Older homes are more likely to have lead paint, pipes, faucets, and fixtures. And poor people tend to live in older homes. Studies done in places like Milwaukee have demonstrated that high-poverty census tracts have lead levels roughly 2 micrograms per cubic meter higher than baseline. This is also true more generally, with some evidence that the effects of lead on childrens' cognition are somewhow exacerbated for low-income families. Children living in redlined communities, of course, are more likely to have high levels of blood lead.
Maternal stress in utero
I don't need to tell you that being poor is stressful. In case you do need me to tell you that, however, there is generally good evidence that poverty is not just associated with stress, but that poverty causes stress. We get this suggestion of causality from studies that use unexpected exogenous (that is, external) shocks causing large drops in income, and seeing how those drops line up with increases in stress. For instance, random weather shocks in Kenya result in increases of the "stress hormone" cortisol in farmers' saliva.
It is this same type of natural experiment that give us good reason to suspect that stress in pregnant women, mediated by increased cortisol levels, causes reduced cognitive skills in the children that they ultimately give birth to. These studies sometimes use natural disasters, like an ice storm in Quebec or an earthquake in Chile: children opposed to an "acute stressor" like a severe earthquake have cognitive test scores around 0.1 standard deviations lower than those who weren't exposed. Perhaps the most convincing kind of experiment compares siblings composed to some such shock to siblings who aren't, allowing researchers to hold family factors constant to estimate the effect of the shock.
Bear in mind that these acute shocks are relatively brief. What is the effect on cortisol levels of prolonged exposure to extreme poverty, or to exposure to severe violence or privation? We don't yet know, but we now have good evidence it could be very bad.
Air pollution
The data on air pollution and cognition has been piling up for years. We have studies from China, information about prenatal exposure, and novel research based on brain games, as well as experimental data and research based on chess tournaments. The facts are in. Air pollution is bad for brains.
Most compellingly for our purposes, however, we have a recent high-quality study that combines satellite and administrative data to estimate the effect of exposure to airborne particulate matter on upward mobility. The authors of that study estimate that as much as a quarter of the Black-white earnings gap is explained by differences in pollution exposure.
This, as well, is no surprise — the poorest areas bear the brunt of air pollution.
Crucially, this is an area where improved regulation can help, because we have good evidence that it already has — a recent study argues convincingly that pollution reductions effected by the Clean Air Act are the primary cause of racial convergence in pollution exposure since 2000.
Concluding
Environmental impacts on cognition are stratified by socioeconomic status, and these cognitive effects propagate throughout the lives of those affected. The legacy of historical racism funnels Black Americans into lower-SES strata where their children's brains will be disproportionately injured by stress and pollutants, producing worse life outcomes as adults.
I think it can sometimes be unclear what people are talking about when they discuss structural racism. I hope this post makes it a little clearer. But I hope it also clarifies something important about life in America: structural racism is only one channel for a kind of environmentally reinforced inequality that affects not just Black or white people but Americans of every possible kind of background. What they share is a uniquely fragile social position and exposure to the worst imaginable kind of inequality: an injustice that constrains children's chances before they are even born.
"Though IQ is heritable, there seems to be no good reason to assume genetic differences in intelligence at the population level: the populations we're talking about here (African-Americans and white Americans) are so heterogeneous that polygenic traits are unlikely to differ meaningfully across groups. "
The average is what matters for average differences in IQ, not whether there's heterogeneity. African-Americans have an average of 80% recent Sub-Saharan admixture, whites about 0%. Differ in plenty of polygenic traits like skin color, eye color, hair color.
"Moreover, the IQ gap between Black and white Americans has decreased substantially in recent years"
https://ideasanddata.wordpress.com/2021/07/06/critiquing-charles-murray-on-changing-iq-gaps/
"Moreover, this also doesn't work for the same reason as above: we see variation in cognitive scores by socioeconomic status, but the groups at each SES level are extremely diverse. Hypothetical differences in IQ by race, ethnicity, or national origin don't make sense as an explanation here, since these characteristics are spread across different SES levels."
It makes perfect sense: IQs are distributed on a bell curve. Different racial means still means there's people of every race in every IQ category.
If IQ differences were caused by low-SES, then comparing within SES would eliminate racial gaps, which is not remotely the case: https://preview.redd.it/82mujmdjpp461.jpg?auto=webp&s=ababc75d2a5637b32426872888519559a157fa2a
Effects of adoption on IQ fade out by adulthood, when the effect of shared environment on IQ is basically 0.
Please read some HBDers who have long since debunked these environmental explanations.
"I think there are compelling .............................................................................. That is to say, adoption into a high-SES household raises cognitive test scores." Your source/evidence ???