Last Updated on November 23, 2022

In a previous post, I argued that the black-white cognitive ability gap is responsible for many of the important social disparities we find between blacks and whites, e.g. disparities regarding income, crime, education, occupational prestige, etc. Therefore, insofar as it is important that we resolve racial disparities in these social outcomes, it is also important that we resolve racial disparities in cognitive ability. In order to solve a problem, we should know the cause of the problem. Therefore, assuming it’s important to resolve racial disparities in these social outcomes, it is also important that we know the cause of the cognitive ability gap. That is the topic of this post.

In this post, I will consider some of the most commonly posited environmental explanations of the black-white cognitive ability gap (by “environmental”, I mean any factors other than genetic factors). First, I consider what I call the Test Bias Hypothesis, which posits that black-white differences in cognitive test scores are the result of biased tests rather than genuine differences in cognitive ability. Second, I consider the Schooling Hypothesis, which holds that the cognitive ability gap is the result of differences in schooling between blacks and whites. Third, I consider the Socioeconomic Hypothesis, which affirms that the cognitive ability gap is the result of the SES gap. Finally, I consider the Racism Hypothesis, which affirms that the cognitive ability gap is the result of societal racism against black people. I will argue that each of these hypotheses fail to adequately account for the cognitive ability gap.

The Test Bias Hypothesis

First, I will consider the hypothesis that the black-white differences in cognitive test scores are not a result of a genuine disparity in cognitive ability. According to this hypothesis, the black-white differences in cognitive ability are the result of an anti-black bias in the tests. I will refer to this hypothesis as the Test Bias Hypothesis.

What does it mean for a test to be biased against black test-takers? I will use the definition provided by Mackintosh (2011) (page 173):

This hypothesis posits that this bias makes the tests unreliable measures of the cognitive ability of black people.

Some points to motivate skepticism

I will begin my assessment of the Test Bias Hypothesis by noting two points that should motivate a reasonable level of skepticism against the hypothesis.

First, the vast majority of experts do not believe that cognitive ability tests are substantially biased against black people. I cite this first because it is the weakest evidence against the Test Bias Hypothesis since it relies more on an appeal to authority rather than on direct empirical evidence. Despite this, the opinions of experts are valuable. If the weight of expert opinion supports a particular claim, then this provides some contributory evidence in favor of the claim (even though it’s obviously not conclusive evidence). That said, multiple studies have shown that most experts in the field of intelligence deny that cognitive ability tests are substantially biased against black people:

• Gottfredson (1997) [archived] reports that “Intelligence tests are not culturally biased against American blacks or other native-born, English-speaking peoples in the U.S” (page 14). This was published in a very brief 3-page statement that outlines conclusions regarded as mainstream by over 50 experts in intelligence and allied fields.
• Reeve and Charles (2008) [archived] examined the opinions of 30 experts in the science of mental abilities about their views on cognitive abilities and cognitive ability testing. The study found a consensus among experts that general cognitive ability “general cognitive ability tests are fair” (76% of experts agree, 10% disagree), that “professionally developed [general cognitive tests] are not biased against members of … minority groups” (73% agree, 13% disagree) (Table 1). Participants in the survey were selected from individuals on the editorial board of the journal Intelligence, from all registered members of the International Society of Intelligence Researchers, and from persons who had published three or more articles in Intelligence over the last 3 years (page 683). Experts were selected from this group by filtering down to “only individuals with a doctorate degree, and having at least five career publications on the topic of intelligence or testing” (page 683). This study was a replication of Murphy, Cronin, and Tam (2003), which found similar results (although this study did not filter to experts specifically).
• Rindermann, Becker, and Coyle (2020) surveyed the opinions of over 100 experts in the field of intelligence about a variety of questions. One of the questions asked researchers about the prevalence of racial/ethnic content bias in cognitive ability tests. The four possible answers were (a) Insignificant amount of content bias, (b) Some content bias, (c) Moderate amount of content bias, and (d) Large amount of content bias. The percentage of experts who selected each answer were 43%, 34%, 20%, and 3%, respectively (Table 3). Thus, only 23% of experts affirmed that there was at least a moderate amount of test bias; the other 77% believed that there is not a moderate amount of such bias. This study was a replication of an earlier survey by Snyderman and Rothman (1987), which also reported “that experts believe there to be some racial bias in intelligence tests, but less than what would be considered a moderate amount” (page 141).

My second point to motivate skepticism of the Test Bias Hypothesis only targets those that affirm a particular form of the hypothesis, namely those who affirm that cognitive ability tests are culturally biased in favor of white people in particular. This particular hypothesis is undermined by the fact that many other non-white groups outscore blacks by large margins as well. Consider the following:

• East Asians (living in Asia or in America) tend to outscore blacks on cognitive ability tests by a substantial margin, typically surpassing whites as well (Neisser et al. 1996, page 92; Gottfredson 1997, page 14; Loehlin 2000, page 180; Rushton and Jensen 2005, page 240; Hunt 2011, page 421; Mackintosh 2011, page 164).
• Hispanics in America also tend to outscore blacks on cognitive ability tests, though to a far smaller degree than Asians and whites (Neisser et al. 1996, page 92; Rushton and Jensen 2005, page 236; Hunt 2011, page 411).

The Test Bias Hypothesis cannot explain why black people tend to score lower on cognitive ability tests than East Asians and Hispanics, unless one posits that cognitive ability tests are culturally biased in favor of Hispanics/Asians and against black Americans (despite the fact that such tests were originally developed for use in European/American culture). While this is not impossible, it is certainly less plausible than the claim that such tests are culturally biased in favor of white people.

Culture-free tests

I will freely admit that past intelligence tests were obviously culturally biased and thus poor measures of the cognitive ability of groups from particular cultural backgrounds. For example, a version of the Binet-Simon Intelligence Scale [archived] from 1916 asked some obviously culturally biased questions. One question asked “What’s the thing for you to do when you have broken something which belongs to someone else?” (page 104). According to the testing rubric, correct answers are those that suggest restitution and/or apology. Some satisfactory answers were “Buy a new one”, “Pay for it”, or “Give them something instead of it.” Incorrect answers were “Feel sorry”, “Be ashamed”, and “Pick it up.”  Another question asked “What’s the thing for you to do if a playmate hits you without meaning to do it?” (page 105). Some correct answers were “I would excuse him”, “I would not blame him for it”, and “Ask him to be more careful.” Some incorrect answers were “I would hit them back”, “Make him excuse himself”, “Make him say he’s sorry”, and “Would not play with him.”

Clearly, these questions are culturally biased. Differences in answers to these questions may reflect differences in cultural upbringing and norms (rather than differences in intelligence). Fortunately, plenty of modern intelligence tests are relatively “culture-free” in the sense that they involve questions that do not require adoption of any particular cultural norms to recognize the “correct” answer. Also, insofar as they do rely on cultural upbringing, this reliance is minimal. For example, Raven’s Progressive Matrices (RPM) measures pattern recognition by requiring participants to find the missing pattern that best fits within a set of given patterns. One might claim that even this question require some cultural background (e.g., perhaps the participant needs to have shown shapes or patterns in their life). But even if this is true, it cannot be doubted that they are relatively culture-free compared to, e.g., vocabulary questions. Thus, if the Test Bias Hypothesis is true, we should expect that the black-white cognitive ability gap will be smaller on relatively culture-free tests.

It turns out that the black-white cognitive ability gap is just as large even when measured with these relatively culture-free tests. Vincent (1991) reviewed several studies to investigate changes in the magnitude of the black-white IQ gap. He cites three studies with four different Raven’s Progressive Matrices tests administered to blacks and whites (Table 1). Large black-white score gaps were found across all four tests. The black-white gap on three of the tests was about one standard deviation (the gap on the other test was just half of a standard deviation). Mackintosh (2011) has also reported that “blacks perform just as badly” on “culture-fair” tests (page 170).

Finally, the black-white cognitive ability gap tends to be greater on more g-loaded tests (Neisser et al. 1996, page 93; Nijenhuis and Van den Hoek 2016, table 3), which are tests that tend to be more cognitively complex regardless of their content (Gottfredson 2002, page 28). For example, if someone is told a series of digits, repeating the digits in forward order (forward digit span) will have a far lower g-loading than repeating the digits in reverse order (reverse digit span), since the latter is more cognitive complex. It turns out that the black-white gap is worse on reverse digit span than on forward digit span (Rushton and Jensen 2005, page 331). The Test Bias Hypothesis cannot explain this finding. Is reverse digit span a more culturally loaded activity than forward digit span? Of course not. Rushton and Jensen (2005) [archived] raised a similar point when addressing accusations of test bias (page 267):

Other nongenetic hypotheses are that standard IQ tests are culturally biased because the test items are not equally familiar and motivating to all groups or that they only measure familiarity with middle-class language or culture. However, despite attempts to equate items for familiarity and culture-fairness, no “culture-fair” test has eliminated the mean group difference. American Blacks actually have higher average scores on culturally loaded tests than on culturally reduced tests, which is the opposite to what is found for some other groups such as Mexican Indians and East Asians. (The mean Black–White group differences are greatest on the g factor, regardless of the type of test from which g is extracted; see Section 4.) Moreover, the three-way pattern of mean Black–White–East Asian group differences occurs worldwide on culture-fair reaction time measures, which all children can do in less than 1 s (see Section 3).

Now, it should be noted that this piece by Rushton and Jensen argues for a “hereditarian” explanation of the black-white cognitive ability gap – i.e. an explanation that posits genetic differences as an explanation of the gap. My post here makes no claims either way on the role that genetic differences play in explaining the gap. I cite the Rushton paper only for support against the Test Bias Hypothesis. I make no claims (in this post) about the causes of those disparities. Mackintosh (2011) has also reported that blacks do just as poorly on culture-free tests without inferring that this indicates that genetic differences are responsible for the cognitive ability gap. He writes that “it is quite false to suppose that just because tests like the block design (or Cattell’s tests or Raven’s Matrices) are non-verbal and embody no superficially middle-class values, they are providing a more direct measure of ‘innate intelligence’ than do verbal tests” (page 170).

Elementary cognitive tasks

Perhaps the least culturally loaded tests of cognitive ability are elementary cognitive tasks (ECTs). ECTs are tasks that measure a participant’s response time as they attempt to solve a very simple problem. Even though the tasks assess a participant’s low level cognitive processes, they are believed to be relevant to general cognitive ability. For example, Carroll (1991) found that “the data strongly suggest that an important aspect of g is speed and efficiency in information processing” (page 435) after analyzing a large number of psychometric tests. Neisser et al. (1996) reports that “Many recent studies show that the speeds with which people perform very simple perceptual and cognitive tasks are correlated with psychometric intelligence” (page 83). More recently, Ravenzwaaij et al. (2011) [archived] reported that “overwhelming empirical evidence” has been gathered in support of Sir Francis Galton’s idea that “general mental ability’ manifests itself by the speed with which people perform elementary cognitive tasks” (page 381). Scholars disagree about why ECTs are associated with cognitive ability. One theory proposed by Rushton and Jensen (2005) [archived] posits that reaction time, one form of ECT, may be related to intelligence because “reaction time measures the neurophysiological efficiency of the brain’s capacity to process information accurately—the same ability measured by intelligence tests” (page 244).

Common low-level processes studied include reaction time (RT) and inspection time (IT). A common RT task is the “odd man out” task, which Neisser et al. (1996) describes as a procedure where “three of the eight lights are illuminated on each trial. Two of these are relatively close to each other while the third is more distant; the subject must press the button corresponding to the more isolated stimulus” (page 83). A common IT test involves showing two lines to a participant for a brief period of time and determining the minimum amount of time needed for the participant to accurately determine which line is longer (Neisser et al. 1996, page 83; Ravenzwaaij et al. 2011, page 388). Note that in the IT task, unlike the RT task, participants are not required to produce an answer as quickly as possible. Performance on these very simple tasks have shown moderate to large associations with cognitive ability:

• Mean inspection time is negatively correlated with cognitive ability. Neisser et al. (1996) report that “Inspection times defined in this way are consistently correlated with measures of psychometric intelligence” (page 84). They cite Kranzler and Jensen (1989), a meta-analysis of a number of studies showing a correlation of –0.54 between IT and IQ among adults after correcting for artifactual sources of error. Their work was replicated in a later meta-analysis by Grudnik and Kranzler (2001), which found a similar correlation (r=–0.51). More recently, Sheppard and Vernon (2008) meta-analyzed 46 reported correlations between general intelligence and mean IT, finding a mean correlation of –0.36 (Table 5).
• Mean reaction time is negatively correlated with cognitive ability. In a review of mental chronometry research, Ravenzwaaij et al. (2011) report that the correlation between mean RT and g tends to range between –0.20 to –0.30 (page 389). Interestingly, they also note that slow RTs are more indicative of low g than fast RTs are indicative of high g (page 384). More recently, Sheppard and Vernon (2008) meta-analyzed hundreds of reported correlations between general intelligence and mean RT across different tasks, finding mean correlations that ranged –0.22 and –0.4 depending on the type of task (Table 1).
• Reaction time variability is negatively associated with cognitive ability. Doebler and Scheffler (2016) meta-analyzed 24 studies and found that that the correlation between RT variability and general intelligence (g) ranges between –0.18 and –0.28. Some have even reported that g is actually more strongly correlated with RT variability than mean RT (Ravenzwaaij et al. 2011, page 385).

Studies show that blacks tend to perform worse than whites across ECTs. That is, blacks tend to have higher mean RTs, higher mean ITs, higher RT variability, and higher IT variability. These are all associated with lower cognitive ability. Consider the following studies:

• Pesta and Poznanski (2007) measured the IQ scores, RTs, and ITs from 139 white and 40 black undergraduates from a large urban university. They found that the black students had significantly lower IQ scores (Table 1), had higher mean RTs and ITs (Table 2), and had higher RT and IT variability (Table 2). The average gap ranged from 0.4 to 0.8 standard deviations, depending on the ECT measured (Table 2). Impressively, controlling for ECTs reduced the relationship between race and IQ by 49%, rendering the relationship statistically insignificant (page 327).
• Sheppard and Vernon (2008) meta-analyzed 14 studies that investigated black-white differences in mental speed (Table 8). Of the 6 studies that investigated differences in RT, 5 found that blacks had significantly higher mean RT (mean effect size = 0.47). Of the 8 studies that investigated differences in general speed of processing, 7 found that blacks had slower speeds (mean effect size = 0.18).

One explanation for the slower times of black people is that they are less motivated than whites or Asians to respond quickly. Rushton and Jensen (2005) have criticized this explanation by noting that blacks have higher reaction times even when they have shorter movement times (page 245):

Jensen (1993) and Jensen and Whang (1994) examined the time taken by over 400 schoolchildren ages 9 to 12 years old in California to retrieve overlearned addition, subtraction, or multiplication of single digit numbers (from 1 to 9) from long-term memory. All of the children had perfect scores on paper-and-pencil tests of this knowledge, which was then reassessed using the Math Verification Test. The response times significantly correlated (negatively) with Raven Matrices scores, whereas movement times have a near-zero correlation. The average reaction times for the three racial groups differ significantly (see Figure 2). They cannot be explained by the groups’ differences in motivation because the East Asian children averaged a shorter response time but a longer movement time than did the Black children.

Again, Rushton and Jensen argues for a “hereditarian” explanation of the black-white cognitive ability gap, but I am making no such claims in this post. I merely cite their arguments and data as evidence against the Test Bias Hypothesis.

Predictive validity

In a review of intelligence research by experts in the field, Neisser et al. (1996) [archived] also criticized the “test bias” explanation of the gap, noting that intelligence tests do do not have a predictive bias against blacks (page 93):

From an educational point of view, the chief function of mental tests is as predictors (Section 2). Intelligence tests predict school performance fairly well, at least in American schools as they are now constituted. Similarly, achievement tests are fairly good predictors of performance in college and postgraduate settings. Considered in this light, the relevant question is whether the tests have a “predictive bias” against Blacks. Such a bias would exist if African American performance on the criterion variables (school achievement, college GPA, etc.) were systematically higher than the same subjects’ test scores would predict. This is not the case. The actual regression lines (which show the mean criterion performance for individuals who got various scores on the predictor) for Blacks do not lie above those for Whites; there is even a slight tendency in the other direction (Jensen, 1980; Reynolds & Brown, 1984). Considered as predictors of future performance, the tests do not seem to be biased against African Americans.

Similar findings were reported in Gottfredson (1997) [archived], which is a very brief statement outlining conclusions regarded as mainstream among over 50 experts in intelligence and allied fields. The statements asserts the following regarding the predictive validity of IQ tests for black people (page 14):

Intelligence tests are not culturally biased against American blacks or other native-born, English-speaking peoples in the U.S. Rather, IQ scores predict equally accurately for all such Americans, regardless of race and social class. Individuals who do not understand English well can be given either a nonverbal test or one in their native language.

In his book Human Intelligence, Hunt (2011) writes that “In sum, the tests are accurate predictors of achievement for the three major racial/ethnic groups in the United States – Whites, African Americans, and Hispanics” (page 424). In his book IQ and Human Intelligence, Mackintosh (2011) has reported that “both in Britain and the USA, that IQ tests predict educational attainment just about as well in ethnic minorities as in the white majority” (page 174) and “there is no evidence that test scores show lower correlations with various indices of job performance in blacks than in whites, and no evidence at all of any systematic underestimate of the performance of blacks” (page 176).

In fact, the predictive validity of cognitive ability tests for black people is high enough controlling for scores on these tests reduces many of the large, stubborn social disparities between blacks and whites. I explained this more in a previous post, but here are a few examples:

• The Armed Forces Qualification Test (AFQT) is a good proxy for IQ (AFQT scores and IQ scores correlate with r=0.8). After controlling for age and AFQT scores, Johnson and Neal (1998) [archived] found that AFQT disparities explain about two-thirds of the wage gap for young men and all of the gap for young women. The wage gap reduces from 24% to 9% for men after controlling for AFQT. For young women, Black women go from a 17% disadvantage to a 5% advantage after controlling for AFQT scores (page 4).
• Pew Research released data [archived] showing that about 75% of Whites born into the bottom quintile of income transition out of that quintile, whereas only about 56% of Blacks do the same (figures 3a and 3b). However, among those with a median AFQT score, there is almost no difference in the likelihood of transitioning out of the bottom quintile – 81% for Whites and 78% for Blacks (figure 13). So, the relative lack of economic mobility for Blacks is mostly the result of the relative lack of cognitive ability.
• Nyborg and Jensen (2001) [archived] have also found that racial differences in occupational status and income are eliminated after controlling for IQ. In fact, blacks actually exceed whites in these outcomes after controlling for IQ. Specifically, for individuals above the 50th percentile in IQ, blacks earn more income than whites. More impressively, blacks have higher occupational statuses than whites at every IQ percentile.

One might respond that these social outcomes are also biased against blacks, therefore the fact that controlling for cognitive ability test scores eliminates disparities in these social outcomes does not show that these tests are not biased (Mackintosh 2011 has also entertained the possibility (page 176)). But this response is implausible. What explains the remarkable correlation between the test score gap and income/occupation gap? The most reasonable explanation is that the test scores gap and the income/occupation gap have the same cause (the alternative requires positing that the correlation between the test score gap and income/occupation gap is just a remarkable coincidence). So what’s the cause of the test score gap and income/occupation gap? Again, the most reasonable explanation is to say that these gaps are both the result of a genuine black-white gap in cognitive ability. If test scores did not reliably measure the cognitive ability of black people, then how could the test score gap correlate with gaps in income and occupation? If test scores did not reliably measure the cognitive ability of black people, how does the market get access to information about test scores of black people? Also, how and why does this information cause differential treatment of blacks and whites by the market (in terms of income and occupation)? Why would the market choose to treat blacks similarly to whites (in terms of their income and occupation) when blacks and have similar test scores? The only non-absurd response to these questions is to posit that agents in the market are motivated to treat workers differently based on their cognitive ability, and that cognitive ability is reliably measured by cognitive ability tests for both blacks and whites.

In summary, the Test Bias Hypothesis fails to explain the black-white cognitive ability gap. The hypothesis fails because these gaps are found on culture-free tests, because the gap is larger on more g-loaded tests, and because cognitive ability tests predict success for blacks equally as well as for whites. Cognitive ability tests are not biased. As Hunt (2011) concluded in his discussion of possible test bias, “the differences in test scores across racial/ethnic groups almost certainly reflect a real difference in the distribution of cognitive skills across racial/ethnic lines” (page 425). The question now is what explains these genuine differences in cognitive ability. That’s where the remaining hypotheses – the Schooling Hypothesis, the Socioeconomic Hypothesis, and the Racism Hypothesis – come in.

The Schooling Hypothesis

Another common explanation of the black-white cognitive ability gap posits that these gaps are the result of black-white differences in schooling. I will refer to this hypothesis as the Schooling Hypothesis.

School quality and funding

The Schooling Hypothesis must predict that there are differences in school quality and that these differences are fairly large (because the cognitive ability gap is substantial). Unfortunately for the Schooling Hypothesis, these predictions do not seem to be true.

From the introductory chapter of The Black-White Test Score Gap, Jencks and Phillips (1998) [archived] note (page 9) some problems with appealing to schooling differences to explain the cognitive ability gap:

Most southern schools desegregated in the early 1970s, and southern black nine-year-olds’ reading scores seem to have risen as a result. Even today, black third-graders in predominantly white schools read better than initially similar blacks who have attended predominantly black schools. But large racial differences in reading skills persist even in desegregated schools, and a school’s racial mix does not seem to have much effect on changes in reading scores after sixth grade or on math scores at any age.

Despite glaring economic inequalities between a few rich suburbs and nearby central cities, the average black child and the average white child now live in school districts that spend almost exactly the same amount per pupil. Black and white schools also have the same average number of teachers per pupil, the same pay scales, and teachers with almost the same amount of formal education and teaching experience.

The authors do note that teachers in black schools tend to have lower test scores than teachers in white schools, but more data is needed to show that this can explain a significant portion of racial achievement disparities. 

More recent studies have corroborated the finding that blacks and whites attend school districts with similar per-pupil spending:

• Rueben and Murray (2008) [archived] find that “differences in spending per pupil in districts serving nonwhite and white students are very small”. Specifically, they report that “In 1972, the ratio of nonwhite to white spending was .98; this trend had reversed by 1982, as spending per pupil for nonwhite students was slightly higher than for white students in most states and in the United States as a whole and has been for the past 20 years” (page 5).
• Richwine (2011) [archived] found that black and white children attend school districts that receive similar amounts of funding per pupil. They report that “Nationwide, raw per-pupil spending is similar across racial and ethnic groups. The small differences that do exist favor non-white students” (page 4).

Some caution must be taken from the above findings. It is possible that, despite attending school districts with similar funding as whites, black children tend to attend schools that receive less funding within their specific district. This is possible, but more data is needed to show that this can explain a significant portion of racial achievement disparities.

Rueben and Murray (2008) [archived] also studied other potential differences between black and non-black schools that may be relevant to differences in cognitive ability. They compared characteristics of newly hired teachers and availability of advanced placement:

• Newly hired teachers for predominantly black schools had similar base-year salaries, had similar years of experience, and were equally as likely to have earned a bachelor’s degree or higher (Table 5). In fact, black schools had a higher percentage of teachers with a master’s degree or higher. Non-black schools had a higher percent of teachers that were certified in primary teaching field, but the difference was not large (the percent of teachers in schools with 90+% black students was 86.8%; the percent for teachers in schools with 0-10% black students was 93.8%). Teachers at predominantly black were substantially more likely to have lower levels of job satisfaction, but this may be an effect of black underachievement (and misbehavior) rather than a cause of it (see here for more detail on the early emergence cognitive and behavioral disparities young between black and white children).
• There were no large differences between the access to advanced placement courses between majority-black and majority-non-black schools. They find that “in 1972, students in schools where at least 90 percent of the students are black were 30 percent less likely to have the opportunity to take AP courses than students in schools where less than 10 percent of the students were black. By 1992, however, these schools had made large strides in course offerings, and black students were about as likely to have AP courses offered at their schools as were white students” (page 11).

The gaps precede schooling

The Schooling Hypothesis also predicts that the magnitude of the cognitive ability gap should grow as children enter school. Otherwise, exposure to schooling would have no impact on the size of the gap, which suggests that school plays no role in causing the gap. Unfortunately, for this hypothesis, this prediction is demonstrably false. The cognitive ability gap appears before children even attend schools, and the size of the gaps (according to tests that measure g) do not grow substantially as children attend school.

• Gottfredson (1997) [archived] reports that “Racial-ethnic differences in IQ bell curves are essentially the same when youngsters leave high school as when they enter first grade” (page 15). This was a claim published in a very brief 3-page statement that outlines conclusions regarded as mainstream among over 50 experts in intelligence and allied fields.
• In the introductory chapter of the bookThe Black-White Test Score Gap, Jencks and Phillips (1998) [archived] report that “African Americans currently score lower than European Americans on vocabulary, reading, and mathematics tests, as well as on tests that claim to measure scholastic aptitude and intelligence. This gap appears before children enter kindergarten (figure 1-1), and it persists into adulthood. It has narrowed since 1970, but the typical American black still scores below 75 percent of American whites on most standardized tests. On some tests the typical American black scores below more than 85 percent of whites” (page 1).
• Brooks–Gunn et al. (2003) examined black-white test score gaps among young children from two data sets. One of the samples involved 315 premature, low birth weight 3 and 5 year olds from the Infant Health and Development Program (IHDP). The cognitive ability of the children were measured using the Stanford–Binet Intelligence Scale at age 3 and the Wechsler Preschool and Primary Scale of Intelligence (WPPSI) at age 5 (page 242). The standard deviations of scores for the two tests were 16 and 15 points, respectively. The black-white gap was about 20 points on the Stanford-Binet and about 15-18 points on the WPPSI (Table 1). This corresponds to gaps of about one standard deviation (page 244).
• Gottfredson (2004) [archived] cites a long list of studies measuring mean IQ differences between blacks and whites (Table 1). Some of the studies report the gaps for children aged 6 and younger. The gaps reported for children in this age group range from 0.6 to 1.23 standard deviations, with a mean gap of about one standard deviation.
• Farkas and Beron (2004) [archived] investigated oral vocabulary (PPVT) scores for black and white children by age. They found (page 477) that “Beginning with the earliest observation at 36 months of age, Whites average significantly higher scores than African-Americans. This pattern is consistent over the full age span, with the White lead remaining significant through 13 years of age. To see how very substantial this vocabulary gap is, note that Whites cross the 40-word level at approximately 50 months of age, whereas African-Americans do not reach this level until approximately 63 months, which puts them 13 months, or more than one year, behind in vocabulary development.” The PPVT is a commonly used measure of intelligence (e.g., see meta-analysis by Protzko 2015 [archived] [table 1]).
• Cottrell, Newman, & Roisman (2015) [archived] examined cognitive ability of children of 1,364 families who participated in the National Institute of Child Health and Human Development (NICHD) Study of Early Child Care and Youth Development (SECCYD). General cognitive ability/knowledge (g) was measured using the math, vocabulary, and reading ability facets of the Woodcock-Johnson Psycho-Educational Battery-Revised (WJ-R). Researchers find that “Black-White gaps in cognitive test scores are large and pervasive, and are already established by 54 months of age” (page 11). They further report that “between 54 months and 15 years of age, this gap did not significantly increase over time” (page 11). The black-white gap in g ranged from around 1.2 and 1.4 standard deviations during this time period (page 11).

See my post here for more on the early emergence of cognitive disparities between blacks and whites. The studies above all show that cognitive disparities appear between black and white children before they reach school with magnitudes approximately equal to the magnitude of the gap found among adults (about one standard deviation). Of the longitudinal studies cited (e.g., Cottrell 2015 and Farkas 2004), they find that the magnitude of the gaps are stable across time. Because the black-white disparities in cognitive ability are in full effect before they reach school, it seems that the the Schooling Hypothesis provides an inadequate explanation of the cognitive ability gap. The true cause of the gap must be factors that emerge before children reach school age.

The Socioeconomic Hypothesis

Another common hypothesis posits that black-white disparities in cognitive ability are the result of black-white differences in socioeconomic status (SES). In other words, this hypothesis posits that the cognitive ability gap is the result of the SES gap. I will refer to this hypothesis as the Socioeconomic Hypothesis.

Measuring SES

Before assessing the evidence for this hypothesis, we must first settle on the measure of socioeconomic status to be used here. The American Psychological Association reports [archived] that SES “is often measured as a combination of education, income and occupation.” A number of other studies have agreed that this tripartite model is the traditional measure of SES (Adler and Newman 2002, Bradley and Corwyn 2002, Winkleby et al. 1992). SES has also been measured using these three components in a number of meta-analyses studying the relationship between SES and a number of social outcomes, including future SES (Strenze 2007, page 406), self-esteem (Twenge and Campbell 2002), well-being (Pinquart and Sörensen 2000), and countless others.

It seems fair to say that SES is traditionally defined as a combination of education, income, and occupation. In fact, the following meta-analyses on parental SES and academic performance also report that these three components are typically used to measure parental SES:

• In a meta-analysis on the relationship between SES and academic achievement, Sirin (2005) [archived] reports that “there seems to be an agreement on Duncan, Featherman, and Duncan’s (1972) definition of the tripartite nature of SES that incorporates parental income, parental education, and parental occupation as the three main indicators of SES” (page 418).
• In another meta-analysis on the relationship between peer socioeconomic status on student achievement, Ewijka and Sleegers (2010) [archived] also states that “There also seems to be agreement on a three-componential view of SES which states that SES can be indicated by either parental education, parental occupation, or parental income” (page 138).
• In another meta-analysis on the relationship between socioeconomic status and college admissions tests and academic performance, Sackett et al. (2009) reports “There is no uniform agreement on measurement of this construct, although most studies have focused on some combination of three measures: parental education, parental income, and parental occupational status” (page 3). 

For the purposes of this post, I will use the traditional measure of SES defined above. Therefore, since this section is concerned with the effects of parental SES, I will test the Socioeconomic Hypothesis by reviewing evidence concerning the alleged relationship between black-white differences in cognitive ability and black-white differences in parental income, education, and occupational status.

Controlling for SES

The Socioeconomic Hypothesis predicts that the cognitive ability gap will be eliminated (or at least substantially reduced) after statistically controlling for parental SES. The preponderance of data indicates that this prediction has been falsified. Statistically controlling for parental SES does not reduce a substantial proportion of the cognitive ability gap.

From the introductory chapter of The Black-White Test Score Gap, Jencks and Phillips (1998) [archived] report that “a two-year reduction of the black-white education gap among mothers would only reduce the IQ gap by about a point for children” (page 22).They further state that “eliminating black-white income differences would reduce the IQ gap by about a point” (page 23).

The findings from Jencks and Phillips regarding parental education are corroborated by recent data [archived] on NAEP test scores among 12th-grade students in 2013 (for the source: if using the archived webpage, click “Focus: racial/ethnic groups” on the left sidebar). Controlling for parental education does not succeed in explaining the test score gap between blacks and whites. In fact, the average NAEP mathematics score of black students with parents who graduated from college was equal to the average score of white students with parents who did not finish high school (both groups score 138 points). The exact same pattern is found for reading scores (both groups score 274 points).

The findings from Jencks and Phillips regarding parental income are corroborated by SAT scores disaggregated by race and income. Black-white differences in income do not explain a significant portion of black-white differences in SAT scores. Consider the following data [archived] on the racial gap in SAT scores in 2005:

Whites from families with incomes of less than $10,000 had a mean SAT score of 993. This is 129 points higher than the national mean for all blacks.

Whites from families with incomes below $10,000 had a mean SAT test score that was 61 points higher than blacks whose families had incomes of between $80,000 and $100,000.

Blacks from families with incomes of more than $100,000 had a mean SAT score that was 85 points below the mean score for whites from all income levels, 139 points below the mean score of whites from families at the same income level, and 10 points below the average score of white students from families whose income was less than $10,000.

The same pattern in SAT scores is replicated by Dixon-Roman et al. (2013) [archived] (Table 2) and a 2008 report [archived] (page 11) by The Journal of Blacks in Higher Education. Note that SAT tests are highly g-loaded (Frey and Detterman (2004) [archived], Koenig et al. (2008) [archived]) so they can be considered to be a reliable proxy for IQ (if not IQ tests themselves) and good measure of cognitive ability.

Rushton and Jensen (2005) [archived] reported that controlling for SES only reduces the mean Black–White group difference in IQ by about a third, around 5 IQ points (page 267). In a review of intelligence research by experts in the field, Neisser et al. (1996) [archived] also criticized SES-based explanations of the gap (page 94).

Gaps at different SES levels

Another way to test the Socioeconomic Hypothesis involves measuring the black-white cognitive ability gap among children from similar SES levels. Unfortunately for this hypothesis, the vast majority of the gap is found at every SES level.

For example, Dixon-Roman et al. (2013) [archived] show that black-white SAT gaps – particularly math gaps – are remarkably stable across all ranges of family income (Table 2). Among families with incomes between $10,000 and $15,000, the black-white SAT math gap was 83 points. Among families with incomes between $40,000 and $50,000, the gap was 79 points. Among families with incomes exceeding $100,00, the gap was 78 points. 

Even worse, some data suggests that cognitive gaps are actually larger among high-SES families:

• Murray and Herrnstein (1994) used data from the National Longitudinal Survey of Youth (NLSY) to show that the IQ gap persists across all SES levels, with the gap the largest at the highest SES levels (page 288). “Parental SES” is measured based on “information about the education, occupations, and income of the parents of NLSY youths” (page 131).
• Gottfredson (2003) [archived] report a finding from Jensen that the black-white IQ is largest among children of families at the highest socioeconomic level (page 18). As we consider children from parents of higher occupational levels and more years of education, the black-white IQ gap tends to increase (Table 2).
• Data [archived] on NAEP test scores among 12th-grade students in 2013 shows that, after controlling for parental education, the gap in academic achievement increases among children with more educated parents. Among children with parents who did not finish high school, the black-white mathematics test score gap is 16 points. Among children with parents who finished high school, the gap is 24 points. Among children with parents who graduated from college, the gap is 32 points. The same pattern is found for the reading test score gaps. A 1999 report [archived] by the College Board had similar findings (Table 2).

The preponderance of evidence suggests that black-white differences in parental income, education, and occupational status fail to explain black-white differences in cognitive ability. The data suggests that the majority of the gap remains after controlling for these SES factors. In fact, some data suggests that the cognitive ability gap is larger among children from high-SES parents. All of this evidence suggests that the Socioeconomic Hypothesis provides an inadequate explanation of the cognitive ability gap.

Confounding

Some of the above studies showed that controlling for SES may reduce the cognitive ability gap by as much as one-third. Some might interpret this as proof that SES differences are responsible for one-third of the cognitive ability gap. This would be an incorrect interpretation. The fact that the cognitive ability gap reduces by one-third only establishes a statistical correlation between racial differences in SES and racial differences in cognitive ability. But, as everyone should know, correlation does not imply causation. In order to demonstrate causation, it is also necessary that we show that the correlation is not spurious. That is, it is also necessary to rule out alternative explanations for the correlation (e.g., perhaps there is a third variable causing racial differences in parental SES and racial differences in cognitive ability). None of the data considered above attempted to rule out alternative explanations, so the reduction in the gap after controlling for SES cannot be assumed to be the result of a causal relationship between the two. Neisser et al. (1996) [archived] (page 94) raised a similar point in their criticism of SES-based explanations of the gap (page 94):

Several considerations suggest that this [SES explanations of the black-white IQ gap] cannot be the whole explanation. For one thing, the Black/White differential in test scores is not eliminated when groups or individuals are matched for SES (Loehlin et al., 1975). Moreover, the data reviewed in Section 4 suggest that–if we exclude extreme conditions–nutrition and other biological factors that may vary with SES account for relatively little of the variance in such scores. Finally, the (relatively weak) relationship between test scores and income is much more complex than a simple SES hypothesis would suggest. The living conditions of children result in part from the accomplishments of their parents: If the skills measured by psychometric tests actually matter for those accomplishments, intelligence is affecting SES rather than the other way around. We do not know the magnitude of these various effects in various populations, but it is clear that no model in which “SES” directly determines “IQ” will do.

A commonly proposed confounder is genetic differences. “Hereditarians” (those who believe that genetic differences are responsible for a large portion of the cognitive ability gap) state that controlling for SES somewhat reduces the cognitive ability gap because controlling for SES implicitly controls for genes (because one’s SES attainment is related to their intelligence, which is itself related to their genes). Thus, the hereditarians argue, the reduction of the gap that occurs after controlling for SES is because the genes associated with SES are causal, but SES itself need not be directly causal. This point has been raised by Rushton and Jensen (2005), perhaps the two most prominent hereditarians (page 267):

The most frequently stated culture-only hypothesis is that the mean IQ differences are due to SES. In fact, controlling for SES only reduces the mean Black–White group difference in IQ by about a third, around 5 IQ points. The genetic perspective does not regard this control for SES as being entirely environmental. It holds that the parents’ socioeconomic level in part reflects their genetic differences in intelligence.

The possibility of genetic confounding is especially important regarding differences in cognitive ability because there is ample data showing that the association between family SES and children’s cognitive development is substantially confounded by genetic factors (Trzaskowski et al. 2014, Plomin et al. 2017). In other words, much of the association between parental SES and children’s cognitive development is not due to a direct causal relationship. Rather, much of the association is due to a third factor – i.e. the shared genetics of the parent and offspring – which causes both parental SES and children’s cognitive development.

This point has even been acknowledged by those less committed to the hereditarian hypothesis. For example, in a review of literature concerning the role of family SES on the black-white score gap among young children, Magnuson and Duncan (2006) [archived] note the need to consider genetic confounding when analyzing the association between racial differences in SES and racial differences in cognitive ability (page 386):

Simply documenting that SES accounts for about .4–.5 of standard deviation of the black–white test score gap does not prove that differences in SES have caused differences in children’s test scores. For example, although Fryer and Levitt (2004) are able to account for virtually all of the racial and ethnic gaps in kindergarten achievement using measures of family background, they lack any measure of genetic endowments and are thus unable to discount the possibility that what appear to be family socioeconomic effects are really caused by other family characteristics.

Duncan and Magnuson (2005) [archived] have reiterated this same point elsewhere. They argue that studies which find that the test score gap is reduced after controlling for SES can, at best, demonstrate an upper-bound for the effect of the variables controlled (page 47):

On average, when black and Hispanic children begin school, their academic skills lag behind those of whites. Accounting studies find that differences in socioeconomic status explain about half a standard deviation of the initial achievement gaps. But because none of the accounting studies is able to adjust for a full set of genetic and other confounding causes of achievement, we regard them as providing upper-bound estimates of the role of family socioeconomic status.

This point has even been appreciated by environmentalists. Nisbett et al. (2012) (page 7) admit that:

It should be acknowledged, however, that at present there is no way of knowing how much of the IQ advantage for children with excellent environments is due to the environments per se and how much is due to the genes that parents creating those environments pass along to their children. In addition, some of the IQ advantage of children living in superior environments may be due to the superior genetic endowment of the child producing a phenotype that rewards the parents for creating excellent environments for intellectual development (Braungart, Plomin, DeFries, & David, 1992; Coon, Fulker, DeFries, & Plomin, 1990; Plomin, Loehlin, & DeFries, 1985). To the extent that such processes play a role, the IQ advantage of children in superior environments might be due to their own superior genes rather than to the superior environments themselves.

In his book Human Intelligence, Hunt (2011) has raised similar objections with socioeconomic explanations for racial disparities in intelligence. However, he has gone further to claim that SES cannot be causal (page 427):

Nevertheless, there are three reasons for not accepting SES as the sole explanation for racial/ethnic disparities in intelligence. The first is statistical: SES measurements do not fully account for the differences in intelligence measures between groups. An SES argument also fails to account for the fact that the differences between groups on cognitive measures tend to be large for “culturally reduced” tests, such as the RPM tests. Second, SES is itself confounded with intelligence. We then get a chicken- and-egg phenomenon. For children, parental SES is confounded with parental intelligence, and since intelligence is heritable to a significant degree, it is not clear whether an effect of parental SES on children’s intelligence is due to social or biological inheritance. Finally, and to me most compelling, socioeconomic status is a statistical abstraction. As such, it cannot cause anything. Intellectual development depends upon physical and social variables within the environment that are correlated with SES.

This further underscores the inadequacy of SES based explanations. Not only does controlling for SES reduce the gap by only one-third, even that one-third reduction may not be caused by controlling for SES. It may be the result of confounding with genetics. Or it may be the result of confounding with other factors correlated with the parent’s intelligence and/or SES (such as, e.g., their parenting practices, how often they speak to the child, etc.). 

The Racism Hypothesis

Another common explanation of the black-white cognitive ability gap posits that these gaps are the result of some sort of societal bias against blacks. Such phenomena have commonly been referred to as “systemic racism” or “institutional racism”. I will refer to this explanation as the Racism Hypothesis.

Other minorities

The first point against the Racism Hypothesis is the fact that many other non-white groups outscore blacks by large margins as well. This is the weakest point, but it is still some evidence against the hypothesis. Consider the following:

• East Asians (living in Asia or in America) tend to outscore blacks on cognitive ability tests by a substantial margin, typically surpassing whites as well (Neisser et al. 1996, page 92; Gottfredson 1997, page 14; Loehlin 2000, page 180; Rushton and Jensen 2005, page 240; Hunt 2011, page 421; Mackintosh 2011, page 164).
• Hispanics in America also tend to outscore blacks on cognitive ability tests, though to a far smaller degree than Asians and whites (Neisser et al. 1996, page 92; Rushton and Jensen 2005, page 236; Hunt 2011, page 411).

The Racism Hypothesis cannot explain why black Americans tend to score lower on cognitive ability tests than East Asian and Hispanic Americans, unless one posits that societal racism has a negligible effect on East Asians but a tremendous effect on black Americans. Further, the Racism Hypothesis cannot explain why Hispanic Americans outscore black Americans, unless one posits that societal racism is significantly stronger against black Americans than Hispanic Americans. While this is not impossible, it is certainly less plausible than the claim that societal racism explains the gap between black and white Americans.

What is the mechanism?

Another problem with the Racism Hypothesis is the fact that black children raised by white parents tend to have significantly higher levels of cognitive ability than black children raised by black parents. In fact, the cognitive ability gap between black children raised by black vs white parents is almost as large as the cognitive ability gap between black and white children.

Willerman et al (1974) [archived] measured the IQ of 129 biracial (black-white) 4-year-olds. They found that biracial children with white mothers had a mean IQ that was 9 points higher than biracial children with black mothers (102 vs 93, Table V). Interestingly, the IQ gap between biracial children raised by black vs white mothers does not seem to be explained by SES differences. Compared to the black-mother/white-father pairs, the white-mother/black-father pairs had an SES index that was only 0.1 SDs higher, they had lower incomes, they had only 0.84 more years of maternal education, and only 0.87 more years of paternal education (page 86). Furthermore, the IQ gap persists even after controlling for marital status. Among married mothers, biracial children with white mothers had a mean IQ of 105 whereas those with black mothers had a mean IQ of 96 (IQ gap = 9 points). Among single mothers, biracial children with white mothers had a mean IQ of 99 whereas those with black mothers had a mean IQ of 88 (IQ gap = 11 points).

Moore (1986) [archived] examined 23 black/biracial children adopted by middle-class white families and 23 age-matched black/biracial children adopted by middle-class black families. When their IQ scores were measured (Table 2 [archived]), the black children adopted by black families had an average IQ of 104, while the black children adopted by white families had an average IQ of 117. The children were aged between 7 and 10 years at the time of testing. Thus, the IQ gap between black children raised by middle-class white vs black families was 13 IQ points, which is nearly as large as the IQ gap between blacks and whites.

These findings are undermined the Racism Hypothesis because they suggest that the cognitive ability gap between black and white children is caused by factors that correlate with parental race other than any factors that directly target the race of the children (as is posited by the Racism Hypothesis). These results are consistent with the results of Arcidiacono et al. (2015), which concludes that “discrimination based on skin color is no longer the first-order concern. We argue instead that disparate outcomes must be operating through characteristics related to maternal race.” These findings also undermine the Socioeconomic Hypothesis because the white families in these studies had similar SES levels as the black families. For more detail on studies of the cognitive ability of black children measured by white families, see my post here.

Controlling for parental cognitive ability and parenting practices

One final problem with the Racism Hypothesis is the fact that the cognitive ability gap is eliminated after controlling for parental cognitive ability and parenting practices.

Mandara, Varner, Greene, and Richman (2009) [archived] examined intergenerational family predictors of the black–white achievement gap among 4,406 adolescents from the National Longitudinal Survey of Youth (NLSY). This study only considered test assessments between the ages of 10-11 and 13-14. Academic achievement was measured with the reading recognition, reading comprehension, and mathematical reasoning subtests of the Peabody Individual Achievement Test (PIAT) (page 871). The black-white test score gap was completely eliminated after statistically controlling for intergenerational family factors (page 874). In particular, the study found that “parenting explained most of the achievement gap” (page 877).

Cottrell, Newman, and Roisman (2015) [archived] examined cognitive ability of children of 1,364 families who participated in the National Institute of Child Health and Human Development (NICHD) Study of Early Child Care and Youth Development (SECCYD). Tests were administered to children at 54 months of age, first grade, third grade, fifth grade, and age 15 (page 7). The math, vocabulary, and reading ability facets of the Woodcock-Johnson Psycho-Educational Battery-Revised (WJ-R) were used to measure general cognitive ability/knowledge (g). Large black-white cognitive test score gaps are found at every point of assessment with gap sizes ranging between 1.24 and 1.39 standard deviations (page 8). The relationship between race and cognitive ability was no longer statistically significant after controlling for a variety of explanatory variables (page 24). The explanatory variables that were statistically significant in explaining the gap (in decreasing order of effect size) were maternal verbal ability/knowledge (uniquely explained 33.5% of the gap), maternal sensitivity (25.2%), learning materials in the home (6.8%), safe physical environment (5.4%), and birth order (5.1%) (table B1 and page 24). Maternal acceptance, birth weight, income, and maternal education were included but were not statistically significant.

These findings undermine the Racism Hypothesis. If societal racism was the cause of the cognitive ability gap, then we would not expect the gap to be eliminated after controlling for parental cognitive ability and parenting practices. After all, if the force of societal racism was impactful enough to cause the cognitive ability gap between blacks and whites, we would expect this force to also hinder the cognitive ability of those black children who have parents with similar cognitive ability and parenting practices as white children. For more detail on studies that attempt to explain the cognitive ability gap, see my post here.

Conclusion

In this post, I considered some of the most commonly posited explanations of the black-white cognitive ability gap: the Test Bias Hypothesis, the Schooling Hypothesis, the Socioeconomic Hypothesis, and the Racism Hypothesis. I argued that each of these hypotheses failed to explain the cognitive ability gap because each of these hypotheses failed to make predictions that were confirmed by the data.

The main points of this post can be summarized as follows. Firstly, the Test Bias Hypothesis fails because black-white gaps are found on culture-free tests, the gap is larger on more abstract and less culturally loaded tests, and because cognitive ability tests predict success for blacks equally as well as for whites. Secondly, the Schooling Hypothesis fails because the black-white differences in school quality and funding are either non-existent or small in magnitude, and because black-white differences in cognitive ability appear in full size before children reach school age. Thirdly, the Socioeconomic Hypothesis fails because a substantial portion of the black-white cognitive ability gap remains after controlling for socioeconomic status. Finally, the Racism Hypothesis fails because the most plausible mechanism through which societal racism might operate (e.g., parental SES, schooling, etc.) fail to account for the gap, because skin color has no or negligible impact on cognitive ability after controlling for important covariates (e.g., after controlling for maternal race), because black children raised by white parents have far higher levels of cognitive ability than black children raised by black parents, and because the cognitive ability gap is eliminated after controlling for parental characteristics.

A true explanation of the black-white cognitive ability gap must therefore lie elsewhere. I will investigate alternative explanations of the cognitive ability gap in later posts.