by Lipton Matthews
Empirical record shows little relationship between school expenditure levels and convergence in black–white cognitive performance.
The Funding Argument and Its Limits
A long-standing argument in education policy holds that disparities in school funding between black and white students are a primary driver of the black–white gap in cognitive test performance. The logic is intuitively appealing: if black children attend under-resourced schools, then directing more money toward those schools should, over time, diminish the achievement and cognitive performance gap. Were this argument correct, we would expect to see sustained convergence in test scores as per-pupil spending became more equitable. The evidence, however, does not support this expectation.
Per-pupil expenditure figures tell an important story. In 1972, the average black student lived in a school district spending $3,261 per pupil (in 1992 dollars), compared to $3,397 for the average white student—a gap of $136. By 1992, those figures had converged almost entirely: $5,387 for the average black student versus $5,397 for the average white student—a difference of just $10. This near-perfect equalization of spending across racial lines represents a remarkable shift in resource allocation. If funding were the decisive variable, measurable convergence in cognitive performance should have followed. It did not.
The failure of increased and equalized funding to produce convergence is illustrated sharply by recent data from Illinois. A 2025 report found that 80 schools in the state scored below proficiency in mathematics despite spending above the state average per pupil. Many of these schools serve predominantly black student populations. However, the pattern is not unique to Illinois; it reflects a national picture in which the assumption that money is a primary driver of performance gaps has been repeatedly tested and found insufficient.
The Evidence from National Assessment Data
One of the most comprehensive and methodologically careful examinations of trends in the black–white cognitive gap uses data from the National Assessment of Educational Progress (NAEP), the large-scale national assessment administered to representative samples of American schoolchildren—often called “The Nation’s Report Card.” The NAEP long-term trend assessments for mathematics and reading, covering the period 1975 to 2008, show a potent test of whether the gap has narrowed over more than three decades of educational investment.
An analysis of combined NAEP mathematics and reading scores from 1975 to 2008 for white 13-year-olds and white, black, and Hispanic 17-year-olds found that the gap between black and white 17-year-olds had not closed over this period. For 17-year-old black students, the mean difference relative to white 17-year-olds remained at more than three years of educational attainment. The analysis further calculated IQ equivalents using the standard mental age formula across 54 years of educational achievement data. From 1954 to 2008, black 13- and 17-year-olds averaged an IQ equivalent of approximately 85—with values for specific years including 86 and 81 in 1954, 87 and 82 in 1966, 75 and 71 in 1975, and 85 and 77 in 2008. These results indicate no narrowing over 54 years in either educational achievement or IQ equivalent scores.
The same analysis reported that data from large investigations conducted in the 1950s and 1960s in Georgia and Virginia, as well as the Coleman Report of 1966—a nationally representative survey of nearly 600,000 schoolchildren from 4,000 schools—documented the same magnitude of black–white educational achievement gap. In the Coleman data, the gap stood at 2.4 years by Grade 9 and 3.3 years by Grade 12. The report also rendered the sober finding that there was not a strong link between schooling resources and student outcomes. The continuity of this gap from the mid-twentieth century through the early twenty-first century, across different datasets, testing instruments, and decades of increased investment, is an observation of considerable significance.
Longitudinal Evidence from the NLSY Children
Similarly, a major longitudinal study using data from the children of women in the 1979 National Longitudinal Survey of Youth (NLSY79) provides further evidence against convergence. The dataset is particularly valuable because it is longitudinal, uses large samples, tracks children regardless of school attendance, includes both achievement and cognitive measures, employs the same testing instruments over time, and contains extensive family background data, including maternal cognitive ability scores.
Across four tests—reading recognition, reading comprehension, mathematics, and the Peabody Picture Vocabulary Test (PPVT-R, a widely used measure of verbal IQ)—scores were examined for children born from the mid-1970s through the mid-1990s. After controlling for child’s age, maternal IQ, maternal education, family income, maternal age at birth, and family structure, the black–white difference did not decline on any of the tests. For reading comprehension, black scores fell over the period. For the PPVT-R, the implied increase in the black–white difference ranged from 0.13 to 0.19 standard deviations per decade, depending on model specification. The findings were consistent across alternative samples and model specifications.
The study acknowledged the existence of a vigorous debate in the literature and discussed findings from IQ test standardization samples that suggested some narrowing between the early 1970s and early 2000s. However, the author proposed a reconciling hypothesis: that any genuine narrowing in cognitive test scores was concentrated among cohorts born before the late 1970s and that what has been observed since amounts to a plateau—a period of no further convergence—precisely the period covered by the NLSY children data. This interpretation is consistent with the pattern observed in the NAEP and SAT data, in which the closest convergence occurred in the late 1980s and early 1990s, after which the gap stabilized or widened slightly. These results are genuinely eye-opening, considering that during these decades, funding gaps steadily closed, and by 1982, spending per pupil in several states actually favored nonwhite students. Interestingly, by 2002, pupil spending for black students was higher in even Southern states like Alabama and Georgia.
A Meta-Analytic Assessment: Estimated Mean IQ and Distributional Data
The most comprehensive recent synthesis of the evidence on racial differences in measured intelligence brings together 139 American studies spanning 1918 to 2017, with a combined sample of over 400,000 individuals. Studies were selected only if they relied on representative U.S. samples, delivered IQ tests with at least three subtests, reported white reference groups, and provided sufficient data to calculate effect sizes. Studies depending on scholastic achievement tests or unrepresentative samples (such as college-only, elite, or city-selected samples) or that lacked within-group standard deviations were omitted.
Using a random-effects meta-analytic model, the overall estimated mean IQ for black Americans was 85. However, the analysis identified strong evidence of small-study effects—a form of bias in which smaller studies, which are more likely to find larger group means for minority groups, disproportionately influence pooled estimates. After correcting for this bias, the estimated mean fell to 81.5. Critically, the largest, most nationally representative datasets produced the lowest black mean IQ estimates. The NLS, NLSY79, NLSY97, ABCD, Project Talent, US Department of Labor, and WW1 enlistee datasets returned black mean IQ values of 81.25, 81.80, 83.19, 79.81, 77.49, and 83.92, respectively. The average across these high-quality large national samples was 81.8—meaningfully lower than the conventionally cited figure of 85 and approximately 18 to 20 IQ points below the white mean.
A moderator analysis concluded that birth cohort, age at assessment, and number of subtests were not statistically associated with the size of the black–white difference, suggesting further evidence against convergence. The analysis found strong evidence against the hypothesis that the gap has receded for cohorts born after 1960. This directly contradicts the premise that equalized or increased school funding would be expected to produce measurable convergence in the cognitive performance of black Americans relative to whites.
The meta-analysis also estimated mean IQ scores for other racial groups: Hispanics at 88.9, Amerindians at approximately 89, Asian Americans at 103 (with a best estimate of 105 when weighted by Asian subgroup population), and Jewish Americans at approximately 107.4. For all groups, the distributions of cognitive scores were found to be approximately normal, with no statistically significant race differences in variance, skewness, or kurtosis once heterogeneity between studies was accounted for.
Table 1: IQ Score Percentiles by Racial Group in the United States
Observed percentile scores from meta-analytic estimates.
Simulated scores based on group means and standard deviations (assuming
normality) are shown in parentheses.
The
percentile data in Table 1 are particularly instructive because they
reveal the distributional implications of the group mean differences
across the full ability spectrum. At the 50th percentile within each
group (the group median), the figures closely track the meta-analytic
means. The gap between black and white medians—approximately 17 IQ
points—is consistent across the observed and simulated values. At higher
percentiles, where the selection of students for cognitively demanding
academic programs and professional roles is concentrated, the
differences become even more consequential in representational terms,
though the size of the gap in absolute IQ points remains broadly
similar.
Conclusion
The convergence of evidence from national longitudinal assessments, large-scale survey datasets, and comprehensive meta-analytic synthesis leads to a clear conclusion: the black–white gap in measured cognitive performance has not dwindled in the period during which school funding between black and white students has been most closely equalized. Per-pupil spending in black students’ districts reached near parity with that of white students’ districts by 1992. Interestingly, NAEP data spanning 1975 to 2008 showed no narrowing over this entire period. Likewise, longitudinal data from the NLSY79 showed no convergence among children born from the mid-1970s to the mid-1990s. And a meta-analysis covering nearly a century of studies found that the gap has remained persistent.
Hence, the continued below-average performance of predominantly black schools in states such as Illinois—even where per-pupil spending exceeds state averages—is therefore not anomalous. It is consistent with a body of evidence accumulated over decades showing that the relationship between school funding and the black–white cognitive performance gap is, at best, indirect and, at the scale of aggregate national data, effectively absent. The funding argument, while politically salient, is not well supported by the empirical record. If anything, these results only provide fodder for hereditarianism.
* * *
Lipton Matthews is a researcher and podcaster. His work has been featured in Mises, The Federalist, Chronicles, American Thinker, Epoch Times, and other publications. He is also author of Busting African Delusions: Institutions, Human Capital, and the Path to Progress.
Source: https://amgreatness.com/2026/04/21/funding-has-not-closed-the-gap-school-expenditure-and-the-persistence-of-black-white-differences-in-cognitive-performance/
No comments:
Post a Comment