Cited by 940School Choice and PerformanceCritical Race Theory in EducationEarly Childhood Education and Development

The DOI matches its Crossref record; 27 of the 27 references it cites were found in a registry or an index; none of the 35 sources its claims cite is checked for support yet.

Understanding the Black-White Test Score Gap in the First Two Years of School

Fryer, Roland G., Levitt, Steven D.·2004·Review of Economics and Statistics

Abstract

In previous research, a substantial gap in test scores between white and black students persists, even after controlling for a wide range of observable characteristics. Using a newly available data set (the Early Childhood Longitudinal Study), we demonstrate that in stark contrast to earlier studies, the black-white test score gap among incoming kindergartners disappears when we control for a small number of covariates. Real gains by black children in recent cohorts appear to play an important role in explaining the differences between our findings and earlier research. The availability of better covariates also contributes. Over the first two years of school, however, blacks lose substantial ground relative to other races. There is suggestive evidence that differences in school quality may be an important part of the explanation. None of the other hypotheses we test to explain why blacks are losing ground receive any empirical backing.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

Using ECLS-K data on children entering kindergarten in 1998, Fryer and Levitt find that the raw Black-White test-score gap in math and reading is essentially eliminated by a small set of observable controls, but Black students lose substantial ground relative to Whites over the first two years of school. The only explanation receiving empirical support is that Black students attend worse schools on average, though the authors characterize this evidence as suggestive rather than definitive.

Key findings

  • In fall kindergarten, the raw Black-White gap is .638 SD in math and .401 SD in reading; with the parsimonious controls (age, birth weight, SES, WIC, mother’s age at first birth, number of children’s books), the estimated gap becomes -.094 SD in math and .117 SD in reading, essentially eliminating the gap.
  • Between fall kindergarten and spring first grade, Black students lose ground relative to Whites: raw losses are .090 SD in math and .128 SD in reading; with controls, losses are .156 SD in math and .188 SD in reading.
  • When school fixed effects are included, the Black-White gap lost over time shrinks to less than one-third of the full-sample estimate and is not statistically different from zero: math difference -.047 (.052), reading difference -.05 (.057).
  • Other tested explanations receive little support: Black students do not suffer larger summer setbacks, teacher assessments mirror test-score patterns, having a Black teacher does not benefit Black students, and adding behavioral proxies does not alter findings.

Methods

  • Data: Early Childhood Longitudinal Study Kindergarten Cohort (ECLS-K), a nationally representative sample of over 20,000 children entering kindergarten in 1998; the analysis sample with nonmissing test scores, race, or age is 13,290 for math and 12,601 for reading.
  • Outcomes: IRT math and reading test scores, normalized to mean zero and SD one, plus subjective teacher assessments; main specifications use weighted least squares with sample weights.
  • Design: Regressions of test scores on race dummies and a parsimonious set of controls, with extensions using an exhaustive set of roughly 100 covariates, school fixed effects, teacher fixed effects, and subsample/sensitivity analyses.

Limitations

  • The authors state that the school-quality explanation is “no more than suggestive”: observable school inputs explain only a small fraction of outcome variation, adding them increases R-squared by only .05, the Black-White gap continues to widen after including them, and Hispanics and Asians also attend worse schools but do not lose ground.
  • The authors note thorny sample-selection issues in within-school comparisons and cannot rule out that Whites attending schools with Blacks are on lower academic trajectories, despite initially scoring better than other Whites.
  • The paper cautions against causal interpretation of covariate coefficients, treats them as proxies for environmental and behavioral factors, and does not directly test explanations such as stereotype threat or “acting white”; the analysis covers only the first two years of school, so extrapolations to later grades are speculative.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1162/003465304323031049

Cites

27 of 27 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

121 Own result · 23 Own interpretation · 13 Sources not found · 7 No basis given · 4 Own method · 3 No full text of the sources · 2 Not checked yet

Show all 113 claims

  • It may state two separate claims as one. 43
  • Probably not a substantial claim. 14
  • The quote only partly states this claim. 8
60 of 60
  • The authors state this as an assumption.

    Children of different races are equally responsive to changes in covariates.

    “The results presented in Tables 2 and 3 maintain the assumption that children of different races are equally responsive to changes in covariates.”
  • The paper relies on this without support.

    The ECLS-K is a nationally representative sample of over 20,000 children entering kindergarten in 1998.

    “The Early Childhood Longitudinal Study Kindergarten Cohort (ECLS-K) is a nationally representative sample of over 20,000 children entering kindergarten in 1998.”
  • The authors state this as an assumption.

    The covariates in the sparse regressions are proxies for a broad set of environmental and behavioral factors.

    “We caution against a causal interpretation of the coefficients on the covariates, which we view as proxies for a broad set of environmental and behavioral factors.”
  • The authors state this as an assumption.

    The parsimonious set of covariates adequately captures a child's environment.

    “to the extent our control variables adequately capture a child's environment, the residual gap after including all the covariates would remain constant.”
  • The authors state this as an assumption.

    Skill in manipulating numbers is a more important input into understanding higher-order math than simple tasks such as counting or evaluating relative size.

    “To the extent that skill in manipulating numbers is likely to be a more important input into understanding higher-order math than are simple tasks such as counting or evaluating relative size”
  • The paper relies on this without support.

    Teacher assessments provide a valid measure of student achievement for testing whether test-score results are an artifact of standardized testing.

    “To assess this hypothesis, we examined the relative performance of children of different races on subject teacher evaluations.”
  • The paper relies on this without support.

    School-fixed effects control for all relevant differences in school quality between schools.

    “With school-fixed effects, the estimated Black-White test score gap is identified off of the relative performance of Blacks and Whites attending the same school, as opposed to across schools.”
7 of 7

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

3 of 3

Citations

Cites

Show all 19 references

Research works printed but not linked: 7

  1. Armor G., 1992, Public Interest, 65
  2. 10.2307/1131822 · DOI
  3. Brooks-Gunn J., Child Development, 63, 736
  4. Goldberg A., 1995, Journal of Economic Literature, 33, 762
  5. Jencks C., 1998, American Prospect, 40, 44
  6. Krohn E., 1989, Journal of Psychology, 27, 59
  7. Naglieri J., 1986, Journal of Social Psychology, 24, 81

Cited by

Sorted by citations

Items 201–300 of 940

Metadata

Authors
Fryer, R. G., & Levitt, S. D.
Affiliations
  • National Bureau of Economic Research (US)
  • American Bar Foundation (US)
Venue
Review of Economics and Statistics
Published
2004-05
Volume
86
Issue
2
Pages
447-464
ISSN
  • 0034-6535 (ISSN-L)
  • 1530-9142
OpenAlex
W3021983229
Cited by
936
References
19
Citations by year
2026: 17 · 2025: 17 · 2024: 12 · 2023: 20 · 2022: 36 · 2021: 31 · 2020: 33 · 2019: 51 · 2018: 47 · 2017: 54 · 2016: 67 · 2015: 77 · 2014: 60 · 2013: 60 · 2012: 50
FWCI
152.19
Type
journal article
Language
en
Open access
closed
Locations
  • The Review of Economics and Statistics (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
Topics
  • Social Sciences › Education › School Choice and Performance
  • Social Sciences › Sociology and Political Science › Critical Race Theory in Education
  • Social Sciences › Education › Early Childhood Education and Development
Keywords
Black-White test score gap, early childhood education, kindergarten, school quality, racial achievement gap, Early Childhood Longitudinal Study, covariate control, Black students

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review