Gender gaps in reading and math

Girls scored higher than boys in reading in 99.9% (95% CI 99.8% to 100.0%)99.9% percent of districts where the reading gap is below zeroEstimandshare of districts (each counted once) where girls' mean reading score exceeds boys', bootstrap interval over statesInterval95% CI 99.8% to 100.0%n12,327 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of the 12,32712,327 districtsEstimandcount of districtsWeightsnone (full count)Populationschool districts with a male-female reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive school districts with a reading estimate, and the average district's reading gap was −0.76 (95% CI −0.82 to −0.70)−0.76 reading gap in grade levels, male minus female (negative means girls scored higher)Estimandmean over districts of the male minus female reading gap, bootstrap interval over statesInterval95% CI −0.82 to −0.70n12,327 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive grade levels. Each gap in this chapter is the mean score of boys minus the mean score of girls in the same district, so a negative gap means girls scored higher. A grade level is the scale of the Stanford Education Data Archive (SEDA), on which one unit is the learning of an average school year, and the scores pool third through eighth grade over the 2009 to 2019 school years.

The math gap has no common sign. Across the 12,20712,207 districtsEstimandcount of districtsWeightsnone (full count)Populationschool districts with a male-female math gap estimate (Virginia has none, DC is absent and Hawaii has no districts)Periodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive districts with a math estimate its mean is −0.031 (95% CI −0.060 to −0.004)−0.031 math gap in grade levels, male minus female (negative means girls scored higher)Estimandmean over districts of the male minus female math gap, bootstrap interval over statesInterval95% CI −0.060 to −0.004n12,207 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive grade levels, boys lead in 43.8% (95% CI 38.1% to 49.4%)43.8% percent of districts where the math gap is above zeroEstimandshare of districts (each counted once) where boys' mean math score exceeds girls', bootstrap interval over statesInterval95% CI 38.1% to 49.4%n12,207 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of districts, and the gap moves with district income. It averages −0.20 (95% CI −0.26 to −0.15)−0.20 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts (equal numbers of districts), bootstrap interval over statesInterval95% CI −0.26 to −0.15n1,213 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the lowest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive grade levels in the tenth of districts with the lowest median household income and +0.08 (95% CI 0.04 to 0.12)+0.08 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts (equal numbers of districts), bootstrap interval over statesInterval95% CI 0.04 to 0.12n1,214 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the highest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive in the tenth with the highest. District income is measured for 2019 to 2023, after the 2009 to 2019 test scores, on the district boundaries of the 2021 to 2022 school year matched to SEDA by district code.

Girls and boys attend school in the same district, so a gap between them is measured inside one set of schools and neighborhoods. It does not come from setting a rich district beside a poor one, as a comparison of average scores across districts does. If a gap has one sign in almost every district, the sign does not depend on the district. If the sign changes with a district's income, the gap varies with place. This chapter applies both readings to reading and to math. The gaps between income groups and between racial groups are the subject of Chapter 2.1, Test scores, class and race.

How the gaps are measured

SEDA builds district estimates from state accountability tests in reading and math for third through eighth graders and places the tests from every state on one scale through the National Assessment of Educational Progress. For each district it reports the mean score of all students, of girls and of boys, and the difference between the last two. This chapter reads the gap estimates for 12,32712,327 districtsEstimandcount of districtsWeightsnone (full count)Populationschool districts with a male-female reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive districts in reading and 12,20712,207 districtsEstimandcount of districtsWeightsnone (full count)Populationschool districts with a male-female math gap estimate (Virginia has none, DC is absent and Hawaii has no districts)Periodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive in math, and the county estimates that SEDA builds the same way.

Weights depend on the question. A statement about districts, such as the share where girls lead, counts each district once. A statement about the average student weights each district by its number of test scores, where a test score is one student's result in one grade and one year. The text names the weight wherever it is not each district once. Intervals come from a bootstrapBootstrap intervalA range for an estimate found by recomputing it on many random resamples of the data and reading the range from the results. It shows how much the estimate would vary from sample to sample. that resamples whole states, because the districts of one state share tests and policies and are not independent draws. The intervals for state means in the state figure are the exception, and the state section says so.

The two gaps across school districts

Figure 1 plots the district gaps. The reading distribution lies almost wholly on the girls' side of zero. The central 90%90% percent of districts between the 5th and 95th percentiles of the gapEstimanddefinition: the central range of district gaps, from the 5th to the 95th percentileWeightsnone (definition)Populationschool districts with a gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of district reading gaps, from the 5th to the 95th percentile, run from −1.18−1.18 reading gap in grade levels, male minus female (negative means girls scored higher)Estimand5th percentile of district gaps, each district counted oncen12,327 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive to −0.36−0.36 reading gap in grade levels, male minus female (negative means girls scored higher)Estimand95th percentile of district gaps, each district counted oncen12,327 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive grade levels, and girls scored higher in 12,31312,313 districtsEstimandcount of districts where girls' mean reading score exceeds boys'Weightsnone (full count)Populationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive districts and boys in 1414 districtsEstimandcount of districts where boys' mean reading score exceeds girls'Weightsnone (full count)Populationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive. The central 90%90% percent of districts between the 5th and 95th percentiles of the gapEstimanddefinition: the central range of district gaps, from the 5th to the 95th percentileWeightsnone (definition)Populationschool districts with a gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of math gaps run from −0.35−0.35 math gap in grade levels, male minus female (negative means girls scored higher)Estimand5th percentile of district gaps, each district counted oncen12,207 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive to +0.29+0.29 math gap in grade levels, male minus female (negative means girls scored higher)Estimand95th percentile of district gaps, each district counted oncen12,207 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive. Weighting each district by its number of reading test scores gives a mean of −0.79 (95% CI −0.85 to −0.73)−0.79 reading gap in grade levels, male minus female (negative means girls scored higher)Estimandreading-test-weighted mean over districts of the male minus female reading gap, bootstrap interval over statesInterval95% CI −0.85 to −0.73n12,327 districtsWeightsnumber of reading test scores in the districtPopulationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive grade levels. SEDA also reports gaps in standard deviations of the national cohortSD units (SEDA)Test scores stated in units of the typical spread of scores around the national average, so that a score reads as the number of such spreads above or below the national average. SEDA also reports scores in grade levels, which are a separate scale., and in those units the district means are −0.228 (95% CI −0.246 to −0.211)−0.228 reading gap in standard deviations of the national cohort, male minus female (negative means girls scored higher)Estimandmean over districts of the male minus female reading gap, bootstrap interval over statesInterval95% CI −0.246 to −0.211n12,327 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive for reading and −0.010 (95% CI −0.019 to −0.002)−0.010 math gap in standard deviations of the national cohort, male minus female (negative means girls scored higher)Estimandmean over districts of the male minus female math gap, bootstrap interval over statesInterval95% CI −0.019 to −0.002n12,207 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive for math.

A district's gap is an estimate with its own error, so a negative estimate does not always mean that girls scored higher. Reading passes this test. The 95% interval, built from the estimate and the empirical-Bayes standard error that SEDA reports, lies below zero in 98.9% (95% CI 98.3% to 99.4%)98.9% percent of districts whose 95% interval for the reading gap lies below zeroEstimandshare of districts (each counted once) whose 95% interval, built from the estimate and its standard error, lies below zero; bootstrap interval over statesInterval95% CI 98.3% to 99.4%n12,254 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of the districts that report a standard error. Math does not. The interval lies below zero in 31.3% (95% CI 25.7% to 37.8%)31.3% percent of districts whose 95% interval for the math gap lies below zeroEstimandshare of districts (each counted once) whose 95% interval, built from the estimate and its standard error, lies below zero; bootstrap interval over statesInterval95% CI 25.7% to 37.8%n12,140 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of districts, above zero in 21.5% (95% CI 17.3% to 26.3%)21.5% percent of districts whose 95% interval for the math gap lies above zeroEstimandshare of districts (each counted once) whose 95% interval, built from the estimate and its standard error, lies above zero; bootstrap interval over statesInterval95% CI 17.3% to 26.3%n12,140 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive and on both sides of zero in 47.2% (95% CI 42.2% to 51.8%)47.2% percent of districts whose 95% interval for the math gap includes zeroEstimandshare of districts (each counted once) whose 95% interval, built from the estimate and its standard error, includes zero; bootstrap interval over statesInterval95% CI 42.2% to 51.8%n12,140 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive. The median standard error of a district's math gap is 0.060.06 standard error of a district's math gap, grade levelsEstimandmedian over districts of the standard error SEDA reports for the male minus female math gapn12,140 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive grade levels, which exceeds the size of the mean math gap.

The math distribution is centered close to zero, with 43.8%43.8% percent of districts where the math gap is above zeroEstimandshare of districts (each counted once) where boys' mean math score exceeds girls', bootstrap interval over statesInterval95% CI 38.1% to 49.4%n12,207 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of districts on the boys' side. Weighting by test scores moves the mean toward girls, to −0.06 (95% CI −0.09 to −0.02)−0.06 math gap in grade levels, male minus female (negative means girls scored higher)Estimandmath-test-weighted mean over districts of the male minus female math gap, bootstrap interval over statesInterval95% CI −0.09 to −0.02n12,207 districtsWeightsnumber of math test scores in the districtPopulationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive grade levels, and districts where boys lead hold 36.5% (95% CI 29.8% to 44.1%)36.5% percent of math test scores in districts where the math gap is above zeroEstimandshare of math test scores that are in districts where boys' mean exceeds girls', bootstrap interval over statesInterval95% CI 29.8% to 44.1%n12,207 districtsWeightsnumber of math test scores in the districtPopulationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of math test scores. The difference comes from the largest districts and not from a steady pattern with size. The 100100 districtsEstimanddefinition: the districts with the most math test scoresWeightsnone (definition)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive districts with the most math test scores average −0.09 (95% CI −0.16 to −0.05)−0.09 math gap in grade levels, male minus female (negative means girls scored higher)Estimandmean of the male minus female math gap, each district counted once, bootstrap interval over statesInterval95% CI −0.16 to −0.05n100 districtsWeightsnone (each district counts once)Populationthe 100 districts with the most math test scoresPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive grade levels, and boys lead in 28% (95% CI 14% to 39%)28% percent of districts where the math gap is above zeroEstimandshare of districts where boys' mean math score exceeds girls', bootstrap interval over statesInterval95% CI 14% to 39%n100 districtsWeightsnone (each district counts once)Populationthe 100 districts with the most math test scoresPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of them. Chicago Public Schools (−0.34 (95% CI −0.35 to −0.33)−0.34 math gap in grade levels, male minus female (negative means girls scored higher)EstimandSEDA's male minus female math gap for the district; the interval uses SEDA's empirical-Bayes standard errorInterval95% CI −0.35 to −0.33n1,628,988 math test scoresWeightsnone (one district)Populationstudents in the Chicago Public Schools District 299Periodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive) and the Philadelphia City School District (−0.33 (95% CI −0.35 to −0.32)−0.33 math gap in grade levels, male minus female (negative means girls scored higher)EstimandSEDA's male minus female math gap for the district; the interval uses SEDA's empirical-Bayes standard errorInterval95% CI −0.35 to −0.32n960,980 math test scoresWeightsnone (one district)Populationstudents in the Philadelphia City SDPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive) are among the ten largest and have gaps well toward girls. The rank correlation between a district's size and its math gap is −0.02 (95% CI −0.09 to 0.05)−0.02 correlation coefficientEstimandcorrelation of the percentile rank of a district's math test scores with the percentile rank of its math gap (a rank correlation), each district counted once, bootstrap interval over statesInterval95% CI −0.09 to 0.05n12,207 districtsWeightsnone (each district counts once)Populationschool districts with a math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive.

Figure 1

Girls scored higher than boys in reading in 99.9% of districts, while math gaps fall on both sides of zero

Distribution across school districts of the male minus female mean test score gap in grade levels, reading and math, third through eighth grade pooled over 2009 to 2019

Show data

Data for Girls scored higher than boys in reading in 99.9% of districts, while math gaps fall on both sides of zero

density	gap	subject
0	−1.90	Reading
1	−1.85	Reading
1	−1.80	Reading
3	−1.75	Reading
4	−1.70	Reading
9	−1.65	Reading
16	−1.60	Reading
24	−1.55	Reading
32	−1.50	Reading
49	−1.45	Reading
89	−1.40	Reading
117	−1.35	Reading
153	−1.30	Reading
226	−1.25	Reading
329	−1.20	Reading
446	−1.15	Reading
593	−1.10	Reading
778	−1.05	Reading
1035	−1	Reading
1247	−0.950	Reading
1375	−0.900	Reading
1510	−0.850	Reading
1588	−0.800	Reading
1578	−0.750	Reading
1525	−0.700	Reading
1406	−0.650	Reading
1290	−0.600	Reading
1139	−0.550	Reading
928	−0.500	Reading
727	−0.450	Reading
567	−0.400	Reading
421	−0.350	Reading
297	−0.300	Reading
195	−0.250	Reading
129	−0.200	Reading
71	−0.150	Reading
44	−0.100	Reading
29	−0.050	Reading
10	0	Reading
6	0.050	Reading
5	0.100	Reading
2	0.150	Reading
2	0.200	Reading
1	0.250	Reading
0	0.300	Reading
0	0.350	Reading
0	0.400	Reading
0	0.450	Reading
0	0.500	Reading
0	0.550	Reading
0	0.600	Reading
0	0.650	Reading
0	0.700	Reading
0	0.750	Reading
0	0.800	Reading
0	0.850	Reading
0	0.900	Reading
0	0.950	Reading
0	1	Reading
0	1.05	Reading
0	1.10	Reading
0	1.15	Reading
1	1.20	Reading
1	1.25	Reading
0	1.30	Reading
0	−1.90	Math
0	−1.85	Math
0	−1.80	Math
0	−1.75	Math
0	−1.70	Math
0	−1.65	Math
0	−1.60	Math
0	−1.55	Math
0	−1.50	Math
0	−1.45	Math
0	−1.40	Math
0	−1.35	Math
0	−1.30	Math
0	−1.25	Math
1	−1.20	Math
0	−1.15	Math
0	−1.10	Math
0	−1.05	Math
0	−1	Math
0	−0.950	Math
0	−0.900	Math
1	−0.850	Math
7	−0.800	Math
13	−0.750	Math
15	−0.700	Math
24	−0.650	Math
55	−0.600	Math
94	−0.550	Math
119	−0.500	Math
185	−0.450	Math
313	−0.400	Math
507	−0.350	Math
748	−0.300	Math
1021	−0.250	Math
1354	−0.200	Math
1700	−0.150	Math
1931	−0.100	Math
2095	−0.050	Math
2058	0	Math
1955	0.050	Math
1662	0.100	Math
1278	0.150	Math
1007	0.200	Math
708	0.250	Math
485	0.300	Math
274	0.350	Math
182	0.400	Math
100	0.450	Math
52	0.500	Math
21	0.550	Math
19	0.600	Math
6	0.650	Math
4	0.700	Math
1	0.750	Math
1	0.800	Math
0	0.850	Math
0	0.900	Math
0	0.950	Math
0	1	Math
1	1.05	Math
0	1.10	Math
0	1.15	Math
0	1.20	Math
0	1.25	Math
0	1.30	Math

The reading gap by state

The reading gap has one sign in every state and a size that differs by state (Figure 2). The test-weighted mean reading gap is below zero in each of the 4848 statesEstimandcount of statesWeightsnone (full count)Populationstates with districts in the income sample (no Virginia, Hawaii or DC)Periodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive states that have both gaps and a district income. It runs from −1.28 (95% CI −1.32 to −1.25)−1.28 reading gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of district gaps in the state (the interval treats districts as independent)Interval95% CI −1.32 to −1.25n234 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the state with the largest reading lead for girlsPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive grade levels in Arkansas to −0.46 (95% CI −0.50 to −0.42)−0.46 reading gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of district gaps in the state (the interval treats districts as independent)Interval95% CI −0.50 to −0.42n326 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the state with the smallest reading lead for girlsPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive in Iowa. Differences between states account for 70% (95% CI 68% to 72%)70% percent of the variance of district reading gaps that lies between statesEstimandR squared of district reading gaps on state, weighted by test scores; interval from resampling districtsInterval95% CI 68% to 72%n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with a reading gap, a math gap and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data ArchiveNoteResampling districts adds a small upward bias to the share, so the interval sits slightly high. of the test-weighted variance of district reading gaps and 36% (95% CI 34% to 40%)36% percent of the variance of district math gaps that lies between statesEstimandR squared of district math gaps on state, weighted by test scores; interval from resampling districtsInterval95% CI 34% to 40%n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with a reading gap, a math gap and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data ArchiveNoteResampling districts adds a small upward bias to the share, so the interval sits slightly high. of the test-weighted variance of district math gaps. Counting each district once, the shares are 47% (95% CI 45% to 49%)47% percent of the variance of district reading gaps that lies between statesEstimandR squared of district reading gaps on state, weighted by districts, each counted once; interval from resampling districtsInterval95% CI 45% to 49%n12,136 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap, a math gap and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data ArchiveNoteResampling districts adds a small upward bias to the share, so the interval sits slightly high. for reading and 24% (95% CI 23% to 26%)24% percent of the variance of district math gaps that lies between statesEstimandR squared of district math gaps on state, weighted by districts, each counted once; interval from resampling districtsInterval95% CI 23% to 26%n12,136 districtsWeightsnone (each district counts once)Populationschool districts with a reading gap, a math gap and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data ArchiveNoteResampling districts adds a small upward bias to the share, so the interval sits slightly high. for math. Reading is the larger share by either count.

The data cannot say why the size differs. States give different tests, SEDA puts them on one scale, and a difference between states in a gap can come from schools, from students or from what a test asks. What the reading comparison establishes is the sign, and that holds in every state. The intervals for the state means treat a state's districts as independent, which the bootstrap used elsewhere in this chapter does not, so they may be too narrow.

Figure 2

The reading gap is below zero in all 48 states, while the math gap is above zero in 20 of them

Test-weighted mean male minus female gap in grade levels by state, reading and math, districts with both gaps and an income, third through eighth grade pooled over 2009 to 2019

• Math
• Reading

Show data

Data for The reading gap is below zero in all 48 states, while the math gap is above zero in 20 of them

conf_high	conf_low	estimate	n	state	subject
0.299	0.228	0.263	326	Iowa	Math
−0.424	−0.499	−0.461	326	Iowa	Reading
0.219	0.095	0.157	41	Utah	Math
−0.722	−0.799	−0.760	41	Utah	Reading
0.164	0.055	0.109	110	Idaho	Math
−0.618	−0.727	−0.672	110	Idaho	Reading
0.144	0.068	0.106	323	Minnesota	Math
−0.573	−0.641	−0.607	323	Minnesota	Reading
0.135	0.073	0.104	509	Oklahoma	Math
−0.495	−0.563	−0.529	509	Oklahoma	Reading
0.132	0.070	0.101	513	Michigan	Math
−0.692	−0.742	−0.717	513	Michigan	Reading
0.134	0.060	0.097	172	Oregon	Math
−0.660	−0.737	−0.698	172	Oregon	Reading
0.151	0.020	0.085	150	North Dakota	Math
−0.685	−0.834	−0.760	150	North Dakota	Reading
0.156	0.011	0.084	240	Nebraska	Math
−0.438	−0.572	−0.505	240	Nebraska	Reading
0.108	0.054	0.081	288	Indiana	Math
−0.901	−0.947	−0.924	288	Indiana	Reading
0.127	0.028	0.077	284	Kansas	Math
−0.479	−0.566	−0.523	284	Kansas	Reading
0.130	0.006	0.068	48	Wyoming	Math
−0.531	−0.651	−0.591	48	Wyoming	Reading
0.100	0.023	0.061	410	Wisconsin	Math
−0.676	−0.749	−0.712	410	Wisconsin	Reading
0.089	−0.017	0.036	210	Montana	Math
−0.747	−0.866	−0.806	210	Montana	Reading
0.064	−0.009	0.027	141	New Hampshire	Math
−0.937	−1.01	−0.973	141	New Hampshire	Reading
0.055	−0.008	0.024	174	Maine	Math
−0.968	−1.05	−1.01	174	Maine	Reading
0.095	−0.063	0.016	147	South Dakota	Math
−0.621	−0.774	−0.698	147	South Dakota	Reading
0.057	−0.028	0.014	158	Connecticut	Math
−0.595	−0.654	−0.625	158	Connecticut	Reading
0.067	−0.048	0.009	16	Delaware	Math
−0.705	−0.872	−0.788	16	Delaware	Reading
0.069	−0.059	0.005	166	Colorado	Math
−0.899	−0.992	−0.945	166	Colorado	Reading
0.020	−0.047	−0.014	67	Florida	Math
−0.558	−0.607	−0.583	67	Florida	Reading
0.004	−0.041	−0.019	607	Ohio	Math
−0.696	−0.734	−0.715	607	Ohio	Reading
−0.018	−0.055	−0.037	1,004	Texas	Math
−0.556	−0.590	−0.573	1,004	Texas	Reading
0.006	−0.089	−0.041	499	Pennsylvania	Math
−0.897	−0.961	−0.929	499	Pennsylvania	Reading
0.059	−0.144	−0.043	16	Nevada	Math
−0.776	−1.06	−0.920	16	Nevada	Reading
−0.015	−0.077	−0.046	492	New Jersey	Math
−0.971	−1.01	−0.992	492	New Jersey	Reading
−0.003	−0.089	−0.046	51	Vermont	Math
−1.02	−1.13	−1.07	51	Vermont	Reading
0.056	−0.150	−0.047	652	New York	Math
−0.699	−0.899	−0.799	652	New York	Reading
−0.021	−0.075	−0.048	276	Washington	Math
−0.916	−0.971	−0.943	276	Washington	Reading
−0.016	−0.084	−0.050	512	Missouri	Math
−0.776	−0.836	−0.806	512	Missouri	Reading
−0.025	−0.096	−0.060	826	California	Math
−0.810	−0.858	−0.834	826	California	Reading
0.014	−0.193	−0.090	47	Alaska	Math
−0.739	−0.964	−0.851	47	Alaska	Reading
−0.056	−0.150	−0.103	55	West Virginia	Math
−1.16	−1.26	−1.21	55	West Virginia	Reading
−0.046	−0.179	−0.113	35	Rhode Island	Math
−1.08	−1.24	−1.16	35	Rhode Island	Reading
−0.081	−0.151	−0.116	268	Massachusetts	Math
−1.12	−1.17	−1.14	268	Massachusetts	Reading
−0.074	−0.167	−0.120	182	Arizona	Math
−0.802	−0.867	−0.834	182	Arizona	Reading
−0.091	−0.165	−0.128	115	North Carolina	Math
−0.480	−0.542	−0.511	115	North Carolina	Reading
−0.062	−0.197	−0.129	754	Illinois	Math
−0.901	−0.994	−0.948	754	Illinois	Reading
−0.125	−0.215	−0.170	180	Georgia	Math
−0.868	−0.941	−0.904	180	Georgia	Reading
−0.131	−0.231	−0.181	140	Tennessee	Math
−0.806	−0.879	−0.843	140	Tennessee	Reading
−0.141	−0.238	−0.189	172	Kentucky	Math
−0.737	−0.853	−0.795	172	Kentucky	Reading
−0.127	−0.265	−0.196	83	New Mexico	Math
−0.957	−1.13	−1.04	83	New Mexico	Reading
−0.156	−0.247	−0.202	78	South Carolina	Math
−0.805	−0.878	−0.841	78	South Carolina	Reading
−0.173	−0.258	−0.215	69	Louisiana	Math
−0.986	−1.05	−1.02	69	Louisiana	Reading
−0.198	−0.287	−0.243	234	Arkansas	Math
−1.25	−1.32	−1.28	234	Arkansas	Reading
−0.153	−0.350	−0.252	24	Maryland	Math
−1.11	−1.21	−1.16	24	Maryland	Reading
−0.244	−0.351	−0.298	137	Alabama	Math
−0.764	−0.854	−0.809	137	Alabama	Reading
−0.301	−0.376	−0.339	135	Mississippi	Math
−0.819	−0.884	−0.852	135	Mississippi	Reading

The math gap by district income

District income here is the mean of tract median household incomes in the district, weighted by households, from the American Community Survey for 2019 to 2023, which is after the 2009 to 2019 test scores. Districts are the boundaries of the 2021 to 2022 school year, matched to SEDA by district code. The income sample holds the 12,13612,136 districtsEstimandcount of districtsWeightsnone (full count)Populationschool districts with a reading gap, a math gap, an all-student score and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive districts with both gaps and an income, which cover 193.8 million193.8 million math test scores (one per student, grade and year tested)Estimandsum of the math test scores behind the district estimatesn12,136 districtsWeightsnone (full count)Populationstudents in the districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive math test scores. Tenths of districts, or deciles, hold equal numbers of districts. Incomes in the lowest tenth are at most $51,26851,268 dollars of district median household incomeEstimandhighest income in the lowest tenth (equal numbers of districts in each tenth)n1,213 districtsWeightsnone (a boundary)Populationthe lowest tenth of districts by incomePeriodhousehold income in 2023 dollars, ACS 2019 to 2023SourceAmerican Community SurveyNoteA household-weighted mean of tract medians, top-coded at 250,001 dollars., and incomes in the highest tenth are at least $115,294115,294 dollars of district median household incomeEstimandlowest income in the highest tenth (equal numbers of districts in each tenth)n1,214 districtsWeightsnone (a boundary)Populationthe highest tenth of districts by incomePeriodhousehold income in 2023 dollars, ACS 2019 to 2023SourceAmerican Community SurveyNoteA household-weighted mean of tract medians, top-coded at 250,001 dollars..

The math gap rises with income (Figure 3). Weighted by test scores, it averages −0.20 (95% CI −0.26 to −0.15)−0.20 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts (equal numbers of districts), bootstrap interval over statesInterval95% CI −0.26 to −0.15n1,213 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the lowest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive grade levels in the lowest tenth and +0.08 (95% CI 0.04 to 0.12)+0.08 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts (equal numbers of districts), bootstrap interval over statesInterval95% CI 0.04 to 0.12n1,214 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the highest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive in the highest, and the two intervals do not overlap. The ten decile means span 0.28 (95% CI 0.22 to 0.35)0.28 grade levels (largest minus smallest of the ten decile means of the math gap)Estimandlargest minus smallest test-weighted mean math gap across the ten income tenths, bootstrap interval over statesInterval95% CI 0.22 to 0.35n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive grade levels from the smallest to the largest. Counting each district once gives −0.16 (95% CI −0.22 to −0.10)−0.16 math gap in grade levels, male minus female (negative means girls scored higher)Estimandmean of the male minus female math gap within a tenth of districts, each district counted once, bootstrap interval over statesInterval95% CI −0.22 to −0.10n1,213 districtsWeightsnone (each district counts once)Populationschool districts in the lowest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive and +0.07 (95% CI 0.05 to 0.11)+0.07 math gap in grade levels, male minus female (negative means girls scored higher)Estimandmean of the male minus female math gap within a tenth of districts, each district counted once, bootstrap interval over statesInterval95% CI 0.05 to 0.11n1,214 districtsWeightsnone (each district counts once)Populationschool districts in the highest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive, the same change of sign. Ranking districts by their average score in math and reading instead of by income gives −0.21 (95% CI −0.29 to −0.18)−0.21 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts ranked by average score (equal numbers of districts), bootstrap interval over statesInterval95% CI −0.29 to −0.18n1,213 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the lowest tenth of districts by average math and reading score of all studentsPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive in the lowest tenth and +0.13 (95% CI 0.10 to 0.17)+0.13 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts ranked by average score (equal numbers of districts), bootstrap interval over statesInterval95% CI 0.10 to 0.17n1,214 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the highest tenth of districts by average math and reading score of all studentsPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive in the highest. Ranking by the poverty rate from the same survey gives +0.12 (95% CI 0.09 to 0.16)+0.12 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts ranked by the share of residents below the poverty line (ACS 2019 to 2023; equal numbers of districts), bootstrap interval over statesInterval95% CI 0.09 to 0.16n1,213 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the lowest tenth of districts by poverty ratePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive in the tenth of districts with the lowest poverty and −0.19 (95% CI −0.24 to −0.14)−0.19 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts ranked by the share of residents below the poverty line (ACS 2019 to 2023; equal numbers of districts), bootstrap interval over statesInterval95% CI −0.24 to −0.14n1,214 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the highest tenth of districts by poverty ratePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive in the tenth with the highest. Assigning each tract to a single district, as the book's tract file does, leaves 9,9939,993 districtsEstimandcount of districtsWeightsnone (full count)Populationschool districts in the income sample that also have an income when each tract belongs to one district onlyPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive districts, and the lowest and highest tenths give −0.21 (95% CI −0.27 to −0.16)−0.21 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts ranked by an income built from each tract's single assigned district (equal numbers of districts), bootstrap interval over statesInterval95% CI −0.27 to −0.16n999 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the lowest tenth of districts by income when each tract belongs to one district onlyPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive and +0.09 (95% CI 0.04 to 0.13)+0.09 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female math gap within a tenth of districts ranked by an income built from each tract's single assigned district (equal numbers of districts), bootstrap interval over statesInterval95% CI 0.04 to 0.13n1,000 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the highest tenth of districts by income when each tract belongs to one district onlyPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive, so the change of sign does not depend on how tracts are split between districts. Income is the ranking used here because an average score is itself a test outcome.

The mix of districts shifts with income as well. Boys lead in math in 23.7% (95% CI 15.3% to 33.1%)23.7% percent of districts where boys' mean math score exceeds girls'Estimandshare of districts (each counted once) where boys' mean math score exceeds girls', bootstrap interval over statesInterval95% CI 15.3% to 33.1%n1,213 districtsWeightsnone (each district counts once)Populationschool districts in the lowest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive of the districts in the lowest tenth, and girls lead in 33.4% (95% CI 25.2% to 40.5%)33.4% percent of districts where girls' mean math score exceeds boys'Estimandshare of districts (each counted once) where girls' mean math score exceeds boys', bootstrap interval over statesInterval95% CI 25.2% to 40.5%n1,214 districtsWeightsnone (each district counts once)Populationschool districts in the highest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive of the districts in the highest. Districts of both signs appear at every income (Figure 4), so income moves the balance between the two signs and does not sort districts into two kinds. Across the income sample, the test-weighted correlation between a district's math gap and its percentile rank in income is +0.40 (95% CI 0.31 to 0.48)+0.40 correlation coefficientEstimandtest-weighted correlation of the district math gap with the district's percentile rank of income, bootstrap interval over statesInterval95% CI 0.31 to 0.48n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive.

The reading gap follows income much less. Its decile means run from −0.79 (95% CI −0.89 to −0.71)−0.79 reading gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female reading gap within a tenth of districts (equal numbers of districts), bootstrap interval over statesInterval95% CI −0.89 to −0.71n1,213 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the lowest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive in the lowest tenth to −0.87 (95% CI −0.97 to −0.78)−0.87 reading gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of the male minus female reading gap within a tenth of districts (equal numbers of districts), bootstrap interval over statesInterval95% CI −0.97 to −0.78n1,214 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the highest tenth of districts by district median household incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive in the highest, and each of the ten lies below zero. They span 0.13 (95% CI 0.06 to 0.24)0.13 grade levels (largest minus smallest of the ten decile means of the reading gap)Estimandlargest minus smallest test-weighted mean reading gap across the ten income tenths, bootstrap interval over statesInterval95% CI 0.06 to 0.24n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data ArchiveNoteThe largest minus the smallest of ten noisy means is biased upward. grade levels, against 0.280.28 grade levels (largest minus smallest of the ten decile means of the math gap)Estimandlargest minus smallest test-weighted mean math gap across the ten income tenths, bootstrap interval over statesInterval95% CI 0.22 to 0.35n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive for the math gap. The correlation of the reading gap with income rank is −0.09 (95% CI −0.21 to 0.04)−0.09 correlation coefficientEstimandtest-weighted correlation of the district reading gap with the district's percentile rank of income, bootstrap interval over statesInterval95% CI −0.21 to 0.04n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive, and its interval includes zero.

Figure 3

The math gap rises from −0.20 grade levels in the lowest-income tenth of districts to +0.08 in the highest, while the reading gap stays below zero

Test-weighted mean male minus female gap in grade levels by tenth of district median household income, reading and math, third through eighth grade pooled over 2009 to 2019

Mean male minus female gap (grade levels)

Show data

Data for The math gap rises from −0.20 grade levels in the lowest-income tenth of districts to +0.08 in the highest, while the reading gap stays below zero

Subject	Income tenth	Median district income (dollars)	Mean gap (grade levels)	Lower limit of the 95% interval	Upper limit of the 95% interval	Districts
Reading	1	46,464	−0.79	−0.89	−0.71	1,213
Reading	2	54,714	−0.80	−0.88	−0.72	1,214
Reading	3	60,094	−0.78	−0.86	−0.70	1,213
Reading	4	64,792	−0.78	−0.85	−0.70	1,214
Reading	5	69,421	−0.74	−0.81	−0.68	1,214
Reading	6	74,142	−0.76	−0.84	−0.69	1,213
Reading	7	80,405	−0.77	−0.86	−0.69	1,214
Reading	8	88,948	−0.80	−0.85	−0.72	1,213
Reading	9	102,837	−0.81	−0.88	−0.73	1,214
Reading	10	136,556	−0.87	−0.97	−0.78	1,214
Math	1	46,464	−0.20	−0.26	−0.15	1,213
Math	2	54,714	−0.14	−0.19	−0.10	1,214
Math	3	60,094	−0.14	−0.20	−0.09	1,213
Math	4	64,792	−0.10	−0.15	−0.05	1,214
Math	5	69,421	−0.05	−0.09	−0.01	1,214
Math	6	74,142	−0.07	−0.11	−0.03	1,213
Math	7	80,405	−0.06	−0.11	−0.02	1,214
Math	8	88,948	−0.06	−0.10	0.00	1,213
Math	9	102,837	+0.01	−0.03	+0.06	1,214
Math	10	136,556	+0.08	+0.04	+0.12	1,214

Figure 4

Districts at one income level differ widely in their math gap, and the decile means rise with income, a correlation of +0.40 with income rank

Male minus female math gap in grade levels against district median household income in dollars (log scale), one dot per district, with test-weighted decile means

Loading data

Within states

Income differs across states, and so do tests and policies, so the national pattern could come from differences between states. A regression of the math gap on a district's income rank with a separate intercept for each state gives a slope of +0.027 (95% CI 0.022 to 0.032)+0.027 grade levels of the math gap per ten points of district income rankEstimandslope of the math gap on the district's percentile rank of income (0 to 100), test-weighted regression with a separate intercept for each state and state indicators, errors clustered on stateInterval95% CI 0.022 to 0.032n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive grade levels for each 1010 points of district income rank (0 to 100)Estimandstep of the income rank in which slopes are reportedWeightsnone (definition)Populationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive points of rank, and the slope without the state intercepts is +0.025 (95% CI 0.019 to 0.031)+0.025 grade levels of the math gap per ten points of district income rankEstimandslope of the math gap on the district's percentile rank of income (0 to 100), test-weighted regression without state indicators, errors clustered on stateInterval95% CI 0.019 to 0.031n12,136 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts with both gaps and a district incomePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive. Of the 4545 statesEstimandcount of states given a slopeWeightsnone (full count)Populationstates with at least the minimum number of districts in the income samplePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive states with at least 2525 districtsEstimandinclusion rule: a state needs at least this many districts in the income sampleWeightsnone (definition)Populationstates given a slope of their ownPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive districts in the sample, the slope is positive in 4545 statesEstimandcount of states where the test-weighted slope of the math gap on income rank is positiveWeightsnone (full count)Populationstates with at least the minimum number of districts in the income samplePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive and its 95% interval lies above zero in 3838 statesEstimandcount of states where the 95% interval (heteroskedasticity-robust) for that slope lies above zeroWeightsnone (full count)Populationstates with at least the minimum number of districts in the income samplePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive. The pattern appears within states and does not come from comparing states.

Where the math gap favors boys

Figure 5 maps the math gap for 2,9962,996 countiesEstimandcount of countiesWeightsnone (full count)Populationcounties with a male-female math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive counties. Boys scored higher in 1,0431,043 countiesEstimandcount of counties where boys' mean math score exceeds girls'Weightsnone (full count)Populationcounties with a male-female math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive of them (34.8% (95% CI 25.9% to 44.7%)34.8% percent of counties where the math gap is above zeroEstimandshare of counties (each counted once) where boys' mean math score exceeds girls', bootstrap interval over statesInterval95% CI 25.9% to 44.7%n2,996 countiesWeightsnone (each county counts once)Populationcounties with a male-female math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive) and girls in 1,9531,953 countiesEstimandcount of counties where girls' mean math score exceeds boys'Weightsnone (full count)Populationcounties with a male-female math gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive. Among the 2,9382,938 countiesEstimandcount of countiesWeightsnone (full count)Populationcounties with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive counties that report a standard error, the 95% interval lies above zero in 599599 countiesEstimandcount of counties whose 95% interval (estimate and standard error) for the math gap lies above zeroWeightsnone (full count)Populationcounties with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive counties, below zero in 1,4001,400 countiesEstimandcount of counties whose 95% interval for the math gap lies below zeroWeightsnone (full count)Populationcounties with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive counties and on both sides of zero in 939939 countiesEstimandcount of counties whose 95% interval for the math gap includes zeroWeightsnone (full count)Populationcounties with a math gap estimate and a standard errorPeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive. The reading gap is below zero in each of the 3,1253,125 countiesEstimandcount of countiesWeightsnone (full count)Populationcounties with a male-female reading gap estimatePeriodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive counties with a reading estimate, so a map of it would carry one color. The tooltip on each county gives both gaps.

The map shows differences between states. The test-weighted mean math gap is above zero in 2020 statesEstimandcount of states whose test-weighted mean math gap is above zeroWeightsnone (full count)Populationstates in the income samplePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive of the 4848 statesEstimandcount of statesWeightsnone (full count)Populationstates with districts in the income sample (no Virginia, Hawaii or DC)Periodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive states, and the 1313 statesEstimandcount of states whose 95% interval for the test-weighted mean math gap lies above zeroWeightsnone (full count)Populationstates in the income samplePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive states whose intervals lie above zero are in the Plains, the Great Lakes states, the Mountain West and Oregon. The count depends on the interval, and a linearized standard error gives 1414 statesEstimandcount of states whose 95% interval for the test-weighted mean math gap lies above zero when the standard error is the Taylor-linearized one for a weighted mean of independent districtsWeightsnone (full count)Populationstates in the income samplePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive, adding Montana. The largest state means favor boys in Iowa (+0.26 (95% CI 0.23 to 0.30)+0.26 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of district gaps in the state (the interval treats districts as independent)Interval95% CI 0.23 to 0.30n326 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the state with the largest math lead for boysPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive) and Utah, and the largest favor girls in Mississippi (−0.34 (95% CI −0.38 to −0.30)−0.34 math gap in grade levels, male minus female (negative means girls scored higher)Estimandtest-weighted mean of district gaps in the state (the interval treats districts as independent)Interval95% CI −0.38 to −0.30n135 districtsWeightsnumber of math and reading test scores in the districtPopulationschool districts in the state with the largest math lead for girlsPeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive) and Alabama. Counties where boys scored higher cluster in the states with intervals above zero, and every county in Alabama, Arkansas, Louisiana, Mississippi and South Carolina is on the girls' side.

SEDA has no math estimates for Virginia and no rows for the District of Columbia. It has no Hawaii districts, but its county file covers Hawaii, and Hawaii's counties are drawn with data except Kalawao. In all, 146146 countiesEstimandcount of drawn counties with no math gap estimateWeightsnone (full count)Populationcounties and county equivalents on the map (Puerto Rico is not drawn)Periodtest scores of third through eighth graders, pooled over 2009 to 2019SourceStanford Education Data Archive map units are drawn as no data, and the tooltip on each gives its reason.

Figure 5

Boys scored higher in math than girls in 34.8% of counties, and those counties cluster in the Plains, the Great Lakes states and the Mountain West

Male minus female mean math score gap in grade levels, county, third through eighth grade pooled over 2009 to 2019

+−Reset

Loading county values

Male minus female math gap (grade levels)

−0.40−0.30−0.15−0.05+0.05+0.15+0.30+0.40

Girls scored higherBoys scored higher

Values outside −0.40 to +0.40 take the end colors.

No data (3,143 counties)

Select a class to show only those places.

Find a county

Hover or tap a county for its value, or search by name. Zoom with the buttons on the map, then drag to move it.

What other studies find

Reardon and colleagues estimated the same gaps with SEDA data for the school years ending in 2009 through 2016.[1] The average district had no math gap and a reading gap favoring girls, and math gaps favored boys more in districts of higher socioeconomic status and in districts with larger gender disparities in adult income, education and occupation. For English language arts they did not find strong associations with those measures. With the 2009 to 2019 data and district income in place of their socioeconomic composite, this chapter finds a math gap near zero on average, a reading gap favoring girls and a math gap that rises with income, and it finds no detectable association of the reading gap with income rank. It adds that the reading gap differs by state.

Hyde and colleagues analyzed state mathematics assessments for grades two through eleven and found average differences between girls and boys close to zero.[2] Fryer and Levitt followed a national panel of children and found no mean difference in math at school entry, with girls losing ground relative to boys over the first years of school.[3] A gap pooled over third through eighth grade, as here, mixes grades at which the gap may differ.

Description, not a causal test

Nothing in this chapter tests why the math gap moves with district income. A district's income goes together with its schools, families, teachers and peers, and a comparison of girls and boys inside the district does not separate those from one another. The sign of the reading gap and the slope of the math gap are descriptions of districts. An explanation would need a design that changes one of these conditions for some children and not for others, and none is used here.

What these data cannot show

The patterns describe school districts and counties. A district's income is not a family's income, and the data do not say that a girl or a boy from a poorer family scores differently from one in a richer family in the same district (ecological inferenceEcological inferenceDrawing conclusions about individuals from data about places or groups. A pattern across counties need not hold for the people inside them.). A district gap is an average over every student the state tested in third through eighth grade, so it cannot show when in school a gap opens.

Each district's gap carries error, and the standard error of a district's math gap is typically larger than the mean math gap, so the decile means, with intervals, carry the income finding and single districts do not. SEDA's estimates are empirical-Bayes estimates, which pull the noisiest districts toward the average and narrow the spread of district gaps. If the shrinkage pulls toward a state or national mean, it could change the shares of variance that lie between states, and the 95% intervals built from the empirical-Bayes standard errors describe the shrunken estimates. The scores come from state tests that differ in content and format, and a gap can reflect what a test asks as well as how girls and boys learned.

District income is built for this chapter and is not part of SEDA. It is a household-weighted mean of tract medians, split across overlapping districts by land area, so it is not the median income of a district's households, and the survey caps the highest incomes. It is measured for 2019 to 2023, after the 2009 to 2019 scores, on the district boundaries of the 2021 to 2022 school year matched to SEDA by district code, so a district whose boundary or code changed may be matched imperfectly or not at all. In all, 3838 districtsEstimandcount of districts left out of the income figuresWeightsnone (full count)Populationschool districts with both gaps but no district income or no all-student scorePeriodtest scores of third through eighth graders, pooled over 2009 to 2019; district income from the American Community Survey, 2019 to 2023SourceStanford Education Data Archive districts with both gaps have no income and are left out of the income figures. Districts that differ in income also differ in other ways, such as the racial composition of their students, and the chapter does not separate those.

SEDA has no math estimates for Virginia, no rows for the District of Columbia and no districts in Hawaii (Hawaii's counties are in its county file and are drawn), and Puerto Rico is not covered. The estimates describe public school students tested in 2009 to 2019 and say nothing about later years. SEDA reports two groups, male and female. Whether gaps in school scores carry into adult outcomes is the subject of Chapter 3.5, Boys and mobility (planned), which compares boys and girls raised in the same neighborhoods.

---

Data sources

• Stanford Education Data ArchiveMale-female test score gaps and all-student scores in reading and math for school districts and counties, third through eighth grade pooled over 2009 to 2019 (version 6.0)Years: 2009-2019 (version 6.0) · Weights: none in the estimates (empirical-Bayes); districts weighted by their number of test scores or counted once, as each figure says · License: CC BY 4.0 (Stanford CEPA); cite SEDA
• American Community SurveyTract median household income and households, 2019 to 2023, apportioned to school districts with the NCES tract to school district crosswalk (2021 to 2022 school year)Years: 2019-2023 (5-year) · Weights: households, with each tract split across districts by land area · License: Public domain (U.S. Census Bureau)

All data sources

Methods

This chapter compares male-female test score gaps in reading and math across school districts and counties, from the Stanford Education Data Archive 6.0, pooled over third through eighth grade and the 2009 to 2019 school years. A negative gap means girls scored higher. District income is not in the archive: it is built from tract incomes of the American Community Survey for 2019 to 2023, split across districts by land area, and it is measured after the scores (2009 to 2019). Means weight a district by its number of math and reading test scores, or count each district once where the text says so.

References

1. Reardon, Sean F.; Fahle, Erin M.; Kalogrides, Demetra; Podolsky, Anne; Zárate, Rosalía C. (2019). Gender Achievement Gaps in U.S. School Districts. American Educational Research Journal 56(6): 2474-2508. doi:10.3102/0002831219843824
2. Hyde, Janet S.; Lindberg, Sara M.; Linn, Marcia C.; Ellis, Amy B.; Williams, Caroline C. (2008). Gender Similarities Characterize Math Performance. Science 321(5888): 494-495. doi:10.1126/science.1160364
3. Fryer, Roland G., Jr.; Levitt, Steven D. (2010). An Empirical Analysis of the Gender Gap in Mathematics. American Economic Journal: Applied Economics 2(2): 210-240. doi:10.1257/app.2.2.210

How to cite

@misc{singh2026gendergapsinreadingandmath,
  author = {Singh, Vishal},
  title = {Gender gaps in reading and math},
  year = {2026},
  url = {https://vishalsingh.org/zipcode/gender-gaps-in-reading-and-math},
  note = {In ZIP Code Destiny. Published 9 October 2026.}
}

Reading paths

• Outcomes that do not follow the gradient