Skip to content
Contents
The Public Mind

    Type to search. Planned chapters match by title only.

    Chapter 12.2

    The diploma map

    Vote-weighted, a county's Democratic share of the two-party presidential vote was 9.2 points higher per 10 points of college share in 2024, against 3.9 in 2000, and the rise came from 2000 to 2008 and from 2012 to 2016, with a change after 2016 of 0.0 for the unadjusted slope and +0.4 to +1.1 for adjusted versions.

    Affiliation
    NYU Stern
    Published

    In 2024 a county's share of residents with a bachelor's degree had a vote-weighted correlationCorrelationA number from minus 1 to plus 1 for how closely two measures move together in a straight-line pattern. Zero means no straight-line pattern; a correlation does not say that one measure causes the other. of +0.66 (95% CI 0.62 to 0.71)+0.66 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.62 to 0.71n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. with its Democratic share of the two-party presidential vote, and a correlation of +0.56 (95% CI 0.45 to 0.66)+0.56 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.45 to 0.66n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. when every county counts once. In 2000 the two were +0.28 (95% CI 0.19 to 0.37)+0.28 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.19 to 0.37n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and +0.10 (95% CI 0.01 to 0.20)+0.10 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.01 to 0.20n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.. As a slope, a county with 1010 points of college shareEstimanddefinition of the slope unitWeightsnone (full counts)Populationunit of the slopePeriod2000 to 2024SourceCounty demographics panel more points of college share had a Democratic two-party share 9.2 (95% CI 8.4 to 10.1)9.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 8.4 to 10.1n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points higher in 2024, vote-weighted, against 3.9 (95% CI 2.5 to 5.1)3.9 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 2.5 to 5.1n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points in 2000. The slope rose from 2000 to 2008 and from 2012 to 2016. Vote-weighted, its change from 2016 to 2024 is 0.0 (95% CI −0.4 to 0.4)0.0 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI −0.4 to 0.4n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points, an interval that includes zero, and counting each county once the change is +0.6 (95% CI 0.2 to 1.0)+0.6 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 0.2 to 1.0n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.. In five adjusted versions of the slope the change after 2016 is positive, between +0.4+0.4 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandlowest of the 2016 to 2024 changes in the slope across five adjusted versions (race and ethnicity held fixed, within states, both, 2000 college share held fixed, 2024 college share held fixed)Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panel and +1.1+1.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandhighest of the 2016 to 2024 changes in the slope across five adjusted versions (race and ethnicity held fixed, within states, both, 2000 college share held fixed, 2024 college share held fixed)Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panel points vote-weighted, so the slope is level after 2016 only in the unadjusted vote-weighted version.

    Both measures describe counties. Education is the share of a county's residents 2525 years of ageEstimandlower age limit of the college shareWeightsnone (full counts)Populationresidents counted in the college sharePeriod2000 to 2024SourceCounty demographics panel and older with a bachelor's degree or more, and the vote is the Democratic share of Democratic and Republican votes, in pointsPercentage pointThe unit for the difference between two percentages. A change from one share to another is stated in points, not in percent, so that absolute and relative change are not confused.. The numbers cannot say how graduates voted, because a county with more graduates also differs from other counties in many other ways, and Chapter 3.2, The diploma divide uses survey respondents for the pattern among individuals. A vote-weighted estimate counts each county in proportion to its two-party votes in that election, so the largest counties count most, and a county-equal estimate counts every county once. The weightings agree on direction and differ in size, most in 2000, and this chapter reports both. Chapter 12.1, Five axes of the county vote ranks education against income, race, age and rurality, and Chapter 12.7, Education, ethnicity and the county shift, 2016 to 2024 takes up the change in county swings from 2016 to 2024. This chapter measures the education gradient and when it appeared.

    County college share is available for the 77 presidential electionsEstimandcount of electionsWeightsnone (full counts)Populationelections with county college sharePeriod2000 to 2024SourceCounty demographics panel presidential elections from 2000 to 2024 in the data used here, and the dating below covers those years, with 3,1143,114 countiesEstimandcount of counties in the estimatesn3,114Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty election returns counties in 2000 and 3,1033,103 countiesEstimandcount of counties in the estimatesn3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns in 2024. The county panel takes college share from the 2000 census sample and from five-year American Community Survey estimates centered on 2010, 2014, 2018 and 2022, and interpolates between those anchors. The elections of 2012, 2016 and 2020 fall midway between two anchors, 2004 and 2008 are interpolated between the 2000 and 2010 anchors, and 2024 takes the 2022 anchor with no extrapolation.The 95% intervals in this chapter come from a bootstrap that resamples whole states, 4949 statesEstimandcount of states, not counting the District of ColumbiaWeightsnone (full counts)Populationstates with counties in these data, Alaska excludedPeriod2000 to 2024SourceCounty election returns states and the District of Columbia, 2,0002,000 replicatesEstimandnumber of bootstrap replicatesWeightsnone (full counts)Populationbootstrap over statesPeriod2000 to 2024SourceCounty election returns times. They show how much a number depends on which states are in the set. The returns are full counts of votes, so the intervals are not sampling error of the vote, and they leave out the sampling error of the college shares. The interval of a change between two elections is computed inside each resample, so it reflects both elections together. The within-state slopes in Figure 7 use HC3 robust standard errorsHC3 robust standard errorsA way of computing standard errors that stays valid when the spread of the outcome differs across respondents, with a correction for small samples. The book uses it for regressions that have no clusters. and treat counties as independent. The last section applies each county's 2000 college share to elections from 1972 to 1996, which ranks counties and does not measure education in those years.

    The slope rose from 2000 to 2008 and from 2012 to 2016

    Figure 1 plots the slope for each election. Vote-weighted, it was 3.9 (95% CI 2.5 to 5.1)3.9 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 2.5 to 5.1n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points in 2000, 5.2 (95% CI 3.9 to 6.4)5.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 3.9 to 6.4n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2004SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2004, 5.9 (95% CI 4.7 to 7.0)5.9 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 4.7 to 7.0n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2008SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2008 and 6.0 (95% CI 4.9 to 7.0)6.0 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 4.9 to 7.0n3,112Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2012SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2012, then 9.2 (95% CI 8.2 to 10.3)9.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 8.2 to 10.3n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2016, 9.3 (95% CI 8.5 to 10.3)9.3 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 8.5 to 10.3n3,105Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2020SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2020 and 9.2 (95% CI 8.4 to 10.1)9.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 8.4 to 10.1n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024. The rise from 2000 to 2008 was +2.1 (95% CI 1.3 to 3.0)+2.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 1.3 to 3.0n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2008SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points, with +1.3 (95% CI 0.9 to 1.7)+1.3 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 0.9 to 1.7n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2004SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 2004 and +0.8 (95% CI 0.2 to 1.4)+0.8 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 0.2 to 1.4n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2004 to 2008SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 2008. The change from 2012 to 2016 was +3.2 (95% CI 2.7 to 3.8)+3.2 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 2.7 to 3.8n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2012 to 2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points, the largest between consecutive elections. The changes from 2008 to 2012, 2016 to 2020 and 2020 to 2024 were +0.1 (95% CI −0.2 to 0.3)+0.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI −0.2 to 0.3n3,112Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2008 to 2012SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., +0.1 (95% CI −0.3 to 0.5)+0.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI −0.3 to 0.5n3,105Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2020SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and −0.1 (95% CI −0.3 to 0.0)−0.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI −0.3 to 0.0n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2020 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., and each interval includes zero. A slope unit of 1010 points of college shareEstimanddefinition of the slope unitWeightsnone (full counts)Populationunit of the slopePeriod2000 to 2024SourceCounty demographics panel points is close to one standard deviation of college share across counties in 2024, 10.310.3 standard deviation, percentage points of college shareEstimandstandard deviation of county college share about its meann3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panel points counting each county once and 11.811.8 standard deviation, percentage points of college shareEstimandstandard deviation of county college share about its meann3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panel vote-weighted.

    Figure 1

    Vote-weighted, the slope of the vote on county college share rose from 3.9 to 9.2 points per 10 points of college share, in steps before 2008 and between 2012 and 2016

    Slope of a county's Democratic share of the two-party presidential vote on the share of its residents 25 and older with a bachelor's degree or more, 2000 to 2024, vote-weighted and county-equal

    Points of Democratic two-party share per 10 points of college share

    Show data
    Data for Vote-weighted, the slope of the vote on county college share rose from 3.9 to 9.2 points per 10 points of college share, in steps before 2008 and between 2012 and 2016
    conf_highconf_lowestimatemeasuren_unitsweightingyear
    5.142.453.87Slope3,114Vote-weighted2000
    6.383.905.18Slope3,113Vote-weighted2004
    7.024.735.94Slope3,113Vote-weighted2008
    7.004.895.99Slope3,112Vote-weighted2012
    10.268.239.23Slope3,113Vote-weighted2016
    10.308.529.35Slope3,105Vote-weighted2020
    10.128.429.22Slope3,103Vote-weighted2024
    2.990.0891.56Slope3,114County-equal2000
    4.581.963.30Slope3,113County-equal2004
    6.154.065.11Slope3,113County-equal2008
    6.363.935.15Slope3,112County-equal2012
    9.366.487.95Slope3,113County-equal2016
    10.037.318.73Slope3,105County-equal2020
    9.797.248.59Slope3,103County-equal2024
    0.3650.1890.282Correlation3,114Vote-weighted2000
    0.4530.2970.378Correlation3,113Vote-weighted2004
    0.4910.3570.427Correlation3,113Vote-weighted2008
    0.4780.3510.415Correlation3,112Vote-weighted2012
    0.6370.5280.577Correlation3,113Vote-weighted2016
    0.6930.5940.637Correlation3,105Vote-weighted2020
    0.7130.6210.663Correlation3,103Vote-weighted2024
    0.1960.0050.099Correlation3,114County-equal2000
    0.3040.1180.211Correlation3,113County-equal2004
    0.3950.2310.310Correlation3,113County-equal2008
    0.3980.2190.305Correlation3,112County-equal2012
    0.5640.3550.460Correlation3,113County-equal2016
    0.6260.4270.530Correlation3,105County-equal2020
    0.6560.4520.559Correlation3,103County-equal2024
    The slope rises from 2000 to 2008, stays level to 2012 and rises again to 2016. Vote-weighted it does not change after 2016, although adjusted versions of the slope rise by +0.4 to +1.1 points in those years (see the sensitivity figure); counting each county once, it is below the vote-weighted slope in every election and the gap between the two is smaller in 2024 than in 2000. The line plots, for each election, the least-squares slope of the county Democratic share on county college share, per 10 points of college share; the solid line counts counties by their two-party votes and the dashed line counts each county once; bands are 95% intervals from resampling states.Note: Election returns are full counts of votes, so they carry no sampling error. The 95% intervals come from a bootstrap that resamples whole states, so they show how much a number depends on which states are in the set. They do not include the sampling error of the county college shares, which are census and American Community Survey estimates interpolated between years.Note: 3,114 counties in 2000 and 3,103 in 2024; Alaska and counties without college share are left out, and in 2024 the votes of Kansas City are added to Jackson County, Missouri. See Methods.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and county demographics (decennial census and American Community Survey estimates interpolated to election years); weighted by two-party votes in the election or county-equal, as stated

    Counting each county once, the slope was 1.6 (95% CI 0.1 to 3.0)1.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 0.1 to 3.0n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000, with an interval that reaches almost to zero, and it rose by +1.7 (95% CI 1.2 to 2.2)+1.7 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 1.2 to 2.2n3,113Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2004SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and +1.8 (95% CI 1.1 to 2.6)+1.8 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 1.1 to 2.6n3,113Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2004 to 2008SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points in the next two elections. It did not change from 2008 to 2012 (0.0 (95% CI −0.4 to 0.4)0.0 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI −0.4 to 0.4n3,112Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2008 to 2012SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.), rose by +2.8 (95% CI 2.3 to 3.4)+2.8 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 2.3 to 3.4n3,113Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2012 to 2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 2016 and by a further +0.8 (95% CI 0.5 to 1.1)+0.8 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 0.5 to 1.1n3,105Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2020SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 2020, and did not change from 2020 to 2024 (−0.1 (95% CI −0.3 to 0.0)−0.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI −0.3 to 0.0n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2020 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.). Both weightings put the largest step between 2012 and 2016. The county-equal slope was still rising from 2016 to 2020, which is why its change from 2016 to 2024 has an interval above zero.

    The slope after 2016 depends on the version of the slope. Figure 2 gives the changes from 2000 to 2008, 2012 to 2016 and 2016 to 2024 in six versions: college share alone, race and ethnicity held fixed, within states, within states with race and ethnicity held fixed, and the slope on each county's 2000 or 2024 college share held fixed for every election. The rises from 2000 to 2008 and from 2012 to 2016 have intervals above zero in all six versions and in both weightings. The change from 2016 to 2024 includes zero only for college share alone, vote-weighted. In the five other versions it is between +0.4+0.4 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandlowest of the 2016 to 2024 changes in the slope across five adjusted versions (race and ethnicity held fixed, within states, both, 2000 college share held fixed, 2024 college share held fixed)Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panel and +1.1+1.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandhighest of the 2016 to 2024 changes in the slope across five adjusted versions (race and ethnicity held fixed, within states, both, 2000 college share held fixed, 2024 college share held fixed)Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panel points vote-weighted and between +0.7+0.7 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandlowest of the 2016 to 2024 changes in the slope across five adjusted versions (race and ethnicity held fixed, within states, both, 2000 college share held fixed, 2024 college share held fixed)Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panel and +1.8+1.8 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandhighest of the 2016 to 2024 changes in the slope across five adjusted versions (race and ethnicity held fixed, within states, both, 2000 college share held fixed, 2024 college share held fixed)Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panel counting each county once, each with an interval above zero. These data therefore date the rises to 2008 and to 2016, and show a smaller change after 2016 whose size depends on the version.

    Figure 2

    Vote-weighted, the slope rose from 2000 to 2008 and from 2012 to 2016 in all six versions, and from 2016 to 2024 it changed by +0.4 to +1.1 in the five adjusted versions and by 0.0 with college share alone

    Change in the slope of the county Democratic share on county college share between elections, six versions of the slope, vote-weighted and county-equal

    2000 to 2008

    2012 to 2016

    2016 to 2024

    Change in the slope (points of Democratic share per 10 points of college share)

    Show data
    Data for Vote-weighted, the slope rose from 2000 to 2008 and from 2012 to 2016 in all six versions, and from 2016 to 2024 it changed by +0.4 to +1.1 in the five adjusted versions and by 0.0 with college share alone
    conf_highconf_lowestimaten_unitsperiodversionweighting
    3.011.282.083,1132000 to 2008College share aloneVote-weighted
    3.782.753.243,1132012 to 2016College share aloneVote-weighted
    0.381−0.381−0.0153,1032016 to 2024College share aloneVote-weighted
    3.731.252.093,1132000 to 2008Race and ethnicity held fixedVote-weighted
    3.882.843.243,1132012 to 2016Race and ethnicity held fixedVote-weighted
    1.250.1770.6403,1032016 to 2024Race and ethnicity held fixedVote-weighted
    2.731.562.143,1132000 to 2008Within statesVote-weighted
    4.032.923.463,1132012 to 2016Within statesVote-weighted
    0.6640.0660.3553,1032016 to 2024Within statesVote-weighted
    3.412.082.673,1132000 to 2008Within states, race and ethnicity held fixedVote-weighted
    4.053.263.643,1132012 to 2016Within states, race and ethnicity held fixedVote-weighted
    1.460.7471.113,1032016 to 2024Within states, race and ethnicity held fixedVote-weighted
    3.071.412.173,1132000 to 20082000 college share held fixedVote-weighted
    4.563.513.983,1112012 to 20162000 college share held fixedVote-weighted
    1.050.3190.6713,1022016 to 20242000 college share held fixedVote-weighted
    2.370.9251.583,1032000 to 20082024 college share held fixedVote-weighted
    3.702.763.203,1052012 to 20162024 college share held fixedVote-weighted
    0.8350.1100.4423,1032016 to 20242024 college share held fixedVote-weighted
    4.482.563.553,1132000 to 2008College share aloneCounty-equal
    3.372.302.803,1132012 to 2016College share aloneCounty-equal
    1.010.2460.6413,1032016 to 2024College share aloneCounty-equal
    5.142.593.573,1132000 to 2008Race and ethnicity held fixedCounty-equal
    3.602.653.083,1132012 to 2016Race and ethnicity held fixedCounty-equal
    1.260.5640.9443,1032016 to 2024Race and ethnicity held fixedCounty-equal
    2.761.682.243,1132000 to 2008Within statesCounty-equal
    3.672.923.293,1132012 to 2016Within statesCounty-equal
    1.010.3160.6733,1032016 to 2024Within statesCounty-equal
    3.752.232.923,1132000 to 2008Within states, race and ethnicity held fixedCounty-equal
    3.712.953.333,1132012 to 2016Within states, race and ethnicity held fixedCounty-equal
    1.070.2800.6903,1032016 to 2024Within states, race and ethnicity held fixedCounty-equal
    4.993.124.093,1132000 to 20082000 college share held fixedCounty-equal
    4.523.363.923,1112012 to 20162000 college share held fixedCounty-equal
    2.241.421.833,1022016 to 20242000 college share held fixedCounty-equal
    3.652.343.023,1032000 to 20082024 college share held fixedCounty-equal
    3.302.252.753,1052012 to 20162024 college share held fixedCounty-equal
    1.941.261.603,1032016 to 20242024 college share held fixedCounty-equal
    The rises to 2008 and to 2016 are positive in every row, and the change after 2016 is smaller than either, with an interval that includes zero only for college share alone, vote-weighted. The six versions are college share alone, with race and ethnicity held fixed, within states, within states with race and ethnicity held fixed, and the slope on each county's 2000 or 2024 college share held fixed for every election; each dot is the change in the slope between two elections, with a 95% interval from resampling states.Note: Intervals resample whole states; see Methods. The 2000 and 2024 college share versions use each county's college share in that year for every election, so they leave out counties without a value in that year.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and county demographics (decennial census and American Community Survey estimates interpolated to election years); weighted by two-party votes in the election or county-equal, as stated

    The correlation does not follow the unadjusted slope after 2016. The vote-weighted correlation was +0.42 (95% CI 0.35 to 0.48)+0.42 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.35 to 0.48n3,112Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2012SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2012, +0.58 (95% CI 0.53 to 0.64)+0.58 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.53 to 0.64n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2016 and +0.66 (95% CI 0.62 to 0.71)+0.66 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.62 to 0.71n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024, a change of +0.09 (95% CI 0.07 to 0.10)+0.09 correlation coefficient, change between two electionsEstimandchange in the correlation of college share with the Democratic share, 2024 minus 2016Interval95% CI 0.07 to 0.10n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. from 2016 to 2024 (the correlation view of Figure 1). A correlation equals the slope per point of college share multiplied by the standard deviationStandard deviationA measure of how far values typically fall from their average. A larger standard deviation means the values are more spread out. of college share and divided by that of the Democratic share across counties. From 2016 to 2024 the vote-weighted standard deviation of college share rose from 11.211.2 standard deviation, percentage points of college shareEstimandstandard deviation of county college share about its meann3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016SourceCounty demographics panel to 11.811.8 standard deviation, percentage points of college shareEstimandstandard deviation of county college share about its meann3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panel points and that of the Democratic share fell from 17.917.9 standard deviation, percentage points of the Democratic share of the two-party voteEstimandstandard deviation of the county Democratic two-party share about its meann3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2016SourceCounty election returns to 16.416.4 standard deviation, percentage points of the Democratic share of the two-party voteEstimandstandard deviation of the county Democratic two-party share about its meann3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns, so the correlation rose while the unadjusted vote-weighted slope stayed level. The slope says how many points more Democratic a county with more graduates was, and the correlation also reflects how much counties differ on each measure.

    A study that combined survey and county data found that a county's share of residents with a bachelor's degree was among the strongest predictors of the change in its vote share between 2000 and 2020 [1]. The estimates here describe the level of the vote in each election and not a county's own change, which Chapter 12.7, Education, ethnicity and the county shift, 2016 to 2024 follows.

    The widening comes mostly from the least educated counties

    Figure 3 ranks the counties of each election into five equal-sized groups by college share. In 2000 the vote-weighted Democratic share was 43.7% (95% CI 42.2% to 45.5%)43.7% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 42.2% to 45.5%n623Weightstwo-party votes in the electionPopulationcounties in fifth 1 of 5 by college share, from least to mostPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in the least educated fifth and 53.0% (95% CI 49.4% to 56.0%)53.0% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 49.4% to 56.0%n623Weightstwo-party votes in the electionPopulationcounties in fifth 5 of 5 by college share, from least to mostPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in the most educated, a gap of 9.3 (95% CI 5.3 to 12.6)9.3 percentage points of the Democratic share of the two-party vote, most minus least educated fifthEstimandDemocratic share in the fifth of counties with the highest college share minus that in the fifth with the lowestInterval95% CI 5.3 to 12.6n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points, and the mean college share of the groups ran from 9.0%9.0% percent of residents 25 and olderEstimandmean county share with a bachelor's degree or more in the fifthn623Weightsnone; each county counted oncePopulationcounties in fifth 1 of 5 by college sharePeriod2000SourceCounty demographics panel to 29.1%29.1% percent of residents 25 and olderEstimandmean county share with a bachelor's degree or more in the fifthn623Weightsnone; each county counted oncePopulationcounties in fifth 5 of 5 by college sharePeriod2000SourceCounty demographics panel. In 2024 the two were 26.4% (95% CI 23.4% to 29.7%)26.4% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 23.4% to 29.7%n621Weightstwo-party votes in the electionPopulationcounties in fifth 1 of 5 by college share, from least to mostPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and 56.1% (95% CI 53.2% to 59.0%)56.1% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 53.2% to 59.0%n621Weightstwo-party votes in the electionPopulationcounties in fifth 5 of 5 by college share, from least to mostPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., a gap of 29.8 (95% CI 26.4 to 33.0)29.8 percentage points of the Democratic share of the two-party vote, most minus least educated fifthEstimandDemocratic share in the fifth of counties with the highest college share minus that in the fifth with the lowestInterval95% CI 26.4 to 33.0n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., with mean college shares from 13.8%13.8% percent of residents 25 and olderEstimandmean county share with a bachelor's degree or more in the fifthn621Weightsnone; each county counted oncePopulationcounties in fifth 1 of 5 by college sharePeriod2024SourceCounty demographics panel to 41.2%41.2% percent of residents 25 and olderEstimandmean county share with a bachelor's degree or more in the fifthn621Weightsnone; each county counted oncePopulationcounties in fifth 5 of 5 by college sharePeriod2024SourceCounty demographics panel. The least educated fifth changed by −17.4 (95% CI −20.7 to −14.0)−17.4 percentage points of the Democratic share of the two-party vote, changeEstimandchange in the Democratic share of a fifth of counties ranked by college share in each electionInterval95% CI −20.7 to −14.0n3,103Weightstwo-party votes in the electionPopulationcounties in fifth 1 by college share in each electionPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points between the two elections and the most educated by +3.1 (95% CI 0.5 to 5.5)+3.1 percentage points of the Democratic share of the two-party vote, changeEstimandchange in the Democratic share of a fifth of counties ranked by college share in each electionInterval95% CI 0.5 to 5.5n3,103Weightstwo-party votes in the electionPopulationcounties in fifth 5 by college share in each electionPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., so most of the change in the gap is in the least educated counties. Over the same years the vote-weighted Democratic share of all counties went from 50.3%50.3% percentage points of the Democratic share of the two-party voteEstimandmean county Democratic share of the two-party vote (the pooled share of the counties when vote-weighted)n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty election returns to 49.2%49.2% percentage points of the Democratic share of the two-party voteEstimandmean county Democratic share of the two-party vote (the pooled share of the counties when vote-weighted)n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns.

    Figure 3

    The most educated fifth of counties voted 29.8 points more Democratic than the least educated fifth in 2024, against 9.3 in 2000

    Democratic share of the two-party presidential vote against the share of residents 25 and older with a bachelor's degree or more, by county, 2024 and 2000, with the mean of five equal-sized groups of counties

    Show data
    Data for The most educated fifth of counties voted 29.8 points more Democratic than the least educated fifth in 2024, against 9.3 in 2000
    County (largest only)College share, 2024 (%)Democratic share, 2024 (%)College share, 2000 (%)Democratic share, 2000 (%)
    Maricopa, AZ36.748.225.944.6
    Pima, AZ36.857.726.754.2
    Alameda, CA52.278.034.974.2
    Contra Costa, CA45.269.635.061.3
    Fresno, CA24.347.717.544.8
    Los Angeles, CA36.067.024.966.2
    Orange, CA44.151.430.842.0
    Riverside, CA25.649.416.646.6
    Sacramento, CA34.160.224.852.1
    San Bernardino, CA23.348.915.949.2
    San Diego, CA43.258.629.547.9
    San Francisco, CA60.383.845.082.4
    San Mateo, CA53.776.039.067.5
    Santa Clara, CA56.570.840.563.8
    Ventura, CA36.157.826.949.5
    Arapahoe, CO46.560.337.045.8
    Denver, CO56.578.834.566.7
    El Paso, CO41.945.031.832.5
    Jefferson, CO52.060.036.545.7
    District Of Columbia, DC64.293.339.190.5
    Brevard, FL34.239.523.645.8
    Broward, FL36.758.624.568.5
    Duval, FL34.449.321.941.5
    Hillsborough, FL38.148.525.148.4
    Lee, FL31.235.721.140.9
    Miami-Dade, FL34.444.221.753.2
    Orange, FL39.356.926.151.0
    Palm Beach, FL40.550.427.763.8
    Pasco, FL29.237.313.150.4
    Pinellas, FL36.847.422.952.1
    Polk, FL23.539.614.945.4
    Volusia, FL28.439.017.654.2
    Cobb, GA50.857.539.838.2
    DeKalb, GA47.482.736.372.4
    Fulton, GA58.572.741.459.2
    Gwinnett, GA39.058.434.133.5
    Honolulu, HI38.561.027.958.0
    Cook, IL42.771.328.070.6
    DuPage, IL51.856.841.743.1
    Lake, IL47.760.538.648.7
    Will, IL37.550.825.548.7
    Marion, IN34.864.125.449.3
    Johnson, KS57.654.347.737.8
    Jefferson, KY37.058.424.850.8
    Anne Arundel, MD45.857.230.646.2
    Baltimore, MD42.362.630.654.7
    Montgomery, MD60.677.454.665.1
    Prince George's, MD36.788.527.281.2
    Essex, MA43.060.431.361.9
    Middlesex, MA59.570.243.667.0
    Norfolk, MA58.864.742.963.8
    Worcester, MA39.755.126.960.4
    Kent, MI40.852.725.839.1
    Macomb, MI27.743.017.651.2
    Oakland, MI51.055.438.250.6
    Wayne, MI27.665.017.270.4
    Hennepin, MN54.171.839.157.7
    Jackson, MO34.059.823.460.6
    St Louis County, MO47.461.935.452.7
    Clark, NV27.851.317.353.4
    Bergen, NJ53.151.738.257.0
    Essex, NJ39.172.827.573.5
    Middlesex, NJ45.654.133.062.4
    Monmouth, NJ51.344.234.652.4
    Ocean, NJ34.431.819.549.1
    Bernalillo, NM38.260.830.551.1
    Bronx, NY22.472.714.688.0
    Erie, NY38.054.824.560.0
    Kings, NY42.172.021.883.7
    Monroe, NY41.559.531.253.4
    Nassau, NY49.747.935.460.1
    New York, NY64.282.449.484.7
    Queens, NY36.162.324.377.4
    Suffolk, NY40.245.027.556.0
    Westchester, NY53.163.140.961.0
    Mecklenburg, NC49.266.737.148.6
    Wake, NC57.263.043.946.4
    Cuyahoga, OH36.565.925.165.2
    Franklin, OH42.164.331.850.5
    Hamilton, OH41.857.529.244.2
    Multnomah, OR48.982.230.769.2
    Allegheny, PA45.860.328.358.4
    Bucks, PA44.850.031.252.2
    Chester, PA57.157.342.545.0
    Delaware, PA41.962.030.056.0
    Montgomery, PA53.061.538.755.0
    Philadelphia, PA35.479.817.981.6
    Shelby, TN34.362.925.357.3
    Bexar, TX31.854.922.746.2
    Collin, TX56.244.347.325.1
    Dallas, TX35.761.327.046.1
    Denton, TX49.443.336.628.2
    Fort Bend, TX50.250.836.939.3
    Harris, TX34.252.826.944.1
    Montgomery, TX38.827.025.322.4
    Tarrant, TX34.847.426.637.7
    Travis, TX56.670.040.647.1
    Williamson, TX48.448.833.629.0
    Salt Lake, UT39.455.327.438.5
    Fairfax, VA65.068.054.849.3
    King, WA56.776.840.063.6
    Pierce, WA30.955.520.653.9
    Snohomish, WA38.259.824.454.2
    Dane, WI55.376.240.665.3
    Milwaukee, WI34.769.623.660.7
    n/a29.126.618.029.2
    n/a33.720.723.125.5
    n/a10.642.610.950.4
    n/a11.717.67.138.8
    n/a15.89.29.628.2
    n/a8.173.07.770.3
    n/a13.538.610.446.6
    n/a20.627.415.241.4
    n/a14.938.39.548.2
    n/a16.512.09.745.7
    n/a16.513.89.932.3
    n/a15.838.09.650.7
    n/a15.041.412.143.9
    n/a14.714.87.835.5
    n/a17.08.09.233.3
    n/a22.520.819.334.4
    n/a21.726.614.150.1
    n/a13.443.09.250.8
    n/a13.928.28.046.9
    n/a17.213.612.233.1
    n/a18.722.611.240.9
    n/a19.79.411.933.7
    n/a20.423.614.031.7
    n/a17.566.413.959.8
    n/a14.112.88.335.5
    n/a25.423.816.628.4
    n/a13.526.710.639.3
    n/a18.122.013.445.3
    n/a14.913.89.240.1
    n/a15.813.19.743.9
    n/a13.211.38.729.6
    n/a15.178.010.580.5
    n/a17.653.48.160.9
    n/a21.824.514.140.7
    n/a23.025.918.429.8
    n/a15.914.010.451.7
    n/a36.255.324.648.4
    n/a12.111.87.837.2
    n/a29.824.018.544.3
    n/a15.218.67.552.6
    n/a42.435.927.939.4
    n/a30.227.816.938.8
    n/a14.468.711.073.6
    n/a21.178.318.887.5
    n/a46.445.434.343.7
    n/a19.548.112.150.8
    n/a13.38.98.040.0
    n/a21.513.913.937.8
    n/a26.041.718.642.9
    n/a11.738.411.842.1
    n/a34.565.528.551.0
    n/a23.323.518.438.4
    n/a15.871.410.069.9
    n/a13.138.39.849.0
    n/a28.037.318.441.8
    n/a18.417.410.039.9
    n/a18.250.89.757.5
    n/a21.417.711.127.5
    n/a46.829.336.821.7
    n/a19.070.712.473.0
    n/a16.533.011.244.9
    n/a20.325.114.138.7
    n/a32.139.724.042.0
    n/a16.613.99.146.3
    n/a17.022.28.645.1
    n/a12.965.810.167.5
    n/a14.38.08.329.6
    n/a16.659.611.368.7
    n/a28.138.318.842.4
    n/a40.360.129.953.6
    n/a18.731.013.945.7
    n/a14.625.711.835.8
    n/a20.929.212.242.9
    n/a13.527.88.741.0
    n/a15.221.99.941.7
    n/a19.141.312.348.8
    n/a22.538.911.949.4
    n/a23.159.415.261.0
    n/a30.332.921.137.5
    n/a16.839.711.843.4
    n/a13.627.612.246.2
    n/a12.825.610.152.3
    n/a19.421.112.840.6
    n/a37.836.120.333.2
    n/a17.217.012.734.4
    n/a15.330.311.954.2
    n/a19.818.57.347.4
    n/a20.033.813.839.3
    n/a18.052.011.764.3
    n/a26.239.519.855.2
    n/a15.318.67.461.0
    n/a18.415.513.941.8
    n/a19.715.710.045.7
    n/a21.931.516.844.4
    n/a19.929.411.549.7
    n/a29.730.820.950.4
    n/a20.020.39.736.8
    n/a17.051.212.855.2
    n/a18.425.79.950.5
    n/a11.335.09.652.1
    n/a12.647.611.163.4
    n/a25.632.817.352.6
    n/a35.033.425.242.7
    n/a14.518.111.044.9
    n/a16.018.310.549.3
    n/a25.331.518.045.3
    n/a21.515.011.043.6
    n/a16.918.910.952.0
    n/a16.329.811.054.7
    n/a17.523.411.252.3
    n/a16.226.311.647.0
    n/a18.119.613.745.6
    n/a19.716.411.752.9
    n/a14.725.210.361.6
    n/a20.160.215.766.9
    n/a17.823.713.147.2
    n/a9.030.59.554.0
    n/a14.517.38.555.3
    n/a7.251.87.366.9
    n/a12.024.57.656.2
    n/a17.222.59.955.8
    n/a16.718.29.442.3
    n/a22.122.614.639.2
    n/a12.920.210.137.8
    n/a14.919.410.439.6
    n/a15.024.112.546.3
    n/a13.733.911.357.8
    n/a12.742.08.459.0
    n/a16.717.88.840.3
    n/a13.829.810.751.0
    n/a15.717.411.832.3
    n/a16.340.312.753.6
    n/a16.520.611.143.8
    n/a15.056.812.465.6
    n/a19.113.010.141.4
    n/a11.417.76.357.9
    n/a17.814.110.933.5
    n/a24.723.219.037.2
    n/a18.216.69.045.6
    n/a37.961.428.155.0
    n/a12.117.510.653.0
    n/a14.250.49.659.4
    n/a29.429.516.440.6
    n/a13.012.78.437.6
    n/a15.013.48.432.0
    n/a23.330.816.639.8
    n/a9.318.69.249.8
    n/a13.918.010.346.7
    n/a15.818.59.843.8
    n/a19.133.014.942.0
    n/a15.019.311.547.9
    n/a35.246.624.543.1
    n/a20.418.715.538.8
    n/a16.233.48.065.4
    n/a14.919.110.948.7
    n/a41.166.328.248.5
    n/a22.435.716.640.3
    n/a31.948.421.840.7
    n/a23.935.617.140.1
    n/a16.435.510.632.5
    n/a21.441.611.040.8
    n/a40.043.826.538.4
    n/a15.132.110.730.1
    n/a33.164.923.051.7
    n/a16.149.610.355.3
    n/a31.548.517.136.0
    n/a18.839.213.537.4
    n/a13.638.210.440.3
    n/a17.849.312.155.2
    n/a13.122.310.729.6
    n/a17.939.312.036.5
    n/a61.182.851.369.4
    n/a31.839.220.237.3
    n/a25.264.020.257.5
    n/a15.347.811.046.5
    n/a24.825.912.424.1
    n/a32.260.628.943.8
    n/a28.265.422.560.7
    n/a39.868.026.457.7
    n/a41.556.326.140.5
    n/a43.245.630.337.8
    n/a22.541.317.535.3
    n/a24.156.517.156.6
    n/a22.749.514.549.4
    n/a39.855.626.743.9
    n/a36.863.829.450.7
    n/a43.978.334.269.2
    n/a23.831.316.631.7
    n/a25.837.617.231.5
    n/a23.340.117.734.1
    n/a28.861.821.459.3
    n/a38.573.928.564.9
    n/a19.644.414.145.7
    n/a20.933.915.332.7
    n/a19.328.611.332.9
    n/a22.445.115.536.6
    n/a16.639.411.537.9
    n/a24.638.816.141.5
    n/a44.868.834.159.4
    n/a17.736.710.336.1
    n/a29.054.617.453.2
    n/a32.444.427.046.2
    n/a40.942.829.032.4
    n/a24.714.014.024.2
    n/a15.730.111.541.7
    n/a64.278.752.457.9
    n/a60.364.7n/an/a
    n/a47.057.024.339.2
    n/a26.110.414.217.9
    n/a57.558.638.849.3
    n/a25.740.814.449.7
    n/a24.957.612.867.7
    n/a12.425.511.937.4
    n/a43.731.526.725.9
    n/a26.631.817.628.1
    n/a28.123.213.528.3
    n/a62.046.451.932.6
    n/a50.962.542.648.6
    n/a36.123.926.627.4
    n/a22.030.313.534.8
    n/a33.651.123.840.1
    n/a52.856.631.252.2
    n/a42.749.634.539.3
    n/a64.765.143.649.4
    n/a47.743.634.937.3
    n/a26.445.616.150.5
    n/a31.321.419.920.2
    n/a20.112.116.122.5
    n/a20.415.315.424.1
    n/a49.558.519.555.1
    n/a49.359.636.444.0
    n/a53.059.139.542.5
    n/a19.343.116.255.8
    n/a26.117.113.223.8
    n/a20.121.114.629.3
    n/a32.337.522.032.3
    n/a51.743.231.236.4
    n/a19.417.912.524.2
    n/a31.339.221.029.3
    n/a28.733.318.730.4
    n/a17.825.113.533.5
    n/a18.236.415.442.1
    n/a56.560.836.835.5
    n/a41.641.530.339.4
    n/a31.517.919.926.4
    n/a62.872.657.161.7
    n/a19.324.211.931.0
    n/a25.647.418.355.9
    n/a24.717.619.519.9
    n/a34.538.118.835.4
    n/a54.264.342.548.5
    n/a31.351.919.651.5
    n/a52.166.743.741.5
    n/a64.975.448.560.5
    n/a27.221.313.430.5
    n/a55.069.248.354.1
    n/a37.931.831.729.8
    n/a24.911.914.320.3
    n/a33.239.221.638.5
    n/a25.316.615.525.5
    n/an/an/a39.954.8
    n/an/an/a29.663.4
    n/an/an/a27.551.6
    n/an/an/a33.859.7
    n/an/an/a27.661.7
    n/an/an/a26.259.5
    n/an/an/a32.857.6
    n/an/an/a19.059.0
    n/a27.551.018.648.6
    n/a39.766.529.561.9
    n/a33.044.516.646.2
    n/a48.860.638.758.1
    n/a17.813.38.229.9
    n/a29.626.117.732.8
    n/a15.221.28.436.2
    n/a13.616.07.742.9
    n/a27.032.917.645.6
    n/a20.626.913.246.1
    n/a28.330.220.125.9
    n/a40.333.427.933.1
    n/a16.824.710.939.1
    n/a10.828.48.443.8
    n/a10.714.66.840.4
    n/a29.740.121.035.9
    n/a31.335.821.252.4
    n/a27.927.912.445.5
    n/a21.365.512.967.1
    n/a15.315.79.436.7
    n/a13.423.29.843.9
    n/a23.522.810.140.3
    n/a11.030.37.344.5
    n/a12.121.78.438.3
    n/a12.230.78.240.6
    n/a21.131.312.751.6
    n/a20.529.513.641.2
    n/a12.59.78.830.3
    n/a33.836.223.140.8
    n/a15.126.812.842.9
    n/a19.240.616.955.1
    n/a9.911.87.232.1
    n/a27.837.616.642.2
    n/a48.961.041.761.1
    n/a16.524.710.644.0
    n/a16.516.37.443.6
    n/a13.935.510.249.8
    n/a35.438.220.845.9
    n/a23.434.113.744.8
    n/a37.134.326.343.9
    n/a38.640.825.550.7
    n/a35.626.318.929.8
    n/a34.828.524.224.6
    n/a15.823.08.947.6
    n/a29.849.315.751.8
    n/a14.525.99.447.4
    n/a50.934.233.133.0
    n/a26.745.415.154.5
    n/a32.024.322.926.1
    n/a40.840.827.446.7
    n/a42.748.231.043.9
    n/a35.331.112.244.3
    n/a16.619.410.533.7
    n/a16.520.08.939.5
    n/a11.115.77.537.6
    n/a23.327.615.746.0
    n/a35.020.916.231.7
    n/a11.517.19.235.9
    n/a11.218.78.434.7
    n/a8.923.06.940.1
    n/a10.413.46.632.2
    n/a16.740.110.759.2
    n/a24.448.916.249.4
    n/a18.610.88.627.6
    n/a21.229.610.931.6
    n/a24.324.414.133.8
    n/a13.033.99.548.4
    n/a15.715.09.437.6
    n/a26.361.321.350.9
    n/a19.822.212.534.3
    n/a14.58.76.230.6
    n/a15.836.611.346.6
    n/a36.131.719.331.0
    n/a32.535.425.438.2
    n/a15.545.39.552.4
    n/a14.427.38.641.6
    n/a13.056.211.759.0
    n/a24.632.116.036.3
    n/a16.026.210.239.1
    n/a23.829.316.534.9
    n/a24.222.113.831.3
    n/a12.321.86.436.4
    n/a37.759.125.049.8
    n/a29.641.725.050.4
    n/a12.017.87.742.8
    n/a41.130.327.024.4
    n/a47.469.339.856.1
    n/a23.453.810.164.7
    n/a22.584.816.666.7
    n/a9.124.210.442.8
    n/a13.927.410.038.4
    n/a15.524.811.433.4
    n/a38.537.132.025.2
    n/a17.926.78.141.8
    n/a36.232.920.629.8
    n/a14.525.06.843.2
    n/a15.337.012.840.8
    n/a18.416.510.932.8
    n/a36.417.218.125.7
    n/a15.138.012.144.8
    n/a17.225.011.640.1
    n/a14.846.19.654.5
    n/a22.870.617.857.6
    n/a30.865.819.237.1
    n/a14.744.312.645.6
    n/a15.78.98.430.7
    n/a27.225.313.630.6
    n/a16.328.29.843.7
    n/a13.527.910.145.9
    n/a14.528.79.039.8
    n/a24.217.510.433.4
    n/a49.948.436.228.9
    n/a25.928.915.838.8
    n/a57.133.334.619.4
    n/a17.813.610.335.8
    n/a24.918.612.931.1
    n/a17.08.06.524.6
    n/a32.436.923.835.2
    n/a18.818.110.633.7
    n/a16.430.810.641.1
    n/a37.835.517.641.8
    n/a24.617.415.826.6
    n/a27.328.018.727.7
    n/a10.367.79.878.5
    n/a17.012.79.035.8
    n/a38.026.821.134.4
    n/a20.822.513.542.9

    Showing the first 500 of 3115 rows.

    In 2024 the group means rise with college share, from 26.4% in the least educated fifth of counties to 56.1% in the most educated, vote-weighted, and from 26.3% to 48.3% counting each county once. Each dot is a county. The lines join the Democratic share of five equal-sized groups of counties ranked by college share in that election (solid, vote-weighted, which pools the votes of the group; dashed, counting each county once), each drawn at the group's median college share; whiskers are 95% intervals from resampling states. The gap between the highest and lowest group is 22.0 points counting each county once. Names are given for counties with 300,000 or more two-party votes in 2024.Note: Intervals resample whole states; see Methods. The number of counties in each group is in the tooltip.Note: Groups are formed within each election, so a county can move between groups from one election to the next. Counties without college share are left out of that election. Dots are all the same size, so the vote-weighted lines, which pool votes, can lie away from the middle of the dots; the file names only the largest counties, to keep it small.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and county demographics (decennial census and American Community Survey estimates interpolated to election years); weighted by two-party votes in the election or county-equal, as stated

    The groups are equal in counties and not in votes. The most educated fifth cast 64.4%64.4% percent of two-party votes in the analysis countiesEstimandshare of the two-party votes cast in the fifthn621Weightsnone (full counts)Populationcounties in fifth 5 by college sharePeriod2024SourceCounty election returns of the two-party votes in 2024 and the least educated 3.6%3.6% percent of two-party votes in the analysis countiesEstimandshare of the two-party votes cast in the fifthn621Weightsnone (full counts)Populationcounties in fifth 1 by college sharePeriod2024SourceCounty election returns, so the vote-weighted figures describe mainly the most educated counties.

    Counting each county once, the 2000 picture is different. The least educated fifth (44.0%44.0% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 42.1% to 46.1%n623Weightsnone; each county counted oncePopulationcounties in fifth 1 of 5 by college share, from least to mostPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.) and the most educated (44.6%44.6% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 41.7% to 47.5%n623Weightsnone; each county counted oncePopulationcounties in fifth 5 of 5 by college share, from least to mostPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.) had similar Democratic shares, and the gap of 0.6 (95% CI −2.6 to 3.9)0.6 percentage points of the Democratic share of the two-party vote, most minus least educated fifthEstimandDemocratic share in the fifth of counties with the highest college share minus that in the fifth with the lowestInterval95% CI −2.6 to 3.9n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points has an interval that includes zero. Both were above the three groups between them, so the 2000 relationship was U-shaped, with the lowest mean, 37.9%37.9% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 34.9% to 41.6%n623Weightsnone; each county counted oncePopulationcounties in fifth 4 of 5 by college share, from least to mostPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., in the fourth fifth. In 2024 the county-equal means were 26.3%26.3% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 23.0% to 30.1%n621Weightsnone; each county counted oncePopulationcounties in fifth 1 of 5 by college share, from least to mostPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., 26.5%26.5% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 24.2% to 29.1%n620Weightsnone; each county counted oncePopulationcounties in fifth 2 of 5 by college share, from least to mostPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and 27.4%27.4% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 24.8% to 30.4%n621Weightsnone; each county counted oncePopulationcounties in fifth 3 of 5 by college share, from least to mostPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in the three least educated fifths, 32.1%32.1% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 29.0% to 35.4%n620Weightsnone; each county counted oncePopulationcounties in fifth 4 of 5 by college share, from least to mostPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in the fourth and 48.3%48.3% percentage points of the Democratic share of the two-party voteEstimandDemocratic share of the two-party vote in a fifth of counties ranked by college share in that electionInterval95% CI 45.3% to 51.1%n621Weightsnone; each county counted oncePopulationcounties in fifth 5 of 5 by college share, from least to mostPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in the most educated, and the gap was 22.0 (95% CI 17.6 to 26.0)22.0 percentage points of the Democratic share of the two-party vote, most minus least educated fifthEstimandDemocratic share in the fifth of counties with the highest college share minus that in the fifth with the lowestInterval95% CI 17.6 to 26.0n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points. The gap changed by +21.4 (95% CI 17.9 to 24.9)+21.4 percentage points of the Democratic share of the two-party vote, change in the gapEstimandchange in the most-minus-least-educated fifth gapInterval95% CI 17.9 to 24.9n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points from 2000 to 2024 counting each county once and by +20.5 (95% CI 16.7 to 24.4)+20.5 percentage points of the Democratic share of the two-party vote, change in the gapEstimandchange in the most-minus-least-educated fifth gapInterval95% CI 16.7 to 24.4n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. vote-weighted.

    The 2000 slope depends on the weighting, and large counties differ from the rest

    In 2000 the vote-weighted slope was 3.9 (95% CI 2.5 to 5.1)3.9 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 2.5 to 5.1n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and the county-equal slope 1.6 (95% CI 0.1 to 3.0)1.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 0.1 to 3.0n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., and in 2024 they were 9.2 (95% CI 8.4 to 10.1)9.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 8.4 to 10.1n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and 8.6 (95% CI 7.2 to 9.8)8.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 7.2 to 9.8n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.. Figure 4 separates large counties, those with 250,000250,000 two-party votes in the electionEstimanddefinition of a large countyWeightsnone (full counts)Populationthreshold for a large countyPeriod2000 to 2024SourceCounty election returns or more two-party votes in the election, from the rest. In 2000 the 7676 countiesEstimandcount of counties in the size classn76Weightsnone (full counts)Populationlarge countiesPeriod2000SourceCounty election returns large counties cast 35.1%35.1% percent of two-party votes in the analysis countiesEstimandshare of the two-party votes cast in the size classn76Weightsnone (full counts)Populationlarge countiesPeriod2000SourceCounty election returns of the votes. They had a mean college share of 30.2%30.2% percent of residents 25 and olderEstimandmean county share with a bachelor's degree or moren76Weightsnone; each county counted oncePopulationlarge countiesPeriod2000SourceCounty demographics panel against 16.2%16.2% percent of residents 25 and olderEstimandmean county share with a bachelor's degree or moren3,038Weightsnone; each county counted oncePopulationother countiesPeriod2000SourceCounty demographics panel in the other counties, and a mean Democratic share of 57.7%57.7% percentage points of the Democratic share of the two-party voteEstimandmean county Democratic share of the two-party voten76Weightsnone; each county counted oncePopulationlarge countiesPeriod2000SourceCounty election returns against 40.8%40.8% percentage points of the Democratic share of the two-party voteEstimandmean county Democratic share of the two-party voten3,038Weightsnone; each county counted oncePopulationother countiesPeriod2000SourceCounty election returns.

    Figure 4

    Counting each county once, the 2000 slope inside large counties and inside other counties each had an interval that includes zero, and by 2024 both were above zero

    Slope of the county Democratic share on county college share inside large counties and inside other counties, with the all-county slope, 2000 to 2024, counting each county once or by two-party votes

    Points of Democratic two-party share per 10 points of college share

    Show data
    Data for Counting each county once, the 2000 slope inside large counties and inside other counties each had an interval that includes zero, and by 2024 both were above zero
    basisconf_highconf_lowestimatemeasuren_unitsseriesyear
    County-equal2.990.0891.56Slope3,114All counties2000
    County-equal4.581.963.30Slope3,113All counties2004
    County-equal6.154.065.11Slope3,113All counties2008
    County-equal6.363.935.15Slope3,112All counties2012
    County-equal9.366.487.95Slope3,113All counties2016
    County-equal10.037.318.73Slope3,105All counties2020
    County-equal9.797.248.59Slope3,103All counties2024
    County-equal4.10−2.890.564Slope76Large counties2000
    County-equal5.94−1.092.37Slope96Large counties2004
    County-equal5.61−0.2232.70Slope110Large counties2008
    County-equal5.30−0.7882.39Slope107Large counties2012
    County-equal7.311.684.69Slope107Large counties2016
    County-equal7.432.705.15Slope139Large counties2020
    County-equal8.053.425.85Slope135Large counties2024
    County-equal1.99−0.7630.688Slope3,038Other counties2000
    County-equal3.620.9602.38Slope3,017Other counties2004
    County-equal5.303.104.23Slope3,003Other counties2008
    County-equal5.432.844.19Slope3,005Other counties2012
    County-equal8.395.266.89Slope3,006Other counties2016
    County-equal9.176.127.72Slope2,966Other counties2020
    County-equal9.136.237.76Slope2,968Other counties2024
    County-equal0.1960.0050.099Correlation3,114All counties2000
    County-equal0.3040.1180.211Correlation3,113All counties2004
    County-equal0.3950.2310.310Correlation3,113All counties2008
    County-equal0.3980.2190.305Correlation3,112All counties2012
    County-equal0.5640.3550.460Correlation3,113All counties2016
    County-equal0.6260.4270.530Correlation3,105All counties2020
    County-equal0.6560.4520.559Correlation3,103All counties2024
    County-equal0.325−0.2320.043Correlation76Large counties2000
    County-equal0.431−0.0840.178Correlation96Large counties2004
    County-equal0.418−0.0170.203Correlation110Large counties2008
    County-equal0.402−0.0560.177Correlation107Large counties2012
    County-equal0.5490.1200.346Correlation107Large counties2016
    County-equal0.5840.2040.403Correlation139Large counties2020
    County-equal0.6360.2580.465Correlation135Large counties2024
    County-equal0.133−0.0430.043Correlation3,038Other counties2000
    County-equal0.2420.0560.148Correlation3,017Other counties2004
    County-equal0.3330.1670.247Correlation3,003Other counties2008
    County-equal0.3380.1500.240Correlation3,005Other counties2012
    County-equal0.5090.2770.393Correlation3,006Other counties2016
    County-equal0.5750.3450.462Correlation2,966Other counties2020
    County-equal0.6100.3790.497Correlation2,968Other counties2024
    Vote-weighted5.142.453.87Slope3,114All counties2000
    Vote-weighted6.383.905.18Slope3,113All counties2004
    Vote-weighted7.024.735.94Slope3,113All counties2008
    Vote-weighted7.004.895.99Slope3,112All counties2012
    Vote-weighted10.268.239.23Slope3,113All counties2016
    Vote-weighted10.308.529.35Slope3,105All counties2020
    Vote-weighted10.128.429.22Slope3,103All counties2024
    Vote-weighted2.96−3.390.150Slope76Large counties2000
    Vote-weighted4.92−1.552.00Slope96Large counties2004
    Vote-weighted4.46−0.6422.14Slope110Large counties2008
    Vote-weighted4.10−0.9291.85Slope107Large counties2012
    Vote-weighted6.231.743.99Slope107Large counties2016
    Vote-weighted6.982.914.82Slope139Large counties2020
    Vote-weighted7.703.565.54Slope135Large counties2024
    Vote-weighted3.751.262.42Slope3,038Other counties2000
    Vote-weighted4.912.293.54Slope3,017Other counties2004
    Vote-weighted5.443.454.43Slope3,003Other counties2008
    Vote-weighted5.323.304.28Slope3,005Other counties2012
    Vote-weighted8.706.687.66Slope3,006Other counties2016
    Vote-weighted9.147.358.25Slope2,966Other counties2020
    Vote-weighted9.227.528.35Slope2,968Other counties2024
    Vote-weighted0.3650.1890.282Correlation3,114All counties2000
    Vote-weighted0.4530.2970.378Correlation3,113All counties2004
    Vote-weighted0.4910.3570.427Correlation3,113All counties2008
    Vote-weighted0.4780.3510.415Correlation3,112All counties2012
    Vote-weighted0.6370.5280.577Correlation3,113All counties2016
    Vote-weighted0.6930.5940.637Correlation3,105All counties2020
    Vote-weighted0.7130.6210.663Correlation3,103All counties2024
    Vote-weighted0.213−0.2570.011Correlation76Large counties2000
    Vote-weighted0.343−0.1180.145Correlation96Large counties2004
    Vote-weighted0.333−0.0500.161Correlation110Large counties2008
    Vote-weighted0.309−0.0670.137Correlation107Large counties2012
    Vote-weighted0.4640.1250.297Correlation107Large counties2016
    Vote-weighted0.5610.2360.396Correlation139Large counties2020
    Vote-weighted0.6110.2920.458Correlation135Large counties2024
    Vote-weighted0.2870.0990.192Correlation3,038Other counties2000
    Vote-weighted0.3670.1750.272Correlation3,017Other counties2004
    Vote-weighted0.4020.2490.325Correlation3,003Other counties2008
    Vote-weighted0.3840.2290.307Correlation3,005Other counties2012
    Vote-weighted0.5860.4320.511Correlation3,006Other counties2016
    Vote-weighted0.6500.5120.586Correlation2,966Other counties2020
    Vote-weighted0.6730.5370.612Correlation2,968Other counties2024
    The two class lines start near zero in 2000 and rise, while the county-equal slope across all counties was 1.6 in 2000. The vote-weighted slope across all counties, 3.9, is higher: the 76 large counties had a mean college share of 30.2% against 16.2% in the other counties, and a mean Democratic share of 57.7% against 40.8%. Large counties are those with 250,000 or more two-party votes in the election; bands, drawn for the two classes, are 95% intervals from resampling states.Note: Intervals resample whole states; see Methods.Note: Large counties: 250,000 or more two-party votes in the election; 76 counties in 2000 and 135 in 2024.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and county demographics (decennial census and American Community Survey estimates interpolated to election years); weighted by two-party votes in the election or county-equal, as stated

    Inside each class, counting each county once, the 2000 slope was 0.6 (95% CI −2.9 to 4.1)0.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the size classInterval95% CI −2.9 to 4.1n76Weightsnone; each county counted oncePopulationlarge countiesPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. among the large counties and 0.7 (95% CI −0.8 to 2.0)0.7 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the size classInterval95% CI −0.8 to 2.0n3,038Weightsnone; each county counted oncePopulationother countiesPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. among the others, and both intervals include zero. Vote-weighted, the 2000 slope was 2.4 (95% CI 1.3 to 3.8)2.4 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the size classInterval95% CI 1.3 to 3.8n3,038Weightstwo-party votes in the electionPopulationother countiesPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. among the other counties, an interval above zero, and 0.2 (95% CI −3.4 to 3.0)0.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the size classInterval95% CI −3.4 to 3.0n76Weightstwo-party votes in the electionPopulationlarge countiesPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. among the large counties, an interval that includes zero. Splitting the vote-weighted covariance of college share and the Democratic share, 61% (95% CI 43% to 87%)61% percent of the vote-weighted covariance of college share and the Democratic shareEstimandshare of the vote-weighted covariance of college share and the Democratic share that lies between the two size classesInterval95% CI 43% to 87%n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. lies between the two classes and 39% (95% CI 13% to 57%)39% percent of the vote-weighted covariance of college share and the Democratic shareEstimandshare of the vote-weighted covariance of college share and the Democratic share that lies inside the size classesInterval95% CI 13% to 57%n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. inside them in 2000, so the difference between large and other counties and a gradient inside the other counties both contribute to the vote-weighted slope of 2000, and the intervals are wide. In 2024 the two shares are 38% (95% CI 28% to 48%)38% percent of the vote-weighted covariance of college share and the Democratic shareEstimandshare of the vote-weighted covariance of college share and the Democratic share that lies between the two size classesInterval95% CI 28% to 48%n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and 62% (95% CI 52% to 72%)62% percent of the vote-weighted covariance of college share and the Democratic shareEstimandshare of the vote-weighted covariance of college share and the Democratic share that lies inside the size classesInterval95% CI 52% to 72%n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.. By 2024 the county-equal slope was 5.9 (95% CI 3.4 to 8.0)5.9 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the size classInterval95% CI 3.4 to 8.0n135Weightsnone; each county counted oncePopulationlarge countiesPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. inside the large counties and 7.8 (95% CI 6.2 to 9.1)7.8 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the size classInterval95% CI 6.2 to 9.1n2,968Weightsnone; each county counted oncePopulationother countiesPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. inside the others, and the correlations, +0.04 (95% CI −0.23 to 0.32)+0.04 correlation coefficientEstimandcorrelation of college share with the Democratic share inside the size classInterval95% CI −0.23 to 0.32n76Weightsnone; each county counted oncePopulationlarge countiesPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and +0.04 (95% CI −0.04 to 0.13)+0.04 correlation coefficientEstimandcorrelation of college share with the Democratic share inside the size classInterval95% CI −0.04 to 0.13n3,038Weightsnone; each county counted oncePopulationother countiesPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000, were +0.46 (95% CI 0.26 to 0.64)+0.46 correlation coefficientEstimandcorrelation of college share with the Democratic share inside the size classInterval95% CI 0.26 to 0.64n135Weightsnone; each county counted oncePopulationlarge countiesPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and +0.50 (95% CI 0.38 to 0.61)+0.50 correlation coefficientEstimandcorrelation of college share with the Democratic share inside the size classInterval95% CI 0.38 to 0.61n2,968Weightsnone; each county counted oncePopulationother countiesPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.. The mean Democratic share of the large counties changed by −0.1−0.1 percentage points of the Democratic share of the two-party vote, changeEstimandchange in the mean county Democratic share, counting each county oncen135Weightsnone; each county counted oncePopulationlarge countiesPeriod2000 to 2024SourceCounty election returns points between 2000 and 2024 and that of the other counties by −9.9−9.9 percentage points of the Democratic share of the two-party vote, changeEstimandchange in the mean county Democratic share, counting each county oncen2,968Weightsnone; each county counted oncePopulationother countiesPeriod2000 to 2024SourceCounty election returns, although the large counties are a different set in 2000 and in 2024, 7676 countiesEstimandcount of counties in the size classn76Weightsnone (full counts)Populationlarge countiesPeriod2000SourceCounty election returns counties and 135135 countiesEstimandcount of counties in the size classn135Weightsnone (full counts)Populationlarge countiesPeriod2024SourceCounty election returns. The change in means is in the other counties.

    With race and ethnicity held fixed the slope is lower and still rises from near zero

    Counties with more graduates differ in racial and ethnic composition from other counties, so the slope of college share alone may carry part of the relation between composition and the vote. Figure 5 adds the non-Hispanic white, Black, Hispanic and Asian shares of residents to the regression, so the slope compares counties with the same four shares, each entering as a straight line. Vote-weighted, the slope goes from 3.9 (95% CI 2.5 to 5.1)3.9 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 2.5 to 5.1n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 2.0 (95% CI −0.1 to 3.4)2.0 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixedInterval95% CI −0.1 to 3.4n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000 and from 9.2 (95% CI 8.4 to 10.1)9.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 8.4 to 10.1n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 7.8 (95% CI 7.0 to 8.9)7.8 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixedInterval95% CI 7.0 to 8.9n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024, and the 2000 interval includes zero. Counting each county once, it goes from 1.6 (95% CI 0.1 to 3.0)1.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 0.1 to 3.0n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 0.8 (95% CI −1.5 to 2.3)0.8 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixedInterval95% CI −1.5 to 2.3n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000 and from 8.6 (95% CI 7.2 to 9.8)8.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 7.2 to 9.8n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 8.5 (95% CI 7.5 to 9.3)8.5 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixedInterval95% CI 7.5 to 9.3n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024.

    Figure 5

    Vote-weighted, the slope with state and race held fixed went from 0.5 in 2000 to 7.6 in 2024

    Slope of the county Democratic share on county college share alone, with race and ethnicity held fixed, within states, and within states with race and ethnicity held fixed, 2000 to 2024

    Points of Democratic two-party share per 10 points of college share

    Show data
    Data for Vote-weighted, the slope with state and race held fixed went from 0.5 in 2000 to 7.6 in 2024
    conf_highconf_lowestimaten_unitsspecificationweightingyear
    5.142.453.873,114College share aloneVote-weighted2000
    6.383.905.183,113College share aloneVote-weighted2004
    7.024.735.943,113College share aloneVote-weighted2008
    7.004.895.993,112College share aloneVote-weighted2012
    10.268.239.233,113College share aloneVote-weighted2016
    10.308.529.353,105College share aloneVote-weighted2020
    10.128.429.223,103College share aloneVote-weighted2024
    3.39−0.1202.043,114Race and ethnicity held fixedVote-weighted2000
    4.721.713.463,113Race and ethnicity held fixedVote-weighted2004
    5.302.824.133,113Race and ethnicity held fixedVote-weighted2008
    5.112.623.963,112Race and ethnicity held fixedVote-weighted2012
    8.326.267.203,113Race and ethnicity held fixedVote-weighted2016
    8.947.007.853,105Race and ethnicity held fixedVote-weighted2020
    8.947.037.843,103Race and ethnicity held fixedVote-weighted2024
    3.621.242.413,114Within statesVote-weighted2000
    5.182.833.973,113Within statesVote-weighted2004
    5.493.554.553,113Within statesVote-weighted2008
    5.373.494.453,112Within statesVote-weighted2012
    8.857.057.903,113Within statesVote-weighted2016
    9.137.438.193,105Within statesVote-weighted2020
    9.097.588.263,103Within statesVote-weighted2024
    1.55−0.6980.4963,114Within states, race and ethnicity held fixedVote-weighted2000
    3.210.9692.153,113Within states, race and ethnicity held fixedVote-weighted2004
    4.172.193.173,113Within states, race and ethnicity held fixedVote-weighted2008
    3.771.882.803,112Within states, race and ethnicity held fixedVote-weighted2012
    7.405.646.443,113Within states, race and ethnicity held fixedVote-weighted2016
    8.236.597.353,105Within states, race and ethnicity held fixedVote-weighted2020
    8.356.877.553,103Within states, race and ethnicity held fixedVote-weighted2024
    5.142.453.873,1142000 college share held fixedVote-weighted2000
    4.802.553.813,1032024 college share held fixedVote-weighted2000
    6.423.845.143,1132000 college share held fixedVote-weighted2004
    5.703.634.763,1032024 college share held fixedVote-weighted2004
    7.274.806.043,1132000 college share held fixedVote-weighted2008
    6.274.365.393,1032024 college share held fixedVote-weighted2008
    7.445.036.213,1112000 college share held fixedVote-weighted2012
    6.454.585.583,1042024 college share held fixedVote-weighted2012
    11.618.9610.193,1112000 college share held fixedVote-weighted2016
    9.797.808.773,1052024 college share held fixedVote-weighted2016
    12.119.6510.793,1032000 college share held fixedVote-weighted2020
    10.128.359.173,1052024 college share held fixedVote-weighted2020
    12.209.7110.863,1022000 college share held fixedVote-weighted2024
    10.128.429.223,1032024 college share held fixedVote-weighted2024
    2.990.0891.563,114College share aloneCounty-equal2000
    4.581.963.303,113College share aloneCounty-equal2004
    6.154.065.113,113College share aloneCounty-equal2008
    6.363.935.153,112College share aloneCounty-equal2012
    9.366.487.953,113College share aloneCounty-equal2016
    10.037.318.733,105College share aloneCounty-equal2020
    9.797.248.593,103College share aloneCounty-equal2024
    2.28−1.510.8143,114Race and ethnicity held fixedCounty-equal2000
    4.100.7632.673,113Race and ethnicity held fixedCounty-equal2004
    5.633.014.383,113Race and ethnicity held fixedCounty-equal2008
    5.663.104.443,112Race and ethnicity held fixedCounty-equal2012
    8.606.437.523,113Race and ethnicity held fixedCounty-equal2016
    9.487.508.513,105Race and ethnicity held fixedCounty-equal2020
    9.357.498.463,103Race and ethnicity held fixedCounty-equal2024
    2.56−0.2031.193,114Within statesCounty-equal2000
    3.731.282.473,113Within statesCounty-equal2004
    4.532.393.433,113Within statesCounty-equal2008
    4.462.243.323,112Within statesCounty-equal2012
    7.835.426.613,113Within statesCounty-equal2016
    8.406.307.333,105Within statesCounty-equal2020
    8.326.257.283,103Within statesCounty-equal2024
    2.13−0.7160.7313,114Within states, race and ethnicity held fixedCounty-equal2000
    3.780.9172.303,113Within states, race and ethnicity held fixedCounty-equal2004
    4.942.533.653,113Within states, race and ethnicity held fixedCounty-equal2008
    4.812.483.583,112Within states, race and ethnicity held fixedCounty-equal2012
    8.075.916.923,113Within states, race and ethnicity held fixedCounty-equal2016
    8.726.917.773,105Within states, race and ethnicity held fixedCounty-equal2020
    8.496.797.613,103Within states, race and ethnicity held fixedCounty-equal2024
    2.990.0891.563,1142000 college share held fixedCounty-equal2000
    2.13−0.1781.013,1032024 college share held fixedCounty-equal2000
    4.752.033.443,1132000 college share held fixedCounty-equal2004
    3.481.192.393,1032024 college share held fixedCounty-equal2004
    6.754.545.653,1132000 college share held fixedCounty-equal2008
    5.003.044.033,1032024 college share held fixedCounty-equal2008
    7.344.736.073,1112000 college share held fixedCounty-equal2012
    5.373.024.243,1042024 college share held fixedCounty-equal2012
    11.468.419.993,1112000 college share held fixedCounty-equal2016
    8.415.516.993,1052024 college share held fixedCounty-equal2016
    12.9010.1411.553,1032000 college share held fixedCounty-equal2020
    9.616.878.303,1052024 college share held fixedCounty-equal2020
    13.1010.5111.823,1022000 college share held fixedCounty-equal2024
    9.797.248.593,1032024 college share held fixedCounty-equal2024
    With state and race held fixed the slope is 0.5 in 2000, an interval that includes zero, and 7.6 in 2024, against 3.9 and 9.2 for college share alone. Vote-weighted, race held fixed lowers the slope by 1.8 in 2000 and 1.4 in 2024, and the difference between those reductions, −0.5, has an interval that includes zero. Each line is the slope of the county Democratic share on county college share per 10 points; race and ethnicity held fixed means the non-Hispanic white, Black, Hispanic and Asian shares of residents are in the regression, and within states means each state has its own baseline. Bands, drawn for two lines, are 95% intervals from resampling states.Note: Intervals resample whole states; see Methods.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and county demographics (decennial census and American Community Survey estimates interpolated to election years); weighted by two-party votes in the election or county-equal, as stated

    Vote-weighted, the adjustment lowers the slope by 1.8 (95% CI 0.8 to 3.5)1.8 points of Democratic two-party share per 10 points of college share, reductionEstimandslope of college share alone minus the slope of the specification (slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed)Interval95% CI 0.8 to 3.5n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points in 2000 and 1.4 (95% CI 0.2 to 2.3)1.4 points of Democratic two-party share per 10 points of college share, reductionEstimandslope of college share alone minus the slope of the specification (slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed)Interval95% CI 0.2 to 2.3n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024, and the difference between the two reductions is −0.5 (95% CI −2.3 to 0.3)−0.5 points of Democratic two-party share per 10 points of college share, change in the reductionEstimandreduction in the slope in 2024 minus the reduction in 2000 (slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed)Interval95% CI −2.3 to 0.3n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., an interval that includes zero, so these data do not show that race matters less to the slope in 2024 than in 2000. Counting each county once, the reduction goes from 0.70.7 points of Democratic two-party share per 10 points of college share, reductionEstimandslope of college share alone minus the slope of the specification (slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed)Interval95% CI −0.2 to 2.5n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. to 0.10.1 points of Democratic two-party share per 10 points of college share, reductionEstimandslope of college share alone minus the slope of the specification (slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed)Interval95% CI −1.0 to 1.2n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and the difference is −0.6 (95% CI −2.2 to 0.0)−0.6 points of Democratic two-party share per 10 points of college share, change in the reductionEstimandreduction in the slope in 2024 minus the reduction in 2000 (slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed)Interval95% CI −2.2 to 0.0n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., an interval whose upper end is at zero. With race held fixed, the rise from 2000 to 2024 is +5.8 (95% CI 4.7 to 8.2)+5.8 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed, later election minus earlier electionInterval95% CI 4.7 to 8.2n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points vote-weighted and +7.6 (95% CI 6.5 to 9.6)+7.6 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed, later election minus earlier electionInterval95% CI 6.5 to 9.6n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. county-equal, and the largest change between consecutive elections is again from 2012 to 2016, at +3.2 (95% CI 2.8 to 3.9)+3.2 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed, later election minus earlier electionInterval95% CI 2.8 to 3.9n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2012 to 2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and +3.1 (95% CI 2.6 to 3.6)+3.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share with the non-Hispanic white, Black, Hispanic and Asian shares of residents held fixed, later election minus earlier electionInterval95% CI 2.6 to 3.6n3,113Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2012 to 2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates..

    A straight-line control can miss a pattern that differs between kinds of counties, so Figure 6 fits the slope separately inside four groups of counties by non-Hispanic white share: under 50%50% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-off for the county groupsPeriod2000 to 2024SourceCounty demographics panel, 50%50% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-off for the county groupsPeriod2000 to 2024SourceCounty demographics panel to under 75%75% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-off for the county groupsPeriod2000 to 2024SourceCounty demographics panel, 75%75% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-off for the county groupsPeriod2000 to 2024SourceCounty demographics panel to under 90%90% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-off for the county groupsPeriod2000 to 2024SourceCounty demographics panel, and 90%90% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-off for the county groupsPeriod2000 to 2024SourceCounty demographics panel or more. In the most white group the county-equal slope was 0.1 (95% CI −2.2 to 2.1)0.1 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the group of countiesInterval95% CI −2.2 to 2.1n1,502Weightsnone; each county counted oncePopulationcounties in the group 90% or more whitePeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000, an interval that includes zero, and 8.7 (95% CI 5.2 to 11.8)8.7 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the group of countiesInterval95% CI 5.2 to 11.8n787Weightsnone; each county counted oncePopulationcounties in the group 90% or more whitePeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024, a change of +8.6 (95% CI 5.4 to 11.3)+8.6 points of Democratic two-party share per 10 points of college share, changeEstimandchange in the slope inside the group, 2024 minus 2000Interval95% CI 5.4 to 11.3n787Weightsnone; each county counted oncePopulationcounties in the group 90% or more whitePeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points. The vote-weighted change is +9.7 (95% CI 6.8 to 12.1)+9.7 points of Democratic two-party share per 10 points of college share, changeEstimandchange in the slope inside the group, 2024 minus 2000Interval95% CI 6.8 to 12.1n787Weightstwo-party votes in the electionPopulationcounties in the group 90% or more whitePeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in that group and +1.8 (95% CI −0.6 to 6.7)+1.8 points of Democratic two-party share per 10 points of college share, changeEstimandchange in the slope inside the group, 2024 minus 2000Interval95% CI −0.6 to 6.7n404Weightstwo-party votes in the electionPopulationcounties in the group Under 50% whitePeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in the group under 50%50% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-off for the county groupsPeriod2000 to 2024SourceCounty demographics panel white, where the 2024 slope was 6.2 (95% CI 4.5 to 7.3)6.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share inside the group of countiesInterval95% CI 4.5 to 7.3n404Weightstwo-party votes in the electionPopulationcounties in the group Under 50% whitePeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.. The change is smallest in the least white group in the point estimates in both weightings, and its interval includes zero in both (+3.7 (95% CI −0.2 to 7.2)+3.7 points of Democratic two-party share per 10 points of college share, changeEstimandchange in the slope inside the group, 2024 minus 2000Interval95% CI −0.2 to 7.2n404Weightsnone; each county counted oncePopulationcounties in the group Under 50% whitePeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. county-equal). The interval of that change nearly touches the interval of the change in the most white group vote-weighted and overlaps it county-equal, so these data do not rank the two changes. In 2024 the slope is above zero, with the whole interval above, in all four groups in both weightings. From the least white group to the most white, the groups held 253253 countiesEstimandcount of counties in the groupn253Weightsnone (full counts)Populationcounties in the group Under 50% whitePeriod2000SourceCounty demographics panel, 635635 countiesEstimandcount of counties in the groupn635Weightsnone (full counts)Populationcounties in the group 50 to under 75% whitePeriod2000SourceCounty demographics panel, 724724 countiesEstimandcount of counties in the groupn724Weightsnone (full counts)Populationcounties in the group 75 to under 90% whitePeriod2000SourceCounty demographics panel and 1,5021,502 countiesEstimandcount of counties in the groupn1,502Weightsnone (full counts)Populationcounties in the group 90% or more whitePeriod2000SourceCounty demographics panel counties in 2000 and 404404 countiesEstimandcount of counties in the groupn404Weightsnone (full counts)Populationcounties in the group Under 50% whitePeriod2024SourceCounty demographics panel, 855855 countiesEstimandcount of counties in the groupn855Weightsnone (full counts)Populationcounties in the group 50 to under 75% whitePeriod2024SourceCounty demographics panel, 1,0571,057 countiesEstimandcount of counties in the groupn1,057Weightsnone (full counts)Populationcounties in the group 75 to under 90% whitePeriod2024SourceCounty demographics panel and 787787 countiesEstimandcount of counties in the groupn787Weightsnone (full counts)Populationcounties in the group 90% or more whitePeriod2024SourceCounty demographics panel in 2024.

    Figure 6

    The slope was higher in 2024 than in 2000 in all four groups of counties by white share, with the largest point change among counties 90% white or more

    Slope of the county Democratic share on county college share inside four groups of counties by non-Hispanic white share of residents at the election, 2000 to 2024, vote-weighted and county-equal

    Points of Democratic two-party share per 10 points of college share

    Under 50% white50 to under 75% white75 to under 90% white90% or more white
    Show data
    Data for The slope was higher in 2024 than in 2000 in all four groups of counties by white share, with the largest point change among counties 90% white or more
    conf_highconf_lowestimategroupn_unitsweightingyear
    7.27−1.224.38Under 50% white253Vote-weighted2000
    8.392.105.89Under 50% white284Vote-weighted2004
    7.142.705.13Under 50% white309Vote-weighted2008
    6.192.044.51Under 50% white339Vote-weighted2012
    7.784.816.57Under 50% white360Vote-weighted2016
    7.514.446.19Under 50% white386Vote-weighted2020
    7.284.456.16Under 50% white404Vote-weighted2024
    6.130.4883.59Under 50% white253County-equal2000
    7.842.915.49Under 50% white284County-equal2004
    7.673.615.63Under 50% white309County-equal2008
    7.652.875.42Under 50% white339County-equal2012
    9.375.137.44Under 50% white360County-equal2016
    9.964.847.56Under 50% white386County-equal2020
    9.364.447.29Under 50% white404County-equal2024
    5.171.953.4950 to under 75% white635Vote-weighted2000
    6.143.514.7150 to under 75% white661Vote-weighted2004
    7.394.455.9350 to under 75% white710Vote-weighted2008
    7.184.235.7150 to under 75% white740Vote-weighted2012
    9.667.138.3650 to under 75% white767Vote-weighted2016
    10.187.678.9950 to under 75% white816Vote-weighted2020
    10.437.709.1450 to under 75% white855Vote-weighted2024
    4.461.122.8550 to under 75% white635County-equal2000
    6.243.434.7850 to under 75% white661County-equal2004
    8.395.857.0150 to under 75% white710County-equal2008
    7.595.076.2350 to under 75% white740County-equal2012
    9.937.098.4550 to under 75% white767County-equal2016
    10.657.939.2550 to under 75% white816County-equal2020
    10.217.718.9550 to under 75% white855County-equal2024
    5.441.413.5175 to under 90% white724Vote-weighted2000
    6.733.135.0875 to under 90% white784Vote-weighted2004
    6.693.245.1175 to under 90% white808Vote-weighted2008
    6.602.955.0075 to under 90% white841Vote-weighted2012
    9.536.508.1175 to under 90% white900Vote-weighted2016
    9.617.258.4375 to under 90% white998Vote-weighted2020
    9.487.068.2575 to under 90% white1,057Vote-weighted2024
    4.291.382.8575 to under 90% white724County-equal2000
    6.183.234.8275 to under 90% white784County-equal2004
    7.584.936.3675 to under 90% white808County-equal2008
    7.634.926.3575 to under 90% white841County-equal2012
    10.138.099.1275 to under 90% white900County-equal2016
    10.628.829.7175 to under 90% white998County-equal2020
    10.358.559.4175 to under 90% white1,057County-equal2024
    2.94−0.5081.1590% or more white1,502Vote-weighted2000
    4.760.3712.4290% or more white1,384Vote-weighted2004
    7.602.715.1090% or more white1,286Vote-weighted2008
    8.693.906.3490% or more white1,192Vote-weighted2012
    12.097.509.8890% or more white1,086Vote-weighted2016
    13.217.5410.3890% or more white905Vote-weighted2020
    13.777.6910.8490% or more white787Vote-weighted2024
    2.13−2.220.06890% or more white1,502County-equal2000
    3.73−0.6331.5490% or more white1,384County-equal2004
    6.451.804.1390% or more white1,286County-equal2008
    7.402.034.7190% or more white1,192County-equal2012
    10.785.358.0390% or more white1,086County-equal2016
    12.055.788.9890% or more white905County-equal2020
    11.815.248.7190% or more white787County-equal2024
    Every panel ends higher than it starts, so the slope did not rise only where counties are mostly white. The change from 2000 to 2024 in the vote-weighted slope is +9.7 in the counties 90% white or more and +1.8 in the counties under 50% white, where the interval includes zero; the two intervals nearly touch vote-weighted and overlap county-equal, so these data do not rank the changes. Each panel plots the slope per 10 points of college share inside one group; the solid line counts counties by their two-party votes and the dashed line counts each county once; bands are 95% intervals from resampling states.Note: Intervals resample whole states; see Methods. Groups are formed from the non-Hispanic white share at each election, so a county can change group.Note: Counties per group in 2000 and 2024: under 50% white, 253 and 404; 50% to under 75%, 635 and 855; 75% to under 90%, 724 and 1,057; 90% or more, 1,502 and 787.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and county demographics (decennial census and American Community Survey estimates interpolated to election years); weighted by two-party votes in the election or county-equal, as stated

    Inside states the slope is lower, and by 2024 it is positive in every state vote-weighted

    A second comparison asks the question inside states, which share candidates, campaigns and election law, by giving each state its own baseline (fixed effectsFixed effectsRegression terms that give each group, such as each country or each year, its own baseline. The other coefficients then compare respondents within the same group.). Vote-weighted, the within-state slope is 2.4 (95% CI 1.2 to 3.6)2.4 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states (state fixed effects)Interval95% CI 1.2 to 3.6n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000 and 8.3 (95% CI 7.6 to 9.1)8.3 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states (state fixed effects)Interval95% CI 7.6 to 9.1n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024, against 3.93.9 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 2.5 to 5.1n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and 9.29.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college shareInterval95% CI 8.4 to 10.1n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. across all counties. Counting each county once it is 1.2 (95% CI −0.2 to 2.6)1.2 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states (state fixed effects)Interval95% CI −0.2 to 2.6n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and 7.3 (95% CI 6.2 to 8.3)7.3 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states (state fixed effects)Interval95% CI 6.2 to 8.3n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates.. The within-state slope is lower than the slope across all counties in every election in both weightings. Vote-weighted, the reduction is 1.51.5 points of Democratic two-party share per 10 points of college share, reductionEstimandslope of college share alone minus the slope of the specification (slope of the Democratic share on college share within states (state fixed effects))Interval95% CI 0.5 to 2.4n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points in 2000 and 1.01.0 points of Democratic two-party share per 10 points of college share, reductionEstimandslope of college share alone minus the slope of the specification (slope of the Democratic share on college share within states (state fixed effects))Interval95% CI 0.5 to 1.5n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024, a difference of −0.5 (95% CI −1.2 to 0.2)−0.5 points of Democratic two-party share per 10 points of college share, change in the reductionEstimandreduction in the slope in 2024 minus the reduction in 2000 (slope of the Democratic share on college share within states (state fixed effects))Interval95% CI −1.2 to 0.2n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., an interval that includes zero. With state and race and ethnicity held fixed together, the vote-weighted slope is 0.5 (95% CI −0.7 to 1.6)0.5 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states with the four race and ethnicity shares held fixedInterval95% CI −0.7 to 1.6n3,114Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000, an interval that includes zero, and 7.6 (95% CI 6.9 to 8.4)7.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states with the four race and ethnicity shares held fixedInterval95% CI 6.9 to 8.4n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2024, a change of +7.1 (95% CI 6.1 to 8.2)+7.1 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share within states with the four race and ethnicity shares held fixed, later election minus earlier electionInterval95% CI 6.1 to 8.2n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2000 to 2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points, and counting each county once the two are 0.7 (95% CI −0.7 to 2.1)0.7 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states with the four race and ethnicity shares held fixedInterval95% CI −0.7 to 2.1n3,114Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. and 7.6 (95% CI 6.8 to 8.5)7.6 points of Democratic two-party share per 10 points of college shareEstimandslope of the Democratic share on college share within states with the four race and ethnicity shares held fixedInterval95% CI 6.8 to 8.5n3,103Weightsnone; each county counted oncePopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates..

    Figure 7 gives the slope inside each of the 3939 statesEstimandcount of states in the within-state comparisonWeightsnone (full counts)Populationstates with at least 20 counties in 2000 and 2024, Alaska excludedPeriod2000 and 2024SourceCounty election returns states with at least 2020 countiesEstimanddefinition of the state setWeightsnone (full counts)Populationminimum number of counties in a state in the within-state comparisonPeriod2000 and 2024SourceCounty election returns counties in both 2000 and 2024, counting counties by their votes. It was positive in 2929 statesEstimandcount of states with a positive slopen39Weightstwo-party votes in the electionPopulationstates with at least 20 countiesPeriod2000SourceCounty demographics panel of them in 2000 and in 3939 statesEstimandcount of states with a positive slopen39Weightstwo-party votes in the electionPopulationstates with at least 20 countiesPeriod2024SourceCounty demographics panel in 2024. Counting each county once the numbers are 2626 statesEstimandcount of states with a positive slopen39Weightsnone; each county counted oncePopulationstates with at least 20 countiesPeriod2000SourceCounty demographics panel and 3636 statesEstimandcount of states with a positive slopen39Weightsnone; each county counted oncePopulationstates with at least 20 countiesPeriod2024SourceCounty demographics panel, and the interval lies above zero in 1212 statesEstimandcount of states whose interval lies above zeron39Weightsnone; each county counted oncePopulationstates with at least 20 countiesPeriod2000SourceCounty demographics panel states in 2000 and 3030 statesEstimandcount of states whose interval lies above zeron39Weightsnone; each county counted oncePopulationstates with at least 20 countiesPeriod2024SourceCounty demographics panel in 2024. In 2000 the county-equal slope was negative in 1313 statesEstimandcount of states with a negative slopen39Weightsnone; each county counted oncePopulationstates with at least 20 countiesPeriod2000SourceCounty demographics panel states: Alabama, Arkansas, Georgia, Indiana, Mississippi, New Jersey, Oklahoma, South Carolina, South Dakota, Tennessee, Texas, West Virginia and Wisconsin. In 44 statesEstimandcount of states whose interval lies below zeron39Weightsnone; each county counted oncePopulationstates with at least 20 countiesPeriod2000SourceCounty demographics panel of them, Oklahoma, South Carolina, Tennessee and Texas, the whole interval lies below zero. In 2024 only Mississippi, South Carolina and South Dakota have a negative county-equal slope, and each interval includes zero.

    Figure 7

    The vote-weighted slope was positive in every one of the 39 states in 2024, against 29 in 2000

    Slope of the county Democratic share on county college share inside each state with at least 20 counties, 2000 and 2024, vote-weighted, with 95% intervals

    • 2000
    • 2024

    Slope: points of Democratic two-party share per 10 points of college share, counties inside the state

    Show data
    Data for The vote-weighted slope was positive in every one of the 39 states in 2024, against 29 in 2000
    hi_2000hi_2024lo_2000lo_2024n_unitsslope_2000slope_2024stateweighting
    4.2512.24−6.921.3467−1.346.79AlabamaVote-weighted
    13.4819.68−12.630.982750.42510.33ArkansasVote-weighted
    11.819.903.165.28587.497.59CaliforniaVote-weighted
    7.6014.37−3.081.89642.268.13ColoradoVote-weighted
    9.3612.150.6014.17674.988.16FloridaVote-weighted
    9.9012.46−2.744.171593.588.32GeorgiaVote-weighted
    8.0513.604.306.56446.1710.08IdahoVote-weighted
    12.6021.38−8.533.361022.0312.37IllinoisVote-weighted
    7.3516.66−9.34−3.4892−0.9956.59IndianaVote-weighted
    5.8612.391.356.24993.619.32IowaVote-weighted
    5.7413.32−5.261.081050.2417.20KansasVote-weighted
    7.2817.28−1.387.641202.9512.46KentuckyVote-weighted
    13.9423.63−8.09−0.026642.9211.80LouisianaVote-weighted
    9.8112.59−7.36−1.57241.225.51MarylandVote-weighted
    4.5710.15−6.201.3483−0.8125.74MichiganVote-weighted
    7.0114.691.688.83874.3411.76MinnesotaVote-weighted
    7.056.86−8.14−2.3582−0.5452.25MississippiVote-weighted
    5.4616.311.8110.601153.6413.45MissouriVote-weighted
    8.0714.05−4.053.24562.018.64MontanaVote-weighted
    11.3020.283.989.52937.6414.90NebraskaVote-weighted
    1.356.84−9.13−4.9621−3.890.942New JerseyVote-weighted
    11.8315.93−2.575.33334.6310.63New MexicoVote-weighted
    15.4614.71−8.09−1.12623.686.79New YorkVote-weighted
    5.9211.890.9596.471003.449.18North CarolinaVote-weighted
    6.7220.34−2.70−1.48532.019.43North DakotaVote-weighted
    7.2716.51−2.124.59882.5710.55OhioVote-weighted
    −2.0618.15−6.058.2177−4.0613.18OklahomaVote-weighted
    19.4519.070.8587.393610.1513.23OregonVote-weighted
    7.9012.77−5.456.25671.239.51PennsylvaniaVote-weighted
    8.058.64−10.77−0.77046−1.363.93South CarolinaVote-weighted
    8.0110.23−2.921.17652.555.70South DakotaVote-weighted
    7.2218.79−5.77−0.834950.7268.98TennesseeVote-weighted
    2.8411.09−5.232.76254−1.206.92TexasVote-weighted
    11.4718.18−13.16−4.0829−0.8447.05UtahVote-weighted
    3.368.880.2385.811331.807.35VirginiaVote-weighted
    14.8611.701.248.65398.0510.17WashingtonVote-weighted
    3.5912.71−4.176.2855−0.2919.50West VirginiaVote-weighted
    15.0017.54−11.78−1.89721.617.82WisconsinVote-weighted
    8.0415.783.216.61235.6211.19WyomingVote-weighted
    1.136.80−8.94−3.2267−3.901.79AlabamaCounty-equal
    1.3210.30−9.57−0.58875−4.124.85ArkansasCounty-equal
    11.9511.526.928.36589.439.94CaliforniaCounty-equal
    6.8912.042.328.40644.6110.22ColoradoCounty-equal
    6.0710.240.2255.77673.158.00FloridaCounty-equal
    1.727.03−5.221.36159−1.754.20GeorgiaCounty-equal
    10.1913.082.404.49446.298.79IdahoCounty-equal
    3.0013.11−1.049.001020.98011.06IllinoisCounty-equal
    3.5311.24−4.225.4192−0.3428.33IndianaCounty-equal
    5.3510.77−0.2036.34992.578.55IowaCounty-equal
    7.1312.340.1824.141053.658.24KansasCounty-equal
    4.5111.15−0.8477.641201.839.39KentuckyCounty-equal
    6.1814.00−5.89−0.050640.1476.97LouisianaCounty-equal
    7.0511.240.0356.30243.548.77MarylandCounty-equal
    4.129.52−2.236.37830.9447.94MichiganCounty-equal
    4.4911.92−0.1477.39872.179.65MinnesotaCounty-equal
    2.143.05−6.67−5.0382−2.26−0.988MississippiCounty-equal
    4.6513.540.4358.001152.5410.77MissouriCounty-equal
    7.9712.09−1.911.13563.036.61MontanaCounty-equal
    6.6211.06−3.30−0.415931.665.32NebraskaCounty-equal
    0.5355.35−6.43−0.76221−2.952.29New JerseyCounty-equal
    11.6012.67−7.842.12331.887.40New MexicoCounty-equal
    10.0111.393.236.91626.629.15New YorkCounty-equal
    4.178.09−2.252.661000.9585.38North CarolinaCounty-equal
    7.7211.88−4.56−9.19531.581.34North DakotaCounty-equal
    4.7912.17−1.846.20881.489.19OhioCounty-equal
    −3.509.28−13.851.9577−8.685.62OklahomaCounty-equal
    16.5616.262.8210.08369.6913.17OregonCounty-equal
    8.3213.511.938.75675.1211.13PennsylvaniaCounty-equal
    −1.122.56−14.77−6.2646−7.94−1.85South CarolinaCounty-equal
    6.135.60−7.79−7.6165−0.828−1.01South DakotaCounty-equal
    −0.87711.72−6.733.3795−3.807.55TennesseeCounty-equal
    −3.146.14−9.291.66254−6.213.90TexasCounty-equal
    10.2912.33−7.185.39291.558.86UtahCounty-equal
    2.699.15−0.4376.471331.137.81VirginiaCounty-equal
    11.4713.431.378.37396.4210.90WashingtonCounty-equal
    4.1010.47−6.097.2855−0.9958.87West VirginiaCounty-equal
    5.2810.39−5.932.5672−0.3276.48WisconsinCounty-equal
    8.1016.434.456.79236.2711.61WyomingCounty-equal
    The filled dot (2024) is to the right of the open dot (2000) in almost every state, and by 2024 the slope is positive in every state shown. Each row is the slope inside one state, counting the counties in it by their two-party votes; whiskers are 95% intervals that treat counties as independent, so they are narrower than an interval that allowed for the similarity of nearby counties. The slope is above zero, with the whole interval above, in 28 states in 2024 and in 12 in 2000.Note: Intervals are HC3 robust and treat counties as independent; see Methods. States with fewer than 20 counties in either year, and Alaska, are left out.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and county demographics (decennial census and American Community Survey estimates interpolated to election years); weighted by two-party votes in the election or county-equal, as stated

    Democratic counties at or below the median college share became rare

    Figure 8 classes each county in each election by two facts: whether its college share is above the median of that election's counties, and whether Democrats won a majority of its two-party vote. The median county college share was 14.5%14.5% percent of residents 25 and olderEstimandmedian county share with a bachelor's degree or moren3,114Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty demographics panel in 2000 and 22.1%22.1% percent of residents 25 and olderEstimandmedian county share with a bachelor's degree or moren3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panel in 2024. Counties at or below the median with a Democratic majority numbered 325325 countiesEstimandcount of counties at or below the median college share, Democratic majorityn3,114Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty election returns in 2000 and 9393 countiesEstimandcount of counties at or below the median college share, Democratic majorityn3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns in 2024. Counties above the median with a Democratic majority numbered 350350 countiesEstimandcount of counties above the median college share, Democratic majorityn3,114Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty election returns and 352352 countiesEstimandcount of counties above the median college share, Democratic majorityn3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns, and counties at or below the median with a Republican majority went from 1,2321,232 countiesEstimandcount of counties at or below the median college share, Republican majorityn3,114Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty election returns to 1,4591,459 countiesEstimandcount of counties at or below the median college share, Republican majorityn3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns. The two classes in which college share and the majority line up were 50.8%50.8% percent of countiesEstimandshare of counties that are above the median college share with a Democratic majority or at or below it with a Republican majorityn3,114Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty election returns of counties and 55.1%55.1% percent of two-party votesEstimandshare of two-party votes cast in the counties that are above the median college share with a Democratic majority or at or below it with a Republican majorityn3,114Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2000SourceCounty election returns of two-party votes in 2000, and 58.4%58.4% percent of countiesEstimandshare of counties that are above the median college share with a Democratic majority or at or below it with a Republican majorityn3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns and 60.8%60.8% percent of two-party votesEstimandshare of two-party votes cast in the counties that are above the median college share with a Democratic majority or at or below it with a Republican majorityn3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns in 2024.

    Figure 8

    Counties at or below the median college share with a Democratic majority fell from 325 in 2000 to 93 in 2024

    Each county by whether its college share is above the median of that election's counties and whether Democrats won a majority of its two-party vote, 2000 to 2024, Alaska excluded

    Loading county values

    College share relative to the median county, and winner

    No data (3,143 counties)

    Select a class to show only those places.

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

    • Alaska draws as no data. Its election returns are reported by election district, and the district codes cannot be matched to boroughs and census areas.
    Show data
    Data for Counties at or below the median college share with a Democratic majority fell from 325 in 2000 to 93 in 2024
    ElectionCountyCollege share (%)Democratic share of two-party vote (%)Two-party votes
    2000Pasco, FL13.150.4138,183
    2000Hidalgo, TX12.961.699,691
    2000Trumbull, OH14.562.592,297
    2000Jefferson, MO12.151.275,382
    2000Muskegon, MI13.955.867,893
    2000Hernando, FL12.751.663,306
    2000Cameron, TX13.454.461,014
    2000Monroe, MI14.352.260,495
    2000Cambria, PA13.752.058,309
    2000Bay, MI14.256.150,401
    2004Hidalgo, TX14.055.0113,300
    2004Trumbull, OH15.261.9107,650
    2004Muskegon, MI15.055.679,584
    2004Fayette, PA12.553.854,165
    2004Wyandotte, KS13.466.152,842
    2004Cumberland, NJ12.753.452,237
    2004Ashtabula, OH11.853.445,098
    2004Laporte, IN15.150.242,030
    2004Cowlitz, WA14.151.641,806
    2004Webb, TX15.157.141,407
    2008Stanislaus, CA15.850.9157,776
    2008Hidalgo, TX15.169.5129,929
    2008Trumbull, OH16.061.5104,309
    2008Jefferson, MO16.051.3104,271
    2008Muskegon, MI16.264.982,966
    2008St. Clair, MI14.951.479,213
    2008Cameron, TX14.664.575,151
    2008Merced, CA12.254.262,735
    2008Cumberland, NJ13.661.057,279
    2008Madison, IN16.553.356,555
    2012Stanislaus, CA16.451.4151,183
    2012Hidalgo, TX16.571.1137,834
    2012Trumbull, OH17.261.799,951
    2012Cameron, TX15.865.776,074
    2012Muskegon, MI17.459.075,320
    2012Merced, CA13.154.560,586
    2012Cumberland, NJ14.262.254,713
    2012Wyandotte, KS15.868.949,798
    2012Webb, TX17.277.248,675
    2012Oswego, NY17.054.143,495
    2016San Joaquin, CA18.756.8191,036
    2016Hidalgo, TX18.471.0167,451
    2016Stanislaus, CA17.150.7152,594
    2016Clayton, GA19.186.190,865
    2016Cameron, TX17.566.888,874
    2016Muskegon, MI18.950.873,431
    2016Merced, CA13.955.359,706
    2016Webb, TX18.176.655,254
    2016Cumberland, NJ15.453.252,224
    2016Wyandotte, KS17.365.444,866
    2020San Joaquin, CA21.057.1282,235
    2020Hidalgo, TX20.058.6218,726
    2020Stanislaus, CA18.750.4209,986
    2020Cameron, TX20.156.6113,095
    2020Clayton, GA20.885.8111,277
    2020Muskegon, MI20.550.390,776
    2020Merced, CA14.755.488,388
    2020Webb, TX20.161.867,718
    2020Cumberland, NJ17.853.161,694
    2020Imperial, CA15.762.555,525
    2024Wyandotte, KS21.262.149,805
    2024Wilson, NC21.750.239,659
    2024Orangeburg, SC19.462.436,582
    2024Apache, AZ16.659.631,667
    2024McKinley, NM13.362.725,075
    2024Edgecombe, NC17.561.424,255
    2024Halifax, NC16.158.923,792
    2024Liberty, GA20.558.822,900
    2024Gadsden, FL21.365.521,698
    2024Jefferson, AR20.160.221,270
    The two classes where college share and the majority line up, above the median and Democratic or at or below it and Republican, were 50.8% of counties in 2000 and 58.4% in 2024. The class that shrank most is the lower left, at or below the median and Democratic, from 325 counties to 93, while the above-median Democratic class held at 350 and 352. The median county college share is 14.5% in 2000 and 22.1% in 2024, so the cut rises with college share across counties. The data table lists the ten largest counties in the shrinking class in each election.Note: Election returns are full counts of votes, so the map draws no intervals. A county's class depends on the national result of the election as well as on its own vote. Counties are drawn on 2020 boundaries and matched on FIPS code; a county with no return or no college share in an election draws as no data. Kansas City's votes, a separate unit in the 2024 returns, are added to Jackson County, Missouri. The data table lists ten counties for each election, not the whole map.Source: County presidential returns, 2000-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, not weighted; college share from the county demographics panel (decennial census and American Community Survey estimates interpolated to election years)

    Of the 325325 countiesEstimandcount of countiesn325Weightsnone (full counts)Populationcounties at or below the median college share with a Democratic majorityPeriod2000SourceCounty election returns counties at or below the median with a Democratic majority in 2000, 66%66% percent of countiesEstimandshare of the counties whose non-Hispanic white share is 50 percent or moren325Weightsnone (full counts)Populationcounties at or below the median college share with a Democratic majorityPeriod2000SourceCounty demographics panel had a non-Hispanic white share of 50%50% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-offPeriod2000 to 2024SourceCounty demographics panel or more, and 208208 countiesEstimandcount of the counties with a Republican majority in 2024n213Weightsnone (full counts)Populationcounties at or below the median college share with a Democratic majority and a non-Hispanic white share of 50 percent or more in 2000Period2000 and 2024SourceCounty election returns of those 212212 countiesEstimandcount of the counties that have 2024 returns and characteristicsn213Weightsnone (full counts)Populationcounties at or below the median college share with a Democratic majority and a non-Hispanic white share of 50 percent or more in 2000, with a 2024 returnPeriod2000 and 2024SourceCounty election returns counties with a 2024 return had a Republican majority in 2024. The other four were Russell County, Alabama, Crittenden County, Arkansas, Wyandotte County, Kansas and St. John the Baptist Parish, Louisiana, and Clifton Forge city, Virginia, has no 2024 return. The 9393 countiesEstimandcount of counties at or below the median college share, Democratic majorityn3,103Weightsnone (full counts)Populationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty election returns such counties in 2024 are mostly in the South (86%86% percent of countiesEstimandshare of the counties in the 17 Census Bureau South states and the District of Columbian93Weightsnone (full counts)Populationcounties at or below the median college share with a Democratic majorityPeriod2024SourceCounty demographics panel), 1717 countiesEstimandcount of the counties in Mississippin93Weightsnone (full counts)Populationcounties at or below the median college share with a Democratic majorityPeriod2024SourceCounty election returns of them in Mississippi, and the rest are in New Mexico, South Dakota, North Dakota, Arizona, Kansas, Montana and Wisconsin. In all of them the non-Hispanic white share is under 50%50% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic white share cut-offPeriod2000 to 2024SourceCounty demographics panel, and 82%82% percent of countiesEstimandshare of the counties whose non-Hispanic Black share is 25 percent or moren93Weightsnone (full counts)Populationcounties at or below the median college share with a Democratic majorityPeriod2024SourceCounty demographics panel have a non-Hispanic Black share of 25%25% percent of residentsEstimanddefinition of a county groupWeightsnone (full counts)Populationnon-Hispanic Black share cut-offPeriod2024SourceCounty demographics panel or more. The map describes a majority in each county and cannot say which residents voted for which party.

    Before 2000 the ranking of counties by college share has a mixed relation with the vote

    The data hold no county college share before 2000, so the education of residents at the time cannot be compared with earlier votes. Figure 9 instead applies each county's 2000 college share to the elections from 1972 to 1996 and to those of 2000 to 2024 alike. College share in 2000 and in 2024 has a correlation of +0.94+0.94 correlation coefficientEstimandcorrelation across counties of college share in 2000 with college share in 2024n3,102Weightsnone; each county counted oncePopulationcounties with college share in both 2000 and 2024Period2000 and 2024SourceCounty demographics panel across counties, so the ranking changed slowly over those years, and the data cannot show whether it changed as slowly before 2000. The 2000 ranking is a different measure from the education of a county's residents in 1980, and a county that gained graduates before 2000 is placed by its 2000 level.

    Figure 9

    Counting each county once, the correlation with 2000 college share was positive in 1972, negative in the point estimates from 1976 to 1996 and positive from 2000

    Correlation of a county's share of residents with a bachelor's degree or more in 2000 with its Democratic share of the two-party vote in each presidential election, 1972 to 2024

    Correlation of 2000 college share with the Democratic share

    Show data
    Data for Counting each county once, the correlation with 2000 college share was positive in 1972, negative in the point estimates from 1976 to 1996 and positive from 2000
    conf_highconf_lowestimaten_unitsweightingyear
    0.3400.1230.2433,107Vote-weighted1972
    −0.094−0.340−0.2133,110Vote-weighted1976
    0.047−0.143−0.0483,110Vote-weighted1980
    0.198−0.0150.1003,112Vote-weighted1984
    0.197−0.0360.0873,112Vote-weighted1988
    0.2460.0050.1333,113Vote-weighted1992
    0.2470.0270.1473,113Vote-weighted1996
    0.3650.1890.2823,114Vote-weighted2000
    0.4390.2830.3633,113Vote-weighted2004
    0.4770.3360.4063,113Vote-weighted2008
    0.4580.3190.3873,113Vote-weighted2012
    0.6160.4970.5513,111Vote-weighted2016
    0.6730.5680.6123,111Vote-weighted2020
    0.6940.5960.6413,111Vote-weighted2024
    0.1780.0280.0993,107County-equal1972
    −0.270−0.387−0.3323,110County-equal1976
    −0.102−0.270−0.1983,110County-equal1980
    −0.016−0.180−0.1053,112County-equal1984
    0.000−0.168−0.0873,112County-equal1988
    0.009−0.179−0.0913,113County-equal1992
    0.028−0.152−0.0663,113County-equal1996
    0.1960.0050.0993,114County-equal2000
    0.3050.1180.2113,113County-equal2004
    0.3990.2310.3133,113County-equal2008
    0.4040.2250.3143,113County-equal2012
    0.5790.3820.4833,111County-equal2016
    0.6390.4630.5543,111County-equal2020
    0.6660.4940.5833,111County-equal2024
    The vote-weighted correlation is +0.24 in 1972, −0.21 in 1976 and −0.05 in 1980 and between +0.09 and +0.15 from 1984 to 1996 in the point estimates, and +0.28 in 2000. County college share is measured in 2000 for every election here, so before 2000 the chart shows how counties that ranked high on college share in 2000 voted, and not the education of their residents at the time. The solid line counts counties by their two-party votes and the dashed line counts each county once; bands are 95% intervals from resampling states.Note: Intervals resample whole states; see Methods. The data hold no county college share before 2000; the 2000 value is applied to earlier elections as a ranking of counties.Note: Counties are matched on FIPS code to the 2000 census; a county with no 2000 value is left out of that election, which removes under one percent of two-party votes before 2000.Source: County presidential returns, 1972-2024 (Algara and Amlani; MIT Election Data and Science Lab), full counts of votes, and the 2000 census county college share; weighted by two-party votes in the election or county-equal, as stated

    Counting each county once, the correlation was +0.10 (95% CI 0.03 to 0.18)+0.10 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI 0.03 to 0.18n3,107Weightsnone; each county counted oncePopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1972SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 1972, close to its 2000 value of +0.10 (95% CI 0.01 to 0.20)+0.10 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI 0.01 to 0.20n3,114Weightsnone; each county counted oncePopulationcounties with returns and a 2000 college share, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., then −0.33 (95% CI −0.39 to −0.27)−0.33 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI −0.39 to −0.27n3,110Weightsnone; each county counted oncePopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1976SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 1976 and −0.20 (95% CI −0.27 to −0.10)−0.20 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI −0.27 to −0.10n3,110Weightsnone; each county counted oncePopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1980SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 1980, and between −0.10−0.10 correlation coefficientEstimandlowest of the point estimates of the correlation across the elections of 1984, 1988, 1992 and 1996Weightsnone; each county counted oncePopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1984 to 1996SourceCounty demographics panel and −0.07−0.07 correlation coefficientEstimandhighest of the point estimates of the correlation across the elections of 1984, 1988, 1992 and 1996Weightsnone; each county counted oncePopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1984 to 1996SourceCounty demographics panel in the point estimates from 1984 to 1996. Vote-weighted it was +0.24 (95% CI 0.12 to 0.34)+0.24 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI 0.12 to 0.34n3,107Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1972SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 1972, against +0.28 (95% CI 0.19 to 0.37)+0.28 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI 0.19 to 0.37n3,114Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod2000SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2000, then −0.21 (95% CI −0.34 to −0.09)−0.21 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI −0.34 to −0.09n3,110Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1976SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 1976 and −0.05 (95% CI −0.14 to 0.05)−0.05 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI −0.14 to 0.05n3,110Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1980SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 1980, and between +0.09+0.09 correlation coefficientEstimandlowest of the point estimates of the correlation across the elections of 1984, 1988, 1992 and 1996Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1984 to 1996SourceCounty demographics panel and +0.15+0.15 correlation coefficientEstimandhighest of the point estimates of the correlation across the elections of 1984, 1988, 1992 and 1996Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod1984 to 1996SourceCounty demographics panel from 1984 to 1996. The two weightings differ in sign from 1984 to 1996, so the earlier relationship, like that of 2000, depends on how counties are weighted. Within 2000 to 2024 the 2000 ranking dates the rise as the election-year college share does. With each county's 2000 college share held fixed, the vote-weighted slope rose by +4.0 (95% CI 3.5 to 4.6)+4.0 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on the 2000 college share of the county, held fixed across elections, later election minus earlier electionInterval95% CI 3.5 to 4.6n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2012 to 2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. points from 2012 to 2016, against +3.2 (95% CI 2.7 to 3.8)+3.2 points of Democratic two-party share per 10 points of college share, change between two electionsEstimandchange in the slope of the Democratic share on college share, later election minus earlier electionInterval95% CI 2.7 to 3.8n3,113Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2012 to 2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. with the election-year share, and the larger step reflects that college share in 2000 is less spread out across counties than in 2016. The vote-weighted correlation with 2000 college share went from +0.39 (95% CI 0.32 to 0.46)+0.39 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI 0.32 to 0.46n3,113Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod2012SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2012 to +0.55 (95% CI 0.50 to 0.62)+0.55 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI 0.50 to 0.62n3,111Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod2016SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. in 2016, and in 2024 it is +0.64 (95% CI 0.60 to 0.69)+0.64 correlation coefficientEstimandcorrelation of the county's 2000 college share with its Democratic share of the two-party vote in the electionInterval95% CI 0.60 to 0.69n3,111Weightstwo-party votes in the electionPopulationcounties with returns and a 2000 college share, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates., near the +0.66+0.66 correlation coefficientEstimandcorrelation of college share with the Democratic share of the two-party voteInterval95% CI 0.62 to 0.71n3,103Weightstwo-party votes in the electionPopulationcounties with returns and county characteristics, Alaska excludedPeriod2024SourceCounty demographics panelNoteInterval: 95% percentile interval from a bootstrap over states, 2,000 replicates. found with the 2024 college share. The interpolated college shares of 2004 to 2020 therefore do not account for the step to 2016.

    What these data cannot show

    Every estimate here describes counties, and a pattern across places can differ from the pattern among the people who live in them, a limit known as ecological inferenceEcological inferenceDrawing a conclusion about individuals from data on groups or places, such as counties. A pattern across counties can differ from the pattern among the people who live in them.. Robinson [2] showed that a correlation across places can have the opposite sign from the correlation across individuals. College share counts residents 2525 years of ageEstimandlower age limit of the college shareWeightsnone (full counts)Populationresidents counted in the college sharePeriod2000 to 2024SourceCounty demographics panel and older with a degree, not voters, and a county with more graduates also has different occupations, ages, densities and histories. The chapter holds only race and ethnicity composition and state fixed. Counties also gain or lose graduates by migration, so a change in the slope can reflect where people live as well as how they vote. A book-length account of the diploma divide among individuals is [3], and Chapter 3.2, The diploma divide uses survey respondents.

    The dates are the dates of elections. College share for 2004 to 2020 is interpolated between anchors, and the 2000 value comes from a census sample, so the college shares carry sampling error, most in small counties, that no interval includes. Counties are not harmonized across boundary changes. In 2024, 99 countiesEstimandcounties with returns and no county characteristicsn3,112Weightsnone (full counts)Populationcounties with 2024 returnsPeriod2024SourceCounty demographics panel of the 3,1123,112 countiesEstimandcount of counties with returnsn3,112Weightsnone (full counts)Populationcounties with two-party votes, Alaska excludedPeriod2024SourceCounty election returns counties with returns have no county characteristics, among them the Connecticut counties, and Kansas City's votes, a separate unit in the 2024 returns, are added to Jackson County, Missouri, whose census figures are county-wide. The vote is the Democratic share of Democratic and Republican votes cast, so it leaves out third-party votes and people who did not vote, and a change in turnout can change a county's share without any voter changing party. The intervals say how much a number depends on which states are in the set, and the within-state intervals treat counties as independent.

    The weightings, the large-county cut, the white-share groups and the median split of the map are rules that move the estimates, and the section on weighting shows how far the weighting moves them. The state comparison and the controls for race and ethnicity do not hold income, age, occupation, density or religion fixed, which Chapter 12.1, Five axes of the county vote and Chapter 12.3, Income and the county vote (planned) take up. The data used here have no measure of attitudes, candidates, campaigns or turnout by group, so they cannot say why the slope rose when it did. They hold no county college share before 2000, so they cannot date the start of the gradient, only describe how the 2000 ranking lined up with earlier votes.

    Data sources

    • County election returnsCounty presidential returns, 1972-2024 (2000-2024 for the estimates; 1972-1996 for the 2000-ranking check): Algara and Amlani (to 2020) and MIT Election Data and Science Lab (2024), repaired fileYears: President 1868-2024 (40 elections); House and Senate 1990-2024 · Weights: none (full counts of votes); estimates weighted by two-party votes in the election, or county-equal · License: The Heideman file is CC BY 4.0 (doi:10.5281/zenodo.22976553), but the state records behind it keep their own terms, and the sites of Florida, Indiana, Colorado and Wisconsin restrict reuse.
    • County demographics panelCounty demographics panel, 2000-2024: share of residents 25 and older with a bachelor's degree or more, and non-Hispanic white, Black, Hispanic and Asian sharesYears: 2000-2024 (election years, interpolated between census and survey anchors) · Weights: none (census and American Community Survey estimates, interpolated to election years; sampling error not carried) · License: US government source data, public domain. The interpolated panel is derived work.

    All data sources

    Methods

    Sample: presidential returns for the 7 elections from 2000 to 2024 (Algara and Amlani for 2000 to 2020, MIT Election Data and Science Lab for 2024), from the repaired county file whose 2024 totals are corrected for 173 counties in South Carolina, Wisconsin and West Virginia. The Alaska rows (election districts) are dropped before any calculation, as are 5 rows with no vote count and 0 rows with no FIPS code or no two-party votes in 1972 to 2024. County characteristics come from the county demographics panel. Its anchors are the 2000 census sample and the American Community Survey five-year estimates for 2008 to 2012, 2012 to 2016, 2016 to 2020 and 2020 to 2024, placed at their midpoints, 2010, 2014, 2018 and 2022, with linear interpolation between anchors: 2004 and 2008 lie 40% and 80% of the way from the 2000 anchor to the 2010 anchor, 2012, 2016 and 2020 lie midway between two anchors, and 2024 takes the 2022 anchor (the nearest anchor; there is no extrapolation), so college share for 2004 to 2020 is an interpolation. The characteristics are the share of residents 25 and older with a bachelor's degree or more (college share) and the non-Hispanic white, non-Hispanic Black, Hispanic and non-Hispanic Asian shares of residents, each at the election year. An election's estimates use the 3,103 to 3,114 counties with returns and all five characteristics; county boundaries are not harmonized, so a few counties drop out in some years (the Connecticut counties in 2020 and 2024, Broomfield County in 2004 and 2008, Shannon County and Bedford city in 2012, Oglala Lakota County in 2024). Kansas City, Missouri is a separate unit (FIPS code 2938000) in the 2024 returns; in every election where the unit appears (2024 only) its votes are added to Jackson County (29095), whose census figures are county-wide, as in the 2000 to 2020 returns. Estimand: the outcome is a county's Democratic share of the two-party vote, 100 times Democratic votes over Democratic plus Republican votes (a slope on the margin is twice a slope on the share; correlations are the same); the gradient is the least-squares slope of that share on college share, per 10 points of college share, or the Pearson correlation of the two. Weights: two-party votes in the election (vote-weighted) or one per county (county-equal). Specifications: college share alone; with the non-Hispanic white, Black, Hispanic and Asian shares entered as straight lines (race and ethnicity held fixed; the omitted remainder is other and multiple races); within states, with each state's weighted mean removed from the share and every regressor (state fixed effects); both; and the slope on each county's 2000 college share, or on its 2024 college share, applied to every election (a county with no value in that year is left out of that version). The reduction in the slope by a control is the slope of college share alone minus the slope in the specification, and the change in the reduction between 2000 and 2024 is computed inside each bootstrap replicate. Groups: five equal-sized groups of counties (fifths) ranked by college share within each election, with ties broken by FIPS order, whose Democratic share is the mean of the county shares weighted as stated (vote-weighted, this is the pooled two-party vote of the group); large counties are those with 250,000 or more two-party votes in the election (slopes inside a size class are shown counting each county once and by two-party votes; the decomposition splits the vote-weighted covariance of college share and the Democratic share into the covariance of the class means, weighted by the classes' votes, and the votes-weighted average of the within-class covariances, and the between-class share is the first over the sum); the four groups by non-Hispanic white share (under 50%, 50% to under 75%, 75% to under 90%, and 90% or more) are formed at each election; the map classes compare a county's college share with the median college share of that election's counties and its Democratic share with 50%. Within-state slopes (the state figure) are county regressions inside each state with at least 20 counties in both 2000 and 2024, with HC3 errors that treat counties as independent. Intervals: 95% percentile intervals from a bootstrap that resamples whole states (49 states and the District of Columbia, 2,000 replicates, seed 12345; every estimate shares the same state draws, Connecticut is a resampling unit with no counties in 2020 and 2024, so a change between elections or specifications is computed inside each replicate, and group membership is held fixed across replicates). The returns are full counts, so these intervals show how much a number depends on which states are in the set; they are not sampling error of the vote and they do not include the sampling error of the college shares. Before 2000: the 2000 college share of each county, matched on FIPS code, is applied to the presidential elections of 1972 to 2024 from the same county file (under one percent of two-party votes in 1972 to 1996 are in counties with no 2000 value); before 2000 this ranks counties and does not measure education at the time. Differences are computed before rounding, so a printed difference can differ by 0.1 from the difference of two printed values. Suppression: none; every estimate rests on counties, and the smallest group has 253 counties.

    References

    1. Zingher, Joshua N. (2022). Diploma Divide: Educational Attainment and the Realignment of the American Electorate. Political Research Quarterly 75(2): 263-277. doi:10.1177/10659129221079862
    2. Robinson, W. S. (1950). Ecological Correlations and the Behavior of Individuals. American Sociological Review 15(3): 351-357.
    3. Grossmann, Matt; Hopkins, David A. (2024). Polarized by Degrees: How the Diploma Divide and the Culture War Transformed American Politics. Cambridge University Press.

    How to cite

    @misc{singh2026thediplomamap,
  author = {Singh, Vishal},
  title = {The diploma map},
  year = {2026},
  url = {https://vishalsingh.org/politics/the-diploma-map},
  note = {In The Public Mind. Published 3 October 2026.}
}

    Same question at another grain

    These chapters take up the same axis with a different unit of analysis: individuals, places or countries.

    Reading paths