Does Attending College Change Political Party Affiliation?
Firoozi (2026) examines whether additional education changes political party affiliation. He analyzes administrative records covering millions of registered voters in California and Florida, along with approximately 250,000 University of California applicants. Using natural experiments involving compulsory schooling laws and college admissions, he finds that an additional year of high school education reduces Republican registration by approximately 1.4–2.5 percentage points. Each expected year at a highly selective university reduces Republican registration by roughly 2.5–2.9 percentage points. These effects persist into adulthood. He finds that interactions with classmates, rather than political persuasion by teachers or professors, likely explain much of this shift.
Why This Article Was Selected for The Policy Scientist
Education has become an increasingly important dividing line in American politics, raising fundamental questions about how schools shape political identities and democratic participation. As partisan differences by educational attainment widen, understanding whether education causes these differences, rather than merely reflecting preexisting preferences, is particularly timely. The findings also illuminate how educational institutions influence political attitudes across generations, with implications for understanding electoral realignment. The study makes an important contribution by using extensive administrative records and credible regression discontinuity designs to establish causal effects.
Full Citation and Link to Article
Firoozi, D. (2026). Education and partisanship. American Economic Journal: Applied Economics. Advance online publication (forthcoming).
Central Research Question
Does education causally influence political party affiliation, and do additional years of schooling and attendance at more selective universities produce lasting changes in political preferences? The study investigates whether education itself contributes to the growing political divide between Americans with different educational backgrounds or whether this relationship primarily reflects differences that existed before individuals entered school. It also examines whether education's political effects arise from interactions with classmates, exposure to instructors and curricula, or subsequent changes in employment and income.
This question has become increasingly important as education has displaced income as a major predictor of political affiliation in advanced democracies. Historically, higher-income voters tended to support conservative parties, while lower-income voters favored parties advocating greater economic redistribution. Increasingly, however, highly educated voters support parties on the political left, even when their earnings place them among society's highest-income groups. The study examines whether educational institutions actively contribute to this political realignment.
Previous Literature
Traditional political economy models emphasize economic self-interest as a principal determinant of political preferences. Romer (1975) and Meltzer and Richard (1981) developed influential models predicting that income differences shape voters' preferences for redistribution. These frameworks generally imply that higher-income individuals should favor parties supporting lower taxation and less redistribution. The increasing political alignment between highly educated, relatively affluent voters and left-leaning parties challenges these expectations.
More recent research documents the growing importance of education in electoral politics. Gethin et al. (2021a, 2021b) demonstrate that educational attainment has become a defining characteristic of political coalitions across Western democracies. Kuziemko et al. (2023) examine the changing relationship between education and American political preferences, while Enke (2020) and Bonomi et al. (2021) emphasize how moral values, social identities, and cultural issues influence political behavior independently of economic interests.
Previous studies examining education's political consequences have produced inconsistent findings. Marshall (2016, 2019) and Campbell and Horowitz (2016) investigate relationships between educational attainment and political attitudes, but observational comparisons struggle to distinguish educational effects from preexisting differences among students. Research by Carrell et al. (2019) and Braghieri (2021) also demonstrates that peer environments can influence social attitudes and political beliefs. The present study extends this literature by exploiting educational policies that generate plausibly exogenous differences in schooling experiences.
Data
The analysis combines extensive administrative records from voter registration systems and higher education institutions. The principal political dataset contains approximately 220 million registered voters across all 50 states and Washington, D.C., covering voter registration and electoral participation between 2012 and 2024. These records identify political party affiliation, registration histories, and selected demographic characteristics.
For the high school analysis, the researcher examines more than five million registered voters born within 70 days of compulsory schooling eligibility thresholds in Florida and California. The underlying voter populations in these two states total approximately 37 million individuals. These states provide contrasting political environments, allowing an assessment of whether educational effects persist across substantially different partisan settings.
The college analysis uses administrative records covering approximately 250,000 University of California applicants between 2007 and 2011. These records include academic qualifications, admissions decisions, and subsequent college enrollment. Individual records are linked to voter registration data to examine political outcomes years after college attendance.
Additional evidence comes from surveys of more than four million incoming college students, approximately 80,000 faculty members, and a separate 2022 survey of 1,105 University of California applicants. These supplementary datasets provide information about political ideology, social interactions, faculty practices, and potential mechanisms connecting education to partisanship.
Methods
The study employs several regression discontinuity designs to estimate the causal effects of education on political affiliation. These methods exploit eligibility thresholds that create differences in educational experiences among otherwise similar individuals. By comparing people immediately above and below these thresholds, the researcher reduces the selection bias that complicates conventional comparisons between individuals with different educational backgrounds.
The first analysis exploits compulsory schooling laws in Florida and California. School entry eligibility depends on whether children reach age five before specified birthdate cutoffs. Combined with minimum school-leaving ages, these rules create differences in educational attainment among students born only days apart. Students born before the cutoff generally complete more schooling before reaching the legal dropout age. Because birthdates near these thresholds are plausibly unrelated to subsequent political preferences, the resulting differences provide a credible basis for causal inference.
The second analysis exploits University of California admissions policies that favor applicants meeting specific academic thresholds. Students immediately above these thresholds receive greater access to highly selective universities than otherwise comparable applicants immediately below them. The researcher examines whether this additional access changes subsequent political affiliation. Separate analyses distinguish the effects of attending a four-year institution from those of attending a more selective university.
The researcher evaluates the credibility of these designs through tests of demographic balance, manipulation around eligibility thresholds, alternative statistical specifications, and placebo comparisons. Supplementary surveys investigate possible explanations for the observed effects. Although the regression discontinuity designs provide strong causal evidence for students affected by the eligibility thresholds, the analyses of underlying mechanisms are primarily descriptive and cannot establish causation with equivalent confidence.
Findings/Size Effects
The central finding is that additional education reduces the probability of registering as a Republican. This relationship appears across high school and university settings, multiple generations, and two states with substantially different political environments.
In the compulsory schooling analysis, an additional year of late-high-school education reduces Republican registration by approximately 1.4 to 2.5 percentage points. These estimates are derived by scaling observed differences at birthdate thresholds using previously established relationships between compulsory schooling laws and educational attainment. Consequently, the precise magnitude depends on assumptions about how much additional schooling the laws produce.
The effects persist into adulthood, including among individuals approximately 50 years old. They are larger in Republican-leaning Florida than in Democratic-leaning California and are especially pronounced among younger generations. For individuals born after 2002, estimated effects are approximately twice as large as those observed among earlier members of Generation Z.
The university analysis produces comparable results. Each expected year of enrollment at a highly selective university reduces Republican registration by approximately 2.49 to 2.94 percentage points. Registration as a Democrat, independent, or third-party voter increases by approximately 5.79 to 8.04 percentage points, with much of the increase concentrated among voters reporting no party preference. These figures describe different registration outcomes and should not be interpreted as a direct one-for-one transfer between political parties.
The political consequences also depend on students' educational alternatives. For students who otherwise might not attend a four-year university, selective college access increases voter registration, particularly Democratic registration. For students who would otherwise attend less selective institutions, the principal effect is movement away from Republican affiliation toward independent registration.
Evidence concerning mechanisms points primarily toward peer socialization. Students attending selective universities encounter classmates with different political, religious, and cultural perspectives. They also spend more time living and interacting with peers rather than family members. Student surveys indicate that classmates and friends exert greater perceived influence over political attitudes than professors. Subsequent career experiences may reinforce these changes, while the evidence provides limited support for deliberate political persuasion by instructors.
Conclusion
The findings indicate that education does more than reflect existing political differences: additional schooling and access to selective universities can causally influence political party affiliation. These effects appear across different educational stages and persist long after students leave school.
The research challenges conventional political economy models that emphasize income as the primary determinant of partisan preferences. Education can increase future earnings while simultaneously reducing Republican affiliation, suggesting that social identities, cultural attitudes, and interpersonal relationships contribute to political behavior independently of economic self-interest.
The study's principal contribution is its use of large administrative datasets and multiple natural experiments to distinguish educational effects from preexisting differences among students. Its findings are relevant to understanding the growing educational divide in American politics and the broader transformation of electoral coalitions. Nevertheless, the causal estimates apply most directly to individuals whose educational experiences change because of specific schooling and admissions thresholds. Whether similar effects occur across all students, institutions, and jurisdictions remains an important question for future research.
Comments