Political Orientations and Civic Engagement
The effect of higher education on civic engagement and political orientations has become highly salient and politicized in contemporary democratic debates, yet existing research struggles to distinguish whether differences among graduates reflect genuine educational treatment effects or self-selection. This paper takes a new approach to addressing this problem through a change in measurement. We measure course exposure rather than degree completion, enabling identification of treatment effects distinct from selection bias.
Our research design leverages panel data from a Californian public university, tracking individual students' political orientations and civic engagement before and after coursework exposure. This relationship is modelled as a dosage-response function using recent advances in difference-in-difference estimations. Courses are differentiated further by enrolment and restricted choice conditions, i.e., general education courses.
Results reveal limited evidence of self-selection bias in outcomes: political orientation liberalization emerges under all constrained selection conditions, with heterogeneous response dosage between the exposures. While civic engagement improvements are slightly attenuated when self-selection is accounted for.
We employ continuous difference-in-differences methods with entropy balancing to estimate dose-response effects of social science, humanities, and cultural and communicative coursework on political orientation and civic engagement. Using an LLM ensemble classification of courses by communicative and cultural competencies, we find that selection operates asymmetrically across outcomes: liberalizing effects strengthen under general education identification (consistent with floor effects from liberal self-selection), while civic engagement effects attenuate (consistent with inflation from engaged self-selection). This asymmetry itself represents a test of validity, a singular source of opposing selection biases.
LACA-weighted measures show linear dose-response relationships with minimal selection influence, providing the strongest causal evidence. Effect sizes remain modest yet are robust to extensive tests on selection and alternative estimations. These findings demonstrate how sophisticated identification strategies can distinguish genuine educational effects from spurious associations, challenging simple narratives about higher education's liberalizing effects while providing a methodological framework for rigorous causal analysis that reveals how exposure intensity and selection mechanisms jointly shape political development.