Politics and Governance

Stephen Ansolabehere ¹, Eitan Hersh ²

Affiliations
1 Department of Government, Harvard University, 1737 Cambridge St., Cambridge, MA 02138, US
2 Department of Political Science, Yale University, 77 Prospect St., New Haven, CT 06520, US

Abstract

In this brief analysis, we use a new dataset of two million voter registration records to demonstrate that gender, race, and age do not correlate with political participation in the ways that previous research has shown. Among Blacks and Latinos, women participate at vastly higher rates than men; many Blacks participate at higher rates than Whites; and the relationship between age and participation is both not linear and varies by race and gender. Survey research is unable to capture the true relationship between demographics and participation on account of survey bias and, more importantly, the non-linearity of effects. As a result, theories of participation, like the dominant resources-based models, have been built on faulty premises and tested with inadequate data. Our evidence calls for a renewed effort to understand election participation by utilizing large datasets, by being attentive to linearity assumptions, and by returning to theory.

Keywords

Big Data, Gender, Race, SES, U.S. Elections, Voter Registration, Voting.

Full Text