A Measure of Nationalization: Subpresidential Elections Are Not Created Equal

• In Event: Novel Applications of Machine Learning

Abstract

Scholars have documented the upward trend of nationalization within American politics, but frequently overlook the significance of local elections. I collect a dataset of presidential election returns at the municipal level and develop a novel, nonparametric measure of nationalization that incorporates a k-means clustering algorithm, which can be applied to most election data. After introducing a configuration of subpresidential elections that distinguishes between subnational and local elections (mayoral and state legislature) with respect to nationalization, I show that nationalization trends at the two levels differ. Congressional and gubernatorial elections have markedly nationalized over time, consistent with existing literature, but local elections have not followed the same rising trend and exhibit considerably lower degrees of nationalization. These findings suggest that conclusions made exclusively based on congressional and gubernatorial elections are not complete and perhaps national issues have not saturated and eclipsed local issues to the extent previously postulated.

Author

• Francesca Tang, Princeton University