A Day in the Life Table: Estimating Turnout Using Survival Analysis in Early Voting States.

A majority of American voters currently have a meaningful opportunity to vote before Election Day, whether in person or by mail. As such, voter mobilization experiments and observational studies of turnout need to consider voting as a dependent variable that occurs over time. This paper presents three statistical methods for estimating turnout across a multi-day voting period: binary logistic regression, multinomial logistic regression, and Cox proportional hazards survival analysis. Ultimately, survival analysis allows for the estimation of treatment effects over time and the potential to test for interactions between covariates and time. A discussion of the implications for researchers and campaign professionals is also included. [ABSTRACT FROM AUTHOR]