Combining Aggregate Data and Exit Polls for the Estimation of Voter Transitions

Our objective is the estimation of voter transitions between two consecutive parliamentary elections. Usually, such analyses have been based either on individual survey data or on aggregated data. To move beyond these methods and their respective problems, we propose the application of so-called hybrid models, which combine aggregate and individual data. We use a Bayesian approach and extend a multinomial-Dirichlet model proposed in the ecological inference literature. Our new hybrid model has been implemented in the R-package eiwild (= Ecological Inference with individual-level data). Based on extensive simulations, we are able to show that our new estimator exhibits a very good estimation performance in many realistic scenarios. Application case is the voter transition between the Bavarian Regional election and the German federal elections 2013 in the Metropolitan City of Munich. Our approach is also applicable to other areas of electoral research, market research, and epidemiology.