Bayesian Analysis of Election Surveys and Forecasts: Learning from the State of Valencia Experience

Election surveys have several purposes, including forecasting election outcomes and studying the distribution of votes as they vary over geographic, demographic, and political variables. Bayesian methods can be useful in the design and analysis of election surveys, with the details depending on speciic features of the data collection and goals of inference. In many cases, the natural Bayesian analyses correspond to variants of standard methods in survey sampling. This article discusses the design and modeling choices for pre-election and exit polls discussed in Bernardo (1996), emphasizing (1) the relation of the analyses to the methods of data collection, (2) the extent to which his Bayesian approach diiers from classical methods, (3) diierences between election studies in Spain and the United States.