Accuracy of Vote Expectation Surveys in Forecasting Elections

Abstract:     
Simple surveys that ask people who they expect to win are among the most accurate methods for forecasting U.S. presidential elections. Vote expectation surveys failed to pick the winner in only 18 (8%) of 214 surveys conducted from 1932 to 2012. Across the seven elections from 1988 to 2012, vote expectation surveys outperformed four established methods (trial-heat polls, prediction markets, econometric models, and experts’ judgment) in predicting election winners and vote shares. Vote expectation surveys are accurate, inexpensive, and easy to conduct. They should be more strongly utilized by election observers as well as researchers.

Number of Pages in PDF File: 27

Keywords: combining forecasts, Iowa Electronic Markets, election forecasting, forecast accuracy, econometric models, FiveThirtyEight

JEL Classification: C53, C42

working papers series