FORECASTING ELECTIONS FROM VOTERS’ PERCEPTIONS OF CANDIDATES’ POSITIONS ON ISSUES
Bucharest Dialogues on Expert Knowledge, Prediction, Forecasting: A Social Sciences Perspective, November 2010.
Abstract. We used the index method to predict U.S. presidential election winners. Our approach draws on information from polls that ask voters which candidate they expect to do a better job in dealing with the issues facing the country. Using a simple heuristic, we expected the candidate who is rated more favorably on a larger number of issues to win the popular vote. This prediction was correct for nine of the ten elections from 1972 to 2008. We then used simple linear regression to relate the candidates’ relative ratings to the actual popular two-party vote-shares. The resulting model yielded out-of-sample forecasts that were as accurate as forecasts from the Iowa Electronic Markets and established econometric models. The model has implications for political decision-makers as it can help to decide which issues to focus on in campaigns.
Click here for the full paper.
