The Political Economy Model: 2016 US Election Forecasts
Michael S. Lewis-Beck, University of Iowa
Charles Tien, Hunter College & The Graduate Center, CUNY
We have been forecasting United States presidential elections since the early 1980s, and stand among the first forecasters of these contests (Lewis-Beck and Rice 1982; 1984). Our approach has been to follow a parsimonious structural model, focusing on a few predictor variables based on strong political science theory and measured well in advance of the election itself. Occasionally, we have attempted to “improve upon” the forecasts of these simple models, with mixed results (Lewis-Beck and Tien 1996; 2008). Based upon our historical testing and outside corroborative evidence, we have decided to go back to our beginnings. Hence, we make a full-fledged return to our initial multivariate effort, a political economy model that expresses the incumbent party vote as a function of presidential popularity and economic growth. Below, we first apply the model to presidential elections, then turn to congressional elections. After working through model estimation and the election forecasts, we interpret their meaning.