State-by-State Political Economy Model: Forecasting the 2020 Election

• In Event: Electoral College Forecasts of the 2020 Election

Abstract

We will use a State-by-State Political Economy (SSPE) model to forecast the 2020 US Presidential Election. A previous version of this model has performed well between 2004 and 2012 at predicting both the Democrat/Republican division of the national suffrage as well as the major candidates’ vote in the Electoral College. In 2016, it succeeded again at predicting the major two-party division of the popular vote but failed to accurately forecast the vote in the Electoral College (See PS 49 (4): 680–86). Not surprisingly, our model did not perform particularly well at predicting the results for battleground states like Florida, Michigan, Pennsylvania, Virginia and Wisconsin.

Our goal is to amend our SSPE model to better capture the impact of regional factors that have produced vote shifts in 2016 undetected by specifications mostly or entirely based on national-level variables. To do that, all the key variables in our model including the presidential popularity and the partisan balance will be measured at the state-level. We believe that this strategy should prove rewarding. The approval record of the outgoing President Obama for instance was significantly below his national level in crucial swing-states including Ohio (-4.5), Pennsylvania (-3.5) and Wisconsin (-1.5). Using a popularity index for the President measured at a national level did not allowed us in 2016 to adequately detect the local signs of weakness in the Democratic vote that eventually proved decisive in key battleground states. Our revised SSPE model will try to address these shortcomings to produce better forecasts in the US 2020 Presidential election.

Authors

• Bruno Jerome, University of Paris II Pantheon Assas

• Veronique Jerome, University of Paris XI-Sud Orsay

• Richard Nadeau, University of Montreal