Strong Theory and Long Lead: Forecasts for the 2020 US Federal Elections

• In Event: Electoral College Forecasts of the 2020 Election

Abstract

Over the course of forecasting many elections we have been calling for strong theory and long lead time. In 2016, we applied these tenets to forecasting the presidential, House and Senate elections. We have argued that presidential elections are referendums on how well the incumbent party has handled major economic and non-economic issues. Using a parsimonious political-economy model with two independent variables that measures the president’s handling of major economic and non-economic issues, we predicted Clinton’s share of the two-party vote within one-tenth of one percentage point. Applying the political-economy model to House and Senate forecasts, we were close to the mark in 2016: our prediction of a three-seat gain for Democrats in the House was close to the actual six seat pick-up. Our Senate forecast of a four-seat gain for Democrats was close to the two seats they picked up.
Our proposal for the 2020 forecasting symposium is to evaluate the parsimonious, political-economy models for forecasting the presidential, House, and Senate elections. We also propose including an Electoral College forecast based on a simple bivariate popular vote model. Two of the last five presidential elections have resulted in a non-majoritarian outcome. What are the chances of another such outcome in 2020?

Authors

• Michael S. Lewis-Beck, University of Iowa

• Charles P. Tien, CUNY, Hunter College & The Graduate Center