The model by Michael Lewis-Beck and Charles Tien is a classic political economy model using multiple linear regression. The model predicts the incumbent two-party popular vote from two variables, namely
- political popularity of the incumbent president, measured through the Gallup Presidential Approval in July of the election year
- economic growth, measured as GNP growth in the first two quarters of the election year
The model was first proposed by Lewis-Beck and Rice (1982) but abandoned after a relatively poor performance in predicting the 1984 election. Lewis-Beck and Tien (2016) revived the model for predicting the 2016 election.
The table below shows the model forecasts and error across the elections from 1984 to 2020. The model’s average error across these ten elections is roughly two percentage points.