The covid-19-driven economic collapse, upheaval over racial politics in the United States and President Trump’s consistently low popularity ratings have given Democrats an opening to win the presidency and both houses of Congress in 2020. Our prediction, based on economic indicators and presidential approval ratings, provides a simple, albeit early, forecast predicting a Democratic win in November. Here’s how we approached these forecasts.
The House forecast
To produce a House forecast for 2020, we built on a well-known forecasting model developed by political scientists Michael Lewis-Beck and Tom Rice. The model is based on three factors: changes in real gross national product in the first and second quarters of the election year; the president’s approval rating in Gallup polls as of July of the election year; and whether it is a presidential or midterm year. That last variable matters because of a well-known historical pattern: In midterm elections, the party that does not hold the presidency gains seats in the House, as happened in 2018.
We applied the Lewis-Beck and Rice model to each election from 1948 to 2018. We adjusted the model by using gross domestic product (GDP) instead of gross national product because GDP excludes foreign production by American-owned firms and is the more commonly used indicator of national economic trends. We believe this approach is reliable, because this model has accurately predicted the winner in 14 out of 17 presidential elections between 1948 and 2012.
We then estimated the model’s predictions for the upcoming 2020 election using the most recent quarter’s GDP and current presidential approval rating. In the first quarter of 2020, GDP fell at the rate of -4.8 percent and Trump’s Gallup Poll rating in June was 39 percent. Important to our prediction is the assumption that current economic and political conditions stay roughly the same.
The model predicts that the Republicans will lose 14 seats in the House, a number that would solidify the Democrats’ majority in the 117th Congress. The accuracy of the model, however, is modest; a plausible outcome ranges from a 24-seat loss to a 53-seat gain for the Democrats.
The Senate forecast
Predicting the Senate outcome is similar, but includes an additional factor: the number of seats the president’s party has up for election. In 2020, this factor favors the Democrats because 23 Senate Republicans — but only 12 Senate Democrats — must defend their seats.
For the Senate, the model predicts that the Republicans will lose seven seats. Given the uncertainty in the forecast, +/- 6 seats, this suggests that both parties have a reasonable chance at a Senate majority come 2021, but the Democrats are clearly favored.
The White House forecast
Of course, it’s very early to make presidential election forecasts — mainly because there are no guarantees that the key factors in our model, the economy and presidential approval ratings, won’t change. But if the economy continues its downward trajectory and the president’s approval rating remains static, our model predicts that Trump will receive only 24 percent of the electoral college vote.
Additionally, because there are fewer presidential elections than congressional elections, models like ours are less accurate than those that predict House gains and losses alone. A White House forecast must make use of fewer election outcomes to make predictions about the future, thus, a plausible percent of the vote won by Trump ranges anywhere from 0 to 54 percent.
A look at the model’s record of prediction may give us a better sense of how much confidence we should place in our prediction. In 2018, we predicted a 38-seat gain in the House for the Democrats; they ended up winning 41 seats. Our Senate prediction was a one-seat gain for the Republicans; they won two seats.
The figure below shows the predictions and actual results for the presidential elections from 1948 to 2016, with the 2020 prediction at the end of the series. As visualized in the model, the predictive record is strong.
Source: Figure prepared by authors, using GDP and Gallup Poll data for predicted vote share, and National Archives records for actual electoral vote share of the incumbent’s party (https://www.archives.gov/electoral-college/results).
The question now is what happens to the economy and Trump’s popularity in the coming months. Given the economic uncertainties surrounding the covid-19 pandemic, the spring protests and the unprecedented stability of Trump’s approval ratings, 2020 may end up challenging some of our model’s assumptions. Yet, considering the model’s historical accuracy, we still believe that these two factors will best predict November’s election results. Without a significant change in Trump’s approval rating and a swift recovery of the pandemic-depressed economy, Republicans seem unlikely to prevail in November.
The TMC newsletter has moved! Sign up here to keep receiving our smart analysis.
Olivia Quinn is a doctoral student in political science at the University of California, Santa Barbara.
Amanda Brush is a doctoral candidate in political science at the University of California, Santa Barbara.
Eric R.A.N. Smith is professor of political science at the University of California, Santa Barbara.