Election Analytics Blog (Gov 1347)

Ella Michaels, Harvard College '22

View the Project on GitHub ellamichaels/gov1347_blog


September 18, 2020

“It’s the economy, stupid.”

Bill Clinton’s campaign manager, James Carville, famously coined this phrase leading up to the 1992 presidential election. Candidate Clinton’s campaign played up the early 90s recession in order to unseat successfully incumbent president George H. W. Bush. This quip only dates back to 1992, but the general wisdom that the economy plays a major role in election outcomes is probably about as old as democracy itself.

But “the economy” is a big concept. It includes indicators like stock market performance or annual gross domestic product (GDP) that measure the health of the national economy. It also includes things like real disposable income (RDI) or local unemployment rates that reveal less about national-level economic health but paint a clearer picture of the state of voters’ pocketbooks. While each measure is certainly related to the others, they tell slightly different stories about election outcomes. Here, we’ll assess the relationship between a variety of different economic indicators and past election outcomes, and what the data suggest about the upcoming 2020 presidential election.

Methodology and Terms

For these purposes, we’ll focus on economic indicators from the second fiscal quarter of presidential election years. Research suggests that voters weigh the year or few months leading up to an election more heavily than the rest of an incumbent’s term in office. (Healy and Lenz, 2014)

We’ll build predictive models around three variables to identify how they impact the popular vote share of incumbent political parties:

If it is true that voters consider each of these indicators at the ballot box, stronger economic performance (i.e. higher GDP and RDI growth and larger decreases in unemployment rates) would track with higher popular vote shares for incumbents.

In order to evaluate the performance of each model, we’ll consider the following tests:

Q2 GDP Growth

Impact of Q2 GDP Growth on Incumbent Party Vote Share


This model predicts that Donald Trump will win 21.26% of the popular vote based on Q2 GDP growth, which hit a historic low. This result is highly mismatched with current polling in which Biden leads Trump 52% to 43%.

Q2 RDI Growth

Impact of Q2 Real Disposable Income (RDI) Growth on Incumbent Party Vote Share


This model predicts that Donald Trump will win 80.33% of the popular vote based on Q2 RDI growth. This result is also highly mismatched and RDI growth, which was much higher in Q2 than normal, might have been distorted by economic stimulus programs in response to COVID-19. It is unlikely to be especially representative of most Americans’ financial situations during the pandemic.

Q2 Unemployment Rate Change

Impact of Change in Q2 vs Q1 Unemployment Rate on Incumbent Party Vote Share


This model predicts that Donald Trump will win a mathematically impossible -28.32% of popular vote. Unemployment rates skyrocketed from just under 4% to around 13% (and counting) between Q1 and Q2 of 2020 due to COVID-19, a historic increase.


While data from past elections demonstrates that there is some relationship between economic performance and incumbent performance, model predictions for 2020 indicate that 1) the economy isn’t everything, and 2) different economic indicators can tell different stories. Pandemic conditions wildly distort economic conditions and may lead voters to different conclusions about the incumbent’s culpability for national performance.