When coronavirus cases started surging the United States at the beginning of March, the Democratic nominee for U.S. president was still in doubt. During the primary, one of the two leading candidates — Sen. Bernie Sanders — had been running on a platform centered on universal health care. One might have expected support for Sanders to surge in the face of a growing epidemic. It did not.
In a recent working paper, we find evidence that coronavirus-induced anxiety prompted a political “flight to safety” that cost Sanders votes. Our findings suggest that if a coronavirus case was reported in a county’s local media market before its primary, that news may have depressed Sanders vote share by between 1.15 and 12.65 percentage points.
Here’s the background
Fourteen states voted March 3, known as Super Tuesday. Although China’s coronavirus outbreak had already hit the news, many Americans weren’t yet especially worried. Just over 100 cases had been recorded nationwide on the eve of the election; the news media were focused on the elections; governors hadn’t yet begun holding daily briefings or issuing stay-at-home orders; and President Trump had said that covid-19, the disease caused by the coronavirus, was soon “going to disappear.”
Just 14 days later, when voters in three states cast their primary ballots, circumstances had changed dramatically. Trump had declared a national state of emergency. Most of the nation’s schools were closed. The stock market had lost more than 20 percent of its value. And many people had begun staying home and practicing “social distancing.”
We wanted to know whether the changes in the medical, social and economic climate between counties before and after the outbreak affected Democratic voters’ choices in the primary.
How we did our research
To examine covid-19’s effect on the primaries, we calculated the change in Sanders’s share of the vote between 2016 and 2020. Across most of our counties, Sanders got fewer votes in 2020 than he had in 2016.
But did he lose even more in areas where voters went to the polls with disease on their minds? We compare counties with similar labor markets, racial and ethnic makeup, educational attainment, and rural/urban populations, but differences only in whether a covid-19 case appeared before their primary election date. We find that Sanders’s support was weaker in counties where a covid-19 case appeared before the primary. We tried to figure out what was driving this result. We redefined the “outbreak date” to compare different groups of states, adjusted the number of reported cases beyond which we considered a county “exposed,” and explored whether the outbreak was worse in areas that voted strongly against Sanders in 2016. In all tests, our core finding persists: Sanders’s vote was already slipping when the outbreak occurred, but the epidemic reduced his share of the vote even more.
Why might the pandemic turn voters away from Sanders?
There are several possible explanations for our findings. Sanders’s support among younger voters is well-documented. Perhaps younger voters were less likely to turn out to vote in areas where the virus had spread. Of course, such a result would be surprising because covid-19 threatens older individuals’ health more than that of the young. Nevertheless, we tested this theory by comparing primary turnout in exposed and insulated counties, and found that the virus did not deter voters from the polls until the March 17 primaries in Arizona, Florida and Illinois, and in any case cannot explain these results — the decline in turnout was similar in counties with more young people and in those with more old people.
Here’s another possibility. In The Washington Post last month, Jared Bernstein, who was Biden’s chief economic adviser in the Obama administration, summarized the stock market’s response to the virus as a “flight to safety,” a well-documented financial phenomenon. Research finds that investors and others tend to respond to uncertainty by moving their money to investments they perceive as less risky. We believe the “flight to safety” applies to Bernstein’s candidate, as well.
Throughout the primary campaign season, Biden has sold himself as an “Obama-Biden Democrat,” someone who will restore the country to an era that Democrats recall as comparatively stable and sane. Sanders, in contrast, has promised to overturn inequities and shake up the system, challenging “the limits of politics as usual,” as his adviser put it, and “taking on the incredibly powerful institutions that control the economic and political life of this country.” Exit polls in a number of states showed that Sanders won a majority of those voters for whom the most important quality in a candidate was “can bring needed change.”
Our findings suggest that primary voters are taking a political flight to safety — from the candidate who promises dramatic change to the candidate who promises a return to “normalcy.” Our results echo similar work by Filipe Campante, Emilio Depetris-Chauvin and Ruben Durante, who found that Ebola-induced fear affected the 2014 U.S. midterm elections. As they put it, “Emotional reactions associated with fear can have a strong electoral impact.” This is consistent with evidence from terrorist attacks, which have demonstrated electoral effects.
The covid-19 effect isn’t the deciding factor in making Biden the likely Democratic nominee. Had there been no pandemic, Sanders probably still wouldn’t have rallied enough support to catch up. But our analysis finds that virus anxiety did have an effect, with voters who felt more intimately endangered by covid-19 were more likely to choose Biden’s promised stability over Sanders’s risky unknowns.
James Bisbee (@JamesBisbee) is a postdoctoral fellow at the Niehaus Center for Globalization and Governance at Princeton University.
Dan Honig (@rambletastic) is an assistant professor of international development at Johns Hopkins School of Advanced International Studies.