SIEPR researchers examine exposure to racial diversity
Not long after social distancing became the recommended response to the COVID-19 pandemic, a disturbing trend emerged: Surveys suggested there were large gaps in how residents of blue and red states were responding — with Democrats staying away from others at far higher rates than Republicans were. Democrats also said they considered the coronavirus to be far more dangerous.
For Stanford economist Matthew Gentzkow, the phenomenon begged a crucial question: Had a “coronavirus culture war” erupted or was there another explanation?
It’s possible, for example, that politics had nothing to do with the disparate responses. The most severe outbreaks were then occurring in major cities populated by Democrats, which means residents — regardless of political affiliation — would be more likely to practice social distancing or be subject to strict shelter-in-place mandates.
On the flip side, regions where the disease wasn’t spreading as fast — and responses were therefore more muted — just happened to be predominantly Republican.
Seeking clarity, Gentzkow and a group of fellow researchers spent the last month delving into GPS data and the results of an online survey they designed.
Their working paper, released this week by the National Bureau of Economic Research, finds that the partisan gap is real. Differences in political beliefs are a factor in why Democrat-heavy regions practice more social distancing and predict more cases of COVID-19 than Republican-dominant areas.
By the third week in March, counties that were predominantly Republican saw 19 percent higher movement than in comparable Democratic counties. “Although it’s not like everyone in red states is going to bars and congregating in big groups on the street, these differences in social distancing behavior are meaningful,” says Gentzkow, a senior fellow at the Stanford Institute for Economic Policy Research (SIEPR).
Gentzkow and his co-authors suggest the divide may be attributable to differences in how political leaders of both parties and media, on the right and left, portrayed the threat.
“Partisan differences in response to COVID-19 aren’t about cheerleading for ‘my team,’” he says. “Instead, the evidence points to real differences among Republicans and Democrats in their beliefs.”
Gentzkow says the consequences could be grave.
“If people have different beliefs about how serious the crisis is and are behaving differently as a result, that suggests that we are not saving as many lives as we could for the amount of social distancing that’s happening,” he says.
Similar situations, different responses
Gentzkow has extensively studied the widening political divide in the United States and elsewhere. Among other studies, he has looked at how Facebook, fake news, and language have contributed to the chasm. He also shown the political rift in the United States has grown faster and larger than it has in other established democracies.
When the COVID-19 contagion struck, Gentzkow says the country lacked the clear and consistent messaging from political leaders and broad public trust necessary to respond effectively. “We have been saying for a long time that deep polarization in the United States will be a real handicap in a major crisis, whether it was a war, environmental catastrophe or, as it turned out, a pandemic,” says Gentzkow.
For their study, Gentzkow and five collaborators — Hunt Allcott from New York University; Levi Boxell and Jacob Conway from Stanford; and Harvard University’s Michael Thaler and David Yang — undertook a two-part experiment.
First, they assembled publicly-available data on anonymized mobile GPS pings from SafeGraph, a location-tracking aggregator. The statistics, covering a ten-week period ending in late March, included measures of traffic to public locations like restaurants, retail stores, movie theaters, and hospitals.
They also collected county-level demographic information on age, race, education, income, and poverty status. Voting patterns from the 2016 election were used to determine political leanings by county.
Gentzkow’s analysis finds strong partisan differences in social distancing. This held even when the scholars controlled for alternative explanations, including population density and the severity of the outbreak.
“If you compare Republican and Democratic counties that look quite alike — say, for example, they are suburban and in the same state with similar income and education levels and have experienced similar outbreaks — you’ll see that people in the Democrat-dominant place are staying home more,” Gentzkow says. When the researchers compared traffic levels to the same period in 2019, they found no evidence of a partisan split.
He cautions that the data sample was not entirely random; it covered only specific apps downloaded to mobile phones and didn’t include users who had restricted location tracking.
A warning about finger-pointing
For the second part of the study, Gentzkow’s team conducted an online survey of 1,665 adults to understand the role that political beliefs played. Among other things, participants disclosed their party affiliation based on a seven-point scale, whether they saw a need to social distance, how much of it they were doing, how many confirmed cases they expected by the end of April, and what they thought President Trump’s approval rating would be at that time.
The results show clear partisan differences. For instance, Democrats reported a 25 percent chance of contracting COVID-19 without social distancing measures, while Republicans placed odds at 20 percent. When asked to predict future COVID-19 cases, Democratic estimates were 13 percent higher than Republican forecasts.
Next, the scholars set out to gauge whether responses reflected toeing the party line or personal beliefs. A portion of participants were told they would be paid $100 if their forecasts turned out to be correct. The idea is to incentivize people to think more deeply about their personal views, as opposed to giving knee-jerk responses that may be more politicized.
“Often, if you pay people a significant amount of money in surveys like these, partisan gaps shrink,” Gentzkow says. “What’s striking here is we saw the opposite: partisan differences got bigger, which suggests that the gap is more likely to be about beliefs.”
Gentzkow advises against jumping to the conclusion that Republicans are making a mistake by not social distancing more or that they don’t care. For some high-risk groups, like the elderly, the need to self-isolate is urgent. For others, like low-income workers, the need to show up for their service jobs outweighs the health risks of doing so.
“I think people need to think very carefully about the possible reasons for those differences in responses so we can all understand better what the consequences are,” says Gentzkow.