How we did our research
In a recent study, we used data that Google collected from Feb. 15 to April 26 from anyone around the world who turned on their cellphone or tablet’s “location history” to use a Google app, no matter what kind of device they have. That allows Google to see how often people visit different types of locations.
But the public data isn’t very fine-grained. It lumps grocery stores and pharmacies into one category, and virtually every other type of business into another. This catchall category includes both businesses governors have closed and “essential” businesses that remain open, such as big box stores, liquor stores, dry cleaners and garden centers. If people are taking stay-at-home orders seriously, they should drastically reduce their visits to even these essential businesses, so we use this general category in our analysis.
The partisan gap in following pandemic recommendations
First, we used the raw daily data to examine how frequently people visited business locations before and after the pandemic hit the United States. We then looked at whether the size of the change was related to Trump’s share of the 2016 two-party vote by county.
The figure below shows that around the country, people significantly changed their daily lives. However, people in heavily Democratic counties (where Trump got 20 percent of votes) restricted their visits much more sharply than did people in heavily Republican counties (where Trump got 80 percent of votes). Before mid-March, people in both types of counties shopped in very similar patterns. But between March 13, when Trump declared a national state of emergency, and March 21, when California Gov. Gavin Newsom (D) issued the first state stay-at-home order, visits to non-grocery businesses in blue counties dropped significantly more than in red counties, until the gap reached about 20 percent, a gap that remained roughly stable for the next five weeks — except on Easter, when Republican counties also stopped moving around.
Of course, Democratic and Republican counties differ along other dimensions as well. But when we include demographic controls in our models — accounting for such things as education, population density, age and race — the difference in behavior between red and blue counties drops by only four percentage points, from 18 to 14 percent.
We even find this gap when counties are hit with more covid-19 cases. In both Democratic-leaning and Republican-leaning counties, shopping drops after a covid-19 outbreak — but a gap remains, with Democratic counties still that are much more likely to stay home.
The gap also holds when we take into account when governors started reacting to the pandemic. We checked whether blue counties are more likely to be in states where governors closed businesses earlier, which might cause the partisan gap. The answer is no.
Party and education lead to different responses to covid-19 cases and governors’ restrictions
So what accounts for these partisan differences? We don’t know for sure. We are investigating various possible explanations, including differences in media consumption and local law enforcement practices. But we did find two noteworthy factors.
You can see the first in the figure below, which shows how shopping in counties tracked the rise of covid-19 cases. Along the x-axis, you can see that for most people, visits to stores decreased as more county residents tested positive for covid-19. But that drop varied both by the county’s partisan leanings, and by its average education level. For the figure, we define “low education” as counties where high school dropouts outnumber college graduates by 5 percent and “high education” as counties where college graduates outnumber high school dropouts by 20 percent.
In Democratic counties, people were more likely to stay home as covid-19 cases rose, no matter what percentage of residents had college degrees. That wasn’t true in Republican counties. As the virus spread, in red counties, people with high education were more likely to stay home, although not as much as people in blue counties. People in Republican counties with low education, however, paid no attention to the growing threat; they continued to go about their daily lives.
Even when a governor issued a directive to stay home, the pattern remained: Pro-Trump counties with low education were much less likely to change their behavior than others.
Our analysis uses county data because Google didn’t release data about smaller units like census blocks. Our findings do not necessarily apply to individuals; Democrats living in heavily red counties may act just like their Republican neighbors, and vice versa in heavily blue counties. If so, people feel pressure to behave as people around them behave, no matter their party. We plan to investigate this possibility.
In the meantime, this analysis suggests partisanship is responsible for the fact that people living in Republican counties have changed their behavior less than those living in Democratic counties. People look to their leaders for cues about how to behave, particularly in times of crisis.
Keena Lipsitz is an associate professor of political science at Queens College, City University of New York and the author of “Competitive Elections and the American Voter” (University of Pennsylvania Press, 2016).
Grigore Pop-Eleches is a professor of politics and international affairs at Princeton University and co-author with Joshua Tucker of “Communism’s Shadow: Historical Legacies and Contemporary Political Attitudes” (Princeton University Press, 2017).