As nations like China and Singapore with more centralized governments and public health systems have been relatively effective in enforcing social distancing and containing coronavirus, the response in the United States has been fragmented and incoherent. California was the first state to issue a stay-at-home order on March 19, but as of April 8 five states still had no order and three others had orders for only some parts of the state.
Even as the last holdouts gradually give in and order their residents to shelter in place, data tracking mobile phone locations suggests that many Americans have not fundamentally altered their behavior. Unacast, a company specializing in the analysis of “human mobility data,” has put together a Social Distancing Scoreboard that measures average distance traveled as well as “nonessential visits” to venues like spas, cinemas, jewelers, and department and clothing stores.
The data is imperfect, but there is a strong signal amid the noise. As of April 7, for instance, the data showed that average mobility in the nation as a whole had fallen by more than 40 percent since February. But that figure obscures tremendous variation in behavior change across US states, and across counties within the same states.
Containing the spread of the coronavirus requires collective, unified action, but data on social distancing makes it clear this isn’t happening everywhere. The question is why. In what kinds of places are residents deciding to move about as if they are immune to the virus that has paralyzed much of the world? What do they look like, and why are they ignoring the calls for social distancing?
To get some hints, I put together several sources of data from US counties focusing on economic and demographic characteristics, voting patterns, civic engagement and social capital, and even attitudes toward climate change from Yale’s Climate Change in the American Mind Survey.
Analyzing the data reveals that social distancing behavior is related to education; race and ethnicity; to political identity and social capital; and to the impact that this virus has already had on the residents of particular counties. And the various sources of data also reveal a larger pattern.
One of the strongest and most robust predictors of social distancing behavior is found in attitudes toward another major challenge facing the United States: climate change. Places where residents are less likely to agree that global warming is happening, that humans are the cause, and that we have an obligation to do something about it are the places where residents haven’t changed their behavior in response to coronavirus. The analysis makes clear that we have a collective action problem much larger than Covid-19.
The uneven geography of social distancing
Almost 2 million people live in Bexar County, Texas, which contains the city of San Antonio and parts of several smaller towns and cities surrounding it. As of April 9, the county had over 500 cases of coronavirus and 18 deaths.
Mobile phone data from Unacast suggests the residents of Bexar County are taking the virus seriously. Travel has dropped by more than 55 percent since February, and nonessential visits have fallen by over 70 percent. Bexar is part of a string of counties in the south of Texas, running northward from San Antonio to Austin, where behavior has changed drastically since the emergence of Covid-19.
In the more sparsely populated counties just south of Bexar, however, residents haven’t scaled back their daily travel nearly as much. Atascosa, Wilson, and Gonzalez counties, all of which have confirmed cases of coronavirus, were each given a D in the Social Distancing Scoreboard. Further east, Harris County, the home of Houston, received a B mainly due to a sharp drop in nonessential visits. In neighboring Liberty County, where 14 cases have been confirmed, the decline in nonessential visits is not nearly as steep, and the county earned a D.
Although the letter grades provide a simple summary that we all understand, it is important to put the numbers in the scoreboard into some context. Social distancing is simply impossible for some segments of the population who need to move around for work or for essential food and supplies, and daily travel looks very different in rural areas where residents are spread out. Yet similar counties exhibit very different patterns of behavior, suggesting that some degree of choice, guided by local norms, is influencing behavioral change. At a moment when a unified effort is crucial to stopping the spread of this new virus, the uneven geography of social distancing may be the biggest threat to our success in containing Covid-19.
What kinds of places are more likely to be social distancing?
Data from the south of Texas shows a sharp divide between urban and rural counties, but the story of social distancing is not a simple story of population density. In fact, the lowest social distancing grades are found in non-metro areas adjacent to larger cities, rather than in rural, sparsely populated counties.
Other measures like the age profile of the county, the total size of the population, the racial and ethnic composition, and even the total number of confirmed cases of Covid-19 do not have a clear relationship with social distancing behavior. Different characteristics turn out to be much better predictors; the strongest predictors are shown in the chart above.
Counties that have experienced at least one death from Covid-19 are more likely to comply with social distancing guidelines, as one might expect. Grades are also higher in counties where residents have higher income, more education, and lower rates of unemployment (measured prior to the crisis).
Perhaps unsurprisingly, politics and civic engagement bear a strong relationship to social distancing behavior. Higher levels of social capital — a combination of voter turnout in federal elections, response rates in the 2010 census, the number of associations and the number of nonprofits per capita — is associated with more social distancing. By contrast, counties with the greatest share of votes for Trump in the 2016 election were least likely to practice social distancing. And the greater the share of residents who disagreed with the statement that global warming is happening, the worse the county’s grade received in the Social Distancing Scoreboard.
All of these characteristics of counties are intertwined in complex ways, which means the patterns in the chart are meaningful only in a superficial way. And some of the variation across counties might simply be driven by differences in behavior in blue and red states, where governors have taken more or less aggressive approaches to encouraging social distancing. To push further, it is useful to consider all of these county factors together, and to analyze variation in social distancing among counties within the same states.
To do so, I carried out a regression analysis with all of the county-level measures and including fixed effects for states — this approach means that the analysis focuses only on variation among counties within the same states. (I should note that the results are virtually identical if I get rid of the state fixed effects and make comparisons among all counties across the nation.)
When all variables are included in the same model, I find that counties that have experienced a death from Covid-19 exhibit higher levels of social distancing, but counties with more cases exhibit lower levels. Counties with larger populations, with more educated residents, and with higher percentages of white and Hispanic residents tend to receive higher grades on social distancing, while the age structure, the median income, and the unemployment rate are no longer associated with social distancing behavior.
Social distancing grades rise with the level of social capital in a county, and grades fall with the percentage of the county voters who cast a ballot for Trump in 2016. Lastly, even after adjusting for all of these other characteristics, counties within the same state where a greater share of residents do not agree that global warming is happening are substantially less likely to change their behavior in response to Covid-19.
In fact, attitudes toward climate change are one of the strongest and most robust predictors of social distancing behavior. In the full model I find that increase of 10 percentage points in the share of residents who do not agree that global warming is happening is associated with a 1 point drop in the county’s social distancing grade — which essentially means shifting from, say, a C to a B- in social distancing behavior.
The same results apply no matter how I run the analysis, and they apply to just about every question asked in the Climate Change in the American Mind Survey. In the places where residents don’t think global warming is real, where they don’t believe humans are responsible, where they don’t think citizens have a responsibility to act, residents are also failing to change their behavior during the coronavirus crisis.
The problem of collective action in the United States
America’s spatial divisions are being exposed and amplified during this period of crisis. Because Covid-19 isn’t bound by the administrative boundaries that divide one county from another, efforts to mitigate the spread of the virus in any given county can be undermined by residents one county over who are ignoring guidelines for social distancing. Divergence in social distancing behavior represents perhaps the most serious risk to our national efforts to overcome this crisis.
But the problem isn’t limited to the coronavirus. When we do get past this crisis, we will face a set of new challenges, along with others that have been around for decades but are becoming more urgent. How does our economy recover from Covid-19? How do we deal with the long-term rise of economic inequality, and the persistence of racial inequality and injustice? What do we do about the opioid epidemic? And how do we respond to climate change?
All these challenges require a collective, unified solution, and difficult decisions to protect the entire society and future generations. Yet we are a nation that has become increasingly divided along spatial lines. For decades, Americans have been encouraged to respond to major challenges like urban decline, social unrest, and environmental degradation by isolating themselves in areas that are increasingly segregated by class and by politics.
Covid-19 is exposing the limits of that response, but it also raises a question that is crucial to answer if the United States is going to prosper in the coming decades: In a divided nation, how do we come together — figuratively, for the time being — to solve collective challenges?
Patrick Sharkey is professor of sociology and public affairs at Princeton University’s Woodrow Wilson School of Public and International Affairs.
Support Vox’s explanatory journalism
Every day at Vox, we aim to answer your most important questions and provide you, and our audience around the world, with information that has the power to save lives. Our mission has never been more vital than it is in this moment: to empower you through understanding. Vox’s work is reaching more people than ever, but our distinctive brand of explanatory journalism takes resources — particularly during a pandemic and an economic downturn. Your financial contribution will not constitute a donation, but it will enable our staff to continue to offer free articles, videos, and podcasts at the quality and volume that this moment requires. Please consider making a contribution to Vox today.