The weeks following the initial outbreak of COVID-19 in the U.S. saw an outpouring of news on increased stress and anxiety levels in citizens. While these feelings appear personal, they are often collective responses to disease outbreaks and the circumstances they create and, historically, it is not unusual for them to find their outlet in social unrest. 

In fact, a recent study revealed that areas visited by epidemics show a 21% increase in their likelihood of seeing a social unrest event (i.e. a riot or protest) and a 57% increase in their likelihood of seeing two social unrest events. 

Statistical modeling

As COVID-19 spread across the globe, the study’s authors, Rebecca Cordell (assistant professor at the University of Texas at Dallas), Thorin Wright (associate professor at Arizona State University), and Reed Wood (reader at the University of Essex), followed the trail of social upheaval it left in its wake. In looking through historical examples of outbreaks, such as the cholera, Ebola, and influenza epidemics, they identified similar patterns. 

Interested, they decided to interrogate the link statistically, using data collected on 48 African countries from 1990 to 2017. The data comes from the Social Conflict Analysis Database (SCAD) and the Center for Research on the Epidemiology of Disasters.

Their approach uses a negative binomial model to estimate the level of unrest in each locality. Because riots and protests are relatively rare events, most of the place-data were assigned zeros.

To limit bias generated from confounding factors, the researchers use the Mahalanobis Distance Matching method. Previous studies have shown that there are a number of factors, aside from the outbreak itself, that can affect the following social outcomes.

For example, population size, healthcare infrastructure, and level of economic development affect a government’s ability to respond to outbreaks, which in turn can affect the public’s satisfaction with the government and their decisions to riot or protest.

To account for these variances, the method pairs similar places together, one that has experienced an epidemic and one that has not, based on shared characteristics — essentially creating control groups, or counterfactuals. Of the 21,756 African localities listed in the datasets, they were able to form matches between 3,148 of them.

The authors also account for preexisting social tensions, which can be exacerbated in times of crises, as a result, for example, of the unequal distribution of medical resources to already marginalized communities.

“[Another] potential pathway for especially violent unrest outcomes is the targeting of specific populations that are associated, usually incorrectly, with the outbreak and are scapegoated,” says Wright. “And those kinds of pathways for really negative social interactions need to be taken into account as well as disparities.”

SCAD’s data is limited geographically to Africa and Latin America. But because of Africa’s diversity, the researchers were able to test their models in a variety of societal structures. 

“Africa is extremely diverse in terms of both economic development and political institutions. And so, it presents in some ways a very good sample for what we're looking at,” says Wright. 

A first step

Like many researchers, this team has been doing their post-COVID-19 research on laptops at home. And although their process was more time-consuming than usual, they were able to run complex analyses using resources like Jetstream.

In this way and after a lengthy geocoding and data formatting process, they found their main result: “When a first-order administrative unit [the highest subcategory below country] in our sample has a disease outbreak, there is an increase in social unrest events — the number goes up,” says Cordell. “We also find a strong relationship for riots as opposed to protests when we break that relationship down.”

The latter finding came as a surprise. It seemed strange that the rate of violent riots would increase, while that of non-violent demonstrations changed little. For what reason would epidemics influence the two unequally?

Although the question requires further research, they offer a possible explanation in their currently unpublished paper: “Unlike grievances that develop at a slower pace, the rapid spread of disease within a community can result in sudden changes to everyday life and individual freedoms (e.g., quarantines, business closures, etc.). These abrupt restrictions are more likely to lead to visceral, violent outbursts by an anxious, uncertain community that is often unprepared and ill-informed about authorities’ motives.”

Altogether, their research makes a large step toward situating some of 2020’s prevalent feelings of fear, anxiety, and distrust within historical patterns. And their identification of many of the theoretical mechanisms, or steps, linking these responses to unrest provides the groundwork for causal investigation:

 “I think what our study does, and it's only a very first step, is to start to showcase what some of those implications [of the link] are,” says Wright. “I think this is a lesson that folks are kind of aware of, especially after the course of the last year, is that it’s not simply a matter of medical treatment and protective measures without social consequences.”