What Drives the Effectiveness of Social Distancing in Combating COVID-19 across U.S. States?

Authors: Mu-Jeung Yang, David Eccles School of Business; Nathan Seegert, David Eccles School of Business; Maclean Gaulin, David Eccles School of Business; Adam Looney, David Eccles School of Business

Since the beginning of the COVID-19 pandemic, the responsibility for directing the U.S. response has been primarily delegated to local governments. As such, lockdown orders varied from mandatory (in California) to voluntary (in Utah) to none at all (in Arkansas), with significant variation in the application and severity of the orders at various levels of government. The state of Utah, for example, did not have a mandatory lockdown, but Salt Lake City did. 

To understand how effective social distancing practices have been at slowing the spread of coronavirus, researchers from the Marriner S. Eccles Institute for Economic Policy and Quantitative Analysis at the University of Utah developed a model to exploit these variations across the U.S. states and find that reliable, trusted government information on case counts is vital to the success of voluntary social distancing measures.

The importance of trust in case count data is reinforced through three main findings:

1. Information-based voluntary social distancing has saved three times more lives than lockdowns for the median state and four times more lives per 100,000 people when controlling for population. On average across states, a 1% increase in publicly reported coronavirus case count results in a 7% increase in social distancing. One of the key factors in the success of voluntary social distancing is that it tended to sharply depress mobility early on, and research shows that the earlier social distancing occurs, the more effective it is in reducing the spread of the disease. The bottom line is that citizens react to rising case counts by distancing themselves from others, and that has economic consequences but saves lives.
2. But there are major differences in social distancing across states, and bad information policies do much more harm than good information policies help. To establish this result quantitatively, the authors compare the number of lives saved by modeling, across all states, the voluntary social distancing responses of West Virginia residents—who were less likely to trust and respond to case count data—compared to the responses of relatively more informed residents of Massachusetts.
   
   The authors find Massachusetts residents increase their social-distancing by 20% for every 1% increase in cases. In West Virginia, social distancing only increases by 2.9% for every 1% in cases — or six times less than Massachusetts.
   
   If the West Virginia approach was applied across all U.S. states, we would have seen more than 240,000 additional fatalities. In contrast, applying the Massachusetts approach across all U.S. states would save an additional 24,000 lives.

• Whether lockdowns are more efficient than information-based voluntary social distancing depends critically on the contagiousness of social interactions. When social interactions are more contagious, the disease spreads faster. This, in turn, leads to more social distancing. In contrast, imposing lockdowns reduces infections in the first place so that subsequent mobility can be higher as the number of confirmed cases is lower. Quantitatively, the authors find that lockdowns that save the same number of lives as voluntary social distancing would have allowed over 24% more mobility for the median state. 

Methodology

To study the effects of social distancing, the authors combine the structural estimation of an infectious disease model (SIR model) with techniques from Machine Learning. This allows them to control for variables that are not often directly available but impact disease dynamics across states—for example asymptomatic transmission, symptom-based testing and quarantining, and the contagiousness of social interactions (i.e. how much disease transmission declines in response to social distancing). The last variable is especially important as it governs the degree of negative health effects from mobility and is a key parameter for policy analysis.

At the heart of the model is the assumption that private citizens balance the benefits from mobility against their expected losses from exposure to the virus, ignoring the unintended negative effects of unwittingly spreading the virus. The authors assume people form expectations about their probability of infection via mobility by using official case and fatality counts.

To quantitatively measure people’s response to case data, they capture the extent of social distancing by using cellphone-location based mobility data from Google. The Google mobility measures provide a daily-frequency comparison of mobility relative to the same calendar day in 2019 to control for general seasonal patterns. The authors focus on economically relevant categories of movement, such as mobility for work, grocery shopping, retail shopping (including restaurants), and transportation (such as public transit). They exclude categories such as visits to parks, since outdoor disease transmission is less common and mobility within parks has increased in some states during COVID-19.