Since the early stages of the COVID-19 outbreak, the spatial spread of the virus across national and regional borders has been of paramount concern (Biancotti et al. 2020, Valsecchi and Durante 2020). Meninno and Wolff (2020) anticipated the implications of the spread, analysing the economic impact of European countries closing their borders. Recently, Eckardt et al. (2020) found that border closures inside the Schengen Zone mitigated the spread of the virus in the early weeks of the pandemic. Unlike the countries in the EU, US states cannot shut down their borders. Yet, they have the freedom to ultimately decide their own social distancing mandates, business closures, large gathering bans, or shelter-in-place orders.
States’ freedom to decide their own lockdown policies, combined with the inability to close their borders, creates a clear problem – a lax policy in one state can exacerbate the outbreak in neighbouring states and in the rest of the country. For example, recent research indicates broad-based interstate spread of COVID-19 following the Sturgis Motorcycle Rally in South Dakota in August 2020 (Dave et al. 2020). While the example of Sturgis is particularly conspicuous, how large is this problem in more general settings? In our recent research (Rothert et al. 2020) we find that it can be substantial.
We analyse the county-level data on confirmed cases and deaths from the COVID-19 Data Repository by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University, as well as state-level data on business closures and mandated social distancing measures compiled by the Institute for Health Metrics and Evaluation (IHME). The IHME metrics include the date on which a state forbade mass gatherings, introduced an initial round of business closures, closed schools, closed all non-essential businesses, and adopted a stay-at-home order. For each day in our time series, we sum the number of restrictions currently imposed within a state to generate a measure of government-imposed behaviour restrictions at the state-level. We call this new variable, taking values 0 through 5, the r-score.
Restrictions vary across states while the virus spreads
Figure 1 displays the transition from no official containment measures in early March (r-score = 0) to universal adoption (of at least some measures) by all states by early April (r-score varies between 2 and 5). The most dramatic change took place between mid-March and early April. 60% of the US population lived in states with no containment measures on 15 March. By 1 April, almost 60% of the population lived in states that had adopted all five factors. At the same time, however, we see that the containment measures vary across the US states. More than 40% of the population lived in states without at least one of the measures described above, and almost 10% in states without at least two of those measures.
While the restrictions varied across states, the virus spread across state borders. Figure 2 displays that spread. Within only a month, COVID-19 transitioned from a fairly sparse outbreak to a widespread epidemic, affecting almost every county in the US.
Illustrative examples of local spillovers
New York was the first state in the US to experience a very significant outbreak, with its epicentre in New York City (NYC). The dynamics of case numbers in and around NYC suggest the presence of interstate spillovers (within the CT-NJ-NY-PA areas). Figure 3 shows a rapid expansion of per-capita case numbers in New York during the second half of March, followed by all other states surrounding the NYC metropolitan area in late March and the first half of April.
We can see similar patterns at the county-level. Figure 4 plots confirmed county-level cases per-capita over time for four metropolitan areas. In each case, the county containing the urban centre appears to trigger the outbreak; for example, Orleans Parish, LA – which contains New Orleans – experienced an early surge that was quickly followed by outbreaks in the surrounding areas.
We first estimate the magnitude of spatial spillovers using a variety of state-of-the-art spatial econometrics models. We find consistent evidence that not only did new cases diffuse across county lines, but that the diffusion across counties was affected by the closure policies of adjacent states. Our estimates suggest total spatial spillover of a one-unit increase of the r-score has a statistically significant and negative effect on county-level cases.
Next, to measure the dynamics of the spatial spillovers we generate ‘spatial impulse response functions’, showing how long a particular county was affected by its neighbours' rate of new cases. We find that the spatial diffusion of new cases is statistically significant and persistent over time, where an increase in a county’s nearest neighbour predicts an increase in cases for at least ten days over our forecast horizon. Likewise, for an increase in the r-score, the growth of new cases in a county declines over the same forecast horizon.
Our results provide an informative picture on the nature of the spatial correlation of the COVID-19 phenomenon. They also confirm previous findings (Deb et al. 2020, Hartl et al. 2020) that social distancing measures work: more stringent state-level restrictions are consistent with a decline in the growth rate of new cases at the county-level.
Next, we analyse the potential impact of local lockdowns on country-wide infections. We develop a spatial epidemiological model where new infections arise from interactions between infected people in one state and susceptible people in the same or in neighbouring states. Social distancing measures can potentially slow that spread down. Model simulations suggest that, had the states with the less restrictive social distancing measures tightened them by one level, the cumulative infections in the remaining states could be about 2% smaller by the end of June 2020, and 5% smaller by December 2021. Figure 5 shows how the percentage changes translate into changes in the number of infected people.
Fragmented by policies, united by outcomes
The states may be fragmented by policies, but they all bear at least part of the consequences of each other’s actions and inactions. Table 1 shows that effect. We divide states into 3 groups, based on the restrictiveness of their social distancing policies as of 30 June 2020. We then ask what would happen if: (1) all states in group 1 adopted the (more restrictive) policies from group 2, (2) all states in group 2 adopted the (more restrictive) policies from group 3, and (3) all states in group 3 adopted the (less restrictive) policies from group 2. We find that a change in social distancing policy within a group of states has a sizeable impact outside of that group.
Table 1 Proportion of population infected by December 2021: Model simulations
Notes: Group 1 – least restrictive policies; Group 3 – most restrictive policies (as of 30 June 2020); numbers in cells are medians within each group.
The presence of inter-state spillovers significantly affected the rate at which COVID-19 spread across the US. A unique feature of the US is that its federal government cannot compel individual states to close their borders nor mandate state-specific lockdown policies; therefore, this result is important when we evaluate the ‘performance’ of different regions in battling the pandemic.
In an environment where states may not coordinate automatically, our findings emphasize the importance of other tools. The federal government can use fiscal tools to promote coordination between states' authorities (Rothert 2020). It can also impact the behaviour of residents nationwide with consistent messaging on precautionary measures, such as mask-wearing (Chernozhukov et al. 2020) and encouraging compliance with social distancing guidance. Given that by the very nature of the problem any action or inaction in response to a viral outbreak creates external effects on surrounding regions, we believe this is a very important area for further research.
Biancotti, C, A Borin, F Cingano, P Tommasino and G Veronese (2020), "The case for a coordinated COVID-19 response: No country is an island", VoxEU.org, 18 March.
Chernozhukov, V, H Kasahara and P Schrimpf (2020), "Mask mandates and other lockdown policies reduced the spread of COVID-19 in the US", VoxEU.org, 15 July.
Deb, P, D Furceri, J D Ostry and N Tawk (2020), "The effect of containment measures on the COVID-19 pandemic", VoxEU.org, 05 June.
Dave, D, A Friedson, D McNichols and J Sabia (2020), “The Contagion Externality of a Superspreading Event: The Sturgis Motorcycle Rally and COVID-19”, IZA Working Paper No. 13670.
Eckardt, M, K Kappner and N Wolf (2020), “Covid-19 across European Regions: The Role of Border Controls”, CEPR Discussion Paper 15178.
Hartl, T, K Wälde and E Weber (2020), "Measuring the impact of the German public shutdown on the spread of COVID-19", VoxEU.org, 14 April.
Meninno, R and G Wolff (2020), “As coronavirus spreads, can the EU afford to close its borders?” Economics in the Time of COVID-19, 87.
Rothert, J (2020), “Optimal federal redistribution during the uncoordinated response to a pandemic”, United States Naval Academy Department of Economics Working Papers 64.
Rothert, J, R Brady and M Insler (2020), “The Fragmented United States of America: The impact of scattered lockdown policies on country-wide infections”, Covid Economics 43: 42-94.
Valsecchi, M and R Durante (2020), "Internal migration and the spread of COVID-19", VoxEU.org, 02 September 2020