Social distancing, involving measures to reduce close contact between individuals, is a key strategy to curb the spread of COVID-19. Past experiences with the 2009 H1N1 influenza and Ebola outbreaks demonstrate its effectiveness in reducing disease transmission. While China implemented large-scale quarantines, the US response was largely decentralized, with states and localities adopting varying social distancing policies, including emergency declarations, work-from-home mandates, school closures, business closures, restrictions on gatherings, and shelter-in-place orders. Although various studies explored the relationship between these policies, mobility changes, and case growth, the specific effectiveness of individual policies remained unclear. This study aims to quantitatively assess the impact of different state-level social distancing policies on both population mobility and subsequent COVID-19 case growth, providing crucial insights for policymakers seeking to balance infection control with economic and social considerations. The availability of anonymized, aggregated mobility data from Google users offers a unique opportunity to conduct this analysis.

Prior research on the effectiveness of social distancing measures in controlling infectious disease outbreaks provided a foundation for this study. Studies on the 2009 H1N1 pandemic and the Ebola outbreak highlighted the positive impact of social distancing on reducing transmission rates (Chowell et al., 2011; Peak et al., 2018). In the context of COVID-19, several studies investigated the correlation between social distancing policies, changes in mobility, and case growth trajectories. For example, Courtemanche et al. (2020) showed the impact of strong social distancing measures in reducing the growth rate of COVID-19 cases. Gao et al. (2020) linked changes in mobility from SafeGraph data to COVID-19 infection rates. Similarly, Badr et al. (2020) used anonymized cell phone data to establish a correlation between reduced mobility and COVID-19 case growth. However, a comprehensive quantitative assessment of the relative effectiveness of individual social distancing policies across the diverse US context remained lacking.

This study utilized a regression discontinuity approach to analyze the impact of social distancing policies on population mobility and subsequent COVID-19 case growth. Data on state-level policies were compiled from official government documents, cross-referenced across multiple sources to ensure accuracy. The primary mobility data source was anonymized and aggregated Google user location data, identical to that used in the publicly available Google COVID-19 Community Mobility Reports. This data included metrics such as time spent away from places of residence, visits to workplaces, retail and recreational locations, grocery stores and pharmacies, parks, and transit stations. Data were aggregated at the county level, and daily changes were compared to a pre-COVID baseline. A regression discontinuity design was employed to estimate the effect of policy implementation on mobility in the week following enactment, compared to the preceding week. County-level data were analyzed using regression discontinuity to estimate the causal effects, while state-level estimates reflected population-weighted averages of county-specific results. Finally, a linear mixed-effects model was used to assess the association between changes in mobility and subsequent changes in COVID-19 case growth rates at the county level, accounting for temporal trends and variations across states. The model incorporated data on case growth rates over multiple weeks following the mobility changes to capture the lagged effect.

The study revealed a strong association between social distancing policies and reduced population mobility, which in turn correlated with decreased COVID-19 case growth. State-level emergency declarations alone led to a 9.9% reduction in time spent away from residences. The implementation of one or more social distancing policies resulted in an additional 24.5% reduction in mobility, with shelter-in-place orders yielding a further 29.0% decrease. A 10% reduction in mobility was associated with a 17.5% decrease in case growth two weeks later. Notably, the study found significant variations in the effectiveness of different social distancing policies. While closing bars and restaurants proved to be the most effective single policy in reducing mobility, the combination of multiple measures resulted in even more pronounced reductions. The impact of policies varied across states and counties, potentially influenced by factors such as pre-existing local policies or variations in adherence to state mandates.

This study provides strong quantitative evidence for the effectiveness of social distancing policies in reducing COVID-19 transmission. The findings confirm and extend previous research by providing precise estimates of the impact of different policy types on mobility and subsequent case growth across a large, diverse population. The significant reduction in mobility observed following the implementation of social distancing measures strongly supports the use of these policies to control the pandemic. The variation in effectiveness across different policies suggests that policymakers should carefully consider the combination and sequencing of policies to maximize their impact. Furthermore, the strong correlation between mobility changes and subsequent case growth highlights the potential of aggregate mobility data as a leading indicator for predicting COVID-19 risk, allowing for proactive interventions.

The study demonstrates a strong link between state-mandated social distancing orders, reduced population mobility, and decreased COVID-19 case growth. The findings underscore the effectiveness of social distancing measures, particularly the combination of multiple policies, in mitigating the spread of the virus. Aggregate mobility data prove valuable as a leading indicator of future COVID-19 risk, suggesting the potential for proactive interventions based on these data. Future research could explore the heterogeneous impacts of policies across different demographic groups and geographic areas, further refining the understanding and optimization of social distancing strategies.

The study's findings should be interpreted in the context of several limitations. First, the mobility data is limited to Google users with Location History enabled, which might not fully represent the population's mobility patterns. Second, differential privacy algorithms were used to protect user anonymity, potentially introducing some level of noise into the data. Third, the analysis focused on state-level policies, while local variations in policy implementation might have influenced the observed effects. Fourth, comparisons across different locations were descriptive rather than causal due to potential confounding factors.