Statistical analysis of effective COVID-19 government response policies: insights from pre-omicron pandemic data

The COVID-19 pandemic has persisted as a monumental tragedy in the history of the United States. While some consequences of the pandemic made an immediate impact on American life, subtle problems such as pandemic policy response have slowly sculpted political separations that have scarred the American political landscape. Even today, uncertainty persists as to which U.S. pandemic policies successfully mitigated the spread of the disease, creating the potential for an unrefined government response in future pandemics. This paper aims to address this potential issue by statistically identifying pandemic response policies that reduced disease transmission when enforced with increased stringency. To find these policies, four regression methods (Ridge, Lasso, Elastic Net, and Bayesian Lasso) were applied to state-level policy data sourced from the Oxford COVID-19 Government Response Tracker and COVID-19 case data gathered by Johns Hopkins University. All four regression models agreed that three key policies proved pivotal in preventing the propagation of COVID-19. A k-means clustering analysis was conducted on the stringency data for these policies, revealing a clear political divide in their enforcement. The application of heat maps to state-level policy data for each identified cluster indicated that one particular cluster, mostly aligned with a specific political party, enforced these key policies with greater rigor. Furthermore, the results of the Mann–Whitney U test confirmed that this cluster experienced a lower average rate of daily COVID-19 cases, indicating that one political party was more successful in mitigating the transmission of COVID-19.