Systematic Biases in Aggregated COVID-19 Growth Rates

The COVID-19 pandemic has emerged as one of the greatest public health challenges of modern times. Policy makers rely on measuring how quickly the disease is spreading to make decisions about mitigation strategies. We analyze US county-level data about confirmed COVID-19 infections and deaths to show that its impact is heterogeneous. A small fraction of counties represent the majority of all infections and deaths. These hot spots are correlated with populous areas where the disease arrives earlier and grows faster. When county-level data is aggregated to create state-level and national statistics, these hot spots systematically bias the growth rates. As a result, infections and deaths appear to grow faster at those larger scales than they do within typical counties that make up those larger regions. Public policy, economic analysis and epidemic modeling have to account for potential distortions introduced by spatial aggregation.