We estimate the impact of U.S. political identity on COVID-19 safety behaviors, cases, and deaths. Our data set merges U.S. county-level data on mask-wearing, cell-phone mobility, vaccination rates, county characteristics, and variables reflecting conservative political identity with COVID-19 cases and deaths from the first 12, 20, and 28 months of the pandemic in the United States. State-level fixed-effect estimations controlling for county characteristics indicate every 10 percentage point increase in the county popular vote for President Trump in the 2020 election to be associated with a 0.36σ reduction in a COVID-safety index, 1,798 additional COVID cases, and 31.9 COVID deaths per 100,000 county residents in the initial 28 months of the pandemic. Further, we ask whether differential behavioral responses during the pandemic can be explained by traditional strains of American conservatism, or are associated with a more specific Trumpian identity. We create state-level indices of traditional libertarian and social conservatism, finding that these indices display little systematic explanatory power over COVID-safety behaviors, cases, and deaths relative to 2020 Trump voter support.