This paper estimates the link between population density and COVID-19 spread and severity in the contiguous United States. To overcome confounding factors, we use two Instrumental Variable (IV) strategies that exploit geological features and historical populations to induce exogenous variation in population density without affecting COVID-19 cases and deaths directly. We find that density has affected the timing of the outbreak, with denser locations more likely to have an early outbreak. However, we find no evidence that population density is positively associated with time-adjusted COVID-19 cases and deaths. Using data from Google, Facebook, the US Census and The County Health Rankings and Roadmaps program, we also investigate several possible mechanisms for our findings. We show that population density can affect the timing of outbreaks through higher connectedness of denser locations. Furthermore, we find that population density is positively associated with proxies for social distancing measures, access to healthcare and income, highlighting the importance of these mediating factors in containing the outbreak.