In a world where externalities are rife, transactions take time, and information is imperfect, market mechanisms alone will not guarantee efficiency. Trust in others' competence and decency, and concern for their well-being, can help economic actors to coordinate their behavior and achieve better outcomes.It is of some concern, then, that Americans feel a decreasing level of trust and increasing hostility along party lines. Gentzkow (2016) reports that, as of 2008, nearly half of Americans classified members of the other party as "selfish", up from around 20% in 1960. The same report shows that 20%-30% of Americans would be upset if their son or daughter married a member of the other party, up from around 5% in 1960. The difference in individual's warmth toward their own party and their warmth toward the other party, each on a scale of 0-100, has increased from around 25 points in 1980 to 45 points today . This hostility may be particularly problematic when members of different parties have to work together to combat a crisis, as has recently occurred during the COVID-19 pandemic. Social distancing poses a classic collective action problem: while the costs of preventative behavior accrue entirely to the individual, the benefits are diffused across a large number of people. It is in precisely this situation that social preferences might help to restore more efficient outcomes. If individuals are altruistic, they internalize some of the utility costs their actions have on other people. When altruism breaks down, society's ability to overcome the collective action problem may be compromised. While this argument has been made in the context of ethnic fractionalization (e.g., Poterba, 1997; Vigdor, 2004), until recently there has been little empirical work examining the consequences of political fragmentation for public goods provision. 1 In this paper, we fill this gap by examining whether political polarization has undermined the response to the COVID-19 pandemic in the United States.A simple comparison of social distancing across more or less polarized individuals or states will not identify the impact of polarization, because both polarization and willingness to comply with social distancing may be related to other, unobserved