Party identity and social-distancing behaviours in Brazil

Does political affiliation of local politicians determine citizens’ compliance to social distancing behaviours? I provide causal estimates of the effect of political identity of municipal mayors on regional differences in engaging in COVID preventive behaviours in the context of Brazil. I employ a sharp regression discontinuity design based on close mayoral elections in 2016 to examine the effects of having a mayor from one of three political parties that President Jair Bolsonaro is closely associated with, by combining Facebook mobility data that tracks regional movement relative to February 2020 levels with electoral data from the 2016 mayoral municipal elections. The methodology compares municipalities that are similar along a wide array of predetermined and observable correlates of the spread of coronavirus, and where the incumbent mayor was selected as-if randomly. I find that residents of Bolsonaro-affiliated municipalities exhibit 60% smaller relative declines in regional movement and are 13% more likely to cross regional boundaries over the months of March, April, and May in 2020. The findings hold for each of the three political parties, for each of the three months since the onset of the pandemic, and after controlling for anti-lockdown measures of 15 March 2020 in Brazil.