Recent research has attempted to document large-scale emotional contagion on online social networks. Despite emotional contagion being primarily driven by in-person mechanisms, less research has attempted to measure large-scale emotional contagion in in-person contexts. In this paper, I operationalize the temporal emotions associated with a particular city at particular points in time using sentiment analysis on Twitter data. Subsequently, I study how emotions converge between seven proximal cities in the state of Virginia, using two-way fixed effect models. I find that positive emotions tend to be synchronous between cities, but that effect is conditional on the level of contact between city residents at that period of time, as indicated by cell phone mobility data. I do not find any synchrony based on other types of emotions or general sentiment. I discourage drawing causal conclusions based on the presumed existence of several unmeasured sources of bias.
Keywords: emotional contagion; sentiment analysis; twitter data.