The Emotional and Cognitive Impacts of Air Pollution: Evidence from Twitter

Abstract

I study whether heightened air pollution leads to emotional and cognitive responses at the infra-marginal level. To address this question, I employ geolocated, timestamped Twitter microdata. Using an original dataset of over 30 million unique Tweets, I observe linguistic responses to varying levels of pollution across the U.S.. I find that Tweets from higher-pollution backgrounds are more negative and aggressive than Tweets from observably similar backgrounds with less pollution. Additionally, I find evidence that higher-pollution Tweets score cognitively lower for some groups, but otherwise have little to no effect. I find that lower cognitive-scoring users Tweet less as air quality worsens, but individual users Tweet at a lower level. I also find that originally negative and cognitively low-scoring Tweets are more vulnerable to air pollution’s negative effects than high-scoring Tweets.