Causal Inference
Causal Effects of Mobility Patterns on Pollution Exposure Mai Waziry* Mai Waziry
Purpose
Individuals move daily, and during this movement, they are exposed to air pollution. Most previous studies relied on static measurements to infer pollution exposure. Individuals’ mobility patterns differ based on lifestyles and the longevity of their activities. This paper focuses on filling the gap in understanding the causal effects of exposure to PM2.5 on the outcome of individuals’ mobility patterns, focusing on outdoor activities. The mobility dataset from Spectus, offered through the Data for Good program, includes almost 8 million observations in Boston and Chicago.
Method
This paper uses a framework with two main steps. First, an instrumental variable analysis is used to close the backdoor between exposure and outcome. Two instrumental variables are used: wind speed and elevation. The second step in the framework is causal mediation analysis, which infers direct and indirect effects between exposure to PM2.5 and the outcome of outdoor activities through the mediator, which is the distance from an individual’s home to the activity location.
Findings
The results from the instrumental variable (IV) analysis using wind speed show that increased daily wind speeds reduce individuals’ PM2.5 exposure, with the relevance and validity assumptions for wind speed as an IV being satisfied. However, elevation, initially considered as an IV, was found to act as an interaction term with pollution exposure rather than a valid IV. The findings for the IV analysis are consistent across both cities. The results provide an understanding of the spatiotemporal patterns of mobility and shed light on how and why responses can play a critical role in disaster mitigation and prevention.