A Causal Inference Method Based on Front-Door Criterion and Difference in Differences for Analyzing Traffic Conditions

In the period of pandemic, various regions implemented distinct control measures, leading to subtle changes in road traffic patterns. These fluctuations were further influenced by external factors such as public risk perception and weather conditions. Although prior research has employed observational methods to assess traffic states and applied causal inference techniques to passive data, the fusion of these approaches in the field of transportation remains relatively unexplored. This study endeavors to address this research gap by investigating the causal impact of pandemic containment policies on urban traffic in multiple cities across China. Our analytical framework combines the Additive Noise Model (ANM) algorithm with rigorous statistical tests designed to capture casual relationship. Specifically, we employ the front-door criterion and a difference-in-differences approach to validate these causal connections. Refutation tests and placebo tests are conducted. Ultimately, our methodology yields a generalized linear model optimized for predicting traffic patterns under varying intervention scenarios. Our research thus paves the way for incorporating causal inference methodologies into the formulation of resilient transportation strategies amid public health crises.