• Congestion pricing reduced taxi trips by 4.56% in the U.S. metropolitan area. • XGBoost model outperformed MLR, showing 50% better accuracy in trip forecasting. • Machine learning enhances policy monitoring; public transit investment is crucial. Urban congestion continues to challenge large metropolitan areas, often exacerbating travel times, air pollution, and economic inefficiencies. In 2025, in an effort to combat these challenges, a major U.S. metropolitan city introduced a congestion pricing policy that targets congestion reduction, emission control, and transit funding. This study used pre- and post-policy taxi trip data to assesses the effectiveness of that policy on taxi travel within the city’s central business district and used a two-prong approach involving Multiple Linear Regression (MLR) and machine learning-based XGBoost models to quantify its impact. Both models were trained on historical data that was gathered prior to the policy’s implementation and incorporated temporal features and external demand signals to forecast trip counts. A comparison of their forecasts with the actual trip counts observed during the policy period revealed a significant reduction in taxi trips within the charged zone, with the XGBoost model providing forecasting accuracy superior to the MLR model. The study also emphasizes the advantages of using machine learning techniques to improve forecasting and evaluations of urban transportation systems policies, providing key insights for cities considering similar interventions.
Javaheri et al. (Thu,) studied this question.