Public transit gains and spatially uneven travel demand changes after NYC congestion pricing

New York City's congestion pricing program, the nation's first cordon-based scheme launched in January 2025, produced a significant rise in bus and subway ridership while overall travel demand fell modestly, according to a new analysis that used machine-learning counterfactuals to isolate the policy's effects [1]. The study, posted to arXiv on June 16, 2026, applied time-series foundation models to generate probabilistic demand forecasts with calibrated uncertainty, sidestepping the difficulty of constructing credible control groups for a system-wide intervention [1][2]. Researchers found that post-policy transit gains extended beyond Manhattan's core, even though reductions in aggregate travel demand were concentrated within the Congestion Relief Zone [1][2]. Spatial heterogeneity was a defining feature of the outcome. "The effects are spatially heterogeneous: while reductions in overall travel demand are concentrated within the Congestion Relief Zone, transit gains extend beyond Manhattan's core," the authors wrote [2]. Socio-demographic breakdowns further revealed uneven adaptation across neighborhoods, raising questions about spatial equity that the paper flags for future policy design [1][2]. The framework itself represents a methodological departure. Because congestion pricing generates spillovers across modes and locations, traditional difference-in-differences approaches struggle to isolate causality. The authors instead built counterfactual no-policy trajectories using foundation models trained on pre-intervention data, then compared observed post-policy volumes against those baselines [1][2]. The technique, they argue, offers a scalable template for uncertainty-aware evaluation of other urban interventions where clean control groups are unavailable [1][2]. New York's program charges vehicles entering Manhattan's central business district, a cordon-based design that differs from the lane-based or facility-specific tolls used in cities such as London, Stockholm, and Singapore. The January 2025 launch followed years of political and legal wrangling, making the city a closely watched test case for U.S. municipalities considering similar pricing structures [1][2]. The paper does not provide raw ridership counts or percentage changes in its abstract, describing the transit increase only as "significant" and the demand decrease as "modest" [1][2]. Full results, including neighborhood-level adaptation patterns, are detailed in the manuscript hosted on arXiv [1].