Understanding the influencing factors of electric vehicle penetration in China: a city-level analysis

Abstract

Rapid electric vehicle (EV) adoption is central to decarbonizing road transport, but national averages obscure substantial city-level differences, limiting understanding of the local conditions that shape market diffusion and its carbon implications. We assembled a city-year dataset covering 296 Chinese cities from 2019 to 2024 and compiled 21 candidate predictors spanning climatic, environmental, socioeconomic, urban, energy-price, infrastructure, and market-maturity dimensions. Following variable screening, 13 substantive predictors were analyzed using a random forest model with city-grouped validation. SHapley Additive exPlanations (SHAP), a 300-replicate city-block bootstrap, region-specific models, and two alternative model specifications were used to interpret predictions and assess uncertainty and sensitivity. EV penetration increased from 4.0% in 2019 to 44.9% in 2024, while substantial spatial disparities persisted. The model generalized well to held-out cities, achieving an R2 of 0.902, an RMSE of 0.048, and an MAE of 0.027. EV stock share was the most important substantive predictor, followed by gasoline price and charging infrastructure stock. In the dynamic specification, the one-year-lagged penetration rate became the dominant predictor, indicating that EV stock share primarily captures adoption momentum and path dependence rather than an independently identifiable peer effect. Predictor rankings also varied across regions, particularly for gasoline prices and charging infrastructure. These findings describe predictive associations rather than causal effects, but they highlight the importance of considering local market maturity, energy prices, and infrastructure conditions when investigating region-specific electrification strategies. The carbon implications of these strategies should also be assessed in relation to regional grid carbon intensity and powertrain composition.