Urbanization inequality: evidence from vehicle ownership in Chinese cities

The study addresses how urbanization in China has produced unequal outcomes in mobility infrastructure, specifically in vehicle ownership, and examines when and whether China might reach vehicle ownership levels comparable to developed countries. Despite rapid growth in total vehicle stocks since 2000, per capita vehicle ownership in China remains much lower than in developed economies, raising sustainability and equity questions. Prior work has emphasized economic inequality, but less is known about inequality in in-use infrastructure stocks such as vehicles. The paper situates vehicles within socioeconomic metabolism as in-use stocks that provide essential services and link material cycles to manufactured capital. Given pronounced regional disparities in China’s economic development and urbanization, the study aims to characterize spatiotemporal patterns and inequalities of vehicle ownership across 283 prefecture-level cities from 2001–2018, identify growth stages, and explore determinants, thereby informing forecasts and sustainability strategies.

City-level research in China has largely focused on determinants of private vehicle ownership using econometric models, highlighting the roles of income/GDP per capita, city scale, road supply, urban form, population density, and built-up area (e.g., Li et al., 2010; Cao and Huang, 2013; Wu et al., 2016; Yang et al., 2017). However, these studies often overlook temporal evolution and spatial disparities in ownership rates. Within the socioeconomic metabolism framework, in-use stocks (vehicles, buildings, infrastructure) are central to service provision and material cycles (Pauliuk and Müller, 2014; Chen and Graedel, 2015). National- and provincial-scale projections of China’s vehicle stocks vary widely due to differing assumptions on saturation levels and historical data, with saturation ranges often assumed at 0.20–0.80 vehicles per capita (Hao et al., 2011; Gan et al., 2019). There is a gap for long-term, city-level analyses that capture the spatial-temporal inequality patterns, growth trajectories, and city stages of motorization to improve forecasts and policy design.

Data: Country-level vehicle ownership and population data were compiled from the UN Statistical Yearbook, World Bank, ACEA, OICA, and national statistical offices (China, Japan, US, UK). In China, vehicle ownership is defined as registered in-use vehicles (passenger, trucks, and low-speed vehicles) with local traffic administrations; registration approximates in-use stocks and includes private and commercial vehicles, excluding hibernating stocks. City-level data for 283 prefecture-level cities (including Beijing, Shanghai, Tianjin, Chongqing) span 2000–2018 and include vehicle ownership, resident population, real GDP (2000 constant price), and urbanization rate from national/provincial/city statistical yearbooks and bulletins. The dataset covers in 2018 about 93% of population, 96% of vehicle ownership, and 97% of GDP. Regions: Cities are grouped into 8 regions (North Coast, Northeast, East Coast, Middle Yellow River, South Coast, Northwest, Southwest, Middle Yangtze River), with additional focus on four urban agglomerations (Beijing–Tianjin–Hebei, Yangtze River Delta, Pearl River Delta, Chengdu–Chongqing). Inequality: The Theil index is used to quantify inequality in vehicle ownership rates (vehicles per capita) across cities, with decomposition into within-region (T_WR) and between-region (T_BR) components. Growth pattern identification: Time-series characterization uses level (vehicles per capita), speed (year-to-year change), and acceleration (change in speed). ARIMA is applied to test stationarity/trends of acceleration and speed, classifying cities into four types: Type-A (linear growth; stationary acceleration around zero), Type-B (parabolic; positive stationary acceleration), Type-C (rapidly increasing acceleration), and Type-D (decelerating; acceleration decreases from positive to negative). Correlation analysis: Pearson correlations (R) between vehicle ownership rate and GDP per capita (2000 CNY) and urbanization rate (%) were computed in R to assess associations and identify cities likely to see faster future growth.

• National and international context: China’s total vehicle stocks grew at an average 15% per year since 2000, reaching 273 million in 2020, the largest globally. On a per-capita basis in 2020, China’s ownership rate was 0.19 vehicles per capita versus 0.83 in the US and 0.47 in South Korea. • Regional distribution and growth: In 2018, vehicle stocks by region were: North Coast 45.8 million, East Coast 37.3 million, Middle Yellow River 30.9 million, Middle Yangtze River 29.0 million, with the Northwest lowest at 6.9 million. Annual growth rates declined around 2009; by 2018 the Middle Yangtze River had the highest growth (15.1%), Northwest the lowest (8.4%), national average 11.3%. • Urban agglomerations: Four major agglomerations (Yangtze River Delta, Beijing–Tianjin–Hebei, Pearl River Delta, Chengdu–Chongqing) held about 40% of national stocks. In 2018, Yangtze River Delta had 37.6 million vehicles; Chengdu–Chongqing had 14.0 million. • City-level variation: Top vehicle-stock cities in 2018 were Beijing (5.7 million), Chengdu, Chongqing, Shanghai, and Suzhou; Ezhou had 68.1 thousand. Growth slowed in megacities (Beijing, Guangzhou, Tianjin, Hangzhou, Shenzhen) due to license-plate restrictions. • Spatial inequality: Vehicle ownership rates expanded but remained uneven, with coastal and provincial capital cities higher. In 2018, highest per-capita ownership was in Hohhot (0.343), Jinhua (0.336), Dongying (0.320), and Urumqi (0.319). Regional gap in ownership rates in 2018: East Coast highest; Middle Yangtze River lowest; difference 0.095. The Theil index rose from 0.28 (2001) to 0.29 (2003) then declined, with within-region disparity (T_WR) consistently higher than between-region (T_BR), which increased slightly 2001–2004 then decreased. All eight regions showed declining Theil indices over time; North Coast had the lowest (most balanced) intra-regional inequality by 2018; Southwest, Northwest, and Middle Yangtze River had higher disparities. • Growth stages and saturation: City-level vehicle ownership rates follow an S-shaped logistic pattern. Classification: Type-A (initial stage; 56 cities; avg 0.10 in 2018), Type-B (take-off; majority; avg 0.15), Type-C (accelerating; 22 cities; avg 0.26), Type-D (slow-down; mostly eastern; avg 0.25). Several Type-D megacities with purchase restrictions (Beijing, Shenzhen, Guangzhou, Tianjin) show low saturation levels (~0.15–0.30). Some Type-D cities (Xiamen, Karamay, Zhuhai) also slow without restrictions. Chinese cities tend to enter slow-down at lower per-capita levels than developed countries, indicating a likely lower saturation level nationally. • Determinants: Vehicle ownership rate correlates positively with GDP per capita (R = 0.80) and urbanization rate (R = 0.63). Type-C/D cities generally have higher GDP per capita and urbanization (>70%), but megacities (Shenzhen, Shanghai, Guangzhou) exhibit lower ownership rates than suggested by GDP/urbanization due to plate restrictions.

The findings reveal pronounced spatial disparities in vehicle ownership rates aligned with China’s uneven urbanization and economic geography, yet inequality (Theil index) has declined nationally and within regions, reflecting effects of regional development strategies and infrastructure investments in central and western regions. City-level logistic growth patterns provide a nuanced understanding of motorization stages, improving forecasts relative to national/provincial aggregates. Evidence from Type-D cities and megacities suggests China’s saturation level will be lower than in developed countries due to high population densities, land constraints, congestion, air pollution concerns, strong public transport development, and policy measures like purchase restrictions. These dynamics have implications for carbon emissions, energy security, air quality, congestion, and end-of-life vehicle management. The results support policy designs that promote balanced regional development, compact urban form, and robust public and shared transport systems to accommodate mobility needs while curbing private car growth. City-level insights can better guide capacity planning for ELV recycling, critical material cycles, and transport decarbonization strategies, particularly as new energy vehicles diffuse.

Using data for 283 prefecture-level cities (2001–2018), the study maps the spatiotemporal evolution and inequality of vehicle ownership in China. Vehicle stocks concentrate in the North and East Coast regions; coastal and provincial capital cities have higher ownership rates. Inequality in vehicle ownership rates has declined at national and regional scales. City-level ownership follows an S-shaped logistic pattern and is strongly associated with GDP per capita and urbanization, while plate restriction policies can cap growth at relatively low levels in megacities, implying a national saturation level likely below that of developed countries. Given many cities remain in initial/take-off stages, continued growth is expected; thus, expanding intercity and intra-city public transport and shared mobility can meet demand with fewer vehicles. The city-level evidence and growth-stage typology enable more accurate forecasts of vehicle stocks, associated energy use, emissions, and material cycles, informing strategies for sustainable, low-carbon transportation. Future work should integrate detailed data on electric vehicles and critical materials, refine city-level projections using heterogeneous saturation levels, and link mobility policies to environmental and equity outcomes.

• Coverage: Some relatively underdeveloped prefecture-level cities are missing, likely leading to underestimation of in-use vehicle stocks and overestimation of per-capita ownership in undercovered regions (notably Northwest and Southwest). • Data quality: Statistical uncertainties (errors and inconsistencies) in city population and GDP series, especially from sample surveys, can introduce noise; comprehensive national censuses occur infrequently. • Measurement: Registered vehicles may diverge from actual in-use stocks due to delays in deregistering end-of-life vehicles and registering newly sold vehicles. • Scope: Lack of disaggregated data (e.g., separate treatment of electric vehicles and other determinants beyond GDP/urbanization) limits analysis of technology-specific dynamics and explanatory factors at the prefectural level. Improved, timely, and detailed datasets are needed.