Urban spatial structure and commute duration: An empirical study of China

ABSTRACT

Urban traffic is embedded in and fundamentally shaped by the spatial pattern of urban land use, such as city size, density, extent of polycentricity, and the relationship between employment and residential locations. Previous evidence, mainly from European and American cities, suggests that the duration of commute trips increases with city size and the spatial separation between jobs and housing. On the other hand, the influences of density and polycentricity are less clear. Using data from 164 cities in China, this study empirically analyzes the relationship between city average commute duration and multiple dimensions of urban spatial structure. Controlling for economic, demographic, and infrastructure characteristics, we find that commute duration correlates positively with city size and jobs–housing separation but negatively with density and polycentricity. As one of the earliest studies on commute cost in the rapidly urbanizing and motorizing Chinese cities, this study can help Chinese decision makers improve urban economic and environmental efficiency through spatial planning and policy making. Specifically, compact, mixed-use, and polycentric spatial development may ease the burden of commute, and thus substitute for unnecessary infrastructure investment and energy consumption during a period of rapid urban expansion in China.

KEYWORDS:

• China
• city size
• commute duration
• jobs–housing balance
• polycentricity
• urban spatial structure

Funding

We thank the National Natural Science Foundation of China (No.41471139), the Shanghai Social Science Foundation (No. 2014BCK003), the Ministry of Education of China Key Research Projects in Social and Humanities Program (No. 11JJDZH004) and the UCLA Ziman Center for Real Estate through its faculty research grant. Opinions, findings, and errors in this article are those of the authors.

Notes

¹ The two dominant modes of public transit are bus and rail in Chinese cities. We reported results including only the bus mode because only 12 cities in our sample had rail transit in 2010 and also because we consider buses per ten thousand people as a general indicator of transit infrastructure/service availability (even cities with rail transit still rely heavily on bus service). In fact, we did try by adding a rail transit dummy variable, but the results across model specifications suggested insignificant effect of rail transit and fairly consistent estimated coefficients of other variables (results available upon request). Thus we chose to report the results without controlling for rail transit availability.