Linking rural road quality and housing investment in China: Insights from AI-based assessment of village view imagery

Public infrastructure, exemplified by rural roads, plays a critical role in promoting rural economic development and shaping household behavior. Recognizing that roads of different functional hierarchies have distinct effects, this study distinguishes inter-village roads and intra-village roads, and examines the differential relationship between them and households’ housing investment behavior. Using data from 97 counties in China, we introduce an AI-based assessment framework that leverages a multimodal large language model to automatically assess village view imagery, and construct a county-level dataset including road quality, rural housing modernization, and other key variables. Multiple regression and mediation analysis show that: (1) Improvements in inter-village roads significantly enhance rural housing quality, but it is not significantly correlated with households’ intention to build new houses. (2) Intra-village roads, as last-mile infrastructure, generate a strong demonstration effect, with a marginal effect on housing quality about 2.2 times that of inter-village roads. (3) Inter-village roads improve housing quality indirectly by fostering intra-village road construction, with the indirect effect via intra-village roads accounting for approximately 58% of the total effect. From a policy perspective, we argue that China and other developing countries should move beyond the single-target approach of roads to every village, adopt differentiated road investment strategies, and leverage the demonstration effect of intra-village roads to mobilize private investment. More broadly, this study demonstrates an AI-based approach to producing rural built environment data, and provides empirical evidence on the micro-level effects of rural public infrastructure.