JAILCD2026-006: Urban Visual Environment and Residential Prices: A Factor Analysis

Abstract

This paper examines how street-level visual environments relate to residential prices in the core districts of Shanghai. Using 7,634 second-hand housing transactions from an online platform, we construct a 500 m walking catchment around each dwelling and link it with street-view images, open geospatial data and housing attributes. Baidu Maps panoramas are processed using a pre-trained DeepLab v3+ model to obtain semantic segmentation, from which we derive visual view indices for elements such as buildings, trees, sky, roads, sidewalks and walls. Together with locational and housing variables, these indicators form a 16-variable system describing the everyday activity space of each dwelling. Factor analysis is then applied to reduce dimensionality and identify core latent dimensions. Four interpretable factors are extracted. Factor 1 is defined by the building view index (Avg_buildi) and distance to the city center, capturing a spatial centrality–façade intensity gradient from highly central, visually dense building frontages to more peripheral areas with weaker façade dominance. Factor 2 reflects green and walkable streets, combining higher tree and sidewalk visibility with lower sky and roadway dominance. Factor 3 represents building characteristics, jointly capturing unit size, building age and elevator provision. Factor 4 measures façade enclosure based on the wall view index. The results suggest that, in dense megacity cores, housing prices are shaped not only by spatial centrality and building characteristics, but also by fine-grained visual and environmental qualities at the street level. Visual environment indicators derived from street-view images provide a useful complement to traditional location and structural variables in urban housing price analysis.