Who you are versus where you are: Revealing the importance of determinants of within-city income inequality in China through an interpretable machine learning approach

Zuge Xing, Canfei He, Jiale Lin, Yuxin Pan,
Who you are versus where you are: Revealing the importance of determinants of within-city income inequality in China through an interpretable machine learning approach,
Applied Geography,
Volume 184,
2025,
103759,
ISSN 0143-6228,
https://doi.org/10.1016/j.apgeog.2025.103759.
(https://www.sciencedirect.com/science/article/pii/S0143622825002541)
Abstract: Within-city income inequality may lead to slower regional growth and social instability. Most existing research attributes within-city income inequality to skill differences, with limited understanding of the combined impact and evolution of individual and city-level factors in the Chinese context. This study uses interpretable machine learning methods to measure within-city income inequality based on census and survey data from 2000 to 2015, and reveals the importance of individual and city-level factors on within-city income inequality using a SHapley Additive exPlanation (SHAP) analysis. We find that within-city income inequality in China is primarily driven by urban-rural gaps rather than skill differences, and individual factors such as gender and age also play important roles. Among city-level factors, housing prices are the main cause of the widening of within-city income inequality. Individual factors have the largest explanatory share in within-city income inequality, but the explanatory contribution of city-level factors is on the rise. The results of this study provide theoretical and methodological contributions to the measurement of the extent of within-city income inequality in China and its driving mechanisms.
Keywords: Within-city income inequality; Census data; Machine learning; SHAP