A novel approach to examining urban housing market segmentation: Comparing the dynamics between sales submarkets and rental submarkets

Abstract

Submarket segmentation outlines an essential prerequisite for monitoring housing market and formulating urban housing policies. Although examining segmentation based on a posteriori knowledge rather than a priori knowledge becomes the mainstream, it follows a data-driven approach without a solid theoretical foundation and involves subjective interventions. Additionally, earlier studies have overwhelmingly examined the dynamics of sales submarkets while overlooking those of rental submarkets. This paper demonstrates a novel approach to segmenting the housing market by integrating the hedonic model, geographically and temporally weighted regression (GTWR), and machine learning, and further applies it to unpack the dynamics of sales submarkets and rental submarkets from 2018 to 2020 in Shanghai, China. More specifically, using the home-fixed panel data of housing sales prices and rental prices for each residential quarter, we first establish a series of hedonic models using GTWR and then aggregate the residential quarters into a number of submarkets using an affinity propagation clustering algorithm based on the produced spatiotemporally explicit coefficients. To validate the identified submarkets, we compare them to the static submarkets delineated by the real estate industry with respect to the performances of hedonic models. Finally, the Jaccard and Rand indices are applied to compare the magnitude of spatiotemporal dynamics of sales submarkets and rental submarkets. Results show that hedonic models based on the identified submarkets through our proposed method perform better than those based on the static submarkets delineated by the real estate industry. We also discover that the submarkets present a complex change over three years, especially in central urban areas. The dynamics between sales submarkets and rental submarkets are of significant differences. In particular, rental submarkets are more stable than sales submarkets. Our approach provides a practical and efficient tool for urban housing market segmentation. Our study highlights that differentiated policies should be formulated for regulating sales submarkets and rental submarkets in order to enhance housing affordability.