A New Approach for Constructing Home Price Indices in China: The Pseudo Repeat Sales Model

Abstract

Due to data and methodology constraints, there is a lack of good quality-controlled residential price indices publicly available in China. New home sales account for quite a large share of total home sales in Chinese cities (87% in 2010). As a result, the standard repeat sales approach cannot be employed, as a new housing units only appears once on the market. The hedonic method may be more suitable in principle, but it is vulnerable to an omitted variables problem which may be more significant in Chinese cities due to extremely dynamic urban spatial structure development and rapid building quality improvement. Taking advantage of the uniquely large scale and homogeneous nature of residential development in Chinese cities, we develop a ―pseudo repeat sale‖ model (ps-RS) to construct more reliable quality-controlled price indices for newly-constructed homes. The new homes are developed in the form of residential complexes. Each complex is developed by a single developer, and often contains several phases and a number of high-rise residential buildings. Each housing unit within the same complex shares the same location and community attributes, as well as similar physical characteristics (such as structure type, architecture style, housing age, etc). Of course, there may still be important differences in unit size, number of bedrooms, floor level within the high-rise, and so forth. Based on specific criteria, we match two very similar new sales within three versions of a defined matching space: within a complex, within a phase of a complex, or within an individual building, respectively. We thus create a ―pseudo-pair‖. We are able to generate a vast number of such pairs, many more than in traditional repeat sales models. By regressing the price differential across time between the two sales in each pseudo-pair onto the within-pair differentials in