Residential real estate price indices in Hong Kong: A pseudo-repeat sales approach

Abstract

Real estate price is one of the core indicators of economic activity of a country or region. Real estate price index, which is designed to reflect the dynamic change in price levels in a property market over time is therefore of great importance. Generally, this index is either appraisal-based or transaction-based. Com-pared with the former which is subject to smoothing problems, transaction-based index has a host of strengths. This type of index is usually constructed by four statistical approaches: mean/median, hedonic, repeat sales and hybrid model (hedonic + repeat sales). Repeat sales approach is the ‘workhorse’ tool to develop real estate price indices. For comparing the change in the price of the “same” property over time, the traditional repeat sale method strictly focuses on the property transaction at least twice within a sample period. However, this method is subject to a number of shortcomings, including but not limited to sample selection bias (sample representativeness), re-duced sample size, constant (unadjusted) quality as well as changing coefficients. Matching is an important nonparametric pre-processing procedure for achieving balance and reducing bi-as as well as model dependence. In 2012, a novel approach based on matching (called matched pair (or pseudo-repeat sales) method), has been proposed to develop price indices by McMillen. This approach has been applied in some real estate markets such as U.S., Mainland China and Singapore. It is perceived to be capable of avoiding many shortcomings of traditional repeat sale method. By relaxing the highly stringent requirement to traditional repeat sales and pairing together individual property sale observations over time according to some criteria, this approach cancels out as much as possible the unobservable at-tributes, and does not restrict the “same” property, but very “similar” property when matching or creating pairs (e.g., within a complex, phase, building). Due to the novelty of match-pair method (coming into ex-istence in less than 5 years), the characteristics of this type of index are unknown to a large extent. In Hong Kong, the most celebrated price index, HKU-REIS, is provided by the Department of Real Estate and Construction, the University of Hong Kong. It adopts a modified repeat sales with a 10-year rolling time window to calculate a new data point. Unlike HKU-REIS, this paper uses 15-year actual residential housing transition data (N>350,000) to produce matched pair index in Hong Kong and compares it with other indices produced by other methods (e.g., HKU-REIS, mean prices per square meter provide by the Rating and Valuation Department (RVD), CCI indicator). This paper finds that this approach is capable of reflecting the dynamic price changes in Hong Kong and may have the potential to be used for construct-ing indices for sub-markets (by size, location or other criteria) and for the non-residential (e.g., commer-cial) market whose price index cannot be developed based on traditional repeat sales method due to very small sample size (caused by thin and lumpy transaction).