Trading Volume and Price Dispersion in Housing Markets

Abstract

The positive volume‐price (return) relationship has been intensively studied and confirmed in both financial and real estate markets, yet their theoretic models offered few direct empirical support. This paper puts forward a liquidity premium model which explains the volume‐price (return) relationship by the volume‐price dispersion relationship. We posit that the extent of price dispersion depends on the level of price information available in the market (measured by the volume of past comparable transactions). The model is tested empirically in two sample periods on the transactions of housing units of an estate in Hong Kong from February 1992 to September 2000 (11,267 transactions), and from May 1991 to May 2008 (18,368 transactions), respectively. The results support our theoretical prediction that the magnitude of price dispersion, as measured by the residuals of a hedonic pricing model, is negatively and significantly related with the volume of transactions in the past 10‐day and 30‐day period windows. It implies that an increase in liquidity reduces pricing error risk, which in turn reduces the required risk premium in buyers' offering price, and thus a positive volume‐price (return) relationship.

Keywords:

• Price dispersion
• trading volume
• volume‐price relationship
• liquidity risk

Acknowledgement

The work described in this paper was fully supported by the Central Research Grant of The University of Hong Kong, Hong Kong Special Administrative Region, China (Project No.: HKU1230/06E) and the internal Research Grant of the Hong Kong Polytechnic University (Project No.: G‐U014). We also thank for all the comments received from the participants of the Hong Kong Real Estate Research Workshop.

Notes

1. For illustration, consider an Ordinary Least Squares (OLS) regression model; Y = X β +D α + ε , in which Y is a N×1 dependent variable matrix of housing price; X is a N×k matrix of independent variables of housing characteristics; D is a N×t matrix of the t time dummies; β is a k×1 matrix of coefficients for housing characteristics and α is a 1×t matrix of coefficients for time dummies. ε is a N×1 matrix of random residuals, representing unmeasured qualities and price dispersion. According to the standard OLS assumptions, the residual terms should satisfy the homoskedastic assumption; i.e. E( εε ′) = σ ² I, else the estimation is inefficient. Heteroskedasticity is referring to the violation of this assumption by E( εε ′) = σ ² Ω ≠σ² I. One of the common correction approaches for known reason of heteroskedasticity is the generalized least square (GLS) estimation with different types of weighting methods. Detailed discussions on the problem and correction methodologies can be found in standard textbooks of econometrics, for example, Greene (2000). In practice, if the price dispersions over time are varying; the problem of heteroskedasticity will be constituted. The OLS estimators are not biased by the violation of the homoskedastic assumption, but the estimators will become inefficient. The estimated variances and co‐variances of the regression coefficients will be biased and inconsistent, and hence tests of hypotheses are invalid. Therefore the detection and correction of heteroskedasticity are crucial for the reliability of the hedonic pricing model.

2. High‐rise housing units are commonly found in Hong Kong, and the data sample are collected from a high‐rise housing estate, where some units are possessed of panoramic (unobstructed) view of the sea from their living rooms. These units are regarded as FULL SEAVIEW in the data set. However, the sea views of some units are obstructed by other blocks of housing in the same estate, or other blocks of building outside the estate. Then they are considered as PARTIAL SEA VIEW flats.

3. By the Conveyancing and Properties Ordinance, Laws of Hong Kong, housing transactions shall be registered within 30‐day after contract. Thus, 30‐day is a time‐line for market information to be officially released to the market participants (at no cost). However, individual real estate agencies also collect transaction information which is normally updated in every 2‐day or 7‐day period. These windows of period can show how fast the volume information affects the price dispersion.