Disposition effect in the housing market : empirical evidence from Hong Kong

Abstract

Disposition effect is one of the most documented trading anomalies studied in financial market. Its presence has been established over time horizons, time periods and market participants. This study will examine such trading behavior in the housing market. Using Mei Foo Sun Chuen estate, one of the largest and most frequent transacted estate in Hong Kong, we show that disposition effect is present in this market.

A major difficulty in the statistical analysis is the presence of censored data problem, which is hard to circumvent in linear regression models. So we adopt a survival analysis approach, which can accommodate the issue and fit the data structure. The other difficulty is the possibility of omitted variables in the analysis. Instead of appealing to instrumental variables model approach, which is widely applied in many research studies on individual behavior but is extremely hard to be justified in a whole market case, we make use of partial identification approach to estimate bounds for the estimates rather than just a point estimate. Even though this seems offer us less precise information, it is still informative, especially when the bounds are narrow, and it is much less vulnerable to the validity of instruments. Besides the above techniques, we use bootstrapping method to estimate the standard errors throughout the analysis in order to make valid inference.

There are three main results we have established in this study. First, the disposition effect is present in Hong Kong housing market. It shows up in both in pre-1997 and post-1997 periods, which suggest that it is a general phenomenon rather than a short-term trading pattern arising from a major macroeconomic event. Secondly, we show that it is the nominal perspective loss that matters, but not real loss. This confirms the validity of basic setup of both Prospect theory and most previous empirical studies on the disposition effect. Thirdly, the disposition effect is more significant for short term owner of less than 3 years. In fact, the disposition effect is absent or even reverses for those flat owners of more than 6 years.

Comparing to the very first study on disposition effect by Shefrin & Statman (1985), our study makes a step forward in understanding trading behavior. We extend it applicability to housing market while their focus is only on financial market even though we are not the first to attempt the extension. We not only show the presence of disposition effect in the housing market, we also show that disposition effect is time dependent. Our results lead further support to disposition effect that its key component, loss, matters only in nominal term, not in real term. Instead of only applying ordinary least squared regression methods, which is widely used in the literature, we apply partial identification approach to tackle the endogeneity issue and apply duration models to deal with censored data and unobserved heterogeneity issues. All this makes our statistical results more robust.