NEW EMPIRICAL EVIDENCE ON HETEROSCEDASTICITY IN HEDONIC HOUSING MODELS : SINGAPORE CONTEXT

Abstract

This dissertation deals with heteroscedasticity in hedonic housing model in the context of Singapore's private residential properties. It examines whether it is appropriate to use aggregate or disaggregate models in getting a best fitted linear regression line to represent private residential property prices. It is understood that property pricing is often a function of various physical attributes. Thus, it makes use of Urban Redevelopment Authority (URA) private properties transactions to study the property prices in the hedonic housing model. It is found out later that heteroscedasticity in hedonic housing model could be greatly reduced at disaggregate level. By disaggregating the properties into their property types and then in their region/location, it is holding certain housing attributes constant (for instance, property type and region/location). Thus, it ensures that the variances of the error term in the multiple linear regression is minimised. However for some regions, even when the properties are disaggregated by property type and by region, price heteroscedasticity is still present. Therefore this dissertation had discovered that unequal price variances cannot be eliminated but minimised.