Determinants of house price: A decision tree approach

Abstract

The hedonic-based regression approach has been utilised extensively to investigate the relationship between house prices and housing characteristics. However, this approach is subject to criticisms arising from potential problems relating to fundamental model assumptions and estimation such as the identification of supply and demand, market disequilibrium, the selection of independent variables, the choice of functional form of hedonic equation and market segmentation. This study introduces and utilises an alternative approach - the decision tree approach, which is an important statistical pattern recognition tool. Using the Singapore resale public housing market as a case study, the article demonstrates the usefulness of this technique in examining the relationship between house prices and housing characteristics, identifying the significant determinants of housing prices and predicting housing prices. The built tree shows that homebuyers are more concerned about the basic housing characteristics of two- and three-room flats or four-room flats such as floor area, model type and flat age. However, homebuyers of five-room flats pay more attention to floor level in addition to the basic housing characteristics. In addition, homebuyers of executive apartments are less concerned about basic quantitative characteristics and have higher housing consumption expectations and pay more attention to quality and service characteristics such as recreational facilities and the living environment.