Automatic Mass Valuation for Non-Homogeneous Housing Markets

In recent years big financial institutions are interested in creating and maintaining property valuation models. The main objective is to use reliable historical data in order to be able to forecast the price of a new property in a comprehensive manner and provide some indication for the uncertainty around this forecast. The need for unbiased, objective, systematic assessment of real property has always been important. This need is urgent now as banks need assurance that they have appraised a property on a fair value before issuing a loan and also as the government needs to know the fair market value of a property in order to determine accordingly the annual property tax. In this study we compare various linear, nonlinear and machine learning approaches. We apply a large set of variables, supported by the literature, describing the characteristics of the real estate properties as well as transformation of these variables. The final set consists of 60 variables. We answer the question of variables selection by extracting all available information with the use of several shrinkage methods, machine learning techniques, dimensionality reduction techniques and combination forecasts. The forecasting ability of each method is evaluated out-of-sample is a set of over 30,000 real estate properties from the Greek housing market which is both inefficient and non-homogeneous. Special care is given on measuring the success of the forecasts but also on identifying the property characteristics that lead to large forecasting errors. Finally, by examining the strengths and the performance of each method we apply a combined forecasting rule to improve forecasting accuracy.