Determining the Implied Prices in Real Estate Properties: an Alternative to the Hedonic Modeling

The hedonic model approach provides a methodology to estimate transaction prices of complex and multiple attributes goods, and have been widely applied in the real estate sector. Artificial intelligence provides new tools, as neural networks, as an alternative to classical econometric modelling. With these NN models it is possible, as it is show in this paper, to generate non linear functions to estimate the hedonic prices in a more realistic way than with classical regression / hedonic models. Also the price estimates are generally more precise using NN and can be applied on a wider range of properties, overcoming the difficulties encountered with the classical approach with high value dwellings, which originate usually large errors. A case study with sample data using over a thousand transactions in a two year period in a middle size city, is presented to show that implicit prices can be estimated advantageously to the hedonic framework, and using a novel procedure based on NN to estimate the hedonic price functions. The results are favourably checked with real market data of several characteristics that can be externally evaluated. The methodology proposed can be easily adapted to different markets and it adapts to a wider range of type of properties in the real estate markets.