Spatial Modelling and Textual Analysis in Real Estate

The provision of residential real estate price indexes requires the combination of index number theory and suitable data. In the case of hedonic indexes, it involves also the use of regression models that allow to impute values of seldom traded houses. To disentangle price changes and quality differences such models need to take physical and location characteristics of a house into account. Unfortunately, many quality attributes are not directly observable. It has been suggested that nonparametric techniques for spatial modelling and/or textual analysis of the remarks section of listings can mitigate this problem.

This PhD would investigate how econometric modelling choices regarding not directly observable quality attributes affect, both, the construction of real estate price indexes and the accuracy of individual real estate valuations. The empirical analysis would build on a large data set of real estate listings and transactions which comprise list prices, transaction prices, spatial information, and information about the marketing process (time on market, texts of remarks section).

The successful applicant is expected to have (or be close to graduating with) an MSc in Economics/Econometrics or in a related area e.g. an MSc in Statistics with significant prior training in economics. Applicants should have an interest in spatial data modelling, machine learning, and urban housing markets. The project requires excellent skills in programming (R and/or Matlab) and data organisation. General knowledge about index number theory, urban economics, and regional economics is desirable.

Applicants interested in this research project should submit a more detailed research proposal that expands on the broad outline given abroad.

Informal inquiries can be made to Dr M Wersing ([Email Address Removed]) with a copy of your curriculum vitae and cover letter indicating your interest in the project and why you wish to undertake it.