Abstract
The need of retrospective predictions arises in situations such as litigations, inheritances or fiscal purposes to provide the fair market value of a dwelling of given characteristics and specific location. In this paper, a Bayesian method to carry out this retrospective analysis is proposed by using a Spatio-Temporal Autoregressive (STAR) model with neighborhood effects and heteroscedastic errors. Moreover, the developed methodology allows the construction of a retrospective price index that shows the evolution of the dwelling price level in a specific geographical area. The methodology is illustrated with an application to the real estate market in the Spanish city of Zaragoza.