Real estate markets exhibit spatial heterogeneity with respect to their dependence on the macroeconomy. Capturing that heterogeneity is important because it drives the mean and the variance of returns under different macroeconomic scenarios. Modelling the evolution of local markets is difficult since they are only partially observed on completion of a transaction.
We begin by modelling the dependence of returns in local markets on the state of the economy, using hierarchical models to generate predictions where little data is available. We then use tools from time series econometrics to predict the future state of the economy. Combining those two approaches yields an engine that is capable of forecasting the returns of spatially distributed real estate portfolios.
SAREB used the resulting forecasts to inform decision making on divesting its real estate portfolio.