Weekly Hedonic House Price Indices: An Imputation Approach from a Spatio-Temporal Model

We develop a method for computing house price indices at a weekly frequency using the hedonic imputation method. The price indices are computed using the Törnqvist formula applied to a matched sample derived from the hedonic... [ view full abstract ]

We develop a method for computing house price indices at a weekly frequency using the hedonic imputation method. The price indices are computed using the Törnqvist formula applied to a matched sample derived from the hedonic model. Typically, with the hedonic imputation method, the model is estimated period-by-period and location is controlled for using postcode dummies. This may not be feasible at a weekly frequency. Instead, we develop a spatio-temporal model represented in a state-space formulation that controls for location with a geospatial spline surface. The model is parsimonious and shadow price parameters are connected over time while retaining the property of non-revisability of the index as new time periods are added to the data set. We show the spatio-temporal specification leads to a modified form of the Kalman filter and a Goldberger's adjusted form of the predictor to obtain the imputations. Using data for Sydney (Australia) we show that weekly house price indices are much more sensitive than quarterly indices to the choice of hedonic method. We then show using a novel criterion that our approach to computing house price indices at a weekly frequency outperforms competing alternatives.