Abstract:
In this paper, we evaluate the forecasting ability of 145 indicators and ten types of forecast combination schemes to predict housing prices and rents in 71 German cities. We are interested in whether local business confidence indicators facilitate substantial improvements of the forecasts, given the local nature of the real-estate markets. The forecast accuracy of different predictors is tested in a framework of a quasi out-of-sample forecasting. Its results are quite heterogeneous. No single indicator appears to dominate all the others for all cities and market segments. However, there are several predictors that are especially useful, namely price-to-rent ratios, the business confidence at the national level, and consumer surveys. We also find that combinations of individual forecasts are consistently selected among the top forecasting models/approaches. However, given a rather small sample size in our recursive forecasting exercise, the optimal combination weights is only possible to detect when using full-sample estimation information. On average, the forecast improvements attain about 20%, measured by a reduction in RMSFE, compared to the naive models. In separate cases, however, the magnitude of improvement is about 40%.