Googling For a New Home: Forecasting the Dynamics of the Dutch Housing Market using Google Search Data

This paper studies whether Google Trends data contains additional information not captured by common macroeconomic variables and, hence, is able to improve the forecasting performance of prediction models for the house price index and the ow of residential mortgages. Real estate related search terms are selected based on a novel approach using the output of Google Correlate in combination with pages of the web directory Startpagina.nl. Evaluating post-lasso OLS estimators for dierent regularisation parameters, the empirical ndings suggest that there is not enough evidence to support the hypothesis that the inclusion of Google search queries signicantly increases the forecasting accuracy across short and long forecast horizons. However, the augmented models prove to perform better than the models comprising the search index of a predened Google Trends category when predicting the residential mortgage ows. Observing specic search terms give insights into the the search behaviour of Dutch citizens in response to policy changes aecting the housing market.