Forecasting Growth-at-Risk of the United States: Housing Price versus Housing Sentiment or Attention

Abstract We examine the predictive power of national housing market-related behavioral variables, along with their connectedness at the state level, in forecasting US aggregate economic activity (such as the Chicago Fed National Activity Index (CFNAI) and real Gross Domestic Product (GDP) growth), as opposed to solely relying on state-level housing price return connectedness. Our results reveal that while standard linear regression models show statistically insignificant differences in forecast accuracy between the connectedness of housing price returns and behavioral variables, quantile regression models, which capture growth-at-risk, demonstrate significant forecasting improvements. Specifically, state-level connectedness of housing sentiment enhances forecast accuracy of the CFNAI at lower quantiles of economic activity, indicative of downturns, whereas connectedness of housing attention is more effective at upper quantiles, corresponding to upturns. The results for GDP growth suggest that, while both sentiment and attention contribute to forecasting performance at lower quantiles, only attention improves forecasting performance at upper quantiles. In terms of statistical significance, the results for GDP growth, however, are less conclusive than those for the CFNAI. Taken together, these findings underscore the importance of incorporating regional heterogeneity and behavioral aspects in economic forecasting.