Forecasting housing investment

Abstract

AbstractThis study applies a model averaging approach to conditionally forecast housing investment in the largest Euro area countries and the Euro area. To account for substantial modeling uncertainty, it estimates a large and diverse number of vector error correction models using a wide set of long‐ and short‐run determinants and applies subset selection based on in‐sample and out‐of‐sample criteria. First, a pseudo out‐of‐sample forecast exercise shows that our model averaging approach consistently beats a battery of distinguished benchmark models, including BVARs, FAVARs, LASSO, and Ridge regressions. This evidences that model averaging provides more accurate forecasts also in the case of housing investment. Second, we find remarkable cross‐country heterogeneity in the drivers of housing investment. Overall, these findings guide forecasters and modelers on improving housing investment models and policymakers on implementing country‐specific housing market policies.