Out-of-sample forecasting of housing bubble tipping points

Abstract

This paper analyzes the information content of statistical tests for bubble detection in the context of international real estate markets. We derive binary indicators from the causal application of five statistical tests to log house prices, and via logit regressions we assess the indicators’ out-of-sample performance in the forecasting of tipping points of housing bubbles and systemic financial crises. In our assessment, three of the indicators - two based on the identification of super-exponential trends and one based on the scaled ratio of the sum of squared forecast errors - exhibit significant out-of-sample results. Combining the indicators via simple threshold-rules yields the most robust and best results.