A historical and econometric investigation of housing bubbles in OECD countries: insights from the GSADF test and machine learning

This paper investigates the existence, causes, and predictive markers of housing price bubbles in 18 OECD countries from 1870 to 2020, thus addressing a significant gap in the understanding of housing market dynamics and their implications for global financial stability. Housing bubbles have substantial impact on economic resilience and have his-torically led to severe financial crises. Employing the Generalized Supremum Augmented Dickey-Fuller (GSADF) test, this study identifies multiple bubble episodes in Australia, Denmark, Germany, Japan, Portugal, and the United States. Furthermore, by using an advanced machine learning approach, Extreme Gradient Boosting (XGBoost), this study statistically confirms the significance of interest rates, loan growth, and population growth as key predictors of housing bubbles. The findings indicate that interest rate variables are the predominant predictors, explaining over 60% of bubble dynamics in Australia and Japan, whereas credit growth and demographic factors are more influential in predicting bubbles in Germany, Denmark, and the United States. This study's originality lies in its comprehensive integration of econometric and machine learning methodologies, offer-ing more accurate, data-driven detection and prediction of housing bubbles than previ-ous research. The study's findings underscore the necessity of coordinated monetary and macroprudential policies, along with proactive demographic and credit market manage-ment, to mitigate future bubble-related risks, presenting significant implications for global policymakers and market participants.