The global financial crisis and the collapse of the collateralized debt obligation (CDO) market have brought increased attention to the proper modeling of housing price co-movements worldwide. We aim at detecting possible contagion effects in international real estate markets while accommodating dependence during extreme tail events. We propose a novel copula based approach incorporating second-moment effects that not only accounts for asymmetric tail dependence, but also allows for time-varying correlation in price movements. Unlike previous studies wherein static copula-based models are utilized, we extend our methodology by employing nonparametric copulas with the adjustment of flexible specification. Common Gaussian or mixture copulas lack the required tail features to capture the empirical stylized facts in housing markets. We proved the lack of monotonicity imposed by parametric methods was evidently not supported by our data. Using monthly data in seven major global markets, we confirm that prices do exhibit correlations that change over time, whilst more importantly their tail dependence structure for extreme losses strengthens in the midst of market turmoil.

We indicated that especially during downturns, CDOs do not provide the level of diversification widely assumed before the subprime crisis. Information on tail dependence would better allow policy makers to anticipate real estate prices on a global scale.