Asset market correlation and stress testing : cases for housing and stock markets

This study aims to achieve a two-fold research objective: first, to econometrically investigate hypothesized linkages between real estate and stock markets by fitting different classes of time-varying volatility model; second, to perform VaR-type stress testing by using the fitted asset price models. In so doing, we use data from Korea and U.S. so that asset price processes, in terms of mean, volatility, correlation, can be compared. In the econometric analyses, we estimate both a multivariate GARCH model that allows contemporaneous and time-varying shock correlations between real estate and stock markets and an univariate GARCH model that does not allow such correlation. Our results indicate that housing price volatilities in both countries are highly time-varying, with the Korean asset markets shown to be more volatile, and that non-consideration of asset market correlation under-predicts the risk embedded in real estate price dynamics. Policy implications of our findings in regard to stress testing and other issues are also discussed