Recent evidence suggests that financial markets experience shifts in volatility (i.e. structural breaks) and these volatility shifts should be accounted for in the models of volatility estimation. This study re-examines volatility dynamics of the US Dollar exchange rate and the US stock market utilizing bivariate GARCH models using daily data from January 2003 to May 2018. The modified iterative cumulative sum of square (ICSS) algorithm is employed to identify shifts in the variance of the two return series. The results show that if volatility shifts are ignored, there is significant volatility transmission from the US stock market to the US Dollar exchange rate but not vice versa, which is consistent with previously documented research. However, after accounting for endogenously determined variance shifts in the bivariate GARCH model, I find no significant volatility transmission across markets. I also show that dynamic risk-minimizing hedge ratios and portfolio weights change substantially when volatility shifts are incorporated into the bivariate GARCH model. My findings point to the presence of cross-market hedging by financial market participants in these markets and indicate an inherent estimation bias in the bivariate GARCH models as they tend to overestimate the volatility spillovers across markets when volatility shifts are present but ignored. I show that my results are robust to using data from other major countries over a longer time series data.
Read full abstract