Thursday 10 May 2012, 1.15PM to 2:15pm
Speaker(s): Yongcheol Shin (University of York)
We extend Evans and Lyons’ (2002a) multi-currency portfolio shifts model to develop a global currency market model using the Global VAR methodology which allows us to combine market-specific models together coherently in a system while avoiding the dimensional problem. Our model accounts for simultaneous and dynamic interactions between exchange rates and order flows from all currency markets, which is often missing in existing studies. Using Reuters D2000-1 daily dataset for eight currency spot markets, we find several important findings. Firstly, the analyses of forecast error variance decompositions and impulse response functions suggest that currency markets are indeed intricately linked with one another but appear to play different informational roles, leaders and followers. Importantly, dollar-buying (selling) pressure in a major market is often absorbed, though partly, by dollar-selling (buying) pressure in another major market. This reflects how investors adjust their portfolios between currency markets, thus affecting exchange rates. Next, our results show that it takes up to four days for trading information to be fully integrated into exchange rates. Finally, based on the forecast error and directional forecast criteria, our GVAR model provides a better forecasting performance than the selected benchmarks. This is attributable to the informational advantage of the our model over the benchmarks through the inclusion of dynamic interactions among currency markets. In general, our model offers a useful approach to understand the informational linkages among currency markets.
Location: Economics Staff Room (EC/202)
Admission: For Staff and Postgraduate students