Abstract
In this paper, we test for structural changes in the conditional dependence of two-dimensional foreign exchange data. We show that by modeling the conditional dependence structure using copulae, we can detect changes in the dependence beyond linear correlation, such as changes in the tail of the joint distribution. This methodology is relevant for estimating risk-management measures, such as portfolio value-at-risk, pricing multi-name financial instruments, and portfolio asset allocation. Our results include evidence of the existence of changes in the correlation as well as in the fatness of the tail of the dependence between Deutsche mark and Japanese yen.
Acknowledgements
We acknowledge useful discussions with Alexander McNeil, and anonymous referees and the journal editor for several detailed comments on an earlier version of this paper.
Notes
In the remaining of the paper, we refer to the daily logarithmic returns simply as the returns.
The values for the kurtosis given are to be compared with a value of 3 for the standard normal distribution.