A Dynamic Vector ARCH Model for Exchange

This paper examines changes in foreign exchange rates with a focus on estimating the variance/covariance structure of the exchange rate time series. The use of a simulation based numerical integration algorithm frees the model speci-cation from restrictions imposed by technical considerations required to obtain analytical solutions. We estimate a vector ARCH (autoregressive conditional heteroskedasticity) model with dynamic parameters and a multivariate Student t sampling distribution. The results connrm the need for joint inference of the multivariate time series and for allowing time dependent variance parameters.