A Test for Long Memory in the Conditional Correlation of Bivariate Returns to Stock and Bond Market Index Futures
22 Pages Posted: 4 Aug 2011
Date Written: August 3, 2011
Abstract
Long memory is found in the conditional volatilities of financial returns measured at daily or higher frequencies, as well as in residual cross-products in bivariate series. We test for long memory in conditional correlations by extending the fractionally integrated GARCH model to include previous conditional correlations and standardized cross-products in a bivariate system. We apply the model to stock-stock and stock-bond futures index returns using Engle’s (2002) two-stage dynamic conditional correlation approach. We find that cross-products are indeed long memory processes, but that this feature arises from long memory in conditional volatilities and not from long memory in conditional correlation.
Keywords: Long memory, FIGARCH, DCC, conditional volatility, conditional correlation
JEL Classification: G15, G11, C15, C34
Suggested Citation: Suggested Citation