Long memory of volatility measures in time series

The authors analyse relations between the long memory parameter of conditional variance and estimates of the long memory in squared residuals in FIGARCH models. The investigations are performed by means of simulations FIGARCH(0, d, 0) and FIGARCH(1, d, 1) models for selected parameters. Simulation results suggest, that estimates of the conditional variance long memory and the long memory in squared residuals can considerable differ. Moreover, only for small d positive relationship between the long memory estimates of squared residuals and the fractional integration parameter d of FIGARCH model can be observed.