Estimation of Generalized Long-Memory Stochastic Volatility: Whittle and Wavelets

Description

We compare the Whittle and a wavelet based Whittle estimator, WWE, for $k$-GARMA and generalized stochastic long-memory volatility models, GLMSV. We show that the decorrelation properties of wavelets of different levels for FI also hold for k-GARMA and GLMSV models. This property is used to derive a wavelet Whittle estimator, which also is shown to be consistent. The small sample properties of Whitcher's(2004) [DWPT, GML], WWE and Whittle's estimator are compared. The WWE clearly dominates Whichter's estimator, and is essentially indistinguishable to Whittle's. Finally, the WWE is illustrated by fitting a GLMSV to Microsoft realized volatilities.

Period	6 Dec 2014 → 8 Dec 2014
Event title	CFE 2014, 8th International Conference on Computational and Financial Econometrics
Event type	Unknown
Degree of Recognition	International