Volatility modeling in a Markovian environment: Two Ornstein-Uhlenbeck-related approaches

TL;DR

This paper extends COGARCH and Barndorff-Nielsen and Shephard models to a Markov-switching environment, incorporating exogenous jumps at regime changes, and analyzes their properties for financial time-series modeling.

Contribution

It introduces Markov-modulated generalizations of key volatility models, enabling regime-switching features and deriving their stationarity, moments, and autocovariance structures.

Findings

Models inherit properties of original models

Capture stylized facts like uncorrelated returns

Allow for exogenous jumps at regime switches

Abstract

We introduce generalizations of the COGARCH model of Kl\"uppelberg et al. from 2004 and the volatility and price model of Barndorff-Nielsen and Shephard from 2001 to a Markov-switching environment. These generalizations allow for exogeneous jumps of the volatility at times of a regime switch. Both models are studied within the framework of Markov-modulated generalized Ornstein-Uhlenbeck processes which allows to derive conditions for stationarity, formulas for moments, as well as the autocovariance structure of volatility and price process. It turns out that both models inherit various properties of the original models and therefore are able to capture basic stylized facts of financial time-series such as uncorrelated log-returns, correlated squared log-returns and non-existence of higher moments in the COGARCH case.