Asymmetric stochastic volatility models: properties and particle filter-based simulated maximum likelihood estimation

Abstract

The statistical properties of a general family of asymmetric stochastic volatility (A-SV) models which capture the leverage effect in financial returns are derived providing analytical expressions of moments and autocorrelations of power-transformed absolute returns. The parameters of the A-SV model are estimated by a particle filter-based simulated maximum likelihood estimator and Monte Carlo simulations are carried out to validate it. It is shown empirically that standard SV models may significantly underestimate the value-at-risk of weekly S&P 500 returns at dates following negative returns and overestimate it after positive returns. By contrast, the general specification proposed provide reliable forecasts at all dates. Furthermore, based on daily S&P 500 returns, it is shown that the most adequate specification of the asymmetry can change over time.