The long memory HEAVY process: modeling and forecasting financial volatility

Abstract

This paper studies the bivariate HEAVY system of volatility regression equations and its various extensions that are directly applicable to the day-today business treasury operations of trading in foreign exchange and commodities, investing in bond and stock markets, hedging out market risk, and capital budgeting. We enrich the HEAVY framework with powers, asymmetries, and long memory that improve its forecasting accuracy significantly. Other findings are as follows. First, hyperbolic memory fits the realized measure better, whereas fractional integration is more suitable for the powered returns. Second, the structural breaks applied to the bivariate system capture the time-varying behavior of the parameters, in particular during and after the global financial crisis of 2007/2008. Keywords Asymmetries • Financial crisis • Forecasting • HEAVY model • High-frequency data • Long memory • Power transformations • Realized variance • Risk management • Structural breaks JEL Classification C22 • C52 • C58 • G01 • G15 B M. Karanasos

Macro‐financial linkages in the high‐frequency domain: Economic fundamentals and the Covid‐induced uncertainty channel in US and UK financial markets

“…Based on the benchmark HEAVY bivariate specification of Shephard and Sheppard (2010), we implement the HEAVY extension introduced by Karanasos and Yfanti (2020), which considers asymmetries (downside risk), power transformations and macro effects. We estimate the macro-augmented model incorporating these features in order to improve the performance of volatility forecasting (see also Karanasos et al, 2021, for a long memory HEAVY extension without macro effects, , for a trivariate AP HEAVY system without macro effects and Yfanti & Karanasos, 2022, for a tetravariate asymmetric HEAVY system without power transformations).…”

“…Market and policy implications: We further illustrate the equity market volatility response to the Covid-19 pandemic shock and the forecasting superiority of the HEAVY extensions during the pandemic-induced market turbulence with a real-world risk management exercise. The widely used daily market risk metric, VaR, denotes the potential loss of a portfolio's value, over a specific holding period, with a given confidence level (see also Karanasos et al, 2021). The VaR calculation's primary input is the 1-day volatility forecast of the portfolio's risk factors.…”