Essays on volatility forecasting, conditional correlations and uncertainty shocks: Evidence from the non-ferrous metals market

This thesis comprises three interconnected topics within the area of finance, contributing to our understanding of volatility by applying novel techniques to forecast and model volatility within the non ferrous metals commodity market, providing valuable insight into their potential usage by industrial users, financial institutions and policy makers for hedging strategies, portfolio diversification, and risk management purposes. Chapter 2 compares the forecasting performance of univariate GARCH models, the GARCH-MIDAS model of Engle, Ghysels and Sohn (2013) with trading volume as a long run component of volatility and the Generalised Autoregressive Score (GAS) model of Creal et al. (2013). We find the standard GARCH model following students-t distribution produces the most accurate forecasts, with trading volume as a predictor variable not able to improve forecasting accuracy. Further VaR analysis corroborates our findings from the MCS test that the standard GARCH model is the most favourable model. In the third chapter, we use various multivariate GARCH models to assess conditional correlations of non ferrous metals with gold, brent crude and the S&P500 index, with likelihood ratio test conducted to select best fitting model to assess hedging effectiveness. Findings highlight that all metals show spikes in conditional correlation for all models in crisis periods, with strong reversals following a crisis period. Results from further wavelet analysis showcase evidence of low comovements between metals at the low frequency horizon. In chapter 4, we explore the effects of geopolitical risk and UK policy uncertainty on non ferrous metals using VAR, SVAR and the TVP-VAR-SV model of Nakajima (2011). Impulse responses at different time horizons produced by selected VAR models exhibit properties of higher non ferrous metal volatility in response to GPR and UK policy uncertainty shocks at short time horizons and during major geopolitical events, although these effects diminish at longer horizons.