Abstract
This thesis is an empirical study of the volatility and correlations among the key commodity markets for metals and crude oils. Both conventional and state-of-the-art models will be applied, which are presented in three main chapters. First, Chapter 2 provides a thorough literature review on commodities that paves the way for the three substantive chapters. It discusses the distinctive characteristics of commodity markets, with special emphasis on gold and crude oil as potential trading options for portfolio diversification. The aim is to present the financial side in the form of a comprehensive and up-to-date survey in commodity investment for stakeholders. Second, Chapter 3 applies the widely-used conventional techniques of Constant Conditional Correlation (CCC) and Dynamic Conditional Correlation (DCC) based on the Generalised Autoregressive Conditional Heteroskedasticity (GARCH) model for volatility in these markets, with particular attention paid to the dynamics of volatility and correlations in the crude oil and gold markets. This study analyses the weekly spot data between 1986[M1:W1] and 2016[M9:W4], with the period being further divided into four significant sub-periods, enabling cross-period comparison. The objective is to use the latest available data for analysis, provide an insight into portfolio diversification, and explore if the markets are of one. The findings suggest that gold can be used to diversify away from crude oil. Interestingly, palladium seems to have low and even negative correlations with other metal commodities, suggesting that palladium may be a new safe option within commodity markets. The correlations between the crude oil markets and the metal markets increased significantly in early 2000s, and continued through financial crisis in 2008 until they peaked between 2009/2010. These observations could be attributed to the surge in trading interest within commodity markets. Third, Chapter 4 presents a state-of-the-art model, the Continuous Wavelet Transform (CWT), using the same three-decade data as the previous chapter. The study begins with examining the co-movements between commodities' returns with a set of continuous wavelet tools, which consist of wavelet variance, covariance, coherency, phase-difference and gain. This is achieved by extending the estimated variables to include higher order coefficients in the time-frequency domain. The findings indicate that gold and oil exhibit low covariance and coherency. In addition, the study also uncovers a set of new stylized facts for commodities (crude oils and metals) that would not be possible to detect with pure time- or frequency-domain methods, nor with the time-frequency domain tools available thus far. While the results support the previous chapter in showing low gold-oil coherency, wavelet analysis offers exceptional level of stylised facts that were not previously available for commodity markets, i.e. frequency analysis of multiple investment horizons. Fourth, Chapter 5 builds on the CWT analysis used in the previous chapter with two major enhancements. In this study, the phase-difference circle is reconstructed into a non-technical and more simplified diagram format to facilitate result interpretation for non-experts. Also, the wavelet working tools are recalibrated so that gain values can be extracted to increase the robustness interpretations of the gain results. Moreover, both spot and future prices are used in this study to examine the Gold-to-Oil Ratio (GOR) and the gold and oil implied volatility indices from 2008[M6:W1] to 2017[M1:W1]. The gain coefficients uncovered interesting findings. For example, the gain frequency for one to four years is always greater than the gain coefficient values of the other two frequencies. Finally, Chapter 6 concludes the three substantive chapters by summarising and contrasting the findings. It also discusses possible directions for future research.