GAS and GARCH based value-at-risk modeling of precious metals

Abstract

We employ 38 VaR model specifications (32 GARCH and - 6 GAS), assuming Gaussian and non-Gaussian distributional innovations. Using the elicitability property of VaR, we further use the Model Confidence Set (MCS) technique, which creates superior set models (SSMs) and ranks them based predictive ability of the VaR forecasts. We employ 4580 daily log-returns of Gold, Palladium, Platinum, and Silver, which span January 01, 2000, to April 04, 2018, which covers turbulent (Eurozone and Global Financial crises periods) and tranquil (post-Global Financial crisis period) market conditions. We find that, for both 1% and 5% VaR forecasts, Platinum exhibits a higher level of heterogeneity among models in contrast with Silver, Gold, and Palladium. Hence, Platinum has the smallest number of models in the SSM. Empirically, the homogeneity in the SSM is suggestive of well-diversified portfolios for the respective metals. Except for a few models, both DQ and CC tests support adequate forecast abilities of the respective 1% and 5% VaR models. This suggests the strength of the MCS procedure to select superior set models as compared to the initial set of 38 models. Our study is important for internal risk modelling, regulatory oversight and may bolster confidence in global investors concerning investments in precious metals.