GLASSO Estimation of Commodity Risks

In this paper we apply the Graphical LASSO (GLASSO) procedure to estimate the network of twenty-four commodities divided in energy, agricultural and metal sector. We follow a risk management perspective. We use GARCH and Markov-Switching GARCH classes of models with different specifications for the error terms, and we select those that best estimate Value-at-Risk for each commodity. We achieve GLASSO estimation exploring the precision matrix of the multivariate Gaussian distribution obtained from a Gaussian Copula, with marginals given by the residuals of the models, selected via backtesting procedure. The analysis of interdependences in the resulting network is carried out by using the eigenvector centrality metric.