Title: Multi-class Vector AutoRegressive Models for Multi-Market Commodity Data
Abstract: Vector AutoRegressive (VAR) models form a special case of multivariate regression models in that the response variables are observed over time and modeled as a function of their own past values. We use VAR models to study dynamics among agricultural, metal and energy commodity returns. As the increasing integration of the world economy suggests commodity dynamics to be comparable for different markets, we aim to jointly analyze these dynamics across markets. To this end, we introduce a sparse estimator of the Multi-Class Vector AutoRegressive model. We jointly estimate several VAR models, one for each market (“class”), to borrow strength across markets. Our methodology encourages effects to be similar across markets, while still allowing for small differences between them. Moreover, we focus on multi-class estimation of high-dimensional VAR models, i.e. models with a large number of time series relative to the time series length. Therefore, our estimate is sparse: unimportant effects are estimated as exactly zero, which facilitates the interpretation of the results.
This is joint work with Luca Barbaglia and Christophe Croux