DYNAMIC PROGRAMMING, MULTIVARIATE GARCH MODELS, AND FUTURES HEDGING

Bent Jesper Christensen

Abstract

Multivariate GARCH models are used to capture the joint dynamics of the volatilities of commodities and futures contracts. This allows the development of optimal time-varying hedge ratios, given as solutions to suitable stochastic dynamic programming problems. To an outside observer, the likelihood of observed hedging strategies is initially degenerate. The way in which the degeneracy is broken determines what can be learned about the stochastic process.