Cointegrated Vector Autoregression Methods: An Application to Non-Normally Behaving Data on Selected U.S. Sugar-Related Markets

The methods of the cointegrated vector autoregression/error correction (VAR/VEC) model are applied to monthly U.S. markets for sugar and for sugar-using markets for confectionary, soft drink, and bakery products. Primarily a methods paper, Johansen and Juselius' methods are applied, with a special focus on addressing well-known issues that preclude statistically normal behavior, and that confront the modelled sugar-based data. In so doing, we illustrate the effectiveness and the benefits of modelling this sugar-related set of markets as a cointegrated system. Perhaps for the first time, cointegrated VEC model results are used to estimate crucial policy-relevant market parameters that drive the markets, as well as to illuminate the dynamic nature of the relationships linking these sugar-based markets.