The Usefulness of the Currency Constraint for Predicting Forex

The research explores the comparative predictive capacities of the base currency and the counter currency in projecting future currency pairs. It employs a system of structural regressions with a framework of two simultaneous linear models, recognising the fundamental attribute of the forex market that a currency pair is composed of a base currency and a counter currency. Furthermore, an investigation is undertaken to ascertain whether the use of two simultaneous models would result in the constrained seemingly unrelated regression (SUR) estimator generating reduced prediction errors in comparison to each individual model. The study’s findings demonstrate that when forecasting currency pairs, (1) the base currency demonstrates superior explanatory power compared to the counter currency, and (2) a system of structural regressions with two simultaneously constrained models yields substantially diminished prediction errors compared to each separate regression.