Exchange Rate Models and the Management of Forex Losses in Ghana: Modelling Exchange Rate Volatilities for Businesses

Abstract

Using the Self-exciting Threshold Autoregressive Model (SETAR_M) and linear models such as the vector error correction model (VECM), and univariate models, this article specifies forecasting models for exchange rate volatilities in Ghana and compares their forecasts accuracy using Diebold–Mariano and Pesaran-Timmermann tests statistics. The relevance of this research is to equip business owners and businesses on managing forex losses and to reduce their impact on profits, productivity and employment in high volatile and unstable currency environments. The research concludes that the non-linear SETAR model is superior to the linear models in predicting short-term volatilities in exchange rates, while the fundamentally based linear model is superior for predicting long-term volatility in exchange rates. Therefore, short-term business commitments or transactions such as raw material purchases, cash expenses or incomes in foreign currencies should be planned or managed using SETAR or a non-linear model, whereas long-term contractual obligations like futures and forward contracts should be planned with a fundamentally based multivariate linear model.