ARFIMA Tests for Random Walks in Exchange Rates in Asian, Latin American and African-Middle Eastern Markets
David Karemera () and
John Cole ()
Additional contact information David Karemera: School of Business, South Carolina State University, Orangeburg, South Carolina 29117, USA
John Cole: School of Business and Economics, North Carolina A&T State University, Greensboro, NC 27411, USA
This article examines fractional processes as alternatives to random walks in emerging foreign exchange rate markets. Sowell's (1992) joint maximum likelihood is used to estimate the ARFIMA parameters and test for random walks. The results show that, in most cases, the emerging market exchange rates follow fractionally integrated processes. Forecasts of exchange rates based on the fractionally integrated autoregressive moving average models are compared to those from the benchmark random walk models. A Harvey, Leybourne and Newbold (1997) test of equality of forecast performance indicates that the ARFIMA forecasts are more efficient in the multi-step-ahead forecasts than the random walk model forecasts. The presence of fractional integration is seen to be associated with market inefficiency in the exchange markets examined. The evidence suggests that fractional integrated processes are viable alternatives to random walks for describing and forecasting exchange rates in the emerging markets.