Exchange-rate forecasts with simultaneous nearest-neighbour methods: evidence from the EMS

In this paper we extend nearest-neighbour predictors to allow for informati on content in a wider set of simultaneous time series. We apply these simul taneous nearest-neighbour (SNN) predictors to nine EMS currencies, using da ily data for the 1st January 1978-31st December 1994 period. When forecasti ng performance is measured by Theil's U statistic, the (nonlinear) SNN pred ictors perform marginally better than both a random walk and the traditiona l (linear) ARIMA predictors. Furthermore, the SNN predictors outperform the random walk and the ARIMA models when producing directional forecasts. Whe n formally testing for forecast accuracy, in most of the cases the SNN pred ictor outperforms the random walk at the 1% significance level, while outpe rforming the ARIMA model in three of the nine cases. On the other hand, our results suggest that the probability of correctly predicting the sign of c hange is higher for the SNN predictions than the ARIMA case. (C) 1999 Elsev ier Science B.V. All rights reserved.