Page 348 - Special Topic Session (STS) - Volume 3
P. 348
STS550 Pierre Guérin et al.
results. Under the null hypothesis of no directional accuracy, one would
expect a success ratio of 0.5. We also report the results of the Pesaran and
Timmermann (2009) test to evaluate the statistical significance of the
directional accuracy results. Across all forecasting approaches, the success
ratios tend to be stronger for forecast horizon ℎ = 1, except for the
JPY/USD. In those cases, the improvements in directional accuracy are
often statistically significant according to the Pesaran and Timmermann
(2009) test. It is also interesting to note that the success ratios are
especially strong at distant forecast horizons for selected currencies, as
high as 72.6 per cent for the CAD/USD and 77.0 per cent for the JPY/USD
in the case of the MS-3PRF with regime changes in the first and third
passes.
Overall, while the differences in predictive accuracy tend to be small
across forecasting approaches in terms of point forecasts, the gains in terms
of directional accuracy are strong with the MS-3PRF approach and typically
statistically significant according to the Pesaran and Timmermann (2009)
1
test.
References
1. Camacho, M., Pérez-Quirós, G., and Poncela, P. (2012). Markov-switching
dynamic factor models in real time. CEPR Discussion Papers 8866, C.E.P.R.
Discussion Papers.
2. Diebold, F. X. and Mariano, R. S. (1995). Comparing Predictive Accuracy.
Journal of Business & Economic Statistics, 13(3):253–63.
3. Greenaway-McGrevy, R., Mark, N., Sul, D., and Wu, J.-L. (2016). Identifying
Exchange Rate Common Factors. Mimeo Notre Dame.
4. Kelly, B. and Pruitt, S. (2015). The three-pass regression filter: A new
approach to forecasting using many predictors. Journal of Econometrics,
186(2):294–316.
5. Pesaran, M. H. and Timmermann, A. (2009). Testing Dependence Among
Serially Correlated Multicategory Variables. Journal of the American
Statistical Association, 104(485):325–337.
6. Rossi, B. (2013). Exchange Rate Predictability. Journal of Economic
Literature, 51(4):1063–1119.
1 We also conducted several Monte Carlo exercises to assess the finite sample performance of
the proposed model. We find that the MS-3PRF performs favourably compared with alternative
modelling approaches whenever there is structural instability in factor loadings.
337 | I S I W S C 2 0 1 9