Title: Comparing out-of-sample forecasts against a random walk: Exact tests with application to exchange rates
Authors: Sermin Gungor - University of Western Ontario (Canada) [presenting]
Richard Luger - Laval University (Canada)
Abstract: An exact inference procedure is developed to test the null hypothesis that a given model's out-of-sample forecasts are no better than a random walk. The proposed Monte Carlo resampling-based procedure accounts for parameter uncertainty and leaves unrestricted the estimation scheme, the forecast horizon, and the forecast evaluation criterion. In contrast to a previously developed Monte Carlo forecast evaluation procedure, our approach is distribution-free and free of nuisance parameters. A simulation study demonstrates the fact that the proposed procedure achieves size control and has good power in comparison to alternative approaches, including various bootstrap methods. We apply the new procedure to test the out-of-sample predictive ability of economic fundamentals in empirical exchange rate models.