2026-07-09http://hdl.handle.net/123456789/534649Two asymptotically valid out-of-sample MSE tests have been developed by Diebold-Mariano (1995) and Stock-Watson (1999). The empirical usefulness of the tests is illustrated through a U.S. wheat model estimated with fixed, recursive and rolling forecasting schemes. Bootstrap methods are adopted to reflect small sample size effect on tests.NEW MSE TESTS FOR EVALUATING FORECASTING PERFORMANCE: EMPIRICS AND BOOTSTRAPText