CMStatistics 2020: Start Registration
View Submission - CFE
Title: Comparing predictive accuracy under unconditional heteroskedasticity Authors:  Yang Zu - University of Nottingham (United Kingdom) [presenting]
Steve Leybourne - University of Nottingham (United Kingdom)
Dave Harvey - University of Nottingham (United Kingdom)
Abstract: The impact of unconditional heteroskedasticity on Diebold Mariano (DM) test for equal forecast accuracy is considered. We analyse the power of the DM test and propose two new powerful DM type tests by exploiting the heteroskedasticity structure in data. Empirical applications to evaluating exchange rate forecasts and the forecasts made by professionals are considered.