This paper develops new methods for testing equal predictive accuracy in panels of forecasts that exploit information in the time series and cross-sectional dimensions of the data. Using a common factor setup, we establish conditions on cross-sectional dependencies in forecast errors which allow...