Note(s): RMSE = root mean squared error; MAE = mean absolute error; $R2$ = out-of-sample coefficient of determination (negative values indicate worse performance than a mean benchmark). Lower RMSE/MAE and higher $R2$ indicate better performance