Predicting Tail-related Risk Measures: The Consequences of Using GARCH Filters for non-GARCH Data

We investigate the consequences for value-at-risk and expected short-fall purposes of using a GARCH filter on various mis-specified processes. We show that careful investigation of the adequacy of the GARCH filter is necessary since under mis-specifications a GARCH filter appears to do more harm than good. Using an unconditional non filtered tail estimate appears to perform satisfactorily for dependent data with a degree of dependency corresponding to actual market conditions.