Improving Expected Shortfall estimations with realized measures

This paper will try to improve the correctness of Expected Shortfall (ES) specification by moving from ’regular’ GARCH to realized GARCH models. This is done with the help of multiple realized measures such as the realized kernel and realized variance. The comparisons between GARCH and real-GARCH ES specifications will be done with the Box-Pierce test by Du and Escanciano (2015) and the Saddle Point approximation test by Wong (2008). Though the results looked promising, they were not conclusive.