The article shows that using realized variance measures in risk models can improve forecasts of Expected Shortfall and Value at Risk for stock indices. The researchers developed a method that outperforms existing models and demonstrated that combining information from different methods can lead to even better forecasts.