Complementing Expected Shortfall with Directly Observable Risk Variables

Abstract: This paper uses daily MSCI Sweden Index data to create one day out ES estimates on the $97.5\%$ percentile level for a SEK 100 portfolio. Thereafter the variables (i) the Riksbank's policy rate, (ii) the Riksbank's balance sheet, (iii) the cyclicality of GDP, and (iv) animal spirits are introduced and compared with ES. As the variables are thought to affect financial markets, they are evaluated in regards to ES to see if the addition of them in an analysis improves estimation of the 2.5% worst losses. If ES does not incorporate signals of risk shown by the variables, increased predictive ability for losses is achieved by including the variables in an analysis. The results indicate that (i), (ii), and (iv) are worth including in addition to ES when estimating the cost of the 2.5% largest losses of a stock index.