Deutche Bank
Risk forecasts by model
One-day 99% value-at-risk and expected shortfall from each model and their average, as the loss on a portfolio worth 100, and model risk: the highest forecast divided by the lowest.
| Measure | Historical simulation | Moving average | EWMA | GARCH(1,1) | t-GARCH(1,1) | Extreme value theory | Average of the models | Model risk |
|---|---|---|---|---|---|---|---|---|
| Value-at-risk | 6.26 | 4.78 | 5.45 | 4.38 | 5.42 | 5.86 | 5.36 | 1.43 |
| Expected shortfall | 8.23 | 5.48 | 6.25 | 5.02 | 7.59 | 9.78 | 7.06 | 1.95 |
Risk over time
One-day 99% forecasts from each model and their average (the thick line), with the range of daily returns below: each bar runs from the worst to the best day in the period. Select a model to pick it out, or a shorter period to see the models apart.
Loading the chart…
Distribution of daily returns
How often returns of each size occurred, against a normal distribution and a Student-t whose degrees of freedom are the tail index of the largest 0.5% of losses since the start of the data.
Loading the chart…
Extreme events
The worst daily fall to expect over different horizons, and the chance of a fall of a given size, from extreme value theory.
Tail index α 2.49 from the latest 1,000 days. The lower it is, the fatter the tails. A Student-t with α degrees of freedom has tails as thick, and for a normal distribution α is infinite.
Loading the chart…
Model diagnostics
Whether returns and their size persist from day to day (autocorrelation, with the 95% band dashed), and how the tails compare with the normal and that Student-t (QQ plot: points on the line mean a good fit; choose the distribution above the chart).
Loading the chart…