8. generating unidimensional scenarios
Guidelines for scenarios by DPG
✓Parallel yield curve shifting by ±100 basis points
✓ Yield curve twisting by ±25 basis points
✓ Each of the four combinations or yield curves shift and
twist
✓ Equity index values changing by ±10 percent
✓ Currencies moving by ±6 percent for marjor currencies
✓ Swap spreads changing by ±20 basis points
9. The OTS*
requiers
to estimate what would happen to
your economic value under parallel
shifts in the yield curve varying
from -400 to +400 basis points
*The office of Thrift Supervision
10. The SPAN system*
The objective:
to identify movements in portfolio values
under a series of scenarios
➙ SPAN searches for the largest loss that a
portfolio may suffer and sets the margin at
that level
*1998, Chicago Mercantile Exchange
11. The SPAN system
✔ it considers only two risk
factors
✘ places the same probability on
most of the scenarios
✘ ignores correlations between
risk factors
15. Factor Push Method
➙ push up and down ➙ compute the changes to the
portfolio
➙ evaluate a worst-case scenario, pushing all
parametres in the «worst loss» position
*completely ignores correlation
*looking at extreme movements may not be appropriate
16. Conditional scenario Method
≈conditional normal VAR modeling
▸ at the core is the covariance matrix
▸ «KEY» variables R* + other variables R
ASL = ∑ w *R *+ ∑ w R
i i i j j j
✘ results are much closer to the actual stress loss
✘ only in case where correlation plays the main role
17. Historical Scenarios
✓ helps to identify
scenarios that may be
outside the VAR window
✓ are useful to measure
joint movements in financial
variables
18. Systematic Scenarios
This analysis provides insight into the vulnerabilities of a
particular portfolio
Maximum-loss criterion
To identify the wors loss through an optimization that
respects the correlations between the risk-factor
movements ∆⨍
ML = min loss (∆⨍ ) subject to ⨍ˈ∑⁻¹ ∆⨍ ≤ C
19. Stress testing model parameters
Sensitivity
We need details!
analysis
Model
Use the same period!
parameters
20. Managing Stress Tests
Relevant scenarios require careful planning
(1) Purchase protection or insurance for the
events in quiestion
(2) Modify the portfolio to decrease the impact of
the event
(3) Diversify
(4) Develop a plan to correct course of action
(5) Prepare sources of alternative funding
21. Conclusions
People may be very bad at pre!cting extreme situations
✓stress test can be viewed as an extension of the historical
simulation method at increasingly higher confidence levels
✓«blind spots» or complement to standard VARmth
✓a complement to VARmth rather than a replacement
✗ can lead to a large amount of unfiltered information
✗ the risk to produce large numbers of scenarios
✗ difficult to interpret