Mildenhall Ch 6: Risk measures in practice Flashcards
Identify 3 ways to select risk measures
- Ad hoc method
- Economic method
- Characterization method
Briefly explain the adhoc method to select risk measures and provide an example
Start with reasonable risk measure & rationalize it by establishing if it has desired properties or argue against it based on its undesirable properties.
Ex: % loading or % of standard deviation
Briefly explain the economic method to select a risk measure and provide an example
Use a rigorous economic theory to select a risk measure.
Ex: utility-based methods that solve optimization problem.
This is the most rigorous selection process but also the hardest to apply in practice.
Briefly explain the characterization method to select risk measures.
Start with a list of desired properties and then determine which risk measures have these properties.
This method is more scientific than ad hoc method and easier to apply than economic method.
Identify 5 important properties of risk measures from practical point of view
- Risk measures must reflect diversification (both COH and CONVEX do)
- When applied in aggregate, risk measure must allow practical allocation to its individual parts (COH does)
- Risk measure must align with theory (COH does)
- Risk measure must be easily explained to users (COH does)
- Backtesting: risk measure should be consistent with observations.
- Elicitability: risk measure can be estimated by regression-like techniques
- Robustness & continuity: small change in inputs should result in small change in measured risk
- MON & TI reflect intuitive concepts of risk & are theoretically and practically sound.
Evaluate VaR based on the desired properties.
VaR does not reflect diversification & ignores tail.
However, it is fairly robust.
If we consider VaR for a range of thresholds, we can somewhat capture the tail.
Evaluate TVaR based on the desired properties.
TVaR reflects diversification but is hard to elicit and difficult to backtest.
List the 5 desirables characteristics of risk margins.
- The less is known about current estimate and trend, the higher the risk margin should be.
- Low frequency higher severity risks should have higher risk margins than high frequency low severity risks.
- For similar risks, longer duration contracts should have higher risk margins than risks with narrower distribution.
- Risks with wide probability distribution should have higher risk margins than risks with narrower distribution.
- Risk margins should decrease with emerging experience since it reduces uncertainty (& vice versa)
List 3 approaches to estimate risk margins
- Outcome methods (VaR or TVaR)
- Cost of Capital
- Discount-related methods
- Explicit assumptions
- Conservative assumptions
List the 8 degrees of tail thickness (from thickest to thinnest)
Thick tailed:
1. No mean
2. Mean, but no variance
3. Limited existing moments
4. Subexponential tail
In between:
5. Exponential tail
Thin-tailed:
6. Superexponential tail
7. Log-concave density
8. Bounded
For which degrees of thickness do law of large numbers and central limit theorem not apply?
Law of large numbers:
No mean only
Central limit theorem:
No mean and Mean, but no Variance
Describe 4 specific uses of risk measures
- Individual risk pricing: quoting and evaluation of market pricing
- Classification ratemaking: setting profit margins and allocation CoC
- Portfolio management: reinsurance purchase, portfolio optimization and ORSA
- Determining risk capital or evaluation held capital
Define intended purpose
Goal addressed by model within context of assignment.
The intended purpose determines the level of detail of modeling (stand alone vs part of portfolio) & appropriate time period for modeling.
Define model
Simplified representation of relationship among:
1. real world variables
2. entities
3. events
using:
a. statistical
b. financial
c. economic
d. mathematical
e. non-quantitative
f. scientific
concepts and equations.
Briefly explain the 3 components of a model
- Information inputs: delivers data & assumptions to model
- Processing: transforms input into output
- Result: translates output into useful business information