15. Risk modelling Flashcards
How would you quantify enterprise risk using dynamic financial analysis?
- Use cashflow models to produce balance sheet and p+l accounts
- Model risks of whole org and relationships between risks
How would you quantify enterprise risk using financial condition reports
- Financial condition report: report of current solvency position and possible future development
- Look at risks exposed ro, project expected level of profits and new business, incl. unusual features
How would you quantify market risk
- VaR
- Tail VaR
- Interest rate models
- Scenario tests
How would you quantify underwriting risk
- Credit risk models for single entity
- Also assessed non-quantitatively e.g. by banks and credit rating agencies
How would you quantify credit risk
- Credit risk models for single entity
- Also assessed non-quantitatively e.g. by banks and credit rating agencies
How would you quantify liquidity risk
- A+L models
How would you quantify operational risk
- Internal and external loss data
- Scenario analysis
- Simulations
How would you quantify environmental risk
- Internal and external loss data
- Scenario analysis
- Simulations
How would you quantify demographic risk
Quantitative methods
Factors to consider when quantifying risk
Extreme events
Data limitations
Interdependence
Unquantifiable risks
Outline what correlation is
- Measures how diff variables relate or associate to each other
- Low level: diversify each other to some extent
- Negative: hedge each other
Name a linear correlation measure
Person’s rho
Merits of Person’s rho
Advantages
Value unchanged under operation of strictly increasing linear transformations
ρ(a+bX,c+dY)=ρ(X,Y) where b and d>0
Disadvantages
Value not unchanged under operation of strictly non-linear increasing transformations
Not well defined where variances are infinte so can’t be used for some heavy-tailed distributions
Independent variables are uncorrelated, r(x,y) = 0, but not all uncorrelated variables are independent (i.e. r(x,y) = 0 doesn’t imply independence just not linear relationship)
Pearson’s rho only valid correlation measure if marginals are jointly elliptical
Given marginals X and Y, and specified rho, won’t necessarily be able to put together joint distribution to combine all info. Value of ro may be one that is unattainable, ie incompatible with marginal distributions**
Name two rank correlation measures
- Kendall’s tau
- Spearman’s rho