VaR Flashcards
Which tail is used for VaR?
Left
When did computers add calculation muscle to risk?
1980
two types of financial institutions that primarily drive risk
private equity and hedge funds
5 steps of enterprise risk management
- risk transparency and insight
- natural ownership and risk strategy
- risk capacity
- risk-related decisions and processes
- risk organization and governance
investment bank that risks the most money
goldmann sachs
when did G10 agree to set minimum capital requirements?
1988
when were banks allowed to use their own proprietary models?
1995
VaR level with 95% confidence
“I am 95% confident that I will not lose more than $xx in one day”
what does a closed form of VaR assume?
- normal distribution
- distribution with a specific mean and standard deviation
what does closed form of VaR explicitly determine?
the p/l probability distribution for a portfolio
what does closed form VaR implicitly determine?
the standard deviation and correlation of p/l
confidence interval recommended by Basle Committee?
99 percent
properties of a normal curve
- perfectly symmetric
- no skew
- mean = 0
- kurtosis = 3 x variance squared
what do fat tails suggest
a higher chance of very high or very low prices
when is fat tails classified
when kurtosis > 3
are options linear or non-linear
options are non linear , price of option doesn’t move 1 to 1 with the price of the underlying
how are portfolios with optionality measured?
by using simulations/estimations of VaR
factors effecting simulations/estimations of VaR
options (delta, theta, gamma)
bonds (complexity and duration)
equities (beta)
curse of dimensionality
multiple risk factors create exponential layers of calculations
models that estimate VaR and open-form equations
Historical and Monte Carlo
historical uses past events to predict the probability of future events
monte carlo incorporates all risk factors into complex calculations
error that occurs with Monte Carlo VaR
specification error = random sampling of risk metrics doesn’t represent true distribution
convergence error = not enough sampling is done
VaR model to use if no optionality
Parametric (variance or covariance) VaR
VaR model with curse of dimensionality
Monte Carlo
distribution kurtosis type for enegry markets? opposite?
Leptokurtotic = woody, concentrated center with fat tails opposite = platokurtotic
two types of VaR model likely to give errors in energy market
parametric and historical
concepts to consider in overcoming VaR limitations
- use stress tests and scenario analysis
- VaR gives probability of a loss that could occur past a certain $xx, but doesn’t specify how far $xx that loss will go
3 types of stress tests
- historical scenarios (using past tail risk events)
- mechanical stress tests (yield curve shifts, forward price curve shifts, changes in volatility curve)
- hypothetical scenarios (user-designed)
shocks that can be used in stress tests
parallel= equal shocks given across the curve twist = equal shocks given to high/low end of curve but with different signs (+, -) curvature = equal shocks given to high/low end of curve but with the same sign. same shock is also applied to the center of the curve but w/ a different sign
problems with stress tests
- subjective
- no probability
- information overload
