Measuring and Managing Market Risk Flashcards
An advantage of statistical factor models
They make minimal assumptions.
However, the interpretation of statistical factors is generally more difficult than the interpretation of macroeconomic and fundamental factor models.
Assumption of CAPM
- Perfect competition - Frictionless and can borrow a the RfR
- Rational, mean-variance optimizer
- Perfect information (same variance and covariance matrix)
What is the purpose of VaR
Value at Risk is to capture market risk.
Equity prices
Commodity prices
Forex
Interest rates
Does not tell about about average loss
How to interpret a one day 95% VaR
95% confidence that we will NOT lose more than … per day
with 95% probability, we will experience a maximum loss of …
How to interpret 5% VaR
The 5% minimum loss of a portfolio over a 1 day period
or…
A expected loss of … to occure every 20 days (depend on duration)
3 different ways to estimate VaR
Parametric method
Historical simulation method
Monte Carlo Simulation
Explain Parametric Method
Variance - Covariance method
Begins with risk decomposition of the portfolio holdings
Assumes return distribution for risk factor is normal distributed
We need expected returns and standard deviation of portfolio
Calculate VaR for parametric method
[Expected return - Z* Portfolio standard dev]*(-1) * Pv
Z = Standard deviation number
Pros and Cons of parametric method
Pro: Simple and straightforward
Con: VaR is very sensitive to expected returns and standard deviation
Difficult to use of portfolio contains options since it threatens normality. Options have a non normal payoff function.
Historical simulation method
We set / construct a portfolio with fixed weights
We measure portfolio return over the observed period
We then rank the portfolio returns from smallest to largest
We then use percentile to find 1,5,10 % VaR
- I we have 500 observation, and we want to find 5% VaR, the 25th observation os our 5% VaR
.
Brief characteristics of historical simulation method
Not constrained by normality assumption
Estimates VaR based on what actually happened
Can handle any kind of financial instruments
Monte Carlo Simulation
Not constrained by any distribution - We can define the distribution
Avoids complexity of parametric method when portfolio has many risk factors
Calculating VaR is the same as historical method
Conditional VaR - CVaR
Relies on a particular VaR measure - Average loss greater than our particular VaR measure.
average loss on the condition that VaR > Cut off
Informs us about average loss
Typically obtained by backtesting
Also known as Expected tail loss or expected shortfall
Incramental VaR - IVaR
How VaR will change if a position size changes relative to the remaining position.
example:
SPY - 80 % weight - > 90% weight
LWC - 20 % weight -> 10% Weight
VaR 2,407,503 -> 2,733,722
IVaR = 2,733,722 - 2,407,503 = 326,192
Marginal VaR - MVaR
Conceptually the same to IVaR, but reflects the effects of a very small change in a position - 1 unit change in position.