A Drawback of VaR

VaR is an attractive measure because it is easy to understand.

It asks the simple question “How bad can things get?”
This is the question all senior managers want answered.
However, when VaR is used to limit the risks taken by a trader, it can lead to
undesirable results.
A Drawback of VaR
The VaR in the two figures are same, but one is much riskier than the other
because a large loss is more likely.
In practice, many trading strategies give a high probability of good returns and a
small probability of a huge loss.
Example: writing out-of-the-money options
◦ Most of the time the trader collects the option premium and does not have to
provide a payoff to the option buyer.
◦ But occasionally the option is exercised in circumstances where the trader takes a big
loss.
Many traders like taking high risks in the hope of realizing high returns.
If they can find ways of taking high risks without violating risk limits, they will do
so.
Expected Shortfall (ES)
Can produce better incentives for traders than VaR.
Also sometimes referred to as:
◦ conditional value at risk,
◦ conditional tail expectation, or
◦ expected tail loss.
VaR asks the question: “How bad can things get?”
ES asks: “If things do get bad, what is the expected loss?”
ES is always larger than the VaR
Expected Shortfall (ES)
ES, like VaR, is a function of two parameters:
◦ T (the time horizon) and
◦ X (the confidence level)
In order to calculate ES it is necessary to calculate VaR first.
ES is the expected loss during time T conditional on the loss being greater than
the VaR.
Example: If X = 99, T = 10 days, and VaR = $64 million,
The ES is the average amount lost over a 10-day period assuming that the loss is
greater than $64 million.
Expected Shortfall (ES)
Setting an ES limit rather than a VaR limit for traders makes it less likely that they
will be able to take the sort of position indicated before.
ES has better properties than VaR in that it always recognizes the benefits of
diversification.
Disadvantages:
ES does not have the simplicity of VaR and as a result is more difficult to
understand.
It is more difficult to back-test a procedure for calculating ES than it is to back-
test a procedure for calculating VaR.
(Back-testing is a way of looking at historical data to test the reliability of a
particular methodology for calculating a risk measure.)
Coherent Risk Measures
Suppose that the VaR of a portfolio for a confidence level of 99.9% and a time
horizon of one year is $50 million.
This means that in extreme circumstances (theoretically, once every thousand
years) the financial institution will lose more than $50 million in a year.
[0.1% = 1/1000]
It also means that if it keeps $50 million in capital it will have a 99.9% probability
of not running out of capital in the course of one year.
Coherent Risk Measures
Suppose we are trying to design a risk measure that will equal the capital a
financial institution is required to keep.
There are a number of properties that such a risk measure should have:
1. Monotonicity: If a portfolio produces a worse result than another portfolio for
every state of the world, its risk measure should be greater.
2. Translation Invariance: If an amount of cash K is added to a portfolio, its risk
measure should go down by K.
3. Homogeneity: Changing the size of a portfolio by a factor λ while keeping the
relative amounts of different items in the portfolio the same should result in the
risk measure being multiplied by λ.
Coherent Risk Measures
4. Subadditivity: The risk measure for two portfolios after they have been
merged should be no greater than the sum of their risk measures before they
were merged.
When we aggregate two portfolios, the total risk measure should either
decrease or stay the same.

Risk measures satisfying all four conditions given above are referred to as
coherent.
VaR satisfies the first three conditions. Does not always satisfy the fourth one.
VaR is not coherent. ES is always coherent.
Coherent Risk Measure [Example]
VaR is not a coherent risk measure whereas ES is a coherent risk measure
Suppose each of two independent projects has a probability of 0.02 of a loss of
₹10 million and a probability of 0.98 of a loss of ₹1 million during a one-year
period.
The one-year, 97.5% VaR for each project is ₹1 million.
When the projects are put in the same portfolio,
there is a 0.02 × 0.02 = 0.0004 probability of a loss of ₹20 million,
a 2 × 0.02 × 0.98 = 0.0392 probability of a loss of ₹11 million,
and a 0.98 × 0.98 = 0.9604 probability of a loss of ₹2 million.
Coherent Risk Measure [Example]
The one-year 97.5% VaR for the portfolio is ₹11 million.
The total of the VaRs of the projects considered separately is ₹2 million.
The VaR of the portfolio is therefore greater than the sum of the VaRs of the
projects by ₹9 million.
This violates the subadditivity condition.
Coherent Risk Measure [Example]
The VaR for one of the projects considered on its own is ₹1 million.
To calculate the ES for a 97.5% confidence level we note that, of the 2.5% tail of
the loss distribution, 2% corresponds to a ₹10 million loss and 0.5% to a ₹1
million loss.
Conditional that we are in the 2.5% tail of the loss distribution, there is therefore
an 80% probability of a loss of ₹10 million and a 20% probability of a loss of ₹1
million. The expected loss is 0.8 × 10 + 0.2 × 1 or ₹8.2 million.
When the two projects are combined, of the 2.5% tail of the loss distribution,
0.04% corresponds to a loss of ₹20 million and 2.46% corresponds to a loss of
₹11 million. Conditional that we are in the 2.5% tail of the loss distribution, the
expected loss is therefore (0.04∕2.5) × 20 + (2.46∕2.5) × 11 or ₹11.144 million.
This is the ES.
Coherent Risk Measure [Example]
Because 8.2 + 8.2 > 11.144, the ES measure does satisfy the subadditivity
condition for this example.
Choice of Parameters for VaR and ES
The Time Horizon: An appropriate choice depends on the application.
When positions are very liquid and actively traded:
◦ It makes sense to use a short time horizon (perhaps only a few days).
◦ If the measure calculated turns out to be unacceptable, the portfolio can be adjusted
quickly.
◦ A longer time horizon might not be meaningful because of changes in the
composition of the portfolio.
When VaR or ES is being calculated by the manager of a pension fund:
◦ A longer time horizon is likely to be used.
◦ Because the portfolio is traded less actively and some of the instruments in the
portfolio are less liquid.
Choice of Parameters for VaR and ES
When the liquidity of a portfolio varies from one instrument to another, the
definition of VaR or ES can be changed so that the changes considered vary from
one market variable to another.
Example: A portfolio consisting of:
◦ Actively traded shares and
◦ A corporate bond that trades fewer than 10 times per year.
It could make sense to calculate a risk measure from:
◦ the change in the price of shares that we are 99% confident will not be exceeded
over 10 days and
◦ the change in the bond price that we are 99% certain will not be exceeded over 60
days.
Choice of Parameters for VaR and ES
Whatever the application, when market risks are being considered, analysts
often start by calculating VaR or ES for a time horizon of one day.
The usual assumption is

These formulas are exactly true when the changes in the value of the portfolio
on successive days have independent identical normal distributions with mean
zero.
In other cases, they are approximations.
Choice of Parameters for VaR and ES
Confidence Level: Likely to depend on a number of factors.
Suppose that a bank wants to maintain an AA credit rating and calculates that
companies with this credit rating have a 0.03% chance of defaulting over a one-
year period.
It might choose to use a 99.97% confidence level in conjunction with a one-year
time horizon when calculating VaR for internal risk management purposes.
If daily portfolio changes are assumed to be normally distributed with zero
mean, we can convert a VaR or ES calculated with one confidence level to that
with another confidence level.
Marginal, Incremental, and Component
Measures
Consider a portfolio that is composed of a number of subportfolios.
The subportfolios could correspond to:
◦ Asset classes (e.g., domestic equities, foreign equities, fixed income, and derivatives).
◦ Business units (e.g, retail banking, investment banking, and proprietary trading).
◦ Individual trades.
Analysts sometimes calculate measures of the contribution of each subportfolio
to VaR or ES.
Marginal, Incremental, and Component
Measures
Suppose that the amount invested in ith subportfolio is xi.
The marginal value at risk for the ith subportfolio is the sensitivity of VaR to the
amount invested in the ith subportfolio.
It is:
Marginal, Incremental, and Component
Measures
The incremental value at risk for the ith subportfolio is the incremental effect of
the ith subportfolio on VaR.
It is the difference between VaR with the subportfolio and VaR without the
subportfolio.

Traders are often interested in the incremental VaR for a new trade.
Marginal, Incremental, and Component
Measures
The component value at risk for the ith subportfolio is:

Marginal ES, incremental ES, and component ES can be defined similarly to

marginal VaR, incremental VaR, and component VaR, respectively.
Aggregating VaRs and ESs
Sometimes a business has calculated VaRs, with the same confidence level and
time horizon, for several different segments of its operations and is interested in
aggregating them to calculate a total VaR.
A formula for doing this is:

This is exactly true when the losses (gains) have zero-mean normal distributions
and provides a good approximation in many other situations.
The same is true when VaR is replaced by ES in the given equation.
Aggregating VaRs and ESs
Suppose the ESs calculated for two segments of a business are $60 million and
$100 million.
The correlation between the losses is estimated as 0.4.
An estimate of the total ES is