6 Value at Risk (VaR)

 The VaR measure

 Properties and limitation of risk measure
 Expected Shortfall
 Choice of parameters for VaR
 Back Testing and Stress Testing
 The Question Being Asked in VaR
2

 “What loss level is such that we are X% confident it will

not be exceeded in N business days?”

 In other words, we are going to make the statement of the

following form:
 “We are X% certain that we will not lose more than V
dollars in the next N business days.”
 The variable V is the Value-at-Risk of the portfolio.
 VaR and Regulatory Capital
3

 Regulators require banks to keep certain capital level

based on VaR:
 The market-risk capital is k times the 10-day 99% VaR
where k is at least 3.0
 Capital for credit risk and operational risk is based on a
one-year 99.9% VaR.
 Advantages of VaR
4

 Historical perspectives on VaR

 Advantages of VaR:
 It captures an important aspect of risk in a single number
 It is easy to understand

 It asks the simple question: “How bad can things get?”

 Example 1, 2
5

 Example 1:
 The gain from a portfolio during six month is normally
distributed with mean $2 million and standard deviation
$10 million. The 1% point of the distribution of gains is
2−2.33×10 = − $21.3 million.
 The VaR for the portfolio with a six month time horizon
and a 99% confidence level is $21.3 million.
 Example 2:
 All outcomes between a loss of $50 million and a gain of
$50 million are equally likely for a one-year project
 The VaR for a one-year time horizon and a 99% confidence
level is $49 million
 Example 3
6

 A one-year project has a 98% chance of leading to a gain of

$2 million, a 1.5% chance of a loss of $4 million, and a 0.5%
chance of a loss of $10 million
 The VaR with a 99% confidence level is $4 million.
 What if the confidence level is 99.9%? Or 99.5%?
 VaR vs. Expected Shortfall
7

 VaR is the loss level that will not be exceeded with a

specified probability.
 When VaR is used in an attempt to limit the risks taken by a
trader, it can lead to undesirable results.
 Expected shortfall is the expected loss given that the loss
is greater than the VaR level (also called Conditinal-VaR
and Tail Loss)
 Two portfolios with the same VaR can have very different
expected shortfalls.
 The limitation of VaR
8

Distributions with the Same VaR but Different Expected

Shortfalls:

VaR

VaR
 Properties of Risk Measures
9

 Monotonicity: If one portfolio always produces a worse

outcome than another, its risk measure should be greater
 Translation invariance: If we add an amount of cash K to a
portfolio its risk measure should go down by K
 Homogeneity: Changing the size of a portfolio by l should
result in the risk measure being multiplied by l
 Subadditivity: The risk measures for two portfolios after
they have been merged should be no greater than the sum
of their risk measures before they were merged
 Coherent Risk Measures
10

 Risk measures satisfying all the previous four conditions

are referred to as Coherent risk measures.

 VaR satisfies the first three conditions but not the fourth
one.
 Expected shortfall satisfies all four conditions.
 Spectral Risk Measures
11

 We can define what is known as a spectral risk measure

by making other assumptions about the weights assigned
to quantiles of the loss distribution.
 VaR assigns all weight to Xth quantile of the loss
distribution.
 Expected shortfall assigns equal weight to all quantiles
greater than the Xth quantile.
 For a coherent risk measure weights must be a non-
decreasing function of the quantiles.
 Example 4
12

 Two $10 million one-year loans each has a 1.25% chance of

defaulting. All recoveries between 0 and 100% are equally
likely. If there is no default the loan leads to a profit of $0.2
million. If one loan defaults it is certain that the other one will
not default.
 What is the 99% VaR and expected shortfall of each project?
(Answers: VaR: $2, $2; Expected shortfall: $6, $6)

 What is the 99% VaR and expected shortfall for the portfolio?
(Answers: VaR: $5.8; Expected shortfall: $7.8)
 Question 1
13

 Suppose that each of two investments has a 4% chance of

a loss of $10 million, a 2% chance of a loss of $1 million,
and a 94% chance of a profit of $1 million. They are
independent of each other.
 (a) What is the VaR and the expected shortfall for one of the
investments when the confidence level is 95%?
 (b) What is the VaR and the expected shortfall for a
portfolio consisting of the two investments when the
confidence level is 95%?
 (c) Show that, in this example, VaR does not satisfy the
Subadditivity condition whereas expected shortfall does.
 Question 2
14

 Assume that portfolio daily returns are independently and

identically normally distributed. Sam Neil, a new quantitative
analyst, has been asked by the portfolio manager to calculate
the portfolio Value-at-Risk (VaR) measure for 10, 15, 20 and
25 day periods. The portfolio manager notices something
amiss with Sam’s calculations displayed below. Which one of
following VARs on this portfolio is inconsistent with the
others?
 a. VAR(10-day) = USD 316M

 b. VAR(15-day) = USD 465M

 c. VAR(20-day) = USD 537M

 d. VAR(25-day) = USD 600M

 Choice of Parameters for VaR
15

 The user must choose two parameters:

 The time horizon
 The confidence level

 The simplest assumption is that daily gains/losses are

normally distributed and independent with mean zero.
 VaR   N 1 ( X )
X : the confidence level
 : the standard deviation of the portfolio change over time
N 1: the inverse cumulative normal distribution
 The Time Horizon
16

 With the i.i.d normal distribution assumption, it is then

easy to calculate VaR from the standard deviation (1-
day 99% VaR=2.33).
 The T-day VaR = T  the one-day VaR
 Regulators explicitly state that banks to the 10-day 99%
VaR can be calculated as 10 times the one-day VaR.
 Impact of Autocorrelation
17

 When daily changes in a portfolio are identically

distributed and independent, the variance over T days is T
times the variance over one day.
 When there is autocorrelation equal to r, the multiplier is
increased from T to
T  2(T  1)r  2(T  2)r2  2(T  3)r3  2rT 1
 Impact of Autocorrelation:
Ratio of N-day VaR to 1-day VaR
18

T=1 T=2 T=5 T=10 T=50 T=250

r=0 1.0 1.41 2.24 3.16 7.07 15.81

r = 0.05 1.0 1.45 2.33 3.31 7.43 16.62

r = 0.1 1.0 1.48 2.42 3.46 7.80 17.47

r = 0.2 1.0 1.55 2.62 3.79 8.62 19.35

 Question 3
19

 Suppose that the change in the value of a portfolio over a

one-day time period is normal with a mean of zero and a
standard deviation of $2 million, what is the five-day
97.5% VaR?
 What difference does it make to your previous answer if
there is first-order daily autocorrelation with a correlation
parameter equal to 0.16?
 Choice of VaR Parameters
20

 Time horizon should depend on how quickly portfolio

can be unwound. Bank regulators in effect use 1-day for
market risk and 1-year for credit/operational risk. Fund
managers often use one month.
 Confidence level depends on objectives. Regulators use
99% for market risk and 99.9% for credit/operational
risk.
 A bank wanting to maintain a AA credit rating will often
use confidence levels as high as 99.97% for internal
calculations.
 Additional Measures
21

VaR Measures for a Portfolio where an amount xi is

invested in the ith component of the portfolio

 Marginal VaR: VaR

xi
Incremental VaR: Incremental effect of a new trade or
closing out an existing trade:
VaR
xi
xi
 Component VaR
22

 The component VaR is approximately the same as the

incremental VaR.
 The total VaR is the sum of the component VaR’s (Euler’s
theorem):
N
VaR
VaR   xi
i 1 xi

 The component VaR therefore provides a sensible way of

allocating VaR to different activities
 Back Testing
23

 Back-testing a VaR calculation methodology involves

looking at how often exceptions (loss > VaR) occur.
 Alternatives:
 a) compare VaR with actual change in portfolio value
 b) compare VaR with change in portfolio value assuming
no change in portfolio composition
 Back Testing: one-tailed tests
24

 Suppose theoretical probability of an exception is p (=1−X ).

 The probability of the VaR limit being exceeded on m or
more exceptions in n days is:
n
n!

k  m k!( n  k )!
p k
(1  p ) nk

 If this probability of the VaR limit is less than confidence level

(say 5%), we reject the hypothesis that the probability of an
exception on any given day is larger than p.
 The probability of m or less exceptions in n days is:
m
n!

k 0 k! ( n  k )!
p k
(1  p ) n k
 Back Testing: two-tailed test
25

 Kupiec (1995) has proposed a relatively powerful two-

tailed test.
 If the probability of an exception under the VaR model is
p and m exceptions are observed in n trials, then
 2 ln[( 1  p)nm pm ]  2 ln[( 1  m / n)nm (m / n)m ]
should have a chi-square distribution with one degree of
freedom.
 Given the confidence level is 5%, we reject the null
hypothesis if Kupiec’s two-tailed test is larger than 3.84.
 Example 5
26

 Suppose we back test a VaR model using 600 days of

data when the VaR confidence level is 99%, which means
p=1%. Find the range of the number of exceptions, m, in
which we can not reject the proposed VaR model at 95%
level.

 Answer:
 The value of the Kupiec’s statistic is greater than 3.84
when the number of exceptions m ≤ 1, or m ≥ 11.
Therefore, we accept the VaR model when 2 ≤ m ≤ 11.
 Question 4
27

 Suppose that we back test a VaR model using 1,000 days

of data. The VaR confidence level is 99% and we observe
15 exceptions. Should we reject the model at the 5%
confidence level? Use Kupiec's two-tailed test.
 Bunching
28

 Bunching occurs when exceptions are not evenly spread

throughout the back testing period
 Statistical tests for bunching have been developed by
Christofferson (1998) (“Evaluating Interval Forecasts”
International Economic Review)
 Stress Testing
29

 In addition to requiring that a model for market risk be

back tested, regulators require market risk VaR
calculations be accompanied by a “rigorous and
comprehensive” stress-testing program.
 Stress testing considers how portfolio would perform
under extreme market moves.
 Key Questions:
 How do we generate and evaluate the scenarios?
 What do we do with the results?
 The Scenarios
30

 Stress individual variables:

 Shifting a yield curve by 100 basis points;
 Changing implied volatility for an asset by 20% of current
value;
 Changing an equity index by 10%;

 Changing the major exchange rate by 6%;

 Choose particularly days when there were big market

movements and stress all variables by the amount they
moved on those days
 Form a stress testing committee of senior management
and ask it to generate the scenarios
 What to do with the Results?
31

 Should managers place more reliance on stress testing

results or VaR results?
 One idea is to ask the stress testing committee to assign
probabilities to scenarios (e.g. 0.05% or 0.2% or 0.5%).
 The stress scenarios can then be integrated with the
historical simulation scenarios to produce a composite
VaR.
 Summary
32

 The VaR measure

 Properties and limitation of risk measure
 Expected Shortfall
 Choice of parameters for VaR
 Back Testing and Stress Testing