AD 468

Financial RISK MANAGEMENT

LECTURE 2
MEASURING MARKET RISK
History of Risk and
Return
Mean-Variance Analysis and Modern Portfolio Theory
The Capital Asset Pricing Model (CAPM)
The Arbitrage Pricing Theory (APT)
Efficient Market Hypothesis and Random Walk
How to Price an Option – Black-Scholes Model
Value-at-Risk and Expected Shortfall
Credit Risk and Derivatives

Hulusi Bahcivan SPRING 2025 2

volatility

Hulusi Bahcivan SPRING 2025 3

Definition of Volatility
Suppose that Si is the value of a variable on day i. The volatility per day
is the standard deviation of ln(Si/Si-1)
Normally days when markets are closed are ignored in volatility
calculations
The volatility per year is 252 times the daily volatility
Variance rate is the square of volatility

Hulusi Bahcivan SPRING 2025 4

Standard Approach to
Estimating Volatility
Define σn as the volatility per day between day n-1 and day n, as
estimated at end of day n-1
Define Si as the value of market variable at the end of day i
Define ui= ln(Si/Si-1)
m
1
 n2  
m  1 i 1
( un  i  u ) 2

1 m
u   un  i
m i 1

Hulusi Bahcivan SPRING 2025 5

Simplifications Usually Made
in Risk Management
Define ui as (Si −Si-1)/Si-1
Assume that the mean value of ui is zero
Replace m−1 by m
This gives
1 m 2
  i 1 un  i
2
n
m

Hulusi Bahcivan SPRING 2025 6

Are Daily Changes in Exchange
Rates Normally Distributed?
Real World (%) Normal Model (%)
>1 SD 25.04 31.73

>2SD 5.27 4.55

>3SD 1.34 0.27

>4SD 0.29 0.01

>5SD 0.08 0.00

>6SD 0.03 0.00

Hulusi Bahcivan SPRING 2025 7

Heavy Tails
Daily exchange rate changes are not normally distributed
◦ The distribution has heavier tails than the normal distribution
◦ It is more peaked than the normal distribution

This means that small changes and large changes are more likely than
the normal distribution would suggest
Many market variables have this property, known as excess kurtosis

Hulusi Bahcivan SPRING 2025 8

Normal and Heavy-
Tailed Distribution

Hulusi Bahcivan SPRING 2025 9

Weighting Scheme
Instead of assigning equal weights to the observations we can set

 i 1 i u
2 m 2
n n i

where
m


i 1
i 1

Hulusi Bahcivan SPRING 2025 10

ARCH(m) Model
In an ARCH(m) model we also assign some weight to the long-run
variance rate, VL:

 VL   i 1  i u n2 i
2 m
n

where
m
    i 1
i 1

Hulusi Bahcivan SPRING 2025 11

EWMA Model
In an exponentially weighted moving average model, the weights
assigned to the u2 decline exponentially as we move back through time
This leads to

2 2 2
 
n n 1  (1   )u n 1

Hulusi Bahcivan SPRING 2025 12

Attractions of EWMA
Relatively little data needs to be stored
We need only remember the current estimate of the variance rate and
the most recent observation on the market variable
Tracks volatility changes
0.94 is a popular choice for λ
◦ Risk Metrics model by JP Morgan

Hulusi Bahcivan SPRING 2025 13

GARCH (1,1)
In GARCH (1,1) we assign some weight to the long-run average variance
rate
2 2 2
n  V  u
L n 1   n 1

Since weights must sum to 1

γ+ α + β =1

Hulusi Bahcivan SPRING 2025 14

GARCH (1,1)
Setting ω= γVL the GARCH (1,1) model is
2 2 2
    u
n n 1   n 1

and

VL 
1   

Hulusi Bahcivan SPRING 2025 15

Example
Suppose

 0.000002  013
2
n . u 2
n 1  0.86 2
n 1

The long-run variance rate is 0.0002 so that the long-run volatility per
day is 1.4%

Hulusi Bahcivan SPRING 2025 16

Example continued
Suppose that the current estimate of the volatility is 1.6% per day and
the most recent percentage change in the market variable is 1%.
The new variance rate is
0.000002  013
. 0.0001  0.86 0.000256 0.00023336
The new volatility is 1.53% per day

Hulusi Bahcivan SPRING 2025 17

Variance Targeting
One way of implementing GARCH(1,1) that increases stability is by using
variance targeting
We set the long-run average volatility equal to the sample variance
Only two other parameters then have to be estimated

Hulusi Bahcivan SPRING 2025 18

Value-at-Risk (VaR)

Hulusi Bahcivan SPRING 2025 19

Question Being Asked in
VaR
“What loss level is such that we are X% confident it will not be exceeded
in N business days?”
Motivation: possibility of bankruptcy as the result of an extraordinary
negative event.
VaR intends to be a measure of the probability of such an event.
It is a unique number that summarizes such possibility.
Maximum value that can be lost in a given period of time for a given
probability.

Hulusi Bahcivan SPRING 2025 20

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?”

Hulusi Bahcivan SPRING 2025 21

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 or
− $21.3 million
The VaR for the portfolio with a six month time horizon and a 99%
confidence level is $21.3 million.

Hulusi Bahcivan SPRING 2025 22

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

Hulusi Bahcivan SPRING 2025 23

Example 3
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

Hulusi Bahcivan SPRING 2025 24

Cumulative Loss
Distribution for Example
3

Hulusi Bahcivan SPRING 2025 25

Expected shortfall

Hulusi Bahcivan SPRING 2025 26

VaR vs Expected
Shortfall
VaR is the loss level that will not be exceeded with a specified
probability
Expected Shortfall (or Conditional VaR) is the expected loss given that
the loss is greater than the VaR level
Although expected shortfall is theoretically more appealing, it is VaR
that is used by regulators in setting bank capital requirements

Hulusi Bahcivan SPRING 2025 27

Distributions with the Same VaR
but Different Expected Shortfalls

VaR

VaR

Hulusi Bahcivan SPRING 2025 28

Coherent Risk Measures
Define a coherent risk measure as the amount of cash that has to be
added to a portfolio to make its risk acceptable
Properties of coherent risk measure
◦ 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
◦ Homogeniety: Changing the size of a portfolio by λ should result in the risk
measure being multiplied by λ
◦ 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

Hulusi Bahcivan SPRING 2025 29

VaR vs Expected
Shortfall
VaR satisfies the first three conditions but not the fourth one
Expected shortfall satisfies all four conditions.

Hulusi Bahcivan SPRING 2025 30

Example 4
Each of two independent projects has a probability 0.98 of a loss of $1
million and 0.02 probability of a loss of $10 million
What is the 97.5% VaR for each project?
What is the 97.5% expected shortfall for each project?
What is the 97.5% VaR for the portfolio?
What is the 97.5% expected shortfall for the portfolio?

Hulusi Bahcivan SPRING 2025 31

Calculating VaR

Hulusi Bahcivan SPRING 2025 32

Historical Simulation
Create a database of the daily movements in all market variables.
The first simulation trial assumes that the percentage changes in all
market variables are as on the first day
The second simulation trial assumes that the percentage changes in all
market variables are as on the second day and so on

Hulusi Bahcivan SPRING 2025 33

Historical Simulation
continued
Suppose we use 501 days of historical data (Day 0 to Day 500)
Let vi be the value of a variable on day i
There are 500 simulation trials
The ith trial assumes that the value of the market variable tomorrow is

vi
v500
vi  1

Hulusi Bahcivan SPRING 2025 34

Example : Calculation of 1-day,
99% VaR for a Portfolio on Sept
25, 2008
Index Value ($000s)
DJIA 4,000
FTSE 100 3,000
CAC 40 1,000
Nikkei 225 2,000

Hulusi Bahcivan SPRING 2025 35

Data After Adjusting for
Exchange Rate
Da Date DJIA FTSE 100 CAC 40 Nikkei 225
y
0 Aug 7, 2006 11,219.38 6,026.33 4,345.08 14,023.44

1 Aug 8, 2006 11,173.59 6,007.08 4,347.99 14,300.91

2 Aug 9, 2006 11,076.18 6,055.30 4,413.35 14,467.09

3 Aug 10, 2006 11,124.37 5,964.90 4,333.90 14,413.32

… …… ….. ….. …… ……

499 Sep 24, 2008 10,825.17 5,109.67 4,113.33 12,159.59

500 Sep 25, 2008 11,022.06 5,197.00 4,226.81 12,006.53

Hulusi Bahcivan SPRING 2025 36

Scenarios Generated
Scenar DJIA FTSE CAC 40 Nikkei Portfolio Loss
io 100 225 Value ($000s)
($000s)
1 10,977.08 5,180.40 4,229.64 12,244.10 10,014.334 −14.334
2 10,925.97 5,238.72 4,290.35 12,146.04 10,027.481 −27.481
3 11,070.01 5,118.64 4,150.71 11,961.91 9,946.736 53.264
… ……. ……. ……. …….. ……. ……..
499 10,831.43 5,079.84 4,125.61 12,115.90 9,857.465 142.535
500 11,222.53 5,285.82 4,343.42 11,855.40 10,126.439 −126.439

11,173.59
Example of Calculation: 11,022.06  10,977.08
11,219.38

Hulusi Bahcivan SPRING 2025 37

Ranked Losses
Scenario Number Loss ($000s)
494 477.841
339 345.435 99% one-day
VaR
349 282.204
329 277.041
487 253.385
227 217.974
131 205.256

Hulusi Bahcivan SPRING 2025 38

The N-day VaR
The N-day VaR for market risk is usually assumed to be N times the
one-day VaR
In our example the 10-day VaR would be calculated as
10 253,385 801,274
This assumption is in theory only perfectly correct if daily changes are
normally distributed and independent

Hulusi Bahcivan SPRING 2025 39

The Model-Building
Approach
The main alternative to historical simulation is to make assumptions
about the probability distributions of the return on the market variables
and calculate the probability distribution of the change in the value of
the portfolio analytically
This is known as the model building approach or the variance-
covariance approach

Hulusi Bahcivan SPRING 2025 40

Daily Volatilities
In most other financial applications, we measure volatility “per year”
In VaR calculations we measure volatility “per day”

 year
 day 
252

Theoretically, σday is the standard deviation of the continuously

compounded return in one day
In practice we assume that it is the standard deviation of the
percentage change in one day

Hulusi Bahcivan SPRING 2025 41

Microsoft Example
We have a position worth $10 million in Microsoft shares
The volatility of Microsoft is 2% per day (about 32% per year)
What is the 99% VaR for the next 10 days?

Hulusi Bahcivan SPRING 2025 42

AT&T Example
Consider a position of $5 million in AT&T
The daily volatility of AT&T is 1% (approx 16% per year)
What is the 99% VaR for the next 10 days?

Hulusi Bahcivan SPRING 2025 43

Portfolio Example
Now consider a portfolio consisting of both Microsoft and AT&T
Assume that the returns of AT&T and Microsoft are bivariate normal
Suppose that the correlation between the returns is ρ=0.3.
Find the 10-day 99% VaR of the portfolio
Recall that a standard result in statistics states that

 X Y   2X  Y2  2 X  Y

Hulusi Bahcivan SPRING 2025 44

Comparison of
Approaches
Model building approach assumes normal distributions for market
variables. It tends to give poor results
Historical simulation lets historical data determine distributions, but is
computationally slower

Hulusi Bahcivan SPRING 2025 45

Back-Testing
Tests how well VaR estimates would have performed in the past
We could ask the question: How often was the actual 1-day loss greater
than the 99% 1- day VaR?

Hulusi Bahcivan SPRING 2025 46

Stress Testing
This involves testing how well a portfolio performs under extreme but
plausible market moves
Scenarios can be generated using
◦ Historical data
◦ Analyses carried out by economics group
◦ Senior management

Hulusi Bahcivan SPRING 2025 47