Risk Management: Introduction
Rangarajan K. Sundaram
Stern School of Business
New York University
TRIUM Global EMBA Program
New York: January 16-20, 2015
Risk & its Management
c
Rangarajan
K. Sundaram
An Overview
This is one of two segments in the module on risk management.
Deals mainly with the instruments for managing market risk and credit
risk, in particular, on
I
I
I
The uses of these instruments.
The risks in these instruments.
The valuation of these instruments.
Professor Ed Altmans sessions focussing on credit risk complement this
material.
Risk & its Management
c
Rangarajan
K. Sundaram
Introduction
What is Risk?
I
Potential that outcomes of an action may differ from those expected
or anticipated.
Ever-present in all economic activity.
I
In normal market conditions.
I
Changes in input prices, exchange rates, interest rates, etc.
Unexpected market dislocations:
I
Financial bubbles, natural disasters, terrorist attacks, political events,
...
Risk & its Management
c
Rangarajan
K. Sundaram
The Management of Risk
From an organizational standpoint, the management of risk requires:
1. Identifying the sources of risks.
2. Where possible, measuring/quantifying these risks.
3. Managing the risks.
I
I
I
Eliminating unnecessary risks.
Transferring risk to markets.
Managing the retained risk.
Risk & its Management
c
Rangarajan
K. Sundaram
The Sources of Risk
Market risk.
I
Difficulty in getting in and out of positions.
Operational risk.
I
Risk that promised payments fail to materialize.
Liquidity risk.
I
Changes in prices in normal market times.
Credit risk.
Lack of proper controls to detect fraudulent activity.
Others:
I
Political, terrorist, catastrophe, reputational, . . .
Risk & its Management
c
Rangarajan
K. Sundaram
Our Focus . . .
. . . is on instruments for managing market and credit risk.
I
The effects of market risk can be exacerbated by the presence of
I
I
As noted, Professor Altmans sessions develop the theme of credit
risk further.
Illiquidity.
Poor operational controls.
The case studies we examine will look at the interplay of these factors.
Risk & its Management
c
Rangarajan
K. Sundaram
2. Measuring Risks
I
Involves specifying a probability distribution over outcomes:
I
I
Set of possible outcomes.
Likelihoods of these outcomes.
What probability distribution should one use?
I
Potential trade-off between using
I
I
distributions that are easy to work with, and
those that fit the data better and/or are more appropriate for the
task at hand.
Common distribution in financial modeling: the Normal or Gaussian.
I
I
Well understood and easy to work with . . .
. . . but, as we discuss below, some important shortcomings from a
risk-management standpoint.
Risk & its Management
c
Rangarajan
K. Sundaram
3. Managing the Risks
Involves:
I
Eliminating unnecessary risks.
I
Transferring risk to markets.
I
I
For example, have minimum creditworthiness standards for advancing
credit.
Derivatives contracts: Futures/forwards, options, swaps, . . .
Hedging versus insurance.
Managing the retained risk.
I
I
Capital to be held against retained risk.
Extent of liquid reserves.
Risk & its Management
c
Rangarajan
K. Sundaram
Potential Problems
Risk-management failures usually because either:
1. Risks are not properly identified.
or
2. Identified risks are inadequately measured.
I
Usually mis-specified and/or excessively optimistic models.
Risk & its Management
c
Rangarajan
K. Sundaram
Risk-Management Failures: Famous Examples
Barings, Sumitomo, Societe Generale.
I
Metallgesellschaft, Aracruz Cellulose.
I
Risk of sharp market moves underestimated.
Amaranth.
I
I
Unidentified operational risk: Rogue trading.
Market and liquidity risks underestimated.
Could not exit positions.
LTCM, AIG, Fannie Mae, Freddie Mac
I
Systemic/correlation risk not captured.
Risk & its Management
10
c
Rangarajan
K. Sundaram
Comment 1: Unmodeled Features
I
Models can only reflect what is put into them.
This can create illusory comfort levels with strategies.
Metallgesellschaft in 1995:
I
I
I
Hedged long-term forward sales with short-term futures.
Cash flow consequences of sharp oil-price drop ignored.
Bankruptcy resulted.
AIG in 2008:
I
I
Built sophisticated models to capture default risk.
Ignored collateral requirements from possible
I
I
Market deterioration short of default.
AIGs own credit-rating deterioration.
Bankruptcy resulted.
Risk & its Management
11
c
Rangarajan
K. Sundaram
Comment 2: Unmodelable Features
Operational risk:
I
I
I
I
Of course, not just a trading/financial markets issue:
I
Nick Leeson and Barings: > $1 billion in losses.
Yasuo Hamanaka and Sumitomo: > $2.5 billion losses.
Jerome Kerviel and Soc Gen: e5 billion losses.
Kweku Adoboli and UBS: $2 billion losses
Enron.
Incentives matter! Where does Harvard Universitys $1 billion+ in
swap-related losses in 2009 fit in?
Risk & its Management
12
c
Rangarajan
K. Sundaram
Comment 3: The Normal Distribution
Key issue: Choosing a distribution that best represents the uncertainty
concerning future prices.
What is meant by best?
The most common starting point: Normal (a.k.a. Gaussian).
Instructive to understand the pros and cons of this choice.
Defined by two parameters:
I
I
The mean : centers the distribution.
The standard deviation : measures dispersion around .
Distribution has the familiar bell-shape (hence bell curve).
Risk & its Management
13
c
Rangarajan
K. Sundaram
The Normal Distribution
Probability
of
observa7on
<
x
Xc
x
-5
-4
Risk & its Management
-3
-2
-1
14
c
Rangarajan
K. Sundaram
The Normal Distribution
The Mean: Centering the Distribution
MEAN
=
0
STD
DEV
=
1
-6
-4
Risk & its Management
MEAN
=
2
STD
DEV
=
1
-2
15
c
Rangarajan
K. Sundaram
The Normal Distribution
The Standard Deviation: Dispersion Around the Mean
MEAN
=
0
STD
DEV
=
1.0
MEAN
=
0
STD
DEV
=
1.50
-6
Risk & its Management
-4
-2
16
c
Rangarajan
K. Sundaram
The Normal Distribution: Properties
Distribution is symmetric around the mean.
Likelihood of an observation depends solely on its distance from the mean
(measured in standard deviations):
Distance from mean
1 standard deviation
1.96 standard deviations
2.58 standard deviation
3 standard deviation
% of all observations
68%
95%
99%
99.73%
Thus, for example, if observations are normally distributed only around 1
in 370 observations should be more than three standard deviations from
the mean.
Risk & its Management
17
c
Rangarajan
K. Sundaram
The Normal Distribution: Properties
One Standard Deviation from the Mean
16%
-5
-4
Risk & its Management
-3
16%
-2
-1
-
s
m
18
1
m+s
c
Rangarajan
K. Sundaram
The Normal Distribution: Properties
1.96 Standard Deviations from the Mean
2.5%
2.5%
-5
-4
Risk & its Management
-3
m
--2
1.96s
0
m
-1
19
m
+
12
.96
s
c
Rangarajan
K. Sundaram
Are Financial Markets Normal ?
Normal distributions have found widespread applications in the physical
and natural sciences.
How well do they fit financial market data?
The main problem: In virtually every financial market, they underestimate
significantly the likelihood of extreme or tail observations.
Example Observations more than 3 standard deviations from the mean
should occur only about 0.27% of the time (roughly once every 370
observations).
I
In practice, they occur far more frequently (see the next several
slides).
Normality is of limited use in estimating tail risk.
Risk & its Management
20
c
Rangarajan
K. Sundaram
S&P 500 Returns: 1950-2015
No
of
Observations:
16,360
Actual
No.
Beyond
Theoretical
No.
Beyond
Actual/Theoretical
Risk & its Management
Period:
3-Jan-1950
to
8-Jan-2015
3
Std
Dev
4
Std
Dev
5
Std
Dev
6
Std
Dev
225
93
44
26
44.2
1.04
0.009
0.00003
5.09
89.74
4,691.2
805,424
21
c
Rangarajan
K. Sundaram
S&P 500 Returns: 1950-2015
No
of
Observations:
16,360
Actual
No.
Beyond
Theoretical
No.
Beyond
Actual/Theoretical
Period:
3-Jan-1950
to
8-Jan-2015
3
Std
Dev
4
Std
Dev
5
Std
Dev
6
Std
Dev
225
93
44
26
44.2
1.04
0.009
0.00003
5.09
89.74
4,691.2
805,424
No
of
Observations:
16,360
Actual
Freq
(Days)
Theoretical
Freq
(Days)
Theoretical/Actual
Period:
3-Jan-1950
to
8-Jan-2015
3
Std
Dev
4
Std
Dev
5
Std
Dev
6
Std
Dev
73
176
372
629
370.4
15,787.2
1,744,278
506,797,317
5.09
89.74
4,691.2
805,424
I Thus, for example, observations 4 standard deviations from the mean occurred
roughly once every seven months, 89.7 times more frequently than predicted by
normality of once every 62.6 years.
Risk & its Management
21
c
Rangarajan
K. Sundaram
6,081
No.
of
Observations:
14,463
Period:
33-Jan-1950
-Jan-1950
to
to
330-Jun-2007
-Dec-2013
Period:
3
Std
Devs
4
Std
Devs
5
Std
Devs
6
Std
Devs
223
41
90
43
26
129
19
10
43.4
0.92
1.02
0.008
0.009
0.00003
39.0
5.14
44.8
88.3
2,291
4,664
350,410
819,397
3.30
S&P 500
Returns: 1950-2007
Actual
No.
Beyond
Theoretical
No.
Beyond
Actual/Theoretical
4,463
Period:
No.
of
Observations:
16,081
Period:
33-Jan-1950
-Jan-1950
tto
o
330-Jun-2007
-Dec-2013
Frequency
(Days)
3
Std
Devs
4
Std
Devs
5
Std
Devs
6
Std
Devs
o.
Beyond(days)
129
178.68
10
Actual
fNrequency
72.11
41
373.98
19
618.50
o.
Beyond(days)
370.4
39.0
15,787.19
0.92
1,744,278
0.008
506,797,332
0.00003
Theoretical
fNrequency
Actual/Theoretical
3.30
88.4
44.8
4,664
2,291
819,397
350,410
Theoretical/Actual
5.14
6,081
No.
of
Observations:
14,463
Period:
3-Jan-1950
to
3-Dec-2013
Frequency
(Days)
3
Std
Devs
4
Std
Devs
5
Std
Devs
6
Std
Devs
618.50
Actual
frequency
(days)
112.12
72.11
352.76
178.68
761.21
373.98
1,446.30
Theoretical
frequency
(days)
370.4
15,787.19
1,744,278
506,797,332
5.14
44.8
88.4
2,291
4,664
350,410
819,397
Theoretical/Actual
3.30
No.
of
before
Observations:
4,463
Period:
deviations
3-Jan-1950
tfrom
o
3-Dec-2013
I Even
2007, 1changes
of more than 4 standard
the mean
Frequency
(Days)
more likely in
3
Sreality
td
Devs than
4
Spredicted
td
Devs
5
Std
Devs
6
Std
Devs
were
45 times
by
normality.
Actual
frequency
(days)
112.12
352.76
761.21
1,446.30
Theoretical
frequency
(days)
370.4
15,787.19
1,744,278
506,797,332
Theoretical/Actual
3.30
44.8
2,291
350,410
Risk & its Management
22
c
Rangarajan
K. Sundaram
USD-EUR Exchange Rates: 2004-2011
No
of
observations:
2085
Actual
Number
Beyond
Theo
Number
Beyond
Actual/Theoretical
No
of
observations:
1042
Actual
Number
Beyond
Theo
Number
Beyond
Actual/Theoretical
No
of
observations:
1043
Actual
Number
Beyond
Theo
Number
Beyond
Actual/Theoretical
Period:
1-Jan-2004
to
31-Dec-2011
3
Std
Dev 4
Std
Dev
5
Std
Dev
24
4
1
5.6
0.13
0.0012
4.3
30.3
836.6
Period:
1-Jan-2004
to
31-Dec-2007
3
Std
Dev 4
Std
Dev
5
Std
Dev
1
0
0
2.8
0.07
0.0006
0.36
0
0
Period:
1-Jan-2008
to
31-Dec-2011
3
Std
Dev 4
Std
Dev
5
Std
Dev
23
4
1
2.8
0.07
0.0006
8.17
60.5
1672.4
I Changes of > 4 standard deviations from the mean are 30 times more likely than
predicted by normality.
I But . . .
Risk & its Management
23
c
Rangarajan
K. Sundaram
Crude Oil Changes: WTI
WTI crude: 2005-2015
No of observations: 2,518
3 Std Dev
43
6.80
6.32
Actual No. Outside
Theoretical No. Outside
Actual/Theoretical
Period: Jan 3, 2005 to Jan 5, 2015
4 Std Dev
5 Std Dev
22
7
0.16
0.001
137.5
4,861
WTI crude: 2012-15
No of observations: 757
3 Std Dev
11
2.04
5.39
Actual No. Outside
Theoretical No. Outside
Actual/Theoretical
Risk & its Management
24
Period: Jan 3, 2012 to Jan 5, 2015
4 Std Dev
5 Std Dev
2
2
0.05
0.0004
41.7
4,608
c
Rangarajan
K. Sundaram
Others
Copper: 2008-13
No
of
Observations:
1515
Actual
Number
Beyond
Theo
Number
Beyond
Actual/Theoretical
Period:
1-Jan-2008
to
31-Dec-2013
3
Std
Dev 4
Std
Dev 5
Std
Dev
17
4
1
4.09
0.1
0.0009
4.2
41.7
1151.3
Brent crude: 2006-2012
No
of
Observations:
756
Period:
1-Jan-2011
to
31-Dec-2013
No
of
Observations:
1798 3
SPeriod:
o
18-Dec-2012
td
Dev1-Jan-2006
4
Std
Dtev
5
Std
Dev
3
Std
Dev3 4
Std
Dev 05
Std
Dev
Actual
Number
Beyond
0
Actual
no
beyond
17
7
1
Theo
Number
Beyond
2.04
0.1
0.0004
Theoretical
no
beyond
4.9
0.114
0.001
Actual/Theoretical
1.5
0.0 970.1 0.0
Actual/Theoretical
3.5
61.5
Risk & its Management
25
c
Rangarajan
K. Sundaram
Non-Normality
The non-normality of equity returns has been documented at least since
1965.
Non-normality is also reflected in implied volatilities obtained from option
prices.
Normality is mainly useful as a benchmark; other distributions may
capture tail-risk better:
I
I
I
I
Students t.
Jump-diffusions.
Cornish-Fisher.
Others.
Risk & its Management
26
c
Rangarajan
K. Sundaram
Comment 4: Historical Behavior
500.00
450.00
Tech:
CSCO
+
INTC
+
MSFT
Fin:
C
+
JPM
+
MS
400.00
350.00
Tech
Stocks
Fin
Stocks
300.00
250.00
200.00
150.00
100.00
50.00
-
1/2/98
Risk & its Management
1/2/01
1/3/04
27
1/3/07
1/3/10
1/3/13
c
Rangarajan
K. Sundaram
The Material to Follow
Derivatives and their role in risk-management:
I
I
I
I
Futures & Forwards.
Options.
Swaps.
Credit derivatives.
Risk & its Management
28
c
Rangarajan
K. Sundaram
Si tacuisses, philosophus mansisses
or: Beware of geeks bearing formulae. (Warren Buffet, 2009)
Alan Greenspan, Federal Reserve Chairman, in 2004:
Not only have individual financial institutions become less
vulnerable to shocks from underlying risk factors, but also the
financial system as a whole has become more resilient.
Joseph Cassano of AIG Financial Products in August 2007:
It is hard for us, without being flippant, to even see a scenario
within any kind of realm of reason that would see us losing $1
in any of those transactions,
Robert Lucas, Nobel Laureate in Economics, in 2003:
The central problem of depression-prevention has been solved.
Risk & its Management
29
c
Rangarajan
K. Sundaram