Professional Documents
Culture Documents
Agenda
`
Risk Management
`
`
`
`
`
`
Risk Management
`
Key observations:
`
Risk
CVaR+
CVaR
CVaRVaR
x
Value-at-Risk
X a loss random variable
VaR ( X ) = min{z | FX ( z ) }
`
`
`
for ]0,1[
Conditional Value-at-Risk
`
CVaR ( X ) =
zdF
X ( z)
where
FX ( z ) = FX ( z )
when z VaR ( X )
1
Conditional Value-at-Risk
CVaR+ (Upper CVaR): expected value of X strictly exceeding VaR
(also called Mean Excess Loss and Expected Shortfall)
( X ) =
FX (VaR ( X ))
1
Conditional Value-at-Risk
`
discontinuous functions
CVaR is convex in X
10
Risk
CVaR+
CVaR
CVaRVaR
x
11
p1 = p 2 = L = p 6 = 16 , =
CVaR = C VaR + =
1
2
f5 +
1
2
f1
Loss
4
6
f6
CVaR
Probability
1
6
2
3
1
6
1
6
1
6
1
6
f2
f3
f4
f5
VaR
CVaR--
1
6
f6
CVaR+
12
Six scenarios,
p1 = p 2 = L = p 6 = 16 , =
1
5
f4 +
f1
f5 +
2
5
f6
CVaR
Probability
1
6
2
5
7
12
1
6
1
6
1
6
f2
f3
f4
1
12
1
6
f5
Loss
VaR
CVaR--
1
2
1
6
f6
f5 + 12 f6 = CVaR+
13
14
Pflug, G.C., Some Remarks on the Value-at-Risk and on the Conditional Value-at-Risk, Probabilistic
Constrained Optimization: Methodology and Applications, (Uryasev ed), Kluwer, 2000
Acerbi, C., Spectral Measures of Risk: a coherent representation of subjective risk aversion, JBF, 2002
15
Payment
Deductible
Premium
Accident lost
PDF
Accident lost
PDF
Deductible
Payment
Premium
Payment
functional
extended sense if:
R1:
R2:
R3:
R4:
`A
functional
is a coherent risk measure in the basic
sense if it satisfies axioms R1, R2, R3, R4 and R5:
R5:
for > 0 (positive homogeneity)
18
A functional
is an averse risk measure in the
extended sense if it satisfies axioms R1, R2, R4 and R6:
R6:
for all nonconstant X (aversity)
A functional
is an averse risk measure in the basic
sense if it satisfies axioms R1, R2, R4, R6 and R5
R2 + R5
(subadditivity)
19
`
`
it is not a
Rockafellar, R.T., Uryasev, S., Zabarankin, M., Optimality conditions in portfolio analysis
with general deviation measures, Mathematical Programming, 2006
24
CounterpartRisk
Measure
where
25
`
`
26
27
CVaR optimization
F ( x, ) = +
1
E{[ f ( x, ) ]+ }
1
Theorem 1:
1.
2.
F ( x, )
is convex w.r.t.
3.
28
CVaR optimization
`
is convex in x
If f(x,) is convex in x then F ( x, ) is convex in x and
`
`
If f(x*,*) minimizes
over X then
is equivalent to
29
CVaR optimization
`
F ( x, ) = + (1 )
k =1
pk [ f ( x, k ) ]+
z + = max { z, 0}
`
1
+
1
p
k =1
k
30
32
is equivalent to
33
min ( E[ X C ]+ + ( 1 1) E[ X C ] ) = CVaR ( X EX )
CR
34
Stability of Estimation
`
`
35
i =1
CVaR ( X ) =
VaR ( X )
z i = E [ X i | X = VaR ( X )] z i
z i
i =1
CVaR ( X )
z i = E [ X i | X VaR ( X )] z i
zi
Estimators of
of
VaR ( X )
zi
CVaR ( X )
zi
36
Assumptions:
Several groups of investors each with
U j ( ER j , Dj ( R j ))
utility function
Rockafellar, R.T., Uryasev, S., Zabarankin, M. Master Funds in Portfolio Analysis with
General Deviation Measures, JBF, 2005
Rockafellar, R.T., Uryasev, S., Zabarankin, M. Equilibrium with Investors using a Diversity of
Deviation Measures, JBF, 2007
37
then
When
then
When
then
41
E i
i =
1 + r0 + i (rM r0 )
1
(E i i (rM r0 ) )
i =
1 + r0
i =
cov(ri , rM )
M2
Generalized CAPM
Consequently:
Each group of investors invests its own Master Fund M
Generalized CAPM
cov ri , QMD
i =
D(rM )
E i
i =
1 + r0 + i (rM r0 )
1
(E i i (rM r0 ) )
i =
1 + r0
D = deviation measure
QMD = risk identifier for D corresponding to M
D( X ) = i CVaRi ( X )
i =1
Q X ( )
D( X ) = CVaR ( X )
QX ( ) = 1{X ( ) VaR ( X )}
X ( )
VaR ( X )
n
D( X ) = i CVaR i ( X )
Q X ( )
i =1
QX ( ) = i 1 X ( ) VaR i ( X )
i =1
1 2 3
+
+
1 2 3
1
+ 2
1 2
1
1
VaR1 ( X )
X ( )
VaR 3 ( X )
VaR 2 ( X )
Data
`
r t=lnSt-lnSt-21
CVaR Deviation
`
D( X ) = j CVaRj ( X )
j =1
D( X ) = CVaR50 % ( X )
Graphs:
35%
30%
25%
20%
15%
10%
5%
0%
91
92
93
MarketIV
94
95
96
97
CAPMpriceIV
98
99 100 101 102 103 104 105 106 107 108 109 110
NoRiskIV
GeneralizedCAPMIValpha=50%
D( X ) = CVaR75 % ( X )
Graphs:
35%
30%
25%
20%
15%
10%
5%
0%
91
92
93
MarketIV
94
95
96
97
CAPMpriceIV
98
99 100 101 102 103 104 105 106 107 108 109 110
NoRiskIV
GeneralizedCAPMIValpha=75%
D( X ) = CVaR85 % ( X )
Graphs:
35%
30%
25%
20%
15%
10%
5%
0%
91
92
93
MarketIV
94
95
96
97
CAPMpriceIV
98
99 100 101 102 103 104 105 106 107 108 109 110
NoRiskIV
GeneralizedCAPMIValpha=85%
Graphs: D( X ) = CVaR95 % ( X )
D( X ) = j CVaRj ( X )
Graphs:
j =1
35%
30%
25%
20%
15%
10%
5%
0%
91
92
93
MarketIV
94
95
96
97
98
CAPMpriceIV
99 100 101 102 103 104 105 106 107 108 109 110
NoRiskIV
CeneralizedCAPMMixed
VaR: Pros
1.
2.
3.
4.
5.
55
VaR: Cons
1.
2.
Risk control using VaR may lead to undesirable results for skewed
distributions
3.
56
CVaR: Pros
1.
2.
3.
4.
5.
6.
57
CVaR: Cons
1.
2.
58
`
`
`
59
four environments:
`
Shell (Windows-dialog)
MATLAB
C++
Run-file
61
matrix of
scenarios
or
covariance
matrix
matrix with
one row
Risk function
(e.g., St.Dev, VaR, CVaR, Mean,)
Linear function
62
PSG: Example
63
Risk functions
Optimization
Sensitivities
Graphics
64
Risk control using VaR may lead to paradoxical results for skewed
distributions
Undesirable feature of VaR optimization: VaR minimization may
increase the extreme losses
P.1
Pb.2
= x11 + ... + xI I
67
(1)
(2)
(3)
(4)
(5)
`
`
`
of the 5 different hedging strategies (e.g., the first raw is for CVaR0.9
hedging strategy)
Each row reports value o different risk functions evaluated at optimal points of the 5
different hedging strategies (e.g., the first raw is for CVaR0.9 hedging strategy)
69
Example:
Chance and VaR constraints equivalence
`
`
`
71
Sharpe ratio for the rebalancing strategy when different risk functions are
used in the objective for different values of the parameter k
72
73
Master Fund Theorem and CAPM can be generalized with the for
different deviation measure.
74
`
`