Combined functionals as risk measures Arcady Novosyolov Institute of computational modeling SB RAS, Krasnoyarsk, Russia, 660036 anov @ icm . krasn . ru http://www.geocities.com/
Jan 12, 2016
Combined functionals as risk measures
Arcady NovosyolovInstitute of computational modeling
SB RAS, Krasnoyarsk, Russia, 660036
[email protected] http://www.geocities.com/
novosyolov/
A. Novosyolov Combined functionals 2
Structure of the presentation
RiskRisk measure
Relations among risk measures
RM: ExpectationRM: Expected utilityRM: Distorted probabilityRM: Combined functional
Anticipated questions
Illustrations
A. Novosyolov Combined functionals 3
Risk
Risk is an almost surely bounded random variable
),,( PBX LXAnother interpretation: risk is a real distribution function with bounded support FF
Correspondence:
,XFX )()( tXtFX P
Why bounded? Back to structure
NextPrevious
A. Novosyolov Combined functionals 4
Example: Finite sample space
Let the sample space be finite:
.|| n Then:
Probability distribution is a vector
),...,,( 21 npppPRandom variable is a vector n
n RxxxX ),...,,( 21
Distribution function is a step function
Back to structure
1
t
)(tFX
1x 2x nx1p
2p np
NextPrevious
A. Novosyolov Combined functionals 5
Risk treatment
Here risk is treated as gain (the more, the better). Examples:
• Return on a financial asset
x 0% 20%
p 0.1 0.9
x -$1,000,00
0
$0
p 0.02 0.98
• Insurable risk
• Profit/loss distribution (in thousand dollars)
x -20 -5 10 30
p 0.01 0.13 0.65 0.21
Back to structure
NextPrevious
A. Novosyolov Combined functionals 6
Risk measure
Risk measure is a real-valued functional
RX:or
.: RFRisk measures allowing both representations with
Back to structure
are called law invariant.
)()( XFX
NextPrevious
A. Novosyolov Combined functionals 7
Using risk measures Certain equivalent of a risk Price of a financial asset, portfolio Insurance premium for a risk Goal function in decision-making
problems
)( XFX
)( XFX
)( XFX
Back to structure
NextPrevious
A. Novosyolov Combined functionals 8
RM: ExpectationXX E)(
)()( ttdFF
Expectation is a very simple law invariant risk measure, describing a risk-neutral behavior. Being almost useless itself, it is important as a basic functional for generalizations.Expected utility risk measure may be treated as a combination of expectation and dollar transform.
Distorted probability risk measure may be treated as a combination of expectation and probability transform.
Back to structure
NextPrevious
Combined functional is essentially the application of both transforms to the expectation.
A. Novosyolov Combined functionals 9
RM: Expected utility
Back to structure
)()( XUXU E
)()()( tdFtUFU
Expected utility is a law invariant risk measure, exhibiting risk averse behavior, when its utility function U is concave (U''(t)<0).
Expected utility is linear with respect to mixture of distributions, a disadvantageous feature, that leads to effects, perceived as paradoxes.
NextPreviousIs EU a certain equivalent?
A. Novosyolov Combined functionals 10
EU as a dollar transform
Back to structure
Value
x1 x2 … xn
Prob p1 p2 … pn
n
kkkU pxUX
1
)()(
:X
n
kkk pxX
1
E
Value
U(x1)
U(x2)
… U(xn)
Prob p1 p2 … pn
:)(XU
NextPrevious
A. Novosyolov Combined functionals 11
EU is linear in probability
Indifference "curves" on a set of probability distributions: parallel straight lines
),()1()())1(( GaFaGaaF UUU ];1,0[a
Back to structure
1p
2p
3p
FGF ,
Expected utility functionalis linear with respect tomixture of distributions.
NextPrevious
A. Novosyolov Combined functionals 12
EU: Rabin's paradox
Back to structure
Consider equiprobable gambles implying loss L or gain G with probability 0.5 each, with initial wealth x. Here 0<L<G. Rabin had discovered the paradox: if an expected utility maximizer rejects such gamble for any initial wealth x, then she would reject similar gambles with some loss L0> L and any gain G0, no matter how large.
Value x-L x+G
prob 0.5 0.5
:),( GLRx
Example: let L = $100, G = $125. Then expected utility maximizer would reject any equiprobable gamble with loss L0= $600. NextPreviou
s
A. Novosyolov Combined functionals 13
RM: Distorted probability
Back to structure
0
0))(1(]1))(1([)( dttFgdttFgX XXg
Distorted probability is a law invariant risk measure, exhibiting risk averse behavior, when its distortion function g satisfies g(v)<v, all v in [0,1].
Distortion function ],1,0[]1,0[: g ,0)0( g 1)1( g
Distorted probability is positive homogeneous, that may lead to improper insurance premium calculation. NextPreviou
s
A. Novosyolov Combined functionals 14
DP as a probability transform
Back to structure
Value
x1 x2 … xn
Prob p1 p2 … pn
XEqxX Q
n
kkkg
1
)(
:X
n
kkk pxX
1
E
Value
x1 x2 … xn
Prob q1 q2 … qn
:X
nkpgpgqn
kii
n
kiik ,...,1,
1
),...,,( 21 nqqqQ NextPreviou
s
A. Novosyolov Combined functionals 15
DP is positively homogeneous
Back to structure
),()( XaaX gg ,0a XX
Consider a portfolio containing a number of "small" risks with loss $1,000 and a few "large" risks with loss $1,000,000 and identical probability of loss. Then DP functional assigns 1000 times larger premium to large risks, which seems intuitively insufficient.
NextPrevious
Distorted probability is a positively homogeneousfunctional, which is an undesired property
in insurance premium calculation.
A. Novosyolov Combined functionals 16
RM: Combined functional
Back to structure
,))(()(1
0
1 dvvFUX XU 1
0
1 )1()()( vdgvFX XgCombined functional involves both dollar and probability transforms:
.)1())(()(1
0
1, vdgvFUX XgU
Discrete case:)()()(
1, XUEqxUX Q
n
kkkgU
Recall expected utility and distorted probability functionals:
NextPrevious
A. Novosyolov Combined functionals 17
CF, risk aversion
Back to structure
Combined functional exhibits risk aversion in a flexible manner: if its distortion function g satisfies risk aversion condition, then its utility function U need not be concave. The latter may be even convex, thus resolving Rabin's paradox. Next slides display an illustration.
NextPrevious
Note that if distortion function g of a combined functional does not satisfy risk aversion condition, then the combined functional fails to exhibit risk aversion. Concave utility function alone cannot provide "enough" risk aversion.
A. Novosyolov Combined functionals 18
CF, example parameters
Back to structure
U(t)
0
1
2
-2 0 2
0),exp(5.0
0),exp(5.0)(
ttt
tttU
33.2)( vvg
g(v)
0.0
0.5
1.0
0.0 0.5 1.0
v
NextPrevious
A. Novosyolov Combined functionals 19
CF: Rabin's paradox resolved
Back to structure
Given the combined functional with parameters from the previous slide (with t measured in hundred dollars), the equiprobable gamble with L = $100, G = $125 is rejected at any initial wealth, and the following equiprobable gambles are acceptable at any wealth level:
L0 G0
$600 $2500
$1000 $4100
$2000 $8100 NextPrevious
A. Novosyolov Combined functionals 20
Relations among risk measures
GeneralizationPartial generalization Back to structur
e
NextPrevious
Legend
A. Novosyolov Combined functionals 21
Legend for relations
Back to structure
NextPrevious
- expectationXE)(XU - expected utility
)(Xg - distorted probability
)(, XgU - combined functional
RDEU – rank-dependent expected utility, Quiggin, 1993
Coherent risk measure – Artzner et al, 1999
A. Novosyolov Combined functionals 22
Illustrations
Back to structure
NextPrevious
Expected utility indifference curves
Distorted probability indifference curvesCombined functional indifference curves
A. Novosyolov Combined functionals 23
EU: indifference curves
Over risks in R2
Over distributions in R3
Back to structure
NextPrevious
A. Novosyolov Combined functionals 24
DP: indifference curves
Back to structure
Over risks in R2
Over distributions in R3NextPreviou
s
A. Novosyolov Combined functionals 25
CF: indifference curves
Back to structure
Over risks in R2
Over distributions in R3NextPreviou
s
A. Novosyolov Combined functionals 26
A few anticipated questions
Why are risks assumed bounded?
NextPreviousBack to structure
Is EU a certain equivalent?
A. Novosyolov Combined functionals 27
Why are risks assumed bounded?
Boundedness assumption is a matter of convenience. Unbounded random variables and distributions with unbounded support may be considered as well, with some additional efforts to overcome technical difficulties.
Back to Risk Back to structure
NextPrevious
A. Novosyolov Combined functionals 28
Is EU a certain equivalent?
Back to EU Back to structure
NextPrevious
Strictly speaking, the value of expected utility functional itself is not a certain equivalent. However, the certain equivalent can be easily obtained by applying the inverse utility function:
X XXUX UU )),(()( 1