1

MS&E 252
Decision Analysis I
Problem Session 5
What concepts do we expect you to master?
· U-Curves
· When do we need them?
· How do we assess them?
· How do we use them?
· Solving problems using
a u-curve
· PIBP and PISP
revisited
· Delta property
· Delta people
· Risk aversion
2
How much do you value a deal?
· When there is no uncertainty. we do not need u-values
at all...
· . But people place diIIerent values on uncertain
deals.
· People react diIIerently due to many Iactors:
their taste Ior risk
their wealth state
When we study risk preIerence.
we study U-curves.
'A persons risk preference (or risk attitude)
describes the relation between an uncertain deal and
that persons PISP and PIBP for the deal.`
TradeoII between greed and Iear
$150.000
-$75.000
0.5
~
?
Greed
Fear
3
Why do we need U-curves?
· U-curves are a good means Ior us to assess the
values people place on uncertain deals.
· Types oI risk preIerence:

risk-averse risk-neutral risk-seeking
Using U-curves
How do we 'roll back¨ a tree?
.4*.95·.6*.32
Uncertainties:
Take e-value of u-values
÷ 0.63
Decisions:
Pick best u-value
÷ 0.57
Max(.4. .57. .63)
0.4
0.6
S
R
0.95
0.32
u
O
I
0.40
0.63
0.57
P
u
4
Using U-curves Finding a Certain
Equivalent graphically.
Example -
Finding the CE oI a deal with 50° chance oI $80
and 50° chance oI $10.
$100 $0 $80 $10
u($10)
u($80)
.5u($80)¹.5u($10)
÷ u(CE)
CE
Using U-curves Why is it important?
· Can also be used to determine the
Value oI Clairvoyance (iterative
construction).
· Reminds you that you should
calculate the e-value oI the u-values.
not the e-value of the dollar
amounts placed on the prospects.
'When faced with a decision tree. unless the decision maker
is risk-neutral. alwavs remember to go to u-space. then take
the e-value. then go back to $-space.`
5
What concepts do we expect you to master?
· U-Curves
· When do we need them?
· How do we assess them?
· How do we use them?
· Solving problems using
a u-curve
· PIBP and PISP
revisited
· Delta property
· Delta people
· Risk aversion
Let us show how we use
u-values and u-curves in an example!
· DA Corp. would like to invest in research on a new drug. 'Clear Voyance.¨
'Clear Voyance¨ would instantly and painlessly cure any eyesight problem.
"What should DA Corp. do?"
· DA Corp. presently has cash and assets worth $49
million in total; investing in research on 'Clear
Voyance¨ is estimated to cost $13 million.
· They estimate that iI they decide to invest in research.
technical success will only be 20° likely. However.
once they have technical success. they have a 90°
chance oI marketing success.
· Total success would bring $72 million proIit.
Marketing Iailure or technical Iailure will lead to $0
proIit.
· DA Corp.`s u-curve is given by u(x) ÷ \x. with x their total wealth.
6
First step Set up the decision tree!
DA Corp. needs to make a decision without knowing about
technical success and marketing success.
0.9
Marketing Success
0.2 $121
Technical Success
10.5 0.1
No Marketing Success
Invest $36
6.9
0.8
No Technical Success
2 $36
7
$49
Don't Invest
$49
$ M
Second step Go to u-space
U-values
11
6
6
7
Forget about the dollar values for now.
0.9
Marketing Success
0.2 $121
Technical Success
10.5 0.1
No Marketing Success
Invest $36
6.9
0.8
No Technical Success
2 $36
7
$49
Don't Invest
$49
$ M
7
Third step Roll back the tree
0.9
Marketing Success
0.2 $121
Technical Success
10.5 0.1
No Marketing Success
Invest $36
6.9
0.8
No Technical Success
2 $36
7
$49
Don't Invest
$49
U-values
11
6
6
7
10.5
6.9
7
Uncertaintv nodes take the e-value of the u-values,
Decision nodes take the max of the u-values.
$ M
Last step Go back to $-space!
0.9
Marketing Success
0.2 $121
Technical Success
10.5 0.1
No Marketing Success
Invest $36
6.9
0.8
No Technical Success
2 $36
7
$49
Don't Invest
$49
U-values
11
6
6
7
10.5
6.9
7
CE ÷ $49M
Who reallv cares about U-values?
$ M
8
More questions on
'Clear Voyance¨ example
· Now. suppose that DA Corp. can
proceed to preliminary tests that
would give them a better sense oI
how likely technical success is.
What is the most thev should be willing to
spend on these tests?
Decision tree to compute
the Value oI Clairvoyance
And what about JoC on both Technical Success AND Marketing Success?
You would go through
the same steps as
beIore. iteratively Ior
diIIerent values oI x.
The x that gives you a
certain equivalent oI
$49M is the value oI
clairvoyance.
$ M
0.9
Marketing Success
$121 - x
Invest
0.1
0.2 No Marketing Success
"Technical Success" $36 - x
1
Don't Invest
$49 - x
0
$36 - x
Invest
0.8
"No Technical Success" $36 - x
1
Don't Invest
$49 - x
9
What concepts do we expect you to master?
· U-Curves
· When do we need them?
· How do we assess them?
· How do we use them?
· Solving problems using
a u-curve
· PIBP and PISP
revisited
· Delta property
· Delta people
· Risk aversion
How do you Iind your PISP
Ior an uncertain deal?
~
Me
¹ s
Me
¹ Deal
~
s ¹ w
x
1
¹ w
p
1
p
2
p
3
x
2
¹ w
x
3
¹ w
÷
u (s + w) _ p
i
u (x
i
+ w)
PISP ÷ s
10
How do we Iind your PIBP
Ior an uncertain deal?
~
Me
Me ¹ Deal
- b
~
w
x
1
¹ w - b
p
1
p
2
p
3
x
2
¹ w - b
x
3
¹ w - b
÷
u (w) _ p
i
u (x
i
+ w - b)
PIBP ÷ b
Example:
Buying and selling a certiIicate.
50
2 1 ) (
x
x u

− =
75 . 20 $ ) 25 . 0 (
25 . 0 ) (
) 5 . 0 ( 5 . 0 ) (
) 0 ( 5 . 0 ) 50 ( 5 . 0 ) (
1
= =
=
=
+ =

u s
s u
s u
u u s u
) 2 1 ( 5 . 0 ) 2 1 ( 5 . 0 0
) ( 5 . 0 ) 50 ( 5 . 0 ) 0 (
50
0
50
50 b b
b u b u u




− + − =
− + − =
$50
$0
0.5
~
s
Selling
$50 - b
$0 - b
0.5
~
$0
Buying
In this case, s ¬ b. But does this alwavs hold?
11
What concepts do we expect you to master?
· U-Curves
· When do we need them?
· How do we assess them?
· How do we use them?
· Solving problems using
a u-curve
· PIBP and PISP
revisited
· Delta property
· Delta people
· Risk aversion
What does it mean
to Iollow the delta-property?
~
x
1
p
1
p
2
p
3
x
2
x
3
x
~
~
p
1
p
2
p
3
x
1
¹ ∆
x
2
¹ ∆
x
3
¹ ∆
?
Do vou want to follow this rule?
We will add an amount ∆ to all possible outcomes.
What happens to vour PISP?
Q
A
If vou are a ∆-person.
then vour PISP should go up bv ∆'
12
Let us examine this in two examples.
~
$100
.5
.5
$0
$45
~
$200
.5
.5
$100
$145
In some cases. vou naturallv want to follow the delta propertv.
Let us examine this in two examples.
~
$100
.5
.5
$0
$45
. but in other cases. it is far from being obvious'
~
$0
.5
.5
$ -100
$ -55?
13
II you choose to Iollow this property. some
other nice properties will Iollow.
· For a delta-person. PIBP oI a deal ÷
PISP oI that deal.
· Much easier to calculate the value
oI clairvoyance: VOC÷VFC-VNC.
· We can characterize your u-curve
by asking a single question.
· Exponential U-curve.
· CE does not depend on initial
wealth or on other existing deals.
∆ ∆∆ ∆
∆ ∆∆ ∆
1) When you satisIy the delta-property. your
PISP÷PIBP Ior any monetary deal.
~
b
x
1
p
1
p
2
p
3
x
2
x
3
By delta-property.
~Add b to both sides¨
Compare:
s ÷ b
~
s
x
1
p
1
p
2
p
3
x
2
x
3
PISP
~
0
x
1
- b
p
1
p
2
p
3
x
2
- b
x
3
- b
PIBP
14
What types oI u-curves satisIy
the delta-property?
¯
∆ + = ∆ +
i
i i
x u p x u ) ( )
~
(
Note that this is the same as the Iollowing tree:
~
p
1
p
2
p
3
x
1
¹ ∆
x
2
¹ ∆
x
3
¹ ∆
∆ + x
~
The u-function must satisfv the following equation.
What types oI u-curves satisIy
the delta-property?
Onlv two tvpes of u-functions satisfv the delta-propertv:
* The proof is in the manuscript.
rho x x
ge f bc a x u
/
) (

+ = + = Exponential
bx a x u + = ) ( Linear
15
2) Delta-people can easily calculate their
VOC.
· To calculate VOC.
· calculate CE without clairvoyance.
· calculate CE with free clairvoyance.
· take the diIIerence:
VOC ÷ VFC - VNC.
· We used this procedure with Jane; now we
can use it Ior Kim as well.
· Both Kim and Jane are delta-people. so we
can Iind their VOC easily.
3) A delta-person`s u-curve can be
characterized by a single number.
We can assess a delta-person`s u-curve
by asking a single question.
This means we need not ask the
multitude oI questions we asked Kim.
Jane. and Mary. We iust need the risk
odds or the risk tolerance.
You can assess a u-curve without
having a particular problem in Iront oI
you.
16
Why are risk odds important?
Recall that delta people must have a linear (risk
neutral) or exponential u-curve.
· Risk neutral:
· Otherwise:
x x u = ) (
x x
r x u br a x u
− −
+ = ¬ + = 1 ) ( ) (
So. if we know r. we know their u-curve.
How do we Iind risk odds?
Definition:
p
1 - p
$1
-$1
~ $0
II
then
p
p
r

=
1
We can also prove:
p
1 - p
$m
-$m
~ $0
II
then
m
m
p
p
r
p
p
r
1
1 1
|
|
.
|

\
|

= ¬

=
17
Determining Risk Odds r
DeIinition:
p
p
rx

=
1
Risk Attitude Relations
r ÷ 1 Risk Neutral
r ~ 1 Risk Averse
r · 1 Risk Seeking
x
-x
p
1-p
0 ~
r
m
= r
n
m
n Property:
ThereIore. iI we know a person`s risk-odds. we
have characterized their entire u-curve!
What is the decision makers u-curve?
and that the decision maker Iollows the delta-property.
0.7
0.3
$2M
$0
~ $1M
Suppose that we are given:
18
ThereIore. iI we know a person`s risk-odds. we
have characterized their entire u-curve!
3 .
7 .
1 = M r
0.7
0.3
$2M
$0
~ $1M
0.7
0.3
$1M
$ -1M
~ $0
A property
000 . 000 . 1
1
3 .
7 .
|
.
|

\
|
= r ThereIore And
This is enough to determine the entire u-curve.
Now we know several things about a
Delta person`s u-curve.
· We know which u-Iunctions satisIy the delta
property.
· We know how to recognize risk attitude Irom
risk odds.
· We know how to assess someone`s whole u-
curve through their risk odds.
. But how do we go back to $-values once we have the
u-value of the Certain Equivalent?
· We know how to roll back a tree to determine
the u-value Ior an uncertain deal.
19
Inversion oI u-curve
Generic exponential u-curve:
u(x) = a + b ⋅ r
−x
ln
u(x) − a
b






= ln r
−x
[ ]
x =
−1
ln(r)
⋅ ln
u(x) − a
b






Case Study Announcements
· Case study reading Irom Yvonne. assignment
available on web
· Group assignments
· Short meeting with TA on Friday. Nov 19
th
· Due date: Thursday. Dec 2
nd

Master your semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master your semester with Scribd & The New York Times

Cancel anytime.