Decision Tree Primer

Craig W. Kirkwood
Department of Supply Chain Management
Arizona State University
COPYRIGHT
c 2002 by Craig W. Kirkwood (1/11/2013 – CC license)
Microsoft is a registered trademark of Microsoft Corporation in the United States of
America and other countries.
Effective January 11, 2013, this work is licensed under a Creative Commons Attribution-
NonCommercial 3.0 Unported License.
See http://creativecommons.org/licenses/by-nc/3.0/ for further details.
Contents

1 Decision Trees . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Decision Trees 1
1.2 Expected Value 5
1.3 Dependent Uncertainties 9
1.4 Sequential Decisions 11
1.5 Exercises 14

2 Risk Aversion . . . . . . . . . . . . . . . . . . . . . 19
2.1 Risk Attitude 19
2.2 Utility Functions 20
2.3 The Exponential Utility Function 22
2.4 Assessing the Risk Tolerance 23
2.5 Certainty Equivalent for an Exponential Utility Function 24
2.6 Exercises 26

3 The Value of Information . . . . . . . . . . . . . . . 27

3.1 Calculating the Value of Perfect Information 27
3.2 The Value of Imperfect Information 28
3.3 Flipping a Probability Tree 30
3.4 Calculating “Flipped” Probabilities 33
3.5 Finding the Expected Value of Imperfect Information 34
3.6 Exercises 34

4 Solutions for Exercises . . . . . . . . . . . . . . . . 39

4.1 Chapter 1 Exercise Solutions 39
4.2 Chapter 2 Exercise Solutions 42
4.3 Chapter 3 Exercise Solutions 45
CHAPTER 1

Decision Trees

T he analysis of complex decisions with signi¯ cant uncertainty can be confusing

because 1) the consequence that will result from selecting any speci¯ ed decision
alternative cannot be predicted with certainty, 2) there are often a large number
of di® erent factors that must be taken into account when making the decision,
3) it may be useful to consider the possibility of reducing the uncertainty in the
decision by collecting additional information, and 4) a decision maker's attitude
toward risk taking can impact the relative desirability of di® erent alternatives.
This chapter reviews decision tree analysis procedures for addressing such com-
plexities.

1.1 Decision Trees

To illustrate the analysis approach, a decision tree is used in the following example
to help make a decision.

Example 1.1

Product decision. To absorb some short-term excess production capacity

at its Arizona plant, Special Instrument Products is considering a short manu-
facturing run for either of two new products, a temperature sensor or a pressure
sensor. The market for each product is known if the products can be success-
fully developed. However, there is some chance that it will not be possible to
successfully develop them.
Revenue of $1,000,000 would be realized from selling the temperature sensor
and revenue of $400,000 would be realized from selling the pressure sensor. Both
of these amounts are net of production cost but do not include development cost.
If development is unsuccessful for a product, then there will be no sales, and the
development cost will be totally lost. Development cost would be $100,000 for
the temperature sensor and $10,000 for the pressure sensor.
 2 CHAPTER 1 DECISION TREES

Development Development Sales Net

Cost Outcome Revenue Profit

Success $1,000,000
$900,000
$100,000
Temperature
Sensor Failure $0
-$100,000

Pressure
Success $400,000
Sensor $390,000
$10,000

Failure $0
-$10,000
Neither

$0

Figure 1.1 Special Instrument Products decision

Question 1.1: Which, if either, of these products should Spe-

cial Instrument Products attempt to develop?

To answer Question 1.1 it is useful to represent the decision as shown in Figure

1.1. The tree-like diagram in this ¯ gure is read from left to right. At the left,
indicated with a small square, is the decision to select among the three available
alternatives, which are 1) the temperature sensor, 2) the pressure sensor, or 3) nei-
ther. The development costs for the \develop temperature sensor" and \develop
pressure sensor" alternatives are shown on the branches for those alternatives.
At the right of the development costs are small circles which represent the uncer-
tainty about whether the development outcome will be a success or a failure. The
branches to the right of each circle show the possible development outcomes. On
the branch representing each possible development outcome, the sales revenue is
shown for the alternative, assuming either success or failure for the development.
Finally, the net pro¯ t is shown at the far right of the tree for each possible com-
bination of development alternative and development outcome. For example, the
topmost result of $900,000 is calculated as $1; 000; 000 ¡ $100; 000 = $900; 000.
(Pro¯ ts with negative signs indicate losses.)
The notation used in Figure 1.1 will be discussed in more detail shortly, but
for now concentrate on determining the alternative Special Instrument Products
should select. We can see from Figure 1.1 that developing the temperature sensor
could yield the largest net pro¯ t ($900,000), but it could also yield the largest
loss ($100,000). Developing the pressure sensor could only yield a net pro¯ t
 1.1 DECISION TREES 3

of $390,000, but the possible loss is limited to $10,000. On the other hand,
not developing either of the sensors is risk free in the sense that there is no
possibility of a loss. However, if Special Instrument Products decides not to
attempt to develop one of the sensors, then the company is giving up the potential
opportunity to make either $900,000 or $390,000. Question 1.1 will be answered
in a following example after we discuss the criterion for making such a decision.

You may be thinking that the decision about which alternative is preferred de-
pends on the probabilities that development will be successful for the temperature
or pressure sensors. This is indeed the case, although knowing the probabilities
will not by itself always make the best alternative in a decision immediately clear.
However, if the outcomes are the same for the di® erent alternatives, and only the
probabilities di® er, then probabilities alone are su± cient to determine the best
alternative, as illustrated by Example 1.2.

Example 1.2

Tossing a die. Suppose you are o® ered two alternatives, each of which
consists of a single toss of a fair die. With the ¯ rst alternative, you will win
$10 if any number greater than 4 is thrown, and lose $5 otherwise. With the
second alternative, you will win $10 if any number greater than 3 is thrown, and
lose $5 otherwise. In this case, since there are 6 faces on a die, the probability
of winning is 2=6 = 1=3 for the ¯ rst alternative and 3=6 = 1=2 for the second
alternative. Since the consequences are the same for both alternatives and the
probability of winning is greater for the second alternative, you should select the
second alternative.
However, the possible outcomes are often not the same in realistic business
decisions and this causes additional complexities, as illustrated by further con-
sideration of the Special Instruments Product decision in Example 1.3.

Example 1.3

Product decision. Suppose that in Example 1.1 the probability of devel-

opment success is 0.5 for the temperature sensor and 0.8 for the pressure sensor.
Figure 1.2 is a diagram with these probabilities shown in parentheses on the
branches representing the possible outcomes for each sensor development e® ort.
(While only the probability of success is speci¯ ed for each development e® ort,
the probability of failure is determined by the rules of probability since the prob-
abilities of development success and development failure must add up to one.)
A study of Figure 1.2 shows that having the probabilities does not resolve this
decision for us. The ¯ gure shows that although the probability of development
success is considerably lower for the temperature sensor than it is for the pressure
sensor (0.5 versus 0.8), the net pro¯ t from successful development of the tempera-
ture sensor is considerably higher than the net pro¯ t from successful development
 4 CHAPTER 1 DECISION TREES

Development Development Sales Net

Cost Outcome Revenue Profit

Success $1,000,000
$900,000
(0.5)
$100,000
Temperature
Sensor (0.5) Failure $0
-$100,000

Pressure
Success $400,000
Sensor $390,000
(0.8)
$10,000

(0.2) Failure $0
-$10,000
Neither

$0

Figure 1.2 Special Instrument Products decision tree

of the pressure sensor ($900,000 versus $390,000). That is, the alternative with
the higher potential payo® has a lower probability that this payo® will actually
be realized.
The resolution of this decision dilemma is addressed in the next section, but
before doing this, De¯ nition 1.1 clari¯ es the notation in Figures 1.1 and 1.2.

De¯ nition 1.1: Decision tree notation

A diagram of a decision, as illustrated in Figure 1.2, is called a decision

tree. This diagram is read from left to right. The leftmost node in a
decision tree is called the root node. In Figure 1.2, this is a small
square called a decision node. The branches emanating to the right
from a decision node represent the set of decision alternatives that are
available. One, and only one, of these alternatives can be selected. The
small circles in the tree are called chance nodes. The number shown
in parentheses on each branch of a chance node is the probability that
the outcome shown on that branch will occur at the chance node. The
right end of each path through the tree is called an endpoint, and each
endpoint represents the ¯ nal outcome of following a path from the root
node of the decision tree to that endpoint.
 1.2 EXPECTED VALUE 5

Your Outcome Payment Net

Payment of Toss to You Profit

Success $15
$6
(2/3)
$9
Accept
Offer
(1/3) Failure $0
-$9

Reject
$0
Offer $0

Figure 1.3 Die toss decision tree

1.2 Expected Value

In order to decide which alternative to select in a decision problem, we need

a decision criterion; that is, a rule for making a decision. Expected value is a
criterion for making a decision that takes into account both the possible outcomes
for each decision alternative and the probability that each outcome will occur.
To illustrate the concept of expected value, we consider a simpler decision with
lower stakes than the Special Instrument Products decision.

Example 1.4

Rolling a die. A friend proposes a wager: You will pay her $9.00, and then a
fair die will be rolled. If the die comes up a 3, 4, 5, or 6, then your friend will pay
you $15.00. If the die comes up 1 or 2, she will pay you nothing. Furthermore,
your friend agrees to repeat this game as many times as you wish to play.

Question 1.2: Should you agree to play this game?

If a six-sided die is fair, there is a 1/6 probability that any speci¯ ed side will
come up on a roll. Therefore there is a 4/6 (= 2=3) probability that a 3, 4, 5, or
6 will come up and you will win. Figure 1.3 shows the decision tree for one play
of this game.
At ¯ rst glance, this may not look like a good bet since you can lose $9.00,
while you can only win $6.00. However, the probability of winning the $6.00 is
2/3, while the probability of losing the $9.00 is only 1/3. Perhaps this isn't such
a bad bet after all since the probability of winning is greater than the probability
of losing.
The key to logically analyzing this decision is to realize that your friend will let
you play this game as many times as you want. For example, how often would you
expect to win if you play the game 1,500 times? Based on what you have learned
6 CHAPTER 1 DECISION TREES

about probability, you know that the proportion of games in which you will win
over the long run is approximately equal to the probability of winning a single
game. Thus, out of the 1,500 games, you would expect to win approximately
(2=3) £1; 500 = 1; 000 times. Therefore, over the 1,500 games, you would expect
to win a total of approximately 1; 000 £$6 + 500 £(¡ $9) = $1; 500. So this game
looks like a good deal!
Based on this logic, what is each play of the game worth? Well, if 1,500
plays of the game are worth $1,500, then one play of the game should be worth
$1; 500=1; 500 = $1:00. Putting this another way, you will make an average of
$1.00 each time you play the game.
A little thought about the logic of these calculations shows that you can di-
rectly determine the average payo® from one play of the game by multiplying
each possible payo® from the game by the probability of that payo® , and then
adding up the results. For the die tossing game, this calculation is (2=3) £$6 +
(1=3) £(¡ $9) = $1.
The quantity calculated in the manner illustrated in Example 1.4 is called the
expected value for an alternative, as shown in De¯ nition 1.2. Expected value
is often a good measure of the value of an alternative since over the long run this
is the average amount that you expect to make from selecting the alternative.

De¯ nition 1.2: Expected Value

The expected value for an uncertain alternative is calculated by multi-

plying each possible outcome of the uncertain alternative by its probabil-
ity, and summing the results. The expected value decision criterion
selects the alternative that has the best expected value. In situations
involving pro¯ ts where \more is better," the alternative with the high-
est expected value is best, and in situations involving costs, where \less
is better," the alternative with the lowest expected value is best.

Example 1.5

Product decision. The expected values for the Special Instrument Prod-
ucts decision are designated by \EV" in Figure 1.4. These are determined
as follows: 1) For the temperature sensor alternative, 0:5 £$900; 000 + 0:5 £
(¡ $100; 000) = $400; 000, 2) for the pressure sensor alternative, 0:8 £$390; 000 +
0:2 £(¡ $10; 000) = $310; 000, and 3) for doing neither of these $0. Thus, the
alternative with the highest expected value is developing the temperature sensor,
and if the expected value criterion is applied, then the temperature sensor should
be developed.
 1.2 EXPECTED VALUE 7

Development Development Sales Net

Cost Outcome Revenue Profit

Success $1,000,000
$900,000
(0.5)
$100,000
EV=$400,000
Temperature
Sensor (0.5) Failure $0
-$100,000

EV=$400,000
Pressure Success $400,000
$390,000
Sensor (0.8)
$10,000
EV=$310,000
(0.2) Failure $0
-$10,000
Neither

EV=$0
$0

Figure 1.4 Special Instrument Products decision tree, with expected values

Figure 1.4 illustrates some additional notation that is often used in decision
trees. The branches representing the two alternatives that are less preferred are
shown with crosshatching (//) on their branches. The expected value for each
chance node is designated by \EV". Finally, the expected value at the decision
node on the left is shown as equal to the expected value of the selected alternative.

Xanadu Traders
We conclude this section by analyzing a decision involving international com-
merce. This example will be extended in the remainder of this chapter

Example 1.6

Xanadu Traders. Xanadu Traders, a privately held U.S. metals broker, has
acquired an option to purchase one million kilograms of partially re¯ ned molyzir-
conium ore from the Zeldavian government for $5.00 per kilogram. Molyzirco-
nium can be processed into several di® erent products which are used in semi-
conductor manufacturing, and George Xanadu, the owner of Xanadu Traders,
estimates that he would be able to sell the ore for $8.00 per kilogram after im-
porting it. However, the U.S. government is currently negotiating with Zeldavia
over alleged \dumping" of certain manufactured goods which that country ex-
ports to the United States. As part of these negotiations, the U.S. government
has threatened to ban the import from Zeldavia of a class of materials that in-
cludes molyzirconium. If the U.S. government refuses to issue an import license
8 CHAPTER 1 DECISION TREES

for the molyzirconium after Xanadu has purchased it, then Xanadu will have to
pay a penalty of $1.00 per kilogram to the Zeldavian government to annul the
purchase of the molyzirconium.
Xanadu has used the services of Daniel A. Analyst, a decision analyst, to help
in making decisions of this type in the past, and George Xanadu calls on him to
assist with this analysis. From prior analyses, George Xanadu is well-versed in
decision analysis terminology, and he is able to use decision analysis terms in his
discussion with Analyst.
Analyst: As I understand it, you can buy the one million kilograms of molyzir-
conium ore for $5.00 a kilogram and sell it for $8.00, which gives a pro¯ t of
($8:00 ¡ $5:00) £1; 000; 000 = $3; 000; 000. However, there is some chance that
you cannot obtain an import license, in which case you will have to pay $1.00
per kilogram to annul the purchase contract. In that case, you will not have
to actually take the molyzirconium and pay Zeldavia for it, but you will lose
$1:00 £1; 000; 000 = $1; 000; 000 due to the cost of annulling the contract.
Xanadu: Actually, \some chance" may be an understatement. The internal
politics of Zeldavia make it hard for their government to agree to stop selling
their manufactured goods at very low prices here in the United States. The
chances are only ¯ fty-¯ fty that I will be able to obtain the import license. As you
know, Xanadu Traders is not a very large company. The $1,000,000 loss would
be serious, although certainly not fatal. On the other hand, making $3,000,000
would help the balance sheet: : :

Question 1.3: Which alternative should Xanadu select? As-

sume that Xanadu uses expected value as his decision criterion.

To answer this question, construct a decision tree. There are two possible
alternatives, purchase the molyzirconium or don't purchase it. If the molyzirco-
nium is purchased, then there is uncertainty about whether the import license
will be issued or not. The decision tree is shown in Figure 1.5. Starting from the
root node for this tree, it costs $5 million to purchase the molyzirconium, and if
the import license is issued, then the molyzirconium will be sold for $8 million,
yielding a net pro¯ t of $3 million. On the other hand, if the import license is not
issued then Xanadu will recover $4 million of the $5 million that it invested, but
will lose the other $1 million due to the cost of annulling the contract.
The endpoint net pro¯ ts are shown in millions of dollars, and the expected
value for the \purchase" alternative is 0:5 £$3 + 0:5 £(¡ $1) = $1, in millions
of dollars. Therefore, if expected value is used as the decision criterion, then the
preferred alternative is to purchase the molyzirconium.
 1.3 DEPENDENT UNCERTAINTIES 9

Cost Import License Revenue Net

Issued? Profit
Yes $8
$3
(0.5)
$5
EV=$1
Purchase
(0.5) No $4
-$1

Don't
Purchase $0

Figure 1.5 Xanadu Traders initial decision tree (dollar amounts in millions)

1.3 Dependent Uncertainties

In this section, we consider an additional complexity that often occurs in business

decisions: dependent uncertainties. Speci¯ cally, we will examine a case that
illustrates this complexity of real world decisions, and review a procedure for
analyzing decisions that include dependent uncertainties.

Example 1.7

Xanadu Traders. This is a continuation of Example 1.6. We now consider

an expanded version of the decision that includes dependent uncertainties and
extend the analysis procedure to handle this new issue. We continue to follow the
discussion between Daniel Analyst and George Xanadu that started in Example
1.6.
Analyst: Maybe there is a way to reduce the risk. As I understand it, the
reason you need to make a quick decision is that Zeldavia has also o® ered this
deal to other brokers, and one of them may take it before you do. Is that really
very likely? Perhaps you can apply for the import license and wait until you
know whether it is approved before closing the deal with Zeldavia.
Xanadu: That's not very likely. Some of those brokers are pretty big operators,
and dropping $1,000,000 wouldn't make them lose any sleep. I'd say there is a
0.70 probability that someone else will take Zeldavia's o® er if I wait until the
import license comes through. Of course, it doesn't cost anything to apply for
an import license, so maybe it is worth waiting to see what happens: : :
 10 CHAPTER 1 DECISION TREES

Cost Import License Still Cost Revenue Net

Issued? Available? Profit
Yes $8
$3
(0.5)
$5
EV=$1
(0.5) No $4
-$1
Purchase

EV=$1 Don't
Purchase
$0
Yes $5 $8
(0.3) $3
Wait
Yes
EV=$0.9
(0.5)
$0 (0.7) $0
EV=$0.45 No
(0.5) No
$0

Figure 1.6 Xanadu Traders revised decision tree, with expected values

Question 1.4: Should Xanadu Traders wait to see if an import

license is issued before purchasing the molyzirconium?

The decision tree for this revised problem is shown in Figure 1.6. The two
alternatives at the top of this tree (\purchase" and \don't purchase") are the same
as the alternatives shown in Figure 1.5. The third alternative (\wait") considers
the situation where Xanadu waits to see whether it can obtain an import license
before purchasing the molyzirconium. This alternative introduces a new analysis
issue that must be addressed before the expected value for this alternative can
be determined. This new issue concerns the fact that there are two stages of
uncertainty for this alternative. First, the issue of an import license is resolved,
and then there is a further uncertainty about whether the molyzirconium will
still be available.

Question 1.5: What is the expected value for the \wait" al-
ternative?

The process of determining the expected value for this alternative involves two
stages of calculation. In particular, it is necessary to start at the right side of the
decision tree, and carry out successive calculations working toward the root node
of the tree. Speci¯ cally, ¯ rst determine the expected value for the alternative
assuming that the import license is issued, and then use this result to calculate
 1.4 SEQUENTIAL DECISIONS 11

the expected value for the \wait" alternative prior to learning whether the import
license is issued.
Examine Figure 1.6 to see how this calculation process works. As this ¯ gure
shows, if the import license is issued, then there is a 0.3 probability that the
molyzirconium will still be available. In this case, Xanadu will pay $5 million
for the molyzirconium, and sell it for $8 million realizing $3 million in net pro¯ t.
If the molyzirconium is not still available, then Xanadu will not have to pay
anything and will realize no net pro¯ t. Thus, the expected value for the situation
after the uncertainty about the import license has been resolved is 0:3£$3+0:7£
$0 = $0:9. This expected value is shown next to the lower right chance node on
the decision tree in Figure 1.6.
From the discussion regarding expected value in Section 1.2, it follows that
this $0.9 million is the value of the alternative once the result of the import
license application is known. Hence, this value should be used in the further
expected value calculation needed to determine the overall value of the \wait"
alternative. Thus, the expected value for the \wait" alternative is given by 0:5 £
$0:9 + 0:5 £$0 = $0:45. This expected value is shown next to the lower left
chance node on the decision tree in Figure 1.6. Since the expected value for the
\wait" alternative is less than the $1 million expected value for purchasing the
molyzirconium right now, this alternative is less preferred than purchasing the
molyzirconium right now. Xanadu should not wait, assuming that expected value
is used as the decision criterion.
The process of sequentially determining expected values when there are depen-
dent uncertainties in a decision tree, as demonstrated in Example 1.6, is called
decision tree rollback. This term is de¯ ned in De¯ nition 1.3.

De¯ nition 1.3: Decision Tree Rollback

The process of successively calculating expected values from the end-

points of the decision tree to the root node, as demonstrated in this
section, is called a decision tree rollback.

1.4 Sequential Decisions

In addition to dependent uncertainties, real business decisions often include se-

quential decisions. This section considers an example that demonstrates how
to address sequential decisions.
12 CHAPTER 1 DECISION TREES

Example 1.8

ABC Computer Company. ABC Computer Company is considering sub-

mission of a bid for a government contract to provide 10,000 specialized computers
for use in computer-aided design. There is only one other potential bidder for
this contract, Complex Computers, Inc., and the low bidder will receive the con-
tract. ABC's bidding decision is complicated by the fact that ABC is currently
working on a new process to manufacture the computers. If this process works as
hoped, then it may substantially lower the cost of making the computers. How-
ever, there is some chance that the new process will actually be more expensive
than the current manufacturing process. Unfortunately, ABC will not be able to
determine the cost of the new process without actually using it to manufacture
the computers.
If ABC decides to bid, it will make one of three bids: $9,500 per computer,
$8,500 per computer, or $7,500 per computer. Complex Computers is certain
to bid, and it is equally likely that Complex will bid $10,000, $9,000, or $8,000
per computer. If ABC decides to bid, then it will cost $1,000,000 to prepare the
bid due to the requirement that a prototype computer be included with the bid.
This $1,000,000 will be totally lost regardless of whether ABC wins or loses the
bidding competition.
With ABC's current manufacturing process, it is certain to cost $8,000 per
computer to make each computer. With the proposed new manufacturing process,
there is a 0.25 probability that the manufacturing cost will be $5,000 per computer
and a 0.50 probability that the cost will be $7,500 per computer. Unfortunately,
there is also a 0.25 probability that the cost will be $8,500 per computer.

Question 1.6: Should ABC Computer Company submit a bid,

and if so, what should they bid per computer?

A decision tree for this situation is shown in Figure 1.7. First, ABC must
decide whether to bid and how much to bid. If ABC's bid is lower than Complex
Computer's, then ABC must decide which manufacturing process to use. If ABC
uses the proposed new manufacturing process, then the cost of manufacturing
the computers is uncertain. The net pro¯ t ¯ gures (in millions of dollars) shown
at the endpoints of the Figure 1.7 tree take into account the cost of preparing
the bid, the cost of manufacturing the computers, and the revenue that ABC will
receive for the computers. For example, examine the topmost endpoint value. It
costs $1 million to prepare the bid, and ABC bids $9,500, which is lower than
Complex Computers' bid of $10,000, and hence ABC wins the contract. Then the
proposed new manufacturing process is used, and it costs $8,500 per computer
to manufacture the 10,000 computers. Therefore, at this endpoint, ABC makes a
net pro¯ t of ¡ 1; 000; 000 ¡ 10; 000 £$8; 500 + 10; 000 £$9; 500 = $9; 000; 000 = $9
 1.4 SEQUENTIAL DECISIONS 13

ABC Complex Manufacturing Cost Per Net

Bid Bid Process Computer Profit
$8,500
$9
(1/4)
New (1/2) $7,500
$19
EV=$22.75
$10,000 (1/4) $5,000
$44
(1/3)
EV=$22.75
$9,500 Current $8,000
$14
EV=$6.92
(2/3) $9,000/$8,000
-$1

$8,500
-$1
(1/4)
New (1/2) $7,500
$9
$10,000
EV=$12.75
/$9,000 (1/4) $5,000
$34
(2/3)
EV=$8.17 EV=$12.75
$8,500 Current $8,000
$4
EV=$8.17
(1/3) $8,000
-$1

$8,500
-$11
(1/4)
New (1/2) $7,500
-$1
$10,000/$9,000
EV=$2.75
$7,500 /$8,000 (1/4) $5,000
$24
EV=$2.75
Current $8,000
-$6

No Bid
$0

Figure 1.7 ABC Computer Company decision tree, with net pro¯ t in millions of dollars

million. Verify the net pro¯ ts shown at the other endpoints so that you better
understand this process.
Calculating the expected values shown on the Figure 1.7 decision tree requires
addressing a new issue, namely what to do when there are multiple decision nodes
in the tree. In this decision, the amount of the bid is the ¯ rst decision, and if
this is lower than the Complex Computers bid, then there is a second decision
involving the type of manufacturing process to use. The calculation procedure
for this situation is a straightforward extension of the calculation procedure that
was demonstrated in the preceding section for dependent uncertainties.
This procedure will be illustrated by considering the topmost set of nodes
in the Figure 1.7 tree. Start at the rightmost side of the tree, and calculate
the expected value for the top rightmost chance node. This is determined as
14 CHAPTER 1 DECISION TREES

(1=4) £$9 + (1=2) £$19 + (1=4) £$44 = $22:75. At the top rightmost decision
node, compare the expected values for the two branches. The expected value for
the top branch of this decision node is $22.75, and (since there is no uncertainty
regarding the cost of the current manufacturing process) the expected value for
the bottom branch is $14. Since the top branch has the higher expected value, it
is the preferred branch. That is, the proposed new manufacturing process should
be used. Hence, the expected value for the \manufacturing process" decision
node is equal to the expected value for the proposed new manufacturing process,
which is $22.75.
Now continue back toward the root of the decision tree by calculating the
expected value for the top leftmost chance node in the tree. Since the expected
value of the manufacturing process decision is $22.75, and there is no uncertainty
about the net pro¯ t if ABC loses the bid, the expected value for the top leftmost
chance node is (1=3) £$22:75 + (2=3) £(¡ $1) = $6:92.
A similar process is used to calculate the expected values for the other three
branches of the root node, and the results are shown in Figure 1.7. These cal-
culations show that an $8,500 bid has the highest expected value, which is $8.17
million. Hence, if ABC uses expected value as its decision criterion, then it
should bid $8,500. In addition, the calculations also show that ABC should use
the proposed new manufacturing process if it wins the contract. The less pre-
ferred branches for each decision node have been indicated on the decision tree
with cross hatching.
The complete speci¯ cation of the alternatives that should be selected at all
decision nodes in a decision tree is called a decision strategy.

De¯ nition 1.5: Decision Strategy

The complete speci¯ cation of all the preferred decisions in a sequential

decision problem is called the decision strategy. The decision strategy
shown in Figure 1.7 can be summarized as follows: Bid $8,500, and if
you win the contract use the proposed new manufacturing process.

1.5 Exercises

1.1 Arthrodax Company has been approached by Ranger Sound with a rush order
o® er to purchase 100 units of a customized version of Arthrodax's SoundScreamer
audio mixer at $5,000 per unit, and Arthrodax needs to decide how to respond.
The electronic modi¯ cations of the standard SoundScreamer needed for this cus-
tomized version are straightforward, but there will be a ¯ xed cost of $100,000
to design the modi¯ cations and set up for assembly of the customized Sound-
Screamers, regardless of the number of units produced. It will cost $2,000 per
 1.5 EXERCISES 15

unit to manufacture the circuit boards for the units. Since Arthrodax has some
short term spare manufacturing capacity, the Ranger o® er is potentially attrac-
tive. However, the circuit boards for the customized units will not ¯ t into the
standard SoundScreamer case, and Arthrodax must decide what to do about ac-
quiring cases for the customized units as it decides whether to accept Ranger's
purchase o® er. An appropriate case can be purchased at $500 per case, but
Arthrodax could instead purchase an injection molder to make the cases. It will
cost $20,000 to purchase the molder, and there is a 0.6 probability that it will be
possible to successfully make the cases using the molder. If the molder does not
work, then the purchase price for the molder will be totally lost and Arthrodax
must still purchase the cases at $500 per case. If the molder works, then it will
cost $60 per case to make the cases using the molder. Regardless of which case is
used, the cost of assembling the SoundScreamer circuit boards into the case is $20
per unit. Unfortunately, there is no way to test the molder without purchasing
it. Assume that there is no other use for the molder except to make the cases for
the Ranger order.
(i) Draw a decision tree for Arthrodax's decision about whether to accept the
Ranger o® er and how to acquire the cases for the customized SoundScream-
ers.
(ii) Using expected net pro¯ t as the decision criterion, determine the preferred
course of action for Arthrodax.
1.2 This is a continuation of Exercise 1.1. Assume that all information given in that
exercise is still valid, except as discussed in this exercise. Ranger now tells Arthro-
dax that there is uncertainty about the number of customized SoundScreamers
that will be needed. Speci¯ cally, there is a 0.35 probability that it will need 100
units, and a 0.65 probability that it will need 50 units. If Arthrodax will agree
now to produce either number of units, then Ranger will pay $6,000 per unit if it
ultimately orders 50 units, and will pay $5,000 per unit if it ultimately orders 100
units. The timing is such on this rush order that Arthrodax will have to make a
decision about purchasing the injection molder before it knows how many units
Ranger will take. However, Arthrodax will only need to purchase or manufacture
the number of circuit boards and cases needed for the ¯ nal order of either 50 or
100 units.
(i) Draw a decision tree for Arthrodax's decision about whether to accept the
Ranger o® er and how to acquire the cases for the customized SoundScream-
ers. Note that this is a situation with dependent uncertainties, as discussed
in Section 1.3.
(ii) Using expected net pro¯ t as the decision criterion, determine the preferred
course of action for Arthrodax.
1.3 This is a continuation of Exercise 1.2. Assume that all information given in
that exercise is still valid, except as discussed in this exercise. Assume now that
Arthrodax could delay the decision about purchasing the injection molder until
after it knows how many units Ranger will take.
16 CHAPTER 1 DECISION TREES

(i) Draw a decision tree for Arthrodax's decision about whether to accept the
Ranger o® er and how to acquire the cases for the customized SoundScream-
ers. Note that this is a situation with sequential decisions, as discussed in
Section 1.4.
(ii) Using expected net pro¯ t as the decision criterion, determine the preferred
course of action for Arthrodax.
1.4 Aba Manufacturing has contracted to provide Zyz Electronics with printed circuit
(\PC") boards under the following terms: (1) 100,000 PC boards will be delivered
to Zyz in one month, and (2) Zyz has an option to take delivery of an additional
100,000 boards in three months by giving Aba 30 days notice. Zyz will pay $5.00
for each board that it purchases. Aba manufactures the PC boards using a batch
process, and manufacturing costs are as follows: (1) there is a ¯ xed setup cost
of $250,000 for any manufacturing batch run, regardless of the size of the run,
and (2) there is a marginal manufacturing cost of $2.00 per board regardless of
the size of the batch run. Aba must decide whether to manufacture all 200,000
PC boards now or whether to only manufacture 100,000 now and manufacture
the other 100,000 boards only if Zyz exercises its option to buy those boards. If
Aba manufactures 200,000 now and Zyz does not exercise its option, then the
manufacturing cost of the extra 100,000 boards will be totally lost. Aba believes
there is a 50% chance Zyz will exercise its option to buy the additional 100,000
PC boards.
(i) Explain why it might potentially be more pro¯ table to manufacture all
200,000 boards now.
(ii) Draw a decision tree for the decision that Aba faces.
(iii) Determine the preferred course of action for Aba assuming it uses expected
pro¯ t as its decision criterion.
1.5 Kezo Systems has agreed to supply 500,000 PC FAX systems to Tarja Stores in 90
days at a ¯ xed price. A key component in the FAX systems is a programmable
array logic integrated circuit chip (\PAL chip"), one of which is required in
each FAX system. Kezo has bought these chips in the past from an American
chip manufacturer AM Chips. However, Kezo has been approached by a Korean
manufacturer, KEC Electronics, which is o® ering a lower price on the chips. This
o® er is open for only 10 days, and Kezo must decide whether to buy some or all
of the PAL chips from KEC. Any chips that Kezo does not buy from KEC will be
bought from AM. AM Chips will sell PAL chips to Kezo for $3.00 per chip in any
quantity. KEC will accept orders only in multiples of 250,000 PAL chips, and is
o® ering to sell the chips for $2.00 per chip for 250,000 chips, and for $1.50 per chip
in quantities of 500,000 or more chips. However, the situation is complicated by
a dumping charge that has been ¯ led by AM Chips against KEC. If this charge
is upheld by the U. S. government, then the KEC chips will be subject to an
antidumping tax. This case will not be resolved until after the point in time
when Kezo must make the purchase decision. If Kezo buys the KEC chips, these
will not be shipped until after the antidumping tax would go into e® ect and the
chips would be subject to the tax. Under the terms o® ered by KEC, Kezo would
have to pay any antidumping tax that is imposed. Kezo believes there is a 60%
 1.5 EXERCISES 17

chance the antidumping tax will be imposed. If it is imposed, then it is equally

likely that the tax will be 50%, 100%, or 200% of the sale price for each PAL
chip.
(i) Draw a decision tree for this decision.
(ii) Using expected value as the decision criterion, determine Kezo's preferred
ordering alternative for the PAL chips.
1.6 Intermodular Semiconductor Systems case. The Special Products Divi-
sion of Intermodular Semiconductor Systems has received a Request for Quo-
tation from Allied Intercontinental Corporation for 100 deep sea semiconductor
electrotransponders, a specialized instrument used in testing undersea engineered
structures. While Intermodular Semiconductor Systems has never produced deep
sea electrotransponders, they have manufactured subsurface towed transponders,
and it is clear that they could make an electrotransponder that meets Allied's
speci¯ cations. However, the production cost is uncertain due to their lack of
experience with this particular type of transponder. Furthermore, Allied has also
requested a quotation from the Undersea Systems Division of General Electrode-
vices. Intermodular Semiconductor Systems and General Electrodevices are the
only companies capable of producing the electrotransponders within the time
frame required to meet the construction schedule for Allied's new undersea habi-
tat project.
Mack Reynolds, the Manager of the Special Products Division, must de-
cide whether to bid or not, and if Intermodular Semiconductor Systems does
submit a bid, what the quoted price should be. He has assembled a project team
consisting of Elizabeth Iron from manufacturing and John Traveler from market-
ing to assist with the analysis. Daniel A. Analyst, a consulting decision analyst,
has also been called in to assist with the analysis.
Analyst: For this preliminary analysis, we have agreed to consider only a
small number of di® erent possible bids, production costs, and General Electrode-
vices bids.
Reynolds: That's correct. We will look at possible per-unit bids of $3,000,
$5,000, and $7,000. We will look at possible production costs of $2,000, $4,000,
and $6,000 per unit, and possible per-unit bids by General Electrodevices of
$4,000, $6,000, and $8,000.
Iron: There is quite a bit of uncertainty about the cost of producing the
electrotransponders. I'd say there is a 50% chance we can produce them in a
volume of 100 units at $4,000 per unit. However, that still leaves a 50% chance
that they will either be $2,000 or $6,000 per unit.
Analyst: Is one of these more likely than the other?
Iron: No. It's equally likely to be either $2,000 or $6,000. We don't have
much experience with deep sea transponders. Our experience with subsurface
towed transponders is relevant, but it may take some e® ort to make units that
hold up to the pressure down deep. I'm sure we can do it, but it may be expensive.
Analyst: Could you do some type of cost-plus contract?
18 CHAPTER 1 DECISION TREES

Reynolds: No way! This isn't the defense business. Once we commit, we

have to produce at a ¯ xed price. Allied would take us to court otherwise. They're
tough cookies, but they pay their bills on time.
Iron: I want to emphasize that there is no problem making the elec-
trotransponders and meeting Allied's schedule. The real issue is what type of
material we have to use to take the pressure. We may be able to use molyalu-
minum like we do in the subsurface towed units in which case the cost will be
lower. If we have to go to molyzirconium, then it will be more expensive. Most
likely, we will end up using some of each, which will put the price in the middle.
Analyst: What is General Electrodevices likely to bid?
Traveler: They have more experience than we do with this sort of product.
They have never made deep sea electrotransponders, but they have done a variety
of other deep sea products. I spent some time with Elizabeth discussing their
experience, and also reviewed what they did on a couple of recent bids. I'd say
there is a 50% chance they will bid $6,000 per unit. If not, they are more likely
to bid low than high|there is about a 35% percent chance they will bid $4,000
per unit.
Analyst: So that means there is 15% chance they will bid $8,000.
Traveler: Yes.
Reynolds: Suppose we had a better handle on our production costs.
Would that give us more of an idea what General Electrodevices would bid?
Iron: No. They use graphite-based materials to reinforce their transpon-
ders. The cost structure for that type of production doesn't have any relationship
to our system using moly alloys.
(i) Draw a decision tree for the decision that Reynolds must make.
(ii) Determine the expected values for each of the alternatives, and specify
which alternative Reynolds should select if he uses expected value as a
decision criterion.
CHAPTER 2

Risk Aversion

E xpected value as a criterion for making decisions makes sense provided that
the stakes at risk in the decision are small enough to \play the long run averages."
The range of decisions for which this is true covers many situations of practical
business interest, but sometimes the stakes are high enough that this is not an
appropriate assumption.

2.1 Risk Attitude

The value of a risky alternative to the decision maker may be di® erent than the
expected value of the alternative because of the risk that the alternative poses
of serious losses. The concept of the certainty equivalent is useful for such
situations, as shown in De¯ nition 2.1.

De¯ nition 2.1: Certainty Equivalent

The certainty equivalent for an alternative is the certain amount that

is equally preferred to the alternative. An equivalent term for certainty
equivalent is selling price.

Example 2.1

Certainty equivalent. Suppose that through a previous business deal you

have come into possession of an uncertain alternative that has equal chances
of yielding a pro¯ t of $10,000 or a loss of $5,000. The expected value for this
alternative is 0:5 £ $10; 000 + 0:5 £ (¡$5; 000) = $2; 500. However, suppose that
you decide that you would be willing to sell this alternative for $500 or more.
Then, your certainty equivalent for the alternative is $500.
Using the concept of the certainty equivalent, it is possible to specify di® erent
attitudes toward risk taking, as shown in De¯ nition 2.2.
20 CHAPTER 2 RISK AVERSION

De¯ nition 2.2: Risk Attitude

If your certainty equivalent for alternatives speci¯ ed in terms of pro¯ ts

is less than the expected pro¯ t for an alternative, you are said to be risk
averse with respect to this alternative. If your certainty equivalent is
equal to the expected pro¯ t for the alternative, then you are said to
be risk neutral. Finally, if your certainty equivalent is greater than
the expected pro¯ t for the alternative, you are said to be risk seeking.
These de¯ nitions are reversed for an uncertain alternative speci¯ ed in
terms of losses. That is, you are risk averse if your certainty equiva-
lent is greater than the expected loss and risk seeking if your certainty
equivalent is less than the expected loss.

Based on our earlier discussion in Chapter 1, if you are risk seeking with respect
to the various decisions that you make, then over the long run you will probably
go broke because on average you will not recover as much from the alternatives
as you are willing to pay for them. This is not typical behavior in business, and
therefore we will not consider risk seeking behavior further. (Note that there are
situations where a risk seeking attitude may make sense in business. For example,
suppose your business is in such serious trouble that it is going to go broke anyway
unless you get lucky. You might as well \pray for rain" in such a situation and
go against the odds. However, this is not a typical business situation.)
It is worth noting that the appropriate attitude toward risk taking can depend
on the asset position of the organization taking the risk. A large Fortune 500
company may be able to play the odds and use expected value as its decision
criterion in situations that would pose serious risks to a small \mom and pop"
business.

2.2 Utility Functions

If certainty equivalents can be determined for the alternatives in a decision prob-

lem, then it is straightforward to determine the preferred alternative|simply
select the alternative with the best certainty equivalent. This section discusses
a procedure to determine certainty equivalents for the decision alternatives. The
theory for how to determine certainty equivalents in a defensible manner has been
developed, and we will present a practical procedure for using this theory that is
appropriate for many realistic business decisions. Readers who are interested in
the theory behind this approach should consult a decision analysis textbook.
Certainty equivalents can be determined using a modi¯ cation of the procedure
that we use to determine expected values. This modi¯ cation involves introducing
a new function, called the utility function. A typical utility function is shown in
Figure 2.1. In this ¯ gure, the evaluation measure scale is shown on the horizontal
axis, and the utility for each evaluation measure level is plotted on the vertical
axis. The range of the evaluation measure in this example is from ¡500 to 2,000,
 2.2 UTILITY FUNCTIONS 21

Evaluation Measure
1.0

0.5

Utility
0.0

-0.5

-1.0
-500 0 500 1000 1500 2000

Figure 2.1 Illustrative utility function

and this evaluation measure might, for example, represent the net pro¯ t from
a business decision in thousands of dollars. Note that the exact numbers on
the vertical scale do not have speci¯ c meanings, except that greater numbers
represent more preferred levels of the evaluation measure. For example, if the
evaluation measure is dollars of pro¯ t, then there is greater utility for an amount
of $2,000 than an amount of $1,000.
The idea underlying the approach to calculating certainty equivalents is to ¯ rst
convert the possible outcomes in a decision problem to utilities using the utility
function, and then calculate the expected value of these utilities for each alter-
native using the same procedure that was used to calculate expected values. After
determined these expected utilities for each alternative, then it is straightforward
to determine the certainty equivalent for each alternative using a procedure dis-
cussed later in this section.

De¯ nition 2.3: Utility Function

A utility function translates outcomes into numbers such that the

expected value of the utility numbers can be used to calculate certainty
equivalents for alternatives in a manner that is consistent with a decision
maker's attitude toward risk taking.

Here is an intuitive explanation of why this calculation procedure using ex-

pected utilities makes sense as a way to take risk attitude into account. Examine
the utility function in Figure 2.1. Note that this function drops o® rapidly as the
level of the evaluation measure becomes worse (more negative), while it grows less
rapidly as the value of the evaluation measure becomes better (more positive).
Intuitively, this is saying that the value that we lose from each unit of decrease of
22 CHAPTER 2 RISK AVERSION

the evaluation measure becomes increasingly great as the level become more neg-
ative. Therefore, if we take an expected value of the utilities over the evaluation
measure, alternatives that have a signi¯ cant probability of yielding bad outcomes
will be penalized more heavily in the calculation procedure than if expected value
were used to evaluate the alternatives. Hence, an alternative with a signi¯ cant
chance of yielding bad outcomes will be downrated using a utility function from
what would be true if expected value was used to evaluate alternatives.

2.3 The Exponential Utility Function

To implement the expected utility approach reviewed above, it is necessary to

¯ rst determine a utility function. Both theory and practical experience have
shown that it is often appropriate to use a particular form of utility function
called the exponential. For risk averse decision makers, in decisions involving
pro¯ ts (more of the evaluation measure is better), this function has the form

u(x) = 1 ¡e¡ x=R

; R>0

where u(x) represents the utility function, x is the evaluation measure, R is a

constant called the risk tolerance, and e represents the exponential function. (The
exponential function is often designated by \exp" on a ¯ nancial calculator or in
a spreadsheet program.)
In situations involving costs where less of the evaluation measure is preferred,
the exponential utility function has the form

u(x) = 1 ¡ex=R ; R>0

and in this case larger values of x have lower utilities.

As noted above, the degree of risk aversion that is appropriate can depend on
the asset position of the decision making entity, and R represents the degree of
risk aversion. As R becomes larger, the utility function displays less risk aversion.
(In fact, when R approaches in¯ nity, the decision maker becomes risk neutral.)
The utility function plotted in Figure 2.1 is an exponential utility function with
R = 1; 000.
 2.4 ASSESSING THE RISK TOLERANCE 23

2.4 Assessing the Risk Tolerance

The following procedure can be used to determine the approximate value of R for
a particular decision maker: Ask the decision maker to consider a hypothetical
alternative that has equal chances of yielding a pro¯ t of ro or a loss of ro =2. Then
ask the decision maker to specify the value of ro for which he or she would be
indi® erent between receiving or not receiving the alternative. (Or, put another
way, ask the decision maker to adjust ro until the certainty equivalent for this
hypothetical alternative is just equal to zero.) When the decision maker has
adjusted ro in this way, then R is approximately equal to ro . Note that the
expected value for this hypothetical alternative is EV = 0:5 £ ro ¡0:5 £ (ro =2) =
0:25 £ ro , and therefore as long as ro is greater than zero the decision maker is
specifying a risk averse utility function.
We will now apply the expected utility approach to the Xanadu Traders deci-
sion.

Example 2.2

Xanadu Traders. This is a continuation of the Xanadu Traders decision

in Example 1.8. We continue to follow the conversation between Daniel Analyst
and George Xanadu.
Analyst: I understand from my previous work with you that ¯ nancial risks of
the size involved in this deal would be uncomfortable but would not sink Xanadu
Traders. If you could, you would buy some insurance against the potential loss,
but you are not going to avoid the deal just because of the possible loss.
Xanadu: That's correct.
Analyst: I recall that you told me in the past that you would be just willing to
accept a deal with a ¯ fty-¯ fty chance of making $2,000,000 or losing $1,000,000.
However, if the upside were $2,100,000 and the downside were $1,050,000, you
would not take the deal.
Xanadu: That's correct.

Question 2.1: Taking into account Xanadu's attitude toward

risk taking, what is the preferred alternative among those shown
in Figure 1.6?

To answer this question, it is ¯ rst necessary to determine Xanadu's utility

function. This can be done using the information in the dialog. Using the concept
of the risk tolerance, ro = $2 million when an uncertain alternative with equal
chances of yielding a pro¯ t of ro or a loss of ro =2 has a certainty equivalent of
0. Hence, R is approximately equal to $2 million. Therefore, Xanadu's utility
function is
u(x) = 1 ¡e¡ x=2 ;
 24 CHAPTER 2 RISK AVERSION

Cost Import License Still Cost Revenue Net Utility

Issued? Available? Profit
Yes $8
$3 0.777
(0.5)
$5
EU=0.064 (CE=0.132)
(0.5) No $4
-$1 -0.649
Purchase

Don't
EU=0.117
Purchase EU=0.000 (CE=0)
$0 0.000
(CE=0.247)
Yes $5 $8
(0.3) $3 0.777
Wait
Yes
EU=0.233
(0.5)
$0 (0.7) $0 0.000
EU=0.117 No
(CE=0.249)
(0.5) No
$0 0.000

Figure 2.2 Xanadu Traders Expected Utility Analysis

where x is in millions of dollars. Using a spreadsheet or calculator, it is easy to ¯ nd

the utilities for each of the endpoint values in the Figure 1.6, and these are shown
in Figure 2.2. In this ¯ gure, the utility numbers shown at the right side of the tree
have been calculated using an exponential utility function with R = $2 million.
For example, the topmost utility number is given by u(x) = 1 ¡e¡ 3=2 = 0:777.
Expected utility numbers are calculated in the same manner as expected val-
ues. For example, the expected utility for the topmost chance node is given by
EU = 0:5 £ (0:777) + 0:5 £ (¡0:649) = 0:064. This is the expected utility for
the \purchase" alternative, and in a similar manner the expected utilities can
be found for the \don't purchase" alternative (EU = ¡1:000) and the \wait"
alternative (EU = 0:117).

2.5 Certainty Equivalent for an Exponential Utility Func-

tion

Expected utility numbers do not have a simple intuitive interpretation, but there
is a speci¯ c certainty equivalent corresponding to any speci¯ ed expected utility.
For an exponential utility function involving pro¯ ts, it can be shown that the
certainty equivalent is equal to

CE = ¡R £ ln(1 ¡EU);
 2.5 CERTAINTY EQUIVALENT FOR AN EXPONENTIAL UTILITY FUNCTION 25

where CE is the certainty equivalent, EU is the expected utility, R is the risk

tolerance, and ln is the natural logarithm. Thus, the certainty equivalent for the
\purchase" alternative in Figure 2.2 is given by CE = ¡2 £ ln[1 ¡0:064] = $0:132
million. The certainty equivalents are shown for all three alternatives in Figure
2.2, and larger certainty equivalents are more preferred.
In situations involving costs, where less of an evaluation measure is preferred
to more, then the certainty equivalent is equal to

CE = R £ ln(1 ¡EU)

and alternatives with smaller certainty equivalents are more preferred in this case.
Since a certainty equivalent is the certain amount that is equally preferred
to an alternative, the alternative with the greatest certainty equivalent is most
preferred for situations where more of an evaluation measure is preferred to less.
Therefore, taking Xanadu's risk attitude into account, the \purchase" alternative
is no longer the preferred alternative, as it was with the expected value analysis.
The \wait" alternative is now most preferred since it has a certainty equivalent of
$0.249 million, and the \purchase" alternative is now the second most preferred
alternative with a certainty equivalent of $0.132 million. The \don't purchase"
alternative continues to be least preferred with a certainty equivalent of $0.
Note that expected utilities can be directly used to rank alternatives in a
decision problem. It can be shown that the alternative with the greatest expected
utility will also have the most preferable certainty equivalent. (Note that this is
true regardless of whether you are dealing with costs or pro¯ ts, provided that
you use the appropriate utility function formula given above.) Thus the three
alternatives in Figure 2.2 could have been ranked directly using the expected
utilities without calculating certainty equivalents. However, it is often preferable
to calculate certainty equivalents since these are easier to intuitively interpret.

Example 2.3

Xanadu Traders. This example completes our study of Xanadu Traders.

A comparison of the expected values and certainty equivalents for the three al-
ternatives in Figure 2.2 is shown in Table 2.1. This demonstrates that the three
alternatives have di® ering risks. There is no di® erence between the expected value
and the certainty equivalent for the \don't purchase" alternative since there is
no uncertainty with this alternative. The di® erence between the expected value
and certainty equivalent is greatest for the \purchase" alternative indicating that
it has the largest risk. This risk reduces the value of this alternative enough
for Xanadu that it is no longer the most preferred alternative. The \wait" al-
ternative also has a lower certainty equivalent than its expected value since this
alternative has some risk. However, this risk is substantially lower than the risk
for the \purchase" alternative, and hence this becomes the preferred alternative
when Xanadu's risk attitude is taken into account.
26 CHAPTER 2 RISK AVERSION

Expected Certainty
Alternative Value Equivalent Di® erence
Purchase 1.000 0.132 0.868
Don't Purchase 0.000 0.000 0.000
Wait 0.450 0.249 0.201

Table 2.1 Comparison of expected values and certainty equivalents

2.6 Exercises

2.1 This is a continuation of Exercise 1.4. Assume that all the information in that ex-
ercise still holds, except assume now that Aba has an exponential utility function
with a risk tolerance of $100,000. Determine Aba's preferred course of action.
2.2 This is a continuation of Exercise 1.5. Assume that all the information in that ex-
ercise still holds, except assume now that Kezo has an exponential utility function
with a risk tolerance of $750,000. Determine Kezo's preferred ordering alternative
using this utility function.
2.3 This is a continuation of the preceding exercise. (That is, assume that Kezo
has an exponential utility function with a risk tolerance of $750,000.) In an
e® ort to attract Kezo's order, KEC Electronics has revised its o® er as follows:
At no increase in price, KEC will now provide Kezo with the right to cancel its
entire order for a 10% fee after the outcome of the antidumping suit is known.
However, KEC will not be able to accept any additional orders from Kezo once
the outcome of the suit is known. Thus, for example, if Kezo has agreed to
purchase 250,000 PAL chips from KEC at $2.00 per chip, Kezo can cancel the
order by paying $50,000. This ability to cancel the order is potentially of interest
to Kezo because it knows that AM Chips would be able to supply PAL chips after
the outcome of the antidumping suit is known in time for Kezo to ¯ ll the Tarja
order. However, Kezo knows that AM will increase the price of its chips if an
antidumping tax is imposed. In particular, if a 50% tax is imposed, then AM will
increase its chip price by 15%. If a 100% tax is imposed, then AM will increase
its chip price by 20%. Finally, if a 200% tax is imposed, then AM will increase
its chip price by 25%. Assuming that all other information given in the preceding
exercise is still valid, determine Kezo's preferred alternative for the initial order
of PAL chips as well as what Kezo should do if the antidumping tax is imposed.
CHAPTER 3

The Value of Information

O ften there is an option in a decision to collect additional information, and this

chapter presents procedures for determining when it is worth collecting additional
information.

3.1 Calculating the Value of Perfect Information

We begin by determining the value of perfect information. Perfect informa-

tion removes all uncertainty about the outcomes for the decision alternatives.
While there is rarely an option in real-world business decisions that would actu-
ally remove all uncertainty, the value of perfect information provides an easily
calculated benchmark about the worth of collecting additional information. If
all the available options for collecting information cost more than the value of
perfect information, then these options do not need to be analyzed in further de-
tail. This is because imperfect information cannot be worth more than perfect
information.

Box 3.1: The Value of Perfect Information

No source of information can be worth more than the value of perfect

information.

The following example illustrates how to compute the value of perfect infor-
mation.

Example 3.1

Xanadu Traders. This is a continuation of the Xanadu Traders decision

that was discussed in Example 1.8. (Figure 1.6 shows this decision.) Suppose a
source of perfect information existed that would let Xanadu know if the import
license would be issued.
28 CHAPTER 3 THE VALUE OF INFORMATION

Question 3.1: How much money would it be worth to obtain

perfect information about issuance of the import license?

Figure 3.1 shows a decision tree with this (hypothetical) source of perfect
information. The topmost three branches of the root node for this decision tree
are the same as the corresponding branches in Figure 1.6. The lowest branch
of the root node is the perfect information alternative. At a quick glance, the
perfect information may appear to be similar to the \wait" alternative, since for
both of these alternatives George Xanadu learns whether the license will be issued
before he purchases the molyzirconium. However, with the perfect information
alternative, information is available immediately about whether the license will
be issued. Therefore, with the perfect information alternative, Xanadu does not
run the risk that a competitor will purchase the molyzirconium before he learns
whether the license will be issued.
Since the probability is 0.5 that the license will be issued, this is the probability
that the perfect information source will report that the license will be issued.
After learning this perfect information, Xanadu then can decide whether or not
to purchase the molyzirconium. Of course, if Xanadu learns that the license will
be issued, then he purchases the molyzirconium, and if Xanadu learns that the
license will not be issued, then he does not purchase the molyzirconium.
By the standard calculation procedure, it is determined that the perfect infor-
mation alternative has an expected value of $1.5 million, and this is shown on the
Figure 3.1 decision tree. Since the best alternative without perfect information
(\purchase") has an expected value of $1 million, the value of perfect information
is $1:5 ¡ $1:0 = $0:5 million. Therefore, this places an upper limit on how much
it is worth paying for any information about whether the license will be issued.
It cannot be worth paying more than $0.5 million for such information, since $0.5
million if the value of perfect information.

3.2 The Value of Imperfect Information

The calculation procedure is more complicated for determining the value of im-
perfect information. This procedure is illustrated by the following example.
 3.2 THE VALUE OF IMPERFECT INFORMATION 29

Cost Import License Still Cost Revenue Net

Issued? Available? Profit
Yes $8
$3
(0.5)
$5
EV=$1
(0.5) No $4
-$1
Purchase

Don't
Purchase
$0
Yes $5 $8
(0.3) $3
Yes
EV=$0.9
(0.5)
Wait $0 (0.7) $0
EV=$0.45 No
(0.5) No
$0
$5 $8
Perfect $3
Information
EV=$3 Purchase
Yes
Don't
(0.5) Purchase
$0
EV=$1.50
$5 $4
-$1
(0.5)
EV=$0 Purchase
No
Don't
Purchase
$0

Figure 3.1 Xanadu Traders decision tree, with perfect information alternative

Example 3.2

Xanadu Traders. Now consider a potential source of imperfect information

in the Xanadu Traders case last discussed in Example 3.1. We continue with the
discussion between Daniel Analyst and George Xanadu.
Analyst: Is there any way of obtaining additional information about the
chances of obtaining a license other than waiting and seeing what happens?
Perhaps there is something that doesn't take as long as waiting for the import
approval.
Xanadu: Well, there's always John S. Lofton. He is a Washington-based
business consultant with good connections in the import licensing bureaucracy.
For a fee, he will consult his contacts and see if they think the license will be
30 CHAPTER 3 THE VALUE OF INFORMATION

granted. Of course, his assessment that the license will come through is no
guarantee. If somebody in Congress starts screaming, they might shut down
imports from Zeldavia. They are really upset about this in the Industrial Belt,
and Congress is starting to take some heat. On the other hand, even if Lofton
thinks the license won't come through, he might be wrong. He has a pretty good
record on calling these things, but not perfect. And he charges a lot for making
a few telephone calls.
Analyst: How good has he been?
Xanadu: He's done some assessments for me, as well as other people I know.
I'd say in cases where the import license was ultimately granted, he called it right
90% of the time. However, he hasn't been so good on the license requests that
were turned down. In those cases, he only called it right 60% of the time.
Analyst: You commented earlier that he was expensive. How much would he
charge?
Xanadu: This is a pretty standard job for him. His fee for this type of service
is $10,000.

Question 3.2: Should Xanadu hire Lofton, and if so, what is

the maximum amount that he should pay Lofton for his ser-
vices?

We know from our earlier analysis of the value of perfect information in Exam-
ple 3.1 that the maximum amount that it could possibly be worth to purchases
Lofton's services is $0.5 million. Since he would only charge $10,000 it is possible
that it would be worth purchasing his services. However, it is clear from the
discussion above that Lofton often makes mistakes, and perhaps Xanadu would
not learn enough to warrant paying Lofton the $10,000.
A partial decision tree for the \Hire Lofton" alternative is shown in Figure 3.2.
To simplify this tree, the possibility of hiring Lofton and then still waiting to see
if the import license is issued has been eliminated from the tree. In this tree, each
of the two subtrees starting from the decision nodes after the outcome of \predict
import license issued?" has the same structure. Each of these subtrees also has
the same structure as the top two branches of the decision tree in Figure 1.6.
However, as we will see below, the probabilities on the \import license issued?"
branches di® er in Figure 3.2 from those in the Figure 1.6 tree.

3.3 Flipping a Probability Tree

In order to complete the analysis of the alternative in Example 3.2, we need

the probabilities for the two branches labeled \predict import license issued?" in
Figure 3.2. Additionally, we need the probabilities for the two sets of branches
under the label \import license issued?" Unfortunately, as often happens in real
problems, the information presented about Lofton's accuracy in his predictions is
 3.3 FLIPPING A PROBABILITY TREE 31

Consultant Predict Purchase Import License Revenue Net

Cost Import License Cost Issued? Profit
Issued?
Yes $8
$2.99
$5
$4
Purchase -$1.01
No
Yes
Don't
Purchase
-$0.01

$0.01
Yes $8
$2.99
$5
$4
Purchase -$1.01
No
No
Don't
Purchase
-$0.01

Figure 3.2 Hire consultant alternative

not in a form that directly provides the required probabilities. Figure 3.3 shows
in probability tree form the information that is given above about the accuracy
of Lofton. The root node on the left side of the tree shows the probabilities for
\import license issued?" speci¯ ed in earlier discussions of this decision problem.
The two chance nodes on the right side of the tree show the probabilities that
Lofton will call the licensing decision right, based on the conversation presented
in Example 3.2 between Daniel Analyst and George Xanadu.
Comparing Figure 3.2 with Figure 3.3, shows that the probabilities in Figure
3.3 are \backwards" from what is needed to assign probabilities to the branches
of the chance nodes in Figure 3.2. That is, the probability of license approval
is known, as well as the probability of Lofton's di® erent predictions, given the
actual situation regarding license approval. However, the decision tree in Figure
3.2 requires the probability of Lofton's di® erent predictions and the probability
of license approval given Lofton's predictions. This is shown in Figure 3.4, where
the probabilities marked A, B, C, D, E, and F are required. If these probabilities
were known, then they could be inserted into the Figure 3.2 decision tree, and
the expected value could be determined for the alternative of hiring Lofton.
This may seem like an odd way to present the information about Lofton's
accuracy, but information about the accuracy of an information source is often
available in the form of Figure 3.3 when there is a historical record about the
accuracy of the source. As an example, suppose that a new test instrument has
32 CHAPTER 3 THE VALUE OF INFORMATION

Import License Predict Path

Issued? Import License Probability
Issued?

Yes
(0.9) 0.45
Yes
(0.5) No
(0.1) 0.05

Yes
(0.4) 0.20
(0.5)
No
No
(0.6) 0.30

Figure 3.3 Accuracy of consultant

been developed to use in testing for defects in the parts that are manufactured on
a production line. How would the accuracy of the test instrument be determined?
Probably by using the instrument on a series of parts that have previously been
tested by other methods. Thus, it would be known whether the parts that are
being tested are good or bad, and hence it would be possible to determine what
fraction of good parts the test instrument correctly identi¯ es as good, and what
fraction of bad parts the test instrument correctly identi¯ es as bad. This is
analogous to the way that the information is presented for Lofton in Figure 3.3.
In a similar manner, the accuracy of a proposed medical diagnostic procedure
for some medical condition is often determined by applying the diagnostic pro-
cedure to patients who are known to either have the condition or not have the
condition. Information from such tests would be in the form of Figure 3.3. Thus,
the form of the information shown in Figure 3.3 is common, and we need to know
how to use such information when analyzing the value of a potential information
source.
To proceed with the analysis of the alternative of hiring Lofton, we need to
\°ip" the probabilities from the tree in Figure 3.3 to determine the probabilities
needed in Figure 3.4.

De¯ nition 3.1: Tree ° ipping

Tree ° ipping is the process of calculating the probabilities for a prob-

ability tree with the order of the chance nodes reversed, as illustrated
by Figures 3.3 and 3.4.
 3.4 CALCULATING \FLIPPED" PROBABILITIES 33

Predict Import License Path

Import License Issued? Probability
Issued?
Yes
C 0.45
Yes
A No
D 0.20

Yes
E 0.05
B
No
No 0.30
F

Figure 3.4 Probabilities needed for the decision tree

3.4 Calculating \Flipped" Probabilities

It is straightforward to determine the probabilities in Figure 3.4. The key to

doing this is to recognize that the paths from the root node to the endpoints
are the same in the Figure 3.3 and Figure 3.4 trees, but they are arranged in a
di® erent order. The probabilities for these paths can be determined in Figure 3.3
by following the multiplication rule for probabilities. Namely, the probabilities on
the branches along a path are multiplied to determine the probability of following
that path. For example, the probability of following the topmost path in Figure
3.3 is determined as 0:5 £0:9 = 0:45.

De¯ nition 3.2: Path probability

A path probability is the probability of a particular sequence of

branches from the root node to a speci¯ ed endpoint in a probability
tree. A path probability is determined by multiplying the probabilities
on the branches included in the path.

Once the probabilities are determined for each path in Figure 3.3, they can be
transferred to Figure 3.4, as shown at the right side of Figure 3.4. (The topmost
and bottommost probabilities are transferred directly from the Figure 3.3 tree
to the Figure 3.4 tree, and the other two path probabilities need to be reversed
when they are transferred.)
Once the path probabilities are known, probabilities A and B can be deter-
mined. Probability A is the probability of a \yes" prediction regarding license
approval, and this occurs only on the two topmost paths in the Figure 3.4 tree.
Therefore, probability A is equal to the sum of the probabilities for the two
topmost paths. That is, A = 0:45 + 0:20 = 0:65. Similarly, probability B is
34 CHAPTER 3 THE VALUE OF INFORMATION

Predict Import License Path

Import License Issued? Probability
Issued?
Yes
(0.69) 0.45
Yes
(0.65) No
(0.31) 0.20

Yes
(0.14) 0.05
(0.35)
No
No 0.30
(0.86)

Figure 3.5 Decision tree probabilities

equal to the sum of the probabilities for the two bottommost paths. That is,
B = 0:05 + 0:30 = 0:35.
Once A and B are known, then C, D, E, and F can be determined using the
multiplication rule. Thus, A £C = 0:45, or C = 0:45=A = 0:45=0:65 = 0:69
(rounded). Similarly, D = 0:20=A = 0:20=0:65 = 0:31 (rounded), E = 0:05=B =
0:05=0:35 = 0:14 (rounded), and F = 0:30=B = 0:30=0:35 = 0:86 (rounded).
Figure 3.5 shows the probabilities ¯ lled in for the Figure 3.4 probability tree.

3.5 Finding the Expected Value of Imperfect Information

The probabilities in Figure 3.5 can now be transferred to the tree diagram in
Figure 3.2, and the expected value can be calculated for the alternative of hir-
ing Lofton by using the same process as in earlier decision trees. The result is
shown in Figure 3.6, where the expected value for this alternative is $1.13 mil-
lion. Figure 1.6 shows that the best alternative without hiring Lofton only has
an expected value of $1 million, and so it is worth hiring Lofton. In fact, it is
worth considerably more than $10,000 to hire Lofton, since the alternative with
hiring him for $10,000 is worth $1.13 million. In fact, it is worth it to hire Lofton
as long as he costs less than $130; 000 + $10; 000 = $140; 000.

3.6 Exercises

3.1 This is a continuation of Exercise 1.4. Assume that all the information presented
in that exercise still holds. Determine the expected value of perfect information
about whether Zyz will exercise its option.
 3.6 EXERCISES 35

Consultant Predict Purchase Import License Revenue Net

Cost Import License Cost Issued? Profit
Issued?
Yes $8
(0.69) $2.99
$5
EV=$1.75
$4
(0.31) -$1.01
Purchase No
Yes
Don't
EV=$1.75
Purchase
(0.65)
-$0.01

$0.01
EV=$1.13
Yes $8
(0.14) $2.99
$5
EV=-$0.45
(0.35)
$4
(0.86) -$1.01
Purchase No
No
Don't
EV=-$0.01
Purchase
-$0.01

Figure 3.6 Hire consultant alternative, with expected value calculation

3.2 For the decision in the preceding exercise, Aba Manufacturing has created a new
option: It can conduct some research and development in an attempt to lower the
¯ xed setup cost associated with manufacturing a batch of the PC boards. This
research and development would not be completed in time to in°uence the setup
cost for the initial batch that Zyz has ordered, but would be completed before
the second batch would have to be manufactured. The research and development
will cost $25,000, and there is a 0.4 probability that it will be successful. If it
is successful, then the ¯ xed setup cost per batch will be reduced by $200,000 to
$50,000. If the research and development is not successful, then there will be no
reduction in the setup cost. There will be no other bene¯ ts from the research and
development besides the potential reduction in setup cost for the Zyz reorder.
(i) Using expected pro¯ t as the decision criterion, determine whether Aba
should undertake the research and development.
(ii) Using expected pro¯ t as the decision criteria, determine the value of learn-
ing for certain whether the research and development will be successful
before a decision has to be made about whether to initially manufacture
100,000 or 200,000 PC boards.
36 CHAPTER 3 THE VALUE OF INFORMATION

3.3 This is a continuation of Exercise 1.5. Assume that all the information presented
in that exercise still holds. Using expected value as the decision criterion, deter-
mine the maximum amount that Kezo should pay for information about whether
the antidumping tax will be imposed if this information can be obtained prior to
making the ordering decision.
3.4 A college athletic department is considering a mandatory drug testing policy for
all its athletes. Suppose that the test to be used will give either a \positive" or a
\negative" indication. From previous testing it is known that if the tested person
is a drug user there is a 0.92 probability that the test will be \positive." In cases
where the tested person is not a drug user, there is a 0.96 probability that the
test will be \negative." Assume that 10% of the athletes to be tested are drug
users.
(i) Determine the probability that a randomly selected athlete will test positive
for drug use.
(ii) Assuming that a randomly selected athlete tests positive, determine the
probability that he or she is actually a drug user.
(iii) Assuming that a randomly selected athlete tests negative, determine the
probability that he or she is actually a drug user.
(iv) In light of the results above, discuss the potential advantages and disad-
vantages of introducing a mandatory drug testing program using this test.
3.5 Intermodular Semiconductor Systems, Part 2|The Value of Infor-
mation. This is a continuation of the case in Exercise 1.6. Assume that all
information presented in that exercise still holds.
Analyst: Would it be possible to get a better handle on production costs
before making the bid?
Iron: As I said earlier, the main issue is what it will cost to reinforce
the electrotransponders to take the pressure. We could make up some material
samples and borrow the high pressure chamber over in the Submersible Systems
Division to do some tests. We'd get some information out of that, but there
would still be a lot of uncertainty. Also, it would be expensive|I would have to
put people on overtime to meet the bid schedule.
The main problem is that we don't have time to do very extensive testing
before the bid is due. We could make up a rack of samples from materials we
have in stock and take some measurements under pressure, but these materials
aren't exactly the same as what we would use in the actual electrotransponders.
Because of this, we would still not know for sure what we will have to do to make
the electrotransponders work.
[This option was discussed at some length. Following this discussion Ana-
lyst summarizes as follows.]
Analyst: As I understand it, the result of doing material tests would be
an indication that the production will either be \expensive" or \inexpensive." If
molyaluminum is going to work, it is more likely that you will get an \inexpensive"
result while if you have to use molyzirconium you are more likely to get an
\expensive" result.
 3.6 EXERCISES 37

Iron: Yes. In previous cases when we have done tests like this and molyalu-
minum ultimately worked, then 80% of the time we had gotten an \inexpensive"
indication. On the other hand, when it has worked out that we needed molyzir-
conium, then 90% of the time we had gotten an \expensive" indication.
Analyst: What about if a mixture worked?
Iron: We haven't gotten very much useful information in those cases. In
cases where a mixture has worked, 60% of the time we had gotten an \inexpen-
sive" indication and 40% of the time it came out \expensive."
Analyst: Based on our earlier discussion, I understand that if molyalu-
minum works the production costs will be $2,000 per unit, if molyzirconium is
needed the costs will be $6,000 per unit, and if a mixture works the costs will be
$4,000.
Iron: That's correct for the 100-unit quantity we are discussing here.
Reynolds: How much would the material tests cost?
Iron: There will be a lot of hand labor. I'll go talk with my people and get
a ¯ gure back to you in a couple of hours.
[Iron leaves the meeting and later reports that it would cost $7,000 to
conduct the material tests.]
(i) Determine the expected value of perfect information about what material
must be used.
(ii) Determine whether it is worth doing the experiment that is outlined above.