TOPIC SEVEN: DECISION THEORY

Introduction
Decision refers to making choices among future uncertain alternatives. From this definition it is
clear that all decision making relates to the future and that decision is a choice between
alternatives in pursuit of an objective. The objective of the firm may be to maximize profits,
maximize sales, or to minimize costs. Where no alternative exists, the manager has no choice,
and therefore, no decision can be made and nothing can be done now that will alter the past.

Decision making is an all-pervasive activity that takes place at every level in an organization
covering both the short term and long-term period. This is because plans are activated by
decisions that require financial and quantitative analysis in order for a rational choice to be made.
Management often relies heavily on producing relevant information for decision making
purposes in various business situations ranging from long term strategic decisions like
acquisitions, launching new products or investing in plants and machinery, and shorter-term
decisions like product planning, make or buy, product pricing, discontinuing product units, etc.
All these decisions rely on data which have been accurately collected, organized, analyzed and
presented in suitable form to permit accurate interpretation by the manager. In most cases,
management decision making is based on quantitative data, but in some cases qualitative data
also forms the basis of decision undertaken by management. For example, consider a manager
who decided to fire the receptionist because she talked arrogantly to the customer. In some cases,
the decision to recruit is also based on qualitative characteristics of the job applicants. The
knowledge of statistics is therefore very important for management decision making purposes.

Steps in Decision Making

The overall decision-making process is sub divided into the following steps;

 Defining the objectives. The decision to be under taken must be in line with the
organization’s objective(s). The decision maker may be pursuing a single objective such
as profit maximization, or pursuing multiple objectives. For the latter case, Terry Lucey
(2003) proposed the need for consistency among these multiple objectives. For example,
a manager cannot pursue cost minimization objective at the same time with the objective
of improving product quality. The objective is always expressible in quantitative term,
but qualitative objectives are also possible. There must be a link between plans,
objectives and decisions since plans are embodiment of the organization’s objectives and
decisions are the implementation of the plans. The organization objectives must be
quantified (objective function) to allow a decision to be taken. The objectives may be
limited by a constraint like shortages of labor, materials, space, machine capacity, finance
or sales demand.
 Considering an alternative. Recall that decision making relates to the future and the
future is uncertain. Decision making always involves predictions when selecting a set of
alternatives from which a final decision will be made. Moreover, management must at
 their disposal, have adequate data about the set of possible alternatives that will be
subjected to scrutiny. These data can be got from both internal and external sources,
quantitative and qualitative, and secondary or primary data. Attractive decisions do not
automatically submit themselves to the decision process, therefore, they must be
subjected to thorough analysis. For the organization to gather enough data for each
possible alternative, the organization must develop effective information systems which
gather information from all sources in order to ensure that opportunities are not
overlooked. In the next chapter, we shall discuss various ways of collecting, organizing,
and classifying statistical data for decision making purposes. In all decision-making
problems, management must be confronted by more than one alternative, even if
management think that they have no choice but to pursue an alternative, say A, there is
always another alternative B, which presents itself in disguise and management always
ignore. That alternative is an option of doing nothing instead of pursuing alternative A
which is the only one alternative for management. The organization may choose the
option of doing nothing now, in order to free resources which will be used to undertake a
more attractive alternative in the future. As a general principle, whenever you do
something and end up getting a worst result (losses), better to choose to do nothing so
that you earn zero profit.
 Evaluating the alternatives. This step involves quantitative comparisons between the
alternatives to provide relevant and correctly specified objective basis for the ultimate
decision. The manager must be familiar with a range of techniques for evaluating
possible alternatives in order for the right decision to be taken. Some of these techniques
such as linear programming, optimization, investment appraisal techniques etc are outside
the scope of this book. Statistical techniques recommended to be applied at this stage, and
discussed at length in this book involve the measures of variability (for assessing the risk
associated with each alternative), Measures of central tendency to estimate the average
outcome for each alternative, probability theory and theory of expectation since we are
making decision under the condition of uncertainty, sampling theory and statistical
inference since we may be evaluating alternatives based on sample information due to
data limitation. Evaluation of alternatives involves the assessment of risk and uncertainty
since decision making is basically concerned with the future where uncertainty is ever
present. Risk may arise from action of competitor, inflation, interest and finance changes,
new government legislation, materials/labor shortages, etc.
 Selecting a course of action. This is the last step in decision making where actual choice
between alternatives is made. According to Lucey (2003), decisions made may be
programmed or non-programmed. Programmed decisions are decision made within a
clearly defined operational area. Decision rules are known and it is a clear cut single
objective e.g. a replenishment decision based on usage and re-order levels in an inventory
control system. Non-programmed decisions are decisions for which decision rules and
procedures cannot or have not yet been devised because they require non-repetitive
 circumstances. Decision making often depends on the individual’s attitude to risk. Thus
individual attitudes may range from risk seeking to risk aversion
Key Elements of Decision
In order to analyze a decision making problem, we need to understand four important elements
associated with decision making. These elements are discussed below.

 Actions: These are the choices available to a decision maker, it is also called alternatives.
For example, an unemployed university graduate comes across a job advert and she has
the choice of either applying for this job, or not applying for the job. Therefore, there are
two possible actions here. However, in some complex decision making situations, there
can be very many possible alternatives or actions available for a decision maker.
 The states of nature: These are uncertain elements of the decision problem; these states
of nature are also called events. From the above example of job advert, the possible state
of nature that this prospective applicant may face is that; either he will be the best
candidate or not the best candidate.
 An outcome: This is the consequence for each combination of both the action chosen and
the state of nature. The possible outcomes in the above example are: (1) misses the job
with burdens (When he applies and he turned not to be the best candidate, he will miss
the job while incurring the costs of application), (2) gets the job and be fine (When he
applies and he turned to be the best candidate), (3) misses the job unnecessarily (When he
does not apply, yet he would be the best candidate), (4) misses the job without burden
( When he does not apply and he would not be the best candidate). The reward or cost
attached to each outcome is called payoff. In business, the payoffs associated with each
outcome are always expressed in monetary terms.
 Probability: the decision maker always tries to assess the likelihood of occurrence of
each state of nature. The chance that an event will occur is called probability.
Fig. 1.1: Illustration of Actions, Events, and Outcomes

Apply Not the best candidate misses the job with cost

The best candidate Gets the job and be fine

The best candidate misses job unnecessarily

Not Apply Not the best candidate misses job without cost

ACTIONS STATES OF NATURE OUTCOMES

The above illustration is called decision tree diagram and it always used to evaluate the
appropriateness of each alternative available to a decision maker. A complete decision tree
diagram will always include probability estimates attached to each branch.
 Decision Scenarios
Decision making involves making choices about alternatives whose outcomes depend on the
state of nature in future. If we know for sure what the state of nature would be exactly, we shall
be sure of the outcomes of each decision, and therefore we shall be making decision under
certainty. If we cannot predict the state of nature with certainty, but we can estimate the
likelihood (probabilities) for each possible state of nature, we shall be making decision under the
condition of risk. If we cannot even predict statistically the probability for the occurrences of
each possible state of nature, we shall be making decision under the condition of uncertainty.

Decision making under certainty:

When we know for sure what the outcome of each decision will be, we are dealing with decision
under the condition of certainty. For example, from the above illustration, if this applicant knew
he would get the job and be fine, he would make a decision to apply for the job. Conversely, if
he knew for sure that he would miss the job with costs, and then he would not apply today.

Example 1

A company is planning to set up a new manufacturing plant in any of the four possible locations:
A, B, C, &D. The net profit from this plant over the period of five years depends on the location
that management has chosen. Management has performed a thorough profitability analysis for
the above four locations and results are as shown below:

Table 1.1: Hypothetical profit projected for different locations of the new plant

Location Profits (UGX)

A 15,000,000

B 12,000,000

C 17,000,000

D 19,000,000

Required: If management is very sure about these outcomes, which location should they
choose? Answer: Location D, because it maximizes profits

Decision making under risk

When decision has more than one possible outcome and we know the likelihood of each
outcome, we shall be making decision under the condition of risk. This problem is illustrated in
the following problem.

Example 2
Referring to the above example, suppose management is convinced that profitability for each
location will depend on three possible states of the economy, that is, Recession, Normal, Boom.
Therefore, management cannot predict outcomes with certainty. However, using the available
macroeconomic data from Uganda Bureau of Statistics, they can make the following probability
estimates and the associated estimated profits for each location.

Table 1.2: Hypothetical Profit Projected for Different Locations of the New Plant under Risk

Locations

State of economy probabilit A(UGX) B (UGX) C (UGX) D (UGX)

y

Recession 0.2 3,000,000 2,000,000 4,000,000 6,000,000

Normal 0.3 8,000,000 6,000,000 5,000,000 7,000,000

Boom 0.5 9,000,000 11,000,000 6,000,000 5,000,000

Required: Determine the location which will maximize expected profits

Solution

In this problem we use the method called expected value criterion. Expected values indicate
what an outcome is likely to be in the long run, if the decision can be repeated many times over.
Using this criterion, we sum up the products of probabilities and the respective outcomes.

A: E(V) = (0.2 × 3,000,000) + (0.3 × 8,000,000) + (0.5 × 9,000,000) = 7,500,000

B: E(V) = (0.2 × 2,000,000) + (0.3 × 6,000,000) + (0.5 × 11,000,000) = 7,700,000

C: E(V) = (0.2 × 4,000,000) + (0.3 × 5,000,000) + (0.5 × 6,000,000) = 5,300,000

D: E(V) = (0.2 × 6,000,000) + (0.3 × 7,000,000) + (0.5 × 5,000,000) = 5,800,000

Decision: Chose location B because it has the highest expected profits. If management faces this
situation many times and always chooses alternative B, it average profits will be higher than for
any other alternative.

Limitations of Expected Values

i. An expected value is a weighted average outcome that will occur in a long run if event
occurs many times over, it is therefore limited in a short run.
ii. The expected value of a decision may be a value that will never occur.
iii. It ignores the extreme values since it is an average.
The expected value criterion is mainly used by the risk-neutral decision maker. A risk neutral
decision maker will ignore the extreme values (extremely high or extremely low) in the
distribution, and is only concerned with achieving the average outcome.

Decision Making Under Uncertainty

This arises when a decision has more than one possible outcome and we do not know the
likelihood of each outcome. When we cannot estimate the probabilities of each possible
outcome, we cannot use expected value criterion.

Example 3

Using data from example 2 above with the exception that we remove probability estimates from
the table of results. We shall have the following results:

Table 1.3: Hypothetical Profit Projected for Different Locations of the New Plant under
Uncertainty

Locations

State of economy A(UGX) B (UGX) C (UGX) D (UGX)

Recession 3,000,000 2,000,000 4,000,000 6,000,000

Normal 8,000,000 6,000,000 5,000,000 7,000,000

Boom 9,000,000 11,000,000 6,000,000 5,000,000

Without the probability estimates for possible outcomes, rational strategies for decision
making are not well defined. However, among several approaches to this decision making
scenario, we shall discuss three approaches.

 Firstly, maximax criterion, here we look for the best possible result. It is an optimistic
approach which is always pursued by risk seekers decision makers. Maximax means
maximizing the maximum profits.
Required: Using Maximax criterion, which location will be chosen?

Solution

Consider the table below

Location Maximum possible profits (UGX)

A 9,000,000

B 11,000,000
 C 6,000,000

D 7,000,000

The Maximum Maximum profit is UGX 11,000,000, therefore using the Maximax criterion,
Location B will be chosen.

 Secondly, the Maximin criterion suggests that a decision maker should select the
alternative that offers the least unattractive worst outcome. This means that the decision
maker chooses the maximum of the minimum outcomes. Using the above example, the
decision maker will choose proceed as follows:

Location Minimum possible profits (UGX)

A 3,000,000

B 2,000,000

C 4,000,000

D 5,000,000

The Maximum Minimum profit is UGX 5,000,000, therefore using the Maximin criterion,
Location D will be chosen.

 Thirdly, the Minimax Regret Criterion, here the decision maker seeks to ‘minimize the
maximum regret’ that there would be from choosing a wrong decision. The regret is the
opportunity loss from taking one decision given that a certain contingency occurs.
Example – Using the previous data in table 1.3

Locations

State of economy A(UGX) B (UGX) C (UGX) D (UGX)

Recession 3,000,000 2,000,000 4,000,000 6,000,000

Normal 8,000,000 6,000,000 5,000,000 7,000,000

Boom 9,000,000 11,000,000 6,000,000 5,000,000

Required: Use the minmax regret criterion to choose the optimal location

Solution
We compute the regret for each option in a given state of economy. The workings below show
how the values marked with * are arrived at.

In recession: The best outcome is UGX 6,000,000, we shall subtract outcomes for each option
from this best option to obtain the regret for that option as follows:

Regret for A = 6,000,000 - 3,000,000 = 3,000,000*

Regret for B = 6,000,000 - 2,000,000 = 4,000,000**

Regret for C = 6,000,000 - 4,000,000 = 2,000,000***

Regret for D = 6,000,000 - 6,000,000 = 0****

The decision is good when its regret is minimum. In fact, for a given state of the economy, the
best decision is the one with regret equals zero. This is the case with location D in a recession
where the regret is zero. We shall perform the same procedure for the Normal state of economy
and the Boom and generate the regret table below.

Locations

State of economy A(UGX) B (UGX) C (UGX) D (UGX)

Recession 3,000,000(w1) 4,000,000(w2) 2,000,000(w3) 0(w4)

Normal 0 2,000,000 3,000,000 1,000,000

Boom 2,000,000 0 5,000,000 6,000,000

The maximum regrets for choosing each locations from the regret table, are as follows:

Locations Maximum Regrets (UGX)

A 3,000,000

B 4,000,000

C 5,000,000

D 6,000,000

Using Minimax regret rule: We choose location A because it minimizes the maximum
regret the decision maker will face.

Task for Students

 (1) An investor is faced with three alternative investment opportunities each of which has
several outcomes with associated probabilities as shown in the table below:
Alternative A

Outcomes (UGX) 8,000,000 10,000,000 15,000,000

Probabilities 0.2 0.6 0.2

Alternative B

Outcomes (UGX) 3,000,000 6,000,000 8,000,000 10,500,000 15,000,000

Probabilities 0.1 0.2 0.4 0.2 0.1

Alternative C

Outcomes(UGX) 8,000,000 12,000,000

Probabilities 0.4 0.6

Using Expected Value rule, advise management on which investment they should undertake.

(2) Consider the pay-off table (in UGX ‘000’) for products X, Y,Z below

Products

Scenarios X Y Z

I 10,000,000 8,000,000 7,000,000

II 9,000,000 12,000,00 7,500,000

0

III (3,000,000) 10,000,00 7,500,000

0

Use Minimax Regret rule to advise management on which product should be produced

Decision Tree Diagram

Decision trees are logical representations of decision problems, chance events, and possible
outcomes in choices under risks and uncertainty. When the decision maker is confronted with
different alternatives which have uncertain outcomes, the decision tree would be a very powerful
analytical tool for evaluating various alternatives. In fact, all the information we have been
analyzing in the standard Newsboy problem can be incorporated in this simple diagram called
the Decision tree. However, we shall rely very heavily on the techniques of expected value and
probability theory which we have already learnt in chapter 1 and from previous sections of this
chapter. In fact, the simple decision tree was already drawn in Chapter one, where the decision
maker was faced with the two alternatives of whether or not to apply for some job. In order to
represent the decision problem on a decision tree, we need to understand the following concepts:

i. The decision nodes

ii. The outcomes nodes
iii. The concept of forward pass
iv. The concept of backward pass to calculate the outcomes
v. In some problems, the knowledge of Bayes’ theorem may be necessary

Figure 7.4: The Basic Structure of a Decision Tree

Outcome X

Cb

A Outcome Y

The node labeled ‘X’ is called the decision node. This is the point where the decision maker has
to make a choice between alternatives, A and B. The outcome node is labeled ‘C’. The outcome
nodes are points where the events depend on probabilities.

Example 1

Consider the information below:

R&D Choice Investment Outcomes Profits Probability

(Million) (Excluding
R&D) (Million)
 Biochemical $100 Success $900 0.7

Failure 500 0.3

Biogenetic $200 Success 2000 0.2

Failure 0 0.8

Example 1

A construction company has a £1million contract to complete a building by 31 March 1995 but is
experiencing delays due to the complex design. The managers have to make a decision now
whether to continue as at present, or to employ specialist engineering consultants at a cost of
£200,000. If the company continuous as at present, it estimates there is only a 30% chance of
completing the building on time, and that the delay could be one, two, or three months, with
equal probability. If the building is late, there are penalties of £100,000 for each month’s delay
(or part of a month). The managers believe that if they employ specialist engineering consultants,
their chances of completing the contract on time will be trebled. But if the building is still late, it
will be only one or two months late, with equal probability.

You are required:

i. to draw a tree diagram to represent this decision problem, using squares for decision
points, circles for random outcomes, and including probabilities, revenues and penalties
ii. to analyze the tree using expected value techniques
iii. to write a short report for the managers, with reasons and comments, recommending
which decision to make
This problem is from Terry Lucey (2002)

Solution

Figure 7.6: Decision tree for the construction project

Revenue

On time 0.9 R=1000

1 month delay 0.05* (100) R= 900

2Months delay 0.05* (200) R= 800

Employ (200)

3 Months delay 0.2333** (300) R= 700

Continue 2 Months delay 0.2333** (200) R=800

 1 Month delay 0.2333** (100) R= 900

On time 0.3 R= 1000

1−0.9
P(1month delay) = = 0.05*
2

1−0.9
P(2months delay) = = 0.05*
2

1−0.3
P(3months delay) = = 0.2333**
3

When they employ consultant:

E(P) = 0.9×1000 + 0.05×900 + 0.05×800 - 200 = £785

When they continue as at present:

E(P) = 0.2333×700 + 0.2333×800 + 0.2333×900 +0.3×1000 = £859.92

Management should not employ the specialist engineer and continue as at the present. However,
any delay in this project completion may not only cost the company in terms of penalties but also
in terms of reputation. It may be difficult for them to get contracts in future if this one is not
completed in time.

Example 2 – Value of Perfect Information

Due to the outbreak of the novel COVID-19 pandemic, the management of Alebtong University
of Science and Technology must choose whether to go ahead with either of the two mutually
exclusive projects, full online teaching or blended online teaching. The expected profits from the
two projects will depend on the intensity of COVID-19 pandemic as summarized in table 7.12
below:

Table 7.12: Profits for teaching alternatives and the rates of COVID-19 infection

Alternative Profit if the rate of Profit if the rate of Profit if the rate of
infection is high infection is moderate infection is low

Full online 40,000,000 12,000,000 -10,000,000

Blended online 15,000,000 10,000,000 5,000,000

Probability of intensity 0.2 0.3 0.5

of COVID-19
Required:

i. By drawing the decision tree, advise management on which option they should take.
ii. There is a Virus Research Institute (VRI) that has offered to provide very perfect
information to the university about the projected rate of infection of COVID-19 at the
cost of UGX 3,000,000. Should the university buy this perfect information from the Virus
Research Institute?

Solution

(i) Consider the decision tree diagram below:

Figure 7.9: Decision tree for the profit of the University from each teaching option

B High (0.2) 15,000,000

8,500,000 Moderate (0.3) 10,000,000

Low (0.5) 5,000,000

Blended

8,500,000 High (0.2) 40,000,000

Full A

6,600,000 Moderate (0.3) 12,000,000

Low (0.5) - 10,000,000

Expected value for Blended, that is; at the outcome node, B

E(P) = 0.2×15,000,000 + 0.3×10,000,000 + 0.5×5,000,000 = 8,500,000
Expected value for full online teaching at the outcome node, A is given by:
E(P) = 0.2×40,000,000 + 0.3×12,000,000+0.5×-10,000,000 = 6,600,000
 It is advisable to pursue blended online teaching because it has a higher expected profit

(ii) We need to determine the value of this perfect information using information from
table 7.12 as follows:
Consider table 7.13 below

Table 7.13: Value of perfect information from the VRI

The true state Option taken Prob. Profit

When the rate of infection is high Full online 0.2 40,000,000

When the rate of infection is moderate Full online 0.3 12,000,000

When the rate of infection is low Blended online 0.5 5,000,000

E(P) = 0.2×40,000,000 + 0.3×12,000,000 + 0.5×5,000,000 = 14,100,000

This is more than the expected profit we obtained from (ii). The difference between the two
expected values is the value for perfect information. Hence, the value of the above perfect
information = 14,100,000 – 8,500,000 = 5,600,000. Since the Virus Research Institute has
charged UGX 3,000,000 for this information, the expected incremental profit from buying this
information is 5,600,000 – 3,000,000 = UGX 2,600,000. It is advisable for management to buy
this information from the Virus Research Institute.

Example 3 – Value of Imperfect information

Mr. Ogwang is trying to decide whether or not to contest for Ajuri County parliamentary seat in
the year, 2026. It is a very competitive contest. A consultancy firm which has always been very
accurate in analyzing the local politics in Ajuri County for a very long period of time has
assessed the probability that Mr. Ogwang will win this election, based on their analysis, as 0.2
and the probability that he will not win as 0.8. It is possible for Mr. Ogwang to hire a political
analyst to conduct a preliminary opinion poll in Ajuri County before he goes for the actual
contest. Many politicians have used this political analyst before and it has been established that if
Mr. Ogwang is really going to win, there is a 0.95 chance that this political analyst will advise
him to contest, but if he is not going to win, there is a probability of 0.1 that the analyst will
advise him to contest. The cost of contesting for this parliamentary seat is $50,000 and he will
incur an additional cost of $20,000, if he chooses to hire a political analyst to conduct a
preliminary opinion poll. If he wins, he will get a consolidated salary of $500,000 during his
term of office. Using the decision tree diagram, determine whether Mr. Ogwang should
participate in this parliamentary election. What is the monetary value of conducting this opinion
poll?

Solution
Note: In this solution, the cost of contesting for the seat, that is; $50,000 has been netted against
the consolidated salary of $500,000, so the net benefit used in the decision tree diagram is
$450,000

P(he will win) = 0.2, P(he will not win) = 0.8

P(if he will win, the analyst will advise him to contest) = 0.95

P(if he will win, the analyst will advise him not to contest) = 1-0.95 = 0.05

P(if he will not win, the analyst will advise him to contest) = 0.1

P(if he will not win, the analyst will advise him not to contest) = 1 - 0.1 = 0.9

P(the analyst will advise him to contest) = ?

P(the analyst will advise him not to contest) = ?

he will win
P( contest ¿ ¿ = ?
the analyst advises him ¿

he will not win

P( contest ¿ ¿ = ?
the analyst advises him ¿

he will win
P( contest ¿ ¿ = ?
the analyst advises him not ¿

he will not win

P( contest ¿ ¿ = ?
the analyst advises him not ¿

The decision tree diagram in figure 7.10 below illustrates the above scenario:

Fig. 7.10a: Decision tree diagram for participating in parliamentary election

Don’t hire the analyst, don’t contest $0

Win 0.2 $450,000

Don’t hire the analyst, contest

Not win 0.8 -$50,000

He will win $450,000

-$20,000 Analyst advises to contest

Hire the analyst will not win -$50,000

 He will not win $0

Analyst advises not to contest

He really will win $0

We compute the missing probabilities as follows: Consider the table below:

Table 7.14: Computation of marginal and joint probabilities

He wins He does not win Total

Analyst advises to contest 19* 8* 27

Analyst advises not to contest 1** 72** 73

Total 20 80 100

The firm has estimated a 0.2 (20%) chance that Mr. Ogwang will win and 0.8(80%) chance that
he will not win. In 20 out of 100 cases that he will win, the analyst will advise correctly in 95%
of those 20 cases; that is: 0.95 ×20 = 19*. Similarly; in 80 out of 100 cases that he will not win,
the analyst will advise incorrectly (advises that he contests when he will not win) 10% of the
cases; that is: 0.1×80 = 8*. Figures marked with ** are balancing figures

Therefore, the missing probabilities are:

19 8 27
P(the analyst will advise him to contest) = + = = 0.27
100 100 100

1 72 73
P(the analyst will advise him not to contest) = + = = 0.73
100 100 100

he wins 19/100
P( contest ¿ ¿= P ¿ ¿ = = 0.7037
the analyst advises ¿ 27/100

he does not win

P( contest ¿ ¿= P ¿ ¿
the analyst advises ¿

8/100
= = 0.2963
27/100

he wins P (he wins ∩advised not ¿ contest )

P( contest ¿ ¿ =
the analyst advises not ¿ p(advised not ¿contest )

1/100
= = 0.0137
73/100
 =
he does not win P (he does not win∩ advised not ¿contest ) 72/100
P( contest ¿ ¿ = =
the analyst advises not ¿ p (advised not¿contest ) 73/100
0.9863

Using the backward pass, we now compute expected values at various outcome nodes of the
decision tree diagram below (fig. 7.10b) as follows:

At At A: E(V) = 0.2 × 450,000 + 0.8× -50,000 = $50,000

At B: E(V) = 0.7037 × 450,000 + 0.2963× -50,000 = $301,850

At C: E(V) = 0.9863 × 0 + 0.0137 × 0 = $0

At D: E(V) = 0.27 × 301,850 + 0.73×0 = $81,500(rounded)

We can now draw a complete decision tree diagram as follows:

Fig. 7.10b: Decision tree diagram for participating in parliamentary election

Don’t hire the analyst, don’t contest $0

Win 0.2 $450,000

Don’t hire the analyst, contest A

Not win 0.8 -$50,000

He will win. 0.7037 $450,000

$61,500 -$20,000 Analyst advises to contest 0.27 B

Hire the analyst will not win 0.2963 -$50,000

D He will not win 0.9863 $0

Analyst advises not to contest 0.73 C

He really will win 0.0137 $0

There are three alternatives at the decision node, E namely:

i. Don’t hire analyst and do not contest, E(V) = $0

ii. Don’t hire analyst and contest, E(V) = $50,000, that is; E(V) at A
iii. Hire the analyst first, decide about contesting later, E(V) = $81,500- $20,000 = $61,500
that is; E(V) at D – the cost of hiring the analyst
Conclusion:
It is advisable for him to first hire the political analyst to conduct the opinion poll before he
contests for the parliamentary seat.

The monetary value of hiring the analyst (conducting the opinion poll) is given by:

E(V) for “Hire the analyst first, decide about contesting later” - E(V) for “Don’t hire analyst and
contest” = $61,500 – 50,000 = $11,500

Table 7.10: Decision and payoff of a Newsboy under imperfect information

Actual Analyst’s Decision Profits Posterior Expected Profits ($)

State Prediction Probabilities
($)

Win Will win Contest 430,000 0.20 ×0.95 = 0.19 0.19×430,000 = 81,700

Win Won’t win Not contest -20,000 0.20 ×0.05 = 0.01 0.01×-20,000 = -200

Not win Will win Contest -70,000 0.80 ×0.1 = 0.08 0.08×-70,000 = -5,600

Not win Won’t win Not contest -20,000 0.80 ×0.9 = 0.72 0.72×-20,000 = -14,400

61,500

Example 7.15

A commercial farmer wants to make a decision between two mutually exclusive projects,
planting soya beans and planting sunflowers. Soya beans have higher yields, and therefore,
higher profits than sunflower when there is enough rain. However, sunflowers are more drought
resistant than soya beans and therefore, have higher yields than soya beans when rain is not
enough. Based on past experience, there is a 0.6 probability that there will be drought and 0.4
probability that there will be rain during the farming season. Below is the summary of the
payoffs(profitability) for the two crops.

Table 7.15: Performance of soyabeans and sunflower

Soyabeans (UGX Sunflower (UGX ‘000’)

‘000’)

Rain 70,000 50,000

 Drought 20,000 40,000

The farmer decided to consult a meteorologist to forecast the state of rainfall during that farming
season. It has been estimated that if there will be drought, there is a 0.8 probability that the
meteorologist will predict it correctly. It is also estimated that if there is rain, there is a 0.9
probability that the meteorologist will predict it correctly.

Required:

i. Represent the above information on decision tree diagram

ii. Determine the value of this imperfect information.
Solution

Consider the decision tree diagram below when there is no information:

Consider the probability tree diagram below:

Rain (0.4) 70,000

40,000
2.1

Soyabeans Drought (0.6) 20,000

44,000

Sunflower Rain (0.4) 50,000

44,000

Drought (0.6) 40,000

 Using expected value criterion

For soyabeans

For the first field: E(V) = (0.4 × 70,000) + (0.6 × 20,000) = 40,000

For sunflower

For the second field: E(P) = (0.4 × 50,000) + (0.6 × 40,000) = 44,000

Decision: With no information, this farmer should grow sunflower and earns an expected profit
of UGX 44,000

With imperfect information, consider the table below for the competition of expected profit:

Actual Weather Decision Profits Posterior Expected Profits (UGX)

State Prediction Crop Probabilities
(UGX)

Rain Rain Soyabean 70,000 0.40 ×0.90 = 0.36 0.36×70,000 = 25,200

Rain Drought Sunflower 50,000 0.40 ×0.10 = 0.04 0.04×50,000 = 2,000

Drought Drought Sunflower 40,000 0.60 ×0.80 = 0.480 0.480×40,000 = 19,200

Drought Rain Soyabean 20,000 0.60 ×0.20 = 0.12 0.12×20,000 = 2,400

48,800

With this imperfect information, the expected profit is UGX 48,800. The value of this imperfect
information is:

Value of Imperfect Information = Expected profit with imperfect information – expected profit
with no information.

Value of imperfect information = UGX 48,800 – UGX 44,000 = UGX 4,800

Note: The decision tree with this imperfect information requires a tree diagram and
incorporation of Bayes’ theorem
 Examination Questions with solutions
Question One

A Japanese automobile manufacturer currently produces its best-selling U.S model in Japan, but the
relative strength of the Japanese Yen versus the U.S dollars has been making the car very expensive for
the U.S market. To ensure a lower and a more stable price, the company is considering the possibility of
manufacturing cars for U.S consumers at one of its American plants. Its payoff table, in millions of
dollars, is given below:

State of nature

Weak dollar Moderate dollar Strong dollar

Probability = Probability = 0.5 Probability = 0.2

0.3

Japanese production 10 15 25

American production 20 18 16

Required:

i. Make a decision tree summary of the table above (2 marks)

ii. Compute the expected payoff for each production (2 marks)
iii. Compute the expected payoff with perfect information (1 mark)
iv. Compute the expected value of perfect information (1 mark)
CPA(U) examination June 2008

Solution

The decision alternatives are: Japanese production and American production

The states of nature are: weak, moderate, and strong dollar

The outcomes are the payoff

 Hence, the decision tree diagram is as follows:

B Weak (0.3) 10m

15.5 Moderate (0.5) 15m

Strong (0.2) 25m

Japanese

18.2 Weak (0.3) 20m

American A

18.2 Moderate (0.5) 18m

Low (0.2) 16m

(ii)Expected payoff are as follows:

E(J) = (0.3 × 10) + (0.5 × 15) + (0.2 × 25) = 15.5m

E(J) = (0.3 × 20) + (0.5 × 18) + (0.2 × 16) = 18.5m

(iii) Expected payoff with perfect information

Strength of dollars Decision Payoff Probability

When dollar is weak Use American plant 20m 0.3

When dollar is moderate Use American plant 18m 0.5

When dollar is strong Use Japanese plant 25m 0.2

E(payoff) = (0.3 × 20) + (0.5 × 18) + (0.2 × 25) = 20m

(iv) Expected value of perfect information = 20m – 18.5m = 1.5m

Question 2

An oil company may bid for drilling one of the two contracts for drilling oil in two different
areas in western Uganda. It is estimated that the profit of $3 million per month would be realized
in the first field and $ 4 million in the second field. These profit figures were determined
ignoring the cost of bidding which amounts $3,500 for the first field and $5,000 for the second
field.

Required:

Advise the oil company on which field they should bid for if the probability of winning a
7 3
contract for the first field was and the second field was .
10 5

CPA(U) December, 2008

Consider the probability tree diagram below:

Winning (0.7) $3m

2.1
2.1

($3,500) )

1st field Loosing (0.3) $0

2395000

2nd Field Winning (0.6) $4m

($5,000) 2.4

Loosing (0.4) $0

Using expected value criterion

For the first field: E(P) = (0.7 × 3000000) + (0.3 × 0) - 3500 = 2096500

For the second field: E(P) = (0.6 × 4000000) + (0.4 × 0) - 5000 = 2395000

Question Three

An outdoor concert featuring a popular musical group is scheduled for a performance on Sunday
at Mandela National Stadium. The promoter is worried about Sunday being a rainy day. He has
contracted experts in weather forecasting, who predict the probability of the Sunday being rainy
to be 0.24. If it does not rain, the promoter is certain of collecting Shs. 100 million and if it rains,
he estimates to collect only Shs. 10 million. An insurance company agrees to insure the concert
for Shs. 100 million against rain at a premium of shs. 20 million.

Required:

Advise the promoter whether he should buy the insurance CPA(U) June, 2009

Solution

Consider the decision tree diagram below:

Rainy (0.24) Shs.100m

100
2.1

(20m) Buy Insurance Not rainy (0.76) Shs.100m

80m

Don’t buy Rainy (0.24) Shs. 10m

78.4

Not rainy (0.76) Shs. 100m

Expected profits when he buys insurance:

E(P) = (0.24 × 100 m) + (0.76 × 100m) - 20m = 80m

Expected profit without insurance

E(P) = (0.24 × 10 m) + (0.76 × 100m) = 78.4m

Advice: They should insure the concert for shs 100m against rain.

Question four

HK restaurant has shs. 100,000 available to prepare either tea or coffee in preparation for a
function that takes place annually.in the location. A decision must be made clearly for planning.
However, the restaurant is unable to predict whether the day will be cold or hot. If tea is prepared
and the day is not cold, the payoff will be shs. 120,000, but if the weather is cold, the payoff will
be shs. 105,000. If coffee is prepared the payoffs for hot and cold weather are shs 110,000 and
shs.125,000 respectively. Past records show that 70% of such functions were hot and 30% were
cold.

Required: Determine the two expected payoffs and arrive at the decision

CPA(U) November 2012

Solution

Consider the decision tree diagram below:

Hot (0.7) Shs.110,000

114500

Coffee Cold (0.3) Shs.125,000

115500

Tea Hot (0.7) Shs. 120,000

115500

Cold (0.3) Shs. 105,000

Expected profits for coffee:

E(P) = (0.7 × 110000) + (0.3 × 125000) = 114500

Expected profit for tea

E(P) = (0.7 × 120000) + (0.3 × 105000) = 115500

Advice: They should prepare tea because it has the higher expected profit.

Question five

Three different projects X, Y and Z are to be evaluated. Each project requires an initial
investment of $12,000 and pays off in the following year. Project X pays $14,000 with certainty,
project Y pays either $10,000 or $ 20,000 with probabilities o.5 in each case, project Z pays
nothing with probability 0.78 and $60,000 with probability 0.22. A summary is shown below:
 Project Payoff ($) Probability

X 14,000 1.00

Y 10,000 0.5

20,000 0.5

Z 0 0.78

60,000 0.22

i. Calculate the expected value for each project

ii. Calculate the expected profit for each project
CPA(U) examination November 2013

Solution

0.5 10,000

Y 0.5 20,000

X 1.000 14,000

Z 0.78 0

0.22 60,000
 Expected value

E(Y) = (0.5 × 10000) + (0.5 × 20000) = 15,000

E(X) = 1.000 × 14000 = 14,000

E(Z) = (0.78 × 0) + (0.22 × 60000) =13,200

Expected Profits

Expected profit for Y = 15,000 – 12,000 = $3000

Expected profit for X = 14,000 – 12,000 = $2000

Expected profit for Z = 13,200 – 12,000 = $1200

Question Six

A company has to make a decision on which of the three mutually exclusive projects to carry
out. The directors believe that the success of the project will depend on the consumer reaction.
There is a 25% chance, a 40% chance and a 35% chance that the consumer reaction will be
strong, good and weak respectively. The company is advised to use expected monetary value
theory to make its decision. The net present value for each possible outcome is as follows:

Consumer reaction Project A Project B Project C

Shs (million) Shs (million) Shs (million)

Strong 1,000 1,600 1,200

Good 250 300 375

Weak 200 140 100

A market research company believes it can provide perfect information on the consumer.

Required:

Calculate the:

i. Expected monetary value for each consumer reaction

ii. Maximum amount that should be paid for the value information from the market research
company
CPA(U) examination December 2021
 Strong (0.25) 1,000

Good (0.40) 250

Weak (0.35) 200

Strong (0.25) 1,600

Project A

Good (0.40) 300

Project B Weak (0.35) 140

Project C Strong (0.25) 1,200

Good (0.40) 375

Weak (0.35) 100

Kisenyi Youth Association has secured funding from Entreprenours Development Fund which
they would like to invest in two projects; mobile money agent or agent banking. The expected
profit or loss for each project and their likelihood are as shown:

Decision Alternatives States of nature

High demand (Shs) Low demand (Shs)

Mobile money 1,250,000 750,000

Agent banking 3,000,000 1,870,500

Probability 0.7 0.3

 Consider the decision tree diagram below:

High demand (0.7) Shs 1,250,000

1,100,0
00

Mobile money Low demand (0.3) Shs 750,000

1,538,850

High demand (0.7) Shs.3,000,000

Agent banking 1,538,8

50

Low demand (0.3) Shs.(1,870,500

The expected monetary values are as follows:

For mobile money

E(V) = (0.7 × 1,250,000) + (0.3 × 750,000) = 1,100,000

For agent banking

E(V) = (0.7 × 3,000,000) + (0.3 × -1,870,500) = 1,538,850

Advice: They should invest in agent banking

The expected value of perfect information – consider the table below:

State of nature Decision Payoff Probability

When demand is high Invest in agent banking 3,000,000 0.7

When demand is low Invest in mobile money 750,000 0.3

 E(payoff) = (0.7 × 3,000,000) + (0.3 × 750,000) = 2,325,000

(i) Expected value of perfect information = 2,325,000 – 1,538,850 = 786,150

THE END OF THE TOPIC