TOPIC II

DECISION THEORY

Decision making is defined as a process which results in the selection of the best course
of action among several alternatives. A problem is a decision problem if it involves more
than one alternative to choose from. The goodness of a selected alternative depends on
the quality of the data used in describing the decision situation. The aim here is to study
the environments under which decision making can take place and determine the
techniques used under such environments.

Terminologies used in decision making

(i) Courses of action: options or alternatives or strategies that are available to the
decision maker to choose from

(ii) States of nature: events or circumstances that influence the outcome of the decision
and for which the decision maker has no control

(ii) Probability: chance or likelihood of the occurrence of the events or states of nature

(iv) Payoff: the consequence of taking a particular action or the net benefit to the decision
maker that accrues from a given combination of decision alternatives and events.

A decision-making environment falls into one of four categories:

(1) Decision-Making under Certainty (Deterministic Process)

This is a situation in which the outcome can be predetermined or deterministic (ie known
with certainty) or the outcome is known in advance. For example, if a company
advertises for tenders they specify what is required and therefore only the one who
meets the specifications is chosen. Decision making under this category is trivial and
not challenging and therefore rarely used in most management problems.

(2) Decision-Making under Risk (Stochastic Process)

This is an environment in which the outcome is not known with certainty (stochastic or
probabilistic). The outcome is based on the element of chance or probability. It is based
on the notion that the higher the probability of success the lower the risk and vice versa.

Expected Value Technique (EV)

1
Decision-making under risk is based on Expected Value criterion, in which the
alternatives are compared based on the maximization of expected gain or minimization of
expected loss.
The Expected Value (EV) of an event is its probability times the outcome or value of the
event over a series of trials (probability x payoff). For example, a course of action that
has 0.3 chance of resulting in a gain of shs. 1200 has EV of shs.360.

Expected Monetary Value(EMV) is the sum of EVs ie multiplying each of the possible
outcome of a course of action with its probability of occurrence and summing the
products. For example, the EMV of a course of action that has 0.6 chance of resulting in
a gain of shs. 600 and a 0.4 chance of resulting in a loss of shs. 400 is shs. 200.

Example 1

Edison has invented a telephone device and has three difficult choices to make. Should
he:

(i) manufacture the device himself

(ii) be paid on a royalty basis by another manufacturer
(iii) sell the rights straight away for a lump-sum payment

The profit that is expected from each level of sales and the probability of the sales levels
are given below.

Profit
High Medium Low
Choices Sales Sales Sales
(i) 75,000 25,000 -10,000
(ii) 20,000 30,000 -5,000
(iii) 15,000 -5,000 30,000

Given that P(High Sales)=0.2, P(Medium Sales)=0.5, P(Low Sales)=0.3

Which option should Edison choose?

Solution

The Expected Monetary Values (EMVs) are:

EMV for (i) = 0.2 x 75,000 + 0.5 x 25,000 - 0.3 x 10,000 = 24,500
EMV for (ii)= 0.2 x 20,000 + 0.5 x 30,000 + 0.3 x -5,000 = 17,500

2
EMV for (iii) = 0.2 x 15,000 + 0.5 x -5,000+ 0.3 x 30,000 = 9,500

Thus based on the expected value rule, he will choose option (i).

Expected Opportunity Loss Technique (EOL)

This is the regret or loss from taking one decision given that there exists a better
alternative. The Expected Opportunity Loss can be used to arrive at the same decision as
Expected Value rule except that where maximum EV is chosen, minimum EOL is
required. To determine the EOL we prepare a Regret table, as shown below for the
example 1 above.

Opportunity Loss or Regret table

High Medium Low

Choices Sales Sales Sales
(i) 0 5,000 40,000
(ii) 55,000 0 35,000
(iii) 60,000 35,000 0

Solution

The Expected Opportunity Loss Values (EOLs) are:

EOL for (i) = 0.2 x 0 + 0.5 x 5,000 + 0.3 x 40,000 = 14,500

EOL for (ii)= 0.2 x 55,000 + 0.5 x 0 + 0.3 x 35,000 =21,500
EOL for (iii) = 0.2 x 60,000 + 0.5 x 35,000+ 0.3 x 0 = 29,500

Thus based on the expected opportunity loss rule, he will choose option (i) .

Exercise 1

A newspaper boy has the following probabilities of selling a magazine:

No of copies sold probability

10 0.10
11 0.15
12 0.20
13 0.25
14 0.30

3
Cost of a copy is shs 30 and sale price per copy is shs 50. He cannot return unsold
copies. How many copies should he order?

DECISION TREES

This is a pictorial method of showing a sequence of interrelated decisions and outcomes;

ie a decision tree is used to show all the possibilities of a sequence of events or a structure
for a problem. It can be used to explore beyond the present decision ie the management
takes into account the impact of decisions that may be made in the future. Frequently the
tree is evaluated using expected value technique.

Note:
 Decision nodes represented by squares are points where a choice exists between
alternatives and decisions are made based on estimates and returns expected.

 Event nodes represented by circles are points where the states of nature occur and
depend on probabilities

 Branches are courses of action emanating from a decision node and random
events from outcome nodes

 Payoffs or outcome values are consequences or results of a sequence of actions

taken and the corresponding events

 Probabilities indicate the relative likelihood of occurrence of the events or

branches from outcome node

4
General Structure

Option B1

Event X1
D2
Option B2
Event X2
Option A1
Event X3
D1

Option A2
Event Y1

Event Y2
Option C1

D3 Option C2

Option C3

Example 2

A manager has a choice between

(i) A risky contract promising shs 7 million when the market is favourable with
probability 0.6 and a loss of shs 4 million with probability 0.4for an unfavourable market
and
(ii) A diversified portfolio consisting of two contracts with independent outcomes each
promising shh 3.5 million for a favourable market with probability 0.6 and shs 2 million
with probability 0.4 for unfavourable market. Use decision tree to solve the manager’s
problem.

5
 Favourable (0.6) 7: EV=8x0.6=4.8million

Risky Unfavourable (0.4)

Contract EMV=3.2m
-4: EV=-4x0.4=-1.6million

Favourable (0.6)
Diversified 3.5:EV=3.5x0.6=2.1million
Potfolio
Unfavourable (0.4)
EMV=2.9m
2: EV=2x0.4=0.8million

Thus the decision is to invest in Risky Contract since it gives higher returns.

Note: Extracting information and putting in a decision tree from left to right is called
FORWARD PASS and solving from right to left is called BACKWARD PASS.

Exercise 2

A firm has developed a new product. They can either test the market or abandon the
project. Testing the market costs shs. 50,000 and the likely outcomes are favourable
(p=0.7) and unfavourable (p=0.3). If favourable they could either abandon or produce,
and the market prices are anticipated to be Low, Medium or High with associated
probabilities and outcomes as shown below:

Low p=0.25 Loss 100,000

Medium p=0.6 Profit 150,000
High p=0.15 Profit 450,000

If the test market indicates failure the project is abandoned. Abandonment at any stage
results in a gain of shs 30,000 from the special machinery used. Draw and evaluate the
decision tree for this problem.

(3) Decision-Making under Uncertainty

This involves alternative actions whose payoffs depend on the random states of nature ie
the decision maker is constrained by luck of knowledge about the probabilities of
occurrence of the outcomes of possible courses of action. For each alternative, only
payoffs are known and nothing is known about the likelihood of each state of nature. This
lack of information led to different persons suggesting several criteria for analyzing a

6
decision problem under such situation; Laplace, maximin, maximax, minimax (Regret),
Hurwitz rule.

The Laplace Criterion is based on the principle of insufficient reason. The alternatives
are evaluated using the assumption that all states are equally likely to occur. We assign
equal probability to each event , multiply by the payoff and sum to get the average for
each option .This can also be determined by summing the payoffs for each option and
dividing by the number of events. For instance if in the example below the sales
1
probabilities were not known, the criteria would assume equal probability of for the
3
three states.
Events

High Medium Low

Choices Sales Sales Sales
A 8,000 6,000 -5,000
B -4,000 12,000 7,000
C 11,000 -6,000 8,000

Thus

1 1 1
Average value for A = x 8,000 + x 6,000 - x 5,000 =3,000
3 3 3
1 1 1
Average value for B = x -4,000 + x 12,000 + x 7,000 =5,000
3 3 3
1 1 1
Average value for C = x 11,000 - x 6,000+ x 8,000 = 4,333
3 3 3

Thus by Laplace rule he should choose Option B.

Maximin (Best of the Worst)- a decision rule based on maximizing the minimum returns
that can occur. Also known as Wald Decision Criterion. From the above example,

Option Worst Payoff

A -5,000
B -4,000
C -6,000

By maximin rule choose Option B.

7
Maximax (Best of the Best)- an optimistic decision rule based on maximizing the
maximum returns that can be gained. From the example above,

Option Best Payoff

A 11,000
B 12,000
C 8,000

By maximax rule choose Option B.

Minimax (Regret)- a decision rule that seeks to minimize the maximum regret that there
could be from choosing a particular strategy. Also known as Savage Decision Criterion.
We construct a regret table for the above example .

High Medium Low

Choices Sales Sales Sales
A 3,000 6,000 13,000
B 15,000 0 1,000
C 0 18,000 0

Option Maximum Regret

A 13,000
B 15,000
C 18,000

By minimax rule choose Option A.

Hurwitz rule- is a weighted average of the best and worst payoffs of each action and is
given by weighted payoff =  x best payoff + (1-  ) x worst payoff where  is a
number between 0 and 1, also known as pessimism-optimism index. The value of 
chosen reflects the decision maker’s attitude to the risk of poor payoffs and the chance of
good payoffs. An  -value above 0.5 implies risk avoidance while an  -value below
0.5 implies risk seeking.

From the example above and taking  =0.7,

weighted average value for A = 0.7 x 8,000 + 0.3 x-5,000 = 45,000

weighted average value for B = 0.7 x 12,000 + 0.3 x -4,000 = 72,000
weighted average value for C = 0.7 x 11,000 + 0.3 x -6,000 = 59,000

The decision is to choose Option B.

8
Exercise 3

A steel manufacturing company is concerned with the possibility of a

strike. It will cost an extra 20,000 pounds to acquire an adequate
stockpile. If there is a strike and the company has not stockpiled,
management estimates an additional expense of 60,000 pounds on
account of lost sales. Should the company stockpile or not if it is to use
(i) Laplace (ii) maximin (iii) maximax (iv) minimax (v) Hurwitz rule
with  =0.4

DECISION MAKING UNDER COMPETITION OR CONFLICT

(GAME THEORY)

The term ‘game’ represents a conflict between two or more parties. A game must meet
the following specifications:
 In each game there are a finite number of participants
 Each participant has a number of possible courses of action (strategies)
 All possible combinations of outcomes of a game can be estimated in advance
(pay-off)

Here, we consider a simple game that involves only two players. The two opponents,
each has a number of alternatives or strategies, and associated with each pair of strategies
is a pay off that one player pays to the other.

Games with two players where the gain by one player equals the loss to the other and
when the two values are added we get zero are known as two-person-zero-sum-games.

There are two types of games:

Pure Strategy Game

Deals with decision situations in which two intelligent opponents have conflicting
objectives and each player plays only one strategy all the time. The outcome of such a
game is known with certainity ( ie deterministic) eg launching advertisement campaigns
for competing products.

In solving two-person-zero-sum-games, one or more strategies for each player is selected

such that any change in the chosen strategies does not improve the pay off to either
player. The games are solved using the maximin or best of the worst creteriuon.

9
Example

Two players A and B sell two brands of a drug. Company A advertises in Radio(A1),
TV(A2), and newspaper(A3). Company B in addition to using Radio(B1), TV(B2), and
newspaper(B3), also mails Brochures(B4). Depending on the intensity of the campaign,
each company can capture a portion of the market from the other. The following table
summarizes the percentage of the market captured or lost by company A.

B1 B2 B3 B4 ROW
MIN
A1 8 -2 9 -3 -3
A2 6 5 6 8 5
A3 -2 4 -9 5 -9
COLUMN 8 5 9 8
MAX

Solution

The solution is based on securing the best of the worst for each player. If A selects A1,
then regardless of what B does, the worst that can happen is that A would lose 3% of the
market share to B. Strategy A2`s worst outcome is for A to capture 5% from B and A3’s
worst outcome is for A to lose 9% to B. To achieve the best of the worst, company A
chooses strategy A2 ie the maximum value in the row min column.

Next, because the pay off table is for A, B’s best of the worst criterion requires
determining the minimax value, which calls for B to select strategy B2. Thus the solution
to the game is that both companies should use TV advertisement. The pay off will be in
favour of A as its market share will increase by 5%. We say that the value of the game is
5% and that A and B are using a pure strategy saddle-point solution. The saddle-point
solution guarantees that neither company is tempted to select a better strategy; e.g if B
moves to say B1,B3 or B4 and A stays at A2, what happens? Similarly what happens if A
moves to A1 or A3?

Mixed Strategy Game

Not all solutions of a game are characterized by pure solutions. Instead, the solution may
require mixing two or more strategies randomly. In other words, the outcomes are not
known with certainity, hence probabilistic or stochastic. In a mixed strategy a manager
follows a certain course of action, say A, until an alternative course of action, say B,

appears to be more profitable. Later should action A appear more attractive again, the
manager switches back. That is there is constant temptation by each player to switch to
another strategy if the current one is not favorable hence a pure strategy is not acceptable.
Instead both players must use random mixes of their respective strategies. The value of
the game is therefore not definite.

10
Exercise 4

Determine the saddle-point solution, the associated pure strategies and the value of the
game for each of the following games. The pay offs are for player A.

(a) (b)
B1 B2 B3 B4
B1 B2 B3 B4 A1 4 -4 -5 6
A1 8 6 2 7 A2 -3 -4 -9 -2
A2 8 9 4 5 A3 6 7 -8 -10
A3 7 5 3 5 A4 7 3 -12 5

11