Decision Theory

Introduction
Decision making can be regarded as an outcome of mental processes (cognitive processes)
leading to the selection of a course of action among several alternatives. Every decision-making
process produces a final choice. The output can be an action or an opinion of choice.
Definitions of key terms
Decision – is a commitment to irrevocably allocate valuable resources. It is a commitment to act
and action is the irrevocable allocation of valuable resources.
Outcome/consequences – A consequence is a result of a course of action or decision taken by
a decision maker.
Decision theory – is a body of knowledge related analytical techniques of different degrees of
formalities designed to help a decision maker choose a set of alternatives in light of possible
consequences. It is a theory that applies in conditions of uncertainty, risks and certainty.
Choice of judgement – it is a tough decision e.g. I will take Maria as my wife. It can be defined
simply as firmness of conviction. These are results arrived at by judges.
Decision makers – A group of persons who make final choice among alternatives.
Alternative – option – this is one of the mutually exclusive courses of action in attaining the
objective.
Course of action – strategy or means available to a decision maker by which the objectives may
be attained.
Objective – something that a decision maker seeks to accomplish or to obtain by means of his
decision. It is short term.
2 0 8 quantitative techniques
STUD Y TEX T
Goal – a general objective. It is long term. Goal is used to denote a very general objective
Cyclic variation - Any change in economic activity that is due to some regular and/or recurring
cause, such as the business cycle or season.
Industry context
We apply decision theory models to political science with the intent of creating a positive theory
of politics. We use models of self-interested individuals to consider various voting mechanisms
used to generate group decisions.
Several important applications of decision theory with uncertainty, applying decision theory to
investments, insurance, and search.
Exam Context
For successful organizations and institutions, good ideas have to be made. In this case the
examiners have previously examined the students as outlined below:
12/06, 6/06, 6/06, 12/05, 12/05, 6/05, 12/03, 12/02, 6/02, 12/01, 12/00
Decision Theory
Almost everything that a human being does involves decisions. We make choices all the time.
Some decisions are easy others are not.
Consider the following and the problem they give:
1. Shall I bring an umbrella today? - This decision depends on something I don’t know,
namely whether it will rain.
2. I am looking for a house, shall I buy this house? This house looks fine, but perhaps I’ll
find a better house if I go searching
3. Shall I smoke the next cigarette? One single cigarette is no problem, but if I make the
same decision sufficiently many times it may kill me.
A decision maker may have more than one objective. The term goal is sometimes used to denote
a very general objective, usually long term.
Decision making process.
Is the process of choosing among alternative courses of actions which are feasible.
Process:
•• Identify the objectives
•• Search for alternative courses of action
•• Gather data about the alternatives
•• Select alternative courses of action
•• Compare actual and planned outcome
•• Respond to divergencies for the plan.
209
STUD Y TEX T
Decision Theory
Decision theory generally involve 4 steps
Example: Consider a manufacturing company that is thinking of several methods to increase its
production.
Steps:
1. List all the viable alternatives for the company considered above. There may be the
following options.
a) Expand the present plant
b) Construct a new plant
c) Subcontract the plan for extra demand.
2. I dentify the expected future events. Often it’s possible to identify most of the events that
can occur i.e. states of nature.
The difficulty is to identify which particular event will occur. For the manufacturing company,
the greatest uncertainty will be about product demand. The future events related to a demand
will be
i) high demand
ii) Moderate demand
iii) Low demand
iv) N o demand
3. C onstruct a payoff table – The decision maker makes a payoff table representation
profits or benefits for each combination course of action of a data and states of nature.
The possible payoffs for the manufacturing company expansion.
Table 6.1
States of nature
Alternatives High Moderate Low Nil
Expand 50,000 25,000 (25,000) (45,000)
Construct 70,000 30,000 (40,000) (80,000)
Subcontract 30,000 15,00 (1,000) (10,000)
4. S elect optimum decisions criteria - Decision makers will choose criteria which result in
target profit. The criteria may be economic, qualitative or quantitative.
Decision making environment
1. Certainty – In this environment, there exists only one state of nature i.e there is complete
certainty about the future. Complete information is also available as to which state of nature
is going to occur. It is thus easy to analyse the situation and make good decisions. The
decision making process is just picking the best alternative.
Unfortunately certainty environment is a very simplistic environment which is rarely applicable
in real life situation.
Examples: Linear programming, transportation and assignment techniques, I/O analysis
and EOQ
Few complex managerial problems ever enjoy the luxury of complete information about the
future and thus decision making under certainty is of little consequential interest.
2 1 0 quantitative techniques
STUD Y TEX T
2. Uncertainty – More than one state of nature exists but the decision maker lacks sufficient
knowledge to allow him assign probabilities with the various states of nature. Uncertain
events are those events that cannot be predicted with statistical confidence i.e the decision
maker does not know the relevant variables nor their probability therefore decision making
in this environment depends on the risk attitude of the decision maker i.e. risk averse or a
risk seeker or risk neutral.
A risk averse avoids risk at all cost; the individual is conservative and assumes the worst
situation will occur.
A risk seeker takes high risks in expectation of high returns. He assumes that the best
outcome will occur.
A risk neutral person is not affected by risks. Such people will make decisions based on
something else but not risk.
3. Risks – Here more than one state of nature exists and the decision maker has sufficient
information to allow him probabilities to each of these states of nature. Therefore it involves
situations or events which may/may not occur but its probability of occurrence can be
calculated statistically. There are frequency of occurrence predicted from first records.
4. Competition – In this environment, the decisions of the firm are affected by the decision of
other firms with opposing interests.
Decision making under uncertainty
When the probability of occurrence of each state of nature can be assessed the expected
monetary
value (EMV) of the expected opportunity loss (EOL) decision criteria are usually appropriate.
When a manager can’t assess the outcome probability with confidence or virtually no probability
data are available other decision criteria may be applied. These include:-
i. Maxmax
ii. Maxmin
iii. Laplace (equally likely/ Criteria of rationality)
iv. Hurcwiz (criteria of realism)
v. Minimax regret criteria (savage criterion)
1. Maxmax
Finds alternatives that maximise the max outcome or consequence of every alternative. The
decision maker first locates the maximum number since this decision criterion locates the
alternatives with the highest possible gain it has been called optimistic decision criterion. The
criterion appeals to risk takers for optimism who are ready to make huge profits if they occur.
Table 6.2
Alternative high medium low N o Action Max. of row
Expand 50,000 25,000 -25,000 -45,000 50,000
Contract 70,000 30,000 -40,000 -80,000 70,000
Subcontract 30,000 15,000 -10,000 -10,000 30,000
Maxmax payoff is sh70,000 corresponding to contract alternative.
211
STUD Y TEX T
Decision Theory
2. Maxmin
Finds alternatives that maximise the min or consequences of each alternative. Under this rule,
the decision maker picks the worst possible outcome under each alternative and then the best
of these worst outcomes. Since this decision maker locates the alternatives that has the least
possible loss it has been called pessimist decision criteria. The criterion appeals to risk averse
decision makers because it’s a criterion of extreme caution. It assumes that the worst outcome
will occur.
Table 6.3
Alternative high medium low N o Action min. of row
Expand 50,000 25,000 -25,000 -45,000 -45,000
Contract 70,000 30,000 -40,000 -80,000 -80,000
Subcontract 30,000 15,000 -10,000 -10,000 -10,000
The maxmin payoff is sh (10,000) to the company. Corresponding to the decision subcontract.
3. Laplace (criterion of rationality)
It holds that if a decision maker does not know the probability of the various states of nature
and has reason to think otherwise then the states of nature will be considered equally likely.
This criterion is based upon the principle of insufficient reason. The criterion assigns equal
probabilities to all the events of each alternative decision and selects the alternatives with the
max expected payoff.
Symbolically if n denotes the no of events and P1 donates the payoffs the expected values of
strategy say S1 = 1/n (p1 +p2 …. +pn)
Table 6.4
Alternative high medium low N o action expected payoff
Expand 50 25 -25 -45 ¼ (50 + 25 -45)=1250
Construct 70 30 -40 -80 (500)
Sub construct 30 15 -10 -10 8,500
NB: Units in ‘000’
The alternative subcontract will be the best to take since it results in max average payoff of sh
8,500.
4. Hurcwicz criterion (criterion of realism / weighted average criterion)
Is a compromise between maxmax and maxmin decision criteria. The criterion is based on
Hurwicz concept of optimism or pessimism. This concept allows the decision maker to take into
account both the maximum and minimum of each alternative and assigns the weight according
to the degree of pessimism or optimism
The alternative which maximises the sum of these payoffs is then selected.
The criterion of realism consists the following steps:
i. A coefficient of realism (alpha) is selected. The coefficient is between zero and one.
When alpha is close to 1 the decision maker is optimistic about the future and when is
it close to zero he is pessimistic about the future.
2 1 2 quantitative techniques
STUD Y TEX T
ii. Determine the max as well as min of each alternative and obtain the following:
iii. P = α (maximum) + 1-α (minimum)
iv. C hoose the alternative that yields the maximum value P using example above
Table 6.5
Alternative high medium low No Action min. of row max p=α(max)+(1-α)min
Expand 50 25 -25 -45 -45 50 31
Contract 70 30 -40 -80 -80 70 40
Subcontract 30 15 -10 -10 -10 30 22
NB: Units in ‘000’
Assume that coefficient of realism, α = 0.8
The best decision under Hurwicz criterion would be to construct since it has the highest weighted
payoff of Ksh 40,000.
Minimax regret criterion (Savage criterion)
Was developed by L.J Savage. He pointed out that the decision has been made and state of
nature has occurred. Thus, the decision maker should attempt to minimise regret by minimising
the maximum opportunity loss of each alternative.
Steps
i. Develop opportunity loss table and then find the max opportunity loss within each
alternative.
ii. S elect the alternative with the min or smallest opportunity loss
Table 6.6
Alternative High Moderate Low Nil Max
Expand 20,000 5,000 24,000 35,000 35,000
Construct 0 0 39,000 70,000 70,000
Subcontract 40,000 15,000 0 0 40,000
The company should expand because this decision will minimise its regret which is Sh. 35,000.
Decision making under risk
In this environment, it is possible to attach some probabilities in the various states of nature.
Several states of nature may occur each at a given probability. The decision maker would either
choose the expected monetary value or the expected opportunity loss (EOL)
These two are similar in the sense that the alternatives that maximise the EMV will minimise
EOL
at the same time.
Expected Monetary Value
Given a decision payoff table with conditional value (payoffs) and probability assessments for all
states of nature, it is possible to determine the expected monetary value for each alternative if the
decision would be expected a large number of times. The EMV of an alternative is just the sum
of possible payoffs of the alternative each weighted by the probability of that pay off occurring
so
that EMV (alternatives) =
213
STUD Y TEX T
Decision Theory
Payoff of 1st state of nature ×probability of 1st state of nature + Payoff of 2nd state of nature
× probability of 2nd state of nature + payoff of last state of nature ×probability of last state of
nature.
Steps
i. C onstruct a payoff table listing the alternative decisions and various states of nature.
Enter conditional profit for each decision event combination along with the associated
probability.
ii. Calculate the EMV for each alternative decision by multiplying the conditional profit by
assigned probability and adding the resulting values
iii. S elect the alternative that yields the highest EMV.
Example
Consider the following table for Thomson Manufacturing Company
States of nature Favourable mkt unfavourable mkt
Alternative $ $
Construct a large plant 200,000 -180,000
Construct a small plant 100,000 -20,000
Do nothing 0 0
Probability 0.5 0.5
Which alternative will give the best EMV?
EMV (large) = 0.5 (200,000) + 0.5 (-180,000) = 10,000
EMV (small) = 0.5 (100,000) + 0.5 (-20,000) =40,000
EMV (Nothing) = 0
The largest expected monetary value results from the 2nd alternative which is to construct a
small
plant.
Expected opportunity loss
An alternative for maximising EMV is to minimise EOL. EOL represents the amounts by which
the maximum possible profit will be reduced under various possible stock action. The cost of
action that minimise these losses is the optimal decisions alternative.
Procedures
i. Prepare the conditional profit table for each decision taken and the associated profit
ii F or each event determine the conditional opportunity loss by subtracting the pay off of
the max pay off of that event.
iii. C alculate the EOL for each decision alternative by multiplying the conditional opportunity
loss by the associated probabilities and then adding the values.
iv. S elect the alternative that yields the lowest value
Alternative favourable unfavourable
Large 0 180,000
Small 100,000 20,000
Do nothing 200,000 0
2 1 4 quantitative techniques
STUD Y TEX T
EOL (large) = 0.5 0) +0.5(180,000) = 90,000
EOL (small) = 0.5 (100,00o) + 0.5(20,000) = 60,000
EOL (nothing) = 0.5 (200,000) + 0.5 (0) = 100,000
The decision to construct the small plant will be selected since it has the lowest EOL
Note: i) The minimum EOL will always result in the same decision as EMV
ii) The minimum EOL= Expected value of project information ( EVPI)
Expected value of perfect information (EVPI)
EV with PI is the expected or average return in the long run. If we have perfect information
before
a decision has been made. To calculate this value we choose the best alternative for each state
of nature and multiply its payoff by the probability of occurrence of that state of nature
The EVPI is the expected outcome with perfect information minus the expected outcome without
perfect information i.e. the maximum EMV.
EVPI = EV, PI – EMU (max) EVPI = Min (EOL)
EVPI = (200,000 × 0.5) + ( 0 x 0.5) – 40,000
EVPI = $60,000 = MIN. EOL
EV PI = is the information which guarantees the future with 100% degree of accuracy.
Multistage decision making
Decision trees
It is a graphic representation of the decision alternatives, states of nature, probabilities attached
to the state of nature and conditional losses.
It consists of a network of nodes. Two types of nodes are used; decision node represented by a
square and states of nature (chance of event) node represented by a circle.
Alternative courses (strategies) originate from the decision node as main branches (decision
branches)
At the end of each decision branch there’s a state of nature mode from which emanates chance
events in the form of sub-branches (chance branches)
The respective pay offs and probability associated with alternative course and chance events
are shown alongside those branches. At the terminal of those branches are shown the expected
value of outcomes.
215
STUD Y TEX T
Decision Theory
Diagram 6.1
The general approach used in decision tree analysis is to work backward through the tree from
right to left compiling the expected value of each chance node and then choosing the particular
branch leaving a decision node which leads to the chance node with the highest expected
value.
Steps in decision tree analysis
1. I dentify the decision point and the alternative courses of action at each decision point
systematically.
2. A t each decision point determine the probability in payoff associated with each course
of action.
3. C ommencing from the extreme right, compute the expected pay off (EMV) from each
course of action
4. C hoose the course that gives the best payoff for each of the decision.
5. Proceed backwards to the next stage of decision point.
6. R epeat above steps until the 1st decision point is reached.
7. I dentify the courses of action to be adopted from the beginning to the end under the
different possible outcomes for the situation as a whole.
Advantages of Decision Tree Approach
i. I t structures the decision process and helps decision making in orderly and systematic
and consequent manner.
ii. I t requires the decision maker to examine all possible outcomes whether desirable or
undesirable.
iii. I t communicates the decision making to others in an easy clear manner illustrating each
assumption about the future.
iv. It displays the logical relationship between the parts of a complex decision and identifies
the time sequence in which various actions and subsequent events occur.
Limitations
i. Decision tree diagram becomes more complex as the number of decision alternatives
increases and more variables are introduced.
ii. I t becomes highly complicated when interdependent alternatives and dependent
variable are present in the problem.
Action A1
Oucome X1
Oucome X2
Oucome X1
Oucome X2
Action A2
Action B1
Action B1
2 1 6 quantitative techniques
STUD Y TEX T
iii. I t analyses the problem in terms of expected value and thus yields an average value
solution
iv. T here is often inconsistency in assigning probabilities of different events
Perfect and imperfect information
Uncertainty about the future can sometimes be reduced by first obtaining more information
about what is likely to occur. Such information can be obtained from various sources like market
research, pilot testing, building a prototype model, etc. Information can be categorised into:
i. Perfect information – is information which guarantees the future with 100% surety.
ii. I mperfect information – is information which, although good and hence better than
having no information at all, could be wrong in its prediction. Also referred to as sample
information.
Expected value of sample information therefore is the maximum that the decision maker should
be willing to spend in order to obtain sampled information.
Decision Trees and Bayes Theorem
There are many ways of getting probability data. The numbers can be assessed by a manager
based on past experience and intuition. They can be derived from historical data or they can be
computed from other available data using Bayes theorem approach. It recognises that a decision
maker does not know with certainty what state of nature will occur. It allows the manager to
revise
his /her initial or prior probability assessment. These are called posterior probability.
This rule or theorem is given by
P(A|B) =
()()
P(B)
PA×PBA
It is used frequently in decision making where information is given in the form of conditional
probabilities and the reverse of these probabilities must be found.
Example 1
In a class of 100 students, 36 are male and studying accounting, 9 are male but not studying
accounting, 42 are female and studying accounting, 13 are female and are not studying
accounting.
Use these data to deduce probabilities concerning a student drawn at random.
Solution:
Accounting A Not accounting
A Total
Male M 36 9 45
Female F 42 13 55
Total 78 22 100
217
STUD Y TEX T
Decision Theory
P(M) = 45 = 0.45
100
P(F) = 55 = 0.55
100
P(A) = 78 = 0.78
100
P(A) = 22 = 0.22
100
P(M and A) = P(A and M) = 36 = 0.36
100
P(M and A) = 0.09
P(F and A) = 0.13
These probabilities can be expressed differently as;
P(M) = P(M and A) or P(M and A )
= 0.36 + 0.09 = 0.45
P(F) = P(F and A) or P(F and A )
= 0.42 + 0.13 = 0.55
P(A) = P(A and M) + P(A and F) = 0.36 + 0.42 = 0.78
P(A)= P( A and M) + P( A and F) = 0.09 + 0.13 = 0.22
Now calculate the probability that a student is studying accounting given that he is male.
This is a conditional probability given as P(A|M)
P(A|M) = 0.80
0.45
0.36
P(M)
P(A and M) = =
From the formula above we get that,
P(A and M) = P(M) P(A|M) ……………….. (i)
Note that P(A|M) ≠ P(M|A)
Since P(M|A) =
()
P(A)
P A a nMd
this is known as the Bayes’ rule.
Bayes’ rule/Theorem
This rule or theorem is given by
P(A|B) =
()()
P(B)
PA×PBA
−
−
−
2 1 8 quantitative techniques
STUD Y TEX T
It is used frequently in decision making where information is given in the form of conditional
probabilities and the reverse of these probabilities must be found.
Example 2
Analysis of questionnaire completed by holiday makers showed that 0.75 classified their holiday
as good at Malindi. The probability of hot weather in the resort is 0.6. If the probability of
regarding holiday as good given hot weather is 0.9, what is the probability that there was hot
weather if a holiday maker considers his holiday good?
Solution
P(A|B) =
()()
P(B)
PA×PBA
Let H = hot weather
G = Good
P(G) = 0.75 P(H) = 0.6 and P(G|H) = 0.9 (Probability of regard holiday as good given hot
weather)
Now the question requires us to get
P(H|G) = Probability of (there was) hot weather given that the holiday has been rated as good).
=
( ) ( ) ( )( )
0.75
0.6 0.9
P(G)
PHPGH
=
= 0.72.
Decision Making Under Competition
Game theory
Competition is an important economic factor. The strategy taken by individuals or organisations
can dramatically affect the outcome of our decision. In the automobile industry, for example, the
strategies of competitors to introduce certain models can dramatically affect the profitability of
car makers.
Today’s business cannot make important decisions without considering what other organisation
or individuals are doing or might do. A game theory is one way to consider the impact of others
on our strategies and outcome. A game is a context involving two or more decision makers each
of whom wants to win.
Game theory is a study of how optimal studies are formulated in conflict situations. In the
business
context, the term game refers to conflicts through time. It is to do with business situations where
usually the success of one person is at the expense of the other.
Competitive game
Properties
i. There’s a finite number of participants. If the number of participants is two, the game is
referred to as a two-person game. If we have M number of participants it is called an
M-person game.
219
STUD Y TEX T
Decision Theory
ii. Each participant has a finite number of possible courses of action
iii. E ach participant must know all the courses of action available for the others but must
not know which of these will be chosen.
iv. A play of the game is said to occur when each player chooses one of its courses of
action. The choices are assumed to be made simultaneously so that no participant
knows the choice of the other until he has decided on his own.
v. A fter all participants have chosen their courses of action their respective game options
are finite.
vi T he game of participants depends on his action as well as those of others
Useful terminology
1. Player – each participant (interested party) is called a player
2. A play of the game results when each player has chosen a course of action. After each
play of the game one player pays the other an amount determined by the courses of
action chosen.
3. S trategy – the decision/role which a player determines his/her course of action is called
strategy. To reach a decision of which strategy to use neither player needs to know the
other’s strategy.
4. Pure strategy – if a player decides to use only one particular course of action during
every play he is said to use a pure strategy. A pure strategy is usually represented by a
number with which a course of action is associated.
5. Mixed strategy – if a player decides in advance to use all or some of his available
courses of action in a fixed proportion he is said to use a mixed strategy. Thus a mixed
strategy is a selection among pure strategies with some fixed proportions.
6. Zero sum game – A game where a gain of one equals the loss of the other is known
as a two person zero-sum game. In such a game, the interest of the two players are
opposed so that the sum of their net gain (sum of the game) is zero. If there are nplayers
and the sum of the game is zero, it is an n-person zero sum game.
Two person zero sum game
Characteristics of two-person zero – sum game
i. Only two players participate
ii. Each player has finite number of strategies to use
iii. Each specific strategy results in a payoff.
iv. T otal payoff to the two players at the end of each play is zero
v. Payoff is the outcome of playing the game. A payoff (game matrix) is a table showing
the amounts received by the players named at the left hand side after all possible plays
of the game. The payment is made by the player at the top of the table.
Fast Forward: Game theory attempts to mathematically capture behaviour in strategic
situations,
in which an individual's success in making choices depends on the choices of others.
Assumptions of Game theory
i. E conomic theory – Each player is assumed to be economically rational; each player
desires to win i.e. sometimes referred to as requirements
ii. I nformation – Both players have the same information set i.e. each player knows the
strategies of the opponent and the consequential payoffs.
2 2 0 quantitative techniques
STUD Y TEX T
iii. Zero Sum of the game – The gain of a given player equals the loss of the other player
so that the game is zero sum in nature
iv. N o indifference – Each player has to play even if they are losing
v. R epeatedness – The game is played expectedly over a long period of time
Example
Consider the following:
S trategy Y1 strategy Y2 min. of row
X1 3 5 3
X2 1 -2 -2
Max. of column 3 5
Determine the strategy that each player will employ.
Both players have dominant or pure strategies therefore the game has a saddle point. Numerical
value of saddle point is the game outcome. Saddle point exists only in pure strategy game. In
this example the saddle point is 3
Note:
•• In reality player X and Y may not see the saddle point.After the game is played for some
time, however, each player will realise that there is only one strategy to be played.
From then on these players will only play one strategy which corresponds to the saddle
point.
•• The value of the game is the average of expected game outcome. If the game is played
in an infinite number of times the value of this game is 3. If a game has a saddle point
the value of the game is equal to its nominal value.
•• The saddle point in this example is the largest number in its column and the smallest
number in its row. This is true for all saddle points.
The minimax procedure is accomplished as follows
1. F ind the smallest number in each row
2. Pick the largest of them. The number is called the lower (maxmin) and the row is the X’s
maxmin strategy.
3. I f the upper value and lower value are the same there is a saddle point which is equal
to the lower and upper value; if not equal, there is no saddle point and if the value of the
game lies between two values such a strategy is called a mixed strategy game.
Note:
•• The game is said to be fair if the maxmin value = minmax value =0 and it is said to be
strictly determinable if the maxmin value = minmax value
•• Always look for a saddle point before attempting to solve a game.
Mixed Strategy Game
When there is no saddle point, players will play each strategy for a certain percentage of the
time.
This is called a mixed strategy game.
Two methods
1. A lgebraic method
2. A rithmetic method
221
STUD Y TEX T
Decision Theory
Algebraic method.
For a 2×2 game an algebraic approach can be used to solve for the percentage of the time each
strategy is played.
Principle of rational expectation
X wants to decide his/her time so that he wins as much when Y is playing Y1 as when Y is
playing
Y2. The following diagram can be used:
Table 6.7
Y
Y1 (p) Y2(1-p)
X
X1(Q)
X2(1-Q)
Where Q, 1-Q equals the proportion time when X plays strategy X1 and X2 respectively P, 1-P
equals the percentage of time when y plays Y1 and Y2 respectively. The overall objective of
each player is to determine the fraction of time that each strategy is to be played to maximise
winnings.
Steps
1. To find X`s best strategy multiply Q and 1-Q times the appropriate game outcome and
solve for Q and 1-Q by setting column 1 equal column 2 in the game.
2. To find Y`s best strategy multiply P and 1-P times the appropriate outcome numbers
and solve for p and 1-p by setting row 1 equals to tow 2 in the game
Arithmetic Method
Steps
1. S ubtract the two digits in column 1 and write them under column 2 ignoring the sign
2. S ubtract the two digits in column 2 and write them under column 1 sign
3. S imilarly proceed to the two rows. These values are called oddments. They are the
frequencies with which the players must use their courses of action in their optimum
strategies.
Table 6.8
Y
Y1 (p) Y2(1-p)
X X1(Q) 4 2 9/11
X2(1-Q) 1 10 2/11
8/11 3/11
2 2 2 quantitative techniques
STUD Y TEX T
Value of the game= 9/11×2 + 2/11×10
=35/11
Note: The arithmetic and algebraic method can only be employed if we are dealing with a 2x2
game matrix i.e. this is a situation where each of the two players has two possible courses of
action.
Principle of dominance
The principle of dominance can be used to reduce the size of the games by eliminating strategies
that will never be played. A strategy or a play can be eliminated if the player can always do as
will
or better than another strategy. In other words, a strategy can be eliminated if all of its outcomes
are the same or worse than the corresponding game outcomes of another strategy.
Note: The objective is to reduce the game to 2x2 so that to use algebraic or arithmetic method.
Example
Using the principle of dominance reduce the size of the following game
y1
y2 y3 y4 min of row
x1
-5 -4 6 -3 -5
x2
-2 6 2 -20 -20
Max of col. -2 6 6 -3
It has no saddle point.
Y will never play Y2 and Y3 as he will lose. He will play Y1 and Y4.
Reduce the following by dominance and find the value of the game.
Non zero-sum game
An assumption of zero-sum game is that the gain of a given player equals the loss of the other
player. This may be possible in practice for some situations e.g. many games or sports like
football; athletics etc. Also a fixed make competition can be modeled as a zero-sum game.
However, in the business world many situations cannot realistically be modeled as zero-sum. In
many cases the players will all gain e.g. cartels in the oil industry. They may all lose e.g. fierce
competition in the matatu industry. Some may gain while others lose, not necessarily by the
same amount.
Chapter Summary
There are many types of decision making
Decision making under uncertainty
This refers to situations where more than one outcome can result from any single decision.
223
STUD Y TEX T
Decision Theory
Decision making under certainty
When only one outcome exists for a decision, we are dealing with this category e.g. linear
programming, transportation assignment and sequencing.
Decision making using prior data
It occurs whenever it is possible to use past experience (prior data) to develop probabilities for
the occurrence of each data.
Decision making without prior data
No past experience exists that can be used to derive outcome probabilities. In this case, the
decision maker uses his/her subjective estimates of probabilities for various outcomes.
Decision making under uncertainty
Several methods are used to make decision in circumstances where only the payoffs are known
and the likelihood of each state of nature are known:
•• Maximin Method
•• Maximax method
•• The Laplace method
•• The Hurwicz method
•• Expected Opportunity Loss (EOL) method
•• The Expected Monetary Value method
•• Minimax regret method
Chapter Quiz
1. …………..theorem recognises that a decision maker does not know with certainty
what state of nature will occur.
2. …………. is also called criterion of rationality.
3. …………. is also called criterion of realism.
4. Payoff is the outcome of playing the game is one of the characteristics of two-persons
zero-sum game.
(a) True
(b) False
5. A n assumption of ……………. is that the gain of a given player exactly equals the loss
of the other player.
6. Which one of the following is not a characteristic of game theory?
(a) E conomic theory
(b) I nformation
(c) Infinite players
(d) Zero-sum of the game
(e) N o indifference
(f) R epeatedness
7. Which principle can be used to reduce the size of the games by eliminating strategies
that will never be played?
2 2 4 quantitative techniques
STUD Y TEX T
Answers to chapter quiz
1. Bayes Theorem
2. Laplace
3. H urcwicz criterion
4. (a) True
5. zero – sum game
6. (c) Infinite players
7. Principle of dominance
questions from previous exams
December 2000 Question 7
Explain the following terms as used in decision analysis.
(a) Decision making under risk versus uncertainty (4 marks)
(b) Decision trees versus probability trees (4 marks)
(c) Minimax versus maximax criterion (4 marks)
(d) Pure strategy versus mixed strategy games (4 marks)
(e) Games with more than two persons versus non-zero-sum games (4 marks)
(Total 20 Marks)
Dec 2001 Question 7
An urban cablevision company is investigating the installation of a cable TV system in urban
areas. The engineering department estimates the cost of the system (in present worth Sh) to be
Sh. 7 million. The sales department has investigated four pricing plans. For each pricing plan, the
marketing division has estimated the revenue per household in present worth Sh. to be:
plan evenue per household (Sh.)
I 150
II 180
III 200
IV 240
The sales department estimates that the number of household subscribers would be
approximately
either 10,000, 20,000, 30,000, 40,000, 50,000 or 60,000.
Required:
(a) C onstruct a payoff table for this problem. (5 marks)
225
STUD Y TEX T
Decision Theory
(b) What would be the company’s optimal decision under the optimistic approach and the
minimax regret approach? (8 marks)
(c) S uppose that the sales department has determined that the number of subscribers will
be a function of the pricing plan. The probability for the pricing plans are given below:
Probability under pricing plan
Number of subscribers I II III IV
10,000 0 0.05 0.10 0.20
20,000 0.05 0.10 0.20 0.25
30,000 0.05 0.20 0.20 0.25
40,000 0.40 0.30 0.20 0.15
50,000 0.30 0.20 0.20 0.10
60,000 0.20 0.15 0.10 0.05
Which pricing plan is optimal?
(d) Briefly explain the main difference between the approaches used in part (b) and (c)
above.
June 2002 Question 7
(a) Define the following terms used game theory:
(i) Dominance (2 marks)
(ii) S addle point (2 marks)
(iii) Mixed strategy (2 marks)
(iv) V alue of the game (2 marks)
(b) C onsider the two-person zero-sum game between players A and B given by the following
payoff table
Player B Strategies
1234
Player A
strategies
1 2 2 3 -1
24326
Required:
(i) U sing the maximin and minimax values, is it possible to determine the value of the
game? Give reasons. (3 marks)
(ii) U se graphical methods to determine optimal mixed strategy for player A and determine
the value of the game. (9 marks)
(Total: 20 marks)
2 2 6 quantitative techniques
STUD Y TEX T
December 2002 Question 7
(a) T he probability distribution showing the number of hours a specially trained accounts
clerk is required to work daily is as follows:
H ours F requency
0 0.4
1 0.3
2 0.2
3 0.1
The accounts clerk is assigned to this work daily based on the expected daily requirement. In
addition, customers pay Sh. 2,500 per hour for this service. The accounts clerk contracted wage
is Sh. 1,500 per hour. If more hours are required than are specified in the contract, the cost is
Sh. 3,000 per hour.
Required:
(i) Determine the hours the accounts clerk should be contracted. (4 marks)
(ii) C alculate the net gain or loss to the company for each possible number of hours if the
contract is written for the expected number of hours. (6 marks)
(b) A student team was planning strategy for a management game being played in one of
their classes. They had to decide whether to price their firm’s products high or low. They
know that their subsequent profits in either case depended on whether the economy
moved up or down – a variable controlled by the instructor. They estimated that a high
price in an upward economy would net Sh. 5 million in profit but a low price would yield
Sh. 3 million profit. If the economy went down, a high price would net Sh. 2 million and
a low price Sh. 1 million in profit.
Required:
Formulate the above information as a game and determine the students’ best strategy.
(4 marks)
(c) A charity is about to launch an appeal for a children’s ward in a national hospital and
has to choose between the following fund raising strategies. Strategy I involves an initial
outlay of Sh. 40,000 and administrative costs of 5 cents for each Sh. 1.00 collected.
Strategy II has no initial outlay but instead involves an extensive door-to-door campaign
with administrative costs of 30cents for each Sh. 1.00 collected. It is estimated that,
whichever strategy is used, the amount collected will be as follows:
A mount (Sh.) 50,000 100,000 150,000 200,000
Probability 0.2 0.4 0.3 0.1
Required:
(i) C onstruct a payoff table (2 marks)
(ii) Determine which strategy the charity should adopt (4 marks)
(Total: 20 marks)
227
STUD Y TEX T
Decision Theory
December 2003 Question 7
a) I n the context of game theory, explain the following statement “winning isn’t everything;
it is the only thing” (3 marks)
b) T he optimal solution of a two-person zero-sum game always represents a saddle point
regardless of whether the players use pure or mixed strategies. Explain.
(5 marks)
c) Joseph Njau and Anne Wairimu can use one strategy some of the time and the other
strategy the rest of the time. Consider the following case of player one (Joseph Njau)
and the opponent (Ann Wairimu). Joseph Njau has two strategies A and B whereas
the opponent, Ann Wairimu has strategies X and Y. Utilities have been assigned as
indicated below:
Ann Wairimu
XY
A 1 -2
Joseph Njau
B -15 2
Required:
Determine the percentage time Joseph Njau plays strategy A and strategy B. Also determine the
percentage time that Ann Wairimu should play strategy X and Strategy Y. (12 marks)
(Total: 20 marks