chapter FOUR

Incorporating Attitudes
Toward Risk

The capital budgeting problem holds a critical place in both theory and
practice of corporate finance. In certain situations, it is a widely accepted
fact that choosing among independent and mutually exclusive investments
based on net present value (NPV) is a consistent approach with owner wealth
maximization. However, an important attribute of real-world decision
making is the risk and uncertainty associated with future outcomes.
In Chapter 3, the concept of expected value, as applied to the eval-
uation of alternative investments, was discussed. The expected value
concept weighs financial consequences by their probabilities, and the
resulting criterion implies the decision maker is impartial to money and
the magnitudes of potential profit or losses. Many analysts consider this
an adequate measure of considering risk. However, risk is not just a
function of the probability distribution of outcomes (reserves or financial
payoffs) but also the magnitude of capital being exposed to the chance of
loss and whether this loss is sustainable by the decision maker. For
example, a person with a total worth of $100,000 may be able to sustain
Project Economics and Decision Analysis
Volume II: Probabilistic Models

a loss of $10,000 or even higher, but a person with a total worth of

$10,000 cannot afford a loss of $10,000. While evaluating competing
investment alternatives with widely varied risk characteristics, it is
important for investors to use a formal and consistent means of risk
policy. In view of this, setting a corporate risk policy is an important
component in managing the evaluation of available investments.1
When the estimated possible loss is high, to the point where a
significant fraction of the capital asset value of the investor is at risk, then
most investors either

• Downgrade the investment because of the total risk threshold that

can be sustained, or
• Share risk by taking less than 100% stake in the investment. This
is done in order to limit potential losses to values below what the
investor considers harmful to its fiscal health.2

Efforts to avoid some of the pitfalls associated with the expected value
concepts lead to the discussion of a fundamental decision science model
known as preference theory, also referred to as expected utility theory. The
theory encompasses the decision makers’ attitude toward risk.
The concepts presented in this chapter are extensions of the expected
value concept, in that the investors’ attitudes about money are incorpo-
rated into a quantitative decision model. The result is a more realistic
measure of value among competing investments characterized by risk and
uncertainty. The preference theory concepts are based on fundamental
and reasonable concepts about rational decision making. This powerful
decision rule can encompass multiple dimensions of value and nonlinear
attitudes toward risk. It can be applied to decision trees using the same
procedure as discussed in the previous chapter.
In this chapter, first the basic theoretical aspects of the expected
utility theory are presented. These are followed by the use of utility
functions and application of the expected utility rule to decision trees.
Once again, PrecisionTree is used to solve decision trees using the
expected utility concepts. In Chapter 5, ways of incorporating risk thresh-
old factors in estimating risk-adjusted value of a project and optimum
participation level of an investor are discussed.

236
 Incorporating Attitudes Toward Risk

THE EXPECTED UTILITY THEORY

The expected value concepts are applicable in many routine,
repetitive decisions where the investor is not exposed to gains or losses
that are not large relative to the investor’s total assets or total worth. In
such situations, it is usually valid to assume a linear preference (risk-
neutral) function. In cases where the stakes are high, relative to the size
of the total portfolio of the investor, real preference functions are often
considered nonlinear. Therefore, it is considered prudent to use the
expected utility (EU) approach rather than the expected monetary value
(EMV) approach discussed in the previous chapter.
In this section, typical risk attitudes of a decision maker, axioms of
utility theory, risk tolerance, risk premium, assessment of utility function,
and mathematical representation of risk preferences are presented.

Typical Attitudes toward Risk

The value of risk is fundamentally a subjective concept. The pleasure
(utility) associated with winning $1000 is generally less than the
displeasure and disappointment of losing the same amount or even less.
Similarly, one derives more pleasure by winning $10 from $30 versus
winning $15 from $1000.
It depends upon such things as the total worth of the decision maker
(individual, department, or an entire company) making the decision. It
determines its security against bankruptcy and the budget level
representing the liquid assets that are at risk of being lost. However, there
is no set rule to measure how risk averse a decision maker ought to be.
Table 4–1 shows outcomes of two projects with their corresponding
probabilities of occurrence. The expected value and standard deviation of
each project is calculated. As evident from Table 4–1, the EMV of Project
A is much higher than the EMV of Project B. Based on the EMV
criterion; investors should select Project A if they are mutually exclusive.
However, the EMV criterion fails to give adequate weight to the decision
maker’s exposure to the chance of a very large financial loss in Project A

237
Project Economics and Decision Analysis
Volume II: Probabilistic Models

(–40 M$) versus Project B (–5 M$). The decision maker knows a loss of
40 M$ with Project A would be much more detrimental to his financial
position. A risk-averse decision maker will obviously choose Project B
with a chance of lower loss. In this situation, the payoffs in monetary units
do not fully reflect the decision maker’s attitude toward risk. Hence, the
comparative payoffs may not be in accord with the decision maker’s
actual risk preferences.

Project A Project B
Probability NPV (M$) Probability NPV (M$)
0.80 80 0.80 30
0.20 -40 0.20 -5
EMV 56 23

Std. Dev. (s) 48 14

Table 4–1 Expected monetary value of two projects

A theoretical utility curve is shown in Figure 4–1. Theoretically, by

analyzing past decisions, a utility curve can be constructed for an indi-
vidual, a department, or an entire company (let’s call it decision maker).
The problem is no one correct utility function can be determined for a
decision maker. It is the one that reflects risk attitudes of previous
decisions. Furthermore, a company’s risk response depends on factors
such as the overall economic climate, changing budgets, management
philosophy, and its fortune over time.

238
 Incorporating Attitudes Toward Risk

Fig. 4–1 Theoretical utility/preference curves

Individuals who avoid risk or are sensitive to risk are called risk-
averse (risk avoiders, risk averters, conservative). Attitude toward risk is
when a decision maker is less likely to choose an alternative with a higher
EMV if it includes proportionately higher risk. The risk-averse behavior
is represented by the utility curve as shown in Figure 4–1, i.e. curved and
opening downward (concave). A decision maker with risk-averse behav-
ior will prefer to invest in a venture having a perceived high chance of
success to a second venture having a low chance of success, even if the
expected value of the second venture is clearly superior.
However, not everyone displays risk-averse behavior all times, and so
utility curves need not be concave for him or her. A convex (opening
upward) utility curve represents risk-seeking (aggressive) behavior. This
attitude is opposite of the attitude of risk-averse decision makers.
Finally, an individual can be risk-neutral. A decision maker who always
selects the alternative with the highest EMV, regardless of the associated
risk, is considered risk-neutral. As shown in Figure 4–1, the diagonal line
in all utility curves represents the risk neutral attitude. Therefore, it is rep-
resented by the EMV and is referred to as the EMV line. The linear relation-
ship means the act that maximizes expected payoffs also maximizes EU.

239
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Therefore, there is no need to use utility numbers for risk-neutral persons.

A risk-neutral decision maker will therefore use EMV for choosing
among alternatives.
Arps and Arps, in an important paper on “prudent risk-taking,” con-
sidered the following two concepts.3

• In successive ventures, the amount of money at risk in each

venture should be less than the amount that would create a
breakeven situation in the longer run. In other words, one should
pay less for a venture than its expected value.
• The amount of money risked on any one venture should not exceed
the amount that would increase the risk of gambler’s ruin beyond
acceptable limits. Arps and Arps defined gambler’s ruin as the
situation where a risk-taker with limited funds goes broke through
a continuous string of failures that exhaust his available funds.

For example, if a decision maker with only $40 million in total

resources invests in Project A of Table 4–1 and the project goes bad, he
will be bankrupt. On the other hand, he can invest the available $40
million in eight projects, each with $5 million cost. This enhances his
chances of meeting the EMV, thus avoiding bankruptcy. Alternatively, the
decision maker may elect to take partial interest in Project A (for
example, 12.5% working interest). By taking partial interest in a venture,
he can reduce his exposure to risk and possible bankruptcy. Chapter 5
presents application of utility theory in arriving at the optimum venture
participation based on the decision maker’s risk tolerance.
Following are some of the important properties of the utility/preference
curves.4

• The vertical scale is dimensionless, representing the relative

desirability of an amount of money. For example, in Figure 4–1,
receiving $1 million is more desirable than receiving $0.5 million
because the utility (0.79) of $1 million is greater than the utility
(0.54) of $0.5 million. The magnitude of the scale is arbitrary and
normally ranges from zero (0) to one (1) as shown in Figure 4–1.

240
 Incorporating Attitudes Toward Risk

However, some curves may represent negative desirability. The

point zero is generally interpreted as the point of indifference. The
positive values show increasing desirability, and negative values
show increasing dislike for the corresponding amounts of money.
• The horizontal axis represents monetary values corresponding to the
various levels of desirability. The monetary values could be NPV,
costs, incremental cash flows, current asset position, CEs, etc.
• The curve is a monotonically increasing function indicating that as
the amount of money increases, the vertical parameter (utility/prefer-
ence) increases in numerical value. It denotes preference of a deci-
sion maker changes as his total wealth or monetary position changes
(getting more is always better).
• The preference values can be multiplied by their probability of
occurrence to arrive at expected preference value or EU for a
decision alternative in the same way as the EMV calculated in the
previous chapter. Mathematically, the EU is given by

(4.1)
N
EU = ∑ pi × U (xi )
i =1

The Axioms of Utility

The mathematical preference theory is based on certain assumptions,
commonly referred to as EU axioms. Mathematical proofs are presented in
literature supporting the following EU axioms. If a decision maker accepts
these axioms as the basis of rational decision, then it is possible his or her
attitudes toward money can be described by a preference/utility curve.

The Transitivity Axiom: According to this axiom, if the decision maker

prefers alternative A to alternative B and alternative B to alternative C, then
he also must prefer alternative A to alternative C (his preference must be

241
Project Economics and Decision Analysis
Volume II: Probabilistic Models

transitive). For example, if a person prefers a Cadillac to Mercedes and

Mercedes to Honda, then he prefers Cadillac to Honda.
The Complete Ordering Axiom: A decision maker is able to order her
preferences or indifference to any two alternatives. For example, if a
decision maker has two alternatives A and B — she either prefers A to B
or she prefers B to A, or she is indifferent between A and B.
The Continuity Axiom: Suppose a decision maker is offered a choice
between two lotteries shown in Figure 4–2a. Lottery 1 offers a no-risk
reward of A, while Lottery 2 offers a reward of B with probability p and
a reward of C with probability of 1 – p. Reward B is preferable over
reward A, and A in turn is preferable over reward C. The continuity axiom
states there must be some value of probability p at which the decision
maker is indifferent between the two lotteries.

Fig. 4–2 Illustration of utility axioms: (a) illustrates the continuity axiom and (b)
illustrates the substitution axiom (after Goodwin and Wright)5

242
 Incorporating Attitudes Toward Risk

The Substitution Axiom: If a decision maker indicates indifference

between the lotteries shown in Figure 4–2a, then according to the
substitution axiom, if reward A appears as a reward in another lottery, it can
always be substituted by Lottery 2 because the decision maker is indifferent
to A and Lottery 2 (meaning both are equally preferable). For example, a
decision maker is indifferent to Lottery 1 and Lottery 2 of Figure 4–2b.
Therefore, according to the substitution axiom, the decision maker will also
be indifferent to Lotteries 3 and 4. These two lotteries are identical except
that in Lottery 4, the $15 of Lottery 3 is replaced by Lottery 2.
Unequal Probability Axiom: If a decision maker prefers reward B to
reward C. Then, according to this axiom, if he is offered two lotteries with
only two outcomes B and C, he will prefer the lottery offering the highest
probability of reward B. For example, in Figure 4–3a the Lottery 1 will be
preferred to Lottery 2.

Fig. 4–3 Illustration of utility axioms: (a) illustrates unequal probability axiom and (b)
illustrates compound lottery axiom (after Goodwin and Wright)5

243
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Compound Lottery Axiom: According to this axiom, a decision maker

will be indifferent to a compound lottery and a simple lottery that offers the
same outcomes with the same probabilities. For example, a decision maker
will be indifferent between the two lotteries shown in Figure 4–3b.
Invariance: All that is needed to determine a decision maker’s
preferences among uncertain events are the payoffs (or consequences)
and the associated probabilities.
Finiteness: No consequences are considered infinitely bad or infi-
nitely good.

Risk Tolerance
In the preference theory approach, the risk tolerance value has a
considerable effect on the valuation of a risky investment. Risk tolerance,
R, is a measure of how much risk a decision maker will tolerate. By
definition, the R-value represents the sum of money at which the decision
makers will be indifferent between a 50:50 chance of winning that sum and
losing half of that sum. The larger the value of R, the less risk averse the
decision maker is. Therefore, a person or company with a large value of R
is more willing to take risks than those with a smaller value of R. The R is
reported in the same units as the monetary value. For example, if revenues
and costs are reported in millions, then R has to be in millions.
The reason for assessing the corporate risk tolerance is one of assess-
ing tradeoffs between potential upside gains versus downside losses. The
decision maker’s attitude about the magnitude of capital being exposed to
the chance of loss is an important component of this analysis.1
A variety of techniques exists for determining R. As defined previously,
R has an intuitive interpretation that makes its assessment relatively easy.
Figure 4–4 provides some insight into the assessment of R in terms of
decision about risky choices. Consider the following gamble

Win $Y with probability of 0.5

Lose $Y/2 with probability of 0.5

244
 Incorporating Attitudes Toward Risk

Fig. 4–4 Assessing risk tolerance (find the largest value of Y at which
alternative A will be preferred)

Would a person be able to take this gamble if Y were 10, 100, 200, or
500? Thinking in terms of investment, how much would a person be
willing to risk ($Y/2) in order to have a 50% chance of tripling the money
(winning $Y and keeping $Y/2)? The decision maker is asked for the
assessment of the value at which her risk becomes intolerable.
The maximum value of Y at which the decision maker would accept
the gamble gives a reasonable estimate of his R. A decision maker, willing
to accept this gamble at only a small value of Y is risk averse, whereas a
decision maker willing to play for larger value of Y is less risk averse.
A decision maker is posed with lotteries of various monetary values
($Y) in order to assess the value to which he will be indifferent. For
example, a decision maker is indifferent at a value of $30 million. This
becomes his risk tolerance level. Any investments costing less than $30
million are accepted, and those higher than $30 million are rejected.
Cozzolino (1977) and Howard (1988) suggest a relationship exists
between certain financial measures (shareholder equity, net income, and
capital budget size, etc.) and the firm’s risk tolerance. Howard suggests
financial statements might be used to develop guidelines for establishing
acceptable R levels, at least in certain industries. Howard suggests certain
guidelines for determining a corporation’s risk tolerance in terms of total
sales, net income, or equity. Reasonable values of R appear to be approxi-
mately 6.4% of total sales, 1.24 times net income, or 15.7% of equity. These

245
Project Economics and Decision Analysis
Volume II: Probabilistic Models

figures are based on observations Howard made in the course of consulting

with various companies.6–7
As a rule of thumb, some companies will take 25% of their total
annual exploration budget as their risk tolerance. For example, if the
annual exploration budget of a firm is $40 million, then it will consider
its risk tolerance level to be approximately $10 million.
Walls and Dyer performed an empirical study to measure the implied
risk tolerance values of the top 50 independent and integrated oil com-
panies from 1981 to 1990. They reconstructed each firm’s annual explo-
ration budget allocations across a set of risky ventures. Based on the
amount each firm was willing to pay for participating in these ventures,
an implied R-value was estimated for each firm in each year. They found
a significant positive relationship between firm size and corporate R, i.e.
the larger the firm the greater was the R value.8

Certainty Equivalent and Risk Premium

The term certainty equivalent (CE) refers to the amount of money
equivalent in a decision maker’s mind to a situation involving uncertainty.
A closely related term, risk premium (RP), refers to the EMV a decision
maker is willing to give up (or pay) in order to avoid a risky decision.
These two concepts are closely linked to the idea of EU. Mathematically,
the RP, EMV, and CE are related by the following equation.

(4.2)
Risk Premium = EMV – CE

The following lottery explains these two concepts.

Win $3000 with probability of 0.50

Lose $200 with probability of 0.50

The EMV of the gamble is $1400. However, if someone wants to buy

the lottery ticket, the owner is willing to sell it for $500. This is a sure
thing and no risk is involved in it. Therefore, the CE of the seller for this
gamble is $500. In the seller’s mind, the lottery is equivalent to a sure

246
 Incorporating Attitudes Toward Risk

amount of $500. If she is unable to sell the lottery at this price then she
would prefer to keep it.
According to Equation (4.2), the RP for this lottery is then equivalent
to $900. This means the owner is willing to give up $900 in expected
value in order to avoid the inherent risk in the lottery. The RP is, therefore,
the premium (lost opportunity) one pays to avoid risk. In any given
situation, the CE, EMV, and RP all depend on the decision maker’s utility
function and the probability distribution of the payoffs.
The EU and the CE can be used for ranking investments. If the CEs
of two investments are same, their EU will also be same and the decision
maker would be indifferent to a choice between the two alternatives. The
alternative with a higher CE is preferred over the alternative with a lower
CE and vice versa. Similarly, the alternative with a higher EU is preferred
over the alternative with a lower EU and vice versa.
For a risk-averse decision maker (concave utility curve) the RP is
positive while for a risk-seeker (convex utility curve) it is negative. The
CE is less than EMV for positive RP, and CE is greater than EMV for
negative RP. The negative RP implies a decision maker would have to be
paid to give up an opportunity to invest.
The following steps are used to calculate the RP of an investment
alternative.

1. Assess the utility function of the decision maker (to be discussed

in the next section).
2. Find the EU of the investment.
3. Find the CE, or the certain amount with the utility value equal to
the EU determined in Step 2.
4. Calculate the EMV of the investment.
5. Subtract the CE from EMV to find the RP.

The higher the RP for an investment, the more risk averse the decision
maker is. A negative premium shows risk-seeker attitude, and a RP of
zero shows risk-neutral attitude.

247
Project Economics and Decision Analysis
Volume II: Probabilistic Models

The CE valuation highlights tradeoffs between potential and uncertain

upside gains versus downside losses with respect to the decision maker’s
risk preference. It also provides the decision makers with a measure of the
amount of expectation they give up by making certain participation choices.
The following example is used to clarify the calculations of RP.

Example 4–1
Use the data of Table 4–1 to calculate the EU, CE, and risk premium
of each project. The utility curve for the decision maker is given in Figure
4–5. Based on the decision maker’s utility function, which project will be
the preferred choice? Use EMV, CE, and EU.

Solution: The utility curve of Figure 4–5, with the payoffs and probabil-
ities given in Table 4–1, are used to solve this example.

Fig. 4–5 Decision maker’s utility curve for the data in Table 4–1 and Example 4–1

248
 Incorporating Attitudes Toward Risk

The following steps are followed to solve the problem.

1. This first step involves converting the dollar payoffs into their corre-
sponding utility values. The utility values for the corresponding
dollar payoffs are read from the utility curve of Figure 4–5 as,

U($80) = 0.8983 U($30) = 0.5756 U(–$5) = –0.1536 U(–$40) = –2.1357

2. The expected utility (EU) for each project is calculated as

EUA = 0.80 0.8983 + 0.20 –2.1357 = 0.2915

EUB = 0.80 0.5756 + 0.20 –0.1536 = 0.4298

3. For EUA = 0.2915 and EUB = 0.4298, the corresponding CE is read from
Figure 4–5. This is done by starting at the vertical axis with the utility
value of 0.2915, reading across to the utility curve, and then dropping
down to the horizontal axis to read the CE. The CEA = 12.06 is shown
by the dotted line in Figure 4–5. Similarly, the CEB is read as 19.66.

4. The EMV of each project is shown in Table 4–1. The expected value
of Project A (56 M$) is greater than the expected value of Project B
(23 M$). Based on EMV criteria, the decision maker will select
Project A. However, based on the EU and CE, Project B is preferred
over project A since EUB > EUA and CEB > CEA. Unlike the expected
value analysis, the EU and certainty equivalent valuations make a
clear distinction between the projects, based on the risk preference of
the decision maker.

5. The risk premium (RP) for each of these projects is calculated as

RPA = EMVA – CEA = $56 – $12.06 = 43.94 M$

RPB = EMVB – CEB = $23 – $19.66 = 3.34 M$

249
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Assessing the Utility Function

A major problem in utilizing preference/utility theory is the assessment
of preference/utility function of the decision maker. Once the utility
function U(x) has been ascertained, finding the optimum investment alter-
native proceeds as before, except utility numbers rather than monetary
payoffs are used. When the utility numbers rather than the monetary payoffs
are used, the preferred act will be the one with the largest expected value of
the outcomes expressed in utility numbers, i.e. the rational decision maker
should select the alternative that will maximize his EU.
The assessment of utility function is a matter of subjective judgment,
just like the assessment of subjective probabilities. The assessment of
utility function involves constructing a mathematical model to represent
the risk preferences (risk attitudes) of the decision maker. This model is
then included in the overall decision analysis process and used to analyze
the situation at hand.
Two approaches, (a) assessment using CEs and (b) assessment using
probabilities, are in common use for the assessment of utility functions.
A disadvantage of the probability approach is that thinking in terms of
probability may be difficult. The most widely used is the certainty equiv-
alence approach, which only requires the decision maker to think in terms
of 50:50 gambles.5
The certainty equivalence method requires the analyst to assess
several CEs. The CEs of the drill versus farm out problem addressed in
Table 3–3 are assessed here. In this problem, a decision maker faces an
uncertain situation. If she chooses to drill the well, the worst case is a dry
hole with –$250,000, and the best case is to have a discovery yielding an
NPV of $500,000. On the other hand, if she farms out the option and the
farmee encounters a producer, her NPV will be $50,000 or she looses
nothing if it is a dry hole. Therefore, there are a variety of options, each
of which leads to some uncertain payoff ranging from a loss of $250,000
to a gain of $500,000. To evaluate the alternatives, the decision maker
must assess her utility for payoffs in this range. The technique presented
here involves eliciting five CE points, thus it is known as the five-point
method of assessing utility function. The following steps are followed.

250
 Incorporating Attitudes Toward Risk

Eliciting Point 1: To start with, identify the worst possible outcome.

In this case, the negative payoff of –$250,000 for the dry hole (worst case)
is taken, and a utility of zero is assigned to it. By assigning zero to the
negative payoff shows negative attitude toward losing money or pre-
ference not to lose money. Therefore,

U(–$250,000) = 0

Eliciting Point 2: Identify a best possible outcome ($500,000 to

encounter a producer). A utility value of one is assigned to it. Therefore,

U(+$500,000) = 1

Eliciting Point 3: Now imagine the decision maker has an oppor-

tunity to play the following lottery, referred to as the reference lottery
or reference gamble.

Win $500,000 with probability 0.5

Lose $250,000 with probability 0.5

The expected value of the previous gamble is $125,000 (0.5 x

$500,000 – 0.5 x $250,000). What is the minimum amount, CE, for which
the decision maker would be willing to sell her opportunity to play this
game? Note a risk-averse decision maker generally trades a gamble for a
sure amount that is less than the expected value of the gamble. Suppose
for this reference gamble, CE of $50,000 is elicited. This means the
decision maker is truly indifferent between $50,000 and the risky gamble.
The utility of this amount must equal the EU of the gamble, which is
calculated as:

U($50,000) = 0.5 U($500,000) + 0.5 U($250,000)

= 0.5 1 + 0.5 0
= 0.5

251
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Eliciting Point 4: Now the CE for a new gamble is elicited, this time
between the U($50,000) calculated in Eliciting Point 3 and the
U($500,000) in Eliciting Point 2.

Win $50,000 with probability of 0.5

Win $500,000 with probability of 0.5

The expected value of the previous gamble is $275,000. Suppose that for
this reference gamble, the elicited CE is $225,000. In this case the utility of
$225,000 is elicited because the decision maker knows that from Eliciting
Point 3 U($50,000) = 0.5 and Eliciting Point 2 U($500,000) = 1.0.

U($225,000) = 0.5 U($50,000) + 0.5 U($500,000)

= 0.5 0.5 + 0.5 1.0
= 0.75

Eliciting Point 5: Now consider the following reference gamble.

Win $50,000 with probability of 0.5

Lose $250,000 with probability of 0.5

The expected value of the previous gamble is –$100,000. For this

reference gamble, the elicited CE is –$100,000. The utility value of this
CE is calculated by using U($50,000) = 0.5 from Eliciting Point 3 and
U(–$250,000) = 0 from Eliciting Point 1.

U(–$100,000)= 0.5 U($50,000) + 0.5 U(–$250,000)

= 0.5 0.5 + 0.5 0
= 0.25

The elicited utility values obtained are now plotted against their cor-
responding CEs. A curve is drawn through the five points.

252
 Incorporating Attitudes Toward Risk

The following five eliciting points are used to construct a utility curve.
The effect of this procedure is to elicit the monetary values in the range of
$0 to $500,000 that have utilities of 0, 0.25, 0.50, 0.75, and 1.0.

Monetary | –$250,000 | –$100,000 | $50,000 | $225,000 | $500,000

Utility | 0.00 | 0.25 | 0.50 | 0.75 | 1.00
The CE values and their corresponding utility functions are plotted as
shown in Figure 4–6. The shape of the curve in Figure 4–6 shows the
decision maker is risk-averse for the range of monetary values in which
the utility is assessed.

Fig. 4–6 Graph of the utility function assessed using the CE approach

The assessments and the graph are checked for consistency. If the
graph is not reasonably smooth, then the assessments are checked, and
some more assessments are made by designing additional gambles. Note
the decision maker’s first response of $50,000 is used in subsequent
lotteries, as both a best and worst outcome. This process is called chain-
ing, and it propagates the very first judgmental error, if any, throughout
the rest of the assessment.

253
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Once the utility functions for the complete range of the monetary
values are assessed and graphed, these are then used in making decisions.
Decision trees can be utilized to solve problems involving utilities. The first
step is to replace the monetary values in a decision tree by their correspon-
ding utility values. If a utility value of a particular monetary value is not
elicited from the five elicited points, it can be read from the utility curve.
The EU for each chance node is then calculated. The alternative with the
highest EU is then selected as a viable option to satisfy the decision maker’s
preference toward risk. The following example clarifies the calculations.

Example 4–2
Rework Example 3–2 (Chapter 3) using the EU criterion. Use the
utility values assessed in the previous section. Based on the EU
maximization, which alternative should be selected?

Solution: The decision tree for the problem is shown in Figure 4–7.

Fig. 4–7 Decision tree for Example 4–2

254
 Incorporating Attitudes Toward Risk

EU (Drill) = 0.65 × U (− $250,000 )+ 0.35 × U ($500,000 )

= 0.65 × 0 + 0.35 × 1

= 0.35

EU (Farm out) = 1.0 × U ($50,000)

= 0.50

Since the EU of the farmout option is higher than the EU of the drill
option [EU(Farmout) > EU(Drill)], the farmout option is selected. The
decision tree calculation procedure is the same as used in Chapter 3. The
only exception is the monetary values are replaced by their corresponding
utility values.

Mathematical Representation of Utility Functions

Utility function is an expression that measures risk by converting
monetary payoffs related to an outcome against their corresponding
utility units. The EU (probability weighted utilities) of one alternative is
then compared to that of another alternative in order to select the
alternative that will maximize the decision maker’s EU.
In a complicated decision problem with numerous possible payoffs, it
might be difficult and time consuming for a decision maker to determine
the EU values corresponding to the possible payoffs from graphical
presentations only. However, if the decision maker is risk-averse or risk-
seeker, then some typical mathematical utility functions can be used as an
approximation of the decision maker’s actual utility function. Mathe-
matical formulas representing the utility functions are as many and as
varied as the individual corporations involved. However, some of the most
common ones are presented in this section.
The utility function curves, such as the one presented in Figure 4–1, are
normally represented by mathematical relationships. Since mathematical
functions give exact forms of expressions, it is usually more accurate and
convenient to fit mathematical functions to utility curves as compared to

255
Project Economics and Decision Analysis
Volume II: Probabilistic Models

drawing the curves by hand. The risk preferences may be represented by a

variety of curves such as exponential, logarithmic, quadratic, linear, power,
and hyperbolic. In this section, some of the most commonly used mathe-
matical relationships representing the utility curves are examined.

Exponential Utility Function: Many variations of the exponential

equations appear in literature, although some of them are mathematically
the same. These various forms of the exponential utility curves are
defined by the following equations.

(4.3)
U (x) = a + be − cx

or (4.3a)
U (x) = 1 − e −x / R

or (4.3b)
U (x) = a + b(1− e −x /R
)
or (4.3c)
U (x) = a − be −x / R

or (4.3d)
U (x) = ae bx

or (4.3e)

U (x ) =
1 (1− e ) −x / R

a
where
x = the dimension of monetary value in currency units
a, b, and c = constants
R = risk tolerance, a constant specified by the decision maker
e = exponential constant (e = 2.71828)

256
 Incorporating Attitudes Toward Risk

The utility curves developed by using Equation (4.3a) are shown in

Figure 4–8. The curve with R = 100 represents a utility function with a
higher degree of risk aversion than the lines represented by R = 200 and
R = 300. This implies the farther from the EMV line the higher the degree
of risk aversion. As R increases, the decision maker becomes more risk
tolerant. At higher values of R, the curve becomes flat. If a constant is
added to a utility curve or if the utility curve is multiplied by a constant,
its properties remain the same.

Fig. 4–8 Representation of exponential utility curves, generated by Equation (4.3a)

As previously mentioned, constructing the utility curve is a subjective

affair. Therefore, use exponential utility when the risk aversion level bounds
the true utility function. Since a majority of the companies are not aware of
their utility functions, bounding or estimating their risk aversion levels may
be a practical approach for deriving the benefits of utility theory.
Not all corporations choose exponential risk aversion, some have
hyperbolic tangent type of risk weighting, others have empirical models
based on prior evaluations of projects and the anticipated value to the
corporate assets, and so on.

257
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Logarithmic Utility Function: The logarithmic utility function is given

by the following equations.

(4.4)
⎛ x⎞
U (x ) = ln⎜1 + ⎟
⎝ R⎠
(4.4a)

U (x) = a log(b + x ) + c
(4.4b)

U (x) = ln(b + x) − c
1
a

Quadratic Utility Function: The quadratic utility function is given by

the following equation.

(4.5)
U (x) = a + bx − cx 2

Linear Plus Exponential Function: This type of utility curve is repre-

sented by the following equation.

(4.6)
U (x) = ax − be −x /R

Power Utility Function: This type of utility curve is represented by

the following equation.

(4.7)
U (x) = a + bx c

258
 Incorporating Attitudes Toward Risk

Hyperbolic Tangent Utility Function: This type of utility function is

given by the following equation. The idea behind the hyperbolic equation
is that there is greater stability in the management of high loss scenarios
than there is with the exponential rule.

(4.8)
⎛x⎞
U (x ) = 1 − tanh ⎜ ⎟
⎝R⎠

Some additional utility curves, representing Equations (4.3d), (4.3e),

(4.4c), and (4.5) are shown in Figure 4–9.

Fig. 4–9 Illustration of utility curves representing Equations (4.3d), (4.3e), (4.4c), and (4.5)

259
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Since the decision maker’s risk preference or utility can be

represented by a mathematical equation, the same equation can be used to
calculate the CE.
CE is the inverse of any equation representing a certain utility curve. For
example the inverse of Equation (4.3 a) and Equation (4.4a) are given by

(4.9)
CE = − R ln[1− U (x)]

(4.10)
⎛ U (x )− c ⎞
CE = Anti log⎜ ⎟−b
⎝ a ⎠

The utility curve presented in Figure 4–5 is generated using Equation

(4.3 a) for R = 35. In addition, the CEA and CEB corresponding to the
EUA and EUB were read from the curve in Figure 4–5 for use in Example
4–1. The same information can be obtained with more accuracy by using
Equation (4.9) and the EUA and EUB from Example 4–1 shown as
follows.
Since EUA = 0.2915 and EUB = 0.4298, then the CEA and CEB using
Equation (4.9) will be

CEA = –Rln[1 – EUA] = –35ln[1 – 0.2915] = 12.0612  12.06 M$

CEB = –Rln[1 – EUB] = –35ln[1 – 0.4298] = 19.6619  19.66 M$

The same values were read from the utility curve in Figure 4–5. The
other equations can be arranged in a similar way in order to calculate the
CE from them.

260
 Incorporating Attitudes Toward Risk

Example 4–3
Rework the data of Table 4–1, assuming the risk tolerance level of the
decision maker is R = 100. Calculate the EU, CE, and risk premium for
each project. Based on these parameters, select the most viable project.
The decision maker’s risk preference can be modeled by exponential
equation of the type 1 – e –x / R.

Solution: First the payoffs of Table 4–1 are converted to their respective
utilities, using the decision maker’s utility curve represented by the equa-
tion 1 – e –x / R.

U ($80 ) = 1 − e −80 / 100 = 1 − e −0.80 = 1 − 0.4493 = 0.5507

U (− $40) = 1 − e 40 / 100 = 1 − e 0.40 = 1 − 1.4918 = −0.4918

U ($30 ) = 1 − e −30 / 100 = 1 − e −0.30 = 1 − 0.7408 = 0.2592

U (− $5) = 1 − e 5 / 100 = 1 − e 0.05 = 1 − 1.0513 = −0.0513

The next step is to calculate the EU of each project.

EU A = 0.80 × 0.5507 + 0.20 × −0.4918 = 0.4406 − 0.0984 = 0.3422

EU B = 0.80 × 0.2592 + 0.20 × −0.0513 = 0.2074 − 0.0103 = 0.1971

Due to the R = 100 versus the R = 35 (used in Example 4–1), the project
preference is changed. Since EUA = 0.3422 > EUB = 0.1971, Project A is
preferred over Project B. The calculations are also shown in Table 4–2.

261
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Project A Project B
NPV Utility NPV Utility
Probability (M$) 1-e-x/100 Probability (M$) 1-e-x/100
0.8 80 0.5507 0.8 30 0.2592
0.2 – 40 – 0.4918 0.2 –5 – 0.0513
EMV & EU 56 0.3422 23 0.1971

Table 4–2 EMV and EU calculations of two projects, Example 4–3

The next step is to calculate the CE of each project, using Equation (4.9) as
CE A = − R ln(1 − EU A ) = −100 × ln (1 − 0.3422) = −100 × −0.4189 = 41.8854

CE B = − R ln(1 − EU B ) = −100 × ln(1 − 0.1971) = −100 × −0.2195 = 21.9525

The certainty equivalent calculations confirm the selection of Project A

over Project B. The last step is now to calculate the risk premium of each
project, using Equation (4.2) as

RPA = EMV A − CE A = 56 − 41.8854 = 14.1146 M$

RPB = EMVB − CE B = 23 − 21.9525 = 1.0475 M$

Approximation to Certainty Equivalent

The examples presented in the preceding sections used exponential
utility function to translate the monetary values of outcomes into their
corresponding utility values. These utility values were then used to cal-
culate the EU. This EU was then fed back into the exponential utility
function (or utility curve) to come up with the certainty equivalent (CE)
in terms of monetary value. This exercise may become tedious when there
are too many outcomes under evaluation.

262
 Incorporating Attitudes Toward Risk

An approximate mathematical correlation is available that can be

used to arrive at the CE. The approximation works well when the
outcome’s probability distribution is represented by normal probability
distribution, i.e. a symmetrical, bell-shaped curve. The following relation-
ship can be used.9–11

(4.11)
2
0.5s
CE ≈ EMV −
R

where s2 is the variance (discussed in Chapter 3) and R is the risk

tolerance. When this is applied to the outcomes of the two investment
alternatives given in Table 4–2, CE is calculated as

0.5[0.80 80 2 + 0.20 (–40) 2 – 56 2 ]

CEA ≈ 56 −
100

0.5 (5,120 + 320 – 3136)

≈ 56 −
100

0.5 2,304
≈ 56 − = 44.48
100

Similarly, CEB = 22.02. The calculated CEA = 41.8854 and CEB = 21.9525
were calculated in Example 4–3. The approximate CE calculated for Project
B is in close agreement with the one calculated in Example 4–3 (22.02 vs.
21.9525). The approximate CE calculated for Project A is not as close to the
one calculated in Example 4–3 (44.48 vs. 41.8854).

Risk Aversion
In the section on Mathematical Representation of Utility Functions,
six different types of mathematical equations and their variations were
presented. Although the shapes of the utility curves generated from these

263
Project Economics and Decision Analysis
Volume II: Probabilistic Models

equations may appear to be quite similar, they signify quite different risk
aversion levels.
To show the important difference between these functions, a term
called risk aversion (RA) is introduced.12

(4.12)
− U ′′(x)
RA =
U ′(x)

where U (x) and U (x) is the second and first derivative of U(x), respectively.
The RA function measures the degree of aversion to uncertainty in a
utility function. The numerator, second derivative of U(x), measures the
curvature (rate of change of slope) of the utility curve. More curvature
means more RA. The EMV line, being a straight linear line, shows an
important concept of the proportionality of decision maker’s risk preference
or utility to the amount of wealth. As previously mentioned, this means that
as the total wealth level of the decision maker increases, he becomes more
risk tolerant. However, this concept may not be generalized.
The RA parameters for some of the equations are derived. In each
case, the expected RA behavior of the decision maker is analyzed. For the
exponential utility function of the form 1 – e–Rx, the RA is

U ′(x) = –(–R)e–Rx = Re–Rx

U ′′(x) = –R 2 e–Rx

– ( –R2e–Rx) R2e–Rx
RA = = =R
Re–Rx Re–Rx

264
 Incorporating Attitudes Toward Risk

Similarly, for the exponential utility function of the form 1/R(1 – e–Rx), the RA is

1
U ′(x) = –(–R)e–Rx = e–Rx
R

U ′′(x) = –Re–Rx

– ( –Re–Rx)
RA = =R
e–Rx

Similarly, the RA for Equation (4.4a) is

1
RA =
b+x

and for Equation (4.5), it is

2c
RA =
b + 2cx

The relationships derived previously show RA is constant for the

exponential utility function, regardless of the total wealth level of the
decision maker. This risk attitude is called constant risk aversion. The
constant risk aversion means that no matter how much wealth a decision
maker has, she will view a particular investment in the same way. The
constant risk aversion may be reasonable for some decision makers, but
most of them might be less risk averse if they had more wealth (the risk
tolerance level increases with increase in wealth). On the other hand, for
the logarithmic and quadratic utility functions the RA decreases as the
level of wealth, x, increases.
The RP profile can also be used to illustrate the constant and decreas-
ing risk aversion properties of exponential and logarithmic utility functions.
It can be shown that the RP for exponential type utility function stays
constant regardless of the increase in wealth while for the logarithmic type
utility the RP decreases as the level of wealth increases.9

265
Project Economics and Decision Analysis
Volume II: Probabilistic Models

If a constant amount is added to all payoffs of an investment and the

decision maker’s preferences show decreasing risk aversion, then the
calculated RP will decrease. On the other hand, if the decision maker’s
preferences show constant risk aversion then the calculated RP will be
constant. Therefore, in the constant risk aversion case the decision maker
will feel the same regardless of increase in his wealth. In the decreasing
risk aversion situation, an increase in the level of wealth will make the
decision maker less risk-averse toward an investment thus resulting in
lower RP. The decreasing risk aversion sounds more logical because, by
nature, a wealthier decision maker can afford to take larger risks.
Example 4–4 shows this concept.

Example 4–4
Consider the payoffs of Project B as shown in Table 4–3. If the decision
maker’s risk preference can be modeled by an exponential utility function
of the form 1 – e–x/R (R = 100), what will be the risk premium if the
decision maker’s wealth is 100 M$, 150 M$, and 200 M$. Repeat the same
problem assuming that the decision maker’s risk preference can be modeled
by a logarithmic utility function of the form 1.73log(74.97 + x) – 3.23.

Project A Project B
Probability NPV (M$) Probability NPV (M$)
0.20 95 0.5 48
0.80 -5 0.5 -18

Table 4–3 Expected monetary value of two projects

Solution: To solve this problem, the existing wealth is added to each

payoff, and the risk premium for the investment is calculated as shown in
Table 4–4. The calculations for the first part are as follows.

266
 Incorporating Attitudes Toward Risk

Wealth Level Outcome A Outcome A

(M$) (M$) (M$)
100 148 82
150 198 132
200 248 182
Probability 0.5 0.5
Risk Tolerance 100
U(x)=1-e-x/100 -100In(1-EU)
Utility, A Utility, B EMV, M$ EU CE, M$ RP, M$
0.7724 0.5596 115 0.6660 109.65 5.3490
0.8619 0.7329 165 0.7974 159.65 5.3490
0.9163 0.8380 215 0.8771 209.65 5.3490

1.73 log (74.97+x)-3.23 10 ( 1.73

)-74.97
EU+3.23

Utility, A Utility, B EMV, M$ EU CE, M$ RP, M$

0.8325 0.5688 115 0.7006 112.11 2.8882
0.9845 0.7765 165 0.8805 162.72 2.2799
1.1109 0.9391 215 1.0250 213.12 1.8839

Table 4–4 Risk premium from exponential & logarithmic functions

The monetary payoffs are first converted into their corresponding utility values

U (A) = 1 − e − (100 + 48 ) / 100 = 0.7724

U (B ) = 1 − e − (100−18 )/ 100 = 0.5596

The EMV and EU for the investment are calculated

EMV = 0.50 × (100 + 48)+ 0.50(100 − 18) = 115 M$

EU = 0.50 × 0.7724 + 0.50 × 0.5596 = 0.6660

267
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Now using the inverse of 1 – e–x/100 , the CE is calculated

CE = −100 ln(1 − EU ) = −100 ln(1 − 0.6660 ) = 109.65

The RP is now calculated as

RP = EMV − CE = 115 − 109.65 = 5.35

The same calculations are repeated, each time adding 150 M$ and
200 M$ to the payoffs. The calculations are shown in Table 4–4, showing
constant risk premium of 5.349 M$.
Similarly, Table 4–4 shows decreasing risk premium when the
logarithmic utility function is used. The detailed calculations are shown in
Excel, Table 4–4.xls on the CD.

Expected Utility Decision Criteria

For all problems, the EMV for each alternative is first calculated. The
alternative with the highest EMV is selected if the decision maker is risk-
neutral. If the decision maker is risk-averse or risk-seeker, then the
investments should be checked to see if the alternatives are sensitive to
changes in risk attitude. If the decision changes due to risk aversion, i.e.,
if the decision is sensitive to risk attitudes (risk-averse or risk-seeker),
then risk attitude should be carefully modeled in the calculations.
The incorporation of risk attitudes involves determining the utility
function for the particular risk attitude and replacing all monetary payoffs
in a decision tree or table by their corresponding utility values. The
alternative with the highest EU or highest expected CE is selected. Both
the highest EU and the highest CE would result in the same decision.
However, since the CE is calculated in monetary values, it is easy to
understand and explain.

268
 Incorporating Attitudes Toward Risk

SPREADSHEET APPLICATIONS
In this section, Excel is used to fit the mathematical equations (pre-
sented earlier in this chapter) to an elicited utility curve and calculate
critical risk tolerance level. PrecisionTree is used to solve decision
problems while incorporating risk attitudes.

Fitting the Utility Curve

In this section, the mathematical equations presented earlier are fitted
to the five points elicited in Figure 4–6 to determine the equation that best
represents the utility curve. The SOLVER option of Excel is used to
perform this task. This should be the standard way of selecting the
particular mathematical equation rather than arbitrarily choosing the
exponential equation as advocated in most literature on the subject. The
following five utility points were generated in the section on Assessing
the Utility Function.

Monetary | –$250,000 | –$100,000 | $50,000 | $225,000 | $500,000

Utility | 0.00 | 0.25 | 0.50 | 0.75 | 1.00
The following systematic procedure is used to fit the quadratic
equation, U(x) = a + bx – cx2, to the data. All other equations are fitted,
not shown here, to the same data as shown in Table 4–5 on the accom-
panied CD. Look at Table 4–5 on the CD to find out why the quadratic
equation has been chosen and how the other equations represent the same
utility curve.

269
Project Economics and Decision Analysis
Volume II: Probabilistic Models

A B C D
2 CE Elicited Predicted
3 (M$) Utility U(x) = a+bx-cx2 SSE
4 – 250 0.00 – 30.7500 945.5625
5 – 50 0.25 – 10.5000 115.5625
6 100 0.50 5.2500 22.5625
7 225 0.75 17.9375 295.4102
8 500 1.00 25.5000 600.2500
9 1,979.3477
10 0.5000
11 0.1000
12 0.0001

Table 4–5 Fitting a utility curve with mathematical equation

Step 1: Enter the elicited CEs in Cells A4:A8 with the corresponding
utilities in Cells B4:B8. The data is arranged as shown in Table 4–5.
Step 2: Enter arbitrary values of a, b, and c in Cells C10, C11, and C12,
respectively. These (a, b, and c) are the three constants of the quadratic
equation.
Step 3: Input the quadratic formula: =C$10 + C$11*A4 – C$12*A4^2 in
Cell C4, to calculate the predicted utility. Copying this formula to the
range C5 to C8 generates predicted utility for the remaining four points.
Step 4: Enter the formula =(B4 – C4)^2 in Cell D4 to calculate the squared
error for the first point. Copying this formula to Cells D5 to D8 computes
the squared error for the other four points on the utility curve. In Column
D, the squared errors for each point on the utility curve are calculated by
squaring the difference between the actual and predicted utility.

270
 Incorporating Attitudes Toward Risk

Step 5: Enter the formula =SUM(D4:D8) in Cell D9 to calculate the sum

of squared errors of all five points.
Step 6: Use SOLVER to find the constants a, b, and c that would best
fit the curve, i.e. choose the values of a, b, and c that will minimize the
sum of squared errors in Cell D9. The following steps are followed for
using SOLVER.

a. Click on the Tools in the Excel toolbar and then click on Solver
in the dropdown menu. If the Solver command is not on the Tools
menu, install the Solver add-in.
b. Enter the cell reference of the target cell $D$9 in the Set Target
Cell box,
c. Click on Min, to have the value of the target cell as small (remem-
ber the objective is to minimize the sum of squared error) as possible.
d. Enter the reference for each adjustable cell: $C$10:$C$12 in the
By Changing Cells box.
e. Click on Solve. As soon as this is done, iterations start and Cell
D9, Cells C4 to C8, and Cells C10 to C12 will be revised with the
final numbers accordingly. The final equation fitted to the utility
curve is now

U(x) = 0.4278 + 0.0015x – 0x2

The final table will look like the one shown in Table 4–6. The fitted
data is plotted on the actual data as shown in Figure 4–10, which shows a
perfect fit. In order to select the equation that best represents the elicited
utility curve, the previous procedure is repeated for all the equations, and
the equation that gives the lowest SSE is chosen for further calculations.

271
Project Economics and Decision Analysis
Volume II: Probabilistic Models

A B C D
2 CE Elicited Predicted
3 (M$) Utility U(x) = a+bx-cx2 SSE
4 – 250 0.00 – 0.0063 0.0000
5 – 50 0.25 0.2658 0.0003
6 100 0.50 0.5031 0.0000
7 225 0.75 0.7359 0.0002
8 500 1.00 1.0060 0.0000
9 0.0005
10 0.4279
11 0.0015
12 0.0000

Table 4–6 Fitting a utility curve with mathematical equation

Fig. 4–10 Illustration of the actual and fitted utility curves

272
 Incorporating Attitudes Toward Risk

Finding the Certainty Equivalent

As presented in the preceding pages, certainty equivalent (CE) can
either be approximated by using Equation (4.11), or it can be back
calculated from the inverse of the utility equation representing the
decision maker’s risk preference. The inverse of one of the exponential
utility equations and one of the logarithmic equations was derived and
presented in Equation (4.9) and Equation (4.10).
However, the inverse of some equations may not be easily derived, or
it may require iterative solution. In addition, once the EU of the desired
alternative (maximum EU) is arrived at, it will be desirable to determine
the corresponding monetary value of the EU. The Goal Seek option of
Excel can be used to perform this function. The following steps are
followed to calculate the CE of EU = 0.6660 in Table 4–4 using expo-
nential utility equation of the type 1 – e–x/100 .

Step 1: Enter the EU for which the CE be calculated in Cell A2.

Step 2: Input a trial value of the trial CE is in Cell B2.
Step 3: Enter the utility equation used to calculate the EU in Cell C2,
while replacing the x in the equation by the Cell reference B2. In this case,
it is = 1 – exp ( –B2 / 100). The cells will look like this.

A B C
1 EU CE U(x)
2 0.6660 50.0 0.3935

Step 4: Click on the Tools on the Excel toolbar and then click on Goal
Seek in the dropdown menu.

273
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Step 5: Specify (a) the Set cell as C2, (b) input 0.6660 in the To value
window, and (c) specify the By changing cell as $B$2 in the Goal Seek
dropdown menu (following Step 4).
Step 6: Click on OK. This will revise Cells B2 and C2 by calculating the
CE and U(x), respectively. The final output will look like this.

A B C
1 EU CE U(x)
2 0.6660 109.661 0.6660

In the same way, CE for Project B (Table 4–2) with EU of 0.1971 is

calculated to be 21.877 M$ (R = 100). This payoff is less than the EMV
of 23 M$ (Table 4–2). Since the utility function exhibits risk aversion, it
is natural the project is valued at an amount less than its expected NPV
(CEB = 21.877 < EMVB = 23).

Critical Risk Tolerance

Critical risk tolerance is the risk tolerance at which the EU of the
alternatives under consideration are equal. For example, in Example 4–3
(Table 4–2) the EU of Project A and Project B was calculated. Project B
was selected because EUA = 0.3422 > EUB = 0.1971 at R = 100 and using
exponential utility equation of the type 1 – e–x/R . The same problem was
solved in Example 4–1, using R = 35. In Example 4–1, Project B was
preferred over Project A (EUB = 0.4298 > EUA = 0.2915). This shows the
risk tolerance at which the decision maker will be indifferent to the two
projects lies between R = 35 and R = 100. Therefore, the critical risk
tolerance for this example is the R at which the EUA = EUB.
The critical risk tolerance can be used for sensitivity analysis
(whether the decision maker’s risk tolerance is above or below the critical
risk tolerance level). If the decision maker’s risk tolerance is above the

274
 Incorporating Attitudes Toward Risk

critical value then she should go for the riskier project; if it is less, then
she should select the less risky project. The critical risk tolerance for the
two projects is calculated as 43.78. Therefore, Project A is selected if the
decision maker’s R > 43.78 and Project B is selected if R < 43.78.
The simplest way to determine the critical risk tolerance is to use
computer in searching for the R-value that will make the EU of the two
projects equal. Once again, the SOLVER or GOAL SEEK options of
Excel can be used to achieve this objective. The following steps and
SOLVER are used here.

Step 1: Input the outcome probability of Project A in Cells A5 and A6 and

the corresponding payoffs in Cells B5 and B6. Similarly, input the outcome
probability of Project B in Cells D5 and D6 with the corresponding payoffs
in Cells E5 and E6. The data is arranged as shown in Table 4–7.
Step 2: Enter a trial value of critical risk tolerance in Cell B9.

A B C D E F
2 Project A Project B
3 NPV Utility NPV Utility
-x/R -x/R
4 Probability (M$) U(x) = 1-e Probability (M$) U(x) = 1-e
5 0.80 80 0.8329 0.80 30 0.4960
6 0.20 – 40 – 1.4935 0.20 –5 – 0.1210
7 EMV & EU 56 0.3726 23 0.3726
8
9 Critical RT 43.78
10 SSE 1.722E – 19

Table 4–7 Calculation of critical risk tolerance

275
Project Economics and Decision Analysis
Volume II: Probabilistic Models

Step 3: Input equation = 1 – exp(–B5 / B$9) in Cell C5. Copy Cell C5 to

Cell C6. Similarly, input equation: = 1 – exp(–E5 / B$9) in Cell F5 and
copy it to Cell F6.
Step 4: Input equation: =SUMPRODUCT(A5:A6,B5:B6) in Cell B7 and
equation: =SUMPRODUCT(A5:A6,C5:C6) in Cell C7. Note the EMV
and EU of Project A are calculated in Cells B7 and C7, respectively.
Similarly, input equation: =SUMPRODUCT(D5:D6,E5:E6) in Cell E7
and equation: =SUMPRODUCT(D5:D6,F5,F6) in Cell F7.
Step 5: Enter equation =(C7 – F7)^2 in Cell B10. This calculates the
critical risk tolerance by minimizing the sum of squared errors between
the EUA and EUB.
Step 6: Use SOLVER to find the R that would make EUA = EUB. The
following steps are followed for using the SOLVER.

a. Click on the Tools in the Excel toolbar and then click on Solver
in the dropdown menu. If the Solver command is not on the Tools
menu, the Solver add-in should be installed.
b. Enter the cell reference of the target cell $B$10 in the Set Target
Cell box.
c. Click on Min to have the value of the target cell as small as
possible.
d. Enter $B$9 in the By Changing Cells box.
e. Click on Solve. As soon as this is done, iterations start and Cell
C7, Cell F7, Cell B9, and Cell B10 will be revised with the final
numbers accordingly. As shown in Table 4–7, the EU of Project
A and B are equal at R = 43.78.

PrecisionTree and Utility Functions

PrecisionTree was introduced in Chapter 3 for solving decision trees.
However, in Chapter 3 it was used for decision analysis based on expected
value concepts. In this chapter, PrecisionTree is used again to solve
problems based on the concepts of CE and EU theory. PrecisionTree has

276
 Incorporating Attitudes Toward Risk

made it simple to incorporate CE or utility in decision trees. The decision

trees developed in Chapter 3 can be easily converted to expected value,
CE or EU basis (desired type of analysis can be easily switched from one
version to the other and vice versa).
PrecisionTree includes built-in exponential (1 – e–x/R ) and loga-
rithmic (ln + R) utility functions. User-defined utility functions can be
added using Excel’s built-in programming language, VBA (Visual Basic
for Applications). Once a utility function is selected, the monetary payoffs
of the decision tree are automatically converted to the desired option
(utility or CE). Optimum paths in a decision tree are selected based on
whether EU or CE is opted for.
The decision tree is first developed in exactly the same way as shown
in Chapter 3. To apply a utility function to a decision tree’s calculations,
the following steps are followed.

1. Click on the Branch name to open the Tree Settings dialog box
as shown in Figure 4–11.

Fig. 4–11 PrecisionTree’s Tree Settings dialog box

277
Project Economics and Decision Analysis
Volume II: Probabilistic Models

2. Change the settings as desired, for example

a. Click the check box Use Utility Function. Leaving this box
blank will make the PrecisionTree perform expected value
calculations. When we click on it, a tick mark will appear in
the box.
b. Select Function (exponential or logarithmic).
c. Input the risk tolerance R.
d. Select Display (EU or CE).
e. Click on OK to start calculations.

PrecisionTree allows defining a different utility function for every

chance node in the tree. The utility function of any node can be changed
at any time during the modeling process. By specifying R = 0 or by setting
decision model to Expected Value will make PrecisionTree assume risk
neutral decision behavior, thus it bases its decisions accordingly.

Example 4–5
Using PrecisionTree, construct a decision tree for the data in Example
4–3 (Table 4–3). As a first pass, calculations are based on expected value
criteria. The same tree is now used with certainty equivalent criteria.
Assume exponential utility function and R = 100.

Solution: The decision tree with monetary payoffs is shown in Figure

4–12. EMV of $15 is calculated for both projects. The decision tree with
certainty equivalents is shown in Figure 4–13. Based on certainty
equivalent criteria, Project B is selected. The same tree can be used with
EU criteria.

278
 Incorporating Attitudes Toward Risk

Fig. 4–12 Decision tree for Example 4–5, showing expected value calculations

Fig. 4–13 Decision tree for Example 4–5, showing certainty equivalent calculations

279
Project Economics and Decision Analysis
Volume II: Probabilistic Models

QUESTIONS and PROBLEMS

4.1 Briefly describe various risk attitudes. Why is it important to
incorporate risk attitudes in investment decision making?

4.2 What is the difference between expected value and EU?

4.3 What is meant by (a) CE, (b) RP, and (c) risk tolerance?

4.4 Do you agree with the statement “the more wealth one has, the
easier it is to take larger risks”?

4.5 Explain the concept of RA, its measure, and how it relates to utility
function.

4.6 Is assessing a utility function of an investor a subjective or objec-

tive process?

4.7 A decision maker’s risk tolerance is $1300 and his risk preference
can be modeled with an exponential utility function of the type
U(x) = 1 – e–x/R .

a. What will be U($1050), U($,750), U(0), and U(–$1200).

b. Calculate the EU for the payoffs in 4.7a if their probability
distribution is 0.35, 0.30, 0.20, and 0.15.
c. Calculate the exact and approximate CEs of the investment.
For each of these, calculate the RP.
d. What will be the approximate CE for investment if its expected
value is $2575 and its standard deviation is $250?

4.8 The best to worst monetary returns for a project are $50,000,
$35,000, $15,000, and –$12,000. If we assign a utility of 1 to
$50,000 and a utility of 0 to –$12,000, assess the utilities of the
remaining two monetary returns.

280
 Incorporating Attitudes Toward Risk

4.9 A utility function for an investor is given as follows.

Payoff, MM$ Utility

12.0 1.00
5.8 0.75
3.5 0.60
2.2 0.45
0.0 0.00

a. Graph the utility function. Based on this graph, how would you
classify the investor’s risk preference?
b. Fit an exponential utility function of the type 1 – e–x/R to this
risk attitude. What is the value of risk tolerance?
c. Also fit logarithmic and quadratic utility functions to this risk
attitude.
d. For each of these utility functions as fitted, which alternative is
preferable and why?
e. Many analysts claim there is insignificant difference between
the exponential utility function and other utility functions;
therefore, exponential utility function can be used in all cases.
Do you agree with this claim?

4.10 Rework Problem 3.10 (Chapter 3) if the investor’s risk preference

can be modeled by an exponential utility function 1 – e–x/R when R
= 50 M$. Calculate the EU, RP and CE for each investment.

4.11 Rework Problem 4.10 to calculate the critical risk tolerance

4.12 Rework Problem 4.10 using PrecisionTree.

4.13 Rework Problem 3.14 (Chapter 3) if the investor’s risk preference

can be modeled by an exponential utility function 1 – e–x/R when
R = 50 M$. Does your answer change if the R value is increased
to R = 75 M$?

281
Project Economics and Decision Analysis
Volume II: Probabilistic Models

REFERENCES
1 Walls, Michael, R., “Corporate Risk Tolerance and Capital
Allocation: A Practical Approach to Implementing an
Exploration Risk Policy,” Journal of Petroleum Technology,
SPE, April 1995, p. 307–311.
2 Lerche, I. and MacKay, J. A., “Energy Exploration &
Exploitation,” Multi-Science Publishing Co. Ltd., 107 High
Street, Brentwood, Essex CM14 4RX, England, Volume 14,
No.2 1996.
3 Arps, J. J. and Arps, J. L, “Prudent Risk-Taking,” Journal
of Petroleum Technology, SPE, v. 26, 1974, p. 711–715.
4 Newendorp, P. D., Decision Analysis for Petroleum
Exploration, PennWell Books, 1421 South Sheridan Road,
Tulsa, OK 74101, USA, 1975.
—— Newendorp, P. D. and Schuyler, J., Decision
Analysis for Petroleum Exploration, 2nd Edition, Planning
Press, 15492 East Chenango Avenue, Aurora, CO 80015,
USA, 2000.
5 Goodwin, P. and Wright, G., Decision Analysis for Man-
agement Judgment, John Wiley & Sons Ltd., Baffins Lane,
Chichester, West Sussex, PO 19 1UD, England, 1991.
6 Cozzolino, J., “A Simplified Utility Framework for the
Analysis of Financial Risk,” SPE paper 6359 presented
at the SPE Hydrocarbon Economics and Evaluation
Symposium, Dallas, Texas, Feb. 20–22, 1977.
7 Howard, R. A., “Decision Analysis: Practice and Promise,”
Management Science, 34, No. 6, 679, 1988.
8 Walls, M. R. and Dyer, J., “Risk Propensity and Firm Per-
formance: A Study of the Petroleum Exploration Indus-
try,” paper 92S–113 presented at the Operations Research
Society of American Institute of Management Science
Joint National Meeting, San Francisco, Nov. 2, 1992.

282
 Incorporating Attitudes Toward Risk

9 Clemen, R. T., Making Hard Decisions: Introduction to

Decision Analysis, 2nd Edition, Wadsworth Publishing Co.,
10 Davis Drive, Belmont, CA 94002, USA, 1996.
10 Pratt, J., “Risk Aversion in the Small and in the Large.”
Econometrica, 32, 1964, pp. 122–136.
11 McNamee, P. and Celona, J., Decision Analysis for the
Professional with Supertree, Scientific Press, Redwood
City, CA, USA, 1987.
12 Samson, D., Managerial Decision Analysis, Richard D.
Irwin, Inc., 1988.

283