Chapter 2

The choice of investments

in an uncertain future

Objectives
After reading this chapter, the student should be able to:

• To understand the difference between the certain environment, the environment

random and the uncertain environment.

• To apply risk consideration techniques in the evaluation of

investment projects:
the reduction of the project's lifespan
the method of certain equivalent

the addition of a risk premium to the discount rate

• To use the probabilistic approach in the selection of investment projects.

• To build decision trees for project comparisons.

• To compare investment projects using the following criteria:

Laplace criterion
Wald Criterion
Maxi Max Criterion
Hurwicz Criterion
Savage criterion
 Chapter 2
The choice of investments
uncertain future

All investments are risky. How to assess the risk related to the decision
to invest? . How to define the risk premium that the investor will require? .
Traditional finance relies on the concept of time. Modern finance has
introduces another dimension into decision-making: risk.
Until now, the risk has only been addressed implicitly, notably through:
To the method of the payback period of the invested capital since indirectly, we
consider that the risk increases over time,
At the choice of a discount rate that includes a risk premium.
I- The investment decision in uncertain futures

1- Empirical methods for considering risk

In the first chapter of this course, we implicitly considered that the
future revenues and expenses were certain. The net present value calculated based on
Cash flow was therefore certain as well.

In reality, future events are random (or probabilistic, meaning that we can
assign them a probability) or even uncertain (that is to say, their probabilities are
unknowns, which we will study in the second section).

The following empirical processes traditionally allow for consideration of

risk in the calculation of the net present value:

The reduction of the project's lifespan,

The method of certain equivalent,

The addition of a risk premium to the discount rate.

 a- The reduction of the project's lifespan
This method consists of updating the cash flows generated by the project on a
duration less than the lifespan of the project. The riskier the project, the more the number of years
The neglected in the calculation is great.

n - p CFN
The risk-adjusted NPV = - I0+ ∑ t
t = 1 (1+ i ) t

With n The lifespan of the project.

p Number of periods neglected to account for risk.
I0 Initial investment.
CFN t : Cash-Flow de la période t.
Discount rate.
b- The certainty equivalent method:
This method involves replacing risky revenues in the calculations with
certain fictional revenues. These certain revenues are lower than the risky revenues. They
are supposed to be equivalent to risky revenues from the decision-maker's point of view.

Equivalent certain = Risky recipe x coefficient.

In the NPV equation, it is a matter of adjusting the net cash flows.

n α CFN
Adjusted VAN =− I +
0 ∑ t tt
t =1
(1 + i)

With n The lifespan of the project.

I0 Initial investment.
CFNtCash flow for the period t.
i Discount rate.

α t is the adjustment coefficient of period t. For all investors with an attitude

risk aversion (when the investor prefers a certain net present value) is
ranging between 0 and 1. It is weaker for a cautious decision-maker than for a decision-maker
audacious.
 c-Addition of a risk premium to the discount rate

This method consists of increasing the discount rate by a premium to reduce it.
the discounted cash flows of risky projects;
Discount rate = risk-free rate + risk premium

n CFN
Adjusted VAN =-I + ∑ t
0
t=1 (1+ p ) t
With n The lifespan of the project.
I0 Initial investment.
CFN t Cash flow for period t.
i Discount rate.

And p = risk-free discount rate + a risk premium.

2 - The probabilistic approach

a- Reminder of statistical mathematics:
Let there be two random variables X1and X2. a and b represent two real coefficients.

E ( a. X) = a. E ( X )

E(a.X1 + b.X2) = a.E(X1) + b.E(X2)

V(a .X) = a2.V(X) .

V ( a. Xl+ b. X2) = a2V ( Xl) +b2. V ( X2) + [Link] ( X1,X2)

= a2. V ( Xl) + b2. V( X2) + 2ab.rδ( Xl )δ( X2 )

r is the correlation coefficient between the two variables.

Particular cases:

If the two random variables Xl and X2are perfectly independent ( r = 0 ) :

V ( a. Xl+ b. X2) = a2. V ( Xl) + b2. V ( X2) .

If the two random variables Xland X2are perfectly correlated ( r = 1 ):

V ( a. Xl+ b. X2) = a2. V ( Xl ) + b2. V( X2) + 2ab.δ( Xl)δ( X2)

 = a2.δ2( Xl) + b2.δ2(X2) + 2ab.δ( Xl)δ( X2 )
=(aδ(X1)+b.δ(X2) )2

δ(a.X1+b.X2) = aδ(X1) + bδ(X2)

b- Arbitrage between expected value and variance

The net present value calculated based on random flows is itself a
random variable characterized by its expected value E (RV) and its variance V
(VAN) or its standard deviation δ(VAN).

The decision-maker seeks to maximize the expected value of NPV and to minimize the
variance of the NPV (synonym for risk).

Example 23:

Let us consider the following projects for which we know E (NPV) and V (NPV).

project E(VAN) V(VAN)

A 12000 9000000

B 15000 7840000

C 17000 8410000
Project A is dominated by projects B and C, which are better from both perspectives.
from the mathematical expectation only from the perspective of risk. Project A will never be preferred.
to projects B or C.
Le projet C est meilleur que le projet B au point de vue de l'espérance ainsi le projet B
takes it for its lower risk. The decision-maker chooses between projects B and C according to his
psychology: the cautious chooses project B and the daring chooses project C.
The investment decision is based on an arbitration between expectation.
mathematics and the variance of NPV.
c- Mathematical expectation of the NPV Let's denote by:
Let us designate by: CFN1CFN2,CFN3the net cash flows for years 1, 2, 3
the cost of capital

I0the invested capital.

E (VAN) = E [ CFN1(1 + i)-1+ CFN2(l + i)-2+ CFN3(1 + i)-3-I0]

Where CFN1, CFN2, CFN3are the random variables Xl, X2and X3.
(1 + 1)-l(1 + 1)-2(1+ 1)-3are the coefficients.

Therefore:

E(VAN) = -I0(1 + i)-1.E (CFN1) +(1 + i)-2.E(CFN2(1 + i)-3.E( CFN3).

The net present value is a random variable (combination of variables

random).
If these variables are independent, the mathematical expectation of the NPV is equal to
the VAN of mathematical expectations.
The mathematical expectation of the NPV constitutes:

- A rejection criterion: any project whose mathematical expectation of the NPV is will be rejected.
negative.
-A selection criterion: between two projects, the one with the higher expected value will be chosen.
the NPV is the highest.
Example 24
An investment of 80 million dinars is expected to generate operating cash flows.
following:

Probabilités Année1 Année2 Année3

Pessimistic hypothesis 0.3 20 25 30

Average hypothesis 0.5 30 40 50

Optimistic hypothesis 0.2 40 50 60

The cost of capital is 10%.

Calculate the expected value of the net present value

Year1
CFNt Pt Pt. CFNt
20 0.3 6
30 0.5 15
40 0.2 8
E(CFN1)=29

Year2
CFNt Pt Pt. CFNt
25 0.3 7.5
40 0.5 20
50 0.2 10
E(CFN2=37.5

Year3
CFNt Pt Pt. CFNt
30 0.3 9
50 0.5 25
60 0.2 12
E(CFN3=46

E(VAN) = -80 + 29 (1 + 0.1)-1+ 37.5 (1+ 0.1)-2+ 46(1 + 0.1)-311.9 million dinars.
d- Variance of the NPV
The criterion of mathematical expectation does not take into account the dispersion of
current net values, and therefore the risk associated with the distribution of probabilities. The more the
the distribution is spread out, the greater the risk. The calculation of the variance or the standard deviation,

indicator of central values, will measure the risk of the distribution.

d.1-Application to time-independent flows

If the CFN1, CFN2, CFN3are mutually independent in time then:

V(VAN) = V(CFN)1(1 + i)-2+ V(CFN2)(1 + i)-4+V( CFN3(1+i)-6With V(Io) = 0

If the net cash flows are constant over time then:

The variance or the standard deviation of the NPV constitutes:

- A rejection criterion: any project whose risk exceeds the established standard will be rejected.

One selection criterion: between two projects, the one with the highest expectation will be chosen.

mathematics is the highest, and the risk is the lowest.

Example 25:

Let's take the example above assuming that the cash flows of a
The years are independent of the flows of other years.
Year1
CFNt Pt Pt. CFN2t
20 0.3 120
30 0.5 450
40 0.2 320
E(CFN1=49

Year2
CFNt Pt Pt. CFN2t
25 0.3 187.5
40 0.5 800
50 0.2 500
E(CFN1=81.25

Year 3
CFNt Pt Pt. CFN2t
30 0.3 270
50 0.5 1250
60 0.2 720
E(CFN1)=124

Calculation of the variance of the NPV:

V(VAN)=49(1+i)-2+ 81.25(1+i)-4+124(1+i)-6166
δ(VAN)= 166 =12.9
d.2-Application to perfectly correlated time flows
If the CFNs1, CFN2CFN3are perfectly correlated then
δ(VAN)=δ(CFN1)(1+i)-1+δ(CFN2(1+i)-2+δ(CFN3(1+i)-3
Example 26:
Taking the previous example assuming that the cash flows of one year
are perfectly correlated with the flows of other years:
Calculation of δ(CFN1),δ(CFN2)setC(FN3)

V(CFN1)=49δ(CFN1)=7
V(CFN2)=81.2δ(CFN2)=9
V(CFN3)=124δ(CFN3=11.1
V(VAN)=490
δ(VAN)=7(1+0.1)-1+9(1+0.1)-2+11.1(1+0.1)-3=22.4= 490

It is noteworthy that, generally, the higher the expected value of the NPV,
the greater the risk increases. The investor will then have to make adjustments favoring the
profitability with a high level of risk, or with a lower risk and more profitability
low. (Returns to example 23)
To compare the risks of two projects, it is relevant to calculate the reduced gap or
the coefficient of variation of the NPV which is equal to Standard deviation / Expected value.
The last criterion allows for measuring absolute risk. The lower this ratio, the lower the risk.
is high. (see example 23).

In the previous example, the absolute risk is equal to 12.9 = 1.08 ( case of
11.9

the independence of cash flows) and 22.14

to = 1.86 ( case of perfect correlation between flows ).
11.9
The risk coefficient in the case of perfect correlation is higher. In this case, the project
is more risky.
e- Limitation of the probabilistic approach:
This method is only applicable when the probability distributions of cash-
flows follow a normal law.

3- Sequential decisions and decision trees:

a- Definition :
Lorsque le projet implique plusieurs décisions d'investissement qui se succèdent dans
time, we represent all decisions and events by a decision tree.
A decision is a choice made freely by the decision-maker.
Un évènement est imposé de l'extérieur au décideur. Il est affecté d'une probabilité.
b- Structure of decision trees:
The decision tree is a directed graph that represents the succession of decisions.
and events. Among the vertices of the graph (or nodes), there are distinguished nodes of
decisions and event nodes.
b.1-The decision nodes
A decision node represents a choice between several decisions. It is represented by
a square. Each decision leads to an event node. The root of the decision tree is
always a decision node.
Décision A

Décision B

b.2-Event nodes
An event node represents an alternative among several events. It is
shaped by a circle.
Each event is associated with a NPV and a probability. The sum of
probabilities assigned to the events of a node equal to 1. For each node, we calculate
the expected value of the NPV and possibly, the variance.

. 1
Event I, probability 1
VAN1
P1.VAN1
Evénement 2, prababilité2
. VAN2 E(VAN)= P2.VAN2
Evénement 3, prababilité3
P3.VAN3
VAN3

b. 3- Procedure
At each decision node, the preferred decision is the one that leads to
event node for which the expected value of the NPV is maximum.
To calculate the expected value of the NPV associated with a decision, it is necessary to
only take into account the probabilities of events occurring after the decision. Indeed,
a decision is only conditioned by future events. It cannot change the
past.
 The calculation is done by going back in time from the end to the beginning. The tree is
progressively modified by eliminating, at each decision node, the decision branches
dominated.

Example 27:
The company Canon has the choice between three solutions to increase its capacity.
production
First project: Expansion of current production units.
Initially, she is considering a low investment of 800, which will
would increase its production capacity by 25%, thanks to productivity gains.
Second project: Construction of a new factory. This investment amounts to 1300
it would allow him to increase his production capacity by 40%.
Third project; do nothing.

The forecast horizon corresponds to the economic lifespan of the projects, namely
ten years. The discount rate used is equal to 10%.

Probability of expansion over 10 years: 50%

Probability of stagnation over 10 years: 40%

Probability of recession over 10 years: 10%.

First project Second project

Annual cash flow Annual cash flow
Expansion 190 360
Stagnation 85 120
Recession -2 -10
Decision tree
E1Expansion (Pr=50%)

E2stagnation(Pr=40%)
E3Recession (Pr=10%)

I=800

E1Expansion (Pr=50%)
D0 E2stagnation (Pr=40%)
I=1300
E3Recession (Pr=10%)

E1Expansion (Pr=50%)
I=0
E2stagnation(Pr=40%)
E3Recession (Pr=10%)

For the investment I=800

Event Cash flow VAN

Expansion −10
1 − (1.1 )
190 × 1167
0.1
Stagnation −10
1 − (1.1 )
85 × =522 -800+50%1167+40%522+10%(-12)=-8.9
0.1
Recession −10
1 − (1.1 )
-2 × -12
0.1
For the investment = 1300 :
 Event Cash flow VAN
Expansion −10
1 − (1.1 )
360 × =2212
0.1
Stagnation 1 − (1.1 )
−10

120 × =737 -1300 + 50%2212 + 40%737 + 10%(-61) = 94.7

0.1
Recession −10
1 − (1.1 )
-10 × -61
0.1

The most economically rational decision is to invest in a unit of

additional production.

Example 28:
A company must choose between the following two alternative decisions:
Invest 6000 immediately.
2- Not investing anything immediately. In this case, at the beginning of year 2, the company could
decide to invest 3500.
Probability of demand:
Year 1:
Probability of strong demand: 40%.
-Probability of low demand: 60%
Years 2 to 5:
If the demand in year 1 was strong, the probability that the demand remains strong is 60%.
If the demand in year 1 was low, the probability that the demand remains low is
90%.

Monetary flows:

1- In case of an immediate investment of 6000, the net cash flows from years 1 to 5 will be
3000 if demand is high or 1000 if demand is low.
2- Without investment, the annual cash flows for years 1 to 5 will be 1000. In case
an investment of 3500 at the beginning of year 2, the net cash flows from years 2 to 5 will be
2500 if the demand is high or 1000 if the demand is low.
Coût du capital: 10%
Construction of the decision tree:
The probabilities of the events are shown on the tree.
When the event is immediately preceded by a decision, the probabilities are
simple probabilities (case of events following the nodes E)land E4).
When there are no decisions between consecutive events, the probabilities of
Subsequent events are compound probabilities (in the case of events that follow the
nodes E5and E6).
Question: determine the choice that must be made.
 E3 Strong demand Pb=0.6
I=3500

Low demand Pb=0.4

Strong demand D2 No investment
Pb=0.4

E1 Strong demand Pb=0.1

E4

I=3500

Weak demand
Pb=0.6 D3 Low demand Pb=0.9

Pass No investment
to invest

D0

Strong demand Pb=0.24

Low demand Pb=0.4
I=6000 E5
Strong demand Pb=0.4

E2
Low demand Pb=0.16

Strong demand Pb=0.06

E6

Low demand Pb=0.6

Low demand Pb=0.54

Year 1 From year 2 to year 5

Calculation of NPVs at the end of each branch

Invest. Net cashVAN at 12% flows.

0 1 1 2 3 4 5
0 3500 1000 2500 2500 2500 2500 4548(1)
0 3500 1000 1000 1000 1000 1000 480
0 0 1000 1000 1000 1000 1000 3605
0 3500 1000 2500 2500 2500 2500 4548
0 3500 1000 1000 1000 1000 1000 480
0 0 1000 1000 1000 1000 1000 3605
6000 3000 3000 3000 3000 3000 4814
6000 3000 1000 1000 1000 1000 -610
6000 1000 3000 3000 3000 3000 3029
6000 1000 1000 1000 1000 1000 -2395

(1) VAN = -3500 (1.12) +-11000 (1.12)-1+ 2500 (1.12)-2+ 2500(1.12) + 2500
-3
(1.12)-4+
-5
2500 (1.12) = 4548.

Calculation of the mathematical expectation of NPVs and elimination of dominated decisions:

NodeD2
Investment decision of 3500: E(V AN) = 4548 x 0.6 + 480 x 0.4 = 2920
Decision not to invest: E(V AN) = 3605
The decision to choose is the one with the highest NPV. The decision to invest 3500 is
eliminated.
NodeD3
Decision to invest 3500: E(NPV) = 4548 x 0.1 + 480 x 0.9 = 896
Decision not to invest: E(NPV) = 3605
The decision to invest 3500 is eliminated.
After eliminating dominated decisions, it is advisable to rebuild a decision tree.
simplified.
NodeD2
Decision not to invest: E(NPV) = 3605
Decision to invest 6000: E(NPV) = 4814 x 0.24 -610 x 0.16 + 3029 x 0.06 + (-2395) x 0.54
-54
The decision to invest 6000 is eliminated.
The decision made is therefore not to make any investments.
Example 29:
A company must choose between 3 projects, evaluated over a period of 4 years, taking into account
count of the following two events:
El The activity remains stable or shows slight progress (probability = 0.6);
E2: The activity tends to decline (probability = 0.4).

First project: Limited investment initially, with possible expansion after a year.
Investment of 6000 at date 0 and 3000 at date 1 if the hypothesis is verified.
Estimated cash flows at the end of the first year: 2000.

Annual cash flows as follows:

-If extension: 3500 if Eland 3000 if E2
If non-extension: 2300 if El and 2000 if E2.
Second project: Global investment from the start: I=8000, expected annual cash flows:
2800 is E1and if E2
Third project: To renounce all investment.
Coût de capital : 10%
Task to do: Determine the choice that must be made.
E1(Pr=0.6)
I=3000 E2(Pr=0.4)
extension
E1(Pr=0.6) E1(Pr=0.6)
D2
I=6000 I=0
E2(Pr=0.4) E(VAN)=
Non-extension
E2(Pr=0.4) 584

I=800 E1(Pr=0.6)
D1
E(VAN)
E2(Pr=0.6) =495
I=0
E(VAN) = 0

weather
II - Choice of investments in an indeterminate universe
The forecasting of cash flows in a random future assumes that one can allocate
a probability for each of the flows. On the contrary, the future is uncertain (or indeterminate)
when we have no idea about the probability of future revenues. In this situation,
the investor is completely unable to assign a probability of realization to a
event. It can only qualify the event as favorable or not.
Example 30:
An investor is planning to embark on an investment project. A study of
the market allows him to predict the expected results, according to the available production capacity
and the expected revenues.
Production capacity Recipe
in thousands of pieces
8000 9000 10000 11000
A :100 -50 -20 0 10
B: 120 -80 -10 50 110

C :140 -110 -30 80 150

What decision needs to be made?

1- Laplace's criterion:
The Laplace criterion consists of calculating the arithmetic mean for each project.
of possible vans and to choose the project for which the average is the highest.
This amounts to calculating, for each project, the expected value of the NPV by assigning
implicitly assigns the same probability to each state of nature. This criterion is therefore a
simplification of the criterion of mathematical expectation.
Let's calculate the arithmetic mean of the results of each production capacity:

A= −50−20+0+10 -15
4

B= −80−10+50+110 =17.5
4

C= −110−30+80+150 = 22.5
4
Decision C is optimal according to Laplace's criterion.
2- Wald Criterion (or Maximin):
The Maximin criterion (maximum of the minima) consists of determining the NPV that is the most

weakness of each project and to choose the project for which this NPV is the lowest.
big.
The Maximin criterion limits the risk of low revenues. It is suited to psychology.
of a cautious decision maker who is mainly afraid of poor results.
Minimum result
Decision A -50
Decision B -80
Decision C -110

According to this criterion, decision A is optimal.

3- Maxi Max Criterion

This criterion favors the riskiest, most optimistic attitude. The decision will be retained.
offering the highest maximum result.
Maximum result
Decision A 10
Decision B 110
Decision C 150

According to this criterion, decision C is optimal.

4- Hurwicz criterion:
The Hurwicz criterion allows for a relativization of decision-making. By using the criteria
previously, it is implicitly considered that the decisions are equiprobable. Hurwicz
It is recommended to weight the probable minimum and maximum results according to a coefficient.
subjective reflecting the degree of optimism or pessimism, the offensive or cautious nature
of the decision-maker. The weighting coefficients are between 0 and 1.
Let's keep the coefficients 0.4 and 0.6 to apply respectively to the maximum values.
and to the minimum values.
 Weighted average
Decision A 0.6x (-50)+0.4x10x10=-26
Decision B 0.6x(-80)+0.4x110= -4
Decision C 0.6x(-110)+0.4x150 =-6
According to this criterion, solution B will be chosen.

5- Savage Criterion (or Minimax Regret Criterion):

The Savage criterion consists of:
1- to identify, for each state of nature, the project that would yield the best result if this
the state of nature would be realized;
2- to calculate, for each assumed realized state of nature, the lost earnings (or regrets)
which would result from the adoption of the other projects than the one identified in 1; this leads to a
matrix of regrets;
3- to determine, for each project, the maximum regret (Max);.
4- to choose the project for which the maximum regret is minimal (Mini Max).
Like the maxi min criterion, the mini max regret criterion corresponds to psychology.
of a cautious decision-maker who seeks to limit their potential loss of earnings.
8000 9000 10000 11000
Project A -50 - (-50) = 0 -10 - (-20) = 10 80-0=80 150-10=140
Project B -50-(-80)=30 -10-(-10)= 0 80-50=30 150-110=40
Project C -50-(-110)=60 -10 - (-20) = 10 80-80=0 150-150=0

According to Savage's criterion, the optimal decision is decision B.

Conclusion

The criteria used lead to different decisions:

criterion decision Production capacity

Laplace C 140

forest A 100

Maxi Max A 140

Savage B 120
 Hurwicz B 120

The investor will make a decision based on behavior, on nature.

clean. However, it should be noted that rarely will an investor operate in a completely
indeterminate. Information about the environment, the industry sector, is often
indispensable.

The concepts to review

Certain environment The variance Wald Criterion
Random environment Independence of flows Maxi Max Criterion
Undefined environment Dependence of flows Savage criterion
Equivalent certain Decision tree Hurwicz criterion
Mathematical expectation Laplace criterion