This action might not be possible to undo. Are you sure you want to continue?

# P.C. Godinho, A.R. Afonso, J.P.

**Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 1
**

On the use of multiple ﬁnancial methods in the

evaluation and selection of investment projects

∗

Pedro Cortes˜ ao Godinho

∗ †

Ant´ onio Ricardo Afonso

∗ ‡

Jo˜ ao Paulo Costa

∗ †

∗

Faculdade de Economia da Universidade de Coimbra

pgodinho@sonata.fe.uc.pt

†

INESC - Instituto de Engenharia de Sistemas e Computadores

‡

Portugal Telecom Inova¸ c˜ ao

Abstract

This paper addresses the analysis and evaluation of investment projects within a mul-

ticriteria framework. We mathematically deﬁne a multicriteria framework and we present

a result that allows the identiﬁcation of redundant methods. Then we try to deﬁne which

ﬁnancial methods can, and which ones cannot, be simultaneously used according to that

framework. We also try to establish a set of guidelines to help decision makers choose the

ﬁnancial methods best suited to their particular situations.

Keywords: Project Evaluation, Project Analysis, Multicriteria Decision Aid

1 Introduction

Analysis and evaluation of investment projects are fundamental activities in most businesses.

In fact, their prosperity depends upon the correct allocation of the capital they raise - if many

unproﬁtable investments are made, the survival of the companies may be in danger. Many

methods for economic evaluation of investment projects, also known as ﬁnancial methods, have

been developed to help decision makers (DMs) assess whether or not an investment will be

proﬁtable, or compare the proﬁtability of diﬀerent investments. Although all these methods

∗

Supported by FCT, FEDER, project POCTI/32405/GES/2000. The authors wish to thank the helpful

comments of an anonymous referee.

c 2004 Associa¸ c˜ ao Portuguesa de Investiga¸ c˜ ao Operacional

2 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

are, directly or indirectly, concerned with the proﬁtability of the investments, they do not

always yield the same results - in fact, diﬀerent methods do sometimes yield contradictory

results when evaluating or comparing the same investments. This fact raises some diﬃcult

questions to DMs in charge of investment selection, concerning:

• which method (or methods) should be used in a particular situation;

• whether one or more method(s) should be used in a particular situation;

• how should the results from diﬀerent methods be aggregated.

To worsen matters, these methods cannot usually account for all the relevant information.

One reason is that some strategic impacts of the investments are too complex to be properly

quantiﬁed in the predicted earnings or cash ﬂows. Another reason is that there are usually

some issues, not related with the proﬁtability of the investment, that DMs want to consider

when they make an investment decision - these issues may refer to prestige, power or ethical

concerns and are relevant to the DMs as individual human beings. This raises the question of

knowing when and how these issues should be considered.

In this paper we will address the analysis and evaluation of investment projects within a

multicriteria framework. This framework will provide a theoretical basis to the aggregation

of the results yielded by diﬀerent methods, and we believe it may also provide a basis for the

aggregation of these with non-ﬁnancial factors. It will also allow the use of decision theory

methods in investment decisions and, hopefully, avoid some decision errors due to an incorrect

aggregation of factors. The framework we use is based upon the work of Bana e Costa [3]

and Roy [13]. In this framework, all the properties, or characteristics, of the investments are

modelled as attributes, and the results yielded by ﬁnancial methods will be called ﬁnancial

attributes. Among the attributes, the decision maker (DM) will build a set of criteria, taking

into account his/her concerns, values and beliefs. Some of the criteria may result from the

aggregation of several attributes.

First, we present the most widely used methods for project evaluation. Using a classiﬁ-

cation based upon [11,12], we divide the most important ﬁnancial methods into ﬁve classes -

equivalent worth, rate of return, ratio, payback and accounting.

In order to have a correct structure for the decision problem, we want the set of criteria to

be a coherent family of criteria [3]. This means that we want the set of criteria to be exhaustive,

cohese and non-redundant. Exhaustiveness means that all relevant criteria are included in the

set of criteria. So, if any two alternatives are equal in all criteria they must be indiﬀerent for

the DM, or else we must conclude that there is at least one relevant issue that is not properly

accounted for by the set of criteria. Cohesion means that if two alternatives, A and B, are

equal in all criteria but one, and A is better than B according to that criterion, then A must

be preferred to B. A set of criteria is non-redundant if the removal of any criterion causes that

set to be no longer both exhaustive and cohese. In section 3, we mathematically deﬁne these

conditions and present a result that can be applied to the identiﬁcation of redundant criteria.

Then, we try to ﬁnd out which ﬁnancial attributes can, and which ones cannot, be used

together as criteria, assuming that we want the set of criteria to be a coherent family of criteria

and to be based on non-contradictory assumptions and concepts. We try to deﬁne whether or

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 3

not ﬁnancial attributes from the same class should be used together as criteria, and whether or

not ﬁnancial attributes from diﬀerent classes should be used together as criteria, according to

our framework. We argue that a DM will usually want to use at most one ﬁnancial attribute

from a single class. We also argue that a DM may use together attributes from diﬀerent classes,

but he/she will usually want to use at most one attribute from both the classes of ratio and

rate of return methods. Afonso et al. [1] perform a similar analysis, using a slightly diﬀerent

classiﬁcation of the ﬁnancial methods.

In section 5 we try to establish a set of guidelines to help DMs choose the ﬁnancial methods

best suited to their speciﬁc situations. First, we characterise a set of decision situations, ac-

cording to the degree of quantiﬁcation, capital availability for the investments, degree of risk

and uncertainty, interdependencies between investments and existence of previously under-

taken investments. We try to ﬁnd out which ﬁnancial method(s) is (are) best suited for each

situation, and how should each characteristic aﬀect the investment selection process. Although

this approach is similar to the approach of Fahrni and Spatig [5], there are some important

diﬀerences between them. One important diﬀerence is that, while Fahrni and Spatig focus on

R&D projects, we try to consider all kinds of projects. Possibly because of this, our charac-

terisation of the decision situations is diﬀerent from theirs. Also, while we aim to suggest one

ﬁnancial method for each situation, Fahrni and Spatig never suggest any particular ﬁnancial

method - they treat ﬁnancial methods as a whole component that shall or shall not be used

according to the situation.

Next, we try a diﬀerent approach. We consider some classes of methods and we try to

deﬁne in which situations should the methods belonging to those classes be used, without

following any particular situation taxonomy.

We conclude that the net present value (NPV), or other method from the equivalent worth

class, is usually the best choice. However, rate of return or ratio methods may be the best

suited for some particular situations. We do not exclude that, in many situations, the DMs

may want to consider other methods, along with the NPV (or the rate of return or ratio

method), due to additional reasons.

2 Financial methods

A large number of ﬁnancial methods is presented in ﬁnancial textbooks and papers. In this

section, we will describe some of the most important ﬁnancial methods and, following [11,12],

we will categorise them into ﬁve classes: equivalent worth, rate of return, ratio, payback and

accounting.

Equivalent worth methods examine the project cash ﬂows and, through discounting or

compounding, resolve them to one equivalent cash ﬂow or to an equivalent series of cash

ﬂows. The most important of these methods is the net present value (NPV). The NPV is the

present monetary value of all the project cash ﬂows (including investment and salvage value)

discounted at the appropriate discount rate. The traditional deﬁnition of the NPV is:

NPV =

T

t=0

CF

t

(1 +r)

t

(1)

where r is the discount rate, CF

t

is the cash ﬂow in period t and T is the horizon period

4 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

(which is often the project lifetime). The NPV is nowadays considered by many authors to

be, in most situations, the best economic proﬁtability measure for investment projects (see,

for example, [4]).

The future worth (FW), the annual worth (AW) and the capitalised worth (CW) are other

equivalent worth methods. The FW is the monetary value of all the project cash ﬂows in a

future date, and it can be found through the compounding of the cash ﬂows to that future

date. The AW is the value of each of the cash ﬂows of an equivalent project (having the same

NPV) with a ﬁnite lifetime (usually identical to the lifetime of the original project or to the

considered horizon period), whose cash ﬂows are constant over its lifetime. The CW is equal

to the AW except that it considers another equivalent project with an inﬁnite lifetime. These

methods can be deﬁned in the following way:

FW =

T

t=0

CF

t

(1 +r)

T0

−t

= NPV (1 +r)

T0

(2)

AW =

T

t=0

CF

t

(1+r)

t

T

t=1

1

(1+r)

t

=

T

t=0

CF

t

(1+r)

t

1

r

−

1

r(1+r)

T

=

NPV

1

r

−

1

r(1+r)

T

(3)

CW =

T

t=0

CF

t

(1+r)

t

∞

t=1

1

(1+r)

t

=

_

T

t=0

CF

t

(1 +r)

t

_

· r = NPV · r (4)

In the deﬁnition of FW, T

0

is the future moment for which the FW is calculated. The remaining

notation was previously deﬁned.

The use of equivalent worth methods must include an adjustment for the risk of the project.

This risk can be seen as the possible deviations from the expected project behaviour, and it can

be divided into systematic and unsystematic risk. The unsystematic risk can be eliminated by

holding a diversiﬁed portfolio of investments, and the systematic risk cannot be eliminated by

diversiﬁcation, and it is the only type of risk that should matter to investors with diversiﬁed

portfolios. Also, ﬁnancial theory prescribes that only the systematic risk shall be incorporated

in the value of ﬁnancial assets and projects. The measure used for this type of risk is the

beta (β) coeﬃcient, deﬁned as the covariance of the asset (or project) returns with the market

returns divided by the variance of the market returns. Using the project betas, it is possible to

incorporate the systematic risk in the project NPV through the use of a risk-adjusted discount

rate, calculated according to the Capital Asset Pricing Model (CAPM). According to the

CAPM, if r

f

is the risk-free discount rate, β is the project beta and E(r

m

) is the expected

market return, then the correct risk-adjusted rate will be:

r = r

f

+β [E(r

m

) −r

f

] (5)

For more details on the calculation of project betas and risk-adjusted discount rates, see [4].

The adjustment of the discount rate works very well for the NPV, but not for the other

equivalent worth methods. In fact, the use of such a risk-adjusted discount rate in the other

equivalent worth methods will not correctly adjust for the project risk. To deal with this

problem we suggest the adjustment of the cash ﬂows, instead of the adjustment of the discount

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 5

rate, by using certainty equivalents of the cash ﬂows ([4], chapter 9). Speciﬁcally, r

f

being the

risk-free discount rate and r being the risk-adjusted discount rate, the period t cash ﬂow is

adjusted to its certainty equivalent,

CE = CF

t

·

_

1 +r

f

1 +r

_

t

(6)

and then the risk-free discount rate is used. For the NPV, it is indiﬀerent to adjust the discount

rate or the cash ﬂows. However, for the other methods, the adjustment of cash ﬂows will allow

the correct comparison of projects with diﬀerent systematic risk.

Rate of return methods measure the rate at which the invested capital will grow if the

project is pursued. The most widely used rate of return method is the internal rate of return

(IRR). This rate can be deﬁned as the discount rate for which the NPV equals zero, and

corresponds to the yield-to-maturity on a bond. It can be calculated by solving the following

equation:

T

t=0

CF

t

(1 +IRR)

t

= 0 (7)

Other methods from the rate of return class include the external rate of return (ERR) and

the marginal return on invested capital (MRIC). Both these methods consider an explicit

reinvestment rate. The ERR is the rate for which the future worth of the initial investment

equals the future worth of the other cash ﬂows compounded, at the reinvestment rate, to the

end of the project. Using I

0

to represent the initial investment, we can deﬁne:

T

t=1

CF

t

(1 +r)

T−t

(1 +ERR)

T

= I

0

(8)

The MRIC is deﬁned in [9]. Two kinds of cash ﬂows are considered - capital cash ﬂows, used

to ﬁnance the project, and operating cash ﬂows, generated by the project. Capital cash ﬂows

are discounted, at the reinvestment rate, to the beginning of the project and operating cash

ﬂows are compounded, at the same reinvestment rate, to the end of the project. The MRIC

is the rate at which the present value of the capital cash ﬂows should be compounded so that

it would equal the future value of the operating cash ﬂows at the end of the project. So, the

MRIC can be deﬁned as:

T

t=0

CCF

t

(1 +r)

t

=

T

t=1

OCF

t

(1 +r)

T−t

(1 +MRIC)

T

(9)

where CCF

t

is the capital cash ﬂow for period t and OCF

t

is the operating cash ﬂow for period

t.

The most signiﬁcant ratio methods can be deﬁned as the quotient between the present

value of the returns and the present value of the investment. The most widely used ratio

method is the proﬁtability index (PI), which is deﬁned as the quotient between the present

value of the future cash ﬂows generated by the project and the initial investment:

PI =

T

t=1

CF

t

(1+r)

t

I

0

(10)

6 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

Other ratio methods can be deﬁned as the ratio between the present value of the returns and

the present value of the investment. The beneﬁt-cost ratio (B/C ratio), for instance, is the

quotient between the present value of all cash ﬂows, excluding the initial investment and the

salvage value, and the diﬀerence between the initial investment and the present worth of the

salvage value. If we use SV

T

to represent the salvage value of the project, we can deﬁne:

B/C =

T

t=1

CF

t

(1+r)

t

−

SV

T

(1+r)

T

I

0

−

SV

T

(1+r)

T

(11)

Payback methods calculate how long it takes to recover the invested capital. These methods

include the payback period and the discounted payback period. The payback period is the

number of years required for the accumulated project cash ﬂows to equal the initial investment.

The discounted payback period is similar to the payback period, except that it considers the

discounted cash ﬂows instead of the raw cash ﬂows. So, we can deﬁne:

Payback = min

_

k :

k

t=1

CF

t

≥ I

0

_

(12)

Discounted Payback = min

_

k :

k

t=1

CF

t

(1 +r)

t

≥ I

0

_

(13)

Accounting methods consider proﬁtability from an accounting perspective. This class includes

the return on original investment (ROOI, a.k.a. original book method) and the return on

average investment (ROAI, a.k.a. average book method), among others. The ROOI is the

quotient between the average yearly accounting proﬁt, which excludes depreciation, and the

investment made in the project. The ROAI is the quotient between the average yearly ac-

counting proﬁt and the average book value (average value of the diﬀerence between investment

and depreciation) during the project life. Using AP

t

to represent accounting proﬁt in period

t and BV

t

to represent book value in period t, we can deﬁne:

ROAI =

T

t=1

AP

t

T

T

t=0

BV

t

T+1

(14)

ROOI =

T

t=1

AP

t

T

I

0

(15)

A further description of most of these ﬁnancial methods can be found in [11,12]. [4] and

[10] also describe some of these methods, discussing its advantages and drawbacks and also

discussing the calculation of the discount or compounding rate needed by some of them.

3 Mathematical deﬁnitions and results

This section provides some mathematical results used to ﬁnd out which attributes can and

which ones cannot be used together as criteria. We consider that, in order to have a correct

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 7

structure for the decision problem, we want the set of criteria to be a coherent family of criteria

[3]. As was said before, this means that the set of criteria must be exhaustive, cohese and non-

redundant. Exhaustiveness means that all relevant criteria are included in the set of criteria.

So, if any two alternatives are equal in all criteria they must be indiﬀerent for the DM, or else

we must conclude that there is at least one relevant issue that is not properly accounted for by

the set of criteria. Cohesion means that if two alternatives, A and B, are equal in all criteria

but one, and A is better than B according to that criterion, then A must be preferred to B. A

set of criteria is non-redundant if the removal of any criterion causes that set to be no longer

both exhaustive and cohese. We will now mathematically deﬁne these conditions and we will

show that two attributes that rank projects identically will be redundant.

Assumptions

Let A={a

1

, a

2

, . . . , a

n

} be the set of projects and F={g

1

, g

2

, . . . , g

m

} the set of attributes

used as criteria. g

k

(a

i

) will be the performance of project a

i

according to attribute g

k

. Let us

also assume, without loss of generality, that a larger value in a given attribute is always better

than a smaller one.

Deﬁnitions The symbols P and I will be used as comparison operators: a

i

P a

j

means that

a

i

is considered to be preferred to a

j

and a

i

I a

j

means that a

i

and a

j

are considered to be

indiﬀerent.

We will consider that, in order to be a coherent family of criteria, the set F must meet the

following exhaustiveness, cohesion and non-redundancy conditions:

(Exhaustiveness)

∀a

i

,a

j

∈A, (g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F) ⇒ a

i

I a

j

(16)

(Cohesion)

∀a

i

,a

j

∈A, ∀g

l

∈F, (g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

l

}∧ g

l

(a

i

) >g

l

(a

j

)) ⇒ a

i

P a

j

(17)

(Non-redundancy)

∀g

p

∈F, ∃a

i

,a

j

∈A: {[(g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

p

})∧(a

i

P a

j

∨ a

j

P a

i

)] ∨

[∃g

l

∈F\{g

p

}:g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

p

,g

l

}∧g

l

(a

i

) >g

l

(a

j

)∧(a

j

P a

i

∨ a

i

I a

j

)]} (18)

The ﬁrst part of expression (18) means that if any criterion g

p

is removed then F will no

longer meet the exhaustiveness condition; the second part of that expression means that if any

criterion g

p

is removed then F will no longer meet the cohesion condition.

Theorem (attribute redundancy):

Let us assume that, for two attributes g

r

and g

s

, we have:

∀ a

i

,a

j

∈A, g

r

(a

i

) >g

r

(a

j

) ⇒ g

s

(a

i

) >g

s

(a

j

) (19)

If the criteria set F includes both g

r

and g

s

then F is not a coherent family of criteria

because, if both the exhaustiveness and cohesion conditions are met, then the non-redundancy

condition is not met. Speciﬁcally, g

r

can be removed from F without the exhaustiveness and

cohesion conditions ceasing to hold. We thus say that g

r

is redundant.

8 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

Corollary: If two attributes rank the projects identically, then they will be redundant (only

one of them should be used as criterion).

Theorem proof:

We will show that (a) if F meets the exhaustiveness condition, then F\{g

r

} also meets the

exhaustiveness condition and (b) if F meets the cohesion condition, then F\{g

r

} also meets the

cohesion condition. This will show that g

r

can be removed from F without the exhaustiveness

and cohesion conditions ceasing to hold, thus g

r

is redundant.

Let us start by proving (a). (19) says that g

r

(a

i

) >g

r

(a

j

) ⇒ g

s

(a

i

) >g

s

(a

j

), thus

g

r

(a

j

) >g

r

(a

i

) ⇒ g

s

(a

j

) >g

s

(a

i

). This means that we may only have g

s

(a

i

)=g

s

(a

j

) when

g

r

(a

i

)=g

r

(a

j

). So:

∀ a

i

,a

j

∈A, g

s

(a

i

)=g

s

(a

j

) ⇒ g

r

(a

i

)=g

r

(a

j

) (20)

Also:

g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

r

} ⇒ g

s

(a

i

)=g

s

(a

j

) (since g

s

∈F\{g

r

})

⇒ g

r

(a

i

)=g

r

(a

j

) (using (20)) (21)

And so:

g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

r

} ⇒ g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F (using (21))

⇒ a

i

I a

j

(22)

(because F meets the exhaustiveness condition)

(22) means that F\{g

r

} meets the exhaustiveness condition, proving (a).

Let us now prove (b). We will prove that, if F meets the cohesion condition (if (17) holds)

then F\{g

r

} will also meet the cohesion condition, meaning that for any projects a

i

and a

j

:

∀g

l

∈F\{g

r

}, (g

k

(a

i

)=g

k

(a

j

),∀g

k

∈F\{g

r

,g

l

}∧g

l

(a

i

) >g

l

(a

k

)) ⇒ a

i

P a

j

(23)

We will start by showing that (23) holds for all g

l

∈F\{g

s

,g

r

}. We will thus prove that,

for g

l

∈F\{g

s

,g

r

}, if g

k

(a

i

)=g

k

(a

j

),∀g

k

∈F\{g

r

,g

l

}∧g

l

(a

i

) >g

l

(a

k

) then a

i

P a

j

.

Since g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

r

,g

l

}, and g

l

∈F\{g

s

,g

r

}, then we have g

s

(a

i

)=g

s

(a

j

) and,

by (20) we have g

r

(a

i

)=g

r

(a

j

). So:

g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

r

,g

l

}∧g

l

(a

i

) >g

l

(a

j

) ⇒ g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

l

}∧g

l

(a

i

) >g

l

(a

j

)

⇒ a

i

P a

j

(24)

(since the cohesion condition holds)

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 9

To complete the proof, we will show that (23) holds for g

l

=g

s

. From (19) we can say that

g

s

(a

i

) ≤g

s

(a

j

) ⇒ g

r

(a

i

) ≤g

r

(a

j

) and, consequently:

g

s

(a

i

) >g

s

(a

j

) ⇒ g

r

(a

i

) ≥g

r

(a

j

) (25)

Using (25) we get, for g

l

=g

s

:

g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

r

,g

s

}∧ g

s

(a

i

) >g

s

(a

j

) ⇒ g

r

(a

i

) ≥g

r

(a

j

) (26)

Let us analyse the expression (26). It says that a

i

and a

j

are equal in all criteria except

g

s

and, possibly, g

r

. Since F meets the cohesion condition, if g

r

(a

i

)=g

r

(a

j

), then we have a

i

P a

j

(because g

s

(a

i

) >g

s

(a

j

). If a

i

is also better than a

j

in g

r

, then a

i

is also considered to be

preferred to a

j

. So:

g

k

(a

i

)=g

k

(a

j

), ∀g

k

∈F\{g

r

,g

s

}∧ g

s

(a

i

) >g

s

(a

j

) ⇒ a

i

P a

j

(27)

We showed that (23) holds for all attributes g

l

∈F\{g

r

}. So, if F meets the cohesion

condition, F\{g

r

} also meets the same condition. (22) shows that if F meets the exhaustiveness

condition, then F\{g

r

} also meets the same condition. This means that, when (19) holds, g

r

will be redundant, so the theorem proof is complete.

The corollary of this theorem can now be used to ﬁnd redundant attributes. Whenever

two diﬀerent attributes rank projects identically, they shall not be used together as criteria,

since they will be redundant.

4 On the simultaneous use of diﬀerent ﬁnancial attributes

In this section we will try to ﬁnd out which ﬁnancial attributes can, and which ones cannot, be

used together as criteria. For that purpose we will initially consider ﬁnancial attributes from

the same class, and then we will consider attributes from diﬀerent classes. In this analysis we

will assume that we want the set of criteria to be a coherent family of criteria, and also to be

based on non-contradictory assumptions or concepts.

We will want the set of criteria to be not only a coherent family of criteria, but also to be

based on non-contradictory assumptions or concepts. We acknowledge that, in the presence of

risk, it may be worthwhile to consider the behaviour of the project under diﬀerent scenarios

(as will be discussed in section 5). However, the data and methods used in each scenario

should not include contradictory assumptions or concepts, so that each scenario represents a

consistent possibility of project behaviour.

In the following discussion, we will only consider the ﬁnancial methods presented in section

2, which include the most common ﬁnancial methods. We believe our conclusions are extensible

to other methods, if these methods can be properly classiﬁed into one of the ﬁve classes we are

considering. In the analysis of ﬁnancial attribute redundancy, we will not discuss whether or

not two attributes happen to rank the projects identically for a particular set of projects. We

10 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

will only consider that two attributes are redundant if they always rank projects identically.

This way, our results will be independent of any particular set of projects, and will still be

valid when some projects are removed from or added to the initial set.

To start with, we will address the simultaneous use of diﬀerent ﬁnancial attributes from

the equivalent worth class. Several constraints must be met by the methods from this class,

so that they can be properly applied. One example: the projects being compared should have

the same discount rate

1

[11]. It is easy to prove that, when properly applied, all equivalent

worth methods rank projects identically. Let us consider the NPV and the CW. We will

assume that the discount/reinvestment rate belongs to the interval ]-1,+∞[, in which it has

economic meaning, and that the projects being compared have the same discount rate, even if

it is necessary to use certainty equivalent cash ﬂows to achieve that. So, the CW will be equal

to the NPV divided by the perpetuity factor,

∞

t=1

1

(1+r)

t

. Since, for r≤0, the perpetuity factor

is +∞, the CW is only deﬁned for r>0, in which case the perpetuity factor equals

1

r

. So, we

must assume r>0, in which case:

NPV(a

i

) > NPV(a

j

) ⇔ NPV(a

i

)·r > NPV(a

j

)·r

⇔ CW(a

i

) > CW(a

j

) (28)

So, the NPV and the CW rank projects identically. Park and Sharp-Bette [10], chapter

7, show that the NPV, the AW and the FW rank projects identically. Thus, according to

the non-redundancy demand of a coherent family of criteria, at most one attribute from the

equivalent worth class should be used as criterion. The interpretation of the NPV provides

some advantages over the other equivalent worth attributes, so the NPV will usually be used.

However, special circumstances may advise the use of a diﬀerent attribute.

We will now consider the rate of return methods. It is well known that diﬀerent rate of

return methods may rank the same projects diﬀerently. That is because the results obtained

depend on the reinvestment assumptions and on the implicit concepts of investment and return

for each method. While the IRR assumes that the proﬁts are reinvested at a rate equal to the

IRR, the ERR and the MRIC consider an explicit reinvestment rate. The ERR diﬀers from

the MRIC on the concepts of investment and return. While the MRIC considers investment to

be all the capital cash ﬂows, the ERR considers investment to be only the initial investment.

As diﬀerent rate of return methods can yield contradictory results, we cannot say that rate

of return attributes are redundant. However, in general only one should be used, chosen

according to the DM’s reinvestment assumptions and concepts of investment and return. If

more than one rate of return attribute is used in the evaluation process, then contradictory

assumptions or concepts will be simultaneously involved.

Like the rate of return methods, ratio methods may also rank the same projects diﬀerently.

That is because they also assume diﬀerent concepts of investment and return. While the PI

assumes investment to be the initial investment and all the other cash ﬂows to be return,

the B/C ratio assumes investment to be the diﬀerence between the initial investment and its

salvage value, and return to be all the other cash ﬂows excluding the salvage value. Thus,

1

The adjustment of cash ﬂows, suggested in section 2, allows us to use the same discount rate for all the

projects, even when the project risk is diﬀerent.

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 11

as diﬀerent ratio methods may rank projects diﬀerently, we cannot say that their results are

redundant. However, only one ratio attribute should be usually used, chosen according to the

DM’s concepts of investment and return.

Accounting methods diﬀer in what they consider relevant about the investment. While the

ROOI assumes that the whole value of the investment (the whole book value) matters, the

ROAI assumes that it is the average book value, after the yearly depreciation, that matters.

So, these methods can yield diﬀerent results, according to the type of depreciation associated

with each project, and thus they are not redundant. However, a DM will usually want to use

at most one accounting attribute, according to what he/she thinks is more signiﬁcant about

the investment: the whole book value or the average book value. Moreover, because these

methods do not take the time value of money into account, it is arguable that a DM will

consider them relevant. Accounting attributes should only be used when accounting issues are

considered important.

Payback methods may rank the same projects diﬀerently, and lead to diﬀerent accept/reject

decisions. That is because while the discounted payback period takes the time value of money

into account, the payback period does not. This means they are not redundant, but usually

at most one of them will be used, according to whether or not the time value of money is

considered important. A DM will usually consider the time value of money relevant and will

thus prefer the discounted payback period.

We will now consider the simultaneous use of attributes from diﬀerent classes. To start

with, we will try to ﬁgure out which perspective, or dimension, of the proﬁtability does each

class address. The equivalent worth class addresses the absolute value of the project, account-

ing methods address the accounting proﬁtability and payback methods address the time it

takes to recover the initial investment, or the liquidity recovery speed. As for the rate of

return and ratio classes, we can see that for each ratio method we can ﬁnd out or deﬁne a rate

of return method that is equivalent in the sense that it ranks projects in the same way and

always leads to the same accept/reject decisions. The opposite is also usually true

2

. As an

example, we can see that the PI is equivalent to the ERR, thus they are redundant. For the

MRIC, we can deﬁne an equivalent ratio method (let us call it Modiﬁed Proﬁtability Index,

MPI) as:

MPI =

T

t=1

OCF

t

(1+r)

t

T

t=0

CCF

t

(1+r)

t

(29)

where OCF

t

is the operating cash ﬂow in period t, CCF

t

is the capital cash ﬂow in period t, r

is the discount rate (equal to the MRIC reinvestment rate) and T is the horizon period. Once

more, if we assume that the rate r belongs to the interval ]-1,+∞[, in which it has economic

meaning, and that this reinvestment rate is equal for both projects, we have:

MRIC(a

i

) > MRIC(a

j

) ⇔ 1 + MRIC(a

i

) > 1 + MRIC(a

j

)

2

For each rate of return method that uses an external rate to resolve all the cash ﬂows into two cash ﬂows,

we can deﬁne a ratio method that is equivalent. This is the case of the ERR and the MRIC.

12 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

⇔

T

¸

¸

¸

¸

¸

¸

¸

_

T

t=1

OCF

t

(a

i

) · (1 +r)

T−t

T

t=0

CCF

t

(a

i

)

(1+r)

t

>

T

¸

¸

¸

¸

¸

¸

¸

_

T

t=1

OCF

t

(a

j

) · (1 +r)

T−t

T

t=0

CCF

t

(a

j

)

(1+r)

t

⇔ (1 +r)

T

T

t=1

OCF

t

(a

i

)

(1+r)

t

T

t=0

CCF

t

(a

i

)

(1+r)

t

> (1 +r)

T

T

t=1

OCF

t

(a

j

)

(1+r)

t

T

t=0

CCF

t

(a

j

)

(1+r)

t

⇔

T

t=1

OCF

t

(a

i

)

(1+r)

t

T

t=0

CCF

t

(a

i

)

(1+r)

t

>

T

t=1

OCF

t

(a

j

)

(1+r)

t

T

t=0

CCF

t

(a

j

)

(1+r)

t

⇔ MPI(a

i

) > MPI(a

j

) (30)

So, the MRIC and the MPI rank projects identically. Park and Sharp-Bette [10], chapter

7, show that some rate of return and ratio methods, including the ERR and the PI, rank

projects identically. It would also be easy to deﬁne a rate of return method equivalent to the

B/C ratio. This relation between rate of return and ratio methods led us to the conclusion that

they would probably address the same proﬁtability dimension. And, in fact, they both address

the relation between the return and the investment. Thus, as both classes address the same

dimension, we should usually use at most one attribute from both these two classes. As for

the other classes, since they address diﬀerent proﬁtability dimensions, attributes from diﬀerent

classes can be simultaneously used as criteria, given that they are not based on contradictory

assumptions or concepts. Table 1 summarises these results.

In this section we dealt with the simultaneous use of diﬀerent ﬁnancial attributes as criteria.

We concluded that a DM will usually want to use at most one ﬁnancial attribute from a single

class, and that a DM may use together ﬁnancial attributes from diﬀerent classes, but he/she

will usually want to use at most one attribute from both the ratio and rate of return classes.

5 On the selection of ﬁnancial methods

In this section we will try to establish a set of guidelines to help DMs choose the ﬁnancial

methods best suited to their speciﬁc situations. First, we will consider that the diﬀerent

decision situations are deﬁned according to ﬁve characteristics: degree of quantiﬁcation, capital

availability for the investments, degree of risk and uncertainty, interdependencies between

investments and existence of previously undertaken investments. We will try to deﬁne how

each of these characteristics shall be considered in the investment selection process. Next, we

will consider diﬀerent ﬁnancial methods, and deﬁne in which situations shall each method be

used.

The degree of quantiﬁcation deﬁnes whether or not ﬁnancial methods can be used to

evaluate the investments. In fact, ﬁnancial methods cannot be used unless quantitative data

about the investment costs and returns are available. The speciﬁc needs depend on the chosen

methods, and can vary from the project cash ﬂows (required by the NPV and IRR, for instance)

to more detailed accounting data (required by accounting methods). So, when quantitative

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 13

Table 1: Simultaneous use of diﬀerent ﬁnancial attributes, assuming that a coherent family of criteria

is wanted.

Class Attributes considered Simultaneous use

(same class)

Simultaneous use

(diﬀerent classes)

Equivalent Worth NPV

FW

AW

CW

Just one attribute. Attributes from all the

other classes can be si-

multaneously used.

Rate of Return IRR

ERR

MRIC

Just one attribute,

chosen according to

the DM’s reinvestment

assumptions and con-

cepts of investment

and return.

Attributes from all the

other classes except ra-

tio can be simultane-

ously used.

Ratio PI

B/C Ratio

Just one attribute,

chosen according to

the DM’s reinvestment

assumptions and con-

cepts of investment

and return.

Attributes from all the

other classes except

rate of return can be si-

multaneously used.

Payback Payback

Discounted Payback

Just one attribute,

chosen according to

the DM’s perceived

importance of the time

value of money.

Attributes from all the

other classes can be si-

multaneously used.

Accounting ROOI

ROAI

Just one attribute,

chosen according to

the DM’s concept of

investment.

Attributes from all the

other classes can be si-

multaneously used.

14 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

data are not available, diﬀerent methods (other than ﬁnancial) must be used to evaluate the

investments. In this paper we will not address the evaluation of investment projects in the

absence of quantitative information.

Capital availability for the investments is, perhaps, the most important characteristic to be

accounted for when selecting the ﬁnancial method(s) to be used (when quantitative data are

available), so we will discuss it in some detail. We will start by making some considerations

about management goals and NPV application. It is usually considered that the management

goal should be the maximisation of the company value. The company value is maximised when

all the investment projects with a positive NPV are undertaken, so many authors consider that

the NPV should be the preferred method to evaluate investment projects (see, for instance, [4]).

However, the use of NPV has some implicit assumptions - it assumes that markets are eﬃcient

and that the intermediate cash ﬂows can be reinvested at the discount rate in investments with

a similar systematic risk. About the ﬁrst assumption, Brealey and Myers [4] say the NPV will

only be weakened when the company owners cannot access an eﬃcient capital market. About

the second assumption, we think that, if the discount rate is properly calculated, it will usually

be met. Even if one of these assumptions does not hold, it is not always certain the existence

of a better method other than the NPV for the considered situation (although in some speciﬁc

cases better methods can be found).

From what was said we can conclude that, when unlimited capital is available for the

investments, the NPV should be the preferred method. Now, the following question can be

made: will any other methods be appropriate for this situation? In the previous section it

was said that all equivalent worth methods will yield equivalent results when properly applied,

so any other equivalent worth method can be used instead of the NPV. Methods from other

classes will not always lead to value maximisation, so they will not usually be as appropriate

for this situation as equivalent worth methods. However, there may be circumstances that do

not advise the use of the NPV nor the use of any other equivalent worth method. One of these

circumstances, which was previously referred in this section, has to do with the unavailability

of proper quantitative data and, when this happens, non-ﬁnancial attributes should be used.

Other circumstances have to do with the NPV assumptions (which are also implicit to the

other equivalent worth methods). For example, if the reinvestment assumption does not hold,

then maybe we can ﬁnd a method from other class that conforms the reinvestment situation

3

.

When it is considered that equivalent worth methods should not be used, rate of return and

ratio methods should be preferred to payback and accounting methods, since the former do

usually lead closer to value maximisation than the latter.

We will now suppose that the available capital for the investments is limited and known.

In this situation, it is seldom possible to undertake all investment projects with a positive

NPV, because the capital needed to ﬁnance all those investments may exceed the available

capital. Thus, the management should usually aim to build a portfolio of projects with an

aggregate NPV as high as possible. If all the available investment projects are known (which

is what usually happens when companies are preparing their annual investment plan), NPV

maximisation may be achieved by solving a mathematical programming problem. The vari-

ables of this problem will be the projects, their coeﬃcients in the objective function will be the

projects NPVs and the constraints will be the capital limitations for the considered periods

4

.

3

Another possible solution to this problem may be the deﬁnition of a new equivalent worth method , based

on an existing one (possibly the NPV), that will assume a diﬀerent reinvestment situation.

4

Such a problem can consider several periods, with capital limitations in each period.

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 15

See [10], chapter 8, for a survey of some mathematical programs for the optimal selection of a

set of projects.

Notice that any other equivalent worth method could be used instead of the NPV to build

this mathematical program. Also notice that, in some circumstances, the NPV assumptions

may not be met and, thus, methods from the equivalent worth class will be rendered inap-

propriate. When equivalent worth methods cannot be applied, one may think the use of a

mathematical program to maximise an aggregate rate of return or an aggregate ratio as the

best approach. However, rate of return and ratio attributes are not additive, resulting probably

in a complex non-linear mathematical programming problem. We think that a heuristic ap-

proach could consist in ordering the proﬁtable projects according to an adequate rate of return

or ratio attribute, and choosing the more proﬁtable projects until no other proﬁtable project

can be chosen without going beyond the available amount of capital. Asquith and Bethel [2]

argue that, if not all positive NPV projects are funded, then the use of such a heuristic may

also be preferable to the use of a NPV-based rule when forecast biases are present in the cash

ﬂow estimates and information about these biases is costly to obtain.

Sometimes it is known that the available amount of capital is not unlimited, but the limit

is not known. An ordering of the investment projects according to their proﬁtability will be

wanted in this situation, with non-proﬁtable projects being commonly excluded from that

ordering. Usually an ordering will be wanted such that, if projects are chosen accordingly, the

aggregate NPV will be maximised. Notice that the NPV will not be a good ordering criterion.

If we were to order projects according to the NPV, a project with a high NPV but needing a

very large investment might be considered better than other projects with only slightly lower

NPVs and much smaller investment required. As the available capital is limited, the latter

projects should be preferred to the former, so the ordering obtained through the use of NPV

would not lead to the maximisation of the aggregate NPV. In this situation, it would be better

to order projects according to an attribute from either the rate of return class or the ratio class.

Since methods from these classes evaluate the relation between the return and the investment,

projects with a higher return for each unit of invested capital will be rated higher. Thus, value

maximisation may be approximated if projects are ordered by using such an attribute.

When the available capital for investment projects is limited (known or unknown), it

may happen that the available amount is to be used not only with the investment proposals

currently known but also with possible investments to be proposed in a future period. As an

example, consider a company that has just raised a reasonable amount of capital and knows it

will be very diﬃcult to raise more capital in the near future. Notice that the uncertainty about

which investment projects will be proposed in the future can occur both when the amount of

available capital is known and when it is not known. An NPV-based ordering and selection

of investment projects will be inappropriate in this situation, and a better selection process

may be achieved with a rate of return or ratio method. We think this process should consist

in the determination of a hurdle rate (or hurdle ratio) and in the selection of all investment

projects having higher rates than that hurdle rate on the ﬁnancial attribute chosen to make the

decision. Historical data about past investment proposals considered by the company should

be used to determine the hurdle rate, when available. This hurdle rate should be periodically

re-examined according to the investments that have been proposed since its last calculation.

The degree of risk and uncertainty is not usually relevant to the choice of the ﬁnancial

methods. However, it will be important to the selection process, both because it may determine

16 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

whether or not risk analysis should be performed and because risk should be incorporated in

the evaluation of the projects. If the situation faced by the DMs is close to certainty, risk

can usually be neglected. On the other hand, when the situation is not close to certainty, risk

analysis should be performed on the investment projects. The risk analysis tools should be

chosen according to the available data. Monte Carlo simulation may give accurate information

about the statistical distribution of the ﬁnancial attributes, but it cannot be used unless there

are available data about the probability distributions of the relevant variables and a detailed

model of the investment projects is built accordingly [4]. Under some assumptions about

the distribution and statistical independence of the relevant variables, the variance can be

calculated for some ﬁnancial attributes, and it can be used as a measure of the total project

risk [8]. Sensitivity, scenarios and break-even analysis demand less data than Monte Carlo

simulation and may provide useful information about the investments [4]. Scenario analysis

may be particularly useful, since it allows the DM to assess how the project will behave

under diﬀerent circumstances. Each scenario will be deﬁned by a set of assumptions, and

will represent a consistent possibility of project behaviour. It is thus possible that diﬀerent

methods make sense in diﬀerent scenarios of the same project, according to the assumptions

of those scenarios.

There are also several methods to incorporate risk in project evaluation. Some of them, like

the certainty equivalent method that was referred in section 2, are based on the adjustment

of the cash ﬂows. Others, like the use of the CAPM, are based on the incorporation of risk

in the discount or compounding rates (when discounting or compounding based methods are

being used), or in the hurdle rate (when rate of return methods are being used). When some

sequential decisions must be made according to the outcomes of various events, decision trees

may be used to represent and evaluate the project. Recently, Option Pricing Theory has gained

a wide acceptance in the evaluation of investment projects. Option Pricing Theory provides

another tool for the calculation of the project NPV when an active management of the project

may inﬂuence its value, for example by reducing losses when the outcome is unfavourable

or by increasing proﬁts when the outcome is favourable. When options are involved, both

the traditional discounted cash ﬂow analysis (as performed by expression (1)) and the use of

decision trees may fail to correctly incorporate risk in the value of the project [14] and option

analysis may provide more accurate project values (see [4,14,15], for example, for more details).

Although both option valuation and decision trees are primarily used for the calculation of

the project NPV, other methods that can be based on the project value may also be adapted

to these types of valuations.

When a multicriteria evaluation of investment projects is being performed, risk also can

be incorporated through one or more attributes. Some ﬁnancial attributes are sometimes used

as risk proxies - in a NPV based evaluation, some rate of return or payback attributes can be

used as risk proxies. For a complete description of some risk analysis and risk incorporation

techniques, see [4,7,15]. For multicriteria models that incorporate risk, see [6] and the chapter

3 of [7].

Interdependencies are usually typiﬁed as synergies (positive or negative), mutual exclusion

and technical dependence. We can say there are synergies between investments A and B when

the ﬁnancial attributes of A change depending on whether or not B is pursued. Investments

A and B are mutually exclusive if the investments cannot be both pursued – that is, if one of

them is pursued, the other cannot be. Finally, we say that A is technically dependent on B if

A cannot be pursued unless B is pursued.

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 17

Interdependencies can be dealt with mathematical programming, in a similar way as the

one referred when we considered that the available capital was limited and known. However,

when we are dealing with non-additive ﬁnancial attributes such a program is complex, and it

may be diﬃcult to solve. When this happens, it is often useful to change all interdependencies

into mutual exclusion. When synergies exist between two investment projects, A and B, we can

turn these two investment projects into three mutually exclusive projects: project A, project

B and project AB, the latter of which consists in pursuing both projects A and B. Also, when

A is technically dependent on B, we can consider two mutually exclusive projects: B and AB,

the latter of which consists in pursuing both projects A and B. All interdependencies can be

thus turned into mutual exclusion, usually easier to deal with. For example, if a selection

process consists in ordering all projects according to an attribute, and then selecting projects

according to that order, it is very easy, after each project is selected, to seek and exclude all

the projects mutually exclusive with the selected project. Combinatorial analysis can also be

a very eﬀective tool to deal with interdependencies.

The selection process may be aﬀected by the existence of a portfolio of previously under-

taken investment projects. If such a portfolio does exist, it may or may not be possible to

abandon the previously undertaken projects.

The existence of previously undertaken projects that can be abandoned may be dealt

with by considering these projects at the same level of the newly proposed projects. When

re-evaluating existing projects, care should be taken not to consider unrecoverable costs or

past beneﬁts, and to properly consider their present salvage values as opportunity costs (the

costs of not selling the assets needed to continue the projects). These present salvage values

correspond to the investment cost in new projects. After this, the process may proceed as if

there were no previously undertaken projects.

If the previously undertaken projects cannot be abandoned, then it is not necessary to

re-evaluate these projects. However, some care should be taken in the selection process,

because it will still be necessary to consider the interdependencies between existing projects

and new projects, as well as the capital requirements of existing projects. Interdependencies

may eliminate some new projects (when there is mutual exclusion), or change the ﬁnancial

attributes of some of them (when there are synergies). When the available capital for the

investments is limited, the capital requirements of existing projects must be subtracted from

the available amount. After this, the selection process may proceed normally.

Thus far we have considered the decision situations deﬁned by a set of ﬁve characteristics,

and we have explained how should each of those characteristics aﬀect the investment selection

process. We argued that equivalent worth attributes should usually be used, and that rate

of return and ratio attributes should be used in particular situations instead of equivalent

worth attributes. Some questions may now be made, concerning whether or not accounting

and payback methods should ever be used. We will now address this issue.

Firstly, we will consider accounting methods. Only by chance these methods will lead to

aggregate NPV maximisation. This means that they should only be used when the DMs’

goal is diﬀerent from that. So, when will DMs use accounting methods ? The only answer

we have to this question has to do with the way the company’s results are made available to

stakeholders, which is usually in the form of accounting statements. If accounting methods are

used, it may be possible to present better accounting results and it is possible that stakeholders

will be happier and, thus, it may be easier for the company to raise money and managers may

18 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

get better rewards. However, there are two severe drawbacks to this strategy. The ﬁrst is that

it is a short term strategy - in the long term, accounting results will have a higher growth when

decisions are made to maximise the company value than when accounting methods are the only

criteria to select investments. The second is that intelligent and well informed stakeholders

will look beyond accounting results to get a clearer picture of the company, and this picture

will be more favourable if the management is aiming to maximise the company value than if

it is aiming to maximise short term accounting results.

We will now turn to payback methods. Some reasons may be presented for the use of these

methods. One of these reasons may be the existence of catastrophic or political risks that

may cause the company to lose, at any moment, the assets on which it invested, and so the

company may want to recover the invested capital as soon as possible. We do not think this

is always valid, because these risks can be accounted for, either through a proper adjustment

of the discount rate or through an adjustment of the predicted cash ﬂows, when equivalent

worth, rate of return and ratio methods are used.

Another reason has to do with the need of quickly obtaining liquidity, when it is diﬃcult

to raise capital either for new investment projects or for ﬁnancial engagements of the company

(the need to pay interests or principal in existing debts, for instance). Notice that such

situations will be related to capital markets imperfections. We think that, in these situations,

the use of mathematical programming to maximise the aggregate NPV according to some

constraints that have to do with liquidity necessities will provide better results than the use

of payback methods, if data are available. However, in these situations, we cannot ﬁnd any

reason to oppose the use of a payback attribute along with an equivalent worth attribute in a

multicriteria evaluation.

We will present another reason that is also related with possible capital markets imperfec-

tions. When the owners of the company may need money at any time and cannot access an

eﬃcient capital market either to raise money for their needs or to sell their stakes in the com-

pany, they may want the company to have high liquidity as often as possible. It can be argued

that, in this situation, payback methods will provide the best results to satisfy the company

owners wishes. Nevertheless we think that the use of mathematical programming to maximise

the aggregate NPV, according to a set of constraints that represent the possible needs of the

company owners, may provide better results. Once again, in this situation we cannot ﬁnd any

reason to oppose the use of a payback attribute along with an equivalent worth attribute in a

multicriteria evaluation.

In this section we have considered that the decision situations are deﬁned by a set of

characteristics, and we have explained how should each of those characteristics aﬀect the

investment selection process. We concluded that equivalent worth attributes should usually

be used, and that rate of return and ratio attributes should be used in particular situations

instead of equivalent worth attributes. Then, we tried to ﬁnd some situations in which the

use of accounting or payback methods might be appropriate. We concluded that accounting

methods might be used to achieve good short-term accounting results, but not to maximise

long-term accounting results. We considered the use of payback methods to account for the

company’s or its owners’ liquidity needs (particularly when capital markets are imperfect) and

also when the company faces political or catastrophic risks. We argued that equivalent worth

methods (eventually within a mathematical programming problem) are best suited for those

situations, but we could not ﬁnd any reason to oppose the use of a payback attribute along

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20 19

with an equivalent worth attribute in a multicriteria evaluation, in those situations.

6 Conclusions

This paper addressed the analysis and evaluation of investment projects within a multicriteria

framework. In this framework, all the properties, or characteristics, of the investments are

modelled as attributes. The decision criteria are chosen from the attribute set. In order to

have a correct structure for the decision problem, we deﬁned that the set of criteria should be

a coherent family of criteria, and should not include contradictory assumptions.

We started with a presentation of the most common methods for project evaluation. In

this presentation, we classiﬁed the methods into ﬁve classes, following a similar classiﬁcation

from [11,12]. Then we mathematically deﬁned the conditions that must be met by the set of

criteria in order to be a coherent family of criteria, and presented a mathematical result that

states that two criteria that order projects identically will be redundant.

The ﬁrst problem we dealt with was to ﬁnd out which ﬁnancial attributes can, and which

ones cannot, be used together as criteria. We concluded that attributes from the same class

either rank projects identically or are based in contradictory assumptions, so a DM will usually

use at most one attribute from a single class. Attributes form diﬀerent classes address diﬀerent

perspectives, or dimensions, of the proﬁtability, with the exception of the rate of return and

ratio classes, which both address the relation between the return and the investment. We also

stated that, for each ratio method, it is possible to deﬁne a rate of return method that ranks

projects identically, and that the converse is also usually true. So, according to our framework,

a DM may simultaneously use attributes from diﬀerent classes, but he/she will usually want

to use at most one attribute from both the classes of ratio and rate of return methods.

Then, we tried to establish a set of guidelines to help DMs choose the ﬁnancial methods

best suited to their speciﬁc situations. We considered a set of characteristics of the decision

situation – degree of quantiﬁcation, capital availability for the investments, degree of risk and

uncertainty, interdependencies between investments and existence of previously undertaken

investments - and analysed how these characteristics aﬀect the investment selection process.

The degree of quantiﬁcation deﬁnes whether or not ﬁnancial methods can be used. When

unlimited capital is available, or when the available capital is limited and known, equivalent

worth methods should usually be used. However, when the available capital is limited and

the limit is unknown, or when the available capital must also be used in future projects not

yet known, rate of return and ratio methods may be a better choice, even if the ultimate

goal is the maximisation of aggregate NPV. In the presence of risk, it will be important to

incorporate the risk in the project evaluation, and to perform risk analysis. We brieﬂy referred

some methodologies for risk analysis and for the incorporation of risk in project evaluation.

We explained that interdependencies could be dealt with through the use of mathematical

programming and combinatorial analysis. Finally we explained that, if a portfolio of previously

undertaken projects exists, these projects should be re-evaluated, and their interdependencies

with new projects and eﬀect in the available capital (if the available capital is limited) must

be considered in the evaluation process.

Since the analysis of the characteristics of the decision situation seemed to never recom-

mend the use of accounting and payback methods, we also tried to deﬁne in which situations

20 P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ c˜ ao Operacional, 24 (2004) 1-20

might these methods be used. We concluded that accounting methods might be used to achieve

good short-term accounting results, but not to maximise long-term accounting results. We con-

sidered some situations in which the use of payback methods seemed appropriate. Analysing

these situations, we always concluded that equivalent worth methods (eventually within a

mathematical programming problem) might be best suited to those situations. However, we

could not ﬁnd any reason to oppose the use of a payback attribute along with an equivalent

worth attribute in a multicriteria evaluation, in those situations.

7 References

[1] Afonso, A.R., Godinho, P.C. and Costa, J.P., Estrutura¸ c˜ ao dos Crit´erios de Avalia¸ c˜ ao e Selec¸ c˜ ao

de Projectos no

ˆ

Ambito de um Sistema de Apoio ` a Decis˜ ao (in Portuguese), Portuguese Journal

of Management Studies, IV (1999) 267-289.

[2] Asquith, D. and Bethel, J.E, Using Heuristics to Evaluate Projects: the Case of Ranking Projects

by IRR, The Engineering Economist, 40 (1995) 287-294.

[3] Bana e Costa, C., Structuration, Construction et Exploration d’un Mod`ele Multicrit`ere d’Aide ` a

la D´ecision (in French), PhD Thesis, Instituto Superior T´ecnico, Lisbon, Portugal (1992).

[4] Brealey, R. and Myers, S., Principles of Corporate Finance, 6th edition, McGraw-Hill (2000).

[5] Fahrni, P. and Spatig, M., An Application-Oriented Guide to R&D Project Selection and Eval-

uation Methods, R&D Management, 20 (1990) 155-171.

[6] Godinho, P.C. and Costa, J.P., A Note on the Use of Bicriteria Decision Trees in Capital Bud-

geting, Global Business and Economics Review, 4 (2002) 147-158.

[7] Hertz, D.B. and Thomas, H., Risk Analysis and its Applications, 1

st

Edition, John Wiley and

Sons (1983).

[8] Hillier, F.S., The Derivation of Probabilistic Information for the Evaluation of Risky Projects,

Management Science, 9 (1963) 443-457.

[9] McDaniel, W.R., McCarty, D.E. and Jessel, K.A., Discounted Cash Flow with Explicit Reinvest-

ment Rates: Tutorial and Extension, The Financial Review, 23 (1988) 369-385.

[10] Park, C.S. and Sharp-Bette, G.P., Advanced Engineering Economics, 1

st

Edition, John Wiley

and Sons (1990).

[11] Remer, D.S. and Nieto, A.P., A Compendium and Comparison of 25 Project Evaluation Tech-

niques. Part 1: Net Present Value and Rate of Return Methods, International Journal of Pro-

duction Economics, 42 (1995) 79-96.

[12] Remer, D.S. and Nieto, A.P., A Compendium and Comparison of 25 Project Evaluation Tech-

niques. Part 2: Ratio, Payback and Accounting Methods, International Journal of Production

Economics, 4 (1995) 101-129.

[13] Roy, B., Multicriteria Methodology for Decision Aiding, Kluwer Academic Press, Dordrecht

(1996).

[14] Trigeorgis, L. and Mason, S.P., Valuing Managerial Flexibility, Midland Corporate Finance Jour-

nal, Spring (1987) 14-21.

[15] Trigeorgis, L., Real Options: Managerial Flexibility and Strategy in Resource Allocation, The

MIT Press (1996).

2

P.C. Godinho, A.R. Afonso, J.P. Costa / Investiga¸ao Operacional, 24 (2004) 1-20 c˜

are, directly or indirectly, concerned with the proﬁtability of the investments, they do not always yield the same results - in fact, diﬀerent methods do sometimes yield contradictory results when evaluating or comparing the same investments. This fact raises some diﬃcult questions to DMs in charge of investment selection, concerning: • which method (or methods) should be used in a particular situation; • whether one or more method(s) should be used in a particular situation; • how should the results from diﬀerent methods be aggregated. To worsen matters, these methods cannot usually account for all the relevant information. One reason is that some strategic impacts of the investments are too complex to be properly quantiﬁed in the predicted earnings or cash ﬂows. Another reason is that there are usually some issues, not related with the proﬁtability of the investment, that DMs want to consider when they make an investment decision - these issues may refer to prestige, power or ethical concerns and are relevant to the DMs as individual human beings. This raises the question of knowing when and how these issues should be considered. In this paper we will address the analysis and evaluation of investment projects within a multicriteria framework. This framework will provide a theoretical basis to the aggregation of the results yielded by diﬀerent methods, and we believe it may also provide a basis for the aggregation of these with non-ﬁnancial factors. It will also allow the use of decision theory methods in investment decisions and, hopefully, avoid some decision errors due to an incorrect aggregation of factors. The framework we use is based upon the work of Bana e Costa [3] and Roy [13]. In this framework, all the properties, or characteristics, of the investments are modelled as attributes, and the results yielded by ﬁnancial methods will be called ﬁnancial attributes. Among the attributes, the decision maker (DM) will build a set of criteria, taking into account his/her concerns, values and beliefs. Some of the criteria may result from the aggregation of several attributes. First, we present the most widely used methods for project evaluation. Using a classiﬁcation based upon [11,12], we divide the most important ﬁnancial methods into ﬁve classes equivalent worth, rate of return, ratio, payback and accounting. In order to have a correct structure for the decision problem, we want the set of criteria to be a coherent family of criteria [3]. This means that we want the set of criteria to be exhaustive, cohese and non-redundant. Exhaustiveness means that all relevant criteria are included in the set of criteria. So, if any two alternatives are equal in all criteria they must be indiﬀerent for the DM, or else we must conclude that there is at least one relevant issue that is not properly accounted for by the set of criteria. Cohesion means that if two alternatives, A and B, are equal in all criteria but one, and A is better than B according to that criterion, then A must be preferred to B. A set of criteria is non-redundant if the removal of any criterion causes that set to be no longer both exhaustive and cohese. In section 3, we mathematically deﬁne these conditions and present a result that can be applied to the identiﬁcation of redundant criteria. Then, we try to ﬁnd out which ﬁnancial attributes can, and which ones cannot, be used together as criteria, assuming that we want the set of criteria to be a coherent family of criteria and to be based on non-contradictory assumptions and concepts. We try to deﬁne whether or

without following any particular situation taxonomy. due to additional reasons. We try to ﬁnd out which ﬁnancial method(s) is (are) best suited for each situation.12]. Possibly because of this.they treat ﬁnancial methods as a whole component that shall or shall not be used according to the situation. we try to consider all kinds of projects. while we aim to suggest one ﬁnancial method for each situation. ratio. we characterise a set of decision situations. rate of return.R. is usually the best choice. First.P. in many situations. In this section. following [11. One important diﬀerence is that. We consider some classes of methods and we try to deﬁne in which situations should the methods belonging to those classes be used. Fahrni and Spatig never suggest any particular ﬁnancial method . Godinho. 2 Financial methods A large number of ﬁnancial methods is presented in ﬁnancial textbooks and papers. the DMs may want to consider other methods. We do not exclude that. according to our framework. along with the NPV (or the rate of return or ratio method). J.P. we will describe some of the most important ﬁnancial methods and. Costa / Investiga¸ao Operacional. Next. In section 5 we try to establish a set of guidelines to help DMs choose the ﬁnancial methods best suited to their speciﬁc situations. or other method from the equivalent worth class. and whether or not ﬁnancial attributes from diﬀerent classes should be used together as criteria. but he/she will usually want to use at most one attribute from both the classes of ratio and rate of return methods. [1] perform a similar analysis. The NPV is the present monetary value of all the project cash ﬂows (including investment and salvage value) discounted at the appropriate discount rate. Afonso. and how should each characteristic aﬀect the investment selection process. 24 (2004) 1-20 c˜ 3 not ﬁnancial attributes from the same class should be used together as criteria.C. capital availability for the investments. we will categorise them into ﬁve classes: equivalent worth. rate of return or ratio methods may be the best suited for some particular situations. while Fahrni and Spatig focus on R&D projects. there are some important diﬀerences between them. Afonso et al. degree of risk and uncertainty. payback and accounting. A. interdependencies between investments and existence of previously undertaken investments. We also argue that a DM may use together attributes from diﬀerent classes. CFt is the cash ﬂow in period t and T is the horizon period . Also. Although this approach is similar to the approach of Fahrni and Spatig [5]. Equivalent worth methods examine the project cash ﬂows and. The traditional deﬁnition of the NPV is: T NPV = t=0 CFt (1 + r)t (1) where r is the discount rate. through discounting or compounding. our characterisation of the decision situations is diﬀerent from theirs. using a slightly diﬀerent classiﬁcation of the ﬁnancial methods. according to the degree of quantiﬁcation. We argue that a DM will usually want to use at most one ﬁnancial attribute from a single class. The most important of these methods is the net present value (NPV). However. We conclude that the net present value (NPV). resolve them to one equivalent cash ﬂow or to an equivalent series of cash ﬂows. we try a diﬀerent approach.

and it can be divided into systematic and unsystematic risk. if rf is the risk-free discount rate. deﬁned as the covariance of the asset (or project) returns with the market returns divided by the variance of the market returns. The future worth (FW). see [4]. This risk can be seen as the possible deviations from the expected project behaviour. Using the project betas. and it can be found through the compounding of the cash ﬂows to that future date. Also. for example. 24 (2004) 1-20 c˜ (which is often the project lifetime). J. whose cash ﬂows are constant over its lifetime. The measure used for this type of risk is the beta (β) coeﬃcient.P. in most situations. then the correct risk-adjusted rate will be: r = rf + β [E(rm ) − rf ] (5) For more details on the calculation of project betas and risk-adjusted discount rates. ﬁnancial theory prescribes that only the systematic risk shall be incorporated in the value of ﬁnancial assets and projects. The adjustment of the discount rate works very well for the NPV. The use of equivalent worth methods must include an adjustment for the risk of the project. and the systematic risk cannot be eliminated by diversiﬁcation. Godinho. β is the project beta and E(rm ) is the expected market return. the annual worth (AW) and the capitalised worth (CW) are other equivalent worth methods. the use of such a risk-adjusted discount rate in the other equivalent worth methods will not correctly adjust for the project risk. and it is the only type of risk that should matter to investors with diversiﬁed portfolios. To deal with this problem we suggest the adjustment of the cash ﬂows.R. calculated according to the Capital Asset Pricing Model (CAPM).4 P. The AW is the value of each of the cash ﬂows of an equivalent project (having the same NPV) with a ﬁnite lifetime (usually identical to the lifetime of the original project or to the considered horizon period). [4]). The NPV is nowadays considered by many authors to be. According to the CAPM. Afonso. it is possible to incorporate the systematic risk in the project NPV through the use of a risk-adjusted discount rate. The CW is equal to the AW except that it considers another equivalent project with an inﬁnite lifetime. The remaining notation was previously deﬁned. The unsystematic risk can be eliminated by holding a diversiﬁed portfolio of investments. the best economic proﬁtability measure for investment projects (see. The FW is the monetary value of all the project cash ﬂows in a future date. T0 is the future moment for which the FW is calculated. but not for the other equivalent worth methods. In fact. These methods can be deﬁned in the following way: T FW = t=0 T CFt (1 + r)T0 −t = N P V (1 + r)T0 T (2) AW = t=0 T t=1 T CFt (1+r)t 1 (1+r)t = CFt (1+r)t t=0 1 1 r − r(1+r)T = 1 r NPV 1 − (3) T r(1+r) CW = t=0 ∞ t=1 CFt (1+r)t 1 (1+r)t T = t=0 CFt (1 + r)t · r = NPV · r (4) In the deﬁnition of FW. Costa / Investiga¸ao Operacional. instead of the adjustment of the discount .C. A.

we can deﬁne: T t=1 CFt (1 + r)T −t (1 + ERR)T = I0 (8) The MRIC is deﬁned in [9]. and operating cash ﬂows. the period t cash ﬂow is adjusted to its certainty equivalent. to the beginning of the project and operating cash ﬂows are compounded.capital cash ﬂows. to the end of the project. This rate can be deﬁned as the discount rate for which the NPV equals zero. which is deﬁned as the quotient between the present value of the future cash ﬂows generated by the project and the initial investment: T CFt (1+r)t PI = t=1 I0 (10) . So. Costa / Investiga¸ao Operacional.P. Two kinds of cash ﬂows are considered . The most widely used rate of return method is the internal rate of return (IRR). and corresponds to the yield-to-maturity on a bond. r f being the risk-free discount rate and r being the risk-adjusted discount rate. by using certainty equivalents of the cash ﬂows ([4]. Both these methods consider an explicit reinvestment rate.P. at the same reinvestment rate. 24 (2004) 1-20 c˜ 5 rate. A. the MRIC can be deﬁned as: T T t=0 CCFt = (1 + r)t OCFt (1 + r)T −t (1 + M RIC)T (9) t=1 where CCFt is the capital cash ﬂow for period t and OCFt is the operating cash ﬂow for period t. chapter 9). Afonso. However. at the reinvestment rate. generated by the project. at the reinvestment rate. The ERR is the rate for which the future worth of the initial investment equals the future worth of the other cash ﬂows compounded. to the end of the project.R. CE = CFt · 1 + rf 1+r t (6) and then the risk-free discount rate is used. Rate of return methods measure the rate at which the invested capital will grow if the project is pursued. Speciﬁcally. Godinho. The most signiﬁcant ratio methods can be deﬁned as the quotient between the present value of the returns and the present value of the investment. Using I0 to represent the initial investment. Capital cash ﬂows are discounted. The MRIC is the rate at which the present value of the capital cash ﬂows should be compounded so that it would equal the future value of the operating cash ﬂows at the end of the project. used to ﬁnance the project.C. It can be calculated by solving the following equation: T CFt (7) t =0 t=0 (1 + IRR) Other methods from the rate of return class include the external rate of return (ERR) and the marginal return on invested capital (MRIC). the adjustment of cash ﬂows will allow the correct comparison of projects with diﬀerent systematic risk. The most widely used ratio method is the proﬁtability index (PI). J. For the NPV. for the other methods. it is indiﬀerent to adjust the discount rate or the cash ﬂows.

24 (2004) 1-20 c˜ Other ratio methods can be deﬁned as the ratio between the present value of the returns and the present value of the investment. The ROOI is the quotient between the average yearly accounting proﬁt.k. is the quotient between the present value of all cash ﬂows. J. A. and the investment made in the project. The discounted payback period is similar to the payback period. among others.P. for instance. 3 Mathematical deﬁnitions and results This section provides some mathematical results used to ﬁnd out which attributes can and which ones cannot be used together as criteria.C.6 P. discussing its advantages and drawbacks and also discussing the calculation of the discount or compounding rate needed by some of them. a.R. excluding the initial investment and the salvage value. These methods include the payback period and the discounted payback period.k. The payback period is the number of years required for the accumulated project cash ﬂows to equal the initial investment. [4] and [10] also describe some of these methods. average book method). So. We consider that.a. Costa / Investiga¸ao Operacional. Godinho. If we use SVT to represent the salvage value of the project. which excludes depreciation. original book method) and the return on average investment (ROAI. we can deﬁne: k Payback = min k : t=1 CFt ≥ I0 k (12) Discounted Payback = min k : t=1 CFt ≥ I0 (1 + r)t (13) Accounting methods consider proﬁtability from an accounting perspective.12]. we can deﬁne: T APt t=1 ROAI = T T (14) BVt t=0 T +1 T APt t=1 ROOI = T I0 (15) A further description of most of these ﬁnancial methods can be found in [11. in order to have a correct . Afonso. a. except that it considers the discounted cash ﬂows instead of the raw cash ﬂows. The ROAI is the quotient between the average yearly accounting proﬁt and the average book value (average value of the diﬀerence between investment and depreciation) during the project life. Using APt to represent accounting proﬁt in period t and BVt to represent book value in period t.a. and the diﬀerence between the initial investment and the present worth of the salvage value. The beneﬁt-cost ratio (B/C ratio). This class includes the return on original investment (ROOI. we can deﬁne: T t=1 CFt (1+r)t − B/C = SVT (1+r)T I0 − SVT (1+r)T (11) Payback methods calculate how long it takes to recover the invested capital.

gr can be removed from F without the exhaustiveness and cohesion conditions ceasing to hold. Godinho. that a larger value in a given attribute is always better than a smaller one. We will now mathematically deﬁne these conditions and we will show that two attributes that rank projects identically will be redundant. ∃ai . Assumptions Let A={a1 . Deﬁnitions The symbols P and I will be used as comparison operators: ai P aj means that ai is considered to be preferred to aj and ai I aj means that ai and aj are considered to be indiﬀerent. gm } the set of attributes used as criteria. . an } be the set of projects and F={g1 . Costa / Investiga¸ao Operacional.P. We will consider that.aj ∈A: {[(gk (ai )=gk (aj ). . A and B. we have: ∀ ai . ∀gk ∈F) ⇒ ai I aj (Cohesion) ∀ai . gr (ai ) >gr (aj ) ⇒ gs (ai ) >gs (aj ) (19) If the criteria set F includes both gr and gs then F is not a coherent family of criteria because.gl }∧gl (ai ) >gl (aj )∧(aj P ai ∨ ai I aj )]} (16) (17) (18) The ﬁrst part of expression (18) means that if any criterion gp is removed then F will no longer meet the exhaustiveness condition. a2 .aj ∈A. . A set of criteria is non-redundant if the removal of any criterion causes that set to be no longer both exhaustive and cohese. cohese and nonredundant. and A is better than B according to that criterion. g2 .aj ∈A. . Cohesion means that if two alternatives. then A must be preferred to B. . A.C. cohesion and non-redundancy conditions: (Exhaustiveness) ∀ai . As was said before. if any two alternatives are equal in all criteria they must be indiﬀerent for the DM.R. ∀gl ∈F.aj ∈A. for two attributes gr and gs . we want the set of criteria to be a coherent family of criteria [3]. Exhaustiveness means that all relevant criteria are included in the set of criteria. So. ∀gk ∈F\{gp })∧(ai P aj ∨ aj P ai )] ∨ [∃gl ∈F\{gp }:gk (ai )=gk (aj ). . (gk (ai )=gk (aj ). We thus say that gr is redundant. without loss of generality. the second part of that expression means that if any criterion gp is removed then F will no longer meet the cohesion condition. ∀gk ∈F\{gp . ∀gk ∈F\{gl }∧ gl (ai ) >gl (aj )) ⇒ ai P aj (Non-redundancy) ∀gp ∈F. or else we must conclude that there is at least one relevant issue that is not properly accounted for by the set of criteria. J. in order to be a coherent family of criteria. if both the exhaustiveness and cohesion conditions are met. are equal in all criteria but one. . gk (ai ) will be the performance of project ai according to attribute gk . then the non-redundancy condition is not met. (gk (ai )=gk (aj ). Theorem (attribute redundancy): Let us assume that. this means that the set of criteria must be exhaustive. Afonso.P. 24 (2004) 1-20 c˜ 7 structure for the decision problem. . the set F must meet the following exhaustiveness. . Let us also assume. Speciﬁcally.

then F\{gr } also meets the exhaustiveness condition and (b) if F meets the cohesion condition. This means that we may only have gs (ai )=gs (aj ) when gr (ai )=gr (aj ). ∀gk ∈F\{gr . then we have gs (ai )=gs (aj ) and. if gk (ai )=gk (aj ).P. Let us now prove (b). then they will be redundant (only one of them should be used as criterion). Let us start by proving (a). for gl ∈F\{gs . (19) says that gr (ai ) >gr (aj ) ⇒ gs (ai ) >gs (aj ).gl }∧gl (ai ) >gl (aj ) ⇒ ⇒ gk (ai )=gk (aj ). Theorem proof: We will show that (a) if F meets the exhaustiveness condition. thus gr is redundant. Since gk (ai )=gk (aj ). proving (a).gr }. So: gk (ai )=gk (aj ). A.gr }. 24 (2004) 1-20 c˜ Corollary: If two attributes rank the projects identically. by (20) we have gr (ai )=gr (aj ). Costa / Investiga¸ao Operacional.gl }∧gl (ai ) >gl (ak ) then ai P aj . ∀gk ∈F\{gr } ⇒ gs (ai )=gs (aj ) (since gs ∈F\{gr }) (21) ⇒ gr (ai )=gr (aj ) (using (20)) And so: gk (ai )=gk (aj ). So: ∀ ai . Godinho. and gl ∈F\{gs . thus gr (aj ) >gr (ai ) ⇒ gs (aj ) >gs (ai ).C. if F meets the cohesion condition (if (17) holds) then F\{gr } will also meet the cohesion condition.8 P. Afonso. gs (ai )=gs (aj ) ⇒ gr (ai )=gr (aj ) Also: (20) gk (ai )=gk (aj ). (gk (ai )=gk (aj ). ∀gk ∈F\{gl }∧gl (ai ) >gl (aj ) ai P aj (since the cohesion condition holds) (24) . We will thus prove that.gr }.∀gk ∈F\{gr .gl }∧gl (ai ) >gl (ak )) ⇒ ai P aj (23) We will start by showing that (23) holds for all gl ∈F\{gs . This will show that gr can be removed from F without the exhaustiveness and cohesion conditions ceasing to hold. We will prove that. then F\{gr } also meets the cohesion condition. ∀gk ∈F ⇒ a i I aj (using (21)) (22) (because F meets the exhaustiveness condition) (22) means that F\{gr } meets the exhaustiveness condition.gl }.R.∀gk ∈F\{gr . meaning that for any projects a i and aj : ∀gl ∈F\{gr }.aj ∈A. ∀gk ∈F\{gr } ⇒ gk (ai )=gk (aj ). J. ∀gk ∈F\{gr .

R. Godinho. the data and methods used in each scenario should not include contradictory assumptions or concepts. if F meets the cohesion condition. but also to be based on non-contradictory assumptions or concepts. they shall not be used together as criteria. so that each scenario represents a consistent possibility of project behaviour.gs }∧ gs (ai ) >gs (aj ) ⇒ gr (ai ) ≥gr (aj ) (26) (25) Let us analyse the expression (26). gr will be redundant. In this analysis we will assume that we want the set of criteria to be a coherent family of criteria. If ai is also better than aj in gr . So: gk (ai )=gk (aj ). This means that. From (19) we can say that gs (ai ) ≤gs (aj ) ⇒ gr (ai ) ≤gr (aj ) and. we will show that (23) holds for gl =gs . So. ∀gk ∈F\{gr . consequently: gs (ai ) >gs (aj ) ⇒ gr (ai ) ≥gr (aj ) Using (25) we get. J. and also to be based on non-contradictory assumptions or concepts. since they will be redundant. The corollary of this theorem can now be used to ﬁnd redundant attributes. be used together as criteria. when (19) holds. we will only consider the ﬁnancial methods presented in section 2. in the presence of risk. 4 On the simultaneous use of diﬀerent ﬁnancial attributes In this section we will try to ﬁnd out which ﬁnancial attributes can.P. if these methods can be properly classiﬁed into one of the ﬁve classes we are considering. Afonso. then F\{gr } also meets the same condition. 24 (2004) 1-20 c˜ 9 To complete the proof. For that purpose we will initially consider ﬁnancial attributes from the same class. We will want the set of criteria to be not only a coherent family of criteria. we will not discuss whether or not two attributes happen to rank the projects identically for a particular set of projects. and then we will consider attributes from diﬀerent classes. and which ones cannot. We . gr . it may be worthwhile to consider the behaviour of the project under diﬀerent scenarios (as will be discussed in section 5). We acknowledge that. Whenever two diﬀerent attributes rank projects identically. F\{gr } also meets the same condition. so the theorem proof is complete. then ai is also considered to be preferred to aj . It says that ai and aj are equal in all criteria except gs and. However. In the analysis of ﬁnancial attribute redundancy. A. which include the most common ﬁnancial methods. for gl =gs : gk (ai )=gk (aj ). possibly. then we have ai P aj (because gs (ai ) >gs (aj ). (22) shows that if F meets the exhaustiveness condition.gs }∧ gs (ai ) >gs (aj ) ⇒ a i P aj (27) We showed that (23) holds for all attributes gl ∈F\{gr }.P. We believe our conclusions are extensible to other methods.C. ∀gk ∈F\{gr . Since F meets the cohesion condition. In the following discussion. if gr (ai )=gr (aj ). Costa / Investiga¸ao Operacional.

24 (2004) 1-20 c˜ will only consider that two attributes are redundant if they always rank projects identically. The interpretation of the NPV provides some advantages over the other equivalent worth attributes. The ERR diﬀers from the MRIC on the concepts of investment and return. Costa / Investiga¸ao Operacional. J. the B/C ratio assumes investment to be the diﬀerence between the initial investment and its salvage value. ∞ is +∞. at most one attribute from the equivalent worth class should be used as criterion. the ERR considers investment to be only the initial investment. even when the project risk is diﬀerent. and will still be valid when some projects are removed from or added to the initial set. we cannot say that rate of return attributes are redundant. for r≤0. we will address the simultaneous use of diﬀerent ﬁnancial attributes from the equivalent worth class. special circumstances may advise the use of a diﬀerent attribute. It is well known that diﬀerent rate of return methods may rank the same projects diﬀerently. A. This way. in which case the perpetuity factor equals 1 . ratio methods may also rank the same projects diﬀerently. suggested in section 2. according to the non-redundancy demand of a coherent family of criteria. show that the NPV.P. the CW is only deﬁned for r>0. Several constraints must be met by the methods from this class. While the MRIC considers investment to be all the capital cash ﬂows.10 P. While the IRR assumes that the proﬁts are reinvested at a rate equal to the IRR. Let us consider the NPV and the CW. Thus. Park and Sharp-Bette [10]. all equivalent worth methods rank projects identically.+∞[. It is easy to prove that. If more than one rate of return attribute is used in the evaluation process. We will now consider the rate of return methods. That is because they also assume diﬀerent concepts of investment and return. One example: the projects being compared should have the same discount rate1 [11]. in which it has economic meaning. in which case: t=1 1 . Like the rate of return methods. We will assume that the discount/reinvestment rate belongs to the interval ]-1. we r must assume r>0. so the NPV will usually be used. our results will be independent of any particular set of projects. in general only one should be used. That is because the results obtained depend on the reinvestment assumptions and on the implicit concepts of investment and return for each method. Thus. Godinho. and return to be all the other cash ﬂows excluding the salvage value.C. and that the projects being compared have the same discount rate. While the PI assumes investment to be the initial investment and all the other cash ﬂows to be return. To start with. So. when properly applied. the CW will be equal to the NPV divided by the perpetuity factor. the AW and the FW rank projects identically. the ERR and the MRIC consider an explicit reinvestment rate. Afonso. allows us to use the same discount rate for all the projects. So. chapter 7. However. then contradictory assumptions or concepts will be simultaneously involved. (1+r)t Since. However. As diﬀerent rate of return methods can yield contradictory results. the perpetuity factor NPV(ai ) > NPV(aj ) ⇔ ⇔ NPV(ai )·r > NPV(aj )·r (28) CW(ai ) > CW(aj ) So. the NPV and the CW rank projects identically. 1 . The adjustment of cash ﬂows. even if it is necessary to use certainty equivalent cash ﬂows to achieve that. chosen according to the DM’s reinvestment assumptions and concepts of investment and return.R. so that they can be properly applied.

As an example. thus they are redundant. these methods can yield diﬀerent results.R. the ROAI assumes that it is the average book value. r is the discount rate (equal to the MRIC reinvestment rate) and T is the horizon period. J. and that this reinvestment rate is equal for both projects. we can see that the PI is equivalent to the ERR. we can deﬁne a ratio method that is equivalent. and thus they are not redundant. So. 2 . the payback period does not. because these methods do not take the time value of money into account. Costa / Investiga¸ao Operacional. it is arguable that a DM will consider them relevant. Accounting attributes should only be used when accounting issues are considered important. that matters. This is the case of the ERR and the MRIC. Once more. We will now consider the simultaneous use of attributes from diﬀerent classes.P.C. according to whether or not the time value of money is considered important.P. 24 (2004) 1-20 c˜ 11 as diﬀerent ratio methods may rank projects diﬀerently. if we assume that the rate r belongs to the interval ]-1. That is because while the discounted payback period takes the time value of money into account. or the liquidity recovery speed. chosen according to the DM’s concepts of investment and return. However. Accounting methods diﬀer in what they consider relevant about the investment. or dimension. This means they are not redundant. a DM will usually want to use at most one accounting attribute. but usually at most one of them will be used. according to the type of depreciation associated with each project. As for the rate of return and ratio classes. in which it has economic meaning. we can deﬁne an equivalent ratio method (let us call it Modiﬁed Proﬁtability Index. MPI) as: T t=1 T t=0 OCFt (1+r)t CCFt (1+r)t MPI = (29) where OCFt is the operating cash ﬂow in period t. A DM will usually consider the time value of money relevant and will thus prefer the discounted payback period. The equivalent worth class addresses the absolute value of the project. after the yearly depreciation. The opposite is also usually true 2 .+∞[. according to what he/she thinks is more signiﬁcant about the investment: the whole book value or the average book value. Afonso. However. we will try to ﬁgure out which perspective. Godinho. Payback methods may rank the same projects diﬀerently. accounting methods address the accounting proﬁtability and payback methods address the time it takes to recover the initial investment. To start with. Moreover. we can see that for each ratio method we can ﬁnd out or deﬁne a rate of return method that is equivalent in the sense that it ranks projects in the same way and always leads to the same accept/reject decisions. CCFt is the capital cash ﬂow in period t. While the ROOI assumes that the whole value of the investment (the whole book value) matters. we cannot say that their results are redundant. A. of the proﬁtability does each class address. and lead to diﬀerent accept/reject decisions. only one ratio attribute should be usually used. we have: MRIC(ai ) > MRIC(aj ) ⇔ 1 + MRIC(ai ) > 1 + MRIC(aj ) For each rate of return method that uses an external rate to resolve all the cash ﬂows into two cash ﬂows. For the MRIC.

J. Thus. in fact. they both address the relation between the return and the investment. In this section we dealt with the simultaneous use of diﬀerent ﬁnancial attributes as criteria. since they address diﬀerent proﬁtability dimensions. This relation between rate of return and ratio methods led us to the conclusion that they would probably address the same proﬁtability dimension. Afonso. ﬁnancial methods cannot be used unless quantitative data about the investment costs and returns are available.P. We will try to deﬁne how each of these characteristics shall be considered in the investment selection process. as both classes address the same dimension. rank projects identically. It would also be easy to deﬁne a rate of return method equivalent to the B/C ratio. So.C. Park and Sharp-Bette [10]. And. A. The degree of quantiﬁcation deﬁnes whether or not ﬁnancial methods can be used to evaluate the investments. and that a DM may use together ﬁnancial attributes from diﬀerent classes. show that some rate of return and ratio methods. degree of risk and uncertainty. attributes from diﬀerent classes can be simultaneously used as criteria. but he/she will usually want to use at most one attribute from both the ratio and rate of return classes. given that they are not based on contradictory assumptions or concepts. we should usually use at most one attribute from both these two classes. Table 1 summarises these results. and can vary from the project cash ﬂows (required by the NPV and IRR. First.12 P. 24 (2004) 1-20 c˜ T OCFt (ai ) · (1 + r)T −t T t=0 T T OCFt (aj ) · (1 + r)T −t T t=0 CCFt (aj ) (1+r)t ⇔ T t=1 CCFt (ai ) (1+r)t >T t=1 ⇔ (1 + r)T T t=1 T t=0 OCFt (ai ) (1+r)t CCFt (ai ) (1+r)t T T > (1 + r)T t=1 T t=0 OCFt (aj ) (1+r)t CCFt (aj ) (1+r)t ⇔ t=1 T t=0 OCFt (ai ) (1+r)t CCFt (ai ) (1+r)t > t=1 T t=0 OCFt (aj ) (1+r)t CCFt (aj ) (1+r)t ⇔ MPI(ai ) > MPI(aj ) (30) So. Next. Costa / Investiga¸ao Operacional. We concluded that a DM will usually want to use at most one ﬁnancial attribute from a single class. Godinho. chapter 7. capital availability for the investments. when quantitative . As for the other classes. for instance) to more detailed accounting data (required by accounting methods). and deﬁne in which situations shall each method be used. 5 On the selection of ﬁnancial methods In this section we will try to establish a set of guidelines to help DMs choose the ﬁnancial methods best suited to their speciﬁc situations. including the ERR and the PI.R. we will consider that the diﬀerent decision situations are deﬁned according to ﬁve characteristics: degree of quantiﬁcation. interdependencies between investments and existence of previously undertaken investments. In fact. The speciﬁc needs depend on the chosen methods. the MRIC and the MPI rank projects identically. we will consider diﬀerent ﬁnancial methods.

Just one attribute. chosen according to the DM’s concept of investment.P. chosen according to the DM’s reinvestment assumptions and concepts of investment and return. Attributes from all the other classes can be simultaneously used. Simultaneous use (diﬀerent classes) Attributes from all the other classes can be simultaneously used. A. chosen according to the DM’s reinvestment assumptions and concepts of investment and return. Class Equivalent Worth Attributes considered NPV FW AW CW IRR ERR MRIC Simultaneous use (same class) Just one attribute. Costa / Investiga¸ao Operacional.P. . Attributes from all the other classes except ratio can be simultaneously used. J. Afonso. assuming that a coherent family of criteria is wanted. Just one attribute.C. Attributes from all the other classes except rate of return can be simultaneously used. Just one attribute. Godinho. Attributes from all the other classes can be simultaneously used. Rate of Return Ratio PI B/C Ratio Payback Payback Discounted Payback Accounting ROOI ROAI Just one attribute. 24 (2004) 1-20 c˜ 13 Table 1: Simultaneous use of diﬀerent ﬁnancial attributes. chosen according to the DM’s perceived importance of the time value of money.R.

A. Godinho. In this paper we will not address the evaluation of investment projects in the absence of quantitative information. rate of return and ratio methods should be preferred to payback and accounting methods. for instance. 4 Such a problem can consider several periods. which was previously referred in this section. NPV maximisation may be achieved by solving a mathematical programming problem. From what was said we can conclude that. When it is considered that equivalent worth methods should not be used. Now. then maybe we can ﬁnd a method from other class that conforms the reinvestment situation 3 . The company value is maximised when all the investment projects with a positive NPV are undertaken. perhaps. so they will not usually be as appropriate for this situation as equivalent worth methods. based on an existing one (possibly the NPV). the following question can be made: will any other methods be appropriate for this situation? In the previous section it was said that all equivalent worth methods will yield equivalent results when properly applied. if the discount rate is properly calculated. non-ﬁnancial attributes should be used. Other circumstances have to do with the NPV assumptions (which are also implicit to the other equivalent worth methods). Methods from other classes will not always lead to value maximisation. The variables of this problem will be the projects. For example. J. that will assume a diﬀerent reinvestment situation. it will usually be met. We will now suppose that the available capital for the investments is limited and known.R. In this situation. Costa / Investiga¸ao Operacional. One of these circumstances. We will start by making some considerations about management goals and NPV application. because the capital needed to ﬁnance all those investments may exceed the available capital. with capital limitations in each period. Another possible solution to this problem may be the deﬁnition of a new equivalent worth method . there may be circumstances that do not advise the use of the NPV nor the use of any other equivalent worth method. the management should usually aim to build a portfolio of projects with an aggregate NPV as high as possible. Thus. when unlimited capital is available for the investments. has to do with the unavailability of proper quantitative data and. so any other equivalent worth method can be used instead of the NPV. diﬀerent methods (other than ﬁnancial) must be used to evaluate the investments.it assumes that markets are eﬃcient and that the intermediate cash ﬂows can be reinvested at the discount rate in investments with a similar systematic risk. It is usually considered that the management goal should be the maximisation of the company value. if the reinvestment assumption does not hold. the use of NPV has some implicit assumptions . Even if one of these assumptions does not hold. the NPV should be the preferred method. However. so many authors consider that the NPV should be the preferred method to evaluate investment projects (see. their coeﬃcients in the objective function will be the projects NPVs and the constraints will be the capital limitations for the considered periods 4 . [4]). we think that.14 P. About the second assumption. 24 (2004) 1-20 c˜ data are not available. it is not always certain the existence of a better method other than the NPV for the considered situation (although in some speciﬁc cases better methods can be found). However. since the former do usually lead closer to value maximisation than the latter. when this happens. Capital availability for the investments is. Afonso.C. About the ﬁrst assumption. so we will discuss it in some detail. 3 . the most important characteristic to be accounted for when selecting the ﬁnancial method(s) to be used (when quantitative data are available). If all the available investment projects are known (which is what usually happens when companies are preparing their annual investment plan). it is seldom possible to undertake all investment projects with a positive NPV.P. Brealey and Myers [4] say the NPV will only be weakened when the company owners cannot access an eﬃcient capital market.

the aggregate NPV will be maximised. chapter 8. Sometimes it is known that the available amount of capital is not unlimited. when available. rate of return and ratio attributes are not additive. in some circumstances. a project with a high NPV but needing a very large investment might be considered better than other projects with only slightly lower NPVs and much smaller investment required. Notice that the uncertainty about which investment projects will be proposed in the future can occur both when the amount of available capital is known and when it is not known. The degree of risk and uncertainty is not usually relevant to the choice of the ﬁnancial methods. This hurdle rate should be periodically re-examined according to the investments that have been proposed since its last calculation. As the available capital is limited. but the limit is not known. An ordering of the investment projects according to their proﬁtability will be wanted in this situation. resulting probably in a complex non-linear mathematical programming problem. consider a company that has just raised a reasonable amount of capital and knows it will be very diﬃcult to raise more capital in the near future. Afonso. if projects are chosen accordingly. A. Historical data about past investment proposals considered by the company should be used to determine the hurdle rate. Godinho. if not all positive NPV projects are funded. Since methods from these classes evaluate the relation between the return and the investment. both because it may determine . with non-proﬁtable projects being commonly excluded from that ordering. When the available capital for investment projects is limited (known or unknown). and choosing the more proﬁtable projects until no other proﬁtable project can be chosen without going beyond the available amount of capital. then the use of such a heuristic may also be preferable to the use of a NPV-based rule when forecast biases are present in the cash ﬂow estimates and information about these biases is costly to obtain. the NPV assumptions may not be met and. As an example. projects with a higher return for each unit of invested capital will be rated higher. thus. Notice that the NPV will not be a good ordering criterion. If we were to order projects according to the NPV. Thus. it may happen that the available amount is to be used not only with the investment proposals currently known but also with possible investments to be proposed in a future period.P. However. Asquith and Bethel [2] argue that. and a better selection process may be achieved with a rate of return or ratio method. the latter projects should be preferred to the former. one may think the use of a mathematical program to maximise an aggregate rate of return or an aggregate ratio as the best approach. Notice that any other equivalent worth method could be used instead of the NPV to build this mathematical program. so the ordering obtained through the use of NPV would not lead to the maximisation of the aggregate NPV. However. Costa / Investiga¸ao Operacional. value maximisation may be approximated if projects are ordered by using such an attribute. J. In this situation. An NPV-based ordering and selection of investment projects will be inappropriate in this situation. for a survey of some mathematical programs for the optimal selection of a set of projects. Also notice that. Usually an ordering will be wanted such that.P.C. We think that a heuristic approach could consist in ordering the proﬁtable projects according to an adequate rate of return or ratio attribute.R. it would be better to order projects according to an attribute from either the rate of return class or the ratio class. 24 (2004) 1-20 c˜ 15 See [10]. it will be important to the selection process. methods from the equivalent worth class will be rendered inappropriate. When equivalent worth methods cannot be applied. We think this process should consist in the determination of a hurdle rate (or hurdle ratio) and in the selection of all investment projects having higher rates than that hurdle rate on the ﬁnancial attribute chosen to make the decision.

and will represent a consistent possibility of project behaviour. If the situation faced by the DMs is close to certainty. decision trees may be used to represent and evaluate the project. the other cannot be. On the other hand. The risk analysis tools should be chosen according to the available data. Others. according to the assumptions of those scenarios. It is thus possible that diﬀerent methods make sense in diﬀerent scenarios of the same project. When a multicriteria evaluation of investment projects is being performed. when the situation is not close to certainty. . Although both option valuation and decision trees are primarily used for the calculation of the project NPV. Afonso. other methods that can be based on the project value may also be adapted to these types of valuations.R. both the traditional discounted cash ﬂow analysis (as performed by expression (1)) and the use of decision trees may fail to correctly incorporate risk in the value of the project [14] and option analysis may provide more accurate project values (see [4. for more details). risk also can be incorporated through one or more attributes. like the use of the CAPM.15]. we say that A is technically dependent on B if A cannot be pursued unless B is pursued. Investments A and B are mutually exclusive if the investments cannot be both pursued – that is. are based on the incorporation of risk in the discount or compounding rates (when discounting or compounding based methods are being used). Finally. Under some assumptions about the distribution and statistical independence of the relevant variables. some rate of return or payback attributes can be used as risk proxies. Monte Carlo simulation may give accurate information about the statistical distribution of the ﬁnancial attributes. Some of them.in a NPV based evaluation. J. Sensitivity.14. Each scenario will be deﬁned by a set of assumptions. We can say there are synergies between investments A and B when the ﬁnancial attributes of A change depending on whether or not B is pursued. but it cannot be used unless there are available data about the probability distributions of the relevant variables and a detailed model of the investment projects is built accordingly [4]. Interdependencies are usually typiﬁed as synergies (positive or negative). There are also several methods to incorporate risk in project evaluation. for example by reducing losses when the outcome is unfavourable or by increasing proﬁts when the outcome is favourable. if one of them is pursued. Godinho. are based on the adjustment of the cash ﬂows. see [4.C. the variance can be calculated for some ﬁnancial attributes. risk can usually be neglected.P. like the certainty equivalent method that was referred in section 2. Costa / Investiga¸ao Operacional. For a complete description of some risk analysis and risk incorporation techniques. since it allows the DM to assess how the project will behave under diﬀerent circumstances. and it can be used as a measure of the total project risk [8].16 P. For multicriteria models that incorporate risk. When some sequential decisions must be made according to the outcomes of various events. risk analysis should be performed on the investment projects. Some ﬁnancial attributes are sometimes used as risk proxies . mutual exclusion and technical dependence. A. Scenario analysis may be particularly useful. Option Pricing Theory provides another tool for the calculation of the project NPV when an active management of the project may inﬂuence its value. scenarios and break-even analysis demand less data than Monte Carlo simulation and may provide useful information about the investments [4]. When options are involved. Option Pricing Theory has gained a wide acceptance in the evaluation of investment projects. Recently.7.15]. for example. see [6] and the chapter 3 of [7]. or in the hurdle rate (when rate of return methods are being used). 24 (2004) 1-20 c˜ whether or not risk analysis should be performed and because risk should be incorporated in the evaluation of the projects.

and to properly consider their present salvage values as opportunity costs (the costs of not selling the assets needed to continue the projects). it may be possible to present better accounting results and it is possible that stakeholders will be happier and. If accounting methods are used. as well as the capital requirements of existing projects. These present salvage values correspond to the investment cost in new projects. However. and that rate of return and ratio attributes should be used in particular situations instead of equivalent worth attributes. which is usually in the form of accounting statements. concerning whether or not accounting and payback methods should ever be used. Only by chance these methods will lead to aggregate NPV maximisation. when we are dealing with non-additive ﬁnancial attributes such a program is complex. the latter of which consists in pursuing both projects A and B. Godinho. After this. 24 (2004) 1-20 c˜ 17 Interdependencies can be dealt with mathematical programming.R. After this. usually easier to deal with. When synergies exist between two investment projects. When this happens. Combinatorial analysis can also be a very eﬀective tool to deal with interdependencies. some care should be taken in the selection process.C. and it may be diﬃcult to solve. in a similar way as the one referred when we considered that the available capital was limited and known. and we have explained how should each of those characteristics aﬀect the investment selection process. We will now address this issue. care should be taken not to consider unrecoverable costs or past beneﬁts. Costa / Investiga¸ao Operacional. However. it may be easier for the company to raise money and managers may . if a selection process consists in ordering all projects according to an attribute. So. the selection process may proceed normally.P. Thus far we have considered the decision situations deﬁned by a set of ﬁve characteristics. the process may proceed as if there were no previously undertaken projects. or change the ﬁnancial attributes of some of them (when there are synergies). A and B. when A is technically dependent on B. A.P. thus. When re-evaluating existing projects. it is very easy. If such a portfolio does exist. the latter of which consists in pursuing both projects A and B. When the available capital for the investments is limited. and then selecting projects according to that order. we can turn these two investment projects into three mutually exclusive projects: project A. it is often useful to change all interdependencies into mutual exclusion. the capital requirements of existing projects must be subtracted from the available amount. We argued that equivalent worth attributes should usually be used. Some questions may now be made. it may or may not be possible to abandon the previously undertaken projects. The existence of previously undertaken projects that can be abandoned may be dealt with by considering these projects at the same level of the newly proposed projects. because it will still be necessary to consider the interdependencies between existing projects and new projects. This means that they should only be used when the DMs’ goal is diﬀerent from that. project B and project AB. we will consider accounting methods. J. we can consider two mutually exclusive projects: B and AB. For example. then it is not necessary to re-evaluate these projects. when will DMs use accounting methods ? The only answer we have to this question has to do with the way the company’s results are made available to stakeholders. Interdependencies may eliminate some new projects (when there is mutual exclusion). The selection process may be aﬀected by the existence of a portfolio of previously undertaken investment projects. All interdependencies can be thus turned into mutual exclusion. after each project is selected. Also. to seek and exclude all the projects mutually exclusive with the selected project. If the previously undertaken projects cannot be abandoned. Afonso. Firstly.

Afonso. Then. when equivalent worth. Notice that such situations will be related to capital markets imperfections. Some reasons may be presented for the use of these methods. when it is diﬃcult to raise capital either for new investment projects or for ﬁnancial engagements of the company (the need to pay interests or principal in existing debts. because these risks can be accounted for. but not to maximise long-term accounting results. there are two severe drawbacks to this strategy. It can be argued that. the assets on which it invested. Costa / Investiga¸ao Operacional. One of these reasons may be the existence of catastrophic or political risks that may cause the company to lose. However. In this section we have considered that the decision situations are deﬁned by a set of characteristics. rate of return and ratio methods are used. When the owners of the company may need money at any time and cannot access an eﬃcient capital market either to raise money for their needs or to sell their stakes in the company. Another reason has to do with the need of quickly obtaining liquidity. in these situations. we tried to ﬁnd some situations in which the use of accounting or payback methods might be appropriate. We concluded that accounting methods might be used to achieve good short-term accounting results. Nevertheless we think that the use of mathematical programming to maximise the aggregate NPV. We do not think this is always valid. in this situation. We considered the use of payback methods to account for the company’s or its owners’ liquidity needs (particularly when capital markets are imperfect) and also when the company faces political or catastrophic risks. We argued that equivalent worth methods (eventually within a mathematical programming problem) are best suited for those situations.in the long term. We will now turn to payback methods. We concluded that equivalent worth attributes should usually be used. but we could not ﬁnd any reason to oppose the use of a payback attribute along . and we have explained how should each of those characteristics aﬀect the investment selection process. accounting results will have a higher growth when decisions are made to maximise the company value than when accounting methods are the only criteria to select investments. A. the use of mathematical programming to maximise the aggregate NPV according to some constraints that have to do with liquidity necessities will provide better results than the use of payback methods. at any moment.P. for instance). We think that. Once again. they may want the company to have high liquidity as often as possible.R.18 P. payback methods will provide the best results to satisfy the company owners wishes.C. 24 (2004) 1-20 c˜ get better rewards. The ﬁrst is that it is a short term strategy . in these situations. The second is that intelligent and well informed stakeholders will look beyond accounting results to get a clearer picture of the company. in this situation we cannot ﬁnd any reason to oppose the use of a payback attribute along with an equivalent worth attribute in a multicriteria evaluation. and this picture will be more favourable if the management is aiming to maximise the company value than if it is aiming to maximise short term accounting results. However. Godinho. and that rate of return and ratio attributes should be used in particular situations instead of equivalent worth attributes. either through a proper adjustment of the discount rate or through an adjustment of the predicted cash ﬂows. and so the company may want to recover the invested capital as soon as possible. J. according to a set of constraints that represent the possible needs of the company owners. We will present another reason that is also related with possible capital markets imperfections. if data are available. may provide better results. we cannot ﬁnd any reason to oppose the use of a payback attribute along with an equivalent worth attribute in a multicriteria evaluation.

When unlimited capital is available. We started with a presentation of the most common methods for project evaluation. it will be important to incorporate the risk in the project evaluation. following a similar classiﬁcation from [11. 24 (2004) 1-20 c˜ with an equivalent worth attribute in a multicriteria evaluation. we also tried to deﬁne in which situations . if a portfolio of previously undertaken projects exists. we deﬁned that the set of criteria should be a coherent family of criteria. according to our framework. capital availability for the investments. and which ones cannot. in those situations. and should not include contradictory assumptions. However. all the properties. and presented a mathematical result that states that two criteria that order projects identically will be redundant. it is possible to deﬁne a rate of return method that ranks projects identically. which both address the relation between the return and the investment. So. or characteristics. these projects should be re-evaluated. so a DM will usually use at most one attribute from a single class. In the presence of risk. rate of return and ratio methods may be a better choice. We considered a set of characteristics of the decision situation – degree of quantiﬁcation. with the exception of the rate of return and ratio classes. or when the available capital is limited and known.and analysed how these characteristics aﬀect the investment selection process. Then. 19 6 Conclusions This paper addressed the analysis and evaluation of investment projects within a multicriteria framework. We brieﬂy referred some methodologies for risk analysis and for the incorporation of risk in project evaluation. degree of risk and uncertainty. and their interdependencies with new projects and eﬀect in the available capital (if the available capital is limited) must be considered in the evaluation process. and that the converse is also usually true. We also stated that.12]. Costa / Investiga¸ao Operacional. We concluded that attributes from the same class either rank projects identically or are based in contradictory assumptions. Attributes form diﬀerent classes address diﬀerent perspectives.C. but he/she will usually want to use at most one attribute from both the classes of ratio and rate of return methods. we classiﬁed the methods into ﬁve classes. of the proﬁtability. Godinho. a DM may simultaneously use attributes from diﬀerent classes. J. The degree of quantiﬁcation deﬁnes whether or not ﬁnancial methods can be used. In order to have a correct structure for the decision problem. even if the ultimate goal is the maximisation of aggregate NPV. or when the available capital must also be used in future projects not yet known. we tried to establish a set of guidelines to help DMs choose the ﬁnancial methods best suited to their speciﬁc situations. of the investments are modelled as attributes. interdependencies between investments and existence of previously undertaken investments . Finally we explained that. when the available capital is limited and the limit is unknown. In this presentation. Since the analysis of the characteristics of the decision situation seemed to never recommend the use of accounting and payback methods. The decision criteria are chosen from the attribute set. Then we mathematically deﬁned the conditions that must be met by the set of criteria in order to be a coherent family of criteria.P. and to perform risk analysis. equivalent worth methods should usually be used. A. Afonso. be used together as criteria. or dimensions. The ﬁrst problem we dealt with was to ﬁnd out which ﬁnancial attributes can. We explained that interdependencies could be dealt with through the use of mathematical programming and combinatorial analysis.P. In this framework.R. for each ratio method.

R&D Management... Instituto Superior T´cnico. We concluded that accounting methods might be used to achieve good short-term accounting results. Payback and Accounting Methods. but not to maximise long-term accounting results. Portugal (1992). B. D. [9] McDaniel. and Bethel.S.. 4 (1995) 101-129. [15] Trigeorgis. [6] Godinho.E. P. and Nieto.. and Jessel. [10] Park. 23 (1988) 369-385.. Advanced Engineering Economics. Construction et Exploration d’un Mod`le Multicrit`re d’Aide a e e ` la D´cision (in French). and Costa. in those situations.. J.C. Multicriteria Methodology for Decision Aiding.S.C. Part 1: Net Present Value and Rate of Return Methods.S.E. J. Kluwer Academic Press. and Mason. 4 (2002) 147-158. The Financial Review.C. M. . Costa / Investiga¸ao Operacional. K.. A Note on the Use of Bicriteria Decision Trees in Capital Budgeting.. Midland Corporate Finance Journal. P. Risk Analysis and its Applications. Godinho. A Compendium and Comparison of 25 Project Evaluation Techniques. The MIT Press (1996). W. Afonso.P. G. McCarty. Dordrecht (1996). 40 (1995) 287-294. Discounted Cash Flow with Explicit Reinvestment Rates: Tutorial and Extension. Global Business and Economics Review. L. Management Science. Portuguese Journal ` a of Management Studies. [14] Trigeorgis. R. John Wiley and Sons (1990). 7 References [1] Afonso.P. 1st Edition.P. and Spatig. A. 20 (1990) 155-171. Principles of Corporate Finance. PhD Thesis. we could not ﬁnd any reason to oppose the use of a payback attribute along with an equivalent worth attribute in a multicriteria evaluation. Real Options: Managerial Flexibility and Strategy in Resource Allocation. Estrutura¸ao dos Crit´rios de Avalia¸ao e Selec¸ao c˜ e c˜ c˜ ˆ de Projectos no Ambito de um Sistema de Apoio a Decis˜o (in Portuguese). D. The Engineering Economist. C.20 P. and Thomas. we always concluded that equivalent worth methods (eventually within a mathematical programming problem) might be best suited to those situations. Spring (1987) 14-21. J. J.P. F.R.B..R. 9 (1963) 443-457. [3] Bana e Costa. D.P. [8] Hillier. and Nieto. A Compendium and Comparison of 25 Project Evaluation Techniques. McGraw-Hill (2000).P. [2] Asquith.R. and Myers.. 1st Edition... and Costa.S. P. The Derivation of Probabilistic Information for the Evaluation of Risky Projects. [7] Hertz. A. 6th edition. [13] Roy. and Sharp-Bette. D. Lisbon. [5] Fahrni. A.A. John Wiley and Sons (1983). International Journal of Production Economics. [12] Remer. Valuing Managerial Flexibility. D. L. Godinho. However. Using Heuristics to Evaluate Projects: the Case of Ranking Projects by IRR. S. C. S... IV (1999) 267-289... e e [4] Brealey. H. Analysing these situations. Structuration. Part 2: Ratio.P. An Application-Oriented Guide to R&D Project Selection and Evaluation Methods. 42 (1995) 79-96. 24 (2004) 1-20 c˜ might these methods be used. [11] Remer. We considered some situations in which the use of payback methods seemed appropriate. International Journal of Production Economics. A.