Cost-Benefit Analysis

Other books by D. W. Pearce published by Macmillan

Cost-Benefit Analysis: Theory and Practice

(with A. K. Dasgupta)
Capital Investment Appraisal (with C. J. Hawkins)
The Dictionary of Modern Economics (editor)
The Economics of Natural Resource Depletion (editor)
The Social Appraisal of Projects (with C. A. Nash)
Cost-Benefit Analysis
Second Edition

D. W.Pearce
Department of Political Economy,
University College, London

M
© D. W. Pearce 1971, 1983

All rights reserved. No part of this publication may be reproduced or transmitted,

in any form or by any means, without permission.

First edition 1971

Reprinted 1973,1977,1978,1981 (twice)
Second edition 1983

Published by
THE MACMILLAN PRESS LTD
London and Basingstoke
Companies and representatives throughout the world
ISBN 978-0-333-35281-6 ISBN 978-1-349-17196-5 (eBook)
OOI 10.1007/978-1-349-17196-5

Typeset in Great Britain by

STYLESET LIMITED
Salisbury, Wiltshire
For Corin and Daniel
Contents

Preface and Acknowledgements ix

1 The Foundations of Cost-Benefit Analysis 1

Value judgements and CBA 4
Money, preferences and 'non-markets' 9

2 The Origins of Cost-Benefit Analysis 14

Appendix: compensation tests and CBA 22

3 The Measurement of Costs and Benefits 25

Benefits and consumer surplus 25
Problems with consumer surplus 27
Costs and forgone benefits 30
Shadow pricing 31
Appendix: consumer surplus when other prices
change 34

4 Time, Discounting and Decision Rules 37

The rationale for discounting 37
Finding the discount rate 40
The inequality of the two rates of discount 43
Adjusting the decision rule 46

vii
 Contents

Ranking projects 51
Discounting and future generations 52
Appendix: deriving a social time preference rate 55

5 Efficiency and Distributive Weights 59

Conventional CBA vs 'revisionism' 60
Deriving distributional weights 64
Problems with weighting procedures 66
Other weighting approaches 68
Conclusions 70
Appendix 1: deriving an income-elasticity weighted
cost-benefit function 71
Appendix 2: deriving a marginal utility of income
weighting procedure for CBA 71

6 Risk and Uncertainty 73

Risk 74
The Arrow-Lind theorem 81
Uncertainty 83
Other approaches to risk and uncertainty 88

7 A Case Study: The Gordon-below-Franklin Dam 90

The project 91
The benefits 91
The costs 97
Irreversibility 98
Estimating required preservation benefits for the
Gordon Dam 102
Epilogue 105

References 106

Index 110

viii
Preface and
Acknowledgements

In 1971 I published a short text on cost-benefit analysis

which, warts and all, has enjoyed some considerable com-
mercial success. It was followed a year later by Ajit Dasgupta's
and my Cost-Benefit Analysis: Theory and Practice (Mac-
millan, 1972), which remains in print. In the late 1970s it
was more than obvious that the small text, which was intended
largely as a student revision book, needed substantial updating
and correction. That project, like Topsy, growed and the end-
result was my text with Chris Nash, The Social Appraisal of
Projects: A Text in Cost-Benefit Analysis (Macmillan,
1981). Equally obvious, however, was the fact that we had
gone well beyond the publisher's instruction to write a suc-
cessor to the small cost-benefit book. Since sales of the
1971 book continued to be reasonably buoyant, it was evident
that there was still a market demand for a brief 'guidebook'
to CBA. This 'new edition' of the 1971 text is that replace-
ment, but it has been completely rewritten. What remains,
however, is the central message of the 1971 text: that there
is no unique way of carrying out cost-benefit studies, nor
should there be, and on that I remain totally unrepentant,
despite the strictures of my colleagues such as Professor Ed
Mishan who have so eloquently argued the opposite but not,
in my view, convincingly. Other features remain, I hope,
unusual enough to attract the same audience that approved
of the 1971 edition.

ix
 Preface and Acknowledgements

My debts to others are, literally, too numerous to mention.

Many of my working partners will see sections that owe
much to their influence. Other colleagues may note passages
in which I have especially gone out of my way to explain
and defend views which I know they disapproved of, and
may still do so. However, while not implicated in the errors
that may remain, I must record enormous gratitude to Chris
Nash for many years of 'internal' debate on cost-benefit,
and to my good friend Jean-Philippe Barde of the OECD,
Paris, for insisting always that technique without relevance
explains rather too much of the disrepute that academic
economists bring on themselves. The text was written during
my last year at Aberdeen University. But for Winnie Sinclair,
an embodied technological revolution in herself, this and so
much else would simply never have appeared. I am forever
grateful.

Aberdeen and London D. W. PEARCE

March 1983

X
 1

The Foundations· of
Cost-Benefit Analysis

Cost-benefit analysis (CBA) excites opmton among econ-

omists and non-economists. Referring to the attempt by CBA
to express all benefits and costs in monetary terms, even
where we have no market in the benefit or cost in question, ·
and, indeed, especially where we have no such markets, Self
(1970) has remarked:

Cost-benefit analysis gets its plausibility from the use of a common

monetary standard, but the common value of the £ derives from
exchange situations. Outside such situations, common values cannot
be presumed, and symbol and reality become easily confused ... To
call these judgements £sis to engage in a confidence trick - to exploit
the ordinary man's respect for the yardstick of money in what are
actually non-monetary situations. (Self, 1970, p. 8)

The same central feature of CBA worried Schumacher ( 1973):

To press non-economic values into the framework of the economic

calculus, economists use the method of cost/benefit analysis. This is

1
 The Foundations of Cost -Benefit Analysis

generally thought to be an enlightened and progressive development,

as it is at least an attempt to take account of costs and benefits
which might otherwise be disregarded altogether. In fact, however, it
is a procedure by which the higher is reduced to the level of the lower
and the priceless is given a price. It can therefore never serve to
clarify the situation and lead to an enlightened decision. All it can
do is lead to self-deception or the deception of others; for to under-
take to measure the immeasurable is absurd and constitutes but an
elaborate method of moving from preconceived notions to forgone
conclusions ... what is worse, and destructive of civilisation, is the
pretence that everything has a price or, in other words, that money
is the highest of all values. (Schumacher, 1973, pp. 41-2)

It would be unfair to the critics to suggest that the sole focus

of all their misgivings is the attempt to put money values on
non-marketed things. There are many other stated objections,
ranging from discrimination against future generations, an
overly narrow definition of what any policy decision should
be about, the alleged neglect of income distribution, the
potential for 'rule by experts' given the complexity of any
rigorously executed CBA, and so on. Many of these objections
will be analysed in the course of this book. The remarks by
Self and Schumacher, however, serve to indicate that some
care needs to be taken in understanding just what the basis of
CBA is.
We may begin with a definition. We define a rational choice
as one in which an individual chooses an option when the
gains from the action in question exceed the losses. For gains
and losses we can use the terms 'advantages' and 'disadvan-
tages', 'pros' and 'cons', or 'benefits' and 'costs'. Further, we
shall leave it to the individual to define what he or she means
by gains and losses. In particular, they need not be gains and
losses to the individual in question. They could already
embody some degree of 'altruism'- concern for others.
Next, we shall produce a second definition: 'society' is noth-
ing more than the collection of individuals who make it up.
There is to be no concern with entities such as the 'state', in
2
 The Foundations of Cost-Benefit Analysis

the sense that shall not regard the 'state' or 'society' as being
something in addition to the sum of people who comprise it.
CBA is a procedure for:
1. measuring the gains and losses to individuals, using money
as the measuring rod of those gains and losses
2. aggregating the money valuations of the gains and losses of
individuals and expressing them as a net social gains or
losses.
Given the definitions of 'rationality' and 'society', we can
therefore say that a rational social decision is one in which
the benefits to society (i.e. the sum of the people in society)
exceed the costs. Note that use of the term 'rational' seems a
little emotive. Few of us would like to think we are not
rational in our choices. But rationality and morality are not
at all the same thing. Judging that action X will give me more
benefits than costs and choosing X as a result is not the same
thing as saying that X is a 'morally correct' action. By depriv-
ing, say, a major charity of money I could otherwise have
given it, my choice may seem distinctly morally unacceptable
to others. In the same way, the summation of a whole set of
choices by many individuals may give a result which the 'state'
or government thinks is not right. As a procedure for aggregat-
ing the preferences of our set of individuals, we can establish
something of fundamental importance at the outset: CBA
makes no claim to produce morally correct decisions.
What CBA produces, and what is morally correct, may
coincide if, and only if, we adopt a further rule, namely that
some aggregated set of preferences of individuals is the morally
correct way of making decisions. In some circumstances the
two may well coincide. In others, government will often
reserve the right to 'overrule' group preferences. In still others,
and these are surely the majority, governments will at least
wish to know what the preferences of the individuals who
make up society are. It is in this sense that CBA is an 'input',
an 'aid', an 'ingredient' of decision-making. It does not
supplant political judgement.
3
 The Foundations of Cost-Benefit Analysis

Now, there can be no doubt that there are unscrupulous

economists, along with unscrupulous politicians, philosophers
and engineers. If, therefore, CBA has been exaggerated in
terms of its role in decision-making, it may be that there are
those who have not practised its tenets properly. Equally, we
must investigate to see why it should be so easy to produce
the kind of result that has clearly irritated and offended the
likes of Self and Schumacher. Their worries arose, in particu-
lar, because of (a) the attempt to apply the market-place
philosophy to non-market situations, and (b) their doubts
about the values expressed in the market-place as a guide to
anything that can be described as morally correct. These are
familiar objections to CBA, and the two separate strands of
concern are frequently confused. lf(b) is correct, for example,
and we cannot ascribe 'morality' to market-place evaluations,
then the valuations obtained by using the same procedures in
contexts where there are no (obvious) markets must also be
immoral. But objection (a) could still be made while (b) is
regarded as morally acceptable. In this case, we are more
likely to be arguing that things not traded in markets are, in
some sense, 'special' and hence not open to valuation in
money terms. Or it could be that the objection is a practical
one as to whether the specific techniques used really are cap-
turing the 'full' value in the non-market situation.

Value judgements and CBA

We have argued that CBA is a technique, as yet undefined,

for aggregating the preferences of individuals. CBA makes no
claim to be morally binding, for the simple reason that what
is moral need not coincide with what people want. That should
be sufficient to establish the role of CBA. It does not make
political judgement redundant because there is no necessary
relationship between those judgements and the wants of
individuals. (We shall not investigate the issue of how political
judgements relate to moral judgements!) Notice that this

4
 Value judgements and CBA

already removes some part of the objections raised by Self

and Schumacher, for it is odd to speak of deceiving the people
whose valuations make up the result obtained by the CBA. If
the objection is that CBA does not in practice reflect individ-
uals' valuations, that is a quite different objection and one
that raises a much larger area of concern about how we 'test'
for the accuracy of any results. We return to that issue later.
In what way, then, does CBA seek to aggregate individual
preferences? It does this by taking the market-place as the
prime context in which those preferences are expressed. The
medium through which they are expressed is money. It is
important to realise the reason for this. It has nothing to do
with being obsessed with money, and everything to do with
the fact that markets are the only contexts in which individ-
uals express millions of preferences daily. The political system
does not begin to compare. We would have to have endless
referendums and elections to get remotely near the complex-
ity of the market-place, whether it be the local fish market,
the Stock Exchange or something as complex as the foreign
exchange market comprising the world's financial institutions
and a very large number of telephones, telex machines and
computer display units.
Within these markets countless individuals express their
preferences for or against goods and services. They vote for
them by buying them and against them by not buying them.
The means that they use to express their votes is, of course,
money. Those votes could be expressed in terms of any
measuring-rod. It so happens that money has evolved as a
convenient measuring-rod. Had it been cowrie shells or camel
bells they would still have been 'money', which is simply a
word for the medium of exchange. In this respect there can
be no objection to a technique which seeks to elicit prefer-
ences expressed in terms of money. If that remains an objec-
tion, we must surely conclude that the critic has not under-
stood at all the evolution of economies. But that cannot be
what worried Self and Schumacher. We get a little closer
when we consider that money is a medium of exchange and a
5
 The Foundations of Cost-Benefit Analysis

store of value in terms of income and wealth. What this means

is that the preferences expressed in the market-place are con-
ditional on the possession of money. That is, those preferences
will be weighted by the market power conferred on individuals
through their possession of money. This is part of what
Schumacher means when he refers to the 'pretence' that
'money is the highest of all values'. As we shall see, there are
ways of adjusting our CBA for this kind of concern, but for
the moment we should pause to see what our analysis has
revealed.
First, it is evident that CBA involves the aggregation of
individuals' assessments of the costs and benefits to them of
a given action, policy, project or programme. This means that
we have implicitly accepted that CBA results will, if properly
derived, reflect individuals' preferences. If CBA is then an
input to the procedure of deciding what a decision-maker
ought to do, then CBA is itself 'normative' and rests on at
least one value judgement or normative statement, namely
that it is a good thing that individuals' preferences should
count. Note that a normative judgement or statement recom-
mends and implies that what is recommended is 'good'. While
we can ask 'why' we should do something, it eventually
becomes rather redundant to ask why something is good. Our
views as to what is and what is not good will diverge.
Second, by looking at the role that money actually pays in
the measure of preferences, we observed that market-places
operate on the basis that those with more money have more
say than those without. Note the contrast here with a political
vote which, in an ideal world, is unrelated to income or
wealth. If we are to leave the aggregated preferences in the
market-place unadjusted, it follows that they will reflect the
structure of market power, or, to put it another way, the
distribution of income. If we are to afford CBA its role in the
decision-making procedure, then, we must add a second value
judgement, namely that the distribution of income used to
weight the preferences of individuals is in some sense the best
one. In short, the existing distribution is good. (Indeed, we
have to go a little further and say it is the best.)

6
 Value judgements and CBA

Now, the two normative judgements that emerge require

restating. They are:
1. individual preferences should count
2. those preferences should be weighted by the existing dis-
tribution of income.
That individual preferences should count implies that social
decision rules reflecting individual preferences are 'good' rules.
This is appealing in that it obviously defines the basis of what
we might call 'simple democracy'. Judgement (1) is thus the
basic requirement of democratic sensitivity, or, as it is best
known in economics, consumer sovereignty. What is morally
appealing about (2)? Perhaps the existing distribution of
income reflects the distribution of effort in the economy and
we might invoke a principle that people deserve a 'proper'
reward for their effort. Against this we might point out that
the existing distribution of income already contains that
reward; what we are arguing about is whether CBA should
also reflect that distribution. After all, if CBA is used to guide
decisions, this is tantamount to saying that those who have
already been rewarded will be rewarded again. Obviously,
the debate over (2) could go on, but one thing is clear. We
do not have to accept (2) in its particular form as stated here.
Equally, we cannot refuse to adopt some judgement such as
(2). For example, if we reject (2) and say that all the prefer-
ences recorded will somehow be 'equalised' to net out the
influence of market power, then we have rejected (2) in its
stated form but have replaced it with a variant of (2). The
only rule we seem to have for selecting one variant of (2)
rather than another is its 'moral appeal'. CBA can be con-
structed in different ways according to whether we combine
(1) with (2), or (1) with some variant of (2). But of course
we could reject (1) on the grounds that individuals are poor
judges of their gains and losses. This 'paternal' argument
would then mean that we would have to substitute another
set of preferences for (1), perhaps the preferences of some
set of experts or those who have the final responsibility for
decision-making. Now, if we accept this line of thought, what

7
 The Foundations of Cost-Benefit Analysis

happens to judgement (2)? We now have to replace that with

some judgement to the effect that each expert's valuation is
equally important or that there is some weighting of judge-
ments (according to seniority, peer group review?).
We conclude that CBA requires TWO normative (value)
judgements. The first states that preferences count, but
requires careful qualification about WHOSE PREFERENCES
are to count. The second must say how the preferences are to
be weighted.
We are now in a position to make some brief, but important,
observations.
First, there are two, and only two, forms of judgement
required. This may be contrasted with the proliferation of
value judgements suggested by some authors (e.g. Peacock,
1973).
Second, if we decide to adjust either value judgement we
have not engaged in any underhand or illicit activity. There
are no rules for choosing between ultimate value judgements.
Choice is determined by 'moral appeal'. Appeals to the exist-
ence of 'virtual constitutions' (Mishan, 1974) cannot be more
than this.
Third, CBA is a normative procedure. Not only is this not
a limitation, it actually reflects the nature of economics itself
and, some would say, all science (Katouzian, 1980). A fre-
quent charge against CBA is that it is 'subjective'. This is a
confusion. As value judgement ( 1) indicates, it is the subjective
preferences of individuals that we are seeking. In that trivial
sense it is indeed 'subjective'. But if the criticism is meant to
be that the analyst himself can influence the outcome in
some arbitrary way, then we have to point out that there is
always the scope for falsification in any analytical technique,
but that scope is not part of the conceptual structure of CBA.
Fourth, and following on from the earlier points, what
value judgements are chosen must be made clear. If they are
hidden in the analysis, then the charge of 'subjectivity' in
practice will have substance. We can go further. Not only
should the judgements be made clear, but the outcome of the

8
 Money, preferences and 'non-markets'

CBA should be recalculated to show the effects of changes in

the value judgements. We should practise 'value judgement
sensitivity analysis' (Nash, Pearce and Stanley, 19 7 5).

Money, preferences and 'non-markets'

So far, then, we have established, in outline form, the philo-

sophical basis of CBA. In so doing we would argue that we
have revealed some sources of misunderstanding in the
quotations given at the outset. And in so far as the objection
to the use of the measuring-rod of money is based on money
as income or wealth, we shall see later how it is possible to
adjust for that objection. But the quotations also indicate a
concern about extending the measurement of preferences to
so-called 'non-market situations'. We need to dwell on this
issue briefly.
The preferences expressed in markets are revealed as offers
of money in exchange for some benefit received. The bid
made by the buyer of the benefit shows up as a willingness to
pay for the benefit, conditional on his ability to pay (income
or wealth). But what is actually paid could well be less than
this willingness to pay. It cannot be more because then the
individual will simply record a preference against the good or
service: he will not buy it. But given that the actual price
paid is determined by the interaction of many buyers and a
few or many sellers, there will be individuals whose willing-
ness to pay exceeds the price they actually pay. As we shall
see, this excess of willingness to pay (WTP) over price is
consumer surplus. Since the sum of money actually paid
involves a loss for the individual, then that loss is the 'cost' of
the purchase. We have a basic element of CBA in this simple
example, for we can write, for the individual:
WTP =Price paid
Individual's net benefit = Consumer surplus

9
 The Foundations of Cost-Benefit Analysis

We shall return to this formulation when we look at the

measurement of benefits.
Now apply this procedure to a situation in which there is
no obvious market in the good or service. To fmd the net
benefit to any individual of a non-marketed benefit such as
peace and quiet, clean air, visual amenity, the preservation of
wildlife, and so on, we need to find WTP and price, but since
there is no market there is no set of actual transactions to
provide us with price. The price, in fact, is zero. Indeed, this
is just how we refer to these types of benefits. They are
'unpriced' or 'zero-priced'. But note that this is entirely
different from saying that WTP is zero. Indeed, it needs only
a moment's reflection to indicate that all the benefits we
have listed have a positive WTP. That is, if there were a market,
individuals would be willing to pay for the benefits obtained.
In fact, the reference often made to 'unmarketed' goods is
slightly misleading. There are no markets in the goods or
services as such, but there are often markets in other goods or
services which are influenced by the valuations placed on the
benefits of the unmarketed goods. These are surrogate mar-
kets. Two examples will suffice. First, while we have no
obvious market in peace and quiet, we do have a housing
market. If people buy and sell houses and are influenced by
the extent to which a specific property is in a quiet location
or not, then we should be able to observe the workings of the
housing market and see if it 'reveals' the WTP for peace and
quiet. As a second example, consider the highly emotive issue
of the 'value of human life'. Resistant though many of us
might be to the idea of valuing life, it is clear that individuals
often do accept sums of money in return for an increase in
the risk of death. One example might be premia added to
wages for working in a dangerous occupation. Notice that it
is not 'life' itself that is being valued by these premia, but the
risk associated with the extra danger. Note also that this is
not an example of WTP but rather 'willingness to accept' for
the increased risk. Since the increased risk is the opposite of a
benefit - i.e. a cost - this measure should strike us as accept-
10
 Money, preferences and 'non-markets'

able and consistent with the procedure so far developed. So, .

our first response to those who argue that there are 'no
valuations' in certain kinds of goods and services must be to
say that the absence of a direct market is not at all the same
thing as the absence of an indirect market, and that the
absence of a direct market does not mean that preferences
and valuations are not made. If the logic of using market-place
valuations is accepted, it is difficult to see how it can be
rejected for surrogate markets.
This leaves us with goods and services whose values are not
revealed in any market, obvious or surrogate. The value of a
blue whale would seem to be a case in point. There is a com-
mercial market in whales, but we would rightly reject the
market price of the whale as oil and blubber as reflecting the
aggregate of individuals' WfPs for the blue whale, for we
know that there are many hundreds of thousands, perhaps
millions, of people who value the whale in its natural state,
even though they may only have seen one on their televisions,
or in photographs. In such contexts we have two options. We
can leave the issue 'unvalued' and say that we have no con-
ceptual technique for working out the WTP for blue whales,
or we can invent a market. To do this we can hypothesise a
situation in which individuals vote for preservation of the
blue whale and ask them what they would be willing to pay
if there was a market in whales which was as open to them as
it was to whalers. We can all envisage the practical difficulties,
but for the moment we are concerned to know if the absence
of an actual market reflecting all our valuations imposes any
conceptual problem on our logic of finding WTP. It would
seem not, and there are indeed a fair number of experiments
in establishing such hypothetical or 'experimental' markets.
How, then, does all this relate to the view that goods and
services which do not have direct markets are somehow 'dif-
ferent' from other goods and services? For these goods to be
so special that we cannot apply even the conceptual logic of
CBA to them requires us to establish that any of the above
means of calculating WTP would seriously mistate the true
11
 The Foundations of Cost-Benefit Analysis

benefits. But that presents us with a pervasive methodological

problem. For we shall not be able to tell whether we have the
'right' valuation unless we have some other measuring-rod
against which to compare our measure of WTP. But if we had
some other measuring-rod, we would surely not need the
WTP approach in the first place. While we shall see that it is
possible to compare the results obtained from surrogate
market analyses with experimental (hypothetical) markets in
certain cases, in most of them we shall have to rely on some
form of judgement as to the extent to which any technique
used captures all the preferences that should be included.
Thus part of what Schumacher had in mind in his own remark
quoted earlier is that economic techniques such as CBA will
fail to identify certain values. Universally, CBA speaks of
values as attributes of things as perceived by humankind. We
make no attempt to include the preferences of blue whales in
our analyses. CBA is 'anthropocentic'. If that is the source of
criticism, then it should be acknowledged, and the point
made earlier, that CBA cannot be presented as something
which is synonymous with moral correctness, must be invoked.
If the criticism is that we have neglected some value of the
good or service to humankind, then it is a charge of omission
and that is important within the rules that CBA has established
for itself. What Schumacher was worrying about was CBA's
neglect of the environment as a system for supporting human
life, and since the environment is invariably the good which
has, at best, surrogate markets and, more usually, no markets,
there is a good reason to agree with Schumacher that there is
a deficiency in the way CBA operates.
As a final comment in this chapter, we may observe from
the way the discussion has progressed that a 'benefit' is any
gain to any individual included in the group in question. That
gain may accrue in some money form, or it may accrue as
some sense of pleasure or happiness. In the language of econ-
omics, we require only that it be some welfare or utility gain.
Similarly, a 'cost' is not defined in terms of a money flow. It
is, in fact, anything that imparts a loss of utility or welfare.
12
 Money, preferences and 'non-markets'

Notice that this will subsume, say, the money value of the
resources used to build a motorway or whatever, because
those resources could have been used to give welfare gains
elsewhere or from some other project. In this sense a cost is
always a forgone benefit. Indeed, it should now be evident
that cost in the context of CBA always means opportunity
cost. One problem which is not readily resolvable in con-
ceptual terms is quite who counts when calculating costs and
benefits. One is tempted to say that a nation's boundary sets
the limits on who should count if we are building a road or
an airport and using up the nation's resources. But that may
not be sound reasoning if the roads and airport are used by
persons from other nations. The use of national boundaries
appears even less defensible if we are looking at the costs and
benefits of a policy to control sulphur dioxide emissions into
the atmosphere and we discover that the cheapest way (in
terms of resources) is to build high chimneys and allow the
sulphur dioxide to travel across to other countries. What may
yield net benefits to the 'emitting' nation may do so only
because that nation has 'exported' the pollution to another
country. And if we reconsider our example of the blue whale,
it would seem necessary to set no national boundaries at all
regardless of where the blue whale is most often seen. There
are in fact no clear rules on setting the 'boundaries' for a
CBA. Most often it will be obvious, but on other occasions it
will not be.

13
 2

The Origins of
Cost-Benefit Analysis

While the underlying theory of CBA can be traced back to

some welfare economics of the nineteenth century, the
practice of CBA can be said to date from the introduction
of the Flood Control Act 1936 in the USA. It seems fair to
say that that Act owed little or nothing to the body of welfare
economics theory that had emerged by the time of the Act's
introduction; indeed, as we shall see, the placing of CBA in a
firm conceptual framework occurred after the Act and par-
ticularly in light of certain theoretical developments at the
end of the 1930s. The Flood Control Act determined that
the control of flood waters was 'in the interests of the general
welfare'. And it went on to state a most general rule to the
effect that the Federal Government 'should improve or
participate in the improvement of navigable waters ... for
flood control purposes if the benefits to whomsoever they
accrue are in excess of the estimated costs' (our italics). Too
much must not be read into the Flood Control Act. First, the
reference to the unrestricted nature of the benefits that
14
 The Origins of Cost-Benefit Analysis

should be taken into account was not matched by a similar

coverage for costs. It is clear that the Act was referring to the
costs of construction and did not embrace the wider idea of a
cost as any loss of welfare. Second, the Act itself makes no
further mention of the 'benefits minus costs' principle and
commentators seem generally agreed that many of the projects
subsequently sanctioned after the Act would not have passed
the benefit-cost rule as it was developed. Third, the meaning
of the term 'benefits' was also not made clear.
The next landmark was the so-called 'Green Book' of 19 50
produced by the US Federal Inter-Agency River Basin Com-
mittee and which attempted to instill some agreed set of rules
for comparing costs and benefits. A further attempt at formal-
isation came with the US Bureau of Budget's Budget Circular
A-47 in 1952. These were early attempts, and they were
followed by the general introduction of economic techniques
into budget management in the USA across many areas of
expenditure. This process was aided by the emerging literature
from the Rand Corporation, which had devoted considerable
time to the development of rules of resource allocation in
military spending. Here the benefits were expressed in terms
of 'national security' or, in more macabre terms, destructive
capability. But the important development was in the use of
procedures for minimising the money cost of a given level of
activity - the beginnings of 'cost-effectiveness analysis'
(CEA), by which the benefit is measured in some physical
units, or is simply stated as a policy objective, and the costs
are expressed in monetary units. At the very least, then, alter-
native means of achieving the same end could be ranked in
terms of the ratio of cost to effectivensss (C/E), and the same
ratio could be used for projects with differing benefits as a
guide to judgement. Thus both CBA and CEA began their
practical lives as aids to government decision-making.
We observed that the Flood Control Act actually preceded
the developments in welfare economics that eventually came
to give CBA its economic foundations. During the 19 50s in
the USA there was a gradual and somewhat unintegrated
15
 The Origins of Cost-Benefit Analysis

convergence of theory and practice. The year 19 58 marked

the publication of three highly important works, by Eckstein
(1958), Krutilla and Eckstein (1958), and McKean (1958),
all ostensibly to do with water resource development in the
USA but each containing important links to the theoretical
literature of welfare economics. The basic linkages were (a)
the construal of a benefit as any gain in welfare (utility)and
a cost as any loss in welfare, (b) the concept of cost as
opportunity cost, so that, strictly, the benefit-cost rule
became one of (ideally) maximising the difference between
measured benefits and the forgone benefits from the project
'displaced' by the chosen project, and (c) the rooting of the
idea of maximising net benefits in the Pareto improvement
rule. The Pareto rule stated a virtual tautology to the effect
that 'society' was better off if at least some of its members
were made better off and no one was made worse off. Note
that the concept of society as the sum of the individuals in it
is explicit in this rule. But since no project is likely ever to
meet this rule, reference was made to the modification of it
introduced by Kaldor ( 1939) and Hicks ( 1939). The so-called
Kaldor-Hicks rule stated that any project should be sanc-
tioned if it improved the welfare of some people, even though
others might lose, provided those who gained could compen-
sate those who lost and still have some benefit left over.
Expressed in this way, the rule, or 'compensation test', is
simple. Imagine that we can put money values on the benefits
and costs, and that benefits accrue to five people and sum to
100 units. Costs accrue to twenty people and sum to 90 units.
Then, in theory, the five beneficiaries can transfer 90 units to
the losers, and they will still have 10 units of net benefit. The
losers will be fully compensated for their initial loss so that
they are no worse off. The compensation principle thus
establishes a potentially attractive way of making the Pareto
rule operational. Notice that it is not affected by the numbers
of people in the gaining or losing groups, nor is anything said
about who the groups are (rich, poor, etc.).
Reference was made in this new literature to one of the

16
 The Origins of Cost-Benefit Analysis

salient criticisms of the Kaldor-Hicks rule, given by Scitovsky

( 1941 ), to the effect that a project sanctioned by the rule
could give rise to a subsequent situation in which those who
had lost could now 'compensate' those who had gained to
move back to the pre-project state. That this could happen
arises from the fact that the project may change the distribu-
tion of income and hence the pattern of relative prices. At
the initial set of prices the project is judged worth while, but
at the new set of prices emerging after the project is under-
taken we can hypothesise a project involving a move back to
the initial position and this project may be sanctioned by the
very same test used to justify the move away from that initial
position. A more formal statement of the compensation test
and Scitovsky's criticism is given in the appendix to this
chapter. While Scitovsky's criticism was acknowledged, it was
largely ignored in what now seemed to be a relentless advance
by CBA and CEA into the appraisal of government expendi-
tures. The context of the techniques- government spending-
was natural because government, as opposed to the private
sector, is supposedly concerned with the wider general welfare
of the population. There is in fact nothing in CBA which
forbids its application to private-sector investments. Indeed,
such procedures are useful indicators of the extent to which
private and socially efficient decisions diverge. But, by and
large, CBA remained, and has remained, in the domain of the
public sector.
The other major American work to emerge in the early
1960s was that produced by the Harvard Water Resource
Program (Maass, 1962), which forged closer linkages with the
underlying welfare theory. Thereafter, countless articles and
books appeared, and CBA also arrived in the United Kingdom
with the application of the technique to the London-Birming-
ham motorway, the Ml (Beesley et al., 1960). In 1967 a UK
Government White Paper gave formal recognition to the exist-
ence of cost-benefit analysis and assigned it a limited role
for nationalised industries (UK Government, 1967). The
essence of the directive was that nationalised industries should

17
 The Origins of Cost-Benefit Analysis

operate on a commercial basis, and should seek an 8 per cent

rate of return on new investment. Returns were, however, to
be measured in terms of financial revenues with the exception
of situations in which 'there are grounds for thinking that the
social costs or benefits do diverge markedly from those
associated with the alternatives' (i.e. alternative investments).
In these cases the government would carry out some type of
cost-benefit analysis.
In the late 1960s CBA was extended to less developed
countries with the publication of a Manual of Industrial
Project Analysis (Little and Mirrlees, 1969). The Manual was
prepared for the Organisation for Economic Co-operation
and Development (OECD) and was updated and revised in
1974 (Little and Mirrlees, 1974), while in 1972 the United
Nations Industrial Development Organisation (UNIDO)
published its own guidelines (Marglin et al., 1972), different
in detail but essentially the same in philosophy. In 1975
came the World Bank's guidelines, which were heavily reliant
on the earlier work of Little and Mirrlees (Squire and van der
Tak, 1975). CBA also gained additional impetus with the
environmental revolution. Since costs and benefits were
defined to include all welfare gains and losses, the complex
task of applying CBA in contexts where measurement prob-
lems were severe attracted many researchers. Unquestionably,
for the comparatively small sums spent on this dimension of
CBA research, the returns were very high. But, arguably, CBA
did little to influence environmental policy as such. As an
example the US Clean Air Act 1970 and the Amendments to
it in 1977 made no mention of costs and benefits. But the
main advance was in establishing a way of thinking about
environmental problems from a rigorous standpoint.
Ironically, it was the advocates of the environmental move-
ment that contributed to the poor image of CBA in the 1970s.
The idea that the environment was to be subject to procedures
for monetary evaluation was, and is, anathema to many who
regard the preservation of environments as some kind of
categorical imperative, not subject to the rules of allocative

18
 The Origins of Cost-Benefit Analysis

efficiency. Wilderness areas, for example, were held to 'beyond

price', a phrase which alternatively seemed to mean that they
possessed infinite values (an absurdity in a world of finite
resources), or that, somehow, they belonged to a category of
things which could not be traded with other things that are
legitimately valued in terms of the measuring-rod of money.
In the United Kingdom, outrage was most marked over the
advice given by a statutory commission (the Roskill Commis-
sion) on the proposed location of London's Third Airport
(Commission on the Third London Airport, 1971 ). Apart
from the fact that the recommendation, based on a very
detailed cost-benefit analysis, was for an inland site in a
context where, politically, a coastal site was preferred, the
motivating force for the condemnation of CBA lay in the
marked difference between the measured costs of noise
nuisance and the costs of time lost by air passengers in the
air and on the ground. The former were, at the most, generous
estimate, only some 0.7 per cent of the latter at the recom-
mended site. In essence this arose from the fact that the value
attributed to savings in travel time were (i) high and (ii) applied
to any time savings whether it was thirty minutes or one
minute. It is arguable that individuals only value time in
'discrete' units of, say, more than five or ten minutes, making
the CBA procedure used in the study illicit. It is not the case
that studies since Roskill suggest any significantly higher
valuations for noise nuisance. More detail of the Roskill
valuation of time and noise procedures is given in Dasgupta
and Pearce (1972). Using the low comparative value of the
environment in relation to saving travelling time, the critics
had a field day, with CBA being described as 'nonsense on
stilts' by one commentator, echoing Bentham on natural
rights (Self, 1970; 1975).
In the United Kingdom, then, CBA fell into some disrepute
in the 1970s, though it continued to flourish in the USA and
elsewhere. Much of the criticism was misplaced, being based
on simple misunderstandings of the role played by 'money' in
the technique. Money as a measuring-rod for benefits and

19
 The Origins of Cost-Benefit Analysis

costs became confused with money as some kind of immoral

or irreligious goal. At the same time, CBA had stolen a march
on other disciplines equally concerned to establish their own
applicability to the evaluation of road and airport expendi-
tures, environmental policy, and so on. Once the groundwork
for the attack on CBA was laid, the way was open for the
advocacy of allegedly superior techniques such as environ-
mental impact assessment. Writing in 1983, it seems fair to
say that the wheel has come full circle. CBA has matured,
and if much of the criticism of it was misplaced it is also the
case that many advocates of CBA 'oversold' its attractions.
Above all, the role of CBA is clearer in that, as Chapter 1
showed, CBA cannot be a unique rule for making decisions,
for it is always open to us to ask whether the particular value
judgements used to construct CBA are value judgements of
which we approve.
One question remains in respect of this historical excursion
into CBA's development. We have seen that CBA actually
achieved its rationale ex post. The practice, as it were, pre-
ceded the theory. The harnessing of welfare theory thus added
a powerful conceptual base for CBA. And yet CBA both
advanced in terms of practice, and in terms of the merging
of theory and practice, at a time when welfare economics
itself had come under the most severe criticism. As we have
seen, Scitovsky ( 1941) found an inherent flaw in the com-
pensation test advanced by Kaldor and Hicks. Little (19 50;
2nd edn 1957) attempted a resurrection of welfare theory
through the explicit treatment of income distribution. A
policy could only be judged a 'good' policy if it met the
compensation test, was not subject to the Scitovsky 'reversal'
procedure, and 'improved' income distribution. De Graaf
(1957) produced a slim volume criticising welfare economics
in such a way that many economists were satisfied that the
subject was quite dead in intellectual terms. At the same time,
the spirit of 'positivism' was afoot (some thirty years after it
had first received its clearest statements in philosophy with
A. J. Ayer's Language, Truth and Logic, 1936), and a good

20
 The Origins of Cost-Benefit Analysis

many economists denounced welfare economics, regardless of

the theoretical criticisms, because of its basis in value judge-
ments. That even the concept of allocative efficiency so basic
to the meaning of economics is itself value-loaded escaped
the positivists (Katouzian, 1980).
How, then, did CBA with its total reliance on welfare
economics survive and flourish? Krutilla (1981) has suggested
one not wholly convincing answer. It was, he says, because
CBA evaluated projects and not policies. In the case of the
former the distributional consequences were insignificant
and hence problems such as those raised by Scitovsky did not
arise. Had CBA attempted to evaluate whole policies, the
distributional consequences would have been significant and
CBA would have been open to an obvious intellectual attack.
There is some truth in this view, but it cannot be the whole
story because, in fact, there were few attempts to assess
whether the distributional consequences of projects were
important or not. Moreover, the kind of attack advanced by
de Graaf had not only included the distributional aspect but
also other criticisms, such as the use of shadow prices based
on marginal cost being erroneous because of second-best
problems. These could not be avoided simply by concentrating
on projects.
To explain the success of CBA thus requires more than
Krutilla's suggestion. Arguably, the explanation is very simple.
Those who practised CBA had a real-world task to attend to.
Someone had to decide on the priorities within any sub-budget
of government expenditure. The niceties of academic inter-
change in the learned journals did little to aid those who had
these tasks. Instead, it seemed that not only did CBA offer a
technique for aiding the evaluative process, albeit subject to
many caveats, it actually offered the only reasoned technique.
Founded, as it is, in a very simple concept of rationality (see
Chapter 1), CBA also had a fundamental attraction of reduc-
ing a complex problem to something less complex and more
manageable.

21
 The Origins of Cost-Benefit Analysis

Appendix: compensation tests and CBA

Figure 2.1 shows a production possibility frontier for an economy. We

select a point P 1 on that frontier and observe that we now have Y1 of
good Y and X 1 of good X. This particular combination of output levels
(X 1 , Y 1 ) can be distributed in various ways between the two individuals,
A and B, who comprise society. To see this, construct the Edgeworth
box OY1P 1X 1 for P 1 and inside it will be a contract curve which is the
locus of all points of tangency of A's and B's indifference curves (e.g.
IA and IB). If we now take the contract curve and show it in Figure 2.2
against the utility levels of A and B (UA and UB), it will appear as
UPC1 -a 'utility possibility curve'. But UPC1 is only one utility pos-
sibility curve relating to the production possibility curve in Figure 2.1.

Figure 2.1

22
 Compensation tests and CBA

Us

UPC1

DL--------------------------------L------------~---
Figure 2.2

For example, if we select P 2 we shall have a new Edgeworth box and a

new contract curve and hence a new utility possibility curve. Let this
be UPC2 as shown in Figure 2.2.
Now consider point C 1 in Figure 2.2 and consider a project that
moves 'society' from C 1 to C2 . We appear not to be able to compare C2
with C1 since C2 has improved B's utility but worsened A's. The strict
Pareto rule does not apply. But at C2 we can redistribute income and
move down UPC2 to point C 3 , which is superior to C1 because both A
and B are better off. Effectively, then, we have changed the move from
C 1 to C2 into a hypothetical move C1 to C3. The move from C2 to C3
involves B (the gainer) compensating A (the loser), but at C3 , B is still
better off than he was at C 1 . Hence C 1 to C2 is an improvement on the
compensation test and we only need to observe that B could compensate
A. He does not actually have to make the transfer.

23
 The Origins of Cost-Benefit Analysis

Now, consider C2 , the situation after the project is implemented.

We now hypothesise the 'reverse' project- i.e. moving from C2 back to
C1 . If the move is made, we see that at C 1 we can move up UPC 1 to C4 ,
which is Pareto superior to C2 • We have an oddity. The move from C1
to c2 is an improvement, but once at c2 the move back to c1 is also
an improvement. This is the Scitovsky 'reversal paradox'.

24
 3

The Measurement of
Costs and Benefits

Chapters 1 and 2 indicated that the basic CBA rule is that an

expenditure is to be judged potentially worth while if its
benefits exceed its costs, where benefits and costs are defmed
to include any welfare gain and loss which occurs because of
the expenditure on the project. But 'cost' has also to be
thought of as opportunity cost, the benefits forgone because
of the project in question. In tum, benefits are measured by
the (aggregate) willingness to pay (WTP) of the beneficiaries.
Expressed in this way, and ignoring the aggregation across
many individuals, we can write:
Net benefits = WTP; - WTPi (3.1)
where i is the project in question and j is the project 'forgone'.
We wish to know how we can go from this general statement
to a more detailed CBA rule.

Benefits and consumer surplus

Consider, first, the expression WTP;. Let us suppose that

project i consists of an increase in the output of product i
25
 The Measurement of Costs and Benefits

(e.g. an extra road, more railways, an additional runway at an

airport, less pollution, and so on). The demand curve fori is
shown in Figure 3.1.
Let the initial situation be that Q1 of i is bought at price
P 1 . The project in question changes the output and price to
Q 2 and P2 respectively. Chapter 2 indicated that the WTP for
Q1 is made up of the amount actually paid ( OQ1 XP1) plus
the excess of willingness to pay over actual price (PI XZ).
Clearly, then, the new total WTP for Q2 is OQ2 YP2 plus
P2 YZ. The WTP for the change in Q (= Q2 - Q1) is thus
Q1 Q2 YX. We require an expression for this area.
We can see that Q1 Q2 YX is made up of an extra amount
actually paid, Q1 Q2 YW, and the triangle WYX. (Note that
this triangle is the consumer surplus on the amount Q1 Q2.
It is not the change in consumer surplus by moving from Q1

Price
z

Demand fori

Figure 3.1

26
 Problems with consumer surplus

to Q1 because this is equal to WYX plus the extra surplus

now enjoyed on the amount OQI.) We can therefore calculate
the change in WTP for i as
AWTP; = QIQ1YW+ WYX
= AQP1 + !Q(PI - P2) (3.2)
The notation 'A' simply means 'change in', so in this case
AQ = Q1 - QI. The second expression on the right-hand side
assumes that the demand curve is a straight line (is 'linear')
and is the formula for the area of a triangle (half base times
altitude). We can rearrange (3.2) as

A WTP; = A Q [ P2 + (PI ; p 1 )]

=!AQ(PI +P2) (3.3)

We now have our formulation for measuring the benefit side
of equation (3.1 ). Benefits are equal to the change in quantity
multiplied by an average of the 'before' (PI) and 'after' prices
(P2 ). Now, if the change in price is very small we shall have,
at the limit, PI = P1, so that (3.3) would become
d WTPi =PdQ (3.4)
where the 'd' now replaces the 'A' because we are thinking of
very small changes in Q indeed. In fact, (3.4) defines the
meaning of 'marginal willingness to pay'. It follows that the
demand curve in Figure 3.1 is a marginal willingness to pay
curve; the first person to observe this important conceptual
linkage, and to coin the phrase 'consumer surplus', was Dupuit
(1844).

Problems with consumer surplus

The CBA procedure requires that the willingness to pay ex-

pression in equation (3.3) be aggregated across all the individ-
uals in question. Additionally, if there is more than one

27
 The Measurement of Costs and Benefits

benefit, we shall have to carry out a similar exercise for the

other gains obtained. We shall take it that, in order to avoid
generating more complicated notation, that both of these
aggregation procedures are undertaken.
But there are problems with the comparatively simple
procedure just described. First, the demand curve may not be
linear. If so, we shall have derived an approximation to WTP,
not the actual amount. Obviously, this is a problem in empiri-
cal technique. If we know the demand curve, then no great
problems are involved.
Second, the demand curve we have drawn in Figure 3.1 is
'Marshallian' - it simply shows quantity demanded against
price. As the price of i falls, the real income of the consumer
rises (he is better ofO. The problem now is that we are meas-
uring an area under a demand curve and calling it a gain in
utility when one of the factors affecting the way we measure
utility, namely income, is itself changing. We are measuring
utility with a money indicator which itself changes as the
price of i falls. What we need is a demand curve in which this
problem is removed, i.e. one in which we can see how utility
varies with price, but in which money income is varied so as
to keep the consumer on the same indifference curve. Our
money measure will then be a unique measure of utility. We
do not pursue this point here save to say that what we require
is not the Marshallian curve in Figure 3.1 but a 'compensated'
demand curve of the kind introduced by Hicks (1943) and
the area under which is termed a 'compensating variation'.
(For a fuller explanation see Pearce and Nash, 1981, pp.
90-1.) How far this adjustment matters - i.e. to what extent
significant errors in our measure of benefit are introduced by
using the Marshallian demand curve to approximate consumer
surplus - is an open question (but for an impressively argued
case that the errors are not of great significance see Willig,
1976).
There is a third problem arising from the fact that we must
aggregate surpluses to obtain an overall measure of benefit.
This arises from the fact that, as the price of the good in
28
 Problems with consumer surplus

question falls, it may well alter the demand curves for other
products which are either substitutes or complements for it.
If so, we need to know whether calculating the change in
surplus for the one good is enough as a measure of the
benefits of the project that give rise to that change in price.
If surpluses change on other products, we should presumably
take account of them. This is the problem of estimating con-
sumer surplus when other prices change. The appendix to this
chapter indicates the nature of the problem. It may be omit-
ted by readers concerned only to note that the problem exists
and must be accounted for in any actual CBA study. An
excellent treatment of the problem is given in Just et al.
(1982). (Indeed, for anyone concerned to pursue the welfare
economics foundations of CBA, this text is highly recom-
mended.)
Clearly, there are problems in actually estimating the con-
sumer surplus relevant to any project. If Willig (1976) is
correct, however, we shall not be dealing with large margins
of error if we adopt the simple measure of surplus as the area
under a Marshallian demand curve. Partly because of empirical
difficulties, and partly because of the view that the errors are
not large, we usually find CBA studies using 'simple' measures
of surplus. In terms of theoretical rigour, however, there is a
reasonable consensus that the measure we should be using is
the area under a 'compensated' demand curve, i.e. one in
which the income effect is removed. In turn, there are two
types of compensated demand curve with the result that we
can choose from three measures of consumer surplus: (a) the
area under the Marshallian curve, or 'simple' consumer surplus;
(b) the area under a demand curve that is adjusted so as to
keep the consumer on his original indifference curve, known
as the 'compensating variation', or CV for short; and (c) the
area under a demand curve that is adjusted to keep the con-
sumer on his subsequent indifference curve (i.e. the one after
any project has been introduced), known as the 'equivalent
variation', or EV for short. For the case we have analysed,
where the project causes the price of the product to fall, it

29
 The Measurement of Costs and Benefits

can be shown that CV will (generally) be less than the simple

measure, which is in tum less then EV. These different meas-
ures of surplus are due to Hicks ( 1943). Debate still occurs
on which to use. We again repeat the view that the error
involved in using the simple measure is not likely to be great
(Willig, 1976; Just et al., 1982). For the record, CV is the
measure most widely recommended in the theoretical liter-
ature (Mishan, 1975).

Costs and forgone benefits

In terms of the simple net benefit measure introduced in

equation (3.1), we now have an approach to its measurement
via the concept of consumer surplus (CV), albeit with some
cautionary notes about how to proceed. Equation (3.1) sug-
gests that we now need to subtract from WTPi the amount
WTPi which is a measure of the surplus forgone on the sacri-
ficed project {j). In fact, CBA does not proceed in this way.
What it does is to substitute for WTPi the costs of implement-
ing project i. Imagine now that the only costs relevant to our
project are what we have called the resource costs, the various
inputs - labour, capital, etc. - used up in project i. This
enables us to ignore other costs such as those imposed on
third parties in the form of, say, pollution (the 'externalities').
We can easily reintroduce them. In this context, then, we
now need to know whether the use of the cost of project i
introduces a significant error into equation (3.1 ), i.e. we need
to know the relationship between the cost ci of i, cj, and
WTPi. If they are not equal we shall have introduced an error
into the basic equation (3.1).
To answer this question we can look at the inputs used in
producing i. Imagine there is only one input, labour. (Our
result will hold equally well for more than one input.) Then
the forgone output in project j is equal to what the labour
used in i could have produced in project j. This will be the
marginal product of labour (MPL) multiplied by the amount
30
 Shadow pricing

of labour used in j (11L ), i.e.

Forgone output in j =MPL 11L (3.5)
and the value of this forgone output is WTP;, or
WTP; =P;MPL 11L (3.6)
Now, the relationship between expression (3.6), which is
WTP;, and C;, the cost of project i, is readily seen, since if
WL =P;MPL (3.7)
where WL is the wage of labour, we can write:
(3.8)
The equivalence between C; and WTP; is assured if, and only
if, labour is paid the value of its marginal product. Generalis-
ing, the equivalence will hold if, and only if, all inputs to i are
priced at their respective values of their marginal products.
This will only be true under perfect competition. Under im-
perfect competition, we know that

P·> WL =MC· (3.9)

I MPL I

where MC; is the marginal cost of producing j. Hence, from

(3 .9), the wage will be less than the value of the marginal
product (indeed, it will equal the marginal revenue product if
the context is one of profit maximisation).

Shadow pricing

The preceding analysis suggests that we know that the equiva-

lence of C; and WTP; does not hold. There will therefore be
an error involved in using C;. None the less, we shall continue
to use Ci for two reasons. First, for any individual project we
cannot engage in the laborious and virtually unlimited activity
of estimating WTP;, whereas we can obtain an estimate of C;.
Second, if we take the view that the error involved is serious,
31
 The Measurement of Costs and Benefits

we can try to calculate the value of the marginal product of

the inputs. The approaches to this are various and often
complex. For our purposes theideaoflooking for the 'proper'
measure· of the opportunity cost of project i in terms of what
the various inputs would produce in a world which obeyed
the efficient pricing principles derived from welfare economics
is the important one: for this is known as shadow pricing. In
the example above, shadow pricing of inputs would consist of
trying to find the true opportunity cost of using the inputs in
project i. Moreover, shadow pricing is generalisable across all
inputs and outputs. Indeed, the idea that we should estimate
the willingness to pay for the output of project i is itself
nothing more than shadow pricing. It is in this context that
we can derive a most important feature of CBA. This is that
the prices we use in order to draw up the CBA need not bear
resemblance to the prices that rule in the market for the vari-
ous outputs and inputs. This is so even when there are markets,
and the rationale for this statement is simply that the ruling
prices may not reflect (marginal) opportunity cost or (mar-
ginal) willingness to pay.
When there are no obvious markets at all, the procedure of
shadow pricing is one that we still pursue. If there is no
obvious market in the product in question - e.g. clean air-
then we can attempt to observe how individuals implicitly
value clean air by looking at surrogate markets, such as the
housing market. We shall be concerned to establish that any
procedure for eliciting values from the observation of such
markets obeys the requirements that they reflect willingness
to pay. In the situation in which we have no surrogate market
either, or in which we wish to test the results obtained from
surrogate market studies, we can resort to experimental meth-
ods such as questionnaires. Once again, the requirement will
be that the values elicited approximate willingness to pay.
It is as well to note that some cost-benefit analysts tend
to use the term 'shadow price', in a more general fashion than
others. Strictly, 'shadow price' relates to the valuation placed
on an input or output in a context of an optimum. That

32
 Shadow pricing

optimum will tend to be constrained by limitations on the

availability of some or all inputs, where the overall constraint
would appear as a limitation on the sum of money available
for investment in projects (a 'budget constraint'). One power-
ful mathematical procedure for analysing problems of opti-
misation in the context of constraints is known as program-
ming. It is from mathematical programming that the term
'shadow price' comes. As a general statement, we can say that
a shadow price measures the gain in the value of some objective
if we increase expenditure on a given project by one unit.
The objective in CBA is our measure of net social benefits.
Hence the shadow price of any output or input is the value of
the increase in net social benefit resulting from a one-unit
change in that output or input. The connection with willing-
ness to pay and opportunity cost should now be evident. The
gain in net social benefit from a one-unit change in output is
precisely what we have been calling 'marginal willingness to
pay'. The loss from increasing inputs into our project will be
the value of the output forgone in the sacrificed projects. (In
programming, shadow prices emerge as the solutions to what
is known as the 'dual' programme.)
We now have the basis for our cost-benefit rule. Instead
of equation (3.1) we shall henceforth use the expression:
Net benefits = WTPi - Ci (3.10)
where it will be noted that we have now substituted Ci for
WTPi.

We close this chapter with three observations:

First, we must reintroduce other costs besides the resource
costs of the project. We can do this by simply remembering
that Ci will be an aggregate of all the costs of project i, includ-
ing the costs borne by people adversely affected by the
project. To find these costs we will need to go through a
further exercise of estimating the CV (compensating variation)
relevant to the costs in question. Thus, if the project creates

33
 The Measurement of Costs and Benefits

pollution, we shall seek the sufferers' willingness to pay to

restore their initial level of welfare before the project, or
what they are willing to accept by way of compensation to
tolerate the pollution.
Second, we need to remember that equation (3.1 0) contains
a set of prices for the outputs and inputs and, regardless of
whether those prices are for outputs and inputs that have
associated markets, the (strictly) relevant prices are the
shadow prices.
Third, our formula is so far 'timeless'. The next chapter
introduces time into the analysis.

Appendix: consumer surplus when other prices change

We wish to know how to aggregate consumer surplus in the (likely)

context where the project we are analysing causes a change in the
demand curves for products which are substitutes or complements for
the product in question. In what follows product X is the one subject
to the initial analysis; its supply is increased because of some investment
project. In terms of the simple analytics of CBA we would normally
concentrate on X alone and ignore any substitute or complement for
X. To see how this affects the analysis, we postulate a substitute
product, Y.
The initial demand curves are shown as D~ and D'} in Figure 3.2.
Let the supply curve of X shift from S~ to S} so that the price of X
falls from J1 toP}. Since Y is a substitute for X, the demand curve for
Y shifts leftwards from D'?-, thus lowering the price of Y. In turn, this
will shift the demand curve for X to the left, and so on. Let the fmal
equilibrium demand curves be D} and D} so that the final equilibrium
prices are J1. andP}. What is the aggregate surplus gain from the initial
price fall in X?
First, we may observe that surplus is not measured by the areas under
the demand curves D~ and D'} taken together. This is what would
appear to be correct if we simply added surpluses. In fact the measure

34
 Consumer surplus when other prices change

Py
Px

s)c

p~

D~
0
Ox Oy

Figure 3.2

of surplus gain is ambivalent. There are three possibilities. These are:

1. the area under D'k (for X) and under D} (for Y)
2. the area under D} (for X) and under D'}- (for Y)
3. assuming that prices move to their new equilibrium at the same
proportionate rate, by the area
(P~ABP}) + (PtCDP})

shown as the shaded areas in Figure 3.2.

Taking measure 3, this is equal to
(P~ - P} )Q't + !(P~ - P} )(Qi- - Q'k) + (P~ - P})Q}
+ !(P~- P})(Q~- Q})
or

The fact that there are three (indeed, many) possibilities of measure-
ment arises because we can choose different paths whereby initial prices

35
 The Measurement of Costs and Benefits

change towards fmal equilibrium prices. The measure of surplus depends

on the choice of path - i.e. there is what is known as path dependence.
In measure 3, for example, the path is linear.
In fact, as long as certain conditions hold, the various measures will
be equal. The prime one is that
3Qx 3Qy
--=--
3Py 3Px
at all sets of prices - i.e. the cross-price derivatives must be equal. This
'symmetry of substitution' will in fact hold if the demand curves in
question are exclusive of income effects (i.e. are 'compensated' demand
curves) but not otherwise.

36
 4

Time, Discounting
and Decision Rules·

Chapter 3 concluded with a 'timeless' CBA formula which

indicated that any project in which the benefits exceed the
costs is a potentially 'good' project. The qualifications are (a)
that the benefits and costs must be valued at shadow prices
unless we have reason to believe that the error involved in not
using shadow prices is small; (b) all costs and benefits, to
whomsoever they accrue, must be accounted for; (c) the rule
as formulated is indifferent between who receives the benefits
and who suffers the costs (a matter we take up in Chapter 5);
and (d) the excess of benefits over costs does not mean that
the project should be undertaken. With respect to point (d)
we shall see later that the introduction of a budget entails
that we must have some ranking procedure.

The rationale for discounting

The costs and benefits in question will occur over time. If,
for example, the benefits accrue at a constant rate over thirty

37
 Time, Discounting and Decision Rules

years, and the costs occur in the first five years, but not there-
after, we could formulate the rule that the project is worth
while if B1 + B2 + ... + B3o is greater than C1 + C2 + ... +
Cs . However, it is unsafe to assume that a benefit in year 2 is
regarded by society in the same way as a benefit in year 1.
Suppose the benefit is £1 in each of these years. If we are to
treat them identically, it will be quite legitimate to add up
the benefits (and costs) in the way we have just done. But
we have two basic interrelated reasons as to why we should
not do this.
First, treating £1 in year 2 as being the same as £1 in year
1 implies that society is indifferent as to when it receives the
£1. But casual reflection will indicate that this is not so. For
by taking the £1 in year 1 we can put it into a savings account
(or some other form of money investment) and earn £1 plus
the rate of interest on £1 by year 2. This means that we shall
value £1 in year 1 at more than the £1 in year 2. We would
rather have the £1 earlier than later simply because there
exists a positive interest rate in the economy. If we argue
that interest rates exist because of the productivity of capital,
then we have one reason for arguing that a given unit of bene-
fit is worth less the further and further it occurs in the future.
Second, we could simply observe the behaviour of indivi-
duals and conclude that, regardless of interest rates, people
do prefer their benefits now rather than later. They could
simply be impatient. Or they might think that, as they will
be richer later on, the £I in the future will mean less to them.
Hence the earlier the £1 accrues, the more they prefer it. On
either, or both, of these criteria, individuals will have positive
time preference - they will prefer now to later.
If we now recall the value judgements underlying our con-
ventional CBA, we observe that the first of these, consumer
sovereignty, dictates that consumer preferences matter. Thus
we cannot logically exclude individuals' preferences about
the incidence of costs and benefits through time. In turn this
means that we must 'discount' future benefits and costs. One
procedure for doing this is indicated by looking at the first
38
 The rationale for discounting

rationale for discounting, the existence of interest rates. £1 in

year 1 would accumulate to £( 1 + r) in year 2 if the interest
rate is r per cent (r is typically expressed as the corresponding
decimal - e.g. 5 per cent would be 0.05, 12 per cent would
be 0.12, and so on). Looked at from the standpoint of year 1,
we can ask the question: 'How much is £1 in year 2 worth to
us in year 1?'The answer will be that it is worth £1/(1 + r),
for the simple reason that if we had this sum in year 1 we
could invest it at r per cent and obtain in year 2
£1
- - x (1 +r) = £1
(1 + r)
In the same way, we see that £1 in year 3 can be expressed as
a value to us in year 1 as follows:
£1/(1 + r) 2
since in year 3
£1
-0-+-r)-=-2 x (1 + r) x (1 + r) = £1

We now have the general formula for discounting. A benefit

B in any year t can be written as Bt, and from the above pro-
cedure we know that this benefit will have a value to us in
year 1 of
Bt
(4.1)
(1 + r/
Notice the procedure whereby we look at future benefits
(and costs, the procedure is the same) from the standpoint of
the present. This is why expressions such as equation (4.1)
are called present values. The procedure for finding a present
value is known as discounting and the rate at which the bene-
fits or costs are discounted is known as the discount rate. In
CBA we are concerned with the costs and benefits to a whole
society. Hence the discount rate we use will be a social dis-
count rate.
39
 Time, Discounting and Decision Rules

Before proceeding to the sources of the social discount

rate it is as well to remove one potential source of misunder-
standing. The discount rate has nothing to do with inflation.
This is because there is always one adjustment to be made
with respect to benefits and costs and this is to ensure that
they are expressed in real terms, i.e. net of any general move-
ments in price levels. This is done by expressing their values
in terms of some base-year set of prices. Since benefits and
costs are in real terms, so is the discount rate, to ensure that
inflation is not relevant to the determination of the discount
rate. In particular, the reasons as to why individuals discount
the future are exclusive of any expectations they might have
about the rate of inflation. This said, it is sometimes the case
that actual CBA studies quote both 'money' and 'real' dis-
count rates. For example, with 7 per cent inflation, a 5 per
cent real discount rate is the same as a 12 per cent money
rate of discount.

Finding the discount rate

In order to analyse further the foundations of a social dis-

count rate, consider Figure 4.1. This shows consumption in
two years, t and t + 1. The function TT' is a transformation
function, or production possibility curve, but, instead of the
familiar form of alternative configurations of the production
of two goods, it shows the possible configurations of produc-
tion between two years. It says, for example, that if resources
are wholly devoted to year t, output will be OT. If they are
wholly devoted to year t + 1 output would be OT'. Notice
that OT' is greater than OT. This is intuitively acceptable
because devoting resources to production in year t + 1 means,
given that we have measured the transformation in terms of
units of consumpton, that OT' can only come about by in-
vesting all the resources that would have been consumed in
year t. That is, if the economy is at position T' it means that

40
 Finding the discount rate

OT was invested in year t and the consumption goods resulting

are OT' in year t + 1.
Also shown in Figure 4.1 is what looks like an indifference
curve. This is a social indifference curve indicating the com-
binations of consumption in period t (Ct) and consumption
in period t + 1 (Ct+d between which society is indifferent.
This is denoted by SS'.
Clearly, society will be in an optimal position if it located
at point X, for then society is able at this point to climb on
to the highest possible social indifference curve given the con-
straints set down by the function TT'. In fact we can find
Cr+1

Figure 4.1
.. T Cr

41
 Time, Discounting and Decision Rules

out just how much investment and consumption occurs if

the economy settles at X. We can read off Ct and Ct+l
immediately and we see that these are given by Ct and
Ct+ 1 . But TT embodies the investment that also takes place,
i.e. the difference between Ct and OT must be the level of
real investment in year t, ft. In tum we see that it is It that
generates the consumption level Ct+1· Using these ideas we
can establish some important equations.
First, observing the level of investment It, we can see that
Ct+ 1 YCt + XY YCt XY
--= =--+- (4.2)
It It It It
But YCt = It because YC t is drawn by constructing the 45°
line YT. Hence
Ct+1 XY
--=1+- (4.3)
It It
Note that the first expression in (4.3) is the gross productivity
of capital, and XY/It is in fact the net productivity of capital,
or its internal rate of return (or marginal efficiency of capital).
Yet this latter concept is precisely the discount rate intro-
duced as the first of the two alternatives, i.e. the interest rate
in the economy, r. Moreover, if we make It very small in
Figure 4.1, we can see that Ct+t!It measures the slope of
TT'. Hence we can rewrite (4.3) as
Slope of TT' =1 + r ( 4.4)
where r is now the marginal rate of return on capital.
Turning our attention to SS' we can proceed in a similar
way. Consider points J and Kin Figure 4.1. These are points
on the same social indifference curve, so that the utility lost
by moving from K to J would be llCtMUt -i.e. the change
('.!:l ') in Ct multiplied by the marginal utility associated with
Ct. The utility gained would be llCt+1MUt+1· Since J and K
are on the same indifference curve, we can write
-llCtMUt =llCt+1MUt+l (4.5)
42
 Inequality of the two rates of discount

and hence
-ACr+l MUr
Slope of SS' = (4.6)
ACt MUr+l
The slope of SS' is simply the ratio of the two marginal
utilities of consumption (which is what we would expect
from our knowledge of indifference curves in general).
Now, as we move along SS' in the direction J to K, society
will tend to require more and more Cr+l to compensate for
a unit loss of Cr. Very simply, ACr+l /ACt > I. Hence we
now have, from (4.6),
MUr
-----'- > 1 (4.7)
MUr+l
Writing the excess of this ratio over unity asS, we have
MUr
-----'- = 1 + s = Slope of SS' (4.8)
MUr+l
We now defines as the social rate of time preference.

The inequality of the two rates of discount

The preceding section has demonstrated that we have two

candidates for the social discount rate. The first is r, which
we saw was the marginal net product of capital, or the rate
of return, or the marginal efficiency of capital. In the litera-
ture on CBA it is also known as the social opportunity cost
of capital, the idea being that we should calculate r for CBA
purposes by looking at the projects that are displaced by any
given investment and seeing what rate of return they would
have earned. The intuitive logic of this approach is very
appealing. To use the rate r for discounting purposes is
equivalent to saying that our project in the public sector
must do at least as well as the projects it displaces, and these
may well be in the private sector.
For example, suppose we are considering investing in a
43
 Time, Discounting and Decision Rules

road project. We have already determined, say, that the rate

of return on marginal projects in the private sector is 8 per
cent. If we use a discount rate of 8 per cent for our public
project and the resulting present value of benefits minus costs
is negative, i.e. costs exceed benefits when discounted, then
what we have said is that the project in question must be
rejected because the funds are better diverted to projects
in the private sector which can earn 8 per cent.
The second discount rate iss. The problem we have is that
sis not observable, at least directly. But suppose we determine
that it is 5 per cent. Then we would apply this rate to all
public projects because it represents the rate at which society
is prepared to trade present for future consumption. This
social time preference rate is obviously consistent with the
value judgement about consumers' sovereignty.
Are r and s likely to diverge? From Figure 4.1 we can see
that there is one circumstance in which they will not diverge,
namely when the economy is operating at X, which is a
(constrained) optimum. If X is an optimal position, then
the level of investment, / 1 must also be an optimum. And at
X the values of r and s are the same because the slopes of
TT' and SS' are the same. Hence we conclude that if the level
of investment in the economy is optimal, s = r and we have
no problem about the selection of the discount rate. We can
use either approach, and since r seems eminently more
observable than s, it would seem sensible to use r. At X,
then, the social time preference rate equals the social op-
portunity cost of capital. Again, this would seem to be an
intuitively sensible result if the capital market operates
perfectly since the rate of return obtained on projects in
the economy should be equal to the rate of interest savers
require, and that rate of interest is in tum determined by
individuals' time preference. That is, the forces that deter-
mine the real interest rate in the economy are the demand
for funds, and this reflects the availability of projects with
positive net capital productivity, and the supply of funds,
which reflects savers' behaviour.

44
 Inequality of the two rates of discount

In practice s does not equal r, and for many reasons. We

can consider one simple explanation. If the public sector
can borrow at s per cent the private sector of the economy
will have to do better than s per cent to attract funds from
investors. This is because of the existence of company taxa-
tion (corporation tax), which means that if a company is to
pay s per cent to its lenders it must in fact earn r per cent,
where
s
r=--
1- t
and t is the tax rate on companies and is less than unity. For
example, if the tax rate is 40 per cent, and s is I 0 per cent,
then companies must earn

I ~·~.4 = 0.167
i.e. 16.7 per cent, for them to be able to pay I 0 per cent to
investors. Hence company taxation necessarily makes r greater
than s (Baumol, 1968).
There are other reasons for supposing that s and r diverge.
Government projects tend to be 'riskless', not because they
are individually subject to less risk than private-sector projects
(often the reverse is the case), but the sheer size of the
public sector means that the risk per project is small because
of the ability to 'pool' risks across many projects, or across
the many people who make up society (Arrow and Lind,
1970). If risk in the private sector is positive, then the value
of r will be increased to reflect the 'risk premium'. In this
way r will diverge even further from s. Last, the value of r
reflects decisions made by individuals acting in their own
interest. The relevant value of s, however, could be argued
to be that which society expresses when it considers projects
from a 'social standpoint'. That is, acting in isolation of one
another, individuals will express one discount rate, but if
they know that their decision to invest is going to be asso-
ciated with the decision of many others to invest they will
45
 Time, Discounting and Decision Rules

tend to quote a lower rate of discount (Sen, 1967). Again,

the effect is to produce a high r relative to the value of s.
All this suggests that, instead of operating at point X in
Figure 4.1, the economy will be operating at a point below
X - i.e. somewhere on the section TX, such as Z. But if it
is doing this then it is on a lower social indifference curve,
such as S 0 as shown. It is thus in a 'second-best' situation.
In such a position we can choose from three options:

1. use r and reject s

2. uses and reject r
3. adjust the decision rule for CBA so that both rands are re-
flected in the decision.
Each option has its adherents in the literature on CBA. The
reasons for using s or r have already been given. Accordingly,
we confine our attention to the issue of how the decision rule
for CBA might reflect both s and r. Before doing that we
should observe that there is no very clear consensus either of
how to compute the value of r and s anyway! Since s is the
more complex of the two, the appendix to this chapter indi-
cates a procedure for deriving values of s. The reader should
not assume, however, that finding values for r is without its
difficulties.

Adjusting the decision rule

The decision rule obtained in Chapter 3 was that benefits

should exceed costs before we consider a project to be a
candidate for acceptance. This chapter has shown that we
need to introduce time and discounting. The formal require-
ment for potential acceptance, then, is now as follows:

E Bt
t=l(I+r)f
-~ Ct
t=l(I+r)t
>0 (4.9)

46
 Adjusting the decision rule

where
Bt is the benefit in time t
Tis the 'time horizon'- i.e. the period over which bene-
fits and costs are calculated
r is the social opportunity cost rate of discount
L means 'sum of'
Note that (4.9) could be rewritten using s instead of r. The
expression involving benefits is the present value of benefits,
PV (B), and PV (C) is the present value of costs. We can thus
write:
NPV = PV(B)- PV(C) (4.1 0)
where NPV reads 'net present value'. Expression (4.9) requires
that NPV> 0.
Before looking at ways in which we can try to account for
the second-best position in respect of r and s, it is as well to
note two alternative formulations of the rule in ( 4.9) which
are common in the CBA literature. Instead of the summation
sign L, we may find use of the integral sign f. This is con-
venient when the discount factor 1/(1 + r)t is expressed in
terms of 'continuous time' - i.e. we do not break time up
into discrete periods such as years. In this case ( 4.9) would
appear as

fi~(Bte-rt- Cte-rt) > 0 (4.11)

(Notice the use of e-rt, which is the value e = 2. 718 ... , the
natural base for logarithms, raised to the power minus rt,
where r is the discount rate; e-rt is the same as I fert .)
Particularly convenient for deriving certain results (as we
shall see) is the fact that when Tis made equal to infmity, so
that we calculate benefits and costs over an 'infinite time
horizon', then the integrals in equation ( 4.11) have the values
B /r and C/r respectively, so that ( 4.11) would read:
B C B-C
--- = - - > 0 (4.12)
r r r
47
 Time, Discounting and Decision Rules

Now, we wish to know how the decision rule is affected by

an attempt to introduce both s and r into the analysis. Various
approaches have been used to re-express the decision rule
in this way. We follow the approach pioneered by Marglin
(1963; 1967). A separate approach is that given by Feldstein
( 1972), but it is possible to demonstrate that, under certain
conditions, Feldstein's approach is the same as Marglin's (see
Pearce and Nash, 1981). The result obtained in (4.12) above
will now prove useful. We shall work with infinite time
horizons, which means that our benefits and costs are per-
petuities. Marglin's approach is generally spoken of as pro-
ducing a 'synthetic' discount rate. In fact it does not derive
a single discount rate at all, but adjusts the NPV(B) formula
in equation (4.9).
The essence of all synthetic approaches is that they dif-
ferentiate the sources of finance for a project. They may also
differentiate types of benefit. Sources may be broadly split
between taxes and borrowing, while benefits may be dif-
ferentiated according to whether they generate cash flows,
which are reinvestable, or benefit flows, which are not. Let
benefits be constant at B per year; then we have
B
PV (benefits)=- (4.13)
s
given that we are working with perpetuities, and s, the social
time preference rate, is the 'fundamental' discount rate.
When looking at costs, however, we need to consider what
proportion of the finance comes from taxes and what from
government borrowing. The general argument is that taxes
are at the expense of forgone consumption so that the cost
is measured in consumption units. Borrowing is held to be at
the expense of forgone private investment. Let the total
capital cost be K and let it all occur in the very first period,
such that
K=I+ Co (4.14)
where I is forgone private investment and Co is forgone
48
 Adjusting the decision rule

consumption. Thus I can be thought of as earning r, and C

can be thought of as earnings. In perpetuity and discounting
at s we then have
Ir Cs Ir
K=- +- =- +C0 . ( 4.15)
s s s
The required inequality for potential acceptance of the
project is now
B Ir
->-+Co ( 4.16)
s s
Note that if.Co = 0, we have
B Ir
->-
s s
or
B
-> r ( 4.17)
r
which, in the simple case under consideration, is a restate-
ment of the social opportunity cost argument.
Now we can consider what happens on the benefit side.
For every £1 of benefit flow, assume a fraction b accrues as
a reinvestable cash flow and the remainder ( 1 - b) as a con-
sumption benefit which cannot be reinvested. Then we can
write:
B(£l)=br+(l-b) ( 4.18)

where it will be seen that the reinvestable fraction earns a

rate of return r when reinvested. Notice again that if all
benefits accrue as non-reinvestable consumption benefits,
b equals 0 and £1 of benefit is simple recorded as £1 of bene-
fit. By allowing for reinvestment we have implicitly attached
a shadow price to £1 of (nominal) benefits. The benefit side
49
 Time, Discounting and Decision Rules

of the equation now reads

B
-[b/r+(l-b)] (4.19)
s
Substituting (4.19) into ( 4.16) brings both the source of
fmance and type of benefit issues together. We obtain

~[b/r + (1 -b)] > [1:' +Co]

or
B
- > -1/r+
- -Cos
--- (4.20)
s s[b/r+(l-b)]
Multiplying both sides by s gives
1/r +Cos
B > ---"-'--'--- (4.21)
b/r+ (1- b)
Equations (4.20) and (4.21) describe the essence of the syn-.
thetic approach. To check its validity, suppose reinvestment
possibilities do not exist. In that case b = 0, and equation
(4.21) will reduce to
B>Ir+Cos
or
B 1/r
->-+Co (4.22)
s s
which is the same as equation (4.16)
We have now derived a CBA rule which incorporates both
sand r. There are problems with the approach which we do
not pursue here (see Pearce and Nash, 1981). The approach
by Feldstein (1972) can also be demonstrated to fit in with
the previous derivation, though once we move from the
hypothetical world of perpetuities the formal equivalence
breaks down.

50
 Ranking projects

Ranking projects

Equation (4.20) gives us a procedure for relating costs to

benefits over time. We see that we need to estimate the flow
of benefits B and the flow of sacrificed consumption Co
(note that Co is not an expression for cost, but instead refers
to consumption). We also need values for r and s and the
fraction, b, of benefits that are likely to be reinvested. Ob-
viously, the use of a 'synthetic' procedure such as that
described in the previous section will begin to impose for-
midable informational difficulties. Largely for this reason,
but also because argument about the use of 'synthetic'
procedures still goes on, most CBA studies make use of the
more simple approach given in equation (4.9) using either
r or s as the discount rate. We revert to that simple procedure
in order to demonstrate a proposition we have so far only
hinted at.
The requirement that B be greater than Cis not sufficient
for us to sanction investment in the project in question. We
can say that B > C is a necessary condition for approval, but
it is not sufficient. This is because we shall invariably face a
limited budget and we cannot undertake all projects where B
exceeds C. We therefore require a ranking procedure. It is
tempting simply to rank by the value of NPV(B). But this is
actually mistaken.
This is easily seen by looking at Table 4.1. This shows
three projects, X, Y and Z, with their present value of their

Table 4.1

Project PV(C) PV(B) NPV(B} PV(B) + PV(C)

X 100 200 100 2.0

y 50 110 60 2.2
z 50 120 70 2.4

51
 Time, Discounting and Decision Rules

benefits, costs and net benefits. Suppose the budget con-

straint is 100 units. Then a ranking by NPV(B) would suggest
X, Z, Y and we would undertake X only with a cost of 100.
The gain to society would be NPY(X) = 100. But casual
inspection shows that we could afford Y and Z, and the NPV
would be NPV(Y) + NPV(Z) = 60 + 70 = 130. Clearly ranking
by NPV does not give us the right answer. This is given by a
ranking of PV(B) divided by PV(C), or the so-called benefit-
cost ratio.

Discounting and future generations

Regardless of how we derive a discount rate, it it likely to

produce a positive value. Suppose now that the project in
question has benefits that accrue for thirty years but costs
which, while small each year, accrue for 500 years. This
might typify, say, the problem of investing in the safe storage
of nuclear fuel waste and other wastes from nuclear power
stations. The problem, in CBA terms, would present itself as
comparing the costs of the storage system (say placing the
waste in glassified blocks, inside metal containers and then
locating them in depositories deep underground) with the
benefits in terms of the reduced risk of exposure to radio-
activity. Clearly, the more we spend on the storage system,
the lower the risks are likely to be. But let us suppose that
we can never quite get to absolute certainty that no adverse
consequence will occur. We then have (probabilistic) costs
over a long period of time and 500 years is a quite reasonable
estimate of how long we would wish to exercise control over
the storage system, provided of course that we care for future
generations. If we engage in conventional CBA we shall dis-
count the benefits and costs. We can now highlight a problem.
Suppose that the event of an accident is known to us with
certainty and that it occurs in year 500 after we begin
storing the waste. Let us be somewhat more heroic and
suppose that we also know what the cost will be if such an
52
 Discounting and future generations

accident occurred, and we set it at £1 0 billion at today's

prices. Our discounting formula requires us to estimate the
present value of the cost of the accident at
£10 billion
(1 +r)soo
Suppose r is 5 per cent. Then we need to calculate the 'dis-
count factor' 1/(1.05) 500 . Reference to the logarithm tables
will show us that this is the reciprocal of 3.9 x 10 10 . Hence
the present value of the accident cost is
£109 xl0 1
3.9 X 10 10 = 3.9 = £0. 25
The £10 billion accident has a cost of only 25 pence in
present value terms!
While the example may be thought to be an extreme one,
it illustrates a basic problem with CBA, namely that the
effect of discounting is to discriminate against the future.
Various views can be taken about this. There are those who
argue that we cannot take account of costs to generations yet
unborn, for to do so is to widen the concept of 'democratic
voting' in an unacceptable way. Those who are alive at the
time of the decision constitute the 'proper' electorate. Others
draw attention to the fact that the kind of 'intergenerational
discrimination' implicit in discounting is an increasing feature
of our society. Examples might be the potential for heating
up the atmosphere through continued burning of fossil fuels
(the 'greenhouse effect'), nuclear power waste problems,
continued and expanding use of toxic metals and chemicals
which do not degrade in the environment, the use of chloro-
fluorocarbons (CFCs) which punch 'holes' in the stratosphere
and increase the amount of ultra-violet rays in certain areas,
perhaps inducing skin cancers, and so on.
It seems fair to say that there is no consensus at all on
what to do about this aspect of CBA. Elsewhere I (Pearce,
1983) have drawn attention to the fact that, whatever the
53
 Time, Discounting and Decision Rules

ethical basis we use for CBA, the discrimination problem

is overcome if we can set up an 'intergenerational compensa-
tion fund'. This would simply apply the Kaldor-Hicks
criterion through time but with actual compensation occurring.
Suppose, for example, that we believe the cost to future
generations in year I 00 will be £I billion. In order to over-
come the discrimination we set aside a compensation fund in
year 1. If the interest rate is 5 per cent (notice this need not
be the same as the social discount rate - we are interested in
investing in a compound interest fund so that only actual
interest rates are relevant), then we know that we can set aside
£1 billion
(1.05)100
in year I and this will grow at 5 per cent for I 00 years to
become
£1 billion 100 _ ..
(l.0 5 ) 100 x (1.05) - £1 billiOn

in year 100. More generally, if the anticipated cost is C1 in

year t, then we need to set aside
X= Cre-rt
as a compensation fund.
The problems with this approach are fairly obvious. We
may not know when the cost will occur, nor the size of it,
nor can we be sure that the value of r will be constant for
I 00 years (indeed, we can be sure it won't be!). But un-
certainty is a fact of life, and while the compensation fund
concept is open to considerable practical difficulties it is
perhaps worth pursuing. We may note that many cost-
benefit analysts would argue that such a fund is not needed
anyway. They would say that future generations are already
compensated by the fact that the project in question will
add to the capital stock of the nation and hence to future
capabilities to invest in solutions to any problems 'shifted'

54
 Deriving a social time preference rate

on to them by current generations and to invest in welfare-

improving activities. It is perhaps this argument that dis-
tinguishes the intergenerational 'hardliners' from those who
would set a constraint on CBA to the effect that no invest-
ment takes place if it constrains the choices of future genera-
tions. Space forbids a detailed treatment, and the reader is
referred to some of the excellent contributions to these
issues (Page, 1977; Goodin, 1982; Barry, 1982).

Appendix: deriving a social time preference rate

Typically, cost-benefit analysts adopt a discount rate for marginal

income of greater than zero, at least partly because the group of per-
sons making decisions about future costs and benefits comprise the
current generation. The current generation may exhibit concern for
future generations, but that concern is not 'wholly altruistic'. As we
have seen, a discount rate for public-sector decision-making can be
obtained by:
1. adoption of a direct estimate of a (social) time preference rate, or
2. adoption of a direct estimate of an opportunity cost rate or
3. some mixture of both.
The foregoing discount rates are generally referred to as 'efficiency'
rates because they have a close relationship to the economist's con-
cept of an efficient allocation of resources both within any time period
and through time. They apply to income, and it is often argued that
the corresponding discount rate for utility is lower. This is because
with economic growth people in future will be richer and may suffer
less from a given small decrease in their income and conversely be
willing to pay more for a given increase in utility. However, efficiency
rates of this kind virtually always include the judgement that the
current generation's preferences shall dominate. If this judgement is
changed so that marginal utility to future generations is placed on an
equal basis ('pure altruism') with damage to current generations, then
the relevant discount rate for utility becomes zero.

55
 Time, Discounting and Decision Rules

A change in income leads to a change in welfare or general well-

being or 'utility'. The source of this utility is consumption, which is
proportional to income. A typical assumption is that the consumption-
utility function has the form

U= U(C) (4.23)
where
dU
- = U'=ac!' (4.24)
de

and a and b are constants. U' is the marginal utility of consumption

-i.e. the extra utility arising from an extra unit of consumption -
and this exhibits a declining level as dC increases (i.e. b is negative).
Now, if consumption can be expected to grow through time, dU/dC
will have a different value at time t than now (t = 0). Thus Cft = a~
and UO = a~. The weight to be given to marginal consumption in
period t is then

u; ad/ d/
w = - = - = - = (1 +c)bt (4.25)
t u;a~ C0

where c is the annual rate of growth of consumption per .head through

time. The value of cis something that can be directly estimated. The
value of b is not observable but has generally been regarded, on indirect
evidence, as having a value of -l to -2. Moreover, the measure of
consumption that is relevant is consumption per capita - otherwise
we would have a situation in which individual welfare rises simply
because more people have the same average level of consumption. Hence
the growth rate 'c' can be decomposed into two parts -the growth
rate of total consumption, k, and the growth rate of population, 1r.
Equation (4.25) would then become

w
t
= (~)b
I+tr
(4.26)

In tum, Wt can be expressed as a discount rate, r, in the following

56
 Deriving a social time preference rate

manner:

1 (1 +k)bt (4.27)
Wt = (1 + r)t = 1 + 1T

From (4.27) we have

(I +ri= ( - -
1+ 1T) bt
1+k
Therefore,

1 +r=
1+1T)b
(- -
1+k
Therefore

r- -(11+1T+
k)-b- 1
-- (4.28)

Now, it is also possible to discount utility, and some writers (e.g.

Olson and Bailey, 1981) reserve the term 'positive time preference'
for utility discounting alone. If we now discount utility at a rate p,
we shall modify equation (4.28) to

- 1 1 (l+k)b
wt=1+r=1+px 1+1T (4.29)

where w r r
and indicate a modified w and r, and is a social time
preference rate which discounts income first because of its declining
marginal utility (giving equation (4.28), and then additionally because
of positive time preference.
Setting

1+p=el'=p_
1 +k=ek =k

1 + 1T =e11 =1T
57
 Time, Discounting and Decision Rules

and modifying (4.29) to read

- (1
l+r= -+7T)b
- x(l+p)
l+k
(4.30)

we have
r=TI.b -!b +f!..
= TI.b - (f + TI.)b +!!..
=!!.- E_b (4.31)
Thus the discount rate for consumption, and hence for income, is made
up of the positive time preference rate {p), and the growth rate of con-
sumption (c) and the marginal utility of income (b) (b is negative, so
the two discount rates are added). There is 'hard' evidence on c, a sort
of consensus on b, while the estimate of p will have to be purely
judgemental. If we adopt the 'pure altruism' approach, p vanishes and
only c and b are relevant. Table 4.2 shows a range of discount rates
that emerge from the analysis assuming economic growth such that
c=2.

Table 4.2
r for values of:
p
-c = 2 ' b = -1 f. = 2' b = -1 .5 £=2,b=-2

0 2 3 4
1 3 4 5
2 4 5 6

Some fmal brief comments are worth making about this derivation
of the social discount rate. If we reject utility discounting on the
grounds that it is irrational (Ramsey, 1928; Pigou, 1952), then p = 0
and the discount rate is determined by cb alone. But the expectation
that c will continue to be positive (for long periods of time) could be
false. Perhaps future real incomes will decline or simply stay constant.
If so, c = 0 and the discount rate will become zero. It is possible, then,
to get to the 'altruistic' discount rate of zero without engaging in
ethical arguments about the rights of future generations.

58
 5

Efficiency and
Distributive Weights

Chapter 1 indicated that there are two fundamental norma-

tive judgements underlying what we might term 'conventional'
CBA which we have formalised in terms of the net benefit
criterion in Chapters 3 and 4. These judgements were (a) that
consumer preferences count, and (b) that the existing distri-
bution of income is, in some sense, 'optimal'. We now need to
investigate this second judgement more closely.
Imagine that society consists of four individuals, A, B, C,
and D. We can for the moment ignore time (and hence the
discount rate). Let A, Band C each gain £100 from a project,
but D loses £200. Our CBA formula tells us that the sum of
the benefits is £300 and the sum of the losses is £200, so
there is a net benefit of £100. The project is potentially
worth while. We could write the net benefit calculation as
follows:
Net benefit= aA BA + asBs + acBc- aDBD
where B A , is the benefit to A, .... The introduction of the
'weights' aA, as, etc., serves to underline an important feature

59
 Efficiency and Distributive Weights

of conventional CBA. This is that it sets a weight of unity to

each £1 of benefit (or cost) regardless ofwho receives that
benefit or who suffers the cost. This is essentially what we
mean by saying that conventional CBA adopts the second
value judgement, namely that the income distribution is
optimal. To emphasise the point, the fact that D could be
very poor and A, B and Call equally rich is immaterial for
conventional CBA.
It is this kind of consideration which has led to the emerg-
ence within the CBA literature of a school of thought which
seeks to vary the weights aA , etc. Typically it will do this in
such a way that, given the statements about the comparative
incomes of A, ... , D in our example, it will make an greater
than aA , ... , Ac. Now, within this school of thought, dubbed
'revisionist' by Mishan (1982), there are two sub-schools. One
would argue that the numerical values of the weights aA , ... ,
an are determined by reference to the political system. That
is, one would look to the decision-maker's objectives and
derive from them a set of weights, and only one set. In this
respect the weights become shadow prices just like any other
shadow price such as the price to be used for labour, the
social discount rate, and so on. Examples of this kind of
approach can be found in Little and Mirrlees (1974) and
Marglin, Sen and Dasgupta ( 1972). It seems fair to say that
this approach has found its major use in projected appraisal
in less developed countries. The other sub-school of the
revisionists would argue that, while derivation of weights from
overt statements by politicians is not ruled out, it is open to
the analyst himself to indicate the way in which the results of
the CBA vary with differing judgements on the weights. For
this approach, then, there will be several sets of results, and
the prodcedure has been called 'value sensitivity analysis' by
Nash, Pearce and Stanley (1975).

Conventional CBA vs 'revisionism'

There are many objections to the use of these distributional

60
 Conventional CBA vs 'revisionism'

weights, and they have been forcefully expressed by Mishan

(1974; 1981; 1982). We can deal with some general ones
before developing procedures for deriving the weights, after
which we can consider some more detailed objections. The
first objection is, effectively, that while conventional CBA
proceeds as if the existing distribution of income is optimal,
it is always open to government to adjust the distribution of
income by wholly separate means, e.g. by making lump-sum
transfers from gainers to losers in just the way that the
Kaldor-Hicks test hypothesises (recall that the application of
the test does not require the actual transfer of money).
Effectively, then, we could argue that while CBA itself
proceeds on the basis of the optimality of the existing in-
come distribution, government itself can ameliorate the
effects of any decision through a wholly separate income
transfer between affected parties. An additional reason for
supporting this view is that cited by Mishan (1974) to the
effect that any weighting procedure will still be consistent
with making the rich richer and the poor poorer. For example,
if we selected, by some procedure, the weights
aA, aB, ac =0.7; and an = 1.0
in our numerical example, it will then be the case that record-
ed benefits will be 210 and costs 200. The project is still
worth while even though Dis the poorest. But if weighting is
consistent with making the rich richer and the poor poorer,
what is the point of the procedure?
The 'revisionist' answer to these objections would be along
the following lines. First, one of the reasons that the Kaldor-
Hicks test deals in hypothetical as opposed to actual com-
pensation is precisely because of the complexities of making
lump-sum transfers. Fiscal measures are influenced by many
factors and it is simply not the case that they are better suited
to 'correcting' the distribution of income than via the use of
project selection (Pearce and Nash, 1981, p. 30). Second, to
say that weighting procedures are consistent with continuing
to make the rich richer is to miss the point. It implies that
61
 Efficiency and Distributive Weights

the purpose of weighting is actually to improve the distribu-

tion of income regardless of the conventional measure of
efficiency. But this has never been the aim of the revisionist
schools of thought. For the weighting procedures used delib-
erately 'mix' efficiency and distributional features. In that
respect it is not at all surprising that we can construct examples
in which projects are selected but in which the distribution is
made worse. All this means is that the distributional weights
do not outbalance the efficiency effects. To make the argu-
ment crystal clear, if we were concerned never to make the
distribution of income worse, then we would always select
weights to ensure that this was the case. In our numerical
example it is simple to find the requisite weight to achieve
this. We calculate the ration of losses to benefits (i.e. 200/
300) to get the weight 0.666. Applying this to the benefits
we obtain the result that any weight less than 0.666 will
achieve the result that the project is rejected because of its
effects on D, the poorest. This is equivalent to adopting a
Rawls-type social criterion whereby we always seek to make
the poorest group better off (Rawls, 1972). But while this is
one social decision rule, it happens not to be one advocated
by the revisionists. We conclude that Mishan has misunder-
stood the purpose of the revisionist exercise, which is to
integrate efficiency and distribution.
Now, having established the function of weighting proce-
dures we need to consider Mishan's next objection, which is
that the values of aA, etc., are 'arbitrary'. Here, again, the ob-
jection is not sustainable. In the case of 'politically deter-
mined' weights, the procedure for deriving them is by refer-
ence to some social objective function, albeit laid down by a
political authority. Such a procedure is not arbitrary - it
derives directly from a given objective function. In the case
of the second group of revisionists, the charge of arbitrariness
is more telling, because we have not as yet determined any
rules for selecting the weights. Instead of 'several' results in
a value sensitivity test, then, we could have a great many.
Mishan's criticism should thus be treated as a caution against

62
 Conventional CBA vs 'revisionism'

proliferating such weights. We shall suggest below various

criteria for selecting weights. But it is worth noting that the
adoption of a set of weights equal to unity is itself arbitrary
unless the distribution of income is held to be optimal. This
observation holds regardless of whether lump-sum transfers
are possible. We argue, then, that it is conventional CBA
which is either arbitrary in its selection of weights, or that
conventional CBA implies the value judgement about the
optimality of the distribution of income.
Before proceeding to the derivation of weights, we may
consider a further objection by Mishan, namely that the
weights chosen are not independent of political authority.
Moreover, because of this, they will vary from one country
to another and from one year to another as political authori-
ties change within any one country. As Mishan ( 1982) states,
'what is presented as an economic calculation has, in fact, no
meaning or sanction independent of the will of the relevant
political authority'. If this is true at all, it is true only of the
'unique-weight' theorists among the revisionists. The objec-
tion does not hold for the 'sensitivity analysts'. In their case,
some of the weights may come from the political authority
but none need in fact do so. However, if they do come from
a political authority, it is surely no surprise that the weights
will vary with changes in that authority. After all, the implicit
prices placed on a great many things change when govern-
ments change. A government wedded to more 'law and order'
places a greater value, 'on behalf of society, on the rights to
life, limb and property than those who might prefer a lower
police profile in the interests of protecting the freedoms of
the individual. Effectively, the 'relative prices' of civil liberties
and individuals' assets change. Mishan's objection only really
holds in the sense that he sees an objection to the subservience
of the economist to political authority. Here we must surely
have sympathy with his viewpoint, and for this reason we
do not wholly subscribe to the unique-weighting school of
thought. In its defence, however, it is worth noting that this
school of thought has its main application in less developed
63
 Efficiency and Distributive Weights

countries where the economy often is very closely planned.

In such a circumstance it is the objective function of the
political authority that matters.

Deriving distributional weights

Now we can consider ways in which weights other than unity

might be obtained. First, consider the difference between
political and economic votes that was alluded to in Chapter 1.
There it was noted that a political vote is not (generally)
'weighted' by an intensity of preference, i.e. we have 'one
man, one vote'. Economic votes are weighted by income (and
wealth). Since the principle of 'one man, one vote' is deeply
embedded in the concept of democracy, one weighting pro-
cedure for CBA would be to adjust each economic vote so
that it accorded with what each person would vote if their
incomes were equalised. A crude procedure for achieving this
result is as follows. We set
y
a·=-
, y.
I

where a; is now the weight to be attached to the ith income

group, Y is the average income (of the nation, community,
etc.) and Y; is the income of the ith group. Table 5.1 sets out
a very simple example of such a weighting procedure. Assum-
ing each group has the same number of persons in it, we see
that the average income is 150. This explains the row showing
the values of a;.
We see from Table 5.1 that what would have been a sanc-
tioned project on the conventional CBA procedure is now
rejected on the 'revisionist' approach. We can generalise the
formula for the new approach as

NSB =:E a; (B;- C;)

i (5.2)

64
 Deriving distributional weights

Table 5.1

Group A B c D Net benefit

Y; 200 250 100 50

B; +100 +100 +100 -100 +200

a; 0.75 0.60 1.50 3.00

a;B; +75 +60 +150 -300 -15

where NSB refers to net social benefit, i refers to the ith

income group, and a; has the meaning given to it in equation
(5.1). In terms of the CBA decision rule incorporating time
and discounting, the new expression looks a little formidable,
and is

(5.3)

where r is the social discount rate, and could be either the

social time preference rate, or the social opportunity cost
rate.
The problem with using equation (5.1) is that a hypotheti-
cal change of income from Y; to Y would in fact be accom-
panied by a change in the individual's expenditure pattern,
determined by the income elasticity of demand for the goods
in question. To put it another way, equation (5.1) fulfils our
requirement of approximating each individual's 'income-
equalised' willingness to pay if, and only if, the income elas-
ticity of demand is unity. In a situation in which the elasticity
is not unity, the proper formula would be

(5.4)

65
 Efficiency and Distributive Weights

where b is now the income elasticity of demand. Equation

(5.1) then becomes a special case of equation (5.4) in which
b = 1. (For more detail see Nash, Pearce and Stanley, 1975,
and appendix 1 to this chapter.)
We now have a perfectly general formula for weighting
costs and benefits. But observe that it is based on the idea of
weights as means of approximating economic votes to the
'one man, one vote' principle. It has nothing to do with what
is or is not 'deserved' by each of the groups in question. If
we wish to adjust for this, either by way of illustration by
the analyst to the decision-maker, or through adoption of
political judgements about what is socially desirable, then we
can introduce a new formula, which would appear as

(5.5)
where v is now some judgemental weighting about 'deserving-
ness'. Indeed, the two procedures in equations (5.4) and (5.5)
could be combined to give
a!'=( Y;y)b+v
I 5.6)
Equation (5.6) is then equivalent to adjusting the 'equalised'
votes again for a factor reflecting deservingness, or need, or
merit, or whatever.

Problems with weighting procedures

So far we have seen that weights can be derived either using

criteria based on 'equalised' votes, or through explicit adoption
of politically determined valuations of one group compared
with others, or through a combination of both. With Mishan,
we would argue that the adoption of the 'v'-type weights
should not be undertaken unless there is good reason, or
should only be carried out to illustrate for decision-makers

66
 Problems with weighting procedures

what the 'sensitivity' of the results of a CBA will be to differ-

ing judgements. The adoption of income-elasticity weights
does not, however, rest on anything except an attempt to
adjust for the fact that economic votes reflect the existing
distribution of income. Hence 'b'-type weights are not
subject to Mishan's criticism about loss of independence by
the economist.
Given the above derivations, we can now consider other
criticisms made by Mishan. In the formulae given above the
value of b was used in order to 'equalise' votes. The basis for
this weighting procedure, then, is based on the value judge-
ment that market votes are unfair. The 'b' weighting approach
is thus ethically derived and the ethical criterion used may or
may not appeal, any more than the use of weights of unity in
conventional CBA may appeal. The value of 'v' could be
politically derived. We have considered Mishan's objections
to politically derived weights and we have argued they are
not wholly valid. But Mishan also objects to 'ethico' weights,
i.e. weights derived from some moral principle. Within this
category he includes weights which reflect the marginal
utility of income to the various individuals. Note that we
have not used this approach at all to derive equations (5.4)
and (5.5). But we could have done. The argument would be
that a benefit of £1 to a rich person is worth less to him than
£1 is to a pauper. If our objective is to maximise the aggre-
gate (cardinal) utility of society, then we would simply set
a; equal to the marginal utility of income of the ith individ-
ual, and so on. In effect, the weights will become the ratios
of the marginal utilities so long as we set one of the absolute
weights equal to unity. Appendix 2 to this chapter explains
how marginal utility weights would be obtained. The effect is
to obtain a set of weights where

(5.7)

where the weight is related to the average income level (but

67
 Efficiency and Distributive Weights

need not be) and e is the elasticity of the marginal utility of

income function.
Now, the objections to such a procedure are several. We
may note that there exists a dispute as to whether the mar-
ginal utility of income is observable or measurable at all. If it
is not, then there is little point in pursuing adjustments of
the kind given in equation (5.7). If it is measurable, then
what we have is to convert the CBA formula into one which
operates with cardinal - i.e. completely measurable - utility.
Mishan's objection to this is that such weights 'have no
social sanction whatever'. They are not ethical weights
because 'in order to qualify as ethical, the set of weights has
to be enshrined in the constitution or else conform with a
virtual constitution, one that effectively represents an ethical
consensus' (Mishan, 1982, p. 38). Mishan's argument is thus
that the conventional CBA approach, in which the weights
are set equal to unity, does have such an ethical consensus.
Mishan is surely right to observe that society offers little by
way of confirmation that maximising aggregate (cardinal)
utility is the objective of society. Equally, it is very hard to
believe that society approves of the existing distribution of
income and has no wish to change it. Mishan's objection is
thus equally valid against his own preferred approach. For
our purposes we need only observe that 'reasonable' (i.e.
likely to command consensus) procedures for weighting need
not involve the use of marginal utility of income weights at
all, as equations (5.4) and (5.5) show.

Other weighting approaches

We conclude that not only has no case been made against

explicit weighting, but that arguments about 'consensus' are
very likely to lead us to adopt procedures such as those given
in equations (5.4) and (5.5). Other weighting procedures have
been proposed. Weisbrod ( 1966) observes the weights implicit
in past government decisions. Assuming that governments

68
 Other weighting approaches

actually do make conscious trade-offs between efficiency

benefits and other objectives, detailed scrutiny of past de-
cisions may permit derivation of the trade-off ratios, i.e. the
weights which reflect social preferences concerning distri-
bution. This general approach has attracted the support of
Maass ( 1966) and empirical studies have been made by
Weisbrod (1966) and McGuire and Gam (1969). Weisbrod
analysed decisions affecting various income and race groups,
and his general results indicated a high relative weight of
+9.3 for low-income non-white families, with the remaining
weights being +2.2 for higher-income white families, -1.3 for
low-income white families and -2.0 for high-income non-
whites. If the results can be accepted at all, they indicate a
high propensity to favour low-income non-white groups, but
thereafter a peculiar ranking in which high-income white fam-
ilies are favoured above low-income white families.
There is clearly room for refinements to this type of
approach. At the very least, they indicate to the decision-
maker the weights implicit in past decisions, providing 'a
general check against absurdities' (McGuire and Gam, 1969).
The problems are that ex post distributional results may not
indicate ex ante government plans. Bonnen ( 1966) has shown
that programmes designed to aid low-income farmers in fact
produced the opposite effect. In this case the observation of
ex post weights should at least indicate to the decision-maker
that the welfare effects of policies are not precisely predict-
able. More important, they should assist the decision-maker
in becoming more conscious of distributional outcomes.
A variant on the implicit-weights approach involves the use
of marginal rates of taxation as weights (Krutilla and Eckstein,
1958). Since the marginal rate tends to rise as income rises, it
would seem that society has implicitly assigned lower weights
to gains to high-income groups than to low-income groups.
The analysis could be generalised to incorporate other
allowances; thus family allowances could be argued to re-
flect society's value weight for families with several children
whose effective income is not so high as that for families with

69
 Efficiency and Distributive Weights

the same nominal income but without children. The marginal

rate of tax can be converted into a surrogate for the marginal
utility of income. The relevant a; would simply be the in-
verses of the marginal tax rates. A rate of 25p in the£ would
mean a relative weight of 4. As income rises, and the marginal
rate rises to, say, SOp in the £, the wieght will fall to 2. The
problems with this approach are several. First, tax rates do
not reflect only society's set of value judgements concerning
equality; they also reflect past decisions of fiscal policy
designed to affect the overall level of income. Second, it
presumes that only some taxes reflect equity judgements.
But indirect taxes may also partly serve this purpose. A full
assessment of weights might therefore require an analysis of
the true incidence of the entire tax system. In general it would
be a stretch of the imagination to suppose that the tax sys-
tem reflected only equity judgements. In practive it reflects
a hotchpotch of influences, not all of which are consistent
with the equity argument. Additionally, using observed rates
of tax incidence to reflect planners' intentions suffers the
same problem noted earlier with Weisbrod's approach.

Conclusions

The approach presented in this chapter is what Mishan has

termed 'revisionist', i.e. it argues that there is no single
correct approach to CBA. Instead, there are several sets of
results depending not only on varying particular values of
things such as the discount rate (if this is not the subject of
consensus) but also on varying the judgements that may be
made about differing weights to be attached to costs and
benefits according to who benefits and suffers. The practical
problem with approaches which do not set weights equal to
unity is the additional information required. This, more than
appeals to the independence of the economist and the credi-
bility of cost-benefit analysts, is the main reason for not
pursuing explicitly weighted procedures. An actual example
70
 Two appendices

of a weighted CBA (which uses marginal utility of income

weights) is to be found in Pearce and Wise ( 1972), which cor-
rects a set of erroneous weighted results given earlier by
Nwaneri (1970).

Appendix 1: deriving an income-elasticity weighted cost-

benefit function

Given a demand curve with the form

Q = kYbpg (5.8)
then the compensating variation for the ith individual is

(5.9)

where P 1 and P 2 are the prices after and before the project in question
(the lower and upper limits of integration). Similarly, the compensating
variation at the average level of income will be

(5.10)

Hence,

CVy =
cvi
(!:)b
Y
(5 .11)

which is equation (5 .4) in the text of Chapter 5.

Appendix 2: deriving a marginal utility of income weighting

procedure for CBA

Let utility be related to income, i.e. U = U(Y), such that the marginal
utility of income function has a constant elasticity. The marginal

71
 Efficiency and Distributive Weights

utility of income function for individual i can then be written

l/., = dU =aY:-e (5.12)

I dY; I

~here -e is now the elasticity of the function. For the average income
Y, we shall therefore have
u' -
-=ay-e (5 .13)
y

and the relative weight for the ith individual would then be

!! =af-e (= y)-e (5 .14)

Y aYje Y;
Notationally (5.14) is similar to (5.11), but the income elasticity of
demand in ( 5.11) is replaced by the elasticity of the marginal utility of
income function.

72
 6

Risk and Uncertainty

In the basic cost-benefit formulae presented so far we have

assumed that the costs and benefits are known with certainty.
In practice, this will not be true, and we can distinguish two
situations. The first is in the context of risk in which we will
know the probability that. the benefit or cost will take on
particular values. For example, a cost in a given year may be
I 00, 200, 500 or even I ,000. To each of these values we can
attach a probability, say 0.2, 0.5, 0.2 and 0.1 respectively.
This means that we judge there is a 20 per cent chance of the
cost being 200, a 50 per cent chance of it being 200, and so
on. We thus defme a risk context as one in which the proba-
bility distribution is known. The second context is one in
which we do not know the probabilities attached to the sizes
of the costs or benefits, but in which we do know the values
that the costs and benefits could take. In this case we would
know, for example, that the cost in a given year is 100, 200,
500 or 1,000, but we do not know what probabilities to
attach to each of these possibilities. This context is defmed as
one of uncertainty. In the literature, risk and uncertainty are
often used interchangeably, but it is convenient to keep them

73
 Risk and Uncertainty

separate as often as we can. As we shall see, the techniques

for dealing with the two different contexts are themselves
different.

Risk

The information in our simple example above can be presented

diagrammatically in Figure 6.1. The vertical axis shows the
probability and the horizontal axis shows the size of the
benefit. (We choose to work with the example of benefit -
it must be recalled that costs are also likely to be probabilistic.)
We can then plot the various possibilities on the diagram as
shown. Now, it may well be that the points shown are all
the information we have. If so, we are said to possess a
discrete probability function. Or it may be that we know
more than this and that we are able to draw the continuous
probability distribution as illustrated by the dashed curve in
Figure 6.1. In practice, we usually do not know the whole
curve, though in some circumstances, such as flood damage
and pollution, we can often use a probability distribution
obtained from various 'simulations' of how floodwaters or
pollution are spread through a given system. These results
come from special mathematical and physical models. When
probabilities are obtained either from past experience of
similar projects or events, or when they come from models of
the kind described, they are called objective probabilities.
But it may often be necessary to rely on judgements about
the probabilities, and these are known as subjective probabili-
ties since they do not derive from objective data but from
assessments by the analyst or some expert. Obviously, quite
often, the probabilities will be a mixture of both subjective
and objective elements. For our purposes we shall not make
the distinction between subjective and objective probabilities
again since our treatment of probability will be the same
regardless of which type of probability we are dealing with.
Now, we obviously cannot leave the information in Figure

74
 Risk

Probability
(p) 0.5 /~
I \
l \
0.4 \
\
I \
I \
0.3
I \
I \
l \
0.2
.._
I
'
I '
0.1

I
I
I
--·
I
0 100 200 300 400 500 600 700 BOO 900 1,000 Benefit (8)

Figure 6.1

6.1 as it is. We need to compress the information as best we

. can into a single indicator which we can then put into our
cost-benefit formula. One fairly obvious way of doing this
presents itself immediately. This is to take some sort of
average from the distribution. In our example the average
could be taken as
(0.2 X 100) + (0.5 X 200) + (0.2 X 500)
+ (0.1 X 1,000) = 320
The figure that we would put into the CBA is then 320 and
this is the mean or expected value. Note that the mean or
expected value is not the same thing as the 'most likely' value
of the benefit. Clearly the value that has the highest probability
of occurring is 200, with a probability of0.5. But this is not
the expected value. (The value of 200 is the mode of the dis-
75
 Risk and Uncertainty

tribution.) Care needs to be taken when looking at actual

cost-benefit studies, since there is widespread use of the
term 'best estimate' for a given benefit or cost, and a 'best
estimate' may be the mean, the mode or even the median
value.
Now, the problem with using the expected value as the
figure to put in our CBA is that it does not reflect the atti-
tude that the public investment agency concerned takes to-
wards risk. To see this consider Figure 6.2, which shows two
distributions, A and B, each with the same expected value.
The distributions shown are normal distributions. The reason
for this is that normal distributions can be described com-
pletely by their expected values and a measure of their 'spread'

BAM IN Expected value (EV) Benefits

Figure 6.2

76
 Risk

or variance. Non-normal distributions require furthermeasures,

particularly of their 'skewness'. As we shall see, there are
problems enough dealing with normal distributions. For this
reason studies which do contain probability functions usually
attempt to approximate them with normal distributions.
It would be surprising if society were indifferent between
the two distributions in Figure 6.2 even though they both
have the same expected values. For example, distribution A
has as its lowest possible value of benefits BA MIN , whereas
distribution B has a lowest possible value of zero. We would
expect society to take account of the greater variance in
distribution B. Such a reaction is said to be one of risk-
aversion. If society were indifferent between the two distri-
butions, i.e. it reacts only to the expected values, then it is
risk-neutral. We omit consideration of a situation in which
there is a positive attitude to large gambles, known as
risk-loving.
Since the expected-value approach does not capture risk-
aversion, and since we assume society is generally risk-averse,
we can approach the problem by saying that the utility which
society attaches to the various outcomes is the relevant entity,
not the size of the outcome on its own. This is sufficient to
enable us to restate the problem. Whereas the expected-value
approach would have given us

the expected utility can be written

EU = p;U(Bl) + P2 U(B2) + ... + Pn U(Bn)

=L: p;U(B;) (6.2)

But we can proceed no further along this line of reasoning

unless we know something about the way in which utility

77
 Risk and Uncertainty

and benefits are related -i.e. something about the utility-

benefit function.
Typically, such a function is assumed to take on the shape
shown in Figure 6.3, i.e. it exhibits diminishing marginal
utility as B increases. We can use this function to derive a
value of B which we can use in our CBA. We proceed as
follows.
Let B1 and B2 be the only possible outcomes of our pro-
ject. Imagine that B1 is certain. In that case PI = 1, and P2 =
0. The value of EU will be given by
1 x U(Bt) + 0 x U(B2) = U(Bt)
which is shown on the vertical axis. Similarly, if B 2 were
certain, the same analysis would give us the point U(B2 ).
u

U(B)

U(B) t - - - - - - - - - - - - - - - - . 1 "

EU (B) t - - - - - - - - - - - - - . r - - - . r l
1//
/I
/ I
// I
/
/
/
/
/
/
U(Bl) - - _J< V

0 8* 8 8

Figure 6.3

78
 Risk

This exercise enables us to fix the end-points of the line XY

in Figure 6.3. We shall soon see the use to which this con-
struction can be put. Now relax the assumption that B 1 or
B2 is certain and return to the fact that our project yields
either B1 or B2, each with a probability greater than zero and
less than 1. Select some arbitrary values for these probabilities,
say 0.4 for B1 and 0.6 for B2. Then the expected value of
the outcome will be 0.4(BI) + 0.6(B2) and this will give us a
point .ii in Figure 6.3. (To find such a point simply set the
ratio B1B/BB2 equal to ptfp2.) Now, the utility that would
correspond to B if it occurred with certainty would be U(B),
which is simply obtained by reading off the function U(B).
But the expected utility will be less because .ii is itself an
expected value of the eventsB1 andB2. In fact, given equation
(6.2), the value of EU is obtained by reference to the straight
line XY, since equation (6.2) is in fact a linear function. In
fact XY has the equation

EU = p 1U(Bl) + P2 U(B2)

Now, another way of looking at EU(B) is to observe that the

same level of utility could be obtained from the certain
occurrence of benefit B*.
Armed with this analysis, we are now in a position to
define some important concepts. We observe that U(Ji) is
greater than EU(B), indicating that the utility from a benefit
that occurs with certainty is greater than the utility that
comes from an expected value of a benefit. Moreover, we
can measure the cost of bearing the risk of the situation in
which B 1 or B2 occurs with degrees of probability. It is given
by the distance .ii - B*, for B* corresponds to the certain
benefit which gives rise to the same utility as the 'gamble'
(Bl and B2 ). We therefore define:

1. B- B* = the 'cost of risk-bearing'

2. B* = the 'certainty equivalent' of the gamble (B1, B2)
which has the expected value B.

79
 Risk and Uncertainty

The relevance for CBA is that we need to find the certainty-

equivalent value of any probabilistic outcome. It is this
certainty-equivalent value that we enter into the CBA. In-
cidentally, this is the same as inserting the expected value
of the benefit and deducting the 'cost of risk-bearing'. To see
this, simply deduct the cost of risk-bearing from the expected
value of the benefit in the above analysis and we obtain
jj- (B- B*) =B*
In our generalised cost-benefit formulae, then, we need to
remember to insert, at all times, the certainty-equivalent
values of costs and benefits. To bring the analysis back to our
observations about means and variances, Figure 6.4 indicates
essentially what is happening. It shows the expected value of
benefits on the horizontal axis and the variance on the vertical
axis. In order to accept higher and higher levels of variance,
we find that society wants a very much greater value of B.
The 'indifference curve' between variance and expected value
thus has the shape shown. But there will be a point on the
benefits axis, B*, which corresponds to a combination of
zero variance and benefits B*. Since the variance is zero, the
· magnitude B* is certain. In other words, B* is the certainty
equivalent of all the combinations on curve i 1 . There will be
a family of such indifference curves. Thus io shows one that
is worse than i1: for any level of B we see that a higher vari-
ance has to be tolerated compared with points on i 1 . Hence
io is a lower indifference curve than i 1 ; and so on.
While we have a satisfactory analytical basis for dealing
with; risk in the CBA context, it will be observed that finding
certainty-equivalent values of benefits and costs involves us
in knowledge of society's utility function with respect to
those benefits and costs. In practice, we have no procedure
for observing these functions. Thus, while we know how we
should take account of risk in an 'ideal' world, we are not
really any more advanced in terms of practical measures for
integrating risk into CBA.

80
 The A"ow-Lind theorem

0 8

Figure 6.4

The Arrow-Lind theorem

We observed that if society could be regarded as risk-neutral,

we could operate with the expected value of probabilistic
benefit or cost estimates. If society is risk-averse, however,
we require an estimate of the certainty-equivalent value of
the cost or benefit. Since the latter procedure requires
further estimation of a function relating money benefits (and
costs) to social utility, however, it presents serious, if not
insurmountable, difficulties for practical CBA. Clearly, it
would be very convenient ifCBA could assume risk-neutrality,
for then we would have no need for an estimate of risk-
bearing costs; we could use expected values of benefits and

81
 Risk and Uncertainty

costs. One argument has been advanced which suggests

exactly this, and it is known as the Arrow-Lind theorem
(Arrow and Lind, 1970). What this theorem argues is that
the larger the group of individuals across whom the risk is
spread, the lower is the risk per head. This seems intuitively
self-evident but a formal proof is complex. For what it
involves is demonstrating that the risk of the project in
question is itself reduced as the number of persons is in-
creased. The Arrow-Lind theorem tells us that as the number
of individuals approaches infinity, so the project risk ap-
proaches zero.
We offer no (formal or informal) proof here. A diagram-
matic approach to the theorem is given in Pearce and Nash
(1981). We satisfy ourselves with making a few observations
on the result of the theorem. First, Arrow and Lind were
concerned with the financial risk of projects in the public
sector. For any nation the population across whom this risk
is 'pooled' will be finite. Indeed, it should really be thought
of as the taxpaying community, so that it might be 20
million in the United Kingdom and 80 million in the USA.
These are large numbers, but obviously do not approach
infinity! Thus the risk remains positive and not negligible.
Second, while financial risk can be pooled, many other
risks that we need to deal with in CBA are not capable of
being pooled even across the taxpaying population. This is
because they are borne by groups of individuals who suffer
the externality caused by the project in question. This may
take the form of exposure to radiation, noise nuisance, air
pollution, visual intrusion, congestion, and so on. The rele-
vant group is thus the group of sufferers, and this will in-
variably be smaller than the nation's taxpayers. Hence the
pooling argument appears to have even less force. One way
of restoring some foundation to the argument is to point
out, however, that compensation for such externalities could
be borne by taxpayers. Thus the pooling of risk would again
be restored.
The third qualification to the Arrow-Lind theorem was

82
 Uncertainty

pointed out by Fisher (1974). If the risk in question shows

up in the form of a public 'bad'- i.e. an externality which
is consumed equally by everyone, and where an increase in
the 'consumption' of the externality by any one individual
does nothing to reduce the consumption by others - then,
however large the population, the risk is in no way reduced
because it is constant across all individuals. Expanding the
numbers of those exposed to the risk will not reduce the
overall risk of the project, as the Arrow-Lind theorem would
lead us to expect. Hence the theorem is applicable only to
private goods and not to public goods or public bads (such
as noise nuisance and the various examples given previously).

Uncertainty

While it is obvious that extensive thought has been given by

economists to the analysis of risk, the preceding sections
suggest that, in practical terms, the analytical procedures
offer little by way of assistance. For this reason we can expect
to find actual cost-benefit studies using simpler and cruder
procedures. However, we need to look at the second major
context in which benefits and costs are not known with
certainty. This is the context of uncertainty in which the
various possibilities are known but the probabilities of their
occurrence are not known. Various rules have been advanced
for dealing with these situations, and these rules emanate
from decision theory. It is important to note that the rules
do not allow us to 'collapse' variable estimates of costs and
benefits into one indicator in the way that our certainty-
equivalence analysis did. Because we have no probabilities
to work with, we shall find that decision theory rules only
really tell us what to do when we have set out the various
net benefits under different assumptions about what will
occur. That is, given various estimates of, say, benefits, we
can estimate the final net benefit total that would accrue
under each different assumption about individual benefit
83
 Risk and Uncertainty

or cost estimates. As we shall see, many CBAs do this anyway,

simply indicating what would happen to net social benefits
if certain values are assumed for the uncertain items. Decision
theory approaches, however, begin with this presentation.
Typically, it is presented as apay-offmatrix. Table 6.1 shows
such a matrix. Reading across, we have one source of the
uncertainty, say the expected rate of economic growth. This
will affect our benefit estimates in, say, a road project, by
affecting our expectations about road traffic demand. The
numbers I to 4 could then be percentage changes in real
national income. Reading down Table 6.1 we have a second
source of uncertainty which is some policy over which govern-
ment has control. Notice that we do not have two 'uncon-
trollable' uncertainties. We have one which is not subject
to control (the growth rate in our example) and one that is,
which we refer to as 'policy'. We consider later what can
be done when there are various sets of uncontrollable un-
certainties.

Table 6.1

Growth
Policy 2 3 4
I 0 3 7 16
2 4 4 4 5
3 0 0 3 3
4 6 10 5 3

The main body of the matrix in Table 6.1 then shows the
resulting net benefit figures. The net benefits do not simply
increase with economic growth.
We may now consider the various rules that might be used.

84
 Uncertainty

These are:

1. The 'maxim ax' criterion

This is based on a very optimistic outlook. The aim is simply

to go for the policy that maximises the net benefits (the 'pay-
off). In Table 6.1 we see that this is policy 1 because that
has the potential for yielding net benefits of 16 units pro-
vided economic growth occurs at 4 per cent p.a. Note that
it is also a very dangerous criterion because 1 per cent econ-
omic growth could occur and then net benefits will be zero.
Since we cannot say what the probabilities are for economic
growth we have no way of assessing the degree of danger in
this approach, but clearly it is one that would only be under-
taken by a confirmed gambler.

2. The 'maximin' criterion

In contrast to maxim ax, the maximin criterion is very cautious.

What it does is to look at the minimum pay-offs under each
strategy, and then choose the largest of these (i.e. it maxi-
mises the minimum pay-offs). In Table 6.1 we see that this
would lead us to ring (mentally) the numbers 0 for policy 1,
4 for policy 2, 0 for policy 3, and 3 for policy 4. We then
choose the maximum of these, which is 4 for policy 2, and
hence we choose policy 2. In the case of the figures in Table
6.1, policy 2 could be argued to have some overall appeal
because it does not have any zero pay-offs in it, and it is
clearly superior to policy 3 under all circumstances. But in
being cautious, policy 2 also missed the chance to go for the
significantly larger benefits under policies 1 and 4.

3. Index of pessimism

Under this approach, we take the best and worst outcomes

85
 Risk and Uncertainty

for each policy. We then apply an 'index of pessimism' to

the worst outcomes, which is in effect a subjective probability
weighted by our feelings of cautions. In Table 6.1 we would
select the following:
Policy 1 = 0,16
Policy 2 = 4,5
Policy 3 = 0,3
Policy 4 = 3,10

Suppose we set the index of pessimism at 0.9 and hence the

index of 'optimism' at 0.1. Then, the calculations are as
follows:
Policy 1 (0.9 x 0) + (0.1 x 16) = 1.6
Policy2 (0.9x4)+(0.lx5) =4.1
Policy 3 (0.9 x 0) + (0.1 x 3) = 0.3
Policy 4 (0.9 x 3) + (0.1 x 10) = 3.7
and the policy with the highest pay-off, i.e. policy 2, is chosen.
The obvious problem with this approach is in setting the
'index of pessimism'. If it is set equal to unity, for example,
it would be equivalent to selecting all the worst outcomes
and would therefore be formally equivalent to the maximin
criterion. If values different from unit are being used, one
must question why this is any less complex than assigning
subjective probabilities.

4. Laplace criterion

If we do not know the probabilities of any of the growth

rates occurring, is this not the same as saying that each one
is equally likely? This is what the so-called 'principle of
insufficient reason' suggests, and what the criterion based
on it, the Laplace criterion, would then suggest is that we
simply assign a probability of 0.25 to each of the four

86
 Uncertainty

growth rates. This would give weighted net benefits of

Policy I = 6.50
Policy 2 = 4.25
Policy 3 = 1.50
Policy 4 = 6.00
and policy 1 would be selected as having the highest pay-off.
The criterion is misplaced, however, because if we do not know
the probabilities we simply do not know them, and we cannot
then deduce from a state of total ignorance something about
the probabilities of events occurring!

5. Minimax regret

Again taking a cautious line we could look and see just what
the cost of making a wrong choice will be. To do this we set
up a 'regret' matrix, and the one corresponding to Table 6.1
is shown in Table 6.2

Table 6.2

Growth
Policy 1 2 3 4
1 6 7 0 0
2 2 6 3 11
3 6 10 4 13
4 0 0 2 13

To obtain the regret matrix, we read down the columns of

Table 6.1. Suppose the growth rate turns out to be 1 per cent
and we chose policy 1. The pay-off is 0. But had we chosen
policy 4 with a growth rate of 1 per cent the pay-off would

87
 Risk and Uncertainty

have been 6. Hence the 'regret' associated with the wrong

choice of policy is 6. Similarly, had we chosen policy 2
the regret would have been 6 - 2 = 4. That is, the regret is
measured as the difference between the highest pay-off, given
that a particular uncontrollable event occurs, and the pay-off
that actually occurs. In this way we build up the regret matrix
in Table 6.2. We then identify the maximum regrets and we
see that these are, for the four policies in sequence, 7, 11, 13
and 13. We see the minimum of the maximum regrets ('making
the best of a bad job') and thus choose policy 1.
Does this analysis help us? We have seen that different poli-
cies are chosen under different rules. Policy 1 was chosen
under the maximax and the minimax regret rule. Policy 2
was chosen under the maximin and index of pessimism ap-
proach. We argued that the Laplace rule was inadmissible, but
if it is used it would also choose policy 1. That different
results are obtained from different rules is not surprising,
however, since each embodies different attitudes to un-
certainty. If society is risk-averse, we would not expect it
to embrace the maximax criterion, leaving us with maximim
and minimax regret. Coincidentally, in our example, mini-
max regret yields the same answer as maximax.

Other approaches to risk and uncertainty

The overriding impression we are left with is that there is

no very satisfactory way of treating either risk or uncer-
tainty in CBA. Typically, in practice, the adjustments made
are of a much cruder nature than those we have suggested.
But, equally, such adjustments are understandable given the
limited usefulness of the techniques so far described. The
main approach used for uncertainty is to make sure that
ranges of estimates are given. In essence this reduces to
presenting the results of a CBA in the form of the pay-off
matrix similar to Table 6.1. However, instead of there being
one controllable and one uncontrollable variable, we have

88
 Other approaches to risk and uncertainty

two uncontrollable ones. For example, we might wish to test

for the variation that will result if we have a range of estimates
for the discount rate and a range of estimates for one item of
benefit. Then a table like Table 6.1 can be constructed and
the pay-offs can be shown. Obviously, uncertainty can extend
to many items, in which case it becomes essential to look at
what would happen with all combinations of assumptions.
This approach is known as sensitivity analysis and is usually
accompanied by the analyst's own 'best-guess' estimates.
A second approach is to apply a 'risk premium' to the dis-
count rate. We observed that the discount rate has the effect
of reducing the value of benefits and costs in terms of their
present values, and the effect is greater the further into the
future those costs and benefits occur. Since the uncertainty
surrounding estimates of costs and benefits is often itself
a function of time, it seems reasonable to make the discount
rate serve two separate functions, the first of reflecting social
time preference, and the second to reduce future money flows
to reflect their uncertainty. In fact the exercise earlier in
the chapter for calculating the cost of risk bearing can be
adjusted to produce an estimate of the requisite risk premium.
The procedure is outlined in Pearce and Nash (1981). How-
ever, even if we knew what addition to make to the social
discount rate, there are some features of risk premia that
make them unattractive. First, if the discount rate is, say,
5 per cent and we decide to add 2 per cent for risk purposes,
we have effectively imposed a particular functional form of
risk. We are saying that it has the form
p = e-0.02t
where p is the premium and 2 per cent is the premium. The
problem is that, set up in this way, we have imposed a time
path on risk which may not be appropriate for the problem
in hand. Second, if the risk in question relates to costs, we
would wish to add to the cost element, not subtract from it.
By adding a premium to the discount rate, however, we will
be reducing costs simply because they are uncertain in size.
89
 7

A Case Study: The

Gordon-below-Franklin Dam

Chapters 1-6 have set out in outline form the main features
of CBA. This chapter shows how some of those principles are
applied in practice and the case study selected is the develop-
ment of the Gordon River system in Tasmania, Australia, for
the purposes of hydro-electric power. The project not only
contains many fascinating issues in the application of CBA, it
has also been the subject of a substantial protest from Aus-
tralians and from the scientific and environmental community
worldwide. The development would destroy an area of wilder-
ness and some areas of outstanding scientific interest, including
some containing evidence of early aboriginal settlement which,
it is claimed, has significant archeological and anthropological
interest. On the other hand, nearly all of Tasmania's electricity
comes from hydro-power, and, at a time of high unemploy-
ment, the project offers the prospect of jobs. We thus have
the classic 'trade-off between the environment and direct
economic gains.
90
 The benefits

The project

Figure 7.1 indicates in the broadest outline the nature of the

scheme. The Franklin River flows into the Gordon River,
which then flows into Macquarie Harbour. The section of the
Gordon between the confluence of the Gordon and the
Franklin is the 'Gordon-below-Franklin' River and the aim
would be to build a dam and hydro-electric power station at
the point marked 'PS' on the outline map. By so doing, the
river waters would back up behind the dam and there would
be flooding in the Gordon River back up to the existing
Gordon Dam (shown as 'GD' in Figure 7.1). The Franklin
and Olga Rivers would also be flooded for considerable
lengths. Part two of the scheme would be to build a dam on
the King River ('KD') to divert waters into the Franklin River
system, and from there to the Gordon system. The additional
effect would then be to flood other parts of the Franklin not
affected by part one of the scheme. The map shows, very
approximately, the areas affected by flooding (shown as the
shaded area) and, of course, the flooded areas then serve as
storage so that the power station can be controlled in terms
of its throughput of water. Some of the areas shown already
contain water storage from an earlier development, the
Gordon River Power Development Stage One, which com-
menced full operation in 1978. The estimated output of the
two new developments is 172 megawatts (one megawatt=
1,000 kilowatts) for the power station on its own, plus a
further 168 megawatts if the King River diversion is included
(Hydroelectric Commission of Tasmania, 1979). The scheme
or schemes would be promoted by the Hydroelectric Commis-
sion of Tasmania, which has responsibility for generating and
distributing Tasmania's electricity supply.

The benefits

The obvious benefit of the Gordon scheme is the output of

electricity. As we saw, the proper way of approaching an
91
A Case Study: The Gordon-below-Franklin Dam •

•
Queenstown

15km ~

Figure 7.1

92
 The benefits

assessment of the benefits of any scheme is to estimate the

consumer surplus involved. Now, in this case, such an estimate
makes no particular sense unless we know what the alternative
is to the hydro-electric schemes. The Hydroelectric Commis-
sion (HEC) considered alternative configurations of control-
ling the river systems, various 'alternative energy' approaches,
such as tidal energy, and a conventional coal-fired power
station of about 400 megawatt size, using imported coal from
New South Wales. Last, the option of constructing a cable
across the Bass Strait which separates Tasmania from Victoria
was also considered, this cable carrying the necessary electric-
ity from the mainland.
We consider the options in terms of a choice between the
'preferred' scheme, as described previously, and the construc-
tion of a coal-fired power station. HEC's argument was that
the hydro scheme could continue to supply electricity at the
same price as that which prevailed prior to the scheme, while
adoption of the coal-fired power station would mean a higher
charge for the electricity from that source. In short, coal-fired
electricity was, it argued, more expensive than hydro-electric-
ity. While some questions remain about the data sources used
by HEC to come up with this general result, we shall accept
it, along with other commentators on the proposal. However,
HEC's analysis of the problem from here on must be replaced
with that of independent economists simply because HEC's
calculation of the extra cost borne by consumers has no
foundation in economic analysis. We therefore follow Saddler,
Bennett, Reynolds and Smith ( 1980) in their approach. This
can be contrasted with the calculations of HEC.
Figure 7.2 shows the situation in respect of the demand
for electricity. D1990 shows the demand curve in 1990 and
this is seen to exhaust the capacity of the electricity system
at price P1, which we take to be the marginal cost of supply-
ing electricity. If demand rises, then, we need to expand
capacity and that is the point of the Gordon-below-Franklin
scheme (or its alternative). Suppose the demand rises to
D2ooo by the year 2000 and that extra capacity is installed.
If HEC is correct, it could supply the extra demand at the

93
 A Case Study: The Gordon-below-Franklin Dam

Price

8 1990 capacity

Quantity

Figure 7.2

same price so long as it is allowed to invest in the hydro

scheme. Hence the quantity of electricity supplied would
change from Q1 to Q2. Now, if we measure consumer surplus
we see that, on the HEC analysis, it would increase from
BWP1 to AXP1 . If the hydro scheme is implemented and
HEC's analysis is correct, surplus in 2000 would be AXP1 .
Now suppose that the hydro scheme is rejected and the coal-
fired plant is built. This has higher costs, given by P2, assumed
equal to the marginal cost of producing electricity from coal.
Demand would now be Q3 and the system would not be at
capacity if the coal-fired station has the same capacity as the

94
 The benefits

hydro project (Ql Q2 ). At P2 and Q3 consumer surplus is

now A YP 2 , which is less than that under the hydro scheme
when, it will be recalled, it would be AXP1 in 2000. By
adopting the more expensive scheme, then, consumers would
apparently 'lose' P2 YXP1. But, as Saddler et al. note, the
amount P 2ZWP 1 is actually a transfer of surplus from con-
sumers to the HEC. The HEC has effectively 'appropriated'
some of the surplus through the higher charge, P2, which
now covers intra-marginal units (OQ1) as well as marginal
units of electricity. Since HEC has gained, there is no net loss
to society with respect to the amount P2ZWP1.
The consumer surplus loss from the more expensive option
is thus ZYXW. This is the loss that Saddler et al. calculate.
They note, however, that HEC's estimate of the cost to con-
sumers is given by ZCXW, i.e. by the extra units of electricity
that would be supplied if price were kept down to P1 (Ql Q2)
multiplied by the change in price (P1P2 ). Clearly, this pro-
cedure exaggerates the consumer surplus loss by an amount
YCX. The HEC failed to take into account the effect of the
higher price of electricity on demand. Notice, of course, that
if the change in price from P 1 to P2 is very small, HEC's
procedure would approximate the true loss.
The correct procedure is now to estimate the true loss of
consumer surplus for each year from 1990 to the relevant
'time horizon', discount it back to a base year and sum the
results to obtain the present value of the consumer surplus
loss from the higher-cost electricity option. This, of course, is
then formally equivalent to the consumer surplus gain from
choosing the lower-cost hydro option. Obviously, the loss in,
say, 1996, will not be the area shown in Figure 7. 2 for the
year 2000. To calculate the loss for each year therefore
requires a projection of the demand for electricity so that
we can estimate the demand curve for each year. This is what
Saddler et al. do and the results are shown in Table 7 .1.
Notice that two discount rates have been used, 5 per cent
and 10 per cent. It is worth noting that assumptions about
the price charged for electricity are not independent of the

95
 A Case Study: The Gordon-below-Franklin Dam

discount rate because the rate of return on capital is an integral

part of the long-run marginal cost. Thus use of a higher dis-
count rate not only reduces consumer surplus, because of the
normal effects of a discount rate, but also because it raises
the price P2 in Figure 7 .2. (It also should raise the price P1.)

Table 7.1 Consumer surplus gain to hydro scheme over coal-

fired alternative ( 1980 Australian $ million)

Type of demand Saddler Saddler HEC

5% 10% 5%

General load 169.6 10.0 249.8

Industrial load 19.5 1.5 95.3

189.1 11.5 345.5

Source: Saddler et al. (1980, p. 57).

Table 7.1 also shows the overall demand broken down into
general (domestic, commercial) demand and industrial de-
mand. This separation is not particularly important for our
purposes, but exists because of the special importance of
selected energy-intensive industries in Tasmania. The use of
10 per cent for a discount rate also reflects the fact that this
is the rate recommended by the Australian Treasury Depart-
ment. By any standards it seems high, and with Saddler et al.
we prefer to focus on the results obtained using the 5 per cent
analysis. Last, we observe that a comparison of the 5 per cent
results by Saddler and those observed by HEC indicate the
substantial nature of HEC's exaggeration of the surplus loss
from adopting the coal-frred alternative. This arises because
the cost differences between the hydro scheme and the coal-
fired alternative are thought to be substantial.
96
 The costs

The costs

The capital and operating costs of the two alternatives were

estimated by both HEC and Saddler et al. For our purposes
we need not investigate the procedures for obtaining these
costs simply because the costs are already implicit in the
prices used to calculate consumer surplus. That is, the benefits
we are interested in have been measured net of the capital
and operating costs. This means that we need only concentrate
on the external costs in the form of loss of wilderness, etc.
If these exceed the surplus gain from the hydro project, the
hydro project is not worth while; and vice versa if they are
less. The relevant figure to bear in mind, then, is the $A 189.1
million in Table 7.1. We need to know if wilderness costs are
greater or less than this.
But there is an obvious problem in that we have no market
in wilderness. We could interview people travelling to the area,
though they are few and far between because of the nature of
the terrain. Indeed, the 'virtue' of wilderness is that it is not
visited by many people, otherwise it would cease to be wilder-
ness. Yet we have noted that CBA, like all economics, is
anthropocentric - it does not recognise values unless they are
the value of human beings. This might suggest that the case
against the hydro scheme in question is hopeless. There are
no people to express any valuations!
The reference earlier to the outcry against the Gordon-
below-Franklin scheme indicates a shortcoming of the view
that people only value things by making some use of them.
Scientists may well wish to visit particular sites of archeological
and anthropological interest. Canoists, trekkers, etc., may well
want to make use of the river systems as they are. Yet much
of the outcry has never had its source among people who
have either visited the area or who might even go there. It is
sufficient for people to express concern about the area and
their preference that it be left as it is for the analyst to identify
the fact that there is a welfare loss to such people. The kind
of valuation that emerges because individuals wish to retain
97
 A Case Study: The Gordon-below-Franklin Dam

an option to make use of a facility at some future date has

been termed option value. It has been demonstrated that this
option value will exceed the individual's expected consumer
surplus from the facility in question. Indeed, option value is
defined as the difference between an option price and the
expected consumer surplus. This form of value was identified
by Cicchetti and Freeman (1971), and a formal derivation of
its relationship to consumer surplus is given in Pearce and
Nash (1981). Other writers claim to have discovered other
valuations as well. These include existence value, the value
placed on the preservation of an asset regardless of wishing to
exercise any option to make use of the asset, and bequest
value, the value placed on an asset as something to be handed
on to children and future generations. One or two studies
claim to have measured the empirical magnitude of option,
existence and bequest value, though it seems fair to say that
substantial controversy surrounds the estimates obtained.
(For an empirical study relating to water quality see Greenley
et al., 1981.)
The above concepts are sufficient to indicate that the
absence of use of a wilderness area by no means implies that
the area is not valued. None the less, it seems evident that
exercises designed to 'reveal' these values will be complex.
None has been undertaken for the Gordon dam example.
Instead, Saddler et al. (1981) made use of a procedure
developed by Krutilla and Fisher (1975). The next section
offers a guide to the procedure developed by these authors
(and others: the literature in question is now substantial).

Irreversibility

The essential point about the Gordon River development is

that if the wilderness area is destroyed it cannot be reinstated.
The cost of the destruction is thus irreversible. What can
economics say about an irreversible loss? We could argue that
no decision should ever be taken that entails an irreversible

98
 I"eversibility

loss, but a moment's reflection will indicate just how stultify-

ing this would be as a decision rule. Instead, we need to
develop some mechanism for comparing costs and benefits.
Equally, we have no obvious way of valuing irreplaceable
assets such as species of wildlife, natural environments, and
so on. What we can do is measure the benefits of the proposed
development and at least ask the question whether the loss of
wilderness, etc., is 'worth' the benefits obtained. But we need
to formalise this simple approach, which, please note, makes
no direct attempt to place a money value on the wilderness
loss. It simply presents the choice to the decision-maker.
We now develop the basic ideas of the Krutilla and Fisher
approach. In doing so, we make use of a synthesis produced
by Porter ( 1982), which is not only an excellent survey of
the work of Krutilla and his colleagues, but actually extends
it in a number of interesting ways. We also use perpetuities
for expositional purposes, just as we did when discussing
discount rates in Chapter 4.
Consider a development project costing only $1. (We resort
to the $ notation since this reflects the American literature
which provides the basis for the approach that follows. It is
also convenient because the$ is also the currency of Australia.)
Let the development benefits from this project be $D per
annum for ever. Then we can immediately write the present
value of this project as

PV(D) = -1 + ~~ De-rtdt (7.1)

where r is the discount rate.

Equation (7 .1) reduces to
D
PV(D) = -1 +- (7.2)
r
Now, we need to compare the development benefits in (7.2)
with the costs of the development. But bearing in mind the
definition of cost as opportunity cost, it will be evident that

99
 A Case Study: The Gordon-below-Franklin Dam

the cost of the project is not simply the $1 expended on

capital and operating costs of the development. It must also
include the forgone benefits of the destruction of the environ-
ment as a natural asset. Let us call these benefits P per annum.
Then the present value of these preservation benefits will be
p
PV(P) =- (7.3)
r
and we shall have to write the net present value of the devel-
opment project as
D p
NPV(D) = -1 +- - - (7.4)
r r
For the development project to be admissible we require that
NPV(D) be greater than zero, i.e. on rearranging (7 4) we will
have
D-P
--> l;or(D-P)>r (7.5)
r
Now we can investigate the nature of the preservation benefits
in a little more detail. First, we can observe that, over time,
the relative price of P is likely to rise as the natural environ-
ment becomes less and less in quantity. Note that this is quite
different from talking about general price rises. Those, as we
saw earlier in the book, are not included in the CBA formula.
But if we have reason to believe that any benefit or cost is
likely to change its price significantly and relative to the
general price level, then we should include that price rise in
the analysis. This leads us to write the preservation benefit in
year t as
Pt =PoeKt (7.6)
where Po is now the initial year's preservation benefit and g
is the growth rate of the price of preservation benefits relative
to the general price level. With Krutilla and Fisher (197 5) we
could go further and argue that our development scheme will

100
 I"eversibility

itself be subject to technological change which will render it

less attractive through time. For example, hydro-electricity
may become less attractive if nuclear power advances as a
low-cost form of electricity. While this particular example,
taken from the work of Krutilla and his colleagues, may seem
suspect now, we use it only to demonstrate the general point
that we may wish to discount the development benefits by a
further factor, k, reflecting the rate of 'technological decay'
of our project. Hence we would write:
Dt =Doe-kt (7.7)
If we now bring (7.6) and (7.7) together with the formula for
NPV(D), we shall have

NPV(D) = -1 + ~: De-(r+k)tdt- i: Pe-(r-g)t dt (7.8)

which looks formidable, but reduces to

D p
NPV(D)=-1+-- - (7.9)
r+k r-g
(The observant reader may note that this will only be finite if
we haver greater than g.)
Porter (1982) shows that equation (7 .9) takes on positive
values if, and only if,
( 7.1 0)
which is a result we take on trust for our purposes. If inequal-
ity (7 .1 0) holds, then we find that the graph of present value
against the discount rate, r, appears as in Figure 7.3. Here we
see that net present value is positive only above a discount
rate ro and below a discount rate r1. In other words, the
development project will succeed only if certain discount
rates are adopted. High rates simply reduce the value of D
in the normal way that discount rates affect benefits. Low
rates tend to give the rate of growth g on the preservation
benefits a chance to influence the choice against the develop-

101
 A Case Study: The Gordon-below-Franklin Dam

ment. Porter then shows that even small values of k and g will
raise the required ratio of D to P if the development project
is to proceed. For example, let P be (arbitrarily) 0.2 and let
k+g=O.Ol. Then inequality (7.10) tells us that, for the
development to be worth while, D has to take a minimum
value given by the inequality
ViJ = v'Q.2 + VQ.Oi
or
D = (0.536) 2 = 0.287
But this means that the ratio of D to Pis 0.287/0.2 = 1.43.
That is, development benefits must be 43 per cent higher
than preservation benefits for the development to be worth
while. The result is therefore very sensitive to the important
introduction of the 'new' features of the CBA, namely k
and g.

Estimating required preservation benefits for the Gordon Dam

The preceding section provides the background to the calcula-

tions carried out by Saddler et al. ( 1980) in their assessment
of what the preservation benefits would have to be for the
Gordon River hydro project to be rejected. Saddler et al.
engage in a more complicated analysis than that presented in
the previous theoretical section. One reason for this is that
wilderness value is unlikely to grow at the rate g for ever
simply because we can expect actual users of the area to in-
crease. Hence the area itself will have some 'limit' to its use
set by congestion. Note that this is true only if we take ac-
count of user values. It is not true if what we believe we are
projecting are option and existence values. Saddler eta!. also
build in the 'technological decay' factor in a manner some-
what different from that indicated by Krutilla and Fisher
(1975) and Porter (1982). They argue that 'technological
change will simply permit an increased availability of manu-

102
 Estimating required preservation benefits for the Gordon Dam

Present value

-1 - -- - - - - - - -- - - - - - -- - - - - - - - -

Figure 7.3

factured goods and setvices from a given source base, while

the supply of the natural environment cannot be increased'
(p. 81 ). This argument they use partly to justify the relative
price increase, which we called g. While this is analogous to
the 'decay' concept in the Krutilla-Fisher model, it is not
quite the same, and indeed seems to suffer less from the
criticism that can be made of the Krutilla-Fisher position,
that the technological change affects the development project's
benefits regardless of what happens to presetvation benefits.
103
 A Case Study: The Gordon-below-Franklin Dam

The net effect is similar, however, in that the compound

growth factor for preservation benefits is augmented.
Various results are reported, depending on differing assump-
tions about the rate of growth of preservation benefits and
differing discount rates. Taking the discount rate to be 5 per
cent, the 'relative price effect' to be 4 per cent p.a. and the
capacity of the region to absorb visitors to be reached in thirty
years gives the present value of $1 of initial-year preservation
benefits as $259.8. Now, to find out how large preservation
benefits must be for the development project to be rejected,
the procedure is as follows. First, we have an estimate of the
consumer surplus gained by adopting the hydro project rather
than the coal-fired power station, and that is $189 million.
Second, we have the present value of $1 of preservation
benefits and this is $259.8. Third, if we divide the former
figure by the latter figure, we shall have obtained an estimate
of what preservation benefits would have to be for them to
equal the benefit of the hydro development. For, on division,
we obtain

189~~~~~000 = 727,483

What this means is that if the initial year's preservation benefits

are, say, $750,000, then the present value of preservation
benefits will be greater than the present value of the develop-
ment benefits. Notice that we do not need to express the
result in terms of benefits in each and every year. All that is
being said is that if the benefits are $0.75 million in the first
year, they will grow in the manner already described and will
be greater than the consumer surplus lost from the choice of
the more expensive coal-fired option.
This is what is meant when we said that preservation bene-
fits did not have to be estimated directly. For all that we
need is to present the results in this manner: do we believe
that the wilderness, wildlife and areas of special interest lost
because of the scheme have a value of $0.7 5 million now? If,

104
 Epilogue

for once, we step outside the cost-benefit analyst's shoes, it

is difficult, surely, to believe that the answer is anything other
than an emphatic 'yes'.

Epilogue

Since the Saddler et al. (1980) study, work has continued on

the Gordon River proposal. No mention here has been made
of employment, and this is a legitimate and worrying concern
of Tasmanians and their government. The proper procedure,
as we have seen, is to shadow price the labour that would be
used in the project. But most likely to outweigh all other
adjustments is the fact that the demand for electricity in
Tasmania has been falling, as with most industrialised coun-
tries. In these circumstances the position of the demand curve
D2ooo in Figure 7.2 has changed and is to the left of what we
have shown. But this means that the consumer surplus loss
from choosing the higher-cost option is also reduced. From
this it follows that the 'required' initial year's preservation
benefits are also less.
In March 1983 Bob Hawke was elected Prime Minister of
Australia. One of his first acts on taking office was to cancel
the Gordon-below-Franklin hydro-electricity project.

105
References

K. Arrow and R. Und (1970) 'Uncertainty and the evaluation of public

investment decisions', American Economic Review, June.
A. J. Ayer (1936) Language, Truth and Logic, Gollancz, London.
B. Barry (1982) 'Intergenerational justice in energy policy', in D.
MacLean and P. Brown (eds), Energy and the Future, Rowman &
Littlechild, Totowa, New Jersey.
W. Baumol (1968) 'On the social rate of discount', American Economic
Review, December.
M. Beesley et al. (1960) The London-Binningham Motorway- Traffic
and Economics, Road Research Laboratory Technical Paper No. 46,
Department of Scientific and Industrial Research, London.
J. T. Bonnen (1966) 'The distribution of benefits from cotton price
supports', inS. Chase ( ed.), Problems in Public Expenditure Analysis,
Brookings Institution, Washington.
C. Cicchetti and A.M. Freeman (1971) 'Option demand and consumer
surplus: further comment', Quarterly Journal of Economics, August.
Commission on the Third London Airport (1971) Report, HMSO,
London.
A. Dasgupta and D. W. Pearce (1972) Cost-Benefit Analysis: Theory
and Prt!ctice, Macmillan, London.
J. Dupuit (1844) 'On the measurement of the utility of public works',
Annales des Ponts et Chaussees, 2nd series, vol. 8, reprinted in English
in D. Munby ( ed.), Transport, Penguin, Harmondsworth, 1968.
0. Eckstein (1958) Water Resource Development, Harvard University
Press, Cambridge, Mass.
M. Feldstein (1972) 'The inadequacy of weighted discount rates', in
R. Layard (ed.), Cost-Benefit Analysis, Penguin, Harmondsworth.

106
 References

A. C. Fisher (1974) 'Environmental externalities and the Arrow-Lind

theorem', American Economic Review.
R. E. Goodin (1982) 'Discounting discounting' ,Journal ofPublic Policy,
vol. 2.
J. V. de Graaf (1957) Theoretical Welfare Economics, Cambridge
University Press, Cambridge.
D. Greenley et al. (1981) 'Option value: empirical evidence from a case
study of recreation and water quality', Quarterly Journal of Econ-
omics, vol. 4, November.
J. R. Hicks (1939) 'Foundations of welfare economics', Economic
Journal, December.
J. R. Hicks (1943) 'The four consumer's surpluses', Review of Econ-
omic Studies, vol. 11, Winter.
Hydroelectric Commission of Tasmania (1979) Report on the Gordon
River Power Development Stage Two, Hobart, Tasmania.
R. Just, D. Hueth and A. Schmitz (1982) Applied Welfare Economics
and Public Policy, Prentice-Hall, Englewood Cliffs, New Jersey.
N. Kaldor (1939) 'Welfare propositions of economic and interpersonal
comparisons of utility', Economic Journal, September.
K. H. Katouzian (1980) Ideology and Method in Economics, Macmillan,
London.
J. V. Krutilla (1981) 'Reflections of an applied welfare economist',
Journal of Environmental Economics and Management, vol. 8.
J. V. Krutilla and 0. Eckstein (19 58) Multiple Purpose River Develop-
ment, Johns Hopkins University Press, Baltimore.
J. V. Krutilla and A. C. Fisher (1975) The Economics of Natural Environ-
ments, Johns Hopkins University Press, Baltimore.
I. Little (1950) A Critique of Welfare Economics, Oxford University
Press, Oxford (2nd edn 1957).
I. Little and J. Mirrlees (1969) Manual of Industrial Project Analysis for
Developing Countries, OECD, Paris.
I. Little and J. Mirrlees (1974) Project Appraisal and Planning for
Developing Countries, Heinemann, London.
A. Maass (ed.) (1962) Design of Water Resource Systems, Macmillan,
New York.
A. Maass (1966) 'Benefit-cost analysis: its relevance to public invest-
ment decisions', Quarterly Journal of Economics, May.

107
 References

M. McGuire and H. Gam (1969) 'The integration of equity and effici-

ency criteria in public project selection', Economic Journal, Decem-
ber.
R. McKean (1958)Efficiency in Government through Systems Analysis,
Wiley, New York.
S. Marglin ( 1963) 'The opportunity costs of public investment', Quart-
erly Journal of Economics, May; and 'The social rate of discount and
the optimal rate of saving', Quarterly Journal ofEconomics, February.
S. Marglin (1967) Public Investment Criteria, Allen & Unwin, London.
S. Marglin, A. Sen and P. Dasgupta (1972) Guidelines for Project
Evaluation, UNIDO, Vienna.
E. J. Mishan (1974) 'Flexibility and consistency in project evaluation',
Economica, February.
E. J. Mishan (1975) Cost-Benefit Analysis, 2nd edn, Allen & Unwin,
London.
E. J. Mishan (1981) 'The nature of economic expertise reconsidered', in
E. J. Mishan,Economic Efficiency and Social Welfare, Allen & Unwin,
London.
E. J. Mishan (1982) 'The new controversy about the rationale of econ-
omic evaluation', Journal of Economic Issues, vol. 16, no. 1, March.
C. A. Nash, D. W. Pearce and J. Stanley (1975) 'An evaluation of cost
benefit analysis criteria', Scottish Journal of Political Economy,
June.
V. C. Nwaneri (1970) 'Equity in cost-benefit analysis: a case study of
the Third London Airport', Journal of Transport Economics and
Policy, September.
M. Olson and M. Bailey (1981) 'Positive time preference', Journal of
Political Economy, February.
T. Page (1977) Conservation and Economic Efficiency, Johns Hopkins
University Press, Baltimore.
A. Peacock (1973) 'Cost benefit analysis and the control of public
expenditure', in J. N. Wolfe (ed.), Cost Benefit and Cost Effective-
ness, Allen & Unwin, London.
D. W. Pearce (1983) 'Ethics, irreversibility, future generations and the
social rate of discount', International Journal of Environmental
Studies, April.

108
 References

D. W. Pearce and C. A. Nash (1981) The Social Appraisal of Projects,

Macmillan, London.
D. W. Pearce and J. Wise (1972) 'Equity in cost-benefit analysis: a
comment',Joumal of Transport Economics and Policy, September.
A. C. Pigou (1952) The Economics of Welfare, Macmillan, London.
P. Porter (1982) 'The new approach to wilderness preservation through
benefit-cost analysis', Journal of Environmental Economics and
Management, vol. 9.
F. Ramsey ( 1928) 'A mathematical theory of saving', Economic Journal,
vol. 38.
J. Rawls (1972) A Theory ofJustice, Oxford University Press, Oxford.
H. Saddler, J. Bennett, I. Reynolds and B. Smith (1980)Public Choice
in Tasmania, Australian National University Press, Canberra.
E. F. Schumacher (1973) Small is Beautiful, Blond, London.
T. Scitovsky (1941) 'A note on welfare propositions in economics',
Review of Economic Studies, November.
P. Self (1970) 'Nonsense on stilts: the futility of Roskill', New Society,
2 July.
P. Self (1975) Econocrats and the Policy Process, Macmillan, London.
A. K. Sen (1967) 'Isolation, assurance and the social rate of discount',
Quarterly Journal of Economics, February.
L. Squire and H. van der Tak (1975) Economic Analysis of Projects,
Johns Hopkins University Press, Baltimore.
UK Government (1967) Nationalised Industries: A Review of Economic
and Financial Objectives, Cmnd 3437, HMSO, London.
B. Weisbrod (1966) 'Income redistribution effects and benefit-cost
analysis', in S. Chase (ed.), Problems in Public Expenditure Analysis,
Brookings Institution, Washington.
R. D. Willig (1976) 'Consumer's surplus without apology', American
Economic Review, vol. 66, no. 4, September.

109
Index

Arrow, K. 45, 81,82 Dasgupta, A. 19

Arrow-Lind theorem 81-3 Dasgupta, P. 60
Ayer, A. J. 20 decision theory 83
and uncertainty 83-8
Bailey, M. 57 diminishing utility of income
Barry, B. 55 78-9
Baumol, W. 45 discounting
Beesley, M. 17 and capital productivity 38,
Bennett, J. 93 42
bequest value 98 defined 37-9
Bonnen, J. 69 and future generations 52-5
Budget Circular A-47 15 rationale for 37-43
and time preference 38
certainty equivalent 79-81 distributional weights 59-72
Cicchetti, C. 98 derivation of 64-6
compensated demand curve 28, and income elasticities 65-6,
29 71
compensating variation 29-30, and income utility elasticity
33 71-2
compensation test 16, 20, 22-4 objections to 60-4
consumer sovereignty 7 problems with 66-8
consumer surplus rationales for 60
and aggregation 28-9 Dupuit, J. 27
and benefits 25-7
defined 9 Eckstein, 0. 16, 69
and the Gordon Dam 94-5 Edgeworth box 22
measurement problems 27-30 environmental impact assessment
and path dependence 36 20
and price changes 2 5-7, equivalent variation 29-30
34-6 existence value 98
cost of risk-bearing 79 expected utility 77-9
cost-effectiveness analysis 15 expected value 7 5

110
 Index

experimental markets 11 . McGuire, M. 69

McKean, R. 16
Feldstein, M. 48, 50 marginal efficiency of capital 42
Fisher, A. C. 83, 98, 99, 100, marginal willingness to pay 27,
102, 103 32,33
Flood Control Act 1936 14-15 Marglin, S. 18, 48, 60
Freeman, A. 98 Marshallian demand curve 28, 29
maximax 85, 88
Garn,H. 69 maximin 85
Goodin, R. 55 minimax regret 87-8
Gordon River Dam 90-1, Mirrlees, J. 18, 60
94-6, 102-5 Mishan, E. J. 8, 30, 60, 61, 62,
Graaf, J. V. de 20, 21 63,66,67,68, 70
'Green Book' 1950 15 morality and CBA 3-4
Greenley, D. 98
Nash, C. A. 28, 48, 50, 60, 61,
Hicks, J. R. 16, 28, 30, 54, 61 66,82,89,98
Hydroelectric Commission of net present value 46-7
Tasmania 91, 93, 94,95 noise nuisance values 19
hypothetical markets 11 nuclear waste 52
Nwaneri, V. 71
index of pessimism 85-6
intergenerational compensation objective probability 74
fund 54 OECD Manual 18
internal rate of return 42 Olson, M. 57
irreversibility 98-9 opportunity cost 13, 16, 25,
and CBA 98-102 30-1
and the Gordon Dam I 02-5 and shadow price 32
and opportunity cost 100 option value 98

Just, R. 29, 30 Page, T. 55

Pareto improvement rule 16, 23
Kaldor, N. 16, 54,61 path dependence (of consumer
Kaldor-Hicks rule 16, 54, 61 surplus) 36
Katouzian, H. 8, 21 pay-off matrix 84
Krutilla, J. V. 16, 21, 69, 98, Peacock, A. 8
100, 102, 103 Pearce, D. W. 19, 28, 48, 50, 53,
60,61,66, 71,82,89,98
Laplace criterion 86-7, 88 Pigou, A. 58
life, valuation of 10 Porter,P. 99,101,102
Lind,R. 45,81,82 preferences 5, 7, 9
little, I. 18, 20, 60 and CBA 5, 6, 7, 9-13
and experts 7
Maass, A. 69 present value 39, 47

Ill
 Index

probability distribution 73-5 synthetic 48-50

programming and shadow prices and utility 55-8
33 social opportunity cost of capital
43
Ramsey, F. 58 social rate of time preference 43,
ranking(ofprojects) 51-2 44,55-8
rational choice, defined 2-3 society, defined 2, 16
Rawls, J. 62 Squire, L. 18
revisionism in CBA 60-4 Stanley, J. 60,66
Reynolds, I. 93 subjective probability 74
risk, defined 73-5 symmetry of substitution 36
risk-aversion 77
risk-bearing, cost of 79 Tak, H. van der 19
risk-loving 77 Third London Airport 19
risk-neutrality 77 time horizon 4 7
risk-pooling 82 time preference 38, 42-3
risk premium 89
Roskill Commission 19 uncertainty, defined 73-4
and decision rules 83-8
Saddler, H. 93, 95, 96, 97, 98, UNIDO Guidelines 18, 60
102 user values 102
Schumacher,E. 1,2,4,5,6, utility 12, 16, 28
12 utility possibility curve 22-3
Scitovsky, T. 17, 20, 21, 24
Scitovsky paradox 17, 20 value judgements 4-9, 21
reversal test 22-4 and discounting 38
Self,P. 1,2,4,5,19 and moral appeal 7
Sen, A. 46, 60 variance 77
sensitivity analysis 63, 89
shadow price 32-3, 34 Weisbrod, B. 68, 69, 70
Smith, B. 93 welfare 12, 16
social discount rate 39, 40-3 welfare economics 20-1
and capital productivity 42 Willig, R. 28, 29, 30
and the Gordon Dam 95-6 willingness to accept 10
and income 55-8 willingness to pay 9, 25, 32
and inflation 40 Wise, J. 71

112