Contemporary issues

in economics
Proceedings of the conference of the
Association of University Teachers of Economics,
Warwick, 1973

edited by Michael Parkin and A. R. Nobay

Manchester University Press

13759892
 ey Rar eee aay

Contents

The contributors page viii

AUTE executive committee ix
© 1975 Association of University Teachers of Economics Editors’ introduction x

Published by Macro-economics
Manchester University Press
Oxford Road, Manchester M13 9PL
Franco Modighani
ISBN 0 7190 0580 9 The life cycle hypothesis of saving twenty years later Z

M. f. Artis, E. Kiernan and }. D. Whitley

The effects of building society behaviour on housing investment 37
Discussion by I. C. R. Byatt 60

Douglas Fisher
Wealth adjustment effects in a macro-economic model 63
Discussion by Marcus H. Miller 76

J. S. Flemming
Wage rigidity and employment adjustment: alternative
micro-foundations 80
Discussion by B. F. McCormick 87

D. I. MacKay and R. A. Hart

Wage inflation and the regional wage structure 88
Discussion by B. A. Corry 117

Fim Taylor
Wage inflation, unemployment and the organised pressure for
Printed in Grea Britain by higher wages in the United Kingdom, 1961-71 120
Adlard & Son Ltd., Bartholomew Press. Discussion by A. P. Thirlwall 139
 a TTT a

Contents Contents vii

Donald D. Hester Albert Fishlow

Inflation and the recent American happening 142 17 Income distribution and human capital: some further results
Discussion by Michael Parkin 167 for Brazil 354
Discussion by C. V. Downton 376
A. F. Preston
Optimal stabilisation policy with a lagged instrument 170 Homi Katrak
Discussion by S. G. B. Henry 191 18 An application of Lancaster’s consumer demand theory to
some recent hypotheses of international trade 380
Allocation and efficiency Discussion by 7. S. Metcalfe 394

Amartya Sen Howard C. Petith

The concept of efficiency 196 19 A new theory of international trade involving joint production 399
Discussions by Fohn Black and John Muellbauer 412
Geoffrey Heal
10 The depletion of exhaustible resources 211 Leonard Dudley and Roger Sandilands
Discussion by C. F. Bliss 222 20 When is foreign aid harmful? The problem of PL 480 417
Discussion by George C. Abbot 429
David }. Mayston
11 Optimal licensing in public sector tariff structures 226 Index of names 432
Discussion by S. C. Littlechild 250 Index of subjects 437

¥. Moreh
12 Investment in human capital and the income tax 252
Discussion by P. R. G. Layard 269

Adrian Ziderman
13 Cost-benefit analysis of Government Training Centres in
Scotland: a simulation approach 271
Discussion by Mark Blaug 288

Til Growth, development, capital and trade

Brinley Thomas
14 A plea for an ecological approach to economic growth 292

G. C. Harcourt
15 The Cambridge controversies: the afterglow 305
Discussion by A. H. Vanags 334

J. R. Sargent
16 The valuation ratio and the growth of the firm 337
 The concept of efficiency : 197

9 The concept of efficiency in both cases is conspicuous not only by what it

says but also by what it does not say. Two efficient points which are distinct
are non-comparable in terms of the dominance relation, i.e. one must be
higher in terms of some component and the other higher in terms of some
The concept of efficiency other component. The notion of efficiency stops short of comparing the gains
and losses in moving from one to the other, and it is this reticence that is
Amartya Sen supposed to account for the wide acceptability of efficiency as a standard of
achievement. Productive efficiency can be ascertained even without bringing
in relative prices of the different commodities, and economic efficiency can
be discussed even without inter-personal comparisons of welfare. The
philosophy underlying the concept of efficiency is to keep out of troubled
waters, and since optimal resource allocation and welfare economics are no
seas of tranquillity, efficiency has been a much-used concept in the literature.
It has also provided the conceptual framework for some of the important
theorems in this area, including the so-called ‘basic theorem of welfare
I
economics’, What I would like to examine is whether this emphasis on the
In the literature of resource allocation two distinct but related notions of concept of efficiency has been justified.
efficiency play important roles. The concept of ‘productive efficiency’
sometimes referred to as ‘technological efficiency’, is concerned with mee
i
modity production. In the set S' of attainable production possibilities with
given resources, a particular commodity bundle » is called efficient if there is I shall take up the question of productive efficiency first. I begin by making a
no other attainable bundle in S such that it produces more of at least one of brief reference to an important problem which I shall not pursue here. The
the commodities and no less of any of the others. Taking inputs as negative identification of the efficient set of production possibilities requires know-
outputs, this definition can be easily extended to variable resources as well ledge of technological possibilities; one has to know the menu first to be able
In contrast to productive efficiency, the term ‘economic efficiency’ is normall to tell which are the efficient elements in the menu. Some of the recent works
used to tefer to the welfare notion of Pareto optimality. Among the site on technology initiated by Kaldor [8] and others would seem to question the
able social states T, a particular social state y is considered to be economicall very notion of a given menu at any point of time. In this view, new techniques
efficient if there is no other attainable state such that in it at least one mond are invented only when there is an active search for such methods. Habakkuks’
would feel better off and everyone would feel at least as well off as 2 y study [7] of the impact of labour scarcity in the United States on the type of
Both these notions have a common analytical structure, viz. they identify technology evolved there provides an important illustration of this pheno-
the ‘maximal’ elements of a set over which a quasi-ordering possibly incom- menon. The idea of a given set of techniques defining a set of production
plete, has been defined in terms of the relation of dominance. Each element possibilities out of which the non-dominated elements are identified as effici-
of the set of production possibilities S is represented by a vector with m ent would seem to run counter to this important approach to technological
components representing the amounts of the m different commodities, and economics.
one production possibility is recognised to be dominant over another if it is While I do feel sympathetic to this approach, I think there is a danger in
larger in one component and no smaller in any component. Any element of taking too ‘dynamic’ a view of the problem of technology. Technology is
S which is not thus dominated by any other element in S is declared to be certainly created by looking for it, but there also exists at any point of time
efficient. Similarly, each social state in T' is represented by a vector with n a collection of already evolved techniques. The fact that the process of search
components representing the individual utility levels of the n different persons and even the process of setting up production structures may alter the stock
and one social state is dominant over another if it is larger in one component of available techniques does not necessarily go against basing production
[aS SE

and no smaller in any component. Any element of T not dominated by an decisions on the already known techniques; one can scarcely ‘choose’ tech-
other element is declared to be economically efficient. To avoid the roblem niques that are yet to emerge and may or may not ever emerge. The question
of utility representation it is possible to define economic efficiency direct is of some importance to development planning, since it has become increas-
in terms of individual preferences. Thus formulated, a state y is ete ingly fashionable to argue that the past menu of techniques is basically
efficient (Pareto-optimal) if, and only if, there is no other attainable pr ‘inappropriate’ and that the developing countries should rely on ‘making’
which someone prefers to y and everyone likes at least as much as ye their own technology by looking for it. This view, while full of healthy vigour,
 198
Amartya Sen
The concept of efficiency 199
is not always a very useful approach to take. : Tolsto y recalls
in hishi Childh
in J ood,
Boyhood and Youth that he had such a total fixatio that more work may be regarded as a cost. For the usefulness of the
n about matter being
created by human mind that he would drive himself concept of efficiency it is important that the frontier of desirability
insane looking rapidly
pas to one side hoping to catch a glimpse of the emptin should not be crossed, but there is in practice considerable traffic across the
ess before his mind
eead time to create matter on that side. There i S perhap
h s a moral inin thisthi story frontier.

IV
Ti
But suppose we abstract from this problem. Let each effect of an activity
The purity of the concept of productive efficiency lies lie entirely on one side or the other of the frontier, i.e. either the more the
supposedly in its
purely technological’ nature. Its alleged independence from better always, or the more the worse always. There still remains the extremely
the problem of
relative valuation of commodities appears to provide a prima serious problem of evaluating the different effects. The use of uniform weights
facie case for
separating out the efficient set as the first step in an exercis in evaluating all alternatives is not really a purely technological question.
e of resource
allocation. This has an obvious bearing on problems of It restricts the problem of evaluation into a relatively narrow box the relevance
planning. Further-
more, even for analysing the working of the market, of which depends on the institutional set-up in question. The point is best
efficiency has been taken
a=

to be a crucial building block, and an import illustrated with a specific example.

ant result in the literature of
resource allocation is concerned with, as Dorfman, Consider two techniques, A and B, such that A can be used only in family-
Samuelson and
Solow
[5] put it, ‘a close relationship between the purely “econ based production (say, because the problem of work supervision that will
omic” notion of
competitive equilibrium and the purely “technologic be involved in its use in the factory system will be unfeasible), and B can be
al” notion of efficiency’
(p. 404). used only in the factory (perhaps because it may require multiple-shift
But is the notion of efficiency in fact a purely techno working, which cannot be employed in the family system). Assume further-
logical one? This can
be so only in a very special sense. It is certainly the more that all productive resources can be used under either system, but that
case that a production
plan which is inefficient will yield less social value of output people prefer to work in family enterprises. Consider now the case in which
than some other
alternative for any set of positive weights, but the entire the family-based production technique, A, requires more labour and the
approach is based
SE

on two underlying assumptions, viz. (a) a watertight same amount of all other resources per unit of output compared with B.
partitioning of the
components of each activity into the production of
A

desirable commodities Obviously the use of A will be inefficient. Should we reject A in any choice
and the using up of inputs, which is treated as a negati in which B is also available? Such a rejection may be quite unwarranted,
ve contribution, and
PeaBeret

(4) confining the valuation problem to the class of since the people involved actually prefer to work under the family-based
rules that use uniform
weights. I believe that the latter is a more serious limitation system. In fact if they demand a higher wage for working under the ‘alienated’
both in theo
and in practice, but I shall take up the former problem first. system of factory production rather than in small family units, it is possible
The distinction between desirable results and undesirable i
ones does not that even in terms of market prices A will involve a lower unit cost than B.
Spot Teeau FS

correspond exactly to that between production of outputs and While B is superior to A in terms of productive efficiency, A can still turn
absorption of
inputs. Industrial waste is an ‘output’ but is not much loved out to be more profitable and may also be thought to be socially more desir-
whereas the
use of one’s latent talents is an input but its use may not able. An exercise that begins by weeding out inefficient techniques can be
be thought to bea
bad thing. This need not, however, in itself be a source quite off the mark.
of major worry.
since it is possible to represent undesired industrial waste In principle, one can salvage the notion of efficiency by taking labour under
as a quantity with
a negative sign in the production vector and one can put a positiv the family system and that under the factory system as two different inputs.
e sign before
the use of people’s latent talents. If people have sensitive preference between different modes of production
This way of getting round the problem will cause no and different systems of organisation, such distinctions would have to be
great hardship as
long as the effect in question is not thought to be desirable introduced for all different techniques. The problem is quite fundamental,
up to a point and
undesirable beyond that, or vice versa. But this, alas, is freque since, while labour is a commodity in some sense, labourers are endowed
ntly enough
precisely the case. An industrial by-product may have some with minds and with ideas and preferences. In view of the fact that these
limited use and
may even command a price in the market if its supply is relativ preferences may include alternative techniques as relevant variables, labour
ely small, but
the same good may be a damn nuisance to get rid of beyond inputs will then have to be thought of as being technology-specific. One can
the level of
saturation. Similarly, in a situation of widespread workle still define the set of efficient production combinations, but the criterion of
ssness more use of
labour may well be thought to be socially desirable up to a point productive efficiency will not then have any cutting power at all, since all
but beyond
techniques will be ‘efficient’ under this definition of inputs. When all the
 Amartya Sen The concept of efficiency 201
ae
qualifications have been put in, productive efficiency can emerge unscathed envisaged by Joan Robinson does frequently arise in economies with multiple
only by becoming an empty box. labour markets. For example, P. N. Dhar and Harold Lydall [4] found in
India that ‘available evidence suggests that small factories use more capital
Vv and more labour per unit of output than larger factories (. 84). How, then,
do the small factories survive? Governmental help, which has been forth-
It is not my intention to argue that no use can ever be made of the concept coming in some cases, is only part of the answer, and the differences in resource
|
of productive efficiency. We use the concept in some limited form all the prices—in particular, differences in labour cost—are extremely important to
i
\

|i time. In coming to this conference in Coventry from London I could easily the survival of small factories. Whether the lower labour costs of the smaller
have considered the following alternative methods of getting here: (a) taking factories reflect a genuine social advantage is an important and complicated
a train from Euston to Coventry, and (6) launching a ship from Southampton, question to investigate, but it should be clear that one cannot get asl:
sinking it in mid-Atlantic, returning to Southampton on a small boat, travel- in this policy decision simply on the basis of the ‘inefficiency’ of a class o
|} ling to Euston, and then taking a train from Euston to Coventry. If the latter
technique were proposed to me, I would have rejected it, and of course it Se efficiency does not, in fact, abstract from the valuation problem;
could be argued that in some sense I would have then used the concept of it uses it implicitly in confining comparisons to the class of uniform prices.
SS
Se

efficiency. ‘This type of efficiency calculation is ingrained in our thinking. This makes the approach particularly unsuitable for guiding resource alloca-
The specification of any technology already involves some efficiency con- tion in situations in which choices between different modes of production
SSS

sideration in this sense. A perfectly possible method of painting a wall is and systems of organisation are involved. ‘The biggest problems arise with
getting a can of paint, dipping one’s shirt in it, throwing away the can of the use of labour, since differential evaluation of work in different technologi-
paint along with the shirt, getting another can of paint and spraying the cal atmospheres is an important prerogative of the working man.
paint on to the wall. The reason why this method, which does lead to painted
eS

walls, is not considered to be a ‘technique’ of painting walls, is that the

definition of a technique already involves some use of efficiency considera- VI
Sasaersae

tions. ‘The problems arise when we try to extend such minimal use of the I should now like to turn to the concept of economic efficiency. Pareto opti-
efficiency criterion to choices between serious technological alternatives. mality has been undoubtedly the most widely used welfare criterion in
And it is here that the misleading simplicity of the concept of efficiency can modern welfare economics. If a change would make someone feel better off
be seriously deceptive. and make nobody feel worse off it is difficult to argue against such a change.
Neoclassical economics has, of course, been much concerned with produc- Problems of the kind encountered with productive efficiency cannot easily
tive efficiency, and the achievement of productive efficiency is one of the many arise with economic efficiency, since the latter is defined in terms of the
alleged virtues of the competitive market. On the other hand, it is by no means indivi ” own preferences.
the case that only neoclassical economists have taken inefficiency to be the Pai ne it has even been argued that if everyone in the society
first thing to eliminate. ‘The so-called Cambridge theories of capital and growth prefers social state x to y, then it must logically follow that « is a better alter-
make considerable use of the notion of efficiency as the first step in analysing native for the society than y. This position is not tenable, for at least two
So

decision-making under capitalism as well as under socialism. In the battle different reasons. First, the fact that some property 1s satisfied by each member
between the two Cambridges efficiency has provided one of the few areas of of a set does not imply that the property must be satisfied by the set itself.
There is no logical reason why the preference of the society must be defined
——-

truce.
In her classic anti-neoclassical textbook, Joan Robinson [14] considers a as the same as the preference of each of its members even if the preferences
hypothetical choice between heavy ploughs and light ploughs for a socialist of all the members coincide. Second, there is a difference in category becyeen
economy, and within the framework of efficiency argues that if ‘a given invest- the type of judgement that is conveyed by saying that person i Degas a be y
ment in heavy equipment not only yields a larger output per man employed and the type of judgement that is expressed by saying that x is a better a -
but also makes a larger contribution to total output’, then ‘clearly, the light native for the society than y. The former is a statement of fact, viz. that the
ploughs are a dud investment and should not be built’ (p. 32). Since light person in question happens to prefer x to y, whereas the latter is a aunt:
ploughs and heavy ploughs may go with very different types of production judgement indicating the recommendation that, given the social choice
organisation and the cultivators involved may have views on the choice between w and y, the former should be chosen. Pareto optimality is a san
between these alternatives, the reality of the choice is, however, considerably criterion proposing a link between people’s actual preferences and optim
less simple. social choice, and the relationship cannot be purely analytic.
The question is not merely of theoretical interest. The type of choice It can, however, be claimed that even though the Pareto criterion may not
 202 Amartya Sen The concept of efficiency 203

have any purely logical force, it is so uncontroversial that i i known (see Koopmans [9]) that there are difficulties in reaching Pareto
accepted without much ado. I have tried to argue Cee RU eerie ae optimality when externalities are present. What is at issue here, however, is
Pareto criterion is not so uncontroversial after all, and in particular that it not whether Pareto optimality is achievable but whether it is desirable to
conflicts with a relatively mild requirement of liberalism. There is no point achieve it. Externalities and interdependent utilities have been recognised
in repeating the whole argument here, but I should just note that the problem for a long time as barriers to reaching Pareto optimality; what emerges from
arises from the fact that what we accept to be socially right cannot, in general discussion of the class of problems illustrated by this example is that Pareto
be ascertained only from people’s preferences, and it is often necessary & optimality may not even be worth reaching, and may in fact be viewed as
bring in the motivation underlying those preferences. An illustration that I socially undesirable under certain circumstances.
used in the earlier discussion concerns the problem of deciding between The argument outlined here raises doubt about viewing Pareto optimality
three alternatives involving the disposal of an allegedly pornographic book even as a necessary condition for social optimization. It is, of course, much
between two persons, viz. P, who is a prude, and L, who is lewd. Consider more widely conceded that Pareto optimality is not very persuasive as a
the case in which the prude prefers most that nobody reads the book; next sufficient condition for social optimisation. A situation can be Pareto-optimal
best that he reads it (he is ready to take the blow on himself rather than lettin and still involve a lot of inequality, and as such the mere attainment of econo-
gullible L be exposed to it), and worst that L reads it. On the other had mic efficiency cannot be taken to be a very great achievement. I have tried to
L considers that the worst case is nobody reading the book, but given the argue that Pareto optimality is not entirely persuasive either as necessary or
choice between either of them reading the book, he prefers that P should read as a sufficient condition for social optimisation, and the example discussed
the book (he is thrilled at the idea of the prude having to go through the stuff) is merely an illustration of the type of problem involved. But I would like to
On liberal grounds it can be argued that if the choice is between nobody devote the rest of this lecture to examining the role of Pareto optimality in
reading the book and the prude P reading it, it is socially better that nobody the standard literature of competitive resource allocation within its own
reads the book, since P himself does not wish to read it and he should ae normative framework, which takes Pareto optimality as a necessary but not
be forced to do so just because L wants him to. Similarly, it can be recom- a sufficient condition for overall optimality.
mended on liberal grounds that, given the choice between nobody readin
the book and lewd L reading it, it is socially better that L should read it
since he does prefer to read it, and the fact that P does not want him to Vil
should not prevent L from having his fun. But both prefer that the prude P The so-called ‘basic theorem of welfare economics’ is concerned with estab-
should read it rather than lewd L. If on Paretian grounds it is now ie ted lishing a two-way relation between competitive equilibrium and Pareto
that it is better for P to read the book rather than L, we now have a nae optimality. The rationale of the price mechanism has been viewed in recent
problem, since each alternative is worse than some other. There is no undomi- years in the light of this dual relationship, and the welfare economic discussion
pone and the mild requirement of liberality conflicts with the of price mechanism has come increasingly to rest on the foundations of this
theorem. As Dorfman, Samuelson and Solow [5] put it (pp. 409-10), ‘More
It is not my purpose here to go into the logical confli recently it has become common to sum up all these in one brief and easily
optimality and liberal values, but only to eshe in Seip ae is understood theorem which contains everything of significance and provides
quite conceivable that some people would recommend the violation of the the backbone of modern welfare economics. This fundamental theorem
criterion of Pareto optimality. While no thundering liberal myself, my own states: Every competitive equilibrium is a Pareto optimum; and every Pareto
inclination in this case would be to recommend that L should read the book optimum is a competitive equilibrium,’
that he wants to read, and P should not have to read the book that he does It is convenient to consider separately the two parts of this analytical result,
not want to. But this view does, of course, imply that the only Pareto- since they have rather different welfare implications and also require rather
inefficient alternative in this set of three is being singled out as socially best different assumptions. That every competitive equilibrium is Pareto-optimal
leaving out the two Pareto-optimal points, one of which is Pareto-superior to I shall call the Direct Theorem, and that every Pareto optimum is a competi-
i eee choice. While both P and L prefer that P reads the book tive equilibrium I shall refer to as the Converse Theorem. Both theorems
rather than L, the motivations underlyi are established by making assumptions such as the absence of externalities,
optimality a less persuasive criterion, et eae but the non-saturation assumptions are somewhat different, and, more
_ Perhaps I should warn against a possible danger in misund i important, the Direct Theorem does not require the convexity assumptions
significance of this problem. In a formal sense it is quite Bes et (e.g. the absence of increasing returns to scale) which are crucial in establishing
the problem arises from the existence of ‘externalities’, even though 4 ve the Converse Theorem.
special type of interdependence is involved in this case. It is, of course, wel Questions can be raised about the realism of these assumptions, but I
 204 Amartya Sen The concept of efficiency 205

do not wish to raise them here. Instead, I should like to examine precisely Direct Theorem. Debreu [3] describes the Converse Theorem as a ‘deeper’
how powerful an argument in favour of the use of the price mechanism is result, which is certainly the case, and Koopmans [9] rightly notes that it is
implied by these theorems within the framework of these assumptions, and this theorem (rather than the Direct Theorem) which ‘is the central proposi-
I would argue that more has frequently been read into the welfare implications tion of the “new welfare economics” ’ (p, 52). No matter how the social
of the price mechanism than the theorems in fact warrant even when all the welfare function is defined, if it is conceded that Pareto optimality is a
simplifying assumptions are fully accepted. necessary condition for social welfare maximisation it follows from the Con-
Consider the Direct Theorem first. That any perfectly competitive equili- verse Theorem that the welfare-maximising state will be a competitive equili-
brium will be Pareto-optimal does not, of course, imply that the outcome will brium, So the very best must lie within the reach of the competitive price
be particularly attractive, since Pareto optimality is consistent with the most mechanism.
intolerable inequities of the distribution of income. In fact it can even be the There is no doubt that the Converse Theorem is a major result in the
case that for every Pareto-optimal situation except one there is a non-Pareto- theory of resource allocation. The question, however, that we have to ask in
optimal situation that is judged to be socially superior. This simple point is the context of our investigation concerns the case for the use of the price
mechanism that is implied by this classic theorem, One way of posing this
U, query is to enquire into the advantages of trying to reach the social optimum
through the price mechanism compared with some other mechanism—say,
a completely centralised allocation.
If the central authority, which we shall refer to as the Planning Commis-
sion, had all the information that would be needed to pick the point of social
optimality, it can obviously proceed to do so even without the benefit of the
price mechanism. But there are at least two real barriers to this: (a) the
Planning Commission does not have all the information, and (b) the individual
economic agents may not have the incentive to execute the chosen plan. It is
on these two points that the price mechanism is supposed to score its major
victories over its rivals. For the working of the competitive mechanism each
participant needs to know relatively little—essentially, his own preferences
and his own production possibilities, plus the prices ruling in the market.
It is this informational economy, as well as the fact that, given the prices,
each agent has an incentive to make the right choice, which have been seen
as the main advantages of the price mechanism. The demonstration of these
advantages has been a major theme in economics for a very long time. As
Fig. 9.1 Arrow and Hahn [1] point out, ‘There is by now a long and fairly imposing
line of economists from Adam Smith to the present who have sought to show
illustrated in the two-person case in fig. 9.1, in which OAB represents the that a decentralised economy motivated by self-interest and guided by price
set of possibilities defined in terms of utility combinations of the two indi- signals would be compatible with a coherent disposition of economic resources
viduals. The social welfare function that is used may be Pareto-inclusive in that could be regarded, in a well defined sense, as superior to a large class of
the sense that a Pareto-superior position must be judged better, but the possible anternative dispositions’ (pp. vi-vii). The theorems in question have
Pareto-non-comparable points can be ordered in any way consistent with the been viewed in this light, and as Koopmans [9] puts it, ‘one can. . . interpret
Pareto quasi-ordering. CD and EF are two typical indifference curves repre- the proposition [the Converse Theorem] as a statement of conditions under
senting two given levels of social welfare. G is Pareto-optimal but is clearly which the simplicity of incentive structure and the economies of information
inferior to H, which is Pareto-inoptimal. Making the appropriate assumptions handling characteristic of a competitive market organization can be secured
of continuity and convexity, we shall find that, with the exception of point without loss in efficiency of allocation’ (p. 53).
I, every other Pareto-optimal state is inferior to some non-Pareto-optimal In summary the position formally appears to be something like this. Let
state. Thus the mere attainment of Pareto optimality can be far from a T be the set of possible social states, and P(7’) the sub-set of Pareto-optimal
thrilling achievement. social states (i.e. the set of states maximal in 7 with respect to the Pareto
Not surprisingly, the case for the use of the price mechanism has more quasi-ordering). Given the objectives of the Planning Commission, let BT)
frequently been seen to be based on the Converse Theorem than on the be the set of states that the commission finds is best in the set T. (B(T) can
206 Amartya Sen The concept of efficiency 207

include one alternative or many.) If the planning Commission’s objectives which must be the case if Pareto optimality is a necessary condition for
include Pareto optimality as a necessary condition, though not necessarily overall optimality. But in order to specify the best set, B(T), from which x
sufficient, B(T) will be a sub-set of P(T’). The Direct Theorem states that is to be chosen the Planning Commission has to know the set 7, and that,
every competitive equilibrium leads to some point in P(T). Such a point alas, does away with the informational economy even before the market
however, need not be in B(T). But the Converse Theorem states that every mechanism is called upon to achieve the chosen best state .
point in P(T’) can be reached by some perfectly competitive equilibrium or
other. Thus every point in B(T’), which is a sub-set of P(T), represents a
Viol
competitive equilibrium as well. Furthermore, once any point in B(7’) has
been picked by the Planning Commission, the appropriate prices can be There remains, however, one final question to examine as far as the informa-
calculated and some initial distribution of resources that will lead to the tional side of possible uses of the Converse Theorem is concerned. In the
equilibrium corresponding to « can be found out (see Debreu [3], pp. 93-4). theory of iterative decentralised planning considerable use is made of the
It will then be in everybody’s interest to do the right thing for« to emerge prices ‘associated’ with an optimal programme (see, for example, Arrow and
(eschewing the problem that can arise from more than one equilibrium Hurwicz [2] and Malinvaud [11]). In the iterative exchange between the
corresponding to the same initial resource distribution), Each economic Planning Commission and the individual firms, hypothetical prices in each
agent need have no information on anything other than his own preferences round serve as ‘prospective indices’ which the Commission sends to the firms.
his own production possibilities and the ruling prices. The proposals of the firms, drawn up on the basis of profit maximisation at
But can the Planning Commission decide on the best set, B('), without the hypothetic prices, serve to reveal to the Commission the relevant informa-
having a lot of information? 'To know what the best points B(T) in T are tion about production possibilities which the Commission takes into account
according to its own objectives, the Planning Commission must know what in initiating the next round. Prices serve here in the systematic exchange of
possibilities are open, i.e. the set 7. But to know the set of possible social information. It is, of course, possible to think of well defined decentralisation
states, with each state representing everybody’s efforts, consumption and procedures leading to convergence to the optimal plan such that the price
production, the Planning Commission has to know the preferences and mechanism is not used. For example, Weitzman [17] has proposed a procedure
production possibilities of all economic agents, and clearly no economy of in which the Planning Commission sends out quantitative plans rather than
information is going to apply as far as the Planning Commission itself is prices, while the firms respond with relevant rates of transformation rather
concerned. Of course, no economic agent other than the Planning Commission than with quantitative plans (see also Korani and Liptdk [10] and Marglin
need look beyond his nose, but to pick the best set B(T’) the Planning Com- [13]). But even these procedures do make substantial use of the duality between
mission has to look at the whole economy in some detail. But if the Planning optimal quantities and associated prices.
Commission does have information in all this detail, then why need it go How does the Converse Theorem contribute to the role of price mechanism
through the market mechanism at all; why can’t it simply make up a quan- in the context of decentralised iterative planning? The fact that there are
titative plan? And if it does not have all this information, then what consola- competitive prices which would be associated with any chosen point in P(T)
tion does the Planning Commission get from a theorem that states that if it and therefore with any chosen point in B(T) does not, of course, imply that
knew B(T), which it does not, and could choose a best point x, which it can- a well defined decentralised procedure necessarily exists that would lead the
not, then the market mechanism would be perfectly capable of establishing Planning Commission to converge iteratively on to the optimal plan. On the
an equilibrium at w? other hand, the existence of such competitive prices does help in overcoming
The picture is a bit more complex than this, as I shall presently discuss an initial hurdle, since there is a close relationship between competitive prices
but it should be clear that there is a simple difference between the use of the and the associated prices with an optimal programme.
Direct ‘Theorem and that of the Converse Theorem as far as the informational Despite this fact, in the existing literature there is no well defined and
advantages of the price mechanism are concerned. The Direct Theorem convergent decentralisation model that has gone fully into the question of
states that every competitive equilibrium is Pareto-optimal and that for equity and income distribution, and the few models that have thrown light
achieving Pareto optimality we do not have to know the set B(7); even the on this question (see, for example, Dreze and de la Vallee Poussin [6] and
Planning Commission need not look beyond its nose. But, as we saw earlier Malinvaud [12]) have been concerned with rather specific issues. It is, in fact,
the welfare content of the Direct Theorem is rather limited, since Barer usually the case that the objective function of the Planning Commission is
optimality is at best a necessary but not a sufficient condition for social welfare defined in terms of total quantities of consumption without bringing in the
maximisation. ‘he Converse Theorem is much more interesting in that once distributional question at all (see the excellent survey of the literature by
the Planning Commission picks its optimal plan—some element « of B(T)— Malinvaud [11]). What are the reasons for this reticence?
the market mechanism can yield x as long as B(T) is a sub-set of P(T), Part of the problem undoubtedly arises from the usual difficulties in having
 208 Amartya Sen
a social welfare function that can go into equity considerations which will
y The concept of efficiency

very substantially the relevance of the Converse Theorem for price-based

209

permit the specification of a best sub-set B(7') of P(T) on distributional optimal allocation of resources.
grounds. But there is also a serious problem of incentives in informational
exchange that make income distributional questions particularly difficult to
bring into the scope of iterative planning, It may be recalled that the Converse IX
Theorem points out that for any chosen best point « out of the set of Pareto- To summarise, I have tried to argue in this lecture that the concept of effici-
optimal states P(7’) there is a competitive equilibrium that will yield x. To ency has less to offer than the concentration on it in the modern theory of
‘obtain that particular competitive equilibrium the initial distribution of resource allocation suggests. As far as productive efficiency is concerned, the
resources between the persons has to be right, and there are procedures for belief that it avoids the problem of relative valuation is true only in a very
calculating the inter-personal resource distribution that will lead to a specified limited sense. Further, even in the absence of externalities and saturation
competitive equilibrium (see Debreu [3], pp. 93-4). At the end of the iterative there may exist a case for choosing a technique that is inferior than another
procedure when the Planning Commission has gathered enough information in terms of productive efficiency because of the workers’ preferences between
to pick a best state, x, the commission will have to redistribute resources in technologies, and while this problem can be avoided by defining labour ina
such a way that « can be achieved through the competitive price mechanism. technology-specific way, this will tend to make all techniques efficient and
Do the individuals have any compelling incentive to play this iterative game? thus productive efficiency will become an empty box.
It is easily checked that frequently enough it will be in the interest of the As far as economic efficiency or Pareto optimality is concered, there are
individual to send in signals that deliberately mislead the Planning Commis- serious reasons for doubting whether it should be a necessary condition for
sion so that a more favourable distribution of resources is achieved by the social welfare maximisation, let alone sufficient. The difficulty arises from
individual compared with what would have happened if honest signals were problems in defining social welfare in terms of individual preferences only,
sent (see Sen [16]). ignoring other factors such as the reasoning and the motivation underlying
It should be noted that even if the decentralisation procedure takes the the preferences. ESN
distribution of initial resources as given, and tries to optimise within that But even if these doubts are eschewed and Pareto optimality is accepted
constraint, it may be in the interest of individuals to send misleading signals, as a necessary condition for social optimization, there are problems in assessing
since relative prices affect the real well-being of people and there exist strategic the relevance of the standard results on Pareto optimality which have come
considerations in playing the iterative decentralisation game. This type of to be known as the ‘basic theorem of welfare economics’. The part of it
influencing through prices is, however, rather indirect, and the possibility which states that every competitive equilibrium is Pareto-optimal (the Direct
may even be somewhat obscure to the individual in question. On the other Theorem) turns out to be not very interesting, and this has been widely
hand, if the initial resources are actually redistributed before the Planning conceded. The other part, asserting that every Pareto-optimal ‘state is a
Commission hands over the system to the competitive mechanism, the indi- competitive equilibrium (the Converse Theorem), is certainly more interesting
vidual may quite easily see that there is a clear relation between his responses but a case for the use of the price mechanism based on this result has to cross
and the lump-sum taxes and subsidies to which he is subjected. For example, some hurdles that Jook extremely difficult to overcome. ‘The informational
for a Planning Commission that is known to be anxious to achieve equality, economy of the competitive mechanism, which does apply to the less interest-
an individual with high productivity but valuing leisure may decide to pretend ing Direct Theorem, turns out to be rather suspect for any use of the Converse
that he is rather incompetent, expecting resource subsidisation by the govern- Theorem for social welfare maximisation. The Planning Commission has to
ment, so that he can take life easy. Problems of this kind can be quite fatal decide on a best set, B(T), for choosing an optimal plan, and if B(T) is a
for iterative decentralised planning involving equity considerations. sub-set of the Pareto-optimal set P(T) any x in B(T) can, of course, be
There are, in fact, two types of incentive problem in the resource allocation reached by the competitive mechanism, thanks to the Converse Theorem.
exercises. The first is concerned with the incentive to reveal one’s hand in But the Planning Commission has to know about the set of possibilities L
the context of iterative informational exchange, and the other is the incentive to be able to choose B(T), and the informational economy that is enjoyed by
to do the appropriate thing once the government has chosen a particular social each participant in the competitive mechanism is not enjoyed by the Planning
state and installed a particular set-up of resource distribution and relative Commission, which has to pick B(T), The particular feature of the Converse
prices. The competitive mechanism gets high marks on the latter score, Theorem that makes it a deeper and more interesting result than the Direct
but has nothing to give on the former. On the other hand, in order to be able Theorem is also the feature that militates against the economy of information
to maximise social welfare through the use of price mechanism in iterative (enjoyed by the uses of the Direct Theorem) when applied to the Converse
decentralised planning the former is quite crucial. The incentive advantages Theorem.
of the competitive mechanism do not extend to this problem, and this restricts An attempt to take care of this difficulty through an iterative decentralisa-
 210 Amartya Sen
tion procedure runs directly into problems of incentives. The

10
Planning
Commission has (a) to collect the information from individuals that will
permit it to choose a welfare-maximising state, «, and the corresponding
competitive equilibrium, and (b) to redistribute initial resources so that the
chosen competitive equilibrium yielding x will emerge. These two roles tend The depletion of exhaustible resources
to conflict with each other in terms of incentives in informational exchange.
The incentive advantages of the competitive mechanism do not extend to
this crucial issue, on which the significance of the Converse Theorem very Geoffrey Heal
largely depends.
These difficulties all arise entirely within the framework of assumptions
in terms of which the concept of economic efficiency has been given a
predominant role in the theory of resource allocation.

References

1 Arrow, K. J., and Hahn, F. H., General Competitive Analysis, Mathematical This paper! addresses a question which is currently very fashionable, which
Economics ‘Text No. 6, Holden-Day and Oliver & Boyd, 1972. is very easy to state, but to which it is unfortunately very difficult to give any
2 Arrow, K. J., and Hurwicz, L., ‘Decentralization and computation in resource definitive answer. The question is ‘Is there any sense in which stocks of
allocation’, in Essays in Economics and Econometrics, ed. R. W. Pfouts, University exhaustible resources are being depleted too fast?? Now that the question
of North Carolina Press, 1960. has been posed, it should be clear why it is difficult to answer: it raises a
3 Debreu, G., Theory of Value, Wiley, 1959. ‘too fast’—
wide range of issues, and contains terms—for example, the phrase
4 Dhar, P. N., and Lydall, H. F., The Role of Small Enterprises in Indian Economic ; aa
Development, Asia Publishing House, 1961.
which are not easily defined.
5 Dorfman, R., Samuelson, P. A., and Solow, R. M., Linear Programming and The way I want to tackle this problem is not by the capital-intensive,
Economic Analysis, McGraw-Hill, 1958. computer-simulation techniques of Forrester, Meadows et al. but | by
6 Dreze, J., and de la Vallee Poussin, D., ‘A tatonnement process for guiding and economic analysis—by analysing the framework within which decisions
financing an efficient production of public goods’, Review of Economic Studies, concerning the rate of resource depletion are made. Essentially one can
vol. xxxviir, 1971. subdivide these decisions into two categories—those made by governments,
7 Habakkuk, H. J., American and British Technology in the Nineteenth Century, and those resulting from market forces. Of course, this distinction is not a
Cambridge University Press, 1962. rigid one: government decisions influence markets, and vice versa. But the
8 Kaldor, N., ‘A model of economic growth’, Economic Journal, vol. uxvit, 1957. distinction is a useful one, and I am afraid that I shall assume it to hold
9 Koopmans, T. C., Three Essays on the State of Economic Science, McGraw-Hill,
1957. ae economic analysis one can carry out is undoubtedly most straight-
10 Kornai, J., and Liptak, T., ‘Two-level planning’, Econometrica, vol. xxxttt, forward in the case of government decisions, so I turn to those first. Where
1963.
11 Malinvaud, E., ‘Decentralized procedures for planning’, in Activity Analysis decisions on the rate of resource-depletion are made by a government agency,
in the Theory of Growth and Planning, ed. E. Malinvaud and M. O. L. Bacharach, then these will normally be the result of an explicit or implicit cost-benefit
Macmillan and St Martin’s Press, 1967. analysis of various alternative policies.2 Now there are many reasons why a
12 — ‘Prices for individual consumption, quantity indicators for collective con- cost-benefit analysis may not produce correct results, and all of them may
sumption’, Review of Economic Studies, vol. xxx1x, 1972. be applicable to the case under consideration. But what I want to concentrate
13 Marglin, S. A., ‘Information in price and command systems of planning’, in on is the role of uncertainty about the future. Of course, this is present in any
Public Economics, ed. J. Margolis and H. Guitton, Macmillan and St Martin’s project evaluation, but one can argue that its presence is particularly obtrusive
Press, 1969.
1'This paper is based on research work conducted jointly by Dr P. S. Dasgupta,
14 Robinson, J., Exercises in Economic Analysis, Macmillan, 1960.
of the London School of Economics, and the present author: a more detailed state-
——

15 Sen, A. K., Collective Choice and Social Welfare, Mathematical Economics ment of these results will be published in due course as “The economics of exhaustible
Text No. 5, Holden-Day and Oliver & Boyd, 1970. resources’, Our understanding of the issues involved has benefited greatly from
16 — On Economic Inequality, Clarendon Press and Norton, 1967. discussions with T. C. Koopmans, E. Malinvaud, J. E. Meade, W. D. Nordhaus
17 Weitzman, M., ‘Iterative multi-level planning with production targets’, Econo- and participants in the economics seminars at Cambridge, the London School of
metrica, vol. xxxvit1, 1970, Economics, Oxford, Sussex and Yale.
2 See, for example, M. V. Posner’s recent book [7].