Wiley

Canadian Economics Association

The Gains from International Trade

Author(s): Paul A. Samuelson
Source: The Canadian Journal of Economics and Political Science / Revue canadienne
d'Economique et de Science politique, Vol. 5, No. 2 (May, 1939), pp. 195-205
Published by: Canadian Economics Association
Stable URL: http://www.jstor.org/stable/137133
Accessed: 23-10-2015 04:39 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].

Wiley and Canadian Economics Association are collaborating with JSTOR to digitize, preserve and extend access to The
Canadian Journal of Economics and Political Science / Revue canadienne d'Economique et de Science politique.

http://www.jstor.org

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
 THE GAINS FROM INTERNATIONAL TRADE

[1] In a recent paper1 the thesis was advanced that while it is not
possible to demonstrate rigorously thatfree trade is better (in some sense)
for a country than all other kinds of trade, it nevertheless can be shown
conclusively that (in a sense to be defined later) free trade or some trade
is to be preferred to no trade at all. I should like here to amplify these
remarks with respect to the last point, that some trade is better than
no trade.
This is by no means a novel proposition. Indeed, it can be traced
back to the beginnings of the Classical theory of international trade.
It has become associated, however, quite unnecessarily in my opinion,
with a labour theory of value, or a "real cost" theory of value, or more
recently, with an opportunity cost theory of value. All of these have
come in for considerable criticism in recent years as restrictive special
cases of the so-called theory of general equilibrium. Those writers who
have insisted on the need for a modern theory of value for a positive
description of behaviour in international trade have in general ignored
the normative aspects of international trade, presumably in the belief
that as soon as one gives up the inadmissible special theories indicated
above, nothing can be said concerning this problem.2 It will be argued
here that this is a mistake, that from the most general theories of equilib-
rium all valid normative propositions can be derived.
[2] It is well to indicate clearly the assumptions under which our
analysis is to take place. We shall consider a single economy consisting
of one or more individuals enjoying a certain unchanging amount of
technological knowledge, so that we may take as data the production
functions relating the output of each commodity to the amounts of
inputs devoted to its production. Any number of commodities is
assumed; there may also be any number of inputs or productive services.
These are not necessarily fixed in amount, but may have supply functions
in terms of various economic prices. Moreover, for our purposes the
differentiation of the factors of production can proceed to any degree;
thus, labour services of the same man in different occupations are not
regarded as the same factor of production unless the provider of these
services is indifferent as between these two uses. Similarly, in order that
the productive services rendered by different individuals may be consid-
1p. A. Samuelson, "Welfare Economics and International Trade" (American
Economic Review, June, 1938).
2A recent exception is provided by P. T. Ellsworth's International Economics (New
York, 1938). However, the problem is posed, not settled. Professor Haberler in his
The Theory of International Trade (London, 1936) does not employ a full general equilib-
rium approach.
195

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
196 The Canadian Journal of Economics and Political Science

ered the same service, it is necessary that in every use they be infinitely
substitutable.
In order to ensure that perfect competition is possible, we rule out
increasing returns, and assume that all production functions show
constant returns with respect to proportional changes of all factors.
Each individual acts as if he were a small part of the markets which he
faces and takes prices as given parameters which he cannot influence by
changes in his own supplies or demands. It is assumed that for each
individual there exists an ordinal preference scale in which enter all
commodities and productive services, and that subject to the restraints
of fixed prices he always selects optimal amounts of each and every
commodity and every productive service (some zero in amount). Each
individual is better off if he receives more of every commodity while
rendering less of every productive service. No attempt is made to
render the "utilities" and "disutilities" of different persons comparable.
[3] Under these conditions, for any assumed set of prices there will
correspond definite demand and supply reactions on the part of every
individual. Moreover, the total outputs of each commodity will be
determined, and the total amounts of productive factors necessary to
produce these outputs will be determined. If the economy is isolated,
it will be necessary as conditions of equilibrium that prices of commodities
and factors of production be such as to equalize the amounts produced
and consumed of each and every commodity, and to equalize the amounts
supplied and demanded of every productive factor.
Under assumed conditions of ownership of the factors of production
and assumed scales of preference for commodities and productive services
on the part of every individual, there will result in general (waiving
possible multiplicities of equilibrium raising problems not peculiar to
international trade) unique equilibrium quantities of consumption goods
and productive services for each and every individual. It is unnecessary
to write down mathematically these equations to deduce the familiar
fact that not enough has been assumed to be able to deduce the absolute
level of commodity and factor prices, but that these are determined
except for a factor of proportionality; i.e., relative commodity and factor
prices are determined. Let us write as follows the equilibrium set of
prices, determined to within a factor of proportionality, which will be
established for our economy when isolated,
Pl . . n, n W?, w2? ... , WS,
2?, *.
with corresponding equilibrium total quantities of the respective commod-
ities and productive services,
XlO, X2,. . ., XO, al?, a2?. . . a. ?.

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
 The Gains from International Trade 197

The total amounts produced of the respective commodities will be

indicated by the barred letters,
X10, X2. X,0

equal respectively in the isolated state to the quantities (unbarred)

consumed.
[4] Before introducing possibilities for trade into our system, it will
be useful in view of the later discussion first to develop certain relations
which must hold in the field of production. Confronted with given
factor prices, firms will combine factors of production in such proportions
as to produce any selected quantity of consumers' goods at the lowest
total money cost. In consequence of this, certain marginal conditions
of equality will be attained (or at least certain inequalities with respect
to finite movements). Although the proof is not given here,3 it can be
shown that this places restrictions on the possible combinations of factors
of production and commodities which can occur. Indeed, it will be
found that the totals of commodities produced and the totals of pro-
ductive services must obey an implicit equation of the following form:
0 [1, 2,2, . . ,, Xn, a1, a2, . . , a]=0. [1
This is capable of the following interpretation: for preassigned values
of all productive services and all but one commodity, this equation gives
the maximum amount of the remaining commodity which can be pro-
duced with the given state of technology. Moreover, with preassigned
amounts of all commodities and all but one productive service, this shows
the minimum amount of this one productive service which is necessary.
Utilizing the well-known law of variable proportions, the following
remarkable theorem can be established. Consider any set of commodity
and factor prices,
Pl', P2', . ., Pn' Wi', 2',. . . , s'.
Since each entrepreneur is trying to maximize his profits, there will result
an optimal set (not unique) of commodities produced and productive
services used, indicated by
Xt, X2t, . . Xnt, al, a 2',. . . ast

satisfying, of course, equation [1]. Our theorem says that for such
preassigned prices the resulting optimal quantities of commodities and
productive services maximize for the economy as a whole the algebraic differ-
ence between total value of output and total factor cost, as compared to any
other commodity and factor combinations satisfying equation [1]. This is
equivalent to the following inequality:
3In a forthcoming paper on the conditions of equilibrium in international trade
I have gone more fully into these and other matters.

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
 ..

198 The Canadian Journal of Economics and Political Science

' . .+<ws'as']l
[P1' gx'+Pa' ~2'+. ?.+P,' g,']--[w' azt+w a:'+.
[Pl' x +P2 x2 + ? . .+Pn n ]-[wl al +w2' as +. . .+w/ a,l, [21
where the unprimed x's and a's represent any point satisfying equation [1].
This inequality merely places certain curvature restrictions on the surface
represented by equation [1], for the various ratios between respective
prices correspond in a well-known manner to the respective slopes (when
they exist) of this surface.4 In figure I are presented typical shapes for

X,. a: X.j

^
..... .. o
FIGUREI

various cross-sections of this surface. In the first diagram is shown the

amount that must be given up of one commodity, xj, in order to get
more of another, xi, with all other variables held constant. This substi-
tution curve must be concave to the origin. The next diagram shows
the amount of one input, ak, which must be added to compensate for
withdrawals of a,, all other variables being held constant. The last
diagram shows the amount of commodity, xj, that can be secured with
additional amounts of ak, with constant levels for the remaining outputs
and inputs.
The above inequality can be written symbolically
zp'x '-Zw'a'_> pp'-Zw'a, [21
where it is always understood that the summations are over the respective
n commodities and s productive services. Of course, a similar inequality
holds for any other preassigned set of prices.
[5] Trade can be introduced very simply into our system without
explicitly dealing with any new country or countries. This is done by
the useful device of supposing that there exists an outside market in
which there prevail certain arbitrarily established (relative) prices at
4The equality sign can hold if all the x's and a's are respectively proportional (or
equal) to the primed x's and a's., In the singular (and rare) case where the preassigned
factor prices are such that all factors of production are used in equal proportions by all
commodities, it is possible for the equality sign to hold. This constant cost case does
not essentially modify the analysis.

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
 The Gains from International Trade 199

which this country can buy or sell various commodities in unlimited

amounts without changing those quoted prices. It does not matter for
the present purposes how, in fact, such prices would be established in
this outside market or source, but rather we are interested in the effects
upon this country of the existence of such quoted prices.
The fact that this outside market will both buy and sell at the new
quoted prices will compel the prices of respective goods in the domestic
economy to assume equivalent ratios, or else corrective arbitrage move-
ments would take place. Obviously, therefore, we have introduced new
forces to determine some of the prices. It is necessary, then, to drop
some of our previous conditions. In particular, we must dispense with
the condition that the amounts of commodities produced domestically
and consumed domestically must be equal. Instead we have the single
condition that the total value of imports must equal the total value
of exports, or
Zpx = Zp. [31
It is clear that for any preassigned prices of internationally tradable
goods there will result certain equilibrium values for all the other vari-
ables, quantities produced and consumed, productive services supplied,
and prices of non-tradable commodities and services.
For one set of prices, namely those proportional to the set [p10, p20,
* P] which would prevail if the economy were isolated, no trade
will result.5 For these particular prices are such as to equalize the
domestic production and consumption of each and every good. For any
other set of prices, some trade will result, and there will emerge new
equilibrium values for all of our unknowns. By assigning appropriate
values to our outside prices, we can obviously reproduce all possible
conditions of trade which could conceivably arise. This is the justifica-
tion for introducing a simplifying device which enables us to ignore the
existence of outside economies. Of course, if we were trying to explain
the actual prices with which our economy will be confronted, it would be
necessary to consider outside conditions.
[6] I first apply our analysis of the effect of introducing relative
prices, different from those which would be established if our system were
isolated, to a simplified case in which all members of our economy are
identical in every respect. That is, the same ordinal preference schedule
relating commodities and productive services is assumed for every
individual, and also the same ownership in the means of production.
5Atrivial exception is provided by the constant cost case mentioned in the previous
foot-note. Here at the isolated state prices there might be an unimportant possibility
of neutral equilibrium as recognized in the Classical theory of international trade.
I adopt the convention of definingtrade to exclude this possibility.

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
200 The Canadian Journal of Economics and Political Science

This does not mean that the utilities of different individuals are com-
parable. Indeed, since all individuals are identical, if one is bettered
(in an ordinal sense) by the introduction of trade, then all will be bettered,
and there will be no necessity for making any welfare comparisons
between individuals.
In these circumstances, the following theorem can be established:
the introduction of outside (relative) prices differing from those which would
be established in our economy in isolation will result in some trade, and as a
result every individual will be betteroff than he would be at the prices which
prevailed in the isolated state. The truth of this has been intuitively
apprehended by a great many economists, but I do not believe that there
exists anywhere in the literature a rigorous proof of this proposition.
To illustrate the difficulties which must be encountered in establishing
this theorem I present a table showing some possible results of the intro-
duction of trade. In the first two columns are respectively the prices
and quantities consumed of three commodities; in the next two columns,
the prices and quantities produced of the same three goods; in the last
two columns, the prices and quantities of two factors of production.
Case I gives a hypothetical set of prices which would prevail in the
isolated state with equal production and consumption of all commodities.
The amounts corresponding to each individual would be some constant
fraction of the total quantities. Although actual prices are given to
avoid the asymmetry of using any one good or service as numeraire,
only relative prices are of importance.
If a new set of relative prices are imposed from without, new equilib-
rium values will be appropriate. Cases II, III, and IV indicate possible
sets of equilibrium values which might emerge, depending on the partic-
ular make-up of tastes of the individuals in question.6 In Case II after
trade is established, it will be noted that more of every commodity is
consumed, while less of every productive service is provided. Obviously,
Case II is an instance of our theorem. But what can be said of Case III?
Here, the same amounts of all productive services are provided, but not
more of every commodity is consumed. More of commodities x2 and X3
will be consumed, but less of commodity xi. In Case IV things appear
to be still worse. Not only does the quantity of some commodity
decrease, but also more of the productive service a2 is provided. Is it
possible to say in the general case that the new situation is better than
the old, or is our theorem false?
It is obvious that a labour theory of value cannot be of any aid in the
6Cases II, III, and IV are alternative and mutually exclusive possibilities. Hence,
although each is consistent with Case I. they are not necessarily consistent with each
other.

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
 The Gains from International Trade 201

analysis of this problem, since two factors of production have been

assumed. The opportunity cost doctrine as presented by Professor
Haberler could be applied only to Case III, where the total amounts of
the various factors of production remain unchanged after trade has taken
place. Contemplation of the behaviour of the terms of trade would
suggest that an improvement has taken place, but it would be easy to
construct examples for which this test would give a spurious result.
None of the usual methods throws any light on the question as to whether
Case IV represents an improvement over the condition which would

TABLE I
p x p x w a
Case I-no trade........ 1 10 1 10 4 5
2 15 2 15 2 20
1 20 1 20

Case II................ 3 11 3 20 9 4
2 17 2 15 3 18
1 23 1 0

CaseIII............... 3 8 3 20 6 5
2 17 2 15 3 20
1 32 1 0

Case IV.............. 3 8 3 20 6 4
2 17 2 15 3 22
1 32 1 0

prevail in the absence of trade. And yet there can be no doubt that the
situation represented in Case IV is the typical case when trade occurs.
If we assume that in the real world there are innumerable commodities
and productive services, it is scarcely conceivable that after trade takes
place more of each and every commodity and less of each and every
productive service will result. The introduction of trade would be
expected to result in less of one or more commodities and more of one
or more productive services.
Still, if the theorem given above is valid, it must follow that we can
very definitely show that all the given cases in which trade takes place
are better than the original situation illustrated by Case I. It remains
only, therefore, to prove our theorem, after which all the illustrative
examples will emerge as special instances. It will be noted that the
proof to be given depends only on the elementary operations of arithmetic:
addition, subtraction, equality, inequality, etc.
To ensure generality, consider any initial set of prices prevailing in
the isolated state,
0, P20,
p2 . . ., Pn

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
202 The Canadian Journal of Economics and Political Science

and the corresponding equilibrium values of the remaining variables,

X10 X2 . X x10 X 20, . . . c, 1, 0, W1 W2O,. .
W . , W20 ., Ws

Now consider any new set of prices leading to trade,

P1', P 2' . . . , P
and the corresponding new equilibrium values
Xll, X2,. . . *Xn f X 1t X^2 .'i . *V?
, ',.al', a2', . . ., as^,Wt, W2', . . , Ws.

From the production inequality of equation [2] we know that

2 p' '-2w'a' > 2p'x? -w'ao. [4]
But from the condition that the total value of imports must equal exports,
or that the total value of goods produced must equal the total value of
goods consumed, a similar inequality will hold if we leave the bars off
the x's and consider goods consumed instead of goods produced. This
gives
p'x' -Zw'a'_ Zp'x?-2w'a?. [5]
I now assert that this condition (barring the unimportant case of
equality sign mentioned in foot-note 4) assures us that each of our
identical individuals is better off in the second case than in the first.
Imagine an individual confronted with commodity and productive
service prices [P1', P2', P. . ', , w2', . . , ws]. Subject to these
prices, his most preferred position with respect to consumption and the
providing of services is shown by his behaviour to be [x1', x2', . ., x,
al', a2, . , a,']. By considerations similar to the economic theory of
index numbers as developed by Pigou, Haberler, Koniis, Staehle, Leontief,
Frisch, et al., it can be shown that this combination is preferred in an
ordinal sense to [x1?, x2?, . . ., x?, a1, a2 , .. ., as0].7 If at the primed
set of prices the individual would have bought the original combination
of goods [X?], and provided the original amounts of productive services
[A?], the total algebraic cost would have been less than that of what he
actually bought and sold [X', A']. In addition, therefore, something
more could have been bought of every commodity, and a little less of
every productive service supplied. This proves that [X', A'] is better
than [X?, A ], for if this were not so, why did not the individual actually
choose [X?, A ], and perhaps a little more of every good and a little less
of every service, in preference to [X', A']? If the individual was in a true
maximum position at the primed prices, it must necessarily follow from
7For many reasons I regard the index number approach as a clumsy device for
solving the problem at hand. A more convenient test as to the ordinal desirability of
two situations is presented in my "Note on the Pure Theory of Consumers'Behavior"
(Economica,March, 1938).

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
 The Gains from International Trade 203

our inequality that [X', A'] is better than [X?, A0]. Thus, our theorem
is proved.
To appreciate the true meaning of this theorem and its proof, the
reader may make the experiment of dropping one or more of our premises
to show how the proof will break down. Such an exercise is provided
by the well-known Graham's Paradox.
No modification in the proof is required by the assumption that there
exist domestic consumers' goods which cannot be traded under any
circumstances. With slight modification transportation costs could be
introduced into the analysis without affecting appreciably the results.
It will be noted that the proof is still valid in the case where there exist
no resources transferable between different production uses. Indeed,
if the conrmodities are not produced at all, but fall from heaven in fixed
amounts per unit time, the theorem still applies. Moreover, the intro-
duction of discontinuities requiring modifications of the usual marginal
analysis is already covered in our theorem.
[7] If, as I have shown, the introduction of outside prices different
from those which would prevail in the isolated state betters all of our
identical individuals, a possible generalization suggests itself. Is it
possible to state that the more prices "deviate" (according to some
convention) from those of the isolated state, the better off all individuals
will be? The answer is in the affirmative. In order not to complicate
the present exposition, I withhold the rigorous proof of this proposition
until a future occasion.
[8] Before going on to consider more realistic cases where individuals
are not all alike, I should like to point out two interesting special cases
covered by the previous theorem. The limiting case of an economy in
which all individuals are exactly alike is that of a single household or
Robinson Crusoe economy consisting of but one unit. Moreover, from
a formal point of view a completely unified economy under perfect
control of some central authorities interested in maximizing some ordinal
preference scale is like a one individual economy. For such single
individual economies, pretending to play the game of perfect competition
is one possible way of arriving at optimal equilibrium values. If self-
sufficiency is not an end in itself, it follows from our previous theorem
(and even under less stringent assumptions) that for an individual or
unified economy trade is always preferable to no trade, although it is
not necessarily true that free trade is the best trading policy.
[9] I now drop the assumption that all individuals are alike with re-
spect to tastes, abilities, ownership of the means of production, etc. The
introduction of changed prices leading to trade cannot, of course, be
expected always to better each and every individual. After trade, the

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
204 The Canadian Journal of Economics and Political Science

prices of items chiefly consumed by a particular individual may have

risen making him worse off than before. (It is not possible, however,
for every individual to be made worse off.)
In order to evaluate the resulting situation, it would be necessary to
have some scale which would take into account comparisons as between
different individuals. For some type of weighting of the fortunes of
different individuals, the result might be judged an improvement. For
some other, such as an egocentric evaluation on the part of those rendered
worse off, the resulting situation might be judged to be worse than that
which prevailed in the isolated state. If nothing more than this could
be said, the problem of the benefits from trade would be of limited
theoretical and practical importance.
Fortunately, definite results which do not depend upon the compari-
sons of the real incomes of different people can be derived. Although it
cannot be shown that every individual is made better off by the intro-
duction of trade, it can be shown that through trade every individual
could be made better off (or in the limiting case, no worse off). In other
words, if a unanimous decision were required in order for trade to be
permitted, it would always be possible for those who desired trade to
buy off those opposed to trade, with the result that all could be made
better off.8 This can be deduced from the fact that as a result of trade
larger (or in the limiting case, equal) amounts of every commodity can
be secured with smaller (or equal) amounts of every productive service.
Without trade the range of possible commodities which are available with
preassigned amounts of all productive factors is given by the implicit
equation [1]. If outside prices are introduced, it will always be desirable
for production policy to be aimed at maximizing the total value of output
at the outside prices, with any preassigned amounts of each and every
productive factor. For this will yield a larger money sum than any
other production policy, and with a larger sum of money more can be
bought of every commodity than with a smaller one. As a result, each
of the following three statements is true: [1] more can be had of every
commodity as of the same totals of all productive services; [2] of the
same preassigned quantities of all consumers' goods, less of every pro-
ductive service need be rendered; [3] after trade, more of every commodity
can be secured with less of every productive service. This ensures us
that by Utopian co-operation everyone can be made better off as a result
of trade.9
8See Professor Viner's interesting remarks in his Studies in the Theory of International
Trade (New York, 1937), pp. 533-4.
9Mathematically, subject to preassigned outside prices and with preassigned
quantities of all productive services, there will result optimal production quantities

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions
 The Gains from International Trade 205

I shall make no attempt to construct a numerical index of the gains

of trade. In the simplest case of a single individual, only an ordinal
preference scale is assumed so that only better or worse comparisons can
be made. Such constructs as consumers' surplus are in general in-
admissible. Even in the singular cases where they are able to be em-
ployed, they are perfectly arbitrary and conventional, adding nothing
to the analysis.
[10] In conclusion, I should like to point out that in the above
exposition an attempt has been made to demonstrate rigorously with
little reliance on intuition the truth of the theorems advanced. Whether
or not this should be done is, of course, a matter of taste. Much more
important than the carrying through of the formal steps of the argument
is the realization that the theorems are true consequences of the premises,
and do not rest on presumption or probability. For in pointing out the
consequences of a set of abstract assumptions, one need not be committed
unduly as to the relation between reality and these assumptions. On
the other hand, in advancing a presumption in favour of an undeducible
proposition, the suggestion is conveyed that the difficult task of interpret-
ing reality has already been performed.
PAUL A. SAMUELSON
Society of Fellows,
Harvard University.
which are functions of the preassigned variables and satisfy the production limitation
of equation [1]. These optimal production quantities will sell for the largest possible
total in the outside market, and hence the expression Zp'x is maximized subject to
equation [11and fixed amounts of productive factors. The resulting money sum will be
sufficient to permit consumption of goods obeying the condition that all imports must
be paid for by exports, or =p'x=maximum lp'x. Because production is optimal, the
result is more (or equal) consumption of every good. Moreover, for sufficiently small
reductions of all production services, it will still be possible to have more of every
commodity, and hence the truth of the third statement follows.

This content downloaded from 128.210.126.199 on Fri, 23 Oct 2015 04:39:29 UTC
All use subject to JSTOR Terms and Conditions