ELSEVIER Journal of International Economics 39 (1995) 201-226
Intra-industry trade: A Heckscher-Ohlin-Ricardo
approach
Donald R. Davis
Department of Economics, Harvard University, Cambridge, MA 02138, USA
Received February 1994, revised version received March 1995
Abstract
The large volume of intra-industry trade is often cited asa critical element favoring
trade theories based on increasing returns and imperfect competition over those with
constant returns and perfect competition. The former provide an elegant account of
intra-industry trade, while the latter, it is often argued, cannot. This paper provides
an account of intra-industry trade basedsquarely on comparative advantage. The key
is to introduce elements of Ricardian trade theory within the Heckscher-Ohlin
framework. This is appropriate, as essential characteristics of intra-industry trade
imply that technical differences matter. Increasing returns, in short, are not
necessary for intra-industry trade.
Key words: Intra-industry trade; Heckscher-Ohlin; Ricardian; Technological differences;
Trade patterns
JEL classification: Fll
1. Introduction
In a recent survey article, Learner (1992) noted that since the Leontief
paradox only one empirical finding fundamentally altered the way economists
view the causes of international trade-the demonstration by Grubel and
Lloyd (1975) of the importance of intra-industry trade. The content of this
change is indicated by Helpman and Krugman (1989), who cite the
prevalence of intra-industry trade as “one of the key empirical reasons for
emphasizing the role of increasing returns and imperfect competition in the
world economy”.
0022-1996/95/$09.50 0 1995Elsevier Science B.V. All rights reserved
SSDI 0022-1996(95)01383-O
202 D.R. Davis I Journal of International Economics 39 (1995) 201-226
The traditional theories of comparative advantage, based on constant
returns to scale and perfect competition, were perceived to be incapable of
explaining intra-industry trade, By contrast, the scale economies model
offers a very simple account: even absent comparative advantage, scale
economies motivate specialization within industries and so intra-industry
trade. The elegance and power of the theory have earned it wide accept-
ance.
Empirical verification of the role of scale economies in giving rise to
intra-industry trade, however, has proven elusive. Tests based on the
Grubel-Lloyd measure of intra-industry trade have consistently shown a
significant negative relation between intra-industry trade and proxies for
scale economies.’ A recent test seeking to account for import shares by
proxies for scale economies found a positive relation by some measures and
a negative relation by another.* The evidence advanced in Helpman’s (1988)
study of fourteen developed countries does not distinguish between a variety
of models with specialization.3 In sum, the direct empirical support of the
scale economies theory is, at best, mixed.
In this paper, I argue that the conditions that characterize intra-industry
trade in fact allow an extremely simple account of this trade based on
comparative advantage. The essential insight is that these conditions bring
to the fore traditional Ricardian determinants of trade. I develop this in a
setting that allows for both Ricardian and Heckscher-Ohlin influences,
providing a unified account of intra-industry and inter-industry trade.
The logic is simply outlined. The key to inducing intra-industry trade is to
have intra-industry specialization acrosscountries. Scale economies provide
one reason for such specialization. Yet they are not the only reason. The
1See Loertscher and Wolter (1980) Caves (1981), Balassa and Bauwens (1987) and Marvel
and Ray (1987). Loertscher and Wolter had anticipated a positive coefficient based on a
contrast between constant returns sectors, lacking intra-industry trade, with scale economies
sectors, where it should be prevalent. Ex post they surmised that if scale economies are too
strong that production might need to be concentrated in one or a few locales, inhibiting
intra-industry trade, a suggestion that has been followed by Caves and Balassa and Bauwens.
Marvel and Ray demurred, arguing that, “It is an interesting commentary on the power of ideas
over evidence that Loertscher and Wolter were willing to reinterpret their variables in light of
their results in order to support the monopolistic competition model.” I view it more
cautiously. Theory suggests that some degree of scale economies is necessary to induce
specialization and trade; beyond this there should be a range in which scale economies are
unrelated to the volume of intra-industry trade; and only when the scale economies force great
concentration of production should it start to reduce this trade. It is perhaps not surprising that
the cross-industry regressions ultimately shed little light on this. In any case, the lack of a
theoretical alternative to the scale economies approach made the reinterpretation of the theory
to conform with the scale economies approach a natural exercise.
‘See Harrigan (1991).
3See Markusen and Wigle (1990) and Learner (1992).
D.R. Davis I Journal of International Economics 39 (1995) 201-226 203
emphasis on specialization is suggestive of Ricardian trade patterns; but
consideration of technical differences as a reason for trade requires some
motivation specific to the problem of intra-industry trade.
The essence of Ricardian theory is that technical differences matter for
trade patterns when expansion of an individual sector does not drive up
marginal opportunity costs. Two characteristics make intra-industry trade
precisely such a setting. The first is the very definition of intra-industry
trade-trade in goods of similar factor intensity. It is evident that substitution
possibilities acrosssuch goods in production will be excellent. The second is
the emphasis within the intra-industry trade literature on the large number
of goods produced and traded. When the number of goods is large relative
to the number of factors, some sectors may be expanded and others
contracted without rising marginal opportunity cost. Both of these characteristics
of intra-industry trade thus suggest the relevance of Ricardian
determinants of trade.
The theoretical problem addressed may also have important implications
for trade policy. The empirical importance of intra-industry trade, and the
perceived inability of the traditional theories to account for it, have been
taken as key pieces of evidence in favor of the increasing returns theory.
Increasing returns, of course, are often associated with imperfect competition.
Yet I show that intra-industry trade could arise, per the traditional
theories, even if returns to scale are constant and markets perfectly
competitive. A determination of the cause of this trade would then have
important implications for appropriate trade policy, which depends critically
on the underlying market structure.
In addition, this account of intra-industry trade provides a sound theoretical
foundation for the large body of empirical work, including CGE
modeling, that has employed the assumptions of constant returns and
perfect competitionP
The organization of this paper is as follows. Section 2 reviews prior
discussions of intra-industry trade and comparative advantage. Section 3
shows that the very definition of intra-industry trade suggeststhe relevance
of Ricardian determinants of trade. This is developed in an ideal case of
trade in what I call “perfectly-intraindustry” goods. Section 4 develops the
4This empirical work could not ignore intra-industry trade; yet neither could it reconcile such
trade with the assumed theoretical structure. The problem was finessed via the Armington
assumption-that consumers perceive otherwise identical goods produced in different locales to
be different goods. Yet it begged the question of why, for example, “french” goods have to be
produced in France. If this account has some plausibility for perfume or wine, it is plainly
incredible as an account of most intra-industry trade. The Armington assumption was an ad hoc
device to ensure intra-industry specialization, so intra-industry trade (see Markusen and Wigle,
1990). By contrast, the model developed here departs from the standard framework only by
adding features intrinsic to intra-industry trade.
204 D.R. Davis I Journal of International Economics 39 (1995) 201-226
simplest possible framework that can give a unified account of inter-industry
and intra-industry trade, while allowing direct comparability with the
standard factor proportions results. Section 5 then examines the significance
of the emphasis in the intra-industry trade literature on the variety and
number of goods produced. This is shown, as well, to suggestthe relevance
of traditional Ricardian determinants of trade. The final section concludes.
2. Intra-industry trade and comparative advantage
In the last decade, there has been wide acceptance of a claim that
traditional trade theories, based on comparative advantage, cannot account
for observed intra-industry trade. Lancaster (1980) emphasized the large
volume of intra-industry trade between similar countries as a puzzle:
‘Intra-industry trade on a large scale, an undeniable fact of trade
between modem industrial economies, is simply not a prediction of
traditional trade theories.’
Helpman and Krugman (1985) stress the trade in goods of similar factor
content as a mystery:
‘... trade patterns seem to include substantial two-way trade in goods
of similar factor intensity. This ‘intraindustry’ trade seems both
pointless and hard to explain from the point of view of a conventional
trade analysis.’
Balassa and Bauwens (1988) make the claim bluntly:
‘Theorists of intra-industry trade hold that economies of scale are a
sine qua non condition of intra-industry specialization; in the absence
of scale economies, all product varieties could be produced domestically
and no intra-industry trade would take place.’
Krugman and Obstfeld (1991) extend this explicitly to the Ricardian model:
‘... the Ricardian model neglects the possible role of economies of
scale as a cause of trade, which leaves it unable to explain the large
trade flows between apparently similar nations ...’
The claims may be synthesized: traditional trade theories have difficulty
accounting for a large volume of trade between countries which are similar
in factor endowments and technology, a large share of which is in “differentiated”
goods of similar factor intensity.
2.1. Previous efforts
Previous efforts to reconcile observed intra-industry trade with the theory
D.R. Davis I Journal of International Economics 39 (1995) 201-226 205
of comparative advantage have fallen into three main areas. The first has
focused on problems of aggregation. Finger (1975) and Chipman (1985)
argued that existing classifications place goods of heterogeneous factor
proportions in a single industry, and so intra-industry trade is unremarkable.
While all observers acknowledge that actual industrial classification does not
mesh neatly with the theoretical demarcation of industries, most would
argue this does not eliminate the puzzle since intra-industry trade is
important down to quite fine levels of disaggregation. A distinct explanation
offered by Chipman (1988) is based on the theoretical properties of
aggregation. Given any division of N goods into fewer than N industries, it
is possible to find a division of world factor endowments between countries
such that all trade is intra-industry trade. This remains possible even as
countries have very similar endowment proportions. Surprisingly, no restriction
on the basisfor aggregation is required. The weaknessof this account is
that as countries come to have more similar endowments, the volume of
trade converges to zero.
The second approach, developed by Falvey and Kierzkowski (1987),
attempts to divorce the question of trade in differentiated products from the
strong assumption that all varieties are produced under identical technical
conditions. In their model, goods are distinguished on the demand side
according to perceived quality, and on the production side by the fact that
high quality goods are produced under conditions of greater capital intensity.
While this accounts for trade in differentiated goods, some such as
Brander (1987) have noted there is “some question as to whether the
phenomenon Falvey models should be regarded as intra-industry trade”. In
effect, the question of trade in goods of similar factor intensity is side-
stepped.
A third strand-the one on which the present paper builds-relies on the
substitution possibilities across goods inherent in intra-industry trade. The
origins of this work may be traced to Samuelson’s (1948) article on factor
price equalization, in which he noted that if both goods have the samefactor
intensity, then the transformation curve is linear. This is continued in the
two-good, two-factor setting by Lizondo, Johnson and Yeh (1981) who
demonstrate that as production functions become more similar, the PPF
becomes “flatter”. Grubel and Lloyd (1975, p.88) recognized the relevance
of the shape of the PPF, and that this would make trade sensitive to
technical differences, but then proceeded to bury the essential insight amidst
a discursion on product cycle and other neo-technology theories of trade.
More recently, Chipman (1985) and Rodgers (1988) have examined this in a
setting of higher dimension of both goods and factors, by grouping goods
into industries based on a Euclidean metric defined on the Cobb-Douglas
production elasticities. They are then equipped to carry out a variety of
comparative statics, including a demonstration that as the production
functions of goods in an industry become closer (by the metric), intra-
206 D.R. Davis I Journal of International Economics 39 (1995) 201-226
industry trade rises. Finally, this strand also includes the important, and
neglected, piece by Vanek and Bertrand (1971). They noted that if there are
more goods than factors, the world transformation surface has flats, which
makes trade sensitive to technical differences. This is discussed further in
Section 5.
In thinking about the sources of intra-industry trade, it is convenient to
begin with an ideal case. As noted above, the theoretical literature defines
intra-~dust~ trade as exchange of goods of similar factor intensity.5
Consider, then, trade in goods which, at any common factor prices, have
identical factor intensity. Call these “perfectly-intraindustry” goodsP The
essential properties of pe~~tly-intr~ndust~ goods may be appreciated by
focusing on the position of the respective isoquants. The key is ‘that the
isoquants of perfectly-intraindustry goods differ by a Hicks-neutral shift. If
we allow cross-country technical differences which are themselves Hicksneutral,
and which are non-u~fo~ across goods, then the ma~itude of
that shift differs across countries. The fact that the isoquants of perfectlyintraindustry
goods differ by a Hicks-neutral shift means that their marginal
rate of transformation is constant. If the magnitude of that shift differs
acrosscountries, then so must the constant marginal rates of tr~sfo~ation.
As this suggests, the paradigmatic example of intra-industry trade is the
textbook Ricardian model.7 Since there is a single factor, the goods are
trivially perfectly-intraindustry goods. Still, two essential and related charac’
See quotes above from Helpman and Krugman, and Brander. Actual industrial classification
includes similarity of characteristics in both production and consumption (cf. Grub& and
Lloyd, 1975, p.84 and Lloyd in Tharakan, 1989a, ~~‘19-22). However, the traditions trade
theories are largely agnostic regarding substitution possibilities in consumption (emphasizing
only homotheticity of preferences). Thus this poses no challenge to the traditional theories.
This is why the literature has focused, as noted, on intra-industry trade as trade in goods of
similar factor intensity.
6Virtually all of the increasing returns literature on intra-industry trade has used the caseof
perfectly-intraindustry goods-either by positing a single factor in this sector or by having all
varieties produced with identical production functions. SeeKrugman (1979), Lancaster (1980),
and Helpman and Krugman (1985, chapters 7 and 8). The latter assumption has at times been
justified by examples of varieties such as trade in red and green pencils. It should be kept in
mind, however, that the empirical work that has motivated this theoretical work has maintained
a significant degree of aggregation. Even at what is considered a highly disaggregated level-say
five digit SITC (e.g. “photocopying appliances “)-the assumption of identical technologies
across varieties is very strong.
‘The Ricardian model is paradigmatic for the constant returns case of intra-industry trade
not only, as emphasized here, because of trade in perfectly-intraindustry goods, but also
because it is a case with more goods than factors, See Section 5,
This can be extended simply to the standard two-good, two-factor,
Heckscher-Ohlin model. As has long been recognized: when both sectors
use the same factor intensity, the transformation curve is linear. With
perfectly-intraindustry goods, factor intensities in the two sectors will always
be the same, at any factor prices, and for any factor endowments. If there
are small Hicks-neutral technical differences between countries, which are
non-uniform across goods, then there must be specialization in at least one
country, and so trade. This insight extends without alteration to the case
with an arbitrary number of goods and factors (Ethier, 1984). Let there be N
goods and M factors. Consider one country’s matrix of optimal technical
coefficients:
D.R. Davis I Journal of International Economics 39 (1995) 201-226 207
teristics will extend to a more general setting: as production is shifted from
one good to the other, there is no rising marginal opportunity cost. Because
of this, technical differences-however small-induce specialization in at least
one of the countries, and trade.*
Each column of A(o) represents the optimal unit input coefficients for a
single good. If the columns of A(w) are linearly independent, then it is
impossible, by definition, to express any one column asa linear combination
of other columns. That is, at constant factor prices (and so unchanged
coefficients), it is literally impossible for the other sectors to release factors
in exactly the proportion in which they are used in an expanding sector.
Expansion of the production of any one good requires that it bid up the
price of factors that it uses intensively, and so yields a rising marginal
opportunity cost of the good in terms of any other set of goods. This is what
gives us the typical bowed-out feature of the PPF.
However, if goods 1 and 2 are perfectly-intraindustry goods, this argument
fails. In this case, A i k hA, for some A. If there are cross-country
technical differences in these goods that motivate the expansion of one of
these sectors, the other can release factors in precisely the proportion used
in the expanding sector, and so without rising marginal opportunity cost.
Thus technical differences-however small-will induce specialization and
trade.
’ Note that, in principle, this implies neither more nor lessspecialization than that associated
with increasing returns technologies.
9See, for example, Samuelson (1948).
208 D.R. Davis I Journal of International Economics 39 (1995) 201-226
4. The Heckscher-Ohlin-Ricardo model
This section develops the simplest possible model that gives a unified
account of intra-industry and inter-industry trade, while allowing a clear
contrast with the standard factor proportions results.” This requires a model
with three goods: two to represent intra-industry goods, and one the other
industry. For comparability with standard results, we require a model with
two factors, capital and labor. We allow there to exist arbitrarily small
cross-country technological differences.
The interaction of factor proportions and technical differences in giving
rise to the trade pattern leads us to call this the Heckscher-Ohlin-Ricardo
model.
4.1. The model
There are three goods, XI, X2 and Y. The first two are perfectlyintraindustry
goods, which means that for all factor price ratios they are
produced under identical factor intensity.” This allows us to unambiguously
make the assumption that the intra-industry goods are capital-intensive
relative to Y. We assume that technologies are identical across countries in
X2 and Y, with a small Hicks-neutral productivity difference in XI, as
reflected in:
Country One:
Country Two:
where we assume A > 1. Let preferences be identical and homothetic.”
Good X2 is the numeraire.
“The extension of the fundamental result-the possibility of accounting for intra-industry
trade between countries with similar endowment proportions-to a world with many countries,
goods and factors is straightforward.
I1Introduction of neutral Ricardian technical differences to a Heckscher-Ohlin framework
has previously been suggested on empirical grounds by Ivfinhas (1962), and for theoretical
purposes by Bhagwati (1964) and Woodland (1982). Ruffin (1988) devised a distinct admixture
of the theories, supposing countries differ in their endowments of “winemakers” and
“clothmakers”, where technical coefficients differ across types, but not across countries.
However, in each case,they were interested in a set of issuesdistinct from intra-industry trade.
I2An assumption maintained throughout this paper.
D.R. Davis I Journal of International Economics 39 (1995) 201-226 209
4.2. Replicating the integrated equilibrium
Let us begin by developing the integrated equilibrium.13 Since there is an
absolute technical advantage in the production of good XI, only the
technology of country One can be used in production of this good. Since the
marginal rate of transformation between our intra-industry goods is constant,
positive production of both goods requires that P,, = l/A. The
relative price P, emerges from the general equilibrium. This gives rise to
factor intensities kx1 = k,, > k,, and establishes a factor price ratio w/r.
The exact allocation of factors to the various production sectors is determined
.
Let v be the world factor endowment, and (VI, V”) = [(K’, L’), (K*, L”)]
be a partition. Let v(i) = [@(&x(i))] be the integrated equilibrium use of
factors in sector i. The A, are shares for country j of the integrated
equilibrium production of good i. The Factor Price Equalization (FPE) set
in this case is:
FPE = [(V’,V*)&,, &y22, Av,,Ay2, 20
such that AxZI+ Ax,* = 1, A,1+ A,* = 1, AxI = 1, AXI = 0,
V’ = xA,,V(i) V* = xAi2V(i)
I I >
This says that FPE is the set of partitions of world factor endowments,
such that when employing the integrated equilibrium techniques:
l Production of the goods in which technologies are identical (X2 and Y)
can be apportioned among the countries.
l Country One can produce the entire integrated equilibrium supply of
the good X, in which it has a technical advantage.
l This is consistent with full employment of factors in each country.
This has a simple geometric interpretation (seeFig. 1). Country One must
have sufficient factors to produce the integrated equilibrium supply of X,
I3The integrated equilibrium has been defined as the resource allocation that would occur if
both goods and factors were perfectly mobile. An obvious element of the integrated
equilibrium is factor price equalization. Consideration of the integrated equilibrium allows us to
construct the set of partitions of world factor endowments that allow countries to attain all of
the benefits of the fully integrated world by trade in goods alone. This will be called the Factor
Price Equalization set. For our purposes, the convenience is that this also allows a simple
analysis of trade patterns. The conception of trade as replicating this ideal construct appeared
as a parable in Samuelson (1949). Dixit and Norman (1980) coined the term, and showed that
the concept could be a powerful analytic tool. Helpman and Krugman (1985) followed this
lead, making it a central element in their extremely lucid presentation of both standard and new
trade theory.
210 D.R. Davis I Journal of International Economics 39 (1995) 201-226
The integrated Equilibrium
kxl = kx2
Fig. 1. The integrated equilibrium.
Eing the integrated equilibrium technique. This is reflected by the vector
V(1) that extends from 0, with slope k,, , reflecting total factor usage in
good X,. Taking this factor requirement as a new vertex for country One,
the equilibrium techniques used in production of goods X2 and Y give rise to
cones for the two countries in factor space. Any division of the world factor
endowment that falls within the parallelogram generated by the intersection
of these two cones allows replication of the integrated equilibrium.
D.R. Davis I Jourrial of International Economics 39 (1995) 201-226 211
4.3. The pattern of trade
Our analysis of the pattern of trade14begins in an area near O,, where an
isoincome line continues to be in the FPE set for all points between the
factor intensity ratios (seeFig. 2). We examine the effects on trade patterns
of movements along the isoincome line.15
Within the FPE set, the factor content of consumption is the intersection
of the diagonal and the isoincome line. So along the isoincome line, the
consumption vector is fixed, country One produces all of the world’s X1, and
there is reallocation of production of goods X, and Y. Consider first if the
endowment is at point A, on the vector with slope k, emanating from 0,.
Country Two produces only good Y, which it exports for goods X1 and X,.
This is the case of pure inter-industry trade. As we move away from A,
towards B, country Two begins to produce someX,, but not yet enough for
its own consumption. It still exports good Y for imports of X, and X,.
Before we reach the diagonal, we reach a factor endowment ratio, at B, at
which country Two just attains self-sufficiency in production of X,.‘” So, at
B, country Two exports Y for imports of X,. As yet, there is no intraindustry
trade. We may refer to this asthe caseof partial inter-industry trade.
As we move from B to C, country Two produces more of X, than it
consumes, and so begins to export it. That is, the labor-abundant country is
now exporting one of the capital-intensive goods. This occurs although this
country has no (absolute) technical advantage in that good, and the trade
pattern is fully determinate. Country two remains a net importer of the
intra-industry goods.
When we reach C, each country is self-sufficient in good Y. However,
country Two must import good X, , which it pays for with exports of X,.
Here, when factor endowment ratios are identical, only intra-industry trade
occurs. We call this the case of pure intra-industry trade.
14The format of the analysis in the following sections, and the statement of the standard
factor proportions propositions follows the account in Helpman and Krugman (1985, chapters 2
and 8). It should be kept in mind throughout that all statements regarding the pattern and
volume of trade pertain only to the factor price equalization set.
” If country Two had a large enough share of world income, movements along the isoincome
line could bring us to the edge of the FPE set before we reach the casesof pure intra-industry
trade or heterogeneous trade. However, this would not change the analysis for points within the
FPE set.
I6The X2 self-sufficiency locus is a ray from 0’ within the FPE set. Note that an expansion
along a ray from the origin by a factor A raises output in each sector for country Two by the
same factor A. With homothetic preferences, this must also raise demand in Two for X, by A.
Thus if the initial point was one of zero excessdemand for X,, then so must the new point be
one in which X, is not traded.
D.R. Davis / Journal of International Economics 39 (1995) 201-226
The Pattern of Trade
Fig. 2. The pattern of trade.
As we move from C to D, country Two begins to import Y aswell asX1,
and pays for them with exports of X2. Thus country One is now the
labor-abundant country, and exports one labor-intensive good and one
capital-intensive good. Its exports of the capital-intensive good Xi are
driven by its technical advantage in this good. However it is now a nef
importer of the intra-industry goods.
D.R. Davis I Journal of International Economics 39 (1995) 201-226 213
At D, country Two produces only X,, which it trades for its entire
consumption of both Y and X,. As is evident, there is substantial trade in
both inter-industry and intra-industry goods. Accordingly, I call this the case
of heterogeneous trade.
The results of the Heckscher-Ohlin-Ricardo model on trade patterns are
very significant for the problem of intra-industry trade. In this example,
technical differences are arbitrarily small;l’ intra-industry trade is in goods
of identical factor intensity. Still, when countries have identical endowment
ratios, we found that 100% of trade is intra-industry trade.
4.4. The volume of trade
I now derive the level curves of trade volume. It is analytically convenient
to separate the effects of factor endowments on the volume of inter-industry
trade from the effects on intra-industry trade. Inter-industry trade is twice
the value of the good Y traded (it being paid for with exports of X1 or X,).
Intra-industry trade, then, consists of the value of goods Xi and X, traded
against one another.” The derivations that follow, of course, hold only
within the FPE set.
The four critical factor ratios noted previously define three regions of the
FPE set. Denote the area bounded by the Y factor intensity line emanating
from 0, and the X, self-sufficiency line as Region I. From there to the
diagonal will be Region II. And from the diagonal to the X, factor intensity
line emanating from 0, will be Region III (see Fig. 2).
4.4.1. The volume of inter-industry trade
Helpman and Krugman (1985, p.23) noted that in the standard Heckscher-Ohlin
setting with factor price equalization, the level curves for total
volume of trade are parallel to the diagonal of the box diagram, with
increasing volume of trade as we move from the diagonal. This is entirely
sensible, as the net factor content of trade is constant along these surfaces,
and the volume of trade can be measured by the value of inputs in traded
goods.
This does not hold in the Heckscher-Ohlin-Ricardo model, since trade in
intra-industry goods may occur with zero net factor content. Instead, within
the FPE set, the lines parallel to the diagonal represent a constant value of
inter-industry trade (see Fig. 3). Suppose, for example, that country Two is
labor abundant. Then it will be the only exporter of Y. The volume of
inter-industry trade is then (superscripts reflecting the country):
“This is not to exclude large technical differences on either theoretical or empirical
grounds-rather it is to say one need not believe they are large for them to have important
effects on trade patterns.
I8This is comparable to the measure used by Grubel and Lloyd (1975, p.20).
214 D.R. Davis I Journal of international Economics 39 (1995) 201-226
Level Curves for Volume of Interindustry Trade
Diagonal
Edge of FPE KX2
Fig. 3. Level curves for volume of inter-industry trade.
VTinter= 2(Y2 - s*q
Within the FPE set Y* and s* are linear functions of endowments, so level
curves here form a set of parallel lines. Since the diagonal is the level curve
with zero inter-industry trade, the other level curves must be parallel.
D.R. Davis I Journal of International Economics 39 (1995) 201-226 215
4.4.2. The volume of intra-industry trade
Now we look at the determinants of the volume of intra-industry trade. It
is simplest if we consider this first in the case of pure intra-industry trade,
i.e. when the endowment is on the diagonal. Inside the FPE set, and with
homothetic preferences, expenditure shares on the goods are fixed and
equal across countries. Along the diagonal, Y is not traded. Thus the only
trade is Two’s exports of X, for imports of X, from One. Balanced trade
implies, then, that we can measure the value of trade as twice the value of
Two’s consumption of X1 (since it produces no X,). Thus the volume of
trade rises linearly with Two’s income, so long as this is consistent with
being in the factor price equalization set. Thus, considering for the moment
only the diagonal, the volume of intra-industry trade within this equilibrium
reaches its maximum at the interior boundary of the FPE set.
Now we investigate the volume of intra-industry trade off of the diagonal.
Recall that the volume of intra-industry trade is defined here asthe value of
direct exchange of goods X, and X,. We have already noted that in Region I
there is no intra-industry trade. So we turn to Region II. We are interested
in finding the level curves for intra-industry trade. In Region II, we can
measure intra-industry trade as twice the value of country Two’s excess
supply of X,. Since within Region 2, country Two’s excesssupply of X, is
fixed on a line parallel to the X, self-sufficiency line, these must be the level
curves of intra-industry trade in this region. Finally, we turn to Region III.
Here intra-industry trade consists of exports of X, from country Two in
exchange for imports of X,. Along an isoincome line, imports of X, are
fixed, and these can only be paid for by exports of X,. Thus the isoincome
line in Region III must itself be the level curve for the volume of intraindustry
trade (seeFig. 4). A direct consequence of the foregoing analysis is
that within the FPE set, the value of intra-industry trade attains a maximum
when countries have identical endowment proportions. The proof is straightforward.
Its geometric sense can be appreciated by examining Fig. 1. We
have already noted that the level curves for intra-industry trade are parallels
to the X, self-sufficiency factor ratio, with increasing volume of trade as we
move toward the diagonal. Since, necessarily, k, < slope of X, self-sufficiency
-=cratio of world endowments, it follows that intra-industry trade
reaches a maximum at the interior intersection of the diagonal and the FPE
set. That is, within the FPE set, intra-industry trade is maximized at a point
where the countries have identical factor endowment ratios.
4.4.3. Aggregate volume of trade
The aggregate volume of trade in the Heckscher-Ohlin-Ricardo model is
readily ascertained. It may be taken as the sum of exports acrosscountries,
or with balanced trade, as twice the exports of either country. We begin by
examining the role of similarity of endowment proportions, and then move
216 D.R. Davis I Journal of International Economics 39 (1995) 201-226
Level Curves for Volume of lntraindustty Trade
kx2
Fig. 4. Level curves for volume of intra-industry trade.
on to consider the role of relative income levels on trade volume {see Fig.
5).
It is convenient to divide the analysis according to the three principal
regions in the FPE set. The pattern of trade has already been derived above.
In Region I, between the k, slope and the X, self-sufficiency line, country
Two exports only Y. Thus the volume of trade in this region is:
VT = 2(Y2 - s2F)
D.R. Davis I Journal of International Economics 39 (1995) 201-226 217
Level Curves for Total Volume of Trade
ge-of FPE kx2
Fig. 5. Level curves for total volume of trade.
A move along the isoincome line towards the diagonal leaves consumption
unchanged, but reduces Two’s production of Y. Thus the trade volume falls
as per the standard theory.
In Region II, between the X2 self-sufficiency line and the diagonal,
country Two exports X2 and Y for imports of X1 only. Thus the volume of
trade is given by:
VT = 2s2y 1
This is clearly constant for any move along the isoincome line. The fall in
218 D.R. Davis I Journal of International Economics 39 (1995) 201-226
inter-industry trade is exactly offset by a rive in intra-industry trade, just as
the countries are coming to have more similar endowment proportions. This
stands in contrast to the standard factor proportions theory, and demonstrates
simply that a process of countries coming to have more similar
endowment proportions can raise intra-industry trade in a world in which
trade arises due to comparative advantage.
In Region III, between the diagonal and the line with slope k,, country
Two is importing Y and Xi, so exports only X2. The volume of trade, then,
is given by:
VT = 2(X; - s’?,)
Clearly, this again is rising as we move along an isoincome line to the edge
of the FPE set, with relatively more dissimilar endowment proportions.
Now we turn to examining the role of relative incomes in determining
trade volume. In Region I there is only inter-industry trade, and so the trade
volume is constant along parallels to the diagonal. In this sense, relative
income does not matter for trade volume in Region I, as in the traditional
theory.
In Region II, the isotrade volume lines are the isoincome lines. Trade
volume rises as the share of country Two in income rises, as per the
equation above. The fact that income matters here is due to the specialization
induced by the special features of the Heckscher-Ohlin-Ricardo
model. The fact that trade volume increases monotonically in country Two’s
share of income arises because of the special assumption that only one
country has a technical advantage in any good.
The level curves for total volume of trade in Region III do not have as
simple an explanation, although the basic tradeoff is clear. Recall that in
Region III, level curves for inter-industry trade are along parallels to the
diagonal, while those for intra-industry trade correspond to isoincome lines.
The level curves for total trade can be shown to depend on the world
endowment ratio, the factor intensity in the X2 sector, the expenditure
shares on each good and the factor price ratio. In any case, the level curves
must have a slope greater than that of the isoincome lines (which have slope
- w/r).
4.5. Contrast with standard Heckscher-Ohlin results
We now contrast the results of the Heckscher-Ohlin-Ricardo model with
the standard results. The Heckscher-Ohlin model yields a simple prediction
on the pattern of trade: a country exports the good that usesintensively its
relatively abundant factor of production. An important corollary is that
countries with identical factor endowment ratios do not trade. When there
D.R. Davis I Journal of Internotional Economics 39 (1995) 201-226 219
are more goods than factors, the model still predicts that the net factor
content of exports reflects relative factor abundance. The caveat is that in
this case, the direction of trade in a particular good is not determinate.
The predictions of the Heckscher-Ohlin-Ricardo model share some of
these properties, although with important qualifications. In some cases,the
predictions diverge strongly. The weaker condition that the net factor
content of trade is predicted by relative factor abundance continues to hold
fully in the new setting. However, there is an ambiguity in the stronger
assertion. The problem is that now there are two capital-intensive goods, X,
and X,. It is quite possible that the labor abundant country exports one of
the capital-intensive goods. Note, though, that unlike the case of more
goods than factors with identical technologies, this has nothing to do with
indeterminacy in the pattern of production. Trade is fully determinate here.
It can arise from two possibilities. First, if One is the labor abundant
country, but has an absolute (so comparative) advantage in production of
good Xi, it produces the world supply of that good, so must export it. This
corresponds to trade patterns in Region III (seeFig. 2). Second, even if One
is the capital abundant country, since it must produce the entire world
supply of X,, its factor endowment net of this may make it the “labor
abundant” country in the residual factor endowments employed to produce
the goods with common technologies. Labor abundant country Two would
then export capital-intensive good X,. This occurs in Region II (seeFig. 2).
None the less, the net flow of capital-intensive goods continues to reflect
factor abundance. In strong contrast to the standard Heckscher-Ohlin
framework, here even countries with identical endowment ratios may
engage in substantial trade.
The standard Heckscher-Ohlin model yields two important predictions
regarding the relation between factor endowments and the volume of
trade.lg The first is that, cet. par., the greater the difference in factor
endowment ratios, the greater the volume of trade. The second is that (in a
sense to be made precise) relative country size has no effect on the volume
of trade .“”
By contrast, in the Heckscher-Ohlin-Ricardo model, these propositions
need not hold. These predictions continue to hold exactly as a prediction of
inter-industry trade. However, when we measure the total volume of trade,
including intra-industry trade, the effects of endowment shifts on trade
volume depend crucially on the region in which the endowments lie. If we
interpret an increase in the difference in factor endowment ratios as a
movement away from the diagonal along an isoincome line, then greater
differences in endowment ratios raise trade volume when in Regions I or III
l9 The world volume of trade is defmed as the sum of exports across countries,
TOSee Helpman and Krugman (1985, section 1.5).
220 D.R. Davis I Journal of International Economics 39 (1995) 201-226
(see Fig. 2). However, such differences in endowment ratios have no impact
on trade volume in Region II. In this region, the rise in inter-industry trade
is fully offset by a fall in intra-industry trade. If we interpret a change in
relative country size as a movement parallel to the diagonal, then relative
country size leaves the volume of trade unchanged in Region I. However, in
Regions II and III, country size does matter.
A striking contrast between the predictions of the standard analysis and
that in the Heckscher-Ohlin-Ricardo model is that the standard model
predicts that countries with identical factor ratios do not trade. In the
present model, such countries have no inter-industry trade. However, there
may yet be intra-industry trade, which then represents 100% of trade. In
fact, within the FPE set, the volume of intra-industry trade reaches a
maximum at a point where countries have identical endowment ratios.
5. The role of product variety and number
An important theme within the literature on intra-industry trade has been
the great variety and number of goods produced and traded.*l This has been
taken as additional confirmation of the importance of increasing returns to
scale in explaining intra-industry trade. A representative statement comes
from Corden (1979, p.9):
‘One can argue that intra-industry trade is visibly a manifestation of
product differentiation, and that product differentiation must make
economies of scale more important.’
However, an important feature of this case has gone almost wholly
unnoted. This is the fact that a rise in the number of traded products may
also bring to the fore Ricardian determinants of trade. As is well known, the
principal results of the Heckscher-Ohlin theory on the pattern of trade are
very sensitive to the number of goods relative to the number of factors
(Bhagwati, 1972; Samuelson, 1953-1954). I will take the emphasis on trade
in differentiated goods to suggestthat the number of goods is large relative
to the number of factors.22In this case, if technologies are identical and
factor prices equalized, then the pattern of trade is indeterminate. The
indeterminacy that arisesunder the stated conditions, however, indicates the
relevance of Ricardian determinants of trade. The indeterminacy arises
‘l For example, Grubel and Lloyd subtitle their book on intra-industry trade, “The theory
and measurement of international trade in ditierentiated products” [emphasis added].
” Such an interpretation is consistent with the frequent use in intra-industry trade theory of
models with a continuum of goods and one or a small hnite number of factors. See, for
example, Krugman (1979). Rader (1979) argues for the relevance of the more-goods-thanfactors
case.
D.R. Davis I Journal of International Economics 39 (1995) 201-226 221
precisely due to the excellent substitution possibilities across goods in
production. To see this clearly, let us return to our matrix of optimal
technical coefficients for a single country:
A(w)=[A, A, . . .
When goods outnumber factors, there is necessarily linear dependence in
the columns of the A(o) matrix-some of the columns may be expressed asa
linear combination of other columns. Then in shifting production between
one good and the corresponding set of other goods, there is no rising
marginal opportunity cost until at least one of the goods is no longer
produced. Thus, technical differences-however small-here also induce
specialization and trade.23
It is straightforward to show that this could give rise to intra-industry
trade. Consider a world with two countries. There are N goods and M
factors, with N > M. The countries share exactly M technologies, and
N - M technologies differ.24 The technical differences are assumed to be
Hicks-neutral. I assume that when employing the integrated equilibrium
techniques, the vectors of factor usage of the M goods with identical
technologies are linearly independent. Preferences are identical and
homothetic.
The Factor Price Equalization Set is given as:
FPE = [(V’, V2))3hij 2 0 such that
c A,= 1 vi EZS )
i
hi1 = 1 Ai, =0 Vi EZ’
Ail=0 A,,=1 ViEZ’
Vi=c Aijv(i) j=l,2
I I
23That indeterminacies in production may arise with N > M hasbeen widely recognized. This
result has largely been regarded as a curiosity or a nuisance, which would be resolved, for
example, by introducing transport costs. However, the essential insight exploited here was first
noted, I believe, by Vanek and Bertrand (1971) in three terse paragraphs at the end of their
article.
*4If we relax the assumption that there are exactly M identical technologies (and so allow
more than M identical technologies), an element of indeterminacy in the pattern of trade of
individual goods with identical technologies may re-emerge. The factor price equalization set
will have full dimension so long as there are M vectors of optimal factor employment among
these M’ goods which are linearly independent.
222 D.R. Davb I Journal of International Economics 39 (1995) 201-226
This says that FPE is the set of partitions of world factor endowments
between the two countries, such that when employing the integrated equilibrium
techniques :
l Production of the goods with identical technologies may be divided
between the countries.
l Each country produces the entire integrated equilibrium supply of the
goods in which it has a technical advantage.
l This is consistent with full employment of factors in both countries.
The geometric intuition carries over to this case. When technologies
differ, only the efficient technology can be employed if we are to replicate
the integrated equilibrium. Thus, each country must have sufficient factors
to produce, using the integrated equilibrium techniques, the integrated
equilibrium levels of output of the goods in which it has an absolute
technical advantage. Taking these initial factor requirements as new vertices,
the A4 technologies which are common to the two countries then
define an M-dimensional cone, reflecting the integrated equilibrium techniques,
for each of the two countries. Then, any division of the world factor
endowment that lies within both cones insures that trade replicates the
integrated equilibrium. If the endowment lies strictly within each cone,
there is diversification in production of all goods with common technologies;
there is specialization in all goods in which technologies differ.
The possibility of explaining intra-industry trade in this setting is immediate.
No restrictions whatsoever have been placed on the factor intensities of
the goods in which production is specialized. It is perfectly possible that
these are goods of similar factor intensity. Since both countries consume the
goods in equal proportions, there must be trade in these goods.
An important special case is when the countries have identical factor
endowment proportions (i.e. their endowment vectors differ by a scalar),
and are within the M-dimensional cones. Of necessity, there is still trade in
all goods in which technologies differ. As well, there is trade in at least some
of the goods in which technologies are identical, except in the knife-edge
casewhere the net factor content of trade in the goods in which technologies
differ is zero.
The key point to recognize is that the emphasis in the intra-industry trade
literature on the large number of goods produced and traded also suggests
an interpretation of this trade based on the traditional Ricardian theory of
comparative advantage. The literature has emphasized that when there are
more goods than factors, factor prices are equalized and technologies are
identical across countries, the pattern of trade is indeterminate. Intraindustry
trade seems pointless. This emphasis, however, has been misplaced.
Rather we should see that this is a setting in which technical
D.R. Davis I Journal of International Economics 39 (1995) 201-226 223
differences may be decisive in determining the pattern and volume of trade.
Here also, substantial intra-industry trade could arise from simple Ricardian
comparative advantage.
6. Conclusion
Growing appreciation of the importance of intra-industry trade has made
its explanation an important theoretical objective. In the last decade there
has been wide acceptance of a claim that increasing returns are necessaryto
account for this trade.25
This paper has shown that-to the contrary-intra-industry trade arises
quite naturally in a constant returns setting. Intra-industry trade is shown to
bear a special relation to traditional Ricardian determinants of trade. These
are relevant because essential elements of the intra-industry setting-the fact
that it is trade in goods of similar factor intensity, and the focus on the large
number of goods produced and traded-give rise to excellent substitution
possibilities across goods in production. In such a setting, small technical
differences induce specialization and trade.
Z5An additional factor often cited in favor of increasing returns models is the normative
impact of trade liberalization (e.g. Helpman and Krugman, 1985, p.3). In a Heckscher-Ohlin
framework, the Stolper-Samuelson theorem suggests that even as trade generates aggregate
gains, that absent redistribution one of the two factors must suffer a decline in real income. If
there is integration among identical countries, no factor need lose, but this is becausethere will
be no trade. However, it is suggested, modem experience belies this. Integration among like
countries has generated nearly universal gains, so engendered little significant opposition;
integration among dissimilar countries, by contrast, has spawned dissension. Scale economies
offer a particularly attractive explanation of the gains from integration of like countries, since it
is possible to reap the gains of scale economies with little or no sectoral reallocation of
resources. The same, however, is true in the general Heckscher-Ohlin-Ricardo model. Gains
here come from two sources: sectoral reallocation corresponding to conventional HeckscherOhlin
gains, and increased efficiency of use of existing resources, as integration allows greater
specialization on Ricardian grounds for a fixed intersectoral distribution. If actual technical
differences were very small, then so too would be the associated gains from integration of
identical countries. However, the assumption of small technical differences was not an
empirical assertion, but rather to establish that small differences are sufficient in the theoretical
model. Moreover, they have an exact parallel in the degree of economies of scale. If this is
sufficiently small, then so too will be the gains from integration of like countries. In the
monopolistic competition framework, some of these gains in the increased efficiency of use of
existing resources may come in the form of gains from increased variety. At times it is also
suggested that this is a reason to prefer this framework to competitive models. However, it
should be evident from the work of Young (1991) that once exact symmetry conditions are
abandoned, either on the cost or demand side, that similar gains from increased variety of
goods are quite consistent with the competitive, constant returns to scale framework. In sum,
the modem experience of integration does not on its face provide a reason to prefer the scale
economies model.
224 D.R. Davis I Journal of International Economics 39 (1995) 201-226
This was first examined in an ideal case-accounting for trade in perfectlyintraindustry
goods. I then developed the simplest model that gives a unified
account of inter-industry and intra-industry trade, the Heckscher-OhlinRicardo
model. Finally, I showed that the emphasis in the literature on the
variety and number of goods produced and traded likewise suggests
Ricardian determinants of trade, and the possibility of substantial intraindustry
trade.
The distinction between the increasing and constant returns theories has
direct relevance for trade policy. Increasing returns are typically associated
with imperfect competition, and so often invite government intervention in
trade. Constant returns are associated with perfect competition, and (other
distortions absent) suggest a policy of free trade.
Increasing returns, in short, are not necessaryto account for intra-industry
trade.26
Acknowledgments
I would like to thank Jagdish Bhagwati, Andre Burgstaller, Ron Findlay,
Mike Gavin, Tatsuo Hatta, Doug Irwin, and two anonymous referees for
helpful comments. I would also like to thank participants in seminars at
MIT, Harvard, Stanford, Pennsylvania, Columbia, Michigan, and the
University of Chicago School of Business. A previous version of this paper
appeared in the International Finance Discussion Papers series of the Board
of Governors of the Federal Reserve, no. 411. The views expressedhere are
not necessarily those of the Federal Reserve.
References
Balassa, B. and L. Bauwens, 1987, Intra-industry specialisation in a multi-country and
multi-industry framework, Economic Journal 97, 923-939.
Balassa, B. and L. Bauwens, 1988, Changing trade patterns in manufactured goods: An
econometric approach (Elsevier, New York).
Bhagwati, J.N., 1964,The pure theory of international trade: A survey, Economic Journal 74,
l-84.
Bhagwati, J.N., 1972, The Heckscher-Ohlin theorem in the multi-commodity case, Journal of
Political Economy 80, 1052-1055.
Brander, James A., 1987, Book review of Greenaway and Milner (1986), Journal of
International Economics 23:1/2, 182-185.
26The fact that increasing returns are not necessaryto account for intra-industry trade does
not imply that these models lack interest. Rather it suggests that efforts to distinguish these
models empirically must move beyond the facts implied by simple specialization.
D.R. Davis I Journal of International Economics 39 (1995) 201-226 225
Caves, Richard, 1981, Intra-industry trade and market structure in the industrial countries,
Oxford Economic Papers 33, 203-223.
Chipman, John S., 1985, Intra-industry trade, factor proportions, and aggregation, Mimeo.
Chipman, John S., 1988, Intra-industry trade in the Heckscher-Ohlin-Lemer-Samuelson
model, Mimeo, University of Minnesota and Universitat Konstanz.
Corden, W.M., 1979, Intra-industry trade and factor proportions theory, in: H. Giersch, ed.,
On the economics of intra-industry trade: Symposium 1978 (Mohr, Tubingen) 3-17.
Dixit, Avinash K. and V Norman, 1980, The theory of international trade (Cambridge
University Press, Cambridge).
Ethier, W., 1982, National and international returns to scale in the modern theory of
international trade, American Economic Review 72:3.
Ethier, W., 1984, Higher dimensional issues in trade theory, in: Handbook of international
economics, Vol. 1 (Elsevier Science, New York) 131-184.
Falvey, R.E. and H. Kierzkowski, 1987, Product quality, intra-industry trade and (im)perfect
competition, in: H. Kierzkowski, ed., Protection and competition in international trade
(Basil Blackwell, New York) 143-161.
Finger, J.M., 1975, Trade overlap and intra-industry trade, Economic Inquiry 13, 581-589.
Greenaway, D. and C. Milner, 1986, The economics of intra-industry trade (Basil Blackwell,
New York).
Greenaway, D. and C. Milner, 1987,lntra-industry trade: Current perspectives and unresolved
issues, Weltwirtschaftliches Archiv 123:1, 39-57.
Grubel, H.G. and P.J. Lloyd, 1975, Intra-industry trade: The theory and measurement of
international trade in differentiated products (Wiley, New York).
Harrigan, James, 1991, Scale economies and the volume of trade, Mimeo, University of
Pittsburgh.
Helpman, E., 1988, Imperfect competition and international trade: Evidence from fourteen
industrial countries, in: A.M. Spence and H.A. Hazard, eds., International competitiveness
(Ballinger, New York).
Helpman, E. and P. Krugman, 1985, Market structure and foreign trade (MIT, Cambridge,
MA).
Helpman, E. and Krugman, P. 1989, Trade policy and market structure, (MIT, Cambridge,
MA).
Krugman, P.R., 1979, Increasing returns, monopolistic competition, and international trade,
Journal of International Economics 9:4, 469-479.
Krugman, P.R. and Maurice Obstfeld, 1991, International economics: Theory and policy, 2nd
edn. (Harper Collins, New York).
Lancaster, Kelvin, 1980, Intra-industry trade under perfect monopolistic competition, Journal
of International Economics 10, 151-175.
Learner, E., 1992, Testing trade theory, NBER Working Paper No. 3957.
Lizondo, J.S., H.G. Johnson and Y. Yeh, 1981, Factor intensities and the shape of the
production possibility curve, Economica 48, 199-202.
Loertscher, Rudolf and Frank Wolter, 1980, Determinants of intra-industry trade: Among
countries and across industries, Weltwirtschaftliches Archiv 116:2, 280-293.
Markusen, James R. and Randall M. Wigle, 1990, Explaining the volume of North-South
trade, Economic Journal 100, 1206-1215.
Marvel, Howard P. and Edward John Ray, 1987, Intra-industry trade: Sources and effects on
protection, Journal of Political Economy 95:6, 1278-1291.
Minhas, Bagicha, 1962,The homohypallagic production function, factor intensity reversals, and
the Heckscher-Ohlin theorem, Journal of Political Economy 70(2), 138-156.
Rader, T., 1979, Factor price equalization with more industries than factors, in: Jerry Green
and Jose Scheinkman, eds., General equilibrium, growth and trade: Essays in honor of
Lionel McKenzie (Academic Press, New York) 347-354.
226 D.R. Davis / Journal of international Economics 39 (1995) 201-226
Rodgers, Joan R., 1988, Intra-industry trade, aggregation and the HOS model, Journal of
Economic Studies 155, 5-23.
Ruffin, Roy J., 1988, The missing link: The Ricardian approach to the factor endowments
theory of trade, American Economic Review 78(4), 759-772.
Samuelson, P.A., 1948,International trade and equalization of factor prices, Economic Journal
LVIII (230) 163-184.
Samuelson, P.A., 1949, International factor price equalization once again, Economic Journal
LIX (234) 181-197.
Samuelson, P.A., 1952,A comment on factor price equalization, Review of Economic Studies
XIX (2), 121-122.
Samuelson, P.A., 1953,Prices of factors and goods in general equilibrium, Review of Economic
Studies 21:2, l-20.
Tharakan, P.K.M., 1989a,ed., Intra-industry trade: Theory, evidence and extensions (Macmillan,
London).
That&an, P.K.M., 1989b, ed., Intra-industry trade (North-Holland, New York).
Vanek, J. and T. Bertrand, 1971,Trade and factor prices in a multi-commodity world, in: J.N.
Bhagwati, R.W. Jones, R.A. Mundell and J. Vanek, eds., Trade, balance of payments and
growth: Papers in international economics in honor of Charles P. Kindleberger (Elsevier,
New York).
Woodland, A.D., 1982, International trade and resource allocation (North-Holland, New
York).
Young, A., 1991, Learning by doing and the dynamic effects of international trade, Quarterly
Journal of Economics 106:2. 369-405.