Professional Documents
Culture Documents
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 support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic
Review.
http://www.jstor.org
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
Trade in Producer Services and in Other
Specialized Intermediate Inputs
By JAMES R. MARKUSEN*
It has long been observed that over half of tant selling the same blueprints to different
international trade occurs in intermediate firms).
goods. Much of this is in foodstuffs, raw The theory of international trade today
materials, and so forth, but much of the consists of a relatively small literature on
trade in intermediates is also in produced intermediate inputs (for example, Kalyan
intermediate manufacturers. A rapidly grow- Sanyal and Ronald Jones, 1982) and a litera-
ing category of trade in intermediates is pro- ture on scale economies and imperfect com-
ducer services, which include such areas as petition in final goods production (many ref-
management consulting, engineering consult- erences are found in Elhanan Helpman and
ing, banking, insurance, marketing, and fi- Paul Krugman, 1985). The two are combined
nancial services. in Wilfred Ethier (1982), Paul Romer (1987),
Many of the intermediate manufactures and Markusen (1988). These authors have a
and producer services that enter into inter- simple model in which intermediate inputs,
national trade are probably characterized produced with increasing retums to scale are
by significant degrees of scale economies assembled to produce final output.
and/or product differentiation. Factor-in- The purpose of this paper is to develop a
tensity data suggest, for example, that inter- model of trade in differentiated intermediate
mediate manufactured goods are on average inputs along the lines proposed by Ethier. I
significantly more capital-intensive than final will be particularly interested in normative
manufactured goods (James Markusen and issues not dealt with by Ethier (1979, 1982)
James Melvin, 1984b). Many producer ser- and Romer (1987). In the model, each of two
vices are both differentiated and knowledge- countries has a competitive sector (Y) and a
intensive. Knowledge intensity in turn sector (X) which produces a composite good
suggests strong scale economies in that from intermediate inputs or services (S).
knowledge must be acquired at an initial The latter are produced with increasing re-
learning cost, after which the knowledge- turns and are complementary in production.
based services can be provided at a very low The results demonstrate that allowing
marginal cost (as in an engineering consul- trade in specialized inputs is superior to al-
lowing trade in final goods only in two dis-
tinct senses. First, free-input trade guaran-
*Department of Economics, University of Western tees that both countries will be made better
Ontario, London, Canada N6A 5C2. Funded under a off relative to autarky while, due to the dis-
grant from The Institute for Research on Public Policy
and the Department of Regional and Industrial Expan- tortion between prices and marginal costs,
sion, Ottawa. The thorough and constructive comments goods trade only does not guarantee that
of an anonymous referee are gratefully acknowledged. free trade will be Pareto-improving. Free-
85
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
86 THE AMERICAN ECONOMIC REVIEW MARCH 1989
input trade guarantees that both countries services or specializedinputs. The produc-
experience an expansion of production in tion functionsin (2) have constantreturnsto
the distorted sector which has been shown to the S for a fixed n.
be a sufficient condition for gains from trade Each S is producedby a single firmand,
when price exceeds marginal cost (Markusen for simplicity, we assume that the produc-
and Melvin, 1981, 1984a). Free trade in final tion of S. requiresonly labor. S producers
goods only may result in the contraction of and X producersare disjoint,and the Sj are
this sector for the smaller or otherwise producedwith increasingreturnsundertech-
" technically disadvantaged" country and nology and behavioral assumptions dis-
thus the sufficient condition for gains fails to cussed below. Given constantreturnsin (2),
hold. Second, free-input trade is superior to we can then aggregatethe individual firm
free trade in goods from the point of view of outputs to arrive at an industryproduction
the world as a whole, although not necessar- function identical to (2) with the i subscript
ily from the point of view of both countries. deleted. In order to keep the problem
This result follows from the complementar- tractable,we assume that the S production
ity of domestic and foreign specialized in- functions are identical and the S are sym-
puts in final goods production, or alterna- metric but imperfectsubstitutesin produc-
tively from the increased division of labor ing X. Assume in particularthat the X in-
supported by trade. With free-input trade, dustry productionfunctionis given by
each country essentially confers a positive
technological externality on its trading part- (3) X- [s#]1 0 < p < 1.
ner.
These results have a good deal of policy Productionof S. uses only labor, and re-
relevance given the current discussions on quires a fixed cost plus a constantmarginal
liberalizing trade in producer services. These cost. For simplicity,the marginalcost of S
services generally face higher trade barriers is assumed to be one in terms of labor.
than do commodities since the former in- Letting Y be numeraire,the cost of produc-
trude on immigration and foreign invest- ing Sj in terms of Y is then
ment policies which are in turn more restric-
tive than goods-trade policies. (4) wSj + wF,
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
VOL. 79 NO. I MARKUSEN: TRADE IN PRODUCER SER VICES 87
relating to S and n are then given as follows. The first-order condition for Si is
(8) ((1-, )/,B)pnaSj =wF. Multiply the first equation through by Si.
Combining the two equations in (14), we get
But from (6) we see that n= (w/p), so that the values of Si and n at the monopolistic-
the solution values are competition equilibrium.
A
(9) Si 1i F; n =( wlp ) /l (15) Si F;
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
88 THE AMERICAN ECONOMIC RE VIEW MARCH 1989
MP _G1 MRT
(16) MRT - My p
MPX n a
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
VOL. 79 NO. I MARKUSEN: TRADE IN PRODUCER SER VICES 89
the remainder of the paper, we assume that profits would be negative from doing so. The
a is sufficiently small, G is sufficiently con- sufficient condition for gains from trade thus
cave, and demand for X is sufficiently high becomes
such that equilibrium occurs on the con-
cave portion of the production frontier in (25) (PgXa-WgLxa)
Figure 1.
= (Pg-wg(Lxa/Xa))Xa <?
II. GainsfromTrade
iff (Lxa/Xa) ? (Lxg/xg).
Consider first a situation in which there is
free trade in goods only. Let subscripts g This result is shown by first noting that
and a denote quantities evaluated at the free (pg - wg(L /Xg))=0 (zero profits). Since
goods-trade and autarky equilibria, respec- X uses ony labor and since X = n7/S,
tively. Cy and C, will denote consumption of (LX/X) falls with increases in X and (25)
goods Y and X, respectively. By the usual holds if and only if Xg Xa. The latter is a
revealed preference criterion, a country gains sufficient condition for gains from trade.
from trade if The "product expansion condition" Xg >
Xa for gains from trade is by now familiar to
(21) Cyg+ PgCxg2 Cya+ PgCxa. most international trade economists. Intu-
itively, with the price of X (private marginal
In autarky, we have market clearing, while in cost) greater than true social marginal cost,
free trade we have balance-of-payments there is a " Harberger trapezoid" associated
equilibrium. with changes in X output, (p - MC) X,
which is positive if A X> 0. In our case,
(22) Cya= Ya, Cxa=Xa; MC = w/MPx = p,, so (p-MC) = p(1-
i) >0.
Cyg+ pgCxg= Yg+ pgXg. It is also reasonably well known that this
condition may not hold in practice. Figure 2
Substituting (22) into (21), we have shows a situation where the small home
country (superscript h) is at a cost disadvan-
(23) yg + PgXg > Ya + PgXa. tage relative to the large foreign country
(superscript f). Relative to autarky equilib-
Now subtract the value of the factor endow- ria at Ah and A1, country h could lose at the
ment at free-trade prices from both sides of free-trade equilibria Ch, Cf (consumption)
(23), and let k denote the return to Y-sector Qh, Qf (production) due to the distortion
capital, between p and the MRT as shown in Figure
2 and as demonstrated by Markusen and
(24) (Yg-wgLyg -kgK) Melvin (1981, 1984a).
Now we consider free trade in specialized
+ ( pgXg - WLxg) inputs versus autarky, assuming countries
trade S and Y but cannot trade X. Note
with respect to (10)-(14) that free trade in S
> ( Ya- Lya- kgK) equalizes the ri across countries. The fact
that the same number of inputs n = nh + nf
+ (PgXa -
wgLxa). is available in each country at the same
prices implies that they are used in the iden-
The left-hand side of (24) is zero since both tical relative amounts in each country. Con-
industries make zero profits in equilibrium. stant returns to the Si in X then imply that
The first term on the right-hand side of (24) the marginal products of additional inputs
is non-positive: the autarky factor propor- are equal across countries and, from (10)
tion (K/Lya) is not the most efficient way to and (14), that the product price ratio p'=
produce Ya at free-trade factor prices, hence r/na must be equal. Factor prices are also
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
90 THE AMERICAN ECONOMIC REVIEW MARCH 1989
nS
cf
Th B
nh
m
nh
u~~~h
p~~~~
f
a~~~~~~~~~~~
A
T T X 0 ff h f
? na nfm ns nf
FIGURE 2
FIGURE3
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
VOL. 79 NO. I MARKUSEN: TRADE IN PRODUCER SERVICES 91
Factor out p(X)(MPx)/n from both the (37) -1< (dnh/dnf) h<O
numerator and denominator of (32) and can-
iff r <17(1-3),
cel. This leaves
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
92 THE AMERICAN ECONOMIC REVIEW MARCH 1989
greater than minus one are that either r1< than -1 or greater than 0 under restrictions
17(1-fl) or nri* > (1/-1). The same re- (37) or (38).
strictions imply that the slope of f 's reac- These slope restrictions along with n h> n h
tion function is less than minus one or posi- and nf > n f imply that the equilibrium
tive. The first restriction will be satisfied if in Figure 3 must involve nS n + nf >
demand is sufficiently inelastic and scale max(n nf ). More inputs are available with
economies are sufficiently weak (1Bclose to free-input trade than in either country in
1), where we use the term "sufficiently" in autarky. Since the average cost of labor in X
the tautological sense of values that satisfy declines with n, and since ps and ws are the
(37). The second restriction (38) was the free-trade values in both countries, the in-
condition that the production frontier be equality in (29) must hold. Both countries
locally concave in the case of goods trade are assured of gains from trade.3
only, as can be seen by referring back to Now consider the comparison of goods
equation (20) (w* = r* since w = fir). Note trade only with free trade in specialized in-
that, unlike the case of no input trade of puts. Input trade is revealed preferred to
Figures 1 and 2, n = (n h + n f ) never ap- goods trade if
proaches zero along the reaction curve, and
thus nrh > (1/fl -1) may hold globally over (41) CYh+ pCxh Cyhg+ psChg
the reaction curve.
Two sets of reaction curves satisfying the By adding and subtracting the same terms,
restrictions in (37) to (40) are illustrated in the goods-trade balance-of-payments con-
Figure 3. n h and nf are the autarky values straint can be written as
of n in countries h and f, respectively. n h
(n f ) is the n that country h(f ) would pro- (42) + (pscxhg- psc)
Ch + pgCxhg
duce if country f(h) were constrained from
producing any n in the free-trade situation. hxg)-
= yh + pgXh + (px h-px
Positive demand for the other country dic-
tates that nh > nh (nf > nf).
The reaction curves, which give the opti- Substituting this into (41) and replacing the
mal n' for one country given the exogenous left-hand side of (41) with the input-trade
ni for the other country, may be both down- balance-of-payments constraint in (27), (41)
ward sloping or both upward sloping (or becomes
combinations therefore, since r1is not a con-
stant). The reaction curves nhnh for the (43) Ys + rsnhS>yh/+p Xh
home country and nf nf for the foreign
country illustrate the former case. The re- -
strictions (37) and (39) imply that nhnh has +(pg ps)( X-Cxg).
a slope between 0 and -1 in this case and
n nf has a slope less than - 1. Subject to Subtract the value of the factor endow-
the economies being of similar size, the reac- ment at input-trade prices from both sides of
tion curves cross and produce a stable equi- (43) and the left-hand side will be zero as
librium at A in Figure 3. before. The condition for superiority of in-
The reaction curves could both slope up-
ward as illustrated by n hn h' and nf nf' in
Figure 3. An equilibrium occurs at a point 3If the restrictions in (37)-(40) fail to hold, the slope
like B, where one or even both countries of h's reaction function could be less than -1 and that
could be specialized (nhnh' becomes flat, or for f between zero and -1. The reaction functions
nf nf' becomes vertical). Combinations of could cross at a value of nh + nf < max(n , n f )) so that
these cases may occur, but the important one country could be worse off. But such an equilibrium
is presumably unstable under some adjustment mecha-
point for our purposes is that the slope of nism, implying an equilibrium at n or nf in which
h's reaction curve must be greater than mi- case the gains-from-trade result goes through. (37)-(40)
nus one, and the slope of f 's must be less are thus not necessary conditions for bilateral gains.
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
VOL. 79 NO. 1 MARKUSEN: TRADE IN PRODUCER SER VICES 93
(pg - ) ( Xgh- C .
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
94 THE AMERICAN ECONOMIC REVIEW MARCH 1989
increasing returns to n. Because consumers the world as a whole does, however, rely on
are identical and face the same prices, the free trade in inputs equalizing commodity
aggregate preference map in Figure 4 is the prices and hence making trade in X-sector
same as with goods trade. Third, the distor- goods redundant.
tion ratio p,8 = MRT holds along TTs as in This will generally not be the case if the X
the case of TTg and the individual country sector uses other scarce factors of produc-
production frontiers. Finally, the slope of tion that are unevenly endowed between the
TTs must be less at S than the slope of the countries. Non-constant returns in final
point on TTg directly below, G. The marginal goods production would also result in non-
product of labor in Y must be less at the equalization of commodity prices. In these
former (Y, > Yg) and the marginal product of cases, free trade in inputs only and free trade
labor in X greater (X, = Xg, but with less in goods only each offers something that the
labor in X at Xs, it follows that nS > ng and others do not. Free trade in inputs offers
hence MPXs> MPXg).The equilibrium along higher technical productivity in final goods
TTs must then be at a point like S' where production than free trade in goods. But if
Xs 2 Xg if X is non-inferior in consumption. free trade in inputs leaves commodity prices
This last result implies that the total num- unequalized, there exist unexploited gains
ber of inputs produced in the input-trade from the exchange of scarce factor services
equilibrium must exceed the total produced embodied in final goods. Goods trade only
in either country in the goods-trade equilib- captures the gains from exchange but not the
rium. This follows from a simple argument technical productivity gains. In the general
by contraction. If, for example, ng > n =n5 case, it is impossible to say which of the two
+ n f, then Xh> Xs (even ignoring output second-best trade regimes dominates.
from f ) whict contradicts the result that Restrictions implying commodity price
Xg = Xf < Xs Thus n_ (n equalization from input trade only are used,
The proof that nS ? (ng, ngf) is in turn all but not needed to prove the superiority of
that we need to prove that the second term free-input trade to autarky. p( X) in the
in (44) is non-positive. Except for the terms- derivation of the reaction curves could refer
of-trade effect (the third term in (44)), both only to domestic price and demand. A spe-
countries are assured of gains at the input- cific factor in X that is unequal between
trade equilibrium relative to the goods-trade countries could be added to prevent the
equilibrium and one must prefer the former. equalization of r from implying the equal-
Note also if we add the inequality in (44) to ization of p. The reaction curve analysis and
the corresponding one for f (identical homo- the expansion of n over autarky levels will
thetic utility makes this meaningful), the continue to go through. The revealed-prefer-
terms-of-trade effects cancel. The world con- ence analysis culminating in (29) will need
sumption bundle at the service-trade equilib- modification for the additional factor but the
rium is preferred to that in goods-trade equi- same result on X-sector expansion and gains
librium. from trade will emerge. Unlike the case of
goods trade only, free trade in the special-
III. Robustness of the Results ized inputs in these circumstances cannot
result in disadvantaging the distorted sector
A few comments on the robustness of the such that there is an inefficient contraction
results may be in order before closing. None of output. The assumption that there are
of the results relies on the fact that the constant returns in X for fixed n is similarly
number of specialized inputs exceeds the unimportant in this respect.
number of goods in the relevant sector. The
X sector could be composed of many goods IV. Summary
with the total number exceeding the number
of inputs n. The result that free-input trade The purpose of this paper is to develop
is preferred to free trade in goods only for and analyze a model of trade in which trade
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions
VOL. 79 NO. I MARKUSEN: TRADE IN PRODUCER SER VICES 95
This content downloaded from 128.252.67.66 on Fri, 27 Nov 2015 02:48:41 UTC
All use subject to JSTOR Terms and Conditions