Krugman 1995

Growing World Trade: Causes and Consequences
Author(s): Paul Krugman, Richard N. Cooper, T. N. Srinivasan

Source: Brookings Papers on Economic Activity, Vol. 1995, No. 1, 25th Anniversary Issue
(1995), pp. 327-377
Published by: The Brookings Institution
Stable URL: http://www.jstor.org/stable/2534577
Accessed: 13/10/2010 10:57
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=brookings.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
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.
The Brookings Institution is collaborating with JSTOR to digitize, preserve and extend access to Brookings
Papers on Economic Activity.
http://www.jstor.org
PAUL KRUGMAN
Stanford University
Growing World Trade:

Causes and Consequences
WHAT ASPECT OF the American economy has changed most in the

twenty-five years since Brookings Papers on Economic Activity first be-
gan appearing? If you took a poll of economic journalists, businessmen,
or policy intellectuals other than professional economists, globaliza-
tion-the growing integration of the United States with the world econ-
omy-would probably top the list. It is now conventional wisdom in
many circles that the growth of world trade and investment has trans-
formed the ground rules for economic policy.
Admittedly, many international economists regard the popular con-
viction that unprecedented globalization has changed everything as con-
siderably exaggerated; Americans are still so taken with the novelty of
extensive international trade that they have yet to acquire a sense of per-
spective about its importance. Even today the shares of imports and ex-
ports in America's GDP are only about half of what they were in the
United Kingdom thirty years ago; the U.S. economy is not now, and
may never be, as dependent on exports as Britain was during the reign
of Queen Victoria. Nonetheless, international trade has certainly in-
creased considerably since the 1960s. In 1960 the share of trade-mea-
sured as the average of imports and exports of goods and services-in
America's GDP was 4.7 percent; in 1994 it was 11.4 percent, an increase
of more than 100 percent. While the growth of trade has not been quite
as dramatic in other advanced countries, it has also been considerable:
the average OECD country had a trade share of 12.5 percent in 1960,
18.6 percent in 1990. And a number of developing countries have seen
I would like to thank T.N. Srinivasan, Richard Cooper, and William Nordhaus for
helpful comments.
327
328 Brookings Papers on Economic Activity, 1:1995
their trade explode: China, virtually isolated from the world economy
before 1978, may now export 25 percent of its GDP.
Why has world trade grown, and what are the consequences of that
growth? These are surprisingly disputed issues. This is true to a limited
extent because of disagreements among professional economists, nota-
bly about the relative importance of trade and technological change in
causing the growing inequality of wages in advanced nations. Even more
striking is the gap between professional opinion and the broader conven-
tional wisdom. There are a number of cases in which the perceptions of
noneconomists who believe themselves to be well-informed about the
world economy are radically at odds with what research seems to indi-
cate. To take a relatively mild example: most journalistic discussion of
the growth of world trade seems to view growing integration as driven
by a technological imperative-to believe that improvements in trans-
portation and communication technology constitute an irresistible force
dissolving national boundaries. International economists, however,
tend to view much, though not all, of the growth of trade as having essen-
tially political causes, seeing its great expansion after World War II
largely as a result of the removal of the protectionist measures that had
constricted world markets since 1913. At least implicitly, therefore, they
also tend to see the trend toward growing integration as potentially re-
versible.
And yet perhaps these disagreements should not be all that surpris-
ing. International trade is, after all, the prime example of a subject in
which it is essential to take account of general equilibrium, in which
everything affects everything else in at least two ways. The general-
equilibrium aspects of international economic issues can cause confu-
sion even among the experts. For example, how should one think about
the effects of technology on income distribution? Should one, as Ed-
ward Leamer appears to believe, model the effects of technological
change by thinking of that change as occurring in isolation in a single
country that faces given world prices? ' Or should one, like most of those
who believe that technological change is the main cause of rising in-
equality, think of it as happening simultaneously in a number of coun-
tries that collectively constitute a more or less closed economy? These
two approaches can give radically different predictions.
1. Leamer (1994).
Paul Krugman 329
And if even the professors can get confused, the broader public-in-
cluding commentators who can sound convincing but do not have the
patience to work through all the implications of their ideas-is all the
more subject to befuddlement. An astonishing amount of the public dis-
cussion of international economic issues, among people who believe
themselves to be sophisticates, involves sheer misunderstanding of ac-
counting identities; and it goes without saying that almost nobody un-
derstands such abstruse concepts as comparative advantage.
In any case, this paper represents an attempt to shed some light on
the causes and implications of growing world trade. The issues dis-
cussed here are the subject of a huge recent literature; the distinctive
feature of the analysis in this paper is an attempt to keep in mind
throughout that world trade must be regarded as the outcome of a proc-
ess in which trade flows, world prices, wages, and employment are all
simultaneously determined.
The paper is in five parts. The first part presents an overview of trends
in world trade; it looks at the long-run evolution of world trade, and tries
to identify those aspects of globalization that are truly new. The second
part asks the question, why has world trade increased? The last three
parts are devoted to the issue that has created the most controversy in
recent discussions of international trade: the effects of exports of manu-
factures from the third world on wages and employment in the first. I
begin by setting out a stylized, minimalist general-equilibrium model of
world trade, wages, and employment, and suggest a set of ballpark
parameters for that model. I then turn to a theoretical and numerical
assessment of the impact of the new phenomenon of low-wage ex-
ports under two different sets of assumptions. First is a "European"
approach, in which the advanced world as a whole is assumed to have
inflexible relative wages, and in which the effects of trade are manifested
in changes in employment. As far as I know, this approach is new-the
rapidly growing literature on trade and wages has consistently assumed
that wages are flexible and that full employment is maintained. It turns
out, however, that the rigid-wage case is not only arguably of consid-
erable empirical relevance, but also has some major advantages in al-
lowing us to interpret the data. Nonetheless, however convenient it may
be to assume that relative wages are fixed, this is obviously not true for
all OECD nations. This analysis is therefore followed by a more prob-
lematic attempt to assess the impact of developing country manufac-
tures exports using an "American" approach, in which wages are

flexible.
The Growth of World Trade: An Overview
It is a late-twentieth-century conceit that we invented the global

economy just yesterday. In fact, world markets achieved an impressive
degree of integration during the second half of the nineteenth century.
Indeed, if one wants a specific date for the beginning of a truly global
economy, one might well choose 1869, the year in which both the Suez
Canal and the Union Pacific railroad were completed. By the eve of the
First World War steamships and railroads had created markets for stan-
dardized commodities, like wheat and wool, that were fully global in
their reach. Even the global flow of information was better than modern
observers, focused on electronic technology, tend to realize: the first
submarine telegraph cable was laid under the Atlantic in 1858, and by
1900 all of the world's major economic regions could effectively commu-
nicate instantaneously. How has world trade evolved since that impres-
sive beginning, and what aspects of the current growth in world trade
are truly new?
Trade as a Share of Output: A Long-term Perspective

Although the volume of world trade has been marked by a steady up-
ward trend since about 1950, a longer-term perspective reveals that such
growing integration is by no means necessary. On the contrary, between
1913 and the early post-World War II years most of the world's econo-
mies turned inward, and the share of world output that entered into in-
ternational trade declined substantially. Much of the growth in trade
since 1950 therefore simply represents simply a recovery to former lev-
els. Indeed, to the extent that it is possible to make comparisons, world
trade as a share of world output does not seem to have recovered to its
1913 level until sometime in the mid-1970s; only the growth since then
truly represents a new degree of integration.
Tables 1 and 2 show some indicative numbers. Table 1, derived from
data assembled by the World Bank, shows estimates of merchandise
trade as a share of world output since the middle of the nineteenth cen-
Paul Krugman 331
Table 1. World Merchandise Exports as Percentage of GDP

Percent
1850 1880 1913 1950 1973 1985 1993
5. la 9.8a I11.9a 7.1 11.7 14.5 17.1
Source: World Bank, (1995).
a. OECD countries only.
Table 2. Trade Shares in the United Kingdom, the United States, and Germany
Percenta
Country 1913 1950 1970 1987

United Kingdom 27.7 13.1 16.6 21.1
United States 3.9 2.9 4.4 7.4
Germany 19.9 9.8 17.4 23.3
Source: Liesner (1989).
a. Merchandise trade, measured as the average of exports and imports, as a share of GDP.
tury.2 The data show a marked increase in trade up to 1913; a substantial

dropoff by 1950; and a recovery that has continued to the present, finally
outstripping 1913 levels sometime between 1973 and 1985.
Table 2 shows merchandise trade, measured as the average of exports
and imports, as a share of GDP for three major economies since 1913.
All three show the same pattern of dip and subsequent rise. Although
the United States had a slightly larger trade share in 1970 than in 1913,
the United Kingdom and Germany were still below their 1913 trade lev-
els; the general picture of world integration that did not exceed early-
twentieth-century levels until sometime well into the 1970s is thus
broadly confirmed.
In the last decade or so, the share of trade in world output has finally
reached a level that is noticeably above its former peak. Nonetheless, it
would be hard to argue that the sheer volume of trade is now at a level
that marks a qualitative difference from previous experience. The
United States, in particular, remains considerably less dependent on
trade than major European countries have been for at least a century.
Does this mean, then, that there is nothing new about the kind of
growth of world trade that has taken place over the last generation? No:
while the overall volume of trade has not increased as much as might
2. Unfortunately, the pre-World War I data are only for OECD countries.
be expected, the aggregates conceal several novel features of modern

international trade.
New Aspects of Trade

It is possible to identify at least four new aspects of modern world
trade-new in the sense that they did not have counterparts in the previ-
ous golden age of the global economy. These are the rise of intra-trade,
trade in similar goods between similar countries; the ability of producers
to slice up the value chain, breaking a production process into many geo-
graphically separated steps; the resulting emergence of supertraders,
countries with extremely high ratios of trade to GDP; and, the novelty
that provokes the most anxiety, the emergence of large exports of manu-
factured goods from low-wage to high-wage nations.
INTRA-TRADE. As already pointed out, the United Kingdom has
been a highly trade-dependent nation since the mid-nineteenth century;
measured as a share of the United Kingdom's output, trade is basically
no higher now than it was in the high Victorian period. There have, how-
ever, been some major changes in the composition of British trade, both
by commodity and region. Table 3 provides indicative numbers compar-
ing British trade in 1913 and 1992. They show that in 1913 Britain was, to
a good approximation, a nation that exported manufactured goods and
imported raw materials, period; and it was a country that largely traded
with raw material producers overseas. By 1992 British imports as well
as exports consisted largely of manufactured goods, and most of the
country's trade was with other European nations-that is, with coun-
tries with similar resources. It is also true that a high proportion of the
trade among industrial countries appears to consist of intra-industry
trade, two-way trade in goods in the same commodity class. And it was
Table 3. Commodity and Geographical Composition of U.K. Trade

Percenta
Exports of Imports of Exports to Imports from
Year manufactures manufactures Europe Europe
1913 75.5 20.2 39.5 44.6
1992 81.9 78.4 63.8 63.7
Source: Mitchell (1988), Barraclough (1978), and Great Britain, Central Statistical Office (1994).
a. Numbers are percentages of the relevant piece of trade. Export columns indicate percentage of total
merchandise exports; import columns indicate percentage of total merchandise imports.
Paul Krugman 333
a striking feature of the growth in trade that followed major trade liberal-
izations among industrial countries, such as the formation of the EEC in
1958 and the United States-Canada auto pact in 1965, that the bulk of
the increase in trade consisted of nearly balanced increases in exports
and imports within three-digit industrial categories. 3
It is probably fair to say that the standard explanation for trade in sim-
ilar products among similar countries is that it is motivated by econo-
mies of scale in the production of differentiated products, although there
remains some skepticism.4 The important point for current purposes,
however, is that the rise of intra-trade depends on some ways in which
the nature of "typical" manufactured products has changed since 1913.
To put it briefly, manufactured goods today are more complex than
those of our great-grandfathers' day; not only are they more finely differ-
entiated, their manufacture involves the use of a much greater variety of
specialized intermediate goods (and intra-industry trade consists largely
of trade in such intermediates). Cotton textiles, the principal British ex-
port in the early stages of industrialization, were a fairly standardized
product; one could not really imagine much two-way trade in bolts of
cloth. Furthermore, production involved only a few steps, from raw cot-
ton to yarn to cloth, leaving little scope for the vertical disintegration of
the industry. By contrast, modern manufactures-take the overused but
inevitable example of automobiles-are highly differentiated, and their
production involves a number of different stages. If these stages take
place in different countries, they become a source of increased trade vol-
ume; so that it is not surprising to see Germans driving Hondas while
Japanese drive BMWs.
SLICING UP THE VALUE ADDED CHAIN. In Detroit's Institute of
Fine Arts there is a remarkable room whose walls are painted with four
stunning murals by Diego Rivera. The Rivera murals, completed in
1933, show in considerable detail the operations of Ford's River Rouge
industrial complex-a giant facility that combined at a single site blast
furnaces, rolling mills, engine casting, body stamping, and assembly of
complete automobiles. The Rouge plant was, in effect, a facility that in-
gested coke and iron ore at one end and extruded passenger cars from
the other.
Although Rivera's murals were intended as a celebration of the power
3. See, in particular, the papers by Grubel (1967) and Balassa (1966).
4. The canonical reference is Helpman and Krugman (1985).
of modern industry (and also, to his patron's dismay, a condemnation of

its brutality), they now have a decidedly archaic feel. Part of that sense
of old-fashioned industry comes from the very degree of integration that
seemed so impressive at the time. What are all those disparate opera-
tions doing in the same facility? Why are they not being done at special-
ized plants scattered around the globe?
It would be interesting to know how many facilities the average iron
atom in a 1995 Ford automobile has passed through (or better yet, to
know how many miles it has traveled, from the time it enters the gate of
the steel plant to the time it rolls off the assembly line). But it is gener-
ally believed (with little hard statistical evidence) that the trend in manu-
facturing has been to slice up the value chain-to produce a good in a
number of stages in a number of locations, adding a little bit of value at
each stage.
Such slicing up could greatly increase the potential volume of interna-
tional trade. In 1913, a given consumer good could, to a rough approxi-
mation, be exported only once. Today it can be exported many times: a
good that is produced in one country may be assembled from compo-
nents produced in other countries, and these in turn may be assembled
from subcomponents produced in yet other countries. As a result, the
trade involved in the global production of a final good may easily be sev-
eral times the value added in all stages of that production.
This increased potential for trade may help explain the next new as-
pect of world trade: the emergence of supertrading nations.
SUPERTRADING ECONOMIES. Globaltradeas a percentageof global
output is, even now, only moderately higher than it was in 1913. The
most trade-oriented economies, however, have much higher trade
shares than ever seen before. As far as the available data indicate, there
was no country in 1913 whose exports exceeded 50 percent of GDP. To-
day there are at least six such countries:5
Exports as percentage
of GDP, 1990
Singapore 174
Hong Kong 144
Malaysia 78
5. World Bank (1994).

Paul Krugman 335
Belgium 70
Ireland 64
Netherlands 52
The emergence of supertrading economies clearly depends on the

ability of modern industry to slice up the value chain, so that the value
of exports can be substantially larger than the value added in the export
industry. This is a fortiori true for Singapore and Hong Kong, where ex-
ports actually exceed GDP (alias, value added in the domestic econ-
omy). But it must also be true for all the other countries shown, since it
is virtually certain that at least 60 percent of the employment and value
added even in small countries is generated in nontradable sectors; thus
a trade share of much more than 30-40 percent can only arise when ex-
ports involve adding a fairly small amount of value to imported interme-
diate goods.
Of the countries listed above, Belgium and the Netherlands represent
something of a special case. They are best thought of as part of an inte-
grated economic region that comprises northern France and, most im-
portant, the Ruhr and nearby areas of Germany; different pieces of a
productive process tend to be spread across this region in much the same
way that manufacturing plants in the modern U.S. automobile industry
tend to be scattered across a fairly wide area of the Midwest.
The other countries in the supertrader class, however, represent
something different: they are all low-wage manufacturing platforms.
That is, they are locations to which semifinished products are sent to
have labor-intensive operations performed on them, and from there
shipped on for further work. Ireland, where wages are high by Chinese
standards but low by European standards, plays this role on a local ba-
sis; the Asian supertraders play this role globally.
One of the effects of the sliced-up value chain in world trade has been
to weaken the traditional link between the population of a country and
its trade share. There is still a strong inverse correlation: it is no accident
that the United States has a lower trade share than any other advanced
nation, or that Japan has the second lowest. But in 1990 Belgium, with
almost ten million people, had a trade share twice that of Iceland, with
only 255,000. The point is that Iceland is not, like Belgium, situated
squarely in the middle of an industrial complex; and Iceland's relatively
high wages, supported by fish exports, ruled it out as a manufacturing
platform along the lines of Ireland (although Reykjavik and Cork are ap-
proximately the same shipping distance from Rotterdam).
LOW-WAGE MANUFACTURING EXPORTERS. Finally, the novel as-
pect of trade that has created the most controversy and concern is the
rapid growth of manufactured exports from low-wage, newly industrial-
izing economies (NIEs).
It seems likely that the rise of NIE exports has something to do with
the slicing up of the value chain. In the early stages of the NIE phenome-
non, when rapid growth was limited to the Asian "tiger" economies, it
was common to hear doubts expressed about the possibilities for such
growth on a really large scale. Surely a second wave of manufacturing
exporters would soon be competing for the same limited markets-ap-
parel, toys, and a few other labor-intensive goods-that were being
served by the "Gang of Four"? And surely there would be an insufficient
range of suitable products to allow rapid growth of manufactured ex-
ports from, to take an unlikely candidate, mainland China!
What has happened, however, is that it has proved possible to find
expanded niches for labor-intensive production by slicing up the pro-
duction of goods traditionally viewed as skill-, capital-, or technology-
intensive and putting the labor-intensive slices in low-wage locations.
To take what has become a classic example, a notebook computer looks
like a high-technology product; but while the American microprocessor
and the Japanese flat-panel display are indeed high-tech, the plastic shell
that surrounds them and the wiring that connects them are not, so the
assembly of notebook computers becomes an NIE industry.
Incidentally, the effect of the sliced-up value chain on low-wage ex-
ports is one area in which the conventional wisdom among business
writers seems to be precisely wrong. It is often said that labor costs are
now such a low share of total costs that low wages cannot be a significant
competitive advantage. But when businesspeople say this, they do not
mean that labor costs have declined as a share of value added: on the
contrary, the division of value added between capital and labor has been
impressively stable over time. What they mean, instead, is that because
of the growing vertical disintegration of industry the value added by a
given manufacturing facility is likely to be only a small fraction of the
value of its shipments; and thus the labor share of that value added is
also a small fraction of costs, which are dominated by the cost of inter-
mediate inputs. But this vertical disintegration, or slicing up of the value
Paul Krugman 337
Table 4. Exports of Manufactured Goods from Newly Industrializing Economies

Percentage of GDP in destination
All industrial
Year economies European Union United States
1970 0.24 0.22 0.28
1990 1.61 1.30 1.91
Source: Exports from UNCTAD (1994); GDP from OECD, Nationial Accountits.
chain, creates a greater, not a smaller opportunity to relocate production

to low-wage locations.
The reasons for the explosion of exports and more generally of eco-
nomic growth in NIEs are deep and deeply disputed questions. What is
clear, however, is that NIE export of manufactures, a phenomenon that
was essentially nonexistent twenty-five years ago, has become a sig-
nificant feature of the world economy. Table 4 shows the growth of man-
ufactures exports from the NIEs, measured as a share of advanced
country GDP. In addition to showing that these exports have effectively
emerged out of nowhere, the table also makes the point that they are still
not very large as a share of total OECD spending.
Why Has World Trade Grown?
When economic commentators try to explain why world trade has

grown faster than world output, they generally offer one of two explana-
tions. The explanation most popular with journalists stresses the effects
of technology: Lower costs of transportation and the growing speed of
communications have made the world a smaller place. International
economists, while paying some attention to this explanation, tend to
stress political factors to at least an equal extent; they point to the role
of the GATT and, more recently, of unilateral liberalizing measures by
developing countries. It is also important to realize that the volume of
international trade depends to some degree on how boundaries are
drawn-for any given geography of world production, the measured
volume of international trade will depend on which shipments are de-
fined as international, and so some of the apparent change in world trade
may simply represent changes in these definitions.
Political Factors
It is difficult to quantify protectionism. Tariff rates can be measured,

although there is still a problem of interpreting ex post data: to take the
extreme case, a country that imposes prohibitive tariffs on some goods
may have a low tariff rate on all of the goods it continues to import.
Nonetheless, there is no question that the general profile of world pro-
tectionism since the early twentieth century has been the inverse of that
of world trade: during the interwar period there was a proliferation of
tariffs, import quotas, and exchange controls; from World War II on-
ward there has been a general trend toward freer trade. The correspon-
dence between the time profile of trade policy and that of trade shares is
prima facie evidence that political factors have played a major, perhaps
dominant role in the growth of world trade since 1950. In fact, a first-
pass story about the growth of world trade might run as follows: the key
technologies that created a global economy were the railroad, the steam-
ship, and the telegraph. Everything since then has represented only a
marginal improvement. Thus very large-scale world trade emerged in
the days of the Pax Britannica. Politics then killed that first global econ-
omy; with the gradual restoration of a relatively free trade environment
under the Pax Americana, the global economy was gradually recon-
structed, and the world economy more or less reached its previous level
of integration by 1980.
It is possible to elaborate on this story in several ways. First, it is
worth pointing out that the pre-World War I world economy was not
exactly characterized by free trade. Indeed, the two largest economies,
the United States and Germany, were frankly protectionist. Thus it
should not be surprising that the closer approximation to free trade that
has been achieved in recent years has brought world trade somewhat
above early-twentieth-century levels, nor that-as shown in table 2-
the once-protectionist United States and Germany now have noticeably
larger trade shares than they did in 1913, while then-free-trading Britain
does not.
Second, it is important to be aware that the progress toward free trade
in the postwar period has not entirely taken place through GATT negoti-
ations, nor for that matter among the OECD countries. In recent years
the most important moves toward free trade have been unilateral actions
by developing countries, many of which have turned away from the im-
Paul Kruigman 339
port-substitution policies of the past. Such moves do not only affect the
developing countries: because trade liberalization increases exports as
well as imports, the move to free trade in the developing world con-
tributes to the growth of developing country exports to the high-wage
nations.
Finally, there is scattered but suggestive evidence that removing for-
mal barriers to trade is not enough to produce full economic integration.
An intriguing study by John McCallum of the Royal Bank of Canada,
using data from the 1988 input-output tables for Canada, finds that Cana-
dian provinces traded far more with each other than they did with Amer-
ican states of comparable population and at comparable. distances.6
Thus Ontario exported more than three times as much to British Colum-
bia, with three million people, as it did to California, with almost thirty
million. When McCallum estimated a gravity equation (see below) for
trade among Canadian provinces and U.S. states, he found that intra-
Canadian trade was a startling twenty times as large as would have
otherwise been expected. What is so dramatic about these findings is
that, although the data predate the Canada-United States Free Trade
Agreement, tariff barriers were already very low in 1988; and the linguis-
tic divide in North America runs through the middle of Canada, not be-
tween Canada and America. Thus this evidence suggests that political
boundaries, even between friendly nations that speak the same lan-
guage, can be serious obstacles to trade. And it therefore helps to con-
firm the belief, which underlies such initiatives as "1992" in Europe, that
there remains substantial room for policy moves to expand international
trade through a process of harmonization of laws and institutions. (Rob-
ert Lawrence has dubbed such moves "deep integration.")
Boundary Issues
McCallum's results on intra- versus extra-Canadian trade notwith-
standing, it is sometimes useful to think about world trade by imagining
that it were possible to take a given geography of world production and
transportation and then draw arbitrary lines on the map called national
borders without affecting the underlying economic geography. If inter-
national trade only includes shipments that cross the borders, it is clear
6. McCallum (1995).
that the volume of that trade will depend quite a lot on where one draws
the lines.
A case in point is the trade of European Union countries. Taken indi-
vidually, they are very open economies, with an average trade share of
28.0 percent in 1990. However, more than 60 percent of their merchan-
dise trade is with each other. If the European Union is taken as a unit,
its merchandise trade with the external world is only 9 percent of GDP,
not much more than that of the United States.7
There has been a fair amount of literal redrawing of boundaries in the
last few years. More important for world trade, however, has been the
changing distribution of world output among existing nations. To see
why this matters, compare the likely share of trade in world output in
two hypothetical cases: a world of two equal-size countries, and one in
which the larger country has 95 percent of gross world product. It seems
obvious that in the latter case the trade share would be much smaller.
This idea can be formalized if we suppose that world trade can be de-
scribed by a simple gravity equation. Gravity equations attempt, with
considerable success, to explain the volume of trade between any pair
of countries with a few variables, usually the GDPs of the countries and
the distance between them. A typical gravity equation is of the form
(1) Tij = kIYYjPD1jy,
where Tij is the trade between two countries, i and j; Yi and Yjare the
GDP of countries i and j, respectively; Di is the distance between the
two countries; and k is a parameter. In practice such equations typically
find the exponents on GDP to be less than one, and a surprisingly strong
effect of distance.8 But in an idealized world in which a buyer is equally
likely, when buying a traded good, to buy it from a supplier anywhere in
the world, a( = B = 1, and -y = 0. And then two results follow. The share
of trade in the GDP of any one country would be
(2) Ti= k(1 -si),

yi
where si is that country's share in gross world product; and the share of
trade in that gross world product would be
7. There are no reliable estimates of the direction of service exports. OECD (1992) has
data on trade shares. Data on merchandise trade by location from European Economy.
8. For some recent estimates, see Frankel, Wei, and Stein (1994).
Paul Krugman 341
(3) Tw = k(1 -
Si2).
Yw
In other words, an individual country's trade would be larger, the

smaller its share of world income; and the overall share of trade in the
world would be larger, the more equal in size its national economies. In
actual fact, the national distribution of gross product among the world's
economies, as measured at nominal exchange rates, has become some-
what more equal over time: the sum of squared shares fell from 0.130
in 1960 to 0.104 in 1993. Or, inverting the ratio, the number of country-
equivalents in the world economy has risen from 7.7 to 9.6. The most
important source of this rise has been the relative decline of the United
States, from 31.6 percent of the gross product of market economies in
1970, to 25.6 percent in 1993.9This may well explain why the rise in the
trade share has been more dramatic for the United States than for other
industrial countries: indeed, comparing the present situation with that in
the 1950s, it may be said that the United States used to trade with a world
that in economic terms was barely larger than itself, but now trades with
a world three times its size.
Technological Change
It is clear that the volume of world trade is not completely determined
by technology: transportation and communication technology were
considerably better in 1950 than they were in 1913, but the world econ-
omy was substantially less integrated. Correspondingly, since much of
the growth of trade since then represents only a return to 1913 levels of
integration, it is hard to argue that technology has been the dominant
factor in that growth. Indeed, it is possible to make a strong antitechnol-
ogy case: Granted that there have been reductions in transport cost and
improvements in the speed and bandwidth of communications, surely
these are marginal improvements on a set of technologies that already
permitted massive long-range trade?'0
9. United Nations (1993).

10. It is worth pointing out, however, that transportation technology in 1995 should
not be compared with that in 1950, but rather with the technology of 1913. Presumably
there was a rise in potential trade during the interwar period, which was suppressed by
political factors and could only manifest itself once a liberal trade regime was reestab-
lished.
The case for the important role of technology must rest on more sub-
tle indicators than the aggregate volume of trade. One indicator is the
emergence of new aspects of trade, especially those associated with the
thinner slicing of the value chain. Arguably, before the widespread
availability of computers and telecommunications, the geographic dis-
persion of a complex production process was too hard to coordinate.
(Even now, the adoption ofjust-in-time production techniques is usually
associated with geographic clustering of production.)
Some support for this view is given, in an indirect way, by estimated
gravity equations. These always show a strong effect of distance on
trade volumes, which is both too large and of the wrong form to be easily
associated with measured transport costs. While it is not clear why dis-
tance plays so strong a role in trade, a common guess is that it proxies
for the possibilities of personal contact between managers, customers,
and so on; that much business depends on the ability to exchange more
information, of a less formal kind, than can be sent over a wire. If this is
true, then we might argue that the advent of such innovations as long-
range passenger jets, cheap intercontinental telephone calls, fax ma-
chines, and electronic mail permit an intensity of long-distance business
relationships that was not possible in 1913. Steamships may have been
quite efficient at transporting bulk commodities, but they were too slow
to allow regular visits to headquarters; telegraphs may have allowed ef-
fectively instantaneous communication of futures prices and interest
rates, but they lacked the bandwidth to allow the home office to transmit
detailed production specifications and the factory to explain why they
would not work.
A final point. There is one aspect of technological progress that has
acted to reduce the share of trade in world output: the faster rate of pro-
ductivity growth in the production of goods than that in services. The
declining relative productivity of the service sector is, according to most
estimates, the main reason why that sector constitutes a growing share
of the ermoloyment and value added in advanced economies (that is, the
elasticity of substitution between services and goods appears to be low,
so a rising relative price translates into a rising share in the economy). "I
Despite some recent growth in service trade, services are by and large
still nontradable. Thus although it has become easier and cheaper to
11. See Fuchs (1968).

Paul Kruigman 343
trade whatever can be traded, a declining share of the economy consists

of tradable goods and services.
A Stylized Model of Global Trade, Employment, and Wages
So far this paper has described the growth of world trade and re-
viewed some possible explanations for that growth. The time has now
come to try to understand the effects of growing trade-bearing in mind
that trade flows cannot be taken as wholly exogenous, but must be
viewed as part of a system in which they, along with a number of other
things, are jointly determined.
What effects of growing trade should be the subject of analysis? It is
a bad idea to try to discuss everything that might occur because of inter-
national integration, all at once. A broad-brush approach may be accept-
able when offering a descriptive survey, but to assess the effects of trade
one needs to create a model; and if the model is to be tractable and com-
prehensible, it must focus on only a few things.
The focus here will be dictated by political controversy. Of the new
aspects of world trade discussed above, the rise of intra-trade has gener-
ally been viewed as benign, and the slicing up of the value chain and the
emergence of supertrading nations have excited interest but little con-
troversy. The controversial new aspect of international trade is the rise
of manufacturing exports from newly industrializing economies.
The rapid growth of NIE exports has more or less coincided with
some disturbing trends in OECD labor markets: a sharp rise in wage in-
equality (especially in the United States) and a sharp rise in unemploy-
ment (mainly in Europe). It is widely believed that the unfavorable labor
market trends and the growth of NIE trade are connected.
This belief has been expressed at greatly varying levels of sophistica-
tion. At one end, there is the phenomenon of a self-made billionaire-
turned-politician, who has declared himself an expert on economics and
launched a campaign to warn his countrymen of the impoverishment
they face as a result of free trade with low-wage nations. I refer, of
course, to Sir James Goldsmith, whose book The Trap has been a Euro-
pean best-seller. While one might dismiss Sir James and his untitled
Texan counterpart as marginal, milder versions of the same warning are
found among highly respected and influential people. Even the Euro-
344 BrookingsPapers on EconomicActivity,1:1995
pean Commission, in its 1993 White Paper Growth, Competitiveness,

Employment, attributed a major share of the rise in European unemploy-
ment rates to the fact that "other countries are becoming industrialized
and competing with us-even on our own markets-at cost levels which
we simply cannot match."'2
Academic research has been far less supportive of the claim that NIE
manufactures exports are a major source of problems in OECD labor
markets. While there are some studies that do claim to find evidence for
substantial pressure from low-wage imports on unskilled labor in ad-
vanced countries, it is probably fair to say that a preponderance of the
research to date suggests that the impact of third world exports on first
world labor markets has been small, or at least elusive. 13
One thing that is conspicuously lacking in the literature to date, how-
ever, is a consistent picture of the interaction between labor market de-
velopments in the high-wage countries and the growth of exports from
the low-wage countries. While some (though not all) of the studies are
based on a consistent underlying model of employment and wages in the
advanced countries, there does not seem to be any effort to show how
wages and employment in the advanced countries, and trade with the
third world might be simultaneously determined. That is, there do not
seem to be any complete general-equilibrium stories out there.
In the remainder of this paper I will try to fill that gap, by developing
a highly stylized model of global trade, employment, and wages. This
model simplifies reality too much to be estimated with or even calibrated
to the data. It is possible, however, to use the results of other peoples'
empirical work to assign a set of ballpark parameters to the model,
allowing what amount to glorified back-of-the-envelope estimates of the
impact of growing NIE trade.
Structure of the Model

In order to focus on the effects of NIE trade in manufactured goods,
I assume a world consisting of only two economies: one that is intended
to represent the OECD as a whole, the other to represent the aggregate
12. Commission of the European Communities (1993, p.4).

13. See, in particular, Wood (1994) and Leamer (1993, 1994) in support of the adverse
effects of NIE exports, and Katz (1992), Bhagwati and Kosters (1994), and Sachs and
Shatz (1994) for the alternative view.
Paul Krugman 345
of NIEs. All transactions within each aggregate are ignored, as is the ex-
istence of other types of countries, like oil exporters.
The aggregate OECD is assumed to produce and consume two goods,
1 and 2, with production of good 1 being skill-intensive. Demand is de-
termined by a utility function in the consumption of goods 1 and 2,
(4) U = U(CI,C2)-
It will be convenient to assume that tastes are homothetic, so that mar-

ginal and average spending have the same composition.
Since the focus of this analysis is on labor market developments in
the OECD, it is necessary to have some explicit treatment of the factor
markets. I will make several strategic simplifications. First, the only
productive inputs will be skilled labor (Ls) and unskilled labor (Lu); capi-
tal will be left out of the story. The main reason for doing this is that the
distribution of income between capital and labor has neither changed a
lot nor been a major source of controversy in the last two decades; the
share of labor compensation in U.S. national income, for example, has
barely changed, actually rising from 73 to 74 percent between 1973 and
1993. It is not clear that this is what one would have expected a priori,
nor is it clear that capital will remain a sort of bystander factor indefi-
nitely. For the purposes of this model, however, all income will be as-
sumed to accrue either to skilled or to unskilled labor.
Second, production, Q, will be assumed to take place under constant
returns to scale, with the production functions
(5) Q,= F(Ls,, LuI)
and
(6) Q2 = G(LS2,LU2)
Economies of scale are widely believed to be important in understanding
both the causes and effects of trade within the OECD, but probably play
a smaller role in NIE trade.
For the same reason, markets are assumed to be perfectly competi-
tive. This is likely to raise some stronger objections. One common story
about the effects of international trade on wages is that it has weakened
the bargaining power of workers; this only makes sense if workers and
employers are struggling over the division of some rent, presumably cre-
ated by the market power of the firm. It is arguable whether such stories
can be a large part of the picture: they seem to predict a shift in the distri-
bution of income between capital and labor, which has not happened,
rather than between different types of labor, which has; and they apply
only to those workers in traded-goods industries, whereas the rise in in-
come inequality has been pervasive throughout the economy. In any
case, for this model competition is assumed to be perfect.
The OECD's trade is the difference between its production and its
consumption. Exports of the skill-intensive good, X,, and imports of the
less skill-intensive good, M2, can be written as
(7) XI = , C,
and
(8) M2= C2-Q,2
How should the OECD's trade with the NIEs be modeled? It is common
in trade theory to work with small economies that face given world
prices; and some writers on the effects of changing world trade still use
this assumption. For the OECD as a whole, however, this is a deeply
unrealistic assumption; worse yet, it is analytically awkward, leading to
excessively "bang-bang" solutions in some cases. It thus makes sense to
regard the OECD as having substantial market power relative to the
NIEs. This can be represented by assuming that the OECD faces a rest-
of-world offer curve,
(9) M2= T(X1).

I will not make any attempt to model the inside workings of the newly
industrializing countries; they will simply be summarized by the offer
curve given by equation 9. The growth and increased integration of the
NIEs with the world economy are then captured simply by an outward
shift in that offer curve. In fact, since their manufactured exports were
negligible in 1970, the effects of their emergence can be approximated
by contrasting an initial period in which the OECD has no external trade
with a subsequent period in which it faces an offer curve that leads to
the observed trade volumes. That is, the model analyzes the effects of
globalization by contrasting the current situation with one of autarky for
the OECD as a whole.
That is the whole theoretical model. Before assigning ballpark param-
eter values, it may be useful to review some of the key mechanics of 2 x 2
production models.
Paul Krugman 347
Figure 1. Relationship between Goods Prices, Factor Prices, and Factor Proportions
WS/WU
P I/P2 LSILU
Source: Samuelson (1949).
The most important relationships are presented in figure 1, which re-

produces the familiar diagram introduced by Paul Samuelson. 14 Figure
1 summarizes the three-way relationship between goods prices, factor
prices, and factor proportions. The right panel shows that given the ratio
of skilled to unskilled wages (WslWu), each industry chooses a ratio of
skilled to unskilled workers in production. The left panel shows the ba-
sic Stolper-Samuelson relationship between the relative price of the
skill-intensive good (P/IP2) and the relative wage of skilled workers that
prevails if both goods are produced.
Putting Numbers to the Model

Quantifying a model like this poses certain conceptual problems.
That is a polite way of saying that it is somewhat difficult to know exactly
how to assign numbers to a model that is, in important respects, bla-
tantly untrue. Not only are there more than two kinds of labor, other fac-
tors besides labor, other regions besides the OECD and the NIEs; it is
not even true that workers of apparently similar skill receive the same
wages in exporting and import-competing sectors. Yet for the model to
be used, it is necessary to assign a set of parameter values that fit to-
gether.
14. See Samuelson (1949).
The following parameters are used in the trade model of the next two
sections:
Initial ratio of skilled to unskilled wages 2

Share of skilled workers in industry 1 employment 0.5
Share of skilled workers in industry 2 employment 0.2
Share of skilled workers in labor force 0.4
Share of wages of skilled workers in industry 2/3
1 value added
Share of wages of skilled workers in industry 1/3
2 value added
Share of good 1 in total expenditure 5/7
-Wage ratio. Adrian Wood, using his definitions, finds a wage ratio
between skilled and unskilled workers in the North of 2.08.15I round this
to 2.
-Employment share in industry 1. Wood also estimates a share of
skilled employment in export-oriented manufacturing of 50.24 percent,
which I round to 50.16
-Employment share in industry 2. Wood does not, for reasons ex-
plained below, estimate the employment composition of Northern im-
port-competing production. Other sources, using different definitions,
have produced estimates. However, in order to make the model inter-
nally consistent, one must meet a constraint that is not met in the actual
data: the difference in average wages between export and import-com-
peting industries must be fully accounted for by the difference in skill
composition. Bela Balassa found that wages in U.S. industries that ex-
port to developing countries were 28 percent higher than those in indus-
tries competing with imports from those countries; by assigning the im-
port-competing industry a 20 percent skill fraction, I get an implied 25
percent average wage ratio, which seems close enough. 17
-Labor supplies. If the OECD is producing both goods, the ratio of
skilled to unskilled workers in the labor force must be between the ratios
in the two industries. The choice of 40 percent skilled workers is arbi-
trary, but has little effect on the results below.
15. Wood (1994, p.403).
16. Wood (1994, p.403).
17. Balassa (1979).
Paul Krugman 349
-Remaining entries. These follow from the previous numbers. In

particular, the share of industry 1 in expenditure is determined by the
requirement that the derived demand for factors equal the supply.
This completes the stylized model of the interaction between NIE
trade and OECD labor markets, together with some semirealistic num-
bers that will allow back-of-the-envelope estimates of effects. The next
step is to apply the model under two alternative assumptions about
OECD labor markets.
The Impact of Low-Wage Exports: A "European" Approach
Most analyses of the impact of NIE trade have had an "American"

flavor, in at least two senses. First, they tend to be based on U.S. data.
Second, to the extent that they try to make an explicit estimate of the
labor market effects, they assume that relative wages are flexible, and
thus that any adverse impact on unskilled workers is reflected in declin-
ing wages rather than increased unemployment. This is a reasonable as-
sumption for the United States, where low real minimum wages, weak
unions, and a very weak social safety net give real wages for unskilled
workers an impressive ability to decline. The European economy, how-
ever, which is roughly comparable to that of America in output and em-
ployment, presents a very different picture. Wage inequality has not in-
creased to anything like the same extent; meanwhile, unemployment
has risen from less than 3 percent at the beginning of the 1970s to double
digits today.
This in itself would suggest that it is worthwhile to look at the impact
of integration under "European" assumptions, with relative wages rigid
and the consequent reduced demand for less skilled workers reflected in
unemployment. In addition, however, the European version of the story
is revealing in other ways: it highlights the importance of a general-equi-
librium approach, and (as the next section shows) there are some sur-
prising quantitative contrasts between the effects of trade in a "Europe-
anized" and an "Americanized" model.
For the moment, then, assume that ws / wu is fixed, so that any pres-
sure on labor demand is reflected in employment rather than wages; and
suppose that a previously autarkic OECD economy is now presented
with an offer curve from a group of newly industrialized economies,
which allows the OECD to export good 1 and import good 2.
Figure 2. Adjustment of OECD Employment under the "European" Model
/ E' E
-- -- -- - - - - -- -- ------
LQ 1
0 Lu
Source: Author's model as described in text.
The consequences may be analyzed in stages.'8 First, note that with

rigid relative wages, the skilled-to-unskilled ratio in each industry is
fixed. These ratios are illustrated by the slopes of rays 1 and 2 in figure
2, which shows employment in the economy. The point E represents the
initial employment of the two factors. Since these levels of employment
must be achieved using the factor proportions implied by the fixed rela-
tive wage, the resources employed in each industry are indicated by Q1
and Q2. Now suppose that there is a fall in the relative demand for the
less skill-intensive good. This cannot be met by a change in relative
wages, so it must be met with a reduction in unskilled employment. The
employment point moves left to E'. And output obeys the Rybczynski
theorem: the resources devoted to industry 2 fall to Q2', while those de-
voted to industry 1 actually rise, to Q1t.
Next, consider the implications for international trade. Figure 3 plots
OECD exports against imports. Point 0 represents autarky (no trade in
manufactures with the NIEs), and slope of the ray OA represents the
relative price of good 1 in autarky. Now the NIEs arrive on the scene,
with their presence summarized by the offer curve OC. What effect does
this have on relative prices? As long as the OECD continues to produce
18. This analysis was inspired by and closely follows the analysis in Brecher (1974).
Paul Kriugman 351
Figure 3. Adjustment of OECD Exports and Imports under the "European" Model
0 x
both goods, it has no effect: the fixed relative wage ties down the rela-
tive price, according to the relationship in the left panel of figure 1. So
the NIEs simply move along their offer curve to the point T, with OT
representing the volume of trade.
How does the emergence of this trade affect the OECD economy? It
must be accommodated by changes in both production and employ-
ment, changes that involve demand as well as supply and even a sort of
multiplier effect. The somewhat surprising logic of this response has
not, to my knowledge, been traced out before; it is illustrated in figure 4.
In this figure, the curve represents the production possibilities of the
aggregated OECD economy, given the initial employment of both fac-
tors. The point A is the equilibrium consumption and production of the
economy in autarky-that is, before the NIEs arrive on the scene. Con-
sumption at the relative prices indicated by the tangent budget line
through A will depend on income; the ray OA represents the income
expansion path.
Now the OECD opens trade with NIE economies that export good 2
and import good 1. If the OECD were a small, price-taking economy, it
would completely cease production of good 2. But because it is not, its
production of 2 falls and its production of 1 rises, with an unchanged rel-
ative price, until the desired trade equals the amount of trade that the
Figure 4. Adjustment of OECD Production and Consumption under the "European"

Model
Q2
A,'
0 Q1
NIEs are willing to do at that given relative price. Figure 4 shows that as
employment of unskilled workers falls, the OECD's production moves
down the "Rybczynski line" AR, which corresponds to the kind of ad-
justment in production shown in figure 2. The value of production falls,
and therefore the budget line shifts in; consumption therefore also
moves down along the expansion path OA. Global trade equilibrium is
reached when the OECD's desired trade vector CQ is just equal to the
NIE desired trade vector OT in figure 3.
The opening of trade with the third world, then-given the assump-
tion of rigid relative wages-leads to a fall in OECD employment. But
how much of a fall? And how would one estimate the employment reduc-
tion in practice?
Interestingly, two popular calculations actually understate the em-
ployment effects of trade. One calculation involves looking at the total
employment embodied in exports and imports. Since the average wage
in export industries is, in fact, higher than that in import-competing in-
dustries, this approach seems to indicate that the number ofjobs created
Paul Krugman 353
Figure 5. Adjustment of OECD Employment and Consumption under the "European"

Model
Ls
E D
-- -- -- -- -- -- - - - - - - -- ---- - -- -- - -- -6-- -- -- -- -- -- -- -- -- - -, A
Xc
0 Lu
by exports falls short of those displaced by imports. The other popular

calculation involves looking only at the unskilled labor embodied in ex-
ports and imports, since it is among the unskilled that employment falls.
Both of these approaches, however, miss the point that as employ-
ment falls, so does income and hence domestic demand. It is clear from
figure 4 that the output of good 1 rises by less than the volume of exports,
and that the output of good 2 falls by more than the volume of imports.
How, then, can the employment effects of trade be determined? One
way is to calculate the new equilibrium and grind out the implied em-
ployment change. There is, however, a shortcut that is possible because
there is no change in relative wages (by assumption) or relative prices
(by implication). That approach is illustrated in figure 5. As in figure 2,
we show the factor content of output, with A the initial employment.
Figure 5, however, also shows OECD demand-the derived demand for
factors embodied in production. In autarky this must be equal to the to-
tal supply. As income falls, this derived demand will fall along the
expansion path OA. Meanwhile, employment of unskilled workers will
fall to a point such as E. Through E, I have drawn a budget line, EC,
whose slope is equal to ws / wu; employment and consumption of factors
must both lie on such a line. Equilibrium involves a situation in which

the production point, E, and the consumption point, C, are such that the
implied trade in embodied factors, EC, equals the difference between
the factors used to produce exports and those that would be needed to
replace imports. 19
It is immediately apparent that the decline in employment of unskilled
labor is greater than the quantity of labor embodied in trade. The net
"import"of unskilled labor is the distance ED, but the actual fall in em-
ployment is measured by the full distance EA.
The Employment Effects of NIE Exports

As figure 5 indicates, the effects of trade on employment in a rigid-
wage economy may be estimated from the factor content of trade, to-
gether with an estimate of the general-equilibrium multiplier effect. Spe-
cifically, the fall in employment is
Net imports of unskilled labor + (Net exports of skilled labor
x Unskilled-to-skilled ratio in OECD economy)
Table 5 shows the results, using the World Bank's latest estimate of
the share of NIE manufactures exports in OECD gross product and the
parameters listed above.
This is a fairly large effect. It is still only a fraction of the actual rise
in European unemployment, but it is far from negligible. And it therefore
seems to suggest considerable reason for concern over the effects of
low-wage manufactures exports on first world labor markets.
This fairly large estimate depends, however, on the assumption of rig-
idly fixed relative wages. Even in Europe, this is an exaggeration; and
relative wages appear to be highly flexible in the United States. How do
19. The relevant factor content of trade here is that in OECD import-substituting pro-
duction; the factors used to produce the goods in the third world are irrelevant. Wood
(1994) has argued that developing countries produce "noncompeting goods" that are no
longer produced in the high-wage nations, and that one must therefore try to estimate what
it would have taken to produce these goods, rather than look at actual OECD industries.
This assertion is, however, problematic. If these really are noncompeting goods, how can
one assess their impact without specifying how they substitute in demand for other goods?
After all, in a two-good model in which the OECD and the NIEs are specialized in pro-
ducing different goods, an expansion of NIE exports would have no effect at all on equilib-
rium relative wages in the OECD.
Paul Krugman 355
Table 5. Employment Effects of Trade under "European" Assumption

Units as indicated
Manufactures importsa 1.75
Net imports of unskilled laborb 0.82
Net exports of skilled laborb 0.41
Unskilled-to-skilled ratio in aggregate 1.5
employment
Fall in employmentc 1.43
Source: Author's calculations based on model described in text and World Bank (1995).
a. Percentage of GDP.
b. Percentage of total employment.
c. Percent.
the results change if we "Americanize" the model, allowing wages to be

flexible and therefore assuming that the effects of trade manifest them-
selves in income inequality rather than unemployment?
The Impact of Low-Wage Exports: An "American" Model
With full employment of both skilled and unskilled labor maintained

by wage flexibility, the effects of opening trade between the OECD and
the NIEs can be represented by a figure so familiar that Ronald Findlay
has dubbed it the "sacred diagram" of international trade. Figure 6
shows how the pieces fit together when relative wages and hence prices
can change. Point A, once again, represents OECD autarky. When trade
is opened, the relative price of good 1 rises; the result is that production
moves to Q, while consumption moves to C. If this is a global equilib-
rium, the NIE offer curve-drawn backward, with its origin at Q-must
also pass through C, so that desired OECD exports equal desired NIE
imports, and vice versa.
But how can we quantify this qualitative picture? At first, it might
seem possible to begin in the same way as in the rigid-wage case, by cal-
culating the factor content of trade; then asking how much these changes
in effective supplies of skilled and unskilled labor affect relative wages,
by making use of some estimated elasticity of substitution. This has in
fact been the approach taken by some studies.20 Unfortunately, it runs
into serious conceptual difficulties. Even the concept of net trade in em-
20. See, in particular, Borjas, Freeman, and Katz (1991).

Figure 6. Adjustment of OECD Production and Consumption under the "American"

Model
Q2
QI
bodied services become hard to make sense of when relative factor

prices change as a result of trade; and the elasticity of substitution be-
tween skilled and unskilled labor will change when an economy is
opened, if it has any meaning at all. (It is possible to rescue the concept
if the economy does not face given world prices, but rather faces a con-
cave foreign offer curve-and this is certainly true for the OECD as a
whole. So one should not be as harsh in condemning studies that attempt
to make inferences from the factor content of trade as some critics, such
as Leamer, have been.21)
Given these conceptual difficulties, several recent studies have at-
tempted to infer the effects of trade on relative wages by looking at
pieces of the mechanism by which the process should work. In particu-
lar, Robert Lawrence and Matthew Slaughter looked for evidence that
the relative prices of less skill-intensive goods have indeed fallen, and
that the industry mix within the United States has shifted toward skill-
21. SeeLeamer(1994).
Paul Krugmnan 357
intensive sectors.22 They could not find any. A follow-up paper by Jef-
frey Sachs and Howard Schatz did find some weak evidence for relative
price changes in the expected direction, but nothing compelling.23
I offer here an alternative approach: with the addition of some further
assumptions to the model already described, it turns into a tiny, comput-
able general-equilibrium model of world trade. I can then ask the follow-
ing question: What changes in relative wages and prices would be con-
sistent with the observed growth of trade? The answer turns out to be
surprisingly small-that is, the same model that predicts fairly large em-
ployment effects with rigid wages predicts quite small effects on relative
wages when they are flexible.
Making the Model Computable

In order to make the model computable in the face of flexible prices
and wages, it is necessary to specify elasticities of substitution in pro-
duction and consumption-in effect, to choose functional forms. Since
this is an illustrative exercise rather than a full-fledged CGE modeling
project, it is sufficient to go with the simplest case (which is not too far
from most empirical estimates) of unitary elasticity. That is, the model
will be made Cobb-Douglas throughout.
In stating the model, it is also convenient to make some simplifying
choices of units. Letting unskilled labor be the numeraire, I choose units
so that in autarky the prices of both goods are one. I also measure skilled
labor in units of half a worker-a trick normalization that implies an au-
tarky relative wage that is also one. To be consistent with this normaliza-
tion, the economy's endowment is assumed to consist of sixty units of
unskilled and eighty revised units of skilled labor.
Given these choices of units, and the parameter values listed above,
output and factor markets in the OECD can be represented by the fol-
lowing equations:
First, letting w be the relative wage of skilled labor (one in autarky)
gives expressions for average cost-which must equal prices if both
goods are produced in the OECD.
(10) P1 = w2/3, P2 =
W-/3
22. Lawrence and Slaughter (1993).

23. Sachs and Schatz (1994).
Next, expressions for the unit input choices for both factors in both sec-
tors may be written
(1 1) a=I = w- w2'3, aS2 =w -2'3, au2 --w1/3

=
3 ul 3 ~~3 32
Given these input coefficients, output is determined by the requirement

of full employment of both factors. These conditions may be written
(12) Ls = asiQi + aS2 Q2 L = auIQI + aU2Q2,

which yield the output equations
(13) QI = D1[aU2Ls - aS2Lu], Q2 = D-[-au,Ls + as,Lu],
where
(14) D = aS aU2 -aS2aul
Given the output and prices of the two goods, it is straightforward to

calculate the implied trade vector. In fact, the relevant number is the
share of OECD-NIE trade in OECD output; this is simply the difference
between the share of good 1 in output and in consumption:
(15) T P1Q1 4
(15) + P2Q2
~~~~~PIQI 7-
Equations 10-15, then, lead from an assumed relative wage to the im-
plied relative prices and share of trade in output. It is also possible,
therefore, to reverse the procedure, and ask how large a change in rela-
tive wages in the OECD might be associated with the emergence of NIE
trade on the scale actually seen. And the answer is that trade on this
scale should be associated with a fairly small wage change-and a very
small change in relative prices.
Table 6 shows the implications of a 3 percent rise in the relative wage
of skilled workers from its autarky level. It turns out that this is large
enough to imply NIE trade of 2.2 percent of OECD gross product; which
is more than the actual share of NIE manufactures in OECD spending.
Yet this wage rise would be associated with a rise of only 1 percent in
the relative price of skill-intensive goods. Admittedly, this exercise is
carried out not only with a highly stylized model, but also on the assump-
tion of unitary elasticities of substitution in production and consump-
tion. If these elasticities were lower, the implied change in relative
Paul Krugman 359
Table 6. Implications of a 3 Percent Rise in Relative Wages of Skilled Workers

Percent
Share of NIE exports in OECD output 2.2
Change in relative price of good 1 1.0
Change in output of good 1 2.8
Change in output of good 2 -6.9
Change in real wages of unskilled workers - 1.7
Source: Author's calculations based on model described in text.
wages would be larger. Nonetheless, the exercise helps to explain why

studies that attempt to infer the effects of trade on income distribution
by looking at prices have failed to find any clear-cut effects: for plausible
parameter values, the change in relative prices associated with the
growth of NIE trade should be well within measurement error.
The Limits to Factor Price Equalization

Some of those who worry about the effects of NIE exports on OECD
labor markets might accept that these effects have been fairly small so
far; but they would argue that these effects will become much larger, as
industrialization spreads. It is possible to make a counterargument: As
newly industrializing countries grow, their comparative advantage may
shift away from products of very low skill intensity. Is it really likely that
skilled labor will be a scarcer commodity in the world economy twenty
years from now than it is today? But it is worth asking how large the po-
tential effects of trade on OECD income inequality could be in a sort of
worst case scenario.
An extreme view would be that growing international trade will lead
to full factor price equalization: that wages for unskilled labor in OECD
countries will be driven down toward their average levels in the world
as a whole. Indeed, it might seem that this is precisely what conventional
trade theory would predict: in the absence of any barriers to trade, isn't
trade a substitute for factor mobility?
This extreme view, however, neglects an important limitation to the
factor price equalization theorem: it only works as long as a country re-
mains nonspecialized. If the change in relative prices is so large that the
OECD no longer produces goods that compete with low-skill imports,
any further reduction in the relative price of these goods will have no
effect on income distribution. (This is why Wood's insistence that NIE

exports are noncompeting is so puzzling: surely this is exactly the case
in which the tendency toward factor price equalization breaks down?24)
In the context of my miniature CGE model, it is straightforward to
find the limits of factor price equalization. At a relative wage of skilled
labor 50 percent above its autarky level, and a corresponding relative
price of the skill-intensive good 14.5 percent above its original level, the
OECD economy becomes completely specialized in the skill-intensive
good. Any further change in the relative price has no effect on relative
wages.
Even this case, however, seems unlikely to occur because it implies
unreasonably large trade volumes. At the point of OECD specialization,
manufactured imports from the NIEs would reach 28.6 percent of
OECD gross product! In the context of this model, this is possible. If
the model were modified to make a reasonable percentage (at least 60
percent) of OECD expenditure fall on nontraded goods and services, the
point of complete specialization in traded goods would be reached after
a substantially smaller change in income distribution.25
The flexible-wage version of the model, then, suggests that NIE trade
can explain only a fraction of the huge increase in income inequality that
has occurred in the United States since the 1970s.26 And while it shows
that larger effects from such trade could occur in the future, it also points
24. See Wood (1994).

25. For example, adding a nontraded sector that receives 60 percent of expenditures,
and assuming that this sector initially uses skilled and unskilled labor in the same propor-
tions as the OECD endowment, the model says that NIE trade can raise the relative wage
of skilled labor by at most 17 percent.
26. The comparison between the European and the American cases may seem to sug-
gest that a little bit of wage flexibility goes a long way, and to make one wonder whether
even Eurosclerotic economies are really unable to adjust relative wages by a few percent.
It is important to be careful about making too much of this. First, the estimated impact of
NIE trade in the European case amounts to roughly 20 percent of the rise in European
unemployment since the early 1970s, while the impact in the American case amounts to
roughly 10 percent of the rise in U.S. wage inequality over the same period; given the num-
ber of ad hoc assumptions involved, this is not a major difference. Second, one should not
fall into the fallacy of imagining that since any adverse shock can be decomposed into
many smaller components, adjustment is always easy! (Achilles can, in fact, outrun the
tortoise.) Rising European unemployment is presumably the result of a number of factors.
Any one of these factors could have been offset by a small change in relative wages; but
to offset them all would presumably have required something comparable to the massive
growth in inequality that has occurred in the United States.
Paul Krugman 361
out that there are limits to the change in relative wages that trade flows
can produce.
Conclusions
The expansion of world trade is a topic that inspires hyperbole. Seem-

ingly sober commentators are easily caught up in a rhetoric that portrays
the growing integration of markets in recent years as an unprecedented
event that changes all the rules for economic analysis and policy. And
many observers, contemplating the rapid growth of manufactured ex-
ports from low-wage countries, have sounded ominous warnings about
the effect of trade on advanced-country labor markets.
The truth is considerably more prosaic. International trade has risen
substantially since World War II. Much of that growth, however, simply
reflects a recovery to levels achieved before World War I. While world
trade is a larger share of world production than ever before, the United
States is still considerably less dependent on trade than most other ad-
vanced countries were a century ago.
It is true that there are new aspects to international trade. Of these,
the most conspicuous and also most controversial is the growth of low-
wage manufactured exports. This growth almost certainly has had some
role in the growth both of unemployment in Europe and of wage inequal-
ity in the United States. A sober assessment does not, however, support
the view that NIE trade is the principal cause of these labor market prob-
lems; nor does it support apocalyptic predictions about the future effects
of that trade.
Perhaps the most important thing to say involves methodology-how
to think about the global economy. One of the disturbing aspects about
much of the popular and even the professional discussion of world trade
has been its casualness. Commentators talk loosely of countries com-
peting like big corporations, or of unskilled workers competing with
their counterparts in the third world; they rarely ask whether their sto-
ries are logically consistent, let alone whether sweeping assertions
about the implications of world trade can be reconciled with the rela-
tively modest magnitudes of actual trade flows and price changes.
There is a way to think systematically about the effects of growing
world trade: it is called general-equilibrium trade theory. Indeed, the hot
current topic of NIE trade in manufactures is almost tailormade for anal-

ysis using the techniques developed decades ago by Jagdish Bhagwati,
T. N. Srinivasan, Ron Jones, and their students. These are textbook
models whose time has come.
Comments
and Discussion
Richard N. Cooper: This is a splendid paper. It is simple, elegant, easy

to read once you get into the framework of the trade theory that Krug-
man uses. It makes a silk purse out of a sow's ear in the sense of turning
a highly simplified and abstract teaching model into a tool to provide
quantitative-not just qualitative-insight about the real world. I will of-
fer two comments on the paper, one relatively brief and one longer. Both
are empirical in nature.
The first point concerns Krugman's view that world trade in the last
half century, at least until the early 1980s, basically represents a return
to 1913. I believe he understates the extent to which barriers to trade
have declined during the whole period. I refer, especially, to technologi-
cal improvements in transportation.
It is true that the steel hull and the screw propeller were known by
1870, and that they were really important innovations in ocean trans-
port. But the average merchant ship in 1950 was between 5,000 and
10,000 tons, compared with modern ships of 150,000 tons and over, such
as supertankers, bulk carriers, container ships, and roll-on-roll-off
ships. These large ships have greatly reduced the cost of ocean trans-
port, particularly for low-value items, such that even iron ore and steam
coal can be shipped economically for long distances. Europe can import
coal from South Africa, Japan from Australia, the United States can im-
port iron ore from Liberia, and so forth. This kind of trade was not gener-
ally economical in 1913. It is true that coal was shipped long distances,
but that was mainly to coaling stations as fuel for ships, and not for use
on land.
Much more significant than those developments, in my view, has
363
been the evolution of air freight with the emergence of the long-distance
jet airliner in 1958 and, especially, of the wide body jet in 1967.
It may come as a surprise, but 29 percent of U.S. exports by value
and 21 percent of U.S. imports traveled by air in 1993. (These figures, of
course, include trade in aircraft.) These shares have been growing stead-
ily. If we exclude trade with Canada and Mexico, much of which is over-
land, then over 40 percent of U.S. overseas exports go by air, and nearly
30 percent of imports.
This represents a staggering change in the modes of international
trade from twenty or thirty years ago. Cut flowers, formerly a local item,
are now tradable over great distances. Israel is a big exporter of cut
flowers to the United States. Air freight also permits the international
organization of production slicing, which Krugman addresses, and com-
bines it with just-in-time. Goods can leave Singapore today and arrive
anywhere in the United States tomorrow.
Also noteworthy, as Krugman acknowledges, is a marked decline in
the legal barriers to trade since the 1930s. Much of this represents a re-
turn to 1913; but we have gone way beyond 1913. Tariff levels on indus-
trial products imported into the industrialized nations are now less than
10 percent of those in 1947, before the first of the eight GATT rounds of
multilateral trade negotiations took place. Trade among industrial coun-
tries, meaning Europe and Japan, was ridden with quantitative restric-
tions at that time. Those are virtually gone now.
Those of us who till the fields of trade policy wring our hands over
the multifiber agreement, antidumping duties, and current quantitative
restrictions on agriculture and so-called voluntary export restraints. But
these are small barriers compared to what existed in Europe and Japan
in the 1950s and even the early 1960s, and in developing countries, as
Krugman points out, as late as the 1980s. The large import liberaliza-
tions of the last decade have not just completed the reductions agreed
on at the Tokyo Round in 1979. Many developing countries have made
unilateral liberalizations beyond these international agreements.
Let me turn to the second part of the paper. Although it is quantita-
tive, it has the great merit of being free of facts. It is thus not possible to
quarrel over the quality of Krugman's data, or the statistical significance
of his econometric equations, which is the normal practice at Brookings
Panel meetings.
This analysis is quite explicitly an illustrative exercise. With Krug-
Paul Krugman 365
man's choice of parameters, which he draws from a combination of the

professional literature, judgment, and the requirement for internal con-
sistency, and under what he calls the European model, which assumes
fixed real wages, all existing North-South trade in manufactured goods
would reduce employment by 1.4 percent in the North.
The level of trade that we actually observe would alter the skilled-
unskilled wage differiential by 3 percent under flexible wages and actu-
ally reduce real wages of unskilled workers by 1.7 percent, compared
with a situation involving no trade.
Those are Krugman's quantitative findings. How do these figures
compare with the growing number of empirical estimates that have been
developed over recent years, some of them in Brookings Papers on Eco-
nomic Activity? I report five for comparison.
George Borjas, Richard Freeman, and Lawrence Katz impute a max-
imum of 15 percent of the growth in the college-noncollege wage differ-
ential over the 1980s to imports. I
Lawrence and Slaughter, coming at this with the perspective of trade
rather than labor economists, find that trade cannot explain any of the
growth in wage differentials. In fact, based on their assessment of the
way commodity prices moved and the way factor prices must move,
they conclude that, if anything, trade "nudged relative wages toward
greater equality."2
Sachs and Shatz, also in a Brookings Panel paper, attribute 6.2 per-
cent of the decline in unskilled employment in manufacturing over the
1980s to imports, on the very strong assumption that trade shares are
frozen at their 1978 levels.3 They criticize Lawrence and Slaughter for
not observing trends, yet they fall into the same error themselves by
making this key assumption. If they permitted a continuation of previ-
ous trends in import ratios, they would find that imports had a smaller
effect.
Wood, who has written an entire book on North-South trade and its
impact on labor markets in the North, attributes 5 percent of the decline
in employment in manufacturing in the OECD area to imports from de-
veloping countries. He then arbitrarily multiplies that number by four
1. Borjas, Freeman, and Katz (1992).

2. Lawrence and Slaughter (1993).
3. Sachs and Shatz (1994).
to allow for services, which he does analyze explicitly, and for what he
considers to be important trade-induced innovation, which on his argu-
ment will be biased against unskilled labor.4 (In that, I believe he is quite
wrong.)
I havejoined the growing crowd in this cottage industry with a partial-
equilibrium analysis.5 I am troubled by the common practice of treating
production workers as unskilled and nonproduction workers as skilled.
That seems much too coarse an assumption. I focus instead on the least
skilled members of the labor force, measured by the wage profile. This
criterion directs attention to the textile, apparel, and leather (TAL) in-
dustries. I conclude that 10 percent of the relative decline in the wages
of unskilled production workers can be attributed to imports into the
United States of unskilled-labor-intensive goods, that is, TAL goods.
I apply the same model to the six largest European countries, of
which the Netherlands is the smallest. Surprisingly, I find that relative
wage movements, although less dramatic than in the United States, gen-
erally have been sufficient to absorb the unskilled labor released by the
TAL industries. The significant exception is France, where the wage
structure was exceptionally rigid over the 1980s. That has to do, inter
alia, with the fact that France has the world's highest minimum wage.
How do I justify a partial-equilibrium analysis and a focus on the tex-
tile, apparel, and leather industries? Here, let me introduce some rele-
vant facts about the United States (similar facts hold for the European
countries), which Krugman's model ignores.
There were thirteen million production workers in U.S. manufactur-
ing in 1990. In the three industries that I focus on, there were 1.9 million.
The decline in the number of production workers in manufacturing over
the 1980s amounted to about one million, of which 40 percent were in the
TAL industries. Thus the bulk of the decline in employment of unskilled
workers, measured by the wage profile, was in these three industries.
Those figures have to be contrasted with the twenty-seven million
production workers in U.S. retail trade, restaurants, and hotels in 1990,
whose number grew by six million over the decade. These include the
McDonald's hamburger flippers that we keep hearing about. Again, by
the wage profile these are the least skilled members of the American la-
bor force.
4. Wood (1994).
5. Cooper (1994).
Paul Krugman 367
The key point is that the employment changes in the tradable manu-
facturing sector were relatively small against the large scale and the
large increase of employment in the nontradable sectors. I calculate the
import-induced reduction in employment in the TAL industries, and in
the in relative wage movements needed to absorb those released work-
ers into the huge nontradable sector, which has a big demand for un-
skilled workers. The result is the 10 percent mentioned earlier. Like
Krugman, I expected to find some unemployment effects in Europe due
to wage rigidities, but I discovered that wage flexibility was sufficient,
except in France, to absorb workers released from the TAL industries
into the retail sector.
Let me close with some remarks on the intellectual framework behind
my analysis because it is rather different from Krugman's. The big dif-
ferences are that he discounts the large nontradable sector, and also re-
lies, in a decisive way, on complete homogeneity of the traded goods.
While imported manufactured goods may be good substitutes for do-
mestic products, they are not completely homogenous, and this is an im-
portant difference. I find, as Lawrence and Slaughter do, that the rela-
tive (value added) prices of textiles, apparel, and leather goods to all
manufactured goods rose over the decade of the 1980s.
How can that happen in the face of stiff import competition? Import
competition from developing countries puts indigenous industries under
competitive pressure, and they respond in two ways. The first is by con-
solidating, shedding labor, and perhaps going out of business. The other
is by upgrading their product. Firms try to differentiate their products.
Under these conditions, it is possible for the price of import goods to fall
even while the price of competing domestic goods rises, because they
are for slightly different products that serve the same function: men's
shirts, for example. On this view, labor is shed from the tradable sector
even with flexible wages. The question then becomes: How good is the
economic system at absorbing the released labor with little decline in
wages? That is where a large nontradable sector plays a critical role.
In summary, Krugman's results do a nice job of bounding the impact
of import competition in the context of classical trade theory and identi-
fying why the plausible effects are small. But two elements of realism
would modify his model. First, because goods are nonhomogenous,
complete specialization in similar but differentiated goods may occur
earlier than Krugman allows. Second, the existence of a large nontrada-
ble sector limits the impact of competition from imports on domestic

wages.
T. N. Srinivasan: I enjoyed reading Krugman's paper and, as always,

was rewarded with penetrating insights and left with envy at his elegant
and parsimonious modeling. His earlier inveighing against vacuous no-
tions of competitiveness and dangerous practices of managed trade,
through its cogent reasoning, put all rational individuals in his debt. He
has now done a great service by analyzing an issue in which emotions,
rather than reasoning, have played a major role.
Richard Cooper, in his masterly way, has surveyed the scene and has
preempted several of my points. I will just add one or two with respect
to the facts relating to trends in international trade. Krugman quotes the
proportion of trade to GDP, going all the way back to 1850. Anyone who
has looked at how trade and GDP data are put together would caution
against placing too much reliance on these numbers. I Even if they were
reliable, it would be hard to interpret them. After all, theory does not
indicate a monotone relation between GDP growth and the trade-to-
GDP ratio. For example, take the standard Cass-Koopmans model of
optimal growth in an economy that is specialized in producing consumer
goods and imports investment goods. Suppose there is no population
growth, capital lasts forever, and there are no intercountry capital
movements. Given a positive discount rate, there exists a unique steady
state in which the investment rate is zero. As such, imports would also
be zero. During the transition to the steady state from an initial capital
stock that is below its steady state value, there would be investment and
hence, imports and a positive trade-to-GDP ratio that converges to zero
in steady state along an optimal growth path. Thus in theory it is possible
for the trade-to-GDP ratio to decline rather than increase as GDP grows.
As Krugman points out, geographical aggregation matters. Think of
a two-country trading world, with each country specialized in its own
export good, and importing what it consumes entirely from the other
country. The trade-to-GDP ratio would be 200 percent for each country,
by definition, because exports and imports both equal the value of GDP.
1. See, for example, Rozanski and Yates (1994) and Srinivasan (1994).
Paul Krugman 369
Now if the two countries combine and form a one-country world, then
all international trade disappears and the ratio falls from 200 percent to
0 percent! Krugman also mentions the changing characteristics of inter-
national trade and draws attention to the increasing importance of intra-
industry trade. Depending upon geographical and temporal aggregation,
as well as aggregation over states of nature, published data could mag-
nify the extent of intra-industry trade.
I agree with Cooper that it is an exaggeration to suggest, as Krugman
does, that all the major technological innovations influencing trade are
those relating to transport and communications, and that these (namely
the railroad, the steamship, and the telegraph) had already taken place
on the eve of World War I. Let me give an example of how a process
innovation could also enable the slicing up of the value added to which
Krugman draws attention. This relates to the steel industry, which fig-
ures in the beautiful mural by Diego Rivera in Detroit mentioned by
Krugman. In this mural, apparently, at one end there are blast furnaces
and at the other end automobiles come out. Now there is a process for
producing steel using natural gas as a reducing agent instead of coke.
This process produces sponge iron which, in turn, is used along with
steel scrap in electric arc furnaces.
With this process there is no need to have a steel plant next to a coal
mine or an iron ore mine, and there are no significant scale economies.
The steel industry became "footloose"-and many "mini" steel plants
of this nature have come up all over the world. Krugman understates the
importance of such process innovations, as well as recent technological
changes such asjumbojets for passenger and cargo transport, container-
ization, computer and communication technology, in the slicing-up of
value added.
Krugman is right in emphasizing that massive reductions in trade bar-
riers since 1913 have contributed to the growth in world trade. However,
the trend toward increasing barriers in the post-World War II era is
more important than Cooper allows. He suggests that the textiles, ap-
parel, and leather industries are more relevant than high technology in-
dustries for the debate on wage trends in the United States and Europe.
But these are the very industries in which the barriers grew after the Sec-
ond World War. The notorious multifibre arrangement (MFA) did not
exist prior to 1960. Initially it was a short-term agreement in cotton tex-
tiles, aimed mainly at exports from Japan; and it was soon extended to
all fibers known to man and god. Besides, anytime a country that was not
under the MFA began exporting a noticeable amount, the industrialized
countries slapped a quota on it and brought it into the MFA! One should
not, therefore, understate this tendency for increasing barriers to trade
when it comes to exports from developing countries.
The Uruguay Round agreement will phase out MFA in ten years. The
agreement also rules out the use of "gray area" measures such as volun-
tary export restraints, and brings greater transparency and discipline on
the antidumping and countervailing duty actions. One hopes that the
agreement will be implemented, and not violated or evaded. Unfortu-
nately the United States has set a bad example with its recent unilateral
decision to double tariffs on selected Japanese automobiles from levels
that had earlier been bound under GATT, and by doing so without wait-
ing for the dispute settlement mechanism of the World Trade Organiza-
tion to decide on its complaint against Japan.
Before I turn to Krugman's stylized model of global trade, let me say
that I agree with Leamer' s withering criticism that many of the empirical
studies, particularly those by labor economists, do not apply the stan-
dard theorems of international trade correctly.2 They do not recognize
that trade is quintessentially an endogenous phenomenon. Besides, ana-
lyzing trade requires general-equilibrium thinking-partial-equilibrium
models of labor economies are inappropriate.3
My skepticism of the empirical literature arises from the fact that ob-
served changes in employment and wages over time are, in principle,
influenced by changes in demand and supply in the relevant product and
factor markets. A well-specified structural model that distinguishes
exogenous forcing variables (possibly taste, technology, and policy
shocks) from endogenous responses, and takes into account the relevant
leads, lags, and expectations, has to be estimated with some plausible
identifying restrictions. As Leamer rightly argues, we are most likely to
make progress if the empirical model is linked clearly with some under-
standable theory. The empirical literature unfortunately fails, by this
test, to be convincing.
Krugman's stylized model certainly provides an understandable the-
ory. Whether it is more than a way of organizing our thinking for doing
2. Leamer (1994).
3. The essays in Bhagwati and Kosters (1994) provide a cogent critique of the liter-
ature.
Paul Krugman 371
Figure Dl. Adjustment under the "European" Model
Q2
Rs
0 Ru Q1
more serious empirical work is arguable. His figures 3, 4, and 5 have

been collapsed into my figure DI. Start from autarky production and
consumption point A. With relative wages and commodity prices (that
is, the slope of the common tangent of the production possibility and
Samuelson social indifference curves at A) fixed, the input coefficients
in production remain unchanged. At A there is full employment of both
factors, unskilled and skilled labor, at the initial endowment levels. Now
keeping the endowment of skilled labor fully employed, if we reduce the
employment of unskilled labor, the production point (at fixed input coef-
ficients) moves down the so-called Rybczynski line ARu. Analogously,
if we keep the endowment of unskilled labor fully employed but reduce
the employment of skilled labor, the production point moves up the Ryb-
czynski line ARs.
Now with the relative commodity price fixed, the only way to adjust
to trade when it is opened is by adjusting in output and employment.
Thus to generate the export of skilled-labor-intensive good 1 in ex-
change for imports of unskilled-labor-intensive good 2 supplied by the
trading partners at the fixed commodity price, production has to move to
Q along the Rybczynski line ARu, and consumption (given homothetic
preferences) to C along the ray OA. CQ has the same slope as the fixed
commodity price ratio. With Q as the origin, the foreign offer curve of
Krugman's figure 3 will pass through C, so that CO' units of good 2 are
imported in exchange for O'Q units of good 1.
The unemployment of unskilled labor resulting from the shift in pro-
duction from A to Q can be read by drawing a line parallel to the Ryb-
czynski line ARs to meet the vertical axis at D. Clearly, by assumption,
along ARs unskilled labor is fully employed, so that by choice of units of
measurement we can make ORs represent the full employment of un-
skilled labor. By the same token, with input coefficients remaining the
same, the employment of unskilled labor at Q is the same as at D, and it
can equal OD by our choice of units. Thus unemployment is DRs. By
drawing a line parallel to ARs through C to meet the vertical axis at B, it
is seen the unskilled labor employment content of consumption is OB.
Thus the unemployment due to the movement of consumption from A
to C is BRs. The unemployment arising from trade is DB, which is less
than the total unemployment, DRs, and DRs/DB is Krugman's multi-
plier effect. This indeed is the message of "European" adjustment,
through output and employment changes but with no price or wage
changes, to the opening up of trade.
I depict the "American" adjustment, which allows price and wage
changes as well, in figure D2, where OC is the foreign offer curve. Under
free trade and full employment of both factors the American, or home,
offer curve is OF, the slope of which at the origin is the autarky price,
OA. Thus with price adjustment, equilibrium trade is at T', instead of at
T as in the "European" case with no adjustment. With the shapes of the
offer curves as drawn, both the relative price of the unskilled-labor-in-
tensive good and the volume of trade are obviously lower, at T'. Thus
allowing prices to adjust reduces the volume of trade to which consump-
tion and production would have to adjust. Of course, as we saw earlier,
the fall in relative price will lower the relative wage of unskilled labor
enough to keep both factors fully employed.
Turning from the stylized model to the numerical simulations based
Paul Krugman 373
Figure D2. Adjustment under the "American" Model

M2
/F
o x
on it, I must admit to some skepticism as to their significance. The nu-

merical calibrations that Krugman uses for his trade model might indeed
capture the initial equilibrium configuration. However, in order to move
from one equilibrium to another one has to make functional form as-
sumptions as well. Krugman is candid in recognizing this when he says,
"the model will be made Cobb-Douglas throughout." I am not as confi-
dent as he that the unitary elasticities of substitution are not "too far
from most empirical estimates." Perhaps for the purposes he has in
mind, namely, to demonstrate nonzero but modest effects of adjusting
to trade 'a la Europe or ai la America, the simulations are adequate. In
any case, compared to most of the empirical estimates in the literature,
which are not adequately grounded in some well-specified economic
theory, Krugman's simulation at least has the virtue of coherence!
General Discussion
A number of panelists questioned the adequacy of Krugman's basic
model for explaining unemployment. Maurice Obstfeld suggested that
perfect labor mobility assumed in Krugman's "European" model leads
it to understate the effect on unemployment. In Krugman's model the

decline in the low-skill-intensive domestic industry only affects the un-
employment of low-skilled workers; high-skilled workers displaced
from this industry are readily absorbed in the expanded skill-intensive
industry. This expansion employs some fraction of the unskilled as well.
Under the extreme assumption that labor is completely immobile and
assuming Krugman's factor intensities, a decline in the low-skill-inten-
sive industry would give rise to 50 percent greater unemployment than in
Krugman's calculations. Krugman discussed whether the nontradable
sector itself is low-skill-intensive. Against Cooper's example of retail
trade, Krugman offered school teachers, doctors, and chiropractors as
types of skill-intensivejobs in the nontradable sector. He cited empirical
research by Borjas, Freeman, and Katz which indicates that traded
goods are slightly less skill-intensive than nontraded goods.
William Nordhaus suggested several feature of unemployment and
the job market that were not captured by Krugman's simple model. Un-
employment in Europe does not appear to be concentrated in one age or
skill group, and net job losses in import-competing industries conceal
significant job creation. Between 1973 and 1988 the apparel industry in
the United States experienced job destruction of 14.4 percent and job
creation of 10.8 percent, according to John Haltiwanger. These facts
demonstrate that trade is only one of several influences on the labor mar-
ket and imply that disentangling trade from factors such as technology
will not be easy.
Nordhaus provided evidence that the ratio of shipments to value
added for U.S. manufacturing establishments is actually smaller today
than in 1930. Unless it turns out that particular tradable industries have
evolved differently than these overall data indicate, this evidence calls
into question Krugman's suggestion that "slicing up the value chain" is
an important factor in explaining the increase in trade. While agreeing
that trade liberalization in developing countries has played a key role in
the dramatic acceleration in the world trade-to-GDP ratio since 1985,
Maurice Obstfeld felt that Krugman gave too little attention to other fac-
tors. Trade in services has expanded as a proportion of overall trade,
reducing the significance of transport costs. And treaties, the develop-
ment of commercial law, greater familiarity with foreign business prac-
tices, and developments in capital markets which reduce the cost of
trade credits have all facilitated trade expansion.
Paul Krugrnan 375
There was some discussion of the difficulty in establishing a baseline

for the growth of trade so as to be able to assess the importance of trade
liberalization and other political changes. Krugman suggested that it is
not obvious that the trade share should rise with income. One trend
working in the opposite direction is that the share of nontradables in
spending tends to rise as GDP rises. He reported research by Vern Hen-
derson showing that single-industry towns in the United States may
have over 50 percent of their employment outside of their main industry,
but in Brazil the proportion is 30 to 40 percent. This suggests that eco-
nomic growth may actually increase the demand for locally produced
goods and services. In the same vein, world growth may spread the dis-
tribution of capital and technology.
The discussion turned to the importance of immigration to wage in-
equality. Benjamin Friedman observed that certain low-skill jobs in the
nontradable sector seem to be performed disproportionately by immi-
grants for example, janitors, hotel staff, and barbers. In the presence of
a shrinking demand for low-skill labor in the tradable sector, this would
imply that immigration policy is important for the trend in wage inequal-
ity. He suggested that a policy favoring high-skill immigrants might
make sense, especially since other policies to deal with inequality ap-
pear to be extremely expensive. For example, citing an estimate by
James Heckman, he noted that a program seeking to amend the falling
fortunes of the lowest income quintile via human capital investments
would have a bill approaching two trillion dollars.
William Brainard suggested that even if economists do not consider
immigration policy, politicians will. Throughout U.S. history, immigra-
tion policy has tended to change with the employment situation. Today
it is certainly an important issue in California. Krugman noted that the
movement of southern European guest workers between jobs in north-
ern Europe has diminished, with a negative net migration from southern
to northern Europe since 1973. However, Richard Cooper thought this
can be explained by the fact that southern Europe has been booming,
and is not necessarily due to policy changes in the north.
References
Balassa, Bela. 1966. "Tariff Reductions and Trade in Manufactures among the
Industrial Countries." American Economic Review 56(3):466-73.
. 1979. "The Changing International Division of Labor in Manufactured
Goods." Banca Nazionale del Lavoro Quarterly Review, 130(Sept.):243-85.
Barraclough, Geoffrey, editor. 1978. The Times Atlas of World History. Lon-
don: Times Books.
Bhagwati, Jagdish, and Marvin H. Kosters, editors. 1994. Trade and Wages:
Leveling Wages Down? Washington: AEI Press.
Bojas, George J., Richard B. Freeman, and Lawrence F. Katz. 1992. "On the
Labor Market Effects of Immigration and Trade." In Immigration and the
Workforce: Economic Consequences for the United States and Source
Areas, edited by George J. Borjas and Richard B. Freeman. Chicago: Univer-
sity of Chicago Press.
Brecher, Richard A. 1974. "Optimal Commercial Policy for a Minimum-Wage
Economy." Journal of International Economics 4(2): 139-49.
Commission of the European Communities. 1993. Growth, Competitiveness,
Employment: The Challenges and Ways Forward into the 21st Century-
White Paper. Luxembourg: Office for Official Publications of the European
Communities.
Cooper, Richard N. 1994. "Foreign Trade, Wages, and Unemployment." Dis-
cussion Paper 1701. Cambridge, Mass.: Harvard Institute of Economic Re-
search (November).
Frankel, Jeffrey A., Shang-Jin Wei, and Ernesto Stein. 1994. "APEC and Re-
gional Trading Arrangements in the Pacific. " Working Paper 94- 1. Washing-
ton: Institute for International Economics (January).
Fuchs, Victor R. 1968. The Service Economy. New York: National Bureau of
Economic Research.
Great Britain, Central Statistical Office. 1994. AnnuialAbstract of Statistics.
Grubel, Herbert. 1967. "Intra-Industry Specialization and the Pattern of Trade."
Canadian Journal of Economics 33:374-88.
Helpman, Elhanan, and Paul Krugman. 1985. Market Structure and Foreign
Trade: Increasing Returns, Imperfect Competition, and the International
Economy. Cambridge, Mass.: MIT Press.
Katz, Lawrence F. 1992-93. "Understanding Recent Changes in the Wage
Structure." NBER Reporter (Winter): 10-15.
Lawrence, Robert Z., and Matthew J. Slaughter. 1993. "International Trade and
American Wages in the 1980s: Giant Sucking Sound or Small Hiccup?"
BPEA, Microeconomics, 2:1993, 161-210.
Leamer, Edward E. 1993. "Wage Effects of a U.S.-Mexican Free Trade
Agreement." In The Mexico-U.S. Free Trade Agreement, edited by Peter M.
Garber. Cambridge, Mass.: MIT Press.
Paul Krugman 377
. 1994. "Trade, Wages and Revolving Door Ideas." Working Paper 4716.
Cambridge, Mass.: National Bureau of Economic Research (April).
Liesner, Thelma. 1989. One Hundred Years of Economic Statistics. London:
Economist Publications.
McCallum, John. 1995. "National Borders Matter: Canada-U.S. Regional
Trade Patterns." American Economic Review 85(3): 615-23.
Mitchell, B. R. 1988. British Historical Statistics. Cambridge: Cambridge Uni-
versity Press.
Organisation for Economic Co-operation and Development (OECD). 1992. His-
torical Statistics 1960-1990. Paris: OECD.
Rozanski, Jerzy, and Alexander Yeats. 1994. "On the (In)Accuracy of Eco-
nomic Observations: An Assessment of Trends in the Reliability of Interna-
tional Trade Statistics." Journal of Development Economics 44(1):103-30.
Sachs, Jeffrey D., and Howard J. Shatz. 1994. "Trade and Jobs in U.S. Manufac-
turing." BPEA, 1:1994, 1-69.
Samuelson, Paul A. 1949. "International Factor-Price Equalisation Once
Again." Economic Journal 59(234): 181-97.
Srinivasan, T. N. 1994. "Data Base for Development Analysis: An Overview."
Journal of Development Economics 44(1):3-27.
United Nations. 1993. Trends in InternationalDistribution of Gross WorldProd-
uct. National Accounts Statistics Special Issue. New York: United Nations.
United Nations Conference on Trade and Development (UNCTAD). 1994.
Handbook of International Trade and Development Statistics 1993. New
York: United Nations.
Wood, Adrian. 1994. North-South Trade, Employment and Inequality:
Changing Fortunes in a Skill-Driven World. New York: Clarendon Press.
World Bank. 1994. World Development Report 1994: Infrastructure for Devel-
opment. New York: Oxford University Press.
. 1995. Global Economic Prospects and the Developing Economies.
Washington: World Bank.

Krugman 1995

Uploaded by

Copyright:

Available Formats

Krugman 1995

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Krugman 1995

Uploaded by

Copyright:

Available Formats

Growing World Trade: Causes and Consequences

Author(s): Paul Krugman, Richard N. Cooper, T. N. Srinivasan

Growing World Trade:

WHAT ASPECT OF the American economy has changed most in the

tures exports using an "American" approach, in which wages are

The Growth of World Trade: An Overview

It is a late-twentieth-century conceit that we invented the global

Trade as a Share of Output: A Long-term Perspective

Table 1. World Merchandise Exports as Percentage of GDP

Country 1913 1950 1970 1987

tury.2 The data show a marked increase in trade up to 1913; a substantial

be expected, the aggregates conceal several novel features of modern

New Aspects of Trade

Table 3. Commodity and Geographical Composition of U.K. Trade

of modern industry (and also, to his patron's dismay, a condemnation of

5. World Bank (1994).

The emergence of supertrading economies clearly depends on the

Table 4. Exports of Manufactured Goods from Newly Industrializing Economies

chain, creates a greater, not a smaller opportunity to relocate production

Why Has World Trade Grown?

When economic commentators try to explain why world trade has

It is difficult to quantify protectionism. Tariff rates can be measured,

(2) Ti= k(1 -si),

In other words, an individual country's trade would be larger, the

9. United Nations (1993).

11. See Fuchs (1968).

trade whatever can be traded, a declining share of the economy consists

A Stylized Model of Global Trade, Employment, and Wages

pean Commission, in its 1993 White Paper Growth, Competitiveness,

Structure of the Model

12. Commission of the European Communities (1993, p.4).

It will be convenient to assume that tastes are homothetic, so that mar-

(9) M2= T(X1).

Source: Samuelson (1949).

The most important relationships are presented in figure 1, which re-

Putting Numbers to the Model

Initial ratio of skilled to unskilled wages 2

-Remaining entries. These follow from the previous numbers. In

The Impact of Low-Wage Exports: A "European" Approach

Most analyses of the impact of NIE trade have had an "American"

Figure 2. Adjustment of OECD Employment under the "European" Model

The consequences may be analyzed in stages.'8 First, note that with

Figure 4. Adjustment of OECD Production and Consumption under the "European"

Figure 5. Adjustment of OECD Employment and Consumption under the "European"

by exports falls short of those displaced by imports. The other popular

must both lie on such a line. Equilibrium involves a situation in which

The Employment Effects of NIE Exports

Table 5. Employment Effects of Trade under "European" Assumption

the results change if we "Americanize" the model, allowing wages to be

The Impact of Low-Wage Exports: An "American" Model

With full employment of both skilled and unskilled labor maintained

20. See, in particular, Borjas, Freeman, and Katz (1991).

Figure 6. Adjustment of OECD Production and Consumption under the "American"

Source: Author's model as described in text.

bodied services become hard to make sense of when relative factor

Making the Model Computable

22. Lawrence and Slaughter (1993).

(1 1) a=I = w- w2'3, aS2 =w -2'3, au2 --w1/3

Given these input coefficients, output is determined by the requirement

(12) Ls = asiQi + aS2 Q2 L = auIQI + aU2Q2,