You are on page 1of 29

The Semiconductor Industry

Author(s): Douglas A. Irwin


Source: Brookings Trade Forum, (1998), pp. 173-200
Published by: Brookings Institution Press
Stable URL: http://www.jstor.org/stable/25063126 .
Accessed: 28/06/2014 08:51

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

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

Brookings Institution Press is collaborating with JSTOR to digitize, preserve and extend access to Brookings
Trade Forum.

http://www.jstor.org

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
DOUGLAS A. IRWIN
Dartmouth College and
National Bureau of Economic Research

The Semiconductor Industry

than a decade has passed since the U.S. semiconductor

More industry was buffeted by a severe recession that pushed many


of its firms to the brink of bankruptcy and forced them out of the

production of memory chips. Threatened in addition


by the growing
market share of Japan's producers, who had received some govern
ment assistance in what was viewed as a closed domestic market, U.S.
firms began issuing
dumping complaints. This move led not to the

imposition of antidumping duties, but to an unusual and controversial


arrangement that set a floor on Japanese export prices.
While the semiconductor industry is vastly different today from
what it wasin the mid-1980s, its experience continues to be instruc
tive about the general relationship between antidumping and com

petition policy, particularly in high-technology industries. One of its


striking lessons is that the mere threat of antidumping actions and
especially the imposition of antidumping measures can seriously re
duce foreign competition and even lead to cooperative producer be
havior.

The Structure of the Semiconductor Industry

Semiconductors are a sophisticated, microelectronic product used


in an increasing multitude of goods, from televisions and microwave
ovens, aircraft and automobiles, computers and calculators to tele

phones and watches. The largest category of semiconductor devices is


the integrated circuit, which accounted for more than 80 percent of
U.S. semiconductor consumption in 1985 and includes several types of
173

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
174 Brookings Trade Forum: 1998

products, notably (for arithmetic


logic chips and decisionmaking func

tions), microprocessors (the central processing unit in computers),


and various application-specific integrated circuits (configured for

particular user needs).


Much of the intense international competition and consequent
trade dispute in the 1980s and thereafter centered on a unique set of

digital integrated circuits?memory chips?which are used primarily


in computers to store and retrieve data in various forms.1 Memory

chips accounted for 18 percent of all U.S. semiconductor purchases in


1985. In the mid-1980s, dynamic random-access memories, the largest
volume of all semiconductor products, made up 7 percent of the total
U.S. market, and electronic programmable read-only memories ac
counted for another 3 percent.
The market for memory chips can be very competitive: DRAMs

(and to a lesser extent EPROMs) are a standardized product and are


almost perfectly interchangeable regardless of which firm produces
them. Product homogeneity means that DRAMs are essentially a com

modity chip, containing fewer proprietary technologies than other spe


cialized types of semiconductor devices, such as microprocessors.
DRAMs are marked by well-defined generations that give rise to dis
tinct, and relatively short, product cycles.2 The fixed costs of producing
semiconductors are very high: expenditures on the specialized (fixed)
capital required to produce semiconductors?mainly in the form of

equipment and production facilities?often run at 15-20 percent of


sales, and the ratio of research and development to sales is on the order
of 10-15 percent (in contrast to about 3 percent across all U.S. indus

tries). Yet the marginal costs, the actual material and labor costs of

manufacturing semiconductors, are quite small. During periods of


weak demand, firms are therefore tempted to undercut the price of ri
vals by pricing below average costs, undermining industry profitability.

1. Random-access memories (RAMs) temporarily store data or information; dy


namic random-access memories (DRAMs) are designed to store large amounts of

data, while static random-access memories are faster but hold less informa
(SRAMs)
tion. Read-only memories store data more permanently than RAMs; eras
(ROMs)
able programmable read-only memories (EPROMs) allow data programs to be easily
erased and reprogrammed.
2. In 1970, the IK random-access memory chip (capable of storing 1,024 bits of

information) was introduced. This was followed by the 4K chip in 1973,16K in 1976,
64K in 1979,256K in 1982, IM in 1985,4M in 1989, and 16M in 1991.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 175

Production is also characterized by dynamic scale economies in


the form of bounded learning by doing.3 This generates steep price
declines after memory chips are introduced. As the produce matures
and firms
achieve a high yield of usable chips, these economies are
exhausted and prices level off. The learning process also creates an
incentive to price below
for firms average static marginal and even
costs early in the production cycle. Current production lowers the
future costs of production, so firms can price below current (static)
average (and perhaps even marginal) costs in order to sustain a

larger initial production volume from which production costs can


be lowered in the future.4 This practice is called "forward pricing,"
a form of dumping in that price may be below average cost for a

period, but it is not necessarily a manifestation of anticompetitive


behavior.

High fixed costs and learning by doing might be thought to imply


that only a small number of firms can survive in this segment of the

industry. Yet for each recent generation of DRAM, roughly fifteen to


twenty firms (mainly in the United States, Japan, Europe, and South

Korea) have undertaken production. Production technology diffuses

rapidly and helps account for the large number of firms in the memory
chip market. Late entry is facilitated
by the licensing of the requisite
production technology (known as second sourcing) from the innovat

ing firms; firms may trade off R&D expenses against licensing ex
penses in determining the timing of market entry.5 Semiconductor
manufacturing equipment can be readily purchased from any of sev
eral suppliers. Some benefits of learning by doing also spill over across
firms, in which case entrants can learn from the production experience
of other firms to reduce their own production costs. Barriers to exit
are insubstantial because rapid technological change quickly depreci
ates capital investments.
There is mixed evidence on the significance of intergenerational

spillovers, in which production experience in one DRAM generation

3. Learning by doing, wherein past production experience provides valuable infor


mation that allows firms to reduce their cost of production, generates dynamic econo
mies of scale in that firms with a large output should have cost
(cumulative)
advantages over other producers. See Irwin and Klenow (1994).
4. See Dick (1991); Flamm (1993b).
5. See Dick (1992).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
176 Brookings Trade Forum: 1998

provides a competitive advantage in the next generation.6 The pres


ence of intergenerational spillovers might allow temporary trade and
industrial policies to have a permanent and persistent effect on the
output of domestic firms. Large semiconductor producers have often
been able to maintain
their competitive position through several gen
erations of DRAMs. At the same time, Micron Technology, a small
semiconductor firm in Boise, Idaho, successfully initiated production
in 64K DRAMs with no previous production experience. Several
other firms (such as Japan's Oki Electric and South Korea's Goldstar)
have either entered the market in later generations or have skipped

production in certain generations of DRAMs. Indeed, several of


South Korea's firms successfully entered the DRAM market in the

early 1980s and became more successful after the imposition of anti

dumping sanctions
against Japanese producers.7
Competition among DRAM producers is particularly robust when
many firms are producing the same range of products in the market,
although competitive conditions can vary over the product cycle.
Firms starting production in a new generation of memory chips may
have some market
power, though their product still competes against
the previous generation and this power diminishes rapidly as other
firms commence production. Table 1 indicates that from 1982 to 1985
the top five Japanese producers accounted for more than 75 percent
of 256K DRAM sales, although as the product matured this propor
tion fell to about 40 percent by 1989. In 1984-86, the top five Japanese
producers accounted for more than 90 percent of IM DRAM sales,
with this share falling to about 45 percent by 1989.

U.S.-Japan Trade Friction in Semiconductors

U.S. semiconductorproducers dominated the market from its ori

gins through the 1970s. Japanese producers made increasing inroads


into the market in the 1970s as demand for semiconductors shifted

6. See Irwin and Klenow (1994).


7. South Korean firms received modest indirect government support. "Without
direct or trade-restricting policies to protect them from foreign competition,
subsidy
Korean firms entered the high-density memory commodity market," ac
successfully
cording toYoon (1988, p. 274).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 111
Table 1. Percentage of Global Production of Dynamic Random Access Memory
Chips, Four Regions, 1979-89_

Percent

Producers

Herfindahl
Hirschman Top five, Other,
Year index Japan Japan US. EC Korean

64KDRAM
1979 0.525 67 0 33 0 0
1980 0.264 59 0 41 0 0
1981 0.178 67 6 28 0 0
1982 0.129 60 6 33 0 0
1983 0.108 53 7 38 2 0
1984 0.092 48 10 38 4 0
1985 0.091 54 7 19 4 4
1986 0.099 56 11 31 5 10
1987 0.106 44 13 20 6 17
1988 0.170 19 22 21 1 37
1989 0.273 4 17 34 n.a. 45

256K DRAM
1979 0.000 0 0 0 0
1980 0.000 0 0 0 0
1981 0.000 0 0 0 0
1982 1.000 100 0 0 0
1983 0.265 92 1 7 0
1984 0.213 89 3 9 0
1985 0.165 82 3 15 0
1986 0.135 77 5 24 2
1987 0.102 55 11 15 9
1988 0.091 46 18 28 7
1989 0.078 43 19 25 10

IMDRAM
1979 0.000 0 0 0 0
1980 0.000 0 0 0 0
1981 0.000 0 0 0 0
1982 0.000 0 0 0 0
1983 0.000 0 0 0 0
1984 0.000 0 0 0 0
1985 0.964 99 0 1 0
1986 0.369 87 0 13 0
1987 0.347 93 6 1 0
1988 0.173 78 10 6 2
1989 0.135 54 16 14 11
1990 0.110 45 15 21 15
Source: Flamm (1996, p. 256).
n.a.Not available.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
178 Brookings Trade Forum: 1998

away from government purchases for the military and space programs
toward commercial applications such as consumer electronics. The rise
of Japan's semiconductor industry was driven by the rapid expansion in
demand for transistors from the domestic consumer electronics indus
try. The different basis for the growth of the American and Japanese
semiconductor industries reflected the considerable differences in
end-use demand for semiconductors in the two markets: in Japan, con
sumer electronics provided 47 percent of semiconductor demand (the
comparable figure in the United States being 8 percent), while data

processing accounted for 44 percent of U.S. demand in 1984.8


Because of the different structure of final demand, firms in Japan
also developed a different structure to meet that demand, namely
through vertical integration. The U.S. semiconductor industry con
sisted primarily of independent merchant producers of modest size
that sold under contract or in arms-length transactions to unaffiliated

purchasers in other industries. In Japan, the major semiconductor


producers were also the major semiconductor consumers (that is,
electronics firms) in large, vertically integrated conglomerates. These
firms also tended to specialize in the production of certain
types of
semiconductors and trade them with another one
on the basis of
long-term contracts or long-standing ties to one another. Japanese
firms are also often affiliated with a large bank that could play a role
in corporate governance through equity ownership and corporate
board participation. Such bank ties not only gave Japanese semicon
ductor firms easier access to capital but allowed them to weather
industry downturns much better than their U.S. counterparts.9 Thus
large and diversified Japanese firms with deep financial resources to
undertake investments and sustain losses were pitted against undiver
sified, medium-size merchant firms in the United States.

Japan encouraged domestic semiconductor


production by protect
ing the home market: before 1975, imports of semiconductors into
Japan were restricted by formal quotas and prior approval require
ments, and foreign investment was essentially forbidden. The Ministry
of International Trade and Industry (MITI) also sponsored the Very

8. See Irwin (1996).


9. Hoshi, Kashyap, and Scharfstein (1990) find evidence that Japanese firms with
bank ties produce and invest more in periods of financial distress (cash flow disrup
tions) than firms without such bank connections.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 179

Large Scale Integration (VLSI) program from 1976 to 1979, an R&D


project that directed $200million in funds (not including interest-free
loans and tax breaks) over four years to several major semiconductor
manufacturers, such as Fujitsu, Hitachi, Mitsubishi, Nippon Electric

Corporation, and Toshiba.

By the late Japan had emerged


1970s, as a major competitor in
memory chips, particularly DRAMs. Japanese firms devoted an aver

age of 28 percent of their sales to capital spending from 1978 to 1985,


while the comparable figure for U.S. firms was 16 percent.10 These
capacity investments translated into a steady increase in Japan's share
of the world semiconductor market. At the end of the 1970s, U.S. firms
accounted for more than 60 percent of the world market and Japanese
firms less than 30 percent; by mid-1985, the market shares of the two
countries were about equal at 45 percent, after which time the Japa
nese took the lead. The rise of the Japanese semiconductor industry
was most pronounced in memory chips, particularly DRAMs, the
natural class of semiconductors for entrants (such as South Korea

later) to begin production, given the relatively straightforward design


technology compared with other semiconductor devices. In the world
wide DRAM market, the United States and Japan "traded places":
the U.S. share decreased from 70 percent in 1978 to less than 20
percent in 1986, while the Japanese share rose from less than 30

percent in 1978to a peak of 75 percent in that same period.11


The rapid growth in the semiconductor market in Japan, served

mainly by Japanese producers, was the fundamental driving force be


hind their increased world market share. The rapid expansion of Japa
nese DRAM production also led to an increase in Japan's exports to
the United States: imports of 64K DRAMs from Japan as a share of
U.S. consumption stood at about
35 percent in 1983,29 percent in 1984,
and 30 percent in 1985.12 Japan's share of the US. market was not fully
indicative of the force of the new competition in DRAMs because, in
an integrated world market, Japanese producers could only capture
market share abroad by forcing the market price downward in world
markets, regardless of which firms were selling in a particular market.

10.OECD (1992, p. 147).


11. See Tyson (1992, p. 106, fig. 42).
12.USITC(1986,p.A42).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
180 Brookings Trade Forum: 1998

The competitive position of the U.S. industry was particularly


harmed during this period by the substantial appreciation of the U.S.
dollar on foreign exchange markets, which hindered exports and gave
a significant cost advantage to Japanese firms. The Semiconductor

Industry Association (SIA) also attributed the investment difficulties


of its members to the high cost of capital in the United States in the
early and mid-1980s compared with the cost in Japan. The lack of
market access in Japan, its members complained, also prevented them
from taking advantage of one of the world's largest and fastest grow
ing markets.13
U.S. producers first considered antidumping action during the in

dustry recession of 1982, when the demand for memory chips slack
ened. The result provides a splendid illustration of the tension
between antidumping and competition policies. Motorola pondered
filing an antidumping petition on 64K DRAMs in early 1982 but
lacked industry support. Instead, it asked the Department of Com
merce to informally monitor Japanese prices. A Commerce official
responsible for administering the antidumping laws warned MITI
that Japanese chip prices in the United States might be monitored.

Japanese firms responded by reducing their exports to the United


States to forestall an antidumping petition, at which point the Depart
ment of Justice launched an antitrust investigation into reports of
Japanese collusion to raise prices in U.S. market!14

Antidumping Actions in the United States

The demand chips has grown rapidly over the past two
for memory
decades owing to the surge in computer production, but it has also
proven to be cyclical. Even a slight slowdown in the growth of demand

13. Compounding these problems were about the quality of U.S. semi
questions
conductors. In a widely publicized paper presented at an industry conference in 1981,
a representative from Hewlett-Packard cited evidence that the firm experienced much
fewer defects on 16K chips from Japanese producers than from U.S. producers. This
U.S. producers heatedly denied, but the perception
(and later acknowledged reality)
of a quality gap allowed long-term supply contracts to shift to Japanese firms.
14. Flamm (1996, p. 149) reports that "within the Japanese semiconductor industry
these reductions are openly acknowledged to have been spurred by MITI
[export]
guidance." Nothing came of the antitrust investigation.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 181

Figure 1. U.S. Domestic Output of Microcomputers, 1975-93

Number of units (millions)

9,

1975 1977 1979 1981 1983 1985 1987 1989 1991 1993

Source: Computer Business Equipment Manufacturers Association, Information Technology Datebook, 1960-2002
(Washington,D.C, 1992), p.94.

can lead to sharp fluctuations in prices, depending upon the state of


current production and inventories. Demand fluctuations are highly
correlated with
dumping complaints and appear to be the proximate
cause of the antidumping actions.
A boom in the semiconductor market in 1983-84 ended with an

industry in 1985. Compared


recession with other downturns, this in
dustry recession was extremely severe and was particularly concen
trated on the memory chip market. While overall semiconductor sales
slumped about 20 percent in 1985, the DRAM market contracted by
about 60 percent. The root cause was a brief slowdown in the com
puter market: after increasing by a factor of five between 1981 and
1984, US. shipments of microcomputers actually fell 8 percent in 1985,
as shown in figure 1. As the U.S. International Trade Commission
(USITC) reported, "By year end 1984 it was increasingly clear that

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
182 Brookings Trade Forum: 1998

Figure 2. Fisher-Ideal Price Indexes, 1972-89

Percent

30

? random access memory


Dynamic chips
20 I
? Erasable programmable read-only
10
memory chips

J_l_l_I_I_I

1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989

Source:Ramm (1993b), p. 273.

demand for personal


computers had fallen far short of forecasts and
expectations, resulting in heavy inventories in producers' ware
houses."15 With chip production and inventories remaining relatively
high, prices collapsed in the face of slumping demand. The price of a
64K DRAM fell from roughly $3.00 in the first quarter of 1984 to
$0.75 by the middle of 1985; the price of 256K DRAMs fell from
$31.00 to $3.00 over the same
time period. Figure 2 shows that DRAM
prices fell by nearly 70 percent in 1985, a decline comparable to that

registered in the 1981-82 industry recession.


As a result, U.S. merchant semiconductor firms suffered unprece
dented losses: pretax income as a percentage of sales fell from more
than 14 percent in 1984 to almost -10 percent in 1985. According to
SIA statistics, capacity utilization among the merchant firms dropped

15.USITC (1986, p. A-44).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 183

from more than 70 percent in 1984 to less than 45 percent in 1985.


Merchant semiconductor employment fell nearly 20 percent (about
55,000 workers) in the first three quarters of 1985. The 1985 industry
recession pushed virtually every U.S. producer out of the DRAM mar
ket: Mostek went bankrupt, and Advanced Micro Devices, Intel, Mo
torola, and National Semiconductor abandoned DRAM production to
concentrate on other semiconductor product lines. Only Texas Instru
ments and Micron remained in the merchant DRAM market, although
IBM and AT&T continued production for their own internal use.

Imports were not a direct cause of the recession: three-quarters of


the fall in revenues of U.S.-based semiconductor companies in 1985
was due to declining overall demand, only a quarter due to lost market
share.16 Indeed, Japanese firms experienced similar losses and layoffs

(prices fell sharply in Japan too) as world demand slumped.While the


majority of USITC commissioners determined that imports contrib
uted to the domestic industry's troubles and that therefore an injury

finding was warranted, two dissenting commissioners argued that the


industry was experiencing a normal cyclical downturn with no evi
dence of injury by means of imports. Import penetration of Japanese
DRAMs actually fell sharply during the two years after 1983: the ratio
of imports from Japan to apparent U.S. consumption dropped to 13.5

percent in 1985, from 23.6 percent in 1984 and 29.3 percent in 1983.

(However, the Japanese share of the U.S. market in 64K DRAMs


alone was roughly flat over this period at just under 30 percent.)
The SIA could never achieve an internal consensus on whether to
file an antidumping action, although this did not prevent any individ
ual member from filing on its own. In July 1985 Micron
Technology
(not a member of the SIA) filed an antidumping petition regarding
64K DRAMs. In October 1985 Intel, Advanced Micro Devices, and
National Semiconductor filed an antidumping petition covering im
ports of EPROMs from Japan. The petitions contended that the home
market sales of Japanese producers were below their costs of produc
tion so that the foreign market value had to be determinedby means
of constructed value. In December 1985 the Department of Com
merce also took the unusual step of self-initiating an antidumping

16. Federal Interagency StaffWorking Group (1987, p. 10). See also the figures for
DRAMs in particular inMersher and others (1986, pp. a-29ff).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
184 Brookings Trade Forum: 1998

case on 256K and future generations of DRAMs without waiting for

private industry to file a suit. This moved the government from the
position of intermediary to one of advocate.
Once the formal
antidumping administrative process was under
way, there were great incentives for the Japanese to settle the dispute
with the petitioners directly rather than to see the antidumping process
through in the hopes of vindication: Department of Commerce investi

gations have almost always found dumping, often at high margins, al

though there is somewhat more uncertainty about whether the USITC


will find final injury.Faced with the likely prospect that duties will be
imposed, mostforeign exporters would prefer to restrict their exports
"voluntarily" to capture some scarcity rents rather than face antidump
ing duties. The SIA also did not want antidumping duties to go into ef
fect just in the United States, which would then become a "high-priced
island" of semiconductors to the detriment of semiconductor-using in
dustries. The SIA wanted an end to dumping worldwide, not just dump

ing in the United States, because its members were adversely affected
by Japan's pricing worldwide. Even though this was technically outside
of the scope of the U.S. antidumping law, a bilateral agreement might be
able to affect such Japanese behavior in third markets.
All forces were therefore moving toward a negotiated settlement
rather than the imposition of antidumping duties. The pressure built
steadily for the two sides to reach a mutually agreeable solution.
Preliminary determinations from the Department of Commerce
found that Japanese firms were pricing at less-than-fair value with
high initial dumping margins, and the USITC reached an affirmative
preliminary determination of material injury to the domestic industry
by means of imports. These rulings made it clear that, barring an

agreement, considerable antidumping duties would remain in effect


with the final determinations.

Evaluating Theories of Japanese Dumping

To assess whether antidumping measures are warranted to pre


serve competition against unfair or predatory trade practices, it is
useful to examine three categories of dumping in light of the semicon
ductor experience: strategic, market-expansion, and cyclical.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 185

Strategie Dumping

In industries where
dynamic scale economies (learning by doing)
and static scale economies (fixed R&D and capital expenses) are

important, a protected home market can allow domestic firms to


increase their scale of production, reduce their costs of production by
spreading them over a larger volume of output, and hence improve
their profitability and competitive export position.17 Strategic dump
ing may deter investments by rivals and possibly create market power
for those undertaking it, but to reap advantages from such dumping
the exit of rivals and injury to consumers through higher prices need
not occur.

At first glance, the semiconductor case has several elements of

strategic dumping. Strategic dumping hinges largely on the interaction


of scale economies and home market protection. A simulation study
of internationalcompetition in 16K DRAMs has illustrated the po
tential for Japanese home market protection to increase the scale of
domestic firms to their advantage in relation to U.S. producers.18

Examining country market shares in DRAM, its authors found that


U.S. producers dominated most
in the world in the early
markets
1980s, except Japan. They that this assumed
exception was due not to

comparative advantage, but to an implicit tariff on imports from the


United States. They then simulated a free trade equilibrium and found
that the assumed home market protection accounted for all of Japa
nese output. That is, without implicit import barriers that allowed

Japanese producers to capture their large home market, Japan would


have produced no memory chips because the scale of output in the
United States reduced the cost of production there.
The
size of Japan's market in the mid-1980s was large enough for
such scale
effects to be potentially important for firms, even though
the market was (and is) divided among five to ten independent com

peting firms. Given the R&D intensity of production and the impor
tance of learning by doing, strategic dumping is plausible but hinges
on the question of home market protection. Formal import restric
tions (such as formal quotas, prior approval requirements, and foreign
investment barriers) were liberalized in 1975, after which few formal

17. See Krugman (1984).


18. Baldwin and Krugman (1988).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
186 Brookings Trade Forum: 1998

governmental trade barriers remained in place. In 1982, as a result of


the first U.S.-Japan Working Group talks on high-technology trade,
both countries set their import tariffs at 4.2 percent, and in February
1985 they eliminated most import tariffs on semiconductors.
The SIA argued that, because the U.S. share of the Japanese mar
ket scarcely budged after 1975, informal trade barriers lingered and
MITI undertook countermeasures to undermine the liberalization.
In June 1985 the SIA filed a petition with the Office of theU.S. Trade
Representative (USTR) under Section 301 of the Trade Act of 1974,
complaining about the lack of market access in Japan as an "unfair"
trade practice. The petition provided circumstantial and other nec

essarily anecdotal evidence of market barriers in Japan. In 1984 the


U.S. industry accounted for more than 83 percent of sales in the U.S.
market, 55 percent in the European market, 47 percent in the other
(Asian) markets, but only 11 percent in the Japanese market. The
U.S. share of the Japanese market, they noted, had remained fixed
near 10 percent for a decade despite the formal liberalization of the
Japanese market in 1975. The SIA also argued that structural barriers
in the Japanese market, such as "Buy Japan" attitudes and reciprocal
trading or tie-in relationships among firms, were an impediment to
U.S. entry.
These actions, according to the SIA, violated U.S. trade agreements
with Japan (Article XI of the General Agreement on Tariffs and Trade

[GATT] the transparency


regarding of trade barriers, and the 1983
bilateral agreement on greater U.S. participation in the Japanese
semiconductor market) and denied U.S. firms "fair and equitable
market opportunities." Therefore they were "unreasonable" under
the meaning of Section 301. The SIArequested relief in the form of
an "equivalence of market participation" in the Japanese market.
Under the 1986 U.S.-Japan semiconductor arrangement that settled
this dispute, Japan agreed to encourage greater foreign sales in Japan
(and, in a sideletter to the agreement, said it would seek a 20 percent
market share target for foreign producers).19
The petition did not fully resolve the question of whether Japan's
market was unfairly closed. The SIA pointed to past government
trade barriers, but not to current practices. The petition made claims

19. For an analysis of the market share target in semiconductors, see Irwin (1994).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 187

about the anti-import policies and attitudes of Japan's government


and firms and argued that the country's market structure was itself a
barrier to trade because of captive production: like IBM and AT&T
in the United States, major Japanese electronics producers also manu
factured their own semiconductors. But by the SIA's numbers, foreign
firms already had nearly 20 percent of the Japanese market if captive

production was excluded from the defined market, as it was in the


United States. The failure of the formal trade liberalization to alter
the U.S. market share was probably due to the difficulties encoun
tered in selling to vertically integrated producers and overcoming
long-term relationships between Japanese firms.20

Concerning country shares in regional markets, an alternative to


the SIA hypothesis was consistent with no Japanese "unfair" prac
tices: U.S. producers dominated the U.S. market, Japanese producers
dominated the Japanese market, and U.S. producers essentially split
European and other markets with other producers, holding a slightly

higher share in Europe owing to long-standing direct investments in

Europe behind the tariff barrier that kept out Japanese imports. The
SIA argued that "these trade [market share] figures, coupled with
Japan's protectionist heritage in microelectronics, strongly suggests
that market barriers still exist in Japan."21 But it could also be argued
that the market share figures were a reflection of Japanese production

efficiency, a hypothesis that could not be easily ruled out. Moreover,

20. The ten largest firms accounting for 80 percent of Japan's semiconductor pro
duction also accounted for 50 percent of Japan's total consumption. The high degree
of captive production in the mid-1980s ranged from 75 percent for Sanyo, 55 percent
for Matsushita, 50 percent for Fujitsu, and down to about 20 percent for NEC, Hitachi,
Mitsubishi, and Toshiba. As Okimoto (1989, pp. 103ff) points out, "The difficulties of
breaking into the organizational nexus are particularly frustrating for foreign produc
ers of high-tech intermediate goods, because the enclosed and nature of
long-term
relations between buyers and sellers alters the character of spot-market, arms-distant
transactions. . . . Such embedded in the structure
nontariff barriers, deeply of the
industrial economy, are not directly connected to Japanese industrial policy. But their
existence, whether by design or by accident, serves basically the same function as
formal measures of home market protection?only more effectively, because they do
not diminish the vigor of market competition between domestic producers."
21. Semiconductor Industry Association and Dewey Ballantine, "Japanese Market
Barriers in Micro-electronics," Memorandum in Support of a Petition Pursuant to
Section 301 of the Trade Act of 1974, as amended, June 14,1984, p. 2.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
188 B roo kings Trade Forum: 1998

Japanese access to the U.S. market in the early 1980s may have been
hindered by a discriminatory distribution system.22
Another piece of evidence
against trade barriers in Japan is the
lack of any distinct price divergence between the Japanese and for

eign markets in the early 1980s. Evidence collected by the USITC


suggests that before 1986 prices of 64K DRAM were lower in Japan
than in the United States, giving rise to arbitragers who bought on the
"grey (informal, or spot) market" in Japan.23 Japanese DRAM pro
ducers would sell some of their output to distributors and trading

companies in Japan, which would then sell them on this grey market.
The USITC noted that one firm described as a "key broker" in the
...
spot market "goes to Japan 'with dollars' and buys heavily at the
end of the month when Japanese DRAM producers unload unsold

inventory at reputedly below-cost prices." This arbitrage activity


would not have been profitable if spot prices in Japan had been higher
than those in the United States, and itwould have tended to keep U.S.
and Japanese prices very close.24
There is no direct evidence, therefore, that Japanese firms were
able to use home market advantages to price discriminate and charge
lower prices in foreign markets. By 1984 one could detect "a subtle
shift in the argument about predatory behavior in semiconductors" by
1983: with the end of overt MITI policies toward the industry and the
abolition of tariffs and quotas, "it could no longer be claimed that

higher prices in the home market


enabled Japanese producers to

persistently price below average cost in foreign markets."25


Even if Japan's market was closed, there is no indication of concen
trated market power or a lack of competition there. Nor is there

22. U.S. semiconductor firms contained Japanese access to the U.S. market by
terminating contracts with distributors who agreed to carry Japanese products. Japa
nese semiconductor firms hadonly one nationwide distributor in the United States

(Marshall Industries) because of the "unspoken ban on Japanese franchises" and the
"dictum that large houses will not take on the Japanese so long as they are supported

by domestic suppliers." See Electronic News, December 9,1985, p. S28.


23. USITC (1986, p.A-72, n. 1).
24. Indeed, Flamm (1993a, p. 167) notes that "prior to 1985, various available
[DRAM] price series are roughly consistent and tend to move relatively closely
together. Significant regional differentials were not important. All this changed after
1987."
25. Flamm (1996, p. 152).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 189

evidence of collusion among


Japanese producers, at least before the
imposition of antidumping sanctions by the United States, with an
intent to monopolize the DRAM market, even though their actual
market position worldwide increased through the early 1980s.

Market-Expansion Dumping

Market-expansion dumping is a form of price discrimination in


which a firm charges a lower net price on its exports than is charged
in the home market. The lack of any substantial price differential
across regional markets for DRAMs before the antidumping actions
of 1985-86 may reflect the fact that, for homogeneous commodity
chips for which transportation costs are negligible, the scope for price
discrimination across markets (absent government trade barriers)
was not
significant.
However, firms may wish to engage in pricing below static marginal
costs to expand output and reduce future costs. This is an inherent
feature of learning by doing: future production costs are dependent
upon cumulative production such that dynamic marginal cost (incor
porating the future cost reduction element of current production) is
lower than static current marginal cost. Therefore, forward-looking
firms setting price equal to dynamic marginal cost may also be setting
price below current marginal costs of production, which may consti
tute dumping.

Forward-pricing dumping ismore likely to occur early in the prod


uct life of a new generation: when a new product is introduced, yields
are initially very low and future production is expected to be large; for
both reasons, dynamic marginal costs are much lower than static costs,
and therefore the gains from increasing yields and decreasing future
costs are substantial. Later in the product life cycle, yields are high and
the expected horizon of future production in the product is short. In
this case, dynamic and static
costs may be equal. Therefore such
"market-expansion" dumping?in the sense of large production lead
ing to below-cost pricing to deter entry or reduce the output of
rivals?can be expected shortly after a new generation of DRAMs is
introduced.
Forward pricing is an unlikely explanation of the 1985 price col
lapse because the 64K DRAM was a mature product in which the

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
190 Brookings Trade Forum: 1998

learning effects on prices had diminished. Furthermore, USITC data


indicated that the 64K DRAM had been quite profitable for the
industry: gross profits by U.S. and Japanese firms amounted to $165
million over 1983-85 with an average margin of nearly 18 percent (the
particularly large margins in 1983 and 1984, reaching 33.2 percent in
1984, compensated for deep losses in 1985).26 The situation was some
what different for 256K DRAMS, which
had just recently been intro
duced and were the subject
of a Department of Commerce
antidumping complaint. In this case, the market-expansion motive for
dumping and forward pricing cannot be so easily dismissed.
It is not clear that this
of dumping is worrisome
type from an
antitrust standpoint. Firms price below cost precisely when their mar
ket power is high, namely, early in the product cycle when market
concentration is at its peak. They are attempting to reduce future costs
associated with the loss of market power when new firms commence
production of the new generation; table 1 confirms that entrants are

evidently not necessarily deterred entering from the market. This

mitigates thestrategic value (in terms of deterrence, but not the


economic value in terms of reducing future production costs) of low
pricing early in the product cycle.

Cyclical Dumping

The high correlation between the dumping complaints voiced by


U.S. producers and industry recessions (in 1981-82, 1984-85, and
1991-92) is strongly suggestive of cyclical dumping, defined as the
lowering of prices in response to a downturn in demand and substan
tial excess production capacity. The price index in figure 2 shows sharp
downward movements in DRAM prices in three years: 1978, 1981,
and 1985. Each year was associated with an industry recession rather
than a particularly different form of Japanese competitive behavior.
In the case of 1985, the underlying cause is perfectly evident: micro
computer output fell after years of double-digit growth. This adverse
demand shock?the computer industry is the major source of demand
for memory chips?in the face of abundant DRAM production capac

26. USITC (1986, pp. 48,53).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 191

ity was clearly the proximate cause of the complaints about dump
ing.27

Although the circumstances do not completely rule out the strate


gic use of this opportunity for predatory actions, there is little evi
dence of collusion among Japanese firms during this period. Prices for
64K DRAMs also did not appear to fall below marginal costs, as might
be consistent with a predatory or (possibly) a market expansion
dumping situation. Prices were
below average costs, however, as most

producers in Japan and the United States suffered losses in 1985. As


a result of their bank ties, major Japanese firms, which were also part
of large conglomerates, were able to continue memory chip produc
tion through the recession, unlike their independent, merchant U.S.
counterparts.

The Consequences of U.S. Antidumping Actions

Under the terms of the suspension agreements signed in Septem


ber 1986, alongside the semiconductor trade arrangement, Japan
agreed to take actions that would end dumping in the United States.
Japanese exporters also agreed to report data on all U.S. sales and on
actualand anticipated costs of production to the Department of Com
merce. On the basis of this information, Commerce would determine
company-specific price floors (foreign market values, or FMVs) each
quarter and provide this information to the Japanese firms them
selves.28 In monitoring the prices and costs of DRAM and EPROM
shipments to the U.S. market, the agreement stipulated, "the Govern

27. Though the majority of USITC commissioners found injury as a result of


imports, their statement (1986, pp. 14-15) lent support for a cyclical dumping position:
"U.S. production of cased DRAMs more than doubled from 1983 to 1984, from 42.2
million units to 106.3 million units. The data concerning indicate that total
capacity
capacity to produce DRAMs has increased the period of investigation.
throughout
These increases in capacity and production reflect the industry's optimism regarding
increased demand and growth during 1983 and 1984. However, total apparent U.S.
consumption did not continue to increase in 1985 as had been Total
anticipated.
apparent U.S. consumption of all cased DRAMs increased 28 percent from 1983 to
1984, from 329.8 million units to 421.9 million units. However, in 1985, total apparent
US. consumption of all cased DRAMs fell by 14 percent, to 361.5 million units."
28. The purpose of company-specific FMVs was to avoid a fixed export price floor
and to allow low-cost producers the freedom to cut prices.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
192 Brookings Trade Forum: 1998

ment ... to
of Japan will take appropriate actions prevent exports at
prices less than company-specific fair value."

Despite the U.S. understanding that Japan was obligated to prevent

dumping in third-country markets as well, the agreement makes no

explicit statement of the government's responsibility for taking action


to prevent such dumping. In a separate section of the agreement on
third markets, the text merely reads: "Both Governments recognize
the need to prevent dumping.... In order to prevent dumping, the
Government of Japan will monitor, as appropriate, costs and export

prices on the products exported by Japanese semiconductor firms


from Japan." There is no mention of the government of Japan taking

"appropriate actions," and MITI later denied responsibility for pre

venting third-country dumping, although the phrase "in order to pre


vent dumping" might be read as implying some measure of

responsibility.29
To prevent dumping, MITI acted to reduce the quantity of semi
conductors exported in the hope of raising export prices sufficiently.30
An "antidumping voluntary export restraint (VER)," or an export
restraint designed to meet a price target rather than a quantitative

target, is inherently more difficult to administer than a quantitative


export restraint. MITI had no statutory authority to force any firm to
comply in reducing output, but bureaucratic delays in approval of
export licenses?also tightened to prevent dumping?could be em
ployed against firms. Initially MITI
recalcitrant had difficulty in get
ting Japanese tofirms
comply with its requests, and it tried to boost
DRAM prices by issuing recommendations (administrative guidance)
in February 1987 to reduce output by 10 percent. One merchant
producer (Micron) argued that a 10 percent reduction was insuffi
cient, but the SIA accused MITI of trying to create "artificial short

ages" and increasing government interference in the market.31


This criticism?that the SIA did not want Japan to sell less, just
wanted it to charge a higher price?ignored the fact that the produc

29. See Irwin (1996).


30. Flamm (1996, pp. 175ff).
31. Andrew Procassini, president of the SIA, stated that "rather than requiring

Japanese semiconductor producers to price devices at or above cost, as the agreement

requires, MITI is trying to drive up prices artificially by creating shortages through


production controls." Electronic News, September 14,1987, p. 1.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 193

tion cuts were a natural outcome of the agreement. As a Department


of Commerce official observed, requiring the Japanese government to
drive up prices in the United States and in third markets

amounted to getting the Japanese government to force its companies


to make a profit and even to impose controls to avoid excess produc
tion?in short, a government-led cartel. For the free-traders of the
United States to be asking Japan to cartelize its industry was the
supreme irony. Yet itwas logical_It was subsequently criticized for
doing so, but it had little choice."32

A stunning U.S. tariff retaliation against Japan in April 1987 for non
compliance with the agreement persuaded Japanese firms to follow
MITFs directives more
closely.
The European
Community, however, strenuously objected to the
third-country provisions of the arrangement and filed a complaint
before a GATT panel. The European Community argued that the
third-market provisions entailed "arbitrary [price] increases" for
European semiconductor consumers.
The GATT panel ruled in 1988
that Japan's monitoring of export prices on third-market sales vio
lated Article XI of the GATT (on governmental quantitative restric
tions), although the market access provisions of the agreement were
not found to violate most-favored-nation treatment. As a result, Japan
announced that it would desist from monitoring sales in third coun
tries.33

Japanese semiconductor
firms benefited from the implicit VER.
The production cutbacks provided a substitute for cooperative indus
try behavior and raised the price of DRAMs on sales abroad, gener
ating large profits for Japanese producers.34 DRAM prices, which
usually fell consistently, shot up dramatically in 1988 when demand
for memory chips sharply rebounded with another rapid expansion in
the computer industry. The price of 256K DRAMs jumped from about
$2.20 at the end of 1986 to $3.50 by the end of 1988, although long
term supply prices in Japan were largely unchanged because Japanese
producers were unconstrained in the prices in their
they charged

32. Prestowitz (1989, pp. 166-67).


33. For an analysis of the European action
Community's antidumping against
Japanese semiconductor producers, see Tharaken (1997).
34. See Flamm (1989, p. 21).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
194 Brookings Trade Forum: 1998

home market. A striking differential in semiconductor prices arose


across international markets, and colorful stories surfaced of arbitrag
ers purchasing chips in Asia, flying them to Canada or Mexico, and
then smuggling them across on IM
the U.S. border. Extra
profits
DRAM sales for Japanese producers amounted to $1.2 billion in 1988
alone, and about $3 billion to $4 billion in 1989 on all devices.35
Japanese producers may have also begun cooperating in a de facto
market-sharing cartel.36 Even after 1989 DRAM prices in the United
States remained substantially above estimates of the FM Vs. As a result
of production controls, which influenced capacity investment choices
by Japanese firms, exporters had to ration output to U.S. consumers,
and this fostered coordination and cooperation, if not outright collu
sion, among them. Whether the price bubble merely reflected a cut
back in excessive production conscious or
collusion by Japanese
producers, or a combination of both, remains difficult to ascertain,
however. Between 1986 and 1987 the share of world output of 256K
DRAMS accounted
for by the top five Japanese producers fell from 77
percent to 55 percent (table 1). Yet the Herfindahl-Hirschman index
indicates that industry concentration remained low and, in fact, contin
ued to fall because of the additional output from smaller Japanese
firms, U.S. firms, and entering South Korean firms. South Korean pro
ducers were not restricted by the arrangement, and the higher DRAM
prices accelerated the entry of Samsung and others.37

35. Flamm (1996, p. 277).


36. See Flamm (1996).
37. This marked the end of U.S. antidumping actions Japan in DRAMs.
against Yet
1992-93 saw a virtual repeat of the semiconductor trade dispute, albeit in a much less
confrontational setting, with South Korea as the defendant. In April 1992, Micron filed
an antidumping petition alleging less-than-fair-value imports of IM DRAMs and

higher from Korea. In October, Commerce announced preliminary dumping margins


(based on petitioner information) against Samsung (87.40 percent), Goldstar (52.41
percent), and Hyundai (5.99 percent). Faced with stiff antidumping duties, the Korean

industry and government proposed in January 1993 a bilateral semiconductor trade

agreement fashioned on the earlier one with Japan. In exchange for a suspension of
the antidumping case, the Korean industry promised to monitor prices of export sales
to the United States. This overture was rejected because Micron strongly opposed
suspension of the case. By March the Korean firms had provided production cost data
to the Department of Commerce, and the final antidumping were
margins drastically
cut: 0.74 percent for Samsung, 4.97 percent for Goldstar, and 7.19 percent from

Hyundai. InMay the ITC split 3-3 on the final material injury,with the result that
duties were imposed. See Irwin (1996).

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 195

Such Japanese collusion, if it did exist, probably could not have


arisen without the governmental trade actions. It seems that "before
the Semiconductor Trade Arrangement of 1986, episodes of success
fully coordinated restraint on exports or output by Japanese produc
ers seem to have occurred only after bureaucrats and politicians
responded to trade friction."38 It has also been suggested that "frame
works imposed or invented in collaboration with politicians and bu
reaucrats in order to deal with trade friction, rather than sudden
changes in market power, seem the more compelling explanation for
historical episodes of successful restrictions on supply. In all cases

prior to 1989, successful producer restraint coincided with active gov


ernment initiatives."39 Because nearly all U.S. producers had exited
the market, the remaining Japanese producers may have possessed
substantial market power once demand for DRAMs again began to
accelerate in 1988.
U.S. semiconductor-using firms criticized Japanese producers for
denying them adequate volumes of semiconductors and hinted that it
was a deliberate attempt at retribution. Japanese firms more likely
allocated their production to firms with which they had long-standing
ties and avoided selling "too much" in the United States.40 No such
bubble appeared in the price of EPROMs, which were also subject to
the FMVs. Intel, AMD, TI, and other U.S. producers did not leave the
EPROM market, and the much lower Japanese market share meant
that production cutbacks on their part would not produce a substan
tial price rise.
The U.S. semiconductor industry was
only a partial beneficiary of
the antidumping measures. In the DRAM
market, only two merchant
firms?Texas Instruments and Micron?continued production.
DRAM sales accounted for as much as 60 percent of Texas Instru

38. Flamm (1993c, p. 280).


39.Hamm (1996, p. 391).
40.As Okimoto (1987, pp. 389-90) points out, just as purchasers of semiconductors
rely on other ties besides price, "From the supplier's standpoint, too, organizational
factors can overwhelm narrowly defined market forces. During of economic
periods
upturn, for example, when supply fails to keep pace with demand, Japanese companies
are not inclined to sell on a neutral, first-come, first-served basis, Rather, they are apt
to allocate limited stocks of semiconductor components to an implicit
according
hierarchy of customers. Loyal, long-standing customers get priority over companies
that buy sporadically on a spot market basis."

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
196 Brookings Trade Forum: 1998

ments's profits in 1988, and Micron's sales rose by a factor of six


between 1986 and 1988.41 The employment effects of the agreement
were probably small: Texas Instruments and Micron had a total
DRAM-related employment of approximately 13,000, and one back
of-the-envelope calculation suggests that for each job retained in
DRAMs, another was lost in downstream computer manufacturing.42
For all the concern about
the strategic necessity of domestic
DRAM production, however, no U.S. producers reentered the market
in the aftermath of the agreement. In June 1989 several firms?AMD,
Hewlett-Packard, Digital Equipment, Intel, LSI Logic, and National
Semiconductor?tried to establish a domestic consortium to produce
DRAMs. Despite even IBM's support, U.S. Memories was stillborn
and collapsed in January 1990. This failure could be interpreted as an
indication of the high up-front costs of production that could consti
tute a barrier to entry. But other factors played a role as well: for
example, other major buyers, such as Apple Computer and Sun Mi
crosystems, were unwilling to commit to specified future purchases
from U.S. Memories.
The clear losers from the higher DRAM prices were semiconduc
tor users, particularly computer manufacturers. Concerted opposition
to the 1986 arrangement first materialized when the price bubble for
256K DRAMs in 1988 severely harmed these users. Three CEOs of
major computer systems firms?IBM, Tandem, and Hewlett
Packard?formed the Computer Systems Policy Project (CSPP) in
early 1989 to end the SIA's monopoly position as the U.S. govern
ment's adviser on semiconductor trade policy. InMay 1989 these firms
invited others?including AT&T, Apple Computer, Compaq Com
puter, Control Data, Cray Research, Digital Equipment, NCR Corpo
ration, Prime Computer, Sun Microsystems, Tektronix, and Unisys

Corporation?to join the CSPP. The CSPP aimed to develop policy


recommendations relatingto the competitive position of the com
puter manufacturers, noting that without its collective action the gov
ernment "may well adopt policies counterproductive, and in some
cases inimical, to the interests of our companies."43

41. Tyson (1992, pp. 116-17).


42. Denzau (1988).
43. Inside US. Trade, June 16,1989, p. 3.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 197

As a result of this coalition of users, trade negotiators at the USTR


no longer faced a single private sector voice on policy. With the
expiration of the 1986 accord nearing, the USTR could not possibly
negotiate a satisfactory agreement in the face of these sharply con

flicting business interests. Rather than mediate, the USTR instructed


the SIA and the CSPP to resolve their policy differences themselves.
After lengthy negotiations, the SIA and the CSPP announced in
October 1990 a joint proposal declaring the antidumping measures a
"success" and suggesting that the Department of Commerce no

longer collect costs or price data or issue foreign market values for
DRAMs and EPROMs. The 1991 renewal agreement accepted this
recommendation.

Conclusions

The semiconductor dumping case of 1985 appears to be a text


book instance of cyclical dumping: world DRAM prices plummeted
sharply in the face of a negative demand shock. The evidence for
market expansion dumping is weak, but that for strategic dumping
has been more difficult to assess and hinges on whether one views
the Japanese semiconductor market of that time as closed and un

competitive.
The semiconductor case clearly illustrates the tensions between
antidumping policy and competition policy. Every formal or informal
dumping complaint by U.S. producers led Japanese producers to limit
competition by reducing exports and production. Formal antidumping
measures failed to preserve competition since they failed to keep
most U.S. producers in the memory chip market. The antidumping
measures did encourage the entry of South Korean producers, who
eventually helped to keep product markets competitive. But the for
mal antidumping measures actually harmed competition because they
facilitated more cooperative production strategies by Japanese pro
ducers. (The 20 percent foreign market share target for Japan in the
1986 semiconductor trade agreement?itself a by-product of the anti
dumping case?may have increased competition there, but also may
have intensified the need for those cooperative production strate
gies.)

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
198 Brookings Trade Forum: 1998

Despite the sporadic complaints about dumping in the 1980s, an


irony of the semiconductor drama was that no actor (with the excep
tion of Micron Technology, then a small memory chip producer in
Idaho) wanted to see the remedy of antidumping duties employed:
Japanese producers preferred a quantitative export restriction, as did
U.S. producers, who feared that U.S.-specific antidumping duties
would make the United States a "high-price island" for memory
chips.
Another irony of the case is that, at the time, domestic memory chip
production was thought to be a "strategic driver" for the semiconduc
tor industry, that loss of domestic production would prove cata
strophic to the competitive position of other product lines and other
high-technology producers. Instead, the U.S. industry flourished in the
1990s by focusing on microprocessors and other semiconductors with
greater proprietary technology. Memory chips have remained just a
commodity chip.

References

Baldwin, Richard E., and Paul Krugman. 1988. "Market Access and Interna
tional Competition: A Simulation Study of 16K Random Access Memo
ries." In Empirical Methods for International Trade, edited by Robert C.
Feenstra, 171-202. Cambridge, Mass.: MIT Press.
Denzau, Arthur T. 1988. "Trade Protection Comes to the Valley?Silicon
Valley." Center for the Study of American Business, Washington Univer
sity.

Dick, Andrew R. 1991. "Learning by Doing and Dumping in the Semicon


ductor Industry." Journal of Law and Economics 34 (April): 133-59.
-. 1992. "An Efficiency Explanation for Why Firms Second Source."
Economic Inquiry 30 (April): 332-54.
Federal Interagency Staff Working Group. 1987. The Semi-conductor Indus
try.Washington, D.C. (November 16).
Flamm, Kenneth E. 1989. "Policy and Politics in the International Semicon
ductor Industry." Paper presented at SEMI ISS seminar, January.
-. 1993a. "Measurement of DRAM Prices: Technology and Market
Structure." In Price Measurements and Their Uses, edited by Murray F.
Foss, Marilyn E. Manser, and Allan H. Young, 157-97. University of Chi
cago Press.

-. 1993b. "Forward Pricing versus Fair Value: An Analytical Assessment

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
Douglas A. Irwin 199
of 'Dumping' in the DRAMs." In Trade and Protectionism, edited by
Takatoshi Ito and Anne O. Krueger, 47-94. University of Chicago Press.
-. 1993c. "Semiconductor Dependency and Strategic Trade Policy."
Brookings Papers on Microeconomics 1:249-333.
-. 1996. Mismanaged Trade? Strategic Policy and the Semiconductor
Industry. Brookings.
Hoshi, Takeo, Anil Kashyap, and David Scharfstein. 1990. "The Role of
Banks in Reducing the Costs of Financial Distress in Japan." Journal of
Financial Economics 27 (September): 67-88.
Irwin, Douglas A. 1994. Managed Trade: The Case against Import Targets.
Washington, D.C.: AEI Press.
-. 1996. "Trade Politics and the Semiconductor Industry." In The Politi
cal Economy of American Trade Policy, edited by Anne O. Krueger. Uni
versity of Chicago Press.
Irwin, Douglas A., and Peter J.Klenow. 1994. "Learning-by-Doing Spillovers
in the Semiconductor Industry." Journal of Political Economy 102 (Decem
ber): 1200-27.
Krugman, Paul. 1984. "Import Protection as Export Promotion: International
Competition in the Presence of Oligopoly and Economies of Scale." In
Monopolistic Competition in International Trade, edited by Henryk
Kierzkowski. Oxford University Press.
Mersher, llene, and others. 1986. 64K Dynamic Random Access Memory
Components from Japan: Determination of the Commission in Investigation
No. 731-TA-270 (Final) under the Tariff Act of 1930, together with the
Information Obtained in the Investigation. Publication 1862. Washington,
D.C.: U.S. International Trade Commission.

Organization for Economic Cooperation and Development (OECD). 1992.


Globalisation of Industrial Activities, Four Case Studies: Auto Parts, Chemi
cals, Construction, and Semiconductors. Paris: OECD.

Okimoto, Daniel 1.1987. "Outsider Trading: Coping with Japanese Industrial


Organization." Journal of Japanese Studies 13 (Summer): 383-414.
-. 1989. Between MITI and theMarket: Japanese Industrial Policy for
High Technology Industries. Stanford University Press.
Prestowitz, Clyde V., Jr. 1989. Trading Places: How We Are Giving Our Future
to Japan and How to Reclaim It. New York: Basic Books.

Tharaken, P. K. M. 1997. "The Japan-EC DRAMs Undertaking?Was It


Justified? What Purpose Did It Serve?" De Economist 145 (April): 1-28.

Tyson, Laura D'Andrea. 1992. Who's Bashing Whom? Trade Conflict inHigh
Technology Industries. Washington, D.C.: Institute for International Eco
nomics.

U.S. International Trade Commission (USITC). 1986. "64K Dynamic Ran

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions
200 Brookings Trade Forum: 1998

dorn Access Memory Components from Japan (731-TA-300, Final)." Pub


lication 1862. Washington, D.C.: USITC.
Yoon, Chang-ho. 1992. "International Competition and Market Penetration:
A Model of the Growth Strategy of the Korean Semiconductor Industry."
In Trade Policy, Industrialization, and Development: New Perspectives, ed
ited by Gerald K. Helleiner, 254-78. Oxford: Clarendon Press.

This content downloaded from 91.223.28.76 on Sat, 28 Jun 2014 08:51:15 AM


All use subject to JSTOR Terms and Conditions

You might also like