ROBERT
Z. LAWRENCE
BrookingsInstitution
Importsin Japan: Closed
Marketsor Minds?
COMPAREDWITH other industrialcountries, Japanimportsan unusually
small share of its domestic use of manufacturedgoods. In 1980, for
example,importsaccountedforjust5.8 percentof Japaneseexpenditures
on manufacturedproducts, comparedwith 9.3 percent of U.S. expenditures.That same year, non-ECimportsaccountedfor 13.9percent of
spendingby the EuropeanCommunity.I
Why Japanimportsso little is a source of greatcontroversy. Popular
explanationsstress the role of bothofficialandunofficialimportbarriers.
The Japanesegovernmentallegedlytakes advantageof the openness of
foreignmarketswhile reservinglocal marketsfor domesticfirms.It once
implementedthis mercantilistpolicy throughformalprotectionistmeasures such as tariffsand quotas. Today it uses administrativeguidance,
discriminatorystandardsand regulations, selective government procurement,the official organizationof domestic firms into cartels, and
weak enforcementof antitrustlaws. Japaneseimportsare also discouragedby unofficialpractices, such as the strongrelationships("invisible
handshakes")betweenlocal suppliersandbuyers, "just-in-time"inventory practices that give nearby suppliers an edge, and an unusually
complex distributionsystem that creates substantialentry barriersfor
newcomers,whetherJapaneseor foreign.2
I amgratefulto MorioKuninori,MarcusNoland,andGarySaxonhouseforcomments;
to GregoryHume, Dale Thompson,and Amy Salsburyfor researchassistance; and to
EvelynTaylorfor text processing.
1. DerekBladesandWendySimpson,"TheOECDCompatibleTradeandProduction
DataBase," WorkingPaper18(Organization
forEconomicCooperationandDevelopment,
January 1985).
2. See DorothyChristelow,"Japan'sIntangibleBarriersto Tradein Manufactures,"
Federal Reserve Bank of New York Quarterly Review, vol. 10 (Winter 1985-86), pp. 11-
517
518
Brookings Papers on Economic Activity, 2:1987
Other observers dismiss allegations of import barriersas based on
biased evidence and stress instead Japanese manufacturingprowess.
The Japanese make high-qualityproducts that both foreigners and
Japanese prefer. While the Japanese invest in long-term strategies to
win foreign markets,foreignersare unwillingto make similarefforts in
Japan.
Many economists also dispute the anecdotal evidence on Japanese
protectionism.They explainJapan's tradestructureby thefundamentals
of its economic situation.Because its citizens have highsavingrates and
diminished domestic investment opportunities, Japan runs a current
account surplus. Because it is poorly endowed in arableland, oil, and
other naturalresources, it achieves this surplusthrougha strong comparativeadvantagein manufacturing.3
In this view, popularperceptions
of Japanesetradepracticesreflectmindsclosed by prejudice.Japanese
industrialpolicies, for example,have actuallybeen relativelyineffective
or have simply overcome some unique Japanese practices, thereby
replicatingthe impactof marketforces.4
In a seminal study, Gary Saxonhouse provided support for the
explanationbased on economic fundamentals.5Saxonhouse explained
the tradeflows of twenty-twocountries,includingJapan,using a model
specified in terms of the Heckscher-Ohlinparadigmand based on 109
18;and WilliamV. Rapp, "Japan'sInvisibleBarriersto Trade,"in ThomasA. Pugeland
Robert G. Hawkins, eds., Fragile Interdependence: Economic Issues in U.S.-Japanese
Trade and Investment (Lexington Books, 1986), pp. 21-45.
3. If Japanesegoods were of poorerquality,a weakeryen ratherthana differenttrade
structurewouldresult.
4. See GaryR. Saxonhouse,"What'sAll This aboutIndustrialTargetingin Japan?"
The WorldEconomy, vol. 6 (September1983), pp. 253-74; PhilipTrezise, "Industrial
Policy Is Not the MajorReason for Japan'sSuccess," BrookingsReview, vol. 1 (Spring
1983), pp. 13-18; and CharlesL. Schultze, "IndustrialPolicy: A Dissent," Brookings
Review, vol. 2 (Fall 1983),pp. 3-12. For a view stressingthe role of industrialpolicy, see
Chalmers Johnson, MITI and the Japanese Miracle: The Growth of Industrial Policy,
1925-1975(StanfordUniversityPress, 1982).
5. See GaryR. Saxonhouse, "The Micro-and Macro-economicsof ForeignSales to
Japan,"in WilliamR. Cline,ed., TradePolicies in the 1980's(Washington,D.C.: Institute
for InternationalEconomics, 1983), pp. 259-304; Saxonhouse, "What's Wrong with
JapaneseTradeStructure?"PacificEconomicPapers,no. 137(July1986);andSaxonhouse
and Robert M. Stern, "An Analytical Survey of Formal and Informal Barriers to
InternationalTrade and Investmentin the United States, Canadaand Japan," paper
preparedfor conferenceon U.S.-CanadianTradeand InvestmentRelationswith Japan
(Universityof Michigan,April2-3, 1987).
Robert Z. Lawrence
519
Table 1. Merchandise Trade and Current Account Balance, Japan and Germany, 1986
Percentof GDP
Category
Japan
Germany
Manufacturedgoods
Exports
Imports
Balance
10.4
2.2
8.2
24.2
14.2
10.0
Othermerchandise
Exports
Imports
Balance
0.3
4.2
- 3.9
2.9
7.1
-4.2
4.3
4.4
5.8
3.9
Total merchandisebalance
Currentaccount balance
Soujrces: Organization for Economic Cooperation and Development, Econtomic Suirvey: 198611987:Germany (Paris:
OECD, July 1987), Statistical Annex, tables B and H; and Bank of Japan, Economic Statistics Monthly (Tokyo:
Bank of Japan, July 1987), pp. 135, 178.
industries.He foundthatJapan'snet exports differedsignificantlyfrom
those predictedby its factor endowmentsand its distance from trading
partnersin industriesthat accountedfor only 6.1 percent of its external
trade.6Saxonhouse'sinfluentialwork has been interpretedas indicating
that Japan'strade structureis conventional and that, in fact, its trade
policy does not matter.7
AlthoughSaxonhouse begins his study citing the unusual natureof
Japanese manufactured goods imports, he actually examines net exports.
And althoughrelativefactor endowmentsmay well explainJapan'snet
trade balance, Japan's trade structure may nonetheless be unusual
becauseboth exports and importsare too small. Manyconfigurationsof
manufacturedexports and imports can generate the current account
surpluscompatiblewith Japan'ssavingrate and resourceendowments.
Saxonhousefails to addressthe crucialquestionof why Japanachieves
its manufacturedgoods surpluswith so few imports.
This issue emerges clearly in a comparison of Japan and West
Germany, as reported in table 1. In some respects the German and
Japanese economies have similar structures. Both nations run large
6. EdwardLeameralso does not findJapanunusualin termsof the Heckscher-Ohlin
model. See Edward E. Leamer, Sources ofInternational Comparative Advantage: Theory
and Evidence (MIT Press, 1984).
7. See L. Alan Winters, "Patternsof World Trade: Does Trade Policy Matter?"
DiscussionPaper160(London:Centrefor EconomicPolicy Research,March1987).
520
Brookings Papers on Economic Activity, 2:1987
Table 2. Intra-Industry Manufacturing Trade Indexes, 1980
Twenty-one
industries
Ninety-four
industries
Australia
Belgium
Canada
Finland
France
0.41
0.87
0.67
0.58
0.88
0.22
0.79
0.68
0.49
0.82
Germany
Italy
Japan
Netherlands
Norway
0.69
0.71
0.30
0.77
0.62
0.66
0.61
0.25
0.78
0.51
Sweden
United Kingdom
United States
0.66
0.82
0.67
0.68
0.78
0.60
...
...
0.48
0.61
Country
Korea
Switzerland
Source: Author'scalculationsusingthe followingformula:
E
[(Xu
+ Mij)
-
I Xij-
Mij I]
Indexj i=
2 (Xi + Mij)
wherei denotesmanufacturing
category,j denotescountry,andx andm are exportsandimports,respectively.This
follows Herbert G. Grubel and P. J. Lloyd, Intra-Industry Trade: The Theory and Measurement of International
Tradein DifferentiatedProducts(New York:HalsteadPress, 1975),p. 22. Data for the twenty-oneindustriesare
from the OECDCompatibleTradeand ProductionData Base. Data for the ninety-fourindustriesare from United
Nations, CommodityTradeStatistics,1980,UN StatisticalPapers,Series D (New York:UnitedNations, 1980).
manufacturedgoods surplusesto offset deficitsin primarycommodities
and services. In 1986, their manufacturingtrade and current account
balances were similarpercentages of GDP. But the Germanexample
demonstratesthatlow levels of manufacturedimportsarenot necessarily
"required" in resource-poorcountries. Despite manufacturedgoods
importsof 14.2percentof GDP, Germanyhas been ableto runsurpluses
in its manufacturedgoods trade that were larger shares of GDP than
those of Japan.
The Heckscher-Ohlintheory assumes standardizedproducts and
predicts that countries will not import and export the same products,
that is, that there will be no intra-industrytrade. In termsof the theory,
intra-industrytrade is a statistical artifact resulting from insufficient
disaggregation.Relying on this view, Saxonhouse has arguedthat his
use of net exports as a dependent variable is permissiblebecause his
data are disaggregated.However, as reported in table 2, the Grubel-
Robert Z. Lawrence
521
Lloyd index of intra-industrytradecalculatedusing dataon ninety-four
manufacturingindustriesindicates that Japanis noteworthyfor its lack
of intra-industrytrade even at the disaggregationlevel used by Saxonhouse. Table 3 furtherhighlightsthe unusualnatureof Japaneseratios
of exportsto imports.For eleven of the twenty-twocategories(accounting for 48.4 percentof OECD manufacturedgoods trade), the Japanese
ratios are far higher than those of any other country. The average
industryratioof Japaneseexports to importsin 1980was 7.6, compared
with thatof the next highestcountry,Finland,which, because of its high
ratioin wood products,had an averageratioof 2.6.
Otherauthorshave examinedimportsdirectly. C. Fred Bergstenand
William Cline regress the ratio of aggregate imports of goods and
nonfactorincome services to GNP againstincome, population,resource
endowments, and distance from tradingpartners.8They find that the
Japanese import ratio does not differ significantlyfrom that of other
countries in the sample, and they conclude that Japan does not have
excessive trade protection.9Bela Balassa, however, using a different
specificationof transportationcosts and specifyingthe Europeancoulitries separately,concludes that Japanis an "outlier" in both its manufacturingand total imports.10The Cline-Bergstenand Balassa studies
have been criticizedby Saxonhouse and MarcusNoland for not being
based on a clear theoreticalfoundation.1IApparently,in addition, the
methodologyyields resultsthatare sensitiveto minorchangesin variable
specification.
Nolandhas triedto remedythese shortcomingsby usinga theoretical
modelpresentedby ElhananHelpmanand Paul Krugmanthat explains
the volume of tradeand allows for intra-industrytrade.12 He concludes
8. C. FredBergstenandWilliamR. Cline, The United States-JapanEconomicProblem
(Washington,D.C.: Institutefor InternationalEconomics, 1985).
9. A similarconclusionis reachedby Luca Barbone, "Is Japanan Underimporter?
SomeContrastingResults" (OECD, 1987).
10. Bela Balassa, "Japan's Trade Policies," Weltwirtschaftliches Archiv., vol. 122,
no. 4, pp. 745-90.
11. As Nolandnotes, we have "two studieswithoutformalmodelsreachingopposite
conclusions,andone studywitha formalmodelof uncertainrelevance." MarcusNoland,
"AnEconometricInvestigationof InternationalProtection"(Washington,D.C.: Institute
for InternationalEconomics,June 1987),p. 6.
12. Elhanan Helpman and Paul R. Krugman, Market Structure and Foreign Trade:
Increasing Returns, Imperfect Competition, and the International Economy (MIT Press,
1985),chaps.7-8.
z
O ',
oo
o k.t
S: w w b t t
i
o
.WX^ote
to
o
moos
ZS,
*W st N
LE :3b
=
4;
t
,,,e
oil
o
^
0
s
t
0Ov o
o
Q
E
'IC C)
o>o "C
St-t
c6 ,Q o
C; C
0o gStO
.os
V-)
V)
mo
Uo
e
E
Z
Robert Z. Lawrence
523
that Japaneseaggregateexports, imports,and total tradeare not out of
the ordinary.AlthoughNoland's work uses a more appropriatetheoretical framework,he explains total merchandisetrade volumes and provides no test of tradein manufacturedproductsalone.13Because manufacturedgoods account for less than a third of Japaneseimports, even
substantialunderimportingof manufacturedgoods is not likely to be
detected in the aggregatespecification.
In sum, the behavior of Japanese manufacturedgoods imports has
not been adequatelyexplained. Althoughthe accounts stressingimport
barriersand those stressingfactor endowmentsare not mutuallyexclusive, a sense of the relative importanceof each is essential for policymaking. If trade barriersare unimportantand Japan's low imports of
manufacturedgoods are the inevitable result of its macroeconomic
behaviorandfactorendowments,policymakerswho committhemselves
to raisingimportsthroughremovalof tradebarrierscould be frustrated
and disappointed.On the other hand, if importbarriersare important,
Japan's adjustmentto recent changes in macroeconomic policy and
exchangerates could occur predominantlyin Japaneseexports, and the
Japanesemarketcould remainrelatively closed. If significantbarriers
are found, theirnatureneeds to be determined.Are the barriersofficial
or private?Are they like quotas, so that imports are unresponsive to
price changes, or are they like tariffs, so that despite their presence,
importvolumes will respondto currencychanges?
In this reportI explore some of these questions. Using a modelbased
on the theory of tradeunderimperfectcompetition,I demonstratethat
Japanese manufacturedimports are about 40 percent lower than one
wouldexpect of a typicalindustrialeconomy. If Japanesemanufactured
importsreflectedmoretypicaltrade,Japan'smanufacturedgoods trade
surpluswould be only about 16 percent smaller,but Japanwould have
considerablymoreintra-industrytrade.
The modelleaves certainquestionsunanswered.It cannotdistinguish
between the impactof Japanesebarrierson foreign importsand that of
foreign barrierson Japanese exports, nor can it differentiatebetween
the effects of importbarriersand the peculiarpreferencesof Japanese
13. Dataconstraints(Nolanddoes not measurefactorendowmentssuchas humanand
physical capitalexplicitly)and the need to assume balancedtrade-clearly violated in
practice-may compromiseNoland'sconclusions.
524
Brookings Papers on Economic Activity, 2:1987
buyers. Because there were few export restraintson Japanesetrade in
early observation years of the data sample, results from these years
provide measures of the impact of importbarriersand unusualpreferences.
In the second part of the paper I examine the price sensitivity of
Japanesemanufacturedimportsandfindthat,for manyproducts,import
price elasticities are as high as those in the United States. Japanese
importbarriersthus operate more like tariffs than like quotas and are
probablydue to unofficialpracticesbased on unusualbuyerpreferences
and monopolies in the distribution system rather than to officially
implementedcartels or quantitativerestraints. I conclude that these
barrierswill not, for the most part, prevent an importresponse to the
strongeryen. Finally, I argue that Japan must take steps beyond the
strongexchangerate to increasethe openness of its markets.
Is Japanese Import Behavior Unusual?
I will use a special case of a model developed by Helpman and
Krugmanto explaintradein differentiatedproducts.Thetheoryexplains
the volumeof importsindependentof comparativeadvantage.It predicts
a relationshipbetween the share of imports in domestic consumption
and the shareof home productionin worldproduction.
In the simplestversionof the model, two countriesof equal size, with
identicalfactorendowmentsandaccess to the sametechnology,produce
the same good: a differentiatedproduct, produced in numerous varieties.'4 Each variety is produced with the same productionfunction,
which exhibits economies of scale. These scale economies are fairly
small, however, so the industryaccommodatesmany producers, each
producinga different variety. In the long run, firms enter the market
untileach earnszero profits.'5Consumersin both countrieshave similar
14. For a rigorouselaborationof this model, see Helpmanand Krugman,Market
Structure and Foreign Trade, chap. 7.
15. Ibid., p. 132. As Helpmanand Krugmanpoint out, "The industry[has]a market
structureknown as monopolisticcompetition;that is, every firmchooses a variety and
pricingstrategyso as to maximizeprofits,takingas given the varietychoice and pricing
strategyof the otherproducersin the industry.In this case every firmends up producing
a differentvarietyof the product."
Robert Z. Lawrence
525
tastes but a preferencefor variety.16 There are no tradebarriersand no
transportationor othertransactionscosts. Tradeis balanced.In equilibrium, there will be n firmsof equal size in each country. With identical
demandcurves for each varietyand cost functionsfor each firm,output
and prices of each firmwill be the same. Consumptionpatternsin each
countrywill be identical,with domesticandforeignfirmsaccountingfor
equalsharesin the purchasesof each consumer.
If countryA doubles in size, A will produce 2n varieties;B, n varieties. Consumers in both countries will allocate two-thirds of their
consumptionto A goods and one-third to B goods. One-thirdof the
productionof A and two-thirdsof the productionof B will be exported.
Similarly,importswill be one-thirdof consumptionin A and two-thirds
of consumptionin B.
Thus, in this frictionless model, a country's share in both national
marketswillbe proportionalto its shareinworldproduction.As Helpman
observes, relative country size is the determinantof trade when all
products are differentiated.'7The larger the country's share in world
production,the largerits share in its home marketand thus the smaller
exportsor importsas a shareof GNP.
What happens in this model if trade is not balanced? Assume, for
example,thatthe economies areof equalsize, each producingn varieties
in similaramounts, but that countryA consumes only half its income
(O.5nif productionof each firm is defined as one unit of income) and
lends the other half to countryB, which consumes 1.5n. If indifference
curves are homothetic (all varieties have unitary income elasticities),
consumers in both countries will consume A and B goods in equal
proportions.A tradesurplusimpliesa greatersharein worldproduction
than in world consumption,but the consumptionproportionsat home
and abroadof home goods remainequal to shares in production.That
16. See Avinash K. Dixit and Joseph E. Stiglitz, "MonopolisticCompetitionand
OptimumProductDiversity,"AmericanEconomicReview, vol. 67 (June1977),pp. 297308. Oneformof the utilityfunction,followingDixit andStiglitz,wouldbe a concave and
symmetricalCES function.The numberof varietiesavailablepotentiallycan be infinite.
However, given fixed productioncosts, a finite numberis supplied in equilibrium.If
varietiesareequallypriced,eachindividualwillconsumeallvarietiesin equalproportions.
See Helpman and Krugman, Market Structure and Foreign Trade, pp. 117-20.
17. ElhananHelpman, "ImperfectCompetitionand InternationalTrade: Evidence
fromFourteenIndustrialCountries,"SeminarPaper304(StockholmUniversity,Institute
for InternationalEconomicStudies,December1984),p. 8.
526
Brookings Papers on Economic Activity, 2:1987
finding is importantbecause it predicts that the share of imports in
consumptiondoes not dependon the tradebalance.
Some of the strongassumptionsin this model can be relaxedwithout
changingthis finding.Increasingthe numberof countriesdoes riotalter
the conclusions. Countriesmay produce both differentiatedand undifferentiatedproductsand may differin factor endowments.18Production
costs could also differbetween countries. Assume, in the two-country
model described above, that A products rise in price by 1 percent. If
demandelasticities are the same worldwide, the share of A in demand
should fall by the same percentage in both markets. Similarly, if the
relativequalityof productsfromA improves,both foreignand domestic
buyersshouldraisetheirpurchasesby similarpercentages.Eachcountry
could specializefully in the productionofjust one variety.Again, shares
in consumptionin each countryandin worldproductionwill correspond.
Three assumptionsare, however, crucialfor this result: similarityin
tastes, absence of trade barriers, and zero transactions costs.19 If
countries have a preferencefor goods made at home, shares of home
goods in domestic consumptionwill exceed those of home goods in
worldproduction.Importbarrierssuch as tariffsor quotaswill raise the
share of home goods in home consumption relative to their share in
world production. Similarly,barriersagainst a country's exports will
lead it to consumerelativelylargesharesof its home production.If there
are internationaltransactions costs, home goods will be relatively
cheaperin the domesticmarketandtheirsharein domesticconsumption
could deviate from thatin worldproduction.20
This analysis suggests that marketshares can be explained by production shares, transportationand transactions costs, trade barriers,
and taste differences:
(la)
(Mij/DUij) = f (Pil/Pi, Tij,Bij),
18. A completetreatmentof this case is providedin Helpmanand Krugman,Market
Structure and Foreign Trade.
19. The assumptionthatutilityfunctionsaresimilarandhomothetic,so thatconsumption patternsare independentof income, is sufficientfor this analysis.If the functionsare
not homothetic,differencesin incomes could affect demandpatternsdespite similarities
in tastes.
20. For an exploration of the impact of transportationcosts, see Helpman and
Krugman, Market Structure and Foreign Trade, pp. 205-09. See also Paul R. Krugman,
"ScaleEconomies,ProductDifferentiationandthe PatternofTrade," AmericanEconomic
Review, vol. 70 (December1980),pp. 950-59.
Robert Z. Lawrence
527
whereM is imports,D U is domesticuse (consumptionplus investment),
P is production,Tis transactionscosts, andB is tradebarriersand taste
differences. The suffix i denotes products andj denotes countries. In
this context tradebarriersincludebarrierson both exports andimports.
If we were to estimateequation la,
(lb)
(MijD Uij) = A + B(P1j/Pi) + C(Tij),
in the frictionlesseconomy, the coefficientC wouldbe zero, A wouldbe
1.0, andB wouldbe - 1.0. In this case, if (Pij/Pi)is zero, the countrydoes
not produce the product, and (Mij/DUij) would be 1-all doinestic use
wouldbe imported.If (Pij/Pi)is unity, the countryaccountsfor all global
production,and (Mij/DUij)would be zero-none of domestic use would
be imported.In the real world, however, in the presence of frictions,
transportationcosts, and nonlinearrelationships, coefficients will not
equal unity. Nonetheless, the coefficients on both the distance and
productionsharesvariablesare expected to be negative.
A SECOND
SPECIFICATION
In the frictionless model outlined above, the shares of a country's
productsin its domesticmarketwill equalits sharesin the foreignmarket.
Thus export shares in foreign consumption should equal the share of
homegoods in domesticconsumption,andforeignmarketsharesshould
offeranalternativemethodforjudgingimportshares.Indeed,thistheory
has some popular appeal. When the semiconductor industry in the
United States argues that Japan follows discriminatorypractices, its
spokesmenpoint to the high shares of U.S. semiconductorproductsin
the United States and other foreign markets in support of their case.
Similarly,those defendingJapanfromallegationsthatit is discriminating
againstforeign goods point to Japanese success in foreign marketsin
electronics and motor vehicles as evidence that Japan simply makes
betterproducts. Again, differencesin tastes, trade barriers,and transactionscosts could influencethe results.
This analysissuggests a second equation:
(2)
(Mij/D Uij) = f(XijlFUi, Tij,Bij),
whereX denotes exports, andFU denotes the use of i in countriesother
thanj.
528
Brookings Papers on Economic Activity, 2:1987
IMPLEMENTATION
With a satisfactoryproxy for internationaltransactionscosts, estimatingequations la or 2 without explicit measures for Bijwill capture
the relationshiptypical of countriesin the samplebetween importsand
productionshares (and export performance)inclusive of the impact of
normaltradebarriersand taste differences. A countrydummyvariable
will not indicatethe total impactof tradebarriersand taste differences.
It indicatesthe impactof only those tradebarriersandtaste differences
that are unusual. A negative dummy variable indicates the aggregate
impact of three kinds of unusual behavior: unusual preferences for
domestic goods, abnormallyhigh importbarriers,and unusualforeign
discriminationagainstthe dummiedcountry'sexports.
For estimatingequations la and 2, I exploit a data set that details
nationalmanufacturingproductionandtradebetween 1970and 1983for
thirteen countries matched at a fairly disaggregatedlevel, with the
manufacturingsector divided into twenty-two industries.21 Domestic
use (consumptionplus investment) for each country in the sample is
estimatedusing the formula
(3)
= Plj - Xlj + MIJ.
~~~~DUlj
I assume that transportationand other transactionscosts (Tij)are
related to the distance of each country from its tradingpartners.The
distances from national economic centers are weighted by shares in
manufacturingproductionanddomestic use. The variableis specifiedin
logarithmicformto capturethe notionthatcosts do not increaselinearly
withdistance.A squaredtermis triedto allowforfurthernonlinearities.22
21. The data must be treatedwith some caution. They are convertedusing current
exchangerates ratherthan purchasing-powermeasures. For a detaileddescription,see
Blades andSimpson,"The OECDCompatibleTradeandProductionDataBase."
22. The distancevariableDjis weightedas
E (2Pk
E -(Pk+
+ 2DUk)
-DUk)
DISTjk
wherePk is the productionof countryk, DUk is the domesticuse of countryk, andDISTjk
is thedistancebetweencountriesjandk. ThisweightingmethodfollowsGarySaxonhouse,
Robert Z. Lawrence
529
RESULTS
Afterdeterminingthe most appropriatefunctionalform(a logarithmic
specificationgave the best results), I entereda dummyvariableequalto
unityin the case of Japan.The results of the twenty-oneregressionsrun
on 1980data for each industryare reportedin table 4.23 The amountof
variance explained excluding the Japanese observation is high (the
averageR2is 0.77). Almost all the coefficients are correctly signed and
most are significantlydifferentfromzero. The coefficienton production
suggestsconsiderablymorebias towardproductionfor the home market
thanthe frictionlessmodelwould suggest. In contrastto the proportional
relationshipin the frictionless model, on average in the twenty-one
industriesused in the estimation,countriessupplied60 percent of their
home marketwhile accountingfor 8 percent of the sampleproduction.
Evaluatedat theirmeans, the typical coefficientof - 0.35 implies that a
1percentagepointrise in productionshare(for example,from8 percent
to 9 percent)lowers the importshareby 1.8 percentagepoints (from40
percentto 38.2 percent).
Equationstandarderrors average 36 percent. In sixteen of twentyone equations the dummy variable on the Japanese observation is
negative;in nineof these the t-ratiois greaterthan 2; in five others it lies
between 1and2. The industrieswith significantandnegativecoefficients
account for 49.9 percent of Japanese manufacturingproduction (and
56.6 percent of Japanese manufacturedgoods trade). Japan's imports
are significantlyhigher than predicted in only one case, nonferrous
metals. No unusual barriersappear (coefficients positive or close to
zero) in the equationsfor aerospace, drugs,chemicals, food, shipbuilding, andpetroleumrefineries.
The export shares specification, reported in table 5, also yields
"What'sWrongwithJapaneseTradeStructure?";PenttiPoyhonen,"A TentativeModel
for the Volumeof TradebetweenCountries,"Weltwirtschaftlichles
Archiv.,vol. 90, nIo. 1
(1960),pp. 93-99;andEdwardE. Leamer,"TheCommodityCompositionof International
Trade in Manufactures:An EmpiricalAnalysis," OxfordEconomic Papers, vol. 26
(November1974),pp. 350-74.
23. Datapeculiarities(a shareof importsin domesticuse greaterthanunityin a large
numberof cases) led to droppingthe industrialcategoryfor miscellaneousmanufacturing
fromthese tests.
Table 4. Tests on Japanese Import Shares in Domestic Use, Production Share
Specification, 1980a
Industry
Aerospace
Office machinery,
computers
Electronic components
Drugs, medicines
Instruments
Electrical machinery
Motor vehicles
Chemicals
Nonelectrical machinery
Rubber, plastics
Nonferrous metals
Other transport
Stone, clay, glass
Food, beverages
Shipbuilding
Petroleum refineries
Ferrous metals
Fabricated metal
products
Paper, printing
Wood, cork, furniture
Clothing, shoes
Distance
Distance squared
-0.60
(- 1.6)
-0.47
(-2.4)
-0.30
(-2.2)
-0.58
(- 3.7)
-0.54
(-3.1)
-0.42
(-4.0)
-0.36
(-2.1)
-6.4
(- 2.2)
-0.33
(-2.5)
-0.45
(-3.1)
5.90
(1.6)
-0.26
(-1.0)
-0.52
(-2.9)
-0.80
(-4.5)
-0.32
(-2.2)
.. .
. ..
. ..
. ..
...
.. .
. . .
0.37
(2.0)
. . .
.. .
-0.46
(-2.0)
...
..
.
. ..
. ..
Production
share
variable
Summarystatistic
Standard
R2
error
-0.10
(-2.1)
-0.24
(-3.4)
-0.28
(-5.4)
-0.45
(-7.5)
-0.27
(-4.6)
0.87
(1.0)
-0.60
(-1.2)
-1.26
(-3.5)
0.58
(1.4)
-0.59
(-1.3)
0.70
0.47
0.40
0.69
0.27
0.79
0.31
0.89
0.35
0.80
-0.37
(-7.9)
-0.22
(-4.1)
-0.35
(-5.9)
-0.38
(-6.6)
-0.37
(-5.7)
-1.10
(-4.0)
-2.75
(-6.5)
-0.48
(-1.3)
-1.20
(-3.5)
0.21
0.90
0.33
0.73
0.26
0.85
0.26
0.85
-1.61
0.29
0.84
-0.09
(-0.9)
-0.28
(-2.8)
-0.30
(-3.3)
-0.10
(-1.0)
-0.50
(-5.6)
1.43
(3.3)
-1.95
(-3.1)
-1.44
(-3.1)
0.32
(0.7)
0.34
(0.9)
0.32
0.90
0.50
0.55
0.36
0.70
0.36
0.71
0.29
0.81
-0.17
(-0.2)
-1.21
(-2.9)
-1.33
(-5.1)
-1.17
(-1.6)
-0.75
(- 1.4)
0.61
0.71
0.32
0.81
0.20
0.95
0.51
0.64
0.35
0.79
-0.55
(-1.5)
0.25
0.88
-14.80
(-2.1)
-0.75
(-4.6)
-0.68
(-6.6)
-14.20
(-2.3)
-15.50
(-3.4)
0.91
(2.0)
. . .
0.87
(2.2)
0.94
(3.2)
-0.43
(-3.2)
-0.30
(-3.8)
-0.43
(-9.0)
-0.17
(-1.0)
0.03
(0.2)
-6.70
(-2.3)
0.40
(2.1)
-0.30
(-5.5)
. ..
Japan
dummy
variable
(-4.3)
Source: Author's calculations. See text description and equation lb. The data are from the OECD Compatible
Trade and Production Data Base.
where M is imports and
a. The dependent variable is market share in each industry i for each country j, MVjIDUijQ
DU is domestic use (consumption plus investment) computed using equation 3. Independent variables are distance,
defined as the distance of each country's national economic center from its trading partners, weighted by shares in
manufacturing production and domestic use; share in production in the world economy, defined as PjjIPi. The Japan
dummy variable equals one for the case of Japan. All variables are in logarithms. Numbers in parentheses are
t-statistics.
Table 5. Tests on Japanese Import Shares in Domestic Use, Export Share Specification,
1980a
Industry
Aerospace
Office machinery,
computers
Electronic components
Drugs, medicines
Instruments
Electrical machinery
Motor vehicles
Chemicals
Nonelectrical machinery
Rubber, plastics
Nonferrous metals
Other transport
Stone, clay, glass
Food, beverages
Shipbuilding
Petroleum refineries
Ferrous metals
Fabricated metal
products
Paper, printing
Wood, cork, furniture
Clothing, shoes
Distance
Distance
squared
-0.79
(-2.0)
-0.67
(-3.2)
-0.53
(-2.6)
-0.91
(-3.7)
-0.79
(-3.9)
. . .
-0.77
(-5.2)
-0.59
(- 2.9)
-9.1
(-2.6)
-0.56
(-3-1)
-0.99
(-4.6)
. . .
. ..
. ..
...
. ..
..
.
0.53
(2.4)
. . .
. ..
Expor-t
share
variable
Summatystatistic
Japan
dummy Standard
var-iable
error
R2
-0.12
(-1.3)
-0.23
(-3.2)
-0.25
(-3.2)
-0.39
(-4.4)
-0.31
(- 3.8)
0.77
(0.8)
-0.58
(-1.1)
-1.01
(-1.8)
0.05
(0.09)
-0.38
(-0.7)
0.78
0.38
0.41
0.67
0.37
0.58
0.47
0.74
0.40
0.74
-0.34
(-5.7)
-0.20
(-3.0)
-0.35
(-4.7)
-0.34
(-4.7)
-0.33
(-3-9)
-0.82
(-2.1)
-2.6
(-4.8)
-0.77
(-4.7)
-1.18
(-2.6)
-1.26
(-2.4)
0.28
0.83
0.40
0.60
0.30
0.83
0.35
0.75
0.37
0.73
-0.01
(-0.1)
-0.22
(-2.1)
-0.24
(-2.7)
0.01
(0. 1)
- 0.49
(-4.0)
1.01
(2.2)
-1.57
(-2.0)
-1.32
(-2.4)
0.22
(0.5)
0.66
(1.2)
0.38
0.85
0.56
0.44
0.40
0.63
0.37
0.68
0.36
0.69
-1.52
(-1.4)
-1.07
(-1.5)
-1.05
(-2.8)
-0.68
(- 1.1)
-0.69
(-1.4)
0.78
0.54
0.45
0.63
0.28
0.90
0.49
0.63
0.36
0.79
-0.56
(- 1.0)
0.38
0.74
-1.3
(-7.0)
-0.53
(-1.9)
-0.82
(-3.7)
-0.82
(-4.4)
-0.46
(- 2.5)
. . .
- 24.57
(-2.7)
-0.96
(-3.9)
-1.19
(-8.0)
-0.78
(-3-1)
-14.56
(-2.9)
1.5
(2.6)
. ..
0.88
(2.7)
-0.53
(-1.8)
-0.24
(-1.6)
-0.42
(-6.1)
-0.38
(-3.0)
-0.02
(-1.1)
-9.33
(2.1)
0.54
(2.0)
-0.27
(- 3-1)
.. .
. ..
...
. . .
. . .
..
.
Source: Author's estimates of equation 2 with data from the OECD Compatible Trade and Production Data Base.
a. The equation is exactly as specified in table 4 except that the production share variable is replaced by an export
share variable (XU/FUij), where X denotes exports and FU is foreign use. All variables are in logs. Numbers in
parentheses are t-statistics.
532
Brookings Papers on Economic Activity, 2:1987
Table 6. Pooled Regressions Explaining Import Share in Domestic Use, 1970, 1980,
and 1983a
Yearand equation
1970
Productionshareequation
Export share equation
Japan ProductionSlummary
dumnmy Production Export Standard
Distance variable
share
share
error
- 0.44
-0.64
(-5.7)
-0.75
(-3.5)
-0.71
(-7.9)
(-3.2)
-0.39
statistic
R2
...
0.67
0.62
-0.27
0.81
0.44
.. .
0.60
0.62
-0.18
(-5.8)
0.72
0.46
. . .
0.65
0.62
-0.15
0.79
0.46
(-15.4)
...
(-8.6)
1980
Production share equation
Exportshareequation
1983
Production share equation
-0.53
-0.55
(-8.0)
-0.72
(-8.9)
(- 3.2)
-0.80
(-4.0)
-0.64
(-7.6)
Export share equation
- 0.58
(-3.2)
-0.93
-0.86
(-9.0)
(-3.9)
- 0.33
(-12.7)
...
- 0.35
(-12.3)
...
(-4.7)
Source: Author's calculations. See tables 4 and 5.
a. Data for the twenty-one industries from the sample of thirteen countries were concatenated to create variables
with 273 observations each. The results reported here, therefore, are summary measures for the results of the twentyone regressions reported in tables 4 and 5. All variables are in logs. Numbers in parentheses are t-statistics.
generally significant coefficients, although it explains less variance
(averageR2 of 0.68) andhas higherstandarderrors(average43 percent).
Negative coefficientson the Japanesedummyoccurin sixteen of twentyone industries,with coefficientsthat are statisticallysignificantin seven
industries(electricalandnonelectricalmachinery;motorvehicles; other
transportation;stone, clay, andglass;rubberandplastics;andfabricated
metals). This test suggests statistically unusual import behavior in
productsaccountingfor 20 percentof production.
The industrydatahave been pooled into singleregressions,each with
273 observations, constraining the coefficients to be similar for all
industries. These equations, reported in table 6 for differentperiods,
have highlysignificantcoefficients. The productionshares specification
explains about 60 percent of the sample variance, the export shares
about 45 percent. The coefficients are relatively stable over time. In
every case, the coefficient of the Japanese dummy is negative and
significant.The typical value on the dummy variable (-0.60) in the
productionsharespecificationindicatesthatJapaneseimportsharesare
unusuallylow by about45 percent.
Data on the membersof the EuropeanCommunityin the samplehave
533
Robert Z. Lawrence
Table 7. Pooled Regressions Explaining Import Share in Domestic Use with European
Community Countries Aggregated, 1980a
Specification
Pooled regression
Total transportb
Total nonelectrical machineryc
Electrical machinery and componentsd
Chemicals and drugse
Instruments
Rubber, plastic
Nonferrous metals
Stone, clay, glass
Food, beverages
Ferrous metals
Fabricated metal products
Paper, printing
Wood, cork, furniture
Clothing, shoes
Summarystatistic
Standard
error
R2
Production
Distance
share
Japan
dummy
variable
-0.20
(-1.3)
-0.38
(-1.2)
-0.17
(-0.5)
-0.24
(- 1.8)
-0.13
(-0.6)
-0.33
(-7.3)
-0.23
(-2.7)
-0.37
(-4.0)
-0.28
(-7.5)
-0.41
(-6.3)
-0.89
(-4.0)
-1.54
(-3.4)
-1.16
(-2.4)
-1.39
(-7.0)
-0.37
(-1.2)
0.66
0.61
0.36
0.80
0.39
0.85
0.16
0.96
0.25
0.94
-0.09
(-0.3)
-0.36
(-1.9)
-2.0
(- 5.4)
-0.19
(-0.8)
-0.39
(- 2.3)
-0.28
(-3.9)
-0.37
(-6.8)
0.23
(1.8)
-0.36
(-5.0)
-0.08
(- 1.4)
-0.96
(-2.2)
- 1.64
(-6.5)
1.01
(2.2)
-1.50
(-4.9)
0.01
(0.02)
0.35
0.85
0.20
0.96
0.37
0.89
0.24
0.92
0.18
0.80
-0.97
(-3.7)
-0.50
(-3.0)
0.48
(0.9)
0.10
(0.3)
-0.12
(-0.4)
-0.21
(-2.7)
-0.39
(-7.9)
-0.26
(-1.3)
-0.03
(-0.2)
-0.31
(-4.2)
-1.24
(-3.6)
-1.58
(-7.0)
-1.17
(-1.5)
-0.61
(-1.0)
-0.59
(-1.6)
0.27
0.93
0.18
0.97
0.63
0.34
0.48
0.02
0.30
0.88
Source: Author'scalculations.
a. Theequationsareas specifiedin table4, exceptthatmembersof the EuropeanCommunityhavebeenaggregated
intoa singleunit,therebyreducingthe countrysampleto eight.Dataforonly fourteenindustrygroupswereavailable;
thusthe pooledregressionequationcontains112observations.All variablesare in logs. Numbersin parenthesesare
t-statistics.
b. Aerospace,motorvehicles, othertransportation,
and shipbuildingcategories.
c. Officemachinery,computers,and nonelectricalmachinery.
d. Electroniccomponentsand electricalmachinery.
e. Chemicals,drugs,and medicine.
been aggregatedinto a singleunit andthe dependentvariableenteredas
the shareof extra-ECimportsin domestic consumption,thus reducing
the sample to eight economies. Data constraintsrequiredaggregating
certainindustrygroupsso that only fourteenindustryregressionscould
be run. As reportedin table 7, the equationpooling industrydata into
variableswith (8 x 14) 112observationsyields a large and statistically
534
Brookings Papers on Econonzic Activity, 2:1987
Table 8. Pooled Regressions Explaining Share of Aggregate Manufactured Imports
in Domestic Use of Manufactured Goods, 1970, 1980, and 1983a
Yearand equation
Distance
Japan Production
dummy Production
share
variable
Export
share
statistic
Summary
Standard
error
R2
1970
Production share equation
Export share equation
-0.42
-0.51
(-3.3)
(-1.7)
-0.66
-0.51
(-3.5)
(-1.1)
- 0.40
. . .
0.26
0.88
-0.44
0.36
0.74
.
0.24
0.83
-0.29
0.31
0.72
.. .
0.24
0.85
-0.26
(-2.9)
0.31
0.74
(-6.9)
..
.
(-4.2)
1980
Production share equation
Export share equation
-0.50
-0.53
-0.29
(-4.2)
(- 1.7)
(-4.8)
..
-0.68
-0.51
(-4.2)
(-1.2)
-0.65
(-4.5)
- 0.87
(-4.7)
-0.61
(-1.9)
-0.55
(-1.3)
.
..
(-3.2)
1983
Production share equation
Export share equation
-0.26
(-4.5)
.. .
Source: Author'scalculationswith aggregateindustrydata.
a. Variablesare as definedin tables4 and 5. All variablesare in logs. Numbersin parenthesesare t-statistics.
significantestimate of Japaneseunderimporting.Regressionsfor single
industriesgive results similarto those in table4.
A second set of summaryequationsexaminesaggregatemanufactured
imports,production,domestic use, and exports. As reportedin table 8,
these equationsyield less precise coefficients but ones similarto those
of the pooled disaggregatedversion. Japanese imports in the 1980
productionshares specificationwere 40 percentlower thanpredicted.
Table 9 reportsthe results of estimatingthe equationsin tables 4 and
5 for differentperiods. The estimates of the Japanesedummiesin these
periodsare similar.
Table 10 reports the dummyvariableson each other country in the
samplewhen these variablesare used in the pooled regressions. In the
1980productionshare specification,only Finland and Japanunderimported significantly.24In the export share specification, the only two
24. The presence of significantcountrydummiescould indicateheteroskedasticity.
As Barbonenotes, "This exercise is questionableon econometricgrounds.The value of
the dummyfor each countryis in fact to be understoodas conditionalon the hypothesis
thatothercountrydummiesarenot significantlydifferentfromzero, a hypothesisrejected
by any othersignificantvalue." Barbone,"Is Japanan Underimporter?"
Robert Z. Lawrence
535
Table 9. Coefficients on Japan Dummy Variables from Import Share Equations,
1970, 1980, 1983a
Productionshare model
Industry
Exportshare model
1970
1980
1983
1970
1.16
(1.1)
0.08
(0.1)
- 1.23
(-2.7)
0.78
(1.4)
-0.58
(-1.0)
0.75
(0.9)
- 0.60
(-1.2)
- 1.26
(-3.5)
0.58
(1.4)
-0.59
(-1.3)
0.92
(1.1)
-0.67
(-1.4)
-1.58
(-4.7)
0.45
(1.1)
-0.56
(-0.9)
1.35
(1.5)
0.28
(0.5)
-0.93
(-1.3)
0.12
(0.1)
-0.41
(-0.6)
0.69
(0.7)
-0.58
(-1. 1)
-1.00
(-1.8)
0.05
(0.09)
- 0.38
(-0.71)
0.90
(0.9)
-0.52
(-0.9)
-1.45
(-3.2)
-0.08
(-0. 1)
-0.37
(-0.5)
-0.70
(-1.8)
Motorvehicles
-2.79
(-4.0)
Chemicals
-0.30
(-0.9)
Nonelectricalmachinery -0.63
(- 1.6)
- 2.06
Rubber,plastics
(-4.1)
-1.10
(-4.0)
-2.75
(- 6.5)
-0.48
(-1.3)
-1.20
(-3.5)
-1.61
(-4.3)
-1.27
(-4.2)
- 2.89
(-6.4)
-0.36
(-1.0)
-1.40
(-4.7)
- 1.65
(-5.0)
-0.39
(0.7)
-2.84
(-3.4)
-0.40
(-0.9)
-0.75
(-1.6)
-1.47
(-2.2)
-0.82
(-2.1)
-2.55
(-4.8)
-0.77
(-1.9)
-1.18
(-2.6)
-1.26
(-2.4)
-1.05
(-2.4)
-2.71
(-4.8)
-0.56
(-1.4)
-1.28
(-3.0)
-1.32
(-2.8)
Nonferrousmetals
0.83
(1.2)
-2.54
(-3.2)
-1.42
(-2.8)
0.21
(0.7)
-1.07
(-1.5)
1.43
(3.3)
-1.75
(-3.1)
-1.44
(-3.1)
0.32
(0.7)
0.34
(0.9)
2.42
(3.7)
-2.86
(-5.4)
-1.20
(-2.6)
0.25
(0.4)
-0.97
(- 1.2)
0.71
(1.1)
-1.62
(-1.6)
-0.99
(-1.6)
0.10
(0.3)
-1.24
(-1. 1)
1.02
(2.2)
-1.57
(-2.0)
-1.32
(-2.4)
0.22
(0.5)
0.66
(1.2)
2.09
(3.3)
-2.32
(-3.4)
-1.11
(-1.9)
0.13
(0.2)
-1.50
(-1.5)
0.27
(0.4)
-0.97
(-2.4)
- 1.16
(-2.4)
- 1.46
(-2.4)
- 1.01
(-3.4)
-0.17
(-0.2)
-1.21
(-2.9)
-1.33
(-5.1)
-1.17
(-1.6)
-0.75
(-1.4)
-0.69
(-0.7)
-0.29
(-0.4)
-1.22
(-5.1)
-1.08
(-1.5)
-0.86
(-2.0)
-0.90
(-1.0)
-0.65
(-0.9)
-0.89
(-1.5)
-0.84
(-1.4)
-0.82
(-2.8)
-1.52
(-1.4)
-1.07
(-1.5)
-1.05
(-2.8)
-0.68
(- 1.1)
-0.69
(-1.4)
-2.44
(-1.9)
-0.17
(-0.1)
-0.99
(-2.4)
-0.59
(-0.9)
-0.82
(-2.2)
-0.88
(-1.9)
-0.55
(-1.5)
-0.69
(-1.9)
-0.35
(-0.6)
-0.56
(-1.0)
-0.70
(-1.2)
Aerospace
Officemachinery,
computers
Electroniccomponents
Drugs, medicines
Instruments
Electricalmachinery
Othertransport
Stone, clay, glass
Food, beverages
Shipbuilding
Petroleumrefineries
Ferrousmetals
Fabricatedmetal
products
Paper,printing
Wood, cork, furniture
Clothing,shoes
1980
1983
Source: Author's calculations.
a. Equations specified as in tables 4 (production share model) and 5 (export share model) for the years indicated.
Numbers in parentheses are t-statistics.
Brookings Papers on Economic Activity, 2:1987
536
Table 10. Coefficients on Country Dummy Variables from Pooled Import Share
Equations, 1970, 1980, 1983a
Production share model
Country
Australia
1970
1980
1983
Export share model
1970
1980
1983
0.74
0.42
0.28
Finland
(3.5)
0.15
(0.9)
0.46
(2.9)
-0.54
(1.9)
0.10
(0.7)
0.33
(2.3)
-0.55
(1.2)
0.07
(0.4)
0.21
(1.3)
-0.45
(5.0)
0.46
(2.2)
0.41
(2.1)
-0.25
(6.9)
0.39
(2.2)
0.19
(1.1)
-0.26
(6.2)
0.41
(2.0)
-0.16
(-0.8)
-0.12
France
(-3.2)
-0.17
(-3.7)
-0.15
(-2.8)
-0.13
(-1.2)
-0.41
(-1.4)
-0.33
(-0.6)
-0.32
(-1.0)
(-1.0)
(-0.8)
(-2.0)
(-1.9)
(-1.6)
0.21
(1.3)
-0.42
(-2.5)
-0.67
0.28
(1.9)
-0.20
(-1.3)
-0.60
0.33
(2.1)
-0.17
(-1.0)
-0.63
0.05
(0.3)
-0.51
(-2.5)
-0.72
0.16
(0.9)
-0.05
(-0.2)
-0.79
(-3.7)
(-3.7)
(-3.5)
(-3.2)
(-3.9)
(-3.8)
0.24
(1.4)
-0.32
(-1.8)
0.09
(0.6)
-0.22
(-1.4)
0.02
(0.2)
-0.31
(-1.8)
0.42
(2.0)
0.12
(0.6)
0.22
(1.2)
-0.10
(-0.5)
0.11
(0.5)
-0.09
(-0.4)
0.27
(1.7)
-0.15
(-0.7)
0.24
(1.3)
0.22
(1.5)
-0.12
(-0.8)
0.58
(3.4)
0.25
(1.6)
-0.13
(-0.8)
0.84
(4.6)
0.38
(2.0)
-0.84
(-3.7)
-0.33
(-1.5)
0.16
(0.9)
-0.37
(-2.1)
-0.31
(-1.5)
0.37
(1.9)
-0.42
(-2.2)
-0.21
(-1.0)
Belgium
Canada
Germany
Italy
Japan
Netherlands
Norway
Sweden
United Kingdom
United States
1.30
1.62
1.60
0.21
(1.0)
0.06
(0.3)
-0.85
Source: Author's calculations.
a. Regressions pooled as in table 6, with a dummy variable equal to one in the case of EC member countries.
significantunderimporterswere the United Kingdomand Japan, with
the coefficienton theJapanesedummytwice thaton the UnitedKingdom.
OnlyJapanis an underimporterin both specifications.25
INTERPRETATION
These resultsrejectthe view thatJapanesemanufacturedimportsare
notunusuallylow. Theyalso indicatethatthe superiorqualityof Japanese
25. A peculiaroverimporterisAustralia,whichis well knownto havehighlyprotective
barrierson manufacturedgoods. The Australiandummyis the resultof havingJapan,a
country of similardistance and much lower import shares, in the sample. When both
JapaneseandAustraliandummiesareintroduced,however,the Australianobservationis
no longersignificant.
Robert Z. Lawrence
537
productscannotexplainJapaneseimports.Japaneseexportvolumes are
too smalltojustify Japan'shighshareof its home market.But the results
leave some issues unresolvedbecause they do not indicatethe relative
importance of export barriers, import barriers, and unusual buyer
preferences.
Because, with the exception of Japanese textiles, Japaneseexports
were not subject to unusual barriersin 1970, it seems reasonable to
interpret the 1970 dummy coefficients as reflecting unusual import
behavior. The relative stability in the aggregate coefficients in the
productionshare specificationbetween 1970and 1980suggests that this
behaviorpersisted. In some industries, though, an upwarddrift in the
coefficientscouldreflectthe impositionof exportrestraints.To interpret
the coefficienton the dummyvariablesas a reflectionof differences in
preferences,we mustassumethatJapanhas differenttastes anddisplays
an abnormalbias for home products.Since most countriesin the sample
have similarper capita GDP income levels, with none deviating more
than 20 percent from the groupmean, nonhomotheticityis an unlikely
cause.26
SIMULATED
IMPACT OF BARRIERS
How differentwould Japan'stradestructurebe if its importbehavior
were normal given its distance from other producers? The dummy
variablecoefficientscan be used to providea roughanswer.
In 1980, Japanese manufacturedgoods imports and exports were
valuedat $31.5 billion and $126 billion, respectively. The coefficient in
table 8 indicatesthat manufacturedimportswould have been higherin
1980by 59 percent, or $18.6 billion, in the absence of unusualbarriers.
Initially,therefore,removingthese barrierswould reduce the manufacturedgoods tradesurplusfrom $94.5 billionto $75.9 billion.
In responseto a rise in imports,however, the yen would weaken. As
impliedby the equationsystemreportedin table 11,to restorethe current
accountto its formerlevel (in yen) with GDP unchangedwould require
a real yen depreciationof about 10.5 percent.27If this depreciation
26. Correlatingthe shares of twenty-two industries in Japanese production and
domesticuse with sharesin the rest of the samplesuggestsnothingunusualabouteither
Japaneseproduction(correlation0.91)or use (correlation0.96).
27. Equationsindicateelasticitieswith respectto the realexchangerateof - 0.91 and
1.07for exportsandimportsof goods and services, respectively.See table 11.
538
Brookings Papers on Economic Activity, 2:1987
Table 11. Aggregate Export and Import Volume Equations, Japan, 1970-85a
Dependent variable
Total exports
Manufactured exports
Total imports
Manufactured imports
World
manufactured
Utilization
Real
Real
exchange
exportsb
ratioc
GDP
rated
1.48
(36.6)
1.39
(31.2)
...
-1.31
(2.6)
-1.59
(-2.9)
...
...
...
...
. ..
0.80
(14.3)
1.84
(13.2)
-0.91
(-6.6)
-0.75
(-5.0)
1.07
(5.6)
1.06
(2.3)
Summarystatistic
S
Standard Durbinerror
Watson
0.03
2.0
0.03
1.7
0.04
2.1
0.09
1.9
Sources: Author'scalculationswithdatafromOECDEconomicOutlook,OECDNationalIncomeAccounts,and
UnitedNations, UN MonthlyBulletitnof Statistics,variousissues.
a. All variablesin logs.
b. Volumeindexof worldexportsof manufactured
goods.
c. Ratioof actualGDP to "normal"GDP, definedas a nine-yearcenteredmovingaverageof actualGDP.
d. Morgan-Guaranty
realexchangerateindex.
occurredwith prices at home and abroadfixed in domestic currencies,
the value of 1980 Japanese manufacturedgoods imports and exports
would then decline to $44.5billionand $123billion,respectively.28Once
these generalequilibriumadjustmentsareallowedfor, removingunusual
characteristicsof Japanese manufacturingimport behavior would increase manufacturedimportsby 41.3 percent and reduce the surplusin
manufacturedtradein 1980by 16.9percent, or $16 billion.
The dummyvariablesestimatedfor individualindustrieswere used
to obtainfirst-roundeffects on importsharesof the removalof uniquely
Japanesetradebarriers.The results, in table 12, show that the structure
of Japanese trade would alter quite appreciably.Although its overall
manufacturingsurpluswouldnot changemuch,Japanwouldhave almost
as muchintra-industrytradeas the other countriesin the sample.
Thefindingsin this sectionindicatethatJapanesemanufacturedgoods
importshave been reduced significantlyby unusualtrade barriersand
preferences.These factors have a large effect relativeto manufactured
importsbuta muchsmallereffect relativeto the totalvalueof all imports.
It is not surprising,therefore, that tests explainingaggregateJapanese
importsmay fail to detect those barriers.29
28. The equations(table 11)indicatereal exchangerateelasticitiesof -0.75 and 1.06
for Japanesemanufacturedexportsandimports,respectively.
29. In 1980, Japan's aggregatemerchandiseimports were valued at $139.9 billion.
Thus the $21.4 billion impactof trade barrierson manufacturedimportswas only 15.3
percentof totalJapanesemerchandiseimports.
Robert Z. Lawrence
539
Table 12. Simulated Values of Japanese Manufactured Imports, 1980
Item
Actual value
Simulated value
Actual averages,
other countriesb
Correlation
between
exports
and importsa
Intra-industry
trade indexa
Mean exportimport 1.atioa
Ratio of imports to domestic use
-0.22
0.30
7.86
0.06
0.55
0.57
2.21
0.10
0.66
0.70
1.31
0.35
Source: Author's calculations. See text description. Simulated values calculated using dummy variables in table 4.
a. Twenty-one manufacturing sectors.
b. Average values of other countries listed in table 2.
Price Responsiveness
By examining the price responsiveness of Japanese imports, it may
be possible to determine the nature of the barriers more precisely. Zero
price elasticities would suggest quotas due to cartels and government
policies such as administrative guidance and prohibitive regulations.
Low (and nonlinear) price elasticities would suggest significant fixed
costs due to entry barriers.30Normal price elasticities, combined with
evidence of underimporting, could indicate tariffs, inadequate competition in the distribution system,31 and unusual buyer preferences.32
Evidence on income elasticities is also useful. Zero price and income
elasticities could indicate fixed quotas. Normal income elasticities and
zero price elasticities would suggest quotas set in terms of market share
rather than in fixed quantities.
30. New entrantswouldbe forcedto recoverthe fixed entrycosts in theirprices. For
small initialchanges they would be much less responsive. In the aggregate,therefore,
importelasticities would be lowered in response to an initial appreciationof the yen.
However,responseswouldnot be symmetrical,linear,or constantover time. See Richard
R. Baldwinand Paul R. Krugman,"PersistentTrade Effects of Large Exchange Rate
Shocks," WorkingPaper2017(NationalBureauof EconomicResearch,August 1986).
31. Monopolistsset prices equalto marginalcost plus a markupequalto (1 + l/ed),
where ed is the elasticity of demand.If competitionto distributeimportedproductsis
weak, markupswill be high, but lower importprices will be reflectedin lower consumer
prices. While highermarkupsreduce the quantitysold, they resemble tariffs (if ed is
constant)andwill not affectresponsivenessto pricechanges.
32. If the utility function (U) is Cobb-Douglas,for example, so that log U equals
AljogH and (1 - AI) log I, where H is the domesticgood, and I the importedgood and
Al > (1 - Al), the shareof income spent on home goods will always be higherthanthat
spenton domesticgoods, butbothwouldhave unitarypriceandincomeelasticities.
O
_
00
Ce
n w0
U
;
.
00
X
.
t
.
O1
X
.
C
O.
o
X
N
N
.
O
C
o.
w.
>0
g~~~~~~~0
.
D
O
o
Q~~~~~~0
=
Q C1
o
Y~~~~~~~~~~~~~~~~~
D
h
?
O*
4
Zs
r.
O
O
O
O
O
C
O0
00-
tn
m
t
U
~~00
m
o
L'
X~~~~~~
;~~~~~-
t:)t:
;
000
:
:
t-
:
00
tl
O)
:
:
)
C1
-\
n
0
"It
CD
:
o
m
OC
to_omXNbmotmN_...
c ?mo
606 W=
:
0\
00
1-
o
o
o
m
U
Q~
o
e
=
X
.?
,
> ce
b
o
c.
b.
?
m.
$
o
oo
b
S
N
?
O
?
o
N
?
-
m
m
4
sElo
s
s
oe
o
o
cq~~~Se.t
.s
Xc|
|
'
t~~~~~~~~~~~~~~~~~~~~
r~~~~~~~~~~~~~~~~~~~~~~,
C)~~~~~~~~~~~~~~~~~~C
t
<>
t-0
-mm
;
b
O_
m
O_
O
O_
oo
O_
t
O_
)
tm
to
o
O_
O
O_
m
O_
N
O_
O
0
4=
Q~~~~~~~~~~~~~~~~~
UC:
O_
~~
~~S ~
_ :zSXt
I
cq
m
~
_
0
N
t-
Z2Y
=. 0
tO
ex~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0
r.
~~~
t
;>o
~~~~E
X2$6~~~~~~~
t
2
.0
X
2:.E
OCC) tr C)
ZSOC
exU
|
09
0Z
=
3~~~~~~~~~~~~~~
o~~~~~~~~~~~~~~~~~~
<
=
a
o 00x
-s*
~
~
m
o
_
3..
~ ~ ~ ~ ~ ~~
:s
c u
IDA
~~~~~~t'
t'c)
?
S
e
C)
C)
E
r~~~~
I=
-Q~~~~~~~~~~~~~
-
Ci
Robert Z. Lawrence
541
To explorethisissue, I haveused the matchedindustrialdatacompiled
by the Commissionof the EuropeanCommunity.These data provide
consistent importvolumes in majorindustriesfor Japanand the United
States. In addition,I matchedimportprices (unit values in the case of
the United States) to domestic wholesale prices for data on ten industry
groupscomprisingabout80 percentof manufacturedimports.
I have estimatedimportdemandfunctions that explainimportquantities between 1971and 1985by real GNP, relativeimportprices, and a
time trend. Variables are expressed in logarithmsso that coefficients
indicateelasticities. Specificationsusing currentand laggedincome and
price termswith andwithoutthe trendtermwere tested. The best fitting
estimatesare reportedin table 13.
Japanesefinishedmanufacturedimportsappearfairly responsive to
price. Indeed, in five of the ten industrialsectors estimatedhere, they
are more responsive than the correspondingimports of the United
States.33The price elasticity estimates and the estimates of the barriers
obtainedin the previous section do not correspond.Chemicals,in which
importsconformedto normalpatterns,have the highestprice responses
(elasticity of - 2.2). Elasticities are between - 1.4 and - 1.5 for under-
importedproducts such as electrical goods, nonmetallicminerals, and
metal products. On the other hand, price elasticities are low and not
significantlydifferentfromzero in threeof the largestsectors of Japanese
imports,basic metals (28.3 percent of manufacturedimports);textiles,
footwear, and clothing(7.3 percentof manufacturedimports);and food
products(13.1 percent).
Theorysuggeststhatlow priceelasticitiesmightindicateadministered
protection. In fact, there is substantialevidence of such protection in
both textiles and food products. Japan maintains quotas on leather
goods. Recession cartels, accompanied by administrativelimits on
imports(and voluntaryexport restraintson Korean goods) have been
33. Earlier studies of disaggregatedJapanese manufacturedimports do not find
unusuallylow importpriceelasticities.See, for example,Joe A. Stone, "PriceElasticities
of Demandfor Importsand Exports:IndustryEstimates for the U.S., the E.E.C. and
Japan," Review of Economics and Statistics, vol. 61 (May 1979), pp. 306-12; William R.
Cline and others, Trade Negotiations in the Tokyo Round: A Quantitative Assessment
(Brookings, 1978). For a comparisonof these results, see Peter A. Petri, Modelling
Japanese-American Trade: A Study ofAsymmetric Interdependence (Harvard University
Press, 1984),p. 54.
542
Brookings Papers on Economic Activity, 2:1987
implemented in the textiles industry.34According to Balassa, food
productsare a principalcase of administrativeguidancein Japan.35
Japanese manufacturedimports are price inelastic in basic metals,
motor vehicles, textiles, and food-products amountingto about half
the total. They are elastic in sectors that have relatively small import
shares. Thus althoughJapanese products generally have higher price
elasticitiesthanU.S. products,when weightedby theirshares,Japanese
importshave an elasticity of - 0.70 while those of the United States are
-1.1.
When the time trends are taken into account, Japanese income
elasticities are generally between a quarter and a third those of the
UnitedStates. Thisdifferenceis somewhatlargerthanmightbe expected
from those countries' growth rates. Between 1970and 1985, Japanese
industrialproduction increased at a 4.5 percent rate, as against 2.7
percentfor the United States and2.6 percentfor all industrialcountries.
Table 14 reports similar regressions for aggregate manufacturing
imports in Japan, the United States, Italy, France, Germany, and the
United Kingdom. The equation for Japan gives results similarto the
disaggregatedversion. Indeed,the unitaryimportpriceelasticityis fairly
close to the weightedaverageof priceelasticitiesfromthe disaggregated
equations.36This Japaneseimportprice elasticity lies in the middle of
the estimatesin table 13.
In sum, therefore,the barriersthat inhibitJapaneseimportsdo not,
in most cases, preventfairlynormalprice responses. These barriersare
thus probablylimitationsin the distributionsystem and differences in
buyer preferences, a conclusion supportedby Christelow'sfindingthat
the distributionsystem marksup importsdifferentially.37
34. For a detaileddescriptionof the 1981recession cartel in textiles and the role of
MITI,see RonaldDore, "StructuralAdjustmentinJapan1970-82"(Geneva:International
LaborOffice).
35. Balassa, "Japan'sTradePolicies."
36. Recentestimatesof Japanesemanufacturedgoods importpriceelasticitiesinclude
Hooper/Helkie(-0.90), the EPA Model (- 1.2), and Spencer (-0.82). See WilliamL.
Helkie and Peter Hooper, "The U.S. ExternalDeficitin the 1980s,"in RalphC. Bryant
and others, eds., Empirical Macroeconomics for Interdependent Economies (Brookings,
1988);NaohiroYashiro, "ExchangeRate Adjustmentand MacroeconomicPolicy Coordination,"DiscussionPaper41 (Tokyo:EconomicPlanningAgency, February1987);and
GrantH. Spencer, "The WorldTradeModel: Revised Estimates," IMF Staff Papers,
vol. 31 (September1984),pp. 469-98.
37. "A (1985)governmentsurvey of distributionmarkupsfor domesticand imported
productsfound that for whiskeys, candies, edible oils, men's overcoats, and footwear,
543
Robert Z. Lawrence
Table 14. Equations Explaining Aggregate Real Manufactured Imports,
Various Countries, 1971-86a
Country
Franceb
Germanyc
Italy
Japanc
United Kingdome
United States
Real
GDP
2.69
(22.5)
2.77
(40.7)
2.72
(16.2)
1.35
(5.2)
2.47
(8.0)
3.36
(19.1)
Relative
price of
imnports
Time
trend
- 1.11
(2.7)
- 0.54
(3.2)
.
-2.13d
.
(3.1)
-1.01
(-2.6)
-0.61
(3.4)
- 1.95d
(5.3)
0.03
1.8
0.02
1.5
0.06
1.5
...
0.09
1.9
0.02
(2.7)
0.02
1.6
0.06
2.0
.
.
Summary statistic
Standard
Durbinerror
Watson
. ..
.
.
.
.
.
MonetaryFund,International
Source: Author'scalculationswithdatafromthe OECDdatabase andInternational
Financial Statistics datatape.
importsin each country.The relativepriceof importsis
a. Annualdata.Dependentvariableis real manufactured
goods. All variablesare in
the ratioof manufactured
importprices to domesticwholesalepricesfor manufactured
logs, except for the time trend.Numbersin parenthesesare t-statistics.
b. 1973-85.
c. Correctedfor serialcorrelation.
d. Currentand one-pefiodlag.
e. 1972-86.
f. Currentand two-periodlags.
Between early 1985and April 1987, as measuredby the multilateral
exchangerate model of the InternationalMonetaryFund (MERM),the
yen appreciated44 percent. Over the same period, Japanese relative
unit laborcosts increasedabout 22 percent. Withthe exception of food
and textiles, the strong yen should increase the importance of the
Japanesemarketas a revenue source for foreign exporters. Not only
will Japan'strade surplusbe reduced, but the amountof intra-industry
tradeshouldrise. Indeed, in the firstfive monthsof 1987,the volume of
Japanesemanufacturedgoods importswas 21.5 percent higherthan in
1985.Becauseimportunitvalues were 24.9 percentlower, however, the
overallyen value had barelychanged.In dollars, manufacturedimports
increasedfrom$36.4billionin 1985to an annualrate of $50.94billion.38
markupson importswere double those on domestic products." Christelow, "Japan's
IntangibleBarriers,"p. 14.
38. For a detailedanalysis of recent Japanesetrade, see Bonnie E. Loopesko and
Robert A. Johnson, "Realignmentof the Yen-DollarExchange Rate: Aspects of the
AdjustmentProcess in Japan," InternationalFinance Discussion Paper 311 (Board of
Governorsof the FederalReserveSystem, August1987).
544
Brookings Papers on Economic Activity, 2:1987
Political Implications of Japan's Trade Structure
Whateverthe causes of Japan's unusual trade structure,its effects
have been dramatic.Few internationalmeetings go by without heated
accusations against Japanand defensive responses. Highly dependent
on its tradingrelationships,Japan has been forced time and again to
bend to externalpoliticalpressures.It has restrainedexports of textiles,
steel, automobiles, and electronic products and ostentatiously proclaimeda series of majorinitiativesto open its markets.
Yet Japancontinues to be the targetof frequentallegationsof unfair
trade practices. Currenttradelegislationdraftedby the U.S. Congress
is aimedprimarilyat Japan.Both the House andSenatetradebillspassed
in 1987would amend Section 301 of the TradeAct to requirethe U.S.
Trade Representativeto identify countries that "maintaina consistent
pattern of market distortingtrade practices," specifically identifying
Japanas such a country.39
Yet from some standpoints,the attentionfocused on Japanappears
peculiar.For its size, Japanis not an extraordinaryexportingnation. In
1980, Japanesemanufacturedexports were 13.2 percent of the exports
of the thirteen-countrysampleused above. By contrast,Japanaccounted
for 18.7 percent of manufacturedgoods use within the sample and 19.9
percent of production.Why is the world able to absorb West German
manufacturedexportswith little or no frictionwhen they are largerthan
those of Japan($217billionas against$204billion in 1986)?Why, in the
early 1980s, was the United States able to export more manufactured
goods thanJapanwithlittlefriction?Andwhy have the tensionsbetween
the United States and Japanbeen so heated in recent years, when the
decline in the U.S. manufacturedgoods tradebalancewith Europe and
Canada($46.5 billion between 1981and 1986)has been largerthan the
decline with Japan($38.4billion)?40
39. Forananalysis,see RobertE. Litan,"Is the UnitedStatesTurningProtectionist?"
Japan Automobile Manufacturers Association Forum, vol. 6 (October 1987), pp. 3-8.
40. Indeed,the declinein the U.S. bilateralbalancewith Japanhas been proportional
to its 1981sharesin manufacturedgoods trade. See RobertZ. Lawrenceand RobertE.
Litan, "TheProtectionistPrescription:Errorsin DiagnosisandCure,"BPEA,1:1987,pp.
289-310.
Robert Z. Lawrence
545
The politicalimplicationsof its tradestructureexplainwhy Japanhas
been singledout. To be sure, Germanyhadthe advantageof establishing
its marketshareduringthe periodof rapidinternationalpostwargrowth,
while since 1973Japaneseexport shares have been growingrapidlyin
more stagnant markets. Germany also has the benefit of virtually
unimpededaccess to EuropeanCommunitymarkets.But a majorreason
for the better reception of Germanmanufacturedexports is the high
value of German manufacturedimports. Similarly, the mutual interpenetrationof European,Canadian,andAmericanmanufacturedgoods
marketsis substantial.In 1986,U.S. manufacturedgoods importsfrom
Europe and Canadawere 70 percent and 30 percent more than U.S.
exports, respectively.41 By contrast, U.S. manufacturedimportsfrom
Japanwere five times as large as U.S. manufacturedgoods exports to
Japan. Intra-industrytrade is conducive to maintainingfree trade because it sets up counterweightpressures. Protection of an industry is
less likely when participantsin that industry have a direct interest in
sellingin the country againstwhich barriersare being sought. Because
Japanimports so little, firms and workers abroadwill not come to its
defense when protectionis sought.
If Japanis to participatefully in the global economy in the decades
ahead,it mustreduce these protectionistconstraintson its trade. Given
its high domestic saving rates, Japanwill tend to have currentaccount
surplusesin the foreseeable future. One response has been to urge the
Japaneseto spendmore of theirincomes. But manyJapanesearguethat
the rapidagingof the populationcompelshighpresentsaving.And since
developingcountrieshave greatdemandsfor capital, it would be useful
for Japan to invest its savings abroad. Thus, a superior long-term
approachwould create the politicalconditionsin which Japanis able to
run whatever currentaccount surplusit chooses. Indeed, as Stephen
Marrishas noted, Great Britain ran a current account surplus that
averaged4 percentof its GNP (quite similarto Japan'ssurplusin 1986)
forfiftyyearsbetween 1860and 1910.42In Britain'screationof a political
structureconducive to such surpluses, its empire and powerful naval
fleet were not inconsequential.For an economic "Pax Nipponica" in
41. In 1981,the UnitedStatesrana surplusof $3.7billionin manufacturedgoods trade
withEuropeand$2.0billionwith Canada.Ibid.
42. Stephen Marris, Deficits and the Dollar: The WorldEconomy at Risk (Washington,
D.C.: Institutefor InternationalEconomics, 1985),p. 92.
546
Brookings Papers on Economic Activity, 2:1987
which it can trademorefreely, Japanwill have to cultivatea hospitable
internationalenvironmentwith more finesse. Even Britain, however,
had an open home market that smoothed relations with its trading
partners.Indeed,althoughits currentaccountwas in surplus,throughout
this period Britainactuallyran deficits in merchandisetrade, offsetting
themby a surplusfromservices transactionsin shipping,insurance,and
financeand a large surplusin net foreignearnings.43
A more open domestic marketand greaterintra-industrytradecould
help relieve the politicalpressuresfaced by Japan.The Japanesehave
not been unmindfulof the need for better political relations. Their
strategyso far has been to use direct foreign investmentand voluntary
exportrestraints."But voluntaryexportrestraintsdelay the openingup
of Japan,andforeigninvestmentmay also provokepoliticalresentment.
Ultimately,as the Britishexamplesuggests, an open Japaneseeconomy
will be crucial.
Conclusions
Japanese trade runs into protectionist constraints more frequently
than that of other countries, in part because of Japan's unusual trade
structure, especially the small value of manufacturedgoods imports.
Overwhelmingly,suchimportsarelow because Japanhas a comparative
advantagein producingmanufacturedgoods and because it is distant
from its tradingpartners. In addition, however, peculiar features of
Japanesetrade reduce manufacturedimportsby a substantialpercentage, but a relativelysmallabsolutevalue.
In 1980, for example, as equations in this study indicate, unusual
barriersandpreferencesreducedJapanesemanufacturedgoods imports
about 40 percent. The removal of such barrierswould have increased
the value of Japanese manufacturedimports about $13 billion dollars
and lowered Japan's trade surplus in manufacturedgoods about $16
billion, or some 17percent. Japanwould have considerablymore intraindustrytrade.
43. See Albert H. Imlah, Economic Elements in the PaxBritannica (Russel and Russel,
1958).
44. See Kiyohiko Fukushima,"Japan'sReal TradePolicy," ForeignPolicy, no. 59
(Summer1985),pp. 22-39.
Robert Z. Lawrence
547
With the exception of basic metals, food, clothing, and textiles,
Japaneseimportsof manufacturedgoods arefairlyresponsiveto relative
price changes. This price responsiveness suggests that the abnormally
low level of manufacturedimportsis due to unofficialimpedimentssuch
as unusualbuyer preferencesand lack of competitivenessin the distribution system. To be able to export freely, Japanneeds to improve its
internationalpolitical environment, open its domestic market, and
restructureits trade.
Two complementary elements in such restructuringwould be a
strongeryen supportedby an easier Japanesefiscal policy. Japancould
temporarilylower its nationalsavingrateby fiscal measuresand initiate
a period of lower current account surpluses. Because imports are
responsiveto relativepricechanges,manufacturedgoods importswould
increase significantly.
A period of a strongyen could have irreversibleeffects. Foreigners
would establishdistributionnetworksand alter distributionstructures,
enablingthemto hold on to marketseven if the yen shouldreverse itself.
And Japanese producers would sink costs into assembly operations
abroadto service the domestic market.
But even with the yen above currentlevels, the Japanesemarketis
likely to remain relatively closed, and the tensions from the unusual
intangiblebarriersto imports are likely to continue. Although import
barriersmay not have a majorimpact on the Japanese manufactured
goods trade surplus, it is still urgent to remove them. Indeed, Japan
cannot affordto have such barriersbecause its fundamentalstructure,
its need to run large manufacturedgoods trade surpluses, makes it
vulnerableto politicalpressures.
It oftenseems expedient,in theface of foreignprotectionistpressures,
for Japan to limit exports rather than increase imports. But such a
response does nothingto create a sustainablestructurefor the future.
Indeed, by avoiding domestic adjustment,it has increased pressures
over the long run. Ultimately, a sustainablestructurerequiresan open
domesticmarket.Accordingly, simply allowingthe currencyto appreciate may not be sufficient.45Japan should take active measures to
45. Thewelfareimplicationsof openingJapan'seconomythrougha strongyen, which
improvesJapan'stermsof trade,mightbe quitedifferentfromremovingimportbarriers,
whichwouldworsenthem.
548
Brookings Papers on Economic Activity, 2:1987
increase the competitiveness of its distributionsystem and to avoid
administrativeguidance and selective governmentprocurementpractices.46
A final considerationhas to do with Japan's agriculturaltrade. If
Japan removes agriculturaltrade barriers, whose existence is not a
matterof dispute, and increases its agriculturalimports, it will have to
shift its spendingpatterns or, more likely, export more manufactured
goods to pay for the imports.Creatinga favorableenvironmentfor these
exportsby enhancingtwo-way tradeshouldbe a majorpolicy objective.
46. For an analysis of the protectionistproblemsassociated with Japanesepolicies
towarddecliningindustries,see RobertZ. Lawrence,"A DepressedView of Policiesfor
Depressed Industries," paper preparedfor conference on U.S.-CanadianTrade and
InvestmentRelationswith Japan(Universityof Michigan,April2-3, 1987).
Comments
and Discussion
Paul Krugman: RobertLawrence has producedan importantpaperone that offers the seal of approvalof the economics establishmentto
the charge that Japan's marketsfor manufacturesare tacitly closed to
imports. His conclusion is that Japanimports about 40 percent fewer
manufacturesthan it would if its marketswere as open as those of the
average non-JapaneseOECD country. That is, with freed markets it
would importabout80 percentmore. Given the imprecisionnecessarily
involved, let's roundit up andsay thatJapanmight,if it would only open
its markets, double its manufacturesimports. We realize immediately
thatthe paperis sayingsomethingvery explosive. Let me say at the start
thatI believe that Lawrenceis very probablyright,althoughsome of the
methodsused are a little shaky. I ammuchless certainaboutthe political
economy offered here, which prescribes trade liberalization as the
answerto Japan'sworseningrelationswith its tradingpartners.Before
I turnto this question, however, let me review the logic of the economic
analysis.
The basic piece of evidence on which the conclusions rest is the
estimationof a numberof import equations based on a differentiatedproductsmodel of tradein manufacturedgoods. Flatteringly,the paper
offers this approach as one based on the monopolistic competition
models of trade volume developed by Elhanan Helpman and me.
However, Lawrenceextrapolatesconsiderablyfrom the highly stylized
analysis that Helpmanand I offered. We used the assumptionof zero
transportcosts as a fundamentalsimplifyingdevice; Lawrence,because
he is interested in reality, is obliged to modify that approachto take
transportationcosts into account.
Now the step from the no-transport-costtheoretical model to the
realistictransport-cost-inclusiveempiricalmodel is not a rigorousone.
549
550
Brookings Papers on Economic Activity, 2:1987
In fact, as anyone who has worked on the problem knows, it is very
difficultto derive a tractablemulticountrymodel of the volume of trade
in the presence of transportcosts. This task is hard in conventional
models; it is if anythingharderwhen we allow for increasingreturns,
because there are all kinds of complicatingeffects that arise from the
fact that countries with large domestic markets tend to be favored
locations for industries subject to large economies of scale. So the
equationsestimatedhere are inspiredby the Helpman-Krugmanmodel
ratherthanderivedfromit.
Whatmakesthis worrisomeis thatthe role of transportcosts is crucial
in assessing Japan's openness to imports. Japan is the only resourcepoor advanced country not in Europe, and the only large industrial
nation with no industrialneighbors. We need to answer whether the
geographicalisolation of Japan is enough to explain its low share of
importedmanufactures,yet we lack a really well-specifiedmodel that
lets us assess this issue. The Lawrencepaperhas what I thinkis a more
plausibletest that makes better use of the availabledata than any other
effort to date, but even it is not completely satisfying because the
empiricalresult is not firmlygroundedin theory.
In spite of this unease, I agree that the rough estimate that Japan
should be importingabout twice as many manufacturesas it does is
plausible. Let me offer two less formalpieces of evidence that seem to
supportthis. First, suppose we compareJapaneseimportsof manufactures with the "nonlocal" manufacturesimports of the United States
and the European Community. In 1984 the United States imported
manufacturesfrom sources other than Canadaequal to 4.8 percent of
GNP; the EC importednon-EC manufacturesequal to 6.5 percent of
GNP; Japan's manufacturesimports were only 2.9 percent of GNP.
Admittedly,Japanis poorerin resourcesthaneitherthe EC or the United
States, which should show up in lower manufacturesimportsas well as
higher exports; but Japanis also smallerthan the other two economic
units, which would lead us to expect a higherimportshare. So it would
not be unreasonableto expect Japan to have a manufacturesimport
share comparable to that of the EC, which would indeed involve a
doublingof those imports.
Second, compareU.S. tradewith Japanand that with Germany.As
Lawrence points out, Germanyand Japanare virtualtwins in their net
tradepatterns.In 1984,however, U.S. manufacturesexportsto Germany
Robert Z. Lawrence
551
were 45 percentof U.S. imports,whereas U.S. manufacturesexports to
Japanwereonly 23percentof U. S. imports.Onceagain,one can suppose
that if the Japanesewere as well behaved as the Germanstheir import
sharewould roughlydouble.
Let me now turn to the problemof political economy. Suppose we
accept thatJapanshouldbe importingtwice as manymanufacturesas it
now does. What we conclude from this depends critically on two
questions, one of which the paperdoes not answer, the otherof which it
answersin a way that I findimplausible.
The questionthe paperfails to answeris whyJapanimportsless than
it should. At the beginningof the papertwo alternativehypotheses are
offered:tacitgovernmentpolicy or the "invisiblehandshake"of Japan's
cartelizedindustrialstructure.It makes a tremendousdifferencewhich
of these you believe is the truth.If governmentpolicy is the villain, then
all a liberal-mindedJapanese governmentneeds to do is reformitself.
Japan'seconomic glasnost mightbe hardto sell to its bureaucrats,but
we might have some hope that forceful leadership could change the
situationquickly.If the problemis insteadrootedin industrialstructure,
then the governmentpresumablywould have to engage in a long-term
programof antitrustand moral suasion to bringdown the barriers.The
papergives us no clue as to which story is the rightone, and thus as to
how patient Japan's trade partnerswill have to be. My readingof the
anecdotal evidence is that industrial structure is unfortunatelyvery
importantin the closure of Japan'smarkets,and thus that even with the
best will Japan's governmentcannot deliver trade liberalizationat all
rapidly.
The 8.96 million yen question, however, is whether opening of
Japanesemarketswould really ease trade frictions. Here I just do not
agree with the paper's premise. What Lawrence does is to compare
Japanwith Germany,which runs a manufacturestrade surplusas large
relative to GNP as Japan's, but with seven times as large an import
share-and which excites little hostility from the United States. However, even a liberalJapanwould not look like Germany-it would still,
on the paper's estimate, export three times as much in the way of
manufacturesas it imports,essentiallybecause Japanis not in the middle
of Europe. Furthermore,I cannot believe that lack of friction with
Germanycan be explained by Germanmanufacturesimports. Recall
that most of the intra-industrytradeof Germanyis within Europe. The
552
Brookings Papers on Economic Activity, 2:1987
United States sells Germanyonly 45 cents' worth of manufacturesfor
every dollar it buys, as compared with Japan's 23 cents. Is it really
plausiblethat the differencein popularperceptioncan be attributedto
the fact that the voice of the worker who loses his job to German
competition is drownedout by the voices of 0.45 exporters, while his
colleague who lost out to Japanesecompetitionhas only 0.23 exporters
to arguewith? I am not convinced.
My own alternative hypothesis is that conflict with Japan is due
primarilyto a differentaspect of its tradeperformance-the rapidgrowth
of Japanese exports and the rising Japanese market share. Unlike the
Germans, the Japanese keep invading new markets that the United
States regards as its private preserves. My guess is that this creates
tension for both rationalreasons, such as the problemof adjustmentto
increased competition, and irrationalones, such as fear and envy. Of
course this is a very pessimisticdiagnosis,since it says thatthe only way
for Japanto have morefriendsis to be less successful. We can only hope
that the paper is right, and that a liberal Japan would be able to run
currentaccount surplusesin peace. I just wouldn'tcount on it.
GeneralDiscussion
A number of participantsquestioned Lawrence's choice of model
with which to analyzetradedifferencesbetween Japanandother OECD
countries. MartinBaily wondered why the Helpman-Krugmanmodel
had been chosen over a model such as Heckscher-Ohlin, in which
comparativeadvantageis critical and specializationoccurs in production. If the Japanesedo not have a taste for productdiversity, then the
latter model is more relevant, and low importsof manufacturedgoods
would be expected because Japan has a comparative advantage in
producingmanufacturedgoods. In lightof this model, WilliamBranson
reasonedthatthe peculiarthingmay be thatJapanimportsany manufactured goods at all. Lawrence conceded that comparative advantage
ideally should be integrated into the model. But with differentiated
goods, Japan nonetheless would have significantimports of manufacturedgoods accordingto the Helpman-Krugmanmodel. HendrikHouthakkerbelieved the model's assumptionof unitaryincome elasticities
did not apply in Japan for U.S. imports and that the model was too
Robert Z. Lawrence
553
restrictivein this and other respects. He also observed that Japanis not
as geographicallyisolated for purposes of trade as its distance from
Europe and the United States would suggest. GeographicallyJapan's
naturaltradingpartnersare the newly industrializedcountries in the
PacificBasin, so one might examine the volume of importsfrom those
countriesfor evidence of importrestraints.
WilliamBrainardelaboratedon PaulKrugman'sremarkin his formal
comment on the paper that the strong conclusions of the HelpmanKrugmanmodelneednot applyonce transportationcosts areintroduced.
In the presence of such costs, the cross-priceelasticities at the product
level wouldbe crucialin determiningimportand export shares. If goods
were close substitutes,even minortransportationcosts would dominate
tradepatterns. He thereforequestionedwhether the regressionresults
presentedby Lawrence,even thoughthey allowfor transportationcosts,
provide a test of the Helpman-Krugmanmodel and whether they can
identifydeparturesfrom normalpatternsof importsand exports in the
presence of transportationcosts. Accepting Lawrence's verdict that
Japanspent relativelyless on importedmanufacturedgoods than other
industrialcountries, Brainardreasoned that the paper had not persuasively demonstratedwhy this was so. In the same vein, Lawrence
Summersemphasizedthat Lawrence's statisticalanalysis did not provide proof of "tradebarriers"in Japan.The low level of importscould
be explainedby differencesin tastes, language,andgeographiclocation.
MarcusNoland questionedwhetherLawrence's results demonstrate
any greaterJapaneseimportrestrictionthanhave previous studies. For
example,the studyby BergstenandClinethatLawrencecites concludes
thatJapandoes not have above-normalprotection.Adjustingthe Bergsten-Clineestimatesto make them applyto total Japanesetradeimplies
that Japan's import restrictiveness is in the range of $7 billion to $15
billion, which is in Lawrence's range. Noland also questioned the
author'srelianceon expansionaryfiscal policy in Japanto help reduce
the tradesurplus.He noted thatestimatesof thateffect are smallrelative
to the magnitudeof Japan's trade surplus, both globally and with the
United States. An analysis by the Bank of Japanconcludedthat over a
two-yearperiod, the government'srecently announcedplan to provide
a $6 trillionyen stimuluspackagewould lower the total Japanesetrade
surplusby $3 billion to $3.5 billion and lower the bilateralsurpluswith
the United States by about $1 billion. Noland's own estimates were
554
Brookings Papers on Economic Activity, 2:1987
somewhathigherin the first case and around$1.3 billion in the second
case. Given such small effects of a fiscal stimulus, Noland concluded
that significantreductionof the trade surpluswould occur only if a high
yen changes the structuralfeatures of the Japanese economy, such as
the distributionsystem, that hampertrade.
Charles Schultze questioned whether an increase in intra-industry
trade would shield Japan from protectionist measures, as Lawrence
hoped. In his view, efforts to obtainprotectionare not organizedalong
broadindustrylines, butratherarisefromcompetitivefearsof producers
representingfairly specific product categories. Greaterintra-industry
trade, as measuredusing SIC codes, would not remove these political
pressuresto stem foreigncompetition.