ROBERT Z. LAWRENCE
Brookings Institution
Towarda BetterUnderstanding
of TradeBalanceTrends:
TheCost-PricePuzzle
The key to the problem of estimatinglong-run [import]elasticity lies, I feel
sure, in the supplyside of the supply-demandpicture.
Even to approximateit will probablyrequirea close survey of the American
economy, with a view to ascertainingthe extent of each industry'svulnerability
to price competition from abroad or, conversely, its ability to raid the foreigner'sshare of our domestic market.-Arnold C. Harberger'
THE INADEQUACIES
of the time series models of import and export
behavior have long been apparent. To justify the omission of the supply
side, the typical Keynesian function has been based upon the assumptions
of imperfect substitution and infinite supply elasticities of importables
both at home and abroad. While a cyclical variable is sometimes introduced to capture nonprice rationing effects such as changes in delivery
times (as well as the particular cyclical composition of demand), the most
common specification makes the demand for imports (exports) a function
of their relative price and an activity variable.
Although such equations have been able to fit historical data with a
Note: I am grateful to the participantsof the Brookings panel for comments on
an early draft. I had helpful conversations with Lawrence B. Krause, James C.
Riedel, and Walter S. Salant. I thank Arthur C. Kupferman and Karen S. Ostrow
for research assistance, Richard B. Thomas and Joseph Tu for computer assistance,
and Michael C. Deppler of the InternationalMonetary Fund for supplyingme with
data.
1. Arnold C. Harberger,"A StructuralApproach to the Problem of Import Demand," American Economic Review, vol. 43 (May 1953, Papers and Proceedings,
1952), pp. 157-58.
0007-2303/7910001-0191$00.25/ 0 ?) Brookings Institution
192
BrookingsPaperson EconomicActivity, 1:1979
reasonabledegree of accuracy,their coefficientshave not been particularly stable.2Yet for the United States,the qualitative implicationsof a
pair of such equationshave proved to be essentiallycorrectfor almost
thirtyyears. In a seminalpaper, Houthakkerand Magee fitted conventionalimportand exportfunctionsto annualdata from the 1951-66 period andestimatedan incomeelasticityfor U.S. importsof 1.5, whichwas
considerablygreaterthanthe estimatedelasticityof 1.0 on foreignincome
in the U.S. exportequation.3These estimatesof importdemandelasticities imply ceterisparibusthat equal rates of growthin the United States
and in the rest of the world will mean a decliningU.S. trade balance.
Assumingtradeis initiallybalanced,U.S. growthwould have to be onethirdless than growthabroadto maintainbalance.4This "HouthakkerMageeeffect"stemsmainlyfromtradein manufacturedgoods. A pair of
equationsfor manufacturedgoods indicates foreign and U.S. long-run
incomeelasticitiesfor U.S. exportsandimportsof manufacturedgoods are
1.3 and3.1, respectively.5Giventhe annualgrowthratesin the U.S. potentialGNP of 3.7 percent(for the 1960-77 period) andin the potentialoutputfor the "restof the world"of 6.0 percent,the coefficientsimplyannual
growthrates of 11.5 percentfor U.S. manufacturedgoods importsbut
only 7.8 percentfor U.S. manufacturedgoodsexports.
As the quotationfromHarbergeraboveindicates,economistshavelong
been concernedabout the purelydemand-sideorientationof these trade
equations.Domesticallyproduced commoditiesare likely to be close
substitutesfor importsof many producergoods, such as chemicalsand
metals,and consumergoods, such as clothingand shoes. In these cases
the coefficienton the income termin an importfunctionwill actuallybe
an excess demandelasticityderivedfrom the home demandand supply
2. For a discussion of the stability of U.S. import and export functions see Peter
Hooper, "The Stability of Income and Price Elasticities in U.S. Trade, 1957-1977,"
International Finance Discussion Paper 119 (Board of Governors of the Federal
Reserve System, June 1978). See also Robert M. Stern, ChristopherF. Baum, and
MarkN. Greene, "Evidenceon StructuralChange in the Demand for Aggregate U.S.
Imports and Exports,"Journal of Political Economy, vol. 87 (February 1979), pp.
179-92.
3. H. S. Houthakker and Stephen P. Magee, "Income and Price Elasticities in
WorldTrade,"Review of Economics and Statistics,vol. 51 (May 1969), pp. 111-25.
4. If growth exceeded this rate, the terms of trade would have to decline to maintain balance.
5. See Robert Z. Lawrence, "An Analysis of the 1977 U.S. Trade Deficit,"BPEA,
1:1978, p. 174.
RobertZ. Lawrence
193
functions.6The coefficienton incomein the conventionalimportfunction
for such undifferentiated
productsis better expressedas a reduced-form
parameterthat picks up the combinedeffects of domestic supply and
demandratherthanas an incomeelasticityof demandin the sensethatit is
used in consumertheory.7And in such cases the relativeprice term is
likelyto be an estimateof the priceelasticitybetweenimportablesand all
other home goods. For standardizedproducts,relative cost schedules,
ratherthan relativeprices, will determinetradeperformance.If the foreign supplyschedulefor importablesshiftsdownwardover time, foreigners will increasetheir sharesof the domesticmarketat any given price.
The conventionalimportspecification,which fails to includecosts in the
equation,will pick up this effectin the incomecoefficient.
In the case in which importsare imperfectsubstitutes,there are also
soundreasonsto suspectthatthe coefficienton the incomevariableis not
a pureincome-demandelasticity.In theory,given preferencesand information,only incomeandrelativepricesaffectdemand.But whena foreign
producerpenetratesa new market,he is likely to invest substantialresourcesin familiarizingthe marketwith his product.It will take time to
establisha service capability,acquirea reputation,and pry consumers
loose from their old familiarhabits.These effectswill not be reflectedin
price,but they shouldshift the demandcurve.It is reasonableto suspect
that the penetrationpatternwill take the form of the familiarlogistic or
S-shapedcurve that characterizesmost adoption processes. The likely
phases are a struggleto obtaina foothold, a periodof rapidgrowth,and
a taperingoff toward a long-runtrend share.8During this penetration
6. If M(Y) = D(Y) - S(Y), where M is the quantity of imports demanded, D is
the home demand for importables, and S is the home supply, this implies that
emy= (D/M)edy - (S/M)e,y, where emyis the import income elasticity; edy, the income elasticityof demand;and e,y, the income elasticityof supply.
7. For a discussion of this distinction and for the derivation of the equation for
the import-income elasticity see Stephen P. Magee, "Prices, Incomes, and Foreign
Trade," in Peter B. Kenen, ed., International Trade and Finance: Frontiers for Re-
search (CambridgeUniversityPress, 1975), pp. 188-92.
8. The entry of foreign automobiles into the United States is a prime example.
Non-Canadianimports had a minuscule share of the U.S. automobile market in the
1950s. The penetration by Volkswagen made a small impression, and by 1964 the
share of foreign automobiles in U.S. purchases was 6.5 percent. In the late 1960s
there was a period of rapid penetration,however, and by 1970 the foreign share had
shot up to 16 percent.Yet in 1975, it was only 16.4 percent and in 1977, 17.2 percent.
See United States International Trade Commission, Automotive Trade Statistics,
1964-77, series B: PassengerAutomobiles, Publication 913 (USITC, 1978).
194
BrookingsPaperson Economic Activity, 1:1979
process,foreignproducersmaybe constrainedby capacityandmayprefer
to rationin orderto build goodwill ratherthan chargewhat the market
will bear. Once these producersgain a foothold in the market,they may
well raise prices. Duringa period of rapidpenetration,the econometric
estimatesof the priceelasticityof demandareunlikelyto detectthese developmentsand will thereforenot indicatetruelong-runcoefficients.If a
strongtrendin importpenetrationoccurs, the coefficienton the income
term,whichitselfhas an upwardtrend,is likelyto be biasedupward.
The conventionaltradeequationswould also misleadin the case when
the entryof foreignproducersinto the domesticmarketdrivesthe prices
of domesticsubstitutesbelow the domesticfirms'long-runaveragecosts.
Short-rundomesticprices might decline, but as firmswere drivenfrom
the markettheir shares would be taken over by new foreign entrants.
Againthe incomecoefficientmightpickup this effect.
The neglectof the supplyside could well lead to erroneousinferences.
As the experiencesof Japanand Germanyillustrate(and the puretheory
of internationaltraderemindsus), growththatis biasedtowardan expansion of exportablesat given termsof tradecould well lead to a tradesurplus ratherthan a deficit. Only by makingthe sectoral compositionof
growthendogenouswill it be possibleto separatesupplyfromdemand.
The foregoingsuggeststhat tradeequationsshouldbe specifiedto account explicitlyfor the supplyside. An ideal model would use the determinantsof costs in explainingtradebehavior.By takingaccountof longrun costs, it would be possible to distinguishthe price declines that
representimprovementsin productivityand outward shifts in supply
schedulesfrom those that representreducedprofitabilityresultingfrom
inwardshiftsin demand.
As a preliminarystep towardundertakingsuch a model, the appropriatelevel of disaggregationmustbe determined.Commonpracticehas
explainedthe movementin manufacturedgoods prices in international
trade by using data for unit costs in the manufacturingsector. In this
paper, I present evidence that brings this practiceinto question. Inferences about relative trade performanceand profitabilitybased upon
manufacturingsectordata are likely to be misleading.In certainforeign
countries,costs in export industrieshave risen considerablyless rapidly
than costs in their manufacturingsectors in general.As a result of this
"dualism,"overall manufacturingcosts provide a misleadingindicator
of true export costs. In other countries,most notablythe United States,
Robert Z. Lawrence
195
there is little evidence of a similar dualism;costs in export industries
seem to have run roughlyparallelto those in manufacturingas a whole.
Fromthisinvestigation,I am ableto drawan importantconclusion.If past
trendscontinue,costs in U.S. manufacturingmust rise less rapidlythan
those in manufacturingabroadfor U.S. productsto remaincompetitivein
worldmarkets.
The Cost-PricePuzzle
Table 1 providestime series on indexes of unit labor costs in manufacturingin the majorindustrialcountries,expressedin U.S. dollars.The
dataaresurprising,particularlybecausethe UnitedStateshasfaredpoorly
in manufacturedgoods trade,while Japan,Germany,and Italyhave done
well. Duringthe 1960-77 period,unit labor costs in U.S. manufacturing
rose at an averageannualrate of 3.3 percent,while those in Japan,Germany,and Italy rose at ratesof 8.4, 8.2, and 6.8 percent,respectively.In
recent periods, this disparityis even greater.From 1970 to 1977, unit
labor costs rose 5.3 percentannuallyin the United Statesand 16.4, 12.8,
and 10.1 percentin Japan,Germany,and Italy, respectively.
These disparitiesin relativeunit costs could simply indicatethat the
absolutecosts (and prices) abroadwere initiallymuch lower than those
in the UnitedStates.But if this explanationwerevalid, one would expect
to see relativeexportprices changingin a similarfashion. Table 2 indicatesthattheyhavenot.
U.S. exportpricescloselyparalleledboth standardunitlaborcosts and
total unit costs until the price hike of the Organizationof PetroleumExportingCountriesin 1973 raised the relativeprice of materialsinputs,
and thus the marginbetweenprices and value added in manufacturing.
By contrast,the ratio of exportprices to eitherstandardunit labor costs
or total unit costs fell persistentlythrough1973 in Germany,Italy, and
Japan.For these countries,smallerbut substantialdownwardtrendsare
also evidentin the ratiosof exportpricesto value addedin manufacturing
and in the ratios of exportprices to wholesaleprices for finishedmanufacturedgoods.
To explorethe disparitybetweenexportpricesof manufacturedgoods
and costs in manufacturing,I estimateda numberof exportprice equations that relate export prices to variables commonlyused to explain
*o0
jON
r- ootn
00 CN
Iz
kt
ia
C)
t W? X> noom
o
m
C4
--o m
-
c-
CD
"-.
IC
T
m
,o
0
N
T-
t
oo
C)
t
C1
sotm
N
t nFXoo
Iz
~~ OOOO~~~~
b ? m
o
o X
O4T- eT-
c float
oo
1A:Z
00
OOOOe14C1
t
"D
C-A CD
T-
-
T-
-
C
a-
-
N
t
oo
oo
C-
C1
^
f
'S
f
o
$
N
m
~ ~
0
oN
Nn
t
'
C,
nr-c,c
trooCoo4
z
X~~~~~0
0 ON CN
Oc
Ch
b~~t
o~~~~~~~~~~~
O oo
0
0~~~~~~~~~~~~~~
mt-
T-
mmt
'O
tr
CN
?
n0
n
m
C
o
0
C
o!bno
.5
COnN0
O
.g~~~~~~~ 2c 00
Q~~~~~~ONa,
"o cD
rA
.E~~~~V
>
-~X
0
QF
VI o
'IO
X
o
(
>
t
o
m
ff
t
c
o
|
00
00
t
.
t
oo oo m m C oo ooS
t-
t-
Q~~~~~c
\ c\
N CN Cs
0~~~~~~~~~~~
._r
t-
c2
t-
t-
t-
t-
t-
t-
CN N N CN CN C ?
m~~~~~~
X
N
o00
3 0
on
tf m
.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~(Nt
C
~~~~~~~~N
o
o o, 01,
C)
O-
ol.
01
4
.0
0
ha
O
U
m0
C)
*5~~~~~~~~~~~
C.x
CNO
CN
1 loX
-1
C
C; ;
V C m N o t oo ? n o t o F ? m. (ev^~~~~~~~~~~~~~~C;
~
m~~~~~~~~~~~C,
00 'I
m oo o t W t
V~~~~~~~~~~~~~~~~~~~~C
.E
X
oo >
N
oo oo
o
oo
oo
m
U
> o o o oo W ^ N fi
o oo o o
o o oo o
o
om 0
o x:
C;
H~~~
U
o og
O
O;u. o~
o
S
o1
Ln
Ers~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
@
.XX
o
>
A
D r ~~~~~~~~~~~~
=n
02Do~~~~~~~~~~~~~~
'C,
Wo
a
oo
o
o
o
o
o
O
O
o
o
o
?
?o
O,, Tx> t
Y~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
A
b
o
o
F
W
N
W
!
11
tv
m
~~00
V--
\J bO
00
'
2
cz
2
Q
Q0
00 CN
ON Q\
I
)=Y
00,Q
V-- CQ
O
;
,
)pM
Qi
0
t
ONgu;
(
a)
~~~~~~~~~~~~~~
=
l
C.
0
BrookingsPaperson Economic Activity, 1:1979
198
them. The conventionalstructuralinterpretationgiven to the equation
for export prices, P, is that of a markupmodel. Firms are assumedto
set exportprices by "markingup" their unit labor costs (correctedfor
cyclicalproductivity),SULC,andunitmaterialscosts, UMC;thatis,
P = k(SULC) + k(UMC).
The markup,k, is in turn assumedto dependupon competitors'prices,
PC, and capacityutilizationat home, Gd,and abroad,Ga.9This leads to
an estimationequationin logarithmicform:
log P = ao + a, log SULC +
a2
log UMC
+
a3 log PC
+ a4log Gd+ a5log Ga.
Estimatingthis equationalone impliesthat SULC, UMC,and PC are
exogenous.In fact, theorysuggeststhatthese variablesare moreproperly
consideredas endogenousto a morefully specifiedequationsystem.
An alternativeformulationtakes domesticprices, DP, as exogenous
and assumesthat exportprices are more stronglyinfluencedby competitors'pricesandhavedifferentcyclicalmovements:
log P
=
bo+ bi log DP + b2 log PC + b3 log Ga + b4 log Gd.
In table 3 several equationsfor export price, based on formulations
suchas these,arereportedfor eachmajorindustrialcountry.The aimhere
is not to selectthe best equation,but ratherto look for unexplainedtrends
in export prices. Some of the equationshave estimatedcoefficientsthat
renderthem of dubiousmerit.In particular,the coefficientson the nominal variables sum to more than unity. But together they convey the
strongimpressionthat in Germany,Italy, Japan,and, to a lesser degree,
France,unit values for exportshave shown significantdownwardmoveAdding a time trendto the
mentsrelativeto the explanatoryvariables.10
equationsfor these countriesleads to a significantnegativecoefficienton
9. For a more complete presentation,see Peter B. Clark, "The Effects of Recent
Exchange Rate Changes on the U.S. Trade Balance," in Peter B. Clark, Dennis E.
Logue, and Richard James Sweeney, eds., The Effects of Exchange Rate Adjustments, proceedings of a conference sponsored by OASIA Research, Department of
the Treasury (Government Printing Office, 1977), pp. 201-36.
10. In the case of Japan, the use of the export price index as a dependentvariable
gave results similar to those of the unit value index for manufacturedgoods.
RobertZ. Lawrence
199
the trend term and an improvementin the estimatedcoefficientsof the
equations.On the other hand,for the United Kingdom,Canada,and the
United States,the trend term is generallyinsignificant.
The equationsdo revealsome furtherinterestingfeaturesof pricingin
exportmarkets.The state of the cycle in the rest of the world does not
influenceexportpricing,andonly in the case of Japanis the cycle at home
a significantfactor. (As a result,only for Japando the equationsuse the
ratioof actualto potentialmanufacturingoutputas an independentvariable.) Competitors'prices,ROWPX, play a majorrole in determination
of export prices, although the noticeable reduction in some of the
ROWPX coefficientswhen wholesale and materialsprices are used as
explanatoryvariablessuggeststhat part of the competitiveeffect could
actuallystemfrom a commonrise in the pricesof globalrawmaterials.
Without the dummyvariable for the price control period, the U.S.
equations exhibit large residualsin 1972 and 1973. Because controls
apparentlyheld down exportprices,the influenceof devaluationon U.S.
exportpricesduringthis periodis underestimated.-'This strongeffecton
export prices from controls is furtherevidence that U.S. export prices
closelyreflectthebehaviorof U.S. domesticprices.
EXPLAINING
THE TRENDS
Several explanationsmight be offeredfor the strong negativetrends
in export prices relative to other prices in Japan, Germany,Italy, and
France. One explanationmight be measurementerror.The weaknesses
of unit values as a pricemeasurearewell known.Yet the serieson export
prices, which indicatestrongcompetitiveimprovementsfor these countries (or smaller competitivedeclines), are more consistentwith their
actualtradeperformancethanare the cost data,whichsuggestlargecompetitive losses. The weaknesscould lie in the data on unit labor costs.
Althoughthe U.S. Bureauof LaborStatisticsstrivesto ensurethat these
data are comparableacross countries,it cautions: "publishedaverage
11. In the state-o.f-the-artpaper on the effects of the exchange rate changes in
1971-73 on the U.S. trade balance, Clark obtains a statistically significant negative
effect of dollar devaluation on U.S. dollar export prices, yet he dismisses it as a
statistical fluke. See his "The Effects of Recent Exchange Rate Changes,"p. 213, note
17.
~
i w S x XFNNoNX
W
F~~C ei
V
b tS NNNG
~~C4 t
c) c;
; C; Cs C4
; C; CS
tn-
*N
o
.4c
O
*
42
W
C4 oo4
uz~~~~~~~O
^~~~~~~~C
Bq
.
X
.
~ ~~~~~
4n
~ ~~~~~~
*
*
*
ocn
A
t- *- 1.
C; C;
.
*
c"
C;
*
N
06
W
CS*l
.
*
.
.
6
o
cl
*
t %
C Fn
C tt
*
*
*
eno
00
O
c
C)
O~
l0
O
e14
r Wo
O
0
CM
CM
*
0R
CM
e
'0
0C
oC
0
0
N
*
N
*
N
0CM
e
ON
00
O
?O O
*
*
00
0
N
C
oC0
oo
oCM
oCM
cn
It
oootn
O
00
"4
oCM
5555o
555 0CM
CC
*
'
0>
0R
eC1i
In
'0
oCM ~;o 55 o;
o
O
"
V-
oW)
o0
o '-4
o
*
00
*
*
*
*
*
*
*
C
oC;
'o
*5
O
CM
C;
-~~~~~~~~~~~~~~~~~~~~~C
cl
oN
W
_
~~~~~~**
ON*
*
*
^
CM m
cr-
t
*
\.
00
*
*
*
*
0-~.r~-t
'I
M-~t0e
*
0
*
*
*
O crA CA
N*
N N*
C
cli
C;
n
cl
t
M-
o
N
o
0CM0^ {
.
*
*
*
^
ONO
0
N*N
~C
*
*-*
b0C0
W
oCA
00
OC
00-.
*
^0
45~~~~~~~~c
n
mN
CS
_n
0
W
*
C)00
C-
N
*
O
*
CS
C-
00
~
Oh
~O
*
CM
o4
N
~
*
N0t
^
C1
00
0OF
00
o
*
d
*
C1
tn~~~~~~~~~~
o0~
CM
N
CA
*
O.
*
* * O
en
**
en
cn
en
N*
W
ON
o
o
oo
*
O
C)
rA
0
C;
cd
C;
rA
V.
Cd
rA
4)
S.
to
Cd
rA
S.
0
rl)
cwd
Cd
tn
w
to-
ZI
21L. *qJ
x0 , 11.)
Cd
C14
O
0
0
C;
C;
C)
.4,
0
>
CZ
-
t3
C)
c;
c;
c;
0 "3
o
w 4) U 'O
4) X 4).
0
Cd -
a
Cd
00
v) r1%0
Cd
t
'O
t3
0
t0
cn tn
en
cd
0
o
cn
0 t0
cn
0
0
OD 0
cd
0
04
cd
0
w
0
,
C, -4
0
>
cd
02
Cd O
.00
Cd
0
Cd
z,-0
0
V--4
"
0
cd
o
a U S
Cd
cn
Cd'
o
V
'Z
cl
0
ten
en
cli
a
It 'n "
B
00
WI
V--4 r-
V--4 in
C'4
N
o
O
Itt
11)
-0
0
R -z
0 0
.0
Cd
0
0
cn
w cd C, 0
Cd
zt
1.0
CN
IC
CN
"O
Ch
cn
1.0
ON
"O
v)
0
IL)
rA
ON
- U% .. Cd
Cd rA =
cl Cd
z
0
Cd
O
11U 11
ci
0t. 0-
oc, a
cd
tZ 0 1
W)
WI
Cd
0
0
0
'r- O
0
CZ C's 0
CZ
..
0
Cd cn
C,3
0
0
0
cd 4
'D U
0
0
0
rA x
'j
0
WI
0
p"
w
o t O,o C, 0
0 a x
o
'o
o
cd
o
cd
r-
cn
., -1:4
w I- 0
w
0
C4
cd
0
0
Cq
cd
Cd 0.2
Cd to -o .
oooo
cd O. S., x
.,
. ,
t.
O
cc
cn
C4
o
-o
cq
V
t.
to
0
0
cc 0
alz
w
to
0
.
0
O
0
cwd
Cd
o
>
>Cd -r
t:
0 A
0
0
to
.7 C;
A
0
x
-0
0
Pl.
0 C
C;
O
Cd
ti
I'D
'v
r.
0
t2
0
0
Z
c
0
Cd
cn
cd
00
O
V.
o
00
cd
S. CA
0
Z",
tt
z
0
Q4
Cd
co
c;
0
0
a
0
0
0
4) C,3
0
-o
o to
cd
C,3
0
c;
L) 0
:
en
C14
cd
0w
Cd
U-
0
0
23
11 n
0
U
Z: 0
11
0
44 cd cd "4
0
;Zi
cd
0 m
ci
0 'R:t
U
T$
Ca
RobertZ. Lawrence
203
hourlyearningsmay not includethe sameitems of labor compensationin
each country."112
But becausethe trendspersistwith alternativeequations
when other prices replace costs, it is difficultto dismissthem simply as
statisticalaberrations.13
If the data are accepted,severalalternativehypothesescan be considered. The first is that firmsmanufacturingfor export in these countries
have been preparedto cut theirprofitmarginscontinuously.This explanationi6 credibleas a short-runresponseto competitivepressures,but it
cannot explain trends over a seventeen-yearperiod. The coefficientsin
the Japanese,Italian, and Germanequationsindicatethat exportprices
fell relativeto costs by between 1 and 3 peroenta year. Profitsas a proportionof value addedwould have had to be implausiblylarge initially
to permitsuchdeclines.In addition,the equationssuggestthatthis reduction in exportprices occurredin excess of that requiredby competitive
pressures.Thereis evidencethat Germanand Japaneseexportfirmshave
had to reduce their profit marginsin recent years. But it is difficultto
believe that the export sector in these economies expandedthroughout
the 1960s in the face of persistent,large declinesin profitmargins.14
Governmentexportsubsidiesare anotherpotentialexplanationthat is
implausiblein view of the persistenceof these price trends.In Germany,
for example,from 1961 to 1973 the exportpriceindex for manufactured
goods declined15 percentrelativeto the wholesalepriceindex for manufacturedgoods. A subsidyof 15 percentto manufacturedgoods exports
would amountto 2.6 percentof the GermanGNP or 14.5 percentof governmentexpendituresin 1973.
A more reasonableexplanationis that the cost data are poor reflections of the costs of manufacturing the particular commodities that are
exported.Accordingto this hypothesis,actualunit costs for exportgoods
12. "EstimatedHourly Compensationof ProductionWorkers in Manufacturing:
Twelve Countries, 1975-1977 and Eight Countries, 1978" (Bureau of Labor Statistics, Officeof Productivityand Technology, October 1978).
13. For a complete discussion of the relative strengthsand weakness of available
competitive indicators,see Irving B. Kravis and Robert E. Lipsey, Price Competitiveness in World Trade, Studies in International Economic Relations, 6 (National
Bureauof Economic Research, 1971), pp. 3-194.
14. The estimates of expansion in domestic export supply relative to income obtained by Goldstein and Khan are 5.6 for Germany, 3.25 for Italy, and 2.6 for
Japan.See Morris Goldstein and Mohsin S. Khan, "The Supply and Demand for Exports: A SimultaneousApproach,"Review of Economics and Statistics,vol. 60 (May
1978), pp. 275-86.
204
Brookings Papers on Economic Activity, 1:1979
have risen less rapidlythan those in other industries,and export firms
have thereforebeen able to lower their relativepriceswhile at the same
time expand profitably.Either productivitygrowth in the "exportsectors"of these economieshas been considerablymore rapidthan productivity growthin the rest of manufacturing,or the export industrieshave
made relativelyintensiveuse of a factor of productionthat has experienceda declinein relativeprice.'5
A DisaggregatedAnalysis
In principleit shouldbe possibleto show by a disaggregatedanalysis
whethercostsof exportproductshaverisenless rapidlythanthoseof other
manufacturedproducts. In practice, however, if export products are
widely dispersedacross firms and industriesor if they are qualitatively
differentfrom productssold domestically,it will be difficultto obtain
sufficientlydisaggregateddatato serve as a usefulproxyfor export costs.
Whereexport productsare concentratedin a few industriesand where
those productcosts are similarto those sold domesticallyfrom the same
industry,disaggregationby industrycan be used to estimateexportcosts.
In this sectionI reportthe resultsof disaggregationcarriedout underthe
assumptionthat export productsand domesticproductsfrom the same
15. Internationaleconomists have often stressedthe "dualisticnature"of modern
economies, although in doing so they have usually emphasized the distinction between traded and nontraded goods. Balassa developed this distinction in his reappraisalof the theory of purchasingpower parity. See Bela Balassa, "The PurchasingPower Parity Doctrine: A Reappraisal,"Journal of Political Economy, vol. 72
(December 1964), pp. 584-96. The Scandinavian theory of inflation relies upon a
similar dualism in its account of the internationaltransmissionof inflation. See, for
example, Odd Aukrust, "Inflationin the Open Economy: A Norwegian Model," in
Lawrence B. Krause and Walter S. Salant, eds., Worldwide Inflation: Theory and
Recent Experience (Brookings Institution, 1977), pp. 107-66. Haberler explained
the marked disparitybetween the Japaneseconsumer price and wholesale and export
prices in terms of the differentproductivitygrowth rates between goods and service
sectors. See Gottfried Haberler, "InternationalAspects of U.S. Inflation,"in Phillip
Cagan and others, A New Look at Inflation: Economic Policy in the Early 1970s
(American Enterprise Institute, 1973), pp. 91-92. See also Ronald I. McKinnon,
Monetary Theory and Controlled Flexibility in the Foreign Exchanges, Essays in
International Finance, 84 (Princeton University Press, International Finance Section, 1971), pp. 21-27. However, in modeling trade behavior, international economists have typically assumed that data for the manufacturingsector were appropriate indicatorsof costs in industriescompeting in exports and imports.
RobertZ. Lawrence
205
industryhave similarcosts, but that costs differacrossindustries.Under
this assumption,the appropriateaggregatecost index for manufactured
exportsis obtainedby a simplereweightingof industrycosts.'6
JAPAN
Table 4 presentsthe resultof reweightingcost datafor individualJapaneseindustriesby theirexportshares.For eachindustry,unitlaborcost,
unit value added, and total unit value indexes were derivedby dividing
the series on labor compensation,net value added, and gross value of
outputby the industrialproductionindex.Theseserieswerethenweighted
by the 1970 shareof each industryin Japaneseexports,and the ratio of
each to the correspondingseries for total Japanesemanufacturingwas
computed.The resultsindicatethatthe costsin Japaneseindustrieswith a
large exportshareincreasedmuch more slowly than those in total Japanese manufacturing,
reflectingthe dualismthatI hypothesizeas the explanationfor the resultsof tables 1 and2.17
The first three columns of table 4 indicate that duringthe 1963-76
period, when weighted by export shares, unit labor costs, unit value
added, and total unit values all declined relative to the corresponding
measuresfor total manufacturingby roughly20 percent.Moreover,as
column4 indicates,in contrastto the strongdownwardmovementin the
ratio of export prices to costs in manufacturingobtained in table 2,
manufacturingunit values weightedby industryexport shares are consistentwith exportunit valuesfor the 1963-71 period.Increasingexport
subsidies,decliningprofitmarginsin exportsales, or measurementerrors
16. The disaggregation used data available in United Nations, Yearbook of
IndustrialStatistics. This source provides time series data on industrial production;
value added in producers'values; value of gross output;and wages, salaries and supplements for three-digitInternational-Standard-Industrial-Classification
(ISIC) manufacturing in several countries. Not all the series are available for each major
industrial country, and for some countries certain series had to be aggregated or
reweighted. These data were matched with export data classified by the StandardInternational-Trade-Classification(SITC). Most industries in the three-digit ISIC
classificationdo correspond in a rough way to two- or three-digit categories in the
SITC data, but the matchingis not perfect.
17. The dualism referredto here is somewhat differentfrom that which has been
widely recognized in Japan. The latter refers to the coexistence of a "large-firmseetor" that has high wages and enjoys access to cheap capital, and a "small-firm
sector," in which wages are lower and capital more expensive.
Brookinigs Papers on Economic Activity, 1:1979
206
Table 4. Ratio of Export-Weighted
to UnweightedUnit LaborCosts, Unit ValueAdded,
and Total Unit Values,and of ExportPrices to Export-WeightedTotal Unit Values
for JapaneseManufacturing,1963-76a
Index, 1970 100
Ratio of export-weighted
to unweighlted
index
Year
Unit
laborcosts
(1)
Uniit
valueadded
(2)
Total
unit values
(3)
Ratio of export
price to exportweightedtotal
unit-valueindex
(4)
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
112
107
105
102
102
100
101
100
98
98
95
92
92
86
111
108
104
103
104
98
99
100
93
96
102
98
84
90
108
105
104
100
100
102
100
100
99
98
97
92
90
90
98
98
97
96
98
98
100
100
102
95
93
104
96
92
Sources: The data for 1970 Japanese exports are taken from United Nations, Yearbookof International
Trade Statistics: 1971 Edition. Unit labor costs for total manufacturing were provided by the Bureau of
Labor Statistics. The deflator for total manufacturing (value added) was supplied by Michael Deppler
of the International Monetary Fund. All other data are from U.N., Yearbook of Industrial Statistics,
vol. 1: GeneralIndustrial Statistics, various issues, tables 4, 9, and 13.
a. The components of the series on unit labor costs were derived by dividing "wages and salaries of
employees" for each industry by the corresponding industrial production index. Similarly, the series on
unit value added and total unit values were derived by dividing "value added" and "gross output,"
respectively, for each industry by the corresponding industrial production indexes. The food products,
tobacco, petroleum refineries, and coal products industries were excluded to make the series conform to
the data on Japanese exports. Each component industry was then weighted by its share in Japanese exports
in 1970. The series presented are the ratio of the export-weighted series to the corresponding (unweighted)
series for total Japanese manufacturing. The export price series is the ratio of the unit value index for
manufactured goods exports to export-weighted total unit values in manufacturing (the ratio of gross
value of output to industrial production).
can be eliminatedas explanationsfor the downwardtrendin exportprices
relativeto totalmanufacturing.Changesin exportproductpricesreflected
the changesin the costs of the industriesin whichtheywere produced.
As mightbe expected,exchangerate modificationsand cyclical developmentsin the 1970s have changedexport prices relativeto costs. The
decline in the ratio of export prices to total unit values (column 4) in
1972 and 1973 followed the appreciationof the yen under the Smithsonianagreementin December1971, while the depreciationof the yen in
RobertZ. Lawrence
207
1974 moved the ratio in the opposite direction.In 1975 and 1976, Japaneseexportersusedforeignmarketsto increasetheircapacityutilization,
andagainpricesdeclinedrelativeto totalcosts.
Reweightingindustrycosts by theirexportsharesassumes that export
and industry-widecost movementsare similar.To explorethe extent to
which exportprices have displayedany trend relativeto their domestic
counterpartswithinparticularindustries,the componentsof the Japanese
exportprice index were matchedwith the correspondingcomponentsof
the Japanesewholesale price index. The estimated annual percentage
trendsin the relativeexportprices over the 1962-77 period are as follows.'8
Industry
Chemicals
Electricalmachinery
Exportprice trend
-1.3
(-2.9)
-0.2
(-1.4)
Generalmachineryand
precisioninstruments
Metalsand relatedproducts
Textiles
Transportequipment
-1.5
(-6.3)
-0.5
(-1.6)
-2.1
(-8.8)
2.8
(10.6)
Statisticallysignificantdownwardtrends are evident in three relative
prices; downwardtrendsthat were not significant,in two; and a significant upwardtrend, in one. At the industrylevel, therefore,there is evidence of a disparityin the trendbehaviorof domesticandforeignprices.
In summary,by appropriatelymatchingthe costs in a particular Japanese industry with exportpricesfor that industry, muchof the puzzlein
the movementof Japaneseexportprices can be explained.However,to
obtaina preciseestimateof the truecosts facingexportfirms,it will probably be necessaryto disaggregatebeyond the industrylevel because, in
18. Here, as elsewhere in this report, the t-statisticsare in parentheses.The data
used in estimatingthe trends are from various issues of the Bank of Japan,Economic
StatisticsAnnual, and the Bank of Japan,Price IndexesAnnual.
Brookings Papers on Economic Activity, 1:1979
208
certainsectors,downwardtrendsremainwhen industryexportpricesare
matchedwith industrywholesaleprices.
OTHER
COUNTRIES
For Germany,Italy,andFrance,the remainingcountriesin whichnegative trendswere evidentin the ratios of exportprices to manufacturing
industrycosts, a reweightingof industryunit costs by exportsharesfailed
to confirmthe resultsreportedfor Japanin table 4. Explainingthe costpricepuzzlein these countrieswill requireconsiderableinvestmentin the
generationof a newdatabase.
A time trend like that found in the aggregateregressionsof table 2
showedup in disaggregatedregressionsexplainingexportprices of nonelectricalmachinery,which accountedfor 23 percentof exportsof German manufacturedgoods in 1976. For this tradecategory,actualexport
price series based upon contractprices are availablefor both Germany
and the United States. It is thereforean ideal candidatefor examining
whethera significantdualismis evident within a particularGermanindustry.Withouta time trend variable,the residualsexhibit strongserial
correlation.Whenthe time trendis added,the serialcorrelationis eliminated,the standarderrorof the equationimproves,and the t-statisticof
all the estimatedcoefficientsalso improves.For the United States, reweightingwas carriedout based both on export sharesand import-competing sharesin differentindustries.Relativeto all manufacturing,costs
and pricesin the exportsectordeclinedslightlybetween 1954 and 1977,
while they rose slightlyin the import-competingsector.But these trends
wereonly on the orderof 0.1 percenta year.
IMPLICATIONS
In summary,this explorationof dualismby calculatingandreweighting
industrycosts has met with mixedresults.In Japan,it is evidentthat the
industrieswith large sharesin Japaneseexportsare also those that have
relativelyslowerincreasesin costs. For the UnitedStates,the reweighting
exerciseindicatedsmall differenceswhen industrycosts are reweighted,
a resultconsistentwith aggregatedata;while in the cases of France,Germany, and Italy, I was unableto find evidenceof significantdifferences,
whichleaves the aggregateresultsunexplained.BecauseI was able to di-
RobertZ. Lawrence
209
vide the manufacturingsector into only twelve "industries,"a more precise disaggregationmightmeet with more success.However,even at the
level of the nonelectricalmachineryindustryin Germany,costs for export
productshave apparentlydeclinedrelativeto costs for outputin general.
These findingsare importantfor model builders.In representingthe
supply side, extreme caution should be used when treating costs in
manufacturingin generalas a proxyfor costs in the "exportsector."Althoughthe use of some variablewith a strongtime trend could provide
an equationwith a good statisticalfit, the coefficientson such a variable
wouldbe biased.Beforethe determinantsof relativecosts in international
tradecan be analyzed,a considerabledegreeof disaggregationis needed.
U.S. Competitiveness
Because foreign manufacturedexports compete with U.S. products
both in the United States and in third-worldmarkets,changes in their
prices relativeto those of U.S. manufacturedgoods are the majordeterAnd becauseof the
minantof U.S. internationalprice competitiveness."9
decliningpriceof exportsrelativeto total manufacturesin some countries
-what I call dualism-and the absenceof such a dualismin the United
States,for U.S. productsto maintaintheirinternationalpricecompetitiveness, the averageprice of its manufacturedgoods has had to rise less
rapidlythanthe averagein othercountries.
The following equationsummarizesthis historicrelationshipfor the
1962-78 period:
PX = 1.03 + 0.97 PWP,
(2.6)
(8.4)
R2 = 0.81; standarderror = 1.5; Durbin-Watson= 2.0,
wherePX is the percentagechangein the ratio of U.S. to foreignexport
unit valuesandPWPthe correspondingratioof U.S. to foreignwholesale
prices of manufacturedgoods.20This equationimpliesthat, if therewere
19. U.S. exports will also compete with domestic manufactured goods in each
foreign market.
20. The denominiatorsin these two ratios are a weighted average taken from
thirteen industrial countries. For a detailed description, see International Monetary
Fund, InternationalFinancial Statistics,vol. 32 (March 1979), p. 416. Data are from
ibid., various issues.
210
BrookingsPaperson Economic Activity, 1:1979
no change in relativeU.S. wholesale prices, relativeU.S. export prices
would have increasedat an annual rate of 1.03 percentagepoints per
year. For relativeU.S. exportpricesto have remainedconstant,the relative prices of manufacturedgoods would have had to decline by 1.06
percentagepointsa year. The relationshipssummarizedby the regression
remainapparentin recentdata.In 1978, for example,U.S. relativeexport
unitvalueswereroughlyat their 1975 levels despitea 6 percentdeclinein
the relativepriceof U.S. manufacturesbetweenthe two years.
Likethe Houthakker-Magee
effectdiscussedat the outsetof this paper,
the equationrelatingrelativeexportprices and relativewholesaleprices
shouldbe interpretedas a statisticalsummaryof a historicrelationship.
The historicaldivergencebetweenthese two endogenousvariablesis an
importantfact underlyingthe decliningcompetitivepositionof the United
States. Just as the Houthakker-Mageeeffect predictsthat equal growth
rateshereandabroadwill lead to a decliningU.S. tradebalance,this equation predictsthat equal inflationin manufacturedgoods prices here and
abroadwill alsoleadto a decliningtradebalance.
The dualismbetweenpricesof exportsand all manufacturedgoods in
the rest of the worldexplainswhy indicatorsof pricesand costs in manufacturinghave not providedan accuratepictureof U.S. pricecompetitiveness. This reporthas drawn attentionto this disparitybetween costs in
exportandotherindustriesin certainforeigneconomies.Furtherresearch
is needed to determinethe contributionof factors such as economiesof
scale, technologicalimitation,embodiedtechnicalchange,and changesin
factorendowmentsas causesof thesedisparities,andto determinewhether
thesedisparitieswill persist.
Discussion
MUCH OF THE DISCUSSIONfocused on
what might explain exceptional
productivitygains in exportindustries.MarinaWhitmansuggestedthat,
even if subsidiesdidnot directlyexplainthe differentialpriceperformance
of exportindustries,they could play a role indirectlyby allowingexport
industriesto reach their technologicalpotentialand to achievesufficient
economiesof scale. JamesDuesenberrydescribedtwo distinctsituations
RobertZ. Lawrence
211
thatmightgive rise to exceptionalproductivitygrowthin some industries.
In the first, countriesmay have developedindustriesor productsthat are
growingrapidlyin worldmarkets.The interestingquestionin this case is
whatcharacteristicsof a countrypermitit to gainleadershipin those new
industries.In the second, a countrymay have a latent comparativeadvantagein an establishedproductthat it utilizesincreasinglyover time to
expandits productionandworldmarketshare.Both situationsinvolvean
interactionof scale effects, demand growth, and technicalchange, and
may well include subsidiesand protection,at least initially.He doubted
that one could understandthese developmentsby estimatingmodelswith
fixed coefficientsand technologies.William Nordhaus added that the
differentialproductivitygrowthamong a country'sindustrieswould determineits exports.The differentialscould be inherentin the country's
basic human and physical resources.The wider these differentials,the
more apparentwould be the dualitythat Lawrenceidentified.
Robin Marrishypothesizedthat similartechnologicaldynamicsapply
both to the successfulexportperformanceof other countriesand to the
failureof U.S. exportsto expand.This dynamismproducesan S-shaped
growthcurve;althoughtechnologicalcatch-upis one elementof the steep
portion of this curve, other qualitativecharacteristicsof economies are
involved that are difficultto model. The United Statesis past this steep
portion,while countriesexportingto it are not. Some developingcountries may now be in this phase, supplantingEuropeaneconomieswhose
exportsexpandedso much duringthe past two decades.
RudigerDornbuschemphasizedthe importancefor policyformulation
of sortingout the reasonsbehindthe erosionin U.S. competitiveness.The
conventionalimport demandequationimplies we need to slow growth
and control domesticinflationin order to improvethe currentaccount.
By contrast,if the differencesin productivitygrowtharedue to economies
of scale or to the openingof new marketsfor new products,channeling
resourcesinto expansionand researchand developmentmight be a superiorpolicyto slowinggrowthor controllingwages.
PeterKenennoted that the dualismhypothesiswas most evidentin the
data for Japan andwas more difficultto identifyfor othercountries.This
led him to hypothesizean explanationrooted in the institutionsof the
Japaneseeconomy. There, the home marketis so thoroughlysheltered
from competitionthat it providesa high marginof profitsfor Japanese
producers;this, in turn, provides an implicit subsidy for exports and
212
Brookings Papers on Economic Activity, 1:1979
permitsJapaneseproducersto operatewith low or variableprofitsat the
marginin their exportbusiness.The dualismhypothesismay thus apply
mainlyto Japanand may reflectthe peculiarcharacteristicsof the Japanese home marketmore than it does other differencesbetweenthe U.S.
economyand its tradingpartners.Lawrencerespondedthat, althoughhe
had been unable to provide disaggregatedevidencefor other countries,
the strengthand significanceof the downwardtrendsin theirexportprices
relativeto their prices for all manufacturescould not be ignored.Also
the dualismeffect was apparentin the equationat the end of the paper,
which used data from thirteen industrialcountries; this showed that
dualismwasnot a uniquelyJapanesephenomenon.