Vous êtes sur la page 1sur 32

The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity

Author(s): Marc J. Melitz

Source: Econometrica, Vol. 71, No. 6 (Nov., 2003), pp. 1695-1725
Published by: The Econometric Society
Stable URL: http://www.jstor.org/stable/1555536
Accessed: 07/12/2008 08:03

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

The Econometric Society is collaborating with JSTOR to digitize, preserve and extend access to Econometrica.

Econometrica, Vol. 71, No. 6 (November, 2003), 1695-1725



This paper develops a dynamicindustrymodel with heterogeneous firmsto analyze
the intra-industryeffects of internationaltrade. The model shows how the exposureto
tradewill induceonly the more productivefirmsto enter the exportmarket(while some
less productivefirmscontinue to produceonly for the domesticmarket)and will simul-
taneously force the least productivefirmsto exit. It then shows how furtherincreases
in the industry'sexposure to trade lead to additionalinter-firmreallocationstowards
more productivefirms.The paper also shows how the aggregateindustryproductivity
growthgeneratedby the reallocationscontributesto a welfare gain, thus highlightinga
benefit from trade that has not been examinedtheoreticallybefore. The paper adapts
Hopenhayn's(1992a) dynamicindustrymodel to monopolisticcompetitionin a general
equilibriumsetting. In so doing, the paper providesan extension of Krugman's(1980)
trade model that incorporatesfirm level productivitydifferences.Firmswith different
productivitylevels coexist in an industrybecause each firmfaces initialuncertaintycon-
cerningits productivitybefore makingan irreversibleinvestmentto enter the industry.
Entryinto the exportmarketis also costly,but the firm'sdecisionto exportoccursafter
it gains knowledgeof its productivity.

KEYWORDS: Intra-industrytrade, firmheterogeneity,firmdynamics,selection.

RECENT EMPIRICALRESEARCHusing longitudinalplant or firm-leveldata from
several countries has overwhelminglysubstantiatedthe existence of large and
persistentproductivitydifferences among establishmentsin the same narrowly
defined industries. Some of these studies have further shown that these pro-
ductivity differences are strongly correlated with the establishment'sexport
status: relativelymore productiveestablishmentsare much more likely to ex-
port (even within so-called "export sectors," a substantial portion of estab-
lishments do not export). Other studies have highlighted the large levels of
resource reallocations that occur across establishmentsin the same industry.
Some of these studies have also correlatedthese reallocationswith the estab-
lishments'export status.
This paper develops a dynamicindustrymodel with heterogeneous firmsto
analyze the role of internationaltrade as a catalystfor these inter-firmreallo-
cations within an industry.The model is able to reproduce many of the most
salient patternsemphasizedby recent micro-levelstudies related to trade. The
model showshow the exposureto trade induces only the more productivefirms
to exportwhile simultaneouslyforcing the least productivefirmsto exit. Both
the exit of the least productivefirmsand the additionalexport sales gained by

'Manythanksto Alan Deardorff,Jim Levinsohn,and ElhananHelpmanfor helpfulcomments

and discussions.This manuscripthas also benefited from commentsby the editor and two anony-
mous referees. Fundingfrom the Alfred P.Sloan Foundationis gratefullyacknowledged.


the more productivefirmsreallocate marketshares towardsthe more produc-

tive firmsand contributeto an aggregateproductivityincrease. Profitsare also
reallocated towardsmore productivefirms.The model is also consistent with
the widely reported stories in the business press describinghow the exposure
to trade enhances the growthopportunitiesof some firmswhile simultaneously
contributingto the downfall or "downsizing"of other firms in the same in-
dustry;similarly,protection from trade is often reported to shelter inefficient
firms. Rigorous empiricalwork has recently corroboratedthis anecdotal evi-
dence. Bernard and Jensen (1999a) (for the U.S.), Aw, Chung, and Roberts
(2000) (for Taiwan), and Clerides, Lack, and Tybout (1998) (for Colombia,
Mexico, and Morocco) all find evidence that more productivefirmsself-select
into export markets.Aw, Chung, and Roberts (2000) also find evidence sug-
gesting that exposureto trade forces the least productivefirmsto exit. Pavcnik
(2002) directlylooks at the contributionof market share reallocationsto sec-
toral productivitygrowthfollowingtrade liberalizationin Chile. She finds that
these reallocationssignificantlycontributeto productivitygrowth in the trad-
able sectors. In a related study, Bernardand Jensen (1999b) find that within-
sector market share reallocations towards more productive exporting plants
accountsfor 20% of U.S. manufacturingproductivitygrowth.
Clearly,these empiricalpatterns cannot be motivatedwithout appealingto
a model of trade incorporatingfirm heterogeneity.Towardsthis goal, this pa-
per embeds firm productivityheterogeneitywithin Krugman'smodel of trade
under monopolistic competition and increasing returns. The current model
drawsheavily from Hopenhayn's(1992a, 1992b)work to explain the endoge-
nous selection of heterogeneous firms in an industry.Hopenhayn derives the
equilibriumdistributionof firm productivityfrom the profit maximizingdeci-
sions of initially identical firms who are uncertain of their initial and future
productivity.2This paper adapts his model to a monopolisticallycompetitive
industry(Hopenhayn only considers competitive firms) in a general equilib-
rium setting.3A contributionof this paper is to providesuch a general equilib-
riummodel incorporatingfirmheterogeneitythat yet remainshighlytractable.
This is achieved by integratingfirm heterogeneity in a way such that the rel-
evance of the distribution of productivitylevels for aggregate outcomes is
completely summarizedby a single "sufficient"statistic-an averagefirmpro-
ductivitylevel. Once this productivityaverage is determined, the model with

'One of the most robustempiricalpatternsemergingfrom recent industrystudies is that new

entrantshave lower averageproductivityand higher exit rates than older incumbents.This sug-
gests that uncertaintyconcerningproductivityis an importantfeature explainingthe behaviorof
prospectiveand new entrants.
'Montagna(1995) also adaptsHopenhayn'smodel to a monopolisticcompetitionenvironment
(in a partialequilibriumsetting),but confines the analysisto a static equilibriumwith no entryor
exit and furtherconstrainsthe distributionof firmproductivitylevels to be uniform.Although it
is assumedthat only the more productivefirmsearningpositive profitsremainin the industryin
futureperiods,the presentvalue of these profitflows does not enter into the firms'entrydecision.

productivityheterogeneityyields identicalaggregateoutcomes as one with rep-

resentativefirmsthat all share the same averageproductivitylevel.
This simplicitydoes not come without some concessions. The analysisrelies
on the Dixit and Stiglitz (1977) model of monopolistic competition.Although
this modeling approachis quite common in the trade literature,it also exhibits
some well-knownlimitations.In particular,the firms'markupsare exogenously
fixed by the symmetric elasticity of substitution between varieties. Another
concession is the simplificationof the firm productivitydynamicsmodeled by
Hopenhayn(1992a). Nevertheless, the currentmodel preservesthe initial firm
uncertaintyover productivityand the forwardlooking entry decision of firms
facing sunk entry costs and expected future probabilitiesof exit. As in Hopen-
hayn (1992a), the analysisis restrictedto stationaryequilibria.Firms correctly
anticipate this stable aggregate environmentwhen making all relevant deci-
sions. The analysisthen focuses on the long run effects of trade on the relative
behaviorand performanceof firmswith differentproductivitylevels.
Another recent paper by Bernard, Eaton, Jenson, and Kortum (2000)
(henceforth BEJK) also introduces firm-level heterogeneity into a model of
trade by adapting a Ricardianmodel to firm-specificcomparativeadvantage.
Both papers predict the same basic kinds of trade-inducedreallocations, al-
though the channelsand motivationsbehind these reallocationsvary.In BEJK,
firms compete to produce the same variety-including competition between
domestic and foreign producersof the same variety.This delivers an endoge-
nous distributionof markups,a feature that is missingin this paper. BEJK also
show how their model can be calibratedto provide a good fit to a combination
of micro and macro US data patterns. Comparativestatics are then obtained
by simulatingthis fitted model. The BEJK model assumes that the total num-
ber of world varieties produced and consumed remainsexogenouslyfixed and
relies on a specificparameterizationof the distributionof productivitylevels.
In contrast, the current paper allows the total range of varieties produced
to varywith the exposureto trade, and endogenouslydeterminesthe subset of
those varieties that are consumed in a given country.Despite leaving the dis-
tributionof firm productivitylevels unrestricted,the model remains tractable
enough to performanalyticalcomparisonsof steadystates that reflect different
levels of exposure to trade. Although the current model only considers sym-
metric countries, it can easily be extended to asymmetriccountries by relying
on an exogenouslyfixedrelativewage between countries.4In this model, differ-
ences in countrysize-holding the relativewage fixed-only affect the relative
number of firms, and not their productivitydistribution.This straightforward
extension is therefore omitted for expositionalsimplicity.

4Thisassumptionis also made in BEJK.See Helpman,Melitz, and Yeaple (2002) for an exten-
sion of the currentmodel that explicitlyconsidersthe asymmetriccountrycase when the relative
wage is determinedvia trade in a homogeneous good sector.

One last, but important, innovation in the current paper is to introduce

the dynamic forward-lookingentry decision of firms facing sunk market en-
try costs. Firmsface such costs, not just for their domestic market,but also for
any potential export market.5These costs are in additionto the per-unittrade
costs that are typicallymodeled. Both surveyand econometricworkshave doc-
umented the importanceof such export market entry costs. Das, Roberts, and
Tybout(2001) econometricallyestimate these costs average over U.S. $1 Mil-
lion for Colombianplants producingindustrialchemicals.As will be detailed
later, surveysreveal that managersmaking export related decisions are much
more concernedwith exportcosts that are fixed in natureratherthan high per-
unit costs. Furthermore,Roberts and Tybout(1977a) (for Colombia),Bernard
and Jensen (2001) (for the U.S.), and Bernard and Wagner(2001) (for Ger-
many) estimate that the magnitudeof sunk exportmarketentrycosts is impor-
tant enough to generate very large hysteresiseffects associatedwith a plant's
export marketparticipation.6


2.1. Demand
The preferences of a representativeconsumer are given by a C.E.S. utility
function over a continuumof goods indexed by co:

u -[fq(w)Pciw - l/p
u=A Pd&)

where the measure of the set Q2representsthe mass of availablegoods. These

goods are substitutes,implying0 < p < 1 and an elasticity of substitutionbe-
tween any two goods of o- = 1/(1 - p) > 1. As was originally shown by Dixit
and Stiglitz (1977), consumerbehaviorcan be modeled by consideringthe set
of varieties consumed as an aggregategood Q _ U associatedwith an aggre-
gate price

_F_ 1-0,
(1) P= / p(w))1-cdw)

I Sunkmarketentrycosts also explainthe presence of simultaneousentryand exit in the steady

state equilibrium.
6Sunk export market entry costs also explain the higher survivalprobabilitiesof exporting
firms-even after controllingfor their higher measured productivity.See Bernardand Jensen
(1999a, 2002) for evidence on U.S. firms.

These aggregates can then be used to derive the optimal consumption and
expendituredecisions for individualvarieties using

q(w) Q[ p(Wi)]
r(w) - R[P )]1

where R = PQ = f; r(w)dw denotes aggregateexpenditure.

2.2. Production
There is a continuum of firms, each choosing to produce a different vari-
ety w. Productionrequiresonly one factor,labor,whichis inelasticallysupplied
at its aggregatelevel L, an index of the economy'ssize. Firmtechnologyis rep-
resented by a cost function that exhibits constant marginalcost with a fixed
overhead cost. Labor used is thus a linear function of output q: 1 = f + q/q'.
All firmsshare the same fixed cost f > 0 but have differentproductivitylevels
indexed by (p> 0. For expositional simplicity,higher productivityis modeled
as producinga symmetricvariety at lower marginalcost. Higher productivity
may also be thoughtof as producinga higher qualityvarietyat equal cost.7Re-
gardless of its productivity,each firm faces a residualdemand curvewith con-
stant elasticity o- and thus chooses the same profit maximizingmarkupequal
to o-/(o- - 1) = 1/p. This yields a pricing rule

(3) wP)=
where w is the common wage rate hereafter normalizedto one. Firm profit is

,g((p) = r((p) - 1(p) = _r((p)

-(p f~

where r(p) is firmrevenue and r(p)/o- is variableprofit.r(p), and hence Vr((),

also depend on the aggregateprice and revenue as shown in (2):

(4) r(SP)=R(PP)U-l

(5) RT(SD) (PPS

7Giventhe form of productdifferentiation,the modeling of either type of productivitydiffer-

ence is isomorphic.

On the other hand, the ratios of any two firms'outputs and revenues only de-
pend on the ratio of their productivitylevels:

(6)() q(P2) )

In summary,a more productivefirm (higher p) will be bigger (larger output

and revenues), charge a lower price, and earn higher profits than a less pro-

2.3. Aggregation
An equilibriumwill be characterizedby a mass M of firms (and hence M
goods) and a distribution i(q') of productivitylevels over a subset of (0, ox).
In such an equilibrium,the aggregateprice P defined in (1) is then given by

p=f p(p)1-'Mt((p)d( p

Using the pricingrule (3), this can be writtenP = M1/(>-,f)p(7),where

(7) = [ f 00


q is a weighted average of the firmproductivitylevels (pand is independent

of the numberof firmsM.8 These weights reflect the relative output shares of
firms with different productivitylevels.9 q also represents aggregate produc-
tivitybecause it completely summarizesthe informationin the distributionof
productivitylevels ,iq') relevantfor all aggregatevariables(see Appendix):

P=M-'T1p((), R=PQ=Mr(p),
Q = Ml/Pq(P), H = M1T(4),

where R = fjor(Gp)Ml((p)d(p and H = j? ir(>p)Ml((p)d(p represent aggre-

gate revenue (or expenditure) and profit. Thus, an industrycomprised of M
firmswith any distributionof productivitylevels bu(p)that yields the same av-
erage productivitylevel q will also induce the same aggregateoutcome as an
industrywith M representativefirmssharingthe same productivitylevel qD=

8Subsequentconditionson the equilibriumbt(sp)must of course ensure that S is finite.

fo=f0x 1[q(p)/q(f
9Using q(p)/q(p) = ('p/p)O' (see (6)), S can be written as ](p)dp
S is therefore the weighted harmonicmean of the ;P'swhere the weights q(p)/q(p) index the
firms'relativeoutput shares.

This variablewill be alternativelyreferred to as aggregateor averageproduc-

tivity. Furthernote that r = R/M and -T= H/M represent both the average
revenue and profit per firm as well as the revenue and profit level of the firm
with averageproductivitylevel SD= S.


There is a large (unbounded)pool of prospectiveentrantsinto the industry.

Prior to entry, firms are identical. To enter, firms must first make an initial
investment,modeled as a fixed entry cost fe > 0 (measured in units of labor),
which is thereaftersunk.Firmsthen drawtheir initialproductivityparameterSD
from a common distributiong(p).10 g(p) has positive supportover (0, oo) and
has a continuouscumulativedistributionG(q().
Upon entrywith a low productivitydraw,a firm may decide to immediately
exit and not produce. If the firm does produce, it then faces a constant (across
productivitylevels) probability8 in every period of a bad shock that would
force it to exit. Although there are some realistic examples of severe shocks
that would constraina firm to exit independentlyof productivity(such as nat-
ural disasters, new regulation, product liability, major changes in consumer
tastes), it is also likely that exit may be caused by a series of bad shocks affect-
ing the firm'sproductivity.This type of firm level process is explicitlymodeled
by Hopenhayn (1992a, 1992b). The simplificationmade in this model entails
that the shape of the equilibriumdistributionof productivity,1q') and the ex-
ante survivalprobabilitiesare exogenouslydeterminedby g(p) and 8. On the
other hand, the range of productivitylevels (for survivingfirms),and hence the
average productivitylevel, are endogenously determined.1"Importantly,this
simplified industrymodel will nevertheless generate one of the most robust
empiricalpatterns highlightedby micro-level studies: new entrants (including
the firmswhose entryis unsuccessful)will have, on average,lower productivity
and a higher probabilityof exit than incumbents.
This paper only considerssteady state equilibriain which the aggregatevari-
ables remain constant over time. Since each firm'sproductivitylevel does not
change over time, its optimal per period profit level (excludingfe) will also re-
main constant. An entering firm with productivitySDwould then immediately
exit if this profit level were negative (and hence never produce), or would pro-
duce and earn wT(p) > 0 in every period until it is hit with the bad shock and is

"0Thiscaptures the fact that firms cannot know their own productivitywith certaintyuntil
they startproducingand selling their good. (Recall that productivitydifferencesmay reflect cost
differencesas well as differencesin consumervaluationsof the good.)
"The increasedtractabilityaffordedby this simplificationpermitsthe detailed analysisof the
impactof trade on this endogenousrangeof productivitylevels and on the distributionof market
shares and profitsacross this range.

forced to exit. Assuming that there is no time discounting,12each firm'svalue

function is given by

v((p) = max{0, (1 - 8)'t P)}| = max{O, IT


where the dependence of 1T(() on R and P from (5) is understood.Thus, * =

inf{p: v(p) > O}identifies the lowest productivitylevel (hereafter referredto
as the cutoff level) of producing firms. Since IT(O) = -f is negative, w((p*)must
be equal to zero. This will be referredto as the zero cutoff profit condition.
Any entering firm drawinga productivitylevel SD<qp*will immediatelyexit
and never produce. Since subsequentfirm exit is assumed to be uncorrelated
with productivity,the exit process will not affect the equilibriumproductivity
distributioni(q(). This distributionmust then be determinedby the initialpro-
ductivitydraw,conditionalon successfulentry.Hence, L(q')is the conditional
distributionof g((p) on [(p*,oc):

(8) 'p)= jG(*) G( (D* if

O otherwise,

and Pi- 1- G(p*) is the ex-anteprobabilityof successfulentry."3This defines

the aggregateproductivitylevel S as a function of the cutoff level 9*:14

(9) S = [1- G(p) f

Slg(S) dS]

The assumption of a finite S imposes certain restrictionson the size of the

upper tail of the distributiong((): the (o- - 1)th uncentered moment of g(p)
must be finite. Equation (8) clearly shows how the shape of the equilibrium
distributionof productivitylevels is tied to the exogenous ex-ante distribution
g(p) while allowingthe range of productivitylevels (indexedby the cutoff Sn*)
to be endogenouslydetermined.Equation(9) then showshow this endogenous
range affects the aggregateproductivitylevel.

12Again, this is assumedfor simplicity.The probabilityof exit 8 introducesan effect similarto

time discounting.Modelingan additionaltime discountfactorwould not qualitativelychangeany
of the results.
"3TheequilibriumdistributionA(tp)can be determinedfrom the distributionof initialproduc-
tivitywith certaintyby applyinga law of large numbersto g(XP).See Hopenhayn(1992a, note 5)
for furtherdetails.
14Thisdependence of p on (p*is understoodwhen it is subsequentlywrittenwithout its argu-

3.1. Zero CutoffProfitCondition

Since the averageproductivitylevel f is completelydeterminedby the cutoff
productivitylevel qp*,the averageprofit and revenue levels are also tied to the
cutoff level (p*(see (6)):

r = r(9)= [ S((p )1
] r((p*), T= [ S?* 1 - -(* )
LP J L j j'

The zero cutoff profit condition, by pinning down the revenue of the cutoff
firm, then implies a relationshipbetween the average profit per firm and the
cutoff productivitylevel:
(10) T(QD*)= r(o*) = of T = fk(p*),
where k(p*) = [-(p*)Ip*10-1 - 1.

3.2. FreeEntryand the Valueof Firms

Since all incumbentfirms-other than the cutoff firm-earn positive profits,
the average profit level X must be positive. In fact, the expectation of future
positive profits is the only reason that firms consider sinking the investment
cost fe required for entry. Let -vrepresent the present value of the average
) t Also Viis the average value of firms,
profit flows: v = =(1- = (1/8)rT.
conditional on successful entry: v-= J7 v(p),1(p)dcp. Furtherdefine ve to be
the net value of entry:

(11) Ve = PinV - fe = 8
T - Tfe

If this value were negative, no firm would want to enter. In any equilibrium
where entry is unrestricted,this value could further not be positive since the
mass of prospectiveentrantsis unbounded.


The free entry (FE) and zero cutoff profit (ZCP) conditions represent two
different relationshipslinking the average profit level X- with the cutoff pro-
ductivitylevel >p*(see (10) and (11)):
T = fk(p*) (ZCP),
(12) 1 fe
1IT= *) (FE).
In (p, ir) space, the FE curve is increasingand is cut by the ZCPcurve only
once from above (see Appendix for proof). This ensures the existence and

(ZeroCutoffProfit) (FreeEntry)

FIGURE 1.-Determination of the equilibriumcutoff (p*and averageprofit I-T.

uniqueness of the equilibriumqp*and 7, which is graphicallyrepresented in

Figure 1.15
In a stationaryequilibrium,the aggregatevariables must also remain con-
stant over time. This requiresa mass Me of new entrantsin every period, such
that the mass of successful entrants, pinMe,exactly replaces the mass 8M of
incumbentswho are hit with the bad shock and exit: pinMe = 8M. The equi-
libriumdistributionof productivity,u(p) is not affected by this simultaneous
entry and exit since the successful entrants and failing incumbents have the
same distributionof productivitylevels. The labor used by these new entrants
for investment purposes must, of course, be reflected in the accounting for
aggregate labor L, and affects the aggregate labor available for production:
L = L p + Le where L p and Le represent,respectively,the aggregatelaborused
for productionand investment(by new entrants).Aggregate paymentsto pro-
ductionworkersLp must match the differencebetween aggregaterevenue and
profit:Lp = R - H (this is also the labor marketclearingconditionfor produc-
tion workers).The market clearingcondition for investmentworkersrequires
Le = Mefe. Using the aggregate stability condition, pinMe = 8M, and the free
entry condition, 7T = 8fe/[l - G(p*)], Le can be written:

Le Mefe =-fe=MiTH.

Thus, aggregaterevenue R = Lp + HI = Lp + Le must also equal the total pay-

ments to labor L and is therefore exogenously fixed by this index of country

15The ZCP curve need not be decreasingeverywhereas representedin the graph. However,
it will monotonicallydecrease from infinityto zero for Sp*E (0, +oo) as shown in the graph if
g(Xp)belongs to one of severalcommonfamiliesof distributions:lognormal,exponential,gamma,
Weibul,or truncationson (0, +oo) of the normal,logistic,extremevalue, or Laplacedistributions.
(A sufficientcondition is that g(Gp)$p/[1
- G(p)] be increasingto infinityon (0, +oo).)

size.16The mass of producingfirmsin any period can then be determinedfrom

the averageprofit level using
(13) M= - =
r o"(ir+f)

This, in turn, determines the equilibrium price index P = M1/(lU-0)p(f) =

Ml/(lU-O)/p, which completes the characterizationof the unique stationary
equilibriumin the closed economy.

4.1. Analysisof the Equilibrium

All the firm-levelvariables-the productivitycutoff (p*and average o, and
the average firm profit Xr and revenue r-are independent of the country
size L. As indicated by (13), the mass of firms increases proportionallywith
countrysize, althoughthe distributionof firmproductivitylevels ,p) remains
unchanged.Welfareper worker,given by

(14) W=P-1 M p1 -

is higher in a larger countrydue only to increased productvariety.This influ-

ence of country size on the determination of aggregatevariables is identical
to that derived by Krugman(1980) with representativefirms. Once < and X
are determined,the aggregateoutcome predictedby this model is identical to
one generated by an economy with representativefirms who share the same
productivitylevel o and profit level ir. On the other hand, this model with het-
erogeneous firmsexplainshow the aggregateproductivitylevel o and the aver-
age firmprofitlevel Xrare endogenouslydetermined and how both can change
in response to various shocks. In particular,a country'sproduction technol-
ogy (referenced by the distributiong(p)) need not change in order to induce
changes in aggregate productivity.In the following sections, I argue that the
exposure of a countryto trade creates preciselythe type of shock that induces
reallocations between firms and generates increases in aggregate productiv-
ity. These results cannot be explainedby representativefirm models where the
aggregateproductivitylevel is exogenouslygiven as the productivitylevel com-
mon to all firms. Changes in aggregateproductivitycan then only result from
changes in firm level technology and not from reallocations.

161tis importantto emphasize that this result is not a direct consequence of aggregationand
marketclearingconditions:it is a propertyof the model's stationaryequilibrium.Aggregate in-
come need not necessarilyequal the paymentsto all workers,since there maybe some investment
income derivedfrom the financingof new entrants.Each new entrantraises the capitalfe, which
providesa randomreturnof iT(<p)(if (p > (p*)or zero (if 'p < 'p*) in everyperiod. In equilibrium,
the aggregatereturnH equals the aggregateinvestmentcost Le in every period-so there is no
net investmentincome (this would not be the case with a positive time discountfactor).


I now examinethe impactof trade in a world (or tradebloc) that is composed

of countries whose economies are of the type that was previouslydescribed.
When there are no additionalcosts associatedwith trade, then trade allowsthe
individualcountriesto replicatethe outcome of the integratedworld economy.
Tradethen provides the same opportunitiesto an open economy as would an
increase in countrysize to a closed economy. As was previouslydiscussed, an
increase in country size has no effect on firm level outcomes. The transition
to trade will thus not affect any of the firm level variables:The same number
of firms in each countryproduce at the same output levels and earn the same
profits as they did in the closed economy. All firms in a given country divide
their sales between domestic and foreign consumers,based on the size of their
countryrelative to the integratedworld economy. Thus, in the absence of any
costs to trade, the existence of firm heterogeneity does not affect the impact
of trade. This impact is identical to the one describedby Krugman(1980) with
representativefirms:Although firmsare not affected by the transitionto trade,
consumersenjoywelfare gains drivenby the increase in productvariety.17
On the other hand, there is mountingevidence that firmswishing to export
not only face per-unit costs (such as transportcosts and tariffs), but also-
critically-face some fixed costs that do not vary with export volume. Inter-
views with managers making export decisions confirm that firms in differen-
tiated product industriesface significantfixed costs associated with the entry
into export markets (see Roberts and Tybout(1977b)): A firm must find and
informforeign buyersabout its product and learn about the foreign market.It
must then research the foreign regulatoryenvironmentand adapt its product
to ensure that it conformsto foreign standards(which include testing, packag-
ing, and labeling requirements).An exportingfirm must also set up new dis-
tributionchannels in the foreign countryand conform to all the shippingrules
specified by the foreign customs agency.Although some of these costs cannot
be avoided, others are often manipulated by governments in order to erect

"7Theirrelevanceof firm heterogeneityfor the impact of trade is not just a consequence of

negligibletrade costs. The assumptionof an exogenouslyfixed elasticityof substitutionbetween
varieties also plays a significantrole in this result. The presence of heterogeneity (even in the
absence of trade costs) plays a significantrole in determiningthe impact of trade once this as-
sumptionis dropped.In a separateappendix(availableupon request to the author),the current
model is modifiedby allowingthe elasticityof substitutionto endogenouslyincreasewith product
variety.This link between trade and the elasticityof substitutionwas studiedby Krugman(1979)
with representativefirms.In the contextof the currentmodel, the appendixshowshow the size of
the economy then affects the aggregateproductivitylevel and the skewnessof marketsharesand
profits across firmswith different productivitylevels. Larger economies have higher aggregate
productivitylevels-even thoughthey have the same firmlevel technologyindex by g(XP).There-
fore, even in the absence of trade costs, trade increases the size of the "world"economy and
induces reallocationsof marketshares and profitstowardsmore productivefirmsand generates
an aggregateproductivitygain.

non-tariff barriers to trade. Regardless of their origin, these costs are most
appropriatelymodeled as independent of the firm'sexportvolume decision.18
When there is uncertaintyconcerningthe exportmarket,the timingand sunk
nature of the costs become quite relevant for the export decision (most of the
previouslymentioned costs must be sunk prior to entry into the export mar-
ket). The strong and robust empirical correlations at the firm level between
export status and productivitysuggest that the export market entry decision
occurs after the firmgains knowledgeof its productivity,and hence that uncer-
tainty concerningthe export markets is not predominantlyabout productivity
(as is the uncertaintyprior to entry into the industry).I therefore assume that
a firm who wishes to export must make an initial fixed investment, but that
this investment decision occurs after the firm's productivityis revealed. For
simplicity, I do not model any additional uncertaintyconcerning the export
markets.The per-unittrade costs are modeled in the standardiceberg formu-
lation, whereby i > 1 units of a good must be shipped in order for 1 unit to
arriveat destination.
Although the size of a country relative to the rest of the world (which con-
stitutes its tradingpartners)is left unrestricted,I do assume that the world (or
trading group) is comprised of some number of identical countries. This as-
sumption is made in order to ensure factor price equalizationacross countries
and hence focus the analysison firm selection effects that are independent of
wage differences.19In this model with trade costs, size differencesacross coun-
tries will induce differencesin equilibriumwage levels. These wage differences
then generate further firm selection effects and aggregateproductivitydiffer-
ences acrosscountries.20I therefore assume that the economy under studycan
trade with n > 1 other countries (the world is then comprised of n + 1 > 2
countries). Firms can export their productsto any country,althoughentry into
each of these export marketsrequiresa fixed investmentcost of fex > 0 (mea-
sured in units of labor). Regardlessof export status, a firmstill incursthe same
overheadproductioncost f.


The symmetryassumptionensures that all countries share the same wage,
which is still normalized to one, and also share the same aggregate vari-

"8Themodelingof a fixed exportcost is not new. Bernardand Jensen (1999a), Clerides,Lack,

and Tybout (1998), Roberts and Tybout (1977a), and Roberts, Sullivan,and Tybout(1995) all
introduce a fixed export cost into the theoretical sections of their work in order to explain the
self-selection of firmsinto the export market.However,these analysesare restrictedto a partial
equilibriumsetting in which the distributionof firmproductivitylevels is fixed.
'9Aswas previouslymentioned,anotherway to abstractfrom endogenousrelativewage move-
ments when countries are asymmetricis to introducea freely traded homogeneous good sector.
See Helpman, Melitz, and Yeaple (2002) for an example incorporatingthis extension.
20inthese asymmetricequilibriawith fixed exportcosts, large countriesenjoyhigher aggregate
productivity,welfare, and wages relativeto smallercountries.

ables. Each firm's pricing rule in its domestic market is given, as before, by
PdAO) = W/pP = l/ptp. Firms who export will set higher prices in the for-
eign markets that reflect the increased marginalcost T of serving these mar-
kets: PjGO) = TIP9? = TPdP). Thus, the revenues earned from domestic sales
and exportsales to any given countryare, respectively,rd(p) = R(Pp(p)f-l and
r.jo) = r`'rd(p), where R and P denote the aggregate expenditure and price
index in every country.The balance of paymentscondition implies that R also
representsthe aggregaterevenue of firmsin any country,and hence aggregate
income. The combined revenue of a firm,r(p), thus depends on its export sta-
rd(O) if the firmdoes not export,
(15) r(p) = rd(p) + nr(>p) = (1 + nrl-rl)rd(p)
if the firmexportsto all countries.

If some firms do not export, then there no longer exists an integratedworld

market for all goods. Even though the symmetryassumptionensures that all
the characteristicsof the goods availablein everycountryare similar,the actual
bundle of goods availablewill be different acrosscountries:consumersin each
country have access to goods (produced by the nonexportingfirms) that are
not availableto consumersin any other country.

6.1. FirmEntry,Exit,and ExportStatus

All the exogenous factors affecting firm entry, exit, and productivitylevels
remain unchanged by trade. Prior to entry, firms face the same ex-ante dis-
tributionof productivitylevels g(p) and probability8 of the bad shock. In a
stationaryequilibrium,any incumbentfirmwith productivity(pearns variable
profitsrx(po)/o- in everyperiod from its exportsales to anygiven country.Since
the export cost is assumed equal across countries, a firm will either export to
all countries in every period or never export.21Given that the export decision
occurs after firms know their productivity(p,and since there is no additional
export market uncertainty,firms are indifferentbetween paying the one time
investment cost fex, or paying the amortized per-period portion of this cost
fx = 8fex in every period (as before, there is no additional time discounting
other than the probabilityof the exit inducingshock 8). This per-periodrepre-
sentation of the export cost is henceforth adopted for notational simplicity.In
the stationaryequilibrium,the aggregatelabor resourcesused in every period

21Therestrictionthat exportcosts are equal across countriescan be relaxed.Some firmsthen

export to some countries but not others-depending on these cost differences. This extension
would also generate an increasingrelationshipbetween a firm'sproductivityand the numberof
its exportdestinations.

to cover the exportcosts do not depend on this choice of representation.22 The

per-period profit flow of any exportingfirm then reflects the per-period fixed
cost fX,which is incurredper export country.
Since no firm will ever export and not also produce for its domestic mar-
ket,23each firm'sprofit can be separated into portions earned from domestic
sales, rTd(p), and export sales per country, i-x (fp),by accountingfor the entire
overhead productioncost in domestic profit:

(16) 7d ( T
rd(_p)-t (<)_ rx(9) f
0J 0J

A firm who produces for its domestic market exports to all n countries if
7rx(wp)> 0. Each firm'scombined profit can then be written: r(<p)=-7Td(P) +
max{0,nrx (>p)}.Similarly to the closed economy case, firm value is given
by v(>p) = max{O, 1T(p)/6}, and (p*= inf>p : v(>p) > 01 identifies the cutoff
productivitylevel for successful entry. Additionally, (p*= inf{>p:(p> (p*and
Wrx(>p)> 01 now represents the cutoff productivitylevel for exporting firms.
If (P*= (0*,then all firms in the industryexport. In this case, the cutoff firm
(with productivitylevel (p*= qp*)earns zero total profit (1T(p*) = lTd(p*) +
frlx(cp*) = 0) and nonnegative export profit (-rx(>p*) > 0). If qp*> p*, then
some firms (with productivitylevels between (p*and qp*)produce exclusively
for their domestic market. These firms do not export as their export profits
would be negative. They earn nonnegative profits exclusivelyfrom their do-
mestic sales. The firmswith productivitylevels above (p*earn positive profits
from both their domestic and export sales. By their definition,the cutoff levels
must then satisfy d(fp*) =0 and rx(p ) = 0.
This partitioningof firmsby exportstatuswill occur if and only if r-lfx > f
the trade costs relativeto the overheadproductioncost mustbe above a thresh-
old level. Note that, when there are no fixed export costs (fx = 0), no level of
variablecost i > 1 can induce this partitioning.However, a large enough fixed
export cost fx > f will induce partitioningeven when there are no variable
trade costs. As the partitioningof firmsby exportstatus (withinsectors) is em-
piricallyubiquitous,I will henceforthassume that the combinationof fixed and
variable trade costs are high enough to generate partitioning,and therefore
that rO-lfx > f. Although the equilibriumwhere all firms export will not be

22inone case, only the new entrantswho exportexpend resourcesto cover the full investment
cost fex. In the other case, all exportingfirms expend resources to cover the smaller amortized
portion of the cost fx = 8fex. In equilibrium,the ratio of new exportersto all exportersis 8 (see
Appendix),so the same aggregatelaborresourcesare expendedin either case.
23Afirmwould earn strictlyhigherprofitsby also producingfor its domestic marketsince the
associatedvariableprofitrd (, ) /vo is alwayspositive and the overheadproductioncost f is already

formallyderived, it exhibits several similarproperties to the equilibriumwith

partitioningthat will be highlighted.24
Once again, the equilibriumdistributionof productivitylevels for incum-
bent firms,bt(p), is determinedby the ex-ante distributionof productivitylev-
els, conditional on successful entry: ,tu(p)= g(p)/l - G(p*)] V8p> (p*.The
ex-ante probabilityof successful entry is still identified by pin = 1- G((p*).
Furthermore,Px = [1 - G(qp*)]/[1 - G(p*)] now representsthe ex-ante prob-
ability that one of these successful firms will export. The ex-post fraction of
firmsthat exportmust then also be representedby Px. Let M denote the equi-
librium mass of incumbent firms in any country. Mx = pxM then represents
the mass of exportingfirmswhile M, = M + nMx representsthe total mass of
varieties availableto consumersin any country(or alternatively,the total mass
of firmscompeting in any country).

6.2. Aggregation
= (p*) and
Using the same weighted average function defined in (9), let Op
Px = P( p*)denote the averageproductivitylevels of, respectively,all firmsand
exportingfirmsonly. The averageproductivityacross all firms, (p,is based only
on domestic market share differences between firms (as reflected by differ-
ences in the firms'productivitylevels). If some firms do not export, then this
average will not reflect the additional export shares of the more productive
firms. Furthermore,neither (pnor s, reflect the proportion r of output units
that are "lost"in export transit. Let 0, be the weighted productivityaverage
that reflects the combined market share of all firms and the output shrinkage
linked to exporting.Again, using the weighted average function (9), this com-
bined averageproductivitycan be written:

1 - ]r-l
ft= f [MO l + nMx(Tls1 )x-]}

By symmetry, t is also the weighted averageproductivityof all firms (domes-

tic and foreign) competing in a single country (where the productivityof ex-
portersis adjustedby the trade cost r). As was the case in the closed economy,
this productivityaverage plays an importantrole as it once again completely
summarizesthe effects of the distributionof productivitylevels ,L(p) on the
aggregate outcome. Thus, the aggregate price index P, expenditure level R,

24Evenwhen there is no partitioningof firmsby export status,the opening of the economy to

tradewill still induce reallocationsand distributionalchangesamong the heterogeneousfirms-
so long as the fixed export costs are positive. In the absence of such costs (given any level of
per-unit costs r), opening to trade will not induce any distributionalchanges among firms,and
heterogeneitywill not play an importantrole.

and welfare per worker W in any countrycan then be written as functions of

only the productivityaverage O, and the numberof varieties consumed

P= Mtl p('t) = Mt -- R = Mtrd(Ot),

(17) POpt
W = -MO-- pOt.
By construction,the productivityaverages , and Ox.can also be used to ex-
press the average profit and revenue levels across different groups of firms:
rd(p) and lTd(cP) representthe average revenue and profit earned by domestic
firms from sales in their own country.Similarly,r,(O,) and ir-(G) represent
the average export revenue and profit (to any given country)across all domes-
tic firmswho export. The overall average-across all domestic firms-of com-
bined revenue, i, and profit, 7r (earned from both domestic and export sales),
are then given by
(18) T = rd(4) + Pxnrx(OPx), Td(P) +

6.3. EquilibriumConditions
As in the closed economy equilibrium,the zero cutoff profit condition will
imply a relationshipbetween the averageprofitper firm - and the cutoff pro-
ductivitylevel sp*(see (10)):
lTd(cP) =0 rld(cP) = fk(fp
fO v(9 x) = xk
where k()p) = [ p 0-1_ 1 as was previouslydefined.The zero cutoff profit
condition also implies that (P*can be written as a function of (p*:

r r((px)
(19 (9x ) fx fX

(19) -p =7 IT

iV=lf T
rd(p*) 0f f

Using (18), Trcan therefore be expressed as a function of the cutoff level p*:
(20) 7r = iTd(p) + pxn7TxG(px)

+ pxnfxk(>p*)
=f k(Gp*) (ZCP),
where p*, and hence Px, are implicitlydefined as functions of p0using (19).
Equation (20) thus identifies the new zero cutoff profit condition for the open

25Inother words, the aggregateequilibriumin any countryis identical to one with M, repre-
sentativefirmsthat all share the same productivitylevel f .

As before, v1= Eto(l - 6)ti7. = 7r/6 representsthe present value of the av-
erage profit flows and ve = pinv - fe yields the net value of entry. The free entry
condition thus remains unchanged: ve = 0 if and only if 7r = 6felpin. Regard-
less of profit differences across firms (based on export status), the expected
value of future profits,in equilibrium,must equal the fixed investmentcost.

6.4. Determinationof theEquilibrium

As in the closed economy case, the free entryconditionand the new zero cut-
off profitconditionidentifya unique (p*and 7r:the new ZCP curvestill cuts the
FE curve only once from above (see Appendixfor proof). The equilibrium *,
in turn, determines the export productivitycutoff fx as well as the average
productivitylevels , fx, t, and the ex-ante successfulentry and exportprob-
abilities pin and Px. As was the case in the closed economy equilibrium,the
free entry condition and the aggregate stability condition, pinMe = 6M, ensure
that the aggregatepaymentto the investmentworkersLe equals the aggregate
profitlevel H. Thus, aggregaterevenueR remainsexogenouslyfixedby the size
of the labor force: R = L. Once again, the averagefirm revenue is determined
by the ZCP and FE conditions: r = rd(() + pxnrx((x) = cr(ir + f + pxnfx).

This pins down the equilibriummass of incumbentfirms,

(21) =R
r L

In turn, this determines the mass of variety available in every country, Mt =

(1 + npx)M, and their price index P = Mtl/(1-U)/ppt (see (17)). Almost all of
these equilibriumconditionsalso applyto the case where all firmsexport.The
only difference is that =- * (and hence Px = 1) and (19) no longer holds.


The result that the modeling of fixed export costs explains the partitioning
of firms,by export status and productivitylevel is not exactlyearth-shattering.
This can be explained quite easily within a simple partial equilibriummodel
with a fixed distributionof firm productivitylevels. On the other hand, such a
model would be ill-suited to address several importantquestions concerning
the impactof trade in the presence of export market entry costs and firm het-
erogeneity:What happensto the range of firmproductivitylevels? Do all firms
benefit from trade or does the impact depend on a firm'sproductivity?How is
aggregateproductivityand welfare affected?The currentmodel is muchbetter
suited to addressthese questions,whichare answeredin the followingsections.
The currentsection analyzesthe effects of trade by contrastingthe closed and
open economy equilibria.The following section then studies the impact of in-
crementaltrade liberalization,once the economy is open. All of the following

analysesrely on comparisonsof steady state equilibriaand should therefore be

interpretedas capturingthe long run consequences of trade.
Let <,*and iPadenote the cutoff and average productivitylevels in autarky.
I use the notation of the previous section for all variables and functions per-
taining to the new open economy equilibrium.As was previouslymentioned,
the FE condition is identical in both the closed and open economy. Inspection
of the new ZCP condition in the open economy (20) relative to the one in the
closed economy (12) immediatelyreveals that the ZCP curve shifts up: the ex-
posure to trade induces an increase in the cutoff productivitylevel (p* > f0*)
and in the average profit per firm. The least productivefirmswith productiv-
ity levels between (Paand q* can no longer earn positive profits in the new
trade equilibriumand therefore exit. Another selection process also occurs
since only the firmswith productivitylevels above p*enter the exportmarkets.
This export market selection effect and the domestic market selection effect
(of firmsout of the industry)both reallocate marketshares towardsmore effi-
cient firmsand contributeto an aggregateproductivitygain.26
Inspection of the equations for the equilibrium number of firms ((13)
and (21)) reveals that M < Ma where Ma represents the number of firms in
autarky.27 Although the numberof firms in a countrydecreases after the tran-
sition to trade,consumersin the countrystill typicallyenjoygreaterproductva-
riety (M, = (1 + npx)M > Ma). That is, the decrease in the numberof domestic
firmsfollowing the transitionto trade is typicallydominatedby the numberof
new foreign exporters. It is nevertheless possible, when the export costs are
high, that these foreign firmsreplace a largernumberof domestic firms (if the
latter are sufficientlyless productive).Although product variety then impacts
negatively on welfare, this effect is dominated by the positive contributionof
the aggregateproductivitygain. Trade-even though it is costly-always gen-
erates a welfare gain (see Appendixfor proof).

7.1. TheReallocationof MarketSharesand ProfitsAcrossFirms

I now examinethe effects of trade on firmswith differentproductivitylevels.
To do this, I contrastthe performanceof a firmwith productivity5o' (Pa before
and after the transitionto trade. Let ra(sp)> 0 and xla((P) : 0 denote the firm's
revenue and profit in autarky.Recall that, in both the closed and open econ-
omy equilibria,the aggregate revenue of domestic firms is exogenously given
by the country'ssize (R = L). Hence, ra(p)/R and r(<p)/R representthe firm's
marketshare (within the domestic industry)in autarkyand in the equilibrium
with trade.Additionally,in this equilibriumwith trade, rd(Gp)/R representsthe

26Because'pt factorsin the outputlost in exporttransit(from r), it is possible for ' , to be lower
than (pa when T is high and f, is low. It is shown in the Appendix that any productivityaverage
that is based on a firm'soutput "at the factorygate"must be higherin the open economy.
27Recallthat the averageprofit X must be higherin the open economy equilibrium.

firm's share of its domestic market (since R also represents aggregate con-
sumer expenditurein the country). The impact of trade on this firm'smarket
share can be evaluatedusing the followinginequalities(see Appendix):
rd((p) < ra(p) < rd(p) + nrX(p) Vp > f

The first part of the inequalityindicates that all firms incur a loss in domestic
sales in the open economy. A firm who does not export then also incurs a
total revenue loss. The second part of the inequalityindicates that a firmwho
exports more than makes up for its loss of domestic sales with export sales
and increasesits total revenues.Thus, a firmwho exportsincreasesits share of
industryrevenues while a firm who does not export loses market share. (The
market share of the least productive firms in the autarkyequilibrium-with
productivitybetween SDa and p*-drops to zero as these firmsexit.)
Now consider the change in profit earned by a firm with productivityp. If
the firm does not export in the open economy, it must incur a profit loss, since
its revenue, and hence variableprofit, is now lower. The directionof the profit
change for an exportingfirm is not immediatelyclear since it involves a trade-
off between the increasein total revenue (and hence variableprofit) and the in-
crease in fixed cost due to the additionalexportcost. For such a firm (p > f),
this profitchange can be written:28

AT(4P) = 7T(Dp) - Ta(4P) =-([rd(Gp) + nrx()] - ra(D)) - nfx

~c~u1[ + n-1
=O u-f (' I ~(c)u-1
('"] nfx'

where the term in the bracket must be positive since rd(p) + nrx(p) > ra(p)
for all q > q'*.The profit change, Ar(Gp),is thus an increasingfunction of the
firm'sproductivitylevel p. In addition, this change must be negative for the
exportingfirmwith the cutoff productivitylevel (pD:29 Therefore,firmsare par-
titioned by productivityinto groups that gain and lose profits. Only a subset
of the more productivefirmswho export gain from trade. Among firmsin this
group, the profit gain increases with productivity.Figure 2 graphicallyrepre-
sents the changes in revenue and profitsdrivenby trade.The exposureto trade
thus generates a type of Darwinianevolution within an industrythat was de-
scribed in the introduction:the most efficient firmsthrive and grow-they ex-
port and increaseboth their marketshare and profits.Some less efficient firms
still export and increase their market share but incur a profit loss. Some even
less efficient firmsremain in the industrybut do not export and incur losses of
both marketshare and profit.Finally,the least efficient firmsare drivenout of
the industry.

28Using rd((p) = ((p/(p*)Of-lof and ra(p) = ( O/<'olf.

29Sinceirx(px) = 0 and rd((p ) < ra((p).



ji . ~ (x


FIGURE2.-The reallocationof marketshares and profits.

7.2. WhyDoes TradeForcethe Least ProductiveFirmsto Exit?

There are two potential channels throughwhich trade can affect the distri-
bution of survivingfirms.The first to come to mind is the increase in product
marketcompetition associated with trade:firmsface an increasingnumber of
competitors;furthermorethe new foreign competitors, on average, are more
productivethan the domestic firms.However, this channel is not operative in
the current model due to the peculiar and restrictivepropertyof monopolis-
tic competition under C.E.S. preferences: the price elasticity of demand for
anyvariety does not respond to changes in the numberor prices of competing
varieties. Thus, in the current model, all the effects of trade on the distribu-
tion of firms are channeled through a second mechanism operating through
the domestic factor marketwhere firms compete for a common source of la-
bor: when entry into new export markets is costly, exposure to trade offers
new profitopportunitiesonly to the more productivefirmswho can "afford"to
cover the entrycost. This also induces more entry as prospectivefirmsrespond

to the higher potential returnsassociated with a good productivitydraw.The

increased labor demand by the more productivefirms and new entrants bids
up the real wage and forces the least productivefirmsto exit.
The current model thus highlights a potentially importantchannel for the
redistributiveeffects of tradewithin industriesthat operates throughthe expo-
sure to export markets.Recent work by Bernardand Jensen (1999b) suggests
that this channel substantiallycontributesto U.S. productivityincreaseswithin
manufacturingindustries.Nevertheless, the model should also be interpreted
with caution as it precludes another potentially importantchannel for the ef-
fects of trade,which operates throughincreases in importcompetition.


The preceding analysiscompared the equilibriumoutcomes of an economy

undergoinga massivechange in trade regime from autarkyto trade. Veryfew,
if any, of the world's current economies can be considered to operate in an
autarkyenvironment.It is therefore reasonable to ask whether an increasein
the exposure of an economy to trade will induce the same effects as were pre-
viously describedfor the transitionof an economy from autarky.The current
model is well-suited to address several different mechanismsthat would pro-
duce an increase in trade exposure and plausiblycorrespondto observed de-
creases in trade costs over time or some specificpolicies to liberalizetrade.The
effects of three such mechanismsare investigated:an increase in the number
of availabletrading partners (resulting, for example, from the incorporation
of additionalcountries into a trade bloc) and a decrease in either the fixed or
variabletrade costs (resultingeither from decreases in real cost levels or from
multilateralagreementsto reduce tariffsor nontariffbarriersto trade). These
three scenarios involve comparativestatics of the open economy equilibrium
with respect to n, T, and f,. The main impact of the transitionfrom autarky
to trade was an increase in aggregateproductivityand welfare generated by a
reallocationof market shares towardsmore productivefirms (where the least
productivefirmsare forced to exit). I will show that increases in the exposure
to trade occurringthroughany of these mechanismswill generate very similar
results:in all cases, the exposureto tradewill force the least productivefirmsto
exit and will reallocate marketsharesfrom less productiveto more productive
firms.The increasedexposureto trade will also alwaysdeliverwelfare gains.30

8.1. Increasein theNumberof TradingPartiners

I firstinvestigatethe effects of an increase in n. Throughoutthe comparative
static analyses,I use the notation of the open economy equilibriumto describe

30Formalderivationsof all the comparativestatics are relegatedto the Appendix.


the old equilibriumwith n countries. I then add primes (') to all variablesand
functionswhen they pertain to the new equilibriumwith n' > n countries.
Inspection of equations (20) and (19) defining the new zero cutoff profit
condition (as a function of the domestic cutoff p*)reveals that the ZCP curve
will shift up and therefore that both cutoff productivitylevels increase with n:
p*` > (p* and fx' > qx. The increase in the number of trading partners thus
forces the least productive firms to exit. As was the case with the transition
from autarky,the increased exposure to trade forces all firms to relinquisha
portion of their share of their domestic market:rd'(p) < rd(Qp),Vp > p*.The
less productivefirmswho do not export (with Sn< Sn*') thus incura revenue and
profit loss-and the least productiveamong them exit.3"Again, as was the case
with the transitionfrom autarky,the firmswho export (with f > ` ) more than
make up for their loss of domestic sales with their sales to the new exportmar-
kets and increase their combined revenues:rd'(p) + n'rx'(p) > rd(p) + nrx(p).
Some of these firms nevertheless incur a decrease in profits due to the new
fixed export costs, but the most productivefirms among this group also enjoy
an increase in profits (which is increasingwith the firms' productivitylevel).
Thus, both market shares and profits are reallocated towards the more effi-
cient firms.As was the case for the transition from autarky,this reallocation
of market shares generates an aggregateproductivitygain and an increase in

8.2. Decreasein TradeCosts

A decrease in the variabletrade cost T will induce almost identicaleffects to
those just described for the increase in tradingpartners.The decrease from T
to T' < T(again I use primes to reference all variablesand functionsin the new
equilibrium)will shift up the ZCP curve and induce an increase in the cutoff
productivitylevel p*'> 5p*.The only difference is that the new export cutoff
productivitylevel 5p' will now be below 'p*.As before, the increased exposure
to trade forces the least productivefirms to exit, but now also generates entry
of new firms into the export market (who did not export with the higher T).
The direction of the reallocationof market shares and profitswill be identical
to those previouslydescribed:all firms lose a portion of their domestic sales,

3lThere is a transitionalissue associated with the exportingstatus of firmswith productivity

levels between (p and (p*'.The loss of export sales to any given country-from rx(p) down to
rx'(fp)-is such that firmsenteringwith productivitylevels between (x and (o*'will not export as
the lower variableprofit rx'((p)/o no longer covers the amortizedportion of the entry cost fx.
On the other hand, incumbentfirmswith productivitylevels in this range have alreadyincurred
the sunk export entry cost and have no reason to exit the export marketsuntil they are hit with
the bad shock and exit the industry.Eventually,all these incumbentfirmsexit and no firmwith a
productivitylevel in that rangewill exportonce the new steady state equilibriumis attained.
32Aspointed out in footnote 26, the productivityaveragemust be based on a firm'soutput "at
the factorygate."

so that the firmswho do not export incur both a market share and profit loss.
The more productivefirmswho exportmore than make up for their loss of do-
mestic sales with increased export sales, and the most productivefirmsamong
this group also increase their profits.As before, the exit of the least productive
firms and the market share increase of the most productive firms both con-
tributeto an aggregateproductivitygain and an increase in welfare.33
A decrease in the fixedexportmarketentrycost f, induces similarchangesin
the cutoff levels as the decrease in r. The increasedexposureto tradeforces the
least productivefirms to exit (sp*rises) and generates entry of new firms into
the exportmarket(qx decreases). These selection effects both contributeto an
aggregateproductivityincrease if the new exportersare more productivethan
the averageproductivitylevel. Although the less productivefirmswho do not
export incur both a market share and profit loss, the market share and profit
reallocationstowardsthe more productivefirms,in this case, will not be similar
to those for the previous two cases: the decrease in fx will not increase the
combinedmarketshare or profit of any firmthat alreadyexportedpriorto the
change in fx-only new exportersincrease their combined sales. However, as
in the previoustwo cases, welfare is higher in the new steady state equilibrium.
Both types of trade cost decreasesdescribedabove also help to explainanother
empiricalfeature, reportedby Roberts, Sullivan,and Tybout(1995), that some
exportbooms are drivenby the entry of new firmsinto the exportmarkets.34

This paper has described and analyzed a new transmissionchannel for the
impact of trade on industry structure and performance. Since this channel
works throughintra-industryreallocationsacross firms, it can only be studied
within an industrymodel that incorporates firm level heterogeneity. Recent
empiricalwork has highlightedthe importanceof this channel for understand-
ing and explainingthe effects of trade on firm and industryperformance.
The paper shows how the existence of export market entry costs drastically
affects how the impact of trade is distributedacross different types of firms.
The induced reallocationsbetween these different firms generate changes in
a country'saggregateenvironmentthat cannot be explainedby a model based
on representativefirms. On one hand, the paper shows that the existence of
such costs to trade does not affect the welfare-enhancingproperties of trade:
one of the most robust results of this paper is that increases in a country's
exposure to trade lead to welfare gains. On the other hand, the paper shows
how the export costs significantlyalter the distributionof the gains from trade

33Seefootnote 32.
340verhalf of the substantialexportgrowthin Colombianand Mexicanmanufacturingsectors
was generatedby the entryof firmsinto the exportmarkets.

across firms. In fact, only a portion of the firms-the more efficient ones-
reap benefits from trade in the form of gains in market share and profit. Less
efficient firms lose both. The exposure to trade, or increases in this exposure,
force the least efficient firms out of the industry.These trade-inducedreallo-
cations towardsmore efficient firmsexplainwhy trade may generate aggregate
productivitygains without necessarily improvingthe productive efficiency of
Although this model mainlyhighlightsthe long-runbenefits associatedwith
the trade-inducedreallocations within an industry,the reallocation of these
resources also obviously entails some short-runcosts. It is therefore impor-
tant to have a model that can predict the impact of trade policy on inter-firm
reallocations in order to design accompanyingpolicies that would address is-
sues related to the transitiontowardsa new regime. These policies could help
palliate the transitionalcosts while taking care not to hinder the reallocation
process. Of course, the model also clearly indicates that policies that hinder
the reallocationprocess or otherwise interferewith the flexibilityof the factor
markets may delay or even prevent a country from reaping the full benefits
from trade.
Departmentof Economics, Harvard University,Littauer Center,Cambridge,
AM 02138, U.S.A.;CEPR;and NBER.
ManuscriptreceivedApril,2002;final revisionreceivedApril,2003.


Using the definitionof 0 in (7), the aggregationconditionsrelatingthe aggregatevariablesto

the numberof firmsM and aggregateproductivitylevel f are derived:

Q= L00 q((p)PM/L(p)dpJ
i 1/p
(by definitionof Q U)

q(p)P (J
/ 'p

=M /pq(>P),

and using the definitionof R and H as aggregaterevenue and profit,

R=jr(40)M/L(0)d40p= r(~)f) Mg4~)dq, =Mr(o),

H= j 7r(p)ML(p) dp =-jr (p)ML (4p)d -Mf=M[--f ]=Mi).

B.1. Existenceand Uniquenessof the EquilibriumCutoffLevel (p*
Followingis a proof that the FE and ZCP conditions in (12) identifya unique cutoff level *
and that the ZCP curve cuts the FE curve from above in ((, ir) space. I do this by showingthat

[1 - G((p)]k((p)is monotonicallydecreasingfrom infinityto zero on (0, oo). (This is a sufficient

conditionfor both properties.)Recall that k(>p)= - 1 where

(B.1) ( 4ff- 1 I( j 6 f-lg(4) dA

as defined in (9). Thus,

f) -C
g W) [ ( SD ) _-1] ~ ( (eD) a-1 (

k((p)g((p) ( - 1)[k((p)+ 1]
1 - G((p) f


(B.2) i((p) = [1 - G((p)]k((p).

Its derivativeand elasticityare given by

(B.3) j'(p) = --( - 1)[1 - G((p)][k(@p)+ 1] < 0,

(B.4) ())=- 1)1+k '- 1.

Since j((p) is nonnegativeand its elasticitywith respect to p is negative and bounded awayfrom
zero, j((p) must be decreasingto zero as p goes to infinity.Furthermore,lim,0 ]j('p)= oo since
limqp,ok(p) = oc. Therefore,j(fp) = [1 - G(>p)]k(@p)decreasesfrom infinityto zero on (0, oc).


C.1. AggregateLaborResourcesUsedto CovertheExportCosts
It was assertedin footnote 22 that the ratio of new exportersto all exporterswas 6, and hence
that the aggregatelabor resourcesused to cover the exportcost did not depend on its represen-
tation as either a one time sunk entry cost or a per-periodfixed cost. As before, let Me denote
the mass of all new entrants.The ratio of new exportersto all exportersis then pxPinMe/PxM.
This ratio must be equal to Uas the aggregatestabilityconditionfor the equilibriumensuresthat
PinMe = 6M.

C.2. Existenceand Uniquenessof theEquilibriumCutoffLevel (p*

Followingis a proof that the FE conditionand the new ZCP conditionin (20) identifya unique
cutoff level * and that this new ZCP curvecuts the FE curvefrom above in (p, ir) space. These
conditions imply 8fe/[l -G(*)] = fk(p*) + pxnfxk(p*), or

(C.1) fj((p*) + nfxj(qp) =


where f0 = T(fx/f)l/(Cf-l)(* is implicitlydefined as a function of * (see (19)). Since j(p) is

decreasingfrom infinityto zero on (0, oo), the left-hand side in (C.1) must also monotonically
decrease from infinityto zero on (0, oc). Therefore, (C.1) identifiesa unique cutoff level * and
the new ZCP curvemust cut the FE curvefrom above.


D.1. Welfare
Using (14), welfareper workerin autarkycan be writtenas a functionof the cutoff productivity

Wa = Ma1 = P(T+f y) a.

Similarly,welfare in the open economy can also be written as a function of the new cutoff pro-
ductivitylevel (see (17)):36

(D.1) w =Mt'-' pft=P (T

) *
Since * > (p*,welfare in the open economy must be higherthan in autarky:W > Wa.

D.2. Reallocations

PROOF THAT rd((p) < ra((p) < rd((p) + nr,(p) = (1 + nT1-0')rd(p): Recall that ra(p) =
((P/p*a) f-1f (V5p> (Pa) inautarkyandthatrd(p) = (p/l*)f-lyf (Vp > *) intheopeneconomy
equilibrium.This immediatelyyields rd((p) < ra((p)since * > (P. The second inequalityis a di-
rect consequenceof anothercomparativestatic involvingr. It is shown in a followingsection that
(1 + nTr`)rd(p) decreases as r increases.Since the autarkyequilibriumis obtained as the lim-
iting equilibriumas r increasesto infinity,ra((p)= lim7+,,, rd((p) = limT+00[,(1 + nr'"-f)rd((p)].
Therefore, ra(fp) < (1 + nT -"f)rd(fp) for any finite r.

D.3. AggregateProductivity
It was pointed out in the paperthat aggregateproductivity t in the open economymaynot be
higherthan (a due to the effect of the output loss incurredin exporttransit.It was then claimed
that a productivityaveragebased on a measure of output "at the factorygate"would alwaysbe
higher in the open economy.Define

(D.2) '= h-1 j r((p)h((p)g((p)d0)

as such an average where h(.) is any increasing function. The only condition imposed on this aver-
age involvesthe use of the firms'combined revenues as weights.37Let 'Pa= h-'((l/R) f0o ()
x h(qp)g((p)d(p) representthis productivityaveragein autarky.Then ' mustbe greaterthan Oa-
for any increasingfunction h(.)-as the distributionr((p)g((p)/R first order stochasticallydom-
inates the distributionra(p)g((p)/R: foJr(()g(()d < fojra(5)g(5)d V8p(and the inequalityis
strictVp > ).38


E.1. Changesin the CutoffLevels
These comparativestatics are all derived from the equilibriumcondition for the cutoff lev-
els (C.1) and the implicitdefinitionof fp as a functionof sp*in (19).

35Using the relationship = r(a)/r(p*) = (R/Ma)/Of = (L/Ma)/Tf.

36Using (st/s(*)a-l = rd(4t)/rd((p*) = (R/Mt)/of = (L/Mt)/f.
37Thisis the standardway of computingindustryproductivityaveragesin empiricalwork.
38This result is a direct consequenceof the marketshare reallocationresult.

Increasein n: = (0/0*)d*ldn
Differentiating(C.1) with respect to n and using d@p*/dn
from (19) yields
d *
1 -fx(P*i('P*)
dn f p*j'((p*) + nfxpji'(qX)

Hence dap*/In> 0 and d@o*/In> 0 since j'(cp) <0VOp (see (B.4)).

Decrease in r: Differentiating (C.1) with respect to - and using dp/aid = fp/i +

(9*/9*)ad*/aT from (19) yields
__ nfj '(P)P <
(E.2) <
dar r7 f(p*jI((p*) + nf4xOi(40*)
since j'(cp)<0VO , and
d_px__ -->(0p) .p 0
dr nfxf'((p) dr
Decreasein fx: Differentiating(C.l) with respectto fx andusingdP*/dfx= (4o*/o*)ado*/dfx
[1/(u - 1)]q*/fx from (19) and j'('p*)'p* = -( - 1)[j(qp) + 1 - G(p*)] from (B.2) and (B.4)
,!i:>p* n[l - G(qp)] <0
afx fj'((p*)+ nfxi'((p) ((*l/(p*)
since j'(cp) <0VOp,and
dap -1 F 1P
- nj](4~p
f(x *) > 0.
dfx= nfxj,( x) dn(P)+t > fx]>?

Welfare: Recall from (D.l) that welfare per workeris given by W = p(L/af)l/(f-l)p*. Wel-
fare must therefore rise with increases in n and decreases in fx or - since all of these changes
induce an increasein the cutoff productivitylevel qo*.

E.2. Reallocationsof MarketShares

Recall that rd(p) = ((/l*) 0- If (Vq > *) in the new open economy equilibrium.rd(p)
therefore decreaseswith increasesin n and decreasesin fx or - since all of these changesinduce
an increase in the cutoff productivitylevel p*. Thus rd(p) < rd((0) Vp > * whenever n' > n,
'r < r, or fx' < fx (since *'` > p*).
The direction of the change in combined domestic and export sales, rd(p) + nrx(p) = (1 +
n,r )rd(p), will depend on the direction of the change in (1 + nrl )/(A*) fl. It is therefore
clear that a firm'scombinedsales will decrease in the same proportionas its domesticsales when
fx decreasessince 1 + nr1-' will remainconstant.On the other hand, it is now shown that these
combined sales will increasewhen n increases or T decreases as (1 + nr1- )/(fp*)0f-l will then

Increasein n: From (E.1),

f* 1 [f (P*]j((*) (P*]j((P*)xl
(p* =-
dnan Lf ) +
fx 1i j((9+)
( n)

ffl(P*? )-1 j(P*?) 'i((. ) + n ji((' ) (using (19))

< 1(T )(7-a 1+ n)] l


since -(j'(()/j(() > a - V8p

1 (see (B.4)) and (p*) -lj((P*)/[((P*)f-lj(9p*)] > 1.39 Hence,

[(,P) z- ] =1 + nr (n19p*d 1]
dn (D(P*)T-1L T-1 + n dn sp*

Decreasein r: From (E.2),

[f *](*)
i + -1]
d_ _ - LnfX f(pj +(1)
[ - G((p*)][k((p*) + 1] 1]g
+ 1I (using (B.3))
= f -1 G((p*)][k((p*) + 1] j

,0a- 0-0-1
[, f?- S1?g(() d +11 (using(19))
+ 1
(SP*a1 1 +nT-ag(T d

since f7?
5t'-g(5t)d5/lf7bo 'tg(5t)dt] > 1 as * <qf. Hence,
IT) ( +1
r ~1 + ,l
0-1 +
- (=r-i)(
+ l- a
- -
a * j

< .

E.3. Reallocationsof Profits

(E.3) n: ) All
'\T(f surviving
= 7'(f firmswho do not export (with f <kf') must incur a profit loss
-(- (f )
since their profitsfrom domestic sales decrease
(rp(<) rd(H))
< and those who would have ex-
ported previously(with the lower n) furtherlose any profits from exporting.Similarly,the firm
with productivitylevel f = 4ox'also incurs a profit loss (althoughthe firm exports,it gains zero
additionalprofits from doing so and still incurs the loss in domestic profits).The profit change
for all exportingfirms(with f > pt') can be written:

((p) - r(p)] -(n'- n)f

1-f[I+ n'ir _I+

- nrl 7] - (n' -n)f .

This profit change increaseswithout bound with xo

and will be positive for all above a cutoff
levelqt> f*'.4O

profit(so) mustbe a decreasingfunctionof rdsince its elasticitywith respect to exis

(u - 1) + 'pj'(Gp)/IGp)
40Theterm in the bracketin (E.3) must be positive as 1 + nrlom/(xo*)rl increaseswith fi

Decreasein r: As was the case with the increase in n, the least productivefirms who do
not export (with sp< sp*')incurboth a revenue and profit loss. There now exists a new category
of firmswith intermediateproductivitylevels (sp`'< sp< sp*)who enter the export marketsas a
consequence of the decrease in r. The new export sales generate an increase in revenue for all
these firms,but only a portion of these firms(with productivitysp> sotwhere sp` < SDt < fp) also
increasetheir profits.Firmswith productivitylevels sp> sp*who exportboth before and afterthe
change in r enjoy a profitincreasethat is proportionalto their combinedrevenueincrease(their
fixedcosts do not change) and is increasingin their productivitylevel sp:

Air(>p)= -[r'(>p) - r(>pfl

= p_ff nr)l + n r'-ff

9/f- -(9*)Of-1 j

where the term in the bracketmust be positive.

E.4. Changesin AggregateProductivity

Any productivityaverage based on (D.2) must increase when n increases or ir decreases as
the new distributionof firm revenues r'(qp)g(qp)/Rfirst order stochasticallydominates the old
one r(qP)g(qP)/R:fo"r'(6)g(6) d f<
r(<)g( ) d6 V,P.4' Note that this propertydoes not hold
when fx decreasesas the revenuesof the most productivefirmsare not higherwith the lower fx.
Nevertheless,the productivityaverage P will rise when fx decreasesso long as the new exporters
are more productivethan the average(p* > P).

Aw, B. Y., S. CHUNG, AND M. J. ROBERTS(2000): "Productivity and Turnover in the Export
Market:Micro-levelEvidence from the Republicof Korea and Taiwan(China),"WorldBank
EconomicReview,14, 65-90.
BERNARD,A. B., J. EATON,J. B. JENSON,AND S. KORTUM(2000): "Plants and Productivity in
InternationalTrade,"NBER WorkingPaperNo. 7688, forthcomingin theAmericanEconomic
BERNARD,A. B., AND J. B. JENSEN(1999a):"ExceptionalExporterPerformance:Cause, Effect,
or Both?"Joumalof IntemationalEconomics,47, 1-25.
(1999b):"Exportingand Productivity,"NBER WorkingPaperNo. 7135.
(2001): "WhySome FirmsExport,"NBER WorkingPaperNo. 8349.
(2002): "The Deaths of ManufacturingPlants,"NBER WorkingPaperNo. 9026.
BERNARD, A. B., AND J. WAGNER (2001): "Export Entry and Exit by German Firms,"
Archiv-Review of WorldEconomics,137, 105-123.
CLERIDES,S. K., S. LACK,AND J. R. TYBOUT (1998): "Is Learning by Exporting Important?
Micro-dynamicEvidencefromColombia,Mexico,and Morocco,"TheQuarterly Joumalof Eco-
nomics, 113, 903-947.
DAS, S., M. J. ROBERTS,ANDJ. R. TYBOUT (2001):"MarketEntryCosts,ProducerHeterogeneity
and ExportDynamics,"Mimeo, PennsylvaniaState University.
DIxIT,A., ANDJ. STIGLITZ (1977): "MonopolisticCompetitionand OptimumProductDiversity,"
AmericanEconomicReview,67, 297-308.
HELPMAN,E., M. J. MELITZ,AND S. R. YEAPLE(2002): "Export Versus FDI," NBER Working
PaperNo. 9439.

4"Again,this is a direct consequenceof the marketshare reallocationresults.


HOPENHAYN,H. (1992a): "Entry,Exit, and Firm Dynamicsin Long Run Equilibrium,"Econo-

metrica,60, 1127-1150.
(1992b): "Exit, Selection, and the Value of Firms,"Joumal of Economic Dynamicsand
Control,16, 621-653.
KRUGMAN, P. R. (1979): "IncreasingReturns, Monopolistic Competition, and International
Trade,"Joumalof IntemationalEconomics,9, 469-479.
(1980): "ScaleEconomies, ProductDifferentiation,and the Patternof Trade,"American
EconomicReview,70, 950-959.
MONTAGNA, C. (1995): "MonopolisticCompetitionwith Firm-SpecificCosts,"OxfordEconomic
Papers,47, 318-328.
PAVCNIK,N. (2002): "TradeLiberalization,Exit, and ProductivityImprovements:Evidence from
Chilean Plants,"TheReviewof EconomicStudies,69, 245-276.
ROBERTS,M. J., T SULLIVAN,AND J. R. TYBOUT(1995): "Micro-Foundations of ExportBooms,"
Mimeo, WorldBank.
ROBERTS,M. J., ANDJ. R. TYBOUT(1997a):"TheDecision to Exportin Colombia:An Empirical
Model of Entrywith Sunk Costs,"AmericanEconomicReview,87, 545-564.
(1997b): WhatMakesExportsBoom? Washington,D.C.: The WorldBank.