27 Pages


Downloading requires you to have access to the YouScribe library
Learn all about the services we offer


Chapter41SpilloversandExportBehavior4.1 IntroductionFacilitating exports by domestic firms today appears as an important priority of policymakers in developed as well as developing countries. In December 2004, the FrenchForeignTradeministerarguedthat“Exportisanationalcauseforwhichthegovernmentismobilized”. Heannounceda setofmeasuresto betaken bythe governmentin orderto “consolidate the international presence of firms already exporting, and broaden thenumberofexporters(...)”.From the economicpoint of view, such interventions are justified in case of marketsfailures,andonepossiblyimportantfailureinthecaseofexportingispositiveinforma-tionexternalities, i.e. exportspillovers. Theunderlyingmechanismofexportspilloversis that the export specific knowledge of firms that are experienced on foreign marketscanbenefitnearbyfirmsand allowthem to start exporting toagiven market. Theexis-tenceofexportspillovershasbeenstudiedbyseveralpapers,buttheempiricalevidence1This chapter is based on Koenig P. (2005), “Agglomeration and the Export Decision of FrenchFirms”,CRESTWorkingPaper#2005-01,submittedforpublication.122CHAPTER4. SPILLOVERSANDEXPORTBEHAVIOR 123is either indirect or questioned. Indeed, the identification of export spillovers requiresdata on a panel of exporting firms, so as to disentangle the export spillovers from thefirm specific effect. And, using a panel of exporting firms, one will need to handle theprobable existence ofasunkexport cost, hencecontrol for thefact ...



Published by
Reads 12
Language English

4.1 Introduction
Facilitating exports by domestic firms today appears as an important priority of policy
makers in developed as well as developing countries. In December 2004, the French
ismobilized”. Heannounceda setofmeasuresto betaken bythe governmentin order
to “consolidate the international presence of firms already exporting, and broaden the
From the economicpoint of view, such interventions are justified in case of markets
tionexternalities, i.e. exportspillovers. Theunderlyingmechanismofexportspillovers
is that the export specific knowledge of firms that are experienced on foreign markets
canbenefitnearbyfirmsand allowthem to start exporting toagiven market. Theexis-
1This chapter is based on Koenig P. (2005), “Agglomeration and the Export Decision of French
is either indirect or questioned. Indeed, the identification of export spillovers requires
data on a panel of exporting firms, so as to disentangle the export spillovers from the
firm specific effect. And, using a panel of exporting firms, one will need to handle the
probable existence ofasunkexport cost, hencecontrol for thefact thatsome firmscon-
tinueto exportwhile someothers start toexport. Finally, becausespilloversare known
to decrease with geographic distance (Jaffe, Trajtenberg and Henderson, 1993), identi-
fyingthem alsorequiresdata on thepresence ofexporters atanadequatelygeographic
Directevidenceofexportsspilloversisprovidedbytwopapers. Aitken,Hansonand
Harrison (1997)find thatthe probability that Mexicanplantsexport in 1986and 1989is
positively linked to the presence of multinational firms in the same state, but uncorre-
latedtoproximitytooverallexporters. Greenaway,SousaandWakelin(2004)showthat
domesticfirmsover1993-1996. IndirectevidenceofexportspilloversisgivenbyLovely,
RosenthalandSharma(2004). Assumingthatexportingrequiresspecializedknowledge
of foreign markets, they show that headquarters of exporting firms are more spatially
concentratedthanthoseofnon-exportingfirms. More,theirresultsassessanincreasing
spatial concentration of exporters’ headquarters with the difficulty of the destination
country. Finally,twopapersunderlinetheabsenceofanyevidenceofexportspillovers.
Barrios, Görg and Strobl (2003) do not find evidence that Spanish firms benefit from
spillovers through export activity of other firms nor from multinationals’ activity be-
tween 1990 and 1998. Bernard and Jensen (2004) find no role for export spillovers on a
panelofU.S. manufacturing firms, be theyfrom nearbyexporters or from export activ-
ity from other firms in the same industry. However, unlike the existing literature, they
handletheasymmetrybetween‘starters’and‘continuers’usingthelaggedexportstatusCHAPTER4. SPILLOVERSANDEXPORTBEHAVIOR 124
pirical literature has only looked for general export spillovers: In the first four papers,
theunderlyingassumption isthatthepresenceofmultinationalsordomesticexporters
impacts the variable cost of a firm at exporting and hence facilitates the overall export
decision. Bernard and Jensen (2004) assume that established exporters can reduce ei-
ther the sunk entry cost or the variable cost, but in both cases the spillovers affect the
overall exportdecision. Inthiscontext, onequestion comestomind concerningthe na-
ture, and thus the identification of export spillovers: What if export spillovers are in
fact destination specific? Indeed it appears quite reasonable to think that the relevant
information thatisabletoinfluenceafirm tostartexporting somewhere isdestination-
specific. Whenlookingforforeignmarketstosellitsproduct, amanufacturerwillwant
to learn details about the preferences of consumers and the structure of distribution
In this chapter, I investigate the presence of export spillovers, precisely allowing
theseeffectstobegeneral,industryspecific,orindustryanddestinationspecific. Iusea
manufacturersanddestinationcountriesbetween1986and1992. TheCustomsdataare
and number of employees. I then estimate a discrete choice model on the probability
that a firm starts to export to a country. In doing so, I handle the existence of a sunk
cost because all firms in the sample have to pay the sunk cost to enter the foreign mar-
ket. Firm fixed effects, and dummies for years and countries allow me to control for
unobserved characteristics of places, firms, years and countries. In the end, I identify
potential export spillovers by studying the effect of the presence of nearby industryCHAPTER4. SPILLOVERSANDEXPORTBEHAVIOR 125
and/or destination specific exporters on a firm’s decision to start exporting to a given
Thechapterisstructuredasfollows. Insection4.2IexplainhowIgetfromtheprofit
of the firm abroad to the estimation of a logit model on the decision to start exporting.
Section 4.3 describes the sources of the data and the variables that will be used in the
estimation. I also emphasize several important aspects of the database, among which
the number of exporters per country. In section 4.4, I comment the results of the logit
estimations, explaining how we get to the preferred specification, and section 4.5 con-
4.2 Theempiricalmodel
Inowdescribethebehaviorofafirmonforeignmarkets. Newtrademodelswithhetero-
firm’s export behavior. With respect to the newtrade models with representative firms
(Krugman, 1980), two new elements are added. First, firms must pay a fixed cost in
order to enter export markets. Second, firms are heterogeneous so that only a subset
of firms is able to overcome the fixed cost and start exporting. In the following I will
profitofthefirmabroad. Thiswillallowmetodescribehowthefirm choosestosellon
foreignmarketsandmodeltheprobabilitytoexport. Beforegettingtothemodel,letus
state the two elements that a rigorous link between the theory and the empirical work
Thefirstpointreferstotheempiricalhandlingofthefixedcostatexporting. Indeed,
letusconsiderafirmifacingthedecisionofexportingornottoacountryj. ByexportingCHAPTER4. SPILLOVERSANDEXPORTBEHAVIOR 126
to that country, the firm is able to make an annual profit equal toΠ . However, if theij
firm has never exported to that country before, it must incur a sunk cost f to cover
the cost of entering the market. For each year and country, there will thus be firms
that continue and firms that start to export to that country, corresponding to firms that
have already paid the sunk cost and those that have not. This asymmetry between the
continuersandthestarters becomesaproblemfortheestimation preciselybecausethis
sunk cost is not observed: we have no information on which firms have paid the sunk
cost and which ones have not. Therefore, we must find a variable to distinguish the
continuers from the starters, because this asymmetry is in itself a potential reason for
which to remain on the export market in year t. If not controlled for, it could bias the
estimates of the firm level variables coefficients. Following Roberts and Tybout (1997),
most of the empirical literature on the export decision uses the lagged export status as
a proxy for those firms. However, as noted by Robert and Tybout (1997) and Bernard
and Jensen (2004), the use of this variable creates substantial econometric difficulties
because the identification of the spillovers also requires the specification to control for
The approach chosen in this chapter and in the following is to consider only the
firms that start to export to a given market. Doing so handles the asymmetry between
the firms, because none of the firms in the remaining sample have paid the sunk cost.
ofthe possibledeterminantsoftheprobabilitytoexport. Ofcourse, theconsequenceof
considering the decision to start exporting is that we are left with the sunk cost in the
tion. Inthefollowing, Iwillpropose awaytomodelthesunkcostthatiscoherentwith
the phenomenon I am willing to identify. Note that in concentrating on the decision+
to enter a foreign market, I also focus on a phenomenon that is more likely to be influ-
encedbyexportspillovers. Indeed,onceafirmhasenoughinformationonacountryin
fromestablishedexportersinyeart 1inordertodecideonherexportbehaviortothat
Thesecondpointtoclarifyconcernsthemodelingofthechannelthrough whichex-
portspilloversaffecttheprofitofthefirmabroad. Intheory,spilloverscouldactthrough
Letusnowmodel afirm’s export behavior. Afirm i starts toexport to acountry j if
thepresentvalueoffutureprofitsislargerthanthesunkentrycost f :j
∞ Πt
f , (4.1)∑ jt1 rt 0
which,assumingnouncertaintyonfutureprofits, canbewritten
f . (4.2)jr
Theprobabilitythatafirmstartstoexporttocountry j isthen:
Pr S 1 Pr Π /r f . (4.3)ij ij j
Equation (4.3) contains two elements that we will now define: the profit of the firm
abroad and the sunk entry cost. The firm is assumed to trade within a Dixit-Stiglitz-
Krugman monopolistic competition framework. Understandardassumptions ofaCES
utilityfunction,thedemandfromcountry j forfirmi’sproductisgivenby:(


x m Y, (4.4)ij j j1 σP
1/ 1 σ1 σwith P p l dl the local price index, σ the elasticity of substitution be-j jl
tweengoods,m theshareofexpendituresdevotedtotherepresentativeindustry,andYj j
theincomein j,whichequalsthenationalaggregateexpenditurelevel.
Thefinal pricepaidbyconsumers, p ,isthe millprice setbythe firm multipliedbyij
thetradecost: p p τ. Tradecostsaread-valoremandassumedtobeofthe‘iceberg’ij ij j
type, i.e. a fraction of the good melts away during the journey. The price set by the
firmisobtainedthroughoptimizationofthefollowinggrossprofit:Π p x a w x .ij i ij i i ij
Production costs for x units of the good correspond to a w , where w is the nominali i i i
wageand a representsthenumberofunitsoflaborusedbythefirm. FollowingMelitzi
(2003),firmsareheterogeneousinthattheydifferbytheirproductivity: Eachfirm i has
σan inverse productivity level a . With mill price p a w , the gross annual profiti i i iσ 1
ofafirmonmarket j writes
1 σ
a w τi i j
Π m Y. (4.5)ij j j1 σσ 1 Pj
In order to complete the description of equation (4.3), let us define the sunk entry
f z , (4.6)j ij
where z is local exporting employment specific to the industry and/or to the des-ij
tination country. Hence, the size of the local labor force working in firms that already
export to market j measures the potential spillovers from neighboring firms that de-+





creasethesunkcostofentryforfirmi onmarket j. Writteninlogs,equation(4.3)gives
Pr S 1 Pr lnΠ lnr ln f , (4.7)ij ij j
lnΠ β β lna β lnw β lnd β lnm β lnP β lnY (4.8)ij 1 i 2 i 3 j 4 j 5 j 6 j
ln f γ γ lnz ε , (4.9)j 0 1 ij ijt
whereε containstheeffectsspecifictofirms,places,countries, andyears.ijt
Pr S 1 Pr β β lna β lnw β lnd β lnm β lnP β lnY(4.10)ij 1 i 2 i 3 j 4 j 5 j 6 j
lnr γ γ lnz ε .0 1 ij ijt
Pr S 1 Pr β β lna β lnw β lnd (4.11)0 1 2 3ij i i j
β lnm β lnP β lnY γ lnz ε4 j 5 j 6 j 1 ij ijt
Assuming ε is distributed logistically, equation (4.11) is estimated using a Logit.ijt
4.3 Data
The final database is constructed using two different sources of data, which are de-
scribed inthefirst sub-section, aswellassomerestrictions thatImadetothesampleof
firms and to the geographical dimension of the database. I then illustrate some salient
features of exporters and markets, in particular the number of starters per destination
4.3.1 Sources
ing French export flows aggregated by firm, year and destination country. Four recent
papers exploit different aspects of this database. The first two explore the relationship
between export behavior and employment structure of individual firms. Biscourp and
Kramarz (2003)look atthe impact ofexporting on the levelof employmentof the firm,
skilled labor because of development and marketing purposes. Two papers by Eaton,
Kortum and Kramarz (2004a, 2004b) provide a detailed “dissection” of trade at the in-
dividual level. Eaton et al. (2004a) establish key features of market penetration by in-
dividual French firms according to destinations, suggesting the existence of different
types of barriers to exporting to national markets. Eaton et al (2004b) develop a Ricar-
dianmodel of trade with firm heterogeneity to explainthese stylized facts infirm level
export behavior. In this paper, I use the same firm-level exports database, and match
the information provided by the Customs with administrative data from the Enquête
Annuelle d’Entreprises (EAE), which contains firm and establishment-level information
on all firms over 20 employees of the manufacturing sector: address and identificationCHAPTER4. SPILLOVERSANDEXPORTBEHAVIOR 131
numberofthefirm(siren),sales,production, employees,wages.
Inordertobuildthefinaldatabase,Irestrict theyearlycustomsdataintwoaspects.
The first restriction is due to the sample of firms. Indeed, the sample of firms retained
to match the export database contains single-plant firms of the manufacturing sector,
with more than 20 employees, that are continuously operating throughout the years
1986-1992. Letusdetailthedifferentstepsoftheconstruction ofthedatabase.
of which I keep and use the manufacturing part, collected by the French ministry of
industry. In1986,themanufacturingEAEcontainsapproximately22,943firms,ofwhich
withfirmsbirthsanddeathsphenomena. Thenextstepistomatchthefirminformation
with the export data. However, the export data gives the identification number of the
firm that exports but does not detail the establishment from where the export flows
originate. Therefore,amongthesefirms,Ichoosetokeepthe6,131singleplantfirms.
sion. 67 destinations are kept in the final database, comprising 27 OECD countries (30
members; France is excluded; Belgium and Luxembourg, and the Czech republic and
Finally, I compute the distance and export spillovers variables. The latter variable
depends on the level of geographic disaggregation of the data. Three levels are avail-
able: the region (22 entities), the département (94 entities) and the “zone d’emploi”
(employment area) (350 entities), the last measure appearing the most adequate as the
employment areas (henceforth areas) are designed by INSEE to fit geographical loca-
tions comprising work and residence of employees. The distance variable is interdist,
the distance between France and each market (see appendix to chapter 3 for more de-