Agglomeration and the Export Decision of French Firms

Agglomeration and the Export Decision of French Firms∗
Pamina Koenig
†
April 30, 2004
Preliminary and incomplete
Abstract
We investigate the determinants of a firm’s export decision by focusing on local industrial characteristics. Proximity to other exporters may increase a firm’s export probability
through two different types of externalities: non-market interactions (export specific knowledge spillovers) and marked-based interactions impacting on the cost of selling abroad. Using
a panel of French manufacturers exporting to OECD countries and followed from 1986 to
1992, we point to evidence of a positive impact of nearby exporters on the probability that
a firm exports. The effect of proximity is increased when considering firms that export to
the same destination market.
J.E.L. classification: F1
Keywords: export decision, export market, agglomeration, spillovers.
∗
I thank Pierre-Philippe Combes, Keith Head, Francis Kramarz, Guy Laroque, and Megan MacGarvie for
very helpful comments and suggestions.
†
CREST and University of Paris I, J360, 15 bd Gabriel Peri, 92245 Malakoff Cedex, France. Email:
[email protected]
1
1
Introduction
The empirical literature on the export behavior of individual firms has now explored various
determinants of the export decision. First, a firm may have positive exports because of its own
ability to be performant: a high productivity lowers the cost of selling abroad (Bernard and
Jensen, 2004; Bernard and Wagner, 1997) and allows to overcome the fixed cost of exporting
(Roberts and Tybout, 1997). Second, exporting may be facilitated by first or second nature
locational advantages (proximity to the sea or to a transport infrastructure). Third, export
success may be enhanced by the industrial environment of the firm, such as the proximity to
multinational firms (Aitken, Hanson and Harrison, 1997; Barrios, Görg and Strobl, 2003).
Among these variables, a question that remains open is the importance of the industrial
environment of a firm in determining its export behavior. Are there externalities from the
presence of nearby firms on a firm’s export decision, and what form do these externalities
take? A firm’s export decision may be impacted by other local exporters through two different
channels: First, proximity to exporters may increase knowledge transmission about the practice
of exporting, and facilitate the flow of information concerning specific destination countries
(non-market interactions). Second, local exporters may, through two distinct market-based
mechanisms, impact on the cost of selling abroad.
In this paper, we consider the export decision of firms and study the extent to which proximity to other exporters is one of its main determinants. Whatever form does the phenomenon
take, if there are positive export externalities1 , then two distinct features, as well as their spatial
consequences, should be observable. First, proximity to other exporters should increase the export probability of a firm. Second, the effect of proximity to exporters on the export probability
should increase with the difficulty to access the export markets. Suppose export externalities
act by reducing the cost of acceding to the export market. Hence, the more difficult it is to
export, i.e. the higher the trade costs, the more these externalities are needed to overcome
the costs of exporting. Finally, if there are positive externalities, the consequences in terms of
the spatial distribution of firms should also be perceivable. Exporters should be more spatially
concentrated then the overall economic activity2 . And, the more the export destination of firms
is “difficult” to access, the more these exporters should be spatially concentrated.
We analyze two situations in which the cost of exporting varies, thus where the role of export
externalities is likely to adjust. We first focus on the evolution of export externalities across time.
Economic integration is a factor that decreases the necessity to benefit from other exporters’
experience abroad. Indeed, access to foreign markets, in terms of trade costs, should decrease
with the lowering of barriers to trade between countries. Then, we compare the determinants
of the export probability of firms that export to different destination countries. Indeed, if there
are export externalities, these are likely to act differently according to where the firm is located
and to where the firm exports. Let us compare two domestic firms, one exporting to a remote
country, the other to a border country. A simple model of the export decision will show that the
export performance of both firms depends, among others, on the proximity to the destination
markets and the proximity to other (exporting or non-exporting) firms. However, it is likely that
these two elements are not valued the same way by both firms. A firm that considers exporting
to a border country will probably value the proximity to the destination market relatively more
than the proximity to other exporters, compared to the firm that exports to a remote market.
Alternatively, we can expect that, in deciding whether to sell to a remote country, a firm will be
1
In the empirical literature, the terms “export spillovers” and “export externalities” are sometimes used
interchangeably, probably because non-market interactions are the main effect on which the papers put emphasis.
However, we will stick to the latter term, as we are conscient that non-market and market interactions are very
difficult to disentangle and that the term “export spillovers” is potentially a misnomer.
2
Lovely, Rosenthal and Sharma (2004)
2
relatively more demanding of informational externalities from current exporters than the firm
exporting to the border country.
Our approach is dived in two steps. We first study the existence of export spillovers in a
dynamic framework, without the country dimension. We seek to identify export externalities
among other sources of spatial concentration of exporters controlling for exogenous advantages
of heterogeneous places. Our dataset is a balanced panel of French manufacturers that are
continuously operating between 1986 and 1992, however not always exporting. We show that
there is a role of other local exporters in explaining the export probability of a given firm.
Then, we analyze the determinants of exporting to each destination market. The full dataset
consists of exports by French manufacturers to a large set of destination countries. As expected,
distance impacts negatively on exporting to a given country. Proximity to other exporters, and
more specifically to manufacturers exporting to the same country, affects positively the export
decision.
Existing empirical work on export externalities has provided two approaches to the issue.
On the one side, some papers directly address the export decision and analyse the importance
of nearby firms in determining the export probability of firms. The evidence is mixed: Aitken,
Hanson and Harrison (1997) consider the export decision of Mexican manufacturing firms in a
static framework, and find that domestic exporters are positively influenced by the proximity
of multinational firms in the same industry, but not by local export activity. Barrios, Görg
and Strobl (2003), on Spanish data, study different types of spillovers on the export decision
and the export intensity of firms. They find that domestic firms’ decision to sell abroad is
positively impacted by the presence of other exporting firms in the same industry, but not
by multinational firms. Bernard and Jensen (2004) study the export decision of US firms
in a dynamic framework. They find no evidence of export spillovers from neighboring firms.
Greenaway, Sousa and Wakelin (2002)...
On the other side, only one paper have concentrated on examining the existence of export
externalities through their consequences in terms of the spatial distribution of exporting firms.
Lovely, Rosenthal and Sharma (2004) study the location of US manufacturing headquarters
in 2000 and test the hypothesis according to which exporters headquarters should spatially
concentrate in order to reduce the cost of obtaining export-specific information. Their results
indicate that exporters headquarters agglomeration increases with the cost of accessing foreign
markets: agglomeration is high in industries that export mainly to difficult countries. Our paper
is not only complementary to both approaches of export externalities. It also provides a new
insight in identifying them, by exploiting the geographical dimension of the export decision.
Knowing the firm’s location is important in being able to work at a very detailed geographical
level for identifying export externalities. Having a variety of distinct destination countries allows
to have distance as an additional explicative variable in the decision to export, and thus to assess
whether the presence of nearby firms is as important a determinant for all destinations.
The paper is structured as follows. Section (2) exposes the model from which we obtain our
estimable equation for the firm’s export decision. Section (3) describes the main characteristics
of the explanatory variables, among which the location of firms. In section () and (), Section
(7) concludes.
2
Theory
We describe a theoretical model that allows to
We describe firms’ behavior using the Dixit-Stiglitz-Krugman monopolistic competition
framework, and display the main equations that allow us to obtain the estimable equation
for the export decision of a firm. We adapt the trade model (Krugman, 1980) to incorporate
3
firm heterogeneity as in Melitz (2003)3 . Firms face the same fixed cost of exporting, but differ
in their marginal cost, which is equal to wi /varphii . wi is the wage of the factor used in the
variable cost and ϕi is the firm’s productivity.
Consumers have the following utility function:
·Z
U = M µ A1−µ
with
N
M=
i=1
σ−1
σ
ci
σ
¸ σ−1
di
(1)
µ ∈ (0, 1) is a constant, M is the consumption of manufactured goods, A of the agricultural
products. ci is the consumption of variety i of the manufactured good, N is the total number
of available varieties. σ > 1 is the elasticity of substitution between two
R Nvarieties.
Consumers maximise utility under their budget constraint Y = i=1 ci pi di . Demand by
consumers from region j for a variety produced in region i is:
ci =
p−σ
ij
Pj1−σ
µYj
(2)
pij is the final price paid by consumers, which contains the mill price multiplied by the advalorem iceberg trade cost: pij = pi τij , with τ > 1. Yj is the income, and Pj is the industrial
price index in region j:
·Z
¸1/(1−σ)
Pj =
p1−σ
d
(3)
i
ij
i
There are two factors: human capital (H) and labor (L). In order to produce xi units of its
variety of the industrial good, a firm uses α units of human capital as a fixed cost and 1 unit of
labor per unit of good. The total cost writes:
T Ci = αwiH +
wi
xi
ϕi
(4)
Each firm maximises the following profit:
wi
Πi = pi xi − wiH α − xi
ϕi
µ
¶
wi
=
pi −
xi − wiH α
ϕi
which yields the pricing rule
pi =
wi σ
ϕi σ − 1
(5)
(6)
The profit can then be written:
Πi =
wi
xi − wiH α
ϕi (σ − 1)
(7)
We follow Roberts and Tybout (1997) and Bernard and Jensen (2004) in the following steps of
the model. The multi-period gross profit is described as
b it = x∗ + δ(Et [Vit+1 (.)|x∗ > 0] − Et [Vit+1 (.)|x∗ = 0]).
Π
it
it
it
(8)
The participation of the firm to the export market is expressed as follows. A firm exports if its
3
Bernard, Eaton, Jensen and Kortum (2003), Yeaple (2003), Helpman, Melitz and Yeaple (2003), Jean (2002),
Head and Ries (2004).
4
current and expected gross profit is larger than variable costs plus an entry cost, N :
½
b it > cit + F (1 − EXPit−1 )
1 if Π
EXPit =
0
otherwise
(9)
Using (6) and (profit), we incorporate the elements affecting the firm’s profit in the participation rule. The reduced form we use to identify the determinants of the export decision can
be described as

 1 if β1 Ait + β2 Dij + β3 Sit + β4 Pjt + β5 Yjt + β6 wit
+β7 Eit − F (1 − EXPijt−1 ) + εijt > 0
EXPijt =
(10)

0
otherwise
Ait is the firm’s productivity. Dij is the distance between firm i and destination country j.
Sit is the size of the firm, and wit its wage. Pjt is the price index in the destination country,
can be understood as an index of the degree of competition in j. Yjt is the country’s GDP. Eit
is the externality variable (...).
2.1
Export Externalities
Following the literature explaining why firms agglomerate, hence also why proximity to other
firms may impact a firm’s own performance, we review the reasons why the presence of a large
number of firms in the same location may influence a firm’s export decision. We expect exporting
firms to be impacted by the presence of other nearby firms through two different mechanisms.
First, non-market interactions may be involved. These are likely to act either on the cost of
producing the goods sold abroad, or on the cost of sending the merchandise to the destination
markets.
Market interactions may also be at stake. Let us distinguish two cases. the first case can be
pictured through the Krugman (1991) setting, in which firms only interact on the labor market
and on the good’s market. The presence of number of firms selling locally and potentially abroad
does not impact another firm’s activity on foreign markets.
The second case
[To be continued ]
3
Data
The data we use consists of information about French firms of 20 employees or more, firm-level
exports, distance data and countries GDPs. Firm-level exports, by destination country, are
collected by the French Customs from 1986 to 1992. We match these data with the Enquête
Annuelle d’Entreprises (EAE), available at the French Statistical Institute (INSEE). The EAE
contains firm and establishment-level information on all firms over 20 employees of the industrial
sector: address and identification number of the firm (siren), sales, production, employees,
wages.
The manufacturing EAE in 1986 contains approximately 23000 firms, of which 12313 are
operating throughout the 1986-1992 period. Our estimations require to use the precise location
of each firm, and to compute a distance variable between the firm and the destination market.
We thus choose to keep only the mono-regional firms, i.e. firms that may have more than one
plant but which geographical extent is limited. Keeping multi-region firms would necessitate to
chose one plant per firm and to attribute all exports to this plant. This would possibly introduce
to much error in the distance calculations. Finally, we drop the firms located in Corsica, as well
5
as two sectors which contained few producers. The remaining 9630 firms pertain to 33 industrial
sectors whose characteristics are summarized in Tables (9) through (15) of the appendix.
3.1
Characteristics of Exporters
Remembering that our sample is restricted to firms of more than 20 employees, it is not surprising to see that the average size of firms across industries is 82 employees. Exporters are in
average larger than non exporters, give higher wages and are more productive. This corroborates descriptive statistics and results on other databases (Bernard and Jensen, 1997; Roberts
and Tybout (1997)). Table (15) desegregates these figures by industry. The share of exporters
in each industry seems relatively high. It ranges from 31 % in “Ceramic and building materials”
to 87 % in “Chemicals”. Table (12) describes the percentage of firms entering the export market
as well as the percentage of firms exiting the export market, for each year. The numbers are
consistent with the transition rates depicted by Bernard and Jensen (2004): between 10 and 12
% of the firms stop exporting from one year to another, while 12 to 14 % begin to sell on foreign
markets.
3.2
Spatial Concentration of Exporters
As emphasized in the introduction, the existence of positive export externalities is likely to be
perceivable in the location of exporting firms. Indeed, if market or non-market interactions are
important to the future exporter, a newly exporting firm will have a higher probability of being
created in an area dense in exporting firms. In addition, we make the assumption that the
importance of nearby firms in providing export-specific support increases with the cost of access
to the export market. In this sens, we should observe, first, a higher spatial concentration of
exporters with respect to non exporters4 , and second, among exporters, a larger concentration
of firms selling to remote markets. In the following we describe the spatial concentration of
exporters and non-exporters, and emphasize the variation in firm agglomeration according to
time and destination markets.
The 9630 firms of our database occupy 4071 communes out of the 36600 existing in France.
The commune is the administrative desaggregation below the departement level. In our sample,
there is in average 2.36 firms by commune, ranging from 1 to 486 (Paris). In Table (1), spatial
concentration indices (Gini coefficients) are calculated, using departements shares of exporting
and non exporting firms in each year. We see that exporters were relatively more concentrated
than non exporters in 1986. This feature inverses through the sample period, as exporters
become less spatially concentrated. Our database contains the same number of firms each year;
we see in Table (13) that the number of exporters increases during the same period. The
decrease in concentration among exporters is thus due to the fact that newly exporting firms,
throughout the years, are located in areas that are less dense in exporting firms.
Table 1: Exporters vs Non Exporters Spatial Concentration, by Year
Exporters
Non Exporters
4
1986
0.131
0.122
1987
0.126
0.128
1988
0.121
0.131
see Lovely, Rosenthal and Sharma (2004).
6
1989
0.112
0.131
1990
0.111
0.130
1991
0.111
0.135
1992
0.107
0.130
Figure (1) depicts spatial concentration coefficients for firms exporting to different countries,
in 1986. The positive relation between the intensity of agglomeration and the distance to the
destination country appears clearly. The further away is the destination country, the more
exporters are agglomerated. This feature is valid for the whole group of countries but also
appears very clearly at the European level: Exporters to the Netherlands are more dispersed
than those that sell to Greece or Finland.
.5
Figure 1: Spatial Concentration of Exporters, by Country
MEX
KO
.4
PL
NZ
Gini coefficient
.3
HU
TU
JA
IRL
N
.2
OE DK P
NL
D
GR
FIN
SW
AUS
CAN
US
SP
I
.1
UK
6
7
8
ln(dist)
9
10
While this agglomeration feature comforts our assumption according to which, the more the
market is difficult to reach, the more exporters need to be located close to one another, we need
to investigate further the data and control for exogeneous advantages of places in order to assert
the existence of such an export externality.
Indeed, this observable fact may also be the outcome of other factors. The presence of
geographical advantages, a good access to transport infrastructure or to a qualified workforce
are different elements that can also account for the spatial concentration of firms selling abroad,
and thus for the positive relation between the export decision and a large number of local
exporters.
4
Estimation results
We now turn to the analysis of the determinants of the export decision. In particular, we want
to investigate the importance of the presence of nearby firms in explaining a firm’s probability
to sell abroad. Table (2) displays the results of the basic regression of the export probability of
a firm i in year t.
Column (1) and (2) show how the probability to export is determined by firm characteristics. As expected, productivity, proxied by sales per employee, impacts positively on the export
decision, as well as the size of the firm, which can be understood as a second proxy for productivity. Column (2) includes the presence of the firm on export markets in the previous year:
the positive and significative coefficient on this variable confirms the role of being already an
exporter as a factor decreasing the cost of exporting for the consequent years. This phenomenon
is present in Bernard and Jensen (2004), which interpret it as a evidence of the existence of sunk
7
Table 2: Logit on the Probability of Exporting, without country dimension
Model :
ln(sales/emp)
ln(wage)
ln(emp)
(1)
1.166a
(0.023)
-0.199a
(0.052)
1.077a
(0.015)
Exported last year
ln(empl in c)
ln(exp. empl in c)
ln(sh. of exp. empl in c)
Dependent Variable: prob(exportit )
(2)
(3)
(4)
(5)
0.760a 1.165a 1.182a 1.156a
(0.032) (0.023) (0.023) (0.024)
-0.038 -0.194a -0.303a -0.334a
(0.074) (0.052) (0.053) (0.054)
0.663a 1.079a 1.042a 0.998a
(0.021) (0.016) (0.016) (0.016)
3.343a
(0.026)
-0.007
(0.005)
0.112a
(0.004)
1.198a
(0.025)
ln(sh. of non exp. empl in c)
Dep. Dummies
N
R2
yes
67410
0.216
Note: Standard errors in parentheses with a ,
at the 1%, 5% and 10% levels.
yes
57780
0.484
b
and
c
8
yes
67410
0.216
yes
67410
0.224
yes
67410
0.278
respectively denoting significance
(6)
1.161a
(0.024)
-0.244a
(0.054)
0.999a
(0.016)
-0.528a
(0.008)
yes
67376
0.272
Table 3: Logit on the Probability of Exporting, without country dimension
Model :
Exported last year
ln(sales/emp)
ln(wage)
ln(emp)
ln(empl in c)
ln(exp. empl in c)
ln(sh. of exp. empl in c)
ln(sh. of non exp. empl in c)
Dep. Dummies
N
R2
Dependent Variable: prob(exportit )
(1)
(2)
(3)
(4)
a
a
a
3.343
3.334
3.280
3.275a
(0.026) (0.026) (0.027) (0.027)
0.758a 0.772a 0.759a
0.748a
(0.032) (0.032) (0.033) (0.033)
-0.030 -0.124c -0.148c
-0.065
(0.074) (0.074) (0.076) (0.075)
0.667a 0.633a 0.602a
0.599a
(0.021) (0.021) (0.022) (0.022)
-0.012c
(0.007)
0.101a
(0.006)
1.084a
(0.029)
-0.466a
(0.011)
yes
yes
yes
yes
57780
57780
57780
57746
0.484
0.488
0.517
0.511
Note: Standard errors in parentheses with a ,
cance at the 1%, 5% and 10% levels.
b
and
c
respectively denoting signifi-
entry costs on export markets. The remaining columns in Table (2) introduce the externality
variables, shown without and with (Table 3) the fixed cost variable. It is interesting to observe
that, while total employment in the commune does not seem to count, exporting employment
in the commune impacts positively on the export decision. The more workers employed in firms
that export, the higher is the probability that a previously non exporting firm starts exporting.
These results are to be compared with results in Table (3), in which the previous experience of
the firm at export has been added in all regressions. Adding this variable clearly increases the
explanatory power of the regression, which doubles between the two Tables. The firm’s productivity, measured by the sales per employee and by the size, while decreasing in magnitude,
are still important determinants of the firm’s export decision. The coefficients on the variables
representing externalities drop slightly in magnitude when using the past experience of the firm,
but stay positive and significative. The share of exporting employment in the region has more
impact on the export probability than the level, i.e the number of individuals that work in an
exporting firm.
Last section showed that firms that start exporting in a given year are, as the years go
by from 1986 to 1992, less and less located in dense areas. We want to study whether this
phenomenon is reflecting the decreasing importance of export spillovers as the process of EU
integration progressively eased exporting to EU markets. [To be continued: importance of
determinants through time]
9
Table 4: Logit on the Probability of Exporting, with country dimension
Model :
lcae
ln(wage)
ln(emp)
ln(dist)
ln(GDP)
Exported last year
Exported last year, same ctr
Exported last year, diff ctr
Dep. Dummies
N
R2
Dependent Variable: prob(exportijt )
(1)
(2)
(3)
(4)
a
a
a
0.872
0.542
0.598
1.027a
(0.006)
(0.008)
(0.009)
(0.007)
a
a
a
0.219
0.269
0.097
0.146a
(0.016)
(0.019)
(0.024)
(0.017)
0.982a
0.741a
0.623a
1.072a
(0.004)
(0.004)
(0.005)
(0.004)
-0.730a
-0.802a
-0.468a
-0.636a
(0.003)
(0.003)
(0.004)
(0.003)
0.535a
0.590a
0.336a
0.467a
(0.002)
(0.002)
(0.003)
(0.002)
a
2.822
(0.013)
3.939a
(0.009)
-1.618a
(0.008)
yes
yes
yes
yes
1469400 1264800 1264800
1469400
0.256
0.346
0.534
0.302
Note: Standard errors in parentheses with a ,
at the 1%, 5% and 10% levels.
5
b
and
c
respectively denoting significance
Exporting to Different Countries
Tables (4), (5) and (6) contain the results of the basic regressions done using the country
dimension. We begin with Table (4), which shows the results of the regression done without
and then with the past experience of the firm. In all regressions, firms’ characteristics have the
expected signs and appear significative. We computed three different variables representing the
firm’s export activity in year t − 1: having exported, having exported to the same country as
the country considered in year t, and having exported, but to a different country. The first
two variables appear with a positive and significative coefficient. It is not surprising to observe
the larger coefficient on the second variable, having exported to the same country. This result
may indicate that the fixed cost of exporting is likely to be heavily linked to the destination
market of the firm. This intuition is confirmed by the last column of Table (4), in which ‘having
exported, but to a different country’, appears with a negative and significative coefficient.
[To be continued]
6
Alternative Estimation Strategies
Table (7) [To be continued]
10
Table 5: Logit on the Probability of Exporting, with country dimension
Dependent Variable: prob(exportijt )
(1)
(2)
(3)
(4)
3.933a
3.897a
3.894a
3.820a
(0.009)
(0.009)
(0.009)
(0.009)
0.601a
0.641a
0.579a
0.648a
(0.009)
(0.010)
(0.010)
(0.010)
b
a
a
0.060
-0.124
0.063
-0.095a
(0.024)
(0.024)
(0.024)
(0.025)
a
a
a
0.605
0.580
0.581
0.628a
(0.005)
(0.005)
(0.005)
(0.006)
-0.469a
-0.399a
-0.474a
-0.246a
(0.004)
(0.004)
(0.004)
(0.004)
0.337a
0.284a
0.342a
0.169a
(0.003)
(0.003)
(0.003)
(0.003)
0.047a
(0.002)
0.213a
(0.002)
0.637a
(0.011)
0.811a
(0.005)
yes
yes
yes
yes
1264800 1264800 1264800 1264800
0.535
0.551
0.539
0.586
Model :
Exported last year, same ctr
ln(sales/emp)
ln(wage)
ln(emp)
ln(dist)
ln(GDP)
ln(exp. empl in c)
ln(exp. empl to same ctr in c)
ln(sh. of exp. empl in c)
ln(sh. of exp. empl, same ctr, in c)
Dep. Dummies
N
R2
Note: Standard errors in parentheses with a ,
at the 1%, 5% and 10% levels.
b
and
11
c
respectively denoting significance
Table 6: Logit on the Probability of Exporting, with country dimension, 13
Model :
Exported last year
ln(sales/emp)
ln(wage)
ln(emp)
common border
ln(dist)
ln(GDP)
ln(exp. empl in c)
ldist*ln(exp. empl in c)
ln(exp. empl to same ctr in c)
ldist*ln(exp. empl, same ctr in c)
ln(sh. of exp empl in c)
ldist*ln(sh exp empl in c)
ln(sh. of exp empl, same ctr, in c)
ldist*ln(sh exp empl, same ctr, in c)
European Economic Area
Dep. Dummies
N
R2
Note: Standard errors in parentheses with a ,
at the 1%, 5% and 10% levels.
Dependent Variable: prob(exportijt )
(1)
(2)
(3)
(4)
a
a
a
2.851
2.887
2.751
2.867a
(0.013)
(0.013)
(0.013)
(0.013)
0.553a
0.597a
0.547a
0.595a
(0.008)
(0.008)
(0.008)
(0.008)
0.236a
-0.009
0.252a
0.021
(0.019)
(0.019)
(0.019)
(0.020)
0.736a
0.701a
0.729a
0.729a
(0.004)
(0.004)
(0.004)
(0.005)
a
a
a
0.416
0.403
0.429
0.319a
(0.010)
(0.010)
(0.010)
(0.011)
a
a
a
-0.759
-0.674
-0.566
-0.430a
(0.011)
(0.008)
(0.005)
(0.006)
0.513a
0.459a
0.514a
0.339a
(0.003)
(0.003)
(0.003)
(0.003)
-0.195a
(0.010)
0.031a
(0.001)
-0.009
(0.009)
0.031a
(0.001)
1.362a
(0.067)
-0.118a
(0.009)
1.430a
(0.029)
-0.073a
(0.004)
Yes
Yes
Yes
Yes
yes
yes
yes
yes
1264800 1264800 1264800 1264800
0.354
0.384
0.356
0.442
b
and
12
c
respectively denoting significance
Table 7: Linear Probability model on the Probability of Exporting, without country dimension
Model :
oneago
ln(sales/emp)
ln(wage)
ln(emp)
(1)
Levels
0.651a
(0.003)
0.080a
(0.003)
-0.006
(0.008)
0.063a
(0.002)
ln(emp. in c)
ln(exp. emp. in c)
ln(sh. of exp. emp in c)
Dependent Variable: prob(exportit )
(2)
(3)
(4)
(5)
FE
FE
FE
FE
0.022a 0.022a 0.019a 0.018a
(0.005) (0.005) (0.004) (0.004)
0.075a 0.075a 0.071a 0.072a
(0.007) (0.007) (0.007) (0.007)
0.054a 0.054a 0.048a 0.047a
(0.013) (0.013) (0.012) (0.012)
0.110a 0.111a 0.088a 0.109a
(0.008) (0.008) (0.008) (0.008)
-0.005
(0.004)
0.093a
(0.002)
0.115a
(0.002)
ln(sh. of non exp. empl in c)
N
R2
57780
0.571
Note: Standard errors in parentheses with a ,
at the 1%, 5% and 10% levels.
57780
0.012
b
and
c
13
57780
0.012
57780
0.056
57780
0.066
respectively denoting significance
(6)
FE
0.020a
(0.004)
0.074a
(0.007)
0.050a
(0.012)
0.10a
(0.008)
-0.070a
(0.001)
57746
0.06
7
Conclusion
[To be continued]
References
Aitken B., G. H. Hanson and A. E. Harrison, 1997, “Spillovers, foreign investment, and
export behavior”, Journal of International Economics, 43(1-2): 103-132.
Barrios S., H. Görg and E. Strobl, 2003, “Explaining Firms’ Export Behaviour: R&D,
Spillovers and the Destination Market”, Oxford Bulletin of Economics and Statistics 65
(4): 475-496.
Bernard A. and J. B. Jensen, 2004, “Why Do Some Firms Export”, The Review of Economics and Statistics, forthcoming.
Bernard A. and J. Wagner, 1997, “Export and Success in German Manufacturing”,
Weltwirtschaftliches Archiv, 133 (1): 134-157.
Clerides A., S. Lach and J. Tybout, 1998, “Is Learning by Exporting Important? Microdynamic Evidence from Colombia, Mexico and Morocco”, Quarterly Journal of Economics,
113 (3): 903-947.
Dixit A. and J. Stiglitz, 1977, “Monopolistic Competition and Optimum Product Diversity”, American Economic Review, 67 (3): 297-308.
Eaton J., S. Kortum and F. Kramarz, 2004a, “Dissecting Trade: Firms, Industries and
Export Destinations”, American Economic Review, forthcoming.
Eaton J., S. Kortum and F. Kramarz, 2004b, “An Anatomy of International Trade:
Evidence from French Firms”, mimeo, CREST, New York University, University of Minnesota.
Greenaway D., N. Sousa and K. Wakelin, 2002, “Do Domestic Firms Learn to Export
from Multinationals?”, GEP Research Paper 2002/11, University of Nottingham.
Head K. and J. Ries, 2003, “Heterogeneity and the DI versus Export Decision of Japanese
Manufacturers”, Journal of the Japanese and International Economies, 17 (4): 448-467.
Krugman P. and A. J. Venables, 1995, “Globalization and the Inequality of Nations”,
Quarterly Journal of Economics, 110 (4): 857-880.
Lovely M., S. Rosenthal and S. Sharma, 2004, “Information, Exporting and the Agglomeration of U.S. Manufacturing Headquarters”, Regional Science and Urban Economics,
forthcoming.
Melitz M., 2004, “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity”, Econometrica, forthcoming.
Roberts M. J and J. R. Tybout, 1997, “The decision to export in Colombia: an empirical
model of entry with sunk costs”, American Economic Review, 87 (4): 545-564.
14
A
Distances and Exit Cities
Internal distance is calculated as the distance between a firm and the exit-city located on the
border between France and the destination market. Each country is associated with one exitcity (table A lists the exit-cities and the corresponding destination country). Internal distance
is computed as the great circle distance between the two locations. The location of a firm
is approximated to the INSEE code of the commune in which the firm resides. There are
approximately 35000 INSEE codes in our database. The great circle formula requires the use
of geographical coordinates; these are all obtained from the Geographic Names Information
System (GNIS) developed by the U.S. Geological Survey (US Department of the Interior).
External distances between the exit-city and the destination country use three different
procedures according to the countries involved. For nearby or particularly large countries (Austria, Canada, Hungary, Italy, Germany, Greece, Netherlands, Poland, Spain, United-Kingdom,
USA), we compute a weighted distance. We calculate bilateral distances between the exit-city
and the country’s regions and weight those distances by the economic size of the regions. The
external distance is then the sum of all regional weighted distances. For the economic size of
regions, we use regional population data at the NUTS 1 level (Nomenclature des Unités territoriales Statistique, Eurostat) for European countries, provided by the Regio database. For
Canada, population for provinces and territories was provided by Statistics Canada. For smaller
countries, we computed a non-weighted distance by using the great circle formula on capital’s
geographic coordinates (Ireland, Denmark, Portugal, Norway, Sweden, Finland, Turkey and
Mexico). For four of our countries, we applied a special procedure (Korea, Japan, Australia
and New-Zealand): We compute great circle distance but we impose a stop-over city because
traditional journeys to these countries are not supposed to travel through the poles.
Table 8: Countries and Exit-cities
Australia
Austria
Canada
Denmark
Finland
Germany
Greece
Hungary
Ireland
Italy
Japan
Korea
Le Havre
Geneva
Le Havre
Lille
Le Havre
Strasbourg
Marseille
Strasbourg
Le Havre
Nice
Le Havre
Le Havre
Mexico
Netherlands
New Zealand
Norway
Poland
Portugal
Spain
Sweden
Turkey
United-Kingdom
USA
15
Le Havre
Lille
Le Havre
Le Havre
Strasbourg
Pau
Perpignan
Le Havre
Marseille
Le Havre
Le Havre
Table 9: Summary Statistics on Firms Variables, 1989
Variable
Wage
Employment
Sales
Sales/empl
Nb obs.
9630
9630
9630
9630
Mean
18.04
82.34
8591.8
95.79
Std. Dev.
5.54
145.94
27892
120.81
Min
8.47
20
134.42
6.11
Max
118.08
5573
1499074
6605.8
Note: Wage, Sales and Sales per employee are in 1000 usd (1989), Employment in nb
of employees.
Table 10: Summary Statistics on Exporters, 1989
Variable
Wage
Employment
Sales
Sales/empl
Nb obs.
5188
5188
5188
5188
Mean
18.4
105.18
11830
109
Std. Dev.
5.43
166.8
28896
153.7
Min
8.47
20
134.4
6.11
Max
91.8
5456
1211615
6605
Note: Wage, sales and sales per employee are in 1000 USD (1989), employment
in number of employees.
Table 11: Summary Statistics on Non Exporters, 1989
Variable
Wage
Employment
Sales
Sales/empl
Nb obs.
4442
4442
4442
4442
Mean
17.6
55.6
4809
80.20
Std. Dev.
5.6
111.15
26171
60
Min
8.47
20
244
7.87
Max
118
5573
1499074
914
Note: Wage, sales and sales per employee are in 1000 USD (1989), employment
in number of employees.
Table 12: Percentage of Entrants and Exits, by Year
% Entrants
% Exits
87
0.1455
0.1078
88
0.1373
0.1092
89
0.1362
0.1055
90
0.1211
0.1204
91
0.1272
0.1124
Note: In percentage of the number of exporting firms in t (entrants),
in t − 1 (exits).
16
92
0.1264
0.1263
Table 13: Number and Percentage of Exporters, by Year
Firms
Exporters
% Exporters
1986
9630
4647
0.482
1987
9630
4852
0.503
1988
9630
5010
0.520
1989
9630
5188
0.538
1990
9630
5192
0.539
1991
9630
5280
0.548
1992
9630
5281
0.548
Table 14: Average Distance to Destination Country (kms)
Netherlands
United-Kingdom
Germany
Ireland
Italy
Austria
Spain
Denmark
Hungary
Poland
Portugal
Norway
637
762
782
981
1047
1056
1165
1240
1284
1358
1461
1703
17
Sweden
Greece
Finland
Turkey
Canada
USA
Mexico
Australia
New Zealand
Korea
Japan
1947
2166
2324
2829
6316
7246
9355
14697
15455
15486
16193
Table 15: Summary statistics on Industries, 1989
Industry
Iron and steel
Steel processing
Metallurgy
Minerals
Ceramic and building materials
Glass
Chemicals
Parachemistry
Pharmaceuticals
Foundry
Metal work
Agricultural machines
Machine tools
Industrial equipment
Mining and civil engeneering eqpmt
Electrical equipment
Electronical equipment
Domestic equipment
Transport equipment
Ship building
Aeronautical building
Precision instruments
Textile
Leather products
Shoe industry
Garment industry
Mechanical woodwork
Furniture
Paper & Cardboard
Printing and editing
Rubber
Plastic processing
Miscellaneous
Total
nb firms
Avg size
(nb empl)
share of
exporters
13
64
32
18
468
57
75
164
71
161
1746
131
186
890
168
272
251
22
301
28
35
213
754
129
157
575
348
326
310
858
84
477
246
487
80
103
283
58
162
128
93
153
119
61
94
66
66
86
88
83
427
168
94
190
79
89
56
125
80
61
79
96
71
127
79
86
0.69
0.56
0.69
0.61
0.31
0.58
0.87
0.80
0.80
0.73
0.46
0.77
0.75
0.48
0.69
0.58
0.51
0.86
0.65
0.32
0.66
0.74
0.65
0.71
0.64
0.44
0.42
0.54
0.57
0.32
0.85
0.70
0.69
18
sales/emp
exporters
(1000 usd)
232.9
136.6
262.7
118.8
84.7
93.2
216.6
155.1
194.3
79.3
82.9
101.9
92.6
115.8
104.2
94.6
103.9
104.1
109.6
174.8
73.9
88.6
122.9
95.5
57.5
117.3
91.3
84.2
131.1
162.4
81.2
118.1
93.8
sales/emp
non exporters
(1000 usd)
205.9
144.3
148.1
144.9
112.8
89.1
180.2
122.1
265.9
72.1
74.5
99.6
69.5
79.9
79.8
76.9
71.0
95.2
87.3
65.6
48.8
62.0
54.3
65.1
42.2
29.7
80.8
71.3
97.9
102.9
54.5
91.6
67.4