Intereconomics, May/June 2008
COMPETITIVENESS
135
DOI: 10.1007/s10272-008-0247-x
Thierry Mayer* and Gianmarco I. P. Ottaviano**
The Happy Few: The Internationalisation
of European Firms
New Facts based on Firm-level Evidence
Policymakers tend to view the internationalisation of firms from the perspective of export,
import and FDI statistics. A lack of statistical information at the firm level has so far
prevented the systematic inclusion of firm-level analysis in the policymaker’s standard
toolbox. Some firm-level datasets are now available, however, and their analysis reveals
some new facts that are simply unobservable at the aggregate level.
Internationalisation is an elusive concept. From the
point of view of a policymaker it refers to the presence
of countries in international markets as measured
by their shares of exports, imports and FDI. From
the point of view of a manager, it refers to the ability
of firms to generate value through international opera-
tions.
Though complementary, the two points of view
are typically considered separately. Policymakers
fret about aggregate exports, imports and FDI. Their
preferred perspective is sectoral. Managers are concerned
that international operations, whether through
exports, imports or FDI, bring additional costs with
respect to domestic activities and these costs generate
barriers that only some firms are able to overcome.
Their preferred perspective is that of their own firms.
The separation between the two perspectives is due
to different objectives and interests but also to different
mindsets. Managers like case studies and exemplary
evidence. Policymakers like statistical information. The
lack of such information at the firm level has therefore
so far prevented the systematic inclusion of firm-level
analysis in the policymaker’s standard toolbox.
This paper argues that the time is ripe to supplement
the policymaking toolbox: firm-level datasets are
now available and provide new information that one
cannot afford to ignore. Interestingly, the statistical
analysis at the firm level reconciles the policymaker’s
and the manager’s points of view.
In particular, the analysis of firm-level evidence reveals
some new facts that are simply unobservable at
the aggregate level:
The evolution of aggregate exports, imports and FDI•
is driven by the changes in two “margins”. The “intensive
margin” refers to average exports, imports
and FDI per firm. The “extensive margin” refers to
the number of firms actually involved in those international
activities (“internationalised firms”, henceforth
IFs).
The “extensive margin” is much more important,•
as the reaction of aggregate trade and FDI flows to
country fundamentals takes place mostly through
that margin. This is impossible to see without firm
level data and thus has not been seen so far.
The “extensive margin” is thin. IFs are rare and their•
distribution is highly skewed, as a handful of firms
accounts for most aggregate international activity.
The “extensive margin” is an exclusive club. IFs are•
different from other firms. They are bigger, generate
higher value added, pay higher wages, employ more
capital per worker and more skilled workers, and
have higher productivity.
To sum up, the international performance of a country
is driven by a handful of high-performance firms.
Hence, from a policy perspective, successful internationalisation
is much more about increasing the
number of firms involved than about increasing the
involvement of already active firms. However, in order
to increase the number of firms involved, policies fostering
firm performance in terms of employment and
* Professor, University of Paris 1; scientific advisor, CEPII, Paris,
France; research affiliate, CEPR, London, UK.
** Professor, University of Bologna, Italy; senior fellow, Bruegel, Brussels,
Belgium; research fellow, CEPR, London, UK.
This paper is based on the EFIM Report 2007, issued under the same
title as Bruegel Blueprint Series Volume III. EFIM is a network of eight
research centres from eight European countries created in 2006 under
the coordination of Bruegel and CEPR to work on policy-relevant
questions concerning the internationalisation of European firms that
are best treated using firm-level trade and FDI data (http://www.
bruegel.org/Public/SimplePage.php?ID=1720). The authors are grateful
to Laurent Eymard and Andrew Fielding for editorial support. Opinions
expressed in this publication are those of the authors alone.
COMPETITIVENESS
Intereconomics, May/June 2008136
productivity are more important than policies fostering
exports, imports or FDI per se.
The paper is organised as follows. Following on
from the introduction, we first show that IFs are rare
and their exclusive club is dominated by a handful of
top firms. We then demonstrate that IFs are different
in that they perform better than other firms. Following
that, we dissect aggregate trade and FDI flows to assess
the relative importance of the intensive and extensive
margins. It is shown that firm-level information
is crucial to understanding aggregate behaviour. Finally,
we summarise the evidence and discuss its policy
implications.
A final caveat. Firm-level data are typically collected
independently either from balance sheets or from
surveys by different public authorities or research institutions
in different countries. The lack of harmonisation
or coordination among the different players is
all but natural. Nonetheless, it prevents the creation
of a homogenous cross-country dataset. The result
is that only very few policy-relevant questions can be
addressed systematically across all countries. Rather
than limiting our attention to those very few questions,
we have chosen to cover a larger range of issues by
selecting for each issue the best available national da-
tasets.
Superstar Exporters
In the following we use firm level data to show that
IFs are few and, among these few, only a handful of
firms account for the bulk of aggregate exports and
FDI.
Let us focus on trade and rank a country’s firms in
terms of their individual exports. Table 1 reports the
contributions of different segments of the ranking to
aggregate exports in the cases of Belgium, France,
Germany, Hungary, Italy, Norway and the UK. The Belgian
and Norwegian samples include all firms and are
therefore exhaustive. The British, German, Hungarian,
and Italian samples cover only relatively large firms and
are therefore restricted. The French data provide both
an exhaustive sample and a restricted sample comparable
to the British, German, Hungarian, and Italian
ones. We mainly use the restricted sample, which provides
more detailed data. Where possible, however,
we also give results from the exhaustive sample.
For each country the columns in Table 1 show the
contributions of the top 1%, 5% and 10% of exporters.
The numbers are striking. In the exhaustive samples,
the top 1% of exporters account for more than
45% of aggregate exports; the top 5% of exporters
account for more than 70% of aggregate exports; the
top 10% of exporters account for more than 80% of
aggregate exports. The results for Germany, Hungary,
Italy and the UK are less extreme. However, comparing
the exhaustive and restricted samples for France
suggests that the focus of those countries’ datasets
on relatively large firms explains such a finding.
This feature of internationalisation is further investigated
in Figure 1 in the case of France using the restricted
sample. The pale grey curve plots the actual
distribution of exports: exporters are ranked from left
to right, starting with the biggest, along the horizontal
axis, with their cumulative contribution to aggregate
exports measured along the vertical axis. The contributions
of the top 1%, 5% and 10% exporters are the
ones already reported in Table 1. As a benchmark, the
straight grey line plots a distribution corresponding
Table 1
Share of Exports of Top Exporters in 2003,
Total Manufacturing
Note: France, Germany, Hungary, Italy and the UK have large firms
only; Belgian and Norwegian data is exhaustive. Numbers in brackets
for France are percentages from the exhaustive sample.
Source: EFIM.
Country of origin top 1% top 5% top 10%
Germany 59 81 90
France 44 (68) 73 (88) 84 (94)
United Kingdom 42 69 80
Italy 32 59 72
Hungary 77 91 96
Belgium 48 73 84
Norway 53 81 91
Figure 1
The Superstar Exporters Phenomenon
(France, restricted sample)
100
percentileofemploymentandexports
0 20 40 60 80 100
percentile of French firms
80
60
40
20
0
uniform distribution actual employment distribution
actual exports distribution
Source: EFIM.
Intereconomics, May/June 2008
COMPETITIVENESS
137
to the case in which all firms export the same value.
Hence, the further away the curve is from the line, the
more concentrated aggregate exports are in the hands
of a few firms. Using the restricted sample, we can plot
a similar distribution for employment (in black) as an
interesting benchmark. Figure 1 shows that the concentration
is high in terms of employment (the black
curve is far from the uniform distribution), but is much
higher in terms of exports.
In addition Figure 2 zooms onto the contributions of
“superstar” exporters by showing what happens within
the club of the top 1% of exporters.1
The picture is
again striking: the top 0.001%, 0.01% and 0.1% of exporters
still account for not much less than 10%, 20%
and 40% of aggregate exports respectively.
For Europe in general, we can summarise the findings
as:
1
As we focus here on a smaller number of firms, we need to use the
exhaustive sample to obtain a representative distribution. The logarithmic
transformation is used to enhance the readability of the picture.
Fact 1 – Aggregate exports are driven by a small
number of top exporters. The top 1%, 5% and 10%
of exporters account for no less than 40%, 70%
and 80% of aggregate exports respectively.
Export Intensity
The fact that only a handful of firms drive aggregate
exports suggests that export status is a mixed bag
containing different types of firms.
Table 2 shows that the share of sampled firms that
export is roughly 65%, 60%, 45%, 75% and 40% for
France, Germany, Hungary, Italy and Norway respectively.
The higher percentages for France, Germany
and Italy reflect the biases of these samples towards
relatively large firms. For each country the table reports
the percentages of firms exporting more than
the given shares of their turnover, and the percentages
of total exports accounted for by these groupings of
firms.
The results for France, Italy and Norway are similar.
They show that, even though only a small subset of
firms exports a major share of their turnover, they still
account for a large fraction of total exports. In France,
Germany and the United Kingdom, around 10% of all
firms export more than 50% of their turnover but they
account for 50% to 75% of total exports. The distribution
can, however, vary substantially across countries.
In this respect, an interesting comparison between
France and Germany exemplifies the potential of firmlevel
data analysis. Germany has a larger proportion
of firms exporting more than 50% of their turnover,
and they represent a much larger share of total exports
than in France. From Table 2 we can see that
the greatest contribution (68%) to total exports in Germany
comes from firms exporting from 50% to 90% of
their turnover. In France on the contrary, the greatest
contribution (46%) comes from firms exporting from
Figure 2
The Superstar Exporters Phenomenon, Logarithmic
Transformation (France, Exhaustive Sample)
100
percentageofFrenchexports
.001 .01 .1 1 2 345 10 50 100
percentile of exporters
1998 2003
80
60
20
40
10
5
N ote: Germany, Italy, Hungary, the United Kingdom and France have large firms only; Norwegian data are exhaustive.
% firms exporting more than % of total exports by firms exporting more than
Country of origin # firms
total mfg
exports
(€ bn)
% export-
ers
5% of
turnover
10% of
turnover
50% of
turnover
90% of
turnover
5% of
turnover
10% of
turnover
50% of
turnover
90% of
turnover
Germany 48325 488.66 59.34 46.89 40.30 11.85 0.96 99.49 98.54 73.57 5.95
France 23691 171.73 67.30 41.16 33.04 9.02 1.39 93.58 95.11 49.22 9.71
United Kingdom 14976 71.46 28.33 22.52 19.27 8.07 1.51 97.60 93.40 65.70 19.00
Italy 4159 58.61 74.44 64.90 57.42 25.58 2.91 99.71 98.53 69.09 7.52
Hungary 6404 30.01 47.53 38.43 34.74 22.19 11.01 99.86 99.64 92.01 69.13
Norway 8125 16.07 39.22 17.98 14.45 5.19 1.26 98.51 97.42 70.27 28.57
Table 2
Distribution of our Sample of Exporters by % of Turnover, 2003
COMPETITIVENESS
Intereconomics, May/June 2008138
10% to 50% of their turnover. France, however, has a
larger proportion of firms entirely “globalised” (selling
more than 90% of their turnover abroad) and the share
of total exports by those is almost twice as large as
for Germany. This echoes other findings showing that
one of the strengths of Germany’s industrial structure
compared to France lies in the larger set of mediumsized
firms heavily involved in exporting.2
Panels (a) and (b) of Figure 3 illustrate this finding
for the entire distribution of firms in two years, 1998 and
2003 respectively. Although this type of cross-country
comparison should be read with great caution, it
seems indeed to be the case that over time the big
divergence between the performance of French and
German exports stems from the middle range of firms.
In 1998, the two distributions look quite similar, with
2
P. Artu s, L. Fontagné: Évolution récente du commerce extérieur
français, rapport n°64 du Conseil d’Analyse Economique, Paris 2007,
la Documentation Française. Note that this finding must be taken with
caution as none of the two French and German datasets we use are
exhaustive. The criteria used by the statistical institutes for sampling
firms seem fairly comparable, however.
France having slightly more of both very small and very
large exporters. In 2003 the picture is quite different,
with Germany outperforming France for middle-size
exporters by a fairly large margin. Whether this change
in distribution can explain the drastic differences in export
performance of the two countries over the same
period is an open question calling for deeper investi-
gation.
For Italy, 3% and 25% of firms export more than
90% and 50% of their turnover and account for roughly
7% and 70% of total exports respectively. For Norway,
around 1% and 5% of firms export more than
90% and 50% of their turnover and account for roughly
30% and 70% of total exports respectively.
Hungary is somewhat different. Around 10% and
22% of Hungarian firms export more than 90% and
50% of their turnover and account for roughly 70%
and 90% of total exports respectively. This reveals
that a large fraction of Hungarian firms is involved in
intense international activity, probably owing to Hungary’s
role as Germany’s industrial backyard.
The previous section implies:
Fact 2 – Only a few firms export a large fraction of
their turnover. Around 5% and 25% of firms export
more than 90% and 50% of their turnover and account
for roughly 10% and 70% of total exports re-
spectively.
Comparing these percentages with the ones reported
in Table 1 reveals that the fraction of firms with top
export intensity is larger than the fraction of top exporters.
Accordingly, top exporters do not necessarily
exhibit top export intensity.
Meet the “Margins”
A handful of firms account for a disproportionate
share of aggregate exports. These firms, however, do
not necessarily export large fractions of their turnover.
Hence, their turnover has to be large. Table 3 provides
additional information on these superstar exporters.
The table refers to France but, as seen in the above,
the different countries in our sample are remarkably
similar once the different compositions across countries
(exhaustive or restricted sample) have been taken
into account.
The top panel of the table reports the percentages
of firms exporting given numbers of products (rows) to
given numbers of markets (columns). The table reveals
a bipolar pattern as the largest percentages of firms
are concentrated in the top left-hand and bottom righthand
cells. In particular, 30% of firms export only one
Figure 3
Export Intensity: France versus Germany
75
shareoffirms
.5 1 5 10 50 75100
share of turnover exported (more than)
Germany France
50
15
5
5
10
50
75
shareoffirms
.5 1 5 10 50 75 100
share of turnover exported (more than)
Germany France
Source: EFIM.
a. 1998
b. 2003
Intereconomics, May/June 2008
COMPETITIVENESS
139
product to only one market while 10% of firms export
more than ten products to more than ten markets.
The bottom panel reports the shares of aggregate
exports due to firms exporting given numbers of products
(rows) to given numbers of markets (columns).
The bipolar pattern is not there: firms exporting more
than ten products to more than ten markets account
for more than 75% of total exports.
Comparing the two panels then yields:
Fact 3 – Top exporters export many products to
many locations. Firms exporting more than ten
products to more than ten markets account for
more than 75 % of total exports.
To summarise, aggregate exports are determined
by a few top exporters that are relatively big and supply
several foreign markets with several differentiated
products. This points to the existence of a process
through which only firms that are large enough and
have a rich enough portfolio of products can withstand
international competition. We shall explore below the
characteristics that make exporters, and a fortiori top
exporters, different from other firms. We shall refer to
such differences as “exporters’ premia”.
As to market coverage, most naturally the larger the
number of markets a firm serves, the larger their average
distance from the firm’s country of origin. Table 3
suggests that distance affects aggregate trade flows
mostly by reducing the number of exporters rather
than by reducing average exports per firm. We shall
compare the two effects in some detail later on. We
shall refer to the former as the adjustment of aggregate
exports along the “extensive margin” and to the
latter as their adjustment along the “intensive margin”.
In this respect, as many trade barriers are typically
correlated with distance, Table 3 suggests that the impact
of trade policy should materialise mainly through
changes in the extensive margin.
The Talent of Internationalised Firms
We shall now show that internationalised firms (IFs)
score better than other firms on various performance
measures.
Table 4 reports employment, value added, wages,
capital intensity and, where available, skill intensity
“premia” defined as the ratios of exporters’ (FDI-makers’)
to non-exporters’ (non FDI-makers’) values.
Share of Exports
(total exports: 314.3 € bn)
Table 3
Distribution of French Exporters over Products
and Markets, 2003
# of countries
# of products 1 5 10+ Total
1 29.61 0.36 0.22 34.98
5 0.76 0.45 0.62 4.73
10+ 0.95 0.89 10.72 18.57
Total 42.59 4.12 15.54 100
Share of Exporters
(total # exporters: 99259)
S ource: EFIM.
# of countries
# of products 1 5 10+ Total
1 0.70 0.08 0.38 1.86
5 0.30 0.08 1.06 1.97
10+ 0.28 0.45 76.3 81.36
Total 2.85 1.55 85.44 100
Table 4
Exporters and FDI-makers Exhibit Superior
Performance
Country of
origin
Employ-
ment
premia
Value added
premia
Wage
premia
Capital
intensity
premia
Skill
intensity
premia
Exporters premia:
Germany
2.99
(4.39)
1.02
(0.06)
France
2.24
(0.47)
2.68
(0.84)
1.09
(1.12)
1.49
(5.60)
United
Kingdom
1.01
(0.92)
1.29
(1.53)
1.15
(1.39)
Italy
2.42
(2.06)
2.14
(1.78)
1.07
(1.06)
1.01
(0.45)
1.25
(1.04)
Hungary
5.31
(2.95)
13.53
(23.75)
1.44
(1.63)
0.79
(0.35)
Belgium
9.16
(13.42)
14.80
(21.12)
1.26
(1.15)
1.04
(3.09)
Norway
6.11
(5.59)
7.95
(7.48)
1.08
(0.68)
1.01
(0.23)
FDI- makers premia:
Germany
13.19
(2.86)
France
18.45
(7.14)
22.68
(6.10)
1.13
(0.90)
1.52
(0.72)
Belgium
16.45
(6.82)
24.65
(11.14)
1.53
(1.20)
1.03
(0.82)
Norway
8.28
(4.48)
11.00
(5.41)
1.34
(0.76)
0.87
(0.13)
Note: The table shows premia of the considered variable as the ratio
of exporters over non-exporters (standard deviation ratio in brackets).
France, Germany, Hungary, Italy and the United Kingdom have large
firms only; Belgian and Norwegian data are exhaustive.
Source: EFIM.
COMPETITIVENESS
Intereconomics, May/June 2008140
Table 5 presents, instead, two measures of productivity
for French exporters. Revenue per worker is recorded
as “apparent labour productivity”. “Total factor
productivity” (TFP) refers to the estimated productivity
of all inputs taken together and is a measure of the
global efficiency of a firm.
The message conveyed by the two tables is clear:
in all countries and on all counts exporters are generally
better performers. The difference is particularly
pronounced for employment and value added. There
is, nonetheless, some variation across countries. For
example, exporters’ premia are significantly lower for
France (2.4 and 2.6) and Italy (2.2 and 2.1) than for Belgium
(9.1 and 14.8) and Norway (6.1 and 7.9). This is
probably due to the fact that the French and the Italian
datasets feature relatively large firms only, which
gives highly selected samples of non-exporters. The
wage premium is, instead, consistently smaller but still
exporters tend to pay wages that are 10-20% higher
than non-exporters.
The employment premium for German exporters is
in line with those of France and Italy. The United Kingdom
employment premium for exporters, on the other
hand, is almost zero, which is a puzzling exception
compared to all other countries and indicators. This
probably derives from the fact that the sample of UK
firms is even more biased than the others in favour
of large firms.3
Given that its sample is also restricted
to large firms, Hungary is an outlier (as it was also in
terms of the percentage of firms that export more than
90% of their turnover). Quite large premia characterise
employment (5.3), value added (13.5) and wages
(1.44). Capital intensity and productivity, on the other
hand, feature rather low premia.
The analysis can be refined by comparing firms that
not only export but also invest abroad with those that
only export or only operate in their domestic markets.4
Figure 4 shows the productivity distributions for the
three types of firms in Belgium. Panel (a) depicts apparent
labour productivity whereas panel (b) refers to
estimated TFP.
3
Sample selection is less likely to explain the cross-country behaviour
of FDI premia as French premia are quite large.
4
In our samples, nearly all FDI-makers are also exporters.
Table 5
French Exporters Exhibit Superior Performance to
French Non-exporters, 2003
Note: The firms considered are manufacturers with more than 20
employees. The table shows premia of the considered variable as the
ratio of exporters over non-exporters. Numbers in brackets are the
ratio of the standard deviation.
Source: EFIM.
Industry
Apparent labour
productivity
Estimated TFP
(Olley-Pakes)
Total manufacturing 1.31 (6.11) 1.15 (4.09)
Food and beverages 1.27 (2.12) 1.21 (1.86)
Textiles 1.53 (3.76) 1.48 (2.94)
Wearing apparel 2.52 (8.04) 1.87 (3.06)
Leather and shoes 1.27 (1.57) 1.06 (1.27)
Wood and wood products 10.37 (497.82) 5.89 (264.51)
Paper and paper products 1.19 (1.25) 1.01 (0.80)
Printing and editing 0.90 (0.17) 1.03 (0.31)
Coke and refined petroleum 6.75 (46.33) 0.47 (0.54)
Chemicals 0.78 (0.44) 0.74 (0.45)
Rubber and plastics 1.08 (0.58) 1.01 (0.58)
Non-metallic minerals 0.98 (1.28) 0.91 (1.27)
Metals 1.19 (1.09) 1.12 (1.03)
Metal products 1.12 (1.11) 1.05 (1.04)
Machinery and equipment 1.11 (1.47) 1.05 (1.38)
Office machines 1.82 (8.23) 1.83 (8.02)
Electrical equipments 1.22 (1.49) 1.11 (1.40)
Radio-TV-communication 1.31 (1.95) 1.17 (1.78)
Precision instruments 1.21 (1.50) 1.10 (1.45)
Motor vehicules 1.23 (1.40) 1.11 (1.59)
Other transport 1.32 (1.73) 1.14 (1.60)
Furniture 1.29 (5.85) 1.21 (3.67)
Recycling 1.01 (0.71) 0.98 (0.94)
Figure 4
Belgian FDI-makers are more productive than
Belgian exporters
Note: Data for Belgium 2004.
Source: EFIM.
Intereconomics, May/June 2008
COMPETITIVENESS
141
For the three types of firm, each panel shows the
share of firms (“density”) that attain each productivity
level. In other words, the panels depict the probability
of picking a firm with a certain productivity level when
the firm is randomly drawn from each type. The two
panels send the same message: a randomly drawn
FDI-maker is likely to be more productive than a randomly
drawn exporter, which in turn is likely to be more
productive than a randomly drawn domestic firm. This
type of finding is not specific to Belgium, and has also
been shown to exist for Italian exporters compared to
domestic Italian firms.5
We have therefore established:
Fact 4 – FDI-makers perform better than exporters
and exporters perform better than non-exporters.
Exporters are generally bigger, more profitable,
more capital intensive, more productive and pay
higher wages than non-exporters. By the same
measures, FDI-makers perform better than export-
ers.
Exporters are also different along an additional dimension.
In particular, Table 6 shows that they are
more likely to be foreign owned. This phenomenon
is more pronounced when the complete population
of firms is available (Belgium) than when only large
firms are sampled (Hungary, Italy or the UK). In Hungary,
where foreign ownership is much more common,
exporters are still four times more likely to be foreign
owned. The associated Figure 5 depicts the evolu-
5
http://www.cepr.org/pubs/PolicyInsights/PolicyInsight8.pdf.
tion of these figures over time. Hungary and the UK
are quite stable in having a very large share of foreignowned
exporters, while foreign ownership is rising fast
in Belgium and Italy.
Hence, we have:
Fact 5 – Exporters are more likely to be foreign
owned.
Learning by Exporting and Investing Abroad?
Exporters are better than non-exporters over a
broad spectrum of performance measures. An interesting
issue is whether their superior performance
predates their access to export markets or, rather,
their performance improves as a result of their access
to export markets.
This chicken-and-egg question is presented for
France and Norway in Figure 6 and Figure 7 respectively.
The figures consider firms in the samples that
Table 6
Share of Foreign-owned Firms among Exporters
and Non-exporters in 2003 (%)
Note: United Kingdom, Italy and Hungary have large firms only, Belgian
data are exhaustive.
Source: EFIM.
Country of origin non exporters Exporters
United Kingdom 18.69 27.94
Italy 4.03 10.26
Hungary 11.47 43.63
Belgium 0.58 12.23
Figure 5
The Rising Foreign Ownership of European
Exporters
(share of foreign-owned exporters in %)
50
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
Belgium Italy Hungary United Kingdom
40
30
20
10
S ource: EFIM.
Figure 6
Compared Performance in Labour Productivity –
Exports (France)
Source: EFIM.
COMPETITIVENESS
Intereconomics, May/June 2008142
became exporters during the period of observation
and that were observed for four years after switching
status (“switchers”). It then compares their behaviour
with that of all other firms (“non-switchers”). In particular,
the comparison is made in terms of value added
per worker a given number of years after the firms first
began exporting.
The two figures show that switchers do move along
steeper trajectories as they perform increasingly better
than non-switchers. This is true no matter whether
they already performed better in the switch year
(France) or not (Norway). Two very different stories
are consistent with those findings. Since we do not
observe what happened before the switch, perhaps
the switchers were already on a better trajectory, so
gaining export status was simply the outcome of an
already promising performance (“selection into export
status”). On the other hand, perhaps the switchers
were no different from other firms before switching,
but gaining export status as a result of some temporary
shock allowed them to learn from international
activity (“learning-by-exporting”). Data for Germany
are also available but only allow one to calculate the
performance ratios of switchers over non-exporters.
We compute those ratios for the three countries and
depict them in Figure 8.
While the labour productivity of firms switching to
exporter status is generally greater than that of nonexporters
one year or more after switching, the pattern
over time is not clear. The advantage increases steeply
for Norway but much less so for France and does not
show any clear trend in the case of Germany.
Only the Norwegian data lend themselves to a study
of the behaviour of firms that start to invest abroad
during the period of observation and that are then observed
for the four next years (“switchers”). Figure 9
compares their behaviour with that of all other firms
(“non-switchers”) in terms of value added per worker
a given number of years after the firms first started to
make FDI. The pattern is U-shaped, with switchers underperforming
in the first three years and overperforming
in the fourth year after switching.
Overall, we have:
2
ratioofvalueaddedperworker
switch year switch+1 switch+2 switch+3 switch+4
Germany France Norway
1.5
1
0.5
0
Figure 8
Compared Performance Ratio in Labour
Productivity for Three Countries – Exports
Source: EFIM.
Figure 9
Compared Performance Ratio in Labour
Productivity – FDI (Norway)
Source: EFIM.
Figure 7
Compared Performance in Labour Productivity –
Exports (Norway)
Source: EFIM.
Intereconomics, May/June 2008
COMPETITIVENESS
143
Fact 6 – There is no clear evidence of firms performing
differently after accessing foreign markets.
While the performance of firms that start exporting
is generally better than that of non-exporters one
year or more after starting to export, the pattern over
time is not clear. The picture is even more blurred in
the case of firms that start to invest abroad.
The Margins of Exports and FDI
This section breaks down aggregate exports and
FDI into their fundamental drivers. It shows that the
most important channel through which these drivers
affect aggregate flows is the “extensive margin”, i.e.
the number of IFs.
The single most robust way to relate aggregate
trade and FDI flows to their fundamental drivers is the
“gravity equation”. This relates the values of flows between
two economies to their sizes and a variety of
trade impediments.6
While this relationship works in
6
The theoretical foundations of this empirical relationship have
emerged late in time compared with the vast number of empirical applications
of gravity. In the last ten years a wide range of theoretical
explanations for gravity have become available (see J.E. Anderson,
E. van Wincoop: Trade Costs, in: Journal of Economic Literature,
Vol. 42, 2004, pp. 691-751 for a survey), and researchers such
as Chaney, Helpman et al. and Melitz and Ottaviano have started to
investigate the importance of firms’ heterogeneity for gravity. (Cf. T.
Ch an ey: Distorted Gravity: The Intensive and Extensive Margins of
International Trade?, in: American Economic Review, 2006, forthcoming;
E. Helpman, M.J. Melitz, Y. Rubenstein: Estimating
Trade flows: Trading Partners and Trading Volumes, in: Quarterly Journal
of Economics, 2007, forthcoming; M. Melitz; G. Ottaviano:
Market Size, Trade, and Productivity, in: Review of Economic Studies,
Vol. 75, 2008, pp. 295–316). On the empirical side, authors such as J.
Eaton , S. Kortu m and F. Kramarz: An Anatomy of International
Trade: Evidence from French Firms, in: Meeting Papers from Society
for Economic Dynamics, No. 197, 2005; and A.B. Ber nard,
J.B. Jen sen, S.J. Redding and P.K. Schott: Firms in International
Trade, in: Journal of Economic Perspectives, Vol. 21, No. 3,
2007, have investigated those issues for US firms and French firms.
the case of both exports and FDI, for ease of presentation
we shall initially focus on trade flows and deal with
FDI later.
Aggregate data show that bilateral trade flows are
positively affected by countries’ sizes and negatively
affected by trade impediments. As some trade impediments
increase with the distance between countries,
this result is reminiscent of Newton’s law of gravitational
attraction, whence the name “gravity equation”.
Through which channels does gravity determine bilateral
trade flows? First of all, gravity may affect the
number of exporters (“firm extensive margin”). Then,
it may affect the average exports per exporter (“firm
intensive margin”). It may also affect the number of
products exported (“product extensive margin”), and
the average exports per firm of each product (“product
intensive margin”). Finally, gravity may affect export
prices (“price margin”) and exported quantities
(“quantity margin”) in different ways. To handle this
complexity in a consistent way, we decompose the
simple gravity equation into increasingly finer detail,
relying on firm-level information. The logic of this decomposition
is visualised in Figure 10.
Let us begin with the decomposition in terms of firm
extensive and intensive margins. In other words, we
ask: do spatial separation, differences in language,
currencies and so on hinder trade flows by limiting the
entry of exporters (“firm extensive margin”) or rather
by constraining the volumes exported by firms (“firm
intensive margin”)?
The decomposition of exports into extensive and intensive
margins can be carried out in a similar fashion
for the French and Belgian data, which both provide
near-exhaustive data for exports over a very comparable
set of years. Furthermore, we are able to compute
for both countries not only the average export value
per firm, but also the number of products exported,
the average quantity (in kilograms) and therefore the
unit value for each product.7
We start with the most simple decomposition exercise,
which contains only distance as a trade impediment.
Figure 11 presents the results.8
The bar chart
represents the contribution of firm extensive (“number
of exporters”) and intensive (“average exports”) mar-
7
We can thus go even further than existing margin decomposition
on US internal (R. Hillberry, D. Hummels: Trade Responses
to Geographic Frictions: a Decomposition Using Micro-Data, in: European
Economic Review, 2005) or external (B. Ber nard et al., op.
cit.) data. Another early paper decomposing trade patterns into the
extensive and intensive margins is J. Eaton et al., op. cit., using
French data for the year 1986.
8
All coefficients are highly significant.
Figure 10
The Margins of Adjustment of Aggregate Exports
Firm extensive margin Product extensive margin Quantity margin
Number of
exported
products
Exports per
product per
exporter
Quantity exported
per product
per exporter
Export price per
product per
exporter
Number of
exporters
Value
exported
per
exporter
Aggregate
value of
exports
Firm intensive margin Product intensive margin Price margin
S ource: EFIM.
COMPETITIVENESS
Intereconomics, May/June 2008144
gins to the overall effects (small diamonds) of three
gravity forces on bilateral exports: the size of the exporting
country (“GDP, ex”), the size of the importing
country (“GDP, im”) and distance (“Dist.”)
The overall effects are extremely standard: close to
one for GDPs and close to -0.9 for distance. In other
words, if country A is 10% larger than country B, then
on average it attracts 10% more exports than B from
other countries. Analogously, country A exports on average
10% more than B to other countries. Moreover,
if A is on average 10% further away from other countries
than B, then it trades 9% less than B with those
countries.
More interestingly, the results of the decomposition
show that the reaction of the firm extensive margin of
trade to gravitational forces is much greater than that
of the intensive margin. For instance, the decrease in
the number of firms accounts for 75 % of the impact of
distance on trade flows. In the same vein, the increase
in trade value associated with the increase in the importing
country’s size comes mostly (60%) from the
increase in the number of exporters to the country in
question. Note also that the entire effect of the exporting
country’s size on trade comes from the number of
its exporting firms.9
More detailed estimates also allow us to identify interesting
differences in the effects of different trade im-
9
This is exactly what should be expected from most theoretical foundations
of the gravity equation and, in particular, from the ones with
differentiated products and imperfect competition, whether with heterogenous
firms (T. Chaney, op. cit.; E. Helpman et al., op. cit.;
M . M elitz, G. Ottaviano, op. cit.) or not. (S. Redding, A. Ven
ables: Economic Geography and International Inequality, in: Journal
of International Economics, Vol. 62, 2004, pp. 53-82).
pediments. Sharing a language increases the number
of exporters and does not affect the average amount
exported. GATT/WTO membership and colonial links
increase the number of exporters and reduce the average
amount exported. This evidence is compatible
with the notion that being a member of GATT/WTO
and having linguistic or colonial links tends to reduce
the fixed costs of exporting rather than the variable
ones.
We have thus established:
Fact 7 – The number of exporters matters the most.
The change in the number of exporting firms accounts
for most of the negative impact of trade
barriers and most of the positive impact of the importing
country’s size on bilateral exports. The increase
in the number of exporting firms accounts
entirely for the positive impact of the exporting
country’s size on bilateral exports.
Product Margins
In datasets where the information is available, a further
decomposition makes it possible to assess how
the number of products exported by firms varies with
different barriers to trade.
Figure 12 displays the results of this new decomposition.
The bar chart represents the contribution
of the firm extensive margin (“number of exporters”),
the product extensive margin (“number of products”)
and the product intensive margin (“average exports
per product by firm”) to the overall effects (small diamonds)
of three gravity forces on bilateral exports.
Strikingly, the results point to an extreme parallelism
Figure 12
Gravity and Aggregate Exports – II
-0.86
0.93
1.05
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
GDP, ex GDP, im Dist.
Number of exporters
Average export per product by firm
Number of products
Overall effect
Source: EFIM.
Figure 11
Gravity and Aggregate Exports – I
Source: EFIM.
1
0.5
0
0.5
1
1.5
GDP, ex GDP, im Dist.
0.86
0.93
1.05
Number of exporters
Average Export
overall effect
Intereconomics, May/June 2008
COMPETITIVENESS
145
in the firm extensive margin and the product extensive
margin. Together, these two margins would imply that
the effect of the size of, and distance between, exporting
and importing countries is much greater than the
estimated total effect. This is because, as shown by
the corresponding parts of the bars, the effect of these
three factors on exports is mitigated by their effect on
average exports per product by firm. Indeed, the average
exports per product by firm fall with GDP and
rise with distance. In particular, a 10% increase in the
GDP of the exporting country leads to an increase of
roughly 10% in both the number of exporters and the
number of products exported as well as a decrease of
roughly 10% in firms’ average export per product. A
10% increase in bilateral distance leads to a 6% fall
in the first two variables and to a 4% increase in the
third.10
These findings establish:
Fact 8 – The number of exported products matters
too. Larger countries have more exporters, export
more products and their exporters have smaller average
exports per product. An increase in bilateral
trade barriers reduces the positive effects of country
size on the number of exporters and products. It
also reduces the negative effect of country size on
exporters’ average exports per product.
The results on the product intensive margin are particularly
interesting. They imply that the indications of
the (net) impacts of GDPs and distance on the firm intensive
margin highlighted in Figure 11 are attributable
to their impact on the total number of exported products,
which is far greater than the impact on average
exports per product.11
We can thus write:
Fact 9 – Firms’ average exports per product matter
less. The changes in the number of exporting firms
and in the number of exported products accounts
entirely for the negative impact of higher trade barriers
and the positive impact of larger countries’
size on bilateral exports.
The finding that the “product intensive margin” falls
with GDP and increases with distance is puzzling at
first sight. Two hypotheses can be proposed to explain
it, one related to “efficiency sorting” and another re-
10
These findings are very similar to the ones by A.B. Ber nard et
al., op. cit., and R. Hillberry, D. Hummels op. cit., for external
and internal US trade flows, respectively.
11
This finding parallels the one by A.B. Ber nard et al., op. cit., for
US exporters.
lated to “quality sorting” of firms over different export
markets. The former refers to the fact that only the
most productive firms from a certain country manage
to export to distant or small foreign markets. This occurs
because only those firms are able to quote low
enough prices but still succeed in exporting large
enough quantities to at least break even. Nearer or
larger markets attract many more exporting firms, and
the proportion of high cost – high price – low quantity
exporters is larger.12
Since the product intensive margin
only considers the average shipment value, such
a composition effect may explain why the effects of
GDP are negative and those of distance are positive.
Alternatively, the puzzling signs of the effects may
have to do with the quality or price/weight ratio exported
to different markets. If firms differ in the quality
of the product exported (or have different qualities
in their portfolio of products), it may be observed that
only the high quality varieties are exported to distant or
small markets, while low quality products can only be
exported to nearer or large markets.13
Distinguishing
between the two alternative explanations is a complex
issue, but the average price of shipments can be used
to shed some light on it.
Price and Quantity Margins
We now turn to the last decomposition, which allows
us to distinguish between the gravity effects on
average quantity and on average price.
A final decomposition of the average exports per
product by firm (product intensive margin) into aver-
12
See M. Melitz, G. Ottaviano, op. cit., for a theoretical formalisation
of this idea.
13
A.B. Ber nard et al., op. cit., conjecture that this second explanation
might be the relevant one to explain their result, but do not investigate
it further.
Figure 13
Gravity and Aggregate Exports – III
Source: EFIM.
COMPETITIVENESS
Intereconomics, May/June 2008146
age quantity per product by firm and average price per
product by firm can be carried out using information
on the value and quantity of shipments measured at
product level. The results of this final decomposition
are reported in Figure 13.
The bar chart in Figure 13 represents the contribution
of the firm extensive margin (“number of exporters”),
the product extensive margin (“number of
products”), the quantity margin (“average quantity
per product by firm”) and the price margin (“average
price per product by firm”) to the overall effects
(small diamonds) of three gravity forces on bilateral
exports.
The chart shows some support for both “efficiency
sorting” and “quality sorting”. The former implies that
firms managing to export to smaller or more distant
markets are on average more productive and therefore
have on average higher volumes of sales. Figure 13
shows that the average quantity exported decreases
with GDP and increases with distance, pointing to the
presence of less efficient exporting firms in larger or
closer markets. The dark areas report the results for
the average unit price of exports. Average prices tend
to increase with distance from the exporting country,
which is consistent with “quality sorting”, as long as
higher quality varieties are the only ones able to reach
distant markets. However, such a mechanism would
certainly predict a negative effect of GDP. Hence, overall
“quality sorting” seems to be a weaker explanation
of the aggregate observed behaviour of the product
intensive margin.14
We have therefore:
Fact 10 – Prices and quantities defy gravity. The
average quantity exported by firms and the average
export price per product are, respectively, smaller
and larger in larger countries. A reduction in trade
barriers leads to a fall in both of them.
14
One must be cautious in interpreting these results since validating
the “quality sorting” hypothesis would imply running the analysis at
the firm level and measuring quality more directly. More generally, the
average price is a mixed bag of all sorts of underlying product prices,
and therefore trade composition effects are likely to blur any story concerning
efficiency or quality sorting at the industry or even firm level.
Those sorting effects could only be properly uncovered through careful
firm or industry level analysis, which goes well beyond the scope
of the present descriptive analysis. See R. Baldwin, J. Harrigan:
Zeros, Quality and Space: Trade Theory and Trade Evidence, in:
NBER Working Paper No. 13214, 2007; and M. Crozet, K. Head,
T. M ayer: Quality sorting and trade: Firm-level evidence for French
wine, in: mimeo, CEPII, 2007 for more detailed hypotheses on this issue.
Deeper investigation is also needed to shed light on additional
issues such as the opposite effects of regional trade agreements (RTA)
on average export price and average export quantity.
Further analysis of French data reveals that former
colonial ties and sharing a common language have a
positive impact on trade flows along both the intensive
and the extensive margins. Moreover, in markets that
are easier to access, such as those of former colonies
and francophone countries, French exporters are more
numerous and on average less efficient, which drives
down the average quantity exported. As colonial ties
and a common language are not directly related to
distance, that suggests that such variables proxy for
lower fixed costs of exporting.
Hence, we can highlight:
Fact 11 – Historical ties and common language
matter. Historical ties such as former colonial links
and a common language foster exports, making it
easier for less efficient firms to export.
Finally, the relationship between market size and
average prices is not as clear as the other three relationships.
This is not unexpected, since this average
price is a mixed bag of all sorts of underlying product
prices.
The Margins of FDI
The gravity model has been primarily devoted to the
study of trade flows, but more recently a fair amount of
research has used the same variables to explain patterns
of bilateral FDI flows or stocks.15
The equilibrium
equation for bilateral capital flows closely resembles
the gravity relation for bilateral trade flows. Nonetheless,
the interpretation of the coefficients is sometimes
very different. Most importantly, in the case of trade
flows the negative coefficient on distance captures the
frictions due to trade costs (including freight costs),
while in the case of FDI flows the same coefficient
captures the frictions due to information and transaction
costs associated with the acquisition or installation
of new capital abroad.
As in the case of bilateral exports, the decomposition
of the margins can be used to highlight the channels
through which gravity forces affect the sales of
foreign affiliates. In Figure 14 each bar chart repre-
15
For example, K. Head, J. Ries: FDI as an Outcome of the Market
for Corporate Control: Theory and Evidence, in: Journal of International
Economics, Vol. 74, Issue 1, 2008, pp. 2-20 have recently developed
a model of FDI where heterogeneous investors bid to obtain
control rights to existing overseas assets. The equilibrium equation for
bilateral capital flows closely resembles the type of trade flow gravity
equation derived with heterogeneous exporters. In the same spirit, A.
Hijzen, H. Görg, M. Manchin: Cross-border mergers and acquisitions
and the role of trade costs, in: European Economic Review,
2007, forthcoming, investigate the role of trade costs in explaining the
increase in the number of cross-border M&As.
Intereconomics, May/June 2008
COMPETITIVENESS
147
sents the contribution of firm extensive (“number of
affiliates”) and intensive (“average sales” per affiliate)
margins to the overall effects (small diamonds) of
two gravity forces: the size of the destination country
(“GDP, im”) and distance (“Dist.”) The decomposition
of the margins is possible for Norway (a), Germany (b)
and Belgium (c), for which we have both the number
and the local sales of foreign affiliates.
Figure 14 shows that, as in the case of exports, the
overall pattern of foreign affiliate sales is overwhelmingly
driven by the extensive margin. The contribution
of the number of affiliates abroad is systematically
higher than the contribution of average sales per affiliate
for all three countries.
The massive positive influence of the GDP of the
country of destination is noteworthy. It highlights the
fact that, at this level of aggregation, FDI is primarily
driven by market access considerations (“horizontal
FDI”) and not cost-saving ones (“vertical FDI”).16
Moreover,
Figure 14 shows that the rise in foreign affiliate
sales associated with the increase in the GDP of the
country of destination comes mostly (65% for Norway,
61% for Germany and 53% for Belgium) from the increase
in the number of foreign affiliates.
More detailed estimations also reveal the key role
of the number of affiliates in transmitting the effects of
other gravity forces: the effect of distance for Belgium,
Italy and Norway; the RTA, language and colonial effects
for Germany and France; the RTA effect for Italy
and the colonial effect for Belgium; the effect of GATT/
WTO membership for Belgium, France, Germany and
Norway.17
Hence, we have established:
Fact 12 – The number of foreign affiliates matters.
Larger countries and lower trade barriers attract
more multinational activities. This attraction
is evident mostly in terms of larger numbers of
foreign affiliates than in terms of more sales per
affiliate.
Conclusions
A lack of statistical information at the firm level has
so far prevented the systematic inclusion of firm-level
analysis in the policymaker’s standard toolbox.
This paper argues that the time is ripe to supplement
the policymaking toolbox: firm-level datasets are
now available and provide new information that one
cannot afford to ignore.
The focus of this paper is on the characteristics
of European firms involved in international activities
through exports or foreign direct investment (“interna-
16
See e.g. G. Barba Navaretti, A. J. Venables: Multinational
Firms in the World Economy, 2004, Princeton University Press; and
B. A. Blonigen : A Review of the Empirical Literature on FDI Determinants,
in: Atlantic Economic Journal, Vol. 33, 2005, pp. 383-403 for
definitions of the two types of FDI and related empirical evidence.
17
In a recent study on the offshoring activities of German firms, C.
Buch, M. Schnitzer, C. Ar ndt, I. Kester nich, A. Mattes, J.
Eaton, C. Mugele and H. Strotmann: Analyse der Beweggründe,
der Ursachen und der Auswirkungen des so genannten Offshoring auf
Arbeitsplätze und Wirtschaftsstruktur in Deutschland, IAW Tübingen
2007, present results for a larger list of determinants, including per
capita income and country ratings. In this study, distance becomes a
significant determinant of German firms’ FDI.
Figure 14
Gravity and Aggregate FDI
S ource: EFIM.
COMPETITIVENESS
Intereconomics, May/June 2008148
tionalised firms”). The analysis of firm-level evidence
reveals some new facts that are simply unobservable
at the aggregate level:
IFs are superstars. They are rare and their distribu-•
tion is highly skewed, as a handful of firms accounts
for most aggregate international activity.
IFs belong to an exclusive club. They are different•
from other firms. They are bigger, generate higher
added value, pay higher wages, employ more capital
per worker and more skilled workers and have higher
productivity.
The pattern of aggregate exports, imports and for-•
eign direct investment is driven by the changes in
two “margins”. The “intensive margin” refers to average
exports, imports and FDI per firm. The “extensive
margin” refers to the number of firms actually
involved in those international activities.
The “extensive margin” is much more important,•
as the reaction of aggregate trade and FDI flows to
country fundamentals takes place mostly through
that margin. This is impossible to see without firmlevel
data and thus has not been seen so far.
We stress six policy implications.
First, promote intra-industry competition. The•
opening up of trade and FDI triggers a selection
process in which the most productive firms replace
the least productive ones within sectors. This is
good for productivity, GDP and wages, even when
it does not lead to sectoral specialisation. Moreover,
precisely because winners and losers belong to
the same sector, the benefits of selection are likely
to be associated with limited social costs of adjust-
ment.
Second, increase the number of exporters and mul-•
tinationals. What matters most for a country’s trade
and FDI performance is, first of all, how many of its
firms engage in exports and FDI. So governments
should focus on policies that broaden the export
base.
Third, forget the incumbent superstars. If the aim is•
to broaden the export base, governments should
not focus on policies that favour existing superstar
exporters and multinationals. Instead, heads of government
should work on lowering barriers to exports
and FDI at home. Trade missions do not generate
trade.18
Fourth, nurture the superstars of the future. Govern-•
ments should provide the conditions for tomorrow’s
superstars to emerge by allowing small exporters
and multinationals to grow.
Fifth, keep up the fight against small-trade costs.•
Small (fixed) costs of internationalisation matter because
they reduce the number of exporters and mul-
tinationals.
Sixth, assess the export and FDI potential of your•
industries. Some industries are more likely than others
to react to shocks through adjustment in the
numbers of exporters and FDI-makers. Hence, they
have greater unexploited export and FDI potential.
These are industries characterised by a larger presence
of small, low-productivity firms. As such, they
are also more likely to react to import competition
through the exit of the worst-performing firms and
therefore also have greater unexploited productivity
gains from selection.
Our findings also leave some questions open. We
prioritise six of them for future investigation. If firms
have to be large to be competitive in international markets,
what is the significance of the size of the internal
market? If superstars dominate international markets,
is there any room for global SMEs? What precisely
does the dominance of the extensive over the intensive
margin imply for policy intervention designed to
promote the internationalisation of European firms?
Do firms improve their performance when exposed
to international competition? Is the fragmentation of
production processes across countries a way through
which firms become more competitive in international
markets? Is the limited internationalisation of European
firms eroding political support for the single market?
Answering these questions requires quality data
at the firm level to be representative and comparable
across European countries. Currently, however, the
overlap among the different national datasets in terms
of several key variables is far from complete at the targeted
level of disaggregation. In this paper we select
different countries depending on the specific issues
addressed. This is clearly a second-best approach
that is nevertheless enough to highlight the benefits
that would come from the creation of a harmonised
European dataset.
18
See J. Head, K. Ries: “Do trade missions increase trade?”, University
of British Columbia, mimeo, 2007.