MEMOIRE Présenté en vue de l'obtention du Master en Sciences économiques, finalité Entreprises Determinants of European R&D offshoring: A gravity model of R&D offshoring flows in Europe Par Sébastien Bouvy Coupery de Saint Georges Directeur: Carine Peeters Assesseur: Pierre-Guillaume Méon Année académique 2010- 2011
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
MEMOIRE
P r é s e n t é e n v u e d e l ' o b t e n t i o n d u M a s t e r e n S c i e n c e s é c o n o m i q u e s , f i n a l i t é E n t r e p r i s e s
Determinants of European R&D offshoring: A gravity model of R&D offshoring flows in Europe
Since the 80s, there have been different waves of manufacturing offshoring all
around the world in order to benefit from the lower labour costs in certain coun-
tries. Indeed, one of the key drivers to offshoring is cost savings. However, there
are several other reasons and advantages such as the access to distinctive skills1
and growing performance from fast-developing economies, particularly in Asia.
Currently, companies are meeting a new round in the offshoring trend: they
are likely to think about other ways to improve their structures in R&D2. The
most recent offshoring is linked to innovation. According to Bardhan (2006),
the globalisation process and the intensification of competition have forced en-
terprises to redesign their management structure and take into consideration all
cost sources, including R&D and innovation-related activities.
More precisely, the international trade theory has a limited consideration as
to the effects of offshoring on R&D activities in the country of origin. Certain
authors such as A. Naghavi and G. Ottaviano (2006) tried to fill the gap in the
international trade literature on the dynamic effects of offshoring on R&D. The
authors determined that, when offshoring reduces the feedback from offshored
plants to domestic labs, it is likely to bring dynamic losses when the countries
of origin are large, and in sectors in which R&D is cheap and product differen-
tiation strong. In their endogenous growth model, offshoring of R&D induces
some coordination problems between the offshored and domestic divisions of a
corporation.1Skills are likely to be available in abundance. For instance, China produces 350,000
engineering graduates each year compared to 90,000 in the U.S.A. - “Offshore bonanza: Smart
firms look beyond mere cost savings”, Strategic Direction (2006), Vol. 22 Issue: 5, pp. 13-15.2The following definition of R&D comes from the OECD summary of Frascati Manual
which helps national experts in OECD countries to collect and issue R&D data: “Research
and experimental development (R&D) comprise creative work undertaken on a systematic
basis in order to increase the stock of knowledge, including knowledge of man, culture and
society, and the use of this stock of knowledge to devise new applications. R&D is a term
covering three activities: basic research, applied research, and experimental development.”
5
Another element to consider is the decision to offshore one’s innovative depart-
ment to a foreign location. Indeed, instead of keeping its research centre in a
domestic location a corporation may decide to set up a foreign affiliate which
will focus on innovation or to subcontract such activities to a foreign partner.
This strategy might be either to benefit from specific factors in a particular area
(a lot of highly-skilled people in a foreign location, capital intensive area, etc.)
or to cut costs by paying lower wages for the same level of skills compared to
the national level.
However, companies should take into account the complementarity of home and
offshored R&D activities to achieve a competitive advantage. As D’Agostino et
al. (2010) suggested, the complementarity between domestic and foreign inno-
vative assets depends on their natures and their complexity. In fact, the home
and offshored R&D activities are complementary if they are not similar as well
as when offshored R&D activity is concerned with modular and less complex
technologies. This finding is based on the geographical technological specialisa-
tion and the reverse knowledge transfer from the offshore locations to the home
regions.
Moreover, when looking at the structural attributes of R&D offshoring, there
are common characteristics to the offshoring of services compared to the off-
shoring of manufacturing activities (see Bardhan (2006)). Actually, R&D off-
shoring and manufacturing offshoring are both more capital intensive than ser-
vices offshoring. In terms of effects on jobs in the home country, manufacturing
offshoring influences contiguous and similar skills and occupations within the
blue-collar workforce, whereas outsourcing/offshoring of services and R&D af-
fects white collar jobs across dissimilar occupations. Manufacturing offshoring
can be viewed as impacting along product lines, whilst, services offshoring is
impacting along occupational/job lines. R&D offshoring is a mix of both. The
development of a new product would initially be part of manufacturing off-
shoring but this kind of activity requires specialised occupations/jobs such as
6
scientists, engineers and so forth. This is the reason why services offshoring
affects other occupational lines compared to R&D offshoring (see Figure 2 in
Appendix B).
The purpose of this paper is to provide a new vision on the dynamics of off-
shoring innovation-related activities. This vision is based on the idea that there
are offshoring flows similar to trade flows between countries and that these move-
ments can be determined by different factors. In this paper, the gravity equation
model is used to set the relation between R&D offshoring flows in Europe and
the most relevant variables. The gravity model was used in several manners
to assess different flows. However, originally, this model was constructed to
analyse international trade and was applied by some pioneers like Tinbergen
(1962), Pöyhöhen (1963), and Linneman (1966). The theoretical basis of the
model came later after many other applications such as in the FDI flows between
countries (Brainard (1997); Mello Sampayo (2009)). The theory which underlies
the gravity model explains that the shorter the distance between two countries,
the greater the intensity of trade activity between those countries. Moreover,
the international trade flows increase with country size and decrease with trade
costs i.e. transportation costs which is represented by distance between nations.
Using the gravity equation model specifications, this study targets to find what
are the relevant determinants of R&D offshoring flows within European coun-
tries. This paper is divided in five other sections where Section 2 clarifies the
difference between offshoring and outsourcing concepts and provides a review of
the literature about R&D offshoring topic. Section 3 explains the basics and the
evolution of the gravity equation model and then provides an empirical back-
ground of this model linked to our subject. Section 4 describes the data, the
sample used, and the econometric aspect of this study. Afterwards, Section 5
brings the results of the estimation. Finally, the paper ends with a conclusion.
7
2 A broad overview on offshoring
2.1 Offshoring vs. outsourcing
Many people do not know the clear difference between offshoring and outsourc-
ing and, very often, use both terms without any deep comprehension of what
they are. The following study is based on the definition in Bardhan’s (2006)
paper: foreign outsourcing is arms length sourcing to suppliers abroad, and
intra-company offshoring is the transfer of production abroad to foreign affili-
ates and subsidiaries of European companies, with the objective of exporting the
output back to the Europe. This definition clarifies the concepts of outsourcing
and offshoring in terms of investment decisions.
A domestic company may decide to invest to create a foreign affiliate so that
the latter conducts a certain activity instead of its parent (e.g. manufacturing
activity, IT services, etc.). This action is called by Lewin et al. (2008) captive
offshoring i.e. the domestic firm keeps the control by owning the majority of the
shares of its foreign subsidiary. On the other hand, the national enterprise can
decide to offshore certain activities by subcontracting with a foreign partner.
This is the offshore outsourcing decision where the foreign partner has the total
control of its supply to the domestic company. The offshoring decision has two
main implications for the concerned company either it decides to offshore and to
outsource its IT services, for instance, or it invests abroad into a subsidiary in
order to offshore and insource its IT services. Hence, we consider two categories
of offshoring: captive offshoring and offshore outsourcing. The key difference
between these two concepts is based on the control from the home company on
its offshored activities.
8
2.2 Literature review and offshoring trends
The typical view on offshoring is always defined by the Northern countries which
offshore some of their basic activities to the Southern countries in order to ex-
ploit a cost advantage in those locations. Antras and Helpman (2004) defined
a theoretical model with two countries, the North and the South, for analysing
the global sourcing strategies. They found that “high-productivity firms acquire
intermediate inputs in the South whereas low-productivity firms acquire them
in the North. Among firms that source their inputs in the same country, the
low-productivity firms outsource whereas the high-productivity firms insource.
In sectors with a very low intensity of headquarter services, no firm integrates;
low-productivity firms outsource at home whereas high-productivity firms out-
source abroad.” Outsourcing can also happen between vertically integrated
firms. Helpman (1984) introduces a model of vertical foreign direct investment
in order to explain the intra-firm trade related to the intra-firm international
outsourcing.
According to PRTM, a large management consultancy firm, and World Trade
magazine survey the first concern for a large number of companies in the US,
Europe and Asia is the offshore transfers and related outsourcing topics. The
survey found also that this issue is not the inherent prerogative of the huge
MNEs3 as many small and medium-size structures are intensively prospecting
offshore opportunities. Moreover, we know that since the 1980s outsourcing
of manufacturing activities to low-cost countries is usually practised (Dunning
(1993); Lee (1986); Vernon (1966)) and even more routinised now. The survey
shows that offshoring decisions are not limited to manufacturing industry but
also apply to a wide range of industries, “[...] from consumer services to high
tech.”
3The acronym MNE refers to “Multinational Enterprise”.
9
Looking at the material outsourcing, a large part of the studies found an in-
creasing extent of international outsourcing of material inputs over time (see
Feenstra and Hanson (1996), Campa and Goldberg (1997), Hummels, Ishii and
Yi (2001), Yeats (2001), Hanson, Mataloni and Slaughter (2004), and Borga and
Zeile (2004)). Additionally, Egger and Egger (2006) show that, for European
countries, there is a negative impact of international material outsourcing on
the productivity of low-skilled workers in the short run, whereas there is a pos-
itive impact in the long run. Empirical evidence in the United States (Feenstra
and Hanson, (1996, 1999)) and the United Kingdom (Hijzen et al., (2002)) show
also that outsourcing of unskilled labour-intensive parts of production processes
from relatively skilled-abundant countries to unskilled-abundant countries leads
to an increase in the relative demand for skilled labour in the skilled-abundant
country and hence increases the skills premium.
For at least a decade, there has been a new trend of globalisation which is
concerned by the internationalisation of services trade and became really im-
portant in the total value of trade around the world. As Dossani and Panagariya
(2005) explained, some developing countries such as in Asia have become large
suppliers of services for developed nations. The increase of this type of trade is
due to more offshoring for this kind of activities and concerns a large range of
services like “back-office services such as payroll; customer-facing services such as
call-centres and telemedicine; design services such as the design of application-
specific integrated circuits; research services such as conducting clinical trials;
software services such as programming; and IT and infrastructure outsourc-
ing such as the managing of corporate e-mail systems and telecommunications
networks.” The same authors argued that the largest growth in offshoring is
happening in business services4.4“Business processes is a general term to refer to the collection white-collar processes that
any bureaucratic structure undertakes in servicing its employees, vendors, and customers such
as human resources, accounting, auditing, customer care, telemarketing, tax preparation, etc.”
- Rafiq Dossani and Arvind Panagariya (2005).
10
With respect to the job reallocation issue, R&D offshoring can lead to some im-
portant consequences in the workplace of both developed and developing coun-
tries. Knell and Rojec (2009) studied the job reallocation issue at the European
level by using the dataset of the publically available European Restructuring
Monitor (ERM). They found that at least half of all European offshoring oc-
curs within Europe. Then, India is larger than China as a source of offshoring,
mainly because of the huge volume of offshoring in the service industries (e.g.
call centres). In order to lower labour costs and have access to well-educated
pool of workers, European offshoring is moving principally to Eastern Europe.
These authors pointed out that offshoring induces the movement of low-skill
jobs out of Western Europe whereas offshoring of innovation-related activities
and the relatively high-skill jobs remain within Western Europe.
The press links offshoring with job losses but Amiti and Wei (2004) show that
there is no evidence to support this assumption. In fact, a large part of developed
nations are not specifically more outsourcing-intensive than many developing
countries. More precisely, many developed regions tend to have surplus i.e., the
rest of the world outsources to them rather than the contrary. The top providers
in services are firstly, the United States and secondly, the United Kingdom. The
authors explained that service outsourcing would not induce a reduction in ag-
gregate employment while it has the potential “to make firms/sectors sufficiently
more efficient, leading to enough job creation in the same sectors to offset the
lost jobs due to outsourcing.”
According to Amiti and Wei (2004), despite the early offshoring of manufactur-
ing activities, the offshoring of high-value adding activities remains a relatively
undiffused practice. In fact, innovation-related activities are still difficult to off-
shore because they imply intangible goods such as the knowledge, the skills, the
education, etc. Furthermore, the domestic firm may have to support a higher
risk in this kind of offshored activities as its product development depends on
the ability and the availability of highly-skilled people in a too distant foreign
11
location to provide the expected results. Consequently, the distance between
two entities depending on each other is important because one needs to sell
new products or new services resulting of an intensive R&D activity to grow
profits and another needs the previous one because its production has not any
value outside their relationship. The information asymmetry can become a huge
problem in the relationship between offshored activity and the domestic parent
or partner as well.
On the other hand, the new wave of offshoring of R&D activities originates
from a change in the business model of firms. As Bardhan and Jaffee (2005)
explain, the individual is experiencing a transformation from a model of pro-
prietary, internal, intra-firm or domestically-based industrial laboratory to an
offshoring model. This change is due to at least one major reason which is the
increasingly global nature of sales of large firms. Indeed, if a firm expands its
market share throughout the world it needs to design its products in line with
local tastes, leading to the strategy to “design to the market” and even to “design
and research to the market” which adds to the previous strategy to “produce
to the market”. According to Bardhan and Jaffee, there is a huge potential of
skilled labour in China and India. In consequence, there is an outward transfer
of R&D activity to India, for instance, in software, bio-technologies, pharma-
ceuticals, engineering design, and development areas.
A large pool of highly skilled workers in emerging countries constitutes a pre-
requisite to offshore innovation-related activities. This is a new key strategic
driver (Bunyaratavej et al., 2007; Deloitte, 2004; Farrell et al., 2006; Lewin &
Couto, 2007; Lewin & Peeters, 2006) and implies more than just the offshoring
of IT activities or business processes. As explained by Manning et al. (2008),
offshoring involves now product development and product design and these phe-
nomena might influence what the authors call the global sourcing of Science and
Engineering (S&E) talent. Based on the annual Offshoring Research Network
survey results, a large part of US and European companies have started to em-
12
ploy S&E skills in different areas in the world. This trend is due to a shift
of clusters providing highly-skilled people from Western countries to emerging
nations such as China and India which have invested more in education and
innovation in order to curb and gradually “reverse” the brain drain.
3 The gravity equation model
For this paper we decided to use the gravity equation framework to assess the
R&D offshoring flows because of the wide empirical history and applications
of this model on bilateral trade flows in the beginning (see Tinbergen (1962)),
and on other flows like FDI between nations later on. This model generally
provides interesting macro-level results about the influences of some factor on
trade-flows. In our case, this is a completely new application of the gravity
equation model which estimates the relationship between offshoring flows and
some determinants. The next sub-section provides the basics about the theoret-
ical aspects of the gravity model and the other sub-section presents an overview
of the empirical literature.
3.1 The classical gravity equation model
In the standard gravity equation, trade flows between a pair of countries are
proportional to their masses (GDPs) and inversely proportional to the distance
between them. Numerous studies used the basic form of this model and showed
relevant empirical results. This form is expressed as following:
Mij = αY βi Yγj N
δi N
εj dµijUij
5 (1)
where Mij is the trade flow of goods or services from country i to country j,
Yi and Yj are GDPs of i and j, Ni and Nj are population of i and j, and dij
is the distance between nations i and j. Usually, we assume that the Uij term
is a lognormal distribution error factor with E(ln(Uij)) = 0. Some authors like5This equation comes from Anderson’s paper (1979) where he explained the theoretical
foundations of the gravity equation model.
13
Anderson (1979) defined the theoretical foundations of this model which had
firstly more empirical specifications.
On the other hand, according to Kimura and Lee (2006), it has been found
that the gravity model can be deducted from different models as Ricardian,
Hecksher-Ohlin and the monopolistic competition model. Indeed, Helpman and
Krugman (1985) have shown the possibility to derive the gravity equation from
the monopolistic competition model with increasing returns to scale. Moreover
Deardorff (1998) found that one can derive a gravity model from a Heckscher-
Ohlin model without assuming product differentiation. A gravity relationship
has been put in evidence by developing a Ricardian model of trade in homoge-
neous goods (see Eaton and Kortum (2002)). As a result, the gravity equation
is part of any model of international trade.
The gravity equation model was used by Frankel and Romer (1999) to assess the
influence of trade on growth by using the same bilateral trade data as Frankel
et al. (1995) and Frankel (1997). This database combines a sample of 63 coun-
tries for the year 1983. The authors drop from their database the observations
where registered bilateral trade is zero. Their findings fit other empirical results
i.e. trade as a fraction of GDP is negatively correlated to distance, is positively
correlated to the size and population of the jth country, etc.
Despite its successful applications and theoretical basis, the gravity equation
from an empirical point of view has some limitations and mismatches when
there is no trade between a pair of countries. Indeed, the majority of empirical
studies log-transformed the bilateral variables (trade, FDI, etc.) in order to have
a consistent log-normal distribution depending on the log-normal distribution
of explaining variables. However, this log-transformation eliminates a part of
the observations on the bilateral dependent variable i.e. for the zero-value. As
a result, the researchers lost some part of the information which may be rele-
vant. To overcome this problem some authors found simple solutions such as
14
adding one to all observations of the dependent variable in order to get, in log-
term, zero-values6. A drastic solution is to drop the pair of countries with zero
trade from the data set and afterwards estimate the remained log-transformed
observations by OLS. Unfortunately, those methods can produce inconsistent
estimators of the parameters of interest.
Another problem with the log-linearisation, quoted by Silva and Teneyro (2006),
is the heteroskedasticity. This leads to have inconsistent estimators and “if er-
rors are heteroskedastic, the transformed errors will be generally correlated with
the covariates.” These authors propose a solution based on a constant elasticity
model to the different problems linked to the log-transformation (for details,
see Appendix A). By conducting a simulation study, they found that a Pseudo-
Poisson-Maximum Likelihood method is the most efficient resolution in com-
parison to other estimation methods (Tobit, NLS and OLS). Indeed, according
to their results, the “income elasticities in the traditional gravity equation are
systematically smaller than those obtained with log-linearized OLS regressions.
In addition, in both the traditional and Anderson−van Wincoop specifications
of the gravity equation, OLS estimation exaggerates the role of geographical
proximity and colonial ties.” Consequently, the regression analysis of this paper
is built on the comparison of different estimation models as PPML in order to
get the best and the most relevant estimators.6Some raw data for the bilateral dependent variable Tij can be equal to zero, so the solution
to take into account in the estimation for such observations is explained as follows:
Adding one to the raw data of Tij variable:
1 + Tij (so, the zero-values take the value one)
In log-term:
log(1+Tij) (then, the observations equal to one (i.e. zero-values, in raw data term) are equal
to zero, in log-term).
15
3.2 Empirical background
The present study does not focus on the corporate-level decision to offshore
its key activities but tends to estimate the importance of some factors on the
R&D offshoring flows in Europe. In doing so, one has to bear in mind the pre-
vious explanation about the two key concepts of this paper (see Section 2.1).
In addition, we conduct a study on an aggregated level i.e. on the bilateral
country flows. Other studies focused on a more disaggregated level about the
relationship between trade and innovation-related activities. Uzagalieva et al.
(2010) used this approach to assess the relationship between innovation expen-
ditures and the intra-industry trade flows in European markets. These authors
concentrated on the imitation and innovation concepts which are important
modes of technological development. They used a gravity equation model to
estimate the potential progress effects of innovation and imitation on a sample
of 20 countries. The results are that the increase in size of the science-based
manufacturing industries leads to a greater intra-industrial trade between coun-
tries which approximates innovation-based technological growth. As usual in
the gravity model, the distance decreases the trade flows. R&D expenditures
have a significant and positive influence on the progress indicator.
Regarding the effect of technological innovation on international trade, Ramos
and Martinez-Zarzoso (2009) find that it has a positive impact on export perfor-
mance but also that it is a non-linear relation. There is a U-shaped relationship
between exportations and creation of technology and between exportations and
diffusion of old technology. However, the relations between exports and diffu-
sion of recent innovations and between exports and human skills are defined
by an inverted U-shaped chart. To overcome the complexity to capture all the
aspects of technologies, they used in their empirical analysis an index called
Technological Achievement Index7 which is based on four dimensions: creation7This composite index was firstly introduced in 2001 by UNDP in its Human Development
Report 2001 - Source: UNDP (2001) Human Development Report 2001, Oxford University
Press, New York.
16
of technology, diffusion of recent innovations, diffusion of old innovations and
human skills.
When assessing whether better information can eliminate the effect of geograph-
ical distance, Loungani et al. (2002) find some heterogeneity between developed
and developing nations. Indeed, within the different determinants of interna-
tional trade technological innovation constitutes “a substitute for distance in
developing countries (better information lowers the effect of distance), whereas
technological innovation and distance are complementary in developed countries
(better information magnifies the effect of distance)”. Furthermore, Fink et al.
(2005) show that communication costs on bilateral trade flows have a significant
effect and they have a greater weight when exchanging differentiated products
compared to exchanging homogeneous products. These empirical results show
that it is important to take into account and to bear in mind the non-linear
influence of technological progress on trade flows.
Dollar and Kray’s (2003) paper show that the quality of institutions consti-
tutes a great determinant of trade flows in our economy. For instance, the rule
of law factor measures the level of corruption in a country and has a clear im-
pact on the level of trade in the concerned country. An exporter have to support
risks linked to the business and corruption might increase it more than other
factors. According to De Groot et al. (2004), the institutional quality has a
clear and positive effect on bilateral trade flows. They used a gravity equation
model to estimate the influence of institutions on trade. Their model shows
that good governance lowers transaction costs for trade between high-income
countries, while trade between low-income countries suffer from insecurity and
transaction costs.
The regional trade agreements (RTAs) have an influence on trade flows between
countries. Some authors studied this kind of determinants within European
countries. According to Stack (2009), the RTAs effects on trade focus on the
17
enlargement process rather than the deepening of trade integration between EU
members. She quotes that in a part of the empirical literature the sign and
significance of trade policy effects can differ. This is due to the existence of bias
because of omitted relevant variables in the analysis. Stack used a dataset of
bilateral flows from 12 EU countries to 20 OECD trading partners between 1992
and 2003. The results show that the positive and significant coefficient estimate
of the European trading bloc dummy variable declines in magnitude with an
increasing degree of heterogeneity in the model. According to these results, it
is difficult to quantify the effect of European integration on trade flows.
Martinez-Zarzoso and Nowak Lehman (2003) studied another free trade agree-
ment between the Mercosur and the European Union by using a gravity equation
model. They used a panel of data from a sample of 20 countries (Mercosur with
Chile and the EU15 bloc of European countries) in order to clarify the time
constant country-specific effects and also to take into account of relationships
between the relevant variables over time. They found that the fixed effect model
is more relevant compared to the random effect gravity model. They added some
variables to the basic gravity equation and the estimation results show that the
infrastructure, income differences and exchange rates are important explana-
tory variables for bilateral trade flows. Specifically, the exporter and importer
incomes have a positive influence on trade between these two blocs of nations.
The tax policy in a specific region can be also an interesting determinants of
trade flows within Europe. Hansson and Olofsdotter (2008) studied the influ-
ence of tax differential on a sample of bilateral FDI flows for the European
Union members over the period 1986-2004. They found that tax differentials
are important determinants explaining FDI flows. Indeed, the marginal effective
corporate tax rates between host and investing countries have a negative im-
pact on FDI flows. De Mooij and Ederveen (2006) argued that tax differentials
influence FDI, but that the magnitude vary substantially and is sensitive to em-
pirical specification as well as time periods and countries considered. Because
18
of those shortcomings, the present paper does not include the taxation issue in
its estimation of R&D offshoring flows in Europe.
4 Data description and econometric aspect
4.1 Data sources
Our bilateral dependent variable data over three periods (2007, 2008 and 2009)
comes from the Eurostat database. This dependent variable is part of the
Balance of Payments (BoP)8 of our sample composed by 15 European coun-
France, Italy, Latvia, Lithuania, The Netherlands, Poland, Romania, and Slo-
vakia). In fact, this is the bilateral transactions in R&D services between resi-
dents and non-residents of a given country i.e. the outward flows (recorded as
total value of credits in the BoP) in R&D services from country i to country j.
This paper focuses on the R&D offshoring flows throughout Europe and, there-
fore, we consider this variable as a proxy of offshoring flows within European
partners. We assume that the transaction flows between a pair of countries
in R&D services is the sum of payments exporting firms, located in country i,
receive from foreign external partners or foreign affiliates/parents, situated in
country j, in delivering offshored (in- or outsourced) R&D services as a result
of the offshoring of innovating-related activities in country i. Our assumption
and proxy variable for R&D offshoring are in line with the statement from Van
Welsum and Reif’s (2005) paper that there does not exist direct official data
measuring the extent of offshoring. These authors take trade in total services8“The Balance of Payments (BoP) systematically summarizes all economic transactions
between the residents and the non-residents of a country or of an economic area during a
given period. The Balance of Payments provides harmonized information on international
transactions which are part of the current account (goods, services, income, current transfers),
but also on transactions which fall in the capital and the financial account.” - Eurostat,
Balance of Payments statistics and International investment positions - Metadata.
19
as a proxy for measuring total services offshoring.
However, in our case it is important to bear in mind that not all trade in R&D
services is linked to offshoring and unfortunately it is not possible to distin-
guish the share of trade in R&D services that is directly related to offshoring.
The sample of countries considered includes 15 countries where each has by
turn the host position and the home position for three periods of time. It is
assumed that the ith country, called “host”, welcomes offshored innovation ac-
tivities and receives payments from the jth country, called “home”, which pays
for offshored R&D services coming from the host country. We take this specific
sample through an inductive process i.e. we selected each country with respect
to the availability of the data for our dependent variable over the considered
time period.
Concerning the databases used to build the independent variables for the regres-
sion analysis, the Eurostat database over three particular years - 2007, 2008,
and 2009 - is taken into account for the share of highly educated people in the
total population (aged 15 to 64 years) i.e. the people who attain at least the first
stage of tertiary education9 (higher education, university degree, etc.) and for
GDP. The reason to consider the first variable is that highly educated popula-
tion constitutes an important factor in offshoring literature (see Bunyaratavej et
al., 2007; Deloitte, 2004; Farrell et al.,2006; Lewin and Couto, 2007; Lewin and
Peeters, 2006) and much more when talking about offshoring complex activities
(see Section 2.2). In order to capture the size-effect on our dependent variable,
we take the GDP of each country of our sample. The level of innovation in
each partner is proxied by the share in GDP of gross domestic expenditure in
R&D whereas the infrastructure level is based on the level of Internet access in
percentage. Both variables may affect the R&D offshoring flows as a country in-9According to the ISCED - the International Standard Classification of Education - UN-
ESCO 1997, the data on highly educated people has a range from the 5th to the 6th level of
education i.e. from the first to the second-level of tertiary education.
20
vesting in innovation and infrastructure is likely to be a favourite location where
companies offshore. The data for the latter variables are taken from Eurostat
as well.
From the GDP data and population data, we calculate the income per capita
disparity variable which is our explanatory variable that capture the effect of
income differences between partners on offshoring flows. This variable is defined
by the difference between GDP per capita of each partner in absolute value. In-
deed, there is an income disparity even in Europe where typically the West is
richer than the East. This gap is likely to have an impact on the choice of each
partner to offshore or not.
The six Kaufmann indicators measuring the quality of institutions10 are part of
our explanatory variables as well (see Appendix C, Table 3 to have a complete
description of each indicator). In line with Dollar and Kray’s paper, governance
may influence R&D offshoring as a country prefers to offshore to a stable econ-
omy with good institutions. Each indicator is linked to a different dimension
of governance. It spreads out from −2.5 to +2.5, the higher the indicator, the
better is the governance. As in Méon and Sekkat (2006), to linearise these indi-
cators and to estimate the elasticities in the regression equation, we added 3.5
to them in order to be able to calculate logarithms.
We built a dummy to capture the membership of both countries taken into
consideration in the Euro Area bloc. The results of Stack’s paper (2009) lead
to add an EMU bloc dummy variable in order to examine the effects of Eu-
ropean integration on R&D offshoring flows. Such a dummy is more relevant
than an EU dummy because our complete sample of countries is part of the10“The governance indicators aggregate the views on the quality of governance provided by a
large number of enterprise, citizen and expert survey respondents in industrial and developing
countries. These data are gathered from a number of survey institutes, think tanks, non-
governmental organizations, and international organizations.” - Kaufmann et al. (2010).
21
European Union whereas the European currency Union dummy evolved over
the chosen time line. Then, the rest of variables were taken from the CEPII
database which provides the distances between capitals and dummy variables
indicating whether two countries are contiguous and share a common official
language. The distance proxies the transportation cost and constitutes a signif-
icant determinant of trade flows. The contiguity and a common language are
respectively geographical and cultural aspects whose effects were studied in a
broad part of the gravity literature and intuitively may have a positive impact
on R&D offshoring. For instance, a company will prefer to offshore a part of its
activities in a close-by location and/or a country with a similar culture in order
to ease communications and keep control on it.
4.2 Statistical discussion
4.2.1 Which are the top favourite locations for offshoring?
If we look at the last column in Table 5 (see Appendix C), the top 5 providers
of services in innovation-related activities are, by descending order: Germany,
Austria, France, The Netherlands, and Italy which have a share in total flows of,
respectively, 30.87 %, 24.67 %, 14.26%, 12.91%, and 9.11%. Therefore, it seems
that lots of firms offshored their innovation centres in those locations in order to
benefit from the highly-skilled labour force and the knowledge from this main
Western European countries. Indeed, these nations compose a large part of the
total highly educated people (see Table 6 in Appendix C) in our sample covering
2007 to 2009. Once exception is Austria which is one of the favourite offshoring
locations but which has not a large highly-skilled labor pool when looking at
its share in the total highly-educated population in our sample (2%). In this
country, the labour factor might be fully exploited in R&D activities and better
than in other countries. For instance, although Poland has a 10% proportion
in the researchers and engineers population of our sample it possesses a small
participation in the R&D offshoring flows. Hence, despite the fact that high-skill
jobs remain currently in Western Europe (see Knell and Rojec (2009)), there is
22
an opportunity for this country to become more and more a favourite offshoring
place thanks to the presence of highly-skilled workers.
4.2.2 Which are the top importers of R&D offshored services?
At the bottom of Table 5 (see Appendix C), we can see that the set of countries
which composed the most favourite locations for offshoring are also more or less
the top importers of innovation services. More precisely, the 5 most important
importers are, by descending order: Germany, France, The Netherlands, Italy
and the Czech Republic which have a share in total flows of, respectively, 40.82
%, 21.31 %, 15.08%, 7.71%, and 4.31%. The difference with the previous set is
that Austria is not present. Austria benefits from R&D centres set up within
its borders by foreign companies and becomes one of the largest net providers of
R&D services. Germany, France, The Netherlands and Italy are on both sides:
on one hand, providers of R&D services and, on the other hand, importers of
R&D services. These nations are most likely to trade together. Hence, there
may be huge intra-offshoring flows within this region as a Western European
company tends to offshore more in other Western European countries than in
other parts of Europe.
4.2.3 The importance of education
A highly-educated population refers to people who attain at least the first stage
of tertiary education (higher education, university degree, etc.) into the age
bracket from 15 to 64 years. Such a population is required in each country
to expand the research in key subjects like biomedicine, biofuels, new business
processes, etc. Consequently, we can assume that the evolution of a highly-skills
population is positively correlated with gross domestic expenditure in R&D.
Indeed, the levels of this type of expenditures as well as the level of innovation are
dependent from the number of researchers and engineers in a given country. This
is the reason why some foreign companies from different countries where there
are not enough well-educated people might offshore their innovation activities
23
to a location with a large pool of engineers or people with a PhD diploma, for
example. If we compare Table 5, 6, and 7 (see Appendix C), the top performers
in terms of R&D offshoring flows have a huge share in the sample in terms of
gross domestic expenditures. For instance, Poland has an important potential
to become one of the favourite destinations to offshore innovation activities from
foreign companies. This country gathers several advantages like a well-qualified
population which is correlated with larger gross domestic expenditure in R&D
than other Eastern European countries. Figure 4 (see Appendix B) shows a
clear relation between gross domestic expenditure in R&D and highly educated
population.
4.2.4 Offshoring flows between blocs
At a more aggregated level, if we consider some blocs of countries such as
Western European countries (Austria, Germany, Denmark, France and The
Netherlands), Southern European countries (Cyprus, Italy and Greece), and
Central and Eastern European countries (Bulgaria, the Czech Republic, Lithua-
nia, Latvia, Poland, Romania and Slovakia) called respectively WEC, SEC, and
CEEC, we would have other interesting results in terms of offshoring flows. The
weight of Western Europe is clearly dominant in our sample by observing Fig-
ure 3 (see Appendix B). This bloc of countries is composed of 4 out of 5 top
providers of services in innovation activities. Southern, Eastern, and Central
Europe seem to be marginalised and have small weights in the total flows. Fo-
cusing only on the Western bloc, we can observe another key element: almost
half of the volume of services provided by the Western countries is done by
Germany. So, Germany is one of the most favourite places to offshore R&D
activities.
Such differences in offshoring flows between these blocs might be explained by
a simple hypothesis that comes from the theoretical foundation of international
trade. More precisely, this assumption, the main one in the gravity equation
24
model, states that the bigger a country the more it trades with other nations.
Previously, we observed that 5 countries which are the biggest in Europe in
terms of GDP (see Figure 5 in Appendix B) provide lots of services in R&D
which means that many companies from other parts of Europe offshore their
innovation activities in these locations. The weight of these top 5 countries in
the offshoring flows and the share of each of them in GDP terms in our sample
are positively correlated. Looking at Figure 5 (see Appendix B), we observe
that Germany has the highest weight in size and offshoring flows. Table 2 (see
Appendix C) summarises the results according to the dimensions of size and
offshoring inflows performance. This table classifies the different countries of
our sample and, as we can see, Austria has an interesting position as a small
country but with a high performance in R&D offshoring flows. So despite its
smaller size than Germany, Austria has nearly the same weight in the sample in
terms of offshoring flows. The other top countries have an intermediate position
in R&D performance and have a different ranks depending on their size. The
rest of our sample is situated in the bottom-left position on the chart (see Figure
5 in Appendix B). However, we can notice that Poland tends to leave this latter
group.
4.3 Econometric specification
To assess the different determinants of the R&D offshoring flows across Europe,
a gravity equation is specified and estimated. The following equation defines
the additive form of the relation between the offshoring flows and these deter-