Determinants of Export Behaviour of German Business Services
Companies

Alexander Eickelpasch and Alexander Vogel


Abstract

The determinants of export behaviour at firm level have been widely investigated for
manufacturing companies. By contrast, what has remained largely neglected is a detailed
investigation in the service sector. As aggregate statistics show, international trade in services
has grown significantly over the last few years. However, it is unclear why some companies
export and others do not. This paper presents some initial results about the German business
services sector at firm level. Using a unique panel dataset of enterprises from the business
services sector (transport, storage and communication, real estate, renting and business
activities) for the years 2003 to 2005, we analysed the impact of several firm-specific
characteristics such as size, productivity, human capital, experience on the national market in
Germany, etc. on the firms’ export performance. Further, we used the pooled fractional probit
estimator, recently introduced by Papke & Wooldridge, an approach that considers both the
special nature of the export intensity variable and in addition unobserved time-invariant
characteristics. When there is no control for fixed enterprise effects the overall results are in
line with previous studies. When there is a control for unobserved heterogeneity, the positive
effects of productivity and human capital disappear, indicating that these variables are not per
se positively related to export performance, but rather to time-constant characteristics that are
unobserved. Size and product diversification still have a positive and significant effect.

Keywords: Business services sector, export behaviour, firm performance
JEL classification: F14, F23, L80


Acknowledgements:
Particular thanks to Joachim Wagner for his readiness to discuss methodological questions, Anja Malchin and
Ramona Pohl from the Research Data Centre in Berlin for providing the data, and Karl-Heinz Pesch from the
Federal Statistical Office for valuable information concerning the specific characteristics of the business services
statistics. To facilitate replication and extensions, the do-files used in this study are available on request.


Alexander Eickelpasch
Alexander Vogel
Deutsches Institut für Wirtschaftsforschung Leuphana
University
Lueneburg
Department of Innovation, Manufacturing, Service
Institute of Economics
Mohrenstraße 58, 10117 Berlin, Germany
PO Box 2440, 21314 Lueneburg, Germany
E-mail: aeickelpasch@diw.de
E-mail: avogel@uni.leuphana.de

---

1
Motivation – Aim

In the last few years, the internationalisation of the economy has continued to increase
undiminished. Accordingly, world trade is growing faster than the individual economies. This
internationalisation is mainly determined by the exchange of goods, but more and more
frequently by the exchange of services. The economies that wish to benefit from the growth of
the world markets have to be successful not only in trading commodities but also services.
The growing internationalisation is mirrored in the German economy. This applies in
particular to trading of products: in 2007, German companies exported 696 billion euros’
worth of goods, according to the balance of payments. This was 62 percent more than in 2000
(in current prices). In addition, services were exported on a large scale. In 2007, the export of
services (not including travel expenses) amounted to 135 billion euros. This was up 86
percent on 2000 (in current prices) and thus represented even greater growth than that of
products.
In comparison with the export of commodities, the export of services makes other
demands on the companies. Services are not generally standardised products: they are mostly
customised and require intensive communication and interaction with clients. For this,
geographical proximity is normally necessary. However, the limitations for export are reduced
by developments in information and communication technologies. Companies are able to
communicate with customers and suppliers long-distance.
Due to the above-mentioned developments, it is highly probable that the export
orientation of service companies has increased over the last few years. However, there has
only been limited information about the export behaviour of service companies, in contrast to
that of manufacturing companies. Information on export behaviour is important in order to
explore the prospects for internationalisation of companies.
To close this gap, this paper contributes to the literature by investigating for the first
time the determinants of export behaviour in German business services enterprises at firm
level. We focus our analysis on enterprises in selected lines of business such as transport,
storage and communication, real estate, renting and business activities. With 680,000 firms, 6
million employees and a total turnover of 700 billion euros in 2005, these sectors are of
particular importance for the German economy. The report is based on the official German
statistics on business services (Strukturerhebung im Dienstleistungsbereich) which was
launched in 2000. This is a unique set of data containing information on, inter alia, export,

2

---

turnover, labour costs, number of persons employed and gross investments. The statistics
cover the period from 2000 to 2005.
We begin our analysis by applying a well-established methodology. We estimate the
export behaviour using cross-sectional probit and fractional probit regressions. The first
estimations investigate the probability of a company exporting or not exporting. The second
approach also captures the export intensity of a company.
Further, we extend the analysis into a panel estimation by means of a recently
introduced pooled fractional probit estimator developed by Papke & Wooldridge (2008) and
rarely used to date. Thus, we are able to consider unobserved time-invariant characteristics of
the enterprises involved in our analyses. This approach also takes into account the specific
nature of the export intensity (percentage of exports to total turnover) as the dependent
variable (Wagner 2008). For these panel econometric analyses, we use a balanced panel data
set of the German business services statistics for the years 2003 to 2005.
Then Section 2 begins with an overview of the literature about the determinants of
export performance. The German business services statistics are described in Section 3, while
Section 4 describes our empirical model and estimation strategy. In Section 5 and 6, the
results of the descriptive and econometric analyses are presented. The final section
summarises the findings and discusses their implications.
2
The determinants of export performance: literature survey

Within the economics literature, determinants of export behaviour (namely the
probability of being an exporter and export intensity, defined as the share of exports in total
turnover) have been widely investigated in the manufacturing sector. Evidence is available,
for example, for Germany (e.g. Arndt et al. 2008, Engelmann & Fuchs 2008, Roper & Love
2002, Wagner 2001, Wagner 2008), the United Kingdom (Bleaney & Wakelin 1999, Roper &
Love 2002, Wakelin 1998), the United States (Bernard & Jensen 1999), Ireland and Northern
Ireland (Roper et al. 2006), Italy (Sterlacchini 2001) and also for developing countries such as
Indonesia (van Dijk 2002) and the Philippines (Dueñas-Caparas 2007). In contrast to studies
of the manufacturing sector, there are only a few economics-based empirical studies about the
determinants of export activities in the service sector (Ebling & Janz 1999 for Germany,
Gourlay et al. 2005 for the United Kingdom, Chiru 2007 for Canada and Love & Mansury
2007 for the United States).
Even if the results differ according to industry (e.g. Wagner 2001), size (e.g.
Sterlacchini 2001) and country, overall innovation, human capital, size and productivity are

3

---

important determinants of export performance as reported in this literature. These
determinants are briefly reviewed below. The product cycle theory (Vernon 1966) and the
technology gap theory (Krugman 1979) suggest that innovation provides countries and
industries with comparative advantages and is therefore the driving force behind exports.
Similar conclusions also emerge from studies at firm level. For the manufacturing sector
overall, a positive effect of innovation (e.g. measured by R&D expenditures or innovative
products) on exporting activities is found in Germany (e.g. Engelmann & Fuchs 2008, Roper
& Love 2002, Wagner 2001) and other developed countries (e.g. Wakelin 1998, Sterlacchini
2001). In this context, capital intensity as an indicator of firm assets embodying past
innovations and reflecting economies of scale is also expected to have a positive effect
(Wakelin 1998). Similar to the manufacturing sector, in the business services sector, too,
innovativeness is predominantly positively associated with the probability of exporting
(Ebling & Janz 1999, Gourlay et al. 2005, Love & Mansury 2007) and the export intensity.
(See Chiru 2007, Gourlay et al. 2005, but, conversely, Love & Mansury 2007 show a negative
effect). Furthermore, a positive effect of human capital on export performance is expected due
to the fact that skills are positive with respect to the technological capabilities of the firm and
that highly educated employees have certain abilities that make it easier to establish and
maintain certain contacts with the foreign market. Because of the high level of interaction
between user and provider, particularly in the service sector, employees must have good
language skills and a high level of intercultural competence (cf. McLaughlin & Fitzsimmons
1996, Winstead & Patterson 1998). Overall, a positive relationship between human capital
and exports is confirmed in the empirical literature on both the manufacturing sector (e.g.
Roper et al. 2006, Wagner 2001, Wakelin 1998) and the business services sector (e.g. Ebling
& Janz 1999, Gourlay et al. 2005, Chiru 2007).
Concerning a positive effect of firm size, it is argued in the literature that larger firms
can, for instance, better absorb the risks associated with internationalisation, have better
opportunities to raise financing and that they have more resources to overcome the fixed or
sunk costs associated with foreign market entry. (See, for example, Aaby & Slater 1989,
Wagner 1995, Bernard & Jensen 1999). To explain the frequently found inverted u-shaped
size effect, it is argued that large firms may be more oriented towards the domestic market if,
for instance, a domestic monopoly gives them no incentive to export (Wakelin 1998), and that
there are limits to the advantage of size because coordination costs increase as the scale of
operation increases, and, at some point, further expansion is not profitable (Wagner 2001).
However, in the business services sector, there is mixed evidence regarding the effect of size

4

---

on export. Concerning the probability of exporting, Love and Mansury (2007) showed a
hump-shaped relationship, Gourlay et al. (2005) found a linear positive effect, and Ebling and
Janz (1999) found no significant effect. Concerning the export intensity, Chiru (2007) showed
a u-shaped relationship, Gourlay et al. (2005) found a hump-shaped relationship, and Love
and Mansury (2007) found no significant effect.
Explanations for the positive effect of productivity on exports are found in the more
intensive competition in international markets as well as in additional costs entailed, for
example, transportation, tariffs, market research, product adaptations and setting up new
distribution networks. Only more productive firms are able to absorb these costs and to
overcome the entry barrier (formally shown by Melitz 2003). A wide rage of empirical studies
document productivity differences between exporting and non-exporting firms for the
manufacturing sector (see Wagner 2007 for a survey) and also for the business services sector
initial evidence shows a higher productivity for exporting firms than for non-exporting firms
(e.g. Jensen 2008, Vogel 2009).
In addition to innovation, human capital, size and productivity, other determinants are
also analysed in the economics literature. Since ownership may also be an important indicator
of a firm’s export potential, for example, by taking advantage of group resources for
marketing or distribution (Roper et al. 2006), a positive effect of foreign ownership on exports
is shown by Roper et al. (2006) for manufacturing firms in Ireland and North Ireland and by
Engelmann and Fuchs (2008) for eastern German establishments. Gourlay et al. (2005)
suggest a positive effect of product diversification on the basis that a more diversified firm is
likely to have more products that will be profitable in foreign markets, but no significant
influence was found. And recent studies show that financially constrained firms are less likely
to export since they may be less able to cover the additional costs related to exporting than
unconstrained firms (e.g. Arndt et al. 2008, Bellone et al. 2008). However, Wagner (2008,
2003) demonstrates the importance of unobserved heterogeneity for the manufacturing sector
in an analysis of the export performance of firms. Thus, it is not the observed characteristics
(such as human capital or R&D intensity) per se that make a successful exporter, but
unobserved time-constant characteristics correlated with these observed characteristics
(Wagner 2008).
There is also a wide range of studies on export performance in the management and
marketing literature. Firm characteristics such as firm performance, size or innovation
activities are important aspects in this literature, too. However, other internal determinants
such as the marketing strategy or management characteristics as well as external determinants

5

---

such as characteristics of the foreign or domestic market seem to be equally important (See
Sousa et al. 2008, Zou & Stan 1998 for an overview). According to traditional models of this
literature, internationalisation is seen as an incremental process that depends on the ability to
accumulate knowledge through exposure to foreign markets. Thus, the step-by-step
internationalisation of firms begins in markets that are similar to the home market and
continues with entry into new markets with successively greater psychic distance (Johanson &
Vahlne 1977, 1990). Roberts (1999) presents evidence that also in the business services
sector, firms progress through various stages in the process of internationalisation.
3
Data source: the German business services statistics

In order to investigate the export behaviour of German business services enterprises,
we use the business services statistics (Strukturerhebung im Dienstleistungsbereich)
established by the German Federal Statistical Office and the statistical offices of the Federal
States (Länder). The statistics were first compiled for the year 2000 on the initiative of the
European Union. This structural survey comprises service activities included in Section I
(“Transport, storage and communication”) and Section K (“Real estate, renting and business
activities”) according to the Statistical Classification of Economic Activities in the European
Community NACE Rev. 1.1 (European Commission 2002). Companies from these lines of
business may be asked to provide information to the statistical offices of the Länder on an
annual basis. This applies to all companies that are subject to turnover tax and to professions
(Freie Berufe) with a turnover of 17,500 euros or more per annum. 15 percent of these receive
a questionnaire from the statistical offices and are asked to participate in the survey. The
companies were randomly selected according to the sample criteria of federal state (Land),
line of business and turnover. Because the same enterprises that participated in 2003 also
participate in 2004 and 2005, it is possible to merge the cross-sectional data sets to a panel
data set that covers the years 2003 to 2005 (Pesch 2007, Federal Statistical Office 2007).
The business services statistics include, among other data, information about the
economic sector, the number of persons employed (not including temporary workers), total
turnover, salaries and wages, and export – defined as turnover for business with companies
located abroad, including exports to foreign affiliates. Unfortunately, the target countries of
exports are not included in the statistics. Also, no information is obtained about other forms of
companies’ activities abroad, such as cooperation, direct investment or imports. Furthermore,
small enterprises with an annual turnover lower than 250,000 euros are given a shorter
questionnaire, so important information, such as information about export activities, is

6

---

missing for these enterprises. As a result, only enterprises with an annual turnover over
250,000 euros are considered for the analyses. For this study, the companies’ responses for
the years 2000 to 2005 were made anonymous and available to the authors by the research
data centres of the Federal Statistical Office and the statistical offices of the Länder. For more
details about the data access, see Zühlke et al. (2004).
In 2005, there were 680,000 companies active in Sections I and K, with 6 million
employees and a total turnover of 700 billion euros. Almost 184,000 of the companies had a
turnover of 250,000 euros per annum or more. These companies had an overall turnover of
625 billion euros, export amounting to almost 38 billion euros and just under 1.5 million
employees.
4
Empirical model

The dependent variable export behaviour is specified in two ways. First, export
behaviour is specified as a binary variable indicating the “export status” of the enterprise (1 if
exporting, 0 if not). In a second variant, export behaviour is captured by the variable “export
intensity” as the percentage of exports to total turnover.
The enterprise characteristics used here to explain the export performance are derived
from the theoretical assumptions and empirical evidence reported in Section 2.
In line with previous studies, we expect size to have a positive relation to the export
behaviour of the enterprises: Large firms have more resources to enter foreign markets than
small companies have. This is mainly due to the fact that there are fixed costs needed for
exporting such as gathering specific information about the respective foreign market, specific
qualifications (languages, soft skills, etc.), marketing, travelling, operating plants, etc. Here,
firm size is measured by the number of employees. However, in order to test for a possible
non-linear relation to the export activity, the second order term of the number of employees
has also been introduced.
Productivity as a determinant for export is widely tested in the literature. Based on the
argument of additional costs caused by exporting that can only absorbed by more productive
enterprises, a positive effect of productivity on export behaviour is expected. The variable is
measured as labour productivity (value added per employed person). The empirical definition
follows the definition applied for the “Structural Business Statistics” of the European
Commission (European Commission 1998).
Human capital is a factor that also has a positive impact on the probability of
companies to export, according to the literature. Most of the studies use per capita wages as a

7

---

proxy for human capital. We use the comprehensive definition of labour costs, made up of
wages, salaries and employers’ social security costs per employee. More appropriate would be
the relation between labour costs and the hours worked. However, the data set does not
contain information on hours worked. In order to control whether using the number of
employees is misleading, we employ available information on the proportion of employees
who work part time. In line with the literature, we expect a positive relationship between
human capital and export propensity. For the control variable part-time work, we expect a
negative relationship with export propensity.
To consider the influence of financial constraints on export activities (e.g. shown by
Arndt et al. 2008), we use the legal status of a firm as an indicator to measure the possibility
of financing business operations by external sources. There are three dummy variables, one if
the firm is owned by a sole proprietor, one if the firm is a business partnership and one if the
company is a limited liability company, such as a stock company or a limited company. Thus,
the liability of the company’s owner is indicated. Limited liability companies are expected to
have a higher probability of exporting since it is easier for them to finance the additional
(sunk) costs related with exporting by external sources compared to companies with a sole
proprietor.
Following Gourlay et al. (2005), we test the potential role of product diversification on
export performance. Based on the argument that a more diversified firm is likely to have more
products that will be profitable in foreign markets, we expect a positive effect. There are
different ways to measure product diversification. One way, which is taken here, is according
to purchases of products or services that are not produced by the company itself but were
purchased explicitly for resale in the same condition as received. We use the share of total
turnover represented by this type of purchase as an indicator.
Our model also incorporates a variable on the market behaviour of companies which
has not been taken into account in other studies to date. Following the idea of the stage model
that regards internationalisation as an incremental process, we argue that for firms that are
experienced in serving the nationwide market, the probability of entering international
markets is higher than for firms only focused on the local or regional market. We capture the
capability of companies to operate nationwide by the number of subsidiaries within Germany.
It is expected that for companies with subsidiaries in Germany, the probability of exporting is
higher than for companies without any subsidiaries.
Furthermore, we consider expectations of growth by including investment activities.
Firms that expect to grow in the coming years and have reached the limits of their capacities

8

---

will invest in machinery, buildings, land and other assets. Although it is not known if the
investments are targeted towards expansion on the domestic or foreign market, export
activities may be either started or expanded. Investment activities are measured in this paper
as the investment intensity, the relationship of gross investment to the number of employees.
We expect a positive impact of investment intensity on export behaviour.
In order to account for regional differences, we include a dummy that indicates if the
enterprise is located in eastern Germany or in western Germany. Taking into consideration
that the eastern German economy, even almost 20 years after German reunification, is still
weaker than the West German economy, a negative coefficient of the eastern German dummy
is expected.
Finally, we control for specific market conditions of companies, including a set of
dummies for the economic activities of the companies by using information about the
companies’ lines of business. To sum up, the above-mentioned variables and their expected
effects are presented in Table 1.
Formally, our model can be expressed as

(1)
Exportit = β0 + β1 Xit + β2 Cit + εit

where i is the enterprise index, t is the index of the year. The dependent variable
Export is either the “export status” or the “export intensity”, as defined. The vector X contains
the explanatory variables, namely the number of employees and its squared value, labour
productivity, the average wage, the share of part-time employees, dummies that indicate the
legal status, the share of goods and services for resale, dummies for nationwide active firms,
and per-capita-investments.1 C indicates the control vector that contains the economic activity
dummies, the region dummy, and, in the case of pooled analyses, a set of year dummies. β0
represents the constant, β1 and β2 indicate the vectors of coefficients, and ε is the error term.

[Table 1 about here]

Our investigation of the export activities of business services firms is separated into
two parts: first, we estimate the determinants of the “export status” (the probability of being

1 To check the robustness of the results, in addition, we estimate a model where all explanatory variables X
are lagged by one period to minimise problems of endogeneity with the dependent variable. Compared to
the model without lagged explanatory variables, the results in terms of signs and significance levels are
equal. However, also in the literature about the learning-by-exporting hypotheses, no clear evidence has
been found that exporting fosters the performance of the enterprises. (See Wagner 2007 for a survey.)

9

---

an exporter) and the determinants of the “export intensity”. To explain the binary variable
“export status” we estimate Equation (1) using a probit regression model. We test for the
years 2003 to 2005 separately and pooled for the respective years.2 Thus, we can compare the
results of our tests with other studies using similar methodology. Equation (1) is then
estimated by a procedure that exhausts all the information about export behaviour by applying
the fractional probit estimator developed by Papke and Wooldridge (1996). Wagner (2001)
points out that, in contrast to a tobit regression or a two-step approach, like a probit regression
followed by a truncated regression, the regression by Papke and Wooldridge considers both
aspects for export behaviour, the fact that a firm does not distinguish between the decision if
and how much it exports and that the export intensity is bounded between one and zero (with
the possibility of observing values at the boundaries) by definition rather than as a result of
censoring.
As a second step, we also control for unobserved time-invariant characteristics that
could be correlated with the explanatory variables, by estimating a fractional response model
for panel data (following Wagner 2008). Papke and Wooldridge (2008) show that in the case
of a balanced panel dataset (with large cross-sectional dimension and only few time periods),
it is controlled for fixed effects by adding the time averages of all explanatory variables to the
fractional probit approach we applied in the first step. In line with this approach, we use a
balanced panel dataset for the years 2003 to 2005. To facilitate the comparison with the
results of the first step, we estimate both a variant without fixed effects that is similar to the
cross-sectional analyses and a second variant where the time averages of all explanatory
variables are added to a pooled form of Equation (1) to control for unobserved heterogeneity.
All models are estimated with robust or, in the case of pooled data, cluster-robust
standard errors. The regressions were run using the Stata program (Version 10). According to
Papke and Wooldridge’s approach, regressions are estimated with the Stata command for
generalised linear models.


2 The pooled analysis is only possible for the time period 2003 to 2005. However, to check the robustness of
the results, we also used the years 2000 to 2002 for cross-sectional analyses. (The results are available in the
Appendix).

10

---

Document Outline

• 1 Motivation - Aim
• 2 The determinants of export performance: literature survey
• 3 Data Source: the German business services statistics
• 4 Empirical model
• 5 Descriptive analysis
  • 5.1 Export behaviour
  • 5.2 Differences between exporting and non-exporting firms
• 6 Estimation results
  • 6.1 Determinants of the export behaviour: cross-section results
  • 6.2 The role of unobserves time-invariant characteristics
• 7 Concluding remarks
• References
• Tables and Figures

---