International Economics and Economic Policy

, Volume 5, Issue 4, pp 363–370

A note on why more West than East German firms export

Authors

    • Institute of EconomicsLeuphana University Lueneburg
Original Paper

DOI: 10.1007/s10368-008-0119-7

Cite this article as:
Wagner, J. Int Econ Econ Policy (2008) 5: 363. doi:10.1007/s10368-008-0119-7

Abstract

Germany is one of the most important exporters of manufacturing goods in the world, but by far not all manufacturing firms in Germany are exporters, and there is a remarkable gap between the share of exporters in all manufacturing firms between West Germany and East Germany. While in West Germany in 2004 about two in three manufacturing plants were exporters, fourteen years after re-unification this share was less than fifty percent in the former communist East Germany. Given that exports play a key role in shaping business cycles and growth in Germany, and the much higher unemployment in East compared to West Germany, promotion of exports by East German firms figure prominently on the policy agenda. However, the reasons for the large difference in the propensity to export between East and West German firms are not yet well understood, not least due to a lack of comprehensive micro data. Using unique new data and a recently introduced non-linear decomposition technique this paper shows that the huge difference in the propensity to export between West and East German plants can only partly be explained by differences in firm size, productivity, and technology intensity.

Keywords

ExportsMicro dataWest GermanyEast Germany

JEL classification

F14

1 Motivation

Germany is one of the most important exporters of manufacturing goods in the world, but not all manufacturing firms in Germany are exporters, and there is a remarkable gap between the share of exporters in all manufacturing firms between West Germany and East Germany. While in West Germany in 2004 about two in three manufacturing plants were exporters, fourteen years after re-unification this share was less than fifty percent in the former communist East Germany. Given that exports play a key role in shaping business cycles and growth in Germany, and the much higher unemployment in East compared to West Germany, promotion of exports by East German firms figure prominently on the policy agenda.

However, the reasons for the large difference in the propensity to export between East and West German firms are not yet well understood, not least due to a lack of comprehensive micro data. Leber (2002) uses establishment level data from a rather small sample of East and West German plants (the so-called IAB Betriebspanel, an establishment level panel data set collected by the research institute of the German Federal Labor Agency) for a descriptive study of export activities. Furthermore, she estimates probit models to document the role of human capital, technology and firm size for participation of plants in export markets, but she does not further investigate the differences between East and West Germany. Using the same data set Loose and Ludwig (2006) focus on East German plants and their export activities only. Comprehensive descriptive evidence on exports by West and East German firms can be found in Institut für Mittelstandsforschung (2007).

Using a unique new data set and a recently introduced non-linear decomposition technique this paper contributes to the literature by investigating the gap in the propensity to export between firms in East and West Germany, and by documenting the share of this gap that is due to observed plant characteristics. The rest of the paper is organised as follows: Section 2 introduces the newly available data. Section 3 reports descriptive evidence. Section 4 outlines the non-linear decomposition technique and presents results from its application. Section 5 concludes.

2 Data

The empirical investigation uses data for plants taken from regular surveys by the Statistical Offices of the German federal states covering all plants from manufacturing industries that employ at least twenty persons in the local production unit or in the company that owns the unit. Participation of plants in the survey is mandated. Late in 2006 these data were matched over all federal states for the first time to form a data set that covers Germany as a whole (for a description of the data see Konold 2007). In this paper the most recent available data for 2004 are used.

Although this data set is comprehensive in respect of its coverage of plants it is weak on detail. It does contain information on the number of employees, domestic turnover and turnover abroad, total amount of wages and salaries, branch of industry, and the location of the unit. But there is no information on variables such as capital stock, research and development activities, innovations, or subsidies received, all of which can be expected to be important determinants of the decision of whether or not to export.

Note that the micro level data are strictly confidential and for use inside the Statistical Office only, but not exclusive. Further information how to access the data is given in Zühlke et al. (2004).

3 Descriptive evidence

As shown in Table 1 the share of exporters in all manufacturing firms1 was much lower in East Germany (46.2%) than in West Germany (65%) in 2004. According to the literature on the microeconomics of export activities participation of firms in export markets is linked to firm size. Firm size is expected to be positively correlated with export activities for various reasons including scale effects, a higher capacity for taking risks in larger firms, and the fixed costs character of various export related costs like retooling and redesigning products for foreign markets (see Wagner 2001). Furthermore, firms from one of the most highly developed industrial countries of the world can be expected to have a comparative advantage in technology intensive products, and the more technologically advanced firms can be expected to have a higher probability to export. The recent literature on the export activities of heterogeneous firms points to the important role of productivity for the participation in export markets. It is argued that only the more productive firms are able to cover the extra costs linked to export activities (like international transport costs, tariffs, or retooling products for foreign markets) and to stay competitive when facing these extra costs (see Wagner 2007b for a survey of studies on exports and productivity, and Wagner 2007c for Germany).

Using the information from the data described in section 2, the following variables are considered as determinants of participation in exporting:2
  • Firm size, measured as the number of employees.

  • Labor productivity, measured as sales per employee. Note that the data set does not include information on value added and capital stock, so neither value added per employee nor total factor productivity can be used to measure productivity. Furthermore, productivity can be considered to be a determinant of exporting, because empirical studies show that there is self-selection of more productive firms into exporting, while learning-by-exporting and higher productivity as a consequence of exporting is rarely found (and not for Germany, see Wagner 2007c). Given that sales per employee often show extremely high or low values for a small number of firms (for reasons that can not be investigated due to the strict confidentiality of the firm level data, but that may or may not be related to reporting errors), all computations were done with and without the top and bottom one percent of firms from the productivity distribution.

  • Technology, measured by two dummy variables indicating that the firm is from an industry classified as high-tech or medium-tech (using the rest of manufacturing as the reference category). The classification is based on the average percentage share of spending on research and development in total sales in an industry. Industries are classified as high-tech if this share is higher than 8.5%, and as medium-tech if it is between 3.5% and 8.5% (for details, see Bundesminsterium für Bildung und Forschung 2002, Annex). Note that the data used here do not include any direct information on research and development activities of the firms (see section 2).

Table 1 documents that in both parts of Germany exporters were larger, more productive, and more often from technology intensive industries than their non-exporting counterparts. All these differences are statistically significant at an error level of one percent or better. Results for probit models show that these links, which are in line with our theoretical priors for export participation, are statistically highly significant ceteris paribus, too.
Table 1

Export participation of manufacturing firms in West and East Germany, 2004

 

West Germany

East Germany

Results for probit estimation of export participation

Sample mean

Results for probit estimation of export participation

Sample mean

Estimated coefficient (p-value)

Exporters Non-exporters

Estimated coefficient (p-value)

Exporters Non-exporters

(1) Results for all firms

Number of Employees

0.00174 (0.000)

179.71

0.00187 (0.000)

114.27

59.71

54.58

Labour productivity

1.30e-7 (0.006)

195588

4.89e-7 (0.000)

169738

159413

130959

High-tech (dummy)

0.501 (0.000)

0.059

0.574 (0.000)

0.074

0.028

0.034

Medium-tech (dummy)

0.492 (0.000)

0.214

0.445 (0.000)

0.212

0.104

0.120

Constant

0.093 (0.000)

   

Number of firms

36729

 

8523

 

Share of exporters (percentage)

65.02

 

46.20

 

(2) Results without firms from top and bottom 1 percent of the productivity distribution

Number of employees

0.00195 (0.000)

179.39

0.00189 (0.000)

113.90

58.04

54.00

Labour productivity

1.15e-6 (0.000)

173703

9.98e-7 (0.000)

151230

136354

120958

High-tech (dummy)

0.500 (0.000)

0.059

0.567 (0.000)

0.074

0.028

0.035

Medium-tech (dummy)

0.505 (0.000)

0.215

0.465 (0.000)

0.213

0.104

0.119

Constant

−0.064 (0.000)

 

−0.476 (0.000)

 

Number of firms

36003

 

8355

 

Share of exporters (percentage)

65.53

 

46.25

 

4 A decomposition of the difference in the propensity to export in West and East German firms

The figures reported in Table 1 reveal a number of differences between West and East German plants regarding the size of the estimated coefficients of the probit models, and in the composition of the samples with regard to these characteristics. Therefore, the question arises to what extent the difference in export participation across space can be explained by differences in characteristics of the firms on the one hand, and by differences in the coefficients on the other hand.

This type of question is familiar from other fields of economics. A case in point is the decomposition of the earnings differential between groups of workers (for example, males and females) into a share that can be explained by differences in characteristics that are related to productivity (years of schooling, years of experience etc.) and the rest that is due to differences in the rates of return to these characteristics (often labelled discrimination). This kind of decomposition is based on earnings functions (linking earnings to its determinants) that are estimated separately for samples of employees from both groups. It was introduced by Blinder (1973) and Oaxaca (1973), has been used in hundreds of empirical studies ever since, and is covered in standard textbooks of labour economics (e.g., Cahuc and Zylberberg 2004: 280ff.).

To illustrate, consider the following hypothetical example related to the question dealt with in this paper: The observed share of exporters in Axistan is 25%, and 5% in Uxistan, leading to a an exporter gap of 20 percentage points in favour of Axistan. To investigate the reasons for this difference between the two countries identically specified empirical models that explain the participation in exporting are estimated using data for firms from each country. If the estimated share of exporters among the firms from the Uxistan sample, using the coefficients calculated for Axistan, is 15%, this means that 10 percentage points (or 50%) of the difference in the share of exporters between the two countries can be explained by differences in the characteristics of firms between Axistan and Uxistan. The other 10 percentage points are due to cross-country differences in the coefficients linking these characteristics to the probability of exporting and due to unobservable or unmeasured factors. In this case, both the characteristics effect and the residual effect are 10 percentage points. If, however, the estimated share of exporting firms for the firms from the Axistan sample, using the coefficients calculated for Uxistan, is 10%, this means that 25% of the difference in the share of exporters between the two countries can be explained by differences in the characteristics of firms between Axistan and Uxistan. Given that the choice of the reference group (Axistan or Uxistan) is arbitrary, we would conclude that between 25 and 50% of the cross-country difference in the share of exporters is due to observed differences between the two groups of firms.

While the Blinder-Oaxaca decomposition technique is easy to apply if the outcome variable is continuous like earnings, a problem arises if the outcome is binary (like exporting or not) and the coefficients are from a (non-linear) probit model because these coefficients cannot be used directly in the standard Blinder-Oaxaca decomposition equations. Fairlie (2006) introduced a decomposition method based on estimates from a non-linear probit model. While a discussion of the details of this method is beyond the scope of this paper, two aspects should be mentioned: First, while the characteristics effect identified in the decomposition represents the part of the difference in export participation that is due to observed differences over the two regions in the characteristics of the firms, the residual effect not only represents the part due to different regression coefficients but captures also the proportion of the difference due to group differences in unmeasured or unobservable factors. Second, each sub-sample can be used as the reference group, and the results usually differ according to the choice of the reference group. Therefore, both variants are computed, and the results are compared.

The type of question answered here is “How high would the share of exporting firms among all manufacturing firms in East Germany have been in 2004 if the firms from the West German sample were located in East Germany, and if the characteristics of these West German firms were linked to the probability of exporting according to the coefficients estimated using the East German sample from 2004?” Results are reported in Table 2, where panel 1 reports results for all firms and panel 2 reports results for a sample that does not include firms from the top and bottom one percent of the productivity distribution.3
Table 2

Decomposition analysis of differences in export participation of manufacturing firms in West and East Germany, 2004

Reference group (percent exporters in sample)

Comparison group (percent exporters in sample)

Difference in participation (percentage points)

Characterisitcs effect (percentage points)

Residual effect (percentage points)

(1) Results for all firms

West

East

   

65.02

46.20

18.82

1.92

16.90

East

West

   

46.20

65.02

−18.82

−2.78

−16.04

Detailed decompositon

  

Characteristic effect (percentage points)

Significance level (p-value)

Reference group: West Germany

 Number of employees

1.80

0.000

 Labor productivity

0.14

0.000

 High-tech (dummy)

0.06

0.000

 Medium-tech (dummy)

−0.08

0.000

Reference group: East Germany

 Number of employees

−1.84

0.000

 Labor productivity

−0.56

0.000

 High-tech (dummy)

0.007

0.649

 Medium-tech (dummy)

−0.39

0.000

(2) Results without firms from top and bottom 1% of the productivity distribution

West

East

   

65.53

46.25

19.28

2.92

16.36

East

West

   

46.25

65.53

−19.28

−3.27

−16.01

Reference group: West Germany

 Number of employees

1.91

0.000

 Labor productivity

1.14

0.000

 High-tech (dummy)

0.005

0.465

 Medium-tech (dummy)

−0.14

0.000

Reference group: East Germany

 Number of employees

−1.84

0.000

 Labor productivity

−0.98

0.000

 High-tech (dummy)

0.02

0.188

 Medium-tech (dummy)

−0.47

0.000

When all firms are included in the calculations and when West German firms are used as the reference group, 10% of the difference in the export participation rate is allocated to observed firm characteristics included in the probit regression. This part is considerably higher (about 15%) when the reference group is formed by East German firms. The detailed decomposition shows that the lion’s share of this characteristics effect is due to the much larger size of West German plants, and that the larger productivity of West German plants matters, too, but to a much smaller amount. The point estimates for the two technology group dummy variables tend to be tiny (and not always statistically significant at a usual level), and for each reference group one of the coefficients has the “wrong” sign. Differences in technology intensity – at least when (due to a lack of better information) measured by average R&D intensities at the industry level – between East and West Germany do not contribute to the explanation of the huge difference in the export participation rate. The picture is very similar if the firms from the top and bottom one percent of the productivity distribution are dropped (see panel 2 in Table 2). When West German firms are used as the reference group, 15% of the difference in the export participation rate is explained by observed firm characteristics from the probit regression, while this part is slightly higher (about 17%) when the reference group is formed by East German firms.

5 Conclusions

According to the results from a new comprehensive data set and a recently introduced non-linear decomposition technique, between 10 and 17% of the large difference in the share of exporting firms in West and East Germany can be explained by the larger firm size and – to a smaller degree – by the larger productivity of West German firms. On the one hand, this is an important result that helps to understand why this difference is still that large many years after re-unification. According to these findings policies that intend to increase the presence of East German firms on international markets should focus on fostering the growth of size and productivity of these firms.

On the other hand, the residual effect – which is at least in part a measure of our ignorance because it not only represents the part due to different regression coefficients but also captures the proportion of the difference due to group differences in unmeasured or unobservable factors – is more than 80%. This clearly demonstrates the need for further research based on more informative plant level data that will hopefully allow the inclusion of more plant characteristics in the decomposition. Furthermore, such data can be used to investigate the difference in the share of exports in total sales between West and East German manufacturing firms. While this share was 18.83% for the West German firms in our sample, it was only 10.89% for the East German firms. These figures clearly show that East German manufacturing firms do not export a similar share of their production but concentrate this production of export goods in a lower number of firms. The different shares of exporters in all firms in East compared to West Germany under-estimate the difference in the propensity to export between West and East Germany. Studies that investigate this difference with more informative data are a prerequisite for a sound discussion of any policy programs that intend to strengthen the export performance of East German firms.

Footnotes
1

In this paper the term firm is used to mean a local production unit, or plant.

 
2

The earlier version of this paper (Wagner 2007a) used wages per employee to proxy human capital as a determinant of exporting. While wages can be expected to be positively correlated with human capital, including wages per employee in a study with a focus on differences between export activities of East and West German firms does not make sense because East German firms are known to use a low wage strategy to establish their competitiveness. Differences in wages per employee between East and West German firms, therefore, do not reflect differences in human capital (alone), but also differences in the rates of return to human capital between East und West Germany. I am grateful to two anonymous referees to point this out.

 
3

Stata 9.2 and the program fairlie.ado were used for computations.

 

Acknowledgement

All computations were done inside the Research Data Centre of the Statistical Office of Berlin. Comments from three anonymous referees that lead to a thorough revision of an earlier version (Wagner 2007a) are gratefully acknowledged.

Copyright information

© Springer-Verlag 2008