Empirical Studies with New German Firm Level Data from Official Statistics: Themenheft Heft 3/Bd. 231 (2011) Jahrbücher für Nationalökonomie und Statistik 9783110510287, 9783828205413

Citation preview

Empirical Studies with New German Firm Level Data from Official Statistics

E d i t e d by Anja Malchin R a m o n a Voshage Joachim Wagner

W i t h C o n t r i b u t i o n s by Boneberg, Franziska, Lueneburg Braakmann, Nils, Newcastle upon Tyne, U.K. Jeßberger, Christoph, Munich Kilian, Stefan, Freising Petrick, Sebastian, Kiel Rehdanz, Katrin, Kiel

Lucius & Lucius · Stuttgart 2 0 1 1

Röder, Norbert, Braunschweig Schiersch, Alexander, Berlin Schmidt-Ehmcke, Jens, Berlin Sindram, Maximilian, Munich Wagner, Joachim, Lueneburg Wagner, Ulrich J., Madrid Zimmer, Markus, Munich

Anschriften der Herausgeber des Themenheftes Anja Malchin Amt für Statistik Berlin-Brandenburg Forschungsdatenzentrum Alt-Friedrichsfelde 6 0 1 0 3 1 5 Berlin, Germany [email protected] Ramona Voshage Amt für Statistik Berlin-Brandenburg Forschungsdatenzentrum Alt-Friedrichsfelde 6 0 1 0 3 1 5 Berlin, Germany [email protected] Prof. Dr. Joachim Wagner Leuphana Universität Lüneburg P.O. Box 2 4 4 0 2 1 3 1 4 Lueneburg, Germany wagner@uni. leuphana .de

Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.d-nb.de abrufbar ISBN 9 7 8 - 3 - 8 2 8 2 - 0 5 4 1 - 3

© Lucius Sc Lucius Verlagsgesellschaft mbH · Stuttgart - 2 0 1 1 Gerokstraße 5 1 , D - 7 9 1 8 4 Stuttgart Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlags unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Ubersetzungen und Mikroverfilmungen sowie die Einspeicherung und Verarbeitung in elektronischen Systemen.

Satz: Mitterweger Sc Partner Kommunikationsgesellschaft mbH, Plankstadt Druck und Bindung: Neumann Druck, Heidelberg Printed in Germany

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2011) Bd. (Vol.) 231/3

Inhalt / Contents Guest Editorial

324-325

Abhandlungen/Original Papers Braakmann, Nils, Joachim Wagner, Product Diversification and Profitability in German Manufacturing Firms Schiersch, Alexander, Jens Schmidt-Ehmcke, Is the Boone-Indicator Applicable? - Evidence from a Combined Data Set of German Manufacturing Enterprises Röder, Norbert, Stefan Kilian, Which Parameters Determine the Development of Farm Numbers in Germany? Petrick, Sebastian, Katrin Rehdanz, Ulrich J. Wagner, Energy Use Patterns in German Industry: Evidence from Plant-level Data Jeßberger, Christoph, Maximilian Sindram, Markus Zimmer, Global Warming Induced Water-Cycle Changes and Industrial Production - A Scenario Analysis for the Upper Danube River Basin . . . . Börnberg, Franziska, The Economic Consequences of One-third Co-determination in German Supervisory Boards

326-335

336-357 358-378 379-414

415-439 440-457

Buchbesprechungen / Book Reviews Bourg, Jean François, Jean Jaques Gouguet, The Political Economy of Professional Sport Klein, Michael W., Jay C. Shambaugh, Exchange Rate Regimes in the Modern Era Weiß, Mirko, Zur Geldpolitik im Euro-Währungsraum: Beschreibung, Auswirkung und Ursachenanalyse von Inflationsunterschieden

458 459 462

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2011) Bd. (Vol.) 231/3

Guest Editorial The German Federal Statistical Office and the statistical offices of the German federal states opened research data centres (described in detail in Ziihlke et al. 2007) in 2001 and 2002. This started a new era for researchers working in empirical economics. Access to confidential individual and firm data for individuals and firms that were collected in surveys performed by the statistical offices became easy (and available at low costs) by using these research data centres (RDC). The number and variety of data sets provided by the RDC increased steadily (see Kaiser/Wagner 2008 for an overview), and so did the use of it by researchers. The high potential of these data as a basis to generate new stylized facts, to motivate assumptions used in formal theoretical models, to test theoretical hypotheses econometrically, and to be used in policy consultation and evaluation, is documented in a large and growing number of publications.1 From their start the RDC offered access to micro level panel data that linked information from various waves of a survey over time. These panel data enormously extended the research potential of data from official statistics by allowing dynamic analyses and control for unobserved heterogeneity via panel econometric methods. Compared to this first generation of firm panel data sets, a second generation of data sets which became available recently has an even higher research potential. These new data combine information for firms gathered in different surveys (or from external sources) that could not be analyzed jointly before. Merging firm level data from different surveys to construct data sets that cover information on a wider range of variables than the ones collected in any of these surveys, one at a time, is the basic idea of the project AFiD. AFiD is an acronym for the German Amtliche Firmendaten für Deutschland (official firm data for Germany). Merging of firm data from different sources of official statistics is legal according to §13a BStatG (Bundesstatistikgesetz, or federal statistics law), and it is technically feasible because an identical firm identifier is used in the different surveys. In the AFiD project, which is in detail described in Malchin and Voshage (2009), several different panel data sets are provided in the RDC, including the AFiD-Panels Agriculture, Industrial Units, Industrial Enterprises, Energy Units, Services and Business Register. For some of these panels the information potential can be enlarged even further by adding variables from the so-called AFID-Modules Earnings, Use of Energy, and the environmental moduls.2 The RDC of the statistical offices of the German federal states hosted a workshop in Berlin in May 2010 to offer researchers who are working with different AFiD-Panels an opportunity to present first results generated with this new type of firm level data and to discuss them with participants from universities, research institutes, statistical offices and policy makers. Selected contributions to this workshop are published in this special issue of the Jahrbücher für Nationalökonomie und Statistik.3 We hope that these papers not only provide important new insights in the fields they are dealing with but also motivate interested researchers that did not work with AFiD data before to do so in the future. 1

2

3

For partial surveys, see Wagner ( 2 0 0 7 a , 2 0 0 8 ) . Recent contributions to this journal based on these data include Görzig et al. ( 2 0 0 7 ) , Ronning ( 2 0 0 8 ) and Wagner ( 2 0 0 7 b , 2 0 0 9 ) . Note that tailor made variants that combine data from various surveys according t o a wish-list provided by a researcher can be prepared on request. All papers went through the usual referee process and were revised accordingly.

Guest Editorial · 325

References Görzig, Β., M . Gornig, A. Werwatz ( 2 0 0 7 ) , Produktdiversifizierung: Konvergenz zwischen ostund westdeutschen Unternehmen? J a h r b ü c h e r für N a t i o n a l ö k o n o m i e und Statistik 2 2 7 : 1 6 8 208. Kaiser, U., J . Wagner ( 2 0 0 8 ) , Neue Möglichkeiten zur Nutzung vertraulicher amtlicher Personen- und Firmendaten. Perspektiven der Wirtschaftspolitik 9: 3 2 9 - 3 4 9 . M a l c h i n , Α., R . Voshage ( 2 0 0 9 ) , Official Firm D a t a for Germany. Schmollers J a h r b u c h / Journal of Applied Social Science Studies 1 2 9 : 5 0 1 - 5 1 3 . Ronning, G. ( 2 0 0 8 ) , Measuring Research Intensity from Anonymized Data: Does Multiplicative Noise with Factor Structure Save Results Regarding Quotients? Jahrbücher für Nationalö k o n o m i e und Statistik 2 2 8 : 6 4 4 - 6 5 3 . Wagner, J . ( 2 0 0 7 a ) , Politikrelevante Folgerungen aus Analysen mit Firmendaten der amtlichen Statistik. Schmollers J a h r b u c h / J o u r n a l of Applied Social Science Studies 1 2 6 : 3 5 9 - 3 7 4 . Wagner, J . ( 2 0 0 7 b ) , Productivity and Size of the E x p o r t M a r k e t . Jahrbücher für N a t i o n a l ö k o nomie und Statistik 2 2 7 : 4 0 3 - 4 0 8 . Wagner, J . ( 2 0 0 8 ) , Die Forschungspotenziale der Betriebspaneldaten des Monatsberichts im Verarbeitenden Gewerbe. AStA - Wirtschafts- und Sozialstatistisches Archiv 2 : 2 0 9 - 2 2 1 . Wagner, J . ( 2 0 0 9 ) , Produktdifferenzierung in deutschen Industrieunternehmen 1 9 9 5 - 2 0 0 4 : Ausmaß und Bestimmungsgründe. J a h r b ü c h e r für Nationalökonomie und Statistik 2 2 9 : 615-642. Zühlke, S., H . Christians, K. Cramer ( 2 0 0 7 ) , Das Forschungsdatenzentrum der Statistischen Landesämter - eine Serviceeinrichtung für die Wissenschaft. AStA Wirtschafts- und Sozialstatistisches Archiv 1: 1 6 9 - 1 7 8 .

Anja Ramona Joachim

Malchin Voshage Wagner

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2011) Bd. (Vol.) 2 3 1 / 3

Product Diversification and Profitability in German Manufacturing Firms By Nils Braakmann, Newcastle upon Tyne, and Joachim Wagner, Lueneburg* JEL D21 ; L60 Product diversification; profitability; Germany.

Summary We use unique rich data for German manufacturing enterprises to investigate the product diversification - firm performance relationship. We find that an increase in the degree of product diversification has a negative impact on profitability when observed and unobserved firm characteristics are controlled for. The effects are statistically significant and large from an economic point of view. This helps to understand the fact that nearly 40 percent of all enterprises with at least 20 employees are single-product firms according to a detailed classification of products, and that multi-product enterprises with a large number of goods are a rare species.

1

Motivation

A cartoon published in The New Yorker shows a manager sitting at his desk when his secretary enters the office saying „Your mother called to remind you to diversify". Mothers' advices, as we all know, are too often ignored („Boy, don't drink that much at the party tonight", etc.). Manufacturing enterprises in Germany are a case in point. Nearly 40 percent of all manufacturing enterprises with at least 20 employees in Germany are single-product firms according to a detailed classification of products, and they do not diversify in product-space. Multi-product enterprises producing a large number of goods are a rare species (Wagner 2009). Mothers' advices, however, are usually derived from life experience, and following these advices might be expected to pay. So why should a firm diversify, i.e. why should a firm produce more than one good and spread activities across markets when it goes for a better performance? According to the resource view (Montgomery 1994: 167 f.) firms that have an excess capacity in productive factors - for example, special knowledge the firm has accumulated through time, and that can be used in other markets without reducing the use in the market the firm is already active in - can reap economies of scope by expanding into different product markets. Alternatively, the firm may sell this specific asset to another firm active in this market. However, it is reasonable to expect that market failure does exist when it comes to trade in intangible assets like knowledge, and this is an incentive to internalize the use of the assets. Furthermore, productive factors of this type are often closely linked to persons who can not simultaneously work for several firms producing * All computations were done in the research data centre of the Statistical Office in Berlin. Many thanks to Ramona Voshage for building the data set and her help in many ways. Helpful comments from two referees and workshop participants are gratefully acknowledged.

Product Diversification and Profitability in German Manufacturing Firms · 327

different products. If a firm owns intangible assets of this type that make it successful in one market, and if these assets can be used in other markets, too, one would expect diversification into other product markets to be positive for firm performance. However, there are extra costs to be considered, too, because producing for a new market usually is connected to costs for developing and introducing the new product, including costs for market research and marketing. A second line of reasoning points to the reduction of risk and uncertainty that can be reached by diversification across product markets (Lipczynski/Wilson 2 0 0 1 : 3 2 4 f.). Demand shocks or new competitors may have a negative impact on sales and profits in a product market in an unpredictable manner. A single-product firm, therefore, is highly vulnerable to adverse shocks that hit their market. A multi-product firm can substantially reduce this vulnerability, especially if the risks on the various product markets are randomly distributed or negatively correlated (for a formal model see Hirsch/Lev 1 9 7 1 ) . Risk reduction will lead to more stable profits. M o r e stable profits may be positively related to growth because they can secure the funds for investment at lower costs, and this may have a positive influence on the level of profits. Again, there are extra costs associated with the serving of different product markets that have to be considered, too. Whether product diversification is good or bad for firm performance, and to which extent, therefore, is an empirical question. Results so far are mixed. Hall ( 1 9 9 5 : 2 6 ) summarizes the findings of a number of studies as follows: „The relationship between diversification and organisational performance has been the subject of numerous studies over the years . . . , with results suggesting: negative relationships ..., positive relationships ..., and lack of relationship .... Regardless of how diversification is measured ..., the corporate diversification literature has failed to reach consensus about the relationship between firm diversification and performance." Similarly, Montgomery ( 1 9 9 4 : 1 7 2 ) argues that the literature surveyed by her „clearly shows that diversification is not a guaranteed route to success." In Germany data on the number of different products produced by a firm 1 and on the turnover realized with each product became available for researchers who are not working inside the statistical agencies only recently. As a first step the so-called producer-product-panel was built that merged information from the cost structure survey and from the survey of products produced for a sample of manufacturing enterprises and for the years from 1 9 9 5 to 2 0 0 1 (see Görzig et al. 2 0 0 5 ) . This data set has been used to compute various measures of diversification for manufacturing industries in the years covered and for comparisons over time (see Zloczysti/Faber 2 0 0 7 ; Görzig et al. 2 0 0 7 a , 2 0 0 7 b ) . Furthermore, descriptive studies investigated the relationship between the expansion and the reduction of the number of goods produced and changes in the profitability of enterprises (see Görzig et al. 2 0 0 7 ; Görzig/Pohl 2 0 0 7 ; Gornig/Görzig 2 0 0 7 ) . Görzig et al. ( 2 0 0 7 ) and Görzig and Pohl ( 2 0 0 7 ) find that enterprises that reduce the degree of product diversification show the largest improvement in profitability. N o t e , however, that these studies do not control for unobserved firm heterogeneity. This paper contributes to the literature by using a unique rich newly built data set for German manufacturing enterprises to investigate the product diversification - firm performance relationship. We find that an increase in the degree of product diversification 1

The expression „firm" is used here to describe either an enterprise (a legal unit) or an establishment (a local production unit). In the empirical investigations data at the enterprise level are used; some of these data were collected at the establishment level and aggregated to the enterprise level.

328 · Nils Braakmann and Joachim W a g n e r

has a negative impact on profitability when observed and unobserved firm characteristics are controlled for. These effects are statistically significant and large from an economic point of view. The rest of the paper is organized as follows: Section 2 introduces the data used. Section 3 presents some stylized facts for product diversification in German manufacturing firms. Section 4 reports the results of our econometric investigation. Section 5 concludes. 2

Data

This study uses a data set that extends the producer-product-panel in three ways: All manufacturing enterprises with at least 20 employees are covered; information from the so-called monthly report of manufacturing establishments (aggregated over all months, and all establishments belonging to an enterprise) is added; and the time frame has been extended to cover the years 1995 to 2004. 2 The focus of this study is on the relationships between product diversification and profitability. Given that information on profitability is available from the cost structure surveys only, the sample of firms used here is limited to the enterprises that took part in these surveys. The annual cost structure survey covers all enterprises from manufacturing industries with 500 and more employees. Smaller enterprises, however, are sampled, and as a rule the samples are replaced after four waves, leading to a rotating panel design. Different from this rule in the period covered by the data set used in this study new samples were drawn in 1995, 1997, 1999 and 2003. Because longitudinal data are needed to investigate the consequences of product diversification for firm performance in the econometric investigations this study uses data from a panel of enterprises that participated in the cost structure survey from 1999 to 2002. 3 To give a first impression on the evidence of product diversification in German manufacturing enterprises, some information is given below. We focus on 2000, a year in the middle of the period considered in the econometric investigations. 4 In 2000 61.25 percent of all 30,955 enterprises covered in the survey of products reported that they produced more than one product. A product here is defined by the most detailed 9digit-level of the manual for the survey of products (Güterverzeichnis für Produktionsstatistiken) used by German official statistics. At this rather detailed level, for example, brandy, whisky, rum, and gin are different products, and the same holds for automobiles with a cubic centimetres stroke volume of up to 1,500, between 1,500 and 2,500, and more than 2,500. Nearly 40 percent of all manufacturing enterprises with at least 20 employees are single-product firms according to this detailed classification. 5 Multiproduct enterprises on average produce 4.35 different goods; firms with a large number of goods, however, are rare - only 3.2 percent of all firms produce more than 10 different goods. Over time the pattern of diversification is rather stable. Among the 17,792 enterprises we have information for in the data set for 1995 to 2004 56.4 (30.9) percent were a multi-product (single-product) enterprise in each year. 2

3 4 5

The data are confidential but not exclusive. They can be used by researchers on a contractual basis via remote data access in the research data centres of the statistical offices in Germany; for details, see Ziihlke et al. (2004). See Fritsch et al. (2004) for a detailed description of the cost structure survey. Detailed descriptive results for 1995 to 2 0 0 4 are reported in Wagner (2009). Note that a firm that produces different brands of a product (for example, a brewery that produces several brands of beer) is classified as a single product firm.

Product Diversification and Profitability in German Manufacturing Firms · 329

3

Descriptive evidence on product diversification and profitability in German manufacturing enterprises

Product diversification is measured in two ways, by the share of sales of the most important product in total sales, and by the Berry-index defined as one minus the sum of squared shares of sales of all products in total sales. By definition, for a single-product firm the share of sales of the most import product in total sales is One, and a decreasing value of this measure shows an increase in diversification. The Berry-index is by definition Zero for a single-product firm, and an increase in its value shows an increase in diversification.

Figure 1 Share of sales of most important product in total sales, manufacturing enterprises in Germany, 2000 1

1

Kernel density estimate with epanechnikov kernel

Berry-Index

Figure 2 Berry-Index, manufacturing enterprises in Germany, 2000 1 1

Kernel density estimate with epanechnikov kernel

330 · Nils Braakmann and Joachim Wagner

To illustrate the distribution of the measures of product diversification in the sample of enterprises used in our econometric investigation figure 1 and figure 2 show kernel density estimates of the share of sales of the most import product in total sales and of the Berry-Index in 2000. Due to the high share of single-product enterprises both distributions are highly skew, and it can be seen that only a small portion of all enterprises is very highly diversified according to both measures. 6 Profitability is measured as a rate of return, defined as gross firm surplus (computed as gross value added at factor costs minus gross wages and salaries minus costs for social insurance paid by the firm) divided by total sales (net of VAT) minus net change of inventories, using information from the cost structure surveys. 7 Figure 3 shows a kernel density estimate of the rate of return (in percentages) for 2000. 8 The distribution is rather symmetric around the positive mean value, and extreme positive or negative values are rare.

Figure 3 Profitability in manufacturing enterprises in Germany, 2000 1 1

6

7

8

Kernel density estimate with epanechnikov kernel

Both measures of diversification are highly positively correlated over time (see Wagner 2009, Table 11), and, therefore, the kernel density estimates look identical for all the years covered. The correlation between the share of sales of the most important product in total sales and the Berry-Index is extremely high in each year; the value for 2 0 0 0 is -0.986 (see Wagner 2009, Table 10). N o t e that the fact that the graph in Figure 1 shows values below one, and that the graph in Figure 2 shows values below zero and above one, for the measure of product diversification is caused by the smoothing technique used in the estimation of the kernel density estimates. The computation of gross firm surplus follows the suggestion of the European Commission (1998: 56). N o t e that the data set does not have any information on the capital stock, or the sum of assets or equity, of the firm, so that it is not possible to construct profit indicators based thereon like return on assets or return on equity. The kernel density estimates look identical for all the years covered in this study.

Product Diversification and Profitability in German Manufacturing Firms · 331

4

Econometric investigation

Our econometric investigation of the relationship between profitability and product diversification uses pooled data for the years 1999 to 2002 and fixed-effects estimators to control for unobserved time-invariant enterprise heterogeneity. 9 Table 1 reports mean values and standard deviations of the variables used in our empirical study. It can be seen that both the profitability and the degree of product diversification vary not only between enterprises (as shown in the figures above) but also over time within the enterprises. Note that the variation in profitability across enterprises is about twice as large as that observed within an enterprise over time, while the variation of both measures of the degree of product diversification across enterprises is more than four times larger than that observed within the enterprises over the four years. Results from fixed effects regressions for profitability are reported in Table 2. Two variants of empirical models are estimated, one that includes only the measure of the degree of product diversification (plus dummy variables for the years, and a constant), and one that adds a number of control variables. 1 0 In all models the fixed enterprise effects control for unobserved firm characteristics that do not vary over time. These fixed effects control for the industry affiliation of the enterprise, too, because only few enterprises tend to change industries between the years; this is important because profitability might be expected to vary between industries due to variation in the intensity of competition or regulation. As can be seen from Table 2 the inclusion of the control variables does not change the results for the estimated link between profitability and product diversification substantially. The regression coefficients for both measures of product diversification are statistically highly significant, and they indicate a negative relationship - the higher the degree of product diversification (i. e., the lower the share of sales of the most important product in total sales, and the higher the value of the Berry-Index), the lower is the profitability, controlling for observed and unobserved enterprise heterogeneity. These findings are in line with the results from descriptive studies using the producerproduct panel (mentioned in section 1) by Görzig et al. (2007) and Görzig and Pohl (2007) w h o report that enterprises that reduce the degree of product diversification show the largest improvement in profitability. Note, however, that these studies do not control for unobserved firm heterogeneity. In a robustness check we tested for a non-linear relationship between the degree of product diversification and profitability by adding a squared term of the share of the most important product in total sales and of the Berry-Index to the empirical model used. All estimated coefficients for the measures 9

10

We experimented w i t h b o t h a propensity score m a t c h i n g a p p r o a c h (that considers p r o d u c t diversification as a binary t r e a t m e n t , w i t h diversified firms as the t r e a t m e n t g r o u p and single-product firms as the control group) and with a generalized propensity score matching a p p r o a c h (that considers p r o d u c t diversification as a c o n t i n u o u s t r e a t m e n t ) . In both cases the a p p r o a c h t u r n e d out to be n o t c o m p u t a t i o n a l l y feasible. M a t c h i n g was never successful, a n d the balancing property w a s not fulfilled. T h e selection of c o n t r o l variables is motivated by the empirical model used to explain the degree of p r o d u c t diversification applied in Wagner (2009) and by the empirical investigation of the link between exports and profitability in G e r m a n m a n u f a c t u r i n g enterprises in Fryges and Wagner (2010). Given t h a t the focus of this p a p e r lies on the relationship between p r o d u c t diversification and profitability and t h a t this relationship turns out to be not affected by the inclusion of control variables (see the results reported below) we do not discuss these control variables and related estimation results in detail.

332 • Nils Braakmann and Joachim Wagner

Ε O Ν Ν 00 00 * E Γη ΓΟ e 1 (Ν ΓΜ τS

IV IV IV m 00 00 00 m m (Ν ΓΜ ΓΜ

IV οο

IV ΟΟ

IV 00

ΓΜ

ΓΜ τ-

ΓΜ

σι σ\ σ\ VD ιο κ rv Gl- Ί-

σι kO m σ> σι σ> VO VO m in ΓΜ r- ο t—

IV 00 IV (Ν m ΓΜ ο ΓΠ rri Ο ΓΜ m (Ν

IV >£> ΐ— m co ÍN (Ν Ο Τ-

m IV 00 00 t ΓΜ 00 σ\ ö 00 cri •"ί ΓΜ m ΓΜ

Ο IV IV m IV (Ν Ö τ—

in σ\ IV τΓΜ ΙΌ σ> Ο VO 00 Ο IV

m

sì

m

m

o

ò) s «I I Έ «¡ (5 m -o Û

m

ΓΜ Ο IV ΙΌ t τ-

ΓΜ σ\ ο ΓΜ ΓΜ

IV m σι

VO ΓΜ ο ο 00 m

IV IV m ο ΓΜ

IV tv IV οο

σι in νο σι οο

οο ΓΜ 00 σί ΓΜ

IV τ— ο ο

m ο ο

ΙΌ Ο Ο

IV τΟ Ò

-Q ni O ίΟ.

υ Q _a¡ •α I« Ι-

α) (Λ U C >, »1 .O c χ— CL E Ζ Q «a 2 4) c OJ ce c ιβ 3 >> LU Ι . E c co Ω_Ο t/1 1> fc 4> o» "ñ¡ S. o^ .2 •- iS •S 1 l-E t m 3 ûj o, -g E ir χ- —c φ • t i"5S >

43 o

Product Diversification and Profitability in German Manufacturing Firms · 333

Table 2 Results from fixed effects regressions for profitability in German manufacturing enterprises, 1999 - 2002 1 Model Exogenous variable Share of sales of most important ß product in total sales Ρ ß Berry-Index Ρ Number of employees ß Ρ Number of employees (squared) β Ρ Share of sales in Germany in β total sales (percentage) Ρ Labour productivity (sales per β employee; €) Ρ Human capital intensity (wages β and salaries per employee; €) Ρ Research and development β intensity (share of employees Ρ in R&D) Year 2000 (Dummy-variable) β Ρ Year 2001 (Dummy-variable) β Ρ Year 2002 (Dummy-variable) β Ρ Constant β Ρ Number of observations Number of firms

1 2.402 0.002

2 2.355 0.003 0.00045 .233 6.20e-11 0.977 -0.041 0.000 1.16e-6 0.300 0.000069 0.001 -0.416 0.865

-0.186 0.013 -1.034 0.000 -1.822 0.000 11.241 0.000 47,699 12,387

-0.285 0.000 -1.201 0.000 -2.057 0.000 12.304 0.000 47,693 12,387

3

4

-2.370 0.001

-2.342 0.001 0.00047 0.222 -3.01 e-11 0.989 -0.041 0.000 1.15e-6 0.300 0.000069 0.001 -0.431 0.860

-0.188 0.012 -1.036 0.000 -1.822 0.000 13.798 0.000 47,699 12,387

-0.287 0.000 -1.204 0.000 -2.057 0.000 14.808 0.000 47,693 12,387

1 ß is the estimated regression coefficient, ρ is the prob-value; robust standard errors of the regression coefficients were adjusted for the firms as clusters.

Table 3 The estimated relation between profitability and product diversification Share of sales of the most important product in total sales .80 .60 .40 .20 Berry Index .20 .40 .60 .80 1

Estimated change in the rate of profitability (percentage points) compared to a singleproduct enterprise1 -0.471 -0.942 -1.413 -1.884 -0.468 -0.936 -1.404 -1.872

The estimates are based on the results reported in column 2 and column 4 of Table 2

3 3 4 · Nils Braakmann and Joachim W a g n e r

of product diversification in these augmented models turned out to be statistically insignificant at any conventional level. To illustrate the economic importance of product diversification for profitability, single product enterprises (with a share of sales of the most important product in total sales on One, and a value for the Berry-Index of Zero by definition) are compared to firms with different degrees of product diversification using the estimated regression coefficients from the empirical models with the control variables. Results documented in Table 3 indicate that a growing degree of product diversification is accompanied by a substantial reduction in profitability. For example, for an average firm in our sample a decrease of the share of sales of the most important product from 100 to 60 percent means a reduction in the rate of profitability by nearly one percentage point, and the same holds when the Berry-Index increases from Zero to 0.40. A question open for discussion is whether the negative ceteris paribus association between profitability and product diversification can be interpreted to indicate a causal negative impact of the degree of product diversification on profitability, or whether there is (instead of this, or additionally to this) a causal effect running from profitability to product diversification. While reverse causality can not be excluded per se in the fixed effects regression framework used in our study, 11 we argue that there are no economic arguments that can explain why the profitability of a firm should have any impact on the number of products produced, or the share of sales of a product in total sales, of an enterprise in the same year. Therefore, we argue that the negative association between profitability and degree of product diversification that results from the fixed effects panel regressions can be interpreted to indicate a negative impact of a higher degree of product diversification on profitability. 5

Concluding remarks

We use a unique rich newly built data set for German manufacturing enterprises to investigate the product diversification - firm performance relationship. We find that an increase in the degree of product diversification has a negative impact on profitability when observed and unobserved firm characteristics are controlled for. The effects are statistically significant and large from an economic point of view. These findings indicate that the extra costs associated with serving different product markets tend to be greater than the extra profits reaped from diversification across these markets. Concentration on a core market pays. This might help to understand the fact that nearly 40 percent of all manufacturing enterprises with at least 20 employees in Germany are single-product firms according to a detailed classification of products, and that multi-product enterprises with a large number of goods are a rare species.

11

Note that using lagged values of the degree of product diversification in the empirical models offers no solution here, since the measures of product diversification are nearly perfectly positively correlated between adjacent years.

Product Diversification and Profitability in German Manufacturing Firms · 335

References European Commission (1998), Commission regulation (EC) no. 2700/98 of 17 December 1998 concerning the definitions of characteristics for structural business statistics. Official Journal of the European Communities L344, 18/12/1998, 4980. Fritsch, M., B. Görzig, O. Hennchen, Α. Stephan (2004), Cost Structure Surveys for Germany. Schmollers Jahrbuch/Journal of Applied Social Science Studies 124: 557-566. Fryges, H., J. Wagner (2009), Exports and Profitability: First Evidence for German Manufacturing Firms. The World Economy 33: 399-423. Görzig, Β., Η. Bömermann, R. Pohl (2005), Produktdiversifizierung und Unternehmenserfolg: Nutzung der Forschungsdatenzentren der Statistischen Ämter. Allgemeines Statistisches Archiv 89: 339-354. Görzig, B., M. Gornig, R. Pohl (2007), Spezialisierung und Unternehmenserfolg im verarbeitenden Gewerbe Deutschlands. Vierteljahrshefte zur Wirtschaftsforschung 76: 43-58. Görzig, Β., M. Gornig, A. Werwatz (2007a), Produktvielfalt und Produktivität der IKT-Produzenten: Eine Analyse unter Nutzung verbundener amtlicher Unternehmensdaten. AStA Wirtschafts- und Sozialstatistisches Archiv 1: 145-161. Görzig, Β., M. Gornig, A. Werwatz (2007b), Produktdiversifizierung: Konvergenz zwischen ostund westdeutschen Unternehmen. Eine Dekomposition mit Mikrodaten der amtlichen Statistik. Jahrbücher für Nationalökonomie und Statistik 227: 168-186. Görzig, B., R. Pohl (2007), Diversifizierungsstrategien deutscher Unternehmen. Auswertung eines Producer-Product-Panels der amtlichen Statistik. AStA - Wirtschafts- und Sozialstatistisches Archiv 1: 179-191. Gornig, Μ., B. Görzig (2007), Verstärkte Spezialisierung deutscher Unternehmen. DIW-Wochenbericht 74: 333-335. Hall, E.H. (1995), Corporate Diversification and Performance: An Investigation of Causality. Australian Journal of Management 20: 25-42. Hirsch, S., Β. Lev (1971 ), Sales Stabilization through Export Diversification. Review of Economics and Statistics 53: 270-277. Lipczynski, J., J. Wilson (2001), Industrial Organisation. An Analysis of Competitive Markets. Harlow. Montgomery, C. A. (1994), Corporate Diversification. Journal of Economic Perspectives 8:163178. Wagner, J. (2009), Produktdifferenzierung in deutschen Industrieunternehmen 1995 - 2004: Ausmaß und Bestimmungsgründe. Jahrbücher für Nationalökonomie und Statistik 229: 615-642. Zloczysti, P., C. Faber (2007), Diversifikationsmaße im Praxistest - Ergebnisse auf der Grundlage von amtlichen Mikrodaten für Deutschland. Vierteljahrshefte zur Wirtschaftsforschung 76: 29-42. Zühlke, S., M. Zwick, S. Scharnhorst, T. Wende (2004), The research data centres of the Federal Statististical Office and the statistical offices of the Länder. Schmollers Jahrbuch/Journal of Applied Social Science Studies 124: 567-578. Dr. Nils Braakmann, Newcastle University Business School, Newcastle University, Ridley Building, Newcastle upon Tyne, NEI 7RU, United Kingdom. [email protected] Prof. Dr. Joachim Wagner, Leuphana University Lueneburg, Institute of Economics, P.O. Box 2440, 21314 Lueneburg, Germany. [email protected]

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2011) Bd. (Vol.) 231/3

Is the Boone-Indicator Applicable? - Evidence from a Combined Data Set of German Manufacturing Enterprises By Alexander Schiersch and Jens Schmidt-Ehmcke, Berlin* JEL L12; L41; D43 Competition; Boone-lndicator; cartels; census data.

Summary Boone (2008a) proposes a new competition measure based on Relative Profit Differences (RPD) that, from its theoretical properties, proves to be more robust than the Lerner-Index. However, the proof of the empirical practicability and robustness of the Boone-Indicator is missing. To fill this gap, we use a rich, newly built, data set for German manufacturing enterprises and test its empirical validity using cartel cases. Since all of the identified cartels significantly restricted market competition, we expect fiercer competition after the uncovering. We asses the validity of the indicators by comparing the indicated competition levels before and after the cartels were uncovered and stopped. The Boone-Indicator is calculated as RPDs and as a beta coefficient of a log-log regression. The Lerner-Index is used as a benchmark. Our analysis finds that the Boone-Indicator, based on a simple regression approach, fails to correctly indicate competition. Since the Boone-Indicator, based on pure RPDs, proves to be inapplicable, we propose an augmented indicator based on size-adjusted RPDs, which performs better. However, our findings suggest that, given the information typically available in census data, the Lerner-Index is still the only measure that correctly indicates competitive changes.

1

Introduction

Studying competition is hampered by the scarcity of appropriate data and, in particular, by the lack of good indicators of the competitive environment with wide coverage. Researchers and policy-makers, for example in antitrust authorities, rely on traditional measures like the Lerner-Index (also called Price Cost Margin, PCM) to assess the competition levels in industries. However, theoretical research raises doubts on the robustness of PCM. Amir (2003), Bulow and Klemperer (1999), Rosenthal (1980) and Stiglitz (1989) show, that there are theoretically possible scenarios in which P C M increases with more intense competition. Despite its theoretical weaknesses, P C M is still a popular measure in empirical research (see, for example Pruteanu-Podpiera et al. 2007; Maudos/Fernandez de Guevara 2006; Aghion et al. 2005; Nevo 2001; Blundell et al. 1999; Klette 1999; Nickell 1999; Geroski 1995; Porter 1990). Boone (2004, 2008a) extends the existing set of competition measures by suggesting a new indicator that relies on profit differences. His approach is based on the notion that * The authors would like to thank two anonymous referees for their helpful comments. Special thanks to the staff of the Research Data Centre Berlin for their patient support in conducting the analysis.

Is the Boone-lndicator Applicable? · 337

competition rewards efficiency. In industries with increasing competition inefficiently operating firms are punished more harshly than more efficient ones. Hereby, efficiency is defined as the possibility to produce the same output with lower costs or, rather, lower marginal costs. Thus, comparing the relative profits between some arbitrarily efficient firm and a firm with greater efficiency contains information about the level of competition within that industry. The more competitive the market is the stronger is the proposed relationship between efficiency differences and performance differences. 1 Two properties make the Boone-lndicator (BI) appealing: First, it has a robust theoretical foundation as a measure of competition (Boone 2008a). It depicts the level of competition correctly both when competition becomes more intense through reduced entry barriers or more aggressive interaction between firms. Second, it has the same data requirements as PCM. Hence, if firm level data are given to calculate P C M , the Boone-lndicator could be calculated as well. The goal of this paper is to evaluate the empirical robustness of the Boone-lndicator as a measure of competition. To accomplish this, we use large cartel cases as natural experiments in combination with a newly constructed data set for German manufacturing enterprises. The intuitive idea behind cartel cases as a natural experiment is that we expect fiercer competition after its uncovering. This should be observed in our data and affect the competition measures. The new data set allows for a much more rigorous analysis than usual firm level data. We proceed in four steps: First, we identify large cartel cases in Germany that sold goods as homogeneous as possible and were large enough to affect competition. Second, we estimate the BI as proposed by Boone (2008a) and show how its performance deviates from theory. We also present the regression approach of the BI as proposed by Creusen et al. (2006) and Boone et al. (2007). Third, we propose an augmented BI correcting for firm size. Finally, we use the cartel cases to evaluate the performance of the regression approach of the BI and the augmented BI in detecting changes in the competitive environment. We compare the results with the traditional P C M as a benchmark. The remainder of this paper is as follows: The next section present related studies. Section 3 presents the Boone-lndicator and the traditional P C M measure and reveals the robustness of the BI compared to P C M . Section 4 gives a detailed description of the data set and present first descriptive statistics. The relevant cartel cases are presented in section 5. Section 6 discusses the Boone-lndicator, propose a modification to control for firm size, conduct the analysis and present the main results. The last section collects the main findings and concludes the paper.

2

Related studies

Despite its theoretical robustness and straightforward data requirements few studies apply the Boone-lndicator to real world data. The only paper published in a refereed journal, to our knowledge, is Bikker and Leuvensteijn (2008). Using data for the Dutch life insurance market, they calculate the Boone-lndicator using three different approximations of the marginal costs: average variable costs, defined as management costs as a 1

The empirical literature on the effects of efficiency on firm performance and the positive impact of competition on efficiency somewhat supports the assumed cohesion between efficiency, firm performance and competition needed for the Boone-lndicator. One of the earliest studies examining the influence of competition on productivity is Nickell (1996).

338 · Alexander Schiersch and Jens Schmidt-Ehmcke

share of the total premium; marginal costs derived from a translog costs function; and scale adjusted marginal cost. Using a least-square dummy variable approach, they regress these variables first one by one on logarithmized relative profits, then, in a second step, on the market share of insurance companies as an outcome variable. Their results point to a weak competition in the Dutch life insurance industry when compared to other industries. However, Bikker and Leuvensteijn do not discuss the robustness of their results. Additionally, the Boone-Indicator is used in a number of reports and discussion papers. The Finnish Ministry of Trade and Industry studied trend changes in the intensity of competition across Finnish business sectors (Maliranta et al. 2007). The report focuses on the service sector and reports the results of nine different measures of competition including traditional measures like Herfindahl, PCM, and the four-firm concentration index as well as six different parameterizations of the Boone-Indicator. Their results suggest an increase in competitive pressure in Finland over the analyzed time interval. However, the outcomes vary a lot with respect to the different parameterizations of the BI. Moreover, PCM and their preferred BI frequently contradict each other in the direction of changes. Griffith et al. (2005) investigate the empirical usefulness of a slightly modified BI based on relative profits. Using data from the annual report and accounts filed by firms listed on the London Stock Exchange over the period 1 9 8 6 - 1 9 9 9 , they compare the relative profit measure with the PCM and the Herfindahl index. Their main results show a positive correlation between the new measure and PCM but no correlation with Herfindahl, which raises questions about the usefulness of the Herfindahl index as a measure of competition. Furthermore, they provide evidence that the relative profit measure is less affected by cyclical changes than the PCM. They also find industries in which PCM and BI contradict each other. However, they cannot derive recommendations on which might be the „correct" measure of competition because „without a prior information about the 'true' degree of competition ... it is difficult to say whether the relative profit measure is empirically better than the price-cost-margin." (Griffith et al. 2 0 0 5 : 14). Creusen et al. (2006) use a similar method to examine the competition in Dutch market sectors during the years 1993-2001 based on firm level data finding a slight decline in the intensity of competition during that time. They also find that PCM and their profit measure frequently contradict each other on the direction of changes. It remains an open question though, which of the indicator points in the right direction. 3

Measuring competition

A common competition measure is the Lerner-Index or Price Cost Margin (PCM). It is based in neoclassic theory where under perfect competition prices (p) equal marginal costs (c). Hence, the PCM, calculated as (p, — c¡)/p¡, takes values greater than zero if competition is not perfect and firms are able to enforce prices above marginal costs. As competition becomes fiercer PCM approaches zero. To evaluate competition on markets or in industries, the industry PCM is calculated as a simple or weighted mean of individual PCMs. The latter is usually derived by calculating it with firm market shares. This ensures that the market power of big firms is adequately captured. The common interpretation is as with firm individual PCMs. It decreases with fiercer competition and increases with weaker competition.

Is the Boone-lndicator Applicable? · 339

However, it is not a robust competition measure. Amir (2003) shows that, under certain conditions, an increase in competition through an increase in the number of firms in a market can result in an increasing average PCM. Given certain circumstances Stiglitz (1989) shows that profits per unit sales can rise in a recession. Thus, even though competition among firms increases during recessions, industry P C M also increases. Another potential source of error can be the reallocation effect. As a result of fiercer competition, the market share of the more efficient firms increases while that for less efficient firms decrease. Thus, the weighted average P C M can increase if the increase in the market share of the more efficient firms overcompensate the decrease of the respective individual PCMs. Therefore, the Lerner-Index is, at least theoretically, potentially misleading. Against this background combined with the difficulty in interpreting popular concentration indices like Herfindahl, Boone (2008a) proposes a new competition measure, called Relative Profit Differences (RPD). Its main idea is that competition rewards efficiency. To get the measure working, Boone postulates some basic assumptions we outline here: First, firms under consideration act in a market with relatively homogeneous goods. Secondly, he assumes symmetry. Hence, firms act on a level playing field ensuring that changes in competition affects firms directly and not indirectly through changes in that playing field. 2 It also implies that „...firm i's profits are the same as firm j's profits would be if firm j was in firm i's situation." (Athey/Schmutzler 2 0 0 1 : 5). Thus, within the theoretical framework of the indicator, this implies equal profit level for two equally efficient companies. Thirdly, firms can be ranked with respect to their efficiency («,). Thereby the efficiency index (N) needs to be one dimensional to ensure transitivity. Given that the production costs are captured by C(q, η) with q as output quantity, the relationship between efficiency and cost is assumed to be: 3 dC(q, n)

ρ q i 3J
η > n'. As Boone (2008a) proves, his measure of competition is robust regarding distortions out of the reallocation effect, which is not discussed here. Instead, the following example illustrates how the reallocation effect works and how it affects both RPD and PCM. Table 1 The reallocation effect and how it affects PCM and RPD 6 d=0.1 d=2

PCM1

PCM2

PCM3

0.950 0.939

0.587 0.385

0.465 0.139

Weighted Industry PCM 0.680 0.717

RPD2 0.262 0.149

We have a simple linear demand function, three firms with constant marginal costs and no entry costs. As shown in Table 1 fiercer competition, simulated by an increase in substitutability of products, results in a rise of the weighted average PCM while the respective RPD is decreasing. Hence, PCM signals a fall of competition while RPD correctly signals fiercer competition. However, if there are more than three firms, comparing RPDs over time for each company is impractical. One convenient way, proposed by Boone (2008a), is to plot the RPDs against normalized efficiencies. This gives a function that is always bounded at one on both axes. Figure 1 presents an example. As in the previous example, we model changes in competition via substitutability of products. The increase in competition leads to lower firm specific RPDs. To measure the change in competition one now calculates and compares the area under both curves. Since we have normalized values the area is bounded between zero and one, with zero implying perfect competition and one the complete absence of competition. The area in our example shrinks and thus correctly indicates fiercer competition.

5 6

Hereby Ν is the efficiency index of n¡ and I is the set of firms in the market (Boone 2008a: 1250 ff.). The demand function is p(x¡,X-\) = a — bx, — dJ2xi a n d t a k e n from Boone (2008b: 590). We

ίφί

apply a=20, b=2 and marginal costs are ci=0.5, C2=5 and C3=7. The substitutability is captured by d, where the quotient d/b = 1 for perfect substitutes. We do not report the RPD for the most and the least efficient firm since they have to be one and zero at both times by definition.

Is the Boone-lndicator Applicable? · 341

o

o α. EC

o ö 0.0

0.2

0.4

0.6

0.8

1.0

normalized efficiency Figure 1 Fiercer competition and R P D 7

4

Data

The data for our analysis is taken from a combined data set, composed of two sources: the German Cost Structure Census (Kostenstrukturerhebung) and the German Production Census (Produktionserhebung). Each data set was gathered and complied by the Federal Statistical Office (Statistisches Bundesamt) of Germany over the period 1995-2006. 8 Plant level data is merged to firm level data using a common identifier. One strength of the data set is its sample coverage and reliability of information. It covers almost all large German manufacturing firms with 500 or more employees over the entire time span. Firms with fewer than 500 employees are included as a random subsample designed to be representative for the small firm segment as a whole in every industry. 9 Only firms with 20 or more employees are covered. 10 The Cost Structure Census contains information on several input categories, namely payroll, employer contributions to the social security system, fringe benefits, expenditures for material inputs, self-provided equipment, goods for resale, energy costs, external 7

8

9 10

We apply the same demand function as in the example of Table 1. We have 20 firms in the beginning with constant marginal costs of c, = if 10. There are no entry costs, a — 20, b — 2 and d increases from 0.1 to 2. The solid line captures the RPDs in situation one, hence with low intense competition due to a d of 0.1, while the dotted line is with d = 2. To overcome the problem that appears if the least efficient firm is assessed, which means dividing by zero, we calculate inverse RPDs, hence: (π(η,θ) — π(η', θ))/(π(η", θ) - π(η',θ)). The normalized efficiency is calculated as: ( « - « ' ) / («" - «') with η" > n > ή . Although the data are confidential, they are not exclusive. Interested scientists can contact the German Research Data Centres for more information about data access (http://www.forschungsdatenzentrum.de/en/index.asp). Samples are drawn in 1995, 1997, 1999 and 2003. In some specific industries, even firms with less than 20 employees are included as a random draw.

342 • Alexander Schiersch and Jens Schmidt-Ehmcke

wage-work, external maintenance and repair, tax depreciation of fixed assets, subsidies, rents and leases, insurance costs, sales tax, other taxes and public fees, interest on external capital as well as „other" costs such as license fees, bank charges and postage, or expenses. 1 1 The sector assignment of a company is given by the activity that generates the highest turnover. Of course that means that a company could belong to sector A if it generates 30 percent of its turnover in market A while the remaining 70 percent are generated in different markets with less than 30 percent in each market (Statistisches Bundesamt 2008). This could bias the results since the effect of 70 percent of the costs and profits of that company are assigned to a market they were not generated in. However, the firm level cost data of the Cost Structure Census are a key part in the analysis. We solve that problem by means of the second data set in our analysis, the German Production Census. The census gives us detailed information about turnover of products, approximated by the nine-digit product classification system (Güterverzeichnis für Produktionsstatistiken) of the Federal Statistical Office. It allows us to create subsamples of companies for the relevant sectors through the ratio of turnover in a specific product and total turnover. In the latter analysis, we only look at companies that have at least three quarters of its overall turnover in the relevant market. Hence, the effect of side activities of a company on total cost and turnover is minimized for the first time. By defining the relevant sectors we are also rigorous. All previously mentioned studies analyzed competition using three digit sector classifications. With our rich data set we are able to use a four digit sector and goods classifications, thus defining the respective markets in even more detail than any previous analysis. Finally, to overcome the outlier problem, we trim the data using the upper and lower one percentile of the respective output variables per year, e.g. the profit-sales ratio. According to Boone (2008a), we calculate profit by subtracting variable costs from revenue. Hereby, we define revenue as revenue out of self produced goods. Hence, it does not include revenues out of other activities like renting or trading operations. The variable costs contain 'consumption of raw materials', 'energy', 'gross wages', 'legal and additional social insurance contributions', 'costs for contract workers' and 'costs of repairs'. Following Boone (2008a) we calculate two different measures for efficiency: i) average variable costs, which we define as total variable costs per sales; and ii) labor productivity, defined as gross value added per employee. Additionally we also use sales per employee as a third measure for efficiency. Descriptive statistics for the variables are presented in Table A l .

5

Cartel cases

In order to evaluate the robustness of the Boone-Indicator we use a natural experiment of three major cartel cases in different sectors. A cartel is defined as an explicit contractual agreement between legally independent companies in order to restrict competition and increase profits. Such contracts define the prices, quantities, markets, etc., for each participating firm. Further the contracts also implement a system of sanctions to ensure that deviant behavior by cartel members is properly punished. Sometimes establishing a cartel 11

For m o r e information a b o u t the Cost Structure Census surveys in Germany, we refer the reader t o Fritsch et al. (2004).

Is the Boone-lndicator Applicable? · 343

includes the formation of an organization that coordinates and monitors participating firms. In addition to explicit cartels, there is also collusive behavior. T h i s is characterized by the absence o f contractual agreements or any form o f record. Instead it often relies on informal, mostly oral, agreements. Although it has the same objective as a cartel, it usually c a n n o t restrict competition as effectively since firms have incentives to deviate from the collusion strategy and the sanction mechanism is missing. 1 2 However, since this way of restraining competition is hard to detect we only focus on cartels. For the purpose o f our analysis cartels have to meet three criteria. Firstly, the cartel must be nationwide. This ensures that it was able to restrict competition all over Germany. Second, it must have been a cartel case of significant size. Hence, the cartel actually must have gained a significant control over the national market. B o t h criteria ensure that the effect can be found in the data. We take the size o f the cartel fine as a proxy for the level of the distortion o f competition. Finally, the assumption o f homogeneous goods was postulated in section 3. T o meet this criterion as efficiently as possible the product o f a cartel should be as homogenous as possible. However, cartels are often established and stable in markets with homogenous good since product diversification as deviant strategy is no real option. Hence, by focusing on cartels we also expect to look at markets with rather homogeneous goods. W h e n such cartels are uncovered and terminated, we expect fiercer competition in subsequent periods. This assumption does not imply that competition changes to perfect competition, nor does it neglect the possibility of future informal oral agreements by the involved companies. However, as previously noted, collusive behavior is less effective than explicit cartels. Moreover, it would not be rational to take the risk of an explicit cartel in the first place if collusion could restrict competition at the same extent from the very beginning. T h e r e f o r e we impose the weak assumption that competition is significantly higher after a cartel was uncovered c o m p a r e d to the cartel period.

Power cable cartel T h e first cartel appearing to be suitable for our analysis is the G e r m a n power cable producers' cartel. It was constructed as a price- and quota-cartel, where producers agreed not only on global market shares but also on shares for every big customer within a precisely defined period and on the respective price range. In order to govern the cartel the „Elektro-Treuhand G m b H " ( E T G ) was founded as a joint venture of all involved producers. T h e mechanism worked in the following way: Every customer query was reported to the E T G . Since E T G also did the cartel accounting, it knew which firms already were at quota during any given time period, and which were not. It passed price- and discount-information to the companies involved, thus ensuring that in the following negotiations those companies scheduled to get the j o b succeeded at the defined price. T h e cartel controlled the entire power cable market for several decades. (Fleischhauer 1 9 9 7 ; Deutscher Bundestag 1 9 9 9 ) . T h e cartel ended in September 1 9 9 6 in a nationwide search and seizure by the Federal Cartel Office. By the end of 1 9 9 7 the cartel office had charged 1 6 companies, t w o cable industry organizations and 2 8 individuals with a fine of 2 8 0 million Deutsch M a r k ( 1 4 3 million Euros) in total: at that time the largest fine in G e r m a n history. All companies, 12

However, there are of course gaming strategies that lead to similar results, given certain well specified condition. See for instance Bester ( 2 0 0 4 : 1 3 8 f f . ) .

344 · Alexander Schiersch and Jens Schmidt-Ehmcke

except one, accepted the fine, thus acknowledging participation in an illegal cartel in order to avoid competition. The organizational structure of the cartel was terminated, including the closure of the ETG and the two cable industry organizations (Deutscher Bundestag 1999). Cement

cartel

The second cartel in our analysis is the German cement cartel. It was created in the aftermath of the German unification in the early 1990s as a price-, quota- and regional cartel covering the entire German market for cement, including importation (Pressemitteilung 19/09). All major players and also medium sized producers took part in the cartel. Due to the physical properties of cement, which leads to excessively high transport costs over 300 kilometers, the market for cement is regional. However, all players agreed on a nationwide organization of regional cartels with explicit organizations for each regional market. The German market was divided into an east, west, north and south submarket. At this level the organizational structure varied. However, in each submarket contractual agreements were made (Oberlandesgericht Düsseldorf 2009). The nationwide monitoring was done by the umbrella association of the German cement producers („Bundesverband der Deutschen Zementindustrie e. V."). In the event that cement was delivered outside a firm's home market, compensatory payments were arranged during ad hoc meetings held in Munich, the so-called „Money-Karussell" (money-carousel). The cartel ended in July 2002 in a nationwide search and seizure of 30 companies. By the end of 2003 the Federal Cartel Offices fined twelve companies and several persons 702 million Euros in total (Deutscher Bundestag 2005). However the companies under suspicion, save for the company acting as principal witness, protested the amount of the fine. The legal disputes lasted until June 2009 when the court finally confirmed all allegations but reduced the fine to 330 million Euros. However, with respect to market effects we can state two things. First, as stated by the court, witnesses, and various experts in the legal case, the consequence of the uncovering of the cartel was a price war that lasted at least until 2005 (Oberlandesgericht Düsseldorf 2009). Second, to gain more information for the court, a second national seizure was carried out in 2004. There was no evidence whatsoever that the cartel still operated. Hence, the market condition changed toward more competition. Blanckenburg and Geist (2009), confirm this, finding fiercer competition after 2002. Ready-mix

concrete

cartel

The last cartel case used in this study is that of the ready-mixed concrete industry. This was actually not one cartel but many regional cartels. This is because the physical properties of ready-mixed concrete limit transport time to roughly 60 minutes after a ready-mix truck is filled. 13 However, the entire German market was governed by regional cartels. The cartels were organized as quota-cartels that specified the share for each participating firms within the regional market. As typical for illegal cartels, regular meetings were established in order to monitor and govern the activities of all involved parties, as for instance proved in the case of the Berlin ready-mixed concrete cartel (Bundesgerichtshof 2005). As established by the courts, cartels in the West formed around 1990, with the ones in East Germany following around 1995 (Deutscher Bundestag 2001, 2003). 13

This information comes from the umbrella association of German ready-mix concrete industry. For further information on the specifics of ready-mixed concrete market see also Syverson (2008).

Is the Boone-lndicator Applicable? · 345 The first cartel was uncovered in May 1999 in Greater Berlin. Additional cartel inquests were opened against companies in the federal states of Sachsen-Anhalt and Niedersachsen. Consequently the Federal Cartel Office initially charged 69 companies in 29 consortiums with a cartel fine of 370 million Deutsch Mark (189 million Euros). With evidence found in these cases and additional information, the cartel office carried out a nationwide search and seizure of 48 companies in March 2000. Moreover, as a result of the cartel inquiry on the cement market, some of the cement companies cooperated with cartel offices and passed further information about the ready-mixed concrete regional cartels to the authorities. This allowed the authorities to open new cases against 70 ready-mixed concrete companies all over Germany. The legal dispute lasted until 2005 when the Federal Supreme Court followed the Federal Cartel Office in all main cases and stated that the participating companies had established and operated illegal cartels, with the last cartel uncovered in 2001. (Deutscher Bundestag 2007; Bundesgerichtshof 2005). Thus, roughly 140 companies were convicted with a fine of approximately 167 million Euros. In the meantime the sector saw major changes. On the one hand, due to high overcapacity many companies closed (Deutscher Bundestag 2007). Two major players, Larfarge and Hanson, exit the market. Moreover, there were 136 mergers between 2002 and 2006 that were seen as a result of fierce competition while the market struggled with overcapacities and declining sales, all approved by the Cartel Offices. Finally, State and Federal Cartel Offices approved structural-crisis-cartels or cartels of small and medium-sized enterprises under supervision of the cartel authorities („Mittelstandskartell") in some regional markets in order to facilitate capacity reduction and the process of adjusting to the new market conditions (Deutscher Bundestag 2003, 2005, 2007). These three cartels meet our defined analysis needs. Each was large enough to influence competition at the national level and included all major suppliers and producers. All three cartels had illegal organizational structures and formal cartel agreements needed to coordinate participating firms. Hence, it is expected that collusion without such a structure is not as effective. Therefore, we expect fiercer competition without such an organizational structure. Each of these sectors is characterized by a relatively homogeneous good. Finally, all were heavily fined due to the extent of the distorted competition.

6

Empirical investigation

We conduct our analysis in three steps. First, we present the regression approach as suggested by Boone et al. (2007) to measure competition and discuss its main drawbacks. Second, we apply the BI, as defined in section 3, to real data, reveal its drawback and propose an augmented indicator correcting for firm size. Finally, we use cartel cases as natural experiments to test for the empirical applicability of the Boone-Indicators and compare its performance with that of the traditional PCM. 6.1 The Boone-lndicator and its modification We start by briefly discussing the applicability of the BI on real word data. Based on the idea that the more efficient a company becomes the greater the profit should be, ceteris paribus, Griffith et al. (2005) were the first to propose a regression of average variable costs on logarithmized profits. Since marginal costs are usually unobservable, average variable costs (AVC), defined as total variable costs divided by sales, are taken as a crude

346 · Alexander Schiersch and Jens Schmidt-Ehmcke

proxy for marginal costs. AVC is also used to assess firm efficiency. The estimated betacoefficient measures the profit elasticity of the respective firms. M o r e precisely, „...it measures the percentage decrease (increase) in firm i's profit if its variable costs (i.e. marginal costs relative to price) increase (decrease) by one percentage point." (Griffith et al. 2005: 6). Since the relationship between profits and average variable costs must be negative, the estimated coefficient needs to be negative. As competition intensifies, the slope of the regression should become even more steeply negative, following the idea that inefficient firms are punished more harshly by fiercer competition. We adopt the regression approach for comparative purpose. However, we follow Creusen et al. (2006) and estimate the elasticity by means of yearly log-log regressions: 1η(π, ;( ) = a — ββ 1η(Λ VQjt) + e,/( with t = year,; = firm and / = market. The coefficient plainly gives the percentage change in profits due to a one percent change in average variable costs. As shown by Boone et al. (2007), using simulations within his theoretical framework, changes in competition are correctly identified by this model. However, measuring competition by means of regression analysis is not without problems. One challenge is the definition of markets. The more precisely we can capture a market, the less other factors or markets will influence the outcome and the better the subsequent competition estimates should be. On the other hand, the more precisely we size a market, the fewer observations we will have. Moreover, while markets with few players are of special interest for competition analysis, fewer observations decrease the stability of regressions. Another problem is related to firm size. As long as we operate under the model's assumptions, the most efficient firm must become the biggest firm in terms of market share and consequently, due to its efficiency level, it also must make the greatest profit. With respect to linear regression analysis we must consider that in reality big firms are not necessarily the most efficient ones and thus, it is possible to find a nonnegative beta-coefficient. 1 4 Therefore, in addition to the regression approach, we try to estimate the RPDs directly. Initially we assess the efficiency of firms in a one-dimensional and transitive way. Following Boone (2008a) we use two different definitions to calculate efficiency: the average variable costs, defined as total variable costs per sales ( T V C / s a l e s ) and labour productivity, defined as gross value added per employee (VA/employee) as efficiency index N. We add sales per employee (sales/employee) as a third measure. Accordingly, for each efficiency measure we calculate the RPDs for market (/') and firm (i) in every year (i) as: RPD,,t(n)

=

n ,

' f n l ~ * i t { f ] with t= 1 , . . . , T, / = 1 , . . . , / a n d / = 1 , . . . ,J, ίΑη ) — πίΑη )

π

(4)

where profits are defined as sales minus total variable costs (πψ(η) = Sijt(rì) — TVCijt(n)). The RPD can only take values between zero and one, with one for the most efficient («") and zero the least efficient firm («'). The efficiency of the firms is normalized via (η — η')/(η" — η'). As the example in Figure 2 reveals, which depicts the RPDs for the cable industry (4 digit classification) in 2006, plainly applying this model on real world data is not sufficient. 1 5 The RPDs are located between roughly 30 and -1, while we should see a scatter plot 14

15

That both, size and few observations can have an impact is not just a theoretical problem as the results of Griffith et al. (2005: 11 ff.) show. In this example the efficiency is defined by average variable costs. However, we can show further examples with labour productivity.

Is the Boone-lndicator Applicable? · 347

co

m CJ ocj α tier

m

0.4

0.6

0.8

1.0

normalized efficiency Figure 2 RPDs for cable industry in 2006 1 6

within the boundaries of zero and one. We do have the most efficient firm located at the coordinate (1,1) as it should be. However, that firm has a profit below many of its competitors and it follows (πψ(η) — πμ(η')) > (π,,(η") - π,,(η')). Thus, the RPD will take values above one for firms with greater profits. At the same time we have a least efficient firm, which has a profit above that of other firms, resulting in negative numerators for these observations. This happens due to firm size. Obviously in the real world there are firms that are really efficient, no matter of how efficiency is captured, but they can be small, at least at certain point in time. On the other hand, large companies may not be as efficient, but because of the larger size, the firms make larger profits. To overcome this problem the RPD must be calculated accounting for firm size. This can be done by means of number of employees or sales. However, applying workforce to normalize profits does not give a good fit. We still have RPDs significantly below zero and above one (Table A2), regardless of the efficiency index used. This might be caused by a weakness of the data set. It lacks information about the number of temporary workers and for how long they stayed in the company, but we know through a costs category that some firms used temporary workers. This biases the profits for the respective firms as well as the efficiency if labour productivity and sales per employee is applied. Therefore we do not use workforce in the calculation of the efficiency or to normalize profits. Instead, in the subsequent analysis profits will by normalized by sales, which is turnover out of the core business without trading or other activities. Thus, the RPD is calculated as:

RPD,jt{n)

16

=

n,,t{n)/sales,jt{n) Kit{n")/sales

- njt(ri) / sales jt{n')

{n") - njt(n>)/sales (ti>)

T h e efficiency was measured by average variable costs.

(5)

348 · Alexander Schiersch and Jens Schmidt-Ehmcke

For the efficiency index we apply the average variable costs. In order to estimate the area under each curve we use Data Envelopment Analysis (DEA). It is a nonparametric method that envelops the scatter plot at its outer boundary. We abstain from presenting the method here and refer the interested reader to Canter et al. (2007) for a concise introduction and to Simar and Wilson (2005) for a detailed discussion. Given the estimated curve we are able to derive the corresponding area by integration. 6.2 Results Given the above defined RPD we proceed with testing the validity of the Boone-Indicator using the cartel cases. For this end we estimate the PCM of each firm as proposed by Boone (2008a), and aggregate them into yearly industry PCMs using market share as the weight. 17 The market share is derived as the share of the firm sales on industry sales in a year. Further, we estimate the BI as beta coefficients of the above outlined regression approach (afterwards also called parametric indicator). 18 Finally, we calculate the modified RPDs and the respective areas as discussed above (also called nonparametric indicator). The change in competition is measured by subtracting the respective indicator in the base period from itself in the reference period. Regardless of the indicator under consideration, a positive result shows an increase in competition between the periods. A negative result on the other hand indicates decreasing competitive pressure over time. We used Welch's t-test for evaluating the significance of changes in PCM since it accounts for unequal variances in two samples. The same test is applied when comparing the beta coefficients. However, the test can only be applied if both of the beta coefficients are significant on their own. Otherwise, we depict just the difference labelling it as not significant. Since the level of competition by means of RPDs is measured as area, tests based on means and variances cannot be applied, since an area has no variance. Therefore, the significance of differences between the areas is calculated applying the nonparametric Wilcoxon rank-sum test on the underlying RPDs. Given the estimates and tests, we can use cartel events to derive the validity of the indicators. As discussed in section 5 we expect fiercer competition in the aftermath of the uncovering of a cartel. Hence, we look at the estimated changes in competition after such an event compared to periods before the event. When possible, we look at the three years after and the three years before the cartel case. The year of the event is not taken into consideration, because the effect on the competition level in that year is not straightforward since we only have annual data. The relevant biannual differences are presented in Table A3 to Table A5. The first cartel we take a look at is the cable cartel. The cartel was uncovered in 1996. Due to time limitations in our data set we can only compare the changes in competition between 1995 and 1997 to 1999. Looking at the Lerner-Index (Table A5), in all of the respective biannual comparisons we see positive differences where two out of three of 17

18

Boone ( 2 0 0 8 a ) follows Nickell ( 1 9 9 6 ) and Aghion et al. ( 2 0 0 5 ) how proposed the approximation of firm i's price cost margin by (revenue, - variable C0StSj)/revenue¡ To overcome the outlier problem, we trimmed the data as described in section 4. Moreover, we also applied the outlier robust MM-estimator introduced by Yohai ( 1 9 8 7 ) on untrimmed data. For a detailed discussion on the MM-estimator and its application see Verardi and Croux ( 2 0 0 9 ) . The results of that approach are in line with the results in Table 4 for which reason we abstain from presenting the results here. However, interested readers can request the results from the authors.

Is the Boone-lndicator Applicable? · 349

these differences are significant. Thus, the PCMs indicate the expected increase in competition. The nonparametric measure (Table A4) also indicates fiercer competition in the years 1997 to 1999 compared to 1995, although just one of the differences is significant. The differences in elasticities (Table A3) on the other hand are positive and negative, thus signalling decrease and increase in competition after the cartel was terminated. However, the betas are not significantly different. This is caused by the fact that the beta coefficient in 1995 is not significant at a 10 percent level. Even if we ignore this fact and test for significant differences in betas, we find all changes to be significant, not only the positive difference. Thus with respect to the aim of this paper we must state that P C M and the nonparametric indicator behaved as expected, indicating fiercer competition in the aftermath of the termination of a cartel. In contrast, the parametric indicator did not behave as expected. Looking at the cement cartel, we find similar results as for the cable cartel. As discussed above, the cartel was uncovered in 2002, thus we evaluate the changes in competition from 1999 to 2001 against 2 0 0 5 to 2006. 1 9 Again, the weighted PCMs signal fiercer competition after the event, where all results are significant. Yet, it is now the parametric indicator signaling fiercer competition without exception and with all changes significant. To a certain degree the area changes also indicate fiercer competition. However, none of the changes are significant. Thus, although we only find positive values indicating fiercer competition, with the absence of significance we must record that the nonparametric indicator shows no change in competition after the cement cartel was terminated. Finally, we look at the ready-mixed concrete cartels, where the first one was uncovered in 1999 while the last one stopped its activity in 2001 as discussed before. We therefore define 1996 to 1998 as base period and 2002 to 2004 as reference period. As presented in Table A5, the PCMs differences again show the expected sign and are all significant. The parametric indicator on the other hand is pointing to the opposite direction. All differences are negative and at least two are significant. If we overlook the insignificants of the 2003 and 2004 betas and test for differences, six out of nine negative differences would be significant. The parametric indicator actually suggests weaker competition in the aftermath of the termination of the ready-mixed concrete cartels. The nonparametric indicator is not performing better. Although seven out of nine biannual comparisons are positive, pointing toward fiercer competition, two are negative and no change is found to be significant. Table 2 Change in competition after the termination of the cartels20 Event Industry (4-digit classification) power cable (3130) cement (2651) ready-mixed concrete (2663)

19

20

PCM

Boon-Indicator

PCM

regression RPD 1996 2002 19992002

r τ*

Î*

τ*

I'

Î" •

Boon-Indicator regression RPD

î î

î

î

I

î î

T h e r e are to few observations in 2 0 0 3 and 2 0 0 4 a f t e r r u n n i n g the outlier detection, so t h a t we could not use this years due t o the data protection rules of the Research D a t a Centres. * m a r k s the direction of changes in c o m p e t i t i o n t h a t take significance into account.

350 · Alexander Schiersch and Jens Schmidt-Ehmcke

Table 2 summarizes the m a i n findings, depicting the direction of changes in competition with and w i t h o u t taking the significance of the changes into account. Here it is especially interesting that the p a r a m e t r i c Boone-Indicator points just once in the expected direction, regardless of significance. The n o n p a r a m e t r i c indicator, o n the other h a n d , fails just once if we ignore significance. Indeed, if we look closely at t h a t unexpected o u t c o m e we find seven biannual differences out of nine signalling fiercer competition, thus pointing into the expected direction. 7

Conclusion

Using a rich, newly built, data set for G e r m a n m a n u f a c t u r i n g enterprises, we test the empirical validity of the Boone-Indicator (Boone 2008a). This is a new competition measure that, f r o m its theoretical properties, proved to be more robust t h a n the Lerner-Index. However, the proof of its empirical practicability and robustness is missing. This paper aims to shed light on that question. To this end we use large cartel cases as events for c o m p a r i n g the competition levels before and after a cartel was uncovered and stopped operating. Since all of the chosen cartels significantly restricted competition, w e expect fiercer competition after such a n event. In the empirical analysis w e c o m p a r e the p e r f o r m a n c e of three competition measures. T h e first is the Lerner-Index as a classical measure of competition. T h e second is the Boone-Indicator calculated as beta coefficient of a log-log regression, as p r o p o s e d by Boone et al. (2007) a n d various other authors. Finally, the Boone-Indicator derived by means of Relative Profit Differences (RPD) is calculated using real data for the first time. O u r analysis reveals t h o u g h , t h a t the latter c a n n o t be applied to real data as theoretically defined. This is because the relationship between the efficiency of a firm and its profit level is not as designed in Boone's theoretical f r a m e w o r k , where the most efficient c o m p a n y is always, by design, the biggest firm in terms of m a r k e t share. O u r results suggest t h a t this relationship does not hold in reality. Therefore w e propose a w a y to account for firm size in the calculation of RPDs. W i t h respect t o the p e r f o r m a n c e of the indicator in the face of uncovered cartels w e note that the Lerner-Index p e r f o r m s as expected. In every case it indicates fiercer competition in the a f t e r m a t h of a cartel with almost all biannual comparisons being significant. Hence, although not theoretically robust, in this analysis it proves its empirical usefulness. This w e c a n n o t state for the t w o Boone-Indicators. The regression based indicator just once indicates fiercer competition, regardless of whether or not we account for the significance of changes and betas. This supports our d o u b t s regarding this a p p r o a c h . T h e Boone-Indicator based o n RPDs also does not p e r f o r m as well as the Lerner-Index. This is mainly because the changes are often not significant. Leaving significance aside, the Boone-Indicator by means of RPD and taking firm size into account p e r f o r m s almost as well as the Lerner-Index. Based on our findings we conclude t h a t the Boone-Indicator, a l t h o u g h theoretically superior, is, at least at this stage, not an empirically robust indicator. T h e Lerner-Index o n the other h a n d indicates changes in competition as expected. However, the results of the RPD based Boone-Indicator are promising. Future research should focus on alternative m e t h o d s t o account for firm size while keep the original variation of the profit levels.

Is t h e B o o n e - I n d i c a t o r A p p l i c a b l e ?

ΓΜ^ΟΓτίΐηΟΟίΝΦΟΟ N M N i n œ o ? ' d d d d d d d d N S S m t i í i m i f i iTNOoin^Nff* roronnmmmrors d o d d o d d d

· 351

IN no m O m

M ^ s O N m m i r N ^ o r o m N o j c o m f s N ' t T m O i N in m r o r m '

T-l-pvfÑ^OOCOOO omoocoin^ccom NN't'tfNOOT-O Tconifiom'Xi'î i f i N O m O c o f ® f í N ^ t T m i r ( N I N r r r r r r

in σι τ- in o h» ^O ΓΝ m ^ ο ο ο ο ο ο ο Φ οτ—ί τ— σ ί οτ—ί ΓΜΓΜΓΜΓΜΓΜΓΜΓΜ Φ ο ο ο ο ο ο ο (1593) *uaujw

ΙΛ Ν CO _ . . . σ * σ ι σ ι σ * σ » ο ο ο ο _ _ _ σ \ σ > σ σ \ σ * ο ο ο ο ο ο ο *-*-*-*-«-ΓΜΓΜίΝ(ΝΓΜ σν σι ο — τ— τ— Τ~ (Ν >< τ- τ

I I I I I I I I I I I

* * *

• * #

* * *

* * *

# * *

* * *

# * *

* #

* * *

# *

* * *

# * *

* *

* * #

* * *

* • *

* #

* * *

* *

* * *

* *

• * *

* • *

* * *

*

* *

* #

# * *

• • •

* * *

* * •

* * *

* # *

* * *

• * •

* • *

* *

* * *

* *

* * *

* * *

* *

* *

* # *

* * *

* • *

+

co οο * ιη σ> Έ)· m σι τ— ιη οο τ- IV ο m IV Ο ΓΜ νο 00 ιη (Ν IV ο τ-^ Ò ο m ο ιη ο

τ— IV σι IV τ— Tt ιη Ο * # *

* *

σ> ΓΜ m ΓΜ νο IV

νο ΓΜ ιη νο ΓΜ ο m IV 00 τ- (Ν σι ο τ— IV co ο t' ο Ο Ο ο ο

* *

*

* * *

* *

* *

ΓΜ Ν νο νο m Ο ο ο ο Ö ο

•

σ> ιη ο ιη ο ο

τ- ΓΜ m ιη νο Ο Ο ο ο ο ο Ο Ο ο ο ο ο (Ν ΓΜ ΓΜ ΓΜ ΓΜ (Ν

* * *

* * *

νο ιη •sf co IV τ— νο 00 (Ν νο σι d ΓΜ οο ^l· m

ιη • τ— TJ- * ιη ΓΜ οο m Ο νο IV νο (Ν νο τ— co V0 o m ο m νο m IV νο ιη 00 m ο νο ο ΓΜ ο ο ο ο ο ΓΜ

τΐ σι ιη Ο ιη ο ο

00 'S- ιη τ— ο 00 νο το ο ο Ο ο ο ο ο

ΓΜ IV •3·ΓΜ ιη TJ- Ο ΓΜ σ\ V0 ΓΜ 00 Ο Ο mΟ Ο Ο ο Ο Ο Ο ο ο Ο Ο ο I

τ— σ\ τ— τ— τ— 00 νο ιη οο 00 •ί- τ— m ιη m τ- ο ΓΟ ο •ΊΙ·ιη τ- ΓΜ Ο ο ο Ο ο ο Ο ο

I

*

ΓΜ ιη ΓΜ (ν V0 ΓΜ 00 ΓΜ νο νο m ο ο ο Ο ο Ο Ο ο ο ο ο ο Ο ο

* * * * φ ΓΜ σ\ 00 00 νο ΓΜ IV m Ti- ιη ιη τ— τ- ΓΜ τ- m (Ν τ- m Ο Ο Ο Ο Ο Ο Ο ι

I I I I I I I I I I I

*

* *

νο τ—

* *

I

m ιη νο (Ν Τ- ΓΜ Ο ΓΜ

* *

*

in vo ΓΜ

* * *

ΓΜ * τ— νο ιη (V) ιη ΓΜ ο ιη co m Ο ο ο Ο

(Ν νο νο ΓΜ ο ΤΟ Ο I

οο σ> οο 00 ΓΜ |ν m ιη ιη ΓΜ m (Ν Ο Ο Ο Ο ο ο

* *

"o Q. ΙΟ -O M

* *

• * *

* * *

• *

Q

*

* * *

0 ΓΜ IV # .Ω 0 νο ΓΜ σ\ "O ο 8 ο m δ 0J ο ο ο ο ò Ν

E o c Í— S 4 vo O (N Q_

* * *

τ— m τ— •Ί- 00 co 00 raτ— 00 m τ— IV m ΟΟ 00 ΓΜ ο ΓΜ τ- ο m τ— ο ο Ο ο ο ο

no ^— σι m ΓΜ Ο Ο Ο

ΙΛ 4) 0) O Q. E (Ν 00 τ— τ- IV Ο m ιη Ο ΓΜ νο 00 ιη ΓΜ IV ΓΜ ΓΜ IV ο τ— Ο ο ni ο ιη ο m Ο V0 ο τ—

* * *

* * *

IV IV 00 τ^ Ο * * *

# *

νο ΓΜ ιη m ίο

* * *

τ— τ— t τ— ιη ΓΜ τ— co νο ΓΜ νο οο VO τ— ο νο m ο ιη νο m IV ιη ιη 00 ΓΜ ο ο ο Ο νο ο ΓΜ Ο ο ΓΜ * * •

νο IV σ\ ΓΜ m νο (Ν ο τ— IV 00 τ- ΓΜ σ\ ΓΜ τ— IV ο Ο Ο ο IV

ιη τ— m νο ο νο (Ν ο

ιη νο IV 00 σι Ο τ- ΓΜ ΓΟ Tt ιη V0 σι σ\ σ\ σν σ> ο Ο Ο Ο Ο ο ο σι σι σι σ> σ> ο ο Ο Ο Ο ο Ο Τ— τ- τ— τ— τ- ΓΜ ÍN ΓΜ ΓΜ ΓΜ ΓΜ ΓΜ

Is the Boone- Indicator Applicable? · 353

*

* *

•

* * *

* * *

•

*

•

*

* *

* * *

* *

* * *

* * *

* * *

T— * IV VO vp in ivi IN ni ni ni *

*

m νο in IN * VO t 00 IN τ— CO o IV IN VO IN ! 00 m Τ- IN in o Ö Ο o τ- τ— τ— r-

ι

ι

ι

ι

ι

ι

ι

ι

ι

ι

ι

ι

*

•

* * *

* *

* * *

* * *

* * *

« * *

*

•

τ— IN m * IN ιη τ— τ— m 00 00 m rn ν ο τ— Τ— ο vo VO σι ο IN IN O

* *

* * *

* * *

*

•

* *

* * *

* * *

ι

ι

ι

ι

ι

00 V0 σ m (Ν Γν ο Ο

*

IV m I ΊΟ

I

*

T-

I

I I

ιη VO IV σι σ\ σ σι σ σι τ— τ— ^—

I

I I

ι

ι

ι

ι

8

(1-593) íuaiuao

00 m 00 00 σ> τ-; ο IN Τ ο V

• * *

ΓΜ IV Ο Ο ι Τ- IN Ο τ—

νο •sf σ m νο ο m IV οο Μ- τιη Ο IV νο Ο ò τ • IN ο ò ni

ιη ιη ιη τ—

* * *

•

* *

•

*

τ— * σ\ σ\ IV IN τ— m ιη οο IN VO τ— νο τ— ο σ νρ IV σ ^ IN IN ni Ο ο

*

•

•

*

ιη ιη ι ιη IN

ι

ι

Τ— IN (Ν (ν m 00 00 νο IN νο νο IN ιη τ— Τ (Ν ni (Ν

I

I

I I I

οο τIN τ-

I

I

I I I

m m νο 00 IN rn (Ν IN ο IN

ιη IN IV ο

ι

ι

τ- * ο ο iv m σ> τ- m σι IN Ν . ιη ιη ο ο in rn ι ni IN Iii ΓΝ IN ιν m νÒ

rv rv νο IV τ, Ο IV r^ I (Ν IN Ó

I

I

I I I

* *

•

m νο I 00 IN

I

I I

I I

I

I

I I I I

I I

I I I I

*

• *

*

σ 00 * * Ο VO I ι ο (Ν τ— ο

I

I

I I I I

σι σ\ œ ιη VD I V IN o

σ\ σ\ SS ι :

I I

I I I I

σ ιη 1 1 οο ο τ—

I

I I I I

I

I I I I

co σ\ in IN m σι σι Ο Ο O o o σ\ σ\ ο Ο Ο O o o τ— τ— (Ν (Ν IN IN IN IN

ιη IV ιη ò

* * *

?

I I

ι

οο σ\ τ— τm IN 1· τ- ιη τ— ^r Τm ni IN Ο * *

• * *

«

m t r^ IN

IV νο VO IV τ— m IV ο m IV οο τ— in VC in τ- VD ιη · οο IV τ— νρ τ— Τ— (Ν rò τ— ο ο

νο σι νο ο (Ν m (Ν ο

τ

ΓΝ τ— * m σι m CO 00 νο σ> Τ- τ— IV Ο Ο ν τ— τ— τ—

ιη 00 (Ν IN IN σ\ Το Ο

•

m in IN σ* ιη m IV I I IN (Ν IV IV IN V V V

00

τ— σ m σ> m τ— ^ ιη ν Τ— (Ν co (Ν (Ν Ο

V0 IV τ-; ni

m IN in rv 00 IN IV οο in in m (Ν IN Γν 00 IV ι in τ— Tf m σι Ο τ— IN ν τ- τ— o ο ο

ιη νο •t IV νο IN IV IV ιη ο Ο rn ο Ο V IN ν

*

σ\ * IN oo Ί· t τ- τ— IN τ- — ι VO vp Τ- I ΙΝ ν ν T— Ο •

'Ιt1— t Ο m o o

• •

IV οο σ\ 00 'ίιη ιη Ο (Ν ν (Ν •

1 1

co σ\ σι τ- σι ιη IN 00 IN IN τ— Ο σ> ιη m VO IV m VO ο V0 in Τ— IN rn ni σΐΓη· IV οό o ιη m

νο m ιη m

PO O o (Ν

Ο τm \0

ο rri σι τ—

m 00 m m

VO ιη νο vo σί (ri o ιη ιη m m

>8 3 «

ΓΜ O O ΓΜ

Ο ιη m VD

IV σί σι Τ—

o oo m m

oo m VO σί o ιη t m

Ε « m ο Ο Ο Ο rOIN 1) 4) CJ 0) o I ° § s 1 1 ja Ό Ό Ό Ό "Ό Ό Ό u •Mg. C C C C C C «J ! T c c « c t 3 J . E . S > « ki: o 3 3 3 3 3 3 C υ 01 Ό £ Λ Λ £ £ ^ j -Q -η -η — ^ Sì τ^ >- -£ "rt"l "ÏTi il > Λ 1 "7Γ5 C SI¿ i m O ν ο s c "δ "δ υ L> s - i i Q_ ι/1 ο U Κ S. S S δ i "TS 0 2 « ΐΕ Χ) ® ï> - C W 2 I - i ρ Κ ί) e Ft E E Sí ° « ^ "(5 1) Ό t¡ 3 3 s » » c (d QJ 3 — ΠΪ o -α (U ΙΛ .. ρi I) S ·. ·. •. a > u (0 M j * O O ¡υ ω u 1 >- u _ β> δ ν g J g a e tí f 5 T i S j o S i ! υ υ S ι ό » ? ι 3 « 5 § ^ -c JJ E c o in in ΓΜ

\$ϋο

402 · Sebastian Petrick, Katrin Rehdanz, and Ulrich J. Wagner

Table A3 W Z 2003 industry codes and according descriptions WZ 2003 code 100* 141 142 140* 151 152 150* 155 156 157 158 160* 171 172 173 174 175 176 177 180* 182 191 192 193 201 202 203 204 205 211 212 221 220* 230* 240* 243 244 245 246 247 251 252 261

Mining and quarrying of energy producing materials and ores (includes 101 - 132) Quarrying of stone Quarrying of sand and clay Mining of chemical and fertilizer minerals, salt and other mining and quarrying n.e.c. (includes 143 - 145) Production, processing and preserving of meat and meat products Processing and preserving of fish and fish products Manufacture, processing and preserving of fruit, vegetables, animal oils and fats (includes 153 - 154) Manufacture of dairy products Manufacture of grain mill products, starches and starch products Manufacture of prepared animal feeds Manufacture of other food products Manufacture of beverages and tobacco products (includes 159 & 160) Preparation and spinning of textile fibres Textile weaving Finishing of textiles Manufacture of made-up textile articles, except apparel Manufacture of other textiles Manufacture of knitted and crocheted fabrics Manufacture of knitted and crocheted articles Manufacture of leather clothes and articles of fur (includes 181 & 183) Manufacture of other wearing apparel and accessories Tanning and dressing of leather Manufacture of luggage, handbags and the like, saddlery and harness Manufacture of footwear Sawmilling and planing of wood; impregnation of wood Manufacture of veneer sheets; manufacture of plywood, laminboard, particle board, fibre board and other panels and boards Manufacture of builders' carpentry and joinery Manufacture of wooden containers Manufacture of other products of wood; manufacture of articles of cork, straw and plaiting materials Manufacture of pulp, paper and paperboard Manufacture of articles of paper and paperboard Publishing Printing and service activities related to printing; reproduction of recorded media (includes 222 - 223) Manufacture of coke, refined petroleum products and nuclear fuel (includes 231 - 233) Manufacture of basic chemicals, pesticides and other agro-chemical products (includes 241 - 242) Manufacture of paints, varnishes and similar coatings, printing ink and mastics Manufacture of pharmaceuticals, medicinal chemicals and botanical products Manufacture of soap and detergents, cleaning and polishing preparations, perfumes and toilet preparations Manufacture of other chemical products Manufacture of man-made fibres Manufacture of rubber products Manufacture of plastic products Manufacture of glass and glass products

Energy Use Patterns in German Industry: Evidence from Plant-level Data · 403

Table A3 W Z 2003 industry codes and according descriptions (continued) WZ 2003 code 262 263 264 265 266 267 268 270* 273 274 275 281 282 283 284 285 286 287 291 292 293 294 295 296 297 300 311 312 313 314 315 316 321 322 323 331 332 333 334 335 341 342

Manufacture of non-refractory ceramic goods other than for construction purposes; manufacture of refractory ceramic products Manufacture of ceramic tiles and flags Manufacture of bricks, tiles and construction products, in baked clay Manufacture of cement, lime and plaster Manufacture of articles of concrete, plaster and cement Cutting, shaping and finishing of ornamental and building stone Manufacture of other non-metallic mineral products Manufacture of basic iron and steel and of ferro-alloys, manufacture of tubes (includes 271 - 272) Other first processing of iron and steel Manufacture of basic precious and non-ferrous metals Casting of metals Manufacture of structural metal products Manufacture of tanks, reservoirs and containers of metal; manufacture of central heating radiators and boilers Manufacture of steam generators, except central heating hot water boilers Forging, pressing, stamping and roll forming of metal; powder metallurgy Treatment and coating of metals; general mechanical engineering Manufacture of cutlery, tools and general hardware Manufacture of other fabricated metal products Manufacture of machinery for the production and use of mechanical power, except aircraft, vehicle and cycle engines Manufacture of other general purpose machinery Manufacture of agricultural and forestry machinery Manufacture of machine tools Manufacture of other special purpose machinery Manufacture of weapons and ammunition Manufacture of domestic appliances n. e. c. Manufacture of office machinery and computers Manufacture of electric motors, generators and transformers Manufacture of electricity distribution and control apparatus Manufacture of insulated wire and cable Manufacture of accumulators, primary cells and primary batteries Manufacture of lighting equipment and electric lamps Manufacture of electrical equipment n.e.c. Manufacture of electronic valves and tubes and other electronic components Manufacture of television and radio transmitters and apparatus for line telephony and line telegraphy Manufacture of television and radio receivers, sound or video recording or reproducing apparatus and associated goods Manufacture of medical and surgical equipment and orthopaedic appliances Manufacture of instruments and appliances for measuring, checking, testing, navigating and other purposes, except industrial process control equipment Manufacture of industrial process control equipment Manufacture of optical instruments and photographic equipment Manufacture of watches and clocks Manufacture of motor vehicles Manufacture of bodies (coachwork) for motor vehicles; manufacture of trailers and semi-trailers

404 · Sebastian Petrick, Katrin Rehdanz, and Ulrich J. Wagner

Table A3 WZ 2003 industry codes and according descriptions (continued) WZ 2003 code 343 351 352 353 350* 361 362 363 360* 366 371 372

Manufacture of parts and accessories for motor vehicles and their engines Building and repairing of ships and boats Manufacture of railway and tramway locomotives and rolling stock Manufacture of aircraft and spacecraft Manufacture of motorcycles and bicycles and other transport equipment n.e.c. (includes 354 -355) Manufacture of furniture Manufacture of jewellery and related articles Manufacture of musical instruments Manufacture of sports goods, games and toys (includes 364 - 365) Miscellaneous manufacturing n.e.c. Recycling of metal waste and scrap Recycling of non-metal waste and scrap

Source: Eurostat 2010, Statistisches Bundesamt (2003). To ensure confidentiality of the data, several sectors have been merged together as noted in the table. They are marked with an *. The total number of sectors has thus been reduced from originally 116 sectors to 98 sectors.

E n e r g y U s e P a t t e r n s in G e r m a n

o £ k. O

00 IV VO m

*»

m ιη ΓΜ o ιη ^f 't σ\ 10 m

I n d u s t r y : E v i d e n c e f r o m Plant-level D a t a

· 405

ο ιη

m t

"t ΓΜ 00 in σ> m

σι ο νο «Ν

ιη

ΓΜ o 00 co Γν m 00 ΓΜ 00 in τ— τ—

σι IV ο τ—

in in ΓΝ τ— IV m νο σι 00 ΓΜ νο 00

ιη m £ νο ι-

ο IV m m

co m 00 σι ΓΝ ιη ΓΜ CO ιη IV m ο (ν σι ΓΜ ιη

ο τ— co m

m O m

σν σ> VO IV

νο IV ΓΜ m

(Ν 00 r-

0

sP \0 0^ m CO 3 σι σί ιη ο m ΓΜ ΓΜ m

NÍ Ο^ ιη V σ\ ΓΝ m 00

ο^ 3 ΓΜ

IV VC tv

ιη ο σι m

(Ν Ο m m m o 00 'Ι- ιη 00 00 ιη ·N 0 0

o Tt

K a t r i n

R e h d a n z ,

a n d

m νο on Ν

ΓΜ i n i t in

m O r-N Ο

m Ο δ O m ΓΜ σ> I S

η τ ιη

tv is

Τ-

m

ΓΜ Τ

νο τ Τ-

U l r i c h

CO Γ Μ 0 0 τ— O 00 t v τ— I S ν ο

J.

W a g n e r

ΓΜ I S 00 νο τ—

is in

IS

σι is

m m

IS O

σ > VO ο ο m 00 o τ— m τ -

Ι" t

οο ο \ IS

o m

τ-

00 i s σ\ νο

S Ν? «*|

00 τ τ -

ο σ\

m m in

m m ΓΜ ι η ι η νο

σν ιη 00

m

ΓΜ «C ΓΜ

5

Ss ra α.

Ν

ÍN

(Π tv in

« ΐ 5 vi : ω S

Β - o

o gg

t

m o o • t τ— *— m τ i n m ÍN

rs

in m τ-

Ο IS Ο

σ\ (ν

τ— Ο ^ οο W

IS O νο ι η 0 0 m τ— ι η r s Τ— I S i s

o

•t

00

m m

οο

is o σι

ΓΜ ΓΜ ι η

σι νο σ\

ΓΜ IS

τ— ι η is m t— ο

Ο m ΓΜ

in co in

σι m τ~

νο IS τ— m

νο i n in σι m IS Τ—

VO o o m 00 οο σν i s οο σι

σι

t3

E S * ν 7TT · >

^

οο I

o

t

οο

N ^ N t ^ I ι η γμ Tt vo

O VO 1*1 VO T t σ> T t ν ο i s σ \ Γ · (Ν O M N i

c —

C 0) _ J?

O υ

0) . c ¡Λ 3 3 S

g - K g O) C O c ί ζ Ul r i 5

ιη m m

I S σ ι τ— ο ν ο ΓΜ σ ι m ν ο τ— I S i s

0 0 ΓΜ i n 00 νο o m τί

m

m m 00 ΓΜ ΓΜ * *

\0 \Ρ ί " · ff^

T l ' τ-

¡£> o\

Is is 9\

σ\

ΓΜ m

m

ιη

(Ν Ν is

m" m νο

σί σι

ο Τ*

ο ο ΓΟ m οο m σι

ΓΜ ο m is ΓΜ τ—

S

is ιη τ-

m νο co

m is m

TI- m CO i s in m

o m

ΓΜ

ο ^ · θ Ν Γ 0 ο ί ο ι ή Ν σ > » · ο ΐ ι ^ - o n ^ - o n o n v o i n N T t T - * '

\ 0s 0 0 0 ΟΝ CO IS σί I s m τ * Tí-

«I W ITI η V IN W >« Mí ΓΝ r M O M i s v o i s i s o o ^ - a i i n ^ o i r M v o i s © T t T - * - o ( r i < j \ i n i n r M r s

in m σ ι τ— ΓΜ Τ -

IS 00 in

νο ΓΜ ιη

γμ

ο νο τ— τ-

τί

σι o σι ο οο νο ι η νο ν ο 0 0 ΓΜ σ ι τ— 0 0 t σν T t m ΓΜ 0 0 I S τ— τ ΐ - ι η o n ι η ΓΝ σ\ m m

m Is 00 00 νο

I S σ> Κ σι rs ΓΜ m οο Ο τ * τ— m τ— τ— τ - τ ιη is t m

ΓΜ IS

νο νο τ» ιη rs

νο IS TT

ΓΜ 00 ID

00

IS o Is

«Ν ι η ι η σ \ 0 0 0 0 Ο σ ι τ— 0 0 i s σ ι σ > 00 on m ΓΜ ΓΜ τ - ο σ ι T t ΓΜ ο ιη ιη is m σ ι 0 0 ΓΜ ι η Τ - ν ο τ~ οο IS on ο ΟΠ τ *

*

00

σ> ΓΜ ιη on

t σ ι I s νο 00 0 0 o d T t τ— VO I S τ— T t τ— ν ο i n τ— τ— τ - τ—

vi

Ρ Ν fN

οο Ο ι η ΓΜ ο ΓΜ 0 0 m Τ- τ - σι

τ—

νο Is IS τ-

νο Ο νο Ν ιη Τ—

Γ Μ τ— m

Ό C πί -Μ C

CL ì5 U

JJ S m τ— O ιη Tt ο ι η IS 00 σ ι

in in

σι IS ΓΜ m ν ο τ— is

00 i s οπ σ ι το

τ—

is. T t ΓΜ ori i n νο m ΓΜ

ΓΜ τ

-

οπ

ΓΜ rs νο

σ ι »J· νο ι η νο ι η ο m m

ιη

ΓΜ o n τ— T t τ ί 00

ιη σι m

Ο 00 ο τ— Γ Μ 0 0 t — τ— < Ν

ΓΜ

o νο t Ο i s ΓΜ S 0 0 ΓΜ 0 0 o 00 νο ο S is ιη ο co IS ι η o ΓΜ 0 0 ΓΜ r s ν ο T T τ—

Ο σ\ νο ΓΜ

ΓΜ

νο

'S· •Φ i n m σι

νΟ 0s·

σι οή Tt ιη Tt ο

ιη m ιη

rs

Ο 00 m on ΓΜ I S

νο

IS rs rs τ-

σι οο οο οο

Tt σι 00 νο ΓΜ m IS

σι

ο Tt ö ΙΛ ΰ

S

•§ ο

ε - S i ο ο_ ο_ g l -σ ·σ s c c « Ι - o r

ιη •sr τ— ο π ΓΜ i s IS i s σ \ ΓΜ

ν ο ΓΜ ΓΜ

0 0 τ— νο Tt 00 i s σ ι ΓΜ

,

Ò ci νο σι νο 00 m τ— σι Is ΓΜ σ ι

ΓΜ

5

ΓΜ

Tt Ν ν ο «Ó in

T - ; ( T \ ' ^ ; o n ^ ; i n v D m T t q

νο

νο σι

•

ο ο σι m ΓΜ

*

I U

ΓΜ 0 1 Τ- ι η Ο νο οο Ο i s ΓΜ

^ δo

(Ν 01

ΓΠ

νο O «sP -> p"s QV "sP O ^ S_0 QN. ν,ρ ν,Ρ

t - E E

υ

ΓΜ τ *

τ-

Tt (Ν

δ υ m (Ν Ρ ^ so £ ¿ fe S

o " i U


- o •

ro IN σ \ νο τ - aco

ë

«

αι C c· m υ 3 ( J bjo ^ (J _ E C Κ oj ν -F, -F,

= 5 «r — ¡ o 3» «S £ ì tα. «υ u£ V -gá C QIJ UCr i— Bài S u VOI Ic IO Q. 3'S > lei> O US xO δ Is § c o P if 3I 2 s ci s i 2 öS ι/ι _i '•Zi+->

Energy Use Patterns in German Industry: Evidence from Plant-level Data • 409

Table A5 Percentiles and relative variation of selected energy variables (median of all sectors,

2006) o in α.

^

Energy intensity (kWh/1000 EUR) Total energy use (kWh) Carbon intensity (g/1000 EUR) Electricity share in energy mix Natural gas share in energy mix

o το.

m •2.

o m o.

in rv α.

o σι

38.6

81.8

162.8

301.6

546.2

o το. ό

σ\

CL·

ο m α. ίη

3 ί: 0) c (υ

& c 0) c c o

3 Β3

υ

ο

Ο.

δ

0.92 0.23 0.20 0.21 -0.34

0.06 0.04 0.04 -0.30

0.96 0.94 -0.11

0.98 -0.12

-0.13

0.59 0.79 0.32 0.34 0.24

0.35 0.52 0.61 0.07

0.55 0.51 0.17

0.97 0.09

0.12

~ΰ ο I-

·*->

•e (β

Ο Ιfcrt Λ Ο

ο k. η

α υ> ο S^

Ο

410 · Sebastian Petrick, Katrin Rehdanz, and Ulrich J. Wagner

Table A7 Users and aggregate use of different fuels in 2006 Ν 44 26 18 2

Electricity Natural gas (incl. LPC) Petroleum products Heat Renewables Coal Other fuels

Aggregate use (TWh)

080 728 786 648 973 378 141

238.0 313.4 166.2 39.0 26.9 154.3 13.2

Own calculations.

400

300

200

100

Metal Recycling (371)

ΛI u

Steam Generatora (283)

Transmutara and Telephones (322)

wem p10-p90 m r m p25-p75

Printing and Reproduction Industrial Process Control (220) Equipment (333)

—

p50

Own calculations. For more information regarding gross output, workforce and payroll see Table A4 in the appendix. For more information on energy intensity for the median of all sectors see Table A5 in the appendix.

Figure A1 The five sectors with the lowest median energy intensity in 2006 (in kWh/1000 EUR) 1000% 900% 800% 700%

600%

500% 400% 300% 200% 100% 0% energy intensity total energy use carbon intensity electricity share natural gas in energy mix share in energy mix • (p75-p25)/p50 • (p90-p10)/p50 Own calculations.

Figure A2 Variation relative to the median of selected energy variables (median of all sectors, 2006 data)

E n e r g y U s e P a t t e r n s in G e r m a n I n d u s t r y : E v i d e n c e f r o m P l a n t - l e v e l D a t a · 4 1 1

3 000.0

2 500.0

2 000.0

1 500.0

1 000.0

500.0

Publishing (221)

Steam Generators (283)

Leather Clothes and Fur (180)

w a s · p10-p90

Medical Equipment (331)

M » p75-p25

Industrial Process Control Equipment (333)

p50

O w n calculations. Note that for some sectors the values for p10 and p90 are not available to protect confidentiality of data. For more information regarding gross output, workforce and payroll see Table A4 in the appendix. For more information on energy use for the median of all sectors see Table A5 in the appendix. Figure A 3 T h e f i v e sectors w i t h t h e l o w e s t m e d i a n e n e r g y use in 2 0 0 6 (in

MWh)

II . 1 1 . IP

70 80

Office Machinery, Computers (300)

Transmitters and Telephones (322)

Steam Generatore (283)

SBMi p10-p90

ä®eip25-p75

Publishing (221)

—

Industrial Process Control Equipment (333)

p50

O w n calculations. For more information regarding gross output, workforce and payroll see Table A4 in the appendix. For more information on carbon intensity for the median of all sectors see Table A5 in the appendix. Figure A 4 T h e f i v e sectors w i t h t h e l o w e s t m e d i a n c a r b o n i n t e n s i t y in 2 0 0 6 (in k g / 1 0 0 0 EUR)

412 · Sebastian Petrick, Katrin Rehdanz, and Ulrich J. Wagner

References A G E B (Arbeitsgemeinschaft Energiebilanzen) ( 2 0 0 9 ) , Energiebilanzen der Bundesrepublik Deutschland, http://www.ag-energiebilanzen.de/viewpage.phpiidpages63, accessed September 4 , 2 0 1 0 . Buttermann, H.-G., B. Hillebrand ( 2 0 0 0 ) , Die Klimaschutzerklärung der deutschen Industrie vom März 1 9 9 6 - eine abschließende Bilanz. Essen. B D E W (Bundesverband der Energie- und Wasserwirtschaft) ( 2 0 0 9 ) , Energiemarkt Deutschland. Berlin. Eurostat ( 2 0 1 0 ) , Statistical Classification of Economic Activities in the European Community, Rev. 1.1 ( 2 0 0 2 ) ( N A C E Rev. 1.1). http://ec.europa.eu/eurostat/ramon/nomenclatures/ind e x . c f m ? T a r g e t U r l = L S T _ C L S _ D L D & S t r N o m = N A C E _ l _ l , accessed September 2 8 , 2 0 1 0 . F D Z (Forschungsdatenzentrum der Statistischen Ämter der Länder) ( 2 0 1 0 a ) , AFiD-Panel Industriebetriebe der Jahre 1 9 9 5 bis 2 0 0 6 . Own calculations. F D Z (Forschungsdatenzentrum der Statistischen Ämter der Länder) ( 2 0 1 0 b ) , AFiD-Modul Energieverwendung der Jahre 2 0 0 3 bis 2 0 0 6 . Own calculations. F D Z (Forschungsdatenzentrum der Statistischen Ämter der Länder) ( 2 0 1 0 c ) , Metadaten AFiDPanel Industriebetriebe, http://www.forschungsdatenzentrum.de/bestand/afid-panel_industriebetriebe/fdz_afid-panel_industriebetriebe_metadaten.pdf, accessed September 8, 2 0 1 0 . F D Z (Forschungsdatenzentrum der Statistischen Ämter der Länder) (2010d), AFiD-Panel Industrieunternehmen. http://www.forschungsdatenzentrum.de/bestand/afid-panel_industrieunternehmen/, accessed November 10, 2 0 1 0 . F D Z (Forschungsdatenzentrum der Statistischen Ämter der Länder) ( 2 0 1 0 e ) , AFiD-Modul Energieverwendung - Metadaten für die On-Site-Nutzung. http://www.forschungsdatenzentrum.de/bestand/afid-modul_energieverwendung/fdz_afid-modul_energieverwendung_metadaten.pdf, accessed September 8, 2 0 1 0 . F D Z (Forschungsdatenzentrum der Statistischen Ämter der Länder) ( 2 0 1 0 f ) , Energy Fuels, Calorific Values and Conversion Factors for Energy Use Statistics. Personal communication with Diane Zabel, M a y 5, 2 0 1 0 . Fritsche, U. R . , L. Rausch ( 2 0 0 7 ) , Bestimmung spezifischer Treibhausgas-Emissionsfaktoren für Fernwärme. Dessau-Roßlau. Frohn, J . , P. Chen, Β. Hillebrand, W. Lemke, C. Lutz, B. Meyer, M . Pullen ( 2 0 0 3 ) , Wirkungen umweltpolitischer Maßnahmen. Heidelberg. Frohn, J . , U. Leichtmann, R. Kräussl ( 1 9 9 8 ) , Fünf makroökonometrische Modelle zur Erfassung der Wirkungen umweltpolitischer Maßnahmen - eine vergleichende Betrachtung. Wiesbaden. Fryges, H., J . Wagner ( 2 0 1 0 ) , Exports and Profitability: First Evidence for German Manufacturing Firms. The World Economy 3 2 : 3 9 9 - 4 2 3 . Girma, S., H. Görg, J . Wagner ( 2 0 0 9 ) , Subsidies and Exports in Germany - First Evidence from Enterprise Panel Data. Applied Economics Quarterly 5 5 ( 3 ) : 1 7 9 - 1 9 5 . Görzig, B., M . Gornig, A. Werwatz ( 2 0 0 5 ) , Produktdiversifizierung: Konvergenz zwischen ostund westdeutschen Unternehmen. Jahrbücher für Nationalökonomie und Statistik 2 2 7 ( 2 ) : 168-186. Görzig, B., R . Pohl ( 2 0 0 7 ) , Diversifizierungsstrategien deutscher Unternehmen - Auswertung eines Producer-Product-Panels der amtlichen Statistik. Wirtschafts- und Sozialstatistisches Archiv 1(3-4): 1 7 9 - 1 9 1 . GWS (Gesellschaft für Wirtschaftliche Strukturforschung) ( 2 0 1 0 ) , ΡΑΝΤΑ R H E I . http:// www.gws-os.com/de/index.php?option=com_content&task=view&id=52&Itemid=38, accessed September 6, 2 0 1 0 . Hillebrand, B. ( 1 9 9 6 ) , Das Modellinstrumentarium des RWI - Konzeption, Erklärungsansätze und Anwendungsmöglichkeiten. Pp. 1 9 3 - 2 1 8 in: S. M o l t , U. Fahl, Energiemodelle in der Bundesrepublik Deutschland - Stand der Entwicklung. Jülich. Hillebrand, B., H . - G . Buttermann ( 2 0 0 3 ) , Prognose der Klimagasemissionen - Unsicherheiten und ihre Konsequenzen. Zeitschrift für Energiewirtschaft 2 7 ( 3 ) : 1 8 1 - 1 9 8 . Hillebrand, Β., K. Lobbe ( 2 0 0 8 ) , Die gesamtwirtschaftliche und sektorale Entwicklung bis 2 0 2 0 . Düsseldorf.

Energy Use Patterns in German Industry: Evidence from Plant-level Data · 4 1 3

IEA (International Energy Agency) ( 2 0 0 8 ) , World energy balances. Paris. IEA (International Energy Agency) ( 2 0 0 9 ) , C 0 2 Emissions from Fuel Combustion. Paris. I S G E P (International Study Group on Exports and Productivity) ( 2 0 0 8 ) , Understanding CrossCountry Differences in Exporter Premia: Comparable Evidence for 1 4 Countries. Review of World Economics 1 4 4 ( 4 ) : 5 9 6 - 6 3 5 . Konoid, M . ( 2 0 0 7 ) , N e w Possibilities for E c o n o m i c Research through Integration of Establishment-level Panel Data of German Official Statistics. Schmollers J a h r b u c h 1 2 7 : 3 2 1 - 3 3 4 . Lutz, C. ( 2 0 0 0 ) , N O x Emissions and the Use of Advanced Pollution Abatement Techniques in West Germany. E c o n o m i c Systems Research 1 2 ( 3 ) : 3 0 5 - 3 1 8 . Lutz, C., B. Meyer, C. Nathani, J . Schleich ( 2 0 0 5 ) , Endogenous technological change and emissions: the case of the German steel industry. Energy Policy 3 3 : 1 1 4 3 - 1 1 5 4 . M a l c h i n , Α., R . Voshage ( 2 0 0 9 ) , Official Firm D a t a for Germany. Schmollers J a h r b u c h 1 2 9 ( 3 ) : 501-513. Mangelsdorf, S. ( 2 0 0 7 ) , Ostdeutsche Industrie holt auf! Entwicklung des Verarbeitenden Gewerbes in Berlin und Brandenburg. Wirtschafts- und Sozialstatistisches Archiv 1 ( 3 - 4 ) : 205-215. Meyer, B. ( 2 0 0 1 ) , C 0 2 - t a x e s , growth, labor market effects and structural change - a n empirical analysis. Pp. 3 3 1 - 3 5 2 in: P . J . J . Weifens (ed.), Internationalization o f the E c o n o m y and Environmental Policy Options. Springer. Meyer, B., G. Ewerhart ( 1 9 9 8 ) , Multisectoral policy modelling for environmental analysis. Pp. 3 9 5 - 4 0 6 in: K. Uno, P. Bartelmus (eds.), Environmental Accounting in Theory and Practice. Kluwer. Schleich, J . , C . N a t h a n i , K. Ostertag, M . Schön, R . Walz, Β. Meyer, C. Lutz, M . Distelkamp, F. H o h m a n n , M . - I . Wolter ( 2 0 0 2 ) , Innovationen und Luftschadstoffemissionen - Eine gesamtwirtschaftliche Abschätzung des Einflusses unterschiedlicher Rahmenbedingungen bei expliziter Modellierung der Technologiewahl im Industriesektor. Karlsruhe/Osnabrück. Statistik der Kohlenwirtschaft e.V. ( 2 0 0 9 ) , Zahlen zur Kohlenwirtschaft. Essen/Köln. Statistik der Kohlenwirtschaft e.V. ( 2 0 1 0 a ) , Kohleneinfuhr nach Lieferländern, http:// www.kohlenstatistik.de/home.htm, accessed April 15, 2 0 1 0 . Statistik der Kohlenwirtschaft e.V. ( 2 0 1 0 b ) , Steinkohlenförderung in den Grenzen der Bundesrepublik Deutschland, http://www.kohlenstatistik.de/home.htm, accessed April 1 5 , 2 0 1 0 . Statistik der Kohlenwirtschaft e.V. ( 2 0 1 0 c ) , Braunkohle im Uberblick. http://www.kohlenstatistik.de/home.htm, accessed April 1 5 , 2 0 1 0 . Statistisches Bundesamt ( 2 0 0 2 ) , Beschäftigung, Umsatz und Energieversorgung der Betriebe des Verarbeitenden Gewerbes sowie des Bergbaus und der Gewinnung von Steinen und Erden 2 0 0 2 . Fachserie 4 , Reihe 4 . 1 . 1 . Wiesbaden. Statistisches Bundesamt ( 2 0 0 3 ) , German Classification of E c o n o m i c Activities, Edition 2 0 0 3 ( W Z 2 0 0 3 ) . Wiesbaden. Statistisches Bundesamt ( 2 0 0 9 ) , Umweltnutzung und Wirtschaft. Tabellen zu den Umweltökonomischen Gesamtrechnungen. Chapter 3 . 2 . Wiesbaden. Statistisches Bundesamt ( 2 0 1 0 a ) , Energieverwendung in der Industrie, Beschäftigte und Umsatz nach ausgewählten Wirtschaftszweigen. http://www.destatis.de/jetspeed/portal/cms/Sites/destatis/Internet/DE/Navigation/Statistiken/Energie/Tabellen.psml, accessed September 2 1 , 2010. Statistisches Bundesamt ( 2 0 1 0 b ) , Beschäftigung und Umsatz der Betriebe des Verarbeitenden Gewerbes sowie des Bergbaus und der Gewinnung von Steinen und Erden. Fachserie 4 , Reihe 4 . 1 . 1 . Wiesbaden. Statistisches Bundesamt ( 2 0 1 0 c ) , Erzeugerpreisindizes gewerblicher Produkte: Deutschland, J a h r e , Güterverzeichnis ( G P 2 0 0 9 2-/3-/4-/5-/6-/9-Steller/Sonderpositionen). Genesis database, https://www-genesis.destatis.de/genesis/online, accessed Novermber 1 2 , 2 0 1 0 . Statistisches Bundesamt ( 2 0 1 Od), Produzierendes Gewerbe. Indizes der Produktion und der Arbeitsproduktivität im Produzierenden Gewerbe. Fachserie 4 Reihe 2 . 1 . Wiesbaden. Statistisches Bundesamt ( 2 0 1 0 e ) , Beschäftigte, Umsatz, Produktionswert und Wertschöpfung der Unternehmen im Verarbeitenden Gewerbe: Deutschland, J a h r e , Wirtschaftszweige (2-1

414 · Sebastian Petrick, Katrin Rehdanz, and Ulrich J. Wagner

3-/4-Steller). Genesis database, https://www-genesis.destatis.de/genesis/online, accessed N o vermber 2, 2010. Umweltbundesamt (2010), Tables with the derived C 0 2 Emission Factors for the German Atmospheric Emission Reporting 1990 - 2008 (Version: EU-Submission 15.01.2010). Dessau. VIK (Verein der Industriellen Kraftwirtschaft) (2010), Statistik der Energiewirtschaft. Essen. Wagner, J. (2000), Firm Panel Data f r o m German Official Statistics. Schmollers Jahrbuch 120: 143-150. Wagner, J. (2006), Politikrelevante Folgerungen aus Analysen mit Firmendaten der Amtlichen Statistik. Schmollers Jahrbuch 126: 359-374. Wagner, J. (2007a), Exports and Productivity in Germany. Applied Economics Quarterly 55(4): 353-373. Wagner, J. (2007b), J o b m o t o r Mittelstand? Arbeitsplatzdynamik und Betriebsgröße in der westdeutschen Industrie. University of Lüneburg Working Paper Series in Economics 47. Wagner, J. (2009a), Offshoring and firm performance: Self-selection, effects on performance, or both? University of Lüneburg Working Paper Series in Economics 153. Wagner, J. (2009b), Produktdifferenzierung in deutschen Industrieunternehmen 1995 - 2004: Ausmaß und Bestimmungsgründe. Jahrbücher für N a t i o n a l ö k o n o m i e und Statistik 229(5): 615-642. Wagner, J. (2010a), The Research Potential of N e w Types of Enterprise Data based on Surveys f r o m Official Statistics in Germany. Schmollers Jahrbuch 130: 133-142. Wagner, J. (2010b), Entry, Exit and Productivity: Empirical Results for German Manufacturing Industries. German Economic Review 11(1): 78-85. Wirtschaftsvereinigung Stahl (2009), Statistisches Jahrbuch der Stahlindustrie 2009/2010. Düsseldorf. World Bank 2009, World Development Indicators. Washington. Sebastian Petrick, Kiel Institute for the World Economy, Hindenburgufer 66, 2 4 1 0 5 Kiel, Germany. [email protected] Katrin Rehdanz, Christian-Albrechts University of Kiel and Kiel Institute for the World Economy, Hindenburgufer 66, 2 4 1 0 5 Kiel, Germany. [email protected] Ulrich J. Wagner, Universidad Carlos III de M a d r i d , Calle M a d r i d 126, 2 8 9 0 3 Getafe (Madrid), Spain. [email protected]

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2011) Bd. (Vol.) 231/3

Global Warming Induced Water-Cycle Changes and Industrial Production - A Scenario Analysis for the Upper Danube River Basin By Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer, Munich* JEL D24; R30; Q01; Q25; Q52; Q53 Environmental decision support system; climate change; water-cycle; river basin management.

Summary Using the environmental decision support system DANUBIA, we analyze the effects of climate change on industry and compare the effectiveness of different adaptation strategies. The observed area covers Germany and Austria up to 2025. Since the main effects of climate change in this region are expected to be caused by changes in the water-cycle, we place a special focus on the exemplary region of the upper Danube catchment area. Industry is the main regional user of water resources. Water is an essential production factor and is used in almost every production process of a manufactured good. We apply estimates of regional production functions, based on AFiD-panel micro-data for Germany, to calibrate regional industrial production and water usage within DANUBIA. Thus, we are able to simulate region-specific effects of climate change and the impact of social scenarios using an innovative model of the reciprocal influences of a huge network of interdisciplinary research areas. Simulation results show wide regional differences in production site reactions as well as between differing scenarios. Comparing scenarios of moderate and serious climate change, we are able to illustrate the severe environmental effects in some regions and to determine considerable economic effects on regional economic growth.

1

Introduction

Is the behaviour of h u m a n society sustainable? Posing this question inevitably relates to the urgent problems of climate change. Climate change affects all three pillars of sustainability: ecological, social and economic stability. It does so in m a n y ways of which the water-cycle is the most crucial. Besides being exceptionally sensitive to climate change, w a t e r is the one scarce good t h a t is w i t h o u t c o m p a r i s o n in its necessity for the functioning of nature and society. Less t h a n 3 % of the earth's w a t e r is fresh-water, of which only V3 is accessible for h u m a n needs at justifiable costs. Climate change will dramatically worsen this situation and as a consequence the natural a n d artificial water-cycles have been gaining increasing attention. O n e of the ten Millennium Development Goals agreed on by the United N a t i o n s a n d world leaders is to cut in half the p r o p o r t i o n of people w i t h o u t sustainable access to safe drinking water by 2 0 1 5 . To this aim, in December 2 0 0 3 the United N a t i o n s General Assembly proclaimed the International *• We want to especially thank Nadine Bartke from the FDZ (Forschungsdatenzentrum) for the tireless assistance in preparing the data.

4 1 6 · Christoph Jeßberger, Maximilian Sindram, and M a r k u s Zimmer

Decade for Action „Water for Life 2 0 0 5 - 2 0 1 5 " (United Nations 2004). In this article we analyze the social, economic and environmental effects of climate change with respect to sustainable access and use of water by applying the environmental decision support system DANUBIA (EDSS DANUBIA) which is part of the GLOWA-Danube project. Siegfried Demuth states in the GLOWA report of the UNESCO International Hydrological Programme and the WMO Hydrological and Water Resources Programme: „ Water management affects our environment, society and culture. Finding solutions to mitigate impacts and adopt to different geographical conditions and climate regions requires an approach that unites sound and unbiased science with social and policy considerations". (IHP HWRP Report on the GLOWA project 2008: 5). In the following analysis we highlight a small portion of the results of DANUBIA. These can help potential stakeholders gain an idea of the capability of the applied web-based environmental decision support system. We investigate the effects of global warming on medium-sized mountainous watersheds as well as on developed societies under temperate climate conditions. By fundamental modelling of the Upper Danube catchment within a simulation region covering Germany and Austria, we aim to answer the following questions: • Climate change is one of the main global issues for sustainable development, but is it also an issue for German and Austrian regions? • What are the regional causes for water scarcity: climate or society? • How can society compensate for climate change and how can differing policy scenarios be evaluated with respect to sustainability? • What are the small-scale effects and regional differences? E. g. how do cities perform in comparison to rural areas? • What are the effects of climate change on economic development given the close interconnection of climate change and the water-cycles? 1.1 Environmental decision support systems focused on climate change and the water-cycle

DANUBIA was the first decision support system of its kind to feature a dynamic, simultaneously interconnected simulation model network. Furthermore, the underlying DeepActor Framework (see section 1.3) allows an analysis of completely new research issues by applying standardized interfaces to systematically connect scientific models of natural, environmental and socio-economic fields of research. There are also an increasing number of similar projects which will be briefly introduced in this section: LANDSCAPE LOGIC (LL) is an environmental decision support system that is currently being developed in Australia. It consists of a research hub with 6 regional organisations, 5 research institutions and is supported by several state agencies in Tasmania and Victoria. 1 It focuses on two current gaps in knowledge. It seeks, firstly, to improve the inadequate understanding of how to organize knowledge and assumptions about dynamic interactions between activities in land management and environmental outcomes. Secondly, LL aims to expand this understanding through studying the historical development of water quality and native vegetation conditions. MODSIM is an EDSS that was developed at Colorado State University. Its aim was to create a generic river basin management decision support system and it is based on a system of simulations of river network flows and reservoir operations (Assaf et al. 1

For further details see http://www.landscapelogic.org.au7

Global Warming Induced Water-Cycle Changes and Industrial Production · 417

2 0 0 8 ) . It has been linked with stream-aquifer models for the analysis of the combined use of groundwater and surface water resources. To control the effectiveness of pollution control strategies, M O D S I M has also been used in water quality simulation models. An important feature of M O D S I M is that it provides an interface to standard geographic information systems (Labadie 2 0 0 5 ) . The web-based decision support system BodenseeOnline (BO) has been developed for Lake Constance to support decision-makers in issues such as water protection and hazardous incidents. Since BO has a strong focus on monitoring the current environmental conditions, it uses measurement stations to deliver current data about wind, water quality, temperature profiles and other important parameters. All this information is processed in a nexus of physical, biological and chemical models. Finally the model output is visualized and published for the stakeholders via the internet (Lang et al. 2 0 1 0 ) . GLOWA-Elbe is another GLOWA project that is very similar to DANUBIA but focuses on the Elbe catchment area. With the aid of the EDSS Elbe-Expert-Toolbox it is possible to analyze and describe scenarios about future changes in water quality and quantity and to assess the consequences of climate change as well as potential adaptation strategies. Contrary to DANUBIA this toolbox only allows limited feedback between the model components during the simulation process. 2 A survey on further water-cycle and climate change related EDSSs can be found in Assaf et al. (2008), but none of the reviewed projects is comparable in depth and scale to DANUBIA. 1.2 GLOWA & DANUBIA - behind the scenes The GLOWA-project originated from an initiative of the German Federal Ministry of Education and Research. Its goal is to study the effects of global change on the water-cycle and to develop an environmental decision support system for the sustainable management of water resources. In this paper we can only present a small selection of results. Most of the background, theoretical foundations, model mechanics and the model calibration have been discussed in detail in corresponding publications. At present there are more than 7 0 0 publications directly related to the GLOWA project, the majority in academic journals. More than 3 0 0 of these publications are connected to the EDSS DANUBIA on which this paper is based. There is a separate online search engine available that enables searches within the GLOWA publications for works relevant to the topics and foundations of this paper. 3 1.3 DANUBIA DANUBIA uses regional climate models to project climate change. Physical and physiological components describe the natural processes (hydrology, hydrogeology, plant physiology, and glaciology). For socio-economic simulations DANUBIA makes use of actor models 4 (farming, economy, water supply, households and tourism) to model decisions 2 3

4

For further details see http://www.glowa-elbe.de/ The online search engine for G L O W A publications is available under: http://www.glowa.org/eng/ literaturliste_eng/literaturliste_eng_suchen.php An actor model simultaneously simulates the actions and activities of many independent actors w h o are able to interact with and perceive each other and their environment. Given their perception of the environment, the actors decide to execute a specific set of actions out of the possible set of actions. This decision will typically maximize the subjective utility of the actor.

418 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

Remote sensing: Integrative environmental monitoring, validation Tourism: Water use and water conflicts

Glaclology: Snow and ice modelling

Meteorology: Mesoscale modelling, atmosphere Human dimensions: Water use, conflicts, market, regulation Hydrology: Evapotranspiration, lateral flows, percolation, runoff

Vegetation: Modelling of the natural vegetation Land use: Farming and forestry

Water management: Surface water, water quality, reservoir management, groundwater, water supply

Source: GLOWA Danube

Figure 1 Interactions inside DANUBIA based on the social structure, the respective general conditions, and individual interests. It enables in particular the simulation of different climate change scenarios and socioeconomic scenarios in conjunction with diverse social and political action and reaction patterns. 5 The objective of this work is a well-founded simulation of the industrial production under climate change conditions and the development and evaluation of climatic and social scenarios of interest with a special focus on the water-cycle. Figure 1 shows a schematic representation of how a real environment is covered by the different sub-models within the simulated environment in DANUBIA. A more detailed description of the different sub-models and of their interaction can be found in the appendix. 1.4 DANUBIA base data and model calibration For the calibration of the industrial model within DANUBIA we use micro data provided by the „Forschungsdatenzentren der Statistischen Ämter des Bundes und der Länder" (Data Research Centres of the Federal and Land Statistical Offices) and estimates of regional production functions which are also based on these micro data. These data are available for research purposes via the AFiD-Panel which can be extended by several modules, e.g. covering environmental statistics. The data contains the economic as 5

The theoretical foundations for the EDSS-DANUBIA as well as the simulation model itself were developed and implemented within the GLOWA Danube project (www.glowa-danube.de) between 2001 and 2010. All components of DANUBIA can run simultaneously on a cost-effective LINUXCluster.

Global Warming Induced Water-Cycle Changes and Industrial Production · 419

well as the environmental characteristics of the companies on the level of the individual firm, of which the industrial water usage is of special interest for our analysis.6 We examine the statistics for the years 1 9 9 8 , 2 0 0 1 and 2004. To generate plausible, small-scale simulation results in DANUBIA, it is indispensable that data is available in a similar spatial resolution. Due to data privacy protection, it is not possible to calibrate the model on the scale of one square kilometre. Therefore, we estimated representative production technologies on the scale of the NUTS 3 regions (the NUTS 3 district classification of the European Union is in size equal to a German Stadtkreis or Landkreis). Since the representative production sites are simulated on the scale of one square kilometre and also the conditions and characteristics that are exchanged with other sub-models vary on the scale of one square kilometre, we observe large differences in the behaviour of producers within the same region. 7 For the simulation results shown in this paper we employ the most restrictive specification estimated in Jessberger and Zimmer (2010), which represents a Cobb-Douglas technology.8 We explicitly model the industrial production and water usage, while we use the output of the existing DANUBIA sub-models as input to our model and vice-versa. Thus, we are able to include the interactions and feedback mechanisms of industrial producers and, for example, social conditions like the labour market and migration or natural conditions like aquifers or river networks. While DANUBIA in general aims at assessing a broad portfolio of problems it was especially tailored around climate change issues. In addition to similar environmental decision support systems, it also accounts for the interaction of the different natural and man-made water-cycles and the atmospheric conditions. The water-cycle plays a major role in climate change, especially if the effects are analyzed on a small-scale. Figure 2 shows the Upper Danube River Basin which is the focus of the investigation of water-cycle effects in this analysis.9 Within this core region the feedback mechanisms of human activities with the natural environment are modelled and calibrated on the micro scale. 10 The entities on which this paper focuses are the industrial producers. These, together with agriculture and tourism, are expected to be most exposed to climate change. 6

This triennial statistic includes all industrial production sites and mining sites that have an annual water demand of at least 1 0 , 0 0 0 m 3 . Therefore it should be noted that the conclusions made here are only valid for the examined sub-group.

7

The full set of characteristics and conditions is only available within the Upper-Danube Catchment of the simulation area since that is the only area that is c o m m o n to all sub-models. For the remaining areas our model employs average characteristics. The regional differences are captured by region dummies. For the Austrian regions of the simulation area we use also data from Statistik Austria but since we did not have access to comparable Austrian micro-data we base our production technologies of the Austrian representative regional industrial producers on estimates of similar Bavarian entities. Thus we impose that Austrian producers can be approximated by Bavarian producers with similar characteristics.

8

9

In the philosophy of EDSSs it is c o m m o n to observe a natural resource within its natural boundaries rather than its administrative borders, examining in particular natural phenomena that occur. Especially when observing the water cycle, it is obvious that the watersheds delimit the dispersion of pollution at least for surface waters. Computer-based EDSSs like D A N U B I A take this into account and typically generate their results to be consistent with the natural borders.

10

In model terms this means on the area of one square kilometre. The remaining regions of the simulation area of this analysis - which covers the remaining parts of Germany and Austria - are mostly based on averages of micro-effects of the core region. Even though the simulations in the remaining regions of the observed area also operate on the area of one square kilometre, the results are much less regionally differentiated for these regions. This further highlights the necessity of small-scale data availability (e.g. the AFiD micro data).

420 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

Source: GLOWA-Danube

Figure 2 The Upper Danube River Basin 1.5 Modelling industrial water use Water is an essential production factor and is used in almost every production process of a manufactured good. Electricity generation, the mining industry and the industries that produce paper products, chemicals and metals require especially large quantities of water. Water is used for cleaning, diluting, transporting products, cooling, heating, generating steam, sanitation and, of course, as a constituent in the final product. In Germany 5.1 billion cubic meters of water were extracted by the public water supply in 2007. Of this only 0.9 billion cubic meters of water were supplied to the industrial sector. This has to be seen in relation to the 27.1 billion cubic meters of water that industry extracts on its own. In the observed area agriculture plays only a very minor role in the water usage, in contrast to regions that are less developed or situated in a warmer climate zone. Agricultural production, forestry and the fishery sector only used a comparably small total amount of 0.2 billion cubic meters of water in 2007 (Federal Statistical Office Germany 2006). For the self-supplied industrial users in Germany, water is easily available and cheap to extract. Even considering possible discharge costs, water is still a relatively cheap factor of production. Nevertheless, the extraction of water is restricted by contingents. These include the sophisticated extraction permits that are enforced by local environmental authorities. Water is often circulated within the production process or used multiple times in consecutive processes. While multiple employment might follow

Global Warming Induced Water-Cycle Changes and Industrial Production · 421 from economic considerations, cycle use is a reaction to regulatory constraints (Egerer/ Z i m m e r 2 0 0 6 a ; Egerer 2 0 0 5 ) . In industrial production processes, water is typically not consumed in the traditional sense. Instead, it is used for production purposes and afterw a r d s returned to the water-cycle. The equivalent to „ c o n s u m p t i o n " is the reduction of the usable amount of the resource for other natural or artificial utilisations. This might for example result from the reduction of the water quality below a critical threshold. Factors that are closer to the traditional interpretation of consumption, for example, if the water resources are evaporated in a cooling process, might also result in an upstream/downstream riparian conflict. In the simulation model, the industry sub-model mimics the decisions of the relevant industrial production sites from an economist's perspective. The decision process is focused on the questions of the optimal production output and on how to produce this output with minimal costs given the technical, regulatory and resource constraints. In accord with the dominant research question in this work, we highlight the use of water resources in the production process. We assume that a representative firm behaves rationally given that its information is limited by its perceptive abilities and its imperfect expectations. 1 1 This means, for example, that rather than incorporating the signal about the sustainability of water-usage directly into the decision process, it is used to determine the a m o u n t of regulation imposed on the production site by the local environmental agencies. 1 2 These also have the means to regulate water usage by extraction or effluent charges or by limiting the a m o u n t of water extraction.

1.6 Inside the industrial producer Depending on the available resources, it is plausible to imagine different approaches for modelling the industrial producer. Typically, the final implementation is based on anecdotal evidence, theoretical considerations or econometric estimates. To construct a representative industrial producer we explored all three of these options. In the theoretical approach, the production function of the firm is modelled and the derived optimal factor demands are used in the simulation. To simulate the production process it is helpful to gather as much information about it as possible. To achieve this we conducted a questionnaire c a m p a i g n , carried out field and telephone interviews and visited actual pro11

12

The conditions influencing the firm's decisions can be categorized into three groups: Factors which the firm perceives as exogenous and thus as not influenceable by its actions, factors that it perceives as being influenced by its decisions, and factors which it can directly determine by choice. In our modelling approach examples for exogenous factors are the technological progress and the condition of the water resources. Among the factors the industrial agent perceives as influenceable are the water-related expenditures (including eventual charges). These are indirectly determined by factors of his direct choice, namely his investments in technologies that reduce the pollution discharge or increase the utilisation factor of the water in the production process. Other important factors of direct choice are the labour employed and the production output. This approach has been identified as preferential since the industrial producers themselves cannot observe the sustainability of their resource usage. While counterintuitive at first glance, this is the consequence of simple information asymmetry. It is indeed true that in reality the production site cannot observe the consequences of its water consumption and that the monitoring of the environmental effects is done by the local environmental authorities. This might not be the case for regions that are less restrictive than Germany and Austria concerning the regulation of environmental pollution. In the observed area production facilities are typically restricted before the environmental effects are obvious to the producer.

422 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

duction sites. 13 The final step was to examine the available data and to draw conclusions on the production technology. As a result we designed the industrial production sites as profit-maximizing entities in a competitive market environment. It was an essential requirement in the model construction to consider the effects of climate change and conservation of the environment. As discussed earlier it is reasonable to model the resulting consequences for the firm as regulatory constraints. Due to the integrative nature of DANUBIA these characteristics influence the macroeconomic and discipline-specific models, which will in turn create a feedback on the industrial producer. Model results are calculated on the scale of a single representative industrial production site on each industrialized square kilometre within the observed area. 14 The characteristics of the representative production site are determined by the local natural environment, the economic conditions and by the econometric estimates of the production technology. Companies minimize the production costs for a given production output. With p„ being the price vector of the production factors employed, the total expenditures E„ of a production site aggregates the costs for the variable production factors X„. min E„ = p'„Xn

s.t.

Yn=f[Xn,T]

The output Y„ of the industrial facility is a function of the vectors of the production factors employed and of the technology used at time T. This black-box, converting multiple inputs into the production output, mirrors the technical production process in a production site. For the simulations featured in this paper we employ estimates of regional Cobb-Douglas production technologies. 15 Details on the estimation procedure and results can be found in Jeßberger and Zimmer (2010). 1 6

13

14

15

16

This participatory process involving the industrial water users was described in depth in Egerer and Zimmer (2006b). For the area of Germany and Austria this corresponds to a total of 16,800 representative agents as identified by analyzing the remote sensing data. Such estimates are based on the known data of factor usage, factor prices and output. A wide range of possible production technologies can be used for estimation. A very good overview of the currently used approaches can be found in Chung (1994). In this work we focus on the Translog production framework. This functional form is especially appealing since it covers many commonly used production functions as special cases, including the Cobb-Douglas specification used in the simulations in this paper. A further reason is that the properties of this function are well known and especially the price and the cross-price elasticities - in which we are ultimately interested for the simulation - are easy to derive from the estimated coefficients. The production technology is assumed to be homothetic and separable from unobserved production factors. Since we have to up- or downscale the technology in various steps of the later simulation it seems sensible to assume constant returns to scale. Further discussion about the theoretical formulation of industrial water usage can be found in Renzetti and Elgar (2002), Gispert (2004) and Dupont and Renzetti (2001). They use the estimation of the primary form of the production function in order to determine the shadow value of industrial water use, the price elasticities of the production factors and the regionspecific dummies which characterize the local production technologies on the scale of the NUTS 3 regions. A comprehensive description of the estimation procedure can be found in Kim (1992) as well as in Eckey et al. (2005) who also describe the calculation procedure for the elasticities of Translog production-functions in detail. Further works focus on assessing the value of water for industrial production include Reynaud (2003), Griffin (2006), Dachraoui and Harcharoui (2004) and Dupont and Renzetti (2003).

Global Warming Induced Water-Cycle Changes and Industrial Production · 423

2

Economic and social scenarios in the context of climate change

Scenario Analysis is a wide field that has been intensively analyzed in many works of different disciplines. A good overview of the theory, guidelines and literature can be found in Alcamo et al. (2008). This book discusses in depth the generation of scenarios for EDSS and also provides many useful examples. The scenarios presented in the following sections have been designed in accordance with the requirements of such a scientifically founded scenario generation process. Air temperature in Central Europe has already increased by up to 1.5 °C compared to the pre-industrial era, and up to 2025 another temperature rise of 1 - 1 . 5 ° C is expected (IPCC, 2007). In terms of DANUBIA this means that droughts will become more common in the summer and water levels will fall strongly. What consequences this implies for the society, in particular for industry, is this paper's focus. Regarding the socio-economic perspective, this paper focuses on the comparison of two opposing social scenarios: A performance scenario representing globalization and a society focused on economic growth and a common public interest scenario with growing environmentalism in the society. These two scenarios span a plausible corridor of adaptation strategies that can be compared with a baseline scenario, which could also been seen as the business-as-usual scenario. Interactions and feedback mechanisms of industrial producers are matched with the storyline of these scenarios. The following sections describe these settings. 2.1 The economy in the baseline scenario Table 1 shows five adjustable scenario parameters that are on their default levels for the baseline scenario. The baseline scenario serves as a benchmark for the other scenarios. The adjustment parameter „investment costs for re-using water", represents investments in water recycling technologies and technologies for water circulation usage, for example. Costs for water extraction pumps and extraction related tasks are summarized in „costs for extracting water". The parameters „subsidies for environmental protection", „cost of capital", and „labour costs" represent policy parameters. Typical governmental interventions like subsidies or taxes are translated to the model by varying these parameters. However, the „cost of capital" parameter also represents conditions of the global capital market and, thus, is interpreted as a proxy for globalization (Kuhn et al. 2008). Table 1 List of scenario parameters for the industrial model Parameter

Parameter declaration

Performance scenario

Common public interest scenario

Change Cost Of Water Reuse Change Cost Of Extraction Change Subsidies

investment costs for reusing water costs for extracting water

constant

decreasing

constant

increasing

subsidies for environmental protection cost of capital

constant

increasing

decreasing

decreasing

labour costs

constant

increasing

Change Cost Of Capital Change Wages

Source: GLOWA Danube scenarios, GLOWA Danube project

4 2 4 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

The assumptions about the different trends in the parameters for the performance nario and common public interest scenario are summarized in Table l . 1 7

sce-

2.2 The industrial model in the performance scenario

The parameter setting of the performance scenario is based on an optimistic view of the sustainability of the water resources. This is motivated by the fact that in Germany a long-term mean of 188 billion cubic meters of water is available per year, whereas total water usage only adds up to 35.6 billion cubic meters. 18 In other words: because about 81 % of the natural water supply is not used, water resource conditions for industrial water usage in Germany seem to be assured today as well as in the future. Nevertheless, the available water is distributed very heterogeneously over space and time. Therefore local water scarcity - especially in the increasingly hot and dry summers - will become more common. This will result in periods of tighter local water usage regulation when conditions become severe. This scenario assumes stable and continual economic growth, and due to limited regulation under regular conditions it assumes that the costs for extracting water stay moderate. Accordingly low investments in water re-usage technologies are expected. Investment costs for re-using water stay at a high level as only few subsidies for environmental protection are assumed. As a theoretical background we use the development of public drinking water prices as a benchmark for future costs for extracting water.19 As public water supply operations are cost-covering, we assume the costs for industrial water consumption to be similar. Additionally we assume moderate development of labour costs. The only parameter that is adjusted is the „cost of capital", representing the reaction to a further globalizing world with decreasing prices on the global capital markets. 2.3 The industrial model in the common public interest scenario

In this scenario, increasing environmental consciousness in society affects governmental policy as well as the behaviour of the industrial sector. Subsidies for environmental investments increase. This is expressed in this scenario as a decrease in the cost of capital for environmental investments. To ensure a balanced national budget, these subsidies are partially financed by higher ancillary labour costs. Moreover, statutory requirements of water usage will become stricter and the costs of water usage will increase. For example, an increase of 5 cents per cubic meter is plausible, assuming that the same extraction fee as in Baden-Wuerttemberg (the so called „water cent") will be established in Bavaria. 20 In this scenario the revenues of the water fees are used to subsidize investments in projects that aim at reducing water intensity.

17 18

19

20

Parameter settings for the different scenarios are listed in appendix A 4 . 3 5 . 6 billion m 3 water are composed of 2 2 . 5 billion m 3 water for thermal power plants, 7 . 7 billion m 3 water for mining, and 5 . 4 billion m 3 water for the public water supply (cf. Federal Statistical Office Germany 2 0 0 6 ) An overview on the current development of drinking water prices and sewage charges c a n be found in appendix A 5 . Appendix A 6 lists the regional differences in water fees in Germany.

Global Warming Induced Water-Cycle Changes and Industrial Production · 425

3

Simulation results

The topic of the following section is the impact of climate change as well as the impact of the socio-economic scenarios. In this paper we present the consequences for the gross regional product and the industrial water demand for the period from 2012 to 2025. Results of the socio-economic scenarios - baseline, performance, and common public interest - are based on the baseline climate trend REMO regional and the baseline climate variant (see Figure 3). The following simulations include the feedback of the majority of the interdisciplinary sub-models, in particular the models Demography, Economy, GroundwaterFlow, Ground-waterTransport, Household, Tourism, and Water-Supply. Exceptions are the models Atmosphere, Farming, Landsurface, Rivernetwork, and Traffic, which are included as pre-calculated scenarios for the corresponding time horizon. The climate trends are assumed to be exogenous. The parameters that influence the industrial model the most are the conditions of the local water resources, water prices (including fees and other regulations), labour market conditions (wages, ancillary labour costs, ...), capital market conditions (interest rates, subsidies,...) and the climate. The full set of feedback and spillover effects between all the sub-models is only implemented inside the upper Danube river basin. Since domestic migration of workers (which serves as an input factor for industrial production) does not stop at this geographic border, we use our best guesses for the remaining areas in Germany and Austria. For this purpose we compute in every simulation step the average value of a parameter inside the upper Danube area which we then use as a proxy for missing parameters outside the upper Danube catchment area. For this reason we observe much more fluctuation and regional distinctions inside this core area. For visual interpretation in a map it is not recommended to show the results on the microscale of the simulated square kilometres. The areas are simply too small and the industrial regions too scattered to allow a sensible interpretation when printed. Therefore we choose to display the NUTS 3 district averages in the following maps. To illustrate the 4th Selection: Policy measure Policy measure 1

IPCC regional

_____—______ 5 warm winter

Policy measure 2

MM5 regional

5 hot summer

Policy measure ...

Foreward projection

5 dry years REMO scaled & bias corrected MM5 scaled & bias corrected

Description of the climate scenarios in W. Mauser, CLOWA Danube Project (2010). Source·. GLOWA Danube

Figure 3 GLOWA Danube scenario matrix

426 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

underlying simulation on square kilometres, Figure 4 shows a cut-out with the micro results of the regions surrounding Munich (München). 3.1 Development of the gross regional product The following results are based on the climate trend REMO regional, the climate variant baseline, and the baseline social scenario. They are compared to a simulation based on an artificial zero climate change climate trend, which is defined by simply conserving today's climate and water-cycle conditions. To assess the consequences of climate change, we use this scenario in order to illustrate the hypothetical case that no further climate change would occur from today onwards.

GLOWA Danube catchment are a

< --0.50% -0.50% - -0,25% • i -0,25% - 0.00% • • 0.00% -
O 3 · Ό c octn — c "C — re ?« Eα J 8 £

υ

1

I» e go Ό

c

> T3 — C re ra

1i ? i

3

3® O re Ì

4 sussi industrial water demand —«—water conditions (one year lagged)

Source: Ifo Institute

Figure 6 Change in industrial water demand in comparison to 2012 and one year lagged ground water conditions in Salzburg For Salzburg the situation stays rather moderate. In the simulation the water quantity conditions worsen once in 2013, but improve again in the following year and remain good in the following years. Consequently the industry reduces the water demand only slowly and does not even reach a 6 % reduction in 2025 relative to 2012 (see figure 6), one reason also being a local increase in the population contrary to a decreasing population in Weiden. 3.3 Comparison of the scenarios for industrial water demand Figure 7 and Figure 8 show the difference in the relative changes of industrial water demand between the socio-economic scenarios performance or common public interest in relation to the baseline scenario for each industrial active square kilometre in 2025 relative to 2012. In other words, we compare one simulation driven by the performance scenario or common public interest scenario with one simulation driven by the baseline scenario and compute the difference of the percentage values for each square kilometre. For this reason, positive values indicate higher industrial water demand in the performance or common public interest scenario than in the baseline scenario. In the performance scenario the changes in water demand range between 2.19 percentage points reduction and 0.75 percentage points increase compared to the baseline scenario (see Figure 7) or between 4.17 percentage points reduction and 0.75 percentage points increase in the common public interest scenario compared to the baseline scenario (see Figure 8). This effect is less negative in areas of high population density like Munich, Innsbruck and Salzburg. Thus, in comparison to rural areas, the relative water demand increases in the cities. Different small-scale effects such as higher migration movements into cities reinforce these regional differences. The observed regional differences can be explained partially by regionally highly divergent water conditions. Water scarcity in Germany and Austria worsens due to climate

Global Warming Induced Water-Cycle Changes and Industrial Production · 429

(values in percent) Source: Ifo Institute

Figure7 Industrial water demand 2 0 1 2 - 2 0 2 5 , baseline scenario vs. performance scenario change. This in conjunction with the social adaptation strategy leads to a reduction in water demand, which is reflected in the light gray and gray districts. Moreover, higher subsidies for investments in sustainable water usage in the common public interest scenario encourage higher water demand reductions of industry when water conditions worsen. To isolate the effect of climate change up to 2 0 2 5 , we again compare results of a local simulation with the zero climate change setting to the results of a local simulation with the „ R E M O regional" climate trend (the similar setting as for the GRP results above).

430 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

1 1

GLOWA Danube catchment are a

§H§ ÜB8 •¡I • •

-4,17*- -2.00% -2,00% - -0,25% -0.25%- 0.00% 0,00% - 0,75%

(values in percent) Source: Ifo Institute

Figure 8 Industrial water demand 2 0 1 2 - 2 0 2 5 , baseline scenario vs. common public interest scenario The changes in industrial water range between 5.58 percentage points reduction and 11.64 percentage points increase up to 2025 relative to 2012 (see Figure 9). The results again show the necessity of regional small-scale simulation and interaction as we can observe large regional differences inside the upper Danube catchment area. These differences reflect the regional divergent effects of climate change due to water condition developments explicitly modelled in the river basin. An important message that we can derive from these results is the fact that the effects of the socio-economic scenarios (performance or in the common public interest) is minor compared to the consequences of climate change.

Global Warming Induced Water-Cycle Changes and Industrial Production · 431

GLOWA Danube catchment are [ _ _ _ ] -5,58% - -2,00% -2,00% - -0,25% -0,25% - 0,00% 0,00%

-11,64%

(values in percent) Source: Ifo Institute

Figure 9 Industrial water demand 2 0 1 2 - 2 0 2 5 , zero climate change, baseline scenario vs. REMO regional climate change, baseline scenario

4

Conclusions

Using the AFiD Panel we calibrated the environmental decision support system DANUBIA and simulated the effects of different climate change and socio-economic scenarios up to 2025. The results show a general decline in water usage accompanied by worsening conditions in natural water-cycles. Thus, climate change is an issue for the sustainable development in German and Austrian regions although it is comparatively moderate. We observe large regional disparities in the extensively analyzed upper Danube river basin. These are mainly caused by climate change but also by society. We further show that

432 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

cities are economically less affected by climate change than rural areas. The results allow the identification of regional hot spots and a quantification of the effects of various policy measures that aim at compensating society f o r climate change with respect to economic and environmental sustainability. However, the potential improvements in the observed socio-economic scenarios are limited since their effect is minor compared to the impact of climate change.

Appendix A1 Sub-model interaction in DANUBIA The U M L diagram in Figure A l presents an overview of the structure of the D A N U B I A framework.21

Actor Demography Economy Farming Household Tourism Watersupply

Landsurtace Biology ChannelFlow Hydraulls Structures Radiation Balance Snow

Soil SoilNitfogen Surface

Groundwater GtountlwatsrTranspor! QroundwalïFlow "f"

Source: GLOWA Danube

Figure A1 Interaction of the models in DANUBIA 21

U M L refers to unified modelling language, a notation convention c o m m o n to computer science applications. Since the general intuition of the illustration is accessable without deeper knowledge of the terminology, w e will abstract f r o m a detailed introduction into object-oriented programming. In general it is sufficient to k n o w that the boxes labelled with the respective superordinated type serve as a container of conformable models representing its classes. The a r r o w s with blank heads mean „ e x t e n d " such that a class or model at the tail of an a r r o w extends the one at the head (or in other w o r d s : the model at the head of the a r r o w is the base of the extending model and its abilities are inherited to the extension). An outgoing line with a circle means „ p r o v i d e " and a semicircle with an incoming line means „ r e q u i r e " .

Global Warming Induced Water-Cycle Changes and Industrial Production · 433

Detailed descriptions of the framework can be found in Barthel et al. (2010), Barthel et al. (2008), Hennicker and Ludwig (2005, 2006) and Barth et al. (2004). A2 Estimation results used for the calibration of the simulation Table A1 Coefficients used for the calibration of the simulation Model

(1) OLS

(2) OLS

(3) OLS

year 0.010** employees 0.549** capital 0.406** extracted water 0.056** region dummies no -13.751* constant R2 0.9850 160 observations

0.012** 0.533** 0.315** 0.054** yes -16.012** 0.9996 160

0.010** 0.490** 0.380** 0.040** yes -15.142** 0.9933 160

Below prices derived from model (3) € 27,698 per year and employee 6.0 % interest rate € 3.07 per cubic meter

Model specification (3) used for the simulations in this paper. Source·. Ifo Institute

The coefficients in Table A l for the calibration of the model were estimated in Jessberger, Zimmer (2010). Due to the log-log specification, the coefficients can be directly interpreted as elasticities. The implicit prices for the production factors that follow from the Cobb-Douglas specification of the production function are listed in the last column of Table A l . As seen in Table A3 in the scenario chapter the German average effluent charges in 2005 were about € 2.28 for each cubic meter of water. Since public water suppliers in Germany charge for their water on a non-profit base according to their extraction and supply costs, we can use their prices as an indicator of the extraction cost of the self-supplied industrial producers. The average costs for public water in 2007 were €1.85, as indicated in Table A4. As expected, the estimated costs of €3.07 per cubic meter lie well below the roughly four euros of the public water supply.

434 • Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer A3 Components of the interdisciplinary global change decision support system DANUBIA Atmosphere: Mesoscale modelling of the atmosphere The mesoscaled atmosphere model MM5 has been integrated in DANUBIA and interconnected with the land surface modelling of the group „Hyd-Fern" by the sub-project group ,,Meterology/MM5". They use a downscaling method, which has been developed within the project by the former project partners „Wirth", to downscale the 45 km grid of MM5 to the 1 km grid of DANUBIA. Meterology/MM5 also employed, processed, and downscaled data of the Al Β scenario of the global climate model ECHAM5 for the years 2001 to 2100. Land surface: a) Plant growth The model Biological simulates plant growth in the context of the DANUBIA decision support system. Biological was designed to assess the role of the vegetation in the cycles of, water, nitrogen, and carbon under climate change conditions using a process based approach (Lenz-Wiedemann et al. 2010). It simulates plant growth, taking into account the influences of radiation as well as the availability of water, nitrogen, and CO2. In the case of agricultural vegetation, Biological interacts dynamically with the farming component. b) Soil Nitrogen Transformation (SNT) The model SNT simulates soil nitrogen transformation in the context of the DANUBIA decision support system. SNT was designed to assess the role of nitrogen transformation processes in the soil in the context of changing cycles of water, nitrogen, and carbon under climate change conditions using a process based approach (Klar et al. 2008). Distinguishing between humus and fresh organic matter, it simulates all relevant turnover processes of ammonia and nitrate pools including nitrate leaching into the groundwater. c) Natural environment Further components modelling soil, land surface and radiation balance.

Actors: The actor component consists of a rich set of models, which are strongly interacted with each other (An interaction is typically an exchange data, e.g. the household model generates the households' water demand, which is used by the water supply model as input in order to decide on the development of additional infrastructure). An overview of the strongest direct interaction paths between the models is given in Figure A2. The models themselves typically consist of several submodels. a) Ground water management and supply In the water supply model, which is implemented as an actor model (Barthel et al. 2008, Barthel et al. 2010), the WaterSupplyCompany actors behave differently in different socioeconomic scenarios (see section „Economic and societal scenarios in frames of climate change"). b) Household The sub-project „environmental psychology" developed an agent based model of lifestyles in the context of the environment and water usage behaviour. Main issues here are drinking water consumption, risk awareness and risk valuation with respect to water, and investments of households in water saving innovations. c) Farming The aim of the sub-project „agricultural economy" is to detect possible changes in agricultural incomes, land and water use and crop management, due to different climate change and socio economic scenarios. Accordingly, a two-step model was developed. At first the process-orientated agricultural sector model ACRE makes plans for farming for the next year on a county level. In a second step the agent model DeepFarming models the daily management decisions per square kilometre based on these agricultural plans. d) Demography The demography model determines the domestic migration movements depending on the socio-economic conditions and amenities of potential destinations and given national demographic trends. It accounts not only for conditions at the specific destination but also for network effects and the conditions in neighbouring regions.

Global Warming Induced Water-Cycle Changes and Industrial Production · 435

d) River network Surface water is modelled with direct interfaces to the atmosphere, ground water, and plant growth. This sub-project illustrates energy and water fluxes inside the upper D a n u b e catchment area and models the water quantities for every square kilometre of a river channel. Thus, it serves as the basis of our industrial model's river water demands. e) Snow cover and glaciers Snow and ice components have been developed from the sub-project „Glaciology" to the needs of other DANUBIA partners like water management, tourism, and other socio-economic issues. As DANUBIA processes are standardized on a l x l km scale (one DANUBIA Proxel), they had to model snow accumulation and snow melt as the development of glacier area and the resulting ice melt by use of subscale parameterization. Groundwater: G r o u n d water balance This sub-project group implemented models to simulate ground water flow and ground water quality (Barthel et al. 2005a, Barthel et al. 2005b, Wolf et al. 2008) as well as the water supply for the upper Danube catchment area.

e) Tourism The sub-project „Tourism" simulates the water d e m a n d of the tourism sector. Therefore several sub-models have been developed for: the operating state of tourism infrastructure (golf courses, ski areas, swimming pools, hotels and gastronomy) and the tourism location attractiveness - measured in the number of bed nights and same day visitors. The model simulates various possible changes in the tourism industry - supply and demand - as conditioned by climate change, e.g. movement of winter guests to more snow-reliable ski areas or the increase of guests during the summer season. f) Economy The „ R I W U " sub-model captures macroeconomic developments and delivers the price levels, wage rates and interest rates that are employed by other sub-models (Langmantel/ Wackerbauer 2003). g) Industry The sub-model which simulates the industrial producers with a spotlight on their usage of water resources is the focus of this paper.

Source: GLOWA Danube

Figure A2 Main paths of interaction in the actor network of DANUBIA

436 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer A4 Trend values of the scenario parameters Table A2 List of adjustable scenario parameters for the industrial model Adjustable parameter

Declaration

Change Cost Of Water Reuse Change Cost Of Extraction ChangeSubsidies

investment costs for reusing water costs for extracting water subsidies for environmental protection cost of capital -0.48% p.a. labour costs -

Change Cost Of Capital ChangeWages

Performance scenario

Common public interest scenario -0.48 % p. a. +0.15% p.a. +0.48% p.a. -0.48% p.a. +0.48 % p. a.

Source: CLOWA Danube scenarios

A5 Development of water prices and sewage charges in Germany Table A3 Sewage charge prices conform to the fresh water benchmark weighted by habitants Old West German states Newly-formed German states Germany

€/m3 2002

€/m3 2005

Change

p.a.

2,05 2,47 2,11

2,16 2,87 2,28

5,4 % 16,2% 8,1 %

1,8% 5,1 % 2,6 %

Source: BDEW (Federal Associationof Energy and Water Management) Table A4 Mean water prices in Germany in 2007 Old West German states Newly-formed German states Germany

€/m3 2001

€/m3 2007

Change

p.a.

1,64 2,05 1,70

1,79 2,15 1,85

9,1 % 4,9 % 8,8 %

1,5% 0,8 % 1,4%

Source: BDEW (Federal Association of Energy and Water Management)

Global Warming Induced Water-Cycle Changes and Industrial Production · 437

A6 Water extraction fees in Germany Table A5 „Water-cent" per each m 3 of extracted water in German states SUte

Watercent

Baden-Württemberg Bayern Berlin

5.1 31

Brandenburg

10.2

Bremen Hamburg

7-8

Hessen MecklenburgWestern Pomerania

1.8

Niedersachsen

5.1

Nordrhein-Westfalen

4.5

Explanations

Yearly payments

since 1988

No label ca. € 55 million

With two times ca. € 20.2 million of increase since 1994 ca. € 0.7 million since 1993, confirmed in of WVU 4/2004 For about. € 3.0 million of 12 years, increased WVU in 12/2005 Abolished in 1 / 03 Updating the ca. € 1 . 7 million water-pfennig of the DDR, confirmed in 1/2003 Confirmed in Ca. € 20 million 12/2004 of the public water supply € 72 million for Since 1.2.2004 drinking water and process water (2005)

Rheinland-Pfalz Schleswig-Holstein

5 - 111)

since 1.1.2004

Saarland

6-7

Proposed To introduce by state-government in 2007

Sachsen

1.5

Designated use of funds

Protection of ground water Realization of WRRL, maintenance of dikes , etc.

For ground water saving arrangements For ground water saving arrangements Realization of WRRL 2 '

ca. € 24.5 million

50 % labelled for different purposes (up to € 3 million) (partially labelled)

ca. € 3.4 million

Labelled for different purposes

Sachsen-Anhalt Thüringen 1)

5 cents for business enterprises as end-consumers if they consume more than 1,500 m 3 of water in time period, 11 cents for all other end-consumers. 2) Possible to apply against expenditures within the farming cooperation. Source: BDEW (Federal Association of Energy and Water Management)

438 · Christoph Jeßberger, Maximilian Sindram, and Markus Zimmer

References Alcamo, J. (ed.) (2008.), Environmental Futures: The Practice of Environmental Scenario Analysis. Elsevier. Assaf, H., E. van Beek, C. Borden, P. Gijsbers, A. Jolma, S. Kaden, M. Kaltofen, J. W. Labadie, D.P. Loucks, N.W.T. Quin, J. Sieber, A. Sulis, W.J. Werick, D . M . Wood (2008), Generic Simulation Models for Facilitating Stakeholder Involvement in Water Resources Planning and Management: A Comparison, Evaluation and Identification of Future Needs. Pp. 229-246 in: A.J. Jakeman, A.A. Voinov, A.E. Rizzoli, S.H. Chen (eds.), Environmental Modelling, Software and Decision Support - State of the Art and New Perspectives. Elsevier. Barth, M., R. Hennicker, A. Kraus, M. Ludwig (2004), DANUBIA: An Integrative Simulation System for Global Change Research in the Upper Danube Basin. Cybernetics and Systems 35: 639-666. Barthel, R., S. Janisch, D. Nickel, A. Trifkovic, T. Hörhan (2010), Using the Multiactor-Approach in GLOWA-Danube to Simulate Decisions for the Water Supply Sector under Conditions of Global Climate Change. Water Resources Management 24: 239-275. Barthel, R., S. Janisch, Ν. Schwarz, Α. Trifkovic, D. Nickel, C. Schulz, W. Mauser (2008), An integrated modelling framework for simulating regionalscale actor responses to global change in the water domain. Environmental Modelling and Software 23: 1095-1121. Barthel, R., D. Nickel, A. Meleg, A. Trifkovic, J. Braun (2005a), Linking the physical and the socio-economic compartments of an integrated water and land use management model on a river basin scale using an object-oriented water supply model. Physics and Chemistry of the Earth 30: 389-397. Barthel, R., V. Rojanschi, J. Wolf, J. Braun (2005b), Large-scale water resources management within the framework of GLOWA-Danube. Part A: The groundwater model. Physics and Chemistry of the Earth, 30: 372-382. Chung, J.W. (1994), Utility and Production Functions. Theory and Applications. Cambridge. Dachraoui, Κ., T.M. Harchaoui (2004), Water Use, Shadow Price and the Canadian Business Sector Productivity Performance. Economic Analysis Research 26. Dupont, D.P., S. Renzetti (2001), The Role of Water in Manufacturing. Environmental and Resource Economics 18: 411-432. Dupont, D. P.,S. Renzetti (2003), The Value of Water in Manufacturing. CSERGE Working Paper ECM 03-03. Eckey, H.-F., R. Kösfeld, M. Tiirck (2005), Regionale Produktionsfunktionen mit Spillover-Effekten für Deutschland - Empirischer Befund und wirtschaftspolitische Implikationen. Schmollers Jahrbuch: Journal of Applied Social Science Studies 125: 239-267. Egerer, M. (2005), Global Change and the Effects on Industrial Water Use in the Catchment Area of the Upper Danube. Conference paper: 7th Nordic Environmental Social Science Research Conference. Göteborg University. Egerer, M., M. Zimmer (2006a), Does Global Change Matter? - The case of Industries in the Upper Danube Catchment Area, Transactions on Ecology and the Environment 98: 75-88. Egerer, M., M. Zimmer (2006b), Weiterentwicklung des Regionalmodells RIWU zu einem „tiefen" Akteursmodell. Pp. 391-436 in: W. Mauser, U. Stasser (eds.), GLOWA-Danube - Integrative Techniken, Szenarien und Strategien zur Zukunft des Wassers im Einzugsgebiet der Oberen Donau, www.glowa-danube.de. Federal Statistical Office Germany (ed.) (2006), Umwelt - Wasserversorgung und Abwasserbeseitigung in der Industrie 19. Gispert, C. (2004), The Economic Analysis of Industrial Water Demand: a review. Environment and Planning C: Government and Policy 22: 15-30. Griffin, R. C. (2006), Water Resource Economics: The Analysis of Scarcity, Policies and Projects. The MIT Press: Cambridge/London. Hennicker, R., M. Ludwig (2005), Property-Driven Development of a Coordination Model for Distributed Simulations. Lecture Notes in Computer Science 3535: 290-305. Hennicker, R., M. Ludwig (2006), Design and Implementation of a Coordination Model for Distributed Simulations. Lecture Notes in Informatics 82: 83-97.

Global W a r m i n g Induced Water-Cycle Changes and Industrial Production · 4 3 9

IHP HWRP Reports (2008), GLOWA - Globaler Wandel des Wasserkreislaufes. Magazine 7. IPCC (2007), Summary for Policymakers. Pp. 1-18 in: S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M.Tignor, H.L. Miller (eds.), Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge/New York. Jessberger, C., M. Zimmer (2010), Estimating the Value of Water for Industrial Production in Germany. Mimeo. Klar, C.W., P. Fiener, V. Lenz, P. Neuhaus, K. Schneider (2008), Modelling of soil nitrogen dynamics within the decision support system DANUBIA. Ecological Modelling 217: 181-196. Kim, Y. H. (1992), The Translog Production Function and Variable Returns to Scale. The Review of Economics and Statistics 74: 546-552. Kuhn, S., R. Barthel, S. Janisch, A. Ernst, T. Krimly, M. Sax, M. Zimmer (2008), DeepActorModelle in DANUBIA. Chapter E3, two pages, in: Universität München (LMU) (ed.), Global Change Atlas, Einzugsgebiet Obere Donau, GLOWA-Danube Projekt. Labadie, J. (2005), MODSIM: River basin management decision support system. Pp. 569-592 in: V. Singh, D. Frevert (ed.), Watershed Models. Florida. Lang, U., R. Schick, G. Schroder (2010), The Decision Support System BodenseeOnline for Hydrodynamics and Water Quality in Lake Constance, Decision Support Systems, Advances in. Vukovar. Langmantel, E., J. Wackerbauer (2003), RIWU - A Model of Regional Economic Development and Industrial Water Use in the Catchment Area of the Upper Danube. International Journal of River Basin Management 1: 1-5. Lenz-Wiedemann, V.I.S., C.W. Klar, Κ. Schneider (2010), Development and test of a crop growth model for application within a Global Change decision support system. Ecological Modelling 221: 314-329. Mauser, W. (2010), Klimavarianten der regionalen Klimamodelle M M 5 und REMO. Chapter S5, eight pages, in: W. Mauser (ed.), Global Change Atlas - Szenarien und Ergebisse. München. Renzetti, S., E. Elgar (2002), The Economics of Industrial Water Use. Northampton. Reynaud, A. (2003), An Econometric Estimation of Industrial Water Demand in France. Environmental and Resource Economics 25: 213-232. United Nations (2004), Resolution adopted by the General Assembly, A/RES/58/217. Wolf, J., R. Barthel, J. Braun (2008), Modeling ground water flow in alluvial mountainous catchments on a watershed scale. Ground Water 46: 695-705. Dr. Christoph Jeßberger, Ifo Institute for Economic Research at the University of Munich, Poschingerstr. 5, 81679 Munich, Germany. [email protected] Maximilian Sindram, Ifo Institute for Economic Research at the University of Munich, Poschingerstr. 5, 81679 Munich, Germany. [email protected] Markus Zimmer, Ifo Institute for Economic Research at the University of Munich, Poschingerstr. 5, 81679 Munich, Germany. [email protected]

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2011) Bd. (Vol.) 231/3

The Economic Consequences of One-third Co-determination in German Supervisory Boards First Evidence for the Service Sector from a New Source of Enterprise Data By Franziska Boneberg, Lueneburg* JEL J50 Co-determination; board-level employee representation; Germany.

Summary In Germany, the establishment of supervisory boards and, therefore, the board-level employee representation are mandatory, depending on the legal form and size of a company. However, the empirical analysis reveals that the bigger part of the companies observed (Limited liability companies with 500 to 2000 employees active in the West-German service sector) does not satisfy the law. This fact has strong impact on research questions in the co-determination field: Many studies have tried to analyze the economic consequences of the German co-determination laws (all examining the 1976 Co-determination Act). However, as the regulations are compulsory, compelling results are difficult to obtain. The bigger part of the studies compares companies that fall into the scope of different co-determination laws. This implies that mainly big companies are contrasted to smaller ones. It is not difficult to see that a comparison of such kind entails further irregularities. The study presented allows better analysis. The data is taken from two sources: the commercial Hoppenstedt Database and official German statistics. Due to the special kind of data it is possible to compare companies of same size, same legal form, active in the same sector that only differ in the existence or non-existence of a supervisory board. Therefore, the study at hand provides more accurate evidence of the economic consequences of the German 2004 Co-determination Act.

1

Introduction

Since their first introduction in the fifties, the G e r m a n co-determination laws have been highly controversial. Especially in the question of implementing w o r k e r s ' participation rights in c o m m o n legal f o r m s within the E u r o p e a n Union, the debate a b o u t their consequences has become very heated. As yet, n o final determination has been m a d e as t o whether the institution brings w i t h it economic advantages for companies, so researchers * Access to the services statistics panel was provided via remote data access at the Research Data Centre of the Statistical Office of Lower Saxony. For more details about the data access, see Ziihlke, Zwick, Scharnhorst and Wende (2004). All calculations were performed using Stata 10. All do-files are available from the author on request. Many thanks go to Nils Braakmann, Christian Pfeifer, Mario Richter, Alexander Vogel, and Joachim Wagner for helpful comments and to Rita Hoffmeister for running the do-files in the Research Data Centre.

The Economic Consequences of One-third Co-determination · 441

continue to disagree about the advantages and disadvantages of co-determination. While advocates emphasize the activating and motivating effects of co-determination, opponents fear loss of efficiency. Several researchers 1 have tried to analyze the possible economic effects o f the German co-determination legislation. M o s t studies have compared companies that fall into the scope of the three different co-determination acts. Since the extent to which the laws can be applied depends on the size of a company, these studies have mainly contrasted big corporations with smaller ones, but such a comparison necessarily entails irregularities. In addition, a lack o f meaningful data and adequate econometric methods has impeded research. T h e present study advances the state of research in several ways. According to the 2 0 0 4 Third Part Act, companies with 5 0 0 - 2 0 0 0 employees must establish a supervisory board and assign a third of the seats on the board to employees. However, previous investigations have shown that, contrary to the law, about half of limited liability companies with 5 0 0 - 2 0 0 0 employees in the western German service sector have not established a supervisory board (Boneberg 2 0 0 9 a ) , so there is no co-determination at the enterprise level in the companies' boards. This means that there are companies of the same size, same legal form and active in the same sector which fall into the scope of the German 2 0 0 4 Codetermination Act due to their size, but differ mainly in the existence or non-existence of a supervisory board. By comparing the economic performance of these firms, more accurate evidence of the economic consequences of the German 2 0 0 4 Co-determination Act can be provided. T h e data base originates from two sources: initial information was collected from the Hoppenstedt Database, a commercial database that provides information on the size, age, legal form and ownership structure of all German companies that employ more than 2 0 0 people and/or have more than 2 0 m Euro in sales volume per year. Because facts on the presence or absence of a supervisory board were not available for all companies from this data source, missing data were collected via telephone calls. However, in order to analyze the economic consequences of the 2 0 0 4 Co-determination Act, details about productivity and profitability of the observed firms are also needed. These are obtained from official German statistics. By merging the information from the Hoppenstedt Database and the official statistics, the performance of companies that have a supervisory board and those that do not can be compared. This paper examines the effects of the existence of a supervisory board on two core performance indicators: productivity and profitability. Because of their size, all observed firms fall into the scope of the 2 0 0 4 Co-determination Act. Due to its increasing relevance, the paper focuses on the western German service sector. While one-third co-determination at the supervisory board level is neither positively nor negatively related to the two indicators for limited liability western German manufacturing firms (see Wagner 2009), the results in the service sector differ. The remainder of the paper is organized as follows: First, the legal and theoretical backgrounds are outlined. N e x t , an overview of the existing empirical studies is given, followed by data description and the empirical results. Finally, the results are discussed and a conclusion is drawn.

1

See section four for further details.

442 · Franziska Boneberg

2

Legal background

In Germany, employee representation is provided at two levels. Workers' participation at the establishment level refers to the works council (Betriebsrat), where employees participate in operational decisions, such as those related to lay-offs. It is the task of the works council to represent the employees' interests with management. Employee representation is also required at the enterprise level, where workers participate in corporate planning and decision-making processes relevant to the company as a whole (see Junker 2006: 442f.). This level of participation is implemented in the supervisory board, on which employees receive a certain number of seats and votes, depending on the legal structure and size of the company. The mission of the supervisory board is primarily to oversee and control the management of the company. In corporations, the chief executive is appointed by the supervisory board, but this is not the case in limited liability companies. Because of the resulting weak position of supervisory boards in limited liability companies, Fuchs and Köstler (2005: 35f.) denote them as solely informational organs. Since 1976, three laws have been implemented that regulate workers' participation on supervisory boards: The M o n t a n Co-determination Act (Montan-Mitbestimmungsgesetz), the 1976 Co-determination Act (Mitbestimmungsgesetz (MitbestG)) and the 2004 Third Part Act (Drittelbeteiligungsgesetz (DrittelbG)). All companies examined in the present analysis fall into the scope of the 2004 Third Part Act, which applies to firms that usually employ 500-2000 workers, whether the firms are corporations (AGs), partnerships limited by shares (KGaA), limited liability companies (GmbHs), mutual insurance associations ( W a G ) , or cooperative, industrial and provident societies (§ 1 DrittelbG) (The working time stipulated by contract is of no relevance when determining company size). The law assigns a third of the seats on the company's board to the employees ( § 4 1 DrittelbG). In contrast to the other co-determination laws, the 2004 Third Part Act does not dictate an exact number of board members, so the provisions of the stock corporation law - which prescribe a board size of at least three members, and thereafter a number divisible by three (§ 95 S. 1, 3 Aktien-Gesetz (AktG)) - are implemented. The 2004 Third Part Act is applied when no statutory regulation indicates that the scope of another co-determination law is more favorable to workers (§ 1 II 1 N o . 1 DrittelbG). The provisions of the 2004 Third Part Act are mandatory, so they cannot be changed by statute or by collective bargaining agreements (see Oetker 2007: 1836). Co-determination at the enterprise level is governed by all three laws. The M o n t a n Codetermination Act, which applies to companies in the coal and steel industry that have more than 1,000 employees, provides equal representation for employees on the company's supervisory board. In addition, a representative of the employees' side can operate as a worker director on the board (see Junker 2006: 452f.; see also Niedenhoff 2005: 382ff. ; Fuchs/Koestler 2 0 0 5 : 2 0 ) . Companies that regularly engage at least 2000 employees fall into the scope of the 1976 Co-determination Act, which also provides equal representation on the supervisory board. However, because of the tie-breaking vote of the chairman, w h o generally sides with the shareholders, the law actually provides „quasiparity" (see Dönges et al. 2007: 15 f.). Only one of the three laws applies to any one company. The provisions of the M o n t a n Co-determination Act have priority (§ 1 II MitbestG), and the 1976 Co-determination Act takes precedence over the 2004 Third Part Act (§ 1 III MitbestG).

The Economic Consequences of One-third Co-determination · 443

3

Theoretical framework

This section outlines the theoretical background concerning the economic consequences of workers' participation. Among economists, there are generally two points of view on the potential effects of co-determination: the Property Rights approach and the Participation Theory approach. 3.1 The Property Rights approach The supporters of the Property Rights approach, such as Furubotn (1985; 1988) and Pejovich (1978; 1990), argue that legal co-determination regulations have primarily negative effects on a company. In their opinion, participation rights reduce the residual decision rights of the owners, which results in less efficient or at least delayed decisions, as well as in delays in the planning and innovation process. In the opinion of Pejovich (1978), shareholders must be able to influence managerial decisions or their willingness to invest capital in the enterprise will decrease. Pejovich (1990: 69) contends that participation rights influence the relationship among employers, shareholders and employees, but that they also alter the roles between risk carrier and benefactor, leading to conflicts of interest that impede efficient solutions. The separation of the position of risk carrier and that of decision-maker has negative impacts on the company's efficiency (see Kraft/Stank 2004: 428). In this context, Pejovich (1990: 69) argues, „Co-determination shifts the responsibility for decisions to a group of people who are not at all affected by the consequences of the decisions" such that shareholders and employees benefit from successful investments, while the consequences of unsuccessful investments fall to the owners alone. Consequently, the owners see lower productivity for their investments and lower incomes, partly because the employees use their increasing influence to participate in the business's profits (Renaud 2007: 691). Pejovich (1976: 18 ff.) points out that the planning horizon and risk tolerance of equity holders, employers and workers varies, resulting in a strong potential for conflict that shareholders can rarely decide in their own interest because of the participation regulations. Instead, the workers are able to maximize their own utility while the shareholders are not. The Principal Agent theory is often used in argumentation against workers' participation. According to this theory, owners transfer their decision-making rights to the management with the mandate to implement decisions based on their best interests. However, the employees' right to participate in decisions at the enterprise level endangers the protection of these interests with the result that the shareholder may not feel well represented. As Weizsäcker (1983: 146) observes, „A company's ability to respond flexibly to changing conditions, to take advantage of innovation opportunities, to balance risks against opportunities, is heavily influenced by its internal organization and decision-making structure. [...] Participation rights divide the decision-making rights in a company and therefore lead to a de facto reduction of its decision-making and coordinating power." 3.2 Participation Theory Adherents of the Participation Theory argue that the benefits of co-determination rights exceed the disadvantages. As the potential conflict that generally defines the relationship between employer and employee eases, satisfaction on both sides increases, and productivity and the acceptance of innovations are augmented (Kraft/Stank 2004: 430).

444 · Franziska Boneberg

Hirshman (1970) as well as Freeman and Medoff (1984: 94) state that participation rights reduce the labor turnover rate, backing their argument with the exit-voice approach, which traces back to Hirshman (1970: 77ff.). This approach assumes that the collective pooling of interests, such as those in trade unions or works councils (voice), helps prevent employees from leaving the company or from reduced performance and motivation that may be due to dissatisfaction (exit). Workers' participation rights help to retain employees because employees generally prefer dialogue to quitting (Freeman/ Medoff 1984: 8). This advantage benefits the employer as the company avoids high turnover costs for reappointments or wage payment with absence of consideration (Dilger 2002: 68ff.). Levine and Tyson (1990: 185ff.) show two effects from workers' participation in decision-making: On one hand, such involvement increases employees' operational readiness and motivation to work. On the other hand, workers' participation rights lead to an activation of knowledge and a better flow of information. Both effects impact productivity positively, raising the efficiency and profitability of the firm. Levine and Tyson (1990: 187f.) also contend that workers' participation enhances their confidence in both the company and the management, leading to a stronger identification with the corporate objectives. Advocates of this point of view also explain that, since even the most detailed contracts can not be explicit on every potentiality (Hart 1995:23ff.), opportunistic behavior or the emergence of an internal prisoners' dilemma may occur. Both, employers and employees have incentives to deviate from their contractual obligations, so mistrust results. Participation rights can lead to long-ranging employment-employee contracts and support cooperative interaction within the company (Dilger 2002: 55f.). Freeman and Lazear (1995: 29) find something true in both lines of argument. They argue that co-determination on the employees' side leads to increased motivation, willingness to invest in firm-specific skills, and acceptance of innovations, all of which increase productivity. At the same time, the workers' bargaining power increases, along with their demands for a greater share in a company's rent. Thus, co-determination affects not only distribution, but also the amount of joint surplus. As explained in section two supervisory boards in limited liability companies only have limited possibilities and power. Nevertheless, in the present case impact on productivity and profitability can be expected: The from the better exchange of information and consultation in supervisory boards resulting raise in efficiency might lead to an increase in productivity. Concerning profitability two effects seem conceivable. If the employees succeed in obtaining a higher share of the rent due to the increased productivity, profitability will decrease. However, it is also thinkable, that their influence is too small in order to gain considerable advantages. In that case profitability would not be affected at all. 4

Empirical evidence

Many publications have dealt with the potential effects of co-determination at the enterprise level.2 The empirical literature has focused on the analysis of changes in productivity, potential effects on shareholder value and profitability. One study analyzed a possible impact on innovation activities of a company, two others have dealt with potential 2

See Addison and Schnabel (2009) for a detailed overview.

The Economic Consequences of One-third Co-determination · 4 4 5

consequences to a firm's employment level. The data mainly is based on publicly available information on publicly-traded companies, which generally does not entail information on the existence or absence of supervisory boards and therefore co-determination at enterprise level. This is why in all studies the existence of the supervisory board has been assumed. 3 In the course of the investigations either co-determined and non-codetermined firms, or companies that fall into the scope of various co-determination acts, are compared in respect to their business metrics. In most studies Ordinary Least Square (OLS) regressions or Difference-in-Difference (DID) estimations are made. There have been no long-term studies on the effects of co-determination (see Renaud 2007: 693). The studies analyzing the potential impacts of co-determination on productivity have reached conflicting conclusions. While FitzRoy and Kraft (2005) as well as Renaud (2007) identified a positive relationship between co-determination and productivity, Gurdon and Rai (1990) as well as FitzRoy and Kraft (1993) found a negative relationship. Svejnar (1982) and Wagner (2009) found no significant impact on productivity. The results of studies dealing with potential impacts on profitability also vary: while Gurdon and Rai (1990) identified negative effects of co-determination at the enterprise level, FritzRoy and Kraft (1993) found positive consequences for profitability. Wagner (2009) found no significant effect on profitability. As to possible effects on stock prices, the results of extant studies provide a more uniform picture: Benelli et al. (1987), Baums and Frick (1999), Vitols (2006) and Frick and Bermig (2009) found no significant effects, and Schmid and Seger (1998) and Gorton and Schmid (2004) calculated a negative impact. The varying results show that no clear conclusions on the impact of co-determination at the enterprise level can yet be reached. Moreover, a direct comparison of the extant work is difficult to draw because of the different methods and data used. Some of the older studies must be regarded with reservations on account of their methodologies. (For additional details, interpretations and criticisms, see Baums/Frick 1996: 5; Dönges et al. 2007: 36ff.; Renaud 2007: 693ff.; Sadowksi et al. 2000:17f.). Because of their improved approaches, Vitols (2005: 26) considers the results of recent investigations more reliable. Taken together, the extant work on the subject makes it clear that, since the existing studies are neither numerous nor definite, the influence of co-determination cannot yet be assumed (see Kraft 2006: 710). Sadowski et al. (2000: 18) summarize the issue: „All in all the studies at hand suggest that the question of potential effects of participation rights in supervisory boards so far is not empirically resolved." The current study takes another approach to the investigation of potential impacts of co-determination, as discussed in the empirical part of the paper. 5

Data and methodological remarks

Initial information for the data base for the present study was collected from the Hoppenstedt Database, a commercial database that provides information on the size, age, legal form and ownership structure of all German companies that employ more than 200 people and/or have more than 20 m in Euro sales volume per year. Information about the existence and allocation of staff on the supervisory board is also usually included in this database, although this information is not available for every company. Missing data 3

Only one study (Wagner 2009) obviously differs in its approach. It follows the same idea the present study does, using the industrial sector.

446 · Franziska Boneberg

were collected via telephone calls. (For detailed data specification and additional inform a t i o n regarding the detailed process of data collection, see Boneberg 2009a.) The i n f o r m a t i o n a b o u t productivity and profitability required to analyze the economic consequences of the 2 0 0 4 Co-determination Act w a s obtained f r o m the official business services statistics (Strukturerhebung im Dienstleistungsbereich) set u p by the G e r m a n Federal Statistical Office a n d the statistical offices of the Federal States (Länder). These statistics include, a m o n g other data, i n f o r m a t i o n o n companies' economic sector, the n u m b e r of employees (not including t e m p o r a r y workers), total turnover, subsidies, salaries and wages of a company. T h e European Union first collected these statistics for the year 2 0 0 0 . The d a t a covers the enterprises and professions (Freie Berufe) of the N A C E divisions I (transport, storage a n d communication) a n d Κ (real estate, renting a n d business activities) with a n annual turnover of at least €17,500. In order to assign the enterprises, a stratified r a n d o m sample is employed. T h e stratification is based o n the federal states, 4-digit industries, and 12 size ranges (in terms of turnover or employees). While the data is generally confidential, it can be utilized by researchers on a contractual basis via controlled remote data access inside the research data centres of the G e r m a n Statistical Offices. (For details, see Z ü h l k e et al. 2004.) Additional i n f o r m a t i o n a b o u t the G e r m a n business services statistics panel can be f o u n d in Vogel (2009). Merging the i n f o r m a t i o n f o r m the H o p p e n s t e d t D a t a b a s e a n d the official statistics makes possible an e x a m i n a t i o n of potential consequences for productivity a n d profitability of companies w i t h a n d w i t h o u t supervisory boards. Merging w a s done using inf o r m a t i o n a b o u t the register n u m b e r and register court of the trade register (Handelsregisternummer und Handelsregistergericht) for an enterprise. This i n f o r m a t i o n is available in both the H o p p e n s t e d t data base and in the official register of enterprises (Unternehmensregister) that w a s linked with the business services statistics data. T h e current study uses the results of the 2 0 0 6 business services statistics; however, profitability a n d productivity can only be determined as proxies because the d a t a set does not include any i n f o r m a t i o n on a company's capital stock or the sum of assets or equity. Therefore, it is not possible to compose profit indicators like return on assets or return on equity. Consequently, profitability is measured as turnover profitability, defined as the rate of return generated as gross firm surplus 4 divided by total sales, minus net change of inventories. Productivity is measured as value added per employee and, for a f u r t h e r robustness check, as turnover per employee. The n u m b e r of employees is based o n the n u m b e r of employed persons and not o n full-time equivalents, since full-time equivalents is n o t included in the dataset. However, this fact does not pose a p r o b l e m because the 2 0 0 4 Third Part Act does not differentiate companies on the basis of full-time equivalents anyway. T h e initial data set used in previous investigations contained 5 0 0 companies (see Boneberg 2 0 0 9 a ) . The present study includes only 173 companies because only these firms are in both the H o p p e n s t e d t D a t a b a s e a n d the official statistics. The official business services statistics comprises solely companies active in sectors I (transport, storage a n d c o m m u nication) and Κ (real estate and renting), while the H o p p e n s t e d t d a t a collection also contains firms f r o m other branches. 4

The definition applied here is in line with the denotation of the European Commission (1998): gross value added at factor costs, minus gross wages and salaries, minus costs for social insurance paid by the firm.

The Economic Consequences of O n e - t h i r d Co-determination · 4 4 7

6

Empirical investigation

This section undertakes the empirical investigation by, first, running t w o different mean tests and, second, running different OLS-estimations. 80 of the companies in the sample have a supervisory b o a r d , and 93 d o not (Table 1). Table 1 Frequencies of firms w i t h / w i t h o u t supervisory boards Frequencies

Supervisory Board 0 1

93 80

Total

Percent 53.76 46.24

173

100

Table 2 reports the results of a t-test on m e a n differences. For productivity, the difference in m e a n value is statistically significant at a 5 % level, so companies with a co-determined supervisory board are, on average, more productive t h a n those w i t h o u t one. The t-test for the value added per employee confirms the result. The o u t c o m e for profitability, on the other h a n d , is not statistically significant. Since the t-test provides inf o r m a t i o n on only one m o m e n t of the productivity or profitability distribution, it is useful to apply the Kolmogorov-Smirnov test as well. This test provides i n f o r m a t i o n on any difference in the w h o l e distribution of productivity and profitability for companies with and w i t h o u t supervisory boards. (For more details concerning this test a n d its application, see Delgado et al. 2002.) According to the Kolmogorov-Smirnov test, both the t w o productivity distributions and the t w o profitability distributions differ for firms with and w i t h o u t w o r k e r s ' participation at the enterprise level. Apparently, firms with a co-determined supervisory board are, on average, significantly more productive t h a n c o m p a nies w i t h o u t a supervisory b o a r d . In contrast, companies with w o r k e r s ' participation at the enterprise level seem t o have a significantly lower profitability t h a n those that d o not. In the next step several OLS-estimations are run, the results of which are shown in Table 3. Models one and four report the o u t c o m e for simple regressions. Here, the variables profitability and the logarithm of the value added per employee are regressed on a d u m Table 2 Differences in m e a n of c o m p a n i e s w i t h / w i t h o u t supervisory boards Codetermined firms

Non-codetermined firms

t-test on mean differences

Mean (Standard deviation)

Mean (Standard deviation)

(p-Values)

Codetermined firms

Noncodetermined firms

73,743 (96,741)

43,646 (54,377)

0.01

0.886

0.000

-.033 (1.166)

0.126 (0.135)

0.19

0.015

0.619

(%) Wages per employee (€)

40,686 (16,334)

25,585 (21,409)

0.00

0.914

0.000

Number of Enterprises

80

93

Value added per employee (€) Profitability

Kolmogorov-Smirnov-Test (p-Values)

448 · Franziska Boneberg

Table 3 Regression results for the enterprise performance of codetermined and not-codetermined firms on board level Performance indicator Exogenous Variable

Codetermination*

Logarithm of Productivity (Value added per Employee in €) Model Model Model (2) (3) (1)

Model (4)

Profitability (%) Model (5)

Model (6)

0.77 (0.000)

0.36 (0.006)

-0.18 (0.190)

Familiy-owned and codetermined enterprise* Sector I (transport, storage and communication)" Sector Κ 70-73 (real estate and renting)" 10.22 Constant (0.000) R-squared 0.179 170 Number of Enterprises'"

-0.16 0.31 (0.022) (0.193) -0.3 e-03 (0.047) 2.05 e-08 (0.056) -4.2 e-07 (0.936) -0.55 (0.002)

-0.15 (0.291) 0.2 e-03 (0.156) -1.31 e-08 (0.239) -0.3 e-04 (0.000) 0.05 (0.758)

0.81 (0.000) 1.02 (0.000) 10.01 (0.000) 0.365 170

0.72 (0.000) 0.98 (0.000) 10.36 (0.000) 0.413 170

0.13 (0.384) -0.11 (0.574) 0.11 (0.213) 0.02 173

0.16 (0.278) 0.13 (0.466) -0.07 (0.623) 0.17 173

Number of employees Number of employees squared Subsidies per employee (€) Familiy-owned enterprise'

0.13 (0.129) 0.01 173

Terms in brackets report the p-value. " Dummy-variable: 1 = yes, 0 = no. Reference category for industry dummies = Κ 74 (business activities). " For confidentiality reasons these values were dropped by the FDZ. "" Due to single computations handed in to the FDZ at a later state, unfortunately three cases were ex post dropped by the FDZ for confidentiality reasons. my variable reflecting the existence or non-existence of a supervisory board. 5 For productivity the R 2 demonstrates that 18 percent of the variance can be explained by the existence or non-existence of a supervisory board. Furthermore, at a significance level of one percent, companies with supervisory boards are 116 percent 6 more productive than firms without them. Looking at the results for profitability, only one percent of the variance can be explained by the presence or absence of a supervisory board. The regression coefficient of the dummy variable that indicates the presence or absence of a supervisory board is not statistically significant. In models two and five the regressions are augmented by 1-digit industry dummies 7 that indicate the sector in which a company is active. Hence, these variables test for industry5 6 7

In order to be able to interpret any changes in productivity and wages as marginal rates, regressions are run with the logarithm of these two variables. To facilitate interpretation, all estimated coefficients have been transformed by 100 (exp(ß) -1). The official business services statistics are comprised only of companies that act in branches I and K. Usually, 5-digit industry identifiers are reported; however, because of the small sample size and the resulting insufficient number of enterprises in single sectors, only 1-digit dummies could be generated for the present study while still preserving confidentiality.

The Economic Consequences of One-third Co-determination · 449

specific s t r u c t u r a l d i f f e r e n c e s a n d s h o c k s ( e . g . , t h e e x t e n t of c o m p e t i t i o n , t e c h n o l o g y of p r o d u c t i o n a n d f l u c t u a t i o n s in d e m a n d , a n d p r o d u c t i o n costs). T h e i n d u s t r i a l s e c t o r is significantly correlated to the value a d d e d per employee; according to the model, prod u c t i v i t y is 4 3 p e r c e n t h i g h e r in c o m p a n i e s w i t h s u p e r v i s o r y b o a r d s t h a n in t h o s e w i t h o u t . In a d d i t i o n , v a l u e a d d e d p e r e m p l o y e e is 1 2 5 p e r c e n t h i g h e r in sector I ( t r a n s p o r t , s t o r a g e a n d c o m m u n i c a t i o n ) , a n d 1 7 7 p e r c e n t h i g h e r in s e c t o r K 7 0 - 7 3 (real e s t a t e a n d r e n t i n g ) , c o m p a r e d t o c o m p a n i e s active in s e c t o r K 7 4 ( o t h e r business activities). H o w ever, t h e c o e f f i c i e n t s a r e n o t statistically s i g n i f i c a n t f o r p r o f i t a b i l i t y . F o r a r o b u s t n e s s c h e c k , m o d e l s t w o a n d five a r e a u g m e n t e d by a d d i t i o n a l c o n t r o l variables. First, t w o v a r i a b l e s i n d i c a t i n g c o m p a n y size - n u m b e r of e m p l o y e e s a n d t h e s q u a r e of t h e n u m b e r of e m p l o y e e s - a r e i n t e g r a t e d . F i r m size s e e m s t o i n f l u e n c e w h e t h e r t h e f i r m h a s a s u p e r v i s o r y b o a r d (see B o n e b e r g 2 0 0 9 a a n d t h e d e s c r i p t i v e statistics in T a b l e A l in t h e A p p e n d i x ) a n d m a y i n f l u e n c e p r o d u c t i v i t y a n d p r o f i t a b i l i t y as well. S e c o n d , a v a r i a b l e r e p o r t i n g s u b s i d i e s p e r e m p l o y e e received by a n e n t e r p r i s e is c o n t a i n e d in t h e e s t i m a t i o n . In t h e o f f i c i a l b u s i n e s s services statistics, subsidies p e r e m p l o y e e a r e d e f i n e d as a n y p a y m e n t s received f r o m local, r e g i o n a l , f e d e r a l or E u r o p e a n a u t h o r i t i e s , w i t h o u t c o n s i d e r a t i o n , in o r d e r t o l o w e r p r o d u c t i o n c o s t s a n d / o r prices of t h e g o o d s p r o d u c e d a n d / o r t o g u a r a n t e e s u f f i c i e n t p a y m e n t s f o r f a c t o r s of p r o d u c t i o n . T h e r e f o r e , subsidies s h o u l d be h i g h e r in f i r m s w i t h l o w e r p r o d u c t i v i t y a n d p r o f i t a b i l i t y . Finally, t w o a d d i t i o n a l d u m m i e s a r e i n c l u d e d i n d i c a t i n g w h e t h e r a f i r m is f a m i l y - o w n e d (and f u r t h e r m o r e c o - d e t e r m i n e d a t t h e e n t e r p r i s e level). Since a n earlier s t u d y by this a u t h o r h a s s h o w n t h a t t h e p r o b a b i l i t y of f a m i l y - o w n e d f i r m s ' h a v i n g a s u p e r v i s o r y b o a r d is l o w e r t h a n t h a t f o r c o m p a n i e s w i t h d i f f e r e n t o w n e r s h i p s t r u c t u r e s (see B o n e b e r g 2 0 0 9 a ) , a n d since f a m i l y - o w n e d f i r m s a r e o f t e n said t o a t t a c h g r e a t e r i m p o r t a n c e t o lasting c o n t i n u i t y t h a n t o m a k i n g p r o f i t s in t h e s h o r t r u n , t h e i n t e g r a t i o n of t h o s e t w o d u m m i e s m a y be of v a l u e . To begin w i t h m o d e l t h r e e , a p a r t f r o m t h e s u b s i d i e s p e r e m p l o y e e all o t h e r v a r i a b l e s i n c l u d e d in t h e e s t i m a t i o n a r e a p p a r e n t l y c o r r e l a t e d t o t h e v a l u e a d d e d per e m p l o y e e . Value a d d e d per e m p l o y e e is 3 6 p e r c e n t h i g h e r in c o m p a n i e s w i t h s u p e r v i s o r y b o a r d s t h a n in t h o s e w i t h o u t . T h e i n d u s t r i a l sector relates s i g n i f i c a n t l y t o t h e v a l u e a d d e d p e r e m p l o y e e , w h i c h is 1 0 5 p e r c e n t h i g h e r in sector I, a n d 1 6 6 p e r c e n t h i g h e r in s e c t o r s K 7 0 - 7 3 t h a n in s e c t o r K 7 4 . M o r e o v e r , a t a s i g n i f i c a n c e level of o n e p e r c e n t , p r o d u c t i v i t y is 4 2 p e r c e n t s m a l l e r in f a m i l y - o w n e d f i r m s . T h e c o e f f i c i e n t s of t h e n u m b e r of e m p l o y e e s a n d its s q u a r e d v a l u e a r e s i g n i f i c a n t , b u t s m a l l ; w h i l e a n a d d i t i o n a l w o r k e r r e d u c e s t h e v a l u e a d d e d p e r e m p l o y e e , a c c o r d i n g t o t h e s q u a r e d v a l u e , this r a t e d i m i n i s h e s t o a min i m u m f i r m size of a r o u n d 7 , 3 1 7 e m p l o y e e s , a f i r m size i r r e l e v a n t f o r t h e p r e s e n t study. T h e m a r g i n a l e f f e c t of a n a d d i t i o n a l w o r k e r o n v a l u e a d d e d p e r e m p l o y e e f o r c o m p a n i e s that employ between 550 and 2 0 0 0 workers a m o u n t s to 0.02 percent. Looking at the c o e f f i c i e n t s in m o d e l six, o n l y subsidies p e r e m p l o y e e relate t o p r o f i t a b i l i t y a n d t h e rel a t i o n s h i p is n e g a t i v e . N o o t h e r v a r i a b l e s a r e statistically s i g n i f i c a n t at a n y c o n v e n t i o n a l e r r o r level. T h e results s h o w t h a t p r o d u c t i v i t y is h i g h e r in c o - d e t e r m i n e d f i r m s t h a n in n o t - c o - d e t e r m i n e d f i r m s , s o it c o u l d be a r g u e d t h a t p r o f i t a b i l i t y s h o u l d a l s o be h i g h e r in these f i r m s . H o w e v e r , t h e results i n d i c a t e t h e o p p o s i t e . O n e possible e x p l a n a t i o n is t h a t t h e e m p l o y e e s a r e a b l e t o d e m a n d h i g h e r w a g e s b e c a u s e of t h e i r h i g h e r p r o d u c t i v i t y , a n d s u c h d e m a n d n e g a t i v e l y i m p a c t s a c o m p a n y ' s p r o f i t a b i l i t y . T h u s , it c o u l d be w o r t h e x a m i n i n g w h e t h e r t h e p r e s e n c e of a c o - d e t e r m i n e d s u p e r v i s o r y b o a r d is l i n k e d t o t h e w a g e levels of a c o m p a n y . T h e results of t h e t-tests a n d t h e O L S - e s t i m a t i o n s a r e pres e n t e d in T a b l e s 2 a n d 4, respectively.

450 · Franziska ßoneberg

Table 4 Regression results for the wage level of co-determined and not-co-determined firms on board level Performance indicator Exogenous Variable Codetermination* Number of employees

Model (7)

Logarithm of wage Model (8)

0.68 (0.000)

0.44 (0.000)

0.42 (0.000) -0.2 e-03 (0.069) 1.49 e-08 (0.105) -4.26 e-06 (0.347) -0.42 (0.005)

9.84 (0.000) 0.206 173

0.46 (0.000) 0.63 (0.000) 9.72 (0.000) 0.305 173

0.38 (0.001) 0.63 (0.000) 10.00 (0.000) 0.351 173

Number of employees squared Subsidies per employee (€) Familiy-owned enterprise" Familiy-owned and codetermined enterprise* Sector 1 (transport, storage and communication) " Sector Κ 70-73 (real estate and renting) * Constant R-squared Number of Enterprises"*

Model (9)

Terms in brackets report the p-value. " Dummy-variable: 1 = yes, 0 = no. Reference category for industry dummies = Κ 74 (business activities). " For confidentiality reasons these values were dropped by the FDZ. *** Due to single computations handed in to the FDZ at a later state, unfortunately three cases were ex post dropped by the FDZ for confidentiality reasons.

Both t-tests and the OLS estimates suggest a positive relationship between wage level and the presence of a co-determined supervisory board. According to the R 2 of model seven, 2 7 percent of the variance in wages can be explained by the presence or absence of a supervisory board; on a significance level of one percent, the wages are ceteris paribus higher in companies with supervisory boards than in companies without them. Running OLS-estimations with the same explanatory variables as before indicates that this relationship is constant. All together it appears that companies with co-determined supervisory boards tend not only to be more productive, but also to pay higher wages. This finding might explain why profitability does not increase as one would expect.

7

Discussion

Because of the very small sample size, it is not the aim of this study to provide comprehensive explanations, but only to compare percentage differentials concerning several control variables. Bartelsman and Doms (2000: 585 f.) point to the difficulty of explaining productivity differences. In this context Griliches and Mairesse (1990:221) say, „The simple production function model, even when augmented by additional variables and further nonlinear terms, is at best just an approximation to a much more complex and changing reality at the firm, product, and factory floor level." The same is true

The Economic Consequences of One-third Co-determination · 451

for profitability. T h e fact that productivity and profitability can only be determined as proxies additionally narrow interpretation. T h e results of both the t-test and the regression models indicate that productivity per employee is, on average, significantly higher in companies with supervisory boards than in companies without them. O n the other hand, at least the Kolmogorov-Smirnov test reports lower profitability in companies with supervisory boards than in those without them. Summing up, then, productivity appears to be higher and profitability likely lower in corporations that have a co-determined supervisory board. This result is congruent with the idea of Freeman and Lazear ( 1 9 9 5 ) , w h o claim that worker participation raises productivity as the employees put more effort into their w o r k , but lowers profitability as highly productive workers exert more influence on the distribution of a c o m pany's profitability. T h e branch of industry is apparently significantly correlated to productivity per employee. C o m p a r e d to companies active in K 7 4 (entries such as consultancy, accounting, cleaning companies and call center), firms in branches I (transport, storage and c o m m u n i c a tion) and Κ (real estate, renting and business activities) achieve greater productivity per employee. Furthermore, family-owned companies tend to achieve less value added per employee. This o u t c o m e is in line with the recent literature as for example Frick ( 2 0 0 4 ) shows that companies run by managing directors are on average more efficient than owner-managed firms. Subsidies per employee are, on average, negatively related to profitability, a reasonable finding since corporations with little profitability tend to receive more subsidies. T h e present study indicates that the presence o f a supervisory board with co-determination relates positively to productivity and wages, but not or only to a small degree to profitability. However, some limitations may constrain the reliability and generalizability of the results. First, the study is based on cross-sectional data. Even though the official business services statistics used here comprises panel data, its advantages c a n n o t be used in the present investigation because whether a firm has a supervisory board does - by construction - not change over time. So studying any fixed effects in order to control for unobserved heterogeneity at the firm level is not possible. Heterogeneity can be caused by differences in management quality, which may be influenced by the presence or absence o f a supervisory board. This correlation leads to biased estimates of the codetermination coefficient. Since it is not known whether management performance is better or worse in companies with supervisory boards, the direction o f the bias is indeterminate. Second, investigating causal effects on firm performance is not possible because of the special kind of data used here. In the model at hand, the presence or absence o f a supervisory board is exogenously determined and fixed, so its status does not change in the course of time. In consequence, working out treatment effects by dividing the firms into those that have been given the treatment (companies with a supervisory board) and those that have not (companies without a supervisory board) is not possible. 8 Finally, a problem arises from controlling for firms' self-selection into co-determination. In the present study whether a firm has a supervisory board or not is assumed as a given. However, the owners or managers o f a c o m p a n y may decide to implement a co-determined supervisory board because they anticipate positive effects on firm performance.

8

See Wagner ( 2 0 0 7 ) for an example of how to design an empirical study of that kind.

452 · Franziska ßoneberg

To determine if this is the case, a variable reflecting the decision in favour or against a supervisory board is needed that is not related to productivity or profitability. Because no such variable is present, an investigation of this kind is not practicable. Another problem of self-selection might occur concerning the abilities and „type of employee" of a firm. It is thinkable that highly productive jobseekers with higher levels of formal qualification select themselves in companies with a supervisory board. In that case it does not seem surprising that productivity and wages in these firms are higher. However, controlling for such a circumstance also is impossible. Although the study has limitations, the paper provides important evidence on the economic consequences of co-determination. The special kind of data appears to be more reliable because it facilitates a direct comparison of firms with and without co-determination at the enterprise level. The study shows that co-determination at the enterprise level is at least not negatively related to the two core performance indicators of productivity and profitability but indicates that productivity is positively affected by the presence of such an institution. This outcome is in line with the results of most recent studies; Wagner (2009) as well as Frick and Bermig (2009) find no positive or negative economic effects of co-determination on a company's productivity and profitability. Considering the research design and methodology of the present study, its findings are not surprising. As explained in section two, the position of the supervisory board in limited liability companies is especially weak. In addition, the companies observed in this empirical investigation fall only into the scope of the 2004 Third Part Act, so employees hold only a third of the seats on the supervisory board. In consequence, the employer always can decide in favor of his own ideas. This fact might be of special interest when it comes to the distribution of rents. Even if a co-determined supervisory board is implemented, it cannot automatically influence the actions of the management, which is why the profitability level in the present study may not be affected by the existence of a co-determined supervisory board. However, the from the better exchange of information and consultation in supervisory boards resulting raise in efficiency may lead to an increase in productivity. Additionally, the workers might feel more motivated by their participation rights to improve productivity. The increased employee satisfaction, as well as the implementation of democratic principles in the economic system, can, depending on the values and ambitions of society, lead to an improvement in social welfare. All together, the results of the present study suggest that the controversial debate about co-determination regulations in Germany and their implementation in common legal forms within the European Union is overstated.

The Economic Consequences of One-third Co-determination · 453

Appendix Table A1a Descriptive statistics for variables included in the regressions of Table 3 Part 1 : Profitability All companies Variable Profitability