Does ownership concentration in Germany result in better performance than
in the United States?
Bachelor Thesis
Amsterdam, June 29 2015
Name and student number: Thomas Appel - 10372539
Supervisor: Torsten Jochem, Ph.D.
Specialisation: Finance & Organisation
Field: Finance (corporate governance and performance)
University: University of Amsterdam
Faculty: Faculty of Economics and Business
Abstract
A lot of research has been done on the relation between ownership and firm performance with mixed results. Most research is based on a particular country or type of firms. In this paper however two countries are compared: Germany where ownership is on average concentrated and the US where ownership is more dispersed. With a panel dataset over the period 2006-2012 it has been investigated if ownership concentration has a significant impact in both Germany and the US. Secondly it has been investigated whether the effect of ownership concentration is different in Germany than in the US. In the regressions of both Germany and the US there was no significant effect of ownership concentration on performance. In the combined regression there is a significant effect that German firms benefit more from ownership concentration than US firms.
2 Statement of Originality
This document is written by Student Thomas Appel who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of contents
1 Introduction 4 2.1 Literature overview 5 2.2 Existing literature 6 2.3 Governance systems 7 3.1 Hypothesis 8 3.2 Methodology 8 3.3 Sample details 10 4.1 General statistics 12 4.2 Results 124.3 Interpretation and analysis 16
4.4 Summary of results 18
5 Conclusion and proposals for future research 18
References 21
4
Does ownership concentration in Germany result in better performance than
in the United States?
1.Introduction
Ownership structure is an important aspect of corporate governance. The debate about corporate governance is based on the separation between ownership and control (management) of corporations. “The goal of corporate governance is to support and ensure a sustainable value creation of a corporation for its shareholders” (Kraus, Britzelmaier, 2011, p. 327). An important aspect of corporate governance is ownership which consists of two parts: ownership concentration and ownership structure (identity). The concentration measures the power of shareholders to influence management, whereas the identity is more about the shareholder’s objectives. A bank owning 10% of the stock has more influence than a 10% block held by dispersed small-stake
investors. The identity can be divided for example in family, private investors, institutional investors, they all have their own incentive to invest (Thomson, Conyon, 2012, p. 123).
At this point there is a notable difference between German and US and UK (Anglo-Saxon) firms. In Germany ownership is on average more concentrated than in Anglo-Saxon countries. The German financial system is in general bank-orientated, German companies often have a good and sustainable relationship with their bank for financing investments. Banks also frequently own part of the shares and/or have a seat on the board. Apart from banks there are also often families and other shareholders who are important owners with large stakes (so called ‘block holders’, in the literature often defined as a 5% or 10% owner or more). The Anglo-Saxon companies on the other hand often have more dispersed ownership with numerous smaller shareholders and is more market-orientated (Thomson, Conyon, 2012, CH 11,13).
The goal of a good corporate governance system is to reduce information asymmetry between owners and managers and to ease monitoring and control (Edwards, Nibler, 2000, p. 239). The way in which this happens can be explained by the ownership structure: a small shareholder will have less incentive to monitor actively, because the costs are borne by the individual shareholder which may outweigh the benefits. There is also a free rider problem, because other shareholders gain from the efforts made by that particular shareholder. Managers can therefore more easily pursue their own interest at the shareholder’s expense compared to a situation in which there is more monitoring via the board (Edwards, Nibler, 2000, p. 240). This is more often the case in Germany because of the more concentrated ownership structure, an institutional investor owning a large stake has more incentive to monitor a board actively, because of the amount of money
5 involved. This more (better) monitoring may have a positive impact on performance, reduce agency costs and managerial discretion (Shleifer, Vishny, 1986).
In this paper I will investigate whether the ownership concentration has a positive impact on firm performance and whether German companies benefit more from their (more concentrated) ownership structure than American companies. The last aspect is interesting since most prior research focused on a particular country, type of firm or type of ownership. I will compare the German firms registered at the HDAX index with a sample of American companies registered at the S&P 500 index. In the next section I will give an overview of the existing literature and results of prior research on ownership and performance. In the third section I will describe the methodology and hypothesis. In the fourth section the results will be interpreted and in the fifth section I will summarize and conclude.
2.1 Literature overview
A lot of research has been done on corporate governance, and ownership in specific, and how it impacts firm performance. The results are mixed, before I will discuss that I will explain the topic of corporate governance and the effect of ownership on performance in theory. In recent years corporate governance has become one of the most discussed topics in business administration, an important reason for this are the balance sheet fraud cases of Enron, Worldcom, Ahold, etc
(Britzelmaier, Kraus, 2011, p. 327). Corporate governance is defined by Thomson and Conyon (2012, p. 4) as ‘the control and direction of companies by ownership, boards, incentives, company law, and other mechanisms’. There are several other definitions, but the central governance problem is about the agency problem caused by the separation of ownership and control as defined by Berle and Means (1932).
This separation can cause misalignment between the shareholder and the management. To make sure that the board is acting in a way that is aligned with the shareholder’s interests, the shareholder should monitor the board to reduce the agency problem. Monitoring is related to an important aspect of corporate governance which is ownership structure and concentration. Large equity ownership by managers or other ‘block holders’ may reduce agency problems and increase firm performance (Seifert, Gonenc, Wright, 2004, p. 173). Owning a large stake in a company will increase one’s incentive to monitor actively, because more money and voting power is involved. A financial institution owning 10% of the shares has in general more incentive and more financial wealth to monitor than many small individual shareholders holding the same 10%. For small shareholders the costs often outweigh the benefits of monitoring.
6 In some cases this holds to a certain level, because when for example managers own a big stake of the company they may feel entrenched and because of that not always pursue profit-maximization. They can for example pay themselves abnormal salaries and secure their own jobs even when they are not doing well. The same holds for financial institutions, they may also pursue other interests than those of the individual outside shareholders (Morck et al, 1988).
2.2 Existing literature
As mentioned, the existing literature shows mixed results to the question whether ownership positively contributes to firm performance. Most prior research is based on the Anglo-Saxon
companies and most literature is from the 1990’s and early 2000, but Lehmann and Weignand (2001) did research on German companies and they found a significant negative impact of ownership concentration on performance (measured as ROA). However when ‘location rights’ (the identity of ownership, for example banks or families, and their control rights) are taken into account the relationship is positive in case that financial institutions are the largest shareholders. The negative effect applies to family and ‘normal shareholder’ ownership.
Another study, from Gorton and Schmid (2000) focused on banks as block holders in industrial firms and their impact. This research showed a significant positive effect of the bank ownership.
As mentioned above when there is a positive effect of ownership concentration this may decrease after a certain level when influence gets too big to pursue profit maximization. Large equity ownership by managers for example may decrease their incentive to pursue profit maximization and pursue empire building and making private benefits more important. Because of the managers’ influence he fears less risk of being removed from his position. This is found by McConnell and Servaes (1990). They found that ownership by insiders between 0 till 40 to 50% has a positive effect on performance, thereafter the relationship gets negative. This is consistent with research done by Seifert, Gonenc, et al. (2004). In this paper performance was measured by Tobin’s Q and the ownership was particularly about insider ownership, so families and managers. Here it was also found that as an insider’s equity increases, firm performance also increases, but at high levels of ownership performance decreases again.
The focus on insider ownership was also a point of discussion in the research done by Han and Suk (1998), who concluded, based on a sample between 1988-1992, that insider ownership is positively related to stock performance. They also found that institutional ownership has a significant positive effect on corporate performance showing the ability to monitor actively. They also found a non-linear relationship between ownership and performance.
7 Others like Demsetz and Lehn (1985) and Demsetz and Villalonga (2001) found no relation between ownership and performance. Their explanation for this is that the optimal ownership varies per firm and in the end the structure is the outcome of optimal decisions made by all shareholders all having their own interests. Most of the existing literature however shows a relation, but not with a clear cut answer whether this is positive or negative and significant, but in many cases the
involvement of institutional ownership seems to have a positive effect on performance.
2.3 Governance systems
The literature tries to categorize the corporate governance systems, a well-known classification is the bank vs. market centred system. These two types of systems are defined by Kaplan (1996). The bank centred system or relationship system is common in Germany and Japan; here banks, families and large corporate shareholders own large stakes in the company and these relationships carry out the monitoring and disciplining. The market based system is common in the Anglo-Saxon countries: the United States and United Kingdom. Here ownership is more dispersed and the involvement of shareholders in corporate governance matters is smaller and companies rely more on equity to finance their investments, whereas in Germany and Japan debt is more common.
These theoretical corporate governance topics have also been investigated, for example by Franks and Mayer (1995), who found that in 1990, 85% of the largest German listed companies had a single shareholder owning at least 25% of the shares. This number was compared to the United Kingdom where it was only 16% of the largest listed firms. So in most large German firms there is not a problem of interest between dispersed shareholders and entrenched managers, but there can be a conflict of interest between controlling and minority owners. Controlling owners can abuse their power at the expense of minority shareholders (Edwards, Nibler, 2000, p. 242). The Franks’- and Mayer research is quite old and in the late 1990’s and early 2000’s a change of bank ownership in Germany began to take place. The bank ownership stake has been reduced from 4.7% in 1997 to 0.6% in 2006 and also their share in the supervisory board fell from 7.6% to 4.3% in the same period (Rapp et al., 2009, in: Thomson, Conyon, 2012, p. 242). But the data for my research (see fourth section) still show a very concentrated ownership in Germany, but the involvement of banks seems to get smaller compared to the past. An explanation for this is the German tax reform in 2000, this reform abolished the tax on capital gains that banks had to pay when they wanted to sell their shares of a company. Before 2000 capital gains were taxed at 52%, this meant that banks would have a huge tax burden in case they wanted to divest, this tax forced them to keep their shares in other
8 3.1 Hypothesis
As described there is not a clear relationship between ownership and performance, and the effect can depend on a lot of other things, like the type of owner, legal system, firm differences and regulation. Based on Gorton and Schmid (2000) my intuitive hypothesis is that there will be a positive effect, since the presence of one or several large shareholder(s) may reduce the agency problem, because that shareholder has more incentive and opportunities to monitor the board. Because of the other factors besides ownership concentration it is hard to judge whether German firms benefit more from ownership concentration than German firms. I expect however a difference between the countries, but considering the mixed results of past literature and the fact that the effect can depend on more things than just the concentration my hypothesis will be two-sided.
The hypothesis will be as follows:
H1: There is an effect of ownership concentration on firm performance (ROA).
H2: Ownership concentration affects the performance of German firms differently than American firms.
3.2 Methodology
To answer my research question an empirical analysis is needed. I will compare the
companies of the German HDAX index with a sample of companies from the American S&P500 index in the period 2005-2012. There are several measures to define firm performance, according to Lehmann and Weigand (2001) the preferred measure is ROA (return on assets), so I will use that as well. The variables frequently used in the literature of industrial organization to determine
profitability are: ownership concentration (OC), absolute firm Size (S), firm growth (G), capital intensity (K), and capital structure (C) (source: German Statistical Office, in: Lehmann and Weigand, 2001, p. 172). The control variables size, growth and capital structure are used in many articles on firm performance and ownership worldwide, for example by Bebczuk (2005) and Corstjens et al. (2006). Capital intensity is less frequent, but also included in some articles, for example by
Himmelberg et al. (1999) in a research on the link between performance and ownership in a random sample of 600 companies available in Compustat.
In the article of Lehmann and Weigand (2001) and in several others the ownership
concentration is expressed as the Herfindahl index. There are other definitions too and because of the limited data availability, especially historic data are not well specified, I defined ownership concentration as the sum of the three largest shareholders in terms of direct or total ownership. As mentioned earlier the effect may decrease after a certain point (not found in all studies), to check
9 this I included a squared-ownership term. To be able to see differences between ownership
structures I indicated the type of largest shareholder to be able to group them. This gives also the opportunity to compare this to the earlier discussed article by Franks and Mayer (1995). I retrieved the ownership data from the Orbis database and manually calculated the ownership because the data were in an unpleasant format in Excel (4 values in one cell) and also because that gave me the opportunity to check for irregularities. Another reason for this approach is that 40% held by ‘other investors’ is not a single investor and should therefore not be included in the sum of the largest shareholders.
The other control variables are firm specific financials which I retrieved from the Bureau van Dijk Amadeus database for the German companies and from Compustat for the American
companies. The first one is firm size (S), this is defined as the natural logarithm of total assets. The second is firm growth (G), in Lehmann and Weigand (2001) it was defined as the logarithmic annual change in turnover, I decided to use the annual percentage change in turnover to avoid that the numbers lie very close to each other and do not show significant changes. The third variable is the capital intensity (K), which is the log of total assets divided by the number of employees. The fourth variable is the capital structure (C), which is defined as the total shareholders’ funds (equity) divided by total capital (shareholder funds + liabilities). Then I also included variables for the index return, because that may be correlated with the total return on assets. In case of the German HDAX listed companies I took the monthly stock index returns of the HDAX and calculated the average per year. For the American S&P500 companies I did the same with the S&P500 index.
To see if the effect of ownership concentration is different for German companies than for American companies an interaction term will be included: ‘German company*ownership
concentration’. German will be a dummy variable which equals 1 if a firm is listed in the HDAX. For the German companies I was able to download the total return on assets as a percentage, of the American companies this was not that clear and the values I found were calculated using different numbers. Therefore I checked how the German ROA was calculated and in the same way I calculated the ROA of the US firms, which was: 𝑝𝑟𝑜𝑓𝑖𝑡 𝑏𝑒𝑓𝑜𝑟𝑒 𝑡𝑎𝑥𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠
All this results in the following model 1:
𝑅𝑂𝐴𝑖𝑡 = β0 + β1Ownership concentrationit+ β2Sizeit+ β3Growthit+ β4Capital intensityit + β5Capital structureit+ β6Return stockindexit+ εit
This regression is a panel regression which I ran in Stata with the “cluster” command, where the firms are clustered, because each firm has several observations and these are not independent. Firm is the ID and the years the time variable.
10 The overall model of all companies looks similar, but with the interaction term included to control for the difference between the two countries, model 2:
𝑅𝑂𝐴𝑖𝑡 = β0 + β1Ownership concentrationit+ β2Sizeit+ β3Growthit+ β4Capital intensityit +β5Capital structureit+ β6Return stockindexit+ β7(country ∗
ownership concentration)it+ β8Countryi+ εit
As mentioned the ownership data are not are not very easy to obtain, apart from this several financials of the control variables were also missing. At the German companies this happened randomly, sometimes data on ownership were missing, in other cases data about turnover or employees, etc. At the American companies the problem was only in the capital structure
information. Companies with too many missing data are dropped from the sample, in case only the first or first 2 years were not complete I kept the firm in the sample, but without those years. The minimum period therefore is 2007-2012. Another thing I did to avoid that too many firms would have been dropped from the sample was estimating the capital structure or intensity in case they were missing. When the capital structure or intensity of a year in the middle of the data was missing I calculated the median of the year before and the year after. In case only the first year was missing I took the average of the next years, if complete, to determine the capital structure or intensity. Note: this has only been done when 1 year was missing and only at the relatively stable variables, i.e. capital structure and intensity. The turnover change is too volatile to estimate, so years with missing values regarding turnover were dropped. The same applies for missing ownership data, those years or companies are also dropped.
3.3 Sample
The companies I investigated are the listed German companies included at the HDAX index. This index consists of the DAX, MDAX and TecDAX and therefore provides a broad picture of the German industry. Also data of listed companies are better available and financials are generated according to general accounting standards. For the American companies I used the S&P 500 firms. Because I calculated the ownership manually I took a sample of the S&P500 index based on industry and size and not all the firms of the S&P500. I wanted to use the data between 2004 and 2012 to have a long period including the times of economic growth and the financial and economic crisis. For the financial data this is not a problem, but the ownership data of the German companies were only available from 2005 onwards and of most of the American companies from 2006 onwards. The German panel regression is therefore based on the period 2005-2012 and the American on
2006-11 2012. The combined regression is based on the period 2006-2012 (with some firms starting in 2007, 17 in total, equally (9 and 8) divided between German and US firms), since most American firms did not have ownership data of 2005. The firms of the S&P500 also provide a broad picture of the American industry, but this of course differs from Germany, for example the S&P500 contains oil companies and a weapon manufacturer. To avoid big industry differences I composed the sample of the American firms on the industries covered by the German HDAX. I first looked up information about all the individual companies of the HDAX and indicated them with a number for each industry, so I used the HDAX as a starting point. Then I knew how much companies were in each industry. With this list I went through the complete list of the S&P500 firms to compose a sample with
approximately the same number of companies in each industry. I also checked the market
capitalization of both the German and US firms and made sure that the differences were not very big. In brief I tried to make the sample of the S&P 500 to be as similar as possible to the HDAX in terms of industry, this has been done via internet research (investing.com and company websites). Because of data unavailability I had to drop several firms of both countries. The original distribution of the companies from both countries is shown in the following table, the numbers of the final distribution are shown between brackets.
Industry HDAX S&P500
Banking 8 (1) 9 (6) Clothing 3 (1) 3 (2) Communication 10 (8) 5 (4) Aerospace 1 (0) 2 (2) Machine/industry building (capital goods) 14 (12) 13 (8) Consumer products 7 (5) 12 (9) Media 1 (1) 1 (1) Chemical, biomedical 23 (19) 12 (8) IT 13 (11) 11 (6) Real Estate 5 (2) 5 (4) Electronics 2 (1) 0 (0) Automotive 4 (4) 5 (3) Energy 7 (5) 8 (2) Construction 6 (6) 3 (2) Private Healthcare 1 (1) 1 (1) Tourism 1 (1) 1 (1) Aircraft 3 (3) 1 (1) Transport 0 (0) 3 (1)
The total sample of Germany consists of 110 companies, and 95 from the S&P500. Because of missing data both have got a bit smaller 81 German companies and 61 American companies. So a total sample of 142. The total list of firms can be found in the appendix.
12 4.1 General statistics
For the panel regressions I tried to include most of the companies of the sample, that means that from some companies one or two years are missing, so instead of dropping a company from the sample it is regressed based on the period 2007-2012 for example. In case more years were missing, i.e. the first observation is from 2008 or later, I dropped the firm. First some general information will be presented about both countries, based on the samples where all data are available.
Germany 81 firms 61 firms
Average % held by largest 3 shareholders
41.75% 24.82%
Average % held by single largest shareholder
26.31% 12.12%
Type largest shareholder most companies
C: Trade & Industry organisation. If this type is ignored: family ownership is most frequent
E: Mutual & Pension fund / Nominee / Trust
% firms having 1 owner with 25% or more 2012
43.04% 3.28%
Average ROA 6.27% 7.23%
Average ln(total assets) 21.88 16.51
Average % change turnover +4.086% +4.93%
Average capital intensity 0.00575 0.0005943
Average capital structure 0.4091 0.3431
Average yearly return stock-index 2006-2012
0.5768% 0.2753%
4.2 Results
First I ran regressions of the individual countries as described by model 1:
𝑅𝑂𝐴𝑖𝑡 = β0 + β1Ownership concentrationit+ β2Sizeit+ β3Growthit+ β4Capital intensityit + β5Capital structureit+ β6Return stockindexit+ εit
All regressions have been run with the “cluster” command in Stata in order to cluster the standard errors on the firm level and to take into account the fact that one firm has several observations and that these are thus not independent from each other. The coefficients of both (individual)
13 regressions are shown in the table below with the robust standard errors between brackets. Previous literature does not have a clear answer to whether the relation between firm performance and ownership concentration is linear. As discussed in some articles is was found that the effect of ownership concentration decreases after a certain point (caused by empire building and
management entrenchment). To control for this potential non-linear effect I added an ownership squared term. The first 2 columns show the linear model, the second 2 the non-linear model.
ROA Coefficients Germany - Linear Coefficients USA - Linear Coefficients Germany – non-linear Coefficients USA – non-linear Size 0.0018 (0.0035) -0.0126** (0.0052) 0.0018 (0.0034) -0.0135** (0.0054) Turnover 0.00014*** (0.00004) 0.1232*** (0.0379) 0.00014*** (0.00004) 0.122*** (0.0376) Capital intensity -0.2314** (0.0976) -8.6012* (4.356) -0.236** (0.0981) -8.682* (4.345) Capital structure 0.2118*** (0.0716) 0.1241*** (0.0361) 0.2115*** (0.0707) 0.1209*** (0.0358) Ownership concentration 0.0519 (0.0385) -0.0254 (0.0295) -0.0612 (0.1209) -0.1229 (0.1363) Average return Hdax/S&P500 0.0805 (0.237) 0.2881 (0.2375) 0.0818 (0.2387) 0.2815 (0.2376) Constant -0.0804 (0.0891) 0.242** (0.0962) -0.0652 (0.087) 0.2733** (0.259) Ownership concentration squared 0.1148 (0.1466) 0.1131 (0.1064) R-squared 0.1236 0.2541 0.1289 0.256 Root MSE 0.10031 0.0776 0.10009 0.0775 Number of observations 616 416 616 416 Number of groups 81 61 81 61 F 54.82 8.42 50.14 7.23
*** significant at 1% level, **significant at 5% level, *significant at 10% level
These results imply that there is no significant association between ownership concentration (measured as the sum of the largest 3 shareholders) and return on assets (ROA) for firms in the HDAX index. In the American regression the ownership concentration is not significant either, apart from this the control variable natural logarithm of total assets is significant compared to Germany. Capital intensity is significant at the 10% level, whereas it is significant at the 5% level in Germany. Also the American regression provides a higher R-squared than the German regression, so the model explains more of the variation in performance for the American firms than for the German firms. The squared ownership concentration term is not significant in both countries, so based on this sample and time frame it cannot be concluded that there is a non-linear effect of ownership concentration on
14 to insider ownership, so this is also dependent on the type of companies and type of ownership in the sample of interest.
The most interesting question is however whether the effect of ownership concentration is different in Germany compared to America. To find this out I ran several regressions based on the second model:
𝑅𝑂𝐴𝑖𝑡 = β0 + β1Ownership concentrationit+ β2Sizeit+ β3Growthit+ β4Capital intensityit +β5Capital structureit+ β6Return stockindexit+ β7(country ∗
ownership concentration)it+ β8Countryi+ εit
First I put all the data into one file and then indicated German firms (firms listed in the HDAX) as 1 and American firms (listed in the S&P500) as 0, so a dummy variable for country. In Stata I defined an interaction term: “sum3xgerman” as the product of “sum3largest” and the German firm indicator. With this interaction term included it can be found out if the effect of ownership concentration is different in Germany compared to the US.
The model has been run with several variables and effects. These will be explained and then the table with all results follows. All regressions have been run with the “cluster” command in Stata in order to cluster the standard errors on the firm level and to take into account the fact that one firm has several observations and that these are thus not independent from each other.
The results are shown in the table on the next page. I ran several regressions, first a basic panel regression with only the ownership concentration variables and the German indicator. In the next model I included all the control variables. Then in model 3 and 4 I did the same but I added fixed year effects to control for business cycles and other year specific events that are not captured by the model. In model 5 and 6 I also added fixed group effects to see if there are differences between groups, these groups are sorted based on the type of the largest shareholder (bank, family, pension fund, etc). The models mentioned in the table are variants of model 2 that I described earlier (with the interaction effect), based on the period 2006-2012.
15 Coefficients of all models with robust standard errors between brackets.
ROA Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
German -0.0375* (0.0206) -0.0371 (0.0236) -0.0402* (0.0207) -0.0377 (0.0238) -0.0457** (0.022) -0.0356 (0.0249) Sum3largest shareholders -0.0524 (0.0408) -0.033 (0.0311) -0.0609 (0.0404) -0.0402 (0.0309) -0.0821* (0.0444) -0.0409 (0.0351) Sum3xgerman 0.0859 (0.0599) 0.0753* (0.0489) 0.0953 (0.0598) 0.0832* (0.0488) 0.1027 (0.0633) 0.082* (0.0499) Ln (total assets) -0.0017 (0.0033) -0.0019 (0.0034) -0.0024 (0.0036) Turnover 0.00014*** (0.000032) 0.00013*** (0.0000364) 0.00013*** (0.000039) Capital intensity -0.2492** (0.113) -0.2666** (0.1092) -0.2572** (0.1006) Capital structure 0.1658*** (0.0374) 0.1637*** (0.0378) 0.1651*** (0.0379) Index return 0.1373 (0.1808) -0.0888 (0.1578) -0.1032 (0.1671) Constant 0.0853*** (0.0128) 0.0522 (0.063) 0.1079*** (0.0143) 0.0758 (0.0632) 0.1155*** (0.0253) 0.09 (0.0671) Fixed year effects
No No Yes Yes Yes Yes
Fixed group effects (type largest shareholder) No No No No Yes Yes R-Sq 0.0089 0.1377 0.036 0.1602 0.0417 0.1649 Prob > F 0.282 0 0 0 . . Number of observations 968 968 968 968 967 967 Number of firms 142 142 142 142 142 142
16 4.3 Interpretation and analysis
Ownership concentration in general is in most cases not significant, only in model 5 it is significant at the 10% level. Furthermore the coefficient of size (natural logarithm of total assets) is not significant in any of the models, whereas it was in the American only regression. The turnover change, capital intensity and capital structure are in all models, in which they are included, significant at 5% and sometimes at 1%. This makes sense since the performance of a company is largely
influenced by turnover (sales) and capital structure has an impact on interest payments and taxes and therefore on profitability.
The coefficient of German should be interpreted as the average difference in ROA between German firms and American firms. In model 1 for example the coefficient German was -0.0375 and significant (at 10%), which means that German firms have, in the time frame of the sample, annual ROA’s that are on average 3.75% lower than those of American firms.
The above models all include an interaction term to see if the effect of ownership concentration is different in Germany compared to the US:
β(German ∗ ownership concentration)
Model 1 is the basic model in which I only looked at the impact of ownership concentration and whether a firm is German on performance. This resulted in a low R-sq of 0.0089 and also the interaction term, sum3xgerman, is not significant. β2 (sum3largest) measures how much ROA changes as a consequence of a 1% increase in the sum of shares of the largest 3 shareholders (ownership concentration). β3 (the interaction term) measures how much the effect of ownership concentration on performance differs when a firm is German. The effect of the interaction term can be seen by taking the derivative of ROA to ownership concentration.
𝜕𝑅𝑂𝐴𝜕𝑂𝐶 = β1 + β3German
In model 1 this equals: -0.0524 + 0.0859(*1) Because β3 is positive, the effect of ownership concentration is higher for a German firm than for an American firm. So the effect of a change in ownership concentration on performance depends no longer just on ownership concentration, but also on whether a firm is German or not. The positive value of β3 is consistent with the theory that the presence of large owners results in better performance because of better monitoring.
Model 1 does not a good job in predicting ROA, but the explanation about the interaction term and German variable apply also to the other models. In model 2 the other control variables are added and this resulted in a coefficient of 0.0753 for sum3xgerman which is significant at 10% in contrast to model 1. Model 2 has a R-sq of 0.1377, so it predicts more of the variation in ROA than model 1 does which is logical since most of the added control variables have significant impact.
17 In model 3 and model 4 I have been running the same regressions as those of model 1 and 2, but I added fixed year effects to control for business cycles and other year specific events that are not captured by the model. These fixed year effects control for the average differences across years for observable or unobservable predictors, so effects that influence ROA within a particular year that are not covered by the model. The financial crisis in 2008-2009 for example will have had an impact on (almost) all firms in the sample in those years, but that effect is not explained by the model. In model 4 sum3xgerman is significant at the 10% level and in model 3 German is significant at the 10% level. Model 4 has a higher R-sq, so the rest of the analysis will be based on model 4. The coefficient of sum3xgerman in model 4 has a value of 0.0832 and is significant at 10%, so German firms benefit more from ownership concentration than American firms, measured by ROA. When ownership concentration increases by 1% the effect on performance is on average 0.0832% higher in Germany than in the US. The value of German equals -0.0377, as I already discussed in the analysis of the first two models, this means that ROA in this sample and time frame is on average 3.77% lower in Germany, but this is not significant. In model 4 two years have a significant coefficient (at 5% level):
ROA Coefficient Standard error t P>|t|
Year 2008 -0.0371 0.0105 -3.53 0.001
Year 2009 -0.0432 0.0094 -4.59 0.000
This means that in the years 2008 and 2009 ROA was lower by 3.71% and 4.32% on average. This clearly shows the effect of the financial crisis, which cannot be covered by the standard regression model. The other year effects were not significant.
In model 5 and 6 I also added fixed group effects to model 3 and 4. In Stata I sorted the groups based on the type (identity) of the largest shareholders, these types are indicated in the Orbis database and there are several ones. The list contains the following types: insurance company, bank, trade & industry organisation, mutual & pension fund, financial company, family, foundation or research institute, employees or managers (insiders), private equity, venture capital and hedge funds. The ‘public’ and ‘nameless private investors’ are also on the list of types, but I did not include them, because these are groups of people and not a single person or institution. The list of which type belongs to which group number can be found in the appendix. The interpretation of the fixed group coefficient is the same as with the fixed year effects, but then for the specific group, so they control for average differences between groups. The coefficient of _Igr2 (largest owner is a bank) for example equals 0.02 in model 6, this means that companies with a bank as largest owner have on average a 2% higher ROA which may be the result of better monitoring. But none of the group effect coefficients are significant in both model 5 and 6, so this conclusion cannot be drawn with certainty. A reason for the fact that none of the fixed group effect show significance may be because the
18 samples get smaller and also because of changes in some years. Ownership is on average quite constant per firm, but my sample covers 6 or 7 years of most companies and then you see several changes within firms. Sometimes the differences are very small, when a bank owns 15% in 2008 and a family relative 14% and a year later the same person of the family owns 16% and the bank still 15% that year belongs to another group, although the change in ownership was relatively small.
Unfortunately the data are not specified enough to determine who has final control rights, the largest owner is not automatically the most important one.
Considering the above analysis model 4 seems the best explanation for ROA, the R-sq equals 0.1602, which is a little bit lower than the R-sq of model 6, which is 0.1649, but the fixed group effects of model 6 are not significant in any of the groups.
4.4 Summary of results
Based on the individual regressions per country there seems no significant effect of ownership concentration on performance in both Germany and America. When both countries are combined into one regression including an interaction term to distinguish the effect on German firms from American firms ownership concentration is still not significant, but there is a significant effect (at the 10% level) between Germany and America. The interaction term sum3xgerman is significant at 10% and it is positive (0.0832), this means that German firms benefit more from ownership
concentration (0.0832% higher ROA than American firms when ownership concentration increases by 1%), which is consistent with the theory of ownership concentration and its effects on corporate governance. The coefficient of German is negative, but not significant in the complete models, so it cannot be said with certainty that ROA is significantly lower in Germany than in America, although the average of the sample shows a 0.97% difference.
5 Conclusion and proposals for future research
In this paper I looked at the influence of ownership concentration on firm performance. Ownership concentration is an important aspect of corporate governance and there exist significant differences in the financial systems over the world. In Germany ownership is historically
concentrated, many companies have block holders owning a significant part of the shares, these are often banks, financial institutions and families. In America ownership is more dispersed and firms are more dependent on the market to attract funds. The data of the sample investigated show that this still holds, the average share of the largest shareholder in Germany is 26.31% and in the US 12.12%. By running several regressions I checked whether ownership concentration has impact on firm performance measured as ROA. In theory more concentrated ownership will increase firm
19 performance because there is less of an agency problem between shareholders and managers. In the regressions per country ownership concentration proved not be significant, but by including an interaction term to distinguish German from American companies to see if ownership has a different effect on German companies than on American firms there is a significant effect (at 10% significance level) that German firms perform better due to ownership concentration. This difference can have many causes, for example differences between the countries in terms of regulation, law, general (firm) differences, economic developments and much more.
However this research has some limitations, first of all the quality of the ownership data. The data available via Orbis show the percentages of ownership and to which type they belong, but sometimes values are missing and most important; based on the numbers it cannot be said who has final control. Owning shares does not automatically mean that someone is committed to a company. Therefore ownership concentration might not be the best way to express ownership, controlling for the identity of owners may be a solution, which I did in the last regression by grouping the firms on their type of largest shareholder. But as already mentioned the samples get sometimes very small then and apart from this there are sometimes odd changes in the type of largest owner between years. Sometimes in a particular year of a family owned company for some or the other reason a bank may be the largest owner in that year. This does not automatically imply that the ownership structure has changed significantly, but because the data do not specify this it is handled as if that year belongs to another category. But more important is the fact that that are a lot of other explanatory factors for firm performance, either observable or unobservable. As mentioned there will be general differences between the countries in terms of regulation and law. There may also be reverse causality, which means that ownership concentration may have influence on performance but also the other way around. For example when a firm is doing well more shares may be in hands of insiders because less money is needed from the public. This reverse causality is for example assumed by Kole (1994), in this article it is argued that when a firm performs well managers want to be compensated with shares. However, others like Morck et al. (1988) did not control for potential reverse causality and assumed that causality only runs from ownership to performance. The literature has not a clear answer to the degree of this problem. If there is an endogeneity problem, Two Stage Least Squares estimation may be a better solution than OLS.
Most prior research focused either on German companies or (in most cases) Anglo-Saxon firms, this research compared both and showed that ownership concentration in Germany has a larger effect on performance compared to America.
Whether ownership concentration has real impact on performance is still a question without a clear answer, so more research to this is needed, especially on the control rights. Another research
20 that would be interesting is to look specific at the effect of the 2000 German tax reform on
performance and ownership concentration (von Beschwitz, Foos, 2013, p. 6). Banks sold of much of their shares in other companies, so the classical bank-orientated governance in Germany seems to get less important, but ownership in general is still very concentrated compared to American firms.
21 Reference list
Bebczuk, R.N. (2005). Corporate Governance and Ownership: Measurement and Impact on Corporate Performance and Dividend Policies in Argentina. Working paper R-516, Latin American Research Network.
Berle, A., Means, G. (1932). The modern corporation and private property. New York: MacMillan.
Von Beschwitz, B., Foos, D. (2013). The causal effect of banks’ equity stakes on their lending. Working paper.
Corstjens, M., Peyer, U., Van Der Heyden, L. (2006). Performance of Family Firms: Evidence from US and European firms and investors.
Demsetz, H., Lehn, K. (1985). The structure of corporate ownership: causes and consequences. Journal of Political Economy, 93, 1155–1177.
Demsetz, H., Villalonga, B. (2001). Ownership structure and corporate performance. Journal of Corporate Finance, 7, 209–233.
Gorton, G., Schmid, F. (2000). Universal banking and the performance of German firms. Journal of Financial Economics, 58, forthcoming.
Edwards, J., Nibler, M. (2000). Corporate governance in Germany: The role of banks and ownership concentration. Economic Policy, 31, 239–267.
Franks, J., Mayer, C. (1995). Ownership and Control, in: Edwards, J., Nibler, M. (2000). Corporate governance in Germany: The role of banks and ownership concentration. Economic Policy, 31, 239– 267
Han, K.C., Suk, D.Y. (1998). The Effect of Ownership Structure on Firm Performance: Additional Evidence. Review of Financial Economics (1998), Vol. 7, No. 2, 143-155.
22 Himmelberg, C. P., Hubbard, R.G., Palia, G. (1999). Understanding the Determinants of
Managerial Ownership and the Link Between Ownership and Performance. The Journal of Financial Economics, 53, 353–384.
Kaplan, S.N. (1996). Corporate governance and corporate performance: comparison of Germany, Japan, and the U.S. Bank of America Journal of Applied Corporate Finance, 9 (4), 86–93.
Kole, S. (1994). Managerial Ownership and Firm Performance: Incentives or Rewards? Working Paper 93–10, University of Rochester, Rochester, New York.
Kraus, P., Britzelmaier, B. (2011). CORPORATE GOVERNANCE AND CORPORATE PERFORMANCE: A GERMAN PERSPECTIVE. International Journal of Management Cases, 2011,13 (3), 327-340.
Lehmann, E., Weignand, J. (2000). Does the Governed Corporation Perform Better?
Governance Structures and Corporate Performance in Germany. European Finance Review, 4, 157– 195.
Morck, R., Shleifer, A., Vishny, R.W. (1988). Management ownership and market valuation: an empirical analysis. Journal of Financial Economics, 20, 293–315
Rapp, M., Schwetzler, B., Sperling, M. (2009) The Disappearing Deutschland AG – An Analysis of Block Trade in German Large Caps, in: Thomson, S., Conyon, M. (2012). Corporate Governance Mechanisms and Systems. Berkshire: McGraw-Hill Higher Education.
Seiferta, B., Gonenc, H., Wright, J. (2004). The international evidence on performance
and equity ownership by insiders, blockholders, and institutions. Journal of Multinational Finance Managment, 2005, 15, 171–191.
Shleifer, A., Visnhy, R.W. (1986). Large Shareholders and Corporate Control. Journal of Political Economy, 94 (3, Part 1 (Jun 1986)), 461-488.
Thomson, S., Conyon, M. (2012). Corporate Governance Mechanisms and Systems. Berkshire: McGraw-Hill Higher Education.
23 Appendix
Fixed group effects sorted by type.
Group Type _Igr_2 B: Bank
_Igr_3 Bank together with another type (same stakes) _Igr_4 C: Trade & Industry organisation
_Igr_5 Trade & Industry organisation with another type (same stakes) _Igr_6 D: Private investor
_Igr_7 E: Mutual & Pension fund / Nominee / Trust / Trustee
_Igr_8 Mutual & Pension fund / Nominee / Trust / Trustee with another type (same stakes) _Igr_9 F: Financial company
_Igr_10 H: Self ownership
_Igr_11 I: Family (one or several relatives) _Igr_12 Family with another type (same stakes) _Igr_13 J: Foundation / Research Institute _Igr_14 M: Managers/employees
_Igr_15 P: Private Equity firms
_Igr_16 S: Public authority/State/Government _Igr_17 V: Venture Capital
24 Firms of final sample (after dropping ones with missing data)
Germany
USA
ADIDAS AG APPLIED MATERIALS INC
ADVA OPTICAL NETWORKING SE AUTOMATIC DATA PROCESSING CARL ZEISS MEDITEC AG AVERY DENNISON CORP
AIXTRON SE BARD (C.R.) INC
BAYER AKTIENGESELLSCHAFT BAXTER INTERNATIONAL INC BECHTLE AKTIENGESELLSCHAFT BOEING CO
BEIERSDORF AKTIENGESELLSCHAFT CSX CORP BAYERISCHE MOTOREN WERKE AKTIENGESELLSCHAFT CA INC
CELESIO AG MOLSON COORS BREWING CO
CANCOM SE DELTA AIR LINES INC
CONTINENTAL AKTIENGESELLSCHAFT DOVER CORP
COMPUGROUP MEDICAL AKTIENGESELLSCHAFT DU PONT (E I) DE NEMOURS
DAIMLER AG PERKINELMER INC
DEUTSCHE BOERSE AKTIENGESELLSCHAFT EATON CORP PLC
DEUTSCHE POST AG EQUIFAX INC
DRILLISCH AKTIENGESELLSCHAFT FORD MOTOR CO DRAEGERWERK AG & CO. KGAA GENUINE PARTS CO
DEUTSCHE TELEKOM AG HASBRO INC
DUERR AKTIENGESELLSCHAFT HEWLETT-PACKARD CO
E.ON SE STARWOOD HOTELS&RESORTS WRLD
EVONIK INDUSTRIES AG ILLINOIS TOOL WORKS
EVOTEC AG INTERPUBLIC GROUP OF COS
FRESENIUS MEDICAL CARE AG & CO. KGAA JOHNSON CONTROLS INC
FREENET AG KIMBERLY-CLARK CORP
FRAPORT AG FRANKFURT AIR- PORT SERVICES
WORLDWIDE LEGGETT & PLATT INC
BILFINGER SE LEGG MASON INC
DMG MORI SEIKI AKTIENGESELLSCHAFT LOCKHEED MARTIN CORP
GERRESHEIMER AG MASCO CORP
HEIDELBERGCEMENT AG CVS HEALTH CORP
HENKEL AG & CO. KGAA MICRON TECHNOLOGY INC HOCHTIEF AKTIENGESELLSCHAFT TENET HEALTHCARE CORP INFINEON TECHNOLOGIES AG NEWMONT MINING CORP JENOPTIK AKTIENGESELLSCHAFT NIKE INC
KONTRON AG NORDSTROM INC
KLOECKNER & CO SE NUCOR CORP
KRONES AKTIENGESELLSCHAFT PACCAR INC
KUKA AKTIENGESELLSCHAFT PALL CORP
LEONI AG PITNEY BOWES INC
DEUTSCHE LUFTHANSA AKTIENGESELLSCHAFT PULTEGROUP INC
LINDE AKTIENGESELLSCHAFT KEYCORP
25 LANXESS AKTIENGESELLSCHAFT ALLEGHENY TECHNOLOGIES INC
METRO AG TORCHMARK CORP
MORPHOSYS AG UNITED TECHNOLOGIES CORP
MERCK KOMMANDIT- GESELLSCHAFT AUF AKTIEN XEROX CORP
MTU AERO ENGINES AG TJX COMPANIES INC
AURUBIS AG ZIONS BANCORPORATION
NORDEX SE CABLEVISION SYS CORP -CL A
NEMETSCHEK AKTIENGESELLSCHAFT CBS CORP
XING AG ALTERA CORP
PFEIFFER VACUUM TECHNOLOGY AG STAPLES INC PSI AKTIENGESELLSCHAFT FUER PRODUKTE UND
SYSTEME DER INFORMATIONSTECHNOLOGIE BIOGEN INC PROSIEBENSAT.1 MEDIA AG TIME WARNER INC
QIAGEN N.V. KOHL'S CORP
QSC AG COSTCO WHOLESALE CORP
RATIONAL AKTIENGESELLSCHAFT EASTMAN CHEMICAL CO RHOEN-KLINIKUM AKTIENGESELLSCHAFT APARTMENT INVST & MGMT CO RHEINMETALL AKTIENGESELLSCHAFT BOSTON PROPERTIES INC RWE AKTIENGESELLSCHAFT GOLDMAN SACHS GROUP INC SMA SOLAR TECHNOLOGY AG FRONTIER COMMUNICATIONS CORP
SAP AG CBRE GROUP INC
STADA-ARZNEIMITTEL AKTIENGESELLSCHAFT STRATEC BIOMEDICAL AG K+S AKTIENGESELLSCHAFT SGL CARBON SE SIEMENS AKTIENGESELLSCHAFT SKY DEUTSCHLAND AG SOFTWARE AKTIENGESELLSCHAFT AXEL SPRINGER SE SARTORIUS AKTIENGESELLSCHAFT SYMRISE AG SALZGITTER AKTIENGESELLSCHAFT SUEDZUCKER AKTIENGESELLSCHAFT MANNHEIM/OCHSENFURT TAG IMMOBILIEN AG THYSSENKRUPP AG TUI AG UNITED INTERNET AG VOLKSWAGEN AKTIENGESELLSCHAFT WACKER CHEMIE AG WIRECARD AG ELRINGKLINGER AG
