Shareholder vs. Efficiency Performance of European Firms
Wesley Kaufmann
RijksUniversiteit Groningen Faculty of Economics Master Thesis for Msc IE&B
Dr. Ir. D. J. Bezemer July 19th , 2006
Shareholder vs. Efficiency Performance of European Firms
Wesley Kaufmann
1ABSTRACT
This thesis is concerned with empirically testing for performance and performance volatility differences between firms from different varieties of capitalism. Based on available literature, the main hypotheses are firstly that Anglo-Saxon, liberal firms will outperform their Rhenish, coordinated competitors when it comes to shareholder performance, while the exact opposite holds for efficiency performance. Secondly, the hypothesis is investigated that liberal firms experience more shareholder performance volatility than their coordinated counterparts, while for efficiency performance volatility no differences are expected between the sample countries. A large sample of 3750 public firms from the UK (Anglo-Saxon, liberal capitalism), Germany (Rhenish, coordinated capitalism) and France (statist capitalism) is retrieved for the period of 1990 – 2004 and the three corresponding 4 year sub-periods. For these sample firms information is retrieved on a range of independent variables and four performance measures that act as dependent variables; stock performance and return on equity performance as proxies for shareholder performance and return on assets performance and asset efficiency performance as proxies for efficiency performance. Additionally, from these performance measures four volatility measures are created in order to investigate the second point of interest for this thesis, which is performance volatility.
The results of the analysis indicate little empirical support for the theoretical predictions. With regard to the hypothesized supremacy of liberal firms’ shareholder performance, I find some empirical support. Regarding the alleged supremacy of coordinated firms on efficiency performance, there is very weak support. For shareholder performance volatility there are no significant differences whatsoever (which is in contrast with the theory), while for efficiency performance volatility there are some significant differences where none were expected (which is also in contrast with the theory).
The main conclusion of my thesis therefore is that no compelling firm-level support in favour of the varieties of capitalism theory has been identified and that more research is required in order to establish if such support exists.
1 I would like to particularly thank Dirk Bezemer for his input and useful comments, Padma Rao Sahib for providing pointers for the methodology section, Hedy Kaufmann for assisting me in obtaining the data for this
CONTENTS
1. INTRODUCTION ... 4
2. LITERATURE REVIEW ... 6
2.1 The liberal vs. coordinated varieties of capitalism... 6
2.2 The Varieties of Capitalism framework by Hall & Soskice (2001)... 8
2.3 Other literature about the Varieties of Capitalism Framework... 8
2.4 Shareholder vs. efficiency performance... 10
2.5 Volatility of performance measures ... 14
3. METHODOLOGY ... 17
3.1 Overview of similar methodologies ... 17
3.2 Sample countries... 18
3.3 Dependent variables ... 20
3.4 Independent variables... 21
3.5 Interpretation of key statistics ... 24
3.6 Methodological approach... 25
3.7 Normality of the residuals ... 31
3.8 Significance of coefficients ... 33
4. RESULTS... 34
4.1 General interpretation of results for log-form analysis... 34
4.2 Performance measures results for log-form analysis ... 35
4.3 Volatility of performance measures results for log-form analysis... 36
4.4 General interpretation of results for standardized analysis ... 37
4.5 Performance measures results for standardized analysis... 38
4.6 Volatility of performance measures results for second analysis... 39
4.7 Summary of results ... 40
4.8 Shortcomings and possible elaborations ... 41
5. IMPLICATIONS FOR THE VARIETIES OF CAPITALISM DEBATE ... 43
6. CONCLUSION ... 48
REFERENCES... 49
1. INTRODUCTION
The varieties of capitalism debate has attracted much attention in the literature for many decades now. Over time the basic distinction between liberal vs. coordinated capitalism (or Anglo-Saxon vs. Rhenish capitalism, or shareholder vs. stakeholder capitalism) has come to the surface as the most widely accepted dichotomy of types of capitalism. Many authors have pointed out the shortcomings of this two-varieties framework and have suggested including other types of capitalism varieties. One of the most common such additions is the ‘statist’
variety of capitalism, of which France is the prime example. Most of the literature concerned with varieties of capitalism has revolved around a description of institutions that are present in a certain type of capitalist country. I intend to use a more empirically ‘firm oriented’
approach. It is my aim to investigate if the variety of capitalism distinction offered in the literature is supported by significant differences in firm performance.
The research aim of this thesis will be two-fold. Firstly, I will investigate differences in firm performance between so-called Anglo-Saxon, Rhenish and statist firms. According to the literature, firms from Anglo-Saxon countries are more shareholder oriented than Rhenish firms. The question then becomes whether or not this shareholder orientation leads to superior shareholder performance; i.e. is the Anglo-Saxon model superior when it comes to shareholder performance? This is an empirical question which has not been satisfactorily answered in the literature as of yet, as can be seen in the article by Fiss & Zajac (2004) in a slightly different context:
‘Future research could extend this focus by examining the consequences of an emerging governance regime for different stakeholder groups. For example, does the adoption of a shareholder value orientation in Germany really lead to greater returns for stockholders?’
The other side of the coin is that Rhenish firms are more long-term, efficiency oriented, hence one would expect to see superior efficiency performance from these firms as compared to Anglo-Saxon firms. Thirdly, statist firms are expected to be somewhere in the middle of these liberal vs. coordinated capitalism extremes, which should be reflected in the shareholder and efficiency performance of these firms accordingly.
Secondly, this master thesis is concerned with the volatility of performance measures for firms from different capitalist models. If one assumes that firms are unable to alter an economy’s fundamentals, then there is no room for Anglo-Saxon firms to consistently outperform its non-Anglo-Saxon rivals with regard to shareholder performance. Managers from these firms may be able to create short-term gains, but these will undoubtedly be offset by similar downturns, as the firm is unable to change the economy’s fundamentals. This will in effect lead to more shareholder performance volatility, but not better shareholder performance. The enhancement of efficiency performance on the other hand is a long term oriented process, whereby the incentives to make short term gains are not nearly as visible.
This leads me to expect that there will be no significant differences in efficiency performance volatility between sample countries.
In order to test the abovementioned research questions, I focus my analysis on the Varieties of Capitalism framework by Hall & Soskice (2001), which is one of the most comprehensive and well-known models dealing with this material. A general sample of 3750 public firms from the UK (Anglo-Saxon, liberal capitalism), Germany (Rhenish, coordinated capitalism) and France (statist capitalism) is collected. For these firms information is retrieved on four performance measures: stock performance and return on equity performance as proxies for shareholder performance, and return on assets performance and asset efficiency performance as proxies for efficiency performance. From these four performance measures four volatility measures are created in order to investigate the second point of interest (performance volatility) of this thesis.
The first part of the thesis is a literature review, firstly focussing on the broad distinction
between liberal and coordinated varieties of capitalism, and later zooming in on the Varieties
of Capitalism framework by Hall & Soskice (2001). The literature review ends with the
introduction of the hypotheses for this thesis. The next step is to begin empirically testing the
abovementioned research questions by introducing the methodology employed. After that the
actual results from the analysis are stated. These analytical results are reviewed in the
following part of the thesis. The thesis is completed with a conclusion.
2. LITERATURE REVIEW
2.1 The liberal vs. coordinated varieties of capitalism
The most basic distinction with regard to varieties of capitalism is the one between liberal and coordinated capitalism. Liberal capitalism is characterized by a strong emphasis on market forces. It is also called the Anglo-Saxon model, due to the fact that Anglo-Saxon countries (such as the UK and the US) fit this model best, or shareholder capitalism, due to the shareholder orientation of firms (see for instance Dore 1999). According to Schmidt (2004), these liberal, Anglo-Saxon capitalist countries are characterized by the following aspects:
‘In such countries, the traditionally ‘liberal’ hands-off state that in the postwar years sought to ensure market-driven inter-firm relations and market-reliant management-labor relations has since the 1980s become even more liberal through the privatization of state-owned firms, the liberalization of the financial markets, and the deregulation of business and of the labor markets.’
In clear contrast with the liberal type of capitalism is the coordinated, Rhenish or stakeholder variety of capitalism. This stereotype of capitalism can be found in countries such as Germany, The Netherlands, Austria, Sweden and Denmark. Schmidt (2004) describes coordinated capitalism as a variety of capitalism where the ‘enabling’ state facilitates collaborative inter-firm relations and cooperative labour management relations.
Summarizing one could say that liberal capitalism is based on competition between firms, while coordinated capitalism is based on cooperation between firms. It should come as no surprise however that the rigid dichotomy between liberal and coordinated capitalism fails to capture all varieties of capitalism which are present in today’s world.
An additional type of capitalism that is often proposed by authors is the so-called statist type of capitalism. This type of capitalism basically comes in somewhere between the liberal and coordinated varieties. Schmidt (2004):
‘In this variety of capitalism, the ‘interventionist’ state that in the postwar period sought to organize inter-firm collaboration, direct business investment, and impose management-labor cooperation has given up its past leadership for a more market-oriented, ‘enhancing’ role. Although the state now also seeks to create and preserve market institutions much as in liberal market economies, this does not
stop it from continuing to intervene strategically where it sees the need, mainly to protect business and/or labour from the worst effects of the markets—whether this means bailing out firms in difficulty, ‘moralizing’ the labor markets through the 35 hour work week, as in France, or engineering corporatist agreements in wage-bargaining and pension reform, as in Italy.’
Many enhancements to the basic dichotomy, such as the one just described by Schmidt (2004) have been proposed by different authors. Hopkin & Blyth (2004) clearly indicate however that elaborating on the dichotomy comes with its own tradeoff:
‘However, the problem with more nuanced approaches is that what is gained in empirical nuance may be lost in analytic leverage. The more attention is paid to institutional specificities, the more difficult it becomes to establish any generalizable conclusions about the choices facing advanced political
economies.’
Hopkin & Blyth (2004) create a model centred around a 2-dimensional matrix, the one being welfare effort and the other economic liberalism. This leaves them with four potential types of capitalism, which is arguable still too many or too few, based on one’s viewpoint.
For the purpose of this thesis I make use of one of the newer frameworks describing varieties of capitalism; the Varieties of Capitalism (VOC) framework by Hall & Soskice (2001). It is in my opinion essential to build my research specifically around one varieties of capitalism model, since otherwise my method of approach would be too divergent and difficult to comprehend. The VOC framework by Hall & Soskice (2001) is used to specifically discuss one example of varieties of capitalism literature, and its empirical merit according to other authors is also reviewed. From time to time however additional literature will be used in order to enhance my analysis; this will be indicated clearly.
The VOC framework is in line with the general liberal vs. coordinated capitalism dichotomy
as outlined above. The framework by Hall & Soskice (2001) has attracted much attention in
the literature and is to be considered one of the most well-known theories dealing with
varieties of capitalism at the moment, which makes it a very suitable starting point for my
own analysis. The next part of this thesis is therefore concerned with describing the VOC
framework.
2.2 The Varieties of Capitalism framework by Hall & Soskice (2001)
In their 2001 article Hall & Soskice introduce the Varieties of Capitalism (VOC) framework.
According to this framework, the most important factor with regard to institutions is strategic interaction between agents. They believe that the degree to which institutions facilitate this interaction is the main point of interest. The approach Hall & Soskice (2001) offer in their article is actor centred; in capitalist economies firms are the crucial actors. The firms in turn are relational in that they seek to develop core competencies and dynamic capabilities in relations with employees and transaction partners. The authors move on to characterize the worlds economies into two broad classifications: the Liberalized Market Economies (LMEs) and the Coordinated Market Economies (CMEs). An LME is defined as a system comprised of competition and hierarchy, while an CME is characterized by more relational contracting, networking, and collaboration. For the CME country strategic interaction is more important than competition, while in the LME country the exact opposite holds. The empirical evidence in the Hall & Soskice article is provided by taking a look at the German case (as a clear CME example) and the US case (as a clear LME example). They use patent data from these two countries to support one of their major VOC predictions, namely that LMEs are more concerned with radical innovation, while CMEs are focussing mainly on incremental innovation.
2.3 Other literature about the Varieties of Capitalism Framework
The Hall & Soskice article has been the starting point of much further research, most of which has been different types of criticism concerning the framework. An article by Allen (2004) explicitly compares the VOC framework to the transaction costs economics, as well as neoclassical economics. The main conclusion of the author is that there appears to be a paradox with regard to the actor-centred approach of VOC. Allen claims that:
“within the (VOC) approach the differences between firms are irrelevant: all firms are assumed to adhere to the ideal type of firm in a CME or an LME and can adapt to become, respectively, incremental or radical innovators. It is the structure of the national economy, not the differences between firms, that largely determines the actions of firms.”
Allen (2004) states that the differences between firms are more important than the VOC
between firms that determine a firms’ actions, not the fact that a certain firm is located in a CME or LME country. An overview of the analysis by Allen is given in table 1.
--- insert table 1 about here ---
Another attack on the VOC framework is made by Taylor (2004). In his article he empirically investigates one of the main hypotheses of the VOC, that LMEs will direct their innovative activity toward radical innovation, while CMEs direct theirs toward incremental innovation.
Taylor finds that this claim does not hold, and furthermore that the US is an LME outlier;
without this outlier there is no support for the abovementioned claim. Even though Taylor uses a similar approach to investigating patents as do Hall & Soskice his findings are quite different. In their article, Hall & Soskice use patent data involving only two countries: the US and Germany. The time period used is only 4 years, which is not long enough, according to Taylor. Furthermore, for Hall & Soskice all patents are considered equal, in that there is no distinction between highly innovative patents and frivolous (minor or incremental) ones. Also, Taylor perceives the US as an outlier, whereas for Hall & Soskice the US is a prime example of an LME. Taylor controls for these perceived shortcomings of the Hall & Soskice article and finds radically different results than do Hall & Soskice. The results themselves are not that much of a surprise perhaps, but the explanation Taylor provides for VOC’s poor performance when it comes to explaining innovation patterns is all the more surprising. The reader may recall from the previously mentioned article that Allen (2004) posed that VOC theory understates the importance of the structure and goals of the individual firm and overstates the influence of the state. Taylor comes up with a conclusion completely opposite to that in order to explain his findings:
“while the firm may be the key actor in capitalist economies, it is difficult to ignore the role of the state in innovation as strongly as VOC’s theory and classification system do.”
Reconciling these opposite conclusions explaining the apparent failure of the VOC framework
is difficult, but clearly indicates that the Varieties of Capitalism framework has received its
fair share of criticism from the academic world. Perhaps one of the more general drawbacks
of the VOC theory is that it leaves out quite a few countries that do not fit nicely into either
the LME or CME category. One of the most notable countries that does not fit the LME nor
CME archetype is France. Clift (2004) discusses the French system and concludes that
although the range of policy areas and institutional characteristics where France exhibits exceptionalism has decreased over time, it is still quite different from the Anglo-Saxon model.
Schmidt (2003) suggests to create a third, statist, type of capitalism, of which France would be the prime example. Yet France is obviously not the only country that fails to fit nicely into the CME / LME framework. Royo (2004) refers to what other authors have called the
‘Mediterrean’ style of capitalism, hence the term MMEs. Royo explicitly looks at the situation in Spain, and reaches the conclusion that at least one more type of capitalism exists other than the two mentioned by Hall & Soskice.
On the whole we see that the VOC framework has sparked much interest in the literature, even extending its scope to the field of European politics (Hix 2005).
Most authors seem to agree that the framework has a certain degree of validity, but much of the actual empirical findings indicate that the VOC framework fails to adequately address actual data. However, in defence of the framework, (as can be noted from the above discussion) most VOC-related literature only deals with either investigating the radical vs.
incremental innovation VOC-claim, or showing that not all countries fit nicely into the framework. It is my aim in this thesis to investigate if there are actual differences with regard to performance measures between firms from different varieties of capitalism. This area of research has been paid scant attention to so far in my opinion, although it is at the heart of the distinction between LME and CME firms. In this thesis I intend to (partly) fill this empirical gap.
2.4 Shareholder vs. efficiency performance
The subject of this thesis is to further investigate a general aspect of liberal vs. coordinated capitalism, namely that liberal (LME) firms focus more on shareholder performance, while coordinated (CME) firms are more concerned with efficiency. It should be noted once more at this stage that although the main model used in this thesis is the VOC framework by Hall &
Soskice (2001), other literature will also be added to the analysis. For the purpose of my own
analysis, I assume that different terminology is interchangeable; hence LME firms is a
synonym for Anglo-Saxon or liberal firms, and CME firms is a synonym for Rhenish or
coordinated firms. In so doing, I am creating an analytical framework which will allow me to
do research based on much of the available literature, as opposed to pinning myself down to
just one such framework.
With regard to differences between orientations of firms, Fiss & Zajac (2004) mention:
‘An important element of much of this research is the comparative analysis of how different countries view the public corporation: as an economic entity whose purpose is to maximize shareholder value versus a social institution whose purpose is to further the interests of the corporation itself, typically considering the interests of multiple stakeholders, including shareholders, employees, creditors, customers, and the society in which the corporation resides. The former view is typically identified as the Anglo-American model of governance, while the latter view is more often found in other parts of Europe and Asia.’
This general distinction between firm orientation from liberal and coordinated capitalism countries can be found in much of the literature. In their 2001 article Hall & Soskice state with regard to LMEs that:
“The relevant contrast is with CMEs, where firms need not be as attentive to share price or current profitability in order to ensure access to finance or deter hostile take-overs.”
Further on it is mentioned that;
‘On the whole, however, the markets for corporate governance in LMEs encourage firms to focus on the publicly assessable dimensions of their performance that affect share price, such as current profitability.’
These statements are right at the core of the liberal vs. coordinated capitalism dichotomy in my opinion when one focuses on actual firm performance, as opposed to describing different countries. From the abovementioned it becomes clear that LME firms are more oriented towards shareholders, while CME firms focus more on other stakeholders, including society.
This assumption is confirmed further by Hall & Soskice (2001) when they discuss the financial system within CMEs:
‘Access to this kind of ‘patient capital’ makes it possible for firms to retain a skilled workforce through economic downturns and to invest in projects generating returns only in the long run.’
The financial systems of LME and CME countries are such that LME firms are encouraged
to focus on shareholder welfare in the short term, while the availability of ‘patient capital’
allows CME firms to focus more on long term goals. As another example in favour of this statement it should be noted that Germany, a classical example of a CME country according to Hall & Soskice (2001) has a strong reputation for not focussing on shareholders. Fiss &
Zajac (2004) mention in this regard:
‘Germany has frequently been cited as the classical case of a non-shareholder orientation, as evidenced by the original German corporate law of 1937, which stated that the company was to be managed for the good of the enterprise and its employees (Gefolgschaft), the common wealth of the citizens (Volk), and the state (Reich)’
Despite its aversion of shareholder orientation, some changes in the way of increased shareholder value are active at the moment (Jurgens et al. 2000). However, due to the central pillars of the German systems (which according to Jurgens et al 2000 are the dominating role of banks, the system of co-determination and the company-centred management system) change towards a shareholder orientation is limited. The stakeholder approach which is relevant to CME firms in turn may imply a lower return to shareholders, according to Cornelius & Kogut (2003):
‘The CEOs of companies should be willing to look shareholders in the eye and tell them: “We are not seeking financial return maximisation and mere compliance. We aspire to a higher, more complex set of values and, yes, there may be a small reduction in shareholder return as a result.’
Additionally it is suggested by Orlitzky et al. (2003) that a stakeholder orientation is associated with an increase in firm efficiency:
‘Furthermore, by addressing and balancing the claims of multiple stakeholders (Freeman and Evan 1990), managers can increase the efficiency of their organization’s adaptation to external demands.’
Dore et al. (1999) compare the German-Japanese capitalist system to that of the Anglo-Saxon over time and state the following with regard to firm efficiency:
‘In the 1920s ‘welfare capitalism, company unions and corporate efficiency’ went hand in hand (Jacoby, 1997, p.21). In what was called the ‘non-union era’, the promise of long-term employment, a greater recognition of seniority, employee welfare and subordination of line management to personnel
departments offered a softer approach to countering the threat of independent unions than the more directly repressive American Plan (Jacoby,1985; Lazonick, 1990,chs 7-8)’
One of the other major influences of the German model came into being during WW II (Dore et al. 1999):
‘industrial mobilization was in the hands of men who believed passionately in organizational efficiency, but not in shareholders’ rights.’
The basic conclusion based on the Varieties of Capitalism framework as well as other relevant literature therefore is that LME firms will focus on shareholder performance, while CME firms focus on firm efficiency. Obviously such a black and white distinction cannot be assumed to hold for every single firm operating in either an LME or CME country, but overall one would expect, based on the relevant literature, that on average significant differences exist between LME and CME firms with regard to these performance measures.
In order to test the abovementioned difference in shareholder vs. efficiency performance of firms, three sample countries are selected; the UK (LME), Germany (CME) and France (neither LME nor CME). An elaboration on the choice of sample countries is provided further on in this thesis. At this stage a few more remarks with regard to the position of France in the VOC framework need to be made however. Hall & Soskice (2001) mention themselves that France does not fit nicely into the LME / CME dichotomy. Schmidt (2003) proposes a third, statist type of capitalism, of which France is a prime example. In this article Schmidt compares the French capitalism to that of Germany and Britain, and states that on the aspect of business relations, France falls in between the positions taken by Germany and Britain, as can be seen in table 2.
--- insert table 2 about here ---
According to Schmidt (2003) the time horizon of business relations in France is a medium-
term view, whereas Britain is short-term oriented and Germany long-term oriented. Similarly,
where British firms focus on profits, German firms focus on firm value and French firms are
said to be geared towards the goal of national-political priorities, which implies that it does
not focus as vehemently on the particular goals that German and British firms deem important. Based on the above discussion, the following hypotheses are constructed:
H1: UK firms will perform better on shareholder performance than German firms H2: UK firms will perform better on shareholder performance than French firms H3: German firms will perform better on efficiency performance than UK firms H4: German firms will perform better on efficiency performance than French firms H5: French firms will perform better on shareholder performance than German firms H6: French firms will perform better on efficiency performance than UK firms
The previous hypotheses are condensed into picture 1.
--- insert picture 1 about here ---
2.5 Volatility of performance measures
A different, yet related point of interest has to do with the volatility of the performance measures used in this thesis. Regardless of whether or not the results of the analysis will indicate that significant differences between the sample countries exist with regard to performance, it is also interesting to establish if certain performance measures are more volatile in one country than the other. At first glance comparing volatility across countries might seem like a straightforward exercise, but in reality this is not the case, as can be seen from the available relevant literature. The vast majority of related literature is concerned with models that are capable of predicting expected volatility and stock returns. Balaban & Bayar (2005) for instance attempt to forecast stock returns by taking expected and unexpected volatility into account. Their sample consists of 14 countries; Germany, Japan, US, UK and emerging financial markets. Their results are inconclusive however and lead them to conclude;
“Given the mixed results and low R2’s the empirical results suggest that forecast standard deviation or variance of returns may not be an appropriate measure of risk.”
Similar findings can be found in the article by Ballie & DeGennaro (1990):
“After estimating a variety of models from daily and monthly portfolio return data, we conclude that any relationship between mean returns and own variance or standard deviation is weak.”
The explanation which Ballie & DeGennaro (1990) offer for this finding is that investors consider some other risk measure to be more important than variance of portfolio returns.
Dellas & Hess (2005) find that the degree of financial development within a country is very important when it comes to explaining stock returns.
“Stock returns are found to be significantly related to the degree of financial development. In general, a deeper and higher quality banking system is associated with lower volatility of stock returns and a greater synchronization in the movements of domestic and world returns.”
A closer investigation of the findings of Dellas & Hess (2005) indicates that the statistical results for France, Germany and the UK are quite similar, which makes intuitive sense since the degree of financial development within these countries can be considered to be similar.
Therefore one would not expect to find significant differences in stock returns between the sample countries based on the degree of financial development.
Evidently different articles come up with different conclusions, and as of yet there is no
consensus as to how best compare volatility across countries. A comprehensive theoretical
framework predicting which type of capitalism would exhibit the most performance volatility
does not exist to my knowledge. This means that the best way of constructing hypotheses
regarding the volatility of performance between the sample countries in my opinion is to go
back to the varieties of capitalism literature itself. With regard to shareholder performance,
the basic assumption is that liberal firms may be able to boost these performance measures in
the short run, but since they are unable to change an economy’s fundamentals, in the longer
run these gains will be followed by losses; and instead of increased shareholder performance,
all we see is increased shareholder performance volatility. The reason for this is that in order
for LME firms to meet short term demands of their shareholders, they are more willing to take
on risky projects than are their CME counterparts. Since an increase in risk implies an
increase in volatility of performance, one can assume that LME firms will experience more
volatility than CME firms. As for the relative position of France in this case (which is neither
an LME nor a CME) the same reasoning as before is applied, based on the article by Schmidt
(2003). Hence I assume that the volatility of French firms will be lower than that of UK firms,
but higher than that of German firms.
H9: UK firms will exhibit more shareholder performance volatility than German firms H10: UK firms will exhibit more shareholder performance volatility than French firms H11: French firms will exhibit more shareholder performance volatility than German firms
The case for efficiency performance is different from shareholder performance. Building efficiency is a long term process, for instance related to the learning curve effect. The learning curve is discussed in more detail on page 23 of this thesis. The long term nature of improving efficiency makes it inherently unlikely that firms will seek short term gains; the entire reasoning behind improving efficiency is long term oriented to begin with. Another indication that differences in efficiency performance volatility are not to be expected is the fact that I have been unable to find any literature with regard to efficiency performance volatility, as opposed to for instance the extensive literature that is available regarding stock performance volatility. This reasoning leads me to expect that there will be no difference in efficiency performance volatility between my sample countries.
H12: UK, French and German firms will exhibit the same efficiency performance volatility
The hypotheses will be elaborated on further in the remainder of the thesis, both with regard
to CME / LME sample countries, as well as the choice of shareholder and efficiency
performance indicators.
3. METHODOLOGY
In this section the methodology with regard to investigating part of the VOC framework is outlined. Firstly a brief overview of methodologies used in similar papers is provided.
Secondly the selection of sample countries is given. Thirdly the selection of dependent variables is given, followed by the selection of independent variables. After this the key statistics of the used variables are provided. Finally the methodological approach, normality of the residuals and significance of the coefficients are discussed.
3.1 Overview of similar methodologies
A thorough examination of studies that have related environmental, strategic and organizational factors to financial performance is given by Capon et al. (1990). In this meta- analysis, they provide an overview of 320 studies, and they summarize the general approaches of these studies. The general theme throughout these studies is that ‘Virtually all studies of financial performance acknowledge the existence of joint causal factors; various multivariate tools (particularly regression analysis) have provided the most common way to establish
“control” of covarying causes by statistical means.’ (Capon et al. 1990). According to the authors a majority of 189 out of 320 studies have used regression analysis, with descriptive statistics (78 studies) coming in at second place. The next section of the article by Capon et al.
(1990) states that of the 320 studies, 73 analyzed at the industry level, 205 at the firm level and 42 at the business level. The high number of firm level oriented studies is again a clear indicator that the techniques described in this article are very relevant to my own study. In order to get a more detailed picture of the analyses used in other articles covering similar subjects I now move on to describe the article by Rennings et al. (2003).
Rennings et al. (2003) compare the stock performance of European firms, by looking at
sustainability criteria. They indicate three different approaches that are dominant in the
literature: portfolio analysis, event studies and panel or cross-sectional studies. Portfolio
studies are mentioned by the authors on page 35: ‘Such studies compare the economic (or
financial) performance of companies with a higher sustainability performance with portfolios
that consist of companies with a lower sustainability performance.’ However, the financial
success of existing funds depends heavily on the ability of the fund management according to
the authors, which makes it difficult to separate different effects. There is a bigger downside
to portfolio analysis however, according to Rennings et al. (2003, page 35):
‘But the main weakness of analyses of fund performance in general is that only the average economic performance of all corporations in the portfolio is considered (see also Wagner 2001). Consequently, the specific form of the influence of sustainability performance variables on the economic (or financial) performance can hardly be separated from other influences (particularly from the ability of the fund management, but also, for example, from the market capitalisation or from regional peculiarities) within this methodology. The identification of isolated effects needs econometric methods that include all relevant variables to explain the economic performance besides the variables of interest (here, the sustainability performance variables).’
They then move on to discuss event studies, which they mention on page 35 to be studies that investigate the effect of news on the performance of single stocks. The authors however are unsure of the explanatory power of ‘special events’, that are the subject of news. Rennings et al. (2003) also provide another disadvantage to event studies on page 36 of their article:
‘But the main weakness of event studies is their short-term character. Thus, short-term over-reactions of stock markets are possible, which may be compensated over time. Consequently, the investigation of the general effect of sustainability performance on economic performance needs long term consideration.’
The authors conclude this section of their article by discussing panel and cross-sectional studies on page 36. ‘These studies investigate the relationship between certain characteristics of companies and their economic performance’. This is their method of choice; according to the authors they decide on panel and cross-sectional studies, since ‘In contrast to event studies, the analysis concentrates on characteristics of companies and not on specific news about the companies. In contrast to portfolio analysis, we do not analyse a portfolio of stocks but single stocks.’ (Rennings et al. 2003, page 36).
Based on the relevant literature and the type of research I am attempting for my thesis, it is logical for me to run a regression analysis, as can be seen later on in this thesis.
3.2 Sample countries
In order to empirically test the VOC framework, i.e. compare CME to LME firms, the first
step is to identify which countries are considered CMEs and which LMEs by Hall & Soskice
(2001). According to the authors, “among the OECD countries six can be classified as liberal
market economies (the USA, Britain, Australia, Canada, New Zealand, Ireland) and another
ten as coordinated market economies (Germany, Japan, Switzerland, The Netherlands,
Belgium, Sweden, Norway, Denmark, Finland and Austria) leaving six in more ambiguous
positions (France, Italy, Spain, Portugal, Greece and Turkey).” Evidently the choice for LME
and CME countries will have to be based on this summation by Hall & Soskice. The most straightforward approach then is to choose one classic LME example, one classic CME example and 1 country which does not fit neatly into the LME / CME dichotomy. Based on the summation given by Hall & Soskice, these sample countries are: the UK (LME), Germany (CME) and France (neither LME nor CME) respectively. Given the approach of this thesis, a first requirement for sample countries is that the country is sufficiently large to offer data on a large sample of firms. This requirement implied that all medium- and small-sized countries were not eligible as sample countries. The four largest countries present in the VOC framework are then the USA and UK as LMEs and Germany and Japan as CMEs. Given this potential set of countries, it was decided to use a ‘European’ approach when selecting the sample countries. Not only does Taylor (2004) conclude that the USA is an LME outlier, but given the fact that the USA and Japan are located on different continents as compared to European countries, when taking these countries into account one runs the risk of comparing geographically different countries, as opposed to varieties of capitalism. Hence, the UK was selected as LME and Germany as CME sample country. Apart from these two sample countries, a country that does not specifically fit into the dichotomy is taken into account.
Schmidt (2003) describes these countries, claiming that there are more types of capitalism than the CME / LME dichotomy:
“Rather than one or two varieties of capitalism, this paper argues that there are still at least three in Europe, following along lines of development from the three post-war models: market capitalism, characteristic of Britain; managed capitalism, typical of Germany; and state capitalism, epitomized by France. While France’s state capitalism has been transformed through market-oriented reforms, it has become neither market capitalist nor managed capitalist. Rather, it has moved from ‘state-led’
capitalism to a kind of ‘state-enhanced’ capitalism, in which the state still plays an active albeit much reduced role, where CEOs exercise much greater autonomy, and labour relations have become much more market-reliant.” (page 526)
Given the large size of the economy, its European location, and its overall importance in the
varieties of capitalism literature, the third and final sample country for this thesis is therefore
France.
3.3 Dependent variables
In order to answer my research questions it is essential to come up with performance indicators that can be used as dependent variables for these different dimensions. Hall &
Soskice (2001) already point in a certain direction in their article when mentioning that LME firms focus on current profitability in order to boost share prices. The most obvious performance indicator with regard to shareholder performance then becomes the firm’s stock performance. Using stock performance as a profitability indicator is common in the literature.
In my thesis stock performance is calculated as the % change in share price for a sample firm over the periods 1990-1994, 1995-1999 and 2000-2004. In order to have a more thorough investigation of the claim made above however, a second indicator has been identified to investigate the shareholders performance, which is return on equity. De Meuse et al. (2004) define return on equity as ‘A company’s return on equity (ROE) is determined by dividing profits by stockholders’ equity. It is similar to ROA, but focuses on the actual financial rate of return to the company’s owners.’ In order to compare actual performance of firms as opposed to certain levels, I use return on equity performance, which is defined as the % change in return on equity for a sample firm over the specified sample periods. The return on equity performance variable is especially relevant as a shareholder performance indicator, as is mentioned by Jurgens et al. (2000, page 57):
‘In Germany, there has been only a minor role for the value orientation which focuses on return on equity and the value behaviour of trading stocks.’
The next step is selecting dependent variables for firm efficiency. According to the article by Bhagat & Black (2000) ‘There is no single ideal measure of long-term firm performance.’
One of the measures they propose however is looking at a firm’s return on assets. Looking at return on assets as an efficiency indicator is done often in the literature. ‘This measure examines the profitability of a company in relation to dollars invested. It is an index of overall return on investment and indicates how efficiently those dollars are utilized.’ (De Meuse et al.
2004). Most articles examining the financial performance of firms use return on assets as a
measure; such as Waddock & Graves (1997), Bhagat & Black (2000), Berman et al. (1999)
and Rennings et al. (2003). It is therefore a widely accepted measure of investigating financial
performance of a firm, as well as firm efficiency. Similar to the other performance measures,
return on assets performance is calculated (the % change in return on assets for a sample firm
over the specified sample periods), as opposed to merely return on assets. Another efficiency measure, according to De Meuse et al. (2004) is asset efficiency; defined as dividing sales by assets. ‘This measure identifies how efficiently a company is using its assets to produce its sales.’ As can be seen from the definition of assets efficiency, this measure is quite similar to ROA, except for the fact that ROA is concerned with a firm’s profits related to its assets base, while asset efficiency is concerned with its sales related to its assets base. Again the asset efficiency data is transformed into an asset efficiency performance measure (the % change in asset efficiency for a sample firm over the specified sample periods), similar to the three other performance measures. Summarizing, there are four primary dependent variables in my thesis;
stock performance and return on equity performance as proxies for shareholder performance, and return on assets performance and asset efficiency performance as proxies for efficiency performance.
As mentioned before, the second block of the analysis is concerned with the volatility of the chosen performance measures. Regardless of potential differences between sample countries in the actual average performance measures, the question whether volatility of these performance measures differs also offers interesting research potential. In order to proxy the volatility of performance, the standard deviation of performance measure for each sample point as compared to the relevant mean is calculated in Excel. As such, a new list containing the standard deviations of each firm compared to the mean is created. This list of individual standard deviations is then added as a dependent variable in the analysis, whereby the same independent variables are used as in the case of the corresponding performance measure.
All data is retrieved from the Thomson One Banker database, which is a composite database of Datastream, Thomson Financial and Worldscope.
3.4 Independent variables
The next step is to establish which independent and control variables are going to be used. A
logical starting point is to look at research of financial firm performance in general, and
establish which criteria are most often used. All the studies reviewed by Capon et al. (1990)
included ‘(1) a dependent variable measuring financial performance; (2) nonfinancial
explanatory factors.’ These studies therefore follow a similar approach to the one I am
following in my thesis. After reviewing all the published studies as mentioned before, the
authors provide a summarizing table. The table shows that all these measures are related to
general industry or firm characteristics that are likely to have an influence on firm financial performance. Keeping in mind the available data from Thomson One Banker and the aim of my thesis, the following measures of this table have been selected: Growth in sales, Assets, Employees Sales and Total debt. A final general measure comes from Waddock & Graves (1997). In this article corporate social performance (CSP) is compared to financial performance. According to the authors ‘size, risk and industry have been suggested in previous articles to be factors that affect both firm performance and CSP.’ Indeed, risk is perceived as an important factor influencing firm financial performance, see for instance Fama & French (1992). The standard way of taking firm-specific risk into account is by calculating its beta (a measure of the sensitivity of a stock’s price to the movement of an index), but for longer periods of time this beta is not widely available. The authors move on to choose the long-term debt to total assets ratio as a proxy for firm riskiness. The long-term debt / total assets ratio is a leverage ratio (Johnson 1997). Fama & French (1992) state that it is plausible that leverage (the mix between debt and equity) is associated with risk and expected return. Brealey et al. (2001) mention this as well on page 490 of their book:
‘If profits rise, the debtholders continue to receive a fixed interest payment, so that all the gains go to the shareholders. Of course, the reverse happens if profits fall…. Because debt increases returns to shareholders in good times and reduces them in bad times, it is said to create financial leverage.
Leverage ratios measure how much financial leverage the firm has taken on.’
This constitutes the same approach that I will be using to proxy firm riskiness. Since all the abovementioned measures are of general importance for firm financial performance according to the literature, they are used as independent variables for all four dependent variables.
Apart from these general independent variables, two additional control variables are also included. They are industry and country dummies. The industry dummies are created based on their Primary SIC code; 8 dummies are created for the first number of industry classes in this code (1 – 8). Secondly, 3 country dummies are created; the first is for German firms, the second for French and the third for the UK. Obviously only two country dummies are included in each analysis. The France and Germany dummies are included for the regressions in which differences between French and German firms vs. UK firms are investigated, while the France and UK dummy are included in order to compare French to German firms.
Evidently these industry and country dummies will also be used for all four dependent
variables.
Not only are there general financial performance measures and the country and industry dummies, there are also some independent variables specific to each dependent variable, based on the literature. Firstly, additional variables are included for stock performance.
According to Fama & French (1992) firm size, market capitalization and book-to-market value ratio offer a surprisingly good explanation of stock performance. Firm size is already taken into account in the general measures, but firm market capitalization (Market Price-Year End * Common Shares Outstanding) is not, and is therefore added as additional independent variable for stock performance. Rennings et al. (2003) also mention the book-to-market value:
‘
The book-to-market factor has a positive coefficient in the stock return regressions. Thus, a higher book-to-market value should coincide with a higher average’ stock return.’My original approach was to include book-to-market ratio as an independent variable as well in the regression. However, the available data from Thomson One Banker proved to be available for UK firms only, hence during the actual regression analysis it became apparent that this variable would have to be dropped. For return on equity, the PE (Market Price-Year End / Earnings Per Share) ratios are also included. According to Easton (2004) the PE ratio is important when investigating ROE, albeit a simplistic approach. For the current study this variable suffices however. For return on assets and asset efficiency there are no additional independent variables, since none have been found in the reviewed literature.
The last general independent variable is the number of years a firm is active. The reason for using this variable is two-fold. Firstly, since I am comparing shareholder performance to efficiency performance of firms, it is essential to have some way of distinguishing between firms that have been in operation for a long period of time, as opposed to relatively new firms.
The reasoning behind this claim is that efficiency-building is a time-consuming process, referred to as the learning curve effect. ‘The learning curve (or experience curve) refers to advantages that flow from accumulating experience and know-how.” (Besanko et al. 2004, page 95). A graphical representation of the learning curve can be found in picture 2.
---- insert picture 2 about here ---
As one can see, as cumulative production increases over time from Q
xto 2Q
x, a firm will
move to the right of the AC-curve due to learning and incur lower average costs. These lower
average costs in turn imply an increase in firm efficiency. Therefore, one expects to find better efficiency performance from firms that have been in operation longer. Secondly, Johnson (1997) mentions that the age of a firm can serve as a proxy for a firm’s reputation.
Knowing that in particular with regard to stock performance the expectations of shareholders are important, having a proxy for reputation is obviously worthwhile.
In order to take into account the age of a firm, its founding year is retrieved from the Thomson One banker database. For each sample period the number of years that a firm has been active is retrieved and calculated by deducting the founding year from the first year of the time period of interest.
The information about the variables of this thesis are summarized in two tables. Table 3 provides an overview of the formal definitions of the used variables, while table 4 shows the key statistics for each of the relevant variables. Please note that table 4 represents the variables as they appear in the log-form analysis which will be carried out later on.
---- insert table 3 about here ----
---- insert table 4 about here ----
3.5 Interpretation of key statistics
With regard to the performance measures, it is interesting to see that stock performance is the only performance indicator which has a positive value for each of the sample periods. ROA performance is consistently negative for each of the periods, while ROE performance is negative in three of the four sample periods. Finally, asset efficiency performance is positive for three sample periods. What is more interesting perhaps is the fact that nearly all of the variables, both dependent and independent, have high standard deviations. In many cases the standard deviation is multiple times the mean. Large standard deviations are an indication that the data points are far from the mean, which reduces the explanatory power of the analysis.
Related to this is the issue of normal distribution of the residuals, to which I come back in
section 3.7. Very high standard deviations can have a significant impact on the final results of
an analysis. Therefore I shall address this issue by running a secondary analysis later on, in
which all of the variables have been standardized and high standard deviations are no longer
an issue.
3.6 Methodological approach
In order to test the hypotheses mentioned earlier on in this thesis, two different regression analyses are performed. Rennings et al. (2003, page 39) for instance also run two econometric approaches in order to get robust results. For both analyses Ordinary Least Squares (OLS) is performed by using the relevant dependent variable and independent / control variables mentioned earlier on in this thesis. In the first analysis outliers of the dependent variable are reduced when performance measures are >1000%, or < -1000% and afterwards the dependent variables are transformed into log-form. My second analysis, which is meant to enhance some of the shortcomings of the first analysis and also to verify the overall results of the first analysis, excludes outliers of the dependent variable with performance > 250% or < -250%
and afterwards all of the variables are standardized.
The choice of these analytical methods merits some further attention at this point. Firstly, it should be noted that the final results of a thesis such as this may be heavily influenced by the choice of sample data, as well as the employed methodology. For this reason I have chosen two distinct analyses, which should make the overall conclusions of my research more robust.
The approach which is used in both analyses of this thesis is two-fold. Firstly, I have manually identified and removed outliers from the dependent variables that would severely affect the outcome of the analysis. Secondly the data is transformed in order to make the analysis more meaningful.
Deciding on which sample points constitute outliers is not a fast-and-hard rule. There is a
trade-off when excluding very large positive or negative points between reducing the
sample’s standard deviation on the one hand, and excluding important information for the
regression on the other. For the log-form analysis I have decided to exclude all values for the
primary dependent variables (stock, ROA, ROE and asset efficiency performance) that are
larger than 1000%. For the volatility performance measures no outliers have been identified,
since they are a carbon copy of the primary dependent variables. For the standardized analysis
a 250% outlier boundary for the primary dependent variables has been identified. Why have I
chosen these performance boundaries? With regard to the 1000% boundary; in my opinion
performance that has improved more than 10-fold in a period of 5 to 10 years can be
considered an outlier, since such a dramatic performance improvement is most unlikely unless
the performance was extremely weak to begin with. Obviously the 1000% boundary is open
to debate, but as an approximation of detecting outliers it suffices for the current purpose. The
1000% boundary which I have used has resulted in the exclusion of the following outliers (table 5). It should be noted that at no point more than approximately 6% of the original sample has been removed.
--- insert table 5 about here ----
The 250% outlier-boundary used in the standardized analysis came about as a result of the very high standard deviations of the first analysis (as mentioned before). There are basically two ways of dealing with high standard deviations; either adjust the sample or adjust the statistical model. In order for me to strongly reduce the high standard deviations of the first analysis I have opted for a combination of these two; hence the outlier-boundary has been severely tightened for the dependent variables, and a different methodological approach used (to which I shall return shortly). It should be noted that by further reducing the outlier- boundary I have introduced a bias in my second analysis. By excluding all the best and all the worst performers, one would expect, based on the hypotheses, that the average performance of UK firms on shareholder performance will artificially decrease (since it is expected to have many well-performing firms excluded), while the performance of German firms will artificially increase (Germany is expected to have many bad-performing firms excluded).
Obviously, we expect the exact opposite to hold with regard to efficiency performance.
Hence, it should be taken into account that an additional bias has been introduced in the second analysis (since obviously the 1000%-boundary of the log-form analysis created a much smaller bias). The bias is merited however in my opinion, since by doing so I am significantly reducing some of the standard deviations of the dependent variables, which is the biggest concern of the first analysis. The 250% boundary which I have used has resulted in the exclusion of the following outliers (table 6). Due to the stricter boundary, significantly more sample firms have been excluded; sometimes up to 20% of the sample for a particular dependent variable.
---- insert table 6 about here ---
For the independent variables I have not manually removed any potential outliers, for which
there are three reasons. The first reason has to do with selecting the ‘objective’ criteria (as
mentioned before there are no clear-cut rules when it comes to excluding outliers) by which
difficult and open to debate. If one were to exclude outliers from the independent variables, different criteria would have to be applied to each independent variable in order to get a result which will yield good results for the researcher (lower standard deviations in my case). The gain of lower standard deviations would be more than offset by the loss incurred due to the arbitrary data selection in my opinion.
The second reason is related to data availability. Despite the relatively large base sample of 3750 firms, the actual regressions that will take place (as mentioned later on in this thesis) are often based on about 100 to 200 sample firms and at times even less. Arbitrarily cutting back on sample points would further reduce the amount of available sample firms for the regression analysis, and thereby weaken the entire analysis.
Thirdly, my standardized analysis will standardize all of the relevant variables; as will be explained further on this significantly reduces the impact of outliers.
Now that the procedures for removing outliers for both analyses have been discussed the transformation of the data for each analysis after removing the outliers is reviewed in turn.
For the log-form analysis, the dependent variables are transformed into a log-form by using the ‘log(dependentvariable)’ function in Eviews. With empirical investigations such as this one, there is often a problem of non-constant variance. According to Anderson et al. (2002) this problem is often corrected by transforming the dependent variable to a different scale. As an example they mention that by using a log scale, the values of the dependent variable will be compressed and thus the effects of non-constant variance diminished. Hence, the choice of transforming my dependent variables into log-form is a logical one.
With regard to the standardized analysis, I have chosen to standardize my variables. As has been mentioned several times before, the biggest drawback of my log-form analysis is the very high standard deviations of my sample. These high standard deviations could have a significant impact on the final outcome of my research, which is obviously an undesirable situation. One way of effectively dealing with high standard deviations is to standardize the data:
“Many researchers have noted the importance of standardizing variables for multivariate analysis.
Otherwise, variables measured at different scales do not contribute equally to the analysis. For example, in boundary detection, a variable that ranges between 0 and 100 will outweigh a variable that ranges between 0 and 1. Using these variables without standardization in effect gives the variable with
the larger range a weight of 100 in the analysis. Transforming the data to comparable scales can prevent this problem. Typical data standardization procedures equalize the range and/or data variability.”2
The standardization procedure which I have used is having the mean subtracted from the series and then divided by the standard deviation; see formula 1.
Formula 1: Standardized variables in Eviews
3YSTD=(Y-MEAN(Y))/SQR(VAR(Y))
Where Y is the series to be standardized
After having standardized in this way, the mean for all variables will be 0 and the standard deviation will be 1. An inspection of the output for the standardized series confirms that the standardization of the variables has been successful. The dummy variables which I use in my thesis (the industry and country dummies) are not standardized, since standardizing these variables would make them meaningless in the regression. These standardized variables will be used in my second analysis.
Before discussing the actual results of the two analyses, a few general comments regarding regression analysis need to be made. Two major issues when running a regression analysis have to do with autocorrelation and heteroskedasticity. Since I am using a panel dataset, autocorrelation is not an issue. As can be seen on page 258 of Carter Hill et al. (2001a);
‘cross-sectional data are often generated by way of a random sample of a number of economic units such as households or firms. The randomness of the sample implies that the error terms for different observations (households or firms) will be uncorrelated.’
The method which is used in the thesis with regard to data selection is the following. The only criterion which I apply is the country of origin. Based on the data in the Thomson One Banker database all public firms that are said to have their base country in France, Germany or the UK are included in a sample firm list (totalling 3750). This list is used to retrieve information for all the variables for each of the sample firms. Naturally not all of the required information
2 Source: http://www.terraseer.com/products/bsr/help/Preparing_data/Why_standardize_variables.htm 3Source: http://www.eviews.com/support/faq/faqdata
