RIJKSUNIVERSITEIT GRONINGEN
Faculty of EconomicsDepartment of International Economics and Business
Master Thesis MSc IE&B
Final Draft
Does the degree of internationalization determine the compliance
rate of firms?
Evidence from listed firms in The Netherlands
Student:
Jan Schumacher
Student ID:
1401696
Date:
August 2008
Supervisor(s):
Professor Hans van Ees (Master Thesis)
Abstract
This article examines whether the degree of internationalization of a firm determines compliance with the best practice provisions and recommendations stated in the Dutch corporate governance code. The study is based on a set of 130 Dutch listed firms in 2004. Except for marginal evidence that cross-listing and sales matter for compliance, results indicate that explanatory variables of internationalization are not significantly related to compliance. However, positive associations between the compliance rate and the control variables industry classification and firm size are found. Hereby, the research adds to and confirms prior research.
1 Introduction & Main Research Question ... 3
2 Theory and Hypotheses ... 5
2.1 Code Characteristics ... 5
2.2 Code Compliance in Europe ... 8
2.2.1 Great Britain... 9 2.2.2 The Netherlands... 11 2.2.3 Greece... 12 2.2.4 Germany... 13 2.4 Research Question ... 15 2.5 Derivation of Hypotheses ... 19
3 Data and Method... 26
3.1 Data ... 26
3.1.1 Dependent Variable... 26
3.1.2 Independent Variables... 27
3.1.3 Control Variables... 28
3.2 Method of Analysis... 30
3.2.1 Problems & Tests regarding the Statistical Analysis... 32
3.2.2 Limited Dependent Variable Analysis... 34
4 Results ... 36
4.1 Testing the model... 37
4.1.1 Descriptive statistics... 37
4.1.2 Testing the assumptions of the general model... 38
4.1.3 Results... 41
4.2 Robustness Checks... 42
4.2.1 Reduced Model... 42
4.2.2 Results... 43
4.3 Simple Regressions and Stepwise Regression Analysis ... 43
1 Introduction & Main Research Question
Today, each member state of the European Union (EU) is required to have a corporate governance code (European Commission 2004, 2005). A code is a set of ‘best practice recommendations’ regarding various aspects of corporate governance, such as specifications relating to the board of directors, relationships with shareholders and top management, auditing, information disclosure and remuneration. “They have been designed to address deficiencies in the corporate governance system” (Aguilera & Cuervo-Cazurra, 2004). Codes are a set of non-binding, informal and self-regulatory recommendations. They are not statutory rules but frequently apply a “comply or explain” principle. Here, compliance itself is not mandatory but disclosure of non - compliance is. This implies that a firm either complies with the recommendations or motivates deviations. In essence this principle is a disclosure obligation (Mac Neil and Li, 2006). Despite this voluntary nature of codes it appears that most stock listed firms correspond to the code recommendations (Gregory & Simmelkjaer, 2002).
Apart from the three factors named above –firm size, cross-listing and industry classification - there are likely to be further aspects that determine the compliance rate. One could be the degree of internationalization of a firm. Firms which are internationally oriented are likely to attract more attention from more parties than national firms, meaning that they also have to consider more interests; for instance those of foreign governments, foreign employees, shareholders and stakeholders in general. They are also more sensitive to (international) media coverage. In order to satisfy all interests and maintain a positive image of the firm one may argue that a higher degree of internationalization leads to a higher degree of compliance with the code. On the other hand it is possible to postulate that a higher degree of internationalization reflects a more complex firm and the necessity to account for various, sometimes conflicting, interests leads to a lower compliance rate.
The twist is that a firm applies the national code of the country of incorporation, regardless of its international orientation. So you may wonder how this affects its compliance behavior. Does the international orientation hinder its compliance, or does it perhaps exert a positive influence? In order to clarify this issue the research addresses the following general research question:
Does the degree of internationalization of a firm determine its compliance rate with the national corporate governance code?
The research analyzes the situation for Dutch stock listed firms in 2004 and examines the link between internationalization and compliance with Tabaksblat, the Dutch code. Thereby it contributes to the general research on corporate governance code compliance and especially adds to research on the Dutch situation.
size is a determinant of code compliance, as well as industry classification. The independent variables directly related to internationalization are however not significant.
The remainder of this paper is structured as follows. Section 2, lays out the theoretical background, reviews compliance behavior in Europe, motivates the research question and derives the hypotheses. Section 3 presents the data and methodology of analysis. Subsequently, section 4 presents the results obtained, to be followed by a discussion in section 5. Lastly section 6 concludes.
2 Theory and Hypotheses
This chapter firstly introduces the development and general characteristics of codes of corporate governance (2.1), specifying the “comply or explain” principle underlying the majority of codes and also the Tabaksblat code. Thereafter it reviews in section 2.2 research on code compliance behavior in Europe. Since codes rely on self-enforcement, it is worthwhile to examine in how far firms actually comply with provisions expressed in the respective national code (Wymeersch, 2005). Section 2.3 then explains the theoretical drivers of compliance. In section 2.4 we motivate the research question and subsequently derive the hypotheses in 2.5.
2.1 Code Characteristics
While national Codes differ to some extent in their design and contents, four general characteristics are underlying them (Seidl, 2007); (1) They are drafted by expert committees who formulate best practice standards, (2) they are formally voluntary and non-binding, (3) they are meant and designed to be flexible and they lastly (4) count on market mechanisms for monitoring and enforcement. We explain them in turn below, focusing on the third aspect.
To begin with, codes are commonly drafted by committees. These committees consist of “experts” who claim their expertise on grounds of practical experience, not on scientific knowledge. As a consequence, committees who predominantly are made up of practioners - such as firm directors, institutional investors, shareholder representatives, lawyers, accountants, etc. (Seidl, 2007) – are often chaired by prominent business men. In The Netherlands for instance, the corporate governance code is also known as the Tabaksblat code, named after the chairman of the committee of corporate governance, Mr. Morris Tabaksblat. Other examples include the various early codes in the UK (Cadbury Report, Higgs Report, Hampel Report) or the German Cromme Code.
Emphasizing the practical experience and expertise of the committees it is not surprising that codes are a collection of best practice recommendations, i.e. they recommend practices which have proved to be valuable over time. Referring to the expertise of the committees and labeling the code provisions as “best practice” is regarded as a means to increase the acceptance of a code (ibid).
however. Therefore, firms are required to indicate where they do not comply and are obliged to provide an explanation why not (Pass, 2006). In the Dutch context the provision of motivated non-compliance constitutes “compliance after approval by the general meeting of shareholders” (Tabaksblat Code, 2003; p.5). Two general concerns underlie the “comply or explain” approach, namely flexibility and the reliance on market forces for monitoring and enforcement.
Flexibility, the third characteristic, is often considered necessary as it is not believed that a “one size fits all” code form is desirable (Davies, 2008). On country level, it is impracticable because every country has an individual set of corporate governance procedures, rooted in the legal and financial system, owed to the economic situation, corporate ownership structure and culture (ibid). On firm level, it is not expected that firms adhere to all code provisions or are able to do so. In contrast, firms are even expected to deviate in aspects where there organizational structure and background denies the possibility to comply. Nowak et al. (2004) identify three classes of reasons for non- compliances, namely industry- or firm specific characteristics such as size and ownership structure, personal matters of management and, as already mentioned, motivated non-compliance. Flexibility of the codes is a means of increasing the responsiveness of the code with regard to the differences of the affected firms. Moreover, it also reduces the complexity of codes, as these do not have to explicitly account for all possible settings but give general recommendations. If firms cannot comply with these, they have the opportunity to apply them in a custom-tailored manor, by disclosing non-compliance.
prices, it is at the same time also an enforcement mechanism (Seidl, 2007). Hence, it is left to capital market players to judge whether individual cases of non-compliance are justified, or not. “The evaluation by those affected means that no particular institutional arrangements have to be made for this purpose” (ibid). Consequently, the complexity of codes is reduced, as well.
2.2 Code Compliance in Europe
Research on corporate governance codes so far analyses the rationale of code adoption (Aguilera and Cuervo-Cazzura, 2004), compares the contents of codes (Gregory and Simmelkjaer, 2002), examines the link between code adoption/compliance and stock valuation (Goncharov et al., 2006; Alves and Mendes, 2004; Fernandez-Rodriguez et al., 2004; Nowak et al., 2004) or describes the compliance behavior of firms with the national code (Akkermans et al., 2007; Werder et al., 2005; Oser et al., 2004; Dedman,2002; Conyon and Mallin, 1997).
Table 1 Code Information Country First Issue Last Issue IssUING Body Comply/Explain principle? Compliance rate Great Britain 1992 2006/08 Financial Reporting Council (FRC) Yes 75% 1 The Netherlands 1997 2003 2 Corporate Governance Committee Yes 86,8 % 3
Greece 1999 2001 Federation of Greek
Industries Yes 70.3%
Germany 2002 2007 Gov. Commission
German Corporate Governance Code
No 94, 7 %
(78,4%)4
2.2.1 Great Britain
A first report on good corporate governance emerged in Great Britain as early as 1992 (Cadbury Committee Report) and was followed by several others throughout that decade (e.g. Greenbury Committee Report, 1995). According to Stiles and Taylor (1993) the Cadbury report was most influential in questioning the effectiveness of voluntary regulation in Britain. In 1998 the gist of these two reports, plus additional provisions thought necessary by the committee, were incorporated in “The Combined Code”.
Its structure consists of two main parts. The first deals with principles of good governance, the second with the relation to institutional investors. The principles are further broken down in main and supporting principles. Naturally, the code follows the “comply or explain” principle. Listing rules require handing in a disclosure statement, commenting on the application of the principles. Additionally, the compliance statement indicates whether the code provisions are adhered to. If not, explanation is necessary.
A first revision of the code occurs in July 2003. It adds several provisions concerning the role of the CEO and chairperson, the composition of the board and its main committees. Furthermore, it stresses the role of non-executive directors and is in particular worried about their independence. Currently, two versions of the code are in effect. The edition of 2006 which applies to accounting periods starting 1 November 2006 or later and the 2008 edition applying to accounting periods beginning 29 June 2008
1
Grant Thornton (2007) find 41% of firms to comply fully and 34 % to comply or explain.
2
The Tabaksblat code was published in December 2003, therefore 2004 was the first year of application.
or later. The revision of 2006 merely features changes of “limited nature” (The Combined Code, 2006) whereas the 2008 update makes two changes. Firstly, the restriction that individuals may not chair more than one FTSE100 firm is removed. Secondly, given that he is considered independent on appointment, listed firms outside the FTSE350 may appoint the chairman to be part of the audit committee.
Grant Thornton (2007) in their corporate governance survey of the UK apply a list of verifiable propositions based on the Combined Code – and including guidance as expressed in Turnbull Report of 2006 – to 306 firms of the FTSE 350. In their year of analysis, 2007, it is revealed that 41% of the firms claim full compliance with the code. However, there are also 4% of firms which provide no explanation at all. The remaining 55% are not fully-compliant but 34% of the total sample embrace the comply or explain principle. In general, Grant Thornton conclude that improvement in compliance behavior and the provision of explanations as compared to 2006 and previous years can be observed. Nevertheless, there is still room for improvement. A number of those firms providing explanations do not make any changes and others merely cosmetic changes to the previous year.
Research of Christopher Pass (2006) examines the degree to which large UK firms comply with the main provisions of the Combined Code of 2003.5 In order to do so, Pass conducts a survey among 50, randomly picked, firms from the FTSE 250 firms list. However, he does not analyze the overall code compliance, but instead focuses on those aspects that are new in the revision of 2003.
The author finds that out of 50 firms 17 (34%) fully comply with the code. This is opposed by a figure of 11 firms (22%) who are in breach of the code, who neither take action to improve their governance nor provide acceptable explanations. The remaining 22 firms (44%) and hence the majority have undertaken efforts to improve governance or provide adequate explanations since the survey was conducted. A major aspect of concern is the status of the chairperson who is not independent of the committees. Moreover, independence in general seems to be a troublesome topic.
5 This research is supposed to be a pilot study for a more comprehensive study, which to our knowledge has
Grant Thornton (2008) also execute a similar study of the Irish case – which we opt to group under Great Britain as Ireland also applies “The Combined Code”. Again, verifiable propositions based on The Combined Code are applied to the firms. These are 32 of the Irish Stock Exchange, representing the majority of listed firms. 31% of them are said to fully comply, the remaining 69% comply with exceptions. There is not any detail as to how many firms – if at all – are non-compliant. Areas of concern in are board balance, the separation of CEO and CFO, independence of board member, committee membership and corporate social responsibility. Consequently, Grant Thornton (2008) conclude that Irish firms are still “a long way from demonstrating best practice I corporate governance”. Nevertheless, significant improvements in the quality of narrative reporting have been found.
2.2.2 The Netherlands
The first set of corporate governance recommendations that was issued in The Netherlands is the Peters Report from 1997. As compliance was rather low, it was replaced by the improved, so called Tabaksblat Code in 2003. As many other, the Dutch code relies on self enforcement and employs the “comply or explain” principle. Its 21 principles cover the following five general areas: 1) compliance with and enforcement of the code, 2) the management board, 3) the supervisory board, 4) shareholders and the annual general meeting (AGM) and 5) financial reporting.
Recent research by Akkermans et al. (2007) investigates the compliance rate of Dutch firms with the Tabaksblat code. Moreover, and this is unique, it analyses the nature of explanations for non compliance with the provisions. The research treats explanations also as compliance.
expected rather firm characteristics to be the dominant factor for non-compliance. Hence, they draw the conclusion that explanations tend to take the form of symbolic-adherence. Three code provisions which receive remarkably little compliance deal with disclosure of internal control mechanisms, the publication of contract details from management board members and finally with independency of supervisory board members. Furthermore, tests reveal that size has a positive relation with compliance.
Lastly, the authors draw the conclusion that the code itself is very complex, yet vague in some parts, especially regarding the required level of information and detail to be provided. Consequently it leaves room for misbehavior. Also, 20 per cent of the provisions cannot be evaluated by means of public information.
2.2.3 Greece
Greece experienced a period of positive economic performance from 1995 to 2000. This trend lasted also in the aftermath of the terrorist attacks of 11 September 2001. Due to its economic success, Greece joined the Economic and Monetary Union (EMU) in 2001. At the same time, the Greek stock exchange in Athens (ASE) grew faster in terms of market capitalization than any other stock exchange in the developed world.
The first code, the so called “Blue Book” was introduced in 1999 and lastly updated in 2001. It follows the “comply or explain” principle and its structure reflects the guidelines and directives of the EU.
Research of Tsipouri and Xanthakis (2004) aims at rating the quality of corporate governance of Greek firms. Their approach is to quantify the compliance of Greek firms with the Blue Book. A questionnaire along the chapters of the Blue Book is created, resulting in 52 questions. Answers are then assigned to several indicators. The five main categories of indicators are the following: 1) the rights and obligations of shareholders, 2) transparency, disclosure of information and auditing, 3) the board of directors, 4) executive management, and 5) corporate governance commitment, the role of stakeholders and social responsibility.
sample set of 120 firms. A particularity of these firms is that they are rarely internationally quoted.
As a result, the researchers find that Greek firms “demonstrate a fairly satisfactory degree of compliance” which amounts to approximately 70.3% (Tsipouri and Xanthakis, 2004). Several aspects are revealed where there is considerably less compliance. These are the role of stakeholders and corporate social responsibility (CSR), disclosure of remuneration and risk management as well as the independency of board members and the organization of corporate governance in general.
2.2.4 Germany
The first German code, known as the “Cromme Code”, was introduced in February 2002. It is its aim to increase the transparency and understandability of corporate management in Germany, as well as enhancing its quality by addressing critique at the German corporate governance (Steinat, 2005). In comparison to other European countries, noting that the international code movement started already during the 1990’s, the introduction may seem rather late. On the other hand, most central aspects that form the contents of foreign codes are already incorporated in German law (Werder et al., 2005). The code entails four kinds of recommendations which are of different relevance for the firms. Firstly, the Code presents existing corporate law with the purpose to facilitate the overview of relevant corporate regulation for (foreign) investors, as these are spread in various statutes.6 These are also the “Must” regulations. Secondly, the Code also interprets these laws; however, firms are free to have a different view. Thirdly, and this is the main body, there are recommendations of “best practice”, indicated by the word “Shall”. Here, firms have to publish non-adherence, but explanations are not required. Lastly, the code entails recommendations about the behavior of management and supervisory board which are not yet best practice, identified by “Can”. Here, firms do not have to publish non-compliance.
For German firms compliance with all but the “Must” regulations -which are essentially law - is voluntary. Nevertheless, it is a prerequisite for listed German firms to disclose annually whether they have been complying and where they deviate. It differs
6 Among others HGB (Handelsgesetzbuch – commercial law), AktG (Aktiengesetz – German Stock
from the “comply or explain” principle with respect to explanation of non-compliance, which is not required in Germany. Therefore, one could say that Germany employs a “comply or disclose” principle (Berg and Stöcker, 2002). Nevertheless, it is expected that numerous firms will offer explanations in order to satisfy informatory needs of shareholders and investors. This expectation is confirmed by Steinat (2005) who finds that indeed several DAX enterprises provide explanations.
It is noteworthy that the German code is revised on an annually basis, making amendments if necessary. In addition, the Berlin Centre of Corporate Governance (which is headed by Werder, A. von) has been given the mandate to execute regular empirical surveys to identify the acceptance of the code and to which extent firms comply with it.
The research by Werder et al. (2005) sets out to analyze the acceptance of the code among German listed firms and consequently studies 864 firms listed at the Frankfurt Stock Exchange. Due to non-responses the sample eventually consists of 408 firms who provided their compliance statements before the end of February 2003. To rate the individual compliances, a questionnaire was constructed of the code recommendations and suggestions allowing for “yes”, “no”, and “in the future” answers.
One finding is that compliance increases with size. Most firms (~ 91,8%) opted for the so called “model of selection” where large parts of the code are complied with but certain provisions are not followed. About 5 % of the firms fully comply with the code and roughly 3,2 % completely reject the code. However, those rejecting the code entirely are not member of any of the four indices (DAX, MDAX, NEMAX7, SDAX). The most firms completely accepting the code are found in the NEMAX index, giving room too speculation that this may be driven by their capital structure. In addition, Nowak et al. (2004) find that industry classification can have an influence, as some, like the banking sector, are more regulated than others. Also, different industries may have different capital market orientation. Furthermore, there is some first evidence that dual-listing in the US can play a role.8 On the other hand, the authors later argue that higher standards of
7 The NEMAX index includes mainly the market segment of new technologies and new start up firms. Due
to the collapse of the new economy it was succeeded by the TecDax in March 2003. Nevertheless, most of those firms listed in NEMAX previously are also in included in TecDax.
8
the US “are not reflected in the Code and thus, the Code is fairly irrelevant” for those cross-listed firms.
Not accounting for those who entirely reject the code, each firm discloses approximately six recommendations it does not adhere to. According to the authors 13 recommendations of the code are labeled “neuralgic” because “[…] they will not be complied with by more than 10% of all firms […]” (Werder et al., 2005). These neuralgic recommendations fall into six categories, namely 1) the personal liability of board members, 2 + 3) configuration and transparency of the remuneration of the executive and supervisory board members, 4) staffing of the boards, 5) structure of the supervisory board and finally 6) accounting requirements.
Overall, the compliance rate with the recommendations accumulates to 94,7 % whereas the compliance rate for the non-binding suggestions is still at 78,4%.
To conclude, it appears that the overall compliance rate in the four countries presented above is fairly high, yet displaying a considerable variation from around 71% in Greece to almost 95% in Germany. The aspect of independence of board members is a general concern, as well as disclosure of remuneration and contract details.
2.4 Research Question
Analyzing compliance research of the four European countries as presented above leads to the conclusion that the compliance rate among the firms is rather high and the respective Codes, despite imperfections, by and large fulfill their purpose. So, the next question to clarify is what determines compliance in general.
borderline between these two concepts in practice being small, it is a useful distinction since it hints at an underlying difference of attitudes towards regulation. May’s results validate the two concepts. He finds that negative motivations are indeed heavily influenced by inspection practices and expected penalties. On the other hand, affirmative motivation is shaped mostly by attitudes and beliefs of the regulatee and his knowledge of the rules.
Ability to comply implies that firms are not always able to comply, even if they are willing to. The diversity of firms in terms of size, industry etc. means that some are more capable to comply then others. Knowledge of the regulations and the financial resources of the firm are the main factors shaping the ability to comply. For instance, small firms in The Netherlands frequently do not have the need to web-cast their annual general shareholder meeting, as their ownership tends to be concentrated, often only in the domestic country.
down into coercive, mimetic and normative forces. First, coercive isomorphism is two-dimensional in the sense that organizations are as well dependent on resources of other parties, as well as they have to account for rules imposed by regulatory agencies or the state. Formal regulation in the Dutch context stems from the fact that the code is incorporated in Dutch law, which requires firms to annually disclose their compliance behavior. Also, there is informal coercive pressure as we learned that shareholders and investors use information about compliance behavior in their investment decision (Seidl, 2007). Moreover, findings by Edelmann (1992) which reveal that larger organizations are inclined to adjust more to public opinion than smaller ones, as they come easier under public pressure, indicate coercive pressures exerted by media. Second, mimetic isomorphism states that the adoption or imitation of best practices increases the legitimization of a firm. Akkermans et al. (2007) find preliminary evidence for mimetic isomorphism in code compliance in the sense that numerous deviations and motivations of non-compliance tend to be similar. Third and last, normative isomorphism means that in terms of educational background, diversity in firms’ management is limited. Hence, one expects management to come to draw similar conclusions and consequently undertake similar actions. This assumption is also implicitly voiced in the conception of codes which state best practices. These best practices which have been applied by successful firms are supposed to be diffused among the corporate landscape of a country, which consequently would lead to high compliance (Seidl, 2007). In conclusion, we can deduct that in contrast to efficiency seeking the concern for legitimacy is shaped and dominated by exogenous forces.
To conclude, four motivations of compliance are identifiable. These are deterrent fears, the duty to comply, the ability to comply and finally social (or: legitimacy) motivations. In the terminology of institutional theory, the latter embraces coercive, mimetic and normative forces.
frequently found and firmly established factor is firm size which exhibits a positive relationship with the compliance rate. Beyond this, some first evidence is found for the importance of industry membership and dual-listing, in particular with the U.S. Thus, there is some hint that the compliance rate can only be expected to vary if firms are diverse in their structure. Hence, the question is what can potentially cause organizational structures to be more diverse and to break up isomorphic pressures? In the wake of globalization –understood as increased economic integration (Khanna et al., 2006) – we argue that one factor could be the degree of internationalization of a firm. Internationalization implies activities abroad, increased attention from foreign media and investors, potentially a more diverse firm leadership and organizational structure in general. Furthermore, two of the above mentioned three factors influencing the compliance rate are to some extent linked to the degree of internationalization of a firm. Larger firms are usually more internationally oriented as entering new markets is the logical consequence at a certain stage if a firm wants to grow further. Also, it is in the nature of the concept of dual-listing that these are international. Nowak et al. (204) finally discard the importance of cross-listing as they state that higher US listing standards are not reflected in the German code. However, this could change if the research is conducted in another context and a code other than the German.
Consequently, as we believe that the degree of internationalization of firms results in more diverse organizational structures, influencing the drivers of compliance we postulate the following general research question:
2.5 Derivation of Hypotheses
Sullivan (1994) argues that “despite its theoretical and practical centrality, estimating the degree of internationalization of a firm remains arbitrary”. According to him literature identifies three components of internationalization, namely performance, structural and attitudinal internationalization (ibid).
Internationalization in terms of performance refers to activities undertaken by multinational enterprises (MNEs) in other than the domestic market. In principal these can be any operation an enterprise carries out at home, too. As examples one can think of research and development, production or marketing activities. Also, sales can be generated in the form of selling output from a foreign plant to the then local market or via exports from the domestic market. Structural internationalization relates to the resources a firm has abroad. For instance the number of foreign subsidiaries or assets placed in an alien country. Lastly, attitudinal internationalization is concerned with the international orientation of a firm’s top management. It can be expressed in the diversity of nationalities or by international experience.
It is the ambition of the present research to amalgamate these three facets of internationalization with the already identified, potential determinants of code compliance and to analyze their relationship towards the compliance rate.
We construct a matrix in Table 2 where we hypothesize how the three facets of internationalization influence the drivers of compliance. A positive relation is marked by “+” and a negative consequently by a “-“sign. If the relationship is arbitrary and could go either way “+/-“ is used. No cahnge is indicated by “0” and if we cannot determine an influence, this is marked by “/”.
Table 2 Effect of internationalization on compliance drivers
Performance internationalization occurs if a firm enters new markets. This happens if the domestic market is saturated or if the firm pursues an expansionary strategy. To identify the effects of performance internationalization on the drivers of compliance, one has to distinguish between the effects on the domestic and foreign market. As affirmative and negative motivation is rooted in the personality of firm leadership, these motivations should be equivalent regardless which market is concerned. 9 Although, negative motivation may increase as a new player in the market and new investments penalties could be especially harmful. Also, this would exert a negative influence on firm image, created in media. The question is, in how far this shows in compliance with the national code.
Due to presumably different regulations abroad, conflicts may arise with national regulations, influencing the ability to comply, especially as the opportunity exists to apply the code by motivated non-compliance. However, our research is only interested in the compliance rate, not the application rate. Hence, the effect remains arbitrary, it can go either way. If foreign rules are less strict, compliance with Tabaksblat should be possible, if they are stricter or pointing to a different direction, this may cause deviations from best practices. Eventually, this depends on the importance of foreign activities in relation to domestic.
Legitimization concerns in the foreign market should certainly increase. From a perspective of coercive forces due to new regulations and mimetic forces are wielded by uncertainty in a new market environment and the wish to conduct business in the appropriate manner. We leave the aspect of normative forces – as the attitude and cognitive frame of firm leadership – for later discussion, because it is strongly related to attitudinal internationalization.
To conclude, it is assumed that with increasing importance of foreign activities the overall compliance with the national code shrinks. Due to a reduced relative importance of the domestic market, the potentially restricted ability to comply and legitimization concerns abroad, that may be detrimental to compliance with the national
9 We assume that firm leadership composition remains unchanged. This assumption is relaxed when we
code, we assume a negative relationship between the degree of performance internationalization and the compliance rate with the national code.
In the line of Sullivan (1994) it is a common approach to account for internationality of performance by the measure “Foreign Sales as a Percentage of Total Sales (FSTS)”. Moreover, the higher this ratio the higher is the importance of foreign activities within the firm, making compliance with the national code relatively less important. Consequently hypothesis 1 states the following:
H1 : The higher the degree of performance internationalization of a firm in terms of FSTS, the lower its compliance rate with the national code.
Structural internationalization is generated by firms establishing themselves in foreign markets and developing “structures” there, such as subsidiaries, plants, divisions or any form of assets in general. Normally, a firm has the following broad options to enter a new market. The most common forms are exporting, licensing, joint venture and finally foreign direct investment (Agarwal and Ramaswami, 1992). These entry modes differ in their level of commitment of a firm, since they exhibit differences in their potential risk and return, their degree of market control and location advantage. At the low end, exporting requires small investments and bear relatively low risks and returns. By exporting a firm stays relatively far off the market, having low market control and location advantage. On the high end, foreign direct investment requires large investments and brings potentially high returns at the cost of a higher risk. Also, market control is high, as well as the location advantage since the firm is directly in the market.
regulations imposed in the foreign environment. In sum this should mean that with higher structural internationalization, compliance with the national code decreases.
The ability to comply is shaped by the knowledge of the rules and the financial and organizational capabilities of a firm. Here we assume that in case of high level commitment entry modes knowledge of the rules and regulations of the country to enter should be given. Moreover, as these are usually firms of larger size, they should dispose of the financial means to comply, too. However, depending on the nature of these rules and regulations, it may be that they are in conflict with those specified by the national code. Similar to the case of performance internationalization we believe that rules and regulations of a lower or comparable level do not negatively influence the compliance rate. If they are of a higher or alien nature, this may occur, however. Thus, the effect remains arbitrary.
Structural performance faces several kinds of influences on the legitimization driver. First, coercive pressures abroad arise, due to rules and regulations there. Moreover, there are mimetic forces which are grounded in uncertainty. As internationalization of a firm is often regarded as an incremental, step-wise process (Johanson and Vahlne, 1977; Anderson, 1993), we believe that uncertainty and therefore mimetic isomorphism should decrease with an increase of structural performance and hence a larger pool of experience gained by activities abroad. Concerning normative isomorphism it can be assumed that firm leadership also generates more diverse insights form doing business abroad and with foreign partners, influencing their cognitive frame and decision making process. Furthermore, if activities abroad gain in importance, so does the legitimization of them. But not only abroad, if foreign business becomes a considerable part of a firm, it has also to be legitimized to and by domestic stakeholders. In conclusion, there are several arguments pro- and contra within the discussion of legitimization aspects, resulting in an arbitrary effect. However, taking all the hypothesized effects together, we assume a negative relationship between the degree of structural internationalization and the compliance rate with Tabaksblat.
require detailed information which firms normally do not make publicly available. As an alternative, the authors believe that a related measure of employees abroad should qualify as a substitute for the proxy of overseas subsidiaries. Furthermore, the more employees work abroad the higher may be media coverage, exercising a higher level of pressure for the legitimization of corporate governance abroad, as well as the internal motivation to attract and retain well suited and qualified personnel abroad. Consequently hypothesis 2 stipulates the following:
H2 : The higher the degree of structural internationalization of a firm in terms of employees abroad, the lower its compliance rate with the national code.
Regarding the attitudinal aspect of internationalization, the international experience of the top management team is often taken as a proxy. Or, also a measure of so called “psychic dispersion” is employed. It stands for the cultural differences between countries/regions where the firm operates in and the cultural background of employees/management (Sullivan, 1994).In consequence, attitudinal internationalization refers to firm leadership characteristics.
Concerning affirmative motivation of management it is reasonable to assume that different nationalities/cultures also possess different attitudes. For instance, some nations are more risk-loving than others. It is possible to deduct this from the classification of business systems where Anglo-Saxon countries are considered to be thinking more in short-term time frames and to be more risk loving than - the classical counter example - Germans, who are rather risk-averse and long term oriented. These attitudes may have an influence on negative motivation, also.
Mimetic isomorphism at the same time should decrease, too. A variety of opinions, experiences and cognitive frames should obstruct ‘copying’. Therefore, we hypothesize in the end a negative relationship between the degree of attitudinal internationalization and the compliance rate with the national code.
The measures ‘Psychic dispersion’ and International experience of TMTs specified by Sullivan are difficult to operate. Information about where firms run their operations is needed in the former and details of the background of management in any case. However, as the board of directors - or management board- can be rated among the “most elemental components of a corporate governance system” (Bhagat et al., 2007), we rely on an alternatively and related proxy to be able to include the concept of attitudinal internationalization in our analysis. Our approach is to take the different nationalities of board members into consideration. Frequently, foreign board members have gained working experience abroad, before joining the board. Moreover, by coming from a different national background they are also likely to possess different cognitive frames. Bhagat et al. (2007) focus their research on the U.S. which commonly features a unitary board structure. In Europe in general and particularly in the Netherlands a two tier structure consisting of a management and supervisory board dominates. As a consequence, supervisory boards are included in the analysis, as well. Therefore, we split hypothesis 3 in part a and b, stipulating the following:
H3a: The higher the degree of attitudinal internationalization of a firm in terms of foreigners on the management board, the lower is its compliance rate with the national code.
Finally, as corporate governance codes apply mainly to stock-listed firms, we would like to account for another form of internationalization, outside the framework of Sullivan. Firms are not only listed at their domestic stock exchanges, but sometimes also cross-list at other stock markets of their interest. One hypothesized reason behind this cross-listing is the bonding hypothesis. The bonding hypothesis states that firms, which are incorporated in a country with comparatively weak investor protection, cross-list on stock exchanges with higher requirements in terms of superior disclosure standards and stricter enforcement. Hence, they “bond” themselves to higher principles. Coffee refers in his research particularly to the U.S. as being a country disposing of high values and consequently being attractive to foreign firms to be cross-listed at (Coffee, 1999 and 2002). A similar line of reasoning may hold for Dutch firms listing abroad, they perform cross-listing in order to upgrade the governance standards of their home country and/or to gain higher credibility by foreign investors. Examining the data, Dutch firms do frequently but not exclusively cross-list in the US. Charitou et al. (2007) set out to examine the relationship between cross-listing and corporate governance. They find that subsequent to cross-listing, firms dispose of more independent boards and audit committees. Hence, cross-listing has an influence on some of the “neuralgic” aspects of non compliance of firms as voiced by Akkermans et al. (2007) or Werder et al. (2005). These are recommendations most frequently not being adhered to and among others the authors identify independence of the supervisory board members as such. In another research, Nowak et al. (2004) find that German firms being cross-listed in the U.S. have a higher compliance rate than those which are not. Therefore, we expect a positive relationship between cross-listing and the compliance rate among Dutch listed firms. Consequently hypothesis 4 stipulates the following:
3 Data and Method
The research centers on an analysis of firms listed at the Dutch stock exchange.
Main reasons for this choice are data availability and the fact that the Netherlands follow the “comply or explain” principle. The latter is in our view better suited for analysis than data from countries where it suffices to identify paragraphs of non-compliance, without explanation. As a consequence of our choice we make use of a purposive judgment selection since sample members are selected to meet specific criteria (Cooper and Schindler, 2003). In our case it is the criteria of being listed on the Dutch stock exchange. Furthermore, the research is focused on one point of time, namely 2004 as the year when the Tabaksblat Code came firstly into action. A counter argument to use later years may be that firms then developed more of a routine of applying the code and had the necessary time to adopt their governance structure. However, if a firm indeed requires time to adopt its corporate governance, they have the opportunity to reveal this in the form of an explanation in their compliance section. This explanation is in line with the “comply or explain” principle then understood and rated as compliance.
3.1 Data
The following subsections explain the dependent, independent and control variables used in the statistical analysis and their measurements (3.1.1 – 3.1.3). The dependent variable as well as the control variable market size is obtained from data sets provided by the RUG. Data for the independent variables is collected from annual reports of the relevant firms by the authors and data for the industry control variables stem from the AMADEUS data base.
3.1.1 Dependent Variable
open-end funds. Due to an insufficient corporate governance structure the 5 funds are excluded from the analysis as are the 15 foreign firms, since these are not obliged to comply with the Dutch code. Thus, in the end a test sample of 130 firms remains.
Firms have the possibility to directly comply with code regulations in so far as they adhere to them. Alternatively they can opt for motivated non compliance, explaining deviations. To determine the compliance rates a set of 246 verifiable propositions was developed out of the code provisions and applied to the individual firms. Hence, the dependent variable will be the percentage of these propositions confirmed by compliance; motivated non-compliance is not accounted for.
3.1.2 Independent Variables
In order to test the effect of internationalization on compliance several variables which are each linked to one of the five hypotheses are employed. These are presented in analogue to the hypotheses. A data set containing the relevant information has been assembled from hand-collected data available from the annual reports of the individual firms. In addition, cross-listing information (H4) stems from AMADEUS data base.
Formulation of H1 necessitates a measurement of international activity. The proxy ‘foreign sales as percentage of total sales’ (FSTS) as suggested by Sullivan (1994) is used. Alternatively, a dummy variable “MNC 10” (for Multinational Corporation) is constructed. Sullivan (1994) uses 10% FSTS as an arbitrary criterion for firms being considered as MNC and the dummy is accordingly MNC10 (value = 1) or not (value = 0). To answer H2 the ratio of number of employees abroad to total employees is used. Alternatively, also full-time equivalents (FTEs) are taken into account. Normally, a firm reports either or but not both. If possible, the figures represent the year average.
To answer H3a a measure of board internationalization is necessary. For this the ratio of foreign management board members to total management board members is used.
To answer H3b we proceed in a similar manner as to answer H2 with the difference that we now focus on members of the supervisory board.
= 1) or not (value = 0). Alternatively, we include a dummy U.S.-listing which specifically determines whether a firm is listed in the U.S. (value = 1) or not (value = 0). This is done to account for the notion by Nowak et al. (2004) who find weak evidence that in the case of German firms’ dual-listing in the U.S. matters.
3.1.3 Control Variables
As mentioned before, our research also accounts for those potential determinants of the compliance rate as identified by prior research. Those which are related to internationalization are included as control variables. Compliance research in the field of corporate governance codes showed that there is a positive relationship between the size of a firm and its compliance rate (e.g. Werder et al., 2005; Akkermans et al., 2007). Moreover, research on internationalization of firms reveals that size is positively related to the probability of firms going international. Thus, we include size as a control variable. Size of a firm can be measured in several ways, for example total assets, number of employees, market value and many more. As the firms under scrutiny operate in many different industries the authors believe that it is not advisable to use the former two measures as they are related to the nature and structure of the firm. On the contrary, market value is a measure which is comparably less sensitive to industry effects. This data is taken from a data set called “Firm Data”, provided by the RUG.
Independent Variables Control Variables Dependent Variable
An organizing framework can be found in Figure 1 which relates the hypotheses and the measurements used for empirical analysis. The terms presented in brackets and in italics represent the abbreviations used in the regression formula and further discussion.
Figure 1 Conceptual Model
Concept Measurement
Code Compliance Compliance Rate
Internationalization of Management Board (IoMB) Internationalization of Activities (IoA) Internationalization of Supervisory Board (IoSB) Internationalization in terms of Listings (CL) Firm Size (MV)
Net Sales Abroad/ Total Sales Percentage of foreign management board members Percentage of foreign supervisory board members Number of Employees Abroad / Total Employees Cross-listings (Dummy) Market Value Industry Classification
(Id_d and Id_k)
Sector Dummy Internationalization of
3.2 Method of Analysis
Data are collected for a certain period of time (2004) over numerous sample units as for instance the compliance rate and the number of cross-listings. Hence, this research uses cross-sectional non-experimental data. To the best of our knowledge research examining the influencing factors of code compliance on the firm level has not been undertaken, yet. However, on the level of national codes the authors’ previous master thesis MSc IB&M (2006/2007) sets out to find the determinants of code content convergence across countries. Here they used multiple regression techniques.
Generally multivariate techniques are preferred to bivariate tests. The main disadvantage of bivariate tests, when for example firm performance is analyzed, is that they omit variables, which (better) explain changes in the dependent variable. Thus, bivariate tests are only useful in cases where one can be certain that there are no other variables influencing the dependent variable. Furthermore, bivariate tests have the drawback that they do not differentiate between the dependent and the independent variables, so causality becomes an issue. However, theory usually guides the analysis and solves this problem.
Consequently, as it is to be expected that various factors determine code compliance, multivariate analysis is used. The aim of multivariate regression is to predict the change in the dependent variable in response to changes in the independent variables. This is often achieved through the statistical tool of least squares (Hill et al., 2001). As we deal with several explanatory (independent) variables the appropriate analysis is the multiple regression, using underlying Ordinary Least Squares (OLS) assumptions, which is executed at the 5% confidence level. Eviews is used as the means of analysis. The general regression equation is the following10:
CR = c + ß1 IoA + ß2 IoWF + ß3 IoMB + ß4 IoSB + ß5 CL + ß6 log (MV) + ß7In_d + ß8 In_k + e
where e
is the error term.
Table 3 provides an overview how to interpret the coefficients of the model and the expected signs they carry.
10 We refrain from constructing four models for individual hypothesis testing, as our conclusions then
Table 3: Interpretation of the regression coefficients & expected signs Coefficient Exp.
Sign Interpretation of coefficients
β 0
The intercept: the value of the function when it crosses the Y line: Reflects the level of the Compliance Rate (CR) when all independent variables are equal to zero.
β
1 -
The change in the level of CR when the level of internationalization of activities changes.
β
2 -
The change in the level of CR when the level of internationalization of the workforce changes.
β
3 -
The change in the level of CR when the level of internationalization of the management board changes.
β
4 -
The change in the level of CR when the level of internationalization of the supervisory board changes.
β
5 + The change in the level of CR due to cross –listing. β
6 + The change in the level of CR when the level of firm size changes. β
7 +
The change in the level of CR due to industry classification D. (manufacturing).
β
8 +
The change in the level of CR due to industry classification K. ( Real estate, renting and business activities)
that the dependent variable could be treated as a limited dependent variable (LDV). Section 3.6.1 explains the concept and means of analysis.
3.2.1 Problems & Tests regarding the Statistical Analysis
As mentioned before, use of the ordinary least square (OLS) multiple regression model is made. To ensure that this model is valid several tests regarding the assumptions have to be performed. Table 4 presents the underlying OLS assumptions.
Before we come to these, we draw two assumptions about the explanatory variables. Firstly, in accordance with Hill et al. (2001), we assume that these are not random variables, in the sense that the values are known to us prior to observing the value of the dependent variable. In the present case it is reasonable to argue that information about the explanatory variables can be obtained from the firms prior to self assessing their compliance rate. For this assumption no formal test is needed. Secondly, none of the explanatory variables is an exact linear function of any other. In consequence we hereby assume that no variable is redundant. If this assumption is violated exact collinearity occurs and as a consequence the least squares procedure fails. Hence, it is necessary to be aware of multicollinearity. Multicollinearity occurs if two or more independent variables in a regression analysis are highly correlated. A high correlation means that the variables essentially contain the same information. Although the individual variable may not be significant individually, taken together they considerably contribute to the model and thus increase the R². In order to detect multicollinearity one has to generate the correlation matrix. If correlations are >0.8-0.9 one should become suspicious of multicollinearity as a strong linear relation is to be expected (Hill et al, 2001). A viable solution to overcome multicollinearity is to delete some variables or combine those which are related, if it makes sense.
Table 4 Assumptions of the Model
Assumptions of the Multiple Regression Model regarding error terms &dependent variable
1 E[et] = 0 Each random error has a probability distribution with zero mean. This assumption entails that the average of all omitted variables, and any errors in the model specification, is zero. Thus, the model is on average correct.
2 Var (et) = σ² The variance σ² is an unknown parameter and it measures the uncertainty in the statistical model. It is the same for each observation, so that for no observations the model uncertainty will be more, less, nor directly related to any economic variable. Errors with this property are said to be homoskedastic.
3 Cov (et, es) = O. The covariance between the two random errors two any two different observations is zero. The size of an error for one observation has no bearing on the likely size of an error for another observation. Thus, any pair of errors is uncorrelated.
4 Random errors have a normal probability distribution. Based on Hill et al (2001), p. 149 f
Assumption 1 is the essential condition that the model is correct. In order to prove this it is necessary to check the validity of the remaining three assumptions. Assumption 2 deals with heteroskedasticity, a common problem in regression analysis. Moreover, it is frequently encountered in analyses using cross-sectional data as the present research does (Hill et al, 2001). Heteroskedasticity means that the errors suffer from varying variances. Although this does not have an influence on the estimated coefficients it does influence the standard errors of them and consequently impacts the t-statistic. An appropriate test to detect heteroskedasticity is the White heteroscedasticity test. It features the null hypothesis that there is no heteroscedasticity. If we can accept the null hypothesis we can confirm the assumption of homoscedasticity. In case heteroscedasticity is present, it is possible to account for it by using the White heteroscedasticity consistent estimation method.
Concerning assumption 4, a formal test to check whether the sample data stems from a normally distributed population is the Jarque Bera test for normality of the residuals. The null hypothesis of this test assumes normality, so it follows that rejection of the null hypothesis corresponds with a non-normal distribution. The Jarque Bera test consists of both the measures of skewness and kurtosis. Skewness refers to the symmetry of values around the mean value and kurtosis refers to the peakedness of the distribution. Non-normal variables can lower the parametric tests’ validity. Nevertheless, should non normality be the case, we hold the view that our sample size of 130 firms11is sufficiently large so that a violation of the normality assumption has practically no consequences (Brooks, 2002).
Finally, a general problem is that of endogeneity which deals with reversed causality. In the present case it is not of concern since it is highly unlikely that the rate of code compliance has an influence on the degree of internationalization of a firm in its various measures.
Summing up, in order for the OLS estimation method to work properly the first three assumptions as presented in the table need to be fulfilled (BLUE). The fourth however is not necessary for OLS to work but it is important for applying t-tests which we eventually do. So, if assumption four is violated our inferences are likely to be invalid.
3.2.2 Limited Dependent Variable Analysis
A final remark concerning our data is the following. In line with earlier research using a comparable or identical dependent variable, the dependent variable is considered to be continuous (Werder et al, 2005; Akkermans et al, 2007).12 However, “[…] every variable we consider […] in econometric is limited in its range” (Maddala, 1983), which brings us to limited dependent model variables.
First we have to clarify what exactly constitutes a limited dependent variable (LDV), before we turn to means of analysis. Generally speaking, there are two categories
11
The RUG provides a data set of 130 firms for the dependent variable “Compliance Rate”. Unfortunately, the hand collected data set for the independent variables exhibits some missing values, leading to a smaller sample depending on the independent variables included in the regression.
of LDV, namely censored and truncated data. Consequently, there are also two kinds of models namely truncated regression models and censored regression models.
To draw the line between these concepts is not easy. A truncated distribution is the part of an untruncated distribution that is above or below some specified value (Greene, 2000). As an example consider an income analysis. If we ware only interested in the distribution of incomes above 80€ we have a subset of the full distribution ranging from zero income to (theoretically) infinity. If the dependent variable is censored it means that values in a certain range are all transformed (or reported as) to a single value. Underlying the censored data is a mixture of a discrete and continuous distribution. In contrast to truncated data “censoring is essentially a defect in the sampling data” (ibid). Censoring can occur in the following three forms. Left side censoring means that a data point is below a certain value, but it is unknown by how much. Similarly right censoring means that a data point is above by a certain value by an unknown amount. Lastly, interval censoring occurs if the data point is to be found within a certain interval (Klein, 2003). The censor points can occur naturally or may be specified by the researcher. The classical study of censored data is household expenditure (Tobin, 1958). Other examples include vacation expenditures (Melenberg and van Soest, 1996).
The compliance rate in the present research yields a percentage which can range from 0% to 100%. Therefore, the dependent variable in the analysis is restricted to values between 0 and 1. In this sense it is a LDV and the data censored in the interval from 0 to 1. However, our data does not stem from a censored sample. In a censored sample information about the regressand (dependent variable) is merely available for a limited number of observations (Gujarati, 1995). Analyzing data from a censored sample by means of OLS yields “biased and inconsistent” outcomes (ibid). We on the other hand have consistent observations for the dependent variable.
researchers frequently “compute OLS estimates despite their inconsistency” where another model specification is more appropriate. On the other hand, Maddala (1983) states that it is not always necessary to apply the more complicated analysis of limited dependent variable models.
Acknowledging the above and after careful consideration we proceed with OLS. This is analogue to previous research with the same data set for CR, which does not stem from a censored sample. However, in addition we run also a preliminary Tobit regression, especially as the multiple and stepwise regression models suffer from a breach of normality. This does not have to mean inappropriateness of the model, but is can be interpreted as a hint that censored least square analysis (CLS) and a Tobit regression is more suitable.
4 Results
Section 4 presents the statistical results of the regression analysis. Section 4.1 applies the statistical tests as expressed in the previous section to the general model. It turns out that the sample size in the general model is relatively low with only 29 observations included and the F-statistic is violated. Hence, an alternative model which removes two variables for the sake of an enlarged sample size is presented as a robustness check and alternative model in 4.2. Furthermore, section 4.3 introduces simple regressions to double check significance of independent variables and finally employs stepwise least square regression techniques to discover the best suited model. In this model, from our point of view the best suited, we alternatively substitute IoA by MNC10 and likewise CL by USCL, to take the alternative measures into consideration.
Finally, as the violation of the F-statistic and normality assumption arouses some concern that OLS is inadequate, we perform three sets of CLS (Censored Least Squares) regressions. The results13 confirm findings from 4.2 and 4.3. The only difference is with respect to the general model. Here, the CLS analysis seems to be valid. However, a more thorough analysis and assessment of the methodologies is left for future research.
4.1 Testing the model
This section begins with selected descriptive statistics of the variables to preliminarily judge the validity of the model. Thereafter, several formal tests regarding the consistency of the model and validation of conditions are performed. Specifically tests regarding heteroskedasticity, autocorrelation and multicollinearity are executed.
4.1.1 Descriptive statistics
The descriptive statistics show also the number of observations we have for each variable. CR is fully available with 130 observations whereas IoWF has the lowest number of observations with 73. If the regression is run, only those firms are included in the analyses which have valid entries for all independent variables. The others are removed. As a consequence, only 29 observations are included as can be seen in Table 8, page 40.
Table 5 Descriptive Statistics
Variable Std. Dev.
Mean Skewness Kurtosis Jarque-Bera Probability Observations
CR 0.134 0.868 -3.759 21.298 2119.633 0.000 130.000 IOMB 0.254 0.167 1.354 3.751 29.279 0.000 89.000 IOSB 0.214 0.169 1.020 3.004 18.565 0.000 107.000 IOA 0.351 0.555 -0.360 1.676 8.335 0.015 88.000 IOWF 0.329 0.596 -0.594 1.991 7.385 0.025 73.000 CL 0.442 0.263 1.078 2.163 26.314 0.000 118.000 ID_D 0.493 0.404 0.393 1.155 19.114 0.000 114.000 ID_K 0.427 0.237 1.238 2.533 30.157 0.000 114.000 MV 2.238 5.943 0.039 2.513 1.022 0.600 101.000
Moreover, Table 5 exhibits that only MV is normally distributed. Consequently, it is important to test for normality in the general model. Should it be violated sample size becomes crucial to assess validity of the model. In case of a sufficiently large sample size this should not be of concern (Brooks, 2002).
