Master’s thesis
Determinants of Subsidiary Autonomy:
Evidence from European Multinational
Corporations
Groningen, November 2008
Author:
Reinout Camphuijsen
S1362682
MSc in International Economics and Business
Faculty of Economics
University of Groningen
Determinants of Subsidiary Autonomy: Evidence
from European Multinational Corporations
ABSTRACT
In this study on the determinants of subsidiary autonomy, we apply three statistical models, OLS, Poisson and negative binomial, to estimate the effect of the included variables on the degree of autonomy of subsidiaries in Europe. The most appropriate model is used to see how the autonomy of these affiliates of European multinational corporations is affected. We also deployed partial regression models to see how the significance of the variables is influenced, and to learn about the interaction between these. We employed characteristics of the parent company, characteristics of the environment, and subsidiary characteristics, for 232 subsidiaries of 10 MNCs from 20 countries. Among other things, out of 13 hypotheses, we found the following variables to significantly affect the degree of autonomy: degree of product diversification, degree of unionization, power distance, subsidiary age, extent of ownership of the subsidiary, and subsidiary size. The remaining, insignificant variables are also discussed in terms of why they do not seem to have an influence on subsidiary autonomy in our sample.
Key words: subsidiary, autonomy, multinational corporations, culture, size, diversification,
Table of Contents
1. INTRODUCTION ... 4
2. THEORETICAL BACKGROUND ... 8
2.1 Parent company characteristics... 8
2.2 The local environment ... 11
2.3 Characteristics of the subsidiary ... 19
3. THEORETICAL MODEL ... 22
3.1 Hypotheses... 22
3.1.1 Parent company characteristics ... 23
3.1.2 The local environment... 25
3.1.3 Characteristics of the subsidiary... 30
3.2 The theoretical models... 33
4. RESEARCH METHODS... 37
4.1 Method of analysis... 37
4.2 Variable definitions ... 37
Explanatory variables ... 38
4.3 Sample and data ... 39
4.4 Variables and measures ... 40
5. EMPIRICAL RESULTS... 43
5.1 Description... 43
5.2 Correlations ... 45
5.3 Findings ... 46
5.3.1 The full model ... 50
5.3.2 The isolated groups ... 52
5.3.3 Combinations of groups ... 53
6. DISCUSSION ... 56
7. CONCLUSIONS, LIMITATIONS AND FUTURE RESEARCH ... 63
REFERENCES... 66
APPENDIX... 70
APPENDIX 1: EXPLANATION OF ALGEBRAIC RESULTS ... 70
List of tables and figures
Table 1 Descriptive statistics……….44
Table 2 Correlations matrix and R2s of auxiliary regressions………45
Table 3 Summary of regression results………..47
Table 4 Summary of partial regression results………...54
Table 5 EVIEWS output of the OLS model………...71
Table 6 EVIEWS output for White heteroskedasticity test...72
Table 7 EVIEWS output of the Poisson model………..73
Table 8 EVIEWS output of the negative binomial model……….74
Table 9 EVIEWS output of the partial regression parent only………..75
Table 10 EVIEWS output of the partial regression environmental only……….75
Table 11 EVIEWS output of the partial regression subsidiary only………76
Table 12 EVIEWS output of the partial regression parent and environmental………76
Table 13 EVIEWS output of the partial regression parent and subsidiary………..77
Table 14 EVIEWS output of the partial regression environmental and subsidiary……….78
Figure 1 Normal distribution of autonomy degree………71
1. INTRODUCTION
Globalization is a phenomenon that has become increasingly important in the past decades. In addition, vast advances in technology have been achieved, together leading to dramatic alterations in the way businesses are organized and how they compete. In light of this, managers are continuously searching for the ‘best’ way to manage their organizations. The subsidiary of the multinational corporation is part of the organization, and subject to a wide array of different practices, stemming from the parent company. For instance, human resources in all its aspects are an increasingly important issue in literature over the past decades, especially regarding the relationship between headquarters and its subsidiaries. Several studies indicated that due to the globalization process the relationship between the parent company and its foreign subsidiaries has become more complicated and sometimes conflictory (e.g. Edwards et al., 2002). Multinationals are becoming a major form of business in the global market, and this of course brings about its own problems. One of these is the delegation of control within the multinational. Those studies have pointed out that the division of decision-making authority between headquarters and the diverse operational units responds to a complex set of factors. As the multinational enterprise becomes more dispersed throughout countries with all very different characteristics than those of the home country, there is a necessity for rethinking the authority channels as practiced by the headquarters. Some markets are sufficiently large or important, or changing that rapid response is needed in order to be able to sustain market presence. White and Poynter (1984) assert that some MNCs grant their foreign affiliates a great deal of independence, whereas other MNCs tend to assume more direct control of their subsidiaries’ activities. In effect, a multinational can use a wide and distinct package of measures to control its operations, but finally it will come down to the question how much autonomy is granted to the subsidiary to carry out its operations how it sees fit. Therefore, many studies have tried to find out where autonomy stems from. Many different facets have been highlighted as possible determinants in some studies, and afterwards many have also been contradicted by different studies. Due to these ambiguous results in the international business literature on subsidiary autonomy, this study wishes to shed more light on the issues at hand. Therefore, we aim to answer the following research question: ‘What aspects determine the degree of autonomy of European multinational
This study is to some extent a follow-up research to the study performed by Vo Van Dut (2007). He conducted research with the same research question as a guideline, but with largely different hypotheses and variables. In summary, he researched the effects on the degree of subsidiary autonomy of the following variables: overall strategic approach of the MNC; the degree of product diversification; parent company size; the degree of institutional embeddedness; subsidiary age; economic performance of the subsidiary; extent of ownership of the subsidiary; subsidiary size; and finally, the relatedness between home and host countries. While keeping some of these variables intact and dropping others, this study adds a number of variables to the equation, too, such as culture. Vo Van Dut (2007), among other things found significant results on all variables but the degree of product diversification and the relatedness between home and host countries. His study has been, thus, an inspiration to us in many ways, but in no way should this study be considered as a somewhat adjusted copy.
In previous literature, different studies have focused on different aspects to find out what determines autonomy in decision-making. According to Björkman (2003), these aspects are to be divided into three groups: characteristics of the subsidiary, characteristics of the parent company, and the characteristics of the local environment. As argued above, the results of this literature are mixed and ambiguous, especially depending on which characteristics have been examined. For example, results regarding the characteristics of the parent company are mixed and no clear understanding has been developed, while the effect generated by the characteristics of the subsidiary on autonomy is found to be more consistent and obvious. In addition, the impact of the local environment on subsidiary autonomy is a research area that has received much less emphasis, and could be seen as somewhat underdeveloped, especially with regard to the host country characteristics and culture. As far as we know, no study has yet examined the combination of the three groups of characteristics mentioned above on the degree of subsidiary autonomy with empirical evidence. Particularly, the present study is novel in combining all characteristics including culture in relation to European MNCs and their subsidiaries in Europe. Hence, in order to answer the research question above, we have developed the following investigative questions:
Do the parent company characteristics affect the degree of subsidiary autonomy? Does the local environment of the host country influence the degree of subsidiary
Do the characteristics of the subsidiary affect its degree of autonomy?
Several reasons for and contributions of this study can be pointed out to examine the degree of subsidiary autonomy in this way. First, we have mentioned above that previous studies have focused on only one or two groups of characteristics, not on the combination of the three. Second, the notion of culture is an underdeveloped issue with regard to literature on subsidiary autonomy. In addition to this, the local environment group in this study includes the degree of unionization of a country, which, as is culture, is an issue that has received little attention. Third, most previous studies focused on a small set of (European) countries, whereas this study includes 232 subsidiaries of 10 multinational corporations in 20 countries in Europe. Fourth, it is expected that the three groups have a substantial impact on the degree of autonomy of foreign affiliates, due to the fact that these factors have a direct influence on the relationship between the overall MNC and its subsidiaries. Fifth, using the method introduced by Hedlund (1981), we have developed our own construct of the degree of subsidiary autonomy, by measuring it as a count variable. This is significantly different from the constructs used in previous literature. Finally, we present an econometric model that evaluates the effect of and provides evidence for the impact of the characteristics pertaining to the three groups mentioned above. Therefore, this study enhances the existing literature with a substantial expansion of the covered area, regarding both the increased data set as well as the ‘full’ body of characteristics.
2. THEORETICAL BACKGROUND
There are many attributes to a company or to an environment that can have an influence on organizational design. Likewise, the body of international business literature on attributes determining the extent to which a subsidiary is autonomous is just as numerous. Some of these characteristics have been researched extensively, whereas others have been merely brought up. Either way, three main areas of characteristics can be found in the international business literature. Goehle (1980) and Hedlund (1981) identified the first two groups, being characteristics of the parent company and those of the subsidiary itself. Later, there were institutionalists such as Soskice and Whitley who identified the third, imperative area. It is the local environment. This last one is, so far, much less developed than the other two. Nevertheless, these are the three main elements that are relevant to our study: the characteristics of the parent company, the local environment, and the characteristics of the subsidiary.
2.1 Parent company characteristics
The characteristics of the parent company dealt with in the business literature are numerous. However, there are some that return in different studies, and there are some that do not receive much attention. Three of those characteristics will be reviewed below. The first two of them are classic characteristics, whereas the last one has been subject to somewhat less attention. They are: the degree of product diversification, the parent company size, and geographic diversification.
The degree of product diversification
authors focused on the overall level of centralization (e.g. Egelhoff, 1982, 1988). In this regard, it can be stated that, in general, rather gross measures of diversification had been used in the process of researching the relationship between parent’s product diversification and subsidiary autonomy (Hedlund, 1981). Perhaps it is therefore, that in literature dealing with this theme, mixed results are found (e.g. Garnier, 1982). Either way, the extent to which the company is diversified in terms of its products has been mentioned by several authors as contributing to an explanation of the variation in the degree of subsidiary autonomy (Hedlund, 1981).
Goehle (1980) finds inconclusive results with regard to diversification and autonomy across functional areas, which is partly caused by the small sample size (ten). Moreover, Taggart and Hood (1999) in their study reviewing the determinants of subsidiary autonomy of German and Japanese subsidiaries in general, find no support for their hypotheses, as their results show up insignificant. In congruence with this, Edwards et al. (2002) test for a sample of subsidiaries taken in Malaysia, whether or not autonomy relates to the specificity of products and they assume that greater autonomy may be granted to subsidiaries that sell specialized products into niche markets, as opposed to commodities into competitive markets. However, no relation was found between the degree of specialization and decision-making and their assumption was rejected.
Vachani (1999) researched product and geographic diversification and its effect on subsidiary autonomy. His theory is broken down into related and unrelated product diversification. The results stemming from his analysis showed support for the hypothesis that increased product diversification as a whole has a negative effect on subsidiary autonomy. However, Vachani (1999) adds that the effect depends on the structure of the multinational organization.
Gate and Egelhoff (1986) performed a study on many possible determinants of subsidiary autonomy. They hypothesized that the degree of centralization is negatively associated with product diversification. Indeed, the result of their research showed that an increase in the number of products exhibits subsidiaries with more autonomy granted. This result is in accordance with the findings of Johnston (2005), who also finds a significant positive relationship between the degree of product diversification and subsidiary autonomy.
Parent company size
variables of structure and autonomy (Child, 1973; Garnier, 1982). Therefore, it has also been considered frequently, especially in early literature (e.g. Pugh et al. 1969; Child, 1973; Garnier, 1982; Gate and Egelhoff, 1986). The general assumptions with regard to MNC size deal with, for instance, the controllability of the corporation. The theory behind this emphasizes that increased size would lead to more structuring of activities which then facilitates decentralization. Furthermore, MNC size forces the parent company manager to decentralize decision-making, as an overload of decisions is pending with increased size.
Garnier (1982), for example, tested the relation between MNC size and subsidiary autonomy for 42 French and 102 Mexican affiliates of US multinationals. In this article, the results made clear that size might have just a minor influence on the division of decision-making power. On the other hand, Hedlund’s (1981) results from his research among Swedish subsidiaries, with regard to parent company size as related to autonomy, were inconclusive. In addition, Gate and Egelhoff (1986) get mixed results after testing for size and autonomy among 50 large US, UK and continental European MNCs. Although the results are as expected for financial centralization, marketing centralization has the opposite sign, which led to the before mentioned mixed results.
Wilkinson et al. (2008) researched the effect of cultural distance to control exerted by headquarters, among Japanese MNCs. They used, among other things, MNC size as a control variable in their analysis and found, as hypothesized, a positive relationship between size and indirect control (i.e. decentralization) methods.
The degree of geographic diversification
possible determinants of subsidiary autonomy. Nonetheless, several authors have addressed this subject in their articles.
Garnier (1982) hypothesized that the larger the multinational group is the more difficult it would be to control. In addition, the more affiliates the MNC has in different countries, so the larger the number of countries it is operating in, the more decentralized control would be, especially because this increased size would make the group less sensitive to risk. However, the results of their investigation showed that the direction was opposite to what was expected. Put differently, the more countries the MNC is operating in, the more centralized control is exercised. Garnier (1982) adds, though, that the variable has little effect on autonomy.
Interestingly, Picard (1977), Goehle (1980), Garnier (1982) and Gate and Egelhoff (1986) hypothesized a negative relationship between subsidiary autonomy and geographic diversification. As they put it, centralization is hypothesized to be positively related to the size of foreign operations of the MNC. They maintained that as the percentage of the firm’s revenues from abroad rises, dependence risk increases. In response, headquarters exercises greater direct control over its subsidiaries. Whereas Garnier (1982) found support for his hypothesis, the tests run by Picard (1977) and Goehle (1980) showed insignificant result, and, finally, Gate and Egelhoff (1986) found contradictory results. In their case, the increased foreign operations variable is negatively related to centralization of decision-making. All together, one can conclude these are mixed results.
2.2 The local environment
The institutional environment
The European neo-institutionalist school, which is many times referred to as the ‘business systems’ approach, predominantly refers to the work of economists such as Whitley (1999), and Hall and Soskice (2001). Whitley (1999), for example, argues that depending on the institutional frame of a particular economy, a customized national business system (NBS) evolves. This national uniqueness affects MNCs in the way that it has consequences for organizational practice. Goehle (1980) highlights that the organizational structure of an MNC should be the product of the adaptation process of the enterprise to the numerous national environments it operates in. Therefore, Birkinshaw and Hood (1998) state that subsidiary autonomy is not only determined by the relationship of the head office to its subsidiaries, but also by the nature of the local institutional environment in which the subsidiary operates. They argue that the nature of local legal conditions, the cultural environment, and the influence of the local authorities will all impinge on the degree to which subsidiaries have local control.
the power resources and strategic choice to resist and influence the implementation of global practices are limited.
Yang and Kang (2005), in their study on Taiwan-based multinationals, find evidence for the increased autonomy of subsidiaries that are highly embedded in their local business system. The higher level of decentralized control in this case was both due to the subsidiaries’ linkages with upstream and downstream players in the same sector, as well as linkages with several institutions, such as educational institutions and labor unions.
On the other hand, Richards (2000) conducted a research to see whether US MNEs provide for differences in autonomy with regard to distinct regions in the world. On basis of personal interviews with managers at 34 US multinational HQ, along with their 23 UK, 26 Thai, 15 Malaysian, and 9 Singaporean affiliates, she found evidence that, in fact, the US multinationals under consideration did not differ in autonomy provided between the UK and the Southeast Asian countries. However, it can be argued that conclusions drawn based on this rather small data set, in both number of countries as well as the number of MNCs, are, to that extent, open to doubt.
Perhaps more importantly, Harzing and Sorge (2003) performed a research with regard to the impact of the country of origin and universal contingencies on corporate control. The study was based on a multi-country data set comparing multinationals from Europe, as a whole, and separately to those from the US and Japan. First of all, they find that concepts that are related in one way or another to control by the HQ over subsidiaries seem to be heavily influenced by the country of origin of the larger concern. Moreover, they conclude that the multinationals from Europe cannot be treated as a homogeneous group, and, therefore, that there is no visible isomorphism in control mechanisms in Europe. In addition, they mention that the nature of internationalization strategy appears possibly to be more decoupled from organizational control than would be the case in a nationally homogeneous enterprise context. Harzing and Sorge (2003) argue that this more general finding, lends more weight to the view that societal institutions constitute ‘different but equal’ practices. From this picture, a notion of universal state-of-the-art practice cannot be ascertained and demonstrated. Societal context and domestic economic strengths appear to define a particular ‘rationale’ of international organizational practice.
national business systems promote a prevalence of sectors and subsectors in national enterprise populations. Additionally, Ferner et al. (2005), in a study on international transfer policies between US multinationals and their subsidiaries, found that, due to differing institutional contexts, transferring policies is not always possible, because the international diversification initiatives of the US MNCs were based on the American environment and institutional settings.
The authors of these articles, thus, highlighted the point that national business systems are highly country specific, with many factors causing differences. Some of the major institutions to cause differences in national business systems are to be found in the rights of employees, constituted in committees of co-determination and collective bargaining, and trade or labor unions (Hall and Soskice, 2001; Dore, 2002; Croucher, Gooderham and Parry, 2006; Geppert and Williams, 2006; Fenton-O’Creevy et al., 2007). Therefore, the notion of the importance of labor unions has earned plenty of attention in the academic literature on national business systems.
Fenton-O’Creevy et al. (2007) have conducted a study on human resource management (HRM) in US subsidiaries in Europe and Australia. The research focused on the determinants of subsidiary autonomy in setting HRM practices of the before mentioned US-parented firms in Europe and Australia, examining both the effect of strategic context as well as the effect of the institutional location of the subsidiary. They investigated the impact of the NBS in which the subsidiary is located and, separately, the impact of labor unions on the multinationals ability to exert centralized control. Fenton-O’Creevy et al. (2007) found that US MNCs show greater centralization of control where the subsidiary faces global markets, and where it faces low union density. If the subsidiary is located in a highly integrated national business system, the MNC will also want more centralized control. This seems at odds with the preceding literature, however, Fenton-O’Creevy et al. (2007) argue that in the context of divergent institutional settings, US MNEs will seek centralized control except when they are dependent on cooperation from labor unions substantially entrenched within the subsidiary.
Other papers (e.g. Ortiz, 1998; Giardini, Kabst, and Müller-Camen, 2005; Singe and Croucher, 2005) also deal with the effects of labor unions on the control headquarters want to exert on their subsidiaries in respectively, Spain, Germany and the UK, albeit not directly related to autonomy.
avoided entering in collective agreements with labor unions, thereby avoiding the link between the structures of the regulatory environment and the firm (Royle, 2000). In these cases, centralization of control by MNCs is significantly more feasible than in cases where subsidiaries have to confront union organizations (Royle, 2000; Fenton-O’Creevy et al., 2007). On the other hand, as reported by Gooderham and Nordhaug (1997), even in less integrated national business systems, labor unions, when present, are able to exert an influence that obliges management to take their view into account.
Summing up, the theory dissertated in literature on national business systems highlights the importance of institutions on the ability of MNCs to seek centralized control. The higher the embeddedness of a subsidiary into the local environment, the more difficult MNCs will find it to impose standardized regulations and control, especially so, where subsidiaries face institutions which are empowered by local regulations. Probably the most important of these institutions are labor unions (as labor unions typically have representatives in co-determination councils, as well as that labor unions provide resources and support to these councils). Following this reasoning, the presence of strong unions will make it even more difficult for parent firms to exert centralized control over their subsidiaries.
The development level of the host country
The development level of the host country is a variable that has not received as much attention in literature on subsidiary autonomy, as have other variables. Nevertheless, it could be an important determinant of the control exercised by the parent firm. The majority of studies on the subject deal with the development of subsidiaries and their relations with headquarters in the advanced economies. Significantly less research has been undertaken into the subsidiaries of MNCs that are operating in emerging economies and in transition economies (Varblane et al., 2005). It is well established that the role of a subsidiary varies according to such contingencies as the local environment, the structural context imposed by the parent, and the entrepreneurial capacity of management (Birkinshaw and Hood, 2000). Therefore, the autonomy ratio of the foreign subsidiaries of MNCs might have a strong relation to the development level of the host country. The more developed the host country the more responsibility could be given to the local unit of the multinational company (Männik et al., 2004).
are economically fast-growing and structurally volatile. Consequently the external networks of subsidiaries in these countries are quickly changing, providing bases for much more rapid change in the capacities and also in their role in internal (corporate) networks (Hoskisson et al., 2000). This is especially true where the MNC internal network mainly consists of subsidiaries that are located in countries with a stable economic environment (Varblane et al., 2005). Therefore, several authors have hypothesized that subsidiaries located in more developed countries will have a higher degree of autonomy than subsidiaries situated in less developed countries.
For instance, Edwards et al. (2002) show that the more developed is the country in which the subsidiary is located, in the sense of demand, existence of potential sourcing partners and level of the national innovation system, the higher is the likelihood that the subsidiary could develop an extensive external network, improve different capacities, and finally gain more autonomy. In addition, Männik et al. (2004) and Varblane et al. (2005) found a significant and positive relationship between the development level of the country in which a subsidiary is located and its autonomy.
The culture of the host and home country
tangible. One of them is the GLOBE project, which is an extension of the Hofstede (1980) framework. In recent years a group of over 150 researchers conceptualized and developed measures of nine cultural dimensions. This could, thus, be seen as a significant addition to the previous framework. However, so far, it has not been as widely used in literature as the Hofstede (1980) framework. Moreover, Leung et al. (2005) have argued that although subsequent work on culture has led to some clarification and refinement, the Hofstede dimensions are robust.
The literature on subsidiary-management constantly considers how culture defies the ability of managers to generalize strategies (Brott, 1984; Herbert, 1999). Differences in culture not only bring into question to what extent conclusions are to be generalized, but also create a certain bias in the development of the tools of analysis (Adler, 1983). It is thus not satisfactory to take theories developed in the US, for example, and just apply them to foreign countries. On the other hand, their applicability abroad should not simply be ignored (Paterson and Brock, 2002). Much work in literature has focused on the relationship between culture and international business phenomena such as entrepreneurship, mode of entry (i.e. mergers and acquisitions), and corporate governance as a whole. There are, however, few studies with an emphasis on differences in culture and its association with control exerted by multinational corporations. Nonetheless the size, some body of literature does exist on the matter, but the studies have produced conflicting findings: one stream of research argues that when cultural distance is greater firms increase their level of control; while the other stream suggests that greater cultural distance is associated with a loosening of control (Wilkinson et al., 2008).
A significant early study in this area was that of Daley et al. (1985), who examined the attitudes of US and Japanese managers and controllers towards aspects of budgeting and control systems design. Daley et al. (1985) found differences in attitudes in respect of some of the budget and control systems practices, but similarity of attitudes in respect of others. However, this study only confines two countries, and, therefore, its generalizability to other countries is limited. Also of particular interest is Newman and Nollen’s (1996) study, which used Hofstede’s national culture dimensions to determine that more employee participation (analogous with subsidiary participation) may actually hinder performance levels in high power distance countries such as those in Latin Europe and East Asia.
dimensions in a regression model to see whether distance affects the control exercised by headquarters for Japanese MNCs. In addition, they hypothesized that the impact of cultural distance on control mechanisms would be more important for newer than for older subsidiaries. They claimed that their hypothesis holds for the sub-dimensions of cultural distance, although not every dimension was statistically significant in the regression models. The strongest effect was found in the ‘uncertainty avoidance’ dimension. In addition, Barr and Glynn (2003) related the Hofstede (1980) cultural dimensions to controllability. Particularly, they hypothesized that the dimension ‘uncertainty avoidance’ (UA) would be an important indicator for the multinationals preference for control, as it affects the standpoint towards threats and opportunities. The results showed that the cultural value of ‘uncertainty avoidance’ had a significant effect: compared to low UA cultures, individuals from high UA cultures were significantly more sensitive to controllability in perceiving strategic issues.
Harrison et al. (1994) have provided probably the most interesting study with regard to the relationship between culture and parent control. For a sample of multinationals and subsidiaries from the US, Australia, Singapore and Hong Kong, they tested how the cultural dimensions developed by Hofstede (1980) influenced, among other things, the control exercised by headquarters, and how this differs between the Asian country sample, represented by Singapore and Hong Kong, and the Anglo-American country sample, represented by the US and Australia. The hypotheses and the regression results were, thus, based on the cultural dimensions and the inherent values possessed by members belonging to a certain cultural group, as proposed by Hofstede (1980). Harrison et al. (1994) found that, compared to the Singapore and Hong Kong, there was a greater use of decentralization in the US and Australia. In contrast, the results showed that there was a greater emphasis on long-term planning and group centered decision-making in Singapore and Hong Kong. One other hypothesis, namely that there is a greater degree of formalization of planning and control in organizations in Singapore and Hong Kong compared to Australia and the US, was found to show insignificant results, although the results did have the right direction.
2.3 Characteristics of the subsidiary
The subsidiary of the MNC has been analyzed in many different contexts. Firstly, there is literature focusing on the roles and strategies of the subsidiary (e.g. White and Poynter, 1984). Second, there is a literature stream dealing with subsidiary management (e.g. Birkinshaw and Morrison, 1995). Nevertheless, probably the most important subject in literature on subsidiary autonomy, in terms of presence, is how the specific characteristics of the subsidiary itself affect its independence (e.g. Hedlund, 1981; Garnier, 1982; Gate and Egelhoff, 1986; Taggart and Hood, 1999; Edwards et al., 2002). Many different factors have in this body of literature been dealt with. Characteristics frequently returning and that are assumed to have an important impact on subsidiary autonomy are: subsidiary age, economic performance, extent of ownership, and size.
Subsidiary age
Since age implies experience, the date of establishment of a subsidiary may be regarded as a key consideration (Young and Tavares, 2004). The experience possessed by an older subsidiary stems from and is displayed in both the host country market as well as the negotiating position within the multinational group. Keeping this in mind, several studies have argued that older subsidiaries have greater autonomy than younger ones (e.g. Goehle, 1980; Garnier, 1982; Taggart and Hood, 1999). However, different authors find different results.
Garnier (1982) found that age has little direct influence on the degree of subsidiary autonomy, as his results showed up insignificant. This was supported by Young et al. (1985), who find no clear link between the age of a subsidiary and its decision-making autonomy, as well as Goehle (1980). Taggart and Hood (1999) also associated age with autonomy. Their results only showed significance in part of the sample: among the full sample of German subsidiaries in the UK, autonomy rises with age; however, among Japanese subsidiaries the opposite was true. In addition, Gate and Egelhoff (1986) in their study on determinants of subsidiary autonomy in general, found that the more autonomous subsidiaries were, in fact, those that were older, but this only held for financial decisions.
and found that most literature supported that experience led to autonomy (e.g. Harzing, 1999). This same result is found by Harzing and Sorge (2003), in their study on the effect of country of origin and ‘universal’ contingencies on corporate control. They showed that the age of the subsidiary had a negative influence on corporate direct control, implying greater decentralization of decision-making. A comparable study by Wilkinson et al. (2008) showed that older subsidiaries receive less direct control than newer subsidiaries. They argue that age has a moderating effect on control and cultural distance.
Subsidiary performance
The performance of a subsidiary is considered to be a potentially important determinant of autonomy (Young and Tavares, 2004). On the one hand, it is argued that performance has to pass a certain threshold before headquarters turn attention to a subsidiary. If it is only moderately bad, the parent will not take action, while if performance is exceptionally good it may attract the attention of headquarters and lead to interference in the firm (Hedlund, 1980). On the other hand, Goehle (1980), for example, argues that successful managers will have more decision-making freedom than their counterparts do. In addition, Birkinshaw (1996), in his work on subsidiary mandates, found that after poor performance, or a lack of distinct value added, a subsidiary might lose its mandate. In a similar vein, Ferner (2000) found that poor performance significantly weakens the political influence of local management.
Extent of ownership
Gate and Egelhoff (1986) argue that the extent of outside ownership of the foreign subsidiary of an MNC is a common source for external complexity. Outside shareholders in the foreign affiliate will primarily be interested in the goals and performance of the individual subsidiary, whereas the MNC group has interests that range much further. Outside equity-holders may seek or prefer short-term profit maximization or non-economic goals. In light of this, it is hypothesized that a higher extent of outside ownership will increase the tendency to decentralize, due to the lower legal dependence of the subsidiary on the parent (Garnier, 1982; Gate and Egelhoff, 1986).
Moreover, in the case of majority ownership, the parent multinational is legitimately capable of imposing centralized control over a subsidiary, whereas in a joint venture or a minority ownership situation the MNC group might lack the authority. This would automatically mean that a lower extent of ownership by the MNC group means more decentralized control (Alsegg, 1971).
Both Garnier (1982) and Gate and Egelhoff (1986) find their hypotheses supported. They confirm that centralization is negatively related to the extent of outside ownership of a subsidiary of an MNC. Männik et al. (2004) hypothesized that more autonomous subsidiaries are among minority-share foreign ownership, since, in this case, there would be less responsibility carried by the MNC. Their results support the hypothesis, and therefore the previous results of other studies, except for the article by Birkinshaw and Hood (2000). They showed that outside ownership is negatively related to its autonomy. Hedlund’s (1981) notion was comparable to Garnier (1982), however, he obtained results that showed no statistical significance and could therefore not draw any conclusions with regard to subsidiary autonomy.
Subsidiary size
studies showing that increasing size leads to more centralization and control. Hedlund (1981) proposed that a large subsidiary would ceteris paribus have greater resources than a small subsidiary, and increasing size should lead to less dependence upon headquarters and thus more autonomy. He argued that the relationship between size and autonomy seems to be an inverted U-shape. This means that autonomy will rise along with size up to a certain cut-off point, after which autonomy will be decreasing as size is increased. The propositions of Hedlund (1981) where supported by empirical research conducted by, for instance, Johnston (2005) and Johnston and Menguc (2007).
Many authors also make a distinction between absolute and relative size of the subsidiary (e.g. Garnier, 1982; Gate and Egelhoff, 1986). With regard to this, Garnier (1982) maintains that the larger the relative size of the subsidiary to its parent firm, the better bargaining position it has. Its bargaining strength can lead to a more decentralized decision-making pattern. Moreover, Garnier (1982) argued that larger absolute size would also increase autonomy because a larger subsidiary would have the ability to support a full management staff.
The findings of the different studies, as mentioned before, varied in directions and support. For example, Picard (1977), Goehle (1980), Garnier (1982), Taggart and Hood (1999) and Richards (2000) all found that their results were insignificant, mixed or even in the opposite direction. This implies that affiliate size does not seem to influence the degree of (de)centralization as imposed by the parent firm. On the other hand, Gate and Egelhoff (1986), Harzing (1999), Männik et al. (2004), Varblane et al. (2005), Johnston (2005) and Johnston and Menguc (2007) have shown in their studies that for their sample a positive relationship between affiliate size and autonomy exists. Some of these also included the inverted U-shape finding.
3. THEORETICAL MODEL
3.1 Hypotheses
mentioned previously, the main goal of this study is to investigate which factors affect the degree of subsidiary autonomy granted by the parent MNC. The hypotheses have been divided into three main groups, according to the different entities affecting subsidiary autonomy, as identified by the literature. Firstly, we will hypothesize the relationships between the characteristics of the parent company and autonomy. Second, we make propositions for the characteristics of the local environment, and, finally, we will come down to hypotheses regarding the characteristics of the subsidiary, in relation to its autonomy.
3.1.1 Parent company characteristics
Degree of product diversification
Both Picard (1977) and Garnier (1982) predicted that product diversity is positively correlated with decentralization. Garnier (1982) argued that the more product diversity there is in an MNC, the greater the probability that some products will have been designed by subsidiaries specifically for their local market. Consequently, more decisions would tend to be made at the subsidiary level. Whereas Garnier (1982) results with this regard were mixed, Picard (1977) argued similarly and his results did support the hypothesis. In a similar vein, Gate and Egelhoff (1986) have argued that, as they relate things to complexity of practices, each new product line that is transferred to foreign subsidiaries requires additional specialization in the foreign subsidiaries and additional coordination between the foreign subsidiaries and headquarters. This results in higher internal complexity for the foreign subsidiaries and headquarters. External complexity also increases, since the marketing of different product lines is apt to force the foreign subsidiaries into different market segments, with different buyers, distribution systems, etc. Therefore, they too expected that the degree of autonomy is positively associated with product diversity. Their hypothesis was supported by their results. Vachani (1999) found that for related product diversification, the autonomy degree would actually decrease, whereas for unrelated diversification he found a positive impact on the level of decentralization.
the parent MNC to exert control. As argued above, the degree of product diversification is such a potential source of complexity, and therefore, in line with other literature, we propose the following hypothesis:
Hypothesis 1: Subsidiaries of multinational corporations with higher product diversification will have higher autonomy.
Parent company size
In the literature section we have seen that the size of the parent company is considered to be one of the most important explanatory determinants of the degree of autonomy possessed by subsidiaries. The general assumptions with regard to MNC size deal with, for instance, the controllability of the corporation (Child, 1973; Garnier, 1982). Increased MNC size makes it more difficult to control the entire range of operations, leading to more decision-making autonomy granted to subsidiary managers. Nonetheless, the results from testing for this relationship have provided no clear conclusions. For example, Hedlund (1981) could not derive conclusions from his results, Gate and Egelhoff (1986) found mixed results and Garnier (1982) found size of the MNC to only have a minor influence on subsidiary autonomy. Most notably, then, were the results of Wilkinson et al. (2008), although they used MNC size as a control variable and the aim was to explain the diminishing effect of cultural distance. They found that increased MNC size leads to a decline in direct control by the parent. Thus, even though the arguments seem clear, it is not evident, due to insignificant previous results, how MNC size affects autonomy. Therefore, we will elaborate on the same argument, but with a different data set, hypothesizing the following relationship:
Hypothesis 2: A positive association exists between parent company size and the autonomy of the subsidiaries.
The degree of geographic diversification
Organizational theorists have not addressed much attention to this possible determinant of autonomy. Even so, directly or indirectly, some authors have discussed this issue and, therefore, we could bundle their theory to develop another.
Prahalad, 1987). In the case of a very large MNC operating in a great number of countries, Garnier (1982) hypothesized that it would be increasingly difficult to control. Consequently, he expects the MNC to decentralize decision-making, but finds no such thing. Conversely, Picard (1977), Goehle (1980), Garnier (1982) and Gate and Egelhoff (1986) hypothesized that centralization is positively related to the size of foreign operations of the MNC. They maintained that as the percentage of the firm’s revenues from abroad rises, dependence risk increases. We argue that this variable is related to the number of countries the MNC is operating in (referred to by Garnier, 1982), since it is most likely that, as an MNC operates in more countries, the share of foreign operations revenue to total revenue will also be larger. Garnier (1982), with regard to the number of countries, found that his results were in the opposite direction of what was hypothesized, whereas Picard (1977), Goehle (1980), Garnier (1982) and Gate and Egelhoff (1986) find insignificant results in case of the first three, and contradictive results in the last. As a result, we expect that, in line with the dependence risk variable, increased foreign operations, in terms of the number of countries, will increase risk and, thus, the tendency to centralize decision-making. We hypothesize:
Hypothesis 3: The number of countries an MNC operates in will negatively influence the degree of subsidiary autonomy.
3.1.2 The local environment
The local institutional environment
However, as mentioned earlier, Ferner et al. (2005), in a study on international transfer policies between US multinationals and their subsidiaries, found that, due to differing institutional contexts, transferring policies is not always possible. Institutional pressures in the host country can force an MNC to drop standardized practices with regard to that country. Particularly, in the face of strong labor unions MNC resistance to institutional pressures is unlikely (Fenton-O’Creevy et al., 2007). CMEs typically have strong and well organized labor unions, in contrast to LMEs. Additionally, firms in CMEs give significantly more importance to the actions of labor unions, especially since this is a legal obligation. As a result, the simplified distinction between CME and LME could be more nuanced in order to create a clarification in how the autonomy levels of the subsidiaries differ. The level of unionization of the country can in this case do the trick. Following the argument of Fenton-O’Creevy et al. (2007), who state that national business system and degree of unionization are significantly correlated, but one does not imply the other, per se, we propose that the existence of strong unions, which will here represent the institutional pressure system, will force MNCs to impose a decentralized strategy on its subsidiaries located in that environment. This leads us to the next hypothesis:
Hypothesis 4: The higher the level of unionization in the country in which a subsidiary is located, the higher will be its degree of autonomy.
The development level of the host country
Moreover, it can be expected that a lesser developed market is more volatile than a mature market. Because of these reasons, the above mentioned studies pointed out that more developed countries will have subsidiaries with a higher degree of autonomy. In line with this, we hypothesize:
Hypothesis 5: Subsidiaries located in more developed countries will have a higher degree of autonomy than subsidiaries situated in less developed countries.
The culture of the host and home country
Culture is considered to be an important factor in international business practices. In this study, the central issue is that of subsidiary autonomy, for it is the dependent variable. Therefore, we need to address in which way culture might influence the degree of autonomy possessed by subsidiaries. In line with other literature (e.g. Harrison et al., 1994) the abstract term of culture will be made more tangent through the use of Hofstede’s culture dimensions in order to relate it to autonomy. As argued in the preceding section, the extensive utilization and recognition of the framework proposed by Hofstede (1980), makes it a useful tool for analysis between both countries and different bodies of literature.
In the original version of the framework, four cultural dimensions were developed, known as: ‘power distance’, ‘individualism versus collectivism’, ‘uncertainty avoidance’, and ‘masculinity versus femininity’. Later it was argued that in order to generalize for global use, an extra dimension had to be added that corresponded with the Asian cultures, but to a much lesser extent to the Western cultures (Hofstede and Bond, 1988). This dimension is known as ‘short-term versus long-term orientation’ or ‘Confucian dynamism’. For the purposes of this study, which range no further than the boundaries of Europe, this fifth dimension is considered to be of no added value and will therefore not be included in the model (Harrison et al., 1994).
Power distance
regard each other as equivalent people, who have equal rights (Hofstede, 1980, p. 122), and subordinates expect to be consulted on actions or decisions that affect them (Hofstede, 1980, p. 110; Child, 1981, p. 327). Other societies, by contrast, are characterized by the acceptance of inequality and the institutionalization of hierarchy- a hierarchy that reflects and underpins inequality. In these high PD societies, superiors are supposed to perform a leading role, to make decisions autocratically and paternalistically, and subordinates are commonly fearsome and unwilling to disagree with their superiors (Hofstede, 1980, p. 110; Child, 1981, p. 327).
Because power distance is directly associated with the acceptance (high PD) or rejection (low PD) of the unequal distribution of power, Hofstede (1980, p. 107) argues that high PD will be associated with greater centralization in organizational design, and low PD with greater decentralization (Harrison et al., 1994). Therefore, we hypothesize the following relationship:
Hypothesis 6: Subsidiaries located in a country scoring high on power distance will have less autonomy than subsidiaries located in a country scoring low on power distance.
Individualism versus collectivism
The cultural dimension of ‘individualism versus collectivism’, shortly ‘individualism’ (IDV), refers to the level of integration of individuals into groups (Hofstede, 1980). In individualist countries (high IDV), the ties between individuals are very loose. Everybody is expected to look after his own interests and at most the interest of immediate family. By contrast, in a collectivistic society (low IDV), the ties between individuals are very strong. People are born into collectivities or groups and people are supposed to look after the interest of that entire group (Hofstede, 1980). An individualist culture is typified by a focus on the individual as having a self-identity which is both unique and whole, and by an emphasis on the individual's interests and attributes, rather than on the interest of the group(s) to which individuals may belong. By contrast, in a collectivist culture, a person is seen as whole only when considered in terms of a group affiliation (Harrison et al., 1994).
desirable. Because decentralization is many times seen as an organizational mechanism promoting these attributes (e.g. Harrison et al., 1994), we propose the following hypothesis:
Hypothesis 7: Subsidiaries situated in countries scoring high on individualism will be granted more autonomy by the MNC than subsidiaries in low individualism countries
Uncertainty avoidance
According to Hofstede (1980), this dimension refers to how a society confronts the uncertainty of the future. A high UA society maintains a shared belief that ‘the uncertainty inherent in life is felt as a continuous threat that must be fought’ (Hofstede, 1980, p. 184) by ‘adhering to strict laws and rules, and safety and security measures’ (Hofstede and Bond, 1988, p. 11). A low UA society is one in which life's uncertainty ‘is more easily accepted and each day is taken as it comes’ (Hofstede, 1980, p. 184). In the organizations of a culture with high UA, the use of technology, rules, and rituals are widely applied to impose order and predictability on an otherwise uncertain environment. UA is, therefore, an indicator of the degree to which a culture values a sense of control (Barr and Glynn, 2003). In addition, Hofstede states that, ‘uncertainty avoidance leads to an escape from ambiguity’ (Hofstede 2001, p. 148). Therefore, Hofstede’s conceptualization of UA embeds the notion of control. In this sense, it can be argued that control is to be seen as formalization or standardization of rules and decision-making processes in order to decrease uncertainty. Consequently, or inversely, decentralization of decision-making would increase uncertainty for the parent company. With this in mind, we propose that high UA cultures will have a preference for centralized decision-making, or, put differently, we hypothesize that:
Hypothesis 8: Subsidiaries of MNCs originating from high UA cultures will have less autonomy than subsidiaries of MNCs from low UA cultures.
Masculinity versus femininity
for the nonmaterial quality of life, for children and for the weak. Feminine cultures define relatively overlapping social roles for the sexes, in which neither men, nor women need to be ambitious or competitive’ (Hofstede, 1984, p. 390). In other words, it represents the degree to which people prefer values of success and competition over modesty and concern for others and therefore it can also be seen as ego enhancement versus relationship enhancement. In this sense, it could be argued that that the ego enhancement country, so a high MAS score, would opt for centralization of decision-making, as it creates an extra feeling of importance and power. Conversely, a country scoring low on MAS would tend to decentralize decision-making, thereby bonding with other important individuals and groups. Therefore, we propose the following hypothesis:
Hypothesis 9: Subsidiaries of MNCs originating from high MAS cultures will have lesser autonomy than subsidiaries of MNCs from low MAS cultures.
3.1.3 Characteristics of the subsidiary
Subsidiary age
Hypothesis 10: Older subsidiaries will have more autonomy than their younger counterparts.
Subsidiary performance
Birkinshaw (1996) found that after poor performance, or a lack of distinct value added, a subsidiary might lose its mandate. Furthermore, Ferner (2000) found that poor performance significantly weakens the political influence of local management. This would imply that negative performance affects the subsidiary autonomy. In contrast, Goehle (1980), for example, argues that successful managers will have more decision-making freedom than their counterparts do and Geppert and Williams (2006) state that managers of subsidiaries with good performance are able to collect negotiating power and the power to protect local best practices and expertise from parent imposed practices. In this case, this would imply a certain independence from the MNC group, and thus that positive performance is a source of decision-making autonomy. This leads us to expect that there is a positive association between subsidiary performance and its autonomy. We propose:
Hypothesis 11: Subsidiary performance and its degree of autonomy are positively related.
Extent of ownership
in its unit, this leads to sufficient power to have actual control and decision-making freedom. This is why we expect that:
Hypothesis 12: The greater the local ownership of a subsidiary, the greater the degree of autonomy granted.
Subsidiary size
As mentioned by several studies, distinction should be made between the absolute size of a subsidiary and its relative size. However, Johnston (2005) notes that the relative size reflects only gross value and not value added. Since value added reflects the real value of the subsidiary, this value contributes to the whole added value of the parent company. It may be that the subsidiaries with the greatest relative size are the most vertically integrated within the MNC and therefore be the least autonomous, in which case it does not reflect variations in autonomy as size changes. Therefore, Johnston (2005) argues that it is more worthwhile when looking at autonomy to use absolute size in terms of number of employees. We will follow this argument and only discuss absolute size.
Gate and Egelhoff (1986), Harzing (1999), Männik et al. (2004), and Varblane et al. (2005) have shown in their studies that for their sample a positive relationship between affiliate size and autonomy exists. We will elaborate on the notion of increased subsidiary size leading to more autonomy, and, hence, we propose:
Hypothesis 13: The degree of subsidiary autonomy and its absolute size will be an inverted
U-shaped relationship.
3.2 The theoretical models
In literature on the subject of subsidiary autonomy and its determinants most studies used point-scale ranking in order to measure the degree of autonomy. As a consequence, the appropriate statistical technique for these studies was a multilinear regression model, using Ordinary Least Squares (OLS) estimation. In contrast, the present study will measure the degree of subsidiary autonomy, hence the dependent variable, as a count variable, due to reasons that will be specified in the section on the variables employed in this research. Following, for instance, Hill et al. (2001) a technique commonly used in the case of count data is the Poisson regression analysis, which’ distribution has one free parameter and does not allow for the variance to be adjusted independently of the mean. Alternatively, in the case of overdispersion (the presence of greater variability in a data set than expected), a mixture of the Poisson model, called the negative binomial model, can be used in order to provide a better fit through an additional free parameter. It is, however, not necessarily the case that either of the three techniques provides for better results, per se. Therefore, in this study, we will perform tests using all three models, with the intention of examining whether or not there are differences in the results between the three models. This implies that not until the proper model has been chosen, can the economic model be specified as it will be eventually. Nevertheless, the theoretical models can be specified. With this in mind, we will now elaborate on each individual model.
First, the multilinear regression model examines the OLS estimate using the subsequent specification:
Where:
Ŷ = the degree of subsidiary autonomy
X1 = the degree of product diversification of the parent company
X2 = the size of the parent company
X3 = the degree of geographic diversification of the parent company (number of countries)
X4 = the degree of unionization of the local environment
X5 = the development level of the host country
X6 = the score of the host country on the cultural value of power distance
X7 = the score of the host country on the cultural value of individualism
X8 = the score of the host country on the cultural value of uncertainty avoidance
X9 = the score of the host country on the cultural value of masculinity
X10 = the age of the subsidiary
X11 = the economic performance of the subsidiary
X12 = the extent of ownership of the subsidiary
X13 = the size of the subsidiary
X14 = the squared size of the subsidiary
β0 = the intercept, the value of Y when all X values are 0
β1,2,3,4,5,6,7,8,9,10,11,12,13,14 = the slope of the regression surface, associated with Xi .
At this point it is imperative to identify whether or not the least square estimations are the Best Linear Unbiased Estimators (BLUE). This is the case if the following assumptions hold Hill et al., 2001):
MR1: yt = β0 + βiXit + et i = 1 14; t = 1 232
MR2: E(yt) = β0 + βiXit E(et) = 0
MR3: var(yt) = var(et) = (No heteroscedasticity)2
MR4: cov(yt, ys) = cov(et, es) = 0 (No autocorrelation)
MR5: The values of Xt are not random and are not exact linear functions of the other
explanatory variables.
MR6: yt ~ N(β0 + βiXit, ) e2 t ~ N(0, )2
squares estimator to estimate the unknown coefficients, then the least squares estimator were still a linear regression and unbiased estimator, but it is no longer BLUE, the standard errors for the least squares estimator are incorrect, and, finally, the use of these standard errors in hypothesis tests and confidence intervals may be misleading (Hill et al., 2001). This brings us to the second model.
The Poisson regression model examines the quasi-maximum likelihood estimator (QMLE) that the subsidiary autonomy is granted by the parent company. The following Poisson distribution is used (Hill et al., 2001):
!
)
(
Y
e
Y
f
Y i
; Y = 0, 1, 2, 3…10 (2) Where: Y is the number of decisions undertaken by the subsidiary and f ( Yi ) indicates the
probability that the variable Y takes non-negative integer values. Y! is Y factorial with Y! = Y x (Y-1) x (Y-2) x (Y-3) x ... x 2 x 1
µ is the mean number of decisions taken, for all subsidiaries, with
i
= exp(β0 + β1Xi1 + β2Xi2 + β3Xi3 + β4Xi4 + β5Xi5 + β6Xi6 + β7Xi7 + β8Xi8 + β9Xi9 + β10Xi10+ β11Xi11 + β12Xi12 +
β13Xi13 + β14Xi14) (3)
The Poisson regression model has the assumption that var(Y) = µ = E(Y) (its variance is the same as its mean value). Therefore, our Poisson regression model will have the following specification:
Ŷi= eβ0+ β1Xi1 + β2Xi2 + β3Xi3 + β4Xi4 + β5Xi5 + β6Xi6 + β7Xi7 + β8Xi8+ β9Xi9+ β10Xi10 + β11Xi11 + β12Xi12+ β13Xi13+ β
14Xi14 + εi (4)
This negative binomial model uses a quasi-maximum likelihood estimator that the degree of subsidiary autonomy is granted by the parent company as well. Therefore, following the theory in Hill et al. (2001), we derive the next distribution function:
N i i i i i i m x Y m x Y Y Y l 1 2 2 2 22 ( , )) ( 1/ )log(1 ( , )) log ( 1/ ) log( !) log (1/ )
log(
,
(5)
Here, m(xi, ) = E(yi|xi, ), and as mentioned above, the negative binomial model brings about
an extra free parameter, , which is a variance parameter to be jointly estimated with the 2
conditional mean parameter β. Since the negative binomial distribution has one more parameter than the Poisson model, the second parameter can be used to adjust the variance independently of the mean. That is why if overdispersion in the data exist, so that v(xi, )>m(xi,
), or, in words, the sample variance is greater than the mean, it is common practice to use the negative binomial model, as long as the following holds:
m(xi, ) = E(yi|xi, ) =exp(xi’ ) var(yi|xi, ) = m(xi, )(1+ m(x2 i , ))
For a fixed , given the above, if we maximize the negative binomial log likelihood we obtain 2
4. RESEARCH METHODS
4.1 Method of analysis
In order to ensure proper results several empirical analyses are conducted in this research. With the software package Eviews we perform univariate, bivariate and multivariate analysis, with the following sequence:
(i) To describe the data and to identify the nature of the distributions we perform univariate analysis. This will, among other things, help us identify and exclude outliers from the sample.
(ii) The bivariate analysis will assist us in determining the correlations between variables, both dependent and explanatory, with the goal of investigating the existence of multicollinearity in the model. In addition, this allows us to decide whether some variables need to be removed from the model.
(iii) Finally, we will use multivariate analysis to estimate the three models, being the multilinear regression model, the Poisson regression model and the negative binomial regression model. This will be done by estimating the models firstly for the parent company characteristics only, then the culture variables will be added, and, at last, the subsidiary characteristics will be complemented to make the model complete. By doing this, we will be able to see how the model changes as more variables are put in.
