The Influence of Home Country Culture on the Relationship
between an MNC’s Board Characteristics and Entry Mode Choice
Master Thesis By
I.D. van Outersterp S2155222
i.d.van.outersterp@student.rug.nl
University of Groningen – Faculty of Economics and Business MSc. International Business and Management
June 21, 2017
2 ABSTRACT
Using arguments from the upper echelon perspective, this research investigates the relationship between the characteristics of the board of directors (BOD) and the MNC’s choice between acquisitions and joint ventures as entry modes. A binary logistic regression analysis draws on data gathered from 200 entry modes by 160 companies with a total of 100 acquisitions and 100 joint ventures in the period of 2010-2015. The board characteristics nationality, age and education are investigated. It is also examined whether the home country cultural variables power distance and uncertainty avoidance influence the relationship between BOD characteristics and entry mode choice. Empirical findings show significant support for nationality and age increasing the likelihood of using an acquisition over a joint venture. Evidence is also provided for power distance and uncertainty avoidance having a moderating effect on the relationship between education level and entry mode choice. This study contributes to the international business literature in that it puts the rarely investigated BOD at the center of attention while combining both demographic characteristics and national culture into a singular study. The results also help managers in their understanding on how the optimize the BOD’s composition.
3 ACKNOWLEDGEMENTS
I dedicate this thesis to my parents, who have always empowered me to fulfill my academic potential and personal ambitions. Not only did I receive high-quality education, but they also offered me a lifetime of experiences and the freedom to work on my future wherever possible. I am grateful for the opportunity of meeting people from all over the world, to learn other cultures, and to have a great student life both in the Netherlands and abroad. Without them I would have never received the opportunity to develop my love for travelling and interest for international business.
As I am close to finishing my two master programs, I have come to realize that the six years well spent at the University of Groningen finally come to an end. This thesis forms the completion of my studies and simultaneously signifies the end of my student life. Without any doubts, my studies could not have been completed without the guidance and support of a number of people. I would therefore like to express my gratitude for the encouragement and support of some people before, during and after my journey at the university.
First of all, I would like to thank my friends and family for their endless love and support. Their unfailing enthusiasm and understanding has provided me with enough space to grow, both as a student and a person. I specifically want to thank Tessa Voors for taking this thesis to the next level. Also I wish to show appreciation to my former roommates. Even though both the house and my student life will make a clean break, the legacy of our house will always pursue in our memories.
Finally, I would like to express my gratitude to Dr. Olof Lindahl, my supervisor, for his diligent supervision and constant support throughout the process of my thesis. His scholarly insights, suggested refinements and professional advice have been important factors in completing this thesis.
4 TABLE OF CONTENTS
LIST OF FIGURES AND TABLES ... 5
ABBREVIATIONS ... 5
1 INTRODUCTION ... 6
2 THEORETICAL BACKGROUND ... 10
2.1 Upper echelons theory ... 10
2.2 Entry mode choice ... 11
2.3 Board characteristics ... 13
2.3.1 Nationality ... 13
2.3.2 Age ... 14
2.3.3 Education ... 14
2.4 Home country culture ... 15
2.4.1 Power distance ... 17
2.4.2 Uncertainty avoidance ... 17
2.5 Board characteristics and entry mode choice ... 18
2.5.1 Nationality ... 18
2.5.2 Age ... 19
2.5.3 Education ... 19
2.6 Culture and entry mode choice ... 20
2.6.1 Power distance ... 20 2.6.2 Uncertainty avoidance ... 23 2.7 Conceptual model ... 26 3 METHODOLOGY ... 27 3.1 Data source ... 27 3.2 Sample ... 28
3.3 Measures and variables ... 28
3.3.1 Dependent variable ... 29 3.3.2 Independent variables ... 29 3.3.3 Moderator variables ... 30 3.3.4 Control variables ... 31 3.4 Data analysis ... 33 4 RESULTS ... 34
4.1 Descriptive statistics and assumptions ... 34
4.2 Control variables ... 35 4.3 Direct-effects analysis ... 36 4.4 Moderation analysis ... 38 4.5 Conclusion ... 40 5 DISCUSSION ... 42 5.1 Direct effects ... 42 5.2 Moderation effects ... 44 5.3 Theoretical implications ... 45 5.4 Managerial implications ... 46
5.5 Limitations and suggestions for further research ... 46
6 CONCLUSION ... 48
7 REFERENCES ... 49
8 APPENDIX ... 56
8.1 Sample ... 56
5 LIST OF FIGURES AND TABLES
Figures
FIGURE 1 Conceptual model 26
FIGURE 2 PDI and UAI of the home countries involved in the sample 31 Tables
TABLE 1 Characteristics of acquisitions and JVs 13
TABLE 2 Cultural dimensions and their definitions 16
TABLE 3 Summary of means, standard deviations and correlations 34
TABLE 4 Hypotheses testing 40
TABLE 5 Summary of logistic regression predicting likelihood of an
acquisition based on the IVs and the interaction variables 41 TABLE 6 Binomial logistic regression predicting likelihood of an acquisition
based on BOD characteristics 59
TABLE 7 Binomial logistic regression predicting likelihood of an acquisition
based on the IVs and the interaction variables 60
ABBREVIATIONS BOD – Board of Directors CQ – Cultural Intelligence CEO – Chief Executing Officer DV – Dependent Variable
HRM – Human Resources Management IV – Independent Variable
JV – Joint Venture
MNC – Multinational Corporation PD – Power Distance
PDI – Power Distance Index R&D – Research & Development UA – Uncertainty Avoidance
6 1 INTRODUCTION
Traditionally, companies have followed the Uppsala model in that they internationalize only in small, incremental steps (Barkema and Drogendijk, 2007). Risk-averse managers have recognized and seized expansion opportunities only in their local foreign environments. Yet, the world has changed. The twenty first century symbolizes an increasing quest for globalization. Multiple dynamics – access to new markets, open economies – have increased the urge for companies to internationalize their business activities. At the heart of this trend are the multinational corporations (MNC) (Staples, 2007). However, whereas a global move can undoubtedly yield benefits, it can also be rather hazardous. Notably, the selection of the appropriate entry mode can have significant and far-reaching effects on the MNC’s overall performance and survival (Brouthers, Brouthers, & Werner, 2000; 2003). An unsuitable entry mode may, for example, result in considerable financial losses. Some companies might even face an exit from the foreign market, with Allianz and McKinsey as well-chosen examples (Mathe and Perras, 1994).
Given this importance, entry mode choice has extensively been researched in the international business literature (Kim and Gray, 2008). Canabal and White (2008) already identified 167 papers related to entry mode choice within the period of 1980 until 2006 and the topic has continued to be studied ever since (e.g. Hollender, Zapkau & Schwens, 2017). This paper also examines this interesting topic, yet, it focuses on the two most academically explored types of entries only: acquisitions and joint ventures (JV). These modes specifically gained significant popularity in the last decades (Beugelsdijk, 2013). Also, both entry modes require a large degree of commitment, investment and control by the MNC – and thus high managerial involvement.
7 unified perspective in the explanation and prediction (Ekeledo and Sivakumar, 2004). The only agreement among scholars is that the resource-based view provides the richest explanation and prediction of entry mode choice (Barney, 1991; Grant, 1991), but application of this theory has been primarily conceptual and descriptive (Ekeledo and Sivakumar, 2004).
Another thing all theories on entry mode choice have in common is that they rarely touch upon the manager’s perceptions associated with various entry modes. Therefore, an objective of this study is to better understand this professedly inconsistent predictor by introducing a third level of analysis: the managerial perspective. Zooming in with a managerial-level lens means that this paper focuses on the managerial characteristics influencing the entry mode choice. Hambrick and Mason (1984) are the founders of the research stream on managerial characteristics, also called the upper echelon (UE) theory. UE theory states that the firm’s organizational outcomes and strategic choices are partially predicted by executives’ background characteristics (Carpenter, Geletkanycz, & Sanders, 2004). Wiersema and Bantel (1992) largely dominate in this field with their study on how certain demographic traits result in more receptivity to change and willingness to take risks.
Most UE literature has solely concentrated on the demographics of newly selected chief executing officers (CEOs) and their link to the firm’s strategy. For instance, Herrmann and Datta (2002; 2006) associate CEO’s cognitive orientations and knowledge base, reflected in their observable demographic characteristics, with a firm’s entry mode choice. However, as many authors argue (e.g. Hambrick, 2007; Hambrick, Cho, & Chen, 1996), attention to the whole board, rather than on one individual, would yield improved and more precise details of organizational outcomes. Kim and Gray (2008) confirm this argument in their recent study by considering the board of directors (BOD) to be the main decision-making organs within an MNC and therefore likely to be greatly involved in important decisions as the entry mode choice. Moreover, Herrmann and Datta (2006) advise that further research should be conducted to examine the effect of board characteristics on entry mode choice. Previous research thus acknowledged the premise that boards are important, or even essential, for strategic decisions at the MNC (Hambrick, 2007).
8 aims to fill the research gap by investigating how the characteristics of the BOD as a whole affect an MNC’s entry mode choice.
This study follows previous literature (e.g. Black, Gregersen, & Mendenhall, 1992) in that it studies observable characteristics that are confirmed to influence strategic decisions on CEO level: nationality, age and education. Yet, the characteristics will now be applied to BOD level. Nationality is interesting to study as existing UE research has barely paid notice to the topic (Heiltjes, Olie, & Glunk, 2003). Age is considered relevant in this aspect due to its association with risk (Herrmann and Datta, 2006), one of the key attributes in the entry mode literature (Erramilli, 1991). Education is researched as previous studies solely focused on one aspect of education (e.g. Tihanyi, Ellstrand, Daily, & Dalton, 2000), whereas this study combines multiple aspects: education level and functional background.
Nevertheless, BOD characteristics cover the discussion only partially. Managerial behavior is argued to be a product of internal characteristics as well as on its external environment (Laurent, 1983). To account for this limitation and thereby account for external forces, the MNC’s home country culture is added as a moderator to this study. Recent studies have agreed on the influence of culture on entry mode choice (Drogendijk and Slangen, 2006). For instance, Erramilli (1991) argues culturally different managers to be distinctive in their willingness to share decision-making with others, and therefore different in their attitude to sharing venture other firms. However, literature has largely disregarded the combined effects of demographic characteristics and national culture) on entry mode choice (Smith and Peterson, 2005). At one side, single-nation studies on demographic effects (e.g. Ely and Thomas, 2001) barely focus on the replication of these effects in dissimilar cultural contexts. On the other side, whilst cross-cultural researchers often identify the potential effects of demographic effects other than nation, they typically deal with these effects as if they are a source of error that needs to be controlled for validity. The error is often controlled for by either matching samples from different nations on demographic criteria or by estimating and partialling out the variance attributable to demographic factors. Thus, only rarely have demographic characteristics and national culture, and thus BOD characteristics and the MNC’s home country culture, been fundamentally studied together.
9 alone suffice to predict entry mode choice or whether external factors like home country culture should also be taken into account. Home country culture is studied by means of two of Hofstede’s (1980) cultural dimensions, namely power distance (PD) and uncertainty avoidance (UA). Why this study focuses on these dimensions specifically is because they can be linked to entry mode choice most evidently. Both dimensions classify countries on their preferences for uncertainty and control and thereby guide an MNC in their selection of hierarchical modes of entry.
This study is designed around the following research questions:
How do MNC’s board characteristics influence the entry mode choice? And how is this relationship moderated by the home country culture?
The aim of this study is to add to the understanding of the relation between an MNC’s BOD characteristics (nationality, age, education) and the choice to enter a foreign market through either an acquisition or a JV. Another goal is to disclose the effect of home country cultural values on this relationship. The theoretical contribution of the study is two-fold: it is not only the first to put the BOD at the center of attention, but it also contributes by merging demographic characteristics and national culture into one single study. The research is also highly relevant in terms of managerial implications on the BOD’s composition.
10 2 THEORETICAL BACKGROUND
This chapter presents the relevant literature to the study. It starts with an introduction to the UE theory and continues with a discussion of the relevant concepts. The list starts off by discussing entry mode choice, followed by an overview of the following BOD characteristics: nationality, age and education. The chapter continues with an outline of the cultural values PD and UA. After that, a dialogue presents the singular combination of the BOD characteristics and the cultural values with entry mode in order to develop the main hypotheses. The chapter concludes with the conceptual model.
2.1 Upper echelons theory
This study builds on the foundations of the upper echelons theory. Originally founded in 1984 by Hambrick and Mason, this theory argues that strategic choices are a reflection of the demographics, values, experiences, personalities and cognitive bases of powerful actors in the firm. The differences between and within characteristics of the upper echelons thus serve as an explanation, or even prediction, with regard to strategic choices.
The ‘upper echelon’ is the dominant coalition at the organizational top who has the power and the responsibility to make important strategic decisions. This could be either the CEO, other executives as the CFO, board members, chairmen and/or committees. The main question thus centers on the players in the game: who are the members of the UE and what do they bring? The question results from the premise of bounded rationality: uncertain situations and complicated information are not objectively understandable but are interpretable (Cyert and March, 1963). One can only grasp why organizations do the things they do, or why they perform the way they do, by looking at the biases and characters of their most powerful actors: the BOD. So far, most literature has solely focused on demographic characteristics of CEO’s. Although the significance of the CEO should be recognized, attention to the entire board (rather than on one individual) would yield improved meaning to organizational outcomes (Hambrick, Cho, & Chen, 1996; Hambrick, 2007). Hence, this study focuses on the BOD as a whole.
11 however more difficult due to the inability to directly measure cognitive bases and restrictions in comparison of indicators. Demographic indicators are thus essential when put to a stringent test. As certain demographic indicators also relate to particular values, beliefs and abilities, for this study it is sufficient to measure demographic characteristics and thereby indirectly measure the less observable, cognitive approach. This study follows previous work (e.g. Black et al., 1992) in that it uses the following observable characteristics that are confirmed to influence strategic decisions on individual level: nationality, age and education.
UE theory likewise studies the influence of BOD characteristics on strategic outcomes in two different ways. One research stream relies primarily on demographic characteristics using their average values (e.g. Van Veen and Elbertsen, 2008). This is based on “the belief that specific demographic characteristics associate with executive perceptions that ultimately lead to certain actions and outcomes” (Tihanyi et al., 2000: 1161). The other stream rather investigates top management team heterogeneity (Hambrick et al., 1996). For the purpose of this paper, we follow the first stream in that it relates demographic managerial characteristics to specific corporate outcomes (Bantel, 1993). It thereby complements prominent studies that embedded the UE perspective to explain how views, backgrounds and experiences of top management influences factors such as management payment (Singh and Harianto, 1989) and firm performance (Certo, Lester, Dalton, & Dalton, 2006).
2.2 Entry mode choice
The strategic outcome investigated is entry mode choice. It refers to the institutional arrangement a firm uses to market its product in a foreign country (Root, 1994). It generally takes a firm three to five years to completely enter a new market. There is no ideal market entry strategy, yet an ill-judged market entry choice in the initial stages of internationalization can threaten a firm’s market activities (Hollensen, 2011). Many authors (e.g. Hitt and Tyler, 1991) therefore deliberate entry mode choice as a significant strategic decision.
12 larger degree of commitment, investment and control by the MNC. As this also requires more BOD involvement, this study focuses on equity-based entry modes only. This choice is justified by the increasing selection and attention of acquisitions and JVs as market entry strategy (Beugelsdijk, 2013).
An acquisition is the purchase of the (total, or at least the majority of the) equity of an existing company in order to establish wholly owned subsidiaries (Dussauge and Garette, 2000). The acquiring firm gains full ownership and control of the foreign firm and directly enters the host country market, without collaborating with a local partner (Glowik, 2009). It enables quick entry and often provides access to an existing customer base, distribution channels and, in some cases, established corporate reputations. Although in rare cases existing management remains to allow the firm to acquire experience, an acquisition often does not account for any uncertainty or risk associated with unfamiliar circumstances in the host country (Dussauge and Garette, 2000).
International JVs are new firms created through a partnership between two or more parties in different countries (Root, 1994). The MNC thus shares supervising power with a local partner company in the host country and relinquishes full control, while decreasing the level of risk and uncertainty associated with doing business in a distant country (Beugelsdijk, 2013). Most JVs are split, where both parties have a 50% ownership stake. Benefits to this partnership are the shared costs and risks between the two partners and the complementary capabilities and resources (e.g. technology, management skills) of both the MNC and the local partner (Dussauge and Garette, 2000). It also clears the road of global operations in research & development (R&D) and productions and provides entry to countries that restrict foreign ownership. Yet, shared ownership and intensive collaboration also fairly complicate the management of a JV.
13 TABLE 1
Characteristics of acquisitions and JVs
2.3 Board characteristics
An MNC’s BOD has a considerable impact on the entry mode choice, as it has the largest stake with regards to strategic decisions in the firm (Bainbridge, 2002). This study hence takes the perspective that board characteristics explain why the MNC chooses a certain entry mode. This section will outline the characteristics nationality, age and education, with the latter divided into education level and functional background.
2.3.1 Nationality
Increasing globalization of organizations and changing demographic patterns make it more important to understand a mixture of different nationalities in leadership positions (Yukl, 2010). Nationality is a surface-level characteristic that is inevitably linked to two deeper levels of influence. First, nationality is linked to the cultural values and traditions of the managers. They could stem from societal norms, specifying acceptable forms of leadership behavior and behavioral conducts (Bass, 1990). The values and traditions considerably influence managerial attitudes and behavior (e.g. Adler, 1997), as they are likely to be internalized by managers who grow up in a particular society.
14 or her home country. In this case, international experience thus refers to the experience gained in a country other than the MNC’s home country. International experience strongly influences the capabilities of managers when acting in an international market (Black, 1988). Experience leads to the creation of a large knowledge base regarding foreign cultures and business practices, which leads managers to be better able to recognize fruitful business opportunities and to feel comfortable in comprehensive foreign business environments.
Section 2.5.2 discusses how the number of different nationalities in the BOD is related to entry mode choices.
2.3.2 Age
One of the most observable board characteristics is the age of the board members. Recent research has mostly focused on age as a control variable (Day, 2014). Yet, this study follows UE theory in that age is a significant independent indicator to characterize personality traits and a firm’s strategic actions (Hambrick and Mason, 1984).
Age represents a person’s willingness to take risks and accept chance (Wiersema and Bantel, 1992). In general, older board members are more risk-averse and resistant to change (Herrmann and Datta, 2006). At the end of their careers, seniors tend to exclude riskier decisions due to their desire of financial and career security (Lebert, 2016). They are also argued to show more devotion to the status quo (Stevens, Beyer, & Trice, 1978). Conversely, Child (1974) links younger board members to organizational growth and encouragement of modernization. They are assumed to be less devoted to the status quo and head towards creating more innovative solutions when dealing with uncertain situations.
Age also indicates the amount of firm experience (i.e. experience with one firm for a longer period) board members possess (Hambrick et al., 1996). This reinforces the argument regarding willingness to take risks. Although an increased amount of experience improves the manager’s problem-solving mindset, more experience is associated with a larger inclination to keep the status quo in doing business and to avoid risks (Finkelstein and Hambrick, 1990).
The focus of this study is the average age of all members within one BOD. Section 2.5.2 discusses how the age of the BOD relates to specific entry mode choices.
2.3.3 Education
15 education. They argue that nationality and age are relations-oriented features, whereas education is a task-oriented feature. Inclusion of both features is crucial for capturing the total effect of characteristics.
Education is defined in two ways: education level and functional background. Education level echoes an individual’s cognitive capacity and skills (Wiersema and Bantel, 1992). Well-educated managers are empowered with superior cognitive skills, decision-making capabilities and awareness levels. Higher education could, for instance, lead to a higher individual cultural intelligence (CQ): an individual’s capability to function and manage effectively in culturally diverse settings (Earley and Ang, 2003).
At the same time, prior research advocates that the study’s functional background reflects an individual’s personality and cognitive style, thereby modeling future views and outlooks (Wiersema and Bantel, 1992). This starts with the assumption that managers develop themselves within one particular business area throughout their careers. The type of functional area creates a paradigm through which managers interpret the environment (Carpenter and Westphal, 2001). Thus, when entering the board, managers bring along their views and standpoints from this particular area. Functional backgrounds are commonly classified into two types of areas: output and throughput. An output functional background entangles areas like marketing, sales, entrepreneurship or R&D (Hambrick and Mason, 1984). People specialized in this area are typically associated with the aptitude to search for new markets and opportunities and are responsible for monitoring and adjusting products. A throughput functional background comprises a background in operational areas like supply chain management, manufacturing and accounting, leading to the ability to optimize efficiency and maintain control. Amongst others, Herrmann and Datta (2002; 2006) concentrated on these two areas while using CEO characteristics to predict entry modes.
Section 2.5.3 discusses how education level and functional background are related to entry mode choices.
2.4 Home country culture
16 turn the strategic decisions made by the firm (Barkema and Vermeulen, 1997). To account for this effect, home country’s culture is measured in this study by adding two moderator variables.
The literature stream around culture often takes a comparative approach: first, universal dimensions are defined, after which unique scores of societies (often countries) are measured on these dimensions. This results into country profiles that are used to classify and compare countries. One prominent framework holds Schwartz’ (1992) cultural value orientations. More contemporary frameworks are the individual-level World Value Survey and the cross-cultural leadership study GLOBE, carried out for the first time in 1981 and 1990 respectively. This research builds on the most widely used approach for culture: Hofstede’s national culture framework (1980; 2010). His model distinguishes cultures on six dimensions (TABLE 2), which each represent independent preferences for one state of affairs over another.
TABLE 2
Cultural dimensions and their definitions
17 dimensions together. Moreover, Hofstede (2010) argues those dimensions to have the biggest influence on cultural traits of associations. The influence of both dimensions on the relationship between board characteristics and specific entry mode choices is discussed in more detail in 2.4.1 and 2.4.2.
2.4.1 Power distance
Hofstede (1980) defines power distance as the extent to which less powerful members of organizations and institutions accept and expect that power is distributed unequally. This dimension thus measures the level of equality within a society (Hofstede, 2003). High scoring countries allow inequalities of power and wealth to grow. The society accepts a hierarchical distribution of power and people understand their “place” in the system. Problems are generally solved by a show of power and autocratic decision-making. People generally depend on each other, yet cooperation among the powerless is difficult to achieve because little faith in people is the norm (Hofstede, 1983). Venezuela (with a score of 81 on a scale of 0-100) is an example.
In contrast, low scoring countries address equal rights and opportunities for everyone (Hofstede, 1983; 2010). Power is widely shared and dispersed and problems are usually solved by flexibility. People are more interdependent on each other, with high cooperation among the powerless based on solidarity. Austria is an example (11).
2.4.2 Uncertainty avoidance
Uncertainty avoidance is defined as the degree to which members of a society are able to cope with future uncertainty without experiencing undue stress (Ueno and Sekaran, 1992). It indicates the extent to which a culture programs its members to feel comfortable in unstructured situations (Hofstede, 1983). It also represents the degree to which someone feels threatened by uncertain or unknown situations – a feeling that is expressed through, amongst other things, nervous stress and a need for predictability (Hofstede, 2003). High scoring countries are risk-averse: their members avoid uncertainty as much as possible, thereby minimizing the possibility of uncommon and unexpected situations. As part of a general concern with security and stability, there is a need for structured rules and social norms (Hofstede, 1983; 2003). Russia (with a score of 95 on a scale of 0-100) is an example.
18 unexpected situations. As members are more comfortable with instability, there is a lesser need for strict rules. Singapore is an example (8).
2.5 Board characteristics and entry mode choice 2.5.1 Nationality
As explained, nationality is correlated with international experience. Including different nationalities in the BOD means that managers are included from countries other than the MNC’s home country. These foreign managers possess individual international experiences that are for instance obtained while living in their home country. The summation of nationalities within a BOD is thus interpreted as the accumulated international experiences of individual board members (Smith, Smith, O’Bannon, Olian, Sims, & Scully, 1994).
In turn, the international entrepreneurship literature has connected the accumulated international experiences to entry mode choice. Black et al. (1992) argue that international experience is positively related to a firm’s global competitiveness. To explain this, we translate individual international experience into competences of transnational managers (Adler and Bartholomew, 1992). Transnational managers possess key competences, or transnational skills, that are indispensable for the success and prosperity of any international strategy. One example is having knowledge about foreign customers and competitors. This results in managers who are better capable of estimating risks and returns in foreign investments (Herrmann and Datta, 2006), who are more comfortable managing complex foreign business environments (Black, 1988) and who are better able to handle highly uncertain situations (Carpenter, Pollock, & Leary, 2003). These transnational skills might therefore be a mean to overcome the “physic distance” that is created by doing business overseas (Melin, 1992). The accumulated international experiences of the board member thus reduce the uncertainty associated with expanding to other countries, which in turn diminishes the need to rely on a local partner (i.e. JV) for support in a foreign venture.
19 2.5.2 Age
To understand the relationship between age and entry mode choice, one should keep in mind that establishing a subsidiary abroad is a risk. The BOD needs to cope with distant (and probably unfamiliar) host country institutions that potentially lead to risky situations. Age differences lead to diverse perspectives to those risks associated. Older board members tend to be more risk-averse and hold a preference for stability and security. This is translated to the entry mode literature in the belief that older BODs may be more prone to use an entry mode that is considered less risky to the MNC. Increased executive age is thus related to lower levels of strategic change (Herrmann and Datta, 2006).
Older board members prefer JVs as they are generally considered as being least hazardous to the firm, due to the lower level of uncertainty they bring to the MNC. Moreover, the shared control of JVs offers more security as the BOD could always resort to the local partner’s knowledge. Contradictory, acquisitions pose greater challenges to the status quo and are therefore more preferred by younger board members (Finkelstein and Hambrick, 1990).
Applying this line of reasoning to entry modes choice, the second hypothesis is created:
2.5.3 Education
20 who have well-developed problem solving and CQ skills would be more confident to adjust correctly to culturally distant situations and thus more prone to choose riskier entry modes.
All things considered, a BOD endowed with high-educated members has a more positive attitude towards undergoing chance and international commitment. They also benefit from their excellent competences in dealing with complex foreign investments, while being more confident and comfortable in taking the riskier strategy. Hence, it is suggested that:
The classification of education types also provides a linkage between functional background and strategic decision-making. Managers from different functional backgrounds develop different viewpoints to entry modes. Managers with an output background are often concerned with progress, improvement and low-cost strategies (i.e. JV) (Herrmann and Datta, 2005). On the other hand, throughput backgrounds are prevalent in industries characterized by high capital intensity or concentration and lower growth (Rajagopalan and Datta, 1996). Moreover, managers with a throughput background emphasize the need to main control and to optimize efficiency – features typically associated with an acquisition. Henceforth the following hypothesis:
2.6 Culture and entry mode choice 2.6.1 Power distance
21 reflected in the organizational structure and thus in the MNC’s ownership policies as well (Erramilli, 1996). Decision-makers from high PD countries prefer individual control and full ownership – features common to an acquisition. Conversely, looser, decentralized hierarchies characterize low- coring countries. There are fewer managers at top level and all organizational members are considered equal, or nearly equal. Subordinates are trusted with more important jobs and will reject leaders whom they perceive as autocratic or patronizing (Sweetman, 2012). Decision-makers from low PD countries favor shared control and ownership – features common to a JV. All things being equal, it is therefore presumed that high PD countries prefer acquisitions to JVs.
22 Further, H2 theorizes age to positively influence the probability of a JV as an entry mode. To control for the influence of PD on this relationship, we follow Khatri’s (2009) argument in that PD determines the criteria on which senior managers are assessed to be influential. Seniors from low PD countries gain respect from employees because of former’s competence. Subordinates expect to be consulted by their seniors, with the ideal boss being a resourceful democrat (Ashkanasy, Wilderom, & Peterson, 2000). Only if the senior comes across as incompetent, he or she will face a loss of influence in strategic decision-making. This might weaken the abovementioned relationship. In contrast, in high PD countries employees gain respect by virtue of age and long tenure in the organization (Beer and Marsland, 1983). Older managers have a disproportionally high influence in strategic choices in comparison to younger board members. Seniors are empowered with unlimited authority and control. Contrariwise, juniors have an unquestioning, even submissive attitude in which they are expected to obey and show respect to seniors. The responsibility for and authority in decision-making is thus vested in the hands of the elderly (or those at least more senior in the hierarchy), whereas juniors rarely participate in decision-making. When a new employee starts working in an organization, he or she would never consider competing with seniors in the organization. This disproportioned high influence of seniors in the decision-making process leads the opinion of the older members to be dominant. As older members tend to be risk-averse, this reinforces the abovementioned relationship:
23 the creation of problem solving skills is dependent on the level of PD in the environment. By way of explanation, Hofstede (1991; 2001) emphasizes PD as a significant factor to team relationships because the team’s PD level influences the managerial approaches to problem resolution. Problem management in low PD teams is based on principles of negotiation and cooperation, whereas problems in high PD teams are primarily resolved by the power holder (Deutsch, 1973). Contributing effectively in problem-based teams might therefore be more challenging for managers with high PD orientations. For instance, individuals who perceive themselves as agents of remote higher authority might come to believe that they are no longer in control of their own actions (Milgram, 1974). This state of affairs may not be beneficial to productive teamwork and thus to the problem solving skills of the team. A high PD level thus weakens the abovementioned relationship. Contrastingly, Couto and Vieira (2004) state that a low PD environment facilitates the creation of a problem solving setting. Managers in low PD organizations delegate more responsibilities to subordinates. That way, they encourage subordinates to enjoy more valuable learning experiences and eventually to gain more problem solving skills. They pay more attention to the managerial problem solving skills and thereby increase the likelihood of managers being able to manage an acquisition. Under those circumstances, PD strengthens the hypothesized relationship. This argument leads to the following hypothesis:
There is not enough scientific evidence to assume a moderating relationship of PD on the relationship between a board member’s functional background and the entry mode choice. 2.6.2 Uncertainty avoidance
24 control. High scoring countries prefer hierarchical structures and to pull the ropes themselves. Clearly, an acquisition best facilitates the ability to exert full control and to implement an own set of policies. Equally important, managers from low scoring countries are more willing to take risks, to delegate, and to be flexible without the need for a comprehensive set of rules (Hofstede, 1983; 2003). Firms are more concerned with the long-term strategy than with current affairs. Organizations are less structured and might even be supervised by shared control. Managers are more tolerant towards collaborating with a business model different from theirs and thus towards each other’s opinion and behavior. These features coincide with a JV – a mode generally entailing more uncertainty, since the home country partner will not be able to exert full control over the firm.
As mentioned, the cultural variables signify the need to check whether the main relationships (H1-H3) hypothesized change for different values in UA. The first hypothesis (H1) holds that more different nationalities in the BOD increase the likelihood of choosing an acquisition as an entry mode. To check for the influence of UA on this relationship, one should understand that this relation is based on the premise that a higher number of nationalities surges the international experience available in the board (Smith et al., 1994). This experience in turn leads to a relative tolerance in dealing with any uncertainty associated with the host country, thereby increasing the probability of entering by an acquisition. Although this premise may be true, in high scoring countries, one tries to avoid uncertainty as much as possible. Indeed, this might result in a downward level of the built up tolerance with uncertainty again. Other said, in high UA countries the experience gained from the inclusion of many different nationalities does not reduce the uncertainty as much as would happen in low UA countries, due to the general intolerance with uncertainty. UA is thus assumed to weaken the aforesaid relationship. Consequently the following is hypothesized:
25 board members prefer stability. This is translated to the entry mode choice literature in that older board members favor a less risky entry mode (i.e. joint venture). A closer look to the features of an older BOD reveals that it highly matches the characteristics of high UA cultures. Both older managers and individuals from high UA countries tend to avoid uncertainty and risk, thereby looking for ways to ensure a culture of controlled ownership and regulated control. Higher UA is hence postulated to cause the hypothesized relationship to be even stronger. Correspondingly, one should realize that lower levels of UA might weaken the relationship as the risk-averseness promoted by older board members may be reduced by the relative tolerance for risk in low UA countries. Following these considerations, we can hypothesize:
The level of UA in the home country might also affect the assumed relationship between education level and entry mode choice. Of the four consequences of education - openness to change, international commitment, problem solving skills and cultural confidence – the first has the largest potential to be effected by the level of UA. Higher education might lead to openness to change; yet managers in high scoring countries are often resistant to change (Hofstede, 2003). For instance, employees rarely switch employer as one tries to avoid uncertainty as much as possible. A high level of UA might thus downward the openness to change and thereby weaken the hypothesized positive influence. As such, the final hypothesis is:
26 2.7 Conceptual model
FIGURE 1 presents the final conceptual model.
27 3 METHODOLOGY
This chapter presents insights into the methodology used to test the hypotheses. It starts with a description of the data and the sample. The section follows with the operationalization of the variables. Lastly, it describes the empirical approach taken to estimate the model.
3.1 Data source
Quantitative secondary data is gathered to measure the influence of BOD characteristics on entry mode choice. The data is obtained from several sources, with Zephyr and firms’ annual reports as main sources. Zephyr is a comprehensive database of deal information, which has already featured several relevant studies (e.g. Craninckx and Huyghebaert, 2011). The data is sourced from Bureau van Dijk’s range of global company information databases with the purpose of everyone being able to compare deals (Bureau van Dijk, 2016). The University of Groningen grants access to this database. Most importantly, Zephyr contains entry mode data on whether firms have selected an acquisition or a JV, complemented with information on important dates around the entry (rumor date, announcement date, completion date), the deal value and status and the name and geography of both the acquirer and the target. For the purpose of this study, the rumor date is used to determine the year in which the BOD characteristics are studied. For example, if the rumor date is 01/05/2013, then further data is collected from the firm’s annual report of 2013. The rumor date is officially defined as the date that the deal was first mentioned (Bureau van Dijk, 2016). The rumor is an unconfirmed report, yet it is assumed to be the closest guess of the year that the BOD planned on entering a foreign country. Other said, the rumor date is the most reliable date to work with as this date is the earliest date which Zephyr researchers can ascertain using evidence. One could argue that the entry mode choice might have been taken before the rumor date, yet this is unlikely to be an issue as board change at a remarkably slow pace with an average tenure of 8.5 years (Spencer Stuart, 2016).
28 nationality, education) is not found in annual reports, information is searched for on Bloomberg.com in order to overcome inaccuracies. This website provides detailed information on individual managers, such as his/her age or career path. Regarding the cultural dimension scores, the geography of the acquirer is used as an indication of the country to be studied. Hofstede’s (2017) website provides information on the scores on PD and UA of this particular country. Finally, data on the control variables is collected by a variety of sources (see 3.3.4).
3.2 Sample
The final sample is based on several criteria. For obvious reasons, only deal types that include either an acquisition or a JV are selected. Also, only completed entry modes are examined so to ensure that the appointed entry mode has truly been used to enter a foreign country. A time period of 2010-2015 is selected because recent data is preferred, yet one should take into account that data on 2016 might still be incomplete. Furthermore only publicly listed acquiring firms are approved, as this allows for a study obtained from secondary sources. All data needs to be available online or in the database, as contacting the firm for additional information will not be possible. Lastly, the general rule of thumb is that to detect reasonable-size effects with legitimate power, one needs 10 observations per parameter. As this study contains 4 independent variables (IV), 2 moderator variables, 1 dependent variable (DV) and 6 control variables, the sample needs at least 130 observations. The objective is therefore to analyze 200 firms.
29 3.3 Measures and variables
3.3.1 Dependent variable
Entry mode choice is operationalized as the MNC’s choice of entry mode when entering a foreign country. Data on the type of entry mode is collected from Zephyr. The variable is scaled based on research by Kogut and Singh (1998), which has proven to be valid and reliable. They created a bivariate variable, also called a dummy-variable, which takes a value of 0 if the market entry is in the form of a JV and a 1 if the market entry is in the form of an acquisition.
3.3.2 Independent variables
The IVs are the MNCs’ board characteristics. Only information about the board members in position at the rumor date is included for each entry. Data is collected from annual reports or Bloomberg.com. Each IV is divided by total board size to enable comparison between different BOD’s and to prevent inaccuracies due to unbalanced magnitudes of the BOD. As such all variables are scale (interval) variables. The variables are operationalized as follows:
Nationality: Nationality is operationalized as the nationality of each individual board member. When a board member had two nationalities, the foreign nationality is selected, since this person might bring influences associated with this dual background (Van Veen and Elbertsen, 2008). If no information is explicitly available, nationality is derived from indirect career information. This method turned out to be reliable in that in a few cases, nationality was discovered at a later stage and consistently corresponded to the appointed nationality. Following Staples (2007), nationality is expressed in a percentage that represents the number of foreign nationalities in the board divided by total board size. The values fall within a continuous range of 0-1.
Age: Age is operationalized as an individual board member’s age expressed in years. The sum of individual ages is divided by total board size to arrive at the BOD’s average age. The values fall within a continuous range of 43–68 years.
30 ‘secondary education’, ‘vocational post-secondary education’, ‘bachelor’, ‘master’, ‘doctor’. The categories differ from the regular framework in that all secondary education categories are grouped together, at it is expected that most board members fall within a higher category (Stanford Graduate School of Business, 2015). All categories are given a score (from 1-6). The number of board members that fall within that specific category multiplies these scores. This number is in turn divided by the total board size in order to get an approximate mean of the education level attained by the board of directors. The values fall within a continuous range of 3.6-6. Although the mean might be an inappropriate value for measures of education, this method is selected as it enables to let SPSS work with an interval data type and to enable comparison between BODs. Second, the type of study of each board member is taken as a proxy for the type of functional background he or she possesses. If the individual has an educational degree in operational areas such as supply chain management, manufacturing, process engineering, law, finance or accounting, he or she has a throughput background. If the individual has a degree in fields like marketing, sales, entrepreneurship or R&D, he or she has an output background. Functional background is scaled in two steps. First, a bivariate variable is created that takes a value of 0 if the board member has an output background and a value of 1 if the board member has a throughput background. The final value used in the analysis is the number of board members with a throughput background divided by total board size. The values fall within a continuous range of 0-1.
3.3.3 Moderator variables
The cultural dimensions are based on the work of Hofstede (1980). The dimensions are not specifically based on the year of rumor, as relative national scores are time invariant. They are operationalized as follows:
Power Distance Index (PDI): Degree of PD in the home country. Uncertainty Avoidance Index (UAI): Level of UA in the home country.
31 FIGURE 2
PDI and UAI of the home countries involved in the sample
3.3.4 Control variables
The control variables control for factors that are known to influence entry mode choice. The variables are aligned with those used in previous research and are based on firm-, industry- and country level respectively.
Firm size: Previous research suggests firm size to significantly affect the type of ownership used to enter a foreign country (Barkema and Vermeulen, 1997). The assumption holds that larger (and sometimes more successful) companies possess more resources. The foreign commitment may thus constitute a relatively small proportion of the overall resources. This often leads to the tendency of larger firms to make great resource commitments in their foreign ventures and to use more integrated ownership structures (i.e. acquisition). Various firm sizes may thus result in different views to the correct entry mode choice (Brouthers et al., 2000). Firm size is hence included as a control variable in this study. It is operationalized by the number of employees a MNC’s parent firm has at the year of the rumor date. The data is gathered from firm annual reports. The values range from 10-60.000 employees.
Firm experience: Evidence shows that firm experience across borders has proven to guide future entry mode choices (Nadolska and Barkema, 2007). Specifically, firms with more international experience tend to choose entry modes associated with higher levels of commitments (i.e. acquisition). Firm experience is therefore incorporated as a control variable in this study. Firm experience is expressed in the number of countries that the firm was active in before the year of rumor. The data is gathered from annual reports. The values range from 0-190 countries.
32 industry to significantly influence entry mode choice. Consequently, this study follows extensive literature (e.g. Kogut and Singh, 1988) in that it controls for industry type. More specifically, Erramilli and Rao (1993) present support for service firms responding differently to entry mode choices than manufacturing firms. Manufacturing expansion requires, in general, great resource commitments than service expansion. This increases the likelihood of a situation of uncertainty wherein the value of the invest opportunity cannot be accurately predicted. Consequently, manufacturing firms may be more willing to keep the initial value low (i.e. JV) while obtaining an option for future investment (Brouthers et al., 2003). In contrast, service firms may be more prone to a riskier entry mode (i.e. acquisition) as their initial resource commitment is relatively small. This study therefore controls for industry type (manufacturing vs. service specifically). A dummy variable is created, which takes the value of 0 if the entry occurs in the service industry and 1 if it occurs in the manufacturing industry. Gross Domestic Product (GDP) Host Country: Several authors (e.g. Zeng, Shenkar, Lee, & Song, 2013) argue host country GDP to be a positive predictor of entry mode choice. Specifically, a less developed market increases the probability of a firm facing risks and uncertainties when entering this market (Barkema and Vermeulen, 1998). As a JV is a commonly used instrument to safeguard the firm against these risks, a less developed market increases the probability of a firm choosing a JV. Host Country GDP should therefore be controlled for in this study. Host country GDP is measured using the two separate control variables size and growth. Although both control variables concern GDP, they are included separately as they may lead to different conclusions. For example, a large country may look attractive in terms of economic size, but at the same time suffer from a large economic decline. An investigation of only a single variable may thus give a one-sided message. The size of the host market is described by GDP per capita in the year of rumor, measured in US dollars. In order to make it fit for statistical analysis, the natural logarithm of these amounts is used. This logarithm variable ranges from 6.32 to 13.42. Furthermore, the growth of the host market is measured by the most recent growth forecast of GDP per capita (expressed as a percentage) in the year of the rumor. The values range from -6.90 to 13.60. Data on both variables is obtained from the databank of the World Bank (World Bank, 2017).
33 and regulations and informally through culture, corruption levels and local business habits. According to Ingram and Silverman (2002: 20), “institutions (e.g. property rights) directly determine what arrows a firm has in its quiver as it struggles to formulate and implement strategy and to create competitive advantage”. For this reason institutional environment is included as a control variable in this study. Following research of Meyer et al. (2009), this research uses freedom of corruption as a proxy for institutional environment. Data is collected based on the economic freedom index from the Heritage Foundation (2017). This index offers fruitful information with regards to market-supporting institutions, thereby focusing on the freedom of the firm and individuals and their business opportunities (Meyer et al., 2009). The index is based on twelve quantitative and qualitative factors, with each category graded on a scale of 0 to 100. A country’s overall score is derived by averaging the twelve scores. The final values fall within a continuous range of 52-90.
3.4 Data analysis
34 4 RESULTS
This chapter demonstrates the empirical results of the binomial logistic regression analysis. It starts with an overview of the descriptive statistics of the variables used in the analysis. After that, the results of three consecutive models are demonstrated. The model is run multiples times with different variables in order to test for the predictive contribution when additional variables or interaction effects are added to the model. This set up also increases the validity and robustness of the results as it enables the researcher to check whether the coefficients do not diverge significantly among the different models. Model 1 comprises only the control variables. Model 2 is the direct-effects model. Model 3 is the final model in that it includes both direct and moderation effects.
4.1 Descriptive statistics and assumptions
The analysis starts with an inspection of the descriptive statistics of the variables used in the analysis. Specifically, TABLE 3 shows the means, standard deviations and the correlations of the DV, the four IVs, the two moderator variables and the six control variables.
TABLE 3
Summary of means, standard deviations and correlations.
35 <. 01), thereby following existing literature in that a strong relationship exists between PD as a cultural trait in the home country and the entry mode choice (Morschett, Schramm-Kleinn, & Swoboda, 2010). As expected, most control variables significantly relate to entry mode. Furthermore, several significant correlations are found between the IVs. Background is significantly related to nationality, age, and education. Also, education is significantly correlated to PD and UA. The latter two are also significantly related to each other. There are also many significantly correlated relationships among the control variables and other variables. As this might become problematic for the analysis, a collinearity diagnostic test is therefore performed on all IVs in order to achieve more reliable results. Fortunately, as no variables show a correlation >.90, a tolerance factor <0.2 or a VIF factor >5 (Leeflang, Bijmolt, Pauwels, & Wieringa, 2014) there are no signs of multicollinearity among the IVs. The significant correlations should thus not pose any problems during the regression.
Seven assumptions need to be satisfied to run the binomial logistic regression for model 1 and 2 (Garson, 2009). Additional assumptions for model 3 are discussed in 4.4. The first four assumptions relate to the choice of study design and measurements. To illustrate, the sample size is sufficient in that maximum likelihood needs at least 15 cases per IVs. In this case, 6 IVs means that 90 cases hold the minimum, being <200. The observations are independent. The DV is measured on a dichotomous scale with mutually exclusive and collectively exhaustive categories. The model includes multiple continuous IV’s. The other three assumptions relate to how the data fits the model. The IVs are not highly correlated to each other. There are no significant outliers, leverages or influential points. Linearity of the continuous variables with respect to the logit DV is assessed via the Box-Tidwell (1962) procedure. A Bonferroni correction is applied using all nine terms in the model resulting in statistical significance being accepted when p < .00555 (Tabachnick & Fidell, 2007). Based on this assessment, all continuous IVs are linearly related to the logit of the DV. Overall, all the necessary assumptions are met.
4.2 Control variables
36 First, the results show overall statistical significance of model 1 (p < .05), meaning that the control variables succeed in predicting the DV compared to no variables. To test for model fit and explained variation, we look at the Hosmer and Lemeshow goodness of fit test and the Nagelkerke R2. The HL-test test is not statistically significant (p > .05), indicating that the model is an adequate fit. The R2 value of .414 indicates that 41.4% of the variance in the DV is explained by model 1.
This analysis is commonly used to forecast whether cases are correctly predicted (i.e. given a correct value of 0 or 1) by means of the variables. Logistic regression estimates the probability of an event (in this case, entering via an acquisition) occurring. To assess the effectiveness of this prediction, the observed and predicted classifications are used. Model 1 correctly classifies 75.5% of the entry modes overall. The control variables thus improve the overall prediction of entry modes into the observed categories of acquisition and JV. The sensitivity, the percentage of entries that are an acquisition and are correctly predicted as an acquisition, is 87.0%. The specificity, the percentage of entry modes that are JVs and are also predicted as JVs, is 64.0%. The positive predictive value (70.7%) is the percentage of correctly predicted acquisitions compared to the total number of acquisitions in the sample. That is, of all cases predicted as an acquisition, 70.7% were correctly predicted. The negative predictive value is the percentage of correctly predicted JVs compared to the total number of JVs in the sample. In our case, 75.5% is correctly predicted.
Amongst the control variables, significant effects are found for firm size and GDP size (p < .05). Firm size and GDP thus significantly influence entry mode. Specifically, this indicates that an increase in size leads to an increase in the likelihood of an acquisition. Note that all these measurements will also be used to assess model 2 and 3. For these models, the contribution of each IV or interaction effect to the model and its statistical significance will also be discussed.
4.3 Direct-effects analysis
The second step is to include the IVs (model 2). Model 2 is thus limited in that it only captures the direct effects of nationality, age, education level and background on the likelihood of an acquisition over a JV. This model enables easy identification of the IVs’ contributions to the model.
37 In where Z is a linear combination of the IVs and coefficients to be estimated:
Here, B0 is the intercept, Bn are the regression coefficients, ε is the error term, and “i” refers to the ith entry mode of 200 entry modes taken into account. X
n are the IVs (X1 = nationalities, X2 = age, X3 = education level, X4 = functional background), C stands for the control variables (C1 = size, C2 = experience, C3 = industry, C4 = GDP size, C5 = GDP growth, C6 = environment).
The results of model 2 are shown in in TABLE 5. The direct-effects binomial logistic regression model is statistically significant, χ2(4) = 88.322, p < .05. We can conclude that there is a significant relation between the IVs and the DV. This also indicates that the joint significance of all variables is not redundant and this model is better than the intercept-only model. According to the HL-test it has a good fit (p > .05). The model explains 47.6% (Nagelkerke R2) of the variance in entry mode, which is 6.2% more than model 1. 78.5% of the cases are correctly classified. Sensitivity is 72.0%, specificity 83.0%, positive predictive value 76.1% and negative predictive value 81.3%.
38 positive coefficient (B = .06), so although this relation is statistically significant, the relation shows a different direction than expected.
4.4 Moderation analysis
A binomial logistic regression also makes it possible to test moderator effects by allowing for interactions between IVs to predict the DV. Henceforth, a third regression is performed to assess the increase in variation explained by the addition of the interaction terms. Model 3 is thus most complete as all direct effects as well as all possible interactions between the cultural variables and the board characteristics are regressed on the DV. The following interactions effects are included in the model: PD x nationality, PD x age, PD x education level, UA x nationality, UA x age and UA x education level. Model 3 results in the following model equation:
In where Z is a linear combination of the IVs and coefficients to be estimated:
Here, B0 is the intercept, Bn are the regression coefficients, ε is the error term, and “i” refers to the ith entry mode of 200 entry modes taken into account. Xn are the IVs (X1 = nationalities, X2 = age, X3 = education level, X4 = functional background, X5 = PD, X6 = UA), XnXn represents the interaction effects and C stands for the control variables (C1 = size, C2 = experience, C3 = industry, C4 = GDP size, C5 = GDP growth, C6 = environment).
39 < .002 (Tabachnick & Fidell, 2007). Based on this assessment, all continuous IVs are linearly related to the logit of the DV. The only critical point is that the variables show multicollinearity (VIF >200). There are two possible reasons for this problem. First, it may be caused by the inclusion of a variable in the interaction that is computed from other variables in the dataset. Second, it can also result from repeatedly using the same variable. Our solution to deal with multicollinearity is based on research by Aiken and West (1991). They suggest to mean center the variables in order to reduce the covariance between the linear and interaction terms, thereby automatically solving for multicollinearity. After mean centering all IVs, we can safely argue that all the necessary assumptions are met.
The results are again shown in TABLE 5. The interaction effects binomial logistic regression model is statistically significant, χ2(19) = 109.456, p < .05. We can therefore conclude that there is a significant relation between the IVs and the DV. The insignificant value of the HL-test indicates the model to have a good fit (p > .05). The model explains 56.2% (Nagelkerke R2) of the variance in entry mode, which is 8.6% more than model 2. The increase in the R2 from Model 2 to 3 indicates that the model with interaction terms included is the best predictor of the DV. The fact that model 3 predicts significantly better than the other models is confirmed by the fact that 80.0% of the cases are correctly classified. Model is henceforth said to have large predictive capacity. Sensitivity is 85.0%, specificity 75.0%, positive predictive value 77.3%, negative predictive value 83.3%. The increase in correct classification indicates that model 3 is best in the overall correct prediction of entry modes in the observed categories acquisition and JV. As model 3 has the best predictive capacity and explained variance, model 3 will be used for interpretation in the remainder of this study.
40 acquisition. Age shows a positive coefficient (B = .13), so although this relation is statistically significant, the relation is in a different direction than expected. The second goal of this model is to check whether the effect of either PD or UA influences the relation between board characteristics and entry mode. Of the six moderating relationships, four of them are significant under the <.05 level: PD x age, PD x education, UA x age and UA x education.
The interpretation is as follows. For PD x age, an increase in one unit to the average level of PD increases the effect of age on entry mode with 1.01. Other said, an increase in PD increases the positive effect of age on the likelihood of exhibiting an acquisition. For PD x education, an increase in one unit to the average level of PD decreases the effect of education on entry mode with .84. That means, an increase in PD weakens the effect of education on the likelihood of exhibiting an acquisition. For UA x age, an increase in one unit to the average level of UA increases the effect of age on entry mode with .99. That means that an increase in UA increases the positive effect of age on the likelihood of a firm exhibiting an acquisition. Lastly, for UA x education, an increase in one unit to the average level of UA decreases the effect of education on entry mode with 1.13. That means that an increase in UA weakens the effect of education on the likelihood of exhibiting an acquisition.
4.5 Conclusion
Having discussed the different models, the following conclusions are drawn: TABLE 4
41 TABLE 5
