1
‘I’M GONNA MAKE HIM AN
OFFER HE CAN’T REFUSE’
A study on the effects of corruption on the entry mode decisions of MNE’s from a
transaction cost perspective
Master Thesis
MSc International Business & Management
University of Groningen Faculty of Economics and Business
March 2018
By
2
Master Thesis
MSc International Business & Management
‘I’m Gonna Make Him An Offer He Can’t Refuse’
A study on the effects of corruption on the entry mode decisions of MNE’s from a
transaction cost perspective
March 2018
By
Marc Ravenhorst
S3023516
Supervisor: Dr. M.H.F. ridder de van der Schueren
Co-Assessor: Dr. D.H.M. Akkermans
Faculty of Economics and Business University of Groningen
Duisenberg Building, Nettelbosje 2, 9747AE Groningen, The Netherlands P.O. Box 800, 9700AV Groningen, The Netherlands
3
Abstract
Corruption is an important factor influencing strategies of multinationals worldwide, especially considering the differences in corruption between countries. Since culture has a high influence on the responses of countries on corruption, the corruption and culture in a particular country go hand in hand. Therefore, this study researches the influence of corruption on the entry mode decisions of firms originating the United States into several countries, both with higher and lower levels of corruption, whereby cultural distance is regarded as the moderator in this relationship. To do so, over 7000 entries into 29 countries are included in the sample. The transaction cost theory and previous studies are used to predict the direction of the proposed relationships. This study finds significant evidence on the effects of high negative corruption distance on the probability of a wholly owned subsidiary, and of low negative corruption distance of joint venture. In addition, high positive corruption distance decreases the probability of joint venture. Medium and high cultural distance moderate in the effects of high and low negative corruption distance, however, no significant results were found for corruption as a moderator in positive corruption distance.
Keywords: Corruption, Culture, Entry Mode, Wholly owned subsidiary, Joint Venture, Transaction cost theory
4
Table of Content
Abbreviations ... 5
Definitions... 5
1. Introduction ... 6
2. Transaction Cost Economics ... 8
3. Hypothesis Building ... 10
3.1 Dependent variable: Entry Mode ... 11
3.2 Independent Variable: Corruption ... 13
3.3 Moderator: Cultural distance ... 16
3.4 Conceptual model ... 20
4. Methodology ... 22
4.1 Dependent variable, independent variable & Moderator ... 22
4.2 Sample ... 25 4.3 Descriptive Statistics ... 26 4.4 Research Methods ... 28 5. Results ... 29 5.1 Control Statistics ... 29 5.2 Hypothesis testing ... 30 5.3 Robustness test ... 32 6. Discussion ... 34
6.1 Conclusion and implications ... 34
6.2 Implications ... 36
6.3 Limitations and suggestions for further research ... 37
5
Abbreviations
TCE Transaction Cost Theory
CD Cultural Distance
JV Joint Venture
WOS Wholly Owned Subsidiary
MNE Multinational Enterprise
CPI Corruption Perception Index
Definitions
Focal firm The leading firm at the top of the value chain; initiator of the international business transaction
Local partner Separate entity not directly associated with the lead firm that starts a joint venture in the host county with the lead firm
Corruption Misuse of public power for private gain; Bribery of governmental officials
Cultural Distance The difference in culture between the home and host country
6
1. Introduction
Despite having been used in a case of blackmailing, the well-known movie quote “I’m gonna make him an offer he can’t refuse” from the Godfather trilogy is often associated with corruption. This tends to be the case due to the mafioso being associated with corruption and the possible of interpretation of ‘offer’ as an illegal financial transaction. On the one hand, this procedure is frightening and disturbing as it undermines the rules and laws of society, on the other hand, the godfather trilogy is fiction and the events tend not to happen to most places in society. However, this does not mean that corruption is not present, as many countries are still dealing with the problems caused by corruption. Therefore, an important question to address is what corruption is and what its effects are.
To explain and understand corruption, imagine being the CEO of a multinational enterprise, you are starting a new project abroad for which you need a permit in the host country. Unfortunately, it is relatively hard to get the permit, as many government officials are not very cooperative. From your local partner, you heard that it might help to give the governmental officials a small present, just so they are more willing to work with you. Would you do this? Most likely you said yes. However, would you also do this if this is regarded as corruption in your home country? And what if it affects your image in your home country? This action, which might be completely normal for the local partner, might cause a major dilemma for the CEO of the focal firm.
7
might have a tremendous effect on the globalization strategy of a multinational, it might even lead to a case whereby corruption has become institutionalized in such a way that the MNE has to choice between either being active in corruption or not entering the country at all (Ashforht & Anand, 2003). Consequently, as corruption is of high influence on multinationals, it is likely to assume that strategies are adjusted to the corruption levels in the host country.
An important decision for MNEs is how to enter the market, whereby the main dilemma is whether to go for full control or to cooperate with a local partner. Considering the entry mode decisions being a business transaction, the transaction cost theory can be used to define entry mode possibilities. This is useful since the theory identifies market- and hierarchy-based transactions, whereby each type of transactions has different entry modes (Williamson, 1985). Moreover, previous scholars have found a significant relationship between the levels of corruption in a country and the entry mode of an MNE. For example, Karhunen and Ledyaeva (2012), among others, concluded that MNEs tend to enter a corrupt market via a joint venture, whereas Tekin-Koru (2006) concluded the opposite. However, most of the scholars conducted their researched bilateral or multilateral using a small number of countries. Providing a research gap for a study using a larger number of countries, as other home or host countries might have been more prevalent in these previous studies. Furthermore, as there are different levels of corruptions among different nations, and corruption is perceived in different ways, it might be suggested that culture is an influencing factor. Several previous scholars have researched the connection between the corruption and culture. For example, Robertson and Watson (2014) who suggested this connection to be especially strong in the cultural values of uncertainty avoidance and masculinity. In addition, Husted (1999) conclude that high uncertainty avoidance, high power distance, and high masculinity tend to increase corruption in a country. Thus, it is argued that culture and corruption are connected. However, a scalable study on the threefold connection between entry mode, corruption and culture has not been conducted.
Consequently, a gap has been identified in the existing research. Therefore, this study investigates how these factors are related. To do so, an answer will be found to the question: “What
8
This paper starts with elaborating on the transaction cost economics and its influence on business strategies in an internationalization context. Then, the TCE is used to elaborate on the entry mode decisions of MNE’s and to develop the hypothesis. Section four continues with an explanation of the research framework, the scope and the framing of the study, and the conceptualization of the dependent variable, independent variable, and the moderator. The data is analyzed, and the results presented, in section five. Finally, in section six the results are concluded, whereby limitations and suggestions for further research are given.
2. Transaction Cost Economics
The transaction cost theory (TCE) states that firms can either choose to conduct a transaction in the market (outsourcing) or in the hierarchy (inhouse). It depends on the transaction which mode is most suitable. The transaction costs are the costs of writing, executing, and enforcing contracts, they vary across markets and depend on the human decision makers and the objective properties of the market (Williamson, 1975). Thus, firms will consider the costs from their transactions, and choose the governance mode with the lowest costs. For example, when a firm has developed a new product, they have to decide whether to produce it inhouse or to outsource the production. Depending on the complexity of the product, it might become so expensive to write and monitor contracts to conduct the transaction at arm’s length, that it becomes cheaper to conduct the transaction inhouse.
9
evaluating whether the contractual compliances have been met (Steenkamp & Geyskens, 2012). Third, the frequency of the transaction explains how often a transaction takes place, whereby it is either one-time, occasional, or frequent (Williamson, 1979).
The human factors of the transaction cost theory are bounded rationality, risk neutrality, and opportunism (Steenkamp & Geyskens, 2012). Bounded rationality implies that, despite the extensive capability of the human mind to see problems and find solutions, it is bounded by rationale set limits. Meaning that outcomes can arise which were previously not expected as they arise outside of the limits of the rationality (Williamson, 1975). Risk neutrality refers to the assumption of the transaction cost theory of decision makers being risk neutral instead of being risk seeking or risk averse (Brouthers, 2013). Opportunism refers to a situation whereby decision makers or actors in the transaction act in self-interest (Williamson, 1975), meaning that an individual might act according to its self-interest while disregarding the consequences for others. In terms of contract enforcements, opportunistic behavior of the partners is often regarded negatively (Demirbag, Tatoglu, & Glaister, 2009).
10
firms using entry modes as predicted by the TCE performed better than otherwise. Thus, providing support for the TCE with regard to entry mode.
In contrast to the support presented for the TCE, it has also received some criticism. For instance, Ghoshal and Moran (1996) claim a misinterpretation of opportunism in the theory. They state that, while the TCE claims opportunism to be strategic behavior, it is not specified how opportunism is created or reduced. More specifically, Williamson assumes opportunistic behavior to be present in transactions and relationships, while he does not specify what causes behavior, making situations in which opportunistic behavior might exist harder to predict. Furthermore, while the TCE relies on transaction costs, it is sometimes hard to actually identify or measure these costs, and even in perfect internalization, there are still transaction costs due to the sourcing of raw materials. As a consequence, the decision to outsource or to internalize based on the TCE is not a simple calculation. Instead, it is more a judgement based on experience and perceptions of risks, which differ from person to person (Jones, 1997).
Taking the fundamentals of the transaction cost theory and the criticism it is has received in consideration, this study will apply the theory to the research question as set in the introduction. Meaning that the TCE will be used to predict the probability of an entry mode in cases of high or low corruption. By considering the criticism the theory has received, the predictions are approached with caution and supplemented with results from previous scholars. Therefore, the TCE will solely be used to define entry mode strategies and to determine the probability of a certain entry mode, instead of finding explanations for the empirical results. Thus, both existing evidence and the TCE are used to determine which entry mode is more likely used in host environments with high corruption, and which in environments with low corruption.
3.
Hypothesis Building
11
process, the host country factors are taken into considerations of which a wide variety exist. Some of these factors are direct stakeholders such as suppliers and potential buyers, others are the environmental or country factors, such as the infrastructure or the government policies. Each factor will somehow influence the entry mode decision and strategy of the multinational. This study is specifically focused on the factors of host-country corruption and cultural distance and will research the influence of these two variables on the probability of a specific entry mode.
3.1
Dependent variable: Entry Mode
As long as firms have internationalized, entry mode decisions have been made. As a result, the academic literature has conducted extensive research in the past on the entry mode decisions and the possible entry modes. Furthermore, the existing literature has been focused on the reasons for firms to follow a certain strategy when internationalizing. For example, the Uppsala-model states that firms internationalize by gradually increasing the foreign commitments (Johanson and Vahlne, 1977), or the research-based view, which evaluates firms as a set of assets and resources, that can be utilized in different locations (Locket & Thompson, 2001). Since this study does not aim to find an explanation of why or how firms expand abroad, the aforementioned theories are not included. Instead, this study will select specific entry modes and strive for empirical evidence on the probable increase of these modes in foreign environments with high or low corruption.
The existing literature makes a distinction between non-equity modes and equity modes of foreign expansion (Pan & Tse, 2000; Brouthers & Nakos, 2004). With regard to host country factors, the non-equity modes tend to have a limited involvement with the host-country (Taylor, Zou, & Osland, 1998). For example, in exporting and licensing, the transactions are made between the home-country firm and a separate host-country firm (Young, 1983). Within the equity modes, a distinction can be made between full control and shared control. Full control usually entails greenfield investments or acquisitions, while shared control implies a joint venture between the focal firm and a local partner (Williams, et al., 2011). Since the equity modes imply an involvement of the focal firm in the host-country, there is a direct influence of the host-country factors on the activities and decisions of the focal firm. Therefore, the current study will further focus on the equity modes and disregards the non-equity modes.
12
transactions. It depends on the efficiency of the modes in combination with the costs involved in writing, executing, and enforcing of contracts, and the human factors, whether a transaction is conducted in the market or in the hierarchy (Williamson, 1975). Meaning that when the transaction costs are too high to conduct the transaction in the market, the transaction will be conducted in the hierarchy. Important to note is that the transactions costs involved arise due to transferring of goods and services between parties in the host country. Thus, the host country factors might increase the costs of writing, executing, monitoring, and enforcing contracts with a local partner to such a degree that it is more beneficial for the focal firm to conduct the transaction without a partner.
Taking the existing literature with regard to non-equity and equity modes, and the transaction cost theory into account, the current study will distinguish joint ventures (JV) and wholly owned subsidiaries (WOS) as possible entry modes. Whereby a JV is regarded as the ‘market’ and a WOS as the ‘hierarchy’ of the TCE. By doing so, this study follows the approach of several previous scholars, such as Taylor et al., (1998), Jung (2004), and Williams et al. (2011).
A joint venture is an entity in which two or more economic groups share assets, risks, and profits. It is common to have joint ventures whereby two partners have a 50/50 or 51/49 distribution of shares, however, any distribution of shares is possible (Young, 1989). Likewise, a wholly owned subsidiary is a corporation in which the focal firm has full control (Young, 1989). The main advantage of a JV from the perspective of the focal firm is the ability to benefit from the knowledge and experience of the local partner. In addition, a JV allows the focal firm to share the costs and risks of entering the market with the partner. However, a major disadvantage is the requirement of the lead firm to share its equity and own knowledge (Hill, 2009). For example, an MNE from a developed country with specific technological knowledge starts a JV with a local partner in order to gain the knowledge of the foreign country, while risking spillover effects of its knowledge. Therefore, the firms will negotiate a contract to prevent spillover effects, and consequently, increase the transaction costs. In contrast, a wholly owned subsidiary has the advantage of maintaining full control over the knowledge and assets, resulting in significant lower spillover effects. However, this implies the that the focal firm is bearing all the risks and costs itself (Hill, 2009). The lead firm will therefore assess the gains and losses of both entry modes to find the most beneficial strategy.
13
3.2
Independent Variable: Corruption
Defining Corruption
Corruption is often defined as ‘the abuse of public power for private gain’ (Transparency International, n.d.) or as the ‘active or passive misuse of powers of public officials for private financial or other benefits’ (OECD, n.d.). Based on these definitions, corruption is explained as a situation in which a government official misuses his powers for private gain, either financially or otherwise. Furthermore, Doh, Rodriguez, Uhlenbruck, Collin, and Ede (2003) explain corruption as a tax which increases the direct and indirect costs for several stakeholders. They state that the direct costs result from the direct interaction between the firm and the government during corrupt activities. The indirect costs result from public-sector failure from missing or weak institutions and failing government policies due to corruption. Furthermore, corruption can be divided in pervasiveness and arbitrariness. Pervasiveness implies the degree to which a firm will encounter corruption in its normal interaction with the government. Arbitrariness is a quantifiable risk over knowable outcomes, meaning that it is derived from the statistical uncertainty involved in corruption (Rodriguez et al., 2005).
Furthermore, the existing literature explains corruption as both a positive and a negative factor. In the positive way, it is often referred to as ‘grease in the wheels of commerce’, whereby it is assumed that corruption will smoothen the way for firms in environments with failing governmental institutions (Duanmu, 2011). The negative perspective is referred to as ‘sand in the wheels of commerce’, as corruption imposes increased costs for firms (Duanmu, 2011). Since extensive empirical evidence exist to support the negative perspective of corruption (e.g. Habib & Zurawicki, 2002; Egger & Winner, 2005), this study will follow the ‘sand in the wheels’ perspective.
Corruption and the transaction cost theory
A variety of studies have been conducted regarding the TCE and corruption. Demirbag, et al. (2009) concluded that corruption results in higher transaction costs as it increased the specific assets needed to safeguard contracts. According to Karhunen and Ledyaeva (2012), the transaction cost theory explains that foreign firms use a strategy of shared ownership to deal with the external uncertainty caused by corruption.
14
and the negotiating, writing, monitoring and enforcing of the contracts, either with the local partner in a joint venture or with partners at arm’s length needed to enter the respective market. Furthermore, since both corruption and anticorruption policies can be institutionalized in a country, corruption influences the transaction costs from an institutional perspective (Tekin-Koru, 2006; Spencer & Gomez, 2011). Meaning that, as corruption becomes more institutionalized in a country, the pressures on the subsidiary to conform to corruption also increases. From the transaction cost perspective, this implies that the costs of monitoring and enforcing the agreements with local firms increases when the home country firm is not willing to conform to corruption. This suggests a higher probability of a WOS in corrupt host countries (Tekin-Koru, 2006). On the other hand, if the home country has corruption institutionalized, it is assumed the firm will not be concerned with protecting its subsidiary in the host country from corruption pressures, thus, the transaction costs of monitoring and enforcements are lower. Consequently, the probability of a JV will increase.
Several other factors of the TCE can be linked with corruption in order to predict which entry mode should be preferred in environments with high corruption and in environments with low corruption. For example, corruption can increase uncertainty over governmental behavior due to opportunistic behavior of governmental officials or due to unclarity over procedures or expectancies of governmental officials (Prasad & Shivarajan, 2015). Furthermore, governments have the monopoly in providing legislation services, making them providers of specific assets. In environments of high corruption, the government officials might only be willing to cooperate with a corporation after receiving a fee. In such a situation the corporation might be pressured to resort to corruption (Prasad & Shivarajan, 2015
Corruption and the entry mode
15
pressurize a joint venture via the local partner, while they are not eager to pressurize a WOS (Spencer & Gomez, 2011). Furthermore, evidence has been found to suggest that corruption, as part of the institutional environment, affects FDI flows (Egger & Winner, 2005). Demirbag et al. (2009) argue and confirm that the level of corruption is an important factor in the decision between entry modes of firms. In cases of high differences in corruption perception, decision makers will perceive a higher uncertainty and a higher risk, thus, a JV will be the appropriate entry mode. Egger and Winner (2005) suggest and confirm a two-fold effect of corruption on FDI. On the one hand, corruption has a negative influence by increasing the costs of a foreign entry as firms: have to pay bribes, are active in resource-wasting activities, and have additional risks because contracts cannot be enforced in courts. On the other hand, corruption has a positive effect on FDI as it smoothens the bureaucratic process and allows firms to gain access from publicly funded projects.
In accordance with the transaction cost theory, Tekin-Koru (2006) showed that when the corruption distance between the home and host country increases, the number of joint ventures decreases. Furthermore, when firms from less corrupt countries enter a more corrupt country, there is a slight preference for a WOS compared to a JV in comparison to firms from more corrupt countries (Duanmu, 2011). However, In a similar study, focused on Russia as the home market, Karhunen and Ledyaeva (2012) found evidence to confirm that MNEs are more likely to use a joint venture when the corruption distance increases. Moreover, Godinez and Liu (2015) concluded that an increase in corruption leads to a decrease in FDI. While previous scholars have found clear evidence of the effects of corruption on the entry mode when the home country of the MNE has less corruption, this effect is less obvious for more corrupt home countries (Duanmu, 2011). However, in a study without making a distinction between less and more corrupt, Habib and Zurawicki (2002), found a negative effect of increasing corruption on FDI flows. Finally, Godinez and Liu (2015) did not find significant evidence on the effects of FDI flows from high corrupt home countries into low corrupt host countries.
Hypothesis building
16
expected that when an MNE from a low corrupt home country expands into a high corrupt host country, the probability of a WOS increases. When the MNE enters a country at a low corruption distance or at a high positive corruption distance, the probability of a JV increases. This leads to the following hypotheses:
Hypothesis 1a: A high negative corruption distance between the home and the host country, results in an increase of the probability of a wholly owned subsidiary.
Hypothesis 2a: A low negative corruption distance between the home and the host country, results in an increase of the probability of a joint venture.
Hypothesis 3a: A high positive corruption distance between the home and the host country, results in an increase of the probability of a joint venture.
3.3
Moderator: Cultural distance
Defining cultural distance
Cultural distance is often summarized as the difference in the national cultural characteristics between a home and a host country (Demirbag et al., 2009). Geert Hofstede, who is a major contributor to the research of cultural values, defined culture as the “collective mental
programming of the human mind which distinguished one group of people from another”
(Hofstede, 2011). These definitions imply that the more cultural distant two countries are from each other, the higher the difference in their organizational characteristics are (Kogut & Singh, 1988). Thus, cultural distance is the difference in mental perceptions between groups of people or nationals, which affects the organizational characteristics.
17
a masculine society are success and money while caring for others and the quality of life are more important in feminine societies. Finally, uncertainty avoidance is the degree to which the people feel threatened by uncertain situations and have institutions to avoid these (Hofstede & Bond, 1984). Since countries can receive a score on a scale from 0 to 100 on each value, many different cultures are identified, meaning that the different degrees of the values result in a large number of unique cultures, whereby each culture has its own effect on many aspects of societies (Hofstede, 1986).
The relevance of cultural distance with regard to the research question can be found on its effects on the behavior and cognitive perceptions of both the individuals and the groups. The individual learns the cultural values of his society as he grows up. As culture determines the knowledge, norms, values, and symbols of the society, and thus, of the individual, it influences the belief of what is good and what is wrong (Verhage, 2004). Therefore, the cultural perception determines whether corruption is regarded as normal, as a positive or as a negative influence. Due to cultural differences, high variations exist between the acceptance of corruption between cultures. Since multinationals operate in a variety of cultures, it is logical to assume that they face different perceptions of corruption. The MNE’s have to determine when it is appropriate to be active in corrupt activities, the degree to do this, and its effect on the MNE in the home countries and other host countries. Consequently, the cultural difference between the home and host country influences how MNE’s respond to corruption forces (Kristjánsdóttir, Guðlaugsson, Guðmundsdóttir, & Aðalsteinsson, 2017; Xiao, Lenzen, Benoît-Norris, Norris, Murray, & Malik, 2017).
Cultural distance and the transaction cost theory
18
costs of writing, monitoring, and enforcing contracts. However, the magnitude of this effect depends on the cultural context.
The transaction costs considered in the context of the cultural distance are the costs of searching, negotiating, writing, monitoring, and enforcing contracts and agreements with partners in a joint venture or at arm’s length. The costs are a cause of the entry into the host country and doing business in the host country. Cultural distance increases the uncertainty in a business transaction, increase the problems of bounded rationality, and as the performance of partners is harder to verify, increases opportunism (Gooris & Peeters, 2014). In addition, Williams et al. (2011) state that the TCE can be applied to the construct of cultural distance since firms face costs when setting up and operating a business abroad. Demirbag et al. (2009) state that the TCE and cultural distance are linked via an increased uncertainty and opportunism when the cultural distance increases. Thus, this implies that the underlying concepts of the TCE due to cultural differences.
Cultural Distance and Corruption
19
cultural values and corruption, this study assumes an interaction effect between cultural distance and corruption. Thus, instead of a direct effect of cultural distance on the degree of corruption in a country, it is assumed that the effects of corruption on the entry mode are influenced by the cultural values of a country.
Cultural Distance and Entry Mode
From the perspective of the TCE, it can be suggested that an increase in the cultural distance results in higher transaction costs, and therefore, in a preference for a wholly owned subsidiary. Several previous authors have researched the connection between cultural distance and entry mode. Evidence has been found to both confirm and deny this expected relationship from the TCE perspective. For example, Hennart and Larimo (1998) argue that an effect of cultural distance on the entry mode decision exists as foreign investors would prefer a joint venture partner with knowledge and experience in dealing with the specific cultural environment in which the investment takes place. Therefore, they assume that when the cultural distance increases, the preference for a joint venture by the focal firm increases. This effect is confirmed by Kogut and Singh (1988), who found support for their hypothesis that when the cultural distance between the home and host county increases, firms tend to choose joint ventures over wholly owned subsidiaries. In contrast, López-Duarte, Vidal-Suárex, and Gonzáles-Diaz (2015) found evidence pointing towards the choice of wholly owned subsidiaries when the cultural distance was relatively large, whereby the power distance and uncertainty avoidance values of the host country were lower, and the individualism and masculinity values of the host country higher.
Hypothesis building
20
instead of an independent variable as previous authors have done, this study proposes an interaction effect between corruption and cultural distance on the entry mode decision of an MNE. Thus, it is proposed that cultural distance will strengthen or weaken the effect of corruption on the entry mode decision. Therefore, the following hypotheses are proposed:
Hypothesis 1b: High cultural distance positively moderates the increasing effect of a high negative corruption perception on the probability of a wholly owned subsidiary.
Hypothesis 1c: Medium cultural distance positively moderates the increasing effect of a high negative corruption perception on the probability of a wholly owned subsidiary.
Hypothesis 2b: High cultural distance positively moderates the increasing effect of a low negative corruption perception on the probability of a joint venture.
Hypothesis 2c: Medium cultural distance positively moderates the increasing effect of a low negative corruption perception on the probability of a joint venture.
Hypothesis 3b: High cultural distance negatively moderates the increasing effect of a positive corruption perception on the probability of a joint venture.
Hypothesis 3c: Medium cultural distance negatively moderates the increasing effect of a positive corruption perception on the probability of a joint venture.
3.4
Conceptual model
21
Figure 1 Conceptual model of hypothesis 1a, 1b, and 1c
The effect of a low negative corruption distance, and the effect of a high positive corruption distance, on the probability of a JV, and moderating effects of high and medium cultural distance, as proposed in hypothesis two and three, are visualized in figure 3. From the figure, it can be seen that in a host country at either a high or a medium cultural distance, the effect of low negative corruption on the probability of a JV is positively influenced. Furthermore, in host countries at a high or medium cultural distance, the effect of high positive corruption distance on the probability of a JV, is negatively influenced.
22
4.
Methodology
4.1
Dependent variable, independent variable & Moderator
Dependent variable
As elaborated in 3.1 (entry mode decision), the dependent variable is defined as the entry mode, which is either a wholly owned subsidiary or a joint venture. Consequently, the dependent variable is conceptualized as a dummy variable. By doing so, this study follows the approach of the transaction cost theory and the methods of existing literature (e.g. Makino & Neupert, 2000; Karhunen & Ledyaeva, 2012; Williams et al., 2011; Brouthers & Nakos, 2004). Furthermore, despite the definition of a wholly owned subsidiary being a 100% ownership subsidiary of the parent firm ( Taylor et al., 1998; Hill, 2009), several scholars (e.g. Lópex-Duarte, et al., 2015; Makino & Neupert, 2000) state that even without a 100% ownership a firm can have such as strong control over another firm, that other shareholders can be neglected. This implies that defining a WOS at 100% is too hard. Therefore, most scholars use a cut-off point of 95%. However, this study assumes that even this point is too high, as this could imply a joint venture with a majority partner owning 94% of the shares. In such a case, the control and added value in terms of either knowledge or assets of the 6% partner can still be neglected. For this reason, this study follows the International Financial Reporting Standard which defines the point at which a minority partner has significant control at 20% (Deloitte, n.d.). Thus, a joint venture is defined as a firm with at least two shareholders, whereby at least on of them has an ownership between 20% and 80%. A wholly owned subsidiary has at least one shareholder who has at least 81% ownership.
Independent variable
23
Since the CPI appoints a low score to a country with high corruption and vis-à-vis, the CPI is inverted to make it more intuitive. The difference in corruption is calculated by subtracting the inverted scores of the host country from those of the home country, whereby a negative score represents a positive distance, and a positive score a negative distance.
The independent variable is conceptualized as a categorical variable with three categories: low, medium and high. Due to the differences arising from positive and negative corruption distance, the sample is split into two sub-samples (see §4.2: Sample). The categories of the variable are created separately for the two sub-samples, as a uniform classification system would not fit the data. The countries in the negative sample at a corruption distance from the home country between 1 and 15 are classified as low, between 16 and 30 as medium, and above 31 as high. The countries in the positive sub-sample are classified as low if the corruption distance is between 1 and 6, medium implies a distance between 7 and 12, and high above 13. A detailed overview of the categories can be found in Appendix A.
Moderator
24
The Hofstede framework uses several cultural values on a scale of 0 to 100 to present the level to which a culture masters that value. When the framework was first developed in the 1970s, it consisted of four values (power distance, uncertainty avoidance, masculinity, and individualism), after a few decades two values (indulgence and long-term orientation) were added. Several countries in the sample are not graded on these newer values. Consequently, only the four original values are used to conceptualize cultural distance. By doing so, the approaches of several previous authors are followed (e.g. Robertson & Watson, 2004; Demirbag, et al., 2009; Boateng, Wang, Wang, & Ahammad, 2017). To measure the cultural distance, the formula developed by Kogut and Singh (1988) will be used. This formula corrects the Hofstede values for the differences in the variance and averages it arithmetically. Meaning that each value contributes to the final measurement of CD equally. By using this formula, this study follows most of the previous authors in this field (Drogendijk & Slangen, 2006).
(1)
In formula 1, CDj is the cultural distance, Iij is the cultural value of the host country, Iiu is the cultural value of the United States, and Vi is the variance of cultural value (Kogut & Singh, 1988). As the formula has been developed a few decades ago, it has received some criticism. For instance, it is said that a variation-based measure should be used instead of the mean-based measure of Kogut and Singh, as firms will not face the average circumstances measured by Kogut and Singh (Beugelsdijk, Maseland, Onrust, van Hoorn & Slangen, 2015). However, according to Drogendrijk and Slangen (2006), almost all existing literature has used the Kogut and Singh formula. Moreover, the variance-based measure proposed by Beugelsdijk et al. (2015) takes inter-country cultural differences into account, which is not in line with the proposed study. Therefore, the formula as developed by Kogut and Singh will be used to measure the cultural distance.
25
medium, and above 1.6 as high. A detailed description of the categories can be found in Appendix B.
4.2
Sample
The hypotheses are tested from the perspective of the United States as the home country, and 29 host countries. Table 1 shows an overview of the host countries. The United States is used as the home country because it has a large economy and has a high number of international market entries. Furthermore, from the perspective of the United States, there are both countries with lower and with higher levels of corruption, making it a suitable reference point to test the hypotheses. The host countries are non-random selected using the criteria of variation in levels of corruption and cultural distance between the host countries. This approach increases the quality of the sample as clusters of similar countries in terms of corruption or cultural distance are prevented. Furthermore, since the aim of this study is to either confirm or to reject the proposed relationships, rather than researching the magnitude, a non-random sample is a suitable fit for this study.
The period under investigation is set as 2015-2017, by using multiple years the possibilities of variations between the number of entries due to external factors over the years is reduced. The data is retrieved using the Orbis database (Bureau van Dijk). This database contains extensive data on corporations worldwide, making it a suitable source of data for this study. A total sample of 7399 entries is compiled, of which 5659 entries into countries at a positive corruption distance, and 1740 at a negative corruption distance.
Sub-samples
To be able to test the hypothesis, a distinction is made between two samples. The first sub-sample exists out of the market entries into the countries at a negative corruption distant and is referred to as the ‘negative sub-sample’. Likewise, the second sub-sample exists out of the entries into the countries at a positive corruption distance, and is referred to as the ‘positive sub-sample’. The distinction between the two sub-samples is made due to the differences in the predicted relationships.
26
Positive sub-sample Negative sub-sample
- Denmark - New Zealand - The Netherlands - Finland - Sweden - Switzerland - Norway - Germany - Luxembourg - United Kingdom - Australia - Belgium - Austria - Russia - Mexico - China - Romania - India - Poland - Taiwan - Hungary - Malaysia - Italy - Brazil - Spain - Latvia - France - Ireland
- The United Arab Emirates
Table 1 Overview of host countries
4.3
Descriptive Statistics
Negative sub-sample
The number of entries from the home countries into the various countries of the negative sub-sample varies heavily, as can be seen in table 2. It can clearly be seen that two countries contribute more than half of the entries, while four countries contribute less than 1% of the entries each, this might suggest flaws in the dataset. In addition, the number of entries is unevenly distributed over the three categories of the categorial variable. It can be seen that, while the category ‘medium’ consists out of 70% of the entries, the other two have around 15% each. Furthermore, the number of wholly owned subsidiaries compared to the number of joint ventures varies between the
Table 2 Descriptive statistics negative sub-sample
Country Total Entries Percentage Percentage of WOS
27
countries. For example, while 95% of the entries into Spain are a WOS, this is only 17% of the entries into the United Arab Emirates. Whereas the fluctuations are less in the categories of the categorical variable, they still exist. These differences might be due to the relationship between the dependent and independent variable. However, they might also be caused by flaws in the dataset. Thus, the data needs to be tested for outliers and goodness-of-fit before the analysis is conducted. This is further discussed in §4.4 (Research Methods).
Positive sub-sample
Similar to the variations in the negative sub-sample, the positive sub-sample also shows notable differences. It can be seen from table 3 that, while 40% of the entries went into the United Kingdom, there are 7 countries with under 5% of the entries. Furthermore, the number of entries is unevenly distributed between the categories of the categorical variable, whereby a similar distribution is seen as in the negative sub-sample, whereby the middle category contains the largest percentage of entries. This might affect the results. Moreover, a variation between the percentage of wholly owned subsidiaries over the joint ventures is clearly visible. For example, 90% of the entries into the
Netherlands are a WOS, while this is only 23% in Australia. These fluctuations are less apparent when looking at the three categories. Interesting to note is that this pattern between the categories is similar in the negative sub-sample, with an increase between low and medium, and a decrease between medium and high, while high remains at a higher percentage than low. These differences might be caused by the relationship between corruption and the entry mode decision as hypothesized. However, as this might also be caused by flaws in the dataset, it needs to be tested for outliers and goodness-of-fit. This is further discussed in §4.4 (Research Methods).
Country Total Entries Percentage Percentage of WOS United Kingdom 2271 40.1% 66% Deutschland 879 15.5% 66% Luxembourg 613 10.8% 59% Netherlands 588 10.4% 90% Australia 446 7.9% 23% New Zealand 341 6.0% 42% Denmark 220 3.9% 62% Belgium 110 1.9% 42% Switzerland 80 1.4% 58% Austria 54 1.0% 61% Norway 31 0.5% 68% Finland 14 0.2% 71% Sweden 12 0.2% 42% Low 610 11% 30% Medium 4462 79% 68% High 587 10% 50%
28
4.4
Research Methods
Since the dependent variable is dichotomous, a binary logistic regression is an appropriate method to test the hypothesis. A logistic regression is a nonlinear model that forces the output to be either 0 or 1, whereby it predicts the probability of the outcome to be 1 (Torres-Reyna, n.d.). Furthermore, the logistic regressions performs a transformation of the output, and due to the nonlinearity of the model, this results in a line that can take many curves. Thus, it fits the data perfectly (Sainini, 2014). The general formula of a logistic regression is Logit= ln ( 𝑝
1−𝑝), whereby P represents the predicted probability. This study uses two formula’s, the first includes the dependent variable and the independent variable, the second also includes the moderator.
𝑝 =
1
1 + 𝑒
−𝛽1+𝛽2𝐶𝑃𝐼𝑝 =
1
1 + 𝑒
−𝛽1+𝛽2𝐶𝑃𝐼+𝛽3𝐶𝐷+𝛽4𝐶𝑃𝐼/𝐶𝐷Assumptions of the binary logistic regression
As explained in §4.3 (Descriptive Statistics), the dataset needs to be tested for goodness-of-fit and outliers due to potential flaws in the data. In addition, several assumptions of the logistic regression have to be satisfied before testing the data (Field, 2009). The first assumption is a linearity between continuous predictors and the logit of the outcome variable. This is a different linearity compared to ordinary regression, as the logistic regression predicts a nonlinear function. Since both the independent variable and the moderator are categorical, the assumption of linearity with the logit of the outcome variable is always assumed (Stoltzfus, 2011). The second assumption is no multicollinearity, which tests whether the independent variable and the moderator correlate, as this might affect the results. Using a regression between the independent variable and the moderator, VIF values of 1.000 are found (see Appendix C). This implies that no multicollinearity exists. To test for goodness-of-fit the Nagelkerke R2 reviewed. This measure provides information on the variance in the dependent variable associated with the independent variable and the
(2)
29
moderator (IBM, n.d.). A high value represents a high variability in the dependent variable (Frost, 2013). The Nagelkerke R2 varies between 0.052 and 0.159, as can be seen in table 4 to 6. This implies that the independent variable and the moderator only explain a small part of the variance in the dependent variable. This implies that the model is a poor fit for the data. However, since statistically significant predictors are found, conclusions can still be drawn Frost, 2013). Finally, to test the data for outliers, normalized residual tables are used (Scäle, 2018), this shows no abnormalities, thus, it is concluded that there are no outliers in the data set. Since the assumptions are met, and since there is no multicollinearity and no outliers, the data is suitable for the analysis.
5.
Results
5.1
Control Statistics
The purpose of the control statistics is to test whether the process is statistically under control, meaning that it is tested whether the observations are equal to a population (OECD, 2003). To determine whether the sample is representative for the wider conclusion and to test the data on the quality, several control tests are conducted. The first test compares the distribution of entries between the countries of the sample with data on outward foreign direct investments (FDI) of the USA. Even though FDI is not the same as foreign entries, it provides a frame of reference to compare the number of entries. The FDI data provides statistics on the investments made from the USA into the countries of the sample. It can be assumed that the percentage of FDI from the USA should resemble the percentages found in the sample. When comparing the total outwards FDI of the USA into the countries of the sample, a distribution of 76/24 is found. This is exactly the same as the distribution found in the sample. Therefore, the overall statistics confirm the quality of the sample. A further description on this can be found in Appendix D.
30
differences, the country factors do not significantly affect the sample. Thus, the data are assumed to be reliable. A further description on this can be found in Appendix D.
5.2
Hypothesis testing
As explained in sections 4.1 (dependent variable, independent variable, and moderator) and in section 4.4 (research methods), the sample is divided into two sub-samples. Hypothesis 1 and 2 are tested using the negative sub-sample, hypothesis 3 is tested using the positive sub-sample.
Negative Sub-sample
Hypothesis 1 predicted that the probability of a WOS increases in a host country with high corruption and that this effect is positively moderated by high or medium cultural distance. Table 4 presents the logistic regression corresponding to hypothesis 1a, 1b, and 1c. As the independent variable and the moderator are entered as a categorical variable, the presented values are the increases or decreases of the category respective to the baseline. The baseline used for both the independent variable and the moderator is the category ‘low’.
For hypothesis 1a to be true, the value should be above 1.000 and significant. As model 1 provides a value of 1.953 (P<0.01), it is
implied that the probability of a WOS is higher in a host country with high corruption than in a host country with low corruption. Thus, hypothesis 1a is supported. However, the Cox and Snell R2 and the Nagelkerke R2 are small, this implies that the model only limitedly explains the variation in the dependent variable. Thus, the hypothesis is supported, with the comment that it only affects the probability of a WOS to a small degree. Hypothesis 1b and 1c are true when the interaction terms, as presented in model 3, are above 1.000 and significant. An interaction term
Table 4 Logistic regression Hypothesis 1
Independent variable 1 2 3
H1A: CPI High 1.953*** 1.692*** 10.946***
Moderator
CD: High 0.743** 0.11***
CD: Medium 1.440*** 0.652
Interaction
H1B: CPI High X CD High X
H1C: CPI High X CD Medium 0.038***
Constant 0.525*** 0.584*** 1.826**
Model Chi-Square 68.307*** 93.044*** 220.456***
Df 2 7
Correctly classified 60.5 58 64.2
Cox and Snell R² 0.038 0.052 0.119
Nagelkerke R² 0.052 0.07 0.159
31
between high corruption distance and high cultural distance has not been found, thus hypothesis 1b is not supported. Furthermore, as the interaction between high corruption distance and medium cultural distance is below zero and significant (0.038, P<0.01), hypothesis 1c is not supported. However, due to its significance, this points towards a decreased probability. Thus medium CD negatively moderates the effect of high negative corruption on the probability of a WOS.
Hypothesis 2 predicted a higher probability of a JV when moving into a country at low negative corruption distance. Furthermore, it was proposed that this effect is reduced when the host country is either at a high cultural distance or at a medium cultural distance. The results of the logistic regression to test the hypotheses are presented in table 5. The presented values of the logistic regression are in comparison to the baseline category. The baseline for testing the independent variable is the category ‘high’, the baseline for the moderator is the category ‘low’.
For hypothesis 2a to be true, the
value in model 1 should be above 1.000 and significant. As the value is 1.953 (P<0,01), it is implied that the probability of a JV is higher in host countries at a low corruption distance compared to countries at a high corruption distance. Therefore, hypothesis 2a is supported. However, due to the low Cox and Snell R and Nagelkerke R, the model only limitedly explains the variation in the dependent variable, which implies the existence of other variables influencing the probability of a JV. Hypothesis 2b and 2c are supported when the interaction values in model 3 are significant and above 1.000. The interaction between low corruption distance and high cultural distance presents a value of 7.142 (P<0.01), whereas the interaction between a low corruption distance and medium cultural distance presents a value of 0.271 (P<0.01). Thus, hypothesis 2b is supported, while
Table 5 Logistic Regression Hypothesis 2
Independent variable 1 2 3
H2A: CPI Low 1.953*** 1.692*** 1.533
Moderator
CD: High 1.346*** 1.278*
CD: Medium 0.695** 5.667***
Interaction
H2B: CPI Low X CD High 7.142***
H2C: CPI Low X CD Medium 0.271***
Constant 0.975 1.012 0.357***
Model Chi-Square 68.307*** 93.044 220.456***
Df 2 4 7
Correctly classified 60.5 58 64.2
Cox and Snell R² 0.038 0.052 0.119
Nagelkerke R² 0.052 0.07 0.159
32
hypothesis 2c is not supported. In addition, the interaction between the low corruption distance and medium cultural distance points towards an interaction in the opposite direction as proposed. Meaning that medium cultural distance negatively moderates the effect of low negative corruption distance on the probability of a JV.
Positive sub-sample
Hypothesis 3 predicted an increased probability of a JV when entering a country at a high positive corruption distance, with a negative moderating effect of high and medium cultural distance. The results of the logistic regression to test the hypotheses are presented in table 6. The presented values of the logistic regression are in comparison to the baseline category. For both the independent variable and the moderator this is the category ‘low.
For hypothesis 3a to be supported, the value in model 1 should
be above 1.000 and significant. However, as the value is 0.474 (P<0.01), the hypothesis is not supported. Instead, this implies a decreased probability of a JV. Due to the dichotomous nature of the independent variable, this implies an increased probability of a WOS in a host environment with high positive corruption distance. For hypothesis 3b and 3c to be supported, the interaction values of model 3 need to be above 1.000 and significant. Since both values are not significant, these hypotheses are not supported.
5.3
Robustness test
To validate the results, a robustness test with five categories for both the independent variable and the moderator is conducted. The division of the countries among the categories can be found in Appendix E. The assumption of the robustness test is that the outcomes of the main analysis do
Table 6 Logistic Regression Hypothesis 3
Independent variable 1 2 3
H3A: CPI High 0.474*** 0.473*** 0.421***
Moderator
CD: High 0.516*** 0.423***
CD: Medium 0.739*** 0.194***
Interaction
H3B: CPI High X CD High 1.082
H3C: CPI High X CD Medium 1.493
Constant 0.99 2.733*** 3.288***
Model Chi-Square 361.199***408.435***435.251***
Df 2 4 8
Correctly classified 66.4 67.4 67.6
Cox and Snell R² 0.062 0.07 0.074
Nagelkerke R² 0.084 0.095 0.101
33
not change when a small adaptation is made. Thus, the hypothesis supported should also be supported in the robustness test.
Table 7 presents the outcomes of the robustness test for hypotheses 1a, 1b, and 1c. Since the value of model 1 is above 1.000 and significant, it confirms hypothesis 1a. No interactions have been found between high CPI and high and medium cultural distance. Thus, the robustness test confirms hypothesis 1a and rejects hypothesis 1b and 1c, this is in accordance with the main analysis.
Table 8 presents the outcomes of the robustness test for hypothesis 2a, 2b, and 2c. Model 1 shows a significant value (P>0,01) for low CPI, as hypothesized in 2a, thus providing support for the hypothesis. From model 3 it can be seen that the interaction between low CPI and high cultural distance is significant and above 1.000, while no interaction was found between low CPI and medium cultural distance. Therefore, support is found for hypothesis 2b, whereas no support is found for hypothesis 2c. These results are in accordance with the main results.
Table 7 Robustness test Hypothesis 1
Table 8 Robustness test Hypothesis 2
Independent variable 1 2 3
H1A: CPI High 4.995*** 9.116*** 18.648***
Moderator
CD: High 0.573*** 0.109***
CD: Medium 1.113 0.003***
Interaction
H1B: CPI High X CD High X
H1C: CPI High X CD Medium X
Constant 0.482*** 0.710** 1.826**
Model Chi-Square 147.979 233.521 332.123
Df 4 8 13
Correctly classified 62.9 62.1 65.8
Cox and Snell R² 0.082 0.121 0.175
Nagelkerke R² 0.109 0.162 0.234 *P<0.1 **P<0.05 ***P<0.01 WOS is predicted Independent variable 1 2 3 CPI: Low 4.995*** 9.116*** 2.683*** Moderator CD: High 1.745*** 1.324** CD: Medium 0.899 43.670*** Interaction
CPI: Low X CD High 6.951***
CPI: Low X CD Medium X
Constant 0.416*** 0.154*** 0.204***
Model Chi-Square 147.979 223.521 332.123
Df 4 8 13
Correctly classified 62.9 62.1 65.8
Cox and Snell R² 0.082 0.121 0.175
Nagelkerke R² 0.109 0.162 0.234
34
Table 9 presents the robustness tests of hypothesis 3a, 3b, and 3c. The value of high CPI in model 1 is not significant, thus, the robustness test does not find support for hypothesis 3a. Furthermore, no interactions were found between high CPI and high and medium cultural distance. Thus, the robustness tests do also not support hypothesis 3b and 3c. The lack of support for the robustness test is similar to the results of the main analysis. However, the main analysis found a significant value for hypothesis 3a, which pointed towards a decrease, instead of the expected increase.
6. Discussion
6.1
Conclusion and implications
Despite extensive efforts to fight corruption globally, it is still present in all economies. Due to the significant differences in the level of corruptions between countries, it is realistic to consider various strategies of multinational enterprises when entering countries with different corruption levels. Therefore, this paper aimed to find significant evidence for a relationship between the corruption distance and the entry mode decision of MNEs. In order to predict the direction of the proposed relationship, the transaction cost theory is used. This theory suggests that a firm will choose to carry out a transaction in the market or in the hierarchy, depending on which leads to the lowest transaction costs. Based on the transaction cost theory and existing evidence of previous scholars (e.g. Duanmu, 2011; Tekin-Koru, 2016), it was expected that an MNE chooses an inhouse strategy in market entries in countries with high corruption. Therefore, by following the existing literature, this study has suggested a tendency towards a WOS when entering a country with high
Table 9 Robustness test Hypothesis 3
Independent variable 1 2 3 CPI: High 0.92 0.467*** 0.456*** Moderator CD: High 0.491*** 0.458*** CD: Medium 0.674*** 0.289** Interaction
CPI: High X CD High X
CPI: High X CD Medium X
Constant 1.076 2.831*** 3.039***
Model Chi-Square 397.896 450.817 453.448
Df 4 8 10
Correctly classified 66.4 67.6 67.6
Cox and Snell R² 0.068 0.077 0.077
Nagelkerke R² 0.092 0.104 0.105
35
negative corruption, and a tendency towards a JV when entering a country at a low negative or high positive corruption distance.
Furthermore, based on differences in corruption perception between cultural differs countries, this study proposed a moderating influence of cultural distance on the relationship between corruption and entry strategies. When considering the transaction cost theory, it might be expected that an increase in cultural distance leads to an increase of transaction costs, and thus, towards a tendency of carrying out the transaction inhouse. Previous scholars found both a direct effect of cultural distance on corruption (e.g. Robertson & Watson, 2004; Jha & Panda, 2017) and a direct effect of cultural distance on the entry mode decision (e.g. Hennart & Larimo, 1998; Kogut & Singh, 1988). Therefore, this study proposed moderating effects of cultural distance on the relationship between corruption and the probability of a JV or a WOS.
To test the hypothesized relationships, around 7000 recent market entries are reviewed. Since this study proposed different effects depending on the corruption distance being either positive or negative, two conceptual models are created. Consequently, two different ways of building the dataset are used, and a division is made between a negative sub-sample and a positive sub-sample. The dependent variable is defined as a dummy variable, whereby one entry resembles the market mode of the transaction cost theory, and the other resembles the hierarchy mode of the theory. Moreover, the independent variable and the moderator variable are both categorical variables with three categories. In addition, a robustness test using five categories is conducted to validate the results. The robustness test confirms the main results.
36
increased (Duanmu, 2011). However, the results are in contradiction with Karhunen & Ledyaeva (2012), who concluded that a JV is more likely to be the entry mode when corruption increases, and in contrast with Demirbag et al. (2009), who found an increase of JVs in higher corrupt host countries. Consequently, by providing support for the proposition that high corruption distance increases the probability of a WOS, this study has a major contribution in this research field. Future authors should further investigate the relationship between corruption and entry mode decisions, and take the results of this study into consideration.
This study has been one of the first to consider cultural distance as the moderator in the relationship between corruption and entry mode decision. It can be concluded that high cultural distance positively moderates the effect of low negative corruption distance on the probability of a JV, while medium cultural distance negatively moderates in this relationship. In addition, it is concluded that medium cultural distance also negatively moderates the effect of high negative corruption distance on the probability of a WOS. However, it can also be concluded that high and medium cultural distance does not moderate the effect of high positive corruption distance on the probability of a WOS, and high cultural distance does not moderate the effect of high negative corruption distance on the probability of a JV. These conclusions have a major implication for existing and future research. As there are confirmed moderating effects, future research has to take cultural distance as the moderator into consideration.
6.2
Implications
37
multiple factors, influence the probability. Thus, policy makers and academics should not only consider corruption and cultural distance when working with entry modes. Despite these variables having an effect, other variables are present and should be taken into consideration. Finally, as no support was found for the hypothesis that high positive corruption distance increases the probability of a JV, and this study being one of the first to research this relationship, there is a major implication when considering the entry mode when moving into an environment with lower corruption. This implies that either the transaction cost theory is not capable of predicting this relationship, or that other factors exist that heavily influence the relationship.
6.3 Limitations and suggestions for further research
This study has several limitations which have to be taken into consideration when applying the results of this study. First, the study was limited by time and money, this enforced a major limitation on the sourcing of the data. Despite using the reliable Orbis database, the possibilities of checking the data were limited. A better data source would have been national statistical databanks of the countries included in the sample. Since finding these sources, checking them, comparing, and adjusting them, would have taken more time than available, this was not possible. As a consequence, the Orbis database has been used as the next best alternative, mostly since it records the data of all countries, in the same way, meaning that no alternations had to be made. Furthermore, as a result of the constraints in time, this study considers the cultural distance as one variable, whereas previous scholars have often used the cultural values as individual variables. As this study aimed to find a moderating effect of cultural distance, this approach is the most suitable approach. However, considering the results, and the direct contrast with previous scholars, it might not be neglected that there is a causation between the approach and the results. Moreover, the constraints in time forced this study to focus on a single country as the home country. As a result, the possibilities to generalize this study to a wider population are limited. By using a single country, home country factors might have affected the results, these factors would have been reduced if more countries had been used as the home country.
38
Considering the results showing contradictions with the TCE in three out of the nine hypotheses, the TCE is acknowledged as a limiting factor in this study. However, the purpose and aim of this study are to find a relationship between corruption and entry mode, and the moderating effect of cultural distance. It does not aim to test or to confirm/reject the TCE. The theory is used to predict a direction of the effect of both the independent variable and the moderator, mostly considering the contradicting results of previous studies. Therefore, despite being a limitation, the TCE is regarded as an appropriate theory to predict the proposed relationships in this study. Future research should either use different theories to substitute the TCE or use the TCE exclusively for prediction purposes. The TCE should not be used to support or reject the proposed relationship.
Third, this study researched the effects of corruption when moving from a higher corrupt country, towards a lower corrupt country, whereby the United States was defined as the home country. The United States ranks number 18 of 176 on the corruption index (Transparency International, 2017), meaning that it is generally considered as a country with low corruption. Thus, the corruption differences between the United States and the host country are relatively low. The results in the positive sub-sample have to be approached with caution. Due to this limitation, the results might not be representative of the situation of a market entry from country ranking high in the CPI index into a country ranking low on the index. Future research should focus on this relationship, in order to expand on the current research domain.
Fourth, it is commonly known that several countries are considered as so-called ‘tax havens’, where the taxation is beneficial for companies (Oxfam International, 2017). This implies that companies might use these countries to coordinate their international expansion. From the observations, it becomes visible that this has happened in the sample used in this study. For example, countries known as tax havens, such as Luxembourg, the Netherlands, Ireland, and Switzerland, have larger percentages of observations than several countries with a larger economy, such as France and Spain. This might have affected the data and the results.
39
the moderating effect of cultural distance on the relationship between corruption and entry mode. This study found significant results for 3 out of 6 of the hypothesis, consequently strong conclusions to confirm or reject the proposition cannot be drawn. Therefore, further research should expand on this exploratory work.
Sixth, this study conceptualized the entry mode decision as either joint venture or wholly owned subsidiary. However, many other entry modes exist, with varying levels of ownership and strategy choices. Despite the binominal nature of the dependent variable fitting the research question, corruption might affect the strategic decision making of MNE’s in a different perspective. Therefore, future research should focus on the effects of corruption on non-equity modes of entry, and on the different strategies of equity modes. For example, mergers, acquisitions, greenfield investments and brown field investments.
