1
MSc Thesis International Economics and Business
Institutional distance and the costs of doing business abroad:
A value-added activity contingency perspective.
University of Groningen Faculty of Economics and Business
Name: R.B. (Bob) Castelein
Student number: S1884913
Student e-mail: r.b.castelein@student.rug.nl
Supervisor: Dr. A.A.J. Van Hoorn
Co-assessor: Dr. P. Rao Sahib
Version: Final
Date: January 5, 2015
2
ABSTRACT
This thesis presents a novel theoretical framework outlining the heterogeneous effects of institutional distance on multinationals’ foreign subsidiaries, differentiated by the characteristics of value-added activities performed at these subsidiaries, thereby refining understanding of institutional distance effects and addressing inconsistent findings and paradoxes in the extent literature. Hypotheses are tested on plant-level data of MNE subsidiaries in 445 unique home-host-country dyads. The diversity in the data allows to control for host country-, home country- and industry-level variation, and effectively isolate institutional distance effects. The results show that while institutional distance in general impacts performance negatively, there is significant heterogeneity among institutional distance effects when made contingent on the different activities performed at MNE subsidiaries.
3
CONTENTS
1. Introduction 4
2. Background 8
2.1 Distance, institutions, and the costs of doing business abroad 8
2.2 MNEs and internationalization 12
2.3 Subsidiary roles 14
3. Theoretical framework and hypotheses development 17
3.1 General framework 17
3.2 The subsidiary’s task environment 21
3.3 Hypotheses development 24
4. Data and method 28
4.1. Data 28 4.1.1. Data sources 28 4.1.2. Operationalization 28 4.2. Method 37 5. Results 40 5.1. Descriptive statistics 41
5.2. Study 1: Institutional distance effects on subsidiary performance 42 5.3. Study 2: Institutional distance effects on activity allocation 44 5.4. Study 3: The moderating role of subsidiary value-
added activities on performance effects of institutional distance 46
5.5. Robustness checks and extensions 50
5.5.1. Introducing alternative distance measures 50
5.5.2. Measurement issues of the key dependent variable 51
5.5.3. Subsidiary activity versus industry contingency 54
5.6. Discussion 59
5.6.1. Discussion of results 59
5.6.2. Limitations 62
5.6.3. Recommendations for future research 64
6. Conclusion 66
References 67
4
1. INTRODUCTION
While the concept of “distance” has much intuitive appeal (Ghemawat 2011; Zaheer et al. 2012) and the gravity equation of international trade is one of the most robust empirical models in economics (Anderson 2010), how exactly distance (broadly construed) affects multinational enterprises (MNEs) remains unclear. A large literature relates different types of distance—geographic, cultural, and institutional—to such variables as entry mode choice and subsidiary performance but generally comes up with mixed findings. More worrisome, different authors have pointed out paradoxes plaguing the literature on (particularly institutional) distance and multinational subsidiaries (Brouthers & Brouthers 2001; Drogendijk & Zander 2010; López-Duarte 2013; Harzing 2003) as well as a general tendency towards trivial effect sizes (Kirkman et al. 2006; Tihanyi et al. 2005; Zhao et al. 2004).
5
effects of institutional distance to a degree that is cannot be achieved using less varied data. Moreover, the plant-level observations allow for measurement at the level at which institutional distance effects can be expected to materialize, namely at the subsidiary level.
Through these contributions, two major theoretical and empirical issues in extant research are addressed. First, this thesis addresses the issue that institutional distance is often too broadly conceptualized to be useful. Institutional distance has been used to explain a variety of outcomes, including those related to foreign direct investment (FDI), such as market selection, entry mode choice, and subsidiary performance (see Bae & Salomon 2010; Hutzschenreuter et al. 2015 for comprehensive reviews). The theoretical link between institutional distance and economic outcomes was established from a transaction-cost perspective with Hymer’s (1960/1976) notion of the “costs of doing business abroad” (the additional costs incurred by a business when operating in a foreign (institutional) environment). This was later refined by Zaheer (1995) who emphasized the “liability of foreignness,” consisting of the hazards arising from institutional distance, as the social costs of doing business abroad, and the most critical type of costs for MNEs (Eden & Miller 2004). The problem with this framework, despite its apparent theoretical clarity, is that it is perhaps too general to adequately account for the outcomes it should explain. For example, when confronted with a high degree of institutional difference between the home and host country environments, a firm can either reduce its risk exposure in an unfamiliar environment by engaging in a joint venture with a local partner, or instead opt for maximum control in order to avoid having to rely on a partner with whom cooperation and communication can be difficult (O’Grady & Lane 1996; Brouthers & Brouthers 2001; Eden & Miller 2004; López-Duarte 2013). The fact that opposite outcomes can be predicted by the same framework yields an intriguing paradox and stresses the need for more accurate conceptualization, which this thesis aims to provide.
6
discussions of institutional distance without the distinction between the two being made (Eden & Miller 2004; Ioanecu et al. 2004; Van Hoorn & Maseland, forthcoming). On the empirical level this produces considerable validity issues when the variables used do not actually measure what researchers intend to measure (Van Hoorn & Maseland 2014). Many studies use institutional profile characteristics to construct their measures of distance (Ioanscu et al. 2004; Bae & Salomon 2010; Slangen & Beugelsdijk 2010; Perkins 2013), yielding confusing results and leaving the question as to the effect of institutional distance as such, separate from other environmental factors, open. Other studies use data on firms from one home country only (Kogut & Singh 1988; Tihanyi et al. 2005; Slagen & Beugelsdijk 2010; Beugelsdijk et al. 2015; Hutzschenreuter et al. 2015), limiting the extent of variation in the data, and making the distance measure strongly correlated with host country institutional profiles (Van Hoorn & Maseland, forthcoming). The large amount of diversity in the data used in this analysis allows for better isolation of the effect of institutional distance by controlling for home country-, host country-, and industry-specific factors at the same time.
7
Dikova 2009; Slangen & Beugelsdijk 2010; Dikova & Rao Sahib 2013), but none of these employ a contingency perspective entirely at the level of the subsidiary.
Taking a “friction lens” to this much-needed contingency perspective, this thesis focuses on the effects of different dimensions of institutional distance and institutional distance hazards at the subsidiary level, differentiated by the type of value-added activities being performed. Theories of institutional distance describe processes at the level of interactions, for example misunderstandings between parties from different institutional contexts and legitimacy issues affecting foreign firms when they are dependent on local stakeholders. Therefore, a more fine-grained understanding of the effects of institutional distance should therefore also focus on the level of interactions, which this study does by differentiating subsidiaries by the value-added activities they perform and the characteristics of the interactions with host country actors these activities entail. The recognition of subsidiary heterogeneity is introduced by drawing from the literature on subsidiary roles (Bartlett & Ghoshal 1989; Gupta & Govindarajan 1991; Rugman et al. 2011), which classifies subsidiaries by their different “roles” within the MNE. This thesis presents a novel perspective by considering institutional distance effects as contingent on the characteristics of subsidiary activities vis-à-vis the host country institutional environment. This allows for a refinement of the concept of institutional distance with the notion of institutional friction as the context-specific ways in which institutional differences may become salient in different degrees, with the subsidiary and its activities as the appropriate locus of institutional friction.
8
2. BACKGROUND
This chapter serves as a background sketch by outlining the three main bodies of literature on which the theoretical framework of this thesis is built, namely the institutional distance and costs of doing business abroad frameworks, theories underpinning internationalization of firms, and theories on subsidiary roles. The most relevant aspects are highlighted, as well as some interlinkages. The next chapter (3) synthesizes and refines these frameworks and presents several testable hypotheses.
2.1. Distance, institutions, and the costs of doing business abroad
Multinational enterprises (MNEs) are firms that “control and manage [plants] located in at least two countries,” (Caves 2007: 1) and are therefore by definition companies that operate across different business environments. MNEs internationalize to exploit the firm-specific advantages that underlie their competiveness, and to access location-specific resources in foreign locations (Dunning & Lundan 2008; Burnstein & Monge-Naranjo 2009; Hennart 2009). Despite this competitive advantage and the potential usefulness of complementary local resources, MNEs typically lack knowledge and familiarity in foreign environments. Moreover, they have to manage tensions arising from being active in multiple different environments at once. In other words, they have to manage distance within their firm boundaries (Zaheer et al. 2012).
9
(Hutzschenreuter et al. 2015). Lastly, the concept of “institutional distance” was developed in the 1990s, based on institutional theory and the New Institutional Economics (DiMaggio & Powell 1983; North 1991; Scott 1995). Institutional distance encompasses cultural distance, and is also broader in that it covers differences between countries in all formal and informal “rules of the game” (North 1991) that govern behavior. Despite the consensus over these definitions, clear conceptualization and operationalization of institutional distance in relation to economic outcomes has proven to be problematic in practice, reflected by paradoxes and ambiguous findings as described above (Bae & Salomon 2010; Hutzschenreuter et al. 2015). However, the clear delineation and comprehensiveness of the frameworks underlying the construct makes institutional distance a useful starting point of refining the conceptualization of distance effects.
Institutions are defined as the “humanly devised [formal and informal] constraints that structure political, economic and social interaction,” (North 1991: 97) or – simply stated – “the rules of the game.” Scott (1995) identifies three main institutional “pillars,” or systems that provide abovementioned constraints or structuration to human interaction, namely the regulatory, normative, and cultural-cognitive (or simply cognitive) institutional pillars. Institutions along the regulatory pillar have the “capacity to establish rules, inspect others’ conformity to them, and, as necessary, manipulate sanctions – rewards or punishments – in an attempt to influence future behavior” (Scott 1995: 52). This process can be formal or informal, but is usually seen as the ways in which formally designated actors and mechanisms enforce formally established laws and rules. The normative pillar pertains to the prevalent norms and values in a society. These “define goals or objectives […] and designate appropriate ways to pursue them” (Scott 1995: 55) in terms of what states of affairs are desirable and what types of behavior are appropriate. The most tacit dimension of institutional environments is the cognitive pillar, defined as the “schemas, frames, and inferential sets which people use when selecting and interpreting information [or] the cognitive structures and social knowledge shared by [people]” (Kostova 1997: 180). Eden & Miller (2004) classify cultural distance as a sub-category of institutional distance, chiefly driven by differences along the normative and cognitive institutional pillars.
10
desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions” (Suchman 1995: 574). In order to be effective in a given environment, an individual or organization has to be perceived as legitimate by the other actors on which it depends in various interactions and exchanges, and must therefore play by the prevailing “rules of the game” (Kostova & Zaheer 1999). Over time, this creates a tendency of increasing homogeneity (or isomorphism) within an institutional field, as actors’ “characteristics are modified in the direction of increasing compatibility with environmental characteristics” (DiMaggio & Powell 1983: 149). Relating this notion to the case of the multinational enterprise, it should first be recognized that an MNE, despite its transnational activities, is shaped by the institutional environment of the country from which it originates – the home country. The practices, underlying the MNEs competitive advantage, that the firm brings along to other (host) countries, are hence heavily influenced by the home country institutional environment (Bloom & Van Reenen 2010). Secondly, when a company sets up operations abroad, the institutional environment there is likely to differ from the one at home. The greater these differences are, the greater the tension between the way the company usually operates and the expectations of actors in the foreign institutional environment. This makes it challenging for the MNE to achieve legitimacy and institutional “fit,” in the host country environment, which also has implications for its effectiveness in this environment.
11
The LOF resulting from home-host country institutional distance creates several, potentially costly hazards for MNEs setting up a subsidiary in a host country. Eden and Miller (2004, distilled from Zaheer (2002)) list these as follows: first, unfamiliarity hazards arise from the lack of experience in and knowledge of the local context, which becomes more severe as the dissimilarity between the home and host environments increases. Secondly, discrimination hazards arise when a foreign firm is perceived and treated differently than local firms, usually in a negative sense. This can for example stem from consumer ethnocentricity or discriminatory treatment by government actors. Thirdly, relational hazards arise from the uncertainty inherent in interactions with local actors whose behavior, motivations, and predispositions are significantly different from those prevalent within the MNE and are - as a result - generally poorly understood. This consists of inter-organizational hazards, in dealing with other organizations in the MNEs local buyer-supplier network, and intra-organizational hazards, in managing operations in which there is significant diversity in institutional backgrounds of people. With (prospective) local employees, customers, and suppliers, a foreign company will have less legitimacy due to its lack of embeddedness. Moreover, there will be a degree of inherent uncertainty in interactions between actors who do not fully understand one another. Internally, this can have implications for the transfer of knowledge and practices between the MNE and its subsidiary (Kostova & Roth 2002). These obstacles to the functioning of the subsidiary are likely to vary in severity with the degree of dissimilarity between the subsidiary’s home country environment and the host country environment. In their 2004 paper, Eden and Miller do not distinguish clearly between hazards arising from institutional distance and hazards arising from the quality of the host country institutional environment (Van Hoorn & Maseland 2015). For conceptual clarity, the hazards discussed above will henceforth be discussed as institutional distance hazards, stemming from the degree of dissimilarity between the home and host country environments – the focus of this analysis. When hazards stemming from the quality of the host country institutions themselves are discussed (such as contract and property right enforcement, rule of law, corruption, level of trust etc.), this will be explicitly addressed and referred to as institutional hazards.
12
firm-specific advantages to the foreign market. Moreover, the firm likely has to rely on these firm-specific advantages in order to compensate for the relative loss of efficiency due to its LOF. This imposes limits on the degree to which the firm can mimic local – host country – practices to increase its local legitimacy (DiMaggio & Powell 1983). Internally, this means that the differences between the institutional background of the firm’s practices and the institutional environment in which they are being implemented (and from which it likely recruits employees) have to be bridged for the transfer of practice to be successful. This dilemma between transferring strategic resources and practices, and adapting to the local environment has prompted scholars (Hennart 1982; Prahalad & Doz 1987; Bartlett & Ghoshal 1989) to visualize this tradeoff as an “integration-responsiveness matrix” with conflicting pressures for global integration and local responsiveness.
2.2. MNEs and internationalization
To elucidate the relationship between institutional distance hazards, subsidiary characteristics, and economic outcomes, one must address why companies internationalize and how they undertake this process.
13
A comprehensive framework concerning the internationalization decisions of multinationals is Dunning’s OLI (or eclectic) framework (Dunning & Lundan 2008). From this perspective, MNEs select their target market and the appropriate entry mode by considering and weighing Ownership, Location, and Internalization advantages. If all three advantages are sufficiently present, the firm in question should engage in foreign direct investment (FDI) and set up a foreign subsidiary. Ownership advantages entail that the stronger the strategic resources and capabilities owned by the MNE, the more likely it will seek to exploit these firm-specific advantages beyond its home market. Secondly, there should be considerations that make the MNE prefer to internalize the foreign activity, rather than outsource it. As discussed, these include a weighing of the costs of coordination vis-à-vis the costs and risks involved with arm’s-length procurement. Lastly, the target country should have local advantages that attract MNEs to this country, such as the presence of resources (human or natural) of superior quality or lower price (for example lower wages) than in other locations. Between the three factors, the location advantages of the target market should make the difference for the firm to choose an FDI strategy over exporting. In other words, there should be country-specific advantages that incite an MNE to set up a subsidiary (Hennart 2009).
14
headquarters and the subsidiary involved, and the flows of goods, resources, and knowledge within the MNE.
These considerations of market selection and entry mode take place in the context of the global strategy that the MNE is pursuing – for example simple exporting, a dispersed multidomestic strategy with little central coordination, or a global strategy with high degree of coordination of dispersed activities (Porter 1986; Oh & Rugman 2012). This MNE strategy in turn is driven by – among other factors – the characteristics of the industry the firm is active in and the sources of the firm’s competitive advantage (Porter 1980; 1986). Of particular relevance are the necessary degree of coordination of activities by the firm’s headquarters and the degree of geographical dispersion of these activities, both of which drive the relationship between the MNE and its subsidiaries and the operational characteristics of the subsidiaries.
Other research emphasizes that FDI is never a one-directional process (the MNE simply taking a local resource, or simply producing for the local market), but an organizational form in which firm-specific advantages of the MNE are strategically bundled together with local country-specific assets (Hennart 2009; Rugman et al. 2011). Liu et al. (2011), when considering offshore outsourcing decisions, show that MNE decisions reflect consideration of both the characteristics of the services activity (in terms of routineness, interactiveness, and complexity of the tasks involved) and the institutional characteristics of the target country. Since the relative resource commitment of FDI is greater than for outsourcing, the same considerations will apply when MNEs set up foreign subsidiaries. The ultimate FDI decision is driven by a careful consideration of the firm’s strategy and resources, as well as the local resources and business environment. Hence when discussing MNEs’ foreign activities, the host country environment should be considered as well.
2.3. Subsidiary roles
15
“Headquarter intensity” is a measure of an industry’s reliance on knowledge-intensive intangible services (Antras & Helpman 2004) to generate value, as opposed to “component intensity,” which is a measure of an industry’s reliance on production services to generate value. The most common measures of reliance on headquarter services are the skill intensity, research and development (R&D) intensity, and marketing intensity of a certain industry (Helpman 1984). Transaction cost theory predicts that firms in industries with a high headquarter intensity will be more inclined to internalize these types of activities, as the production and transfer of knowledge-intensive intangible services is characterized more by hold-up risks (incomplete contracts, idiosyncratic investments, etc. (Williamson 1979)) than production and transfer of tangible products (Gorodnichenko et al. 2015). Even though headquarter intensity as an industry measure is less useful when considering subsidiary-level characteristics, it does provide a useful framework to think about these types of activities and their transaction cost considerations, as they can be used to characterize activities of MNE subsidiaries themselves. Ibarra-Caton (2015: 3) argues that “if a firm can generate competitive advantage through cooperative headquarter-subsidiary relations in these high value services, the advantage is likely to improve the cost efficiency of their foreign operations.” The perspective is clearly at the firm, rather than the subsidiary level, but the key point is that subsidiary performance in these activities depends on how well the MNE can transfer its practices in these knowledge-intensive, intangible services to the subsidiary.
In the literature on “subsidiary roles”, authors address the fact that there is often considerable variation in the types of activities that are performed at subsidiaries of the same MNE. This entails complex within-firm networks and resource flows that simple descriptions in terms of horizontal or vertical integration fail to capture. Over time, several frameworks have emerged to classify subsidiaries by their roles, each emphasizing different characteristics by which subsidiaries can be differentiated.
16
black holes. Rugman et al. (2011) raise the criticism that this typology is rather generic, and disregards the increasing complexity of MNEs as differentiated networks with fragmented value chain activities across different locations. They expand Bartlett and Ghoshal’s subsidiary roles framework by further differentiating subsidiaries by the type of value chain activity they perform. Simplifying Porter’s (1985) typology of value chain activities, they broadly classify these activities as innovation, production, sales, and administrative support. They reason that each of these activities requires a different bundling of firm-specific competencies and local assets, and hence have different implications for the subsidiary’s role within the MNE. Their framework yields sixteen role/activity combinations, specifying the activities being performed, as well as the importance of local resources, subsidiary capabilities, and the subsidiary’s importance for the MNE.
To develop their own typology of subsidiary roles, Gupta and Govindarajan (1991; 2000) reason that within an MNE there are capital, product, and knowledge flows (based on Williamson 1985). The formal and informal structures of governance relations and coordination mechanisms within the firm – as well as the firm’s global strategy – determine how these resources flow between subsidiaries and headquarters (Chandler 1962; Bartlett & Ghoshal 1989). Depending on the direction and magnitude of resource flows, subsidiaries have different roles within the firm. Gupta and Govindarajan (1991; 2000) emphasize that in most MNEs this results in a complex within-company network of capital, product, and knowledge flows. Of these different flows, they consider knowledge flows to be the most critical to the MNEs competitive advantage. When an MNE first decides to set up a subsidiary, they expect to achieve a certain “intended knowledge flow” between the subsidiary and the rest of the firm for their strategic purposes, but the performance of the subsidiary depends on the extent to which these intended knowledge flows are realized. The effectiveness of knowledge transfers directly impacts the performance of subsidiaries that are dependent on these flows, and any impediments in the transfer of knowledge result in suboptimal performance.
17
3. THEORETICAL FRAMEWORK AND HYPOTHESES DEVELOPMENT In this section, first a general comprehensive framework is presented, integrating the three streams of research outlined above. Subsequently, a more specific framework is presented, essentially zooming in on the key part of the general framework, regarding the subsidiary’s task environment. From this more focused framework, several testable hypotheses are distilled.
3.1. General framework
The tradition in the institutional distance literature so far has been to theorize on or measure the effect of institutional distance on certain economic outcomes straightforwardly, without taking into account moderating factors (Hutzschenreuter et al. 2015). This perspective has left the costs of doing business abroad defined too broadly, as a general disadvantage relative to local firms in the host country, driven primarily by institutional distance and a consequent lack of local legitimacy and embeddedness (Zaheer 1995; Eden & Miller 2004). However, heterogeneity in subsidiaries and foreign activities is overlooked. This heterogeneity is addressed specifically in the “subsidiary roles” frameworks, but from this perspective the interaction between the subsidiary and the host country environment is usually overlooked. A common criticism of the OLI-framework and internationalization theory is that, while they form an important underpinning for location choice and subsidiary roles considerations, they overlook the importance of institutions (Nölke & Taylor 2010).
18
consider the effect of institutional distance hazards as contingent on the interactions in which institutional differences actually produce problems.
Also the notion that institutional distance should drive a competitive disadvantage relative to local firms – the common definition of LOF – is problematic. In terms of legitimacy and embeddedness in the local institutional environment, this “disadvantage” perspective makes sense. On the other hand, the competitiveness of the subsidiary relative to local firms may or may not be relevant to the MNE (for example, in case of a fully vertically integrated subsidiary, its competitive position in the host country is less relevant). As mentioned, there are many viable MNE subsidiaries in institutionally distant markets. A common explanation is that these firms can effectively leverage firm-specific advantages or employ an isomorphic strategy to overcome the LOF (Zaheer 1995), but again this includes the uncomfortable assumption that LOF is an equally relevant and equally problematic factor for al firms operating across a given institutional distance.
19
Figure 1. Foreign direct investments, subsidiary roles, and institutional distance effects – a comprehensive framework.
20
When recognizing the heterogeneity of subsidiary roles, one must conclude that institutional distance and LOF do not necessarily produce a fixed level of costs, disadvantages, or uncertainty for MNEs. Just as subsidiaries differ considerably, so should the effects of (institutional) distance on these subsidiaries. Institutional distance and LOF should rather be seen as drivers of potential costs, disadvantage, or uncertainty. The extent to which institutional differences become salient and institutional distance hazards materialize depends on the context of subsidiary operations. Therefore the concept of “institutional friction” is introduced, as the extent to which given institutional differences actually hinder operations of the subsidiary. Institutional distance and associated hazards are one factor contributing to institutional friction, others factors being the characteristics of the subsidiary involved and its interactions with the local environment. These factors interact to yield a level of friction that depends both on institutional differences and the context in which these become salient at the subsidiary level. Assuming that the institutional distance between two countries is given, this generates uncertainty and risks for the MNE in the host country environment, summarized as the four institutional distance hazards of LOF. The nature of the subsidiary’s operations determines the extent to which the subsidiary is exposed to these institutional distance hazards, and the extent to which these hazards may be detrimental to performance.
21
is more exposed to intra-organizational hazards than inter-organizational hazards. On the other hand, an expatriate-staffed subsidiary using locally procured materials to produce for local corporate clients is more exposed to inter-organizational hazards than intra-organizational hazards. Lastly, the subsidiary’s exposure to discriminatory hazards depends on the degree to which it needs to reach local (potentially ethnocentric) consumers or is otherwise dependent on local stakeholders, notably local government officials.
When confronted with institutional distance between their home country and a (prospective) host country, MNEs are not simply passively affected by the problems institutional differences may produce. Based on the hazards the firm perceives or already experiences, it can strategically adjust the characteristics of its subsidiary to mitigate the exposure to these hazards and the potential institutional friction (Luo et al. 2002). This can for example be done by adjusting its ownership strategy, increasing local responsiveness strategies, adjusting the way the subsidiary is managed, or adjusting the mix of activities taking place at the subsidiary. Formally, in its behavioral response to these hazards, the MNE will minimize the sum of production and transaction costs incurred through the subsidiary, considering tradeoffs between accessing host country-specific assets and institutional hazard exposure if necessary. Even if the subsidiary has already been established and managers perceive subsidiary performance to be suffering from exposure to certain hazards, the firm can still make adjustments in its strategy with regard to the subsidiary, feeding back into changes in the subsidiary role. This is visualized in Figure 1 in the feedback loop between CDBA, performance, strategic adjustment, and subsidiary roles.
In summation, the effect of institutional distance on MNE subsidiaries is not straightforward, but moderated by the characteristics of the subsidiary and its activities that makes it more or less susceptible to institutional friction stemming from one or more institutional distance hazards.
3.2. The subsidiary’s task environment
22
Figure 2. The subsidiary task environment.
23
While institutional distance generates certain potential institutional distance hazards, actual hazard exposure and the extent to which these hinder operations (institutional friction) depend on a complex interplay of factors that constitute the subsidiary’s task environment. The effects of institutional distance on the subsidiary’s hazard exposure are moderated by the characteristics of this task environment, made up of the activities being performed by the subsidiary, the relevant stakeholders with whom interaction is required in these activities, and the transactional characteristics of the activities in terms of interactiveness and complexity.
As can be seen in Figure 2 above, the degrees of exposure to different types of institutional distance hazards are partially driven by the subsidiary’s interaction with different types of local stakeholders. Table 1 below visualizes this perspective in terms of the institutional distance hazards that stem from interactions with local stakeholders, and the activities that are most likely to expose subsidiaries to these hazards.
Table 1. Institutional distance hazards arising from interactions with different host country stakeholders in exemplary value-added activities.
Local stakeholder Hazard Exemplary value-added activities
Employees Intra-organizational All relevant subsidiary activities
Local partner firm Intra-organizational All relevant subsidiary activities
Other firms Inter-organizational Procurement, business-to-business sales and service,
logistics (if involving arm’s length transactions), marketing
Consumers Discrimination Sales and service
Government Discrimination Procurement (utilities, licenses etc.),
business-to-government sales
General public Discrimination Al relevant subsidiary activities
The relevant local stakeholders determine potential institutional distance hazard exposure, depending on the type of activities being undertaken at the subsidiary, but ultimately the extent of hazard exposure and institutional friction – at a given level of institutional distance – is driven by the transactional characteristics of internal and external interactions.
24
location based on firm-specific advantages and location-specific advantages, but adjust the allocation of subsidiary activities based on (potential) institutional hazards affecting the operations of the subsidiary. Of key interest is the extent to which the allocation of activities to a certain location – the behavioral response – is affected by institutional distance (apart from industry- or location-specific factors). Secondly, the performance of subsidiaries should be considered. As discussed above, it is not always useful to measure subsidiaries’ competitive position relative to local firms, as local competitors may or may not be relevant for the firm. Rather, drawing on Gupta and Govindarajan’s (1991) notion of intended knowledge flows versus realized knowledge flows, one should consider the extent to which the subsidiary operates at the desired level of efficiency. While firms should – in theory – function optimally as a result of continuous adjustments at the margin towards optimal efficiency, Leibenstein (1966) suggests that persistent deviations from optimal efficiency (“X-inefficiency”) within firms can be explained by incomplete labor contracts, limited knowledge of the production function (bounded rationality), and market imperfections in the market for inputs. This notion ties in well with the transaction costs considerations outlined above. The greater the institutional distance, the greater the uncertainty in interactions with local stakeholders and the greater the risk of incomplete contracts. Similarly, foreign firms that lack knowledge of and embeddedness in the local environment are likely to see their operational efficiency suffer more from market imperfections in the host country.
3.3. Hypotheses development
Based on the framework outlined above in Figure 2, several expectations can be formulated. First a general hypothesis of the effect of institutional distance: one would expect institutional friction to increase when institutional distance increases, simply because there are more and/or greater differences that can translate into friction.
Hypothesis 1: If the home-host institutional distance increases, institutional friction increases (and performance decreases) – all else being equal.
25
Figure 1). In this process, allocation of an activity will be more sensitive to institutional distance when the hazard exposure of the activity is greater. Accordingly, subsidiaries in institutionally distant markets will likely to a lesser extent perform activities that particularly expose the subsidiary to institutional distance hazards. Secondly, it can be expected that subsidiaries performing activities with a greater associated institutional distance hazard exposure will see their performance deteriorate more as a result of institutional distance than subsidiaries with a relatively limited hazard exposure. These expected relationships are visualized in Figure 3 below.
Hypothesis 2: The effect of home-host country institutional distance on institutional friction increases (becomes more positive) when the subsidiary’s exposure to institutional distance hazards increases – all else being equal.
Hypothesis 2a: The more a certain activity exposes the subsidiary to institutional distance hazards, the more the subsidiary’s reliance on these activities decreases as institutional distance increases.
Hypothesis 2b: The more a certain activity exposes the subsidiary to institutional distance hazards, the more the subsidiary’s performance decreases as institutional distance increases.
Figure 3. Visualization of Hypotheses 2, 2a, and 2b.
Now several more specific hypotheses can be formulated, based on Hypothesis 2. As discussed, one should disaggregate the subsidiary’s task environment into its constituent
26
elements, namely the activities involved, the stakeholder interactions this entails, and the degrees of external interactiveness and internal and external complexity of the subsidiary’s task environment – all of which drive the subsidiary’s hazard exposure and institutional friction. First, when the subsidiary’s interactiveness with local stakeholders increases, its exposure to discrimination and inter-organizational hazards increases, at a given level of institutional distance. This can be expected to increase the institutional friction experienced by the subsidiary, as outlined in Hypothesis 3a. Second, when the complexity of the task environment increases, there is more uncertainty inherent in the activities being performed by the subsidiary, aggravating any pre-existing potential for institutional friction due to a given level of institutional distance. Hence increased complexity can be expected to increase the effect of institutional distance on institutional friction (Hypothesis 3b). Complexity can stem from two factors, namely the complexity of internal transactions and the complexity of external transactions. External complexity is determined by the characteristics of the subsidiary’s transactions with local external stakeholders, specifically the risks of incomplete contracts. The smaller the part of the transaction that can be explicitly codified and covered by contracts, the greater the reliance has to be on mutual trust, communication, and tacit understanding (Aghion et al. 2013). The greater the home-host institutional distance, the more difficult it is to deal with contract incompleteness in these ways, and hence greater institutional friction can be expected. The same argument can be made for the complexity of intra-firm transactions with local partners and employees. Moreover, internal complexity also stems from the subsidiary’s reliance on tacit knowledge-intensive activities and/or intra-MNE knowledge flows (Gupta & Govindarajan 2001; Antras & Helpman 2003; Liu et al. 2011). Transfer of tacit knowledge and practices becomes more difficult as institutional distance (particularly along the cognitive and normative pillars) increases between the parties involved in the transfer (Kostova & Roth 2002), generating a disincentive to perform these activities in institutionally distant locations, and reduced performance of subsidiaries performing these activities across greater home-host institutional distances. Based on these considerations, the following hypotheses follow from the general Hypothsis 2:
Hypothesis 3 The effect of home-host country institutional distance on institutional friction increases (becomes more positive) when – all else being equal: Hypothesis 3a - the external stakeholder interactiveness of the subsidiary’s activities
increases.
27
Hypothesis 3c - when the subsidiary’s reliance on tacit-knowledge intensive activities and/or intra-MNE knowledge flows increases.
Hypothesis 3d - when the transactions with external stakeholders are characterized by more incomplete contracts.
Generally, the more complex a transaction or task is, the greater the risks of contract incompleteness (Williamson 1979). This applies to any task involving any stakeholder, whether tacit-knowledge intensive tasks performed by employees, or transactions with other firms or consumers with a high degree of asset specificity. As the uncertainty stemming from contract incompleteness is inherent to the transaction, uncertainty stemming from regulatory differences will not add significantly to the overall uncertainty. On the other hand, the tacit understanding, effective communication, and mutual trust required for these transactions are likely more affected by differences along the normative and cognitive pillars. Therefore, one expects that more complex transactions are affected relatively strongly by normative and cognitive differences:
Hypothesis 4: When the complexity of subsidiary activities and transactions with internal and external stakeholders increase, the effects of institutional distance will be driven more by cognitive and normative differences, relative to regulatory differences – all else being equal.
28
4. DATA AND METHOD
4.1. Data
4.1.1. Data sources
The sample used to test the hypotheses consists of 3006 MNE subsidiaries in 34 different host countries, with parent companies from 44 different home countries (see Appendix A for a complete list), making up 445 unique home-host country dyads. Three waves – 2002, 2004, and 2005 – of the Business Environment and Enterprise Performance Survey (EBRD & World Bank 2015) are compiled into a cross-sectional dataset of MNE subsidiaries. In the BEEPS surveys, questionnaire responses from a large number of firms – predominantly in Central and Eastern Europe and the Former Soviet Union – are recorded on questions relating to the business environment and the functioning of companies. Per country, the administrators of the survey aim to accurately reflect the industry composition of the country in the composition of firms surveyed, and to have at least 10% of the firms surveyed be foreign-owned (Fries et al. 2003). Although every wave is slightly different from the others in the exact composition and phrasing of the questions, the three waves selected are analogous on the questions relevant for this study. Essential factual information, such as home country, industry, ownership, and nationality of foreign owners is recorded similarly for all firms in the surveys.
The data on countries’ scores on Hofstede’s cultural dimensions is taken from Hofstede’s own books and website (2001; 2015; Hofstede et al. 2010). The various other institutional distance measures are taken from the CEPII “GeoDist” (Mayer & Zignago 2011) and “Language” (Melitz & Toubal 2012) datasets, and the “The Quality of Government” dataset by LaPorta et al. (1999).
4.1.2. Operationalization
Dependent variables:
29
intensity” measures) with different levels of interactiveness and complexity will be used. The activities with their approximated transactional characteristics are listed in Table 2.
Table 2. Value-added activities with their transactional characteristics.
Value-added activity Interactiveness Internal complexity External complexity
Host country sales High Low Unknown
Host country procurement High Low Unknown
Research and development Low High Low
Marketing Moderate High Moderate
High-skilled labor intensive operations Low High Low
Low-skilled labor intensive operations Low Low Low
The subsidiary’s reliance on these activities is operationalized as follows:
- Host country sales is obtained from the 2002 survey using question q14a1, and from
the 2004 and 2005 surveys using question q7a (“What percentage of your firm’s sales are sold domestically?”).
- Host country procurement is obtained from the 2002 survey using question q24a and
from the 2004 and 2005 surveys using q15a (“What percent of your establishment’s material inputs and supplies are purchases from domestic sources?”)
- R&D intensity is obtained from the 2002 survey using q83: “On average since 1998
how much has your company spent on the following expressed as a per cent of the average annual sales of your firm over the same period?,” with the option “research and development” (q83b). For the 2004 and 2005 waves the R&D intensity is calculated somewhat differently using q58b (“Could you please tell me how much your firm [is projected to] spend in 2004 on each of the following”) with the option “research and development.” Dividing total sales (recorded for q57a: “[…] could you tell me the [projected] estimate of your firm’s total sales [to the end of 2004]?”) by expenditure on R&D yields the R&D intensity – an equivalent measure to the one in the 2002 survey.
- Marketing intensity is obtained similarly to R&D intensity using q83 from the 2002
survey, recording spending on “advertising and marketing” (q83c). For the 2004 and 2005 data the marketing intensity is also calculated using q58, with regard to spending on “advertising and marketing” (q58c), and the total sales (q57a).
- High-skilled labor intensity (or high-skill intensity) is obtained using question q92
30
are:” including the categories “managers” (q92a) “professionals” (q92b) and “skilled workers” (q92c). Together these three categories make up the share of skilled employees in the subsidiary’s personnel. The percentages recorded for these three questions are summed to obtain the measure for skill intensity of the subsidiary. From the 2004 and 2005 surveys, with nearly identical wording of the question, the percentages recorded for q68a1, q68a2, and q68a3 are used.
- Low-skilled labor intensity (or low-skill intensity) is obtained from the 2002 survey
using q92d and from the 2004 and 2005 surveys using q68a4 (“What percent of your current permanent, full time workers are […] unskilled workers?”).
MNE subsidiary performance – operational efficiency. In research on the effect of institutional distance on MNE subsidiary performance different measurements have been used to capture performance (Hutzschenreuter et al. 2015; López-Duarte et al. 2015). The most used measures at the subsidiary level included affiliate sales, profitability, survival, and managerial perceptions of affiliate effectiveness. In this study, subsidiary performance is operationalized using the latter type of measurement. Managerial satisfaction with subsidiary performance (Dikova 2009) or perceptions of the magnitude of problems affecting the subsidiary (Vachani 2005) have an advantage over financial performance measures (sales, profits etc.) in that performance is assessed independently from transfer pricing and accounting practices. These factors can severely distort other measurements of firm performance, especially when there is a high degree of vertical integration and a high share of intra-firm trade between the subsidiary and the MNE’s other subsidiaries or headquarters.
Using the BEEPS, managerial evaluation of performance is operationalized using questions q90a (2002 survey) and q65a (2004 and 2005): “In your judgement, what is your firm’s current output in comparison with the maximum output possible using its facilities/man power […]?” Responses to this question (given in % of reported capacity utilization) reflect how efficiently the subsidiary operates. If the firm encounters any institutional friction from internal or external hazards, this should be reflected in sub-optimal efficiency levels achieved by the subsidiary.
31
reported by managers, that increases with the severity of the friction. Likewise, the smaller the friction is in subsidiary operations, the closer the capacity utilization will be to 100%.
Due to the subjective nature of the variable and the associated reliability issues, two robustness checks will be conducted, one using different sub-samples and one using an alternative measure of performance.
Independent variables:
Regulatory institutional distance. Often measures of institutional quality are used to measure regulatory distance (Ioanscu et al. 2004; Bae & Salomon 2010; Slangen & Beugelsdijk 2010; Salomon & Wu 2012; Perkins 2013). The problem with these measures is that, while they do measure the degree of uncertainty stemming from regulatory factors, they do not measure regulatory distance: differences between home and host countries are not meaningful when only considering the quality of the host country institutional environment. Any foreign firm will reap the benefits of a high-quality regulatory, political, and legal institutional environment (rule of law, economic freedom, impartial justice system etc.), even if its home country has a less favorable institutional environment. Eden & Miller (2004) argue that in the case of corruption (or what they call “corruption distance”) firms from countries with higher levels of corruption know better how to deal with corruption than firms from countries with lower levels of corruption. While this may seem a reasonable conjecture, their idea of “corruption distance” is problematic. Low-quality regulatory institutions are usually in themselves characterized by a large degree of uncertainty – potentially obscuring any separate effect that institutional distance along the regulatory pillar may have on uncertainty. Moreover, they also usually entail a large degree of arbitrariness as to which firms are affected the most and how.
32
home and host country dummies are used to capture unobserved country-level variation (for example regarding institutional quality), this variable should help isolate the effect of regulatory distance.
Normative institutional distance. Normative distance is operationalized using Kogut and Singh’s (1988) index, based on Geert Hofstede’s (2001) original four dimensions of national cultures:1
- Power Distance Index (PDI): the extent to which it is accepted in a society that power is distribute unequally among people. A high PDI indicates a higher tolerance of inequality, whereas a low PDI characterizes a society with more egalitarian inclinations.
- Individualism versus Collectivism (IDV): A higher IDV index indicates a higher degree of individualism in a society.
- Masculinity versus Femininity (MAS): The higher a society scores on MAS, the stronger its preference for, and emphasis on “masculine” values such as achievement, competition, and assertiveness, whereas a low score indicates a stronger preference for “feminine” values such as cooperation, care, and consensus decision-making.
- Uncertainty Avoidance Index (UAI): UAI indicates the extent to which a society tolerates ambiguity and uncertainty, with a higher score indicating more pragmatism, and a lower score indicating rigidity and orthodoxy.
Although Hofstede calls these “dimensions of culture,” several authors stress the point that this characterizes predominantly normative aspects of culture, in that the dimensions reflect norms and values in a society in terms of what states of affairs are desirable over others (Scott 1995; Lucas 2006). Recognizing that culture is made up of normative as well as cognitive (how people select, organize and interpret information, and assign meaning) dimensions, Hofstede’s cultural framework captures normative, rather than cognitive differences. Despite its shortcomings, the Hofstede framework does provide a comprehensive set of normative indicators, making it a suitable measurement tool (Drogendijk & Slangen 2006).
1
33
Kogut and Singh (1988) combine these four dimensions into a composite index using the following general formula:
(𝟏) 𝐼𝑁𝐷𝐸𝑋𝑗𝑏 =1 4∑ { (𝐼𝑖𝑗− 𝐼𝑖𝑏)2 𝑉𝑖 } 4 𝑖=1
Where the index of normative distance between countries j and b (in Kogut and Singh’s 1988 article this base country b is simply the US) is calculated using both countries’ index score I on the ith Hofstede dimension, and total variation Von every index Ii. For every dimension,
the countries’ distance on this dimension is squared (to achieve consistently positive values, regardless of which country is considered the reference country) and divided by the overall variance on the dimension index. These values are subsequently summed and divided by four to obtain a normative institutional distance index.
Cognitive institutional distance. So far, relatively few studies have attempted to measure cognitive distance (Bae & Salomon 2010; Hutzschenreuter et al. 2015), most of which have used custom surveys of people within firms (Kostova 1997; Kostova & Roth 2002). Ioanscu et al. (2004) used differences in educational attainment levels and computer and internet usage to approximate cognitive distance. The problem with this latter measurement is analogous to that of most regulatory distance measures in that it measures host country qualities rather than distance. A well-educated workforce and well-developed communications infrastructure are general advantages to companies, but do not reflect differences in the “schemas, frames, and inferential sets which people use when selecting and interpreting information [or] the cognitive structures and social knowledge shared by [people]” (Kostova 1997: 180) that constitute cognitive distance.
34
language.” Their variable indicates the probability that two random people from the two countries can understand each other in (any) shared language (that is spoken by at least 4% of the population). Generally, the greater this probability is, the easier it is to bridge the general cognitive distance between the home and host country or to find ways to achieve this. It should be noted that, when specified this way, this is a measure of proximity rather than distance. For the purpose of this study, it is transformed into a measure of linguistic distance by subtracting the probability of mutual understanding from 1, hence obtaining a measure of the probability that two people from the countries in the dyad do not understand one another. As this probability approaches one, linguistic distance increases, constituting a measure of cognitive distance.
As one of the hypotheses (4) entails a comparison of the effect sizes of the different types of institutional distance, these institutional distance variables will be standardized (mean=0, standard deviation=1) before analysis - with the exception of the indicator variable for regulatory distance.
Control variables
35
Plant size. For several theoretical considerations, it is important to consider the size of the subsidiary. For example, when a subsidiary is large, the probability is greater that it will undertake some ‘headquarter services’ or business support services itself. Therefore, the size of the subsidiary is controlled for, using responses on the BEEPS question “How many full-time employees work for this company [today]?” (s4a in all waves). The responses are coded on an ordinal scale, ranging from 1 (2-10 people) to 7 (1000-9,999 people). Even though this is not as informative as the exact number of employees, it does provide a suitable proxy of plant size.
Foreign ownership share: Typically, foreign ownership share has been used as a dependent variable in institutional distance research, and a considerable amount of research has been conducted on how entry mode and ownership strategy are affected by institutional distance (see Eden & Miller (2004) for a theoretical framework; Hutzschenreuter et al. (2015) for a literature review). Especially when considering the effect of institutional distance on activity allocation, there should be controlled for how the involvement of a foreign partner (reflected by the share of the firm not owned by the foreign multinational) may affect the allocation of activities. Furthermore, when considering the effects of institutional distance on subsidiary performance being moderated by the subsidiary’s activities, there should be controlled for the uncertainty and intra-organizational institutional hazards potentially arising from involvement of a local partner (Eden & Miller 2004). The foreign ownership share is measured using the responses on questions s4c1 (2002 survey) and s5a (2004 and 2005 surveys) from the BEEPS: “What percentage of your firm is owned by […] private foreign company/organization?”
36
which can be conceived of as being home or host country- or industry-specific. Other studies, using data on MNEs from only one home country (usually the United States, see for example Kogut and Singh 1988; Slangen & Beugelsdijk 2010; Gorodnichenko et al. 2015) lack the degrees of freedom in their samples to include country dummies to capture between-country variation, as these would correlate perfectly with the distance measures between the host countries and the United States. With the great diversity of home and host countries in the BEEPS data, country dummies can be used to capture unobserved between-country variation. Similarly, industry dummies are included to capture unobserved differences between industries.
Additional variables for robustness checks
Alternative distance measures. To verify whether the measures used adequately isolate the effects of institutional distance as separate from other forms of distance, a robustness check is conducted using alternative distance measures. Drawing from the GeoDist dataset (Meyer & Zignago 2011) these alternative measures include dummy variables indicating a shared border and colonial ties within a dyad, and the geographic distance (in 1000km) between the country capitals in a dyad.
37
4.2. Method
Hypothesis 1, regarding the general effect of institutional distance on subsidiary performance
is tested using the following model:
(𝟐) 𝐶𝐴𝑃𝑈𝑇𝑖𝑗 = 𝛽1+ 𝛽2𝑅𝐸𝐺𝐷𝐼𝑆𝑇𝑗+ 𝛽3𝑁𝑂𝑅𝑀𝐷𝐼𝑆𝑇𝑗+ 𝛽4𝐶𝑂𝐺𝐷𝐼𝑆𝑇𝑗+ 𝛽5𝐸𝑋𝑃𝐸𝑅𝑖 + 𝛽6𝑆𝐼𝑍𝐸𝑖+ 𝛽7𝑂𝑊𝑁𝑖+ 𝛿𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌 + 𝛾𝐻𝑂𝑀𝐸 + 𝜑𝐻𝑂𝑆𝑇 + 𝜀𝑖𝑗
Where CAPUTij (capacity utilization) is the measure of operational efficiency of subsidiary i
from home-host country dyad j, β1 is a constant, and REGDIST, NORMDIST, and COGDIST
are regulatory, normative, and cognitive institutional distance, respectively, as outlined above.
EXPER, SIZE, and OWN are firm-level control variables controlling for host country
experience, plant size, and the ownership share of the multinational, respectively. INDUSTRY,
HOME and HOST are vectors of industry, home country, and host country dummy variables,
controlling for unobserved variation at these levels, in order to isolate the effect of institutional distance. ε is an error term.
Secondly, the hypothesis regarding activity allocation (2a) is tested using regression models along the following specification:
(𝟑) 𝐼𝑁𝑇𝐸𝑁𝑆𝐼𝑇𝑌𝑎𝑖𝑗 = 𝛽1+ 𝛽2𝑅𝐸𝐺𝐷𝐼𝑆𝑇𝑗+ 𝛽3𝑁𝑂𝑅𝑀𝐷𝐼𝑆𝑇𝑗+ 𝛽4𝐶𝑂𝐺𝐷𝐼𝑆𝑇𝑗+ 𝛽5𝐸𝑋𝑃𝐸𝑅𝑖 + 𝛽6𝑆𝐼𝑍𝐸𝑖+ 𝛽7𝑂𝑊𝑁𝑖+ 𝛿𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌 + 𝛾𝐻𝑂𝑀𝐸 + 𝜑𝐻𝑂𝑆𝑇 + 𝜀𝑖𝑗
Where INTENSITYaij is the intensity of activity a performed at subsidiary i from home-host
country dyad j, for the six types of subsidiary activity outlined in Table 2. The other variables are the same explanatory and control variables as used in equation (2). Again, ε is an error term. As outlined in the hypotheses in chapter 3, one would expect the effects of institutional distance to be stronger for the activities with higher degrees of interactiveness and complexity. Moreover, for these activities the effects of cognitive and normative distance are expected to be stronger relative to regulatory distance.
38
(𝟒) 𝐶𝐴𝑃𝑈𝑇𝑖𝑗 = 𝛽1+ 𝛽2𝐼𝑁𝑇𝐸𝑁𝑆𝐼𝑇𝑌𝑎𝑖𝑗 + 𝛽3𝑅𝐸𝐺𝐷𝐼𝑆𝑇𝑗+ 𝛽4𝑁𝑂𝑅𝑀𝐷𝐼𝑆𝑇𝑗+ 𝛽5𝐶𝑂𝐺𝐷𝐼𝑆𝑇𝑗 + 𝛽6(𝐼𝑁𝑇𝐸𝑁𝑆𝐼𝑇𝑌 ∗ 𝑅𝐸𝐺𝐷𝐼𝑆𝑇) + 𝛽7(𝐼𝑁𝑇𝐸𝑁𝑆𝐼𝑇𝑌 ∗ 𝑁𝑂𝑅𝑀𝐷𝐼𝑆𝑇)
+ 𝛽8(𝐼𝑁𝑇𝐸𝑁𝑆𝐼𝑇𝑌 ∗ 𝐶𝑂𝐺𝐷𝐼𝑆𝑇) + 𝛽9𝐸𝑋𝑃𝐸𝑅𝑖 + 𝛽10𝑆𝐼𝑍𝐸𝑖 + 𝛽11𝑂𝑊𝑁𝑖
+ 𝛿𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌 + 𝛾𝐻𝑂𝑀𝐸 + 𝜑𝐻𝑂𝑆𝑇 + 𝜀𝑖𝑗
Equation (4) will be estimated separately for every activity to isolate the activity-contingent effect of institutional distance. In this equation, β1 is a constant, CAPUTij (capacity utilization)
is the reported operational efficiency of subsidiary i from home-host dyad j, INTENSITYaij is
the subsidiary’s reliance on each of the six different activities outlined in Table 2. The regulatory distance variables, firm level controls, and the industry, home country, and host country dummies are the same as in equations (2) and (3) above. Also included are three interaction terms that make the effect of institutional distance on subsidiary performance contingent on the type of value-added activities being performed at the subsidiary. ε is an error term.
A key assumption underlying the standard ordinary least squares regression method is that the dependent variable is (approximately) normally distributed (Hill et al. 2012). When assessing the distribution of the dependent variables (six measures of subsidiary’s reliance on certain value-added activities as per Table 2, and the capacity utilization of the subsidiary), their histograms (Appendix 2) show highly skewed distributions, with the greatest concentration of observations at an extreme end of the distribution – in this case at zero or one, as the variables are fractions. Both a skewness-kurtosis test and a Shapiro-Wilk test lead to a rejection of the null hypothesis of normality (p<0.01) for all dependent variables. Unfortunately, any variable transformation would still retain the greatest concentration of observations at the high and low ends of the distribution. Four types of transformations were attempted (natural logarithm, square root, logarithm with base 10, logarithm with base 2 – see Cleveland (1984)) but in all cases the skewness-kurtosis test and the Shapiro-Wilk test indicated a significantly non-normal distribution.
39
40
5. RESULTS
5.1. Descriptive statistics
Before estimating the models as outlined above, some descriptive measures of the variables are presented. Table 3 shows the descriptive statistics of the variables used in the analyses in their raw (untransformed) form. The measures for normative distance (the Kogut-Singh Index) and cognitive distance (linguistic distance) are standardized (mean=0, standard deviation=1) before analysis to facilitate comparison of effect sizes. The variables regarding activity intensity and capacity utilization are kept untransformed, as their scales (from zero to one) are already the same, and more meaningful in their present form.
Table 3. Summary statistics (before transformation).
Observations Mean Std. Dev. Min Max
Host Country Sales 3005 0.712 0.366 0 1
Host Country Procurement 2920 0.494 0.400 0 1
R&D Intensity 1616 0.011 0.041 0 0.7 Marketing Intensity 1801 0.013 0.037 0 0.5 High-Skill Intensity 2961 0.754 0.255 0 1 Low-Skill Intensity 2961 0.131 0.202 0 1 CapacityUtilization 2828 0.712 0.307 0.01 1 Legal Distance 3006 0.778 0.416 0 1
Kogut Singh Index 2280 2.291 1.739 0 12.489
Linguistic Distance 2856 0.740 0.267 0.007 1 Shared Border 3005 0.221 0.415 0 1 Colonial Ties 3005 0.159 0.365 0 1 Geographic Distance (1000km) 3005 2.587 2.754 0.060 13.159 Plant Size 3006 3.139 1.706 1 7 Experience 3006 14.444 17.882 3 202 Foreign Ownership 3006 77.096 26.919 1 100 Price-Cost Margin 2474 24.404 14.776 0 120
41
Furthermore, Table 4 shows the correlation matrix of the variables that are used in the regressions:
