On embeddedness and subsidiary performance

(1)

1

On embeddedness and

subsidiary performance

Name: Ioana Ruxandra Bedreaga

Student No. 2552590

Email address: ioana.bedreaga@yahoo.com

Supervisor: Dr. Raquel Ortega-Argiles

Co-assessor: Prof. Dr. Steven Brakman

Date: 7-07-2014

(2)

2

List of tables

Tables to be inserted in the text

Table 1: Overview of literature review ... 65

Table 2: Distribution of subsidiaries across home and host countries ... 66

Table 3: Construction of the variables ... 67

Table 4: Descriptive statistics and correlation matrix... 68

Table 5: Estimation results of embeddedness proxies on performance ... 69

Table 6: Geographic distance non-linear effect and firm-specific moderating effect ... 71

Table 7: Moderating effect of firm size and experience on institutional distance ... 72

Table 8: Moderating effect of firm size and experience on cultural distance ... 73

Table 9: Moderating effect of firm size and experience on geographic distance ... 74

Table 10: Moderating effect of experience on institutional distance ... 75

Table 11: Moderating effect of experience on cultural distance ... 76

Table 12: Moderating effect of experience on geographic distance ... 77

Table 13: Estimation results using Driscoll-Kraay standard errors ... 78

Tables in the Appendix Table A: Sample decomposition by home-host country pairs ... 79

Table B: Sample decomposition by industry ... 84

Table C: Eurostat indicators of high-tech and knowledge-intensive industries ... 85

List of figures

Figure 1: Overview of conceptual framework ... 13

Figure 2: Framework of testable hypotheses ... 25

(4)

4

Abstract

The emergence of the multinational and the relocation of the corporate activities in foreign countries have made the differences in national social, institutional and cultural systems particularly relevant in the business exchange. As firms expand beyond domestic markets, they encounter different regulations, norms and values that constrain their organizational behaviour and challenge their ability to gain legitimacy in foreign markets. This thesis builds upon the embeddedness paradigm existing in economic sociology to investigate the effect of the external environment in which subsidiaries are operating on their performance. The main hypothesis is that foreign affiliates tend to perform better when their chances to adapt to the local idiosyncrasies and embed in the host environments are higher. The regression results obtained from the analysis of a panel data of 10825 subsidiaries world-wide offered empirical support to the theorized hypotheses.

(5)

5

1. Introduction

Many scholars have tried to understand the development of internationalization patterns, particularly in terms of foreign direct investment as they assume great additional costs, financial risk and strategic impact (Hymer, 1976). Theoretical mechanisms mainly imply a positive relationship between internationalization strategies and performance based on economies of scale and scope, (Ghoshal, 1987; Kogut 1985), location-specific advantages (Dunning, 1980), new opportunities for learning and value creation (Barkema and Vermeulen, 1998) as well as spread of risks and costs over larger markets (Contractor et al, 2003). However, at an empirical level, the cost-benefit trade-off of internationalization has yielded mixed and often contradictory results (Bausch and Krist, 2007), suggesting that the relationship between internationalization and performance is not universalistic, but instead might be moderated by contextual factors that can produce differential performance effects.

In general, operational performance has been theorized to depend on the interplay between the internal resources that are under the “direct, real time control of the manager” and the environmental variables that “are not subject to direct or positive real time control” (Luthans and Stewart, 1978:686-7) and thus, “the best way to organize depends on the nature of the environment to which the organization relates” (Scott 1992:89). In line with this argument, this thesis contends that internationalization is a successful undertaking to the extent to which firms are able to adjust their operational behaviour to the exogenous characteristics of the foreign environments in which they operate. The concept of “embeddedness” existent in the economic sociology literature is particularly relevant for this purpose. Generically, embeddedness entails that organizational behaviour cannot be understood independent of the social and institutional contexts surrounding the firm. Thus, this thesis suggests that embeddedness acts as a mechanism to obtain strategic fit in various settings and consequently, organizational performance is dependent on the opportunities of the firms to embed in the socio-political environments of the host countries. As each individual host country carries different potential for embedding their foreign affiliates, it is expected that organizational performance also varies depending on the idiosyncrasies of the local environments in which firms operate.

(6)

6 essence, the concept has captured the integration of the economic action into the broader social and institutional environment, yet up to this point its applications have been flexible and spanned across multiple levels of analysis depending on the purpose of the study. Theoretical frameworks have ranged from social, cultural and institutional embeddedness (Zukin and DiMaggio, 1990; Granovetter, 1985), to geographic, spatial and temporal forms of embeddedness (Halinen and Tornroos, 1998; Henderson et al, 2002) and multiple other variations exist along the embeddedness spectrum depending on the academic subfields in which the concept has proliferated (Krippner and Alvarez, 2007). Moreover, the phenomena of internationalization and the emergence of the multinational - defined as a company which “engages in foreign direct investment (. . .) and owns or controls value-adding activities in more than one country” (Dunning, 1993:3), turned “embeddedness” into a “spatial concept” (Hess, 2004:165) with different points of analysis that added to the theoretical incongruities. For instance, economic geography scholars have used “embeddedness” to investigate how foreign direct investment affects local development through additional employment and technological spillovers (e.g. Phelps et al, 2003), while in the business literature the focus has been on the multinational and the relocation of the corporate activities in different national contexts (Heidenreich, 2012).

(7)

7 theoretical framework that touches upon different aspects of embeddedness simultaneously. This framework is subsequently employed as an analytical tool to answer the question “Which exogenous country-specific factors affect subsidiary performance?”. The factors identified would therefore be useful in explaining the contextual nature of performance within internationalization processes. The effects of these factors on performance have been validated using a panel data analysis of 10825 subsidiaries world-wide, thereby hoping to make a valuable contribution to the empirical work on embeddedness and performance. In addition, the mediating role of subsidiaries’ resources on embeddedness has also been tested, suggesting that firms can mitigate the influence of external factors on performance and set the pace for embeddedness, however the results have shown that this is a rather long-term process that requires significant investment.

This paper is divided as follows: the second section gives an overview of the existing theoretical literature and empirical work on embeddedness. Section three details the conceptual framework and the testable hypotheses developed to answer the research question. Section four summarizes the data selection methodology and empirical specification, and section five continues with the interpretation of the empirical results. Finally, section six presents some concluding remarks and section seven offers some suggestions for future research.

2. Literature review

2.1.Theoretical development of “embeddedness”

(8)

(9)

9 studies (Turok, 1993; Phelps et al, 2003; Phelps and Fuller, 2000). Embeddedness is assumed to be beneficial both for the host regions – as local firms access new technological knowledge, create additional employment and promote economic growth – and for the investors, as multinationals gain access to knowledge of local market conditions, as well as to new and complementary resources that would enhance their overall performance (Henderson et al, 2002). In this sense, network embeddedness can be a key factor in regional economic growth and in capturing global opportunities (Henderson et al, 2002). However, the extent to which organizations can exploit these benefits depends on their internal organizational trajectory or what Halinen and Tornroos (1998) defined as “temporal embeddedness”: “companies are bound to past, present and future modes of time […] they have their own histories during which they have evolved; they are also in the midst of their own present, and have objectives and expectations about the future which affect their present decisions and actions” (p. 195).

(10)

10

2.2.Empirical work on “embeddedness”

Intuitively, embeddedness in its various forms has been assumed to affect organizational performance. However, the empirical work on embeddedness has been rather scarce. Partly responsible for this has been the “confusing variety of meanings” (Hess, 2004:176) that has prevented researchers from operationalizing the concept and bridging it to experimental – be it qualitative or quantitative - conditions. Nevertheless, some empirical evidence exists, although most of the work has been built upon insights from the embeddedness paradigm rather than explicitly operationalize the concept of embeddedness on its own and in addition, all the studies have focused solely on one dimension of embeddedness.

Uzzi (1997) was the first to empirically study and operationalize embeddedness, although his work was restricted to the relational dimension of the concept. Based on his field research, Uzzi (1997) identified three main aspects that underlined the quality of structural embeddedness: trust, fine-grained information transfer and joint problem-solving arrangements that in conjunction promoted economies of time, integrative agreements, allocative efficiency and complex adaptation. While Uzzi (1997) specifically emphasized the benefits of embeddedness on economic outcomes, he also acknowledged the risks of “overembeddedness”, underlying the lock-in effects of embeddedness that can frustrate performance “if the social aspects of exchange supersede the economic imperatives” (p.59). Subsequent studies also highlighted the costs of embeddedness which might arise from over-reliance on a limited number of business partners, persistence of redundant ties and continuous exchange of homogenous information (Gargiulo and Benassi, 2000; Masciarelli et al., 2009; Rowley, 2000).

(11)

11 by their relationships with suppliers, customers and competitors, and explicitly excluded larger institutional structures and non-business partners from their analysis. In a comparable manner, Johannisson et al (2002) used a survey-based methodology to operationalize embeddedness in terms of an actor’s relationships both with related business partners and with more general non-economic, institutional and political constituents and proved its “leverage for business creation”. The authors particularly stressed the benefits of institutional elements as “without institutions the embeddedness of the business community is incomplete since a considerable number of business persons remain disconnected, i.e. they are not directly related” (p. 304).

Hardy et al (2005) employed a case-study approach in their evaluation of embeddedness, yet unlike previous studies that concentrated specifically on firm network embeddedness, the authors shifted the emphasis on cultural and political embeddedness and evaluated how dissimilar political and cultural contexts convey different choices of strategic fit. In a similar manner, without referring specifically to the concept of embeddedness, other researchers have considered the impact of cultural or institutional factors on business strategy – particularly choice of entry modes - and more rarely on organizational performance (e.g. de Jong and van Houten, 2014; Halkos and Tzeremes, 2008; Antia et al, 2007; Hutzschenreuter and Voll, 2008; Ionascu et al, 2004; Hennart and Larimo, 1998; Delios and Beamish, 1999; Meyer, 2001; Meyer et al, 2009).

(12)

12 customers, but their findings have been rather inconsistent (e.g. Javorcik, 2004; Barrios et al, 2011; Crespo and Fontoura, 2007; Giroud et al, 2012).

Table 1 summarizes the existing theoretical literature on embeddedness along with examples of relevant empirical work.

[INSERT TABLE 1 ABOUT HERE]

3. Conceptual framework and hypotheses development 3.1. Conceptual framework

The discussion in the previous section has made it evident that embeddedness remains a fuzzy concept (Oinas, 1997; Markusen, 1999), with different meanings in different contexts and capturing its complexity within a single approach is rather challenging. Consequently, the degree to which the empirical studies enumerated above captured the actual potential of the term is debatable. By focusing on single dimensions of embeddedness, these empirical studies made its operationalization imprecise, given the multi-dimensionality of the concept: “Embeddedness refers to the social, cultural, political, and cognitive structuration of decisions in economic contexts. It points to the indissoluble connection of the actor with his or her social surrounding” (Beckert, 2003:769).

(13)

13 existing literature, this would be defined by cognitive and cultural characteristics that enable individual actors to find common symbolic understandings of their actions and provide them with the mechanisms to cooperate and coordinate in economic transactions (Saka-Helmhout and Geppert, 2011). Therefore, this level of analysis looks at the factors that assist the formation of inter-personal relationships across individuals operating in the business environment and that ultimately facilitate the alignment of their economic interests, i.e. the degree of relational

embeddedness. To summarize, a three-stage framework as developed in this section (see Figure 1)

would provide a consistent and systematic tool to explore the multifaceted aspects of embeddedness and would reconcile the various theorizations currently existing in the literature, thus ensuring theoretical consensus and empirical validity.

Figure 1: Overview of conceptual framework

3.2.Hypotheses development

Some researchers have decried the static character of the existing theses on embeddedness and suggested that insights from the agency theory are essential in conceptualizing its dynamic nature (Heidenreich, 2012; Hess, 2004). The concept of “embedded agency” (Garud et al, 2007:961) clearly evidences this point; embeddedness is not the result of an inescapable pressure from institutions, it is rather a two-way process because firms can adjust to and influence the institutional environment in which they operate – “firms have an important agency role (. . .) they must make sense of, manipulate, negotiate, and partially construct their institutional environments” (Kostova et al, 2008:1001). This view puts into perspective embeddedness as a strategic alternative that multinationals can exploit, rather than a taken-for-granted situation. Multinationals opt for “embedded” strategies in their global undertakings depending on the degree of “fit” between the

Meso-level embeddedness

Overall level of embeddedness:

“The indissoluble connection of the actor with his or her social surrounding” (Beckert, 2003:769)

Macro-level embeddedness

Captures the relationship between foreign organizations and elements specific to national formal institutions, i.e. institutional embeddedness

Captures the inter-personal relationships across individuals operating in the local business environment, i.e. relational embeddedness

Captures the relationship between foreign organizations and local business partners such as suppliers, customers, competitors etc., i.e. network embeddedness

(14)

14 firms’ needs and business purposes and their regional conditions (Mattes, 2010; Teece, 2006). Thus, insights from the agency theory underlie the potential of embeddedness as a contingent factor in explaining organizational performance. The extent to which organizations manage to strategically involve in local business networks and adapt to the institutional settings of their host countries is an important determinant in their success (embedded agency).

In this respect, embeddedness seems to be a phenomenon largely dependent on the ability of firms to gain and retain legitimacy in the markets in which they are operating. Legitimacy is defined as “a generalised perception or assumption that the actions of an entity are desirable, proper or appropriate within some socially constructed system of norms, values, beliefs and definitions” (Suchman, 1995:572). Legitimacy thus reflects firms’ recognition from outside parties that validates their interaction with institutional and market actors and subsequently, their ability to exploit agency roles that facilitate embeddedness in the external environment. In international business, cross-national differences can challenge the ability of firms to obtain legitimacy in foreign markets because they have difficulties in dealing with local idiosyncrasies (Dikova, 2009; Dow and Larimo, 2009). When firms enter new foreign markets, they have to overcome the “liability of foreignness” that steams from unfamiliarity with the local environment and comprises informational disadvantage, higher transactional uncertainty and lack of connections with local agents that lead to higher managerial costs and decreased subsidiary performance (Zaheer, 1995; Zaheer and Mosakowski, 1997). Foreign affiliates need to adjust to a complex set of rules and values that dictate local behavioural norms and business routines (Johansen and Vahlne, 1977; Kostova and Roth, 2002) in order to embed in the new socio-institutional context and minimize the costs born out of the liability of foreignness. If firms fail to adapt their organizational structures to be consistent with the isomorphic pressures from the local environment, they cannot earn the operational legitimacy necessary to integrate and succeed in the new context (Meyer and Rowan, 1977). Therefore, the basic expectation of this thesis would be: Higher levels of embeddedness in

the host country leads to higher performance of the subsidiary.

3.2.1. Institutional embeddedness

(15)

15 studies, which move beyond “rationalist and functionalist concerns of efficiency and performance” (Clark and Geppert, 2011:396) and evaluate the moderating effect that different institutional frameworks have on firms’ strategic choices and performance (Peng et al, 2008; Zhou et al, 2006). Generally, formal institutions are comprised of “tangible rules, laws and regulations, as formalized guidelines that shape behaviour” (Scott, 1995:35) and define the proper functioning of the market exchange mechanisms. Thus, formal institutions are a critical factor in enhancing firm performance because they provide the means to decipher the rules and norms of economic exchange, affect the transaction costs of doing business specific to each country and support the effective functioning of a market by minimizing costs and associated risks (North, 1990).

(16)

16 As countries have different levels of development in domestic institutional systems, there tends to be a large cross-national variance in institutional endowments (Meyer and Peng, 2005; Hoskisson et al, 2000) which unevenly affect a firm’s ability to create competitive advantage abroad (Ingram and Silverman, 2002). From this perspective, institutional embeddedness reflects the extent to which subsidiaries adapt their profit-maximization strategies to the macro-institutional framework prevalent in each host country. Nonetheless, the ability of the firms to integrate within existing frameworks and predict formal institutional developments depends on the institutional knowledge that they carry when they decide to operate in a foreign environment, i.e. how familiar they are with the foreign institutional environment (Xu and Shenkar, 2002; Xu et al, 2004). It is easier for organizations to adopt the behaviour and business practices deemed as legitimate in environments where they have some understanding of the underlying market structures and formal rules and regulations. On the contrary, operating in unfamiliar institutional settings raises transaction costs, increases uncertainty and prevents firms from engaging in complex operations (Meyer, 2001). It is expected then, that institutional embeddedness is reinforced by the institutional proximity between the origin and the host country, while larger dissimilarities may delay, or even preclude, embeddedness.

(17)

17 partners and affects their pursuit for legitimacy in the host environment. In this respect, organizations must develop managerial and commercial practices that suit the variations in institutions and the higher the variation, the more difficult the adaptation (Xu and Shenkar, 2002; Luo, 2001; Meyer, 2001). In conclusion, institutional proximity enhances institutional embeddedness in the foreign environment, increases the chances that subsidiaries align their business operations to the foreign institutional environment and ultimately, implement the most efficient profit-maximization strategies. On the contrary, the higher the distance between the home and the host institutional environments, the more difficult firms find it to decipher the rules of transaction, the more inefficient their undertakings and the lower their performance levels are expected to be. The first hypothesis then reads as follows:

Hypothesis 1: The higher the institutional distance between home and host countries, the lower

the performance of the subsidiaries.

3.2.2. Social network embeddedness

(18)

18 Nahapiet and Ghoshal, 1998; Oh et al, 2004; Tsai and Ghoshal, 1998). Of these, the most prevalent has been the effect that social capital has on the flow of information and exchange of knowledge, since external networks host unique information and resources relevant for organizational learning and the creation of new knowledge (Granovetter, 1985; Uzzi, 1997; Lin, 2001; Dyer and Singh, 1998).

Many scholars have contended that the reason why social networks and the accompanying social capital are developed is intrinsically related to the level of trust inherent in the individuals that are part of the network. Coleman (1988) equated trust with social capital and Nahapiet and Ghoshal (1998) argued that social networks and interpersonal relations cannot exist without trust. Mayer et al (1995) offered a comprehensive definition of trust that has long been accepted among researchers – “the willingness of a party to be vulnerable to the actions of another party based on the expectation that the other will perform a particular action important to the trustor, irrespective of the ability to monitor or control that other party” (p. 712). Therefore, trust is necessary for social order and human interaction to be possible (Misztal, 1996) and it is the main social mechanism that “discourages malfeance” and lubricates social embeddedness (Granovetter, 1985).

(19)

19 Buren, 1999), reducing enforcement and monitoring actions and enhancing innovation (Nahapiet and Ghoshal, 1998). In network theory, trust appears to be the most important mechanism in coordinating intra- and inter-firm organizational exchanges (Adler, 2002) because it facilitates social exchange which helps solve coordination and cooperation problems (Nahapiet and Ghoshal, 1998). Specifically, trust has been proven to be a precondition in shaping inter-firm network embeddedness and building relations with suppliers and customers (Day et al, 2013; Chen et al, 2009; Barney and Hansen, 1994). High-trust relationships provide essential benefits for partners due to better cooperation (Palmatier et al, 2007) and lower governance costs (Dyer and Chu, 2003) that enhance inter-firm learning (Fawcett et al, 2013) and increase firm performance (Johnston et al, 2004).

For foreign affiliates, trust is at the core of social interaction and network embeddedness with local actors. First, operating in high trust societies is of particular benefit in lowering uncertainty of doing business in unknown environments, as trust enhances the formation of social links which can carry valuable information about market characteristics, production technologies and potential trade partners. Second, trust encourages risk taking, which is essential in the context of business transactions between new, illegitimate foreign subsidiaries and local partners (Nahapiet and Ghoshal, 1998). Lastly, trust is indispensable for ensuring cooperation among strangers and individuals that encounter each other irregularly (LaPorta et al, 1996). In high-trust societies people develop generalized expectations about the trustworthiness of other people and engage in trust-based social interactions regardless of the level of familiarity between actors (Rotter, 1971). This is a vital benefit for foreign entrants in new markets, as they are lacking market legitimacy and it provides the primary means to gain recognition from local actors and pursue business goals with unfamiliar customers and suppliers. All in all, general trust makes information exchange easier, facilitates the establishment of business networks, ensures an easier adaptation to the external environments and encourages a cooperative behaviour among firms.

Hypothesis 2: The higher the societal level of trust in the host environments, the higher the

performance of the subsidiaries.

3.2.3. Relational embeddedness

(20)

(21)

21 inter-personal tensions, foreign actors must adapt their behaviour to the local cultural habits (Newman and Nollen, 1996). However, individuals have only limited capacity to assimilate new cultural values as they are normally subconsciously embedded and their tacit nature prevents their transfer across foreign individuals (Hayward, 2002). Consequently, the more unfamiliar they are with the host cultures, the more difficult the recalibration and adaptation processes (Kostova and Zaheer, 1999).

Diverging cultural frameworks also carry differences in cognitive structures and regularities of mental processes that dictate reasoning (Zukin and DiMaggio, 1990) and help individuals make sense of the knowledge and information encapsulated in systems of codes, rules and symbols (Nooteboom, 2000). Individual cognitive processes are constructed by the environments in which people have grown and the distinct life paths that help them interpret, understand and evaluate the world (Nooteboom, 2000). To the extent that these path-dependent processes have been different, people are separated by cognitive distance. For individuals to achieve a common purpose, they need to share certain basic perceptions and interpretation systems that align their competencies and motives (Weick, 1995). On the contrary, “conflicting rule systems of signification or legitimation” (Mense-Petermann, 2006:306) can prevent recipients from effectively deciphering the information exchanged and impede the formation of relationships that warrant the transfer of knowledge among individuals (Hutzschenreuter and Voll, 2008).

(22)

22 Hypothesis 3: The higher the cultural distance between home and host countries, the lower the

performance of the subsidiaries.

3.2.4. The relevance of geographic distance

The most basic concept of cross-country distance refers to geographic distance (Ghemawat, 2001; Deadorff, 1998) and consequently, its impact on the relationship between embeddedness and performance cannot be ignored. The effects of geographic distance on foreign direct investment and business internationalization has been researched both at macro and at micro-levels, and some scholars even explored geographical aspects as a determining factor of cultural similarity between partner countries, as countries that are geographically closer tend to share cultural similarities as well (see Ragozzino (2009) for a review). Intuitively, theoretical mechanisms would suggest that geographic distance is deemed to have a negative impact on foreign subsidiary performance mainly from the lenses of the transaction costs theory because the cost of managing and coordinating foreign subsidiaries increases with the spatial distance between countries of origin where parent firms are established and the host country of the foreign affiliates (Zaheer, 1995). A larger geographic distance inevitably assumes increasing costs in transportation and communication (Dow, 2000) and induces information asymmetries that exaggerates the adaptation costs that arise from the liability of foreignness (Malloy, 2005). In addition, geographic distance creates barriers to direct communication and interactions (Hinds and Bailey, 2003) that ultimately make it more difficult for foreign affiliates to achieve hands-off experience and understanding of the local environments. Geographic distance also challenges the transfer of information and operational knowledge that would otherwise increase business efficiency, particularly as the stickiness of valuable information increases with geographic dispersion (Kumar et al, 2009).

(23)

23 levels and increased operational costs induced by geographic distance are expected to reduce performance, yielding the following hypothesis:

Hypothesis 4: The higher the geographic distance between the home and the host countries, the

lower the performance of the subsidiaries.

3.2.5. The moderating effect of firm size and experience

While organizations are confined to existing societal orders, they can exploit their agency role in order to minimize the constraining role of the environment and enhance their organizational performance (Saka-Helmhout and Geppert, 2011). The agency role of the firms resides in their ability to actually assimilate the cultural systems and regulations that make up the institutional framework of the host country and mediate the negative impact of cross-country distance. In this respect, the agency perspective should be complemented with insights from the resource based theory. The resource based view emphasizes the central role of the firms in building competitive advantage, regardless of the nature of the external environment (Barney, 1991). Among these, size and experience have been considered the main resources that firms can exploit within their international expansion (Bausch and Krist, 2007). Unlike small firms, larger firms have a large availability of resources to be invested in assets necessary to acquire knowledge of the host environment and overcome the difficulties induced by the liability of foreignness (Coviello and McAuley, 1999). In addition, large firms also benefit from large financial resources that give them the leverage to develop efficient business networks and higher lobbying power in creating points of contact with policy makers and regulators (Kayalica and Lahiri, 2007; Baum and Oliver, 1992). Thus, larger size implies that firms dispose of more resources to assimilate the local idiosyncrasies faster and even exploit institutional framework to their advantage, speed up the embeddedness process and moderate the negative impact caused by institutional and cultural distances.

(24)

24 effects of distance (Shenkar, 2001). In addition, the more foreign firms operate in the host environment, the more they start interacting and building ties with local external parties such as customers, suppliers and regulators so that they gradually become accepted as legitimate actors (Zaheer and Mosakowski, 1997; Li et al, 2010). As foreign firms accumulate greater local knowledge and gain legitimacy, the uncertainty induced by institutional and cultural distances reduces, giving them the opportunity to find and implement the most efficient profit-maximizing strategies (Salomon and Wu, 2012). The more experienced a firm becomes in a foreign market, the more it learns to do business there and higher profitability is expected (Johanson and Vahlne, 1977). Firms can then exploit experience as a source of competitive advantage as the resource based theories predict. Based on these arguments, it is expected that subsidiaries’ size and experience play a significant role in moderating the effect of institutional and cultural distance on performance leading to the following sub-hypotheses:

Hypothesis 1a: Conditional on the level of institutional distance, larger subsidiaries have higher

levels of performance than smaller subsidiaries.

Hypothesis 1b: Conditional on the level of institutional distance, more experienced subsidiaries

have higher levels of performance than less experienced subsidiaries.

Hypothesis 3a: Conditional on the level of cultural distance, larger subsidiaries have higher levels

of performance than smaller subsidiaries.

Hypothesis 3b: Conditional on the level of cultural distance, more experienced subsidiaries have

higher levels of performance than less experienced subsidiaries.

(25)

25 Figure 2: Framework of testable hypotheses

4. Research design and methodology 4.1.Sample selection and data collection

In international business, the largest share in performance studies has focused on the corporate as the unit of analysis (Li, 2007). The aim of this thesis however, is to analyse performance contingent on the level of embeddedness in various host environments. Since it is expected that performance variations occur dependent on host country’s environmental characteristics, this study shifts the focus of attention on subsidiary performance rather than the multinational as a whole. In addition, as multinationals operate in dissimilar environments (de Jong et al, 2011) with uneven opportunities for embeddedness, at an aggregate level the individual effects of asymmetric embeddedness on performance may cancel out and the results are likely to be biased. Consequently, for the purpose of this study a multi-level database was constructed that contains

H3a: + Size H3b: + H4: - H3: - H2: + H1: - H1b: + H1a: +

Formal institutional distance (proxy for institutional

embeddedness)

Level of societal trust (proxy for network social

embeddedness)

Cultural distance (proxy for relational

(26)

26

performance indicators and the corresponding variables to operationalize embeddedness at the subsidiary level. The firm-level data was obtained from Bureau van Dijk’s (BvD 2014) Orbis database. Orbis is the most appropriate database for this research because it contains detailed information on corporate performance and other financial indicators for many private companies world-wide.

(27)

27

Table 2 summarizes the distribution of the firms across the home and host countries. A more detailed analysis of the distribution of the subsidiaries across the home-host pairs of countries is available in Table A in the Appendix.

[INSERT TABLE 2 ABOUT HERE] 4.2.Measures

4.2.1. The dependent variable

(28)

28

4.2.2. Independent variables

Formal institutional distance

The empirical proxy for institutional embeddedness is defined by the distance between the home and host country institutional characteristics. Country-level data regarding institutional environment were retrieved from the World Bank World Governance Indicators database developed by Kaufmann et al (2010). The database provides annual aggregate indicators for 215 countries over the period 1996-2012, for six dimensions of governance: (1) Voice and

accountability, reflecting “the extent to which a country's citizens are able to participate in

selecting their government, as well as freedom of expression, freedom of association, and freedom of media”, (2) Political stability and absence of violence/terrorism, capturing “the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including politically-motivated violence and terrorism”, (3) Government effectiveness, summarizing “the quality of public and civil services and their independence from political pressures, as well as the quality of policy formulation and implementation”, (4) Regulatory quality, reflecting the “ability of the government to implement policies and regulations that allow the development of the private sector” (5) Rule of law, i.e. the extent to which agents have confidence in and abide by the rules of society, and in particular “the quality of contract enforcement, property rights, the police, and the courts” and lastly, (6) Control of corruption, reflecting the extent to which “public power is exercised for private gain, including both corruption and the “capture” of the state by private interests”.

As all the six indices have been relevant in the international business literature, particularly in the study of foreign direct investment flows (Buckley et al, 2007; Globerman and Shapiro, 2003), and given the high correlation across the dimensions, this study uses a measure for institutional distance that employs all the six indices into one variable. The composite measurement for institutional distance has been developed using the Eucledian distance formula averaged over six: 𝐼𝐷ℎ,𝑓,𝑡 = 1/6 ∑ (𝐼𝑖,ℎ,𝑡− 𝐼𝑖,𝑓,𝑡)

2 /𝑉𝑖,𝑡 𝑖=6

(29)

29 database (2014). As detailed in the section above, it is expected that the performance of the subsidiaries operating in more dissimilar institutional environments will be lower due to their difficulty to adapt to extraneous institutional rules and bureaucratic norms that should otherwise assist the development of daily operational activities. Hence, a higher value of the composite institutional distance index is expected to return lower values of performance.

Host country societal trust

As explained before, trust is the core mechanism that facilitates human interaction and social exchange. Social embeddedness cannot occur if a reasonable level of trust is absent from the host societies where subsidiaries are operating. At societal level, trust is represented by a common psychological state of a social unit whose members’ behaviour and actions are controlled by the expectation that all the members of the unit are trustworthy (Lewis and Weigert, 1985). Aggregate data on generalized trust is largely available at the country level from secondary sources, such as World Value Survey (WVS) and regional barometers. The empirical data on trust available in these datasets has been collected by ASEP/JDS and compiled into a trust index specific to each country, so that the ASEP/JDS research project contains information on aggregated trust levels for over 100 countries, covering nearly 90% of the world’s population (Medrano, 2014)1. The trust index has been obtained from individual information aggregated at a macro-level, and it refers to generalized forms of trust – i.e. the preconceptions and generalized beliefs about individuals that allow people from different identifiable groups to cooperate with each other in order to achieve common goals (Bottazzi et al, 2011). Thus, it can be expected that individuals living in higher-trusting societies are more likely to develop social connections with foreign actors, both at a personal level and within business-related networks. In line with this argument, this thesis uses the ASEP/JDS trust index specific to the host countries to predict the effect of social embeddedness on performance; subsidiaries performing in higher-trusting societies are expected to have higher levels of performance due to their enhanced chances for network embeddedness. In addition, using societal trust as a proxy for network embeddedness corresponds to existing empirical studies that similarly used measures of trust from World Values Survey/Eurobarameter etc. to construct indicators that

(30)

30 estimate the impact of social networks on economic performance (Knack and Keefer, 1997; Zak, 2001; LaPorta et al, 1996). While it is expected that trust changes over time (Morrone, 2009), the trust index has been developed using the 1995-2009 time frame, which matches the period of observation of the study sample well and therefore, hinders any potential bias carried by temporal incongruity between the firm-level observations and changes in country-specific societal trust.

Cultural distance

It has been argued above that the closest proxy for relational embeddedness resides in the cultural and cognitive similarity between individuals, as cultural and cognitive factors are the main mechanisms through which individuals make sense of the surrounding environment and create points of connection with other people. In the business literature, the most fundamental study on cultural aspects has been Hofstede’s (1980) framework. Hofstede identified four cultural dimensions referred to as (1) Power distance: ‘‘the extent to which the less powerful members of institutions and organizations within a country expect and accept that power is distributed unequally’’ (p. 28), (2) Uncertainty avoidance: ‘‘the extent to which the members of a culture feel threatened by uncertain or unknown situations’’ (p. 113), (3) Individualism versus collectivism: referring to the degree to which ‘‘the ties between individuals are loose’’ or “integrated into strong, cohesive in-groups’’ (p. 51) and (4) Masculinity versus femininity: reflecting to what extent “social gender roles are clearly distinct or overlap’’ (p. 82). While Hofstede’s conceptualization of culture has been widely applied in empirical analyses of cultural diversity (see Kirkman et al (2006) for a review), there has also been considerable criticism of its dimensions (Kirkman et al, 2006; Shenkar, 2001) and related studies have suggested the advantage of alternative constructs (Schwarz and Bilsky, 1990; Tsui et al, 2007; House, 2004). Notwithstanding the on-going debate on the adequacy of cultural constructs, this thesis employs Hofstede’s cultural framework because it has been subject to most validity checks (Shane, 1995; Hofstede, 2001) and in addition, it covers the largest sample of countries.

The index developed by Kogut and Singh (1988) is used to quantify aggregate cultural distances based on Hofstede’s dimensions following the formula: 𝐶𝐷ℎ,𝑓 = 1/4 ∑𝑖=4𝑖=1(𝐼𝑖,ℎ− 𝐼𝑖,𝑓)

2

(31)

31 dimension have been retrieved from Hofstede (2001) and although the dimensions steam from a different period than the one used to construct the sample, this should not affect the results because culture is assumed to be persistent over time (Roland, 2004).

Besides the Hofstede’s (2001) framework, language is another cultural factor that has been extensively researched in international business literature. Considering the purpose of this study, language is crucial in defining embeddedness because it is the main means of inter-personal and inter-organizational interaction. Research on linguistic differences has shown that thinking is affected by language (West and Graham, 2004) and operating across linguistic boundaries frequently involves misunderstandings (Marschan-Piekkari et al, 1999). In addition, language is a prime factor of cognitive proximity which allows for the formation of a common platform for knowledge creation and information exchange, which occurs “through the existence of shared knowledge and vocabulary and through the sharing of collective narratives (Nahapiet and Gshoshal, 1998:253). Through common language actors share the same frame for deciphering and understanding context and consequently, find it easier to align business purposes and create successful partnerships in cross-national settings (Feely and Harzing, 2003). In line with previous research, a dummy variable that captures the existence of a common language between partner countries (Srivastava and Green, 1986; Arora and Fosfuri, 2000) is the second cultural explanatory variable of the model.

Geographic distance

To investigate the effects of geographic distance on performance, an indicator calculated via the great circle formula using latitudes and longitudes of the most important urban agglomerations available in the CEEPI database (Mayer and Zignago, 2011) has been introduced as the last explanatory variable of the analysis (see also Peeters and Gooris, 2013). Insights from transaction costs theory and regionalization studies suggest a negative relationship between geographic distance and performance, so it is expected that the measurement for spatial distance as described here also has a negative impact on subsidiary performance.

(32)

32 Figure 3: Summary of variables and main testable hypotheses

Theoretical background

Hypothesis Variable Variable degree Potential for embeddedness Expected effect on performance Formal institutional embeddedness H1 Institutional distance High Low _- Low High + Social network embeddedness

H2 Trust High High ₊

Low Low _-

Relational embeddedness

H3 Cultural distance High Low -

+

Low High

Additional factors _H4 Geographic distance High Low _-

Low High ₊

4.2.3. Control variables

To avoid bias from other effects on performance, several control variables were included in the model (see also de Jong and van Houten, 2014; Mezias, 2002). First, in line with the resource-based view, firm characteristics and organizational capacities – usually reflected by the firm size - are critical factors in determining performance. Large firms have the financial means and capital resources to exploit economies of scale through specialization and capital outlays (Chao and Kumar, 2010), have higher chances to diversify risk and may engage in extensive R&D activities that enhance learning (Nooteboom, 1994). Consequently, the log of the number of employees was included in the model specification to control for firm size (Johnston and Menguc, 2007). Second, firm age - measured as the difference between the year of observation and the year of incorporation – can affect performance as well. Generally, age is expected to have a positive effect on performance, as the older firms get, the more they advance along the learning curve. However, it is likely that age has a non-linear effect on performance, as older firms tend to perform worse due to conservative management practices, inflexible structures and old technology (Sapienza et al, 2003; Sorenson and Stuart, 2000). Third, sales growth is also expected to affect performance, because high sales growth normally implies a better utilization of the existing capacity and spreads of costs over more revenue (de Jong et al, 2011). The sales growth control variable was calculated as 𝑆𝐺_𝑖,𝑡 = (𝑆_𝑖,𝑡− 𝑆_{𝑖,𝑡−1})/𝑆_{𝑖,𝑡−1}, where 𝑆𝐺_𝑖,𝑡 represents sales growth in year of observation 𝑡, and 𝑆𝑖,𝑡 represents the amount of sales in year 𝑡 for firm 𝑖.

(33)

33

and industry-specific life cycle products (Porter, 1981; Hawawini et al, 2001). In particular, such industry characteristics appear to be mostly related to the degree of knowledge and technology intensity involved in the production process (Utterback, 1974). In line with this argument, industry-specific effects were captured by dummy variables that stratify the manufacturing sectors in which the subsidiaries operate according to the Eurostat (2014) classification into low-technology (e.g. food products, beverages and tobacco), medium-low low-technology (such as rubber and plastic products), medium-high technology (for instance machinery and equipment) and high technology (such as pharmaceuticals, computers and electronics)2_.

Lastly, country-specific characteristics can also influence firm performance and define the chances for embeddedness. For instance, economic factors and political policies such as inflation, level of GDP, level of unemployment, extent of privatization, openness to trade, labour regulations, level of education, degree of industrialization, infrastructure development, access to financial capital and technology penetration etc. can have a significant effect on the performance levels of firms operating in a specific country (de Jong et al, 2011; Kumar, 2001; Lopez-Claros et al, 2005; Dunning, 2009). In addition, political studies have shown that countries with similar levels of economic development share comparable characteristics along these dimensions. For instance, unlike their richer counterparts, poor and developing countries generally have a high level of inflation, low GDP per capita, high levels of unemployment, low access to financial markets, low degrees of industrialization and poor infrastructure (Todaro and Smith, 2009; Schaffner, 2013). Consequently, in order to control for host country-specific effects, dummy variables were included that categorize each host country in the sample as developed (“high-income”) and developing (“middle and low income”) at the cut-off level of a GNI/capita of $12,615 according to the country-group income classification developed by World Bank (2014). However, besides host-country characteristics, home-country idiosyncrasies can also influence the performance of the subsidiaries. Research has shown that home-country characteristics have a great impact on internationalization strategies, managerial decisions, operational activities and patterns of organizational behaviour (e.g. Wan and Hoskisson, 2003; Bausch and Krist, 2007). Therefore,

(34)

34

dummy variables were created to control for the effects of each individual country of origin used in the sample.

Table 3 summarizes the construction of the relevant variables. [INSERT TABLE 3 ABOUT HERE] 5. Empirical specification and results

5.1.Model description and estimation

Having constructed the key measurements, it is now possible to estimate the impact of embeddedness on the subsidiary performance using regression analysis. To this end, the full empirical specification reads as follows:

ln(𝑅𝐸𝑉𝐸𝑁𝑈𝐸)𝑖,𝑡 = 𝛽0+ 𝛽1𝐼𝑁𝑆𝑇_𝑑𝑖𝑠𝑡𝑖,𝑡+ 𝛽2𝑇𝑅𝑈𝑆𝑇𝑖+ 𝛽3𝐶𝑈𝐿𝑇_𝑑𝑖𝑠𝑡𝑖+ 𝛽4𝐿𝐴𝑁𝐺𝑈𝐴𝐺𝐸𝑖+ 𝛽5𝐺𝐸𝑂𝐺𝑅𝐴𝑃𝐻𝐼𝐶_𝑑𝑖𝑠𝑡𝑖+ ∑4𝑥=1𝛽𝑥𝑋𝑖,𝑡+ 𝛽6𝐻𝑂𝑆𝑇_𝑖𝑛𝑐𝑜𝑚𝑒𝑖+

∑𝑗−1𝑗=1𝛽𝑗𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌𝑖,𝑗+ ∑𝑧−1𝑧=1𝛽𝑧𝑌𝐸𝐴𝑅𝑧+ ∑ℎ−1ℎ=1𝛽ℎ𝐻𝑂𝑀𝐸𝑖,ℎ+ 𝑒𝑖,𝑡 (1)

where the subscripts stand for subsidiary (𝑖) and year of observation (𝑡), 𝛽s are the estimated parameters, 𝑋𝑖,𝑡 is a vector of time-variant firm-specific control variables (including 𝑆𝐼𝑍𝐸𝑖,𝑡, 𝑆𝐴𝐿𝐸𝑆_𝑔𝑟𝑜𝑤𝑡ℎ𝑖,𝑡, 𝐴𝐺𝐸𝑖,𝑡 and 𝐴𝐺𝐸_𝑠𝑞𝑟𝑡𝑖,𝑡, where the squared term of age has been introduced to control for the non-linear relationship between age and performance), 𝑒𝑖,𝑡 is the specific error term and 𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌𝑖,𝑗 and 𝑌𝐸𝐴𝑅𝑧, are industry and time dummies that control for performance differences across manufacturing sectors and time-specific economic conditions, and lastly, 𝐻𝑂𝑆𝑇_𝑖𝑛𝑐𝑜𝑚𝑒𝑖 and 𝐻𝑂𝑀𝐸𝑖,ℎ are country dummies that control for host and home country characteristics. The descriptive statistics of the variables are available in Table 4.

The hypotheses were estimated using six different specifications of the model, examining first the effects of the control variables only (Model 1). The basic model was subsequently complemented with each of the main explanatory variables, so that Models (2) – (5) sequentially estimate the independent effects of institutional distance, trust, cultural distance and language, and geographic

distance on subsidiary performance. Lastly, in order to capture the full effect of embeddedness on

(35)

35 characterizes cross-sectional time-series data and which instead is captured by fixed effects (FE) and random effects (RE) estimators (Greene, 2002). Hence, while the results from POLS are reported for comparison purposes, post-estimation tests revealed the adequacy of RE and FE respectively, so the focus of this analysis will be on the estimations obtained from the RE and FE specifications. More precisely, considering the time-invariant nature of the key explanatory variables and in accordance with the results of the Breusch and Pagan Lagrangian multiplier tests, Models (3) – (6) were evaluated using the RE estimators. With respect to Models (1) and (2), where time-changing explanatory variables were regressed, the Hausman test has suggested the superiority of the FE model over RE and POLS. Nevertheless, the variables of interest generally maintained their sign and significance levels whether estimated with POLS, RE or FE estimators, which builds confidence in the reliability of the results. In addition, the estimated coefficients of the main effects variables kept their signs and levels of significance both when regressed separately and when regressed together in Model (6). Consequently, Model (6) seems to provide reliable and consistent estimations for all the key variables and it will be the preferred econometric specification for evaluating the complete effect of embeddedness on performance using RE generalized least squares (GLS) estimates as summarized in equation (1).

(36)

36 value of the standard errors and inflating the t-statistics (Cameron and Trivedi, 2005). Thus, autocorrelation of the disturbances could severely bias the parameters estimates of the coefficients and consequently, the interpretation of the relative impact of the embeddedness proxies on performance might be incorrect. Indeed, a more detailed examination of the data revealed a highly significant autocorrelation of errors for the data sample (the computed Wooldridge F-stat for first order autocorrelation equalled 1948.559, p-value<0.001 for the benchmark Model (1) and has remained highly significant for all the remaining regressions). Consequently, in order to control both for heteroskedasticity and for within-group autocorrelation of disturbances, all the econometric specifications – POLS, RE and FE – were estimated using clustered robust standard errors as suggested by White (1980), Rogers (1993), Wooldridge (2002, 2013), Arellano (1987, 2003) and Stock-Watson (2008)3. The clustered robust standard errors by firm identifier was particularly attractive in this case, given the short nature of the unbalanced panel – short fixed time frame of ten years with an average of 6 observations available per firm and large number of firms (n=10825).

[INSERT TABLE 4 ABOUT HERE] 5.2.Discussion of the results

5.2.1. Regression results for the main explanatory variables

Table 5 reports the estimation results for the six different econometric configurations. The fit parameters show that together with the control variables, the embeddedness model seems to explain a significant share of the variance in the levels of subsidiary performance (the value of R2 being approximately 0.7). In all the regressions, the results for the control variables at the firm level are highly significant at the 1% level and have the expected positive signs, revealing that subsidiary size (β=.571, p-value<0.001) and sales growth (β=.023, p-value<0.001) foster

(37)

37 subsidiary performance. In addition, the theorized non-linear relationship between age and performance seems to be valid. The coefficients of age and its corresponding quadratic term are both highly significant at the 1% significance level (p-values<0.001) and have opposite signs (βAGE=.621; βAGE_sqrt= -.005), as firms tend to perform better the older they get, yet after a certain time period performance starts to level off, possibly because old firms tend to be more inflexible and conservative and employ outdated technology and management strategies (Sorenson and Stuart, 2000). Similarly, in all the regressions all the time-dummies, industry-dummies and home-country dummies are significant at the 5% level (p-values<0.05) and have positive values. One notable exception is Italy, whose dummy estimated parameter turned out to be insignificant in some of the regressions, suggesting that the characteristics of the Italian country have no significant effect on the performance of the foreign subsidiaries, at least when compared to the effect that the benchmark country, Germany, has on its subsidiaries (results available upon request).

(38)

38 of transactions and hinders firms from implementing the most efficient profit-maximization strategies (Meyer et al, 2009; Henisz, 2003).

(39)

39 pointed out that organizational operations are influenced by cultural embeddedness in as much as ”shared understandings and meanings come to give form to organization activity, structure and process […] and shape strategies and goals” (Dacin et al, 1999:328). Similar cultural-cognitive factors cause actors to employ systems of shared understandings in the social exchange process (Kostova et al, 2008). Consequently, firms facing important cultural differences find it difficult to develop the common values and beliefs necessary to coordinate daily operational activities within and across operational entities because individuals have different frames of reference in the interpretation of the business purpose. The more dissimilar the cultural frameworks are, the more likely it is that foreign affiliates will face misunderstandings and miscommunication in the exchange processes that will decrease their ability to successfully operate in the host environments (Gomez-Mejia and Palich, 1997).

The regression results corresponding to the effect of geographic distance on subsidiary performance are rather surprising, revealing a significant and positive relationship (0<0.001, β=.290) that is largely in conflict with theoretical underpinnings. However, a more careful inspection of the sample data used in the study suggested that there might be a non-linear relationship between performance and geographic distance. Therefore, the effect of spatial distance was re-estimated by including the squared term of the geographic distance variable as a regressor in order to control for a potential non-linear relationship between performance and home-host country spatial distance4. Indeed, the regression results not only confirmed a non-linear relationship between performance and geographic distance, but also revealed a negative effect in the level of geographic distance, suggesting that assuming a strictly linear relationship between spatial distance and subsidiary performance would most likely result in biased estimators. The regression results controlling for the non-linear relationship are reported in Table 6. The estimated

(40)

40 coefficients of the geographic distance variable are significant at the 1% level (p-values<0.001) and reveal an initial negative impact on the subsidiary performance (βGEOGRAPHIC_dist= -2.986), but this effect subsides after a certain point, allowing firms to follow a positive performance trajectory regardless of the spatial distance (βGEOGRAPHIC_dist_squared=.225). The negative estimated coefficient of the linear term provides empirical support to the transaction cost-based arguments that geographic distance induces information asymmetries during transactions, creates barriers to face to face communication and direct interactions and thwarts the efficient coordination and collaboration between business partners that reduces their chances for embeddedness in the host environments and negatively affect performance.

(41)

41 smaller firms. In a similar manner, the interaction term between geographic distance and experience is highly significant and negative (p-value<.0001, β= -.079), which implies that when foreign affiliates are located at larger geographic distance, the younger firms tend to perform worse than their more experienced counterparts. In the light of the non-linear relationship between geographic distance and performance, the fourth hypothesis of the study receives only partial empirical support.

5.2.2. Regression results for firm size and experience moderating effect

(42)

42 (Henisz and Delios, 2001) and gain higher legitimacy (Zaheer and Mosakowski, 1997), so that the liability of foreignness born out of cross-country distance may decrease in time.

The coefficient of the interaction term between size and institutional distance is highly significant at the 99% level of confidence (p-value<0.001) and has a positive sign (β=.088), thus supporting the argument that the leverage and the resources that large subsidiaries benefit from are useful means to ease adaptation to institutionally distant environments and cause them to perform better than smaller firms, everything else equal. However, size does not seem to have a significant impact in mediating the negative effect of cultural distance. The RE-estimated interaction term between size and cultural distance has a positive sign (β=.027), but it is statistically insignificant. Cultural differences seem to exert an overwhelming pressure on the ability of foreign affiliates to integrate and perform well on the local markets. Unlike formal institutions which are rather transparent, codified and can be learned, cultural and cognitive knowledge is tacit and difficult to comprehend so that it cannot be easily assimilated by foreigners. Thus, overcoming cultural differences seems to be a long-term process that can be achieved only with higher participative experience in the local environment.

[INSERT TABLES 7 AND 8 ABOUT HERE]

5.2.3. Robustness checks

As a test of robustness, the impact of institutional distance and cultural distance on subsidiary performance were re-estimated controlling for potential non-linear effects. Following the logic of this analysis, it is expected that given the exogenous nature of the variables, the effect of cultural distance and institutional distance remains negative at any point, everything else constant. Indeed, the results of the regressions when adding the squared terms to control for potential non-linear effects validate a continuously negative relationship on performance for both variables (see Table 9). The parameter estimates for cultural distance and for institutional distance maintain their negative signs both for the linear and for the squared forms of the variables, however the RE and the FE estimators suggest a negative linear – rather than non-linear - relationship between cultural distance and performance and institutional distance and performance respectively, as the estimated parameters for the squared forms of the variables are statistically insignificant.

(43)

43 Additional robustness tests were run to confirm the significant impact that firm experience seems to have on enhancing embeddedness and mitigating cross-country distances. Therefore, the moderating effect of experience was re-estimated controlling for the non-linear relationship between age and performance. Two dummy variables were created for this purpose – OLD (set to 1 if the firm is older than 10 years, the value corresponding to the 75th_{percentile in the dataset)} and YOUNG (set to 1 if the firm is younger than 4 years, the value corresponding to the 25th percentile in the dataset). Thus, the range between 4 and 10 years would be the baseline group for comparison. The results of the robustness tests (summarized in Tables 10-12) confirm the moderating effect of experience on cross-country distances. All the estimated parameters for the coefficients corresponding to the interaction terms between experience and institutional, cultural and geographic distances are significant at the 1% level (p-values<0.001) and have a negative sign for young firms and a positive sign for old firms, suggesting that conditional on the cross-country distance, younger firms tend to perform worse than medium-aged firms, while mature and more experienced firms tend to perform better.

[INSERT TABLES 10 - 12 ABOUT HERE]

(44)

44 explanatory variables have remained highly significant (p-values<0.001) and kept their expected signs (see Table 13). Unfortunately, the Driscoll - Kray standard errors can only be calculated within the context of POLS and FE specifications, such that individual specific effects of the time-invariant variables (normally estimated using the RE estimator) cannot be captured in the analysis. However, the fact that the results of the POLS Driscoll-Kray robustness test has returned consistent results is only reassuring with respect to the validity of the main model.

6. Concluding remarks

This thesis has investigated the effect that embeddedness has on foreign subsidiaries’ levels of performance. Researchers have developed various theoretical mechanisms to evidence the relevance of embeddedness in business operations, and some additional studies have contributed to the literature body by indirectly drawing upon insights from the embeddedness paradigm. However, the exact conceptualization of the term has been highly debated, as numerous interpretations have focused on single context-dependent aspects of the concept. This study has compiled the different streams of the literature into a coherent conceptual framework that allowed for the empirical investigation of the various dimensions of embeddedness simultaneously. The framework so-developed contains three levels of analysis that altogether comprise all the aspects of embeddedness: (1) macro-level embeddedness, reflecting the dynamics between firms’ operational behaviour and the formal institutional frameworks specific to the countries in which they operate; (2) meso-level embeddedness, referring to the networks and social relationships established across organizations that share the same business setting and (3) micro-level embeddedness, which analyses the extent to which individuals operating in the same business environment are able to establish personal relationships that warrant the alignment of their economic purposes. As these three dimensions have been identified based on the various perspectives on embeddedness that exist in the current literature, this conceptual framework is a first attempt to achieve a comprehensive understanding of the embeddedness paradigm.

(45)

On embeddedness and subsidiary performance