Cultural Distance and Firm Internationalization: A Meta-Analytical Review and Theoretical Implications.

This paper presents the most comprehensive review and meta-analysis of the literature on cultural distance and firm internationalization to date. We analyze the effects of cultural distance on key strategic decisions throughout the entire process of internationalization. For the preinvestment stage, we examine the decisions on where to invest (location choice), how much to invest (degree of ownership), and how to organize the foreign expansion (entry and establishment mode). For the postinvestment stage, we examine the decisions of how to integrate the foreign subsidiary into the organization (transfer of practices) as well as the performance effects of cultural distance at both the subsidiary and the firm level. We find that firms are less likely to expand to culturally distant locations but if they do, they prefer greenfield investments and integrate subsidiaries more through transfer of management practices. Cultural distance does not seem to affect how much capital firms invest and whether they enter through a joint venture or full ownership. Interestingly, cultural distance has a strong negative effect on subsidiary performance but no effect on the performance of the whole multinational company. In addition, we find that the effects of cultural distance are not sensitive to time, but they are sensitive to the cultural framework used (e.g., Hofstede vs. Global Leadership and Organizational Behavior Effectiveness) and the home country of the company (developed vs. emerging market). Based on our study, we feel confident to offer some theoretical insights, recommendations for improving the validity and reliability of cultural-distance research, and ideas for future research.

When internationalizing, firms are faced with several critical decisions, such as where and how much to invest and how to organize and govern the foreign venture for maximizing benefits and minimizing risks and losses (Dunning & Lundan, 2008; Marano, Arregle, Hitt, Spadafora, & van Essen, 2016). Theories of internationalization explaining these processes and strategies have been at the core of the field of international business (Andersen, 1993; Johanson & Vahlne, 1977; Vernon, 1979). Central to this research is the proposition that due to the cross-border condition, multinational companies (MNCs) are different from domestic firms not only in degree but also in kind, as they are simultaneously embedded in multiple and diverse social contexts. This uniquely affects their strategies and organization and creates distinct challenges and opportunities that need to be carefully managed (Bartlett & Ghoshal, 1998; Hymer, 1976; Johanson & Vahlne, 1977; Kostova, Roth, & Dacin 2008; Kostova & Zaheer, 1999; Westney & Zaheer, 2009).

To better understand the essence and the impact of the cross-border condition, international business scholars have introduced the concept of distance (i.e., difference between countries) and have applied it to a wide range of topics. Distance has been found to affect various organizational processes and outcomes in MNCs, including location choices, entry mode, standardization of practices, transfer of knowledge, performance, and others (Johanson & Vahlne, 1977; Kogut & Singh, 1988; Kostova, 1999; Kostova & Zaheer, 1999; Tihanyi, Griffith, & Russell, 2005; Xu & Shenkar, 2002). The centrality of this condition has led some to conclude that “essentially, international management is management of distance” (Zaheer, Schomaker, & Nachum, 2012: 19; italics in original).

Reflecting the different domains of contexts, scholars have studied different types of distance, including geographic (e.g., Eden & Miller, 2004), economic, administrative (e.g., Ghemawat, 2001), institutional (e.g., Kostova, 1996, 1997; Kostova & Roth, 2002), linguistic (e.g., Dow & Karunaratna, 2006), or combinations of the above (e.g., Beugelsdijk, Nell, & Ambos, 2017). Despite such proliferation, cultural distance, that is, the difference in cultural values between two countries, remains the most widely used type of distance in international business (Beugelsdijk & Mudambi, 2013; Shenkar, Luo, & Yeheskel, 2008; Tihanyi et al., 2005), perhaps owing to the centrality of cultural values in shaping individual and organizational behaviors (Hofstede, 2001; House, Hanges, Javidan, & Dorfman, 2004; Kirkman, Lowe, & Gibson, 2006; Schwartz, 1994). Despite serious critique on both conceptual and methodological grounds (most notably, Shenkar, 2001), the construct of cultural distance continues to be widely used. With over 5,000 citations, the original article that introduced cultural distance and provided a measurement instrument for it—the so-called Kogut and Singh cultural-distance index (Kogut & Singh, 1988)—is among the most cited papers in management (Harzing & Pudelko, 2016). In fact, in a review of Hofstede-based research, Kirkman et al. (2006: 299) concluded that “most research examined the impact of cultural distance on organizational and country level outcomes.” Similarly, Lopez-Duarte, Vidal-Suarez, and Gonzalez-Diaz (2016) found that more than 80% of the articles on culture and firm internationalization focused on cultural distance.

Given the vast amount of work on cultural distance and firm internationalization on the one hand and the serious points of critique raised about the cultural-distance construct on the other, we believe that there is a need for a critical assessment of the current state of this research. First, internationalization is an increasingly common strategy for all types of companies around the world and understanding the impact of cultural differences on the survival and success of these endeavors is vital. Pressured by growing global competition, Western companies are internationalizing at unprecedented levels and are often expanding into rather “distant” developing and emerging host countries. Likewise, emerging market firms are aggressively internationalizing to distant Western countries (Boston Consulting Group [BCG], 2014; Gubbi, Aulakh, Ray, Sarkar, & Chittoor, 2010; Guillén & García-Canal, 2009; Luo & Tung, 2007). In this context, the original view that cultural distance is a deterrent in international expansion may need to be reassessed. Does cultural distance continue to be an important factor concerning internationalization decisions, and does it matter what the home base of the firm is—developed or emerging market country? Second, in our review we found that with few exceptions, researchers tend to apply the “blanket” logic of negative effects of distance on internationalization and rarely provide an in-depth or nuanced explanation of its multifaceted impact. How does distance affect the different stages of the internationalization process? Is it equally salient in the pre- and post-expansion period? Which particular outcomes associated with firm internationalization are most affected by cultural distance? Third, in light of some of the existing critique (Kirkman et al., 2006; Shenkar, 2001; Tung & Verbeke, 2010), is it worth testing for methodological contingencies related to operationalization and measurement of cultural distance? For example, which cultural frameworks (e.g., Hofstede or Global Leadership and Organizational Behavior Effectiveness [GLOBE]) have the most salient impact on firm internationalization? Are perceptual or “objective” measures of cultural distance (e.g., based on Hofstede and GLOBE indexes) equally potent? How has economic globalization impacted the importance of cultural distance for firm internationalization? For example, is distance less important now than it was 25 years ago?

Accordingly, the objective of our paper is threefold: (a) take stock of the growing literature on cultural distance and the process of firm internationalization, (b) synthesize and analyze this literature identifying robust findings, and (c) develop new theoretical insights on the effects of cultural distance on the firm internationalization process. Such a combined approach of review, analysis, and theory expansion is particularly important for areas of research that have experienced massive growth and may have produced inconsistent and inconclusive results, as is the case for the work on internationalization. Moving forward requires making sense of what has been already done in an informed and rigorous way and laying out ideas about future research steps in this area of inquiry.

Our study seeks to make a distinct contribution beyond the existing reviews and the six prior meta-analyses on cultural distance and internationalization (Magnusson, Baack, Zdravkovic, Staub, & Amine, 2008; Morschett, Schramm-Klein, & Swoboda, 2010; Reus & Rottig, 2009; Stahl & Voigt, 2008; Tihanyi et al., 2005; Zhao, Luo, & Suh, 2004). It is more comprehensive and detailed at the same time because we assess the impact of cultural distance on the various stages of the entire internationalization process. This is different from previous work, which has focused only on specific aspects of internationalization (e.g., examining entry mode but ignoring location choice) or has aggregated various aspects into one internationalization construct. We distinguish between pre- and post-investment stage. Preinvestment decisions include (a) location choice (Dunning & Lundan, 2008; Rugman & Verbeke, 2009), that is, which host country to enter; (b) entry mode (e.g., Brouthers, 2002; Kogut & Singh, 1988), that is, whether to enter through a joint venture (JV) or a wholly owned investment (WOS); (c) establishment mode, that is, whether to enter through acquisition or greenfield; and (d) degree of ownership (e.g., Chan & Makino, 2007; Madsen, 2009), that is, the size of the investment or the amount of capital invested, which reflects the level of commitment to the host country (Ghemawat, 1991). Postinvestment decisions concern (a) the integration of the foreign operations through practice transfer from the parent company to the subsidiary (e.g., Ahuja & Katila, 2001; Sarala & Vaara, 2010; Slangen, 2011) and (b) performance results of internationalization at both subsidiary and firm level (e.g., Barkema, Bell, & Pennings, 1996).

To ensure rigor, parsimony, and confidence in our findings, we use a meta-analytic methodology (Duran, Kammerlander, van Essen, & Zellweger, 2016), pulling together a large number of independent studies of cultural-distance effects on various stages of the internationalization process. This technique also allows us to examine certain contextual and methodological contingencies that could be viewed as boundary conditions of the underlying theoretical model, for example, the measurement approach used for computing cultural distance or the type of home country of the MNC—developed versus emerging market.

We have reviewed and coded a total of 156 papers published in a wide range of management and international business journals in the period 1988 to 2015. Our coding protocol is extensive, assessing both different stages of the process of firm internationalization and different approaches to conceptualizing and measuring cultural distance. This much bigger sample compared to previous meta-analyses (with sample sizes between 14 and 61 papers) provides the necessary statistical power to more precisely assess the various stages and outcomes of the process of firm internationalization. It also covers a more diverse set of countries with greater variation of cultural values and level of economic development, which allows us to explicitly test many of the conjectures suggested by critics of the cultural-distance literature (Shenkar, 2001; Tung & Verbeke, 2010).

The picture that emerges from our study is that cultural distance has a differential effect on the various stages of the internationalization process. It is a significant factor in the ex-ante decisions about location choice (a high cultural distance reduces the probability of investment in a country) and establishment mode (a high cultural distance is associated with firms preferring a greenfield and not an acquisition) but does not directly affect the degree of ownership invested. Regarding the postinvestment stages, cultural distance is associated with greater transfer of home country practices, most likely as a way to bring the parent company and the foreign subsidiary closer together. Interestingly, we find that cultural distance makes transfer of practices more difficult, but firms that do so benefit from it. Finally, the performance implications of cultural distance are also nuanced. It has a negative impact on subsidiary performance (consistent with the liability of foreignness argument) but has no effect or even a marginally positive effect on the performance of the whole MNC. We also find that effects can depend on the particular way in which cultural distance is measured (Hofstede, GLOBE, Schwartz, or perceptual measures).

Cultural Distance and Firm Internationalization

National Cultural Distance

Theoretically, the argument on the role of national cultural distance in firm internationalization is a core element of the Uppsala model (Johanson & Vahlne, 1977) and can even be traced back to Beckerman (1956). As these authors suggested, cultural distance, that is, the difference between the cultures of the home and host countries, is an important consideration in internationalization strategies. When internationalizing, firms first expand to culturally and/or geographically close countries and move gradually to culturally and geographically more distant countries as they learn from their international experiences. Implicit here is the idea that cultural distance creates difficulties and challenges for firms due to lack of knowledge and understanding of how the host country works, as well as the perceived “foreignness” or “psychic distance” that creates barriers for collaboration and cooperation.

Cultural distance affects all stages of the internationalization process, including the preinvestment stage when the company has to make a decision whether to invest in a particular market, what entry mode to use, and how much to invest, as well as the postinvestment stage when the decisions revolve around the degree of integration of the foreign location through common practices as well as the performance outcomes of the international investment. Appendix A (see online supplement) presents a set of quotes (at least one for each stage and associated strategic decision of the internationalization process) that illustrate these effects. Although the particular arguments about the impact of cultural distance vary by stage and decision, the overarching rationale is that cultural distance leads to higher complexity and costs of doing business abroad.

Empirically, cultural distance was first operationalized by Kogut and Singh in their 1988 article, where they used the construct to explain entry mode choice. Using Hofstede’s multidimensional culture framework, Kogut and Singh (1988) introduced a Euclidean distance measure to capture cross-country cultural differences in one index. The Euclidean distance index takes the difference on the national score on each of Hofstede’s cultural dimensions (Hofstede, 1980) and then aggregates these differences in one overall index. Cultural distance is calculated as the distance to a single country. The vast majority of cultural-distance studies follow this approach in operationalizing and measuring cultural distance (Kirkman et al., 2006; Kirkman, Lowe, & Gibson, 2017). As seen in Figure 1, the number of cultural-distance studies published in management journals has steadily increased since 1988.

Figure 1

Distribution of Cultural-Distance Papers Over Time

Despite its proliferation, cultural-distance research has been criticized on multiple grounds (Beugelsdijk, Kostova, & Roth, 2017; McSweeney, 2002; Shenkar, 2001, 2012; Tung & Verbeke, 2010): (a) for presenting an overly simplistic way of using the cultural-distance construct in theory building—assuming an equivalent (negative) effect of cultural distance on different organizational outcomes (location choice, entry and establishment mode, governance, performance); (b) for ignoring important statistical properties of the index, for example, assuming uncorrelated cultural dimensions; and (c) for using almost exclusively the possibly outdated Hofstede’s data in computing the index of cultural distance. Finally, it has been suggested that distance effects are possibly conflated with level effects depending on the sample structure (Brouthers, Marshall, & Keig, 2016; Harzing & Pudelko, 2016; van Hoorn & Maseland, 2016). Cultural-distance studies that include one home (host) and multiple host (home) countries may not be able to attribute the effect of cultural distance to cultural differences (and, in fact, find a level effect), depending on the absolute score of the single home (host) country on the cultural dimensions. Van Hoorn and Maseland (2016) show that this is particularly problematic for cultural-distance studies using the United States as a reference country.

Adding to this growing literature, in this paper we examine the differential effects of cultural distance on various decisions related to the different stages of the internationalization process, recognizing that these effects can differ in strength and also in terms of underlying theoretical explanations. Thus, we aim to address the critique that cultural distance has been used as a blanket, “catch-all” treatment of country differences and the myopic view that it affects all phenomena of cross-border nature in a similar and negative way. In testing the relationship between cultural distance and location choice, entry and establishment mode, degree of ownership, transfer of practices, and performance, we take into account these critical observations.

The Process of Firm Internationalization

As depicted in Figure 2, the process of firm internationalization has been conceptualized as a set of several key decisions—on location (whether a company should invest into a particular host country), entry mode, how much it should invest, and how the foreign operation should be controlled and managed. These are strategically important decisions and making a mistake in any of them can have a detrimental impact on performance, including a potential failure of the foreign operation altogether. Expanding the company’s operations abroad is far more challenging than doing it in a domestic setting. Abroad, firms face difficulties and incur additional costs due to political and economic risks in the host country (Alvarez & Barney, 2005; Maitland & Sammartino, 2015a) as well as legitimacy challenges (Kostova & Zaheer, 1999) and the so-called liability of foreignness (Eden & Miller, 2004; Hymer, 1976; Zaheer, 1995). This is due to lack of familiarity with the host country and the ways of organizing and conducting business, limited information about opportunities and risks on operating in a foreign country, lack of adequate organizational capabilities to deal with those risks, and common discrimination by local constituents against “foreign” entities (Zaheer, 1995). These difficulties permeate all stages and aspects of a firm’s expansion and operation abroad and can only be addressed, at least to some extent, with appropriate internationalization strategies.

Figure 2

The Firm Internationalization Process Unpacked

As we describe below, various theories have been proposed to explain different outcomes associated with the stages of a firm’s internationalization process. Rather than being comprehensive in our review of this vast literature, our goal is to sketch the totality of approaches and the central themes and findings to build a basic understanding of the firm internationalization process, which can then provide the necessary foundation for our examination of the role of cultural distance.

Location choice

Location-choice theories of firm internationalization are classified into two main types (Buckley, Devinney, & Louviere, 2007; Kim & Aguilera, 2016). The first is rooted in the economic tradition (Kindleberger, 1969; Vernon, 1966), whereby the choice of a specific location for foreign investment is based on a rational process of decision-making based on a set of clear criteria (Buckley & Casson, 1976). In this perspective, internationalization motives typically include market seeking, efficiency seeking, natural-resource seeking and knowledge or strategic-asset seeking (Dunning, 1980; Dunning & Lundan, 2008; Hymer, 1976). Firms choose to invest in a specific location because of the related growth opportunities and/or cost advantages. This is a calculative rational economic decision.

The second perspective takes a more behavioral approach. Grounded in Cyert and March (1963) and Penrose (1959), it emphasizes the gradual learning that happens as firms internationalize, which then expands firms’ horizons for future internationalization. This perspective on internationalization is captured by the so-called Uppsala model (Barkema & Drogendijk, 2007; Johanson & Vahlne, 1977, 1990, 2009). Here, location choices are viewed as a sequence that builds on previous foreign expansions and the associated organizational learning. Each subsequent foreign expansion is likely to be to a market that is somewhat similar to the existing locations of the company’s operations. Although it has been suggested that location choice is best explained by a combination of both rational economic approach and capability process based approach (e.g., Makino, Lau, & Yeh, 2002), these two internationalization theories continue to be generally seen as distinct archetypes of firm location-choice theories (Buckley et al., 2007).[1]

Theoretically, location-choice studies typically explain the decision to expand to a specific host country based on the anticipated communication, coordination, and control costs. Accordingly, they predict that firms will first locate in countries that are culturally close and may move to more distant countries later, after they gradually learn how to do business internationally (Johanson & Vahlne, 1977). Similar arguments stressing the costs of doing business abroad have been advanced by scholars following the economics perspective (Buckley & Casson, 1976; Ramachandran & Pant, 2010). Some recent research provides evidence for the limitations of this prediction, as companies seem to be motivated to enter culturally (and otherwise) distant host markets due to their strategic and economic appeal. For example, many emerging market firms from China, South Korea, and other Asian countries are boldly investing in Western (culturally distant) hosts to be closer to technology centers, strong competitors, and demanding customers who would help them develop further their innovation and organizational capabilities (BCG, 2014; Guillén & García-Canal, 2009; Luo & Tung, 2007). Although this work does not explicitly suggest that the large cultural distance is the reason for such location decisions, it implies that cultural-distance concerns can be outweighed by other factors that create benefits for the firm. Hence, it provides an argument for considering boundary and contingency conditions in studying cultural-distance effects on the process of firm internationalization.

The empirical evidence on cultural distance and location choice is mixed. Holburn and Zelner (2010) find a significant negative effect; Delios, Gaur, and Makino (2008) a significant positive effect; and Rose and Ito (2008) do not find any significant effect. Despite the broad interest in cultural distance and firm internationalization, location choice studies are relatively scarce, and there is no meta-analysis on this topic to date. Anecdotal evidence and consulting reports acknowledge cultural differences as a factor that should be taken into account when firms decide whether to enter a specific host country, but only after market size, growth opportunities, legal constraints, market stability and costs of production (KPMG, 2016). This is consistent with Sethi, Guisinger, Phelan, and Berg’s (2003: 319) observation that MNCs may “be compelled to ignore the greater cultural distance of developing countries in favor of their low-wage advantage.” More robust evidence is provided by Buckley et al. (2007), who show in a series of experiments that managers rank culture 16th in importance as a factor of foreign location choices (return on investment ranks 1st). All in all, the existing evidence on location choice suggests that cultural differences may be relevant to location choice, but only after key economic indicators suggest that a location is attractive.

Entry and establishment mode

The next step in the firm internationalization process concerns the decision about the specific organizational form of the operation. This literature distinguishes between entry mode and establishment mode (see Dikova & Brouthers, 2016, for an overview), with the former referring to JV versus WOS and the latter to acquisition versus greenfield (Brouthers & Hennart, 2007; Martin, 2013; Slangen & Hennart, 2007). The term entry mode often is used to refer to both (Klier, Schwens, Zapkau, & Dikova, 2017).

The primary theoretical perspective that has been employed in studying entry and establishment mode is transaction-cost economics (TCE; Williamson, 1985), with some variations depending on whether a JV should be classified as a form of hierarchical control (Hennart, 1988, 1991) or a hybrid organizational form between hierarchy and market (e.g., Anderson & Gatignon, 1986; Erramilli & Rao, 1990). In this view, the choice of a specific entry mode (JV vs. WOS is most commonly studied; Brouthers & Hennart, 2007) is based on the anticipated cost of transactions, which are in turn determined by the firm’s asset specificity (e.g., research and development intensity) or the uncertainty of the transaction (both internal uncertainty, such as international experience, and external uncertainty, such as country risk).[2] The transaction-cost perspective overall has provided high explanatory power to studying entry-mode decisions, as shown in a meta-analysis on the topic (Zhao et al., 2004).

In addition, some entry- and establishment-mode research has employed the resource-based view (RBV; Barney, 1991), which focuses on firm resources (e.g., experience) in explaining the choice between JV and WOS (Delios & Henisz, 2000; Madhok, 1997) and between acquisition and greenfield (Klier et al., 2017). In general, the RBV perspective on entry-mode choice suggests that the greater the resource base of the MNC, the higher the likelihood that it will select more complex organizational arrangements (Brouthers, Brouthers, & Werner, 2008; Brouthers & Hennart, 2007), a finding in line with the key predictions of the transaction-cost theory. In addition to TCE and RBV, entry mode studies have also used institutional theory (Martin, 2013), whereby the main idea has been that firms mimic others from their organizational class, that is, they select a particular entry mode because other firms in the same industry and/or country tend to use that entry mode (e.g., Lu, 2002). In a study combining the transaction-cost perspective with institutional theory, Yiu and Makino (2002) showed that both perspectives are robust in explaining firms’ preference for JV or WOS.

Theoretically, most of this work views cultural distance as a source of uncertainty, complexity, and additional costs (see Appendix A in the online supplement) and suggests that greater distance increases the need to collaborate with a local partner familiar with the host country culture, thus predicting a JV (Anderson & Gatignon, 1986). From a transaction-cost perspective, “cultural distance increases information asymmetry and consequently leads to increased monitoring costs. Accordingly, internalized foreign activities would be more efficient” (Morschett et al., 2010: 62), and further, “Transferring a company’s capabilities to a culturally dissimilar host country is difficult and it is linked to high learning costs in the unfamiliar environment. . . . A cooperative entry mode can serve as a risk-reduction strategy” (Morschett et al., 2010: 61). Therefore, cultural distance is associated with a JV rather than WOS entry mode.

Interestingly, the same theoretical perspective has been used to argue exactly the opposite (e.g., Hennart, 1988)—that when cultural distance is significant, firms should limit interaction with foreign partners and do it by themselves, that is, choose a WOS entry mode. High cultural distance increases uncertainty, and because of that, a firm may want to limit interaction and collaboration with a local partner. Postacquisition integration requires interaction between employees from different cultures, potentially causing conflict and misunderstandings (Reus & Lamont, 2009). Also, working with another partner “would involve ‘double-layered’ acculturation whereby the company expanding abroad would have to cope with the foreign culture of customers and, moreover, with the different corporate culture of a cooperative partner, thus enhancing complexity” (Morschett et al. 2010: 62; see also Barkema et al., 1996). When cultural distance is high, it is “difficult for MNCs to integrate into their corporate network acquisitions made in culturally distant countries, as the practices of MNCs and acquired firms are likely to be incompatible and difficult to transfer in such cases” (Drogendijk & Slangen, 2006: 365). In acquisitions, the acquired company may strongly resist knowledge transfer to the acquiring company (Hennart, 1991). This line of reasoning predicts a lower probability of acquisitions and a higher probability of greenfield investments and WOS when cultural distance increases. As Anderson and Gatignon (1986: 18) note, “transaction costs analysis suggests both views are correct.”

The empirical findings on cultural distance and entry and establishment mode are inconclusive. In a comprehensive review of culture research in international business, Kirkman et al. (2006) state that “the most glaring need . . . is to explain the conflicting findings regarding the effects of cultural distance on various organizational decisions such as entry mode choice” (Kirkman et al., 2006: 302). Specifically, Morschett et al. (2010) find no significant relation between cultural distance and entry mode, defined as cooperative (e.g., JV) versus WOS. Zhao et al. (2004) establish a small negative effect of cultural distance on entry mode operationalized as ownership mode (though it is unclear whether this refers to JV, WOS, acquisition, or greenfield. They also find that this effect is moderated by whether the reference country is the United States or not (Zhao et al., 2004: 531-532), which is in line with the earlier observation that sample structure may matter for cultural-distance effect. Other meta-analyses on cultural distance and mode choice show inconclusive results (Magnusson et al., 2008; Morschett et al., 2010; Reus & Rottig, 2009; Tihanyi et al., 2005). One particular challenge with entry mode studies, including these meta-analyses, is that entry mode choice is usually defined broadly and mode decisions are explained by estimating logistic models on several binary choices between modes. Martin (2013) observes that scholars compare not only JV versus WOS but for example also JV versus acquisition and JV versus greenfield, and combinations of these different modes. This is problematic to the extent that any finding on a possible determinant of entry- or establishment-mode choice (e.g., cultural distance) is “contingent on the heterogeneous aggregation or exclusion of some modes of entry” (Martin, 2013: 36). As a result, the reference category shifts across studies. We tackle this empirical challenge in our meta-analysis by clearly distinguishing between entry (JV vs. WOS) and establishment (greenfield vs. acquisition) mode.

Degree of ownership

Research on cultural distance and degree of ownership (or level of commitment) has been usually integrated with entry-mode studies and similarly has produced inconclusive findings. There appears to be no consensus regarding the effects of cultural distance on amount of capital invested (often operationalized by ownership share for cooperative entry modes). Some studies report a negative relationship, suggesting less ownership shares under large cultural distance (e.g., Malhotra, Sivakumar, & Zhu, 2011; Wilkinson, Peng, Brouthers, & Beamish, 2008), whereas others find a positive relationship (e.g., Padmanabhan & Cho, 1996). In their meta-analysis, Tihanyi et al. (2005) do not find a significant direct effect of cultural distance on the degree of ownership. We note though that in Tihanyi et al.’s study, the degree of ownership is pooled with other high-equity entry modes such as WOS, acquisition, and JV (Tihanyi et al., 2005: 274), making it hard to directly attribute these results to a particular measure of amount of capital invested.

Integration of foreign operation

Having decided on location, entry and establishment mode, and degree of ownership, MNCs need to address the question of how to manage the foreign operation, what is the proper governance arrangement between the parent company and the foreign unit that would provide the best integration, and coordination and control (Bartlett & Ghoshal, 1998; Kostova, Nell, & Hoenen, 2016). Different models require different levels of control and coordination between the headquarters and the subsidiary (Bartlett & Ghoshal, 1998; Prahalad & Doz, 1987); they vary with regard to allocation of assets and decision-making authority and the degree to which different units in the MNC use standardized organizational practices and structures (Kostova, Marano, & Tallman, 2015). Transfer of practices is an essential element in all MNC models, although the direction and the drive of this process might vary across models (Kostova, 1999). Although research on transfer of practices within MNCs has employed a number of theoretical perspectives, such as information-processing theory (Szulanski, 1996) and social-capital theory (Nahapiet & Ghoshal, 1998), the majority of the work in this area is based on institutional theory (Kostova, 1999; Powell & Dimaggio, 1991; Sanders & Tuschke, 2007; Scott, 1995).

Theoretically, the relationship between cultural distance and integration of the foreign subsidiary into the MNC is complex. On the one hand, cultural distance is expected to negatively affect the degree and ease of integration because it is associated with different organizational practices and ways of doing business at the parent company and the foreign operation, difficulties in communication due to language barriers and distinct communication patterns, and a general lack of trust between the two sides as a result of the perceptions of foreignness. Several studies in international management have theorized and proposed such negative effects on various aspects of integration, including control, coordination, transfer of practices, and agency relationships between headquarters and subsidiaries (Kostova, 1999; Kostova et al., 2016).

On the other hand, the strategic motivation for investing abroad when distances are considerable is often accompanied by a belief that the MNC possesses firm-specific competences that, if transferred to the foreign location, will create value or that it can learn from the host country and leverage its competences worldwide (Bartlett & Ghoshal, 1998). This could explain the paradox of emerging market firms investing aggressively in developed economies and, vice versa, developed economy MNCs investing boldly in distant and less developed countries where they see economic advantages and a potential benefit of organizational upgrades of the foreign operation.

Thus, on the one hand, cultural distance makes it more beneficial for the company to integrate the foreign operation through best practices and establishing organizational control and coordination systems; on the other hand, cultural distance makes such integration more challenging and difficult compared to locations that are culturally proximal. The empirical literature is reflective of this complex picture. Extant meta-analyses have not explored this aspect of firm internationalization. As concluded by Stahl and Voigt (2008: 161), “integration process variables . . . have not been examined with sufficient frequency in previous research to be considered” in their meta-analysis. Theoretically, it may be important to distinguish between the amount and benefits of transfers. Research would benefit if scholars could capture this distinction between the potential value of or need for integration versus the potential difficulty in achieving integration.

Performance

The dominant view in the literature is that cultural distance has negative performance consequences because of the complexity and uncertainty of doing business in a distant host country (see Appendix A in the online supplement). Complexity results in higher transaction, communication, coordination, and control costs as well as in increased difficulty to integrate the foreign operation through common practices (Kostova et al., 2016). Uncertainty further exacerbates such costs and risks and drives down a company’s commitment to a certain location. Recently, a few studies have suggested a positive effect of cultural distance due to the potential benefits of learning from a more distant counterpart that is likely to have different competences and capabilities and also more creative decision making (Gomez-Mejia & Palich, 1997; Morosini, Shane, & Singh, 1998). Reus & Lamont (2009) show that firms that have chosen to acquire a foreign firm and possess integration capabilities are able to mitigate the negative performance effects of cultural distance.

The empirical evidence on this relationship is mixed. Magnusson et al. (2008) report a small negative effect of cultural distance on performance. A meta-analysis of performance effects in international joint ventures (IJVs) shows that “empirical findings for a direct effect of cultural distance on IJV performance are inconclusive” (Reus & Rottig, 2009: 610). Tihanyi et al. (2005: 276) find that “the estimate of the multivariate relationship indicated that cultural distance was not meaningfully related to firm performance.” A possible reason for the inconclusive results regarding performance (besides sample size differences as suggested by Tihanyi et al., 2005) may be the fact that none of the extant meta-analyses have distinguished between the MNC and subsidiary level of analysis and very few (e.g., Reus & Lamont, 2009) have explored additional moderating conditions where the performance effect of distance turns positive.

Research Questions

In summary, our review of the literature on cultural distance and the process of firm internationalization shows that scholars have employed an “envelope” of theories and theoretical perspectives (transaction-cost theory, RBV, institutional theory; Dunning, 2000) to explain different outcomes associated with various aspects of the firm-internationalization process. Furthermore, the findings on cultural-distance effects have been inconclusive (positive, negative, or insignificant results for the same outcome) and the research approach has been typically partial and incomplete (e.g., focusing on only one stage as opposed to all stages, pooling firm and subsidiary performance, and/or pooling mode choices). In our effort to synthesize and further advance this literature, we address several research questions, some concerning the base relationship between cultural distance and various aspects of the firm internationalization process, others addressing additional contingences (moderating factors) that could help explain the inconclusive findings in past research. Under the broad research question of our study about the relationship between cultural distance and the process of firm internationalization, we address the following specific research questions:

Research Question 1: How does cultural distance affect the different stages of the firm internationalization process? Does the effect vary depending on the particular aspect of the internationalization process—location choice, entry and establishment mode, degree of ownership, and transfer of practices? Does the performance effect vary between subsidiary and MNC?

Research Question 2: Given the criticism on the measurement of cultural distance, do the relations uncovered under Research Question 1 depend on the particular operationalization and measurement of cultural distance used in the respective studies?

Research Question 3: Are the effects of cultural distance on the various aspects of internationalization contingent on the type of home and/or host country studied? Specifically, does the developed versus emerging market country condition moderate these relationships?

Research Question 4: Are cultural-distance effects stable or possibly diminishing over time, as a result of globalization and cross-country integration of the world economy and firms’ increasing international experience?

Methodology

Sample

To address our research questions, we conducted a meta-analytical study that followed recently established guidelines for developing rigorous meta-analytic research in management and international business (Buckley, Devinney, & Tang, 2013; Marano et al., 2016). To identify the highest number of articles investigating the effects of cultural distance on firm internationalization, we followed a sequence of five search strategies. First, we read several narrative reviews (e.g., Kirkman et al., 2006, Shenkar, 2001) and existing meta-analyses (Klier et al., 2017; Magnusson et al., 2008; Morschett et al., 2010; Reus & Rottig, 2009; Stahl & Voigt, 2008; Tihanyi et al., 2005; Zhao et al., 2004) concerning the relationship between cultural distance and aspects of the process of firm internationalization (none of these address the whole process in an integrated way). Second, we searched three major electronic databases (Business Source Complete, Google Scholar, and Web of Science) by using the following search terms: distance, cultural distance, cultural differences, and internationalization. Third, after the initial sample of studies was completed, we conducted a manual search in 15 journals across the disciplines of economics, management, and international business that have published articles on cultural distance, including: Journal of International Business Studies, Journal of Management, and Academy of Management Journal. Fourth, we continued our search by using the snowballing technique, which entails exploring references lists and Google Scholar citations of the articles in our initial pool. Finally, we reached out to researchers whose studies we had identified but we were not able to access through the above channels. This systematic approach reflects best practice for conducting meta-analysis because it minimizes the chance of missing important papers and increases the validity of the findings.

Our search process yielded a final dataset consisting of 156 studies published in the period 1988 to 2015 from various fields, including international business, strategy, human resource management, entrepreneurship, marketing, economics, and finance. We note that studies using country-level foreign direct investment (FDI) data were not included in the sample because our paper is about firm internationalization, which is difficult to derive from country- level statistics. As other scholars have pointed out, such country-level FDI studies do not specifically capture the foreign-value-adding activity of MNCs (Beugelsdijk, Hennart, Smeets, & Slangen, 2010).

A full list of all primary studies is included in Appendix B (see online supplement). Out of all 156 studies in the sample, 153 are published and three are working papers or doctoral dissertations. The primary studies published between 1988 and 2015 included in our sample cover the period 1968 to 2011, in which firms made internationalization decisions. These include both developed and emerging markets from all regions of the world. Our data concerning the cultural-distance performance relationship consist of 218,106 bivariate observations and 698,589 partial observations. This is a significant increase from the previous meta-analyses on the cultural-distance internationalization relationship by Tihanyi et al. (2005) based on 7,848 bivariate observations, Magnusson et al. (2008) based on 35,005 bivariate observations, Reus and Rottig (2009) with 22,460 bivariate correlations, and Stahl and Voigt (2008) with 9,396 bivariate observations. The larger sample size ensures the necessary statistical power to derive findings and implications for the various aspects of the internationalization process. We add to the previous literature by examining the distance effects on multiple outcomes related to internationalization, distinguishing between different entry and establishment modes, and examining performance impact at both subsidiary and MNC levels. Finally, we apply more advanced meta-analytical techniques leveraging the progress made in this area of research (Kirca & Yaprak, 2010; Stanley & Doucouliagos, 2012). One extension is that we use partial correlation as effect sizes, allowing us to incorporate samples from disciplinary results such as economics, in which Pearson product–moment correlations are not normally reported (van Essen, Heugens, Otten, & van Oosterhout, 2012) and control for the influence of the control variables contained in the z-vector (Marano et al., 2016). Table 1 summarizes the differences between our study and previous similar meta-analyses, including the work on foreign market entry mode (Morschett et al., 2010; Zhao et al., 2004), performance (Reus & Rottig, 2009; Stahl & Voigt, 2008), and entry mode and performance (Tihanyi et al., 2005; Magnusson et al., 2008).

Table 1

Comparison of Meta-Analyses on Firm-Level Consequences of Cultural Distance

Variable	Our Meta-Analysis	Zhao et al. (2004)	Tihanyi et al. (2005)	Magnusson et al. (2008)	Stahl and Voigt (2008)	Reus and Rottig (2009)	Morschett et al. (2010)
No. studies included / k	156 / 437	14 / 15	55 / 66	61 / 72	16 / 31	40 / 37	14 / 37
Effect size data	Pearson’s r and partial correlation rxy.z	Pearson’s r	Pearson’s r	Pearson’s r	Pearson’s r	Pearson’s r	Vote count
Time window	1988–2015	1988–2002	1992–2002	1991–2005	NA	1997–2007	1992–2008
Location choice	r- and rxy.z-based mean: −.023**	NA	NA	NA	NA	NA	NA
Scale of investment	r- and rxy.z-based mean: –.006				NA	NA	NA
Entry mode	r- and rxy.z-based mean: .003	r-based mean: −.029***	r-based mean: −.064	r-based mean: −.036*	NA	NA	Vote count mean: –.473 (ns)
Establishment mode	r- and rxy.z-based mean: −.050***	NA	NA	NA	NA	NA	NA
Amount of practices transferred	r- and rxy.z-based mean: .011	NA	NA	NA	NA	NA	NA
Benefit of practices transferred	r- and rxy.z-based mean: .148***	NA	NA	NA	NA	NA	NA
Performance	r- and rxy.z-based mean: -.032***	NA	r-based mean: −.035	r-based mean: −.040**	r-based mean: .01 (ns)r-based mean: −.03**	r-based mean: −.028***	NA
Moderators influencing the effect of CD on DVs	(Single) home countryU.S.; developed; emerging(Single) host countryU.S.; developed; emergingFirm identitySubsidiary; MNCTimeUntil median year; after median year	Host countryU.S.; non- U.S.Home countryU.S.; non- U.S.Industry typeManufacturingService; non-specified	Home countryU.S.; non- U.S.Host countryDeveloped; developingIndustryHigh tech; othersTime1980s; 1990s	Home countryU.S.; non- U.S.; Europe; AsiaTimePrior to 1990; 1990 to1995; after 1995	Degree of (industry) relatednessLow; medium; high	Host countryChina; non-China	Industry typeService; manufacturingTimeEarly; late
Methodological artifacts	Published study; median year; panel design; endogeneity check	NA	NA	NA	NA	NA	NA
Model specification artifacts	Study controls for other distances; study controls for performance; partial correlation dummy	NA	NA	NA	NA	NA	NA
Statistical artifacts: CD	KSI/Euclidean distance; Mahanalobis distance; cultural zone distance; perceptual distance; other distance operationalizationHofstede data (plus separation of dimensions); GLOBE data; Schwartz; Trompenaars; Ronen and Shenkar data; other data source	Secondary data; survey	Euclidean distance; other	Individual measurement; national measurement	National; organizational	KSI; subjective CD; other	NA
Statistical artifacts: DVs	PERFAccounting performance; market performance; survey performance; survival; innovation; other	ACI/EMBinary; equity ownership; categorical	NA	NA	PERFAnnouncement effects; longer-term effects; target firms; acquiring firms	PERFObjective performance; subjective performance	NA

Note: ACI/EM = amount of capital invested/entry mode; CD = cultural distance; DV = dependent variable; GLOBE = Global Leadership and Organizational Behavior Effectiveness; KSI = Kogut and Singh cultural-distance index; MNC = multinational company; NA = not available or not tested; PERF = performance; U.S. = United States.

p < .10.

p < .05.

p < .01.

We proceeded by reading all articles and by developing a coding protocol (Lipsey & Wilson, 2001) to extract data on all relevant variables and study characteristics. Two authors coded all the data, while a third author coded a subsample of 270 randomly selected effect sizes to assess the degree of agreement in terms of extracting information from primary studies (Stanley & Doucouliagos, 2012). We had a high degree of interrater agreement—(Cohen’s kappa of 0.98; Cohen, 1960).

Meta-Analytic Procedure

We used two methodological procedures, Hedges-Olkin-type meta-analysis (HOMA) and meta-analytic regression analysis (MARA), which help achieve distinct analytical objectives.

HOMA procedure

We use HOMA in order to determine the mean size of the effect of cultural distance on the outcomes associated with the different stages of internationalization. We used Pearson product–moment correlations (r) and partial correlation coefficients (r) as effect sizes. The latter represents the relationship between those variables when keeping a certain set of variables (z) constant. Like r, r is an easily interpretable and scale-free measure of linear association. It can be computed from the t statistics and degrees of freedom reported in the primary studies (Greene, 2003). We performed our computations using random-effects HOMA, which accounts for potential heterogeneity in the effect size distribution and is more conservative than fixed-effects HOMA (Kisamore & Brannick, 2008; Raudenbush & Bryk, 2002).

When multiple measurements of the focal effect were reported in one study (for example, due to the reporting of results for different operationalizations of cultural distance), we included all of them in our analyses. Monte Carlo simulations show that procedures using the complete set of measurements outperform those representing each study with a single value in areas like parameter significance testing and parameter estimation accuracy (Bijmolt & Pieters, 2001). To accurately account for differences across effect sizes, we weighted each effect size by its inverse variance weight w, the inverse of the squared standard error (Hedges & Olkin, 1985).[3] Next, we used these weights to compute the standard error of the mean effect size and its corresponding confidence interval.[4]

MARA procedure

We use MARA to test the robustness of our model against a number of control variables. In the MARA analyses, the dependent variable is neither cultural distance nor any of the independent variables (e.g., entry mode or performance) but an estimate of the associational strength of the focal relationship in a given sample (e.g., cultural distance and performance), such that all independent variables in the regression equation are modeled as moderators of the focal relationship (van Essen, Otten, & Carberry, 2015). MARA is a weighted least squares technique, which seeks to model previously unexplained variance in the effect size distribution (Lipsey & Wilson, 2001). We used weighted regression to account for differences in precision across effect sizes. The statistically preferable weighting variable is, once again, w (Hedges & Olkin, 1985).

Following current standards in the meta-analytic literature (Geyskens, Kirshnan, Steenkamp, & Cunha, 2009), we used random-effects estimation methods in the MARA analyses, which are more conservative than conventional fixed-effects methods. Specifically, this yielded the following regression equation:

where R is the correlation between cultural distance and each of the outcomes for the different stages of the firm internationalization process (i.e., location choice, entry mode, establishment mode, degree of ownership, transfer of practices, and performance), y0 is the constant term, D is a vector of measurement artifacts, S is a vector of methodological study characteristics, R is the set of firm characteristics, and u is the random component.

Operationalizing Firm Internationalization and Cultural Distance

Stages of internationalization

As described above, primary studies have related cultural distance to various decisions associated with the firm internationalization process. Consistent with the literature, we operationalize them in the following way:

Location choice (Dunning & Lundan, 2008; Rugman & Verbeke, 2009), that is, in which host country to invest. The choice to invest in a country is typically measured using a binary variable, with the MNC-host country-year as the unit of analysis. The variable takes the value of 1 if the MNC invests in a certain host country in a given year and 0 otherwise. Since the unit of analysis is the MNC-host country-year, the primary studies focusing on the choice to invest are based on a sample size that is considerably higher than that of other studies;

Entry mode, operationalized through a binary variable, which is equal to 1 when the MNC opts for a wholly-owned foreign subsidiary (WOS) and to 0 when it chooses a joint venture (JV) with a local or international partner;

Establishment mode (e.g., Brouthers, 2002; Kogut & Singh, 1988), that is, whether the company enters the foreign market through acquisition or greenfield investment. Following extant literature (e.g., Barkema & Vermeulen, 1998; Slangen, 2011), we operationalize investment mode through a dummy variable, taking the value of 1 for acquired subsidiaries and 0 for those established through greenfield investments;

Degree of ownership (e.g., Chan & Makino, 2007), that is, the size of the foreign investment, which determines the level of commitment (Ghemawat, 1991) in the host country. The scale of investment is rarely measured in absolute terms, that is, in terms of the absolute amount of capital employed by the MNC when investing in a certain host country. Consequently, we use a proxy that captures scale of investment in relative terms, that is, the equity stake of the parent company in the foreign investment (e.g., Chan & Makino, 2007; Xu, Pan, & Beamish, 2004);

As discussed above, we operationalize the integration of foreign operations as both the amount of practices transferred to the foreign subsidiary and the benefits of the practice transfer. The amount of practices transferred is measured by (a) whether a transfer event has occurred (e.g., Hansen & Løvås, 2004; Xia, 2011), (b) the number of transfers (e.g., Drogendijk & Slangen, 2006; Slangen, 2011), and (c) the actual amount of transferred practices, such as those incorporated in the patents of an acquired subsidiary (e.g., Ahuja & Katila, 2001). The benefit of the practice for the recipient foreign subsidiary is measured as the unit’s perceived organizational learning as a result of the transfer (e.g., Lane, Salk, & Lyles, 2001; Minbaeva, Pedersen, Björkman, Fey, & Park, 2003; Sarala & Vaara, 2010);

Firm performance. For a broader account of the internationalization strategy, we examine performance effects at the multinational enterprise (MNE) and the subsidiary level (e.g., Barkema et al., 1996). Specifically, we use (a) accounting performance, including return on assets (ROA), return on investment (ROI), return on sales (ROS), and return on equity (ROE) (e.g., Barkema & Vermeulen, 1998; Luo, 2005); (b) market performance, including earnings per share, market to book value, Tobin’s Q, and cumulative abnormal returns on the stock (e.g., Aybar & Ficici 2009; Reuer, 2001); (c) subsidiary longevity (e.g., Lu & Beamish, 2006) or survival (e.g., Delios & Beamish, 2004); (d) innovation performance reflected in the innovation output of the firm, for example in terms of patents (e.g., Ahuja & Katila, 2001). All other measures of performance (e.g., sales growth, market share) are included in the Other category.

Cultural distance

Since cultural distance has been measured in different ways, we distinguish between the various measures and data sources. We test for a possible moderating effect of the operationalization and measurement approach by creating dummy variables indicating whether cultural distance was measured through one of the following measures:

Kogut and Singh’s (1988) cultural-distance index (KSI), measured as the Euclidean distance (using normalized scores on culture dimensions), that is, the square root of the sum of the squared differences in cultural value dimensions between home and host country. We coded this dummy as 1 when a study used this measure of cultural distance, and 0 otherwise. Typically, KSI is based on the four dimensions of Hofstede’s (1980) culture framework.

Mahanalobis distance, introduced in the distance literature by Berry, Guillen, and Zhou (2010). This measure, unlike the Euclidean distance, takes into account the correlation between the cultural dimensions used in the measurement. In the absence of correlation between the culture dimensions, this measure is identical to KSI based on Euclidean distance (Beugelsdijk, Nell, et al., 2017). The dummy takes the value of 1 when the Mahalanobis technique is used to calculate cultural distance.

A dummy variable indicating whether the host country is located in a cultural cluster different from the home country of the firm. Typically, studies that use this approach rely on the cultural clusters identified by Ronen and Shenkar (1985, 2013). We coded this dummy as 1 when a study used cultural clusters to measure cultural distance.

Perceptual (or “psychic”) distance, which is managers’ perception of the cultural distance between home and host country. This measure typically employs primary data collected through questionnaires among managers involved in the internationalization process and does not involve scores from both home and host country. We coded the dummy as 1 if a study used perceptual measures of cultural distance.

Other measures of distance include, for example, stepwise cultural zone distance (Barkema et al., 1996) and sum of cultural distance between the home country and the host countries weighted by number of subsidiaries in each host country (Beamish & Kachra, 2004). The dummy takes the score 1 if such other operationalizations of cultural distance are used.

Cultural-distance data source

We also examine the impact of the source of cultural-distance data sources used by the primary studies in our sample. For an extensive description of the dimensions included in each of these frameworks, we refer to the original publications and overviews, such as Kirkman et al. (2006). Specifically,

Most studies rely on the cultural framework developed by Hofstede (1980). In his study of how values in the workplace are influenced by culture, Hofstede analyzed a large amount of primary data collected at IBM between the late 1960s and early 1970s and identified the following cultural dimensions: power distance, individualism, masculinity, uncertainty avoidance, and long-term orientation. Hofstede, Hofstede, and Minkov’s (2010) recent addition of a sixth dimension (indulgence versus restraint) is too recent to have been included in primary studies considered. We would also note that the correlation between the fifth and sixth dimensional distance metric is very high.

National scores on cultural dimensions from the GLOBE Project (House et al., 2004). The cultural dimensions identified in the study are performance orientation, assertiveness, future orientation, humane orientation, institutional collectivism, in-group collectivism, gender egalitarianism, power distance, and uncertainty avoidance.

National scores on cultural dimensions based on Schwartz (1994, 1999, 2004). The author identifies three key issues that societies confront and derives three corresponding dimensions for cross-country cultural analysis: embeddedness versus autonomy, hierarchy versus egalitarianism, and mastery versus harmony.

Trompenaars (1993) developed a framework that includes seven cultural dimensions: universalism, individualism, neutral vs. affective, specific vs. diffuse, achievement vs. ascription, attitudes with regard to time, attitudes with regard to the environment. Although these data are not publicly available, they have been included in a small subset of studies.

Cultural clusters identified by Ronen and Shenkar (1985, 2013). These authors reviewed and synthesized eight studies on cross-country cultural differences and identified eight relatively distinct cultural clusters: Anglo-Saxon, Germanic, Nordic, Latin European, Latin American, Near East, Far East, Arabic.

Primary data, which overlaps with the perceptual measurement. These data refer to surveys in which managers are asked to indicate the (perceived) cultural distance to a particular country. These data are study specific (e.g., Luo, 2002).

Control variables

When performing the MARA analysis, we included several control variables that have been continuously raised by the critics, aimed to account for the effect of various artifacts on the relationships of interest.

We controlled for the moderating effect of firm identity on the effect of cultural distance on performance. As discussed in the Theory section, there is a reason to believe that the effect of cultural distance on performance differs between subsidiary and MNC.

In order to test for the moderating effect of methodological artifacts, we controlled, first, for the “file drawer problem” (Meyer, van Witteloostuijn, & Beugelsdijk, 2017; Rosenthal, 1979), by including a dummy variable denoting whether a study was published (1) or not (0). Our sample predominantly includes published studies which may limit the possibility to detect selection bias. However, the file drawer problem does not appear to affect correlation tables in published versus unpublished papers (Dalton, Aguinis, Dalton, Bosco, & Pierce, 2012), and since we provide both the results of the bivariate as well as the partial correlation coefficients, we have no reason to suspect a major bias of our result because of the selection bias. Second, we controlled for the sample median year to test whether the base relationship has changed over time. Third, we included a panel (1) or cross-sectional (0) data dummy. Fourth, we included an endogeneity check dummy to test if endogeneity is driving our results or not, taking value of 1 if the effect is estimated while controlling for potential endogeneity or not (0).

Since a significant part of our sample is based on U.S. companies, and it has been suggested that using a developed country, specifically the United States, as a single reference country may affect the results, we included a dummy that takes value of 1 when cultural distance is measured from or to the United States and 0 otherwise.

We included a dummy variable indicating whether the home or host country is developed or an emerging market.

We also controlled for model specification artifacts, which are all dummy variables. Specifically, we controlled for whether the effect is measured as a partial (1) or a bivariate correlation (0). Two dominant extensions of the cultural-distance construct are the CAGE framework (Ghemawat, 2001) and the institutional distance construct (Kostova, 1999). In order to control for potential effects of alternative types of distance, we included in the MARA analyses a binary variable taking value of 1 when the primary study includes other types of distance (i.e., economic, institutional/administrative, or geographic) in the estimated models. We also controlled for whether the primary study includes other performance controls, normally lagged performance measures.

Results

HOMA Results

Tables 2 through 10 show results of our HOMA. We only show the bivariate and partial correlation coefficients when the number of effect sizes is based on a minimum number of effect size (k) of 3 (Lipsey & Wilson, 2001) consisting of at least two studies (Valentine, Pigott, & Rothstein, 2010). Table 2 reports the results of a number of r- and r-based HOMA analyses of the effect of cultural distance on the decision to invest in a foreign country (location choice). We find that cultural distance has a negative and statistically significant effect on the choice to invest in a particular host country (mean effect size = −0.02, p = .03). Our distinction between measurement techniques shows that this negative relation is driven by two studies using the Mahalanobis technique to calculate cultural distance (Berry et al., 2010; Zhou & Guillén, 2015). For the Hofstede-based studies using the standard Kogut and Singh index of cultural distance, we find no significant effect on location choice. The use of the Mahalanobis technique is fairly recent. It is thus no surprise that the relationship between cultural distance and location choice becomes more negative over time. As the number of studies that have used the Mahalanobis technique is still very limited, we interpret this result with care. More location choice studies applying the Mahalanobis technique are required to corroborate this finding.

Table 2

HOMA Meta-Analytic Results: Cultural Distance to Location Choice

	Pearson Product–Moment Correlation (r) and Partial Correlation Coefficients (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I 2
Cultural distance to location choice	34	2,441,680	−.02 (.03)**	.01	8086.78***	1.00
Measurement of cultural distance
Kogut and Singh index	26	1,147,466	−.02 (.20)	.02	5663.34***	1.00
Mahanalobis	8	1,294,214	−.04 (.03)**	.02	2328.68***	1.00
Cultural-distance data source
Hofstede	28	1,651,546	−.02 (.08)*	.01	7306.44***	1.00
Time
Until medium year	21	618,377	.00 (.67)	.01	284.89***	0.93
After medium year	13	1,823,303	−.05 (.00)***	.02	4731.84***	1.00

Note: Location choice is measured as the 0/1 measure to invest in a particular country. I2 = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect size; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

Table 10

MARA Meta-Analytic Results: Cultural Distance to Performance

Variable	Model 1	Model 2	Model 3
Measurement of cultural distance
Kogut and Singh index	−.25 (.00)***		−.36 (.00)***
Mahalanobis distance	−.32 (.00)***		−.37 (.00)***
Dummy variable	−.24 (.00)***		−.05 (.21)
Perceptual measures	−.45 (.00)***		−.44 (.00)***
Absolute distance	−.26 (.00)***		−.37 (.00)***
Other (reference group)
Cultural-distance data source
Hofstede		.20 (.00)***	.12 (.17)
Berry		.12 (.02)**	.04 (.78)
GLOBE		.27 (.00)***	.20 (.03)**
Ronen and Shenkar		.20 (.00)***	−.20 (.06)*
Primary data (reference group)
Firm performance definition
Accounting measures	.09 (.00)***	.08 (.00)***	.08 (.00)***
Market performance	.05 (.06)*	.10 (.00)***	.02 (.48)
Survey measures	.04 (.01)**	.06 (.00)***	.03 (.10)
Survival	.07 (.00)***	.10 (.00)***	.08 (.00)***
Innovation	.10 (.00)***	.06 (.04)*	.11 (.00)***
Other (reference group)
Performance evaluation
MNC	−.13 (.00)***	−.14 (.00)***	−.13 (.00)***
Subsidiary	−.15 (.00)***	−.15 (.00)***	−.15 (.00)***
Methodological study artifacts
Published study	.03 (.14)	.04 (.05)*	.03 (.19)
Median year of sample window	.00 (.67)	.00 (.30)	−.00 (.25)
Panel design	.01 (.73)	.05 (.01)***	−.02 (.18)
Endogeneity check	.00 (.92)	−.06 (.02)**	.01 (.71)
Home country type
Developed markets	−.01 (.77)	−.03 (.12)	.00 (.96)
Emerging markets	.04 (.11)	−.00 (.93)	.05 (.06)*
Host country type
Developed markets	.07 (.01)**	.03 (.31)	.06 (.03)**
Emerging markets	−.05 (.04)**	−.09 (.00)***	−.04 (.08)*
Model specification artifacts
Distance controls	−.01 (.12)	−.02 (.07)*	−.00 (.70)
Performance controls	−.06 (.02)**	−.11 (.00)***	−.07 (.00)***
Partial correlation	−.00 (.88)	.02 (.21)	−.01 (.44)
k	437	437	437
r 2	.29	.23	.32
Qmodel (p)	420.93 (.00)	311.74 (.00)	473.10 (.00)
Qresidual (p)	1015.56 (.00)	1064.04 (.00)	997.14 (.00)
V(*100)	.66	.71	.64

Note: The table shows estimated coefficients and p values in parentheses. GLOBE = Global Leadership and Organizational Behavior Effectiveness; MARA = meta-analytic regression analysis; MNC = multinational corporation.

p < .10.

p < .05.

p < .01.

Table 3 reports the results of the r- and r-based HOMA analyses of the effect of cultural distance on entry mode decision. We find that overall the relationship between cultural distance and entry mode decision is not statistically significant. However, this result varies across cultural-distance data sources. Specifically, results based on Hofstede’s data on four cultural dimensions suggest a negative and statistically significant effect of cultural distance on the likelihood of WOS (mean effect size = −0.02, p = .06), whereas results based on GLOBE’s (mean effect size = 0.08, p = .001) and Schwartz’s (mean effect size = 0.17; p = .00) data show a positive and statistically significant relationship. The effect of cultural distance changes over time, being negative and statistically significant in earlier years and positive and statistically significant in more recent years. This change in effect over time coincides with the use of GLOBE and Schwartz (versus the use of Hofstede) in more recent years. The number of studies that have unpacked the overall Hofstede-based cultural-distance measure in its different cultural dimensions is limited. The findings do suggest that especially the individualism–collectivism dimension drives the negative overall effect of cultural distance. This is not surprising given the generally acknowledged relevance of individualism as one of the key dimensions of national culture (Triandis, 1995).

Table 3

HOMA Meta-Analytic Results: Cultural Distance to Entry Mode

	Pearson Product–Moment Correlation (r) and Partial Correlation Coefficients (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I 2
Cultural distance to entry mode	119	92,923	.01 (.81)	.01	931.57***	.87
Measurement of cultural distance
Kogut and Singh index	86	80,022	−.01 (.24)	.01	686.78***	.88
Cultural-distance data source
Hofstede	99	74,347	−.02 (.15)	.01	796.19***	.88
Four dimensions	69	60,135	−.02 (.06)*	.01	401.99***	.83
Five dimensions	7	3,370	.01 (.89)	.10	197.15***	.97
Power distance dimension	5	2,221	−.03 (.67)	.07	34.09***	.88
Uncertainty avoidance dimension	5	2,221	.02 (.32)	.02	7.96*	.50
Individualism dimension	5	2,221	−.10 (.05)**	.05	17.95***	.78
Masculinity dimension	5	2,221	.01 (.95)	.04	12.00***	.67
GLOBE	14	17,244	.08 (.00)***	.02	85.45***	.85
Schwartz	5	1,194	.17 (.00)***	.03	6.75	.41
Time
Until medium year	63	36,495	−.06 (.001)***	.017	537.90***	.88
After medium year	56	56,428	.07 (.000)***	.01	258.37***	.79

Note: Entry mode is operationalized as wholly owned investment taking a 1 (joint venture = 0). Results for perceptual measures and primary data are based on similar primary studies; for reasons of completeness, we have included them in both the measurement and the data category. GLOBE = Global Leadership and Organizational Behavior Effectiveness; I2 = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect sizes; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

Table 4 reports the results of the r- and r-based HOMA analyses of the effect of cultural distance on establishment mode. Consistent with the extant literature (e.g., Barkema & Vermeulen, 1998; Kogut & Singh, 1988), we find a negative and statistically significant effect of cultural distance on the likelihood of acquisition (mean effect size = −0.05, p = .00). This result is consistent when using perceptual measures (mean effect size = −0.10, p = .01). These negative effects become insignificant when Schwartz data are used (mean effect size = −0.08, p = .40).

Table 4

HOMA Meta-Analytic Results: Cultural Distance to Establishment Mode

	Pearson Product–Moment Correlation (r) and Partial Correlation Coefficients (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I 2
Cultural distance to establishment mode	95	46,184	−.05 (.00)***	.01	735.30***	.87
Measurement of cultural distance
Kogut and Singh index	67	36,261	−.07 (.00)***	.02	550.61***	.88
Dummy variable	16	6,528	−.01 (.79)	.03	65.06***	.77
Perceptual measures	3	630	−.10 (.01)**	.04	5.62*	.64
Cultural-distance data source
Hofstede	66	36,882	−.07 (.00)***	.02	557.97***	.88
Schwartz	6	1,232	−.08 (.40)	.09	50.04***	.90
Primary data	3	630	−.10 (.01)**	.04	5.62*	.64
Ronen and Shenkar	16	6,528	−.01 (.79)	.03	65.06***	.77
Time
Until medium year	53	37,437	−.05 (.00)***	.02	439.51***	.88
After medium year	42	8,747	−.05 (.07)*	.03	295.23***	.86

Note: Establishment mode is operationalized as acquisition taking a 1 (greenfield = 0). Results for perceptual measures and primary data are based on similar primary studies; for reasons of completeness, we have included them in both the measurement and the data category. I = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect size; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

Table 5 reports the results of the r- and r-based HOMA analyses of cultural-distance effect on degree of ownership. We find no significant relationship between the two, and this finding is stable across different cultural-distance measures and data sources.

Table 5

HOMA Meta-Analytic Results: Cultural Distance to Degree of Ownership

	Pearson Product–Moment Correlation (r) and Partial Correlation Coefficients (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I 2
Cultural distance to degree of ownership	90	463,008	−.01 (.49)	.01	1930.77***	.95
Measurement of cultural distance
Kogut and Singh index	58	444,796	−.01 (.60)	.01	1845.58***	.97
Dummy variable	16	13,816	.00 (.97)	.01	31.96***	.53
Perceptual measures	12	1,476	.00 (.94)	.05	32.72***	.66
Cultural-distance data source
Hofstede	61	341,295	−.01 (.48)	.01	1826.10***	.97
Four dimensions	47	315,066	.00 (.97)	.01	1314.07***	.96
Five dimensions	4	721	.05 (.16)	.04	1.23	.00
Uncertainty avoidance dimension	4	7,472	−.08 (.30)	.08	84.38***	.96
Individualism dimension	4	7,472	−.04 (.75)	.12	214.59***	.99
Primary data	12	1,476	.00 (.94)	.05	32.72***	.66
Ronen and Shenkar	16	13,816	.00 (.97)	.01	31.96***	.53
Time
Until medium year	46	392,760	−.002 (.83)	.01	1240.67***	.96
After medium year	44	70,248	−.009 (.53)	.02	591.12***	.93

Note: Degree of ownership measures the size of the foreign investment. Results for perceptual measures and primary data are based on similar primary studies; for reasons of completeness, we have included them in both the measurement and the data category. I2 = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect size; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

Table 6 reports the results of the r- and r-based HOMA analyses of the effect of cultural distance on amount of practice transfer, showing no statistically significant relationship overall (mean effect size = 0.01, p = .44). However, we find variation depending on the particular cultural-distance measures used. Specifically, Hofstede-based measures show a positive and statistically significant effect of cultural distance on amount of practice transfer (mean effect size = 0.04; p = .001), whereas perceptual measures show a strong negative relationship (mean effect size = −0.61; p = .02). It should be noted, though, that the results for perceptual measures are based on only two studies (Cho & Lee, 2004; Drogendijk & Slangen, 2006). Furthermore, as shown in Table 7, it seems that the opportunity perceived in cultural distance turns into actual benefits for MNEs. The results of the r- and r-based HOMA analyses show that cultural distance has a positive and statistically significant effect on benefits of practice transfer (mean effect size = 0.15, p = .00), and these results are consistent across cultural-distance data and over time. The effect size is also very high, suggesting a strong relationship between cultural distance and the benefits of practice transfer.

Table 6

HOMA Meta-Analytic Results: Cultural Distance to Amount of Transfer

	Pearson Product–Moment Correlation (r) and Partial Correlation Coefficients (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I 2
Cultural distance to amount of transfers	47	171,990	.01 (.44)	.04	1522.60***	.97
Measurement of cultural distance
Kogut and Singh index	42	171,420	.04 (.00)***	.01	1287.00***	.97
Perceptual measures	5	570	−.61 (.02)**	.27	147.56***	.97
Cultural-distance data source
Hofstede	40	170,928	.04 (.00)***	.01	1285.73***	.97
Four dimensions	37	157,848	.05 (.00)***	.01	1210.22***	.97
Primary data	5	570	−.61 (.02)**	.27	147.56***	.97
Time
Until medium year	32	153,792	.05 (.00) ***	.02	1199.32***	.97
After medium year	15	18,198	−.13 (.00)***	.04	259.01***	.95

Note: Amount of transfers deals with the amount of knowledge that has been transferred or acquired. Results for perceptual measures and primary data are based on similar primary studies; for reasons of completeness, we have included them in both the measurement and the data category. I = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect size; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

Table 7

HOMA Meta-Analytic Results: Cultural Distance to Benefits of Transfers

	Pearson Product–Moment Correlation (r) and Partial Correlation Coefficients (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I 2
Cultural distance to benefits of transfers	18	3,589	.15 (.00)***	.03	50.50***	.66
Measurement of cultural distance
Kogut and Singh index	15	3,232	.17 (.00)***	.03	34.92***	.60
Perceptual measures	3	357	−.02 (.48)	.05	3.88	.48
Cultural-distance data source
Hofstede	5	1,380	.08 (.00)***	.03	8.90	.55
GLOBE	10	1,852	.23 (.00)***	.02	7.55	.00
Primary data	3	357	−.02 (.54)	.05	3.88	.48
Time
Until medium year	12	2,008	.20 (.00)***	.03	26.88	.59
After medium year	6	1,581	.08 (.03)**	.03	9.02	.45

Note: Benefits of transfers deals with the degree to which a (knowledge) transfer has been beneficial for the vocal entity. Results for perceptual measures and primary data are based on similar primary studies; for reasons of completeness, we have included them in both the measurement as well as the data category. GLOBE = Global Leadership and Organizational Behavior Effectiveness; I2 = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect size; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

The r- and r-based HOMA results for the relationship between cultural distance and firm performance are reported in Table 8. We find that cultural distance has a negative and statistically significant effect on firm performance (mean effect size = −0.03, p = .00). The variance in effect size distribution is substantial (Q = 7126.47, I2 = 0.94), suggesting that the mean effect is best interpreted as an average rather than a common true correlation value, implying that further robustness analyses are needed.

Table 8

HOMA Meta-Analytic Results: Cultural Distance to Performance

	Pearson Product–Moment Correlation (r) and Partial Correlation Coefficients (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I2
Cultural distance to performance	437	913,260	−.03 (.00)***	.01	7126.47***	.94
Measurement of cultural distance
Kogut and Singh index	267	821,834	−.04 (.00)***	.01	4140.89***	.94
Mahalanobis distance	6	42,269	−.02 (.00)***	.01	2.98	.00
Dummy variable	51	8,199	−.05 (.03)**	.02	212.29***	.76
Perceptual measures	40	7,673	−.21 (.00)***	.05	639.67***	.94
Cultural-distance data source
Hofstede	324	839,576	−.02 (.00)***	.01	5885.53***	.95
Four dimensions	240	815,150	−.03 (.00)***	.01	4061.10***	.94
Five dimensions	20	5,444	−.08 (.00)***	.02	30.86**	.38
Power distance dimension	11	2,575	−.02 (.28)	.02	9.85	.00
Uncertainty avoidance dimension	14	3,409	−.04 (.03)**	.02	18.24	.29
Individualism dimension	12	3,305	−.00 (.93)	.03	26.83***	.59
Masculinity dimension	11	2,575	−.04 (.26)	.03	20.26**	.51
GLOBE	9	3,680	.02 (.71)	.04	23.94***	.67
Ronen and Shenkar	57	12,993	−.04 (.04)**	.02	255.48***	.78
Trompenaars	2	264	−.21 (.00)***	.06	2.99*	.67
Primary data	40	7,673	−.21 (.00)***	.05	639.67***	.94
Performance types
Accounting performance	60	84,578	−.02 (.03)**	.01	311.21***	.81
Market performance	72	17,232	.03 (.38)	.04	1595.45***	.96
Survey performance	119	130,697	−.05 (.00)***	.01	477.07***	.75
Survival	95	410,861	−.05 (.00)***	.01	986.74***	.90
Innovation	39	176,750	.03 (.06)*	.02	2016.16***	.98
Other	52	93,142	−.11 (.00)***	.02	975.42***	.95
Firm identity
MNC	157	303,590	.02 (.11)	.01	4369.65***	.96
Subsidiary	231	569,163	−.07 (.00)***	.01	2610.52***	.91
Home country type
United States	52	57,951	−.03 (.29)	.03	2318.39***	.98
Developed markets	185	651,779	−.04 (.00)***	.01	3605.40***	.95
Emerging markets	31	19,152	.01 (.42)	.02	83.18***	.64
Host country type
United States	21	5,667	.06 (.06)*	.03	101.53***	.80
Developed markets	41	12,224	.04 (.09)*	.02	189.31***	.79
Emerging markets	109	28,214	−.11 (.00)***	.02	964.90***	.89
Time
Until medium year	234	766,672	−.02 (.00)***	.01	3429.88***	.93
After medium year	203	146,588	−.04 (.00)***	.01	3546.59***	.94

Note: GLOBE = Global Leadership and Organizational Behavior Effectiveness; I2 = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect size; MNC = multinational corporation; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

Table 8 also reports robustness tests of the HOMA results. With the exception of the GLOBE data, the results are similar across different measures and data sources of cultural distance as well as over time. We would note that the effect size for the distance measure based on Trompenaars’ framework is large compared to the other data sources, but that this should be interpreted with care given the limited number of studies using Trompenaars. Furthermore, results are largely robust across different performance measures, and most of the subsample analyses yield effect sizes consistent with the overall mean. We observe a very large effect size for perceptual measures compared to the nonperceptual measures (−.21 versus −.04 for the Kogut-Singh index). A similar result for perceptual measures on performance was obtained by Reus and Rottig (2009) in their meta-analysis of performance of international JVs.

The only two performance measures that do not show a significant negative relationship with cultural distance are market performance and innovation. While the mean effect size is not significant for market performance, the mean effect size for innovation is positive and statistically significant (mean effect size = 0.03, p = .06). Interestingly, this result seems consistent with our findings about the effect of cultural distance on the amount and benefit of practice transfer and suggests that cultural distance may represent an opportunity for organizational learning and, as a consequence, the innovativeness of the firm.

One unexpected finding that HOMA analysis reveals is that cultural distance impacts only subsidiary performance and not the performance of the whole MNC. This differential effect suggests that the risks and costs associated with investments to culturally distant countries may be offset by the overall benefits of internationalization, which seem to be reaped at the level of the MNC as opposed to the level of a specific host-country subsidiary.

Furthermore, we find that cultural distance has a negative effect on performance (mean effect size = −0.11, p = .00) for emerging markets but a positive effect (mean effect size = 0.04, p = .09) for developed host countries. This might suggest a potential learning effect of internationalization, especially for firms coming from emerging countries. Finally, the HOMA analysis shows no significant performance effect of cultural distance for U.S. firms. This could be explained perhaps by the higher degree of internationalization and greater international experience of American firms accentuating the learning effects.

Table 9 reports the analytical results for both Pearson bivariate correlation and partial correlation coefficients. It shows that cultural distance has a negative and statistically significant effect on performance using both techniques (r-based mean = −0.03 and p = .002; r-based mean = −0.03 with p = .00). However, there are a few noteworthy differences. First, when using Pearson correlations, cultural distance has a negative effect on subsidiary performance but has no significant effect on MNC performance. Results are slightly different for the partial correlation technique, where cultural distance shows a negative and statistically significant impact on subsidiary performance and a positive and significant effect on MNC performance. This is possibly due to the potential organizational learning opportunities of internationalization achieved at the level of the entire MNC network. Second, Pearson correlation technique does not yield significant results with regard to firm origin, and the partial correlation technique shows negative and statistically significant results for firms from developed countries (r-based mean = −0.05, p = .00) and positive and statistically significant results for firms from emerging markets (r-based mean = 0.04, p = .09). Third, r-based estimations do not provide evidence of a significant effect of cultural distance from developed host countries. However, consistent with our HOMA results, r-based estimations show a positive and statistically significant effect. Overall, we can conclude that we find less significant results when using the Pearson bivariate correlation technique of meta-analysis than when using the partial correlation technique (which, as stated above, keeps other variables constant), probably as a result of different sample sizes in the two analyses.

Table 9

HOMA Meta-Analytic Results: Cultural Distance (CD) to Performance

	Pearson Product–Moment Correlation (r)						Partial Correlation Coefficient (rxy.z)
Predictor	k	N	M (p value)	SE	Q test	I 2	k	N	M (p value)	SE	Q test	I 2
CD to performance	162	216,961	−.03 (.00)***	.01	3140.59***	0.95	275	696,299	−.03 (.00)***	.01	3984.33***	0.93
Measurement of CD
Euclidean (Kogut and Singh) distance	105	185,530	−.03 (.05)*	.01	2971.85***	0.97	162	636,304	−.04 (.00)***	.01	1168.35***	0.86
Mahalanobis distance	3	20,967	−.03 (.00)***	.01	0.75	0.00	3	28,302	−.02 (.00)***	.01	1.01	0.00
Cultural zone distance (dummy)	6	420	−.03 (.51)	.05	1.63	0.00	45	7,779	−.05 (.04)**	.02	210.60***	0.79
Perceptual distance	16	6,192	−.18 (.00)***	.03	33.07***	0.55	24	5,381	−.23 (.00)***	.07	605.37***	0.96
Cultural-distance data source
Hofstede	131	190,012	−.02 (.07)*	.01	3000.91***	0.96	193	649,564	−.01 (.02)**	.01	2884.58***	0.93
Four dimensions	89	179,394	−.03 (.04)**	.02	2935.38***	0.97	151	635,756	−.03 (.00)***	.01	1124.61***	0.87
Five dimensions	10	2,966	−.04 (.01)**	.02	15.04*	0.40	10	2,478	−.13 (.00)***	.02	6.17	0.00
Power distance dimension	7	1,328	−.04 (.17)	.03	6.05	0.01	4	1,247	−.00 (.92)	.03	3.10	0.03
Uncertainty avoidance dimension	9	2,110	−.02 (.43)	.02	5.48	0.00	5	1,299	−.05 (.31)	.05	10.34**	0.61
Individualism dimension	8	2,058	−.03 (.51)	.04	15.94**	0.56	4	1,247	.04 (.49)	.06	9.53**	0.69
Masculinity dimension	7	1,328	.01 (.69)	.03	9.02	0.33	4	1,247	−.09 (.00)***	.03	4.31	0.30
GLOBE	5	3,108	−.01 (.76)	.05	13.19**	0.70	4	572	.01 (.86)	.07	8.71**	0.66
Ronen and Shenkar	7	645	.03 (.42)	.04	6.47	0.07	50	5,381	−.04 (.03)**	.02	247.76***	0.80
Primary data	16	2,292	−.18 (.00)***	.03	33.07***	0.55	24	12,348	−.23 (.00)***	.07	605.37***	0.96
Performance type
Accounting performance	40	75,171	−.04 (.00)***	.01	189.54***	0.79	20	9,407	.01 (.68)	.03	113.23***	0.83
Market performance	20	2,607	−.01 (.76)	.02	17.99	0	52	14,625	.04 (.31)	.05	1541.81***	0.97
Survey performance	66	39,665	−.06 (.00)***	.02	253.73***	0.74	53	91,032	−.04 (.00)***	.01	204.93***	0.75
Survival	6	32,902	.05 (.10)	.03	48.96***	0.90	89	377,959	−.06 (.00)***	.01	870.37***	0.90
Innovation	8	29,436	.12 (.15)	.08	1491.35***	1.00	31	147,314	.01 (.35)	.01	127.07***	0.76
Other	22	37,180	−.08 (.00)***	.02	270.97***	0.92	30	55,962	−.16 (.00)***	.02	703.59***	0.96
Firm identity
MNC	63	123,987	.00 (.84)	.02	2349.96***	0.97	94	179,603	.03 (.02)**	.01	2013.72***	0.95
Subsidiary	84	90,702	−.06 (.00)***	.01	728.21***	0.89	147	478,461	−.08 (.00)***	.01	1880.86***	0.92
Home country
United States	21	24,562	.01 (.95)	.07	1950.98***	0.99	31	33,389	−.04 (.03)**	.02	116.58***	0.74
Developed markets	58	153,247	−.01 (.54)	.02	2648.95***	0.98	127	498,532	−.05 (.00)***	.01	955.57***	0.87
Emerging markets	17	14,154	−.01 (.76)	.02	32.16***	0.50	14	4,998	.04 (.09)*	.02	27.45**	0.53
Host country
United States	7	3,063	−.02 (.74)	.05	47.81***	0.87	14	2,604	.09 (.00)***	.02	14.91	0.13
Developed markets	23	9,169	−.00 (.89)	.03	79.53***	0.72	18	3,055	.07 (.07)*	.04	77.42***	0.78
Emerging markets	49	8,465	−.10 (.00)***	.02	231.07***	0.79	60	19,749	−.12 (.00)***	.03	729.78***	0.92
Time
Until medium year	83	163,928	−.01 (.45)	.02	2775.79***	0.97	145	597,023	−.01 (.00)***	.00	595.55***	0.76
After medium year	79	53,033	−.05 (.00)***	.01	297.58***	0.74	130	99,276	−.04 (.02)**	.02	3289.25***	0.96

Note: GLOBE = Global Leadership and Organizational Behavior Effectiveness; I2 = scale-free index of heterogeneity; k = number of effect sizes; M = mean effect size; MNC = multinational corporation; N = total sample size; Q = Cochran’s homogeneity test statistic; SE = standard error of mean correlation.

p < .10.

p < .05.

p < .01.

MARA Results

MARA results (Table 10) further confirm the importance of controlling for methodological and model specification artifacts and variable operationalization.

Consistent with the HOMA results, the relationship between cultural distance and performance is more negative when cultural distance is operationalized through perceptual measures based on primary data. As already noted for HOMA analysis, performance is more positively influenced by cultural distance when operating in developed host countries and more negatively influenced by cultural distance when operating in emerging markets. Moreover, the impact of cultural distance on performance is more positive for firms from emerging markets (in Model 3, β = 0.05, p = .06). Also, consistent with the HOMA results, the MARA analysis indicates that cultural distance has a negative effect on performance when it is measured at the subsidiary level (in all models, β = −0.15, p = .00). Among the methodological artifacts, panel design of a study shows a significant positive effect on the cultural distance to performance relationship (in Model 2, β = 0.05, p = .01). This indicates that panels tend to yield more positive effects of cultural distance on firm performance. Also, the focal relationship tends to be more negative when potential endogeneity issues are addressed in the primary study (in Model 2, β = −0.06, p = .02).

Discussion

Our objective in this paper was to bring additional clarity on the role of cultural distance in the process of firm internationalization. Despite the wide use of cultural distance in the global strategy literature, results on its effects on the firm internationalization process have been inconclusive. Based on our review and analysis, we believe that a major reason for this lack of consistency is that this research has been often done in a rather broad-brush manner. Scholars have either tended to generalize the construct of internationalization a bit too much without sufficient attention to its different stages, aspects, or outcomes, or have narrowly focused on a specific decision or outcome without an attempt to integrate findings across related outcomes. In addition, different cultural-distance measures have been used without proper explanation of their reliability or relative advantages.

To remedy these limitations, we adopted a comprehensive view of the process of firm internationalization, examining all key stages and strategic decisions related to this process, even adding to the discussion its performance consequences. We followed a similar approach to cultural distance, considering a wide range of studies that employed different operationalizations and measures of cultural distance. We were able to maximally leverage existing research by conducting the largest meta-analysis of primary cultural distance studies to date. Furthermore, we employed the most advanced meta-analytical methodology for our analysis. As a result, we feel confident that our review and analysis of the substantial literature on this topic were both comprehensive and rigorous, and thus provide a solid foundation for drawing a number of important theoretical insights and ideas for future research in this area. Figure 3 summarizes our key findings. Below we relate the key findings to the four research questions that we posited and put our results in perspective.

Figure 3

Summary Findings on Cultural Distance and the Process of Firm Internationalization

Note: The constructs in each of the four blocks are all dependent variables studied in our meta-analysis. Constructs in bold show significant relationships between cultural distance and the specific aspect of firm internationalization. We report only significant coefficients. The coefficient refers to the Hedges-Olkin-type meta-analysis results reported in the different tables, and their relative effect size can be interpreted in an absolute way.

Research Question 1 asked whether cultural distance affects key decisions in the various stages of firm internationalization, and if so, how. We find that cultural distance affects internationalization, but in a more intricate way than commonly assumed. In sum, firms tend to stay away from culturally distant countries, which is consistent with mainstream theories of location choice and FDI. If they invest in such countries, firms prefer greenfields over acquisitions. Although transaction-cost theory has been used to predict both an acquisition (as it provides learning possibilities) and a greenfield (to minimize friction with host country nationals), our meta-analytic results suggest the latter effect dominates. Firms also benefit significantly from the transfer of practices to such culturally distant locations. However, going to a culturally distant host country negatively impacts the performance of the subsidiary there. Figure 3 provides a visual summary of cultural-distance effects on the various stages of firm internationalization. Unpacking these stages shows the differential effects of cultural distance and underscores that studies of internationalization would benefit from more fine-grained analysis by stage. As seen in Figure 3, the effect size of cultural distance is largest for the integration stage of practice transfer (|.15|), followed by the negative subsidiary performance effect (|.07|), the preference for greenfield versus acquisition (|.05|), and lastly, location choice (|.02|).

The differential-performance effect of cultural distance (subsidiary vs. MNC) is one of our most interesting findings. The negative effect of distance on subsidiary performance is in line with existing theories including both classic MNC views and the behavioral view. According to the classic view (Hymer, 1976; Kogut & Zander, 1992), negative performance effects are due to costs exceeding the benefits of internationalization. In the behavioral view (Foss & Lindenberg, 2013; Maitland & Sammartino, 2015a; Powell, Lovallo, & Fox, 2011), they result from underestimation of the true costs associated with internationalizing to culturally distant countries (Dibbern, Winkler, & Heinzl, 2008; Larsen, Manning, & Pedersen, 2013). Thus, while the two perspectives suggest different explanatory mechanisms, the result is the same—subsidiaries are impacted negatively by large cultural distance. Intriguingly, cultural distance does not affect the performance of the MNC as a whole. This finding is intuitive, as companies would not be internationalizing if it were otherwise. But understanding how firms can compensate at the corporate level for the negative subsidiary performance in distant (and thus risky and high-cost) host countries is a fascinating question. The data from our sample did not allow us to tease out these complex dynamics, but we are excited about the opportunity to study this question further in the future. It seems that companies perhaps make these location decisions in the context of their overall strategic portfolios of international operations rather than with regard to a specific host market. Theories of internationalization should be catching up with this possible view.

Another set of findings worth noting is the mixed effect of cultural distance on amount of practice transfer, coupled with a positive impact on the benefits of practice transfer. In fact, this was the strongest effect of cultural distance among all outcomes that we examined. The first part of this finding is rather straightforward as companies are reluctant to engage in such efforts given the very different context in which the subsidiary is placed; hence the anticipated difficulties of transferring the practice and the meaning behind it, in particular, which is essential for its successful adoption (Kostova, 1999; Kostova & Roth, 2003). The second part of the finding, although a bit counterintuitive, is not surprising either. It is consistent with the concept of the transnational organization (Bartlett & Ghoshal, 1998), which suggests that companies benefit from a more integrated model of sharing knowledge and best practices among the portfolio of operations that might be located in very diverse national settings. As scholars have shown, there are mechanisms through which companies can overcome the challenges of cultural distance in such integration efforts through common practices. For example, Kostova and Roth (2002) found that social capital reflected in trust, commitment, and identification of the subsidiary with the corporate headquarters facilitate practice transfer by closing of the gap between the two sides. Although our study did not allow us to test these ideas on a larger scale, future research to identify the most effective ways in which organizations may overcome cultural distance for the benefits of organizational integration and when such benefits outweigh the costs related to distance is warranted.

Research Question 2 concerned the sensitivity of the distance effects to the particular measurement and operationalization of cultural distance. We find that studies using perceptual measures tend to have more pronounced effects compared to the cultural distance index based on secondary data. One explanation may be that the number of studies using perceptual measures is still limited. This does not however hold for performance studies, of which a sufficient number exists and for which we still find a large negative effect, much larger than any of the other cultural distance measures. The reason why the perceptual measures have a stronger performance effect (relative to other measures of cultural distance) may be that they possibly capture other perceived differences and difficulties in the respective host country (beyond culture), but respondents attribute these negative perceptions to cultural distance. As a result, perceptual measures may overstate the role of cultural distance. In fact, the early Uppsala model of internationalization was based on psychic (i.e., perceptual) rather than cultural distance.

We also showed that results are not always consistent across different cross-cultural frameworks. For example, regarding entry mode decisions, we found an insignificant effect of the Hofstede-based cultural distance, but a positive effect of cultural distance based on GLOBE or Schwartz. Similarly, the negative effect of cultural distance on establishment mode using Hofstede turns positive when using GLOBE. Finally, the negative effect of cultural distance on performance using Hofstede turns insignificant when using GLOBE. This raises the question of which cross-cultural framework to use in cultural distance studies. It is beyond the scope of this paper to discuss the numerous methodological differences between these cross-cultural frameworks and their pros and cons (Beugelsdijk & Maseland, 2011; Schaffer & Riordan, 2003; Smith, 2006). Moreover, many management scholars are users of these frameworks and tend to be methodologically indifferent regarding which framework to use to measure cultural distance. Instead of suggesting which framework may be the preferred one, we take a more pragmatic approach.

One practical recommendation is to run the cultural-distance analysis using multiple frameworks (e.g., Drogendijk & Slangen, 2006). This however raises the follow-up question of what to do if results are not consistent, and how to attribute these findings to differences between these frameworks. An alternative solution is to treat these frameworks as complementary, each capturing part of the overall variation in cross-national cultural values (Steenkamp, 2001). Without a prior normative position on which framework is best, the optimal approach is then to combine all these frameworks in one composite cultural distance index. Beugelsdijk, Kostova, and Roth (2017) have calculated such a composite cultural-distance index using the Mahalanobis correction to control for the correlation between the dimensions of the three frameworks. They find that the resulting cultural-distance scores match the cultural classification of countries in specific cultural zones (Ronen & Shenkar, 2013). One key advantage of such a composite cultural-distance index is that it simplifies the discussion what framework to use, and limits the possibility for researchers to shop for the result that best supports their hypothesis. Of course, the disadvantage of this approach is that it can only be used for a generic cultural-distance argument and not to test distance effects on a specific cultural dimension. Our literature review showed that most studies are interested in exploring such a general cultural-distance effect and not the effect of distance on specific dimensions.

Research Question 3 concerned the possible contingences of home and host countries being developed or emerging markets. Our results suggest that cultural distance effects are very sensitive to sample structure. When the home country is an emerging market (e.g., India, Brazil, or China), the negative effect of cultural distance on performance turns positive and insignificant. Alternatively, if the host country is an emerging market, the negative relation between cultural distance and performance becomes even more negative. In contrast, when the host country is a developed market (especially when it is the United States), the relation between cultural distance and performance turns positive. There are two possible explanations for the moderating effects of home or host and emerging or developed country. It might be that emerging market MNEs actively seek distant locations to invest because they are usually associated with more competitive environments where these nascent global players can acquire technological capabilities and learn best practices. The benefits of operating in such distant places can outweigh the risks and costs of dealing with cultural differences. While our data again did not allow us to fully explore these alternative explanations, the results at least highlight this interesting contingency and present opportunities for future research.

Another possible explanation is methodological in nature. It might be that studies of cultural distance conflate distance with direct or level effects and the results that we see are not due to the difference between home and host country but are instead caused by the conditions (cultural or institutional) in the home or the host country. For example, going to a “failed state” or an emerging market with “institutional voids” might lead to negative outcomes regardless of whether the home country is culturally similar or distant from the host country. Distinguishing between distance and direct (or “level”) effects is particularly problematic when distance studies include only one home or one host country (Brouthers et al., 2016). Thus, the suggestion to ideally have multiple home and hosts as to make sure level and distance effects are not conflated.

Research Question 4 concerned the impact of time. Our sample included primary studies published between 1988 and 2015, which allowed us to indirectly explore the longitudinal performance effects of cultural distance. We found that the effect of cultural distance on firm internationalization is relatively stable over time. This is in sharp contrast to the conclusion by Taras, Steel, and Kirkman (2012) (using different data and a different method) that Hofstede data are less and less able to explain cultural differences because they are outdated. Their interpretation is however at odds with our finding that results on cultural distance do not consistently depend on the use of Hofstede data or more recent data from alternative culture frameworks. More important perhaps is that our finding on the relatively stable effect of cultural distance over time does not imply that cultures do not change. As long as cultures change on parallel trajectories (as shown by Inglehart & Baker, 2000), cross-country cultural distances are relatively stable (Beugelsdijk, Maseland, & van Hoorn, 2015). Here the question was whether the effect of cross-country cultural distance changes over time, and results suggest it does not, at least not in a significant way and not in a consistent direction.

Jointly, these findings provide a solid foundation and interesting insights for future research in this area. In addition to the ideas that we discussed in the previous paragraphs, we would note the following. First, our review showed an interesting gap in the literature on cultural distance—there appears to be very little work examining its effects on managerial and organizational aspects of internationalization, both in an absolute sense, but also especially compared to the large number of studies on entry mode, establishment mode, and performance. For the few studies that address management aspects such as the benefits of the transfer of practices, we find large effect sizes of cultural distance. Although we cannot rule out the possibility that these large effect sizes are (partly) driven by the low number of primary studies available, we see this as a very promising area of future cultural distance research. We recommend future work to address an array of management-related outcomes related to internationalization. We discussed transfer of practices but there are many other important aspects of integration and control of foreign operations that warrant attention, including use of cross-cultural teams, organizational learning and innovation across subsidiaries and parent companies, and management of agency problems between parent companies and foreign operations (Kostova et al., 2016). Second, we would encourage more in-depth studies of different types of distance on internationalization, particularly institutional and economic distance, and an examination of the relative salience of different types of distance for different outcomes. It is possible that cultural distance, for example, is more tied to postinvestment management integration while institutional distance is more critical in the preinvestment stages. Such extensions of research in this area will bring more definitive understanding of what particular context matters for what organizational outcomes. Finally, some of our findings raise questions (if not provide insights) about the limitations and boundary conditions of existing internationalization theories. Do they apply equally to developed and emerging market MNCs, is the notion of distance possibly bound by the perspective of industrialized countries, and what are the remedies to distance in MNC management?

To conclude, cultural differences continue to be a serious consideration for managers and companies as they expand internationally. Understanding when and for which aspects of the internationalization process cultural differences really matter is a necessary step in learning how to manage and possibly leverage such differences.