Introduction
Firms’ foreign market entry mode[i] choice is one of the most researched topics in international business (e.g. Anand/Delios, 1997; Benito/Pedersen/Petersen, 2005; Brouthers/Hennart, 2007; Chen/Hennart, 2004; Datta/Herrmann/Rasheed, 2002; Delios/Henisz, 2003; Erramilli/Agarwal/Kim, 1997; Hennart, 1991; Kim/Hwang, 1992; Kogut/Zander, 1993; Madhok, 1997; Malhotra/Agarwal/Ulgado, 2003; Martin/Salomon, 2003; Tse/Pan/Au, 1997; Yiu/Makino, 2002) . Yet, despite the considerable attention devoted to this topic, most studies still refer to a specific mode exerted by a firm in a given foreign market – be it joint ventures, wholly owned greenfield subsidiaries, or acquisitions – and often with reference to a specific value chain activity, such as when the choice is between exports, manufacturing subsidiaries, and licensed production. This simplified view of entry modes, while convenient and useful for theory building and empirical investigation, stands in contrast to the variety of combined entry modes that can be observed in real-world firms.
The general approach in extant literature has been to view each geographic area-value chain activity combination independently, thereby disregarding additional areas and activities. However, managerial decisions on such entry modes are not independent but are rather interdependent. For example, transaction cost concerns (Buckley/Casson, 1976) may motivate the firm to standardize its use of entry modes across geographies and activities, while the firm’s search for diverse knowledge (Zahra/Ireland/Hitt, 2000) on the other hand may motivate it to diversify its entry modes. Thus, the usefulness of analyzing a specific entry mode at the activity and country level, without regarding the overall set of entry modes a given firm may have, might be quite limited (Asmussen/Benito/Petersen, 2009; Buckley/Hashai, 2004, 2005; Hill/Hwang/Kim, 1990; Petersen/Benito/Welch/Asmussen, 2008).
The aim of this paper is to expand extant foreign market entry mode research by switching the unit of analysis from activity- and location-specific entry mode to the analysis of multiple
entry modes of a firm across its value chain and across foreign markets. This approach implies that entry modes decisions are likely to be interdependent across host markets and value chain activities, and are not taken independently of each other as implicitly assumed by extant literature.
More specifically, we aim to investigate how firms’ level of technological knowledge intensity affects their foreign entry mode diversity,definedas their propensity to vary entry modes across locations and activities. The direction of such an effect is not clear as different theoretical perspectives predict contradictory effects. On the one hand, internalization theory (Buckley/Casson, 1976, 1998; Rugman, 1981); as well as knowledge transfer efficiency considerations (Kogut/Zander, 1993; Martin/Salomon, 2003) essentially imply that greater technological knowledge intensity limits entry mode diversity. On the other hand, greater technological knowledge intensity is also associated with a capacity for learning due to greater absorptive capacity (Cohen/Levinthal, 1990). The exposure to different types of technological learning through multiple types of entry modes is likely to leverage diverse technological knowledge and skills in foreign markets, thus leading firms with greater technological knowledge intensity to engage in more diverse entry modes. By empirically comparing these two contradictory theoretical predictions we provide a partial answer to the question why some firms use diverse entry modes while others apply only few modes.
The structure of the paper is as follows. In the next section we conceptualize entry mode diversity according to different levels of analysis (area level, activity level, and corporate level) and derive hypotheses as to how technological knowledge intensity may affect entry mode diversity at these three levels. The hypotheses derived from our conceptual framework are tested on unique data of entry modes used by a sample of Israeli-based firms. This is followed by an analysis and a discussion of the results. Finally we suggest further research avenues and conclude.
Conceptualizing Entry Mode Diversity
While extant research on foreign market entry mode mostly refers to a firm’s entry mode decision as a general decision at the location-activity level[ii], a few studies have indicated the importance of referring to a firm’s variety of entry modes (Asmussen et al., 2009; Benito/Petersen/Welch, 2009; Petersen et al., 2008). Some of these studies have theoretically advanced the conceptualization of internationalizing firms as a locus of value chain activities to which firms simultaneously determine the location and entry mode in order to minimize their overall costs (Buckley/Casson, 1998; Buckley/Hashai, 2004, 2005; Casson, 2000). Other studies have empirically shown that firms often do not stick to one particular entry mode, but instead simultaneously employ a variety of entry modes at the value chain activity level (Benito/Welch, 1994; Fina/Rugman, 1996; Petersen/Welch, 2002). Taken together, it is therefore implied that firms may often simultaneously use multiple entry modes in different locations and value chain activities.
Petersen et al. (2008) used an entry mode matrix to illustrate this point. Assuming that an international firm operates in I host markets and has J identifiable activities in its value chain, its entry mode matrix at a given point in time can be denoted M=(mij), where i=1…I indexes host markets and j=1…J indexes value chain activities. Each cell in the matrix (mij) may then contain one or multiple entry modes under which the given activity is performed in the given host market. The general form of the entry mode matrix is presented in Figure 1.
[Insert Figure 1 about here]
The matrix depicts three levels of aggregation in which entry mode diversity can be discussed[iii]:
- Area-level diversity refers to different entry modes exerted by a firm within a given foreign area – country or region – and can therefore be evaluated by looking at a row vector of activity-level decisions. The larger the variation in entry modes within this vector, the higher the area-level diversity.
- Activity-level diversity is about how a specific value chain activity is performed in different geographical areas (countries or regions), as measured by each column in the matrix. Activity-level diversity for a given activity is therefore described by a column vector of the form: .
- Corporate diversity represents the variety of entry modes in the entire matrix M, as represented by all the combinations of area-level and activity-level entry mode decisions.
Petersen et al. (2008) discuss what may potentially affect the diversity within the entry mode matrix. However, to our knowledge no study has attempted to develop and test hypotheses about the predictors of entry mode diversity. One such hypothesis may pertain to the firm’s technological knowledge intensity, which has been emphasized in foreign market entry mode research as a distinctive variable affecting foreign market entry mode choice (e.g. Anand/Delios, 1997; Delios/Henisz, 2003; Erramilli et al., 1997; Gatignon/Anderson, 1988; Padmanabhan/Cho, 1999; Tan et al., 2001; Tse et al., 1997; Yiu/Makino, 2002). Technological knowledge intensity represents the level of technological knowledge contained in each unit of output that the firm produces (Almor/Hashai/Hirsch, 2006; Hashai/Almor, 2008; Jones, 1999). Since technological knowledge intensity, often measured as the ratio of research and development (R&D) expenditures to sales, has been shown to affect the entry mode decision of firms, it is quite likely that the diversity of firms’ entry mode portfolio across countries and value activities is affected by this variable as well. Hence, we aim to investigate what is the likely impact of this variable on foreign market entry mode diversity. As mentioned above, internalization theory and organizational learning theory constitute two perspectives that may inform us about this relationship.
Internalization theory and entry mode diversity
Internalization theory explains the existence and growth of multinational enterprises (Buckley/Casson, 1976; Rugman, 1981, Teece, 1986a). The theory highlights firms’ technological
knowledge intensity as a dominant determinant of internalization and externalization decisions. This stream of literature is primarily focusing on the impact of failures in the market for firm-specific know-how (most often referring to technological know-how) on firms’ choice between licensing and wholly owned subsidiaries. In essence, the major prediction of this school of thought is that higher levels of technological knowledge imply a higher risk of market failure and hence lead an internalized mode of operation in foreign markets, i.e. wholly owned entry modes. Consequently, if a knowledge-intensive firm were to engage in diverse entry modes, it would presumably face higher transaction costs. This is the result of information asymmetry (difficulty of evaluating and transferring high levels of technological knowledge, see Arrow 1982; Davidson/McFetridge, 1984) between the focal firm and potential collaborators coupled with the high uncertainty of managing and coordinating multiple entry modes for highly technology-intensive firms (Contractor, 1990; Kim/Hwang, 1992; Osborn/Baughn, 1990; Williamson, 1975). It therefore follows that higher levels of technological knowledge intensity are likely to be associated with lesser entry mode diversity.
A complementary view refers to the relationship between the relative efficiency of technological knowledge transfer for internationalizing firms using different types of entry modes. The technological knowledge developed by highly technology-intensive firms is often complex, hard to codify and to teach, and, hence, is relatively difficult to transfer (Hashai/Almor, 2008; Kogut/Zander, 1992, 1993; Martin/Salomon, 2003; Teece, 1977). Externalization of such knowledge is likely to result in knowledge dissipation costs associated with the misappropriation of transferred knowledge, and with higher control and monitoring costs to protect technological knowledge, as well as higher negotiation and litigation costs (Martin/Salomon, 2003).
Greater technological knowledge intensity often implies greater complexity of coding and decoding the transferred knowledge (Kogut/Zander, 1992, 1993; Martin/Salomon, 2003). Higher entry mode diversity is therefore likely to result in greater costs of transferring complex knowledge for highly technological knowledge intensive firms, since it requires tight
coordination of knowledge transfer between multiple parties engaging in different contractual arrangements. For example, if a technology-intensive firm were to use a mix of sales agents, licensing agreements, joint ventures, and wholly-owned subsidiaries in a given foreign market, it would presumably have to incur large costs and efforts in order to manage, organize, and transfer knowledge across these diverse arrangements while avoiding the appropriation of its knowledge by other firms.
Overall, the above views imply that greater levels of technological knowledge intensity are expected to be associated with lower entry mode diversity. We therefore hypothesize that:
Hypothesis 1: Technological knowledge intensity is negatively associated with entry mode diversity.
Organizational learning theory and entry mode diversity
While the above hypothesis mainly draws on internalization theory, there have been recent calls to incorporate a resource-based view into entry mode research (Madhok, 1997; Zhao/Luo/Zhu, 2004) and thus complement the (transaction) cost minimization concern of internalization theory with a value generation perspective. Indeed, highly technology-intensive firms are arguably dependent on having diverse technological knowledge in order to create and sustain their competitive advantage. Strategic management research has shown that a firm’s ability to draw on diverse knowledge is an important source of competitive advantage (Kilduff/Angelmar/Mehra, 2000; Milliken/Martins, 1996). This is so since knowledge diversity stimulates problem-solving and enhances innovation by providing multiple viewpoints (Page, 2007). Highly technology-intensive firms are likely to obtain the capability to observe and mobilize new types of knowledge due to their high absorptive capacity because — as noted by Cohen and Levinthal (1990), Autio, Sapienza and Almeida (2000) and others — greater levels of technological knowledge intensity are associated with a greater capacity for learning. Yet, a firm’s ability to benefit from this
absorptive capacity is contingent on the availability of external learning opportunities (Cohen/Levinthal, 1989).
In the case of internationalizing firms which operate across national boundaries, exposure to diverse technological knowledge is particularly pronounced (Ghoshal, 1987; Zahra/Ireland/Hitt, 2000). Firms which already possess strong technological capabilities are motivated to seek out technological knowledge abroad, in order to enhance their knowledge diversity (Cantwell/Janne, 1999; Chung/Alcácer, 2002). This may in turn affect a given firm’s choice of entry mode portfolio, since its entry modes constitute its organizational interface with different host country environments. All else being equal, a more diverse set of entry modes allows sourcing from a more diverse pool of technological knowledge. Interaction with different types of partner firms through multiple types of organizational arrangements is likely to leverage diverse technological knowledge and skills in foreign markets through which firms can source knowledge to facilitate and strengthen their competitive advantage (Vermeulen/Barkema, 2002). In fact, a review of the entry mode literature suggests that different types of entry modes – e.g. market-based, contractual, jointly owned, and wholly owned – convey different learning experiences for internationalizing firms.
Market-based entry modes such as arms length relationships with sales agents and distributors enable firms to learn from these local agents about technologies that are specific to their markets (Almor et al., 2006; Hirsch, 1989; Zahra/Ireland/Hitt, 2000). Porter (1998) suggested that technological innovation is propelled by having a “window on the market”, by benchmarking against technologically advanced competitors and by targeting the preferences of sophisticated customers in knowledge-intensive locations. A cost-effective way for foreign firms to acquire these benefits may be to interact with local agents, who have extensive experience with the market and broad knowledge of local technological developments (Canabal/White III, 2008; Petersen/Pedersen, 2002). Agents may also act as “filters” through which the R&D-intensive entrant firm can evaluate the local applicability and relevance of its own proprietary technologies.
Contractual entry modes (strategic alliances, OEM agreements, etc.), on the other hand, enable firms to gain deeper technological understanding from their partners and acquire complementary competencies directly from them (Hamel, 1991; Teece, 1986b). Indeed, recent observations indicate that firms operating in high tech industries are those which are most likely to engage in multiple contractual agreements through which they combine their technological capabilities with complementary technological capabilities of partner firms as means of fostering their competitive advantage (Dyer/Singh, 1998; Kale/Dyer/Singh, 2002; Lavie, 2006).
Partly owned entry modes (joint ventures) enable internationalizing firms to learn from their partners and acquire the type of operational and tacit technological knowledge that can only be transferred by close collaboration and supervision (Barkema/Shenkar/Vermeulen/Bell, 1997; Reuer/Tong, 2005). In particular, technological knowledge which is teachable but not codifiable (Kogut/Zander, 1993) could be effectively appropriated through a jointly owned arrangement, since this provides the opportunity to work alongside a local firm’s employees in a common organizational framework.
Finally, wholly owned entry modes facilitate “learning by doing” where specific knowledge about host country technologies, their operational competency requirements, and their complementarity and compatibility with the entrant firm’s proprietary technology, are revealed through trial and error (Arora/Fosfuri, 2000). Where market based entry modes enable relatively broader technological learning, wholly owned entry modes enable a much deeper learning as a result of doing business in a particular foreign setting (Almor et al., 2006; Hirsch, 1989). Nevertheless, acquiring both broad and deep knowledge is likely to be the most powerful way for a firm to enhance its technological competitive advantage.
At the value chain activity level, increased entry mode diversity should thus enable highly technology-intensive firms to learn from multiple foreign partners with whom they interact in different contractual ways. At the host market level, increased entry mode diversity of technology-intensive firms may be motivated by the learning opportunities arising from
simultaneously conducting R&D, production, distribution, and servicing activities under different modes in a given host country. This is so since a firm’s technological knowledge is likely to have an effect not only on the R&D function but on all value chain activities. Firms with greater absorptive capacity are likely to have a greater capacity to learn from such diverse entry modes. Since greater technological knowledge intensity is associated with greater absorptive capacity it therefore follows that highly technology-intensive firms are likely to engage in more diverse entry modes which will serve as a vehicle for obtaining more diverse technological knowledge through the use of market based-, contractual, partly owned, and wholly owned entry modes. We therefore hypothesize that:
Hypothesis 2: Technological knowledge intensity is positively associated with entry mode diversity.
Data and methods
Our hypotheses were tested on data obtained through a survey of Israel’s leading publicly traded industrial firms. The data was collected for the years 1995 and 1999. A time span of four years was considered long enough so that changes in, and additions to, the firms’ entry modes could be observed, while not long enough as to introduce a large amount of entries and exits (Shyam-Kumar, 2009). The dataset is quite unique as it portrays different entry modes at both the activity and area levels. This refined level of aggregation on entry modes data does not exist, to the best of our knowledge, in publicly available secondary datasets and is essential for testing hypotheses relating entry mode diversity.
The original list included Israel’s one hundred and fifty largest industrial firms. Combined exports by these 150 firms represented about 80 percent of Israel’s industrial exports in 1999. The list was based on data received from Israel’s Ministry of Industry and Trade and data provided by Dun & Bradstreet (2000). After eliminating foreign affiliates, conglomerates and firms which were not publicly traded we were left with a sample of 101 firms. To obtain a balanced panel we
further eliminated all firms with missing data for any variable for either of the years of 1995 and 1999. Hence, the final sample consisted of 67 firms that provided useable information, including questionnaire data[iv]. Comparisons between the 67 participating firms and the 34 non-participating firms did not show evidence of any response bias in terms of firm sales, number of employees, age, industrial classification and percentage of foreign sales.
As noted above, the chosen dataset is unique compared to traditional datasets as it includes data on the specific entry modes of firms in specific host markets, and is elaborated for four value chain activities (R&D, production, distribution and customer support) and six major regions (United States, Rest of America, European Union, Rest of Europe, South East Asia, and Rest of the World). Since entry mode data collection on per country and per value chain activity level is extremely complex we decided to focus on region-specific entry modes at the value chain activity level. This approach is quite common in extant literature (e.g. Almor et al., 2006; Kim/Hwang/Burgers, 1993; Rugman/Verbeke, 2004; Yeung/Poon/Perry, 2001) and reflects the tendency of firms to configure their operations at a regional, rather than at a country level. Such an approach is especially feasible for small and medium-sized firms which are resource constrained. As shown later, this firm size characteristic fits our sample well.
Dependent variables
A firm is defined as having a foreign entry if it performs, or have other organizations performing on its behalf, value chain activities in a certain foreign location. Hence, for each value chain activity and each region it was assessed whether one or more of the following categories of entry modes could be assigned:
- Market based (e.g. arms length transactions with an agent/distributor who performs distribution activities for the firm)
- Contractual (e.g. formalized strategic alliance or original equipment manufacturer (OEM) relationship with local firm)
- Partly Owned (e.g. joint venture with local firm)
- Wholly Owned (e.g. wholly owned greenfield or acquired subsidiaries conducting R&D, production, sales, or customer support).
Overall, there were 204 market based entry modes in our sample in 1995 (and 251 in 1999), 38 (102) contractual entry modes, 5 (32) partly owned entry modes, and 297 (427) wholly owned entry modes. This entry mode classification served as the basis for computing the three measures of entry mode diversity, following the aggregation levels of diversity suggested in the conceptual framework.
Area-level diversity describes the variation in entry modes across value chain activities within a given location. For each area (region) in which the firm had at least one foreign entry, we calculated an entropy measure of its entry modes, defined as, where mi is the share of the firm’s entry modes in that area that fall into category i as defined above. These area-level entropy values were then averaged over the number of regions in which the firm had activity to arrive at its overall area-level entry mode diversity. Entropy is commonly used to measure diversity (e.g. Hitt/Hoskisson/Kim, 1997). In the context of this study it has the advantage that it does not only take into account the number of different entry modes used by the firm but also the distribution of entries across these entry modes. Nevertheless, we also tried a simple count measure and got very similar results from our model (not reported here). This is not surprising as the two measures were highly correlated.
Activity-level diversity describes a firm’s tendency to vary its entry modes of a specific value chain activity across locations (regions). For each value chain activity, we therefore measured the diversity of entry modes worldwide using an entropy measure similar to the one defined above, and these activity-level entropy measures were then averaged over the number of
activities in which the firm had foreign entries to arrive at the firm’s overall activity-level entry mode diversity.
These two variables – area- and activity-level diversity – capture variations along the two dimensions of the entry mode diversity matrix (cf. Figure 1). For example, a firm which always uses joint ventures for production and always wholly-owned subsidiaries for R&D would have a higher degree of area-level diversity than of activity-level diversity as it does not standardize its governance form within the individual locations. Conversely, a firm using wholly-owned subsidiaries for all activities in Europe and joint ventures for all activities in Asia would have a higher degree of activity-level diversity than of area-level diversity as it does not distinguish between different value chain activities in its governance forms.
Finally, corporate entry mode diversity uses the entry modes found in the entire entry mode diversity matrix of the firm as an input to calculate an entropy measure of diversity. This captures variations along both the area and business activity dimensions.
Independent Variable
Technological knowledge intensity, as defined earlier, represents the level of technological knowledge embodied in the firm’s output. Following earlier studies (e.g. Almor et al., 2006; Cohen/Levinthal, 1990; Hashai/Almor, 2008; Jones, 1999), we measured this variable as the ratio of R&D expenditures to sales. This ratio reflects the investment share directed towards the creation and absorption of technological knowledge and hence is a major source of firms’ technological knowledge (Hashai/Almor, 2008). Naturally, not all R&D investments are likely to result in increased technological knowledge. However, on average, higher outlays (as a proportion of total sales) on the creation of technological knowledge are expected to result in higher levels of such knowledge. R&D expenditures were used by Cohen and Levinthal (1989, 1990) as an indication of firms’ absorptive capacity – a concept which our second hypothesis builds upon. The R&D per sales ratio in our sample was heavily skewed to the left, so we
performed logarithmic transformations on it in order to bring skewness values down from above 3 to below 0.5.
Control Variables
We also used several control variables to ensure that our results really captured the effect of technological knowledge intensity on entry mode diversity and not any spurious relation caused by, for example, differential learning needs caused by technology-intensive firms being smaller or larger, younger or older, more or less internationalized, or performing more or less value chain activities than other firms. One may argue that a number of “liabilities” unrelated to technological intensity would lead firms to rely more heavily on learning from their agents and partners, and thereby influence their foreign entry mode diversity. This argument is relevant for relatively small firms (facing a liability of smallness), for young firms (a liability of newness), as well as for firms that are relatively less internationalized and would need to overcome their liability of foreignness (Contractor/Kundu/Hsu, 2003; Coviello/Munro, 1997; Lu/Beamish, 2004). Such learning needs often do not relate to technological aspects but rather to local market information and knowledge in foreign countries, yet they are likely to result in greater engagement in multiple collaborations of different types and hence in greater entry mode diversity. We therefore need to control for the possible effects of firm size, age, and level of internationalization when analyzing the relationship between technological knowledge intensity and entry mode diversity[v].
We controlled for firm size, measured as total revenues (in USD) in a given year. As was the case for R&D intensity, firm size was heavily skewed to the left and therefore transformed with logarithms. The year of establishmentof the firm—effectively, the inverse of firm age—was used to control for the impact of accumulated managerial experience on entry mode diversity. Internationalization level was measured by the international diversity of the firm’s foreign operations, operationalized with an entropy measure based on its sales distribution across the
different foreign regions, i.e. as 
where pi is the share of the firm’s international sales generated in region i.
We also controlled for the firm’s foreign value chain scope, based on an expectation that firms performing a larger variety of value chain activities in foreign countries also have an opportunity to use a greater variety of entry modes. We therefore counted the number of activities with entry modes in each region where the firm operates (ranging from 1-4), and averaged this count over the number of regions in which the firm operates.
Dummy variables were used to control for industry effects (such as: per industry regulation, industry-specific transaction costs, and industrial organization) on entry mode choice and hence on entry mode diversity. Our sample did not include conglomerates (all firms operated in a single industry), so we could classify the firms in our sample into the following industries: (1) chemicals; (2) food & beverage; (3) metal; (4) rubber, plastic, wood & paper; (5) textile & clothing; (6) electronics and computer hardware; (7) software; (8) telecommunication; (9) pharmaceuticals and (10) other. After controlling for other effects five of these industries were identified as having relatively more diversified entry modes than other industries: Rubber, plastic, wood & paper, textile and clothing, electronics and computer hardware, telecommunication and metal. Industry dummies for these five industries were therefore used as control variables.
Table 1 depicts the descriptive statistics and correlations of our sample. The mean establishment year of the firms in the sample was 1975. The average sales revenue was USD 128.0 million (92.3 million in 1995 and 163.6 million in 1999), and R&D expenditures constituted 13 percent of revenue (12 percent in 1995 and 14 percent in 1999). This implies that the firms in our sample are typically small to medium sized, but with high growth rates, and that many of them can be considered R&D-intensive. These firms have a slightly higher level of activity-level entry mode diversity than of area-level entry mode diversity; note however that there are high correlations between the three measures of diversity. Overall entry mode diversity (corporate level) increased from 0.42 to 0.56 between 1995 and 1999.
We used panel data models to analyze our sample. Panel data models allow estimation of cross-sectional (firm) effects, time effects, or both. Initially we estimated all three types of models to evaluate the importance of each of these two dimensions. The two-way models with both time and firm effects were almost identical to the one-way cross-sectional models, and the time effect was insignificant for all dependent variables except corporate entry mode diversity where it was only significant at the p=0.05 level. Therefore, we concluded that incorporating time-varying intercepts or errors would not justify the resulting decline in parsimony and degrees of freedom, and we proceeded to estimate a series of one-way models with only firm-specific effects.
For each of the three dependent variables we developed three models: a pooled OLS regression, a fixed effects model allowing for firm-specific intercepts, and a random effects model treating the error term as firm-specific. Each of these models is reported both with and without the control variables, i.e. firm size, international diversity, value chain scope of foreign operations, industry, and firm year of establishment. Note that the traditional fixed effects estimator does not allow time-invariant control variables (industry and firm year of establishment) since these are perfectly collinear with the firm dummies. Hence, to include these variables in the fixed effects model we used the unit effect vector decomposition technique developed by Plumper and Troeger (2004).
In this approach the estimated firm-specific intercepts are regressed on the time-invariant variables and the residual from this regression is used as a predictor in a pooled OLS regression along with the time-varying and time-invariant variables. This effectively decomposes the firm-specific fixed effect into two orthogonal components: one which is explained by the time-invariant variables – in our case, an industry-specific and age-related component – and a residual component of firm effects not explained by these variables (and hence caused by other, unobserved variables). While it produces the same R-square, the technique is more efficient than the fixed effects model, especially if the time-varying independent variables are “almost time-invariant” and if the sample is small (as in our case). It has also been shown in Monte Carlo simulations to outperform the pooled OLS, random effects, and Hausman-Taylor instrumental variables approaches in terms of consistency and unbiasedness (Plumper/Troeger, 2004).
Results
The results of our panel data models regressions are presented in Tables 2 – 4. For each model we present the regression results with and without the control variables. The reported coefficients are standardized betas, which makes us able to compare the impact of different variables. The interpretation is such that one standard deviation change in an independent variable leads to β standard deviations change in the dependent variable, where β is the coefficient reported in the table. Overall we use 134 observations (one observation per year (1995, 1999) for each of the 67 firms).
For all dependent variables, adding the fixed firm-specific effects to the pooled OLS regression increases the variance explained from about 30 percent to about 90 percent. The F-test confirms that these group effects are significant, which implies that the pooled OLS regression without group effects may be biased. The pooled OLS regression is also rejected by the significance of the LM statistic, which in all cases favors the random effects model (Breusch/Pagan, 1979).
For all three diversity measures, the Hausman m-value is insignificant, implying that the estimates produced by the fixed and random effects models are similar and that the random effects model is not biased (Hausman, 1978). A casual comparison of the coefficients confirms this. The somewhat lower significance for the fixed effects coefficients can be attributed to the lower efficiency of this model and the large share of variance captured by the firm dummies, which reflects the general advantage of using a random effects specification in small samples. Alternatively, the vector decomposition model (model 4 in all three tables) is similar to the fixed effects model but more efficient. The results of all the entry mode diversity models are generally robust to different model specifications. Variance inflation factors are reported for model 4, and as they are all quite low (much lower than the recommended threshold of 10), multicollinearity can be assumed not to significantly bias the results (Neter/Wasserman/Kutner, 1990).
Overall, the results of all the models presented in Tables 2 – 4 indicate that R&D intensity is positively correlated to the three measures of entry mode diversity, although this correlation seems to be more significant for area-level and corporate entry mode diversity than for activity-level entry mode diversity (model 4 in Table 3 is significant, but models 2 and 6 are not). Hence, hypothesis 2 is strongly supported while hypothesis 1 is rejected for area-level and corporate level entry mode diversity, indicating that technological learning considerations have a greater impact on these types of entry mode diversity than internalization and knowledge transfer efficiency considerations. Hypothesis 2 is weakly supported for activity-level entry mode diversity, implying that technological learning has less pronounced impact on the benefit of differentiating entry modes across geographic regions than within geographic regions.
As for the control variables, firm size is positively correlated to all entry mode measures. Firm age is negatively correlated to all entry mode diversity measures. The impact of these two control variables in terms of the standardized coefficients is quite similar in magnitude to that of R&D intensity. The results for international diversity are inconsistent across the three entry mode diversity measures and value chain scope is positively correlated to area-level (with the largest impact in term of its coefficient) and corporate entry mode diversity, but negatively correlated to activity-level entry mode diversity.
Finally, industry effects indicate that relative to other industries area-level entry mode diversity is higher in the metal, textiles & clothing, electronics & computer hardware, and telecom industries, whereas activity-level entry mode diversity is relatively higher in the electronics & computer hardware and telecom industries but lower in the rubber, plastic, wood & paper industry. Corporate entry mode diversity is relatively higher in the electronics and telecom industries. These industry effects imply that industries in which the value chain relatively easy can be split into distinct value chain activities tend to have higher area-level diversity (e.g. the textiles industry which is characterized by multiple stages of R&D, production and marketing) whereas more technology oriented industries (e.g. telecom) are characterized by multiple entry modes per value chain activity. Since telecom is traditionally considered to be R&D intensive, the latter finding further supports our arguments and findings regarding the positive association between R&D intensity and entry mode diversity.
Discussion
Our analysis of entry mode diversity at the area, activity and corporate levels reveals several interesting findings. The high correlations among our dependent variables indicate that the three entry mode diversity types are strongly interrelated[vi]. Furthermore, the empirical analysis shows that the entry mode diversity types are more or less determined by the same organizational characteristics. As a scale, the three items have Cronbach’s alpha of 0.94 and they all load on the same factor in a post-hoc confirmatory factor analysis. This could indicate that entry mode diversity is indeed a firm-level construct.
We were generally able to support the strong positive relationship between technological knowledge intensity and entry mode diversity. The finding that firms with high technological knowledge intensity pursue more diverse entry modes is by no means a trivial one in a theoretical perspective since entry mode diversity is predicted to provide both increased costs and increased benefits to technology-intensive firms. Our results indicate that the learning opportunities that might be derived from entry mode diversity for highly technology-intensive firms overrule transaction cost and knowledge efficiency transfer effects. Moreover, these learning opportunities seem to overrule the impact of the ”not invented here” syndrome which leads firms to prefer self- developed technological knowledge on the expense of others knowledge (Katz/Allen, 1982). One explanation for this might be the fact that most of the entry modes in our sample are either market based or wholly owned, hence implying a relatively lesser extent of technological learning from partner firms through alliances and joint ventures.
While greater levels of R&D intensity is often associated with an increased propensity for internalization (due to the information asymmetry, uncertainty and knowledge transfer complexity effects detailed above, see Williamson, 1985; Kogut/Zander, 1993) our results suggest that the ability of highly technology-intensive firms to build on their high absorptive capacity (Cohen/Levinthal, 1990) and garner multiple technological learning opportunities drives them to diversify their foreign market entry modes. Of course, these predictions need not be mutually exclusive, as R&D-intensity may well lead toentry modes that are both more diversified and—on average—more internalized. Nevertheless, our conclusion presents a challenge to the existing foreign market entry mode literature, which traditionally has seemed more preoccupied with the question which particular entry mode gives the “best” learning opportunity for firms. Such an approach bears the implicit assumption that learning is an outcome of an either-or choice of a specific entry mode which may lead to greater or lesser learning (e.g. Barkema/Bell/Pennings, 1996). In contrast, this study proposes that different types of complementary learning can be combined by having a diverse foreign entry mode portfolio, thus leading firms with the capacity to conduct such learning (for instance, technology-intensive firms) to pursue greater entry mode diversity.
In this respect it is noteworthy that while this study has focused on the role of technological learning in affecting entry mode diversity, other types of learning and in particular learning about specific foreign market traits in order to overcome the liability of foreignness may also have an impact. Our results reveal that firm age is negatively related to entry mode diversity, reflecting the role of the “liability of newness” in generating a need for learning from agents and partners through multiple and differing entry modes. On the other hand, contrary to our expectations we found that size is positively correlated with entry mode diversity. This may merely reflect that large firms with more diverse operations can have higher diversity between those operations than can firms with only a limited scope of activity. We did not find any clear impact of firms’ level of internationalization (proxied by their international diversity) and their entry mode diversity.
While our conclusions are general to the entry mode diversity construct, our results also reveal interesting differences between the different types of diversity, where greater technological knowledge intensity has a weaker association with activity-level diversity across geographical areas. Our interpretation of this result is that cross-national/regional differences (culture, language, laws and regulations, etc.) have a considerable impact on the learning opportunities faced by technology-intensive firms. Perhaps intra-regional learning is relatively easier than inter-regional learning – an assumption which is consistent with extant literature on the regional spread of multinational firms (Rugman/Verbeke, 2004). Alternatively, it may be interpreted to mean that firms can learn just by performing a certain activity in multiple locations, without necessarily diversifying its entry modes across these locations, and that other factors than technological knowledge intensity thus drive activity-level entry mode diversity. Taken together, this implies that future research on entry mode diversity should aim to explicitly incorporate factors of cultural distance and host market institutional differences as explanatory variables.
Overall, the findings suggest that managers of technology-intensive firms should consider their entry mode decisions by taking an overall view of their specific value chain activities and their worldwide dispersion rather than taking such decisions in isolation for each entry mode. Such a change in the unit of analysis is likely to have considerable implications on managers’ choice of foreign market entry modes and in particular on the implications of engaging in multiple entry modes as means of fostering organizational learning.
A key contribution, but also a potential limitation, of this study is that it analyses an understudied population of firms originating in Israel. Our sample differs from traditionally analyzed samples of firms from the United States or Europe, since the small domestic market of Israeli firms may lead them to larger foreign markets comparatively early and rapidly in their life span (Hashai/Almor, 2004). The latter observation is especially true for technological knowledge-intensive firms that need to exploit and explore technological advantages in a world where product life cycles are getting shorter. The fact that young, inexperienced, knowledge-intensive firms need to rapidly expand their foreign market presence could lead them to seek more diverse modes of operations in comparison to US or European knowledge-intensive firms that usually exhibit high levels of internalization (Buckley/Casson, 1976, 1998). Such diverse entry modes enable this type of firms to share costs (mainly marketing related costs), build on indigenous foreign markets familiarity of their partners as well as to learn about foreign complementary technologies. Thus, our results may be at least partially driven by our sample characteristics, and additional studies with larger samples of firms from multiple countries are required in order to enhance the external validity of our results. Future research on this subject may therefore analyze the entry mode diversity of firms originating in different countries and learn about their association with technological knowledge intensity.
There are several other avenues for future research on entry mode diversity. First of all, more research is required in order to analyze the impact of additional factors on entry mode diversity. Also, while our suggested conceptual framework is not expected to be time-specific, it may help to analyze entry mode diversity for more recent time periods and over larger spans of time. Exploring the dynamics in entry mode diversity (Benito et al., 2009; Petersen/Welch, 2002) is a potentially important line of research, as understanding how and why firms change their entry mode diversity should garner further insights on the process of entry mode selection. From a dynamic perspective, the finding that age is negatively correlated with entry mode diversity is particularly interesting. This may imply that younger firms are in greater need for learning through their agents and partners through multiple entry modes (as discussed above). It could also be interpreted to mean that in their early years in a certain foreign market, firms experiment with different types of entry modes, but that after a period of trial and error a relatively narrow set of the most efficient entry modes is chosen within that specific location.
Furthermore, while we have attempted to control for a large number of variables that may influence entry mode diversity, we realize that we cannot completely rule out the existence of alternative explanations of our results. We see our framework as a complementary perspective that does not invalidate well-known determinants of the entry mode choice, such as considerations of market failure, risk sharing, and managerial or financial resource constraints. By leading firms to favor certain entry modes over others, these factors would also have an indirect impact on entry mode diversity – an effect that we are not able to tease out with our current data set.
Another possible limitation of the current study is the focus on regions rather than on specific host countries; hence data collection on entry modes at the host country level may further garner insights as to entry mode diversity. Future research may also incorporate other important variables which may potentially affect entry mode diversity but which were not included in our study. For instance, our finding that firm size is positively associated with entry mode diversity implies that there are economies of scale in entry mode diversity. Here also the significance of this result is lower for activity-level diversity, thus indicating that larger firms benefit more from having diverse entry modes across value chain activities (within a specific region) than from having diverse entry modes across host markets (for the same value chain activity). Evidently, firms are able to derive greater economies of scale from entry mode diversity when they are vertically integrated than when they pursue a division of labor for specific value chain activities. This might imply that the international strategy of firms (international, multidomestic, global and transnational, see Bartlett/Ghoshal, 1989) is likely to be associated with different levels of entry mode diversity as a function of value chain disaggregation across host markets and hence is another avenue for future research.
While technological knowledge intensity was found to be significant in explaining entry mode diversity, it might still be that, beyond a certain threshold of R&D intensity, transaction costs reduce the learning benefits of increased diversity. In fact, since we use a logarithmic transformation of R&D intensity, the functional form of our model contains exactly such an effect: at higher levels, a larger (untransformed) increase in R&D intensity is needed to induce a given increase in entry mode diversity. Still, similar studies relating to larger firms and to firms with a more diverse range of R&D intensity may help us to strengthen the external validity and functional form of the linkage between technological knowledge intensity and entry mode diversity. Furthermore, since our findings are not fully consistent for area-level diversity and activity-level diversity there seems to be much room for studies analyzing the impact of firm-specific characteristics versus region-specific institutional and cultural characteristics on entry mode diversity.
Finally, a plausible avenue for future research will be to explore the relationship between entry mode ownership level and entry mode diversity. When looking at the various entry modes chosen for different value chain activities in different host countries we may not only calculate various diversity measures but also refer to an “average” degree of ownership. Entry mode ownership level can be thought of as the “mean” degree of ownership or internalization across a given firm’s value chain, where entry mode diversity can be thought of as the “variance” of such ownership degrees. Both the ownership and diversity of entry modes are potentially important factors as they enhance our conceptualization of foreign market entry modes from an ordinal, categorical variable to a continuous variable which may be characterized by its mean and variance. This should also pave the way for investigating the performance implications of entry mode ownership and diversity. Unravelling the relationship between entry mode ownership and diversity and performance has several empirical complexities (Shaver, 1998), but is of utmost importance for better understanding the normative implications of the foreign entry mode choice.
