University of Groningen Faculty of Economics and Business
Research Master Thesis First supervisor: Dries Faems Second supervisor: Gjalt de Jong
Alliance Portfolio Diversity and Technological
Diversification within Firms:
An Empirical Exploration of the Pharmaceutical Industry
- Final version -Brenda Bos S1605186 31 July 2012
ABSTRACT
Whereas existing studies have examined the impact of alliance portfolio diversity on firms’ technological performance (i.e. number of patent applications), this study will analyse its effect on firms’ technological diversification (i.e. variety of patent applications in terms of technological classes). In this way, we are able to evaluate the impact of alliance portfolios on the quality of technology portfolios rather than its impact on the quantitative output. Merging data from the Thomson SDC database and the EPO database, we create a unique panel dataset (i.e. between 1995 and 2002) for the 43 largest Pharmaceutical firms. The results show that alliance portfolio industry diversity has a positive impact, while alliance portfolio functional diversity has a U-shape relationship with firms’ technological diversification. Jointly, these findings indicate that alliance portfolios can be used as a strategic tool to diversify firms’ technological capabilities.
INTRODUCTION
Firms increasingly use open innovation models in which they rely on strategic alliances, or ‘any voluntary initiated interfirm cooperative agreement that involves knowledge sharing, or co-development’ (Gulati, 1995, p. 619), to complement their internal innovation activities (Hagedoorn, 2002; Cassiman and Veugelers, 2006; Faems et al., 2010). Although the literature initially focused on individual alliances, various scholars have stressed the importance of evaluating all operational alliances of a firm together as a portfolio (George et al., 2001; Lavie, 2007; Wassmer, 2010; Wassmer and Dussauge, 2011). At the portfolio level, synergies and constraints between alliances can be identified (Parise and Casher, 2003), which can subsequently help to increase the overall value of a firm’s alliance portfolio (Wassmer and Dussauge, 2011).
Hence, the goal of this study is to explore the relationship between alliance portfolio diversity and technological diversification. Based on the organizational learning theory and the knowledge-based view of the firm, we hypothesize that access to a variety of knowledge contributes to the creation of new linkages and combinations, and therefore, has a positive impact on technological diversification. Following Jiang et al. (2010) and Duysters and Lokshin (2011), we argue that alliance portfolio diversity is a multi-dimensional construct. More specifically, this study distinguishes between the effect of alliance portfolio industry diversity, alliance portfolio functional diversity and alliance portfolio geographical diversity. These dimensions are motivated by different mechanisms and strategic reasons.
In order to test our hypotheses, a unique panel dataset is created, comprising archival data on alliance activities from the Thompson SDC database and data on patent applications from the EPO database for the 43 largest Pharmaceutical firms. For each firm, we subsequently created alliance portfolios and patent portfolios for a time period of eight years (1995 – 2002). Based on the results of a Hausman test, we estimated our models by conducting fixed effects regression analyses.
The results of our study show that alliance portfolio diversity has a significant impact on the technological diversification of firms. More specifically, we found support for a positive effect of alliance portfolio industry diversity. For the other hypotheses, the nature of the relationships diverted from what was hypothesized. Alliance portfolio functional diversity is shown to have a U-shape relation with firms’ technological diversification, whereas the relation between alliance portfolio geographical diversity and technological diversification is found to be insignificant.
In this way, this research shows that access to a variety of knowledge sources not only facilitates the development of new ideas and knowledge, as argued by the organizational learning theory and the knowledge-based view, but it also broadens the scope of this new knowledge. This is an interesting contribution to the existing literature, because it indicates that firms can rely on alliance portfolios as a strategic tool to diversify their technological portfolio. Our study also emphasizes the need to make an explicit distinction between different dimensions of alliance portfolio diversity, showing that alliance portfolio industry diversity, alliance portfolio functional diversity, and alliance portfolio geographical diversity have a differential impact on technological diversification.
This paper will now proceed as follows; the next section will discuss the theoretical background of this study by relating alliance portfolio diversity to technological diversification. We also elaborate on the different dimensions of alliance portfolio diversity and subsequently derive the hypotheses. Thereafter, the methodology is described that is used to conduct this study and the empirical results are presented. The remaining sections will discuss the results in more depth, suggest avenues for further research and conclude the main findings of this study.
ALLIANCE PORTFOLIO DIVERSITY AND TECHNOLOGICAL PORTFOLIOS
A portfolio of different types of alliances exposes companies to many and a variety of ideas and knowledge, which contributes to the development of innovations (George et al., 2001). In accordance, Duysters and Lokshin (2011) find that alliance portfolios of innovators are more diverse in terms of type of partners and nationalities than firms that do not innovate. On the other hand, diverse portfolios are also more complex to manage and more vulnerable to conflicts. This increases coordination and managerial costs (Jiang et al., 2010).
While the majority of the existing literature (e.g. Luo and Deng, 2009; Sarkar et al., 2009; Faems et al., 2010; Sivakumar et al., 2011) measures alliance portfolio diversity as a single dimension, recently scholars have started to study diversity in more detail by developing multi-dimensional constructs (Jiang et al., 2010; Duysters and Lokshin, 2011). Duysters and Lokshin (2011) extend the type of partner with a distinction between national and international alliances. In addition, rather than only looking at the characteristics of the alliance partners, Jiang et al. (2010) consider the characteristics of the alliances themselves, such as their functional purposes and governance structures.
Recent research has examined whether alliance portfolio diversity increases the size of firms’ technology portfolios, i.e. the generation of patents. Although these studies operationalize the construct ‘alliance portfolio diversity’ in different ways, they reach the same conclusion, namely that the relationship is found to have an inverted U-shape (Nooteboom et al., 2007; Luo and Deng, 2009). This indicates that a moderate level of alliance portfolio diversity leads to a higher level of technological performance in comparison to a low or high level of diversity. An overview of the existing literature that examines this relationship is provided in table 1.
a patent portfolio received, rather than the effect on its size. The results of their study show an inverted U-shaped relation with the number of citations that a technology portfolio received.
Table 1: The existing literature on the link between alliance portfolio diversity and technology portfolios Article Theoretical
background Independent variable Operationalization Dependent variable Operationalization Result Nooteboom et al. (2007) Resource-based view; organizational learning; absorptive capacity Cognitive distance Average of the correlations between the focal firm’s technology profile and that of each of its alliance partners Technological performance Number of patents, with a distinction between more exploitative and explorative patents Inverted U-shaped relation, with a stronger effect in exploration than in exploitation Luo and
Deng (2009) Social network; organizational learning; organizational ecology and institutional perspectives Similarity Percentage of similar partners in a focal DBF’s alliance portfolio
Innovation Number of patents issued to the firm in a year
Inverted U-shaped relation
Phelps
(2010) Recombinatory search; social network Network technological diversity Knowledge distance between a focal actor and each of its partners and the distances among the partners
Exploratory
innovation New patent citations divided by total patent citations for each year per firm (7 year window) Positive effect Vasudeva and Anand (2011) Absorptive capacity; organizational learning Technological
diversity Pooling of the patents of all alliance partners in each focal firm’s portfolio Knowledge utilization from portfolio Citation-weighted patent count in the lagged five-year window following an observation year Inverted U-shaped relation
amount of new patents, which advocates composing an alliance portfolio with a moderate level of diversity.
While the existing literature concentrates on the economies of scale, i.e. the generation of more patents (citations), the performance advantage in R&D activities appears to lie in economies of scope (Quintana-García and Benavides-Velasco, 2008). Rather than counting the number of (cited) patents, examining the impact on the diversity of the patents, i.e. the technological diversification of firms, can yield interesting results. Moreover, Antonelli and Calderini (2008) state that it is important to evaluate both the quantity as well as the quality of technology portfolios in order to get a real understanding of the technological competences of firms. However, to the extent of our knowledge, the relationship between alliance portfolio diversity and technological diversification has remained unexplored. Next, the concept of technological diversification will be briefly discussed, where after we will elaborate on its relationship with alliance portfolio diversity.
THE RELEVANCE OF TECHNOLOGICAL DIVERSIFICATION
Quintana-García and Benavides-Velasco (2008) define technological diversification as ‘the diversity in the knowledge system and principles underlying the nature of products and their methods of production’ (p. 492), or, to put it differently, it is the ability of firms to patent in different technological domains. It can also be seen as a reflection of a firm’s knowledge portfolio.
portfolio have a higher market value, measured by stocks, compared to companies with a more focused portfolio. Moreover, the study of Belderbos et al. (2010) shows a peak in the financial performance of firms, when their technology portfolio is balanced in terms of explorative and exploitative activities, i.e. a high level of technological diversification. Technological diversification can also enhance a firm’s investments in R&D as found by Garcia-Vega (2006).
Furthermore, owning patents in a large variety of technological fields facilitates the creation of new patents as technological diversification is found to increase the number of (cited) patents (Garcia-Vega, 2006; Quintana-García and Benavides-Velasco, 2008; Phelps, 2010). On the other hand, the results of Leten et al. (2007) show an inverted U-shaped relation between technological diversification and the number of patent applications, indicating that a moderate level of technological diversification leads to the highest technological performance. They argue that besides the positive effects of a higher level of technological diversification, there also additional coordination and integration costs.
Finally, Quintana-García and Benavides-Velasco (2008) have examined how technological diversification affects different types of innovation, i.e. exploration and exploitation. Although positive and significant results are found for both types, technological diversification has a larger effect on exploratory patents. Moreover, these authors stress that it is important for firms to develop capabilities to balance exploratory and exploitative activities. The positive effect of technological diversity on exploratory innovation is also confirmed by Phelps (2010). In his study, technological diversification is found to increase the relative share of new patent citations.
Table 2: The existing literature on the effect of technological diversification Article Theoretical
background Independent variable Operationalization Dependent variable Operationalization Result Garcia-Vega (2006) No specific theory is addressed Diversification 1 – Herfindahl index of concentration of a patent portfolio Innovative
activity (1) R&D intensity (R&D / sales) (2) Number of patents Positive relations Leten et al. (2007) No specific theory is addressed Technological
diversification The spread of the patent portfolio over technological classes
(Herfindahl index)
Patents Yearly number of firm EPO applications Inverted U-shaped relation Antonelli and Calderini (2008) Economics of
knowledge Knowledge compositeness Weighted measure of differentiation that provides a fair account of the knowledge compositeness of company’s portfolio.
(Technological)
Market share (1) Relative share of patents owned by each car maker over the patent production in a given year. (2) Markets shares
One and two year positive lagged effects Quintana-García and Benavides-Velasco (2008) Evolutionary theory and technological trajectories; organizational learning; absorptive capacity Technological
diversification Herfindahl index: 1 - sum(pi2),
where p is the proportion in a firm of patents in technological field i. Innovative
competence Number of patents granted by the firm in a year Positive relation (larger impact on exploratory innovative competence than on exploitative innovative competence) Belderbos et al. (2010) Absorptive
capacity Technological activities Share of explorative technological activities (situated in a technology domain that is new or unfamiliar to the firm in the past 5 years)
Financial
performance Tobin’s Q, which is a ratio of the market value of a firm and the book value of the firm’s assets An inverted U-shape Chen and Chang (2010) No specific theory is addressed
HHI of patents Herfindahl index of patent classes of a firm
Market value Average stock price of a company in a given year multiplied by the number of common stock shares outstanding Negative relation Phelps
(2010) Recombinatory search; social network
Technological
diversification The spread of the patent portfolio over technological classes
(Herfindahl index)
Exploratory
innovation New patent citations divided by total patent citations for each year per firm (7 year window)
portfolio diversity and technological diversification has not been studied so far. This is an interesting gap, because both alliance portfolios and technological diversification are shown to be valuable for firms. Moreover, existing research has shown that alliance portfolio diversity has a significant impact on the quantity of firms’ technology portfolio, whereas it is also interesting to examine whether it affects the quality of technology portfolios. Next, we will explore the relationship between alliance portfolio diversity and technological diversification in more depth. In this current study, we also aim to examine the effect of different dimensions of alliance portfolio diversity. Three dimensions will be considered to have an influence, namely alliance portfolio industry diversity, alliance portfolio functional diversity and alliance portfolio geographical diversity.
THE RELATIONSHIP BETWEEN DIFFERENT TYPES OF ALLIANCE PORTFOLIO DIVERSITY AND TECHNOLOGICAL DIVERSIFICATION
According to organizational learning theories, access to heterogeneous contexts facilitates the creation of new knowledge. A diverse alliance portfolio exposes a firm to variety of knowledge, ideas and experiences (Levitt and March, 1988; Levinthal and March, 1993; George et al., 2001), which facilitates the creation of new linkages between internal knowledge and incoming, i.e. external, knowledge (Cohen and Levinthal, 1990; Lin, 2011). The above standing process of combining internal and external knowledge can result in the creation of patents in unfamiliar technological domains, which subsequently increases the diversity of a firm’s technology portfolio.
development of new ideas. This access will also increase the probability that new incoming information is related to existing knowledge base, and therefore provides a more robust setting for learning to take place (Cohen and Levinthal, 1990; Powell et al., 1996). Additionally, there is a higher probability that at least some of the obtained knowledge will lead to a successful innovation outcome as argued by Leiponen and Helfat (2010).
Although Cohen and Levinthal (1990) state that learning is difficult in novel domains, knowledge diversity can increase the absorptive capacity of a firm (Quintana-García and Benavides-Velasco, 2008; Phelps, 2010; Lin, 2011), which is the ability of firms to absorb, learn and use external knowledge (Cohen and Levinthal, 1990). As shown by Powell et al. (1996), firms with a higher level of absorptive capacity learn more from their collaborative activities.
The knowledge-based view of the firm also acknowledges the advantages of having access to a diverse set of knowledge. This perspective focuses on the fit between product domain and the knowledge domain of a firm (Grant, 1997). Most knowledge is subject to economies of scale, i.e. low replication costs of knowledge, and economies of scope, i.e. a specific piece of knowledge can be used for the production of many products (Grant and Baden-Fuller, 2004). Strategic alliances provide access to the relevant knowledge for a specific product without the costs of owning the knowledge (Grant, 1996) and in this way optimize the economies of scope of knowledge.
technological fields in which a firm patents. Next, we will elaborate on the effects of the different dimensions of alliance portfolio diversity, and subsequently derive hypotheses.
Alliance portfolio industry diversity
Firms can form alliances with partners who are active in different industries. Collaborating with similar partners, i.e. from the same industry, reduces coordination costs and eases the transfer, assimilation and use of knowledge due to the overlap in prior related knowledge (Cohen and Levinthal, 1990, Lui and Ngo, 2005; Luo and Deng, 2009). However, if all alliance partners of a firm are similar, an additional alliance with a similar partner will not yield extra value for the development of innovations (Luo and Deng, 2009). Accessing similar capabilities and knowledge can cause problems of redundancy, in which new combinations of existing knowledge are exhausted (Vasudeva and Anand, 2011). Additionally, if a company only cooperates with similar partners, it is unlikely that it will explore new technological domains.
Collaboration with partners outside the industry of the focal firm reduces this problem by providing new knowledge, capabilities, trends, and practices (Vasudeva and Anand, 2011). This can accordingly be used to make novel associations and linkages with a firm’s internal knowledge activities (Cohen and Levinthal, 1990; Lin, 2011). In addition, allying with firms that are already active in a particular industry, market entry to this new unfamiliar industry is facilitated by providing legitimacy (Eisenhardt and Schoonhoven, 1996; Duysters and Lokshin, 2011).
diverse set of partners, will increase the variety of technology fields in which a firm patents, i.e. its technological diversification. This leads to the following hypothesis:
H1: Alliance portfolio industry diversity increases firms’ technological diversification.
Alliance portfolio functional diversity
Firms can also form alliances in different stages in the value chain of a product (Rothaermel and Deeds, 2004; Jiang et al., 2010). These different stages require different types of searches for knowledge and therefore the formation of different types of alliances (Rothaermel and Deeds, 2004). In the early stages of product development, the exploration of new combinations of knowledge plays a central role, which advocates the use of R&D alliances (Jiang et al., 2010). These upstream alliances are usually formed with universities and other research institutes (Rothaermel and Deeds, 2006). In later stages, where the focus shifts to the exploitation of the created innovation, firms will employ marketing, manufacturing and distribution alliances (Jiang et al., 2010). These alliances are also called downstream alliances (Hess and Rothaermel, 2011). The above standing studies indicate that exploration alliances are formed with different types of partners than in case of exploitation alliances. For instance, in the pharmaceutical industry, exploration alliances will mostly be formed with biotechnology firms, whereas exploitation alliances are usually formed with large pharmaceutical companies (Rothaermel and Deeds, 2004). Moreover, Chesbrough (2003) states that many firms concentrate on one of the three types of activities, namely funding, generating or commercializing innovations.
activities is essential for the continuation of firms. This indicates that firms forming alliances with different functional purposes during the different stages of the value chain have access to a wide variety of external knowledge. This access facilitates the development of new linkages and combinations, which can accordingly increase the number of technological fields in which a firm patents.
Concluding, balancing exploration and exploitation activities can increase the technological competences of firm by simultaneously exploiting familiar technological domains and exploring unfamiliar technological domains. Accordingly, firms can apply for patents in different stages of the value chain, resulting in an increase in the technological diversification of firms. Therefore, the following hypothesis is stated:
H2: Alliance portfolio functional diversity increases firms’ technological diversification.
Alliance portfolio geographical diversity
Finally, firms can form alliances with companies located in different countries across the world. This dimension becomes increasingly relevant, as technological knowledge is more and more dispersed over the world (Archibugi and Iammarino, 2002; Duysters and Lokshin, 2011). Moreover, Inkpen (1998) argues that international strategic alliances can create unique learning opportunities for firms by providing access to different skills, knowledge bases and organizational cultures. The creation of new knowledge due to the exposure to new contexts is also supported by the organizational learning theory.
collaborating with foreign partners can provide access to specialized technological expertise, which is not available locally (Duysters and Lokshin, 2011). Because the knowledge from international partners is less available, it is also more valuable than knowledge acquired from domestic alliances (Zhang et al., 2010).
Hence, we argue that if a firm allies with partners from different regions across the world, it has better access to new and specialized knowledge compared to companies that only cooperate locally. This broader access to knowledge facilitates the development of new ideas and subsequently the creation of unfamiliar patents. Moreover, we expect that these firms are able to patent in a wider variety of technological classes. This leads to the following hypothesis:
H3: Alliance portfolio geographical diversity increases firms’ technological diversification.
METHODOLOGY
To test the hypotheses, a panel dataset has been constructed containing the alliance portfolios and patent portfolios of 43 pharmaceutical firms for a time period of eight years (1995 – 2002). These firms have been selected from the ‘2004 EU Industrial R&D Investment Scoreboard’ as the top R&D spending pharmaceutical firms.
To construct the alliance portfolios, we used the Thomson SDC Platinum database. This database is based on press releases in which new alliances are announced. For each of the 43 pharmaceutical firms, we downloaded all announcements for the selected years. Next, the downloaded text files were transformed to an Excel-format, after which they were imported and merged in the software program STATA.
Several modifications had to be done to make the dataset usable. The original dataset consisted both of alliance-level variables and firm-level variables. The former were only listed for one of the partners in the alliance and therefore had to be duplicated for the other partner(s). Moreover, some variables were coded as strings, e.g. ‘Yes’ or ‘No’. These variables were recoded in numerical values in order to be able to include them in regression analyses. Moreover, we analyzed whether the names of the companies were consistent and adjusted them if necessary.
Finally, if a company resulted from a large merger after 1995, it was only included in the sample after its foundation, which makes the dataset slightly unbalanced1. In total, the data set contains 294 firm-year observations.
Dependent variable
In line with prior research (Leten et al., 2007; Quintana-García and Benavides-Velasco, 2008), as can also be seen in table 2, we measure technological diversification of firms as the variety of technological classes in which a firm has applied for a patent in a particular year. The patent data is collected from the European Patent Office (EPO). Each granted patent is assigned at least to one of the IPC (International Patent Classification) classes. For our diversification measure, a distinction is made between 625 different 4-digit IPC classes.
Based on this classification, a variation of the Herfindahl index is used to construct the dependent variable, namely the Blau Index of Variability (Blau, 1977). According to Quintana-García and Benavides-Velasco (2008) and Leten et al. (2007), the Blau Index has been widely used to measure the heterogeneity of categorical variables. The Index can be calculated as follows: D = 1 – Σpi2, where D is the degree of diversity, p is the proportion belonging to given category (i.e. IPC class), and i is the number of total number of categories in which a firm patents. The range of the dependent variable is between 0 and 1, where 0 represents zero technological diversification and 1 indicates a fully balanced patent portfolio in which patents are equally distributed across the classes in which a firm patents.
Independent variables
As argued in this paper, alliance portfolio diversity can be measured in different ways. We focus on three different dimensions, namely alliance portfolio industry diversity, alliance
1 The following firms resulted from a larger merger after 1995 and were included in the dataset after their
portfolio functional diversity and alliance portfolio geographical diversity. As a first step, the characteristics of each alliance were coded into different categories. Next, we will discuss which categories were used to construct the three diversity measures.
First, alliance portfolio industry diversity focuses on the overlap in the four-digit SIC
codes of the focus firm and its partner(s) in a particular alliance (Keil et al., 2008). In this way, alliance portfolio industry diversity represents the variety of operational alliances in terms of industry relatedness. An alliance was coded as ‘0’, when the SIC codes were unrelated; as ‘1’, when they shared the same one- or two-digit SIC code indicating a related industry; and finally, as ‘2’, when an alliance was formed between partners with the same three- or four-digit SIC codes referring to an intra-industry alliance.
Second, alliance portfolio functional diversity concentrates on the variety of
operational alliances in terms of functional activities. While Jiang et al. (2010) code functional activities as R&D, marketing, manufacturing or other type of alliances, we argue that this categorization is too broad. Following the existing literature, we distinguish between explorative and exploitative purposes. The functional activity of an alliance is coded as follows: ‘1’ for R&D alliances, indicating explorative activities, and ‘2’ for manufacturing, marketing and other type of alliances2, indicating exploitative activities.
At last, alliance portfolio geographical diversity represents the variety of operational
alliances in terms of geographical coverage. Geographical diversity was defined as the overlap in the world regions in which the focal firm and its partner(s) were located. Each partner firm was coded based on its location in one of the six categories: ‘1’for North America, ‘2’ for South America, ‘3’ for Africa, ‘4’ for Europe, ‘5’ for Asia, and ‘6’ for Australia.
Likewise, the Blau Index of Variability was used to construct the three portfolio diversity measures. To clarify, we provide a brief example how we calculated this index for one of these dimensions. In our example, firm A has an alliance portfolio that consists of 20 partners, namely 4 North American, 3 South American, 1 African, 8 European, 2 Asian and 2 Australian partners. By squaring the proportions of each of the six categories, the Blau index is calculated as follows:
D = 1 – Σpi2 = 1 – [(4/20)2 + (3/20)2 + (1/20)2 + (6/20)2 + (2/20)2 + (4/20)2] = 0,795. A value close to zero represents a low level of diversity, whereas a higher value indicates that a firm’s alliance portfolio is more balanced with equal proportions for each category.
In addition, squared terms of each dimension of alliance portfolio diversity were created. The existing literature that examines the impact of alliance portfolio diversity on the number of patents finds an inverted U-shape relation, indicating that firms face a certain cognitive limit that constrains learning and the transfer of knowledge (Levitt and March, 1988; Cohen and Levinthal, 1990; Levinthal and March, 1993; Grant, 1996; Grant and Baden-Fuller, 2004). By adding squared terms, we are able to investigate whether the hypothesized relationships are non-linear or marginally increasing rather than linear.
Control variables
Second, some firms differ have experience in conducting R&D activities than others. This can be seen as the variations in the return on R&D or the efficiency of converting R&D efforts into patents. Hence, we control for these differences by including the patent propensity. This is measured by dividing the number of patent applications that a firm received in particular year by its R&D expenditures in that year.
Third, we control for the alliance experience of firms. Existing literature (e.g. Sivakumar et al., 2011) shows that companies with more experience in managing past alliances have a higher innovation performance. Alliance experience is measured as a logarithm of the number of operational alliances.
In addition, the larger a firm is, the higher its innovation performance (Rothaermel and Deeds, 2004, Luo and Deng, 2009, Sivakumar et al., 2011). Duysters and Lokshin (2011) argue that larger firms have more resources available for the management of alliance portfolios in comparison to smaller firms, which often face financial and managerial constraints (Duysters and Lokshin, 2011). We control for this by including an indicator for firm size, namely a logarithm of the number of employees. The data on R&D expenditures, sales, and employees is gathered from corporate annual reports and financial databases (i.e. Worldscope and Compustat).
Finally, time dummies are included to account for time-varying effects that are related to firms’ technological diversification, but are identical for all firms in the sample. These could, for instance, be industry-related or macroeconomic effects. The year 1995, the first year in the sample, is used as the reference category.
Analytical method
unobserved individual effects, which are present in panel data. Fixed effects regression assumes that the explanatory variables are related to these unobserved individual effects. Moreover, this method aims to identify differences within groups. The random effects regression, on the other hand, assumes that there is no relation between the explanatory variables and the unobserved individual effects. Because the dataset contains repeated observations for the same groups (i.e. firms), this assumption is not likely to hold. A Hausman test has been conducted to test whether the fixed effects estimators and random effects estimators were systematically different. The test statistic advocates the use of the fixed effects method to estimate our proposed models (Model 8 ߕଵହଶ = 36.56, p = 0.0015). In addition, the fixed effects regression results provide significant evidence of a correlation between the explanatory variables and the unobserved individual effects (F = 7.31, p = 0.0000).
EMPIRICAL RESULTS
Table 3 presents the descriptive statistics and correlations. The technological diversification of firms slightly increased over time, as the average value was 0.7882 in 1995 and 0.7939 in 2002. Also the average number of applied patents increased from 55 per year (in 1995) to 90 per year (in 2002). The sample consisted of large firms with on average 28174 employees, annual sales of $9088796 million and spending $1133632 million on R&D (in 2002). Additionally, table 3 shows that the alliance portfolios of the selected pharmaceutical firms were quite diverse for each of the dimensions.
larger in size also have more alliances, which seems plausible. Moreover, the VIF measures are all below a value of 4 suggesting that multicollinearity of the variables is not a problem in our analyses. We also tested whether our main independent variables were endogenous, i.e. whether technological diversification also affects the alliance portfolio diversity of firms. However, we did not find evidence of endogeneity for each of the three dimensions3.
Table 3: Descriptive statistics and correlations Variables Mean SD 1 2 3 4 5 6 7 1 Technological diversification 0.7824 0.1069 2 Patent propensity 0.0124 0.0142 0.2520 3 Log (employees) 9.2549 1.4366 0.6118 0.0226 4 Log (alliances) 2.6779 1.0216 0.4835 0.0139 0.7824 5 R&D intensity 0.1777 0.4350 -0.1598 0.0607 -0.3402 0.0109 6 Industry diversity 0.4263 0.2067 0.3553 0.0864 0.3812 0.4448 -0.1434 7 Functional diversity 0.3705 0.1290 0.3247 -0.0099 0.2961 0.4294 0.1174 0.3356 8 Geographical diversity 0.4670 0.1694 0.2472 0.1010 0.2147 0.2631 -0.1019 0.4279 0.4031 The results of the fixed effects regression analyses are provided in table 4. Model 1 is the
baseline model by including only the control variables. The number of operational alliances has a positive, significant effect on the technological diversification of firms (p < 0.10). For the other control variables we did not find any significant results. The linear terms of industry, functional and geographical diversity are added in Model 2, 4 and 6 respectively. Subsequently, the squared terms were added in model 3, 5 and 7. Finally, we jointly test the effects of the three dimensions of alliance portfolio diversity in model 8.
Hypothesis 1 predicted a positive relation between alliance portfolio industry diversity and technological diversification, which will be analyzed based on the results of model 2 and model 3. Model 2 shows a positive and significant root term for industry diversity (p = 0.074),
3
We conducted several Hausman tests in order to investigate whether our three main independent variables were endogenous. Under the null hypothesis, the coefficients from the OLS and the IV regression are similar. The alternative hypothesis states that only the IV estimators are consistent. As an instrument for ‘industry diversity’, we used the other two dimensions (functional and geographical diversity). The approach for the other two dimensions was the same. The results show that industry diversity (Model 2 ߕଵଷଶ= 17.98, p = 0.1583), functional
providing support for hypothesis 1. Further analyses were conducted to analyze the existence of a non-linear relationship. The result of a joint F-test of the root and the squared terms in model 3 is insignificant (p = 0.1925). Additionally, a more formal ‘utest’ was conducted, developed by Lind and Mehlum (2010). This test evaluates whether the relationship is U-shaped or inverted U-U-shaped by determining whether the inflection point is between the minimum and the maximum value of the explanatory variable. The results of this test show that the extreme point is outside the interval, indicating that the relation between alliance portfolio industry diversity is linear and positive. Hence, we accept hypothesis 1.
Alliance portfolio functional diversity was also predicted to have a positive impact on technological diversification as stated in hypothesis 2. Model 4 shows a negative coefficient for functional diversity, however it is marginally insignificant (p = 0.1052). Subsequently, a squared term is added in model 5. The root and the squared terms are jointly significant (p = 0.022). The negative coefficient of the root term and the positive coefficient of the squared term indicate the presence of a U-shaped relation. The ‘utest’ of Lind and Mehlum (2010) provides a test value of 1.39, which is significant at the 10% level (p = 0.0828). Concluding, because the findings show a U-shaped relation between alliance portfolio functional diversity and technological diversification, we fail to accept hypothesis 2.
Table 4: Regression analyses4
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8
Patent Propensity 0.221 0.333 0.300 0.176 0.352 0.310 0.319 0.554 (0.659) (0.659) (0.668) (0.657) (0.656) (0.660) (0.668) (0.657) Logarithm of # Employees 0.020 0.022 0.022 0.017 0.023 0.018 0.018 0.025 (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) Logarithm of # Alliances 0.026* 0.016 0.016 0.036*** 0.042*** 0.035** 0.035** 0.035** (0.013) (0.014) (0.014) (0.015) (0.015) (0.015) (0.015) (0.015) R&D Intensity -0.013 -0.011 -0.011 -0.013 -0.014 -0.015 -0.015 -0.013 (0.017) (0.017) (0.017) (0.017) (0.016) (0.017) (0.017) (0.015) Industry Diversity 0.065* 0.091 0.090** (0.036) (0.089) (0.037)
Industry Diversity Squared -0.042
(0.133)
Functional Diversity -0.082 -0.374*** -0.382**
(0.051) (0.139) (0.154)
Functional Diversity Squared 0.545** 0.553**
(0.242) (0.260)
Geographical Diversity -0.051 -0.042 -0.027
(0.035) (0.098) (0.040)
Geographical Diversity Squared -0.013
(0.1 )30 Intercept 0.511*** 0.487*** 0.491*** 0.535*** 0.488*** 0.524*** 0.525*** 0.462*** (0.177) (0.177) (0.178) (0.177) (0.177) (0.177) (0.178) (0.176) Number of Observations 294 294 294 294 294 294 294 294 Number of Firms 43 43 43 43 43 43 43 43 Model F 2.376 2.468 2.277 2.414 2.658 2.362 2.172 2.736 R-Squared Within 0.098 0.110 0.111 0.108 0.127 0.106 0.106 0.148 Between 0.442 0.481 0.474 0.382 0.415 0.397 0.398 0.442 Overall 0.388 0.417 0.411 0.347 0.368 0.363 0.363 0.395
Finally, we estimate the joint effect of the three dimensions of alliance portfolio diversity. Based on the previous findings, we include, next to the root terms of industry, functional and geographical diversity, a squared term for functional diversity in model 8. The joint F-test is highly significant (p = 0.0091). This finding shows that alliance portfolio diversity influences the technological diversification of firms. Because we found diverging results for the effects of the different dimensions, we cannot infer anything about the nature of the relationship between both constructs.
When looking at the individual results of the three dimensions in model 8, we observe that their significance levels have changed. Industry diversity has a stronger, positive effect on technological diversification (p = 0.016) as well as the U-shaped effect of functional diversity (p = 0.040). The relation between geographical diversity and technological diversification remains insignificant (p = 0.498). A further discussion of our results and their implications will be provided below.
DISCUSSION AND CONCLUSION
findings provide support that alliance portfolio diversity significantly influences firms’ technological diversification, the nature of this relationship varies between the three dimensions of alliance portfolio diversity.
First of all, a positive, significant effect is found for alliance portfolio industry diversity, indicating that collaborating with partners from a diverse set of industries facilitates the creation of new linkages and associations (Vasudeva and Anand, 2011), and accordingly provides firms the opportunity to patent in new technological domains. In addition, no evidence was found of the negative effects of high level of industry diversity. This finding suggests that pharmaceutical companies should both form intra- and inter-industry alliances in order to increase their technological diversification.
Still, we observe that some firms apparently are able to be successful in terms of technological diversification while only conducting exploitative alliances. Because firms increasingly concentrate on a particular type of functional activity, as argued by Chesbrough (2003), some firms might decide to conduct their explorative activities in-house, while collaborating on exploitative activities. This might allow firms to concentrate their alliance activities, while being able to patent in a variety of technological fields. However, this statement has not been studied so far, and might be an interesting avenue for future research. Overall, if a pharmaceutical company wants to increase its technological diversification, an intermediate level of functional diversity is found to be the least successful.
Third, our study did not find a significant relation between geographical diversity and technological diversification. Although the literature states that knowledge is more and more dispersed over the world (Duysters and Lokshin, 2011), which would advocate the formation of alliances with partners in different world regions to be able to access this information, we did not find support for this statement. This is an interesting result, because this might indicate that the same knowledge is available everywhere (in the pharmaceutical industry) or that the location of an alliance partner does not matter. Future research could examine this finding in more depth in order to analyze whether this insignificant relation also holds in other industries.
research finds an inverted U-shape between alliance portfolio diversity and the size of a technology portfolio, indicating that the effect of alliance portfolio diversity becomes negative after a certain threshold is reached. Our results, on the other hand, do not find negative effects at a high level of diversity and therefore clearly show the benefits of creating of a highly diverse alliance portfolio. This indicates that it is important to distinguish between the effect on the size and the composition firms’ patent portfolios.
Furthermore, as advocated by the organizational learning theory and the knowledge-based view of the view (Cohen and Levinthal, 1990; George et al., 2001; Grant, 1996; Lin, 2011), our study acknowledges the importance of being able to access a diverse set of knowledge for the development of new knowledge and innovations. As a new contribution to the existing literature, this study shows that access to a variety of knowledge also broadens the scope of the knowledge developed by the firm. This indicates that firms can rely on alliance portfolios as a strategic tool to diversify their technological portfolio.
Besides, only recently, studies have started to operationalize alliance portfolio diversity as a multi-dimensional construct, e.g. Jiang et al. (2010). Our results confirm that diversity is not a single dimensional construct by showing that three dimensions of alliance portfolio diversity have a differential impact on the technological diversification of firms. Hence, this study aims to contribute to this new development of analyzing the effects of alliance portfolio diversity in more depth.
with whom to form an alliance and for which purpose. The results of this study suggest that it is important for firms to map their alliance portfolio to increase their ability to evaluate the added value of a potential alliance to the portfolio. In this way, managers can analyze how this potential addition changes the diversity of their alliance portfolio, which can accordingly impact the technological diversification of their firm.
For policymakers it is important as well to realize that they can increase the innovative capabilities of their country by stimulating the collaboration within and between industries. For instance, the Dutch government has developed a new policy program in order to become one of the top 5 knowledge economies in the world. To achieve this goal, 9 ‘top sectors’ have been identified, in which governments, companies, university and research institutes agreed to collaborate. However, this program mainly focuses on the collaboration within industries, while the results of our research also stress the importance of collaborating between industries. This is an issue that needs attention from the Dutch government if they want achieve their goal.
This study knows several limitations. First of all, our empirical analyses concentrated on the pharmaceutical industry. Although the existing alliance portfolio research largely focused on this industry, it might be interesting to examine other industries as well. This will allow us to identify differences and similarities across industries. Furthermore, the 43 largest firms in terms of R&D expenditures were selected. A future study could extend this sample by also including smaller firms, which would provide a more complete overview of this industry. A third concern is the time frame of eight years, which could be extended in future studies.
future studies could examine the type of knowledge, i.e. tacit or explicit, that is transferred between the different partners. In this way, we might identify which type of knowledge is most valuable to increase the technological diversification of firms. Finally, although this study combined several theoretical perspectives, a clear theoretical framework in this field of research is missing. Future research should therefore aim to strengthen the theoretical underpinnings of alliance portfolio research.
REFERENCES
Antonelli C, Calderini M. 2008. The Governance of Knowledge Compositeness and Technological Performance: The Case of the Automotive Industry in Europe. Economics of Innovation and
New Technology 17(1-2): 23-41.
Archibugi D, Iammarino S. 2002. The Globalization of Technological Innovation: Definition and Evidence. Review of International Political Economy 9(1): 98-122.
Belderbos R, Faems D, Leten B, Van Looy B. 2010. Technological Activities and Their Impact on the Financial Performance of the Firm: Exploitation and Exploration within and between Firms.
Journal of Product Innovation Management 27(6): 869-882.
Blau PM. 1977. Inequality and Heterogeneity: A Primitive Theory of Social Structure. Free Press: New York.
Cassiman B, Veugelers R. 2006. In Search of Complementarity in Innovation Strategy: Internal R&D and External Knowledge Acquisition. Management Science 52(1): 68-82.
Chen YS, Chang KC. 2010. The Relationship between a Firm’s Patent Quality and its Market Value – The Case of US Pharmaceutical Industry. Technological Forecasting & Social Change 77(1): 20-33.
Chesbrough HW. 2003. The Era of Open Innovation. MIT Sloan Management Review 44(3): 35-41. Cohen WM, Levinthal DA. Absorptive Capacity: A New Perspective on Learning and Innovation.
Administrative Science Quarterly 35(1): 128-152.
Duysters G, Lokshin B. 2011. Determinants of Alliance Portfolio Complexity and Its Effect on Innovative Performance of Companies. Journal of Product Innovation Management 28(4): 570-585.
Eisenhardt KM, Schoonhoven CB. 1996. Resource-based View of Strategic Alliance Formation: Strategic and Social Effects in Entrepreneurial Firms. Organization Science 7(2): 136-150. Faems D, de Visser M, Andries P, Van Looy B. 2010. Technology Alliance Portfolios and Financial
Performance: Value-Enhancing and Cost-Increasing Effects of Open Innovation. Journal of
Product Innovation Management 27(6): 785-796.
George G, Zahra SA, Wheatley KK, Khan R. 2001. The Effects of Alliance Portfolio Characteristics and Absorptive Capacity on Performance: A Study of Biotechnology Firms. Journal of High
Technology and Management Research 12(2): 205-226.
Grant RM. 1996. Toward a Knowledge-Based Theory of the Firm. Strategic Management Journal 17(SI): 109-122.
Grant RM. 1997. The Knowledge-Based View of the Firm: Implications for Management Practice.
Long Range Planning 30(3): 450-454.
Grant RM, Baden-Fuller C. 2004. A Knowledge Accessing Theory of Strategic Alliances. Journal of
Management Studies 41(1): 61-84.
Gulati R. 1995. Social Structure and Alliance Formation Patterns: A Longitudinal Analysis.
Administrative Science Quarterly 40(4): 619-652.
Hagedoorn J. 2002. Inter-firm R&D Partnerships: An Overview of Major Trends and Patterns since 1960. Research Policy 31(4): 477-492.
Hess AM, Rothaermel FT. 2011. When are Assets Complementary? Star Scientists, Strategic Alliances, and Innovation in the Pharmaceutical Industry. Strategic Management Journal 32(8): 895-909.
Inkpen AC. 1998. Learning and Knowledge Acquisition through International Strategic Alliances.
Academy of Management Executive 12(4): 69-80.
Jiang RJ, Tao QT, Santoro MD. 2010. Alliance Portfolio Diversity and Firm Performance. Strategic
Management Journal 31(10): 1136-1144.
Keil T, Maula M, Schildt H, Zahra SA. 2008. The Effect of Governance Modes and Relatedness of External Business Development Activities on Innovative Performance. Strategic Management
Journal 29(8): 895-907.
Lavie D. 2007. Alliance Portfolios and Firm Performance: A Study of Value Creation and Appropriation in the U.S. Software Industry. Strategic Management Journal 28(12): 1187-1212.
Leten B, Belderbos R, Van Looy. 2007. Technological Diversification, Coherence, and Performance of Firms. Journal of Product Innovation Management, 24(6): 567-579.
Levinthal DA, March JG. 1993. The Myopia of Learning. Strategic Management Journal 14(S2): 95-112.
Levitt B, March JG. 1988. Organizational Learning. Annual Review of Sociology 14: 319-340.
Lin BW. 2011. Knowledge Diversity as a Moderator: Inter-firm Relationships, R&D Investment and Absorptive Capacity. Technology Analysis & Strategic Management 23(3): 331-343.
Lind JT, Mehlum H. 2010. With or Without U? The Appropriate Test for U-Shaped Relationship.
Oxford Bulletin of Economics and Statistics 72(1): 109-118.
Lui SS, Ngo HY. 2005. An Action Pattern Model of Inter-firm Cooperation. Journal of Management
Studies 42(6): 1123-1153.
Luo X, Deng L. 2009. Do Birds of a Feather Flock Higher? The Effects of Partner Similarity on Innovation in Strategic Alliances in Knowledge-Intensive Industries. Journal of Management
Studies, 46(6): 1005-1030.
March JG. 1991. Exploration and Exploitation in Organizational Learning. Organization Science 2(1): 71-87.
Nooteboom B, Van Haverbeke W, Duysters G, Gisling V, Van den Oord A. 2007. Optimal Cognitive Distance and Absorptive Capacity. Research Policy 36(7): 1016-1034.
Parise S, Casher A. 2003. Alliance Portfolios: Designing and Managing Your Network of Business-Partner Relationships. Academy of Management Executive 17(4): 25-39.
Phelps CC. 2010. A Longitudinal Study of the Influence of Alliance Network Structure and Composition on Firm Exploratory Innovation. Academy of Management Journal, 53(4): 890-913.
Powell WW, Koput KW, Smith-Doerr L. Interorganizational Collaboration and the Locus of Innovation: Networks of Learning in Biotechnology. Administrative Science Quarterly 41(1): 116-145.
Rosenkopf L, Almeida P. 2003. Overcoming Local Search Through Alliances and Mobility.
Management Science 49(6): 751-766.
Rothaermel FT, Deeds DL. 2004. Exploration and Exploitation Alliances in Biotechnology: A System of New Product Development. Strategic Management Journal 25(3): 201-221.
Quintana-García C, Benavides-Velasco CA. 2008. Innovative Competence, Exploration and Exploitation: The Influence of Technological Diversification. Research Policy 37(3): 492-507. Sampson RC. 2007. R&D Alliances and Firm Performance: The Impact of Technological Diversity
and Alliance Organization on Innovation. Academy of Management Journal 50(2): 364-386. Sarkar MB, Aulakh PS, Madhok A. 2009. Process Capabilities and Value Generation in Alliance
Portfolios. Organization Science 20(3): 583-600.
Sivakumar K, Roy S, Zhu J, Hanvanich S. 2011. Global Innovation Generation and Financial Performance in Business-to-Business Relationships: The Case of Cross-Border Alliances in the Pharmaceutical Industry. Journal of Academy of Marketing Science 39(5): 757-776. Tushman ML, O’Reilly CA. 1996. Ambidextrous Organizations: Managing Evolutionary and
Revolutionary Change. California Management Review 38(4): 8-30.
Vasudeva G, Anand J. 2011. Unpacking Absorptive Capacity: A Study of Knowledge Utilization from Alliance Portfolios. Academy of Management Journal, 54(3): 611-623.
Wassmer U. 2010. Alliance Portfolios: A Review and Research Agenda. Journal of Management 36(1): 141-171.
Wassmer U, Dussauge P. 2011. Value Creation in Alliance Portfolios: The Benefits and Costs of Network Resource Interdependencies. European Management Review 8(1): 47-64.
