Does size matter?
A two-dimensional comparative research on the impact of alliance
portfolio size on firm innovativeness
An empirical study of MNEs in the global automotive industry
Master Thesis
byMarcel Zirpins S3458377
m.zirpins@student.rug.nl
M.Sc. International Business & Management University of Groningen, The Netherlands
Faculty of Economics and Business
Department of Global Economics and Management
Date of submission: June 18th, 2018
Supervisor: Assistant Prof. Dr. Viacheslav Iurkov Co-Assessor: Associate Prof. Dr. Miriam Wilhelm
Does size matter?
A two-dimensional comparative research on the impact of alliance
portfolio size on firm innovativeness
An empirical study of MNEs in the global automotive industry
ABSTRACT
The field of alliance portfolio configuration has received a lot of attention within literature over the last few decades but falls short on explicitly explaining the interaction between two of its main pillars, namely number of partners and number of alliances. This study is designed to analyze the combined impact of these two parameters and measure their impact on innovation performance (number of granted patents per year) among the 14 largest automotive manufac-turers in the world. To do so, first the sole impacts of both number of unique partners and number of alliances on innovation performance has been measured. Thereafter, by creating a ratio of the two (APS-Index), their combined effect on innovation performance has been evalu-ated. Following former research on alliance portfolio size it was hypothesized that the rela-tionship between number of partners, number of alliances and APS-Index as independent var-iables on the one side and innovation performance as dependent variable on the other side would follow an inverted shaped relationship. The results indeed disclosed an inverted U-shaped relationship between number of partners/number of alliances and firm innovativeness. The relationship between the APS-Index and firm innovativeness, however, follows a standard U-shape. Building on different theories this study not only complements extant research on alliance portfolio size but also gives both theoretical suggestions for future research and prac-tical implications for managers how to set-up an efficient alliance portfolio.
Keywords: Alliances; Partners; Alliance Portfolio Size; Alliance Portfolio Configuration;
TABLE OF CONTENTS
LIST OF FIGURES AND TABLES ... I LIST OF ABBREVIATIONS ... II
1 INTRODUCTION ... 1
2 ALLIANCE PORTFOLIO CONFIGURATION AND HYPOTHESES ... 2
2.1 Firm Innovativeness ... 5
2.2 Alliance Portfolio Configuration: Number of Alliances ... 7
2.3 Alliance Portfolio Configuration: Number of Alliance Partners ... 9
2.4 APS-Index ... 11 3 METHODOLOGY ... 13 3.1 Sample ... 13 3.2 Data Collection ... 14 3.3 Measurement... 15 3.3.1 Dependent Variable ... 15 3.3.2 Independent Variables ... 16 3.3.3 Control Variables ... 18 3.4 Analytical Method... 19 4 RESULTS ... 20 4.1 Descriptive Statistics ... 20
4.2 Regression Results and Hypotheses Testing ... 23
5 DISCUSSION ... 27
5.2 Managerial Implications ... 32
5.3 Limitations and Future Research ... 33
6 CONCLUSION ... 34
7 ACKNOWLEDGMENTS ... 35
LIST OF FIGURES AND TABLES
Figure 1: Concepts regarding different alliance portfolio configurations ... 4
Figure 2: Conceptual model of APS-Index and innovativeness ... 12
Figure 3: Data collection process of APS and patents ... 14
Figure 4: Average APS-Index and number of patents between 1999 and 2008 ... 22
Figure 5: Average number of partners and of alliances between 1996 and 2005 ... 23
Figure 6: Predicted inverted U-shaped relationship between number of unique partners and number of patents ... 25
Figure 7: Predicted inverted U-shaped relationship between number of alliances and number of patents ... 26
Figure 8: Predicted U-shaped relationship between APS-Index and number of patents ... 27
Table 1: Descriptive statistics and correlations... 21
LIST OF ABBREVIATIONS APS: Alliance Portfolio Size
BMW: BMW Group (Bayerische Motorenwerke) BYD: Acronym for Build Your Dreams
Daimler: Daimler Group
GM: General Motors Corp.
Honda: Honda Motor Co. Ltd. Hyundai: Hyundai Motor Co. Ltd.
LL: Log-pseudolikelihood value
MNE: Multinational Enterprise NBREG: Negative Binomial Regression OEM: Original Equipment Manufacturer Orbis: Bureau van Dijk’s Orbis database R&D: Research and Development
SDC: Thomson Reuters Securities Data Corporation SIC: Standard Industrial Classification
SME: Small and Medium Enterprises Toyota: Toyota Motor Corp.
USPTO: United States Patents and Trademarks Office
1 INTRODUCTION
“Coming together is a beginning, staying together is a progress, working together is a success.”
This quote by Henry Ford, inventor of the modern assembly line and founder of the Ford Motor Company, reflects how important successful collaboration is when doing business – today more than ever before. With faster information transmission, extended global product reach, compa-nies tapping into new unknown markets all while offering more complex products, the automo-tive industry is facing fundamental changes in the next years. The product portfolio of original equipment manufacturers (OEM) will likely change from traditional fuel-operated vehicles to electrified, connected and self-driving products in the next decade. Simultaneously, the auto-motive industry has to deal with new competitors like Tesla, but also with companies from outside the industry like Apple and Google which are currently working to develop their own self-driving cars.
As of today, 80% of a manufactured vehicle consists of parts that are produced exter-nally and not by the company which is in charge of assembling the final product (Zellner, 2010). These sub-products are usually purchased in large quantities or first have to be invented in collaboration with other companies. Although the transaction cost theory (Williamson, 1975) would suggest an integration of the process into the company, i.e. manufacturing the parts within the focal company instead of buying it on the market, it is financially not feasible for a firm to take care of the whole production process itself by integrating the complete supply chain into the company. Electrified and self-driving vehicles, for instance, require cutting-edge inno-vations and components (e.g. batteries, sensors, board-computers) as well as resources and know-how all of which a traditional automotive manufacturer generally does not possess com-pletely. Alliances can be seen as a solution to these deficiencies. These are the reasons why alliances between suppliers and OEMs are so common and play a decisive role in the automo-tive industry today and will so in the future. They provide access to new product markets, make required resources and know-how available to the firm and share operational risks among part-ners.
automotive manufacturers in the world. Configuration of alliance portfolios has been investi-gated from a partner perspective (focusing on different dimensions of diversity among alliance partners) as well as from a size perspective (focusing on the impact of total number of alliances). In his comprehensive review about existing alliance portfolio research, Wassmer (2010) calls for a combination of these two perspectives by conducting a comparative research. Although there have been some first approaches (Kim & Choi, 2014), this study is designed to fill this gap by drawing on the resource-based view, network theory and organizational learning theory to examine to what extent different settings of number of alliances and number of partners have an impact on a firm’s innovation performance. To answer this question a ratio of number of partners and number of alliances – called Alliance Portfolio Size-Index (APS-Index) – has been created. As a result, the following research questions has been put forward:
What is the combined impact of number of partners and number of alliances in a port-folio on firm innovativeness?
The answer to this question is supposed to help to better understand why different portfolio configurations are more effective in the innovation process than others. According to Deeds & Hill (1996) the reason why firms usually don’t have the optimal level of alliances lies within human decision-making. Although managers intend to decide rationally, their decision-making is often rationally bounded and subject to cognitive limits, ambiguity, uncertainty and complex-ity. With this study I hope to shed more light on this relatively young field of research and give both managers and researchers more insight of what makes up an effective configuration of an alliance portfolio in the automotive industry in regard to innovation performance.
The study is organized as follows. First, the underlying theoretical framework of this study and the hypotheses will be explained in detail. Second, the chosen methodological ap-proach and data collection process will be described, followed by the analytical results section. The study will be finalized with a discussion and implications of the results, and further men-tions limitamen-tions and possibilities for future research.
contracts with different firms simultaneously across the globe, leading to the emergence of al-liance portfolios. While there is general agreement among researches what strategic alal-liances are – namely partnerships between at least two independent firms to jointly work on a specific project or activity by coordinating required skills and resources to bypass a make or buy-deci-sion (Dussage & Garette, 1999) – the existing literature on the conceptualization of alliance portfolios is less consistent, gives more variation and is partly confusing (Wassmer, 2010). Next to approaches based on network theory (Baum et al., 2000; Ozcan & Eisenhardt, 2009) and learning theory (Hoang & Rothaermel, 2005; Anand & Khanna, 2000), the most common way to define alliance portfolios is the additive approach. Definitions reach from simple wordings to more precise ones. Whereas Hoffmann (2007) defines alliance portfolios as all alliances of a focal firm, Reuer et al. (2002) refer to it as a firm’s accumulated international joint venture experience. For this study I follow Lavie (2007, p. 1188) to define an alliance portfolio as a “firm’s collection of direct alliances with partners.”
Existing research on alliance portfolios can be categorized into three dimensions, namely (1) the emergence, (2) the configuration, and (3) management, and builds upon a num-ber of different theories like social network theory (Goerzen & Beamish, 2005; Gulati, 1999; Stuart, 2000), organizational learning theory (Deeds & Hill 1996; Lavie & Miller, 2008), or resource-based view (Ahuja, 2000a, 2000b; Lavie, 2006; Zaheer & Bell, 2005). To better un-derstand which dimension this study can be attributed to, I will next focus on a short explanation of only the first two dimensions.
The emergence of alliance portfolios as the first stream of research aims at answering why firms build alliance portfolios and if they do, how do they tackle this challenge. The an-swers to these questions can stem from firm level perspective or individual level perspective. The first explains the motivation to build alliance portfolios through rational strategic reasons like the access to valuable resources (Das & Teng, 2000; Chung et al., 2000), reduction of transaction costs (Kogut, 1988), acquisition of knowledge from partners (Inkpen, 2000), or the improvement of the competitive positioning (Silverman & Baum, 2002; Gimeno, 2004). The individual level perspective, mainly based on agency theory, explains the emergence of alliance portfolios as a result of managers’ attempt to maximize their personal utility function (Reuer & Ragozzino, 2006).
performance (Zheng & Yang, 2015). The perhaps most eminent dimensions that research about alliance portfolio configuration has focused on are size, structure, relations between individual alliances within the portfolio, and the partner dimension (Wassmer, 2010). Among these, the size dimension has received most attention but was primarily tackled with a one-dimensional approach, i.e. adopt either number of alliances or number of partners as a count variable to measure their impact on performance (Kim & Choi, 2014). Existing research gives evidence that both measures seen separately are essential to better understand this impact, but it specifi-cally lacks to give a combined view of the two and their impact in innovativeness. A differen-tiation has to be made between portfolios consisting of a high number of alliances with a high number of partners, compared to portfolios consisting of a high number of alliances with only a few partners (Wassmer, 2010). Figure 1 illustrates these two concepts. While Firm A has a total of n = 8 alliances (sum of the numbers left to the blue lines) with six partners (number of green circles), Firm B also has a total of n = 8 alliances but with only three partners.
Figure 1: Two concepts of firms with the same number of alliances n but different numbers of partners k
Differentiating between these two concepts is important as their distinct structures bring along different costs and benefits to the firm. For instance, having less alliance partners in a portfolio can essentially decrease a firm’s transaction costs in terms of reduced search and mon-itoring costs, as well as diminished uncertainty and possibility for opportunism of partners (Wil-liamson, 1985). Likewise, different configurations can also have a critical impact on innova-tiveness of a firm. Therefore, firms should be highly interested in the optimal relation between number of partners and number of alliances their alliance portfolio consists of. In his review about alliance portfolios, Wassmer (2010) called for a combination of both size dimensions (i.e. number of alliances, number of partners) and suggested a comparative research because the existing one-dimensional approaches to measure portfolio size are of limited use. Kim & Choi
Firm A Firm B
2 1 1 1 2 1 3 3 2
(2014) followed this call and conducted research on the Korean pharmaceutical industry, meas-uring the impact of number of alliances, partners and the spanning of structural holes on per-formance (growth rate of revenue, growth rate of profit). Zheng & Yang (2015) studied the effect of alliance partner repeatedness on breakthrough innovations in the US biopharmaceuti-cal industry.
In the next sections I will explain why both measures for themselves are important in regard to the innovativeness of a firm, what constitutes innovation, and give further theoretical background. Then, I will create a two-dimensional approach by setting the number of partners and the number of alliances in an alliance portfolio into relation and hypothesize their combined impact on a firm’s innovativeness.
2.1 Firm Innovativeness
In order to keep up with the steadily increasing pace of technological development and shorter product life cycles, firms are forced to utilize and commercialize both knowledge and know-how in a more prompt and cost-efficient manner (Lin et al., 2012). Thus, in order to protect and maintain their long-term growth and survival, firms are required to deepen and broaden their innovative capabilities (Sampson, 2007). Innovation is based on the idea of ‘newness’, is contingent on interfirm collaboration (Gupta, Tesluk, & Taylor, 2007), and fundamentally con-tributes to a firm’s value creation by generating novel technologies, enabling to access new markets and to come up with new products and services (Cui & O’Connor, 2012). For this, the fusion of new and diverse resources, knowledge, and information with internally existing ones is indispensable (Iansiti & West, 1997). To assess a firm’s innovativeness, many studies relied on patents as a measure. This is a common procedure to examine innovativeness in different industries, e.g. telecommunication equipment (Sampson, 2007), chemicals (Ahuja & Katila, 2001) or biotechnology industry (Lin et al., 2012).
has been accessed and learned from partners – among other parties in the network of the firm. The process of sharing may take place through transfer of equipment and technology, or through more informal interactions of employees, e.g. through day-to-day work.
Innovation can further be differentiated by types and methods. Oerlemans et al. (2013) distinguish between two types of innovation, namely incremental and radical, which both “re-quire different type, depth and variety of knowledge.” In the case of incremental innovation, a product, product line, a dominant design or technological standard already exists which is why innovation is targeted at further upgrading and improvement. Consequently, firms engaged in incremental innovations tend to cooperate with a more internationalized and more diverse port-folio of external partners, as a study in the telecommunication industry showed (Feller et al., 2007). Radical innovations, on the other hand, relate to the generation or application of com-pletely new technologies. The necessary knowledge for their development is often possessed by specific and specialized external parties. This causes a firm to partner up with only a few, specialized companies to access their scarce capabilities and expertise which in turn leads to a smaller, less diverse portfolio of alliance partners (Laursen & Salter, 2006). A second, more inherent reason leading to smaller and less diverse alliance portfolios is the absorptive capacity of a firm. Absorptive capacity is the ability to recognize, assimilate and apply novel and external knowledge and to use it as means of commercialization (Cohen & Levinthal, 1990). The chances of achieving high innovation performance and subsequently competitive advantages increase with greater absorptive capacity (Lane & Koka, 2006; Camison & Fores, 2010).
knowledge) and codified knowledge (i.e. explicit knowledge) in a repeatable manner. Thus, open innovation advises firms to turn away from solving problems internally in favor of finding answers externally (Billington & Davidson, 2013).
2.2 Alliance Portfolio Configuration: Number of Alliances
One of the most prominent discussions within the literature about alliance portfolio size con-cerns the number of alliances a firm is engaged in. For over two decades researchers have fo-cused on the effects of the number of alliances on various outcome measures like financial performance or innovativeness. Despite the broad range of studies concerning these issues, there have been different results but no general consent about the final effects.
Deeds & Hill (1996) were among the first ones to propose a positive, but non-linear relationship between number of alliances and rate of new product development. Following the resource-based view of Barney (1991) they argue that extending the alliance portfolio, that is increasing the number of alliances, will give the firm access to more external resources which can complement the already existing internal resources owned by the firm. These resources may not be easily imitated and therefore lead to competitive advantages over other competitors (Ei-senhardt and Schoonhoven, 1996). External resources, or network resources (Lavie, 2007), comprise the partner’s tangible and intangible assets, e.g. human resources, financial assets or reputation, that are vital to generate, store and commercialize knowledge in order to develop new products (Mowery, Oxley, and Silverman, 1998; Rothaermel, Hitt, and Jobe, 2006). More-over, a firm can create additional value by combining network resources of different partners and generate synergies that individual partners in the portfolio are unable to access and exploit (Lavie, 2007).
alliances in a firm’s portfolio will grant access to a greater base of diverse external knowledge, previously unexplored by the firm, which can be utilized to solve problems. Hence, adding more alliances to a portfolio is likely to improve innovation performance of a firm (Lahiri & Narayanan, 2013). These positive relationships between number of alliances and both firm per-formance and innovativeness may be one reason why there has been a significant increase of firms’ engagement in alliances in the 1990s (Kale and Singh, 2009).
Nonetheless, increasing the number of alliances in a portfolio does not exclusively in-hibit benefits but also comes with costs. In their study about new product development of 132 biotechnology firms in the U.S., Deeds & Hill (1996) specify three main reasons for diminish-ing and even negative returns when excessively increasdiminish-ing the number of alliances. First, simply following the economic “law” of diminishing returns, they state that the marginal contribution of each additional alliance a firm engages in is relatively minor compared to the contribution of the previous alliance. Second, accessing external resources and knowledge may encompass risks for the firm. Complementary resources that are to be combined with internal resources may be of poor match, they may contradict the promise made by the partner, or the partner may even act opportunistically by exploiting the firm’s know-how while only giving little in return. Additionally, a firm’s ability to extract and process important and valuable know-how is limited and also underlies diminishing returns (see “absorptive capacity”, Dyer & Singh, 1998). The reason for these problems is the negative correlation between number of alliances and manage-rial effectiveness of a firm. The more a firm extends its alliance portfolio, the timelier and more complicated the information processing measures will be, while simultaneously the quality of partner search and their proper monitoring will decline due to limited management resources dedicated to these operations. Even though diverse external resources are favorable in leverag-ing internal resources (Levinthal and March, 1993), the challenge of effectively exploitleverag-ing these resources intensifies with their increasing diversity. Similarly, internal routines are important to create products and processes (Nelson and Winter, 1982), but have to be adapted simultane-ously to multiple partners in a portfolio which will lead to an increase in transaction costs the more diverse the resources are.
positive effect at all. Although this section gave only a glimpse of the existing literature on alliance portfolio size, it demonstrates that to date there is no common sense of the influence of number of alliances in a portfolio and their effect on outcome measures, and that there are still many gaps to be filled. This problem of generalizability in my opinion occurs because it is impossible to find a sole relationship for all types of businesses, ranging from small local com-panies to large global players, and from traditional corporations to entrepreneurial start-ups. For the purpose of this paper I will follow Deeds & Hill (1996) and propose the following:
Hypothesis 1a: There is a non-linear, inverted U-shaped relationship between number of
alli-ances and firm innovativeness.
2.3 Alliance Portfolio Configuration: Number of Alliance Partners
The second, yet less analyzed part of alliance portfolio size is the dimension of alliance partners (Wassmer, 2010). The bulk of the studies on this dimension are based on the resource-based view (Leeuw, Lokshin, Duysters, 2014; Jiang, Tao, Santoro, 2010), network theory (Kim & Choi, 2014), organizational learning theory (Zheng & Yang, 2015), or a combination of these (Pangarkar & Wu, 2013; Goerzen, 2007), and focus on the diversity or characteristics of part-ners. Research on the number of alliance portfolio partners are comparably scarce (Kim & Choi, 2014). Most of the theories used when analyzing alliance partners encourage an increase in alliance partners, at least to a certain point.
the flow of new, innovative ideas into the group (Burt, 1992). The ‘familiarity trap’ (Zheng & Yang, 2015) emerges when structured routines become too rigid, leading partners to filter out novel ideas, and center their attention on exploitation rather than exploration (Koza & Lewin, 1998) which hinders innovativeness. This in particular can be observed among older, more tra-ditional firms – for instance in the automotive industry – that have developed rigid structures and routines they worked with for decades. Over-embeddedness with partners therefore may lead to decreasing firm performances (Uzzi, 1997). Equally, repeated equity-based alliances clearly have a negative impact on firm performance, especially in environments characterized by high technical uncertainty (Goerzen, 2007). Management teams that put too much emphasis on the reduction of transactions costs simply to maximize management efficiency run into dan-ger of locking out newcomers that can provide required cutting-edge technologies to the firm (Gulati, 1999).
at the wrong time or place (Koput, 1997) that are potentially not taken seriously (attention al-location problem). Therefore, according to network theory and resource-based view, too many and too diverse partners negatively affect firm performance, leading to an inverted U-shaped relationship between number of alliance partners and both firm performance and innovativeness (De Leeuw, Lokshin, & Duysters, 2014; Kim & Choi, 2014; Zheng & Yang, 2015). For the purpose of this paper I will follow Kim & Choi (2014) and propose the following:
Hypothesis 1b: There is a non-linear, inverted U-shaped relationship between number of
unique partners and firm innovativeness.
2.4 APS-Index
Following Wassmer’s (2010) call to conduct a comparative two-dimensional approach on alli-ance portfolio size (including both number of allialli-ances and number of partners), I created a ratio of these two variables (called Alliance Portfolio Size-Index or APS-Index) to measure their combined effect on innovativeness. Referring to Figure 1, in which Firms A and B hypotheti-cally both had eight alliances but with six and three partners, respectively, it is much likely that different configurations of alliance portfolios in terms of the ratio between number of unique partners and number of total alliances have diverging effects on innovation performance of a firm. The ratio r which I created – which will be further explained in the next section – can be obtained by dividing the number of unique partners in a certain year (in the following number of partners) by the total number of alliances in a firm’s alliance portfolio in the same year (in the following number of alliances). Taking Figure 1 as an example, Firm A therefore has a value of r = 0.75 and Firm B a value of r = 0.375. The creation of the ratio r facilitates measuring their combined effect on innovativeness of a firm. Hypothetically, r can take values between 0 and ¥. In reality, however, I expect values for r not to fall short of 0.3 or to exceed 3.0 for large firms because having three-times as many alliances as partners (or vice versa) cannot be ob-served and is not advisable. Three values or (hypothetical) value ranges can be differentiated.
For values of r > 1, a firm’s alliance portfolio consists of multi-partner alliances, leading to a higher number of partners than overall alliances. With increasing values of r the inflows of new information and diverse knowledge to a firm rise exponentially and may quickly turn into an overflow. The inflow is strongly limited by the firm’s absorptive capacity. Therefore, if r surpasses the value of 1, the rising number of partners can quickly lead to the exceedance of absorptive capacity because the firm is not able to process the overwhelming amount of new ideas, possibly paying too little attention to important information that would lead to cutting-edge innovations. Thus, for values of r > 1, innovation performance should decrease.
For moderate levels of the APS-Index (e.g. r » 1), the number of partners and the num-ber of alliances is (almost) equal. It means that a firm enters into alliances with multiple partners or repeated partnerships only rarely which on the other side limits its options to split risks and create synergies within alliances consisting of multiple partners. Radical innovations often re-quire extensive capital input, and bear high risks and uncertainty for the innovators, which is why large-scale R&D projects are usually executed by several partners. However, moderate levels of the APS-Index should maximize a firm’s innovativeness because excessive focus on both repeated partnerships or multi-partner alliances lead to diminishing innovation perfor-mances as described earlier. Accordingly, I propose the following:
Hypothesis 2: There is a non-linear, inverted U-shaped relationship between APS-Index and
firm innovativeness with innovativeness reaching its maximum at moderate values of the APS-Index.
Figure 2: Conceptual model of the relationship between APS-Index and Innovativeness
Alliance Portfolio Configuration Number of unique Partners
Alliance Portfolio Configuration Number of total Alliances
3 METHODOLOGY
In the following sections I will describe more in depth (1) the sample and industry that has been used for this study, (2) what specifically has been measured and how the data has been collected, (3) show and explain descriptive statistics as well as (4) run the analysis. The methodological choices that have been made are justified within this section. A longitudinal approach has been chosen to provide insights into the impact of different values of APS-Index on firm’s innova-tiveness over a period of time and to be able to eventually test the three hypotheses.
3.1 Sample
The empirical setting of this study is the automotive industry. This industry is characterized by high-density alliance networks, especially among the largest original equipment manufacturers (OEMs). Long-term supply or R&D alliances are extremely useful in this industry due to the high complexity of the final product, i.e. the automobile. Today, about 80% of an automobile consist of parts that have been purchased externally from suppliers. The remaining 20% are produced or developed internally (Zellner, 2010). OEMs in an automotive supply chain are companies that eventually assemble all intermediary- and sub-products into a final product, in this case a car, truck or bus. On average, an automobile is comprised of 10,000 different parts, which makes it really vulnerable for production faults. To minimize these faults, OEMs have to cooperate closely with their various suppliers. Besides that, the recent push both from gov-ernments and societies for more ecofriendly automobiles, and the decreasing reserves of fossil fuels put high pressure on OEMs and their suppliers to develop new, innovative concepts for the mobility of the future including electric and self-driving cars. That’s why many manufac-turers enter into long-term alliances with partners from different levels of the value chain. In 2010, the German Daimler Group with over 1,500 suppliers, for instance, built a joint venture with the largest battery and accumulator producer in the world, BYD Co Ltd. from China, to develop a new electric vehicle. Although the need for and overall number of alliances in the automobile industry decreased over the last three decades – influenced by the emergence of the internet and open innovation (Billington & Davidson, 2013) – OEMs and suppliers are still largely dependent on alliances. This makes the automotive industry an interesting field of study for alliance portfolio research.
study. Of those 14 groups two are based in the U.S., and six are from Europe and Asia, respec-tively. The relevant data have been collected over a time period of ten years and were retrieved from different databases which will be further explained in the next section.
3.2 Data Collection
For the purpose of this study I collected data for the years 1992-2008, sourcing information from various databases. First, the Thomson Reuters Securities Data Corporation (SDC) data-base on Strategic Alliances was used to collect general data of alliances in the automobile in-dustry (SIC codes 3711, 3713, 3714, 3715, 3751) for the given time period. The alliances have then been traced back to the ultimate parent name of the company which then allowed to count the number of alliances of firms in my sample for each single year. The problem of mergers and acquisitions that took place between 1992 and 2008, therefore creating different ultimate parent names, was accounted for by manually assigning companies and their alliances to the present ultimate parent name. For instance, the present Daimler AG was called Daimler-Benz AG until 1998 and then DaimlerChrysler AG between 1999 and 2007 before receiving its cur-rent name. Thus, all number of alliances of the three ultimate pacur-rent names have been accumu-lated and assigned to the Daimler AG. This process has been carried out in all cases of mergers and acquisitions. However, in most cases, the SDC database is incomplete concerning ending dates of alliances. Nevertheless, ending dates are crucial to know as they indicate if an alliance has to be included in a portfolio or not. As a consequence, I assume an alliance to last five years according to previous studies from Stuart (2000) and Sampson (2007). Thus, for instance, the number of alliances or number of partners in the portfolio of a firm (as part of APS) in 1996 are both obtained by accumulating all alliances of the focal firm for the years 1992-1996 (cf. Figure 3).
Figure 3: Concept of collecting data of alliance portfolio size (APS) over five years and assigning to patents with a three-year time-lag.
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
APS
1996 1997APS 1998APS 1999APS 2000APS 2001APS 2002APS 2003APS 2004APS 2005APS
Patents
Second, the received list from the SDC database containing all alliances was then im-ported into the UCINET software (version 6.652). This software is used to count the number of unique partners an ultimate parent firm had in their alliance portfolio in a given year.
Third, the accumulated number of patents of a firm for a specific year has been collected from the Bureau van Dijk’s Orbis database. The patent data in Orbis is based on the patents registered at the United States Patents and Trademarks Office (USPTO). As with the SDC da-tabase, at first all names the sample companies possessed between 1992 and 2008 had to be identified. This has been conducted manually by searching annual reports and corporate web-sites. After assigning all names to the present company name these firms’ BvD ID numbers were determined in the Orbis database. For each BvD ID number only the accumulated number of granted patents have been searched for in the Orbis database for specific years in the needed timeframe. Due to a time-lag of three years between the year an alliance becomes effective and the year patents have been granted, the respective period of time is between 1999 and 2008. For instance, the granted patents that the Toyota Motor Corp. issued in 1999 have to be assigned with a three-year time-lag to the alliance portfolio size of 1996 (cf. Figure 3). The COM-PUSTAT database has been used to obtain all required information for the control variables that have been collected for this study (firm age, total assets, R&D intensity, debt-equity ratio). Finally, all obtained data for each company in the sample have been merged into a single Excel-file and eventually imported into the STATA15 software for further analysis which will be described in the following.
3.3 Measurement
The following subsections will explain the dependent variable, independent variables and con-trol variables, and their respective measurements that were used in this study.
3.3.1 Dependent Variable
(Deeds & Hill, 1996), to R&D intensity (De Leeuw et al., 2014), to number of patents (Lin et al., 2013; Lahiri & Narayanan, 2013), and evaluating a firm’s innovativeness through industry-surveys (Oerlemans et al., 2013; Cui & O’Connor, 2012). Most research regarding alliance portfolio and innovativeness that have been reviewed for this study measure innovativeness as the number of patents. According to Acs & Audretsch (1989) patents are a valid measure of innovation activity and can be directly related to innovation performance (Pakes & Griliches, 1984). Focusing only on patents when conducting research on innovativeness among firms in the automotive industry is sufficient because alternative measures such as rate of new product development or R&D expenses have great overlap with number of patents (Hagedoorn & Cloodt, 2003). Rather than taking into account all patents of a company, for this study only the granted patents have been used because they can be directly linked to innovativeness (Ahuja & Katila, 2001). High-tech industries like software or computer-industries are characterized by a high number of patent outputs of which only a small share is eventually granted. Therefore, considering granted patents allows a more precise effect measurement on a firm’s innovative-ness. The number of granted patents is a count variable which only can take integer values in this study.
3.3.2 Independent Variables
The number of alliances refers to the total number of alliances n that a firm i has in its alliance portfolio in a specific year t of the sample period. This measurement is needed to test hypothesis 1a and hypothesis 2 in the following sections. The number of total alliances in a firm’s portfolio is aggregated in the following way:
!",$ = ' !$() *
)+, where n is the total number of alliances,
i is a focal company in the sample, and t Î {1996, …, 2005}.
alliance partners using the UCINET software. The number of unique alliance partners is essen-tial to the following analysis since repeated partnerships with the same partner firm will there-fore not bias the outcome. This measure is needed for testing hypothesis 1b later in the analysis. The number of unique alliance partners k of a firm i’s alliance portfolio in a specific year t of the sample period has been obtained as follows:
-",$ = ' -$() *
)+, where k is the number of unique partners,
i is a focal company in the sample, and t Î {1996, …, 2005}.
The APS-Index is a measure that has been newly created as part of this study about alliance portfolio configurations and includes both dimensions of alliance portfolio size, i.e. number of alliances and number of partners, to test hypothesis 2. Following Wassmer’s (2010) proposal to conduct a two-dimensional comparative measure, creating a ratio r of these two variables seemed appropriate when studying their combined effect on innovation performance. This ratio r for a firm i’s alliance portfolio in the year t of the sample has been measured as follows:
.",$ = !/012. 34 /!56/2 78.9!2.: 5! 9ℎ2 8<<58!=2 73.943<53",$ 9398< !/012. 34 8<<58!=2: 5! 9ℎ2 8<<58!=2 73.943<53",$ =
-",$
!",$ , where r represents the APS-Index,
i is a focal company in the sample and t Î {1996, …, 2005}.
In theory, r can take all values between 0 and ¥. As mentioned earlier and for the pur-pose of this study, three values or value ranges can be differentiated that are important to iden-tify for the following analysis:
0 < r < 1 for repeated partnerships,
3.3.3 Control Variables
Firm Size will be controlled for several reasons. The size of a firm influences the propensity of engaging in collaboration with other firms (De Leeuw, 2014). Larger firms are apt to have more abundant resources and thus may be more likely to better manage multiple technology collab-orations and various innovation objectives simultaneously (Cohen & Klepper, 1996; Belderbos et al., 2006). Additionally, firm size may lead to scale effects and more market power or posi-tional advantages (Leiblein, Reuer, & Dalsace, 2002), it can facilitate the access to lower cost of capital and at the same time lower risk (Chang & Thomas, 1989), thus boosting firm perfor-mance and effecting innovation perforperfor-mance. Firm size was measured as a firm’s total assets (Lahiri & Narayanan, 2013). The values of all total assets of the 14 companies have been stand-ardized to US-$ using historical exchange rates.
Firm Age as the second control variable was used due to several reasons. It is associated with survival and determination rates (Levinthal, 1991). Older firms are likely to perform better due to developed social networks, brand value, established reputation and recognition (Zaheer & Bell, 2005). Moreover, their greater knowledge stock can help across different innovations (Git-telman & Kogut, 2003). However, there are two sides to firm age: younger firms have less rigidity and thus have learning advantages of newness but at the same time suffer from liability of newness (Pangarkar, 2013). Firm age was measured as the number of years between the year of foundation and the respective year in the sample.
R&D intensity as the third control variable needs to be considered because innovation perfor-mance is believed to significantly influenced by R&D intensity. R&D engagement enhances a firm’s absorptive capacity leading to better processing of external knowledge from partners (Cohen & Levinthal, 1990). R&D intensity was measured by dividing a firm’s R&D expenses by net sales.
Capital Structure of a firm is argued to affect performance (Jensen, 1989), thus also having an impact on innovativeness. Consequently, capital structure needs to be controlled for and is measured through debt-to-equity ratio, i.e. debt divided by equity.
only be used as a control variable for testing hypothesis 2. Alliance portfolio size is measured as number of a firm’s alliances in a specific year (Baum et al, 2000) and therefore equals the independent variable number of alliances.
3.4 Analytical Method
As mentioned before, this study uses a longitudinal approach to better understand the impact of alliance portfolio configuration on firm innovativeness, represented by the three hypotheses. The dependent variable, innovativeness (number of patents), is a count variable. It can only take non-negative integer values. Following Hausman et al. (1984), it is suggested to use the Poisson regression model when the dependent variable is a count variable. However, when the count data in the model is over-dispersed the Poisson model provides inconsistent results (Hausman et al., 1984). Over-dispersion is present in the model if the value of the standard deviation of the dependent variable exceeds the value of its mean value. Table 1 shows that the mean inno-vation performance (number of patents) is 929.044, whereas the standard deviation is higher at 1200.477. If over-dispersion is present in the model, Park et al. (2015) recommend using a negative binomial regression model instead. In order to test for the hypothesized curvilinear effect between the three independent variables and the dependent variable, a quadratic function has been used in the following way:
> = 8 + 1@∗ B + 1C∗ BC + 1
D∗ E@. . . +1GHC∗ EG+ 2
where y is the number of patents, a is the constant,
X Î {Alliances; Partners; APS-Index}, Z represent the control variables.
To control for potential endogeneity and unobserved heterogeneity between firms a fixed-effect model has been chosen. Fixed-effect models are used to justify within-firm varia-tion in innovativeness over time, whereas inter-firm variavaria-tion is explained by random-effect models (Lavie, 2007). In order to clarify which of the two models fits this study appropriately, both a Hausman test as well as Akaike’s information criterion have been conducted. The results of the Hausman test (Model 2: c2 = 33.39; p = 0.0000) and the Akaike’s information criterion
binomial regression (NBREG) does not automatically have to be favored over a Poisson regres-sion if the standard deviation exceeds the mean value. All three models have also been tested with Poisson regressions to control for robustness. Still, the NBREG showed better and more significant results than the Poisson regression for the independent variables. Due to the long panel dataset an unconditional effects model has been favored over a conditional fixed-effects model. Thus, an unconditional fixed-effect NBREG has been used for the analysis to test the three hypotheses. In Table 2, the log-pseudolikelihood (LL) and pseudoRC ratios of model 2, model 3 and model 4 have been used to assess improvements between these hypothe-ses-models and the adjusted baseline model (model 1), which only included control variables and unconditional fixed-effects (Lavie, 2007). The results of this analysis will be explained more in-depth in the next section.
4 RESULTS
In this section the results of the analysis are presented. First, the descriptive statistics are clari-fied more in detail. Thereafter, the results of the negative binomial regression for all three hy-potheses will be described.
4.1 Descriptive Statistics
Table 1: Descriptive Statistics and Correlations Variable Mean S.D. 1 2 3 4 5 6 7 8 1 Patents 929.044 1200.477 1.0000 2 Number of Partners 40.70714 42.73349 -0.1137 1.0000 (0.1811) 3 Number of Alliances 35.72143 40.59983 -01525 0.9905 1.0000 (0.0721) (0.0000) 4 APS-Index 1.136737 0.2321649 0.3169 -0.0420 -0.1144 1.0000 (0.0001) (0.6225) (0.1785) 5 Firm Age (ln) 4.374832 0.2468863 -0.6036 0.2289 0.2874 -0.3739 1.0000 (0.0000) (0.0065) (0.0006) (0.0000) 6 Firm Size 110,072.8 101,855 -0.1126 0.5713 0.5864 -0.0802 0.2619 1.0000 (0.1853) (0.0000) (0.0000) (0.3459) (0.0018) 7 R&D Intensity 0.0368359 0.0118447 0.0910 0.2205 0.2284 -0.3250 0.0020 0.2424 1.0000 (0.3126) (0.0135) (0.0104) (0.0002) (0.9826) (0.0065) 8 Capital Structure 1.441059 1.483624 -0.2416 0.2993 0.3451 -0.2489 0.3038 0.5850 -0.0279 1.0000 (0.0040) (0.0003) (0.0000) (0.0030) (0.0003) (0.0000) (0.7572)
There is high variation in number of patents, reaching from 1 granted patent for Fiat Chrysler in 2005 up to 8.593 granted patents for Hyundai Motor in 1999. The average number of patents also strongly differs between continents. Asian firms had an average of 1,637 patents granted whereas the numbers were 391 and 313 for Europe and North America, respectively. The aver-age number of patents in the sample was lowest in 2002 equaling 757 and peaked with 1,229 in 2006 (see Figure 4). The APS-Index has a mean value of 1.1367 and a standard deviation of 0.2321. The lowest APS-Index was reached by Honda in 2008 with 0.50, whereas Hyundai reached an APS-Index of 2.25 in 2000. The average values for Asia are an index of 1.1599, and 1.1045 and 1.0698 for Europe and North America, respectively. The lowest industry average of the APS-Index was reached in 2008 with 1.0495, whereas 2001 marked the highest with 1.1884 (see Figure 4).
Figure 4: Average number of patents and average APS-Index for each of the 14 sample companies per year, adjusted with a time-lag of three years, between 1999 and 2008.
Table 1 also indicates that a firm’s alliance portfolio – on average – is comprised of a higher number of partners (! = 40.7071) than of number of alliances (" = 35.7214). The average number of unique partners per year are significantly higher for North American firms (k = 83) compared to European (k = 28) and Asian firms (k = 23). The lowest number of unique partners was reached by Honda in 2008 (k = 1), whereas General Motors had 247 unique partners in 1996. Similarly, the lowest number of alliances was reached by BMW in 2005 (n = 2), whereas General Motors had a total of 228 alliances in 1996. Both the number of unique partners and
1,00 1,05 1,10 1,15 1,20 0 200 400 600 800 1000 1200 1400 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
on average, a firm’s alliance portfolio was comprised of 63 alliance partners and 56 alliances, whereas in 2005 the numbers dropped to 16 and 16, respectively (see Figure 5).
Worth mentioning are the comparably high average values for firm size (total assets = 283.098) and capital structure (debt-equity ratio = 2.7115) for North American firms. On aver-age, European firms had total assets of 87.230 and a debt-equity ratio of 0.9871 during the sample-period, while Asian firms had 79.232 and 0.9378, respectively.
Figure 5: Average number of partners and average number of alliances of firm alliance portfolios between 1996 and 2005.
4.2 Regression Results and Hypotheses Testing
The results of the fixed-effects negative binomial regression which has been used to test all three hypotheses are provided in table 2. Four different models have been used for the analysis.
Model 1 – the adjusted baseline model – includes only the five control variables and measures their impact on innovation performance, i.e. number of patents. Table 2 indicates that firm size (# = 6.66e-06) influences innovation performance and is significant at the 1%. As mentioned earlier, the log-pseudolikelihood value (LL) has been used for comparison. The first model has a log-pseudolikelihood value of LL = -813.63304. 0 10 20 30 40 50 60 70 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Table 2: Negative Binomial Regression Results with Unconditional Fixed-Effects
Variable Model 1 Model 2 Model 3 Model 4
Hypothesis 1a Hypothesis 1b Hypothesis 2
Number of unique Partners 0.0153573***
(0.006041)
Number of unique Partners (squared) -0.0000343**
(0.0000174)
Number of Alliances1 0.0071309 0.0156767*** 0.0076181 (0.0053099) (0.0051167) (0.0052406) Number of Alliances (squared) -0.000036**
(0.0000187) APS-Index -1.841159** (0.8518244) APS-Index (squared) 0.7048521*** (0.227761) Firm Age -0.0011879 0.0218705 0.0175497 -0.0008417 (0.0702733) (0.0665502) (0.064949) (0.0705925) Firm Size 6.66e-06*** 6.12e-06*** 5.87e-06*** 6.95e-06*** (1.63e-06) (1.43e-06) (1.41e-06) (1.63e-06) R&D Intensity 12.64011 17.34679 15.47129 12.13.02065 (25.81886) (25.01511) (26.05108) (25.8592) Capital Structure -0.0261152 -0.0303778 -0.0335314 -0.027103 (0.0492915) (0.0482829) (0.0506283) (0.0511925) Constant 4.945139 1.828455 2.476471 5.963878 (7.092833) (6.693872) (6.442759) (7.290721) Number of observations 125 125 125 125 Number of firms 14 14 14 14 Log-pseudolikelihood -813.63304 -812.96124 -812.73177 -812.29502 Pseudo R2 0.1649 0.1656 0.1658 0.1663 Significance: * p<0.1; ** p<0.05; *** p<0.01; Standard errors in parentheses
Model 2 includes the control variables and adds the independent variables number of unique partners and number of unique partners (squared). This model is used to test hypothesis 1a. Hypothesis 1a predicts an inverted U-shaped relationship between the number of unique partners and a firm’s innovation performance. For this model the number of alliances as control variable has been removed because of strong correlation with number of unique partners (p = 0.9905). Adding the independent variable and its squared term to the model improves the re-gression compared to the first model (LL = -812.96124). Firm size remains significant at the 1%-level. Number of unique partners is significant at the 1%-level (# = 0.0153), while Number of unique partners (squared) is significant at the 5% level (# = -0.0000343). Their positive and negative coefficients predict an inverted U-shape relationship between number of unique part-ners and innovation performance (see Figure 6). Therefore, hypothesis 1a is supported.
Figure 6: Predicted inverted U-shaped relationship between number of unique partners and number of patents.
Model 3 is used to test hypothesis 1b which predicts an inverted U-shaped relationship between number of alliances and innovation performance. For this model number of partners and number of partners (squared) have been exchanged by number of alliances and number of alliances (squared). It includes all control variables except number of alliances. The control variable firm size (# = 5.87e-06) remains significant at the 1%. The log-pseudolikelihood value
0 1000 2000 3000 4000 Pre d ict e d N u mb e r o f Pa te n ts 0 100 200 300
(LL = -812.73177) has further increased compared to the previous two models. The two inde-pendent variables for this model, namely number of alliances and number of alliances (squared) are significant at the 1% and 5%-level, respectively. The positive coefficient for number of alliances (# = 0.0156) and the negative coefficient for number of alliances (squared) (# = -0.000036) predict an inverted U-shaped relationship with innovation performance (see Figure 7). Thus, hypothesis 1b is supported.
Figure 7: Predicted inverted U-shaped relationship between number of alliances and number of patents.
Model 4 is used to test hypothesis 2 which predicts an inverted U-shaped relationship between APS-Index and innovation performance. For this model number of alliances and num-ber of alliances (squared) have been exchanged by APS-Index and APS-Index (squared). The model also contains all control variables, including number of alliances which adds to the sig-nificance of the independent variables. Firm size (# = 6.95e-06) is still strongly significant at 1%. This model exhibits the highest log-likelihood value (LL = -812.29502) of all tested mod-els. The two independent variables APS-Index (# = -1.8411) and APS-Index (squared) (# = 0.0748) are strongly significant at the 5%-level and 1%-level, respectively. The negative coef-ficient of APS-Index and the positive coefcoef-ficient of APS-Index (squared) do not justify the pre-dicted inverted U-shape relationship with innovation performance. This model rather predicts a U-shaped relationship (see Figure 8). Therefore, hypothesis 2 is rejected.
The results show that firm size positively influences the number of granted patents in all models at a 1%-significance level. An increase in firm size (i.e. total assets) therefore entails an increase in innovation performance (i.e. number of granted patents). Furthermore, the outcomes disclose that both number of unique partners and number of alliances seen separately positively influence innovation performance at significance levels of 1%, up to a certain point after which the influence turns negative (inverted U-shape). Against expectation, their combined effect – namely the APS-Index – first negatively influences innovation performance to a certain point after which the impact turns positive (U-shape). The results and possible explanations will be discussed more in-depth in the next section.
Figure 8: Predicted U-shaped relationship between APS-Index and number of patents.
5 DISCUSSION
The following section will discuss the results of the analysis more in-depth and focus on expla-nations of the results. It will also attempt to give both theoretical and managerial implications that might be useful for both researchers as well as managers of alliance portfolios. Limitations of this study and suggestions for future research will be given at the end of this section.
To date, existing research on alliance portfolio configuration mainly involved finding the optimal set-up of alliance portfolios that trigger the best performance for the focal firm. However, separately conducting research on either number of partners or number of alliances was not crowned with success so far as there is still no common understanding on their influ-ences. This study was designed to extend and complement existing research on alliance portfo-lio configuration by combining both pillars of portfoportfo-lio size (namely number of partners and number of alliances) and assess their combined effect on firm innovativeness. The findings of this study both support existing research as well as pave the way for a potential new stream of research that requires further and deeper investigation.
Today, it is commonly acknowledged that alliances contribute to firms’ long-term suc-cess in many industries and therefore require sufficient managerial know-how and attention in order to make large portfolios most effective and efficient. In the automotive industry the pop-ularity of entering alliances peaked in the late 1980’s and beginning of the 1990’s (Kale and Singh, 2009) after which the average number of both alliances and partners per firm steadily decreased among the largest OEMs (see Figure 5). This decrease may be explained by the emergence of open innovation which facilitated communication and increased speed and amount of information transmission. The development of the internet and new innovative tech-nologies enabled firms to access required knowledge and (human) resources not solely through contractual long-term agreements but to source critical information and know-how online, thereby complementing internal resources (Billington & Davidson, 2013).
up with the ever-increasing pace of technological change and innovation. These new techno-logical advances require resources and know-how that large traditional manufacturers usually do not possess. To access or acquire necessary resources to develop, for instance, connected and self-driving vehicles OEMs are repeatedly entering alliances with firms outside the auto-motive industry, sometimes partnering up with unconventional firms like Google or small start-ups. These facts demonstrate that – although the overall number of alliances decreased over the last three decades – the demand for innovative and unconventional alliances in the automotive industry is growing from which OEMs hope to achieve first-mover advantages in the fields of electric and autonomous driving.
Building on the resource-based view this study first adds to extant literature about alli-ance portfolios in a way that the results indicate an initial positive relationship between number of alliances and firm innovativeness for small portfolios. Increasing the number of alliances in a portfolio will give the firm access to more external resources which can complement the al-ready existing internal resources owned by the firm. The access to tangible and intangible assets owned by the partner enhances a firm’s innovation performance (Hoffmann, 2007) and enables a firm to generate, store and commercialize knowledge in order to develop new products (Rothaermel, Hitt, and Jobe, 2006). However, excessively extending the alliance portfolio just for the good of the company automatically incurs costs as well, and these costs will eventually exceed the benefits of additional alliances after a certain point. The costs for managing large alliance portfolios increase exponentially while the benefits of each new alliance are subject to the economic “law” of diminishing returns, implying minor advantages of each additional part-nership the firm engages in. The results of this study therefore are in line with extant literature (e.g. Deeds & Hill, 1996) and imply an inverted U-shaped relationship between number of al-liances and firm innovativeness among OEMs in the automotive industry (see Figure 7).
their partner search to the automotive industry but rather expand alliances across different in-dustries and other levels of the value chain. However, as with number of alliances, increasing the number of partners does not come without costs. The benefits each additional partner brings to the alliance is a function of the capabilities and resources dedicated to the management of a focal firm’s alliance portfolio. A portfolio consisting of many highly diverse partners requires substantial management resources for monitoring and controlling and consequently may lead to an information overflow due to limited absorptive capacity of the focal firm. For that reason, the results of this study, in line with previous research (e.g. De Leeuw, Lokshin, & Duysters, 2014; Zheng & Yang, 2015), demonstrate that too many and too diverse partners will have negative effects on firm innovativeness and eventually lead to an inverted U-shaped relation-ship between number of partners and firm innovativeness among OEMs of the automotive in-dustry (see Figure 6).
and know-how through strong organizational routines and sustainable relationships. It is a good example of how repeated partnerships can increase innovativeness, as Toyota was recently ranked the world’s most admired automotive company for the fourth consecutive year by For-tune Magazine (Toyota, 2018).
High values of the APS-Index on the other hand indicate that the focal firm for the most part enters into alliances consisting of multiple partners (multi-partner alliances). Although these kinds of alliances can produce tremendous challenges in terms of effective information processing and excessive management costs, the firms of the analyzed sample are likely to pos-sess measures to tackle these challenges. Each of the 14 investigated companies is a global player and therefore is likely to have an organizational unit that acts as a focal point and that is dedicated to implement routines that help to codify, learn and utilize alliance know-how. These dedicated alliance functions facilitate the codification of existing and new alliance-management knowledge into guidelines and manuals which can help managers to select partners, negotiate and formulate contracts, and terminate alliances (Kale, Dyer & Singh, 2002). They also support and assist in managing large alliance portfolios by filtering incoming knowledge and providing channels for its distribution within the firm (Grand & Baden-Fuller, 2004). The observed large OEMs of this sample are also likely to possess outstanding absorptive capacity. Absorptive capacity is a function of the size of firm and its disposable resources. Thus, larger firms can dedicate more financial and human resources to the management and processing of new ac-quired knowledge compared to average sized companies. The alliance data retrieved from the SDC database also reveal that there is a high share of multi-partner alliances in the sample period. In 1992 for instance BMW and VW formed a joint venture with four other firms from different industries to develop an advanced information service for road traffic in Germany. Similarly, GM, Daimler and BMW formed a joint venture in 2006 to develop and manufacture hybrid engines. Thus, against expectations, the results of this study indicate a U-shaped rela-tionship between the APS-Index and firm innovativeness which implies that a firm can be most innovative at either low or high values of the APS-Index and should avoid a balance (moderate values) between number of partners and number of alliances in its portfolio (see Figure 8).
5.1 Theoretical Implications
has only scarcely put attention to examining the interplay of both pillars of portfolio size (e.g. Kim & Choi, 2014) to fill the theoretical gap mentioned by Wassmer (2010). The results of this study demonstrate that number of partners and number of alliances, as well as their combined effect represented by the APS-Index have a significant effect on firm’s innovativeness. Further-more, the R&D intensity of OEMs in the automotive industry seem to have no significant im-pact on firm innovativeness in any of the models which contradicts previous findings (e.g. Lin et al., 2012), whereas firm size significantly influences firm innovativeness throughout the whole sample, which aligns with prior research (e.g. Cui & O’Connor, 2012). Lastly, this study also illustrates that a combined approach of alliance portfolio size, incorporated through the APS-Index, yields better results than conducting research on both pillars separately. This im-plies that research should put more attention towards a combined approach as suggested by Wassmer (2010).
5.2 Managerial Implications
5.3 Limitations and Future Research
Indeed, using patents as a count measure for innovativeness assigns the identical weight for each single patent, regardless of their importance to the company and thus may eventually not reflect the actual innovativeness of a firm. Following Hagedoorn & Cloodt (2003), future re-search may broaden the assessment of firm innovativeness by adding and combining different innovativeness measures like patent forward citations, new product announcements or R&D inputs.
Overall, the presented concepts and models are subject to some limitations but simulta-neously provide an interesting base for future research in the field of alliance portfolio config-uration and firm innovativeness.
6 CONCLUSION
7 ACKNOWLEDGMENTS
I see this thesis as the final step to successfully completing the Master of Science degree in International Business & Management at the University of Groningen. Although the past four months have been intense and exhausting I am thankful for this experience as I look at it as a good preparation for my future career. And even though this project sometimes pushed me to my limits, I learned a lot about myself and what I am capable to accomplish.
8 REFERENCES
Acs, Z., & Audretsch, D. (1989). Patents as a measure of innovative activity. Kyklos 42: 171– 180.
Ahuja, G. (2000a). Collaborative networks, structural holes, and innovation: A longitudinal study. Administrative Science Quarterly, 45: 425-455.
Ahuja, G. (2000b). The duality of collaboration: Inducements and opportunities in the for-mation of interfirm linkages. Strategic Management Journal, 21(Special Issue): 317-343. Ahuja, G., & Katila, R. (2001). Technological acquisitions and the innovation performance of acquiring firms: A longitudinal study. Strategic Management Journal, 22(3), 197-220.
Anand, B. N., & Khanna, T. (2000). Do firms learn to create value? The case of alliances. Strategic Management Journal, 21: 295-315.
Baum, J., & Calabrese, T., & Silverman, B. (2000). Don’t go it alone: Alliance network com-position and start-ups’ performance in Canadian biotechnology. Strategic Management Jour-nal, 21: 267-294.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of Manage-ment, 17(1): 99–120.
Belderbos, R., Carree, M., & Lokshin, B. (2006). Complementarity in R&D Cooperation Strat-egies. Review of Industrial Organization, 28(4), 401–426.
Billington, C., & Davidson, R. (2013). Leveraging Open Innovation using Intermediary Net-works. Production and Operations Management, 22(6), 1464-1477.
Burt, R. (1980). Autonomy in a social topology. American Journal of Sociology, 85(4), 892– 925.
Burt, R. (1992). Structural Holes: The Social Structure of Competition. Harvard University Press: Cambridge, MA.
Camison, C., & Fores, B. (2010). Knowledge Absorptive Capacity: New Insights for its Con-ceptualization and Measurement. Journal of Business Research, 63, 707–715.
Chang, Y., Thomas, H. (1989}. The impact of diversification strategy on risk-return perfor-mance. Strategic Management Journal 10(3): 271–284.
Chesborough, H. (2005). Open Innovation: The New Imperative for Creating and Profiting from Technology. Harvard University Press, Boston.
Chung, S., Singh, H., & Lee, K. (2000). Complementarity, status similarity and social capital as drivers of alliance formation. Strategic Management Journal, 21: 1-22.
Cohen, W., & Klepper, S. (1996). A Reprise of Size and R&D. Economic Journal, 106(437), 925–951.
