1
The Concept of Strategic Groups and Interfirm
Collaboration: Effects on Firm Innovation Performance
Master’s Thesis MSc. Business Administration – Strategic Innovation Management
July 2017
Supervisor: Dr. Charles Carroll
Co-Assessor: Dr. Pedro de Faria
Melanie Evelyn Flennert
S2981238
Langestraat 130
9712 MG Groningen
m.flennert@student.rug.nl
University of Groningen
Faculty of Economics and Business
2 ABSTRACT
Theoretically, this study links strategic management literature and industrial organization economy with the relational view. Practically, it investigates the concept of strategic groups and interfirm collaboration and its effects on firm-level innovation performance within the North American pharmaceutical and biotechnology industry. Contrary to prior strategic group research, this research focuses on innovation performance instead of financial firm performance. Following the structure-conduct-performance (SCP) paradigm, these three components were examined successively. By conducting a cluster analysis, the structure of this current research was found to be a five-cluster-solution. These clusters also showed – as expected – differences in their firm level innovation performance. In the next step, the conduct of firms within and between these strategic groups have been studied. Interfirm collaboration as a means to deal with resource scarcity, growing complexity and high failure rates therefore represents exactly this conduct. Yet, this collaborative engagement can take place within (i.e. intragroup) and between (i.e. intergroup) strategic groups, but also among firms whose one partner is not part of the underlying sample (i.e. extragroup). Results have confirmed that firms are more likely to form complementary alliances (i.e. intergroup or extragroup) compared to pooling alliances. Additionally, it has been found that complementary interaction is also more likely to result in higher firm-level innovation performance compared to pooling alliances.
Keywords: Structure-conduct-performance (SCP) paradigm, strategic groups, alliances, interfirm collaboration, alliances, innovation performance
3
Table of Content
ABSTRACT ... 2
1. INTRODUCTION ... 4
2. THEORY AND HYPOTHESES DEVELOPMENT ... 7
2.1 Underlying Theories ... 7
2.1.1 The Structure-Conduct-Performance Paradigm ... 7
2.1.2 The Relational View: The Extended Resource-Based View ... 8
2.2 The Concept of Strategic Groups and its Rationales ... 8
2.2.1 Strategic Groups: The Structure within an Industry ... 8
2.2.2 Firms’ Conduct and Strategic Behavior ... 11
2.2.3 Firm-and-Group-level Performance ... 16
3. RESEARCH METHODOLOGY ... 19
3.1 Research Setting ... 19
3.2 Data Collection ... 19
3.3 Variable Description ... 21
4. ANALYSIS AND RESULTS ... 25
4.1 Descriptive Statistics ... 25
4.2 Cluster Analysis ... 25
4.3 Interfirm Collaboration and Innovation Performance ... 29
5. DISCUSSION AND CONCLUSION ... 36
REFERENCES ... 41
4
1. INTRODUCTION
Firms strive for the best strategic positions within their industry and aim to outperform competitors. Analyzing industry structures, firms’ strategic behavior and conduct can help to understand these interrelationships and their effects on firm performance. Likewise, it has long been a major interest and endeavor of both academics and practitioners to study these relationships (Short, Ketchen, Palmer and Hult, 2007).
Coined in the 1970s, strategic groups have become a famous concept within strategic management to better understand exactly these structures and the progress within industries (Hunt, 1972). Strategic groups are an intriguing and rather abstract phenomenon. Put simply, a strategic group can be regarded as a set of firms that is more homogeneous within groups and more heterogeneous between groups (Porter, 1979). The concept helps to understand and describe how firms within an industry relate to each other and behave to influence their strategic positions and thereby their performance (Porter, 1979; Ferguson, Deephouse and Ferguson, 2000).
Strategic group research brought together two fields of research, industrial organization and strategic management. But it also combines two levels of analysis, the industry-level and the individual firm-level. The result of this consolidation is a middle ground level of analysis in between the two other levels: the strategic-group level (Porter, 1980; Leask and Parker, 2006).
Since the initial exploration of this concept, strategic group research blossomed. Some scholars focused on the different dimensions and industry-specific variables that can be applied to define and determine strategic groups (Cattani, Porac and Thomas, 2017). Others examined the effect of the strategic group structure on rivalry, which, in turn, affects performance (Cool and Dierickx, 1993). Peteraf and Shanley (1997) studied the relationship between strategic group identity and its consequences for reputation.
In general, strategic group research designates a rather fragmented field revealing quite conflicting and controversial findings. Some scholars have harshly criticized the whole field of research (Hatten and Hatten, 1987). Claiming that the methodologies are insufficient and inappropriate to prove the existence of strategic groups, they even suggested to abandon strategic group research (Barney and Hoskisson, 1990).
5 economics, this paradigm provides a rationale to examine these relationships between industry structure, firms’ conduct and performance.
According to this paradigm, strategic groups represent the structure within an industry. Studying the structure of strategic groups and how firms strategically (inter)act within and between groups, can help to explain firm performance differences among groups. Determining this structure-performance relationship by focusing on the group-membership-performance relationship has also been one of the major research interests among strategic management and industrial organization theorists (e.g. Cool and Schendel, 1987, 1988; Mehra, 1996; Leask and Parker, 2006; Chittor and Ray, 2007). Yet, research has mainly focused on the financial dimension of performance (Zúñiga-Vicente, de la Fuente-Sabaté and González, 2004).
Fast changing technologies, growing complexity and globalization, amongst other factors, have led to a remarkable increase in competition in many industries. This demands firms to further adapt and improve to be able to be competitive enough to not go bankrupt and to outperform others. Additionally, some industries depend on cost-intensive and scarce research and development (R&D) activities to serve the increasingly complex and difficult customer needs (Roijakkers and Hagedoorn, 2006). Thus, their competitive advantage is to a large extent determined by their innovation performance and without being able to innovate these firms will fall on hard times. Pharmaceutical and biotechnology firms, for instance, fall within that category of firms. For them, innovation performance is a necessary evil to be successful in financial terms and to be able to outperform industry competitors (Leask and Parker, 2006; Rebière and Mavoori, 2016).
But how do these firms manage these challenges? It is that “c”, the “conduct” of the SCP paradigm that matters and can make the difference. Having enough time, money and luck, one can do everything by oneself. Yet, all these three are short in supply, especially in a globalized world. Firms need to cooperate to be successful, otherwise they are too slow, it is too expensive or the firms simply do not know how to be successful (Ohmae, 1989). Thus, interfirm collaboration and alliance formation has become essential to firms’ strategies and conduct (Roijakkers and Hagedoorn, 2006).
6 their complex R&D processes (Roijakkers and Hagedoorn, 2006). This shows that the “conduct” of firms plays an important role.
Since firms within the pharmaceutical and biotechnology industry depend to a large extent on their innovation performance, this growing interfirm collaboration can also be observed within these industries (Roijakkers and Hagedoorn, 2006). Firms can pool their resources to achieve economies of scales or to complement each other’s unique resources and skills and recombine them in a novel way to create innovations (Baum, Calabrese and Silverman, 2000; Lavie, 2006; Wuyts and Dutta, 2014).
By linking strategic management and industrial organization economics with the relational view, this study examines the concept of strategic groups within the North American pharmaceutical and biotechnological industry. Within that research setting, the phenomenon of interfirm collaboration within and between these strategic groups as well as its consequences and effects on firm innovation performance is investigated. Accordingly, this thesis aims to contribute to extant strategic group and SCP literature by answering the following research questions:
1. How does interfirm collaboration take place within and between strategic groups? a. Do strategic groups differ in their innovation performance?
b. What is the effect of interfirm collaboration within and between strategic groups on firm-level innovation performance?
Figure 1: Structure of the Underlying Thesis 5. Discussion & Conclusion
Findings Managerial & Research Implications Future Research Study Limitations 4. Analysis & Results
Descriptive Statistics Cluster Analysis ANOVA & T-Tests 3. Research Methodology
Research Setting & Data Collection Variable Description & Measurements 2. Theory & Hypotheses
SCP Paradigm The Relational View SG & its Rationales Hypotheses Development 1. Introduction
7
2. THEORY AND HYPOTHESES DEVELOPMENT
This second chapter provides the theoretical background the underlying study hinges on. The first section (section 2.1) introduces the SCP paradigm and the relational view. The second section (2.2) is divided into three sub-section structured according to the SCP paradigm. In 2.2.1 the concept of strategic groups as the structure of an industry and its relevant tenets are introduced. The subsequent sub-section (2.2.2) describes the issue of interfirm collaboration representing the conduct of firms within an industry. Linking the structure-conduct relationship leads to the last sub-section (2.2.3) in which this research examines the effect of firms’ conduct based on the industry structure on performance of strategic group members.
2.1 Underlying Theories
2.1.1 The Structure-Conduct-Performance Paradigm
The endeavor to examine industry structures and study the determinants of competitive forces to explain economic returns has been a major issue of industrial management economics (Panagiotou, 2006). This so-called industry structure view explains performance with a firm’s membership in an industry that, in turn, shows specific structures, such that the level of analysis is set at the industry-level (e.g. Rumelt, 1991; Schmalensee, 1985).
Industrial organization’s economists developed the structure-conduct-performance (SCP) paradigm. This paradigm, that has also been adopted by strategic management theorists, helps to study how the underlying industry structure relates to and impacts a firm’s conduct, that, in turn, affects its market performance. It is therefore a well-known tool to examine and explain economic returns. According to the SCP paradigm, structure affects conduct, which affects performance (Bain, 1951, 1956; Mason, 1949; Lipczynski and Wilson, 2004).
Under the extreme conditions of either perfect competition or monopoly, the SCP paradigm is rather deterministic. Since the structure of an industry explains firms’ conduct and the firm’s conduct, in turn, explains performance, it can be assumed, by transivity, that performance is determined by the industry structure (Cool and Schendel, 1987; Newman, 1978). This rather deterministic and parsimonious assumption allows to exclude the “C” and therefore to ignore the behavior of firms (conduct), when applying the SCP model. However, this convenient approach can only be applied under the described conditions of perfect competition and monopoly and not under conditions such as oligopolistic conditions.
8 conduct and that could induce changes in the industry structure (Phillips, 1976; Clarke, 1985). While that is an ongoing debate, strategic group literature has acknowledged that all three components of the SCP are interrelated and have an impact on each other (Lawless, Bergh and Wilsted., 1989; Flavián and Polo, 1999; McNamara, Luce and Tompson, 2002; Panagiotou, 2006). Although interrelations among the three components might be the case, this will not represent the focus of the underlying research.
2.1.2 The Relational View: The Extended Resource-Based View
Traditionally, strategic management theorists and the representatives of the resource-based view (RBV) proposed to keep resources and capabilities within firm boundaries to sustain their competitive advantage. These scholars take a firm-level view and argue that it is not a firm’s membership in an industry, but a firm’s unique bundle of resources and capabilities that explains superior firm returns. Firms have therefore different strategies to invest in certain resources and even in resource bundles that are costly and difficult to imitate. “Isolating mechanisms”, “barriers to imitation”, “causal ambiguity” represent such instruments to keep resources within firm boundaries and thereby sustain the firm’s competitive advantage. These mechanisms render imitation strategies of competition less beneficial or even useless (e.g. Wernerfelt, 1984; Barney, 1991).
Later on, Dyer and Singh (1998) introduced their so-called relational view claiming that a competitive advantage and rents cannot only be achieved by keeping resources and capabilities within the firm boundaries, as the resource-based view claims. Indeed, the latter suggest that relational rents can also be generated by engaging in interfirm collaboration and alliance formation with other firms while crossing firm boundaries. When studying the sources of performance, this research follows this latter view assuming that cooperation among firms is rather beneficial and that firms’ resources and capabilities can also be the source of superior rents (Dyer and Singh, 1998).
2.2 The Concept of Strategic Groups and its Rationales
2.2.1 Strategic Groups: The Structure within an Industry
9 in terms of key strategies and determinants of competitive advantage and heterogeneous between strategic clusters (Short, Ketchen, Palmer and Hult, 2007; McGee and Thomas, 1986; Barney and Hoskisson, 1990; Nohria and Garcia-Pont, 1991).
Strategic groups help to describe and analyze these performance effects, however by studying the intra-industry structure (Cattani, Porac, Thomas, 2017). Thus, from a theoretical lens, strategic groups can be regarded as an intermediate level of analysis or a compromise between the industry-level and the firm-level: the strategic-group level (Hunt, 1972; Fiegenbaum and Thomas, 1990; McGee and Thomas, 1986; Short, Ketchen, Palmer, Hult, 2007). In this study, this intermediate level of analysis is applied to explain firms’ performances. Referring to the SCP paradigm, strategic groups represent the “S” and thereby the structural setting of current research. The following sections provide the necessary theoretical and historical background concerning strategic groups.
The Problem of “Infinite Dimensionalities”
Following the SCP, it first needs to be determined what actually represents the “S”, the structure that will be examined in this study. In this case this means, one needs to determine what are the strategic groups and how are they defined. As mentioned above, prior literature proposes different dimensions and variables of strategic choices to cluster the industry our study. Yet, these strategic choices differ from industry to industry and firm to firm and from strategic group to strategic group (McGee and Thomas, 1986).
10
Table 1. Overview - Determinants of Strategic Group Clustering
Variable Literature
Firm-specifications
(Firm size, age, …) Primeaux, 1985; Porter, 1973; Hatten, 1974; Hatten and Schendel, 1977; Caves and Pugel, 1980; Ramsler, 1982; e.g. Hatten, 1974; Hatten and Schendel, 1977; Hatten, Schendel and Cooper, 1978; Hergert, 1983; Hawes and Crittenden, 1984; Hatten and Hatten, 1985
Financial strategy
(current ratio, return on assets, …)
e.g. Baird and Sudharsan, 1983; Primeaux, 1985; Ryans and Wittink, 1985; Hatten, Schendel and Cooper, 1978
Marketing variables
(Promotion, target market, number of brands, …)
e.g. Hatten, 1974; Hatten and Schendel, 1977; Hergert, 1983; Hawes and Crittenden, 1984; Hatten and Hatten, 1985
Customer-specifications
(Customer needs and groups)
Howell and Frazier, 1983; Hayes, Spence and Marks, 1983
Source: own figure show overview of different determinants of strategic group clustering, applied by prior studies and based upon literature review by McGee and Thomas, 1986
Conflicting Views on Strategic Groups
Any strategic group study hinges on the a priori determination of strategies that represent the variables for clustering firms. Again, referring to the problem of infinite dimensionality confirms the diversity of findings and controversies. However, the discussion is not only about theory. Indeed, it is especially the methodological approach that has been criticized (Barney and Hoskisson, 1990; Hatten and Hatten, 1987).
As a common methodological approach to identify strategic groups within an industry, most researchers conduct cluster analyses based on different characteristics and variables (Leask and Parker, 2006). A cluster analysis is a method that groups together different sets of similar observations (Ketchen and Shook, 1996). Therefore, it is not evident whether these clusters result in truly existing groups found through that method or whether these clustered subsets are a more even distribution of firms that have been created through the application of that method.
This has even been criticized as an “embarrassment to strategic management” (Ketchen and Shook, 1996, p. 442), especially due to a lack of significance tests for cluster analyses (Barney and Hoskisson, 1990; Ketchen and Shook, 1996; Hatten and Hatten, 1987). According to these critics, lacking significance tests weaken the meaningfulness of the research outcomes that apply such cluster analyses for identifying strategic groups and do not sufficiently prove their actual existence (Barney and Hoskisson, 1990; Hatten and Hatten, 1987).
11 research has fallen on hard times. This problem has led to a major debate in strategic group research and split the extant body of research into two fundamentally different and conflicting views on strategic groups: the interdependent and the independent view (Carroll, 2017).
The independent view refers to the concept of strategic groups more as a man-made analytical tool or even an analytical convenience to analyze and better understand the structures within an industry. Hatten and Hatten (1987, p. 329) regard it as a tool “to segment industries into sets of companies whose competitors, actions, and results are relevant to each other”. They argue that it might be convenient to cluster an industry into subsets of firms that are more analyzable and that make it easier to understand the industry structure. They argue that similar groups might be likely to make similar strategic choices. Strategic groups could help to predict how firms react to their environment and explain performance differences across strategic groups (Hatten and Hatten, 1987; Carroll, 2017). However, from this perspective it is assumed that firms within and across these strategic groups are not interdependent, in the sense that they are not necessarily interacting among each other. Hatten and Hatten (1987) claim that researchers have the tendency “to anthropomorphize the group and forget that it is our invention, not a real feature of industrial life discovered by fundamental research” (Hatten and Hatten, 1987, p.334).
The interdependent view, on the other hand, argues that strategic groups are not independent, but rather interdependent subsets of firms that are likely to interact among each other. If firms belong to the same strategic group they are likely to follow similar strategies and even target similar customers and suppliers. This implies that firms within one group might pay attention to their environment and their rivals to reduce uncertainty and cope with bounded rationality (Peteraf and Shanley, 1997). To maximize their own profits, firms might consciously observe each other’s actions and strategies and then react accordingly. This mutual awareness among firms might even reinforce the level of rivalry within a group. Thus, according to the interdependent view, strategic groups might even reflect the structure of rivalry within an industry (Tang and Thomas, 1992; Caves and Porter, 1977).
2.2.2 Firms’ Conduct and Strategic Behavior
12 The interdependent view makes it possible to study the behavior of firms, but the “C” of the SCP paradigm that is brought into play here may still take several forms. By relating this to the relational view, which has been introduced before, it is assumed that firms can also achieve superior rents by engaging in interfirm relationships, or put simply, by collaborating with other firms (Dyer and Singh, 1998). Inter-group performance differences could then be the result of an idiosyncratic combination of both collaborative and competitive intra-group interaction (Dranove, Peteraf and Shanley, 1998).
Set into that context of strategic groups, collaborative engagement takes place in the form of interfirm-alliance formation. Since strategic groups cause that firms appear as clustered subsets, this interfirm collaboration can occur in two different forms: (1) among firms within the same group (i.e. intragroup interaction) and (2) among firms belonging to different groups (i.e. intergroup interaction).
The alliance data has been collected for this study in such a way that only one partner needed to be part of the regular company dataset. Likewise, the alliance data includes worldwide alliance data (for further details and restrictions, see chapter 3). Due to this, it is possible that an alliance is formed between any of the focal firms from the sample and an alliance partner that is not included in the main company sample. Such an alliance can therefore not be assigned to intergroup interaction nor to intergroup interaction. For the sake of clarity and practical convenience, the term extragroup interaction was coined that indicates such an alliance with one partner in the company dataset of this study and one partner that is not included in the sample and can additionally also be operating in any other industry. From an alliance perspective, extragroup interaction resembles intergroup interaction, because both interaction types designate an alliance formation across strategic groups.
Competitive Forces within Industries
Firms’ behavioral and strategic actions and interactions with other firms such as alliance formation depend to some extent on the degree of competition within the environment they are operating in. Many researchers have studied competition in relation to strategic groups. An interesting question that emerged in extant literature is whether competitive forces are greater between or within strategic groups (Dranove, Peteraf and Shanley, 1998; Porter, 1979, 1980; Hatten and Hatten, 1987; Nicolau-Gonzálbez and Ruiz-Moreno, 2014).
13 competitive environment. Furthermore, the high degree of similarities is likely to develop some kind of mutual understanding and awareness among the group members that can develop some sort of a cognitive, intra-industry group. Peteraf and Shanley (1997) term this latter phenomenon a strategic group identity. Rooted in diverse social and isomorphic forces and process, this group identity also increases the degree and the strength of intra-group trust (Smith, Grimm, Wally and Young, 1997). This common fate and trust allows group members to anticipate each other’s strategies and behavioral actions and to reduce uncertainty and bounded rationality (Caves and Porter, 1977; Nohria and Garcia-Pont, 1991).
Other scholars argue that firms with a higher degree of similarities and commonalities are more likely to cause a higher degree of competition among the firms within strategic groups. Likewise, firms of one group are more likely to compete for the same target market and customers (Hatten and Hatten, 1987). Yet, whether firms perceive each other as competitors also depends on other factors such as the industry, the strategic position, overarching goals and environmental constraints (Mas-Ruiz and Moreno, 2011; Nicolau-Gonzálbez and Ruiz-Moreno, 2014).
Thus, competitive forces are likely to occur within strategic groups as well as between strategic groups. Interfirm collaboration, intergroup or intragroup, is therefore characterized by a competitive and a collaborative component and causes different implications for the alliance outcome depending on various factors (Bengtsson and Kock, 2000; Estrada, Faems and de Faria, 2016). Brandenburger and Nalebuff termed this a “coopetitive relationship” between two allying firms (Brandenburger and Nalebuff, 1997; Bengtsson and Kock, 2000).
So, why do firms engage in such risky and tense relationships? There are many and non-straightforward factors that make firms engage in these coopetitive relationships. For instance, anticipating environmental uncertainty and reducing bounded rationality plays an important role in managing coopetitive forces and can be achieved through interfirm collaboration. Furthermore, firms’ resource bases, resource scarcity and fast changing technologies and growing complexity require collaboration among firms to stay competitive. It can also aim at actually reducing the competitive component of these coopetitive relationships and engage in a trustworthy and beneficial alliance, which in turn allows to better anticipate competitors’ strategies and actions.
14
R&D and Knowledge Generation
As explained above, R&D and hence innovations are of utmost importance within the two industries under study: pharmaceutical and biotechnology (Leask and Parker, 2006). To successfully innovate and come up with new ideas is a difficult task which requires that many factors need to be fulfilled. Innovations are often the result of the recombination of existing knowledge and resources (Wuyts and Dutta, 2014). Complementary assets, resources, and strategies are thus an effective means to increase the likelihood of developing a unique, novel recombination of already existing resources. This recombination allows, in turn, to create superior products and innovations. Firms are often facing resource scarcity and possessing a diverse portfolio of resources and skills to come up with novel recombination by themselves is too time-consuming, cost-intensive and thus less effective (Lavie, 2006; Wuyts and Dutta, 2014; Sampson, 2007).
Thus, diversity in firms’ resources bases increases the level of novelty and unique knowledge recombination. However, literature suggests that the degree of diversity should not exceed a moderate level (considering dyad as well as alliance portfolios), mentioned here only as an incidental remark (Sampson, 2007; Wuyts and Dutta, 2014). The two research-driven industries that are examined in this research are highly dependent on new product developments and innovations to stay competitive (Rebière and Mavoori, 2016; Cool and Dierickx, 1993; Leask and Parker, 2006).
There is little question that R&D and innovation constitute one of the key sources of competitive advantage and therefore for success in the pharmaceutical and biotechnology industry. The molecular biology has opened up new opportunities for the production and discovery of new drugs, but it has also accelerated competitive behavior. Considering this renders innovation performance nowadays a key driver to success, more than ever before (Cockburn, Henderson, Orsenigo, and Pisano, 1999; Bottazzi, Dosi, Pammolli and Riccaboni, 2000). As introduced in the introduction, firms need to engage in interfirm collaboration to be effective and to manage these challenges.
Interfirm Collaboration
As explained above, collaborative, and thus coopetitive engagement, can occur in this study in three different forms: intragroup, intergroup and extragroup.
15 that partner firms collaborating intragroup face similar resource bases and strategic options, which in turn has predetermined outcomes for their collaboration.
First, and probably most important, an intragroup collaboration enables partner firms to pool their (similar) capabilities and resources to achieve higher efficiency through greater scale and scope economies, production rationalization and the convergence of technologies. This enables them to enhance their competitive and strategic position in the industry. Thus, firms in this research setting, can form alliances to specialize in certain drug areas and then jointly compete against rivals through the benefits of these synergistic and more cost-efficient effects (Lavie, 2006; McCutchen and Swamidass, 2004). Making use of such synergism to leverage existing knowledge, resources and skills to enhance the production and its efficiency is also known as exploitation. The effects of such exploitative activities become even stronger when having a portfolio of multiple alliances with similar resource bases (e.g. intragroup alliances) (Levinthal and March, 1993; Lavie and Rosenkopf, 2006; Faems, van Looy and Debackere, 2005).
Second, intragroup interfirm collaboration can help overcome competitive forces for a higher purpose. Group members could collaborate for reasons such as to jointly defend themselves and their strategic positions from other groups or to enhance their industry-wide reputation by suing their group-specific capabilities and resources (Nohria and Garcia-Pont, 1991; Peteraf and Shanley, 1997).
The reasons and implications might differ and depend upon various factors. Yet, firms within the underlying industries depend upon new knowledge creation. Allying with partners of the same group might help firms within manufacturing to enhance their productivity and efficiency, but might not lead to the intended outcome of creating novel combinations and enhancing firms’ innovation performance (Lavie, 2006). It is therefore expected to be less important and thus likely to take place within this specific research setting.
16 which the strategies of partners are complementary and hence, reflects the potential offered by an alliance to support the partners’ respective strategic thrusts” (Hamilton, 1990, p. 150).
Considering most firms are facing resource restrictions, firms cannot easily pursue various knowledge trajectories and specializations (Chirstensen, 1997; Rosenkopf and Almeida, 2003). Thus, such types of collaborative efforts are a more efficient, flexible and less cost-intensive way of creating new knowledge and resulting in a knowledge recombination that leads to superior products and innovations (Lavie, 2006; Lavie and Rosenkopf, 2006; Sampson 2007; Wuyts and Dutta, 2014).
The fact that the number of interfirm relations (in general) is remarkably increasing is not surprising and supports the presented arguments (Hagedoorn, Roijakkers and Kranenburg, 2006; Rebière and Mavoori, 2016; Cool and Dierickx, 1993; Leask and Parker, 2006). Thus, the more experienced, older, but more rigid, large pharmaceutical companies and the smaller and younger, agile and more specialized biotechnological firms with their low degree of similarities are likely to be members of different strategic groups. Such complementary alliance partners are more likely to create superior products and innovations compared to two similar firms belonging to one strategic group (Baum, Calabrese, Silverman, 2000; Diestre and Rajagopalan, 2012). For this research setting, it is thus expected and hypothesized that:
Hypothesis 1 (H1): The phenomenon of interfirm collaboration is more likely to aim at complementary alliances and thus, more likely to take place intergroup or extragroup than intragroup.
2.2.3 Firm-and-Group-level Performance
Managers knowing and understanding their firm’s group membership and position within the industry can serve as a kind of reference point to analyze their competition and respond to each other’s strategic actions to enhance their performance (Fiegenbaum, Hart and Schendel, 1996). A plethora of strategic group researchers analyzed the relationship of strategic group membership and firm or group-level performance (e.g. Cool and Schendel, 1987; Cool and Schendel, 1988; Fiegenbaum and Thomas, 1990; Hatten and Hatten, 1987; Peteraf and Shanley, 1997; Short, Ketchen, Palmer and Hult, 2007). From these studies, one can conclude that performance differences among strategic groups might exist.
17 issue of defining it in a manner that it is still a meaningful measure of the underlying construct to be studied (Short et al., 2007).
Yet, the occurrence of performance differences hinges upon various factors. For instance, the different theoretical rationales such as the industrial organization theory (IO) or the resource-based view (RBV) that underlie the different studies or the particular industry under study influence the determination and investigation of performance differences among groups. Performance differences are thus possible, but not guaranteed.
For instance, this study has already explained that the RBV explains performance differences among firms through resource heterogeneity (Barney, 1991). According to the IO, the expectation that there exist performance differences between strategic groups has its roots in the concept of mobility barriers. Assuming that business opportunities are not evenly distributed across an industry, some areas are more profitable than others (resource heterogeneity). Such mobility barriers represent firm strategies that can then hinder other firms to change groups due to the risks of higher costs than benefits and thus explains why between-group performance variance is greater than within-group variance (Mascarenhas and Aaker, 1989; Short et al., 2007; Caves and Porter, 1977; Hatten and Hatten, 1987).
Table 2. Overview - Performance Measures
Performance Measure Literature
Return on Assets e.g. Dess and Davis, 1984; Cool and Schendel,
1988; Lawless, Bergh and Wilsted, 1989; Nair and Kotha, 2001; Mehra, 1996; Zúñiga-Vicente, de la Fuente-Sabaté and González, 2004; Short et al., 2007
Return on Equity Lawless, Bergh and Wilsted, 1989; Mehra, 1996;
Zúñiga-Vicente, de la Fuente-Sabaté and González, 2004; Short et al., 2007
(Adjusted) Return on Sales Nair and Kotha, 2001; Cool and Schendel, 1987, 1988; Lawless, Bergh and Wilsted, 1989
Annual Sales Growth Dess and Davis, 1984
Tobin's Q Thomas and Waring, 1999; Short et al., 2007
Source: own table, overview of variables used by prior research to assess performance differences between strategic groups
18 Considering the underlying industries and the importance of R&D and innovation to outperform other firms, this study examines firm performance in terms of innovation performance. Most extant research paid attention on how firms perform financially (see Table 2). Yet, for some industries such as the pharmaceutical and biotechnological industry, financial performance is determined by their innovation performance. Being successfully in terms of new knowledge creation and R&D activities is very likely to result in superior financial performance
Therefore, it is interesting, especially when considering the biotechnological sector, to pay attention to innovation performance, as a prior step to financial performance. This study therefore intends to investigate strategic groups and interfirm collaboration in relation to their innovation performance. Considering the presented arguments, it is expected that firms belonging to one group do not only face similar levels of financial performance, but also similar levels of innovation performance and therefore hypothesize the following:
Hypothesis 2 (H2): Firms belonging to the same strategic group have a more similar level of innovation performance than firms belonging to different strategic groups. It is the presented conduct that can help firms to influence their strategic position. Firms can ally with firms belonging to the same strategic group, but also with firms belonging to a different group (or are even outside the underlying sample). Through complementary alliances, firms are able to recombine their existing with the partner firms’ new knowledge and thus create superior products and innovations. All three presented types of interaction (intragroup, intergroup, extragroup) show advantages and can be beneficial.
Yet, it is especially intergroup and extragroup interaction that are likely to enhance firms’ innovation performance. Given this particular research setting and based upon these arguments as well as those made in section 2.2.2 it is expected that intergroup and extragroup interaction are more likely to enhance firms’ innovation performance. Put differently, it is therefore hypothesized that:
19
3. RESEARCH METHODOLOGY
Since the business phenomenon of strategic groups is theoretically conceptualized one, a theory testing approach was applied for this study (van Aken, Berends, and van der Bij, 2012). The following sections describe each of the methodological steps that were necessary to conduct this study and test the hypotheses.
3.1 Research Setting
The hypotheses were tested on a sample of 121 North American firms of which 59 are operating in the pharmaceutical and 61 in the biotechnological industry. These two industries were chosen for the following reasons: (1) they have a high degree of similarity in terms of market dynamics, structure, uncertainty, wherefore they compete for similar resources and capabilities; (2) interaction (competition, collaboration, coopetition) constitutes an essential strategy for firms to perform within these industries and is likely to take place across industry borders (Rothaermel and Deeds, 2006). Therefore, these industries constitute a representative business environment to answer the research questions of this study.
A longitudinal study was conducted and a relatively large time frame was used to ensure the inclusion of sufficient alliance data that outperform expectations. Therefore, the main observation period is from 2008-2012. This time frame has been chosen for rather practical reasons such as 2012 being the most recent available year in the SDC Thomson Reuters Alliance database and that five-years are assumed to be the average duration of an alliance (Sampson, 2007). For the alliance data collection, this time frame was extended by the years 2003-2007 since the first round of data collection (2008-2012) did not reveal a sufficient number of observations. Expecting that an alliance formed in 2003 will still be existent in 2008, this approach ensured that there was a more comprehensive alliance dataset (Sampson, 2007). All other company data, however, were collected only for the observation period 2008-2012.
3.2 Data Collection
20 The first step was taken with the help of the Orbis database. North America (United States and Canada) served as the geographical scope for this study and only firms that are incorporated in this region, were selected. As explained above, the pharmaceutical and the biotechnological industry were chosen as the underlying industries. Their similarities are also reflected in their Standard Industry Classification (SIC) codes that are rather close, indicating a similar and even overlapping business environment (pharmaceutical firms: SIC 2834 and biotechnological: SIC 2836). Since a comprehensive dataset was necessary to have sufficient data, company data within that industry and geographical scope was identified. Furthermore, all very large and large, medium-sized, small and active companies were selected in Orbis and included in the dataset. Consequently, only inactive firms were excluded. As already mentioned, the period of observation is 2008-2012, wherefore only companies with an incorporation date up to and including 2012 were chosen.
Initially, this search strategy resulted in a dataset of 372 companies. In the second step, this dataset was complemented with information of the Compustat database. Since these data needed to be merged with the Orbis data, the unique company ticker numbers were used as a means of identification to ensure that the dataset stays consistent during the data collection process. Compustat served as a source, especially for financial variables, that were not completely available or missing in Orbis. After the two datasets were merged and missing or outlier variables were excluded, 121 companies remained.
These 121 firms represent the final sample size of this study. Discrepancies in numbers and information among the two databases have been double checked by conducting an extensive internet recherché. If that did not result in a consistent finding, the research relied upon the Compustat data since this is a more reliable source for financial variables.
21 The FDA does not provide a complete database, wherefore this research first needed to merge the data files the FDA provides on its webpage (original file names: “applications.txt” and “submissions.txt”) to create a new database with the information about the NDA per company between 2008-2012. The resulting dataset was relatively large and to avoid human error, a code in Python was created. With this code, data was automatically sorted per company and by year to then retrieve the number of NDAs per company (see Appendix A).
The last step in the process involved the alliance data collection through the Thomson Securities Data Corporation (SDC) Platinum database on Alliances. This comprehensive database contains worldwide alliances and an extensive set of information on these. I identified the alliances for the time frame and for both industries using the primary SIC codes (2834 and 2836). This resulted in a dataset with all alliances formed between 2003 and 2012 and where at least one alliance partner’s operation was in one of the defined industries.
The focus was on two-partner-relationships, multi-partner alliances were therefore split and observed as dyads. Since the dataset is quite large and does not exclusively contain parent firms, some alliances needed to be consolidated and added to the parent firm’s alliances (see Appendix B for clarification). By filtering the dataset for each firm in the sample, it was possible to identify and export all alliances that the companies have formed into a separate file. That way, this study divided between total, intragroup, intergroup or extragroup alliances per firm and group.1
3.3 Variable Description
Strategy Variables
The identification of strategic variables always depends on the industry under study, in this case the North American pharmaceutical and biotechnological industry (Zúñiga-Vicente et al., 2004; Leask and Parker, 2007; Cool and Schendel, 1988). As explained in section 2, this selection process always entails a tradeoff between the collection of industry-specific variables to adequately cluster the firms and the endeavor to be as generalizable as possible. However, this is a necessary evil to be able to explore this interesting research area and one cannot avoid some flaws and restrictions in terms of generalizability. In a first step, an extensive, preliminary research of the underlying industries was conducted and studies analyzed that have already conducted strategic group research within the pharmaceutical and biotechnological industry.
This resulted in a theory-based identification of more than 20 strategic variables. In practice, some more restrictions were necessary. First, to maximize validity and reliability of the
22 data and to reduce bogus results, longitudinal data was collected. Therefore, all variables that were simply not accessible or not available for the five-year time frame (2008-2012) have been eliminated. For a cluster analysis, one should exclude or reduce the amount of categorical and dummy variables and – as for any study – highly correlated data.
For the analyses, the natural logarithm (ln) for the three strategy variables was calculated and used: current total assets, cost intensity and firm age. To account for a positively skewed distribution, values of 0 have been transformed to a value close to 0, namely 0,005. Thus, the final set of strategy variables included the following six variables: (1) the ln of the five-year average of current total assets, (2) the year average of cost-intensity, (3) the ln of the
five-year average of research & development intensity, (4) dummy variable: wholesale as a main firm activity; (5) dummy variable: service as a main firm activity and (6) the ln of a firm’s age.
Dependent variables
This study applies innovation performance instead of firm financial performance and included three variables (three dependent variables) to measure this: (1) total number of patents, (2) innovation efficiency and (3) total number of new drug applications (NDAs). These three measurements are all indicators for innovation performance, however, they are not similar enough to be taken as three measures for the same construct. This is shown be the value of the Cronbach’s alpha. This is a measure of internal consistency, that is, how closely related a set of items are as a group. Technically speaking, Cronbach’s alpha is then a coefficient of reliability (consistency). The result shows a Cronbach’s alpha of 0.449. This relatively low score indicates a relatively low internal consistency indicating that it would not be reliable to measure the construct of innovation performance by taking a mean of the three measures. Therefore, all three measures are separately considered in each of the analyses.
Independent variables
To test the hypotheses, the concept of strategic groups was connected with alliance literature to observe the phenomenon of interfirm collaboration within and between strategic groups. Therefore, the three independent variables were collected through the Thomson SDC Platinum database on Alliances.
23 Total number of intergroup alliances. The total number of intergroup alliances designates the sum of all alliances formed by a focal firm with another firm that is not a member of the same group as the focal firm, but included in the sample.
Total number of extragroup alliances. The total number of extragroup alliances designates the sum of all alliances formed by a focal firm with another firm that is not a member of the same group as the focal firm and not included in the sample.
24
Table 3. Overview – Variables and Measurements
Variable Frame Time Data Source Measurements Literature Strategy Variables 2008-
2012
Current total assets Compustat Ln of (five-year average of current total
assets), in million U.S.$ Nohria and Garcia-Pont, 1991
Cost intensity Compustat Ratio of five-year average of sales/five-year average of costs of goods sold, measured in million U.S.$
Segars and Grover, 1994 R&D intensity Compustat Ln of (Ratio of five-year average of
R&D expenditure/five-year average of sales), measured in million U.S.$
Cool and Schendel, 1988; Leask and Parker, 2007
Wholesale Orbis Dummy variable; 1=wholesale is one of
the firm’s main activities; 0=wholesale is not one of the firm’s main activities
Leask and Parker, 2006
Service Orbis Dummy variable; 1=service is one of
the firm’s main activities; 0=service is not one of the firm’s main activities
Leask and Parker, 2006
Firm age Orbis Ln of (2012-incorporation date),
measured in years
Ferguson et al., 2000
Alliances 2003- 2012
Intragroup SDC Sum of total number of intragroup
alliances of ten years -
Intergroup SDC Sum of total number of intergroup
alliances of ten years -
Extragroup SDC Sum of total number of extragroup
alliances of ten years -
Innovation Performance
2008-2012
New drug
applications (NDAs)
Drugs@FDA Total number of new drug applications of the focal firm that has been approved within the five years
Bierly and Chakrabarti, 1996; Chittoor and Ray, 2007 Innovation efficiency Compustat,
Orbis
Ratio of number of patents over five years/R&D expenditures, in million U.S.$
Martens, 1988
Patents Orbis Total number of patents the focal firm
25
4. ANALYSIS AND RESULTS
4.1 Descriptive Statistics
Table 4 presents the variables and data that has been used for both the cluster analysis and the follow-up analyses (frequency tabulations, ANOVA, t-tests). Variables 1 – 6 describe the strategy variables that have been used for the cluster analysis, variables 7,8 and 9 constitute the performance variables and the last three variables determine the three alliance types that have been used for the hypotheses on interfirm collaboration. The table shows some high correlations such as NDA – assets (r=0,700; p<0.01) and extragroup alliances – intragroup alliances (r=0,690: p<0.01) which indicate that multicollinearity could exist. Since these are the highest correlations and still fall below r=0.8, which is commonly regarded as a point at which multicollinearity can become problematic, those variables were left in the sample and therefore included in the analysis (Ketchen, Combs, Russell, Shook, Dean, Runge, Lohrke, Naumann, Haptonstahl, Baker, Handler, Honig, Lamoureux, and Beckstein, 1997).
4.2 Cluster Analysis
A hierarchical cluster analysis was carried out on the six identified strategy variables by applying a comprehensive algorithm coded in SPSS (Carroll, 2017). The firms were labeled as cases using Ward’s hierarchical cluster technique and squared Euclidean distances. The permutation test as well as the Monte Carlo test used 999 iterations to generate the null distributions of the clustering statistic (total within-group variance). Including two significance tests makes us much more confident in accepting consistent findings and increases the validity of our results significantly (Carroll, 2017; Brewer and Hunter, 2006). No-z-scores values were used since no disturbing overemphasis of large values was found. However, the natural logarithm of some variables was used (see description in section 3.3) to account for some minor skewness.
26
Table 4. Descriptive Statistics and Pearson's Correlations
Variable Min. Max. Mean St.D. 1 2 3 4 5 6 7 8 9 10 11 12
1 Assetsa -1,68 10,95 4,36 2,12 1 2 Cost intensity 0 16,99 2,04 3,37 0,541** 1 3 R&D intensitya -5,29 8,20 -0,11 2,49 -0,173 -0,293** 1 4 Service 0 1,00 0,14 0,35 -0,077 -0,003 0,16 1 5 Wholesale 0 1,00 0,31 0,46 0,191* 0,224* 0,02 -0,165 1 6 Firm agea 0 4,83 2,68 0,72 0,453** 0,229* -0,062 -0,033 0,177 1 7 NDAa 0 5,79 0,63 1,35 0,700** 0,543** -0,259** -0,144 0,226* 0,392** 1 8 Patentsa 0 6,76 2,30 1,79 0,677** 0,257** 0,04 -0,025 0,112 0,428** 0,384** 1 9 Innovation efficiencya -5,84 2,11 -0,88 1,24 -0,491** -0,340** 0,079 0,029 -0,034 -0,071 -0,365** -0,009 1 10 Intragroup alliances 0 5,00 0,16 0,68 0,479** 0,145 -0,125 -0,127 0,081 0,347** 0,519** 0,340** -0,240** 1 11 Intergroup alliances 0 8,00 0,26 0,90 0,343** 0,146 -0,015 -0,03 -0,009 0,093 0,325** 0,316** -0,187* 0,449** 1 12 Extragroup alliances 0 67,00 3,41 9,09 0,675** 0,246** -0,148 -0,135 0,047 0,487** 0,588** 0,467** 0,398** 0,690** 0,601** 1
N=121; Correlation is significant at * p<0.05 and **p<0.01
27
Table 5. Cluster Analysis with Significance and Kink Tests
Significance test Kink test
Groups Ward’s Criteria Permutation Probability Monte Carlo Probability Kink Permutation Percentile Monte Carlo Percentile 12 290.879 0.001 0.001 -- -- -- 11 316.313 0.001 0.001 1.015 0.145 0.121 10 349.087 0.001 0.001 1.010 0.279 0.266 9 389.235 0.001 0.001 1.009 0.426 0.345 8 437.872 0.001 0.001 1.002 0.688 0.602 7 493.513 0.001 0.001 1.026 0.405 0.187 6 570.688 0.001 0.001 0.995 0.877 0.816 5 656.553 0.001 0.001 1.154 0.001 0.002 4 871.713 0.001 0.001 1.042 0.128 0.383 3 1206.457 0.001 0.080 0.027 1.000 1.000 2 1547.863 0.001 0.146 1.383 0.001 0.140 1 2746.024 1.000 0.598 -- -- --
The following describes and names each of the five strategic groups and its members separately and according to their strategies (strategic variables, see Table 6). This description ensures a better understanding and in turn, allows conclusions on their collaborative engagement and their innovation performance. An overview of all groups and their firms can be found in Appendix E1-E5.
Table 6. Overview - Strategy Variables per Group
Strategic Group Assetsa Cost
intensity
R&D
intensitya Service Wholesale
Firm agea BT/P- proportionb Strategic Group 1 7,59 5,15 -1,84 0,26 0,11 3,31 0,37 Strategic Group 2 6,32 13,67 -1,41 0,33 0,50 2,86 0,33 Strategic Group 3 3,91 1,01 0,02 0,11 0,32 2,56 0,52 Strategic Group 4 3,22 0,06 3,44 0,14 0,41 2,64 0,64 Strategic Group 5 2,27 0,00 -5,29 0,00 0,22 2,26 0,56
Table shows group averages of strategy variables; a indicates that for this variable the natural logarithm has been used and is shown
28 Strategic Group 1: “Large, Established Firms”
The first group consists of 19 firms marked by the highest mean score of total assets. On the other hand, that group also shows the second lowest average R&D intensity compared to the other groups indicating that these firms spend a relatively low amount of their money for R&D. Notably, this group also includes the oldest firms among all identified groups with an average firm age of 40,89 years. The amount of biotechnology firms is rather small in that group compared to the other ones. Main activities of these firms tend to be more likely service than wholesale. Representing few, but very large multinational companies that are not exclusively operating within the pharmaceutical or biotechnological industry, this group can be characterized as powerful global players.
Strategic Group 2: “Small, Wealthy Firms”
With six group members, this is the smallest and the second oldest group of our sample and the one with the lowest percentage of biotechnological firms. Yet, this group has the second highest value of total assets and by far the highest cost intensity ratio revealing a supernormal amount of sales. On the other hand, group 2 shows the second lowest R&D intensity among the five groups. Considering that this group includes the firms with the highest proportion of service being their main activity and very low R&D intensity, it emphasizes that most firms of group 2 tend to rely upon service activities instead of focusing on R&D activities. Approximately one third of the firms in this group are biotechnological.
Strategic Group 3: “The Mainstream Firms”
Group 3 is by far the largest group with 65 member firms. The average of the total assets of these firms is only half of that of group 1. Group 3 has a rather low value for cost intensity indicating that the group average of sales and the costs of goods sold are almost equal. Interestingly, the firms within that group indicate a comparatively high R&D intensity. The firms engage more in wholesale than in service activities. With an average firm age of 15,46 years, this group is relatively young. The proportion of biotechnology and pharmaceutical firms is equal, 50% of the one industry and 50% of the other. So far, this group does not show very remarkable characteristics and was therefore named accordingly.
Strategic Group 4: “Innovation-Biotech Youngsters”
29 however, the cost intensity is close to zero indicating that the firm did not really sell its products sufficiently to compensate for the costs of the goods sold. Remarkably, group 4 is a very research-intense group revealed by its very high R&D intensity, which could be caused by the high percentage of (mostly young and dynamic) research-oriented biotechnological firms. Due to these characteristics, this group was named “innovation-biotech youngsters”. Furthermore, this group seems to be more active in wholesale activities than in service activities. Concerning its average firm age of 14,68 years, this group is the second youngest group.
Strategic Group 5: “The Young Innovators”
The last group is also rather small and shows the lowest amount of total assets. An average firm age of 10,89 renders this one the youngest group in the sample. Slightly more than 50% of the firms within that group are biotechnological and only approximately one quarter of the firms is active within wholesale as a main activity and interestingly, none within service. A cost intensity of zero indicates that the firms of group 5 do not generate any sales and can therefore also not cover the costs of their goods. Furthermore, the group shows an R&D intensity of zero. Yet, that does not necessarily mean that these firms are not spending money on R&D, but it can also indicate that they are not having any or, at least only very low amount of sales. Indeed, it is the latter that is true for this group. Considering that firms of group 5 are still quite young and focusing on R&D, they need (in the future) more sales to compensate for these investments.
4.3 Interfirm Collaboration and Innovation Performance
Frequency of Alliances
In the first hypothesis (H1), it was stated that the phenomenon of interfirm collaboration is more likely to take place intergroup or extragroup. Since firms within this research setting are aiming at enhancing their innovation performance, it was expected that they will more likely engage in complementary alliances (i.e. intergroup and extragroup) and less likely form pooling alliances (i.e. intragroup). To further examine this, a frequency table was set up depicting the number of alliances. Table 7 shows the sums of the alliances formed by each group, split into the four categories: total, intragroup, intergroup and extragroup alliances.
30 belonging to the same strategic group (i.e. intragroup interaction) has been observed 19 times. Interestingly, most of these 19 intragroup alliances, namely 16, were formed solely by group 1. This latter group is also the only one that shows a higher frequency of intragroup interaction compared to intergroup interaction.
Taken together, these findings confirm the expectations and suggest that firms are more likely to search for complementary resources (i.e. intergroup or extragroup alliances) instead of similar resources (i.e. intragroup alliances). It follows that the first hypothesis (H1) is supported. Nevertheless, further research should provide supplementary examinations into this phenomenon to confirm the findings and test it on a larger scale.
Table 7. Overview - Alliances per Group
Strategic Group Total Intragroup Intergroup Extragroup
Strategic Group 1 303 16 13 274 Strategic Group 2 16 0 2 14 Strategic Group 3 121 3 11 107 Strategic Group 4 22 0 5 17 Strategic Group 5 1 0 0 1 Sum 463 19 31 413
Table shows total number of alliances per strategic group, split into total, intragroup, intergroup and extragroup alliances.
Table 8. Overview - Performance Variables per Group
Strategic Group New drug applicationsa Patentsa Innovation efficiencya
Strategic Group 1 1,14 2,93 -2,49
Strategic Group 2 2,03 1,98 -2,10
Strategic Group 3 -4,32 0,58 -1,46
Strategic Group 4 -5,29 0,37 -1,60
Strategic Group 5 -5,29 0,09 -0,70
Table shows group averages of performance variables,
a indicates that for these variables the natural logarithm has been used.
Innovation Performance
31 et al., 2004; Leask and Parker, 2007; Cool and Schendel, 1988). Since especially one dependent variable had many values of zero for which one cannot calculate the natural logarithm, the natural logarithm of 1 was chosen for these values. This resulted in values of zero instead of dismissing those values. In Addition, that ensures that the distribution of the values is still comparable and close to the original distribution. Table 9 shows significant results for this test on all three dependent variables indicating a potential impact of strategic group membership on firm innovation performance.
To understand where exactly the differences between the groups lie, the results of the Games-Howell post hoc test needed to be interpreted. The multiple comparison reveals which means in performance show statistically significant differences between particular groups. Results for groups 1 and 2 do not show any significant differences for any of the dependent variables indicating that changing group membership between these groups might not significantly affect their level of innovation performance. Group 1 and group 4 (p=0.034) and group 1 and group 5 (p=0.004) show statistically significant differences in their mean performance in number of patents. For innovation efficiency, the Games-Howell post-hoc test does not show any significant differences in groups’ mean performance. Due to the presented reasons, results were only interpretable to a limited extent and hypothesis 2 can only be partially supported.
Table 9. ANOVA
PerformanceVariable
Sum of Squares df Mean Square F Sig.
NDA Between Groups 102,661 4 25,665 25,484 0
Within Groups 116,825 116 1,007 Total 219,486 120 P Between Groups 59,339 4 14,835 5,326 0,001 Within Groups 323,117 116 2,785 Total 382,457 120 IE Between Groups 25,55 4 6,387 4,621 0,002 Within Groups 160,342 116 1,382 Total 185,892 120
NDA=New drug application, P=Patents, IE=Innovation efficiency
Complementary Alliances and Firm Innovation Performance
32 types were conducted: intergroup interaction and extragroup interaction. Additionally, intragroup alliances were examined and therefore an independent t-test for intragroup alliances was conducted.
Intergroup Interaction. In the sample were 18 firms forming intergroup alliances and accordingly 103 firms that did not engage in intergroup interaction. The following results can be observed (see Table 10a and 10b): for new drug applications and patents, firms show a higher mean innovation performance compared to the firms that did not form intergroup alliances. For their innovation efficiency, on the other hand, counts exactly the opposite. Yet, only the results for the number of patents shows a high significant difference, with firms having intergroup alliances performing much better compared to firms that do not engage in interfirm collaboration.
Extragroup Interaction. Most firms in the sample engaged in (at least) extragroup interaction. In total, 65 out of the 121 firms engaged in at least one extragroup alliance. If firms formed extragroup alliances, the chance was higher that they have also formed other alliances (i.e. intergroup or intragroup). Results (see Table 11a and 11b) show similar patterns as for intergroup interaction. There is a significant difference in the means of innovation performance for the number of new drug applications (NDAs), number of patents (P) and innovation efficiency (IE). For the number of NDAs, firms that engage in extragroup interaction show a higher innovation performance compared to firms that do not engage in that type of alliances. For the number of patents, the results show that firms that form extragroup alliances also perform better compared to firms without alliances. Looking at the third performance variable, innovation efficiency, one can see the same result as for intergroup interaction: firms that do not engage in interaction achieve a significantly higher innovation performance compared to firms with extragroup alliances.
33
Table 10a. Intergroup: Independent Samples Test - Group Statistics
Intergroup Alliance N Mean S.D. S.E. Mean
NDAa yes 18 1,25 2,18 0,51 no 103 0,52 1,13 0,11 Pa yes 18 4,00 1,58 0,37 no 103 2,00 1,65 0,16 IEa yes 18 -1,21 1,48 0,35 no 103 -0,82 1,20 0,12
a indicates that for these variables the natural logarithm has been calculated and used,
NDA=New drug application, P=Patents, IE=Innovation efficiency
Table 10b. Intergroup: Independent Samples Test
Levene's Test for
Equality of Variances T-test for Equality of Means
F Sig. t df Sig. (2- tailed) M.D. Difference S.E. 95% Confidence Interval of the Difference
Lower Upper
NDAa Equal variances assumed 17,197 0 2,152 119 0,033 0,7325 0,34039 0,05849 1,40651
Equal variances not assumed 1,395 18,637 0,179 0,7325 0,52516 -0,36812 1,83311
Pa Equal variances assumed 0,39 0,534 4,768 119 0 2,00104 0,41965 1,17009 2,83199
Equal variances not assumed 4,935 24,028 0 2,00104 0,40551 1,16417 2,83791
IEa Equal variances assumed 3,078 0,082 -1,218 119 0,226 -0,38645 0,31733 -1,01478 0,24189
Equal variances not assumed -1,051 21,087 0,305 -0,38645 0,36773 -1,151 0,37811
a indicates that for these variables the natural logarithm has been calculated and used,
