Innovation performance in R&D alliances: The moderating role of Absorptive Capacity
Master Thesis | Master of Business Administration Thesis Proposal – Strategy Track - 6314M0304Y
Thesis supervisor | Nathan Betancourt Rosanne Noumon | 10156267 18 | 08 | 2017
Abstract
In the current study, we explore the moderating role of absorptive capacity in the relation between R&D alliances and incremental and radical innovation performance. Drawing from previous research, we propose that organizations with high knowledge acquisition, knowledge assimilation, and knowledge exploitation capabilities are better able to enhance their innovation performance when engaging in R&D alliances. Concretely, three indicators R&D intensity, knowledge distance, and the number of previously approved drugs are studied with regards to their moderating effect in the relation between R&D alliance frequency and incremental and radical innovation performance. We Securities Database Corporation Platinum (Thomson Reuters), Compustat and the Food and Drug Administration database to collect data on 147 firm-alliance observations. Our results indicated that R&D alliances were positively related to incremental and radical innovation performance. However, no support has been found for the moderating effects of R&D intensity, knowledge distance and the number of previous patents. Furthermore, the practical and theoretical implications of our findings are thoroughly discussed. Lastly, some limitations and suggestions for further research are proposed.
Index
1. Introduction……….. 4 2. Literature Review………...………. 7
2.1 The innovative effect of R&D alliances
2.2 The distinction between radical and incremental innovations 2.3 R&D alliances and the Absorptive Capacity perspective 2.4 R&D intensity and innovation performance in R&D alliances 2.5 Knowledge distance and innovation performance in R&D alliances 2.6 Previous patents and innovation performance in R&D alliances
3. Method……….……. 18
3.1 Sample
3.2 Data collection: Alliance data 3.3 Data collection: Patent Data
3.4 Measurements 3.4.1 Dependent Variables 3.4.2 Independent Variables 3.4.3 Moderating Variables 3.4.4 Control Variables 4. Results……….…………..…… 23 4.1 Descriptive statistics 4.2 Correlation Analysis 4.3 Regression Analysis 4.3.1 Model Explanation
4.3.2 Negative Binomial Regression analysis
4.3.2.1 R&D intensity and innovation performance in R&D alliances
4.3.2.2 Knowledge Distance and innovation performance in R&D alliances 4.3.2.3 The number of previous drug approvals and innovation
performance in R&D alliances
5. Discussion………..………... 33 6. References………..……..……. 37
1. Introduction
Innovation plays a vital role in the growth, long-term survival, and success of organizations (Schumpeter, 1942; Christensen, 1997). In the face of new and emerging technologies, shortening product life cycles and changing market pressures, companies are often forced to develop and exploit innovations in a more timely and cost-effective manner (Lin et al., 2012). Consequently, strategies aiming to enhance these innovative capabilities, are gaining popularity (Verona & Ravasi, 2003). The Research & Development (R&D) alliance between industry peers is one of the strategies that has provoked substantial academic and managerial attention (Estrada, Feams & Faria, 2014; Brandenburger & Nalebuff, 2011; Tsai, 2002; Bengtsson & Kock, 2000; Nalebuff Brandenburger & Maula, 1996). By providing access to external sources of knowledge and skills, the R&D alliance helps the participating firms to deepen and broaden their innovative capabilities (Sampson, 2007; Lin et al. 2012). However, there is no empirical consensus on how and to what extend R&D alliances are able to enhance innovation performance (Lin et al. 2012).
When explaining the variance of innovative performance in R&D alliances, previous studies have been directing towards absorptive capacity as one of the most important moderating factors (George & Zahra, 2001; Liao, Fei & Chen, 2007; Nooteboom et al. 2007; Lin et al. 2012). It is argued that firms are better able to reap the innovative benefits of their R&D alliances when they possess over a high-level absorptive capacity; allowing for efficient acquisition, assimilation, and exploitation of the externally obtained knowledge (Cohen & Levinthal, 1990). However, when looking at the moderating role of absorptive capacity in the relation between R&D alliances and innovation performance, the notion of an universalistic approach to innovation performance might be inappropriate (Downs & Mohr, 1976). Innovations vary widely in their degree of newness to the relevant unit of adoption (Dewar & Dutton, 1986). Therefore an important distinction has been made between radical and incremental innovations (Tushman & Anderson, 1986). Radical innovations entail extensive changes from existing products and technologies and constitute the basis for products and services that are entirely new to the market, while incremental innovations only incorporate minor changes and modifications to existing products or technologies are regarded to as incremental (Tushman & Anderson, 1986). Because, radical innovation requires a higher amount of ''new'' knowledge, compared to incremental innovations, we suggest that absorptive capacity has a different moderating effect on the relation between R&D alliances and both forms of innovation. Because data on radical and incremental
innovation is difficult to obtain, the majority of studies focusing on innovation in R&D alliances do not distinguishing between these two main forms innovations (Estrada, Feams & Faria, 2014; Brandenburger & Nalebuff, 2011; Tsai, 2002; Bengtsson & Kock, 2000; Nalebuff Brandenburger & Maula, 1996). Nevertheless, in the effort to gain further understanding of how and to what extend R&D alliances are able to enhance innovation performance, we aim to fill this theoretical gap.
Moreover, in the process of gaining further insights on how absorptive capacity enhances radical and incremental innovation performance in R&D alliances, we focus on a second (empirical) gap. Although absorptive capacity is broadly recognized to be a multidimensional construct, involving knowledge acquisition, knowledge assimilation, and knowledge exploitation (Zahra & George, 2002), previous studies on this topic often use unilateral measures to operationalize the construct (Cohen & Levinthal, 1990; Mowery, Oxley, & Silverman, 1996). Because of the intangible nature of the construct, the process of developing separate operationalization's for the three dimensions, in addition to gathering a sufficient amount of data on these measures has proven to be a challenging and time-consuming process (Zahra & George, 2002; Jansen et al., 2005; Lewin et al., 2011; Roberts et al., 2012). Hence, single measures like R&D spending (Cohen & Levinthal, 1990), and technological distance (Mowery, Oxley, & Silverman, 1996) have been previously used to operationalize the construct, even though they do not render distinct insights concerning the individual dimensions. However, firms might not reach the same capacity level for all three dimensions; the capacity level of one dimension does not guarantee the same capacity level for the other dimensions (Zahra & George, 2002). To illustrate, a firm might possess over highly developed acquisition capabilities that allow the firm to acquire external knowledge effectively. However, the same firm may not be as effective when it comes to exploiting this newly obtained knowledge. Furthermore, we argue that in order to fully understand the moderating role of absorptive capacity in the relation between R&D alliances and radical and incremental innovation performance, it is of great importance to utilize separate operationalizations for each of the three dimensions. Hence, we gain insight into the individual moderating effect of each of the dimensions.
The current study will assist in correcting the theoretical deficiency in the literature by making a clear distinction between the two forms of innovation (e.g. radical and incremental) when looking into the relation between R&D alliances and innovation performance. Simultaneously, we aim to address a second empirical gap from an academic research perspective, by studying the individual moderating effect of the three dimensions of
absorptive capacity; knowledge acquisition, knowledge assimilation, and knowledge exploitation. These insights will be of great relevance to managers that aspire to maximize their incremental and radical innovation performance by engaging in R&D alliances.
The remainder of the paper will consist out of four sections. Firstly, we'll provide a theoretical overview of the existing research in this field, focusing on R&D alliances, radical and incremental innovation performance and absorptive capacity. Furthermore, we'll propose a conceptual framework and hypothesis. Moreover, we present our methodological approach, results, and analysis. Subsequently, we will elaborate on the theoretical and managerial implications, while acknowledging the main limitations of our study. Finally, we will provide some suggestions for future research.
2. Literature Review
2.1 The innovative effect of R&D alliances
In the current study, we define innovation as 'a multi-stage process whereby organizations transform ideas into new or improved products, service or processes, in order to advance, compete and differentiate themselves successfully in their marketplace’ (Baregheh, Rowley & Sambrook, 2008). Historically, organizations aiming to enhance their innovation performance used to solely focus on their internal R&D activities (Mowey, 1983). However, when the innovative advantages of sole internal R&D expenditure were perceived to be declining, the locus of innovation shifted from the early Schumpeterian model, towards a broader view of collaborative alliances (Lin & Chen, 2007; Lin et al., 2012; Chesenbourgh, 2003). More frequently, organizations started to engage in alliances with industry peers to promote purposeful inflows of new knowledge, which helped to accelerate their internal innovation capability (e.g. Rigby & Zook, 2002; Christensen et al., 2005). As Chesenbourg (2003) acknowledged: "Not all of the smart people work for us, so we must find and tap into the knowledge and expertise of bright individuals outside of our company."
In previous studies, the knowledge-based perspective has been used to help understand why R&D alliances in specific might help to enhance innovation performance (Sampson, 2004). The knowledge-based perspective argues that knowledge is the firms single most valuable resources (DeCarolis & Deeds, 1999). According to this theory, superior knowledge allows firms to enjoy a sustained competitive advantage in general, and innovative benefits in specific. Based on this theory the development, integration, and transfer of knowledge should be regarded as one of the most critical aspects of organizational strategy making (Johanson & Vahlne, 2003). Hence, R&D alliances can be seen as a valuable tool to help to create new knowledge combinations that allow for enhanced performance (Das & Teng, 2000).
In accordance, various studies have confirmed that R&D alliances, in particular, enable organizations to enhance their innovation performance (Sampson, 2007; Hagedoorn, Kranenburg & Osborn, 2003; Brandenburger & Nalebuff, 2011; Tsai, 2002; Bengtsson & Kock, 2000; Nalebuff, Brandenburger & Maula, 1996). For example, Sampson (2007) demonstrated that R&D alliances contribute significantly to firm innovation. In addition, Lin and colleagues (2012) found strong support for the positive relation between inter-firm R&D alliances and innovation performance. Hagedoorn, Kranenburg, and Osborn (2003) also argued that inter-firm R&D partnerships increase organizational involvement in patenting.
2.2 R&D alliances and the Absorptive Capacity perspective
Although R&D alliances are often viewed as a great strategy to pursue innovation, previous studies suggest that an estimate of 50% of all alliances fail to live up to their expectations (Dyer, Kale, & Singh, 2001). Some researchers have argued that the variance in the ability of an R&D alliance to enhance innovation performance can be partially explained by the construct absorptive capacity (George & Zahra, 2001; Liao, Fei & Chen, 2007; Nooteboom et al. 2007; Lin et al. 2012). These studies have proposed that for an R&D alliance to generate innovative results, the focal firm needs to have a sufficient level of absorptive capacity. Absorptive capacity has been classically defined as ‘'the ability of a firm to recognize the value of new external information, assimilate this information and effectively apply it to commercial ends'' (Cohen & Levinthal, 1990). A higher level of absorptive capacity will enable an organization to acquire, assimilate and utilize external knowledge more effectively, which increases the chance of enhanced innovation performance in R&D alliances (Zahra & George, 2002). Some researchers even state that when the focal firm is not able to identify, assimilate, and apply the ''new'' external knowledge, they won't have the capability to leverage any innovative benefits from the R&D alliance by any means (Liao, Fei & Chen, 2012).
Previous studies have provided us with valuable insights on the moderation effect of absorptive capacity in the relation between R&D alliances and innovation performance (Ritala & Hurmelinna-Laukkanen, 2012; Tsai, 2001; Escribano, Fosfuri & Tribo, 2009). Ritala & Hurmelinna-Laukkanen (2012) found that absorptive capacity has a significant positive moderating effect on the relationship between R&D alliances and innovation performance. Higher levels of absorptive capacity were associated with an increased number of innovations created in R&D alliances. Additionally, the study by Tsai (2001) showed that when R&D business units hold a higher level of absorptive capacity, they are more likely to leverage increased innovation performance compared to R&D business units with a lower level of absorptive capacity. Furthermore, Escribano, Fosfuri, and Tribo (2006) revealed that firms with higher levels of absorptive capacity could manage external knowledge flows more efficiently, which stimulated their innovative outcomes.
Although absorptive capacity is broadly recognized to be a multidimensional construct, involving knowledge acquisition, knowledge assimilation, and knowledge exploitation (Zahra and George, 2002), previous studies on this topic often use unilateral measures to operationalize the construct (Cohen & Levinthal, 1990; Mowery, Oxley, & Silverman, 1996). Ritala and Hurmelinna-Laukkanen (2012) utilized a 4-item scale to
measure absorptive capacity. Although this measure proved to have a high reliability (r = .76), the items examined only the acquisition dimension of absorptive capacity and collected no information on the focal firm's ability to assimilate or exploit knowledge. Furthermore, Tsai (2001) measured absorptive capacity using R&D intensity (defined as R&D expenditure divided by sales). Although R&D intensity is widely used to operationalize absorptive capacity, it has been noted that the measure mainly provides insides into the firm's ability to acquire knowledge (Lane, Koka & Pathak, 2006). In their critical review, Lane Koka and Pathak (2006) emphasize that by solely focusing on R&D related measures two-thirds of Cohen and Levinthal's definition (1990) is ignored. They even state that this has led to ''an overemphasis on the importance of firms being able to acquire the external knowledge at the expense of the capability needed to assimilate en exploit this knowledge''.
The nature of the construct absorptive capacity is intangible, which makes the process of developing separate operationalization's for the three dimensions, in addition to the gathering of a sufficient amount of data on these measures, a challenging and time-consuming process (Zahra & George, 2002; Jansen et al., 2005; Lewin et al., 2011; Roberts et al., 2012). However, by using unilateral measures to operationalize the construct, it is not possible not render distinguishable insights concerning the individual dimensions. These insights might particularly be interesting since firms might not reach the same capacity level for all three dimensions; the capacity level of one dimension does not guarantee the same capacity level for the other dimensions (Zahra & George, 2002). To illustrate, a firm might possess over highly developed acquisition capabilities that allow the firm to recognize and acquire external knowledge efficiently. However, the same firm may not be as effective when it comes to exploiting this newly obtained knowledge. Furthermore, we argue that in order to fully understand the moderating role of absorptive capacity in the relation between R&D alliances innovation performance, it is of great importance to utilize separate operationalizations for knowledge acquisition, knowledge assimilation as well as for knowledge exploitation. Hence, we are able to gain insights into the individual moderating effects of each of the three dimensions of absorptive capacity.
2.3 The distinction between radical and incremental innovations
While a broad range of earlier research has demonstrated that R&D alliances positively enhance innovation performance, there is a lack of research examining the distinctive role of R&D alliances in the creation of both incremental and radical innovations (Sampson, 2007; Hagedoorn, Kranenburg & Osborn, 2003; Brandenburger & Nalebuff, 2011; Tsai, 2002;
Bengtsson & Kock, 2000; Nalebuff Brandenburger & Maula, 1996). In 1986, Dewar and Dutton criticized innovation research for adopting the assumption that an universalistic theory of the innovation process could be developed that applies to all types of innovations. Furthermore, their paper describes the notion that fundamental differences are found between different forms of innovations, which makes the thought of an universalistic approach somewhat inappropriate (Downs & Mohr, 1976). In one of their later studies, Downs, and Dutton (1986) further build on this idea by categorizing innovations based on their degree of newness to the relevant unit of adoption (Dewar & Dutton, 1986). Based on this view a distinction can be made between radical innovations, innovations that encompass large changes from existing products and technologies and constitute the basis for products and services that are completely new to the market, and incremental innovations, which concern minor changes and modifications to existing products or technologies (Tushman & Anderson, 1986).
Because of the inherent differences between both types of innovation, it is proposed that differences can be found between the relation between R&D alliances and incremental innovation performance and the relation between R&D alliances and radical innovation performance. This difference is believed to be present because there is a different amount of ''new'' knowledge required to generate incremental in comparison to radical innovations (Ritala & Hurmelinna-Laukkanen, 2012). Because incremental innovations concern minor changes and modifications to existing products or technologies, a relatively small amount of ''new'' knowledge is required (Dewar & Dutton, 1986). Therefore, it is expected that firms that engage more frequently with industry peers through R&D alliances are better able to create incremental innovations. The R&D alliance will provide the focal firm with ''new'' external knowledge that will assist them in realizing minor changes and modifications to their existing products or technologies.
Radical innovations require a relatively large amount of ''new'' knowledge because significant changes from the existing product need to be made in order to generate a products or technology that is completely new to the market (Dewar & Dutton, 1986). Hence, it is also expected that in the case of enhancement of radical innovation performance the R&D alliance will have a positive effect, by allowing the focal firm access to ''new'' external knowledge. However, the relation between R&D alliance frequency and radical innovation performance between industry peers is expected to be smaller for radical innovations as compared to incremental innovations. As explained by Ritala and Hurmelinna-Laukkanen (2012), the knowledge that is exchanged between the industry peers engaging in an R&D alliance is
typically quite similar, and might therefore not be as useful when creating radical innovations. The creation of radical innovations requires knowledge that is significantly different from the focal firm's existing knowledge base (Emden, Calantoe & Droge, 2006). Therefore it is expected that the relation between R&D alliance frequency and radical innovation performance is more positively related to incremental innovation performance, compared to radical innovation performance.
Concludingly, it is proposed that the more frequently organizations engage in R&D alliances the more it will positively enhance incremental and radical innovation performance. However, this moderating effect is expected to be stronger for incremental innovations than for radical innovations. In line with the above discussion, the following hypotheses are formulated:
H1 | R&D frequency is positively related to Incremental Innovation performance
H2 | R&D frequency is positively related to Incremental Innovation performance
H3 | R&D frequency is more positively related to incremental Innovation performance, compared to radical innovation performance
2.4 R&D intensity and innovation performance in R&D alliances
The first identified dimension of absorptive capacity is knowledge acquisition (Zahra & George, 2002). Knowledge acquisition is related to an organization's ability to locate, identify, value and acquire external intellectual assets that might provide the company with new competitive benefits (Lane & Lubatkin, 1998; Zahra & George, 2002). In the context of R&D alliances, knowledge acquisition capabilities help the focal firm to recognize and select potential R&D partners more accurately, by being better able to locate, identify, and evaluate whether they hold knowledge that is complementary to their current knowledge base. The ability of a firm to successfully acquire external knowledge has been positively related to innovation performance (Adams et al., 1998; Moorman & Rust, 1999; Chandrashekaran et al., 1999).
R&D intensity (defined as R&D expenditure divided by sales) has proven to be a strong determinant of acquisition capabilities (Cohen & Levinthal, 1990; Mowey, 1984). Research shows that firms with a high R&D intensity are often more successful when it comes to identifying and evaluating external knowledge (Cassiman & Veugelers, 2006). A
high R&D expenditure helps firms to develop a broader and deeper knowledge base, which provides them with a stronger ability to recognize and evaluate the potential value of the knowledge base of potential R&D partners (McGrath, 1999). Following this argument, it is proposed that the higher the R&D intensity of the focal firm the higher its knowledge acquisition capacity in R&D alliances.
Based on the insights of previous studies, we aim to explore the role of R&D intensity in the relation between R&D alliance frequency and innovation performance. Firstly, previous research has indicated that a high R&D intensity might help the focal firm in better selecting a suitable R&D alliance partner (Emden, Calantone, & Droge, 2006). It is acknowledged that, when selecting an appropriate collaboration partner, a significant amount of relevant previous knowledge is needed, which helps to assess the potential partner's technological capability, resource complementarity, and overlapping knowledge bases. Higher levels of R&D intensity have been previously related to firms having a broader and deeper knowledge base, which improves their ability to locate, identify, value and acquire a promising alliance partner (McGrath, 1999). Furthermore, we might expect that a high R&D intensity helps the focal firm in selecting a suitable R&D alliance partner, which increases the change of enhanced innovation performance in R&D alliance. Secondly, a similar reasoning applies to the focal firm's ability to acquire premium knowledge from the R&D alliance partner. It is shown that increased R&D intensities provide organizations with a deeper and broader knowledge base (McGrath, 1999). A well-developed existing knowledge base will enable firms to locate, identify, value and acquire relevant knowledge from their R&D partners more accurately (Cohen & Levinthal, 1990). In line with these arguments, we might expect that a high R&D intensity helps the focal firm to locate, identify, value and acquire superior knowledge, which increases the change of enhanced innovation performance in R&D alliance. Lastly, previous research has shown that a higher R&D intensity reduces inefficiencies associated with external knowledge acquisition (Veugelers, 1997; Arora & Gambardella, 1994; Rosenberg 1990). Firms with a low R&D intensity are shown to be less equipped when it comes to seeking and recognizing relevant external knowledge. Therefore the process of acquiring external knowledge is more time consuming (Arora & Gambardella, 1994; Rosenberg 1990). Additional research shows that memory development is self-reinforcing in that the more objects, patterns, and concepts that are stored in memory, the more readily new information about these constructs can be acquired (Bower & Hilgard, 1981). In line with these arguments, we suggest that a high R&D intensity will help to make
the knowledge acquisition process more efficient, which is expected to enhance innovation performance in R&D alliance.
Following these arguments, it is proposed that R&D intensity positively moderates the relation between R&D alliance frequency and incremental as well as the relation between R&D alliance frequency and radical innovation performance. However, because of the inherent differences between incremental and radical innovation (Dewar & Dutton, 1986), a divergent strength is expected for the moderating effect. Both forms of innovations require a different amount of ''new'' knowledge. Because incremental innovations concerns only minor changes and modifications to existing products or technologies, a relatively small amount of ''new'' knowledge is required (Dewar & Dutton, 1986). While the creation of radical innovations demands a relatively large amount of ''new'' knowledge because significant changes from the existing product need to be made in order to generate a products or technology that is completely new to the market (Dewar & Dutton, 1986). Based on the previous arguments we predict that a higher R&D provides the focal firm with better knowledge acquisition capabilities. However, the knowledge that is acquired might not be as useful in the creation of radical innovations when compared to incremental innovations. Although, high R&D intensity might help to locate, identify, value and acquire superior R&D partners and superior external knowledge that is often related to the existing knowledge base, while the creation of radical innovations requires knowledge and skills that are completely new to the organization (Crossan & Inkpen, 1995; McGrath, 2001).
Therefore, the role of R&D intensity in the creation of innovations in R&D alliances is expected to be smaller for radical innovations as compared to incremental innovations.
H4 | R&D Intensity has a positive moderating effect on the creation of incremental innovations in R&D alliances.
H5 | R&D Intensity has a positive moderating effect on the creation of radical innovations in R&D alliances.
H6 | The moderating effect of R&D intensity is stronger in the creation of incremental than of radical innovations in R&D alliances.
2.5 Knowledge distance and innovation performance in R&D alliances
The second dimension of absorptive capacity that is identified by Zahra and George (2002) is knowledge assimilation. Knowledge assimilation is related to organizational routines and processes that allow the organization to analyze, process, interpret and understand information obtained from external sources (Zahra & George, 2002). Previous research has regarded knowledge assimilation as a critical factor in the creation of innovations in alliances (O'Leary, 1999; Tsai, 2001; Phene & Almeida, 2008). Phene and Almeida (2008) found that host country firms that were capable of effective knowledge assimilation experienced a higher quality and scale of innovation. Additionally, Tsai (2001) showed that organizational units produced more innovations when they got access to new knowledge developed by another unit. However, this was largely dependent on the focal unit's knowledge assimilation capabilities.
Previous studies show that the ability to assimilate new knowledge is mainly dependent on a firm’s knowledge distance, which is a measure of the ease of transition from one knowledge system to another (Liyanage & Benard, 2003; Cohen & Levinthal, 1990; Mowery et al. 1998; Lane & Lubatkin, 1998). Zahra & George (1995) argued that a smaller knowledge distance makes it easier to comprehend, process and internalize new knowledge which may result in better performance. Based on the insights of previous research, we aim to explore the role of knowledge distance in the relation between R&D alliance frequency and innovation performance. Lane and Lubatkin (1998) showed that firms with a similar knowledge base, are more likely to form an alliance and leverage increased performance as a result. Additionally, it has been shown that in alliances where the new external knowledge is embodied by heuristics dissimilar from the ones used by the focal firm, firms are often confronted with a delay in the interpreting and understanding of the information, and therefore making the assimilation process less effective (Leonard-Barton, 1995).
However, there are some studies that suggest that a small knowledge distance might not be as beneficial to the innovation process as proposed in the studies mentioned above. When the knowledge distance is too small, there might not be a sufficient amount of new knowledge to be able to enhance the innovative capabilities of either of the firms involved (Lin, et al. 2012). Empirical studies have shown that organizations are less likely to reap performance benefits from an alliance when there is high knowledge similarity between the firms (Letterie et al., 2008; Cassiman et al., 2005). Nevertheless, a third body of research can be found that aim to combine these seemingly contradicting viewpoints. Nootebook (1999) and Ahuja and Katila (2001) argue that the moderating effect of knowledge distance in the
relation between collaborations and innovation takes the form of an inverted U-shape. They indicate that the innovative benefits as a result of collaboration are highest when the knowledge distance between the participating firms is moderate. In this situation a firm is as still able to effectively analyze, process, interpret and understand the information obtained from the external source, because of the partial knowledge overlap. However, there is a sufficient amount of new knowledge present that helps to enhance innovation capabilities.
Based on this knowledge we propose that knowledge distance moderates the effect of R&D alliance frequency on innovation performance in an inverted U-shaped way. However, we expect a different moderating effect for incremental innovations as supposed to radical innovations. Research shows that radical innovations represent revolutionary changes in existing technology (Duchesneau, Cohn, & Dutton, 1979), while incremental changes only require minor adjustments in current technology (Munson & Pelz, 1979). The newness of an innovation largely depends on the degree of novel knowledge that gets embedded (Dewar & Dutton, 1986). Therefore it is suggested that a large knowledge distance is needed to spark radical innovation compared to incremental innovation. Furthermore, the u-shaped moderating effect of knowledge distance is expected to be larger in incremental innovations compared to radical innovations.
H7 | Knowledge Distance has an inverted-U curvilinear moderating effect on the creation of incremental innovations in R&D alliances
H8 | Knowledge Distance has an inverted-U curvilinear moderating effect on the creation of radical innovations in R&D alliances
H9 | The inverted-U curvilinear moderating effect of knowledge distance is stronger in the creation of incremental than of radical innovations in R&D alliances.
2.6 Previous patents and innovation performance in R&D alliances
The third and final dimension of absorptive capacity is knowledge exploitation (Zahra & George, 2002). Knowledge exploitation is defined as an 'organizational capability that is based on the routines that allow firms to refine, extend, and leverage existing competencies or to create new ones by incorporating acquired and transformed knowledge into its operations.' A previous by Spencer (1996) has shown that knowledge exploitation is largely associated with the creation novel goods, systems, processes, knowledge or organizational forms. Additionally, Rotherarmel and Deeds (2004) demonstrate that a focus on knowledge exploitation in alliances allow firms to bring a larger amount of product on the market.
Previous studies have indicated that there is a great learning component to an organization's ability to exploit knowledge. Previous experience in knowledge exploitation helps organizations to develop routines, which allow them to be more efficient and effective when it comes to introducing new products and services to the market (Lavie & Rosenkopf, 2006; Levinthal & March, 1993; Lenart, 2015). Consequently, Lenart (2015) argues that the primary predictor of knowledge exploitation is the quantity and frequency of services or products that the organization has previously introduced to the market. Similarly, Jean-Pierre Noblet and colleagues (2011) had demonstrated that the number of patents or products that a firm previously launched functioned helped to predict their exploitative capabilities. Based on this reasoning, we propose that in the case of pharmaceutical companies a higher number previous drug approvals positively moderates the relation between R&D alliance frequency and innovation performance.
Since the process of patenting radical innovations is proven to be more difficult and time-consuming, we expect a different moderating effect for incremental innovations as supposed to radical innovations. Research shows that the process of patenting radical innovations is often more risky, costly and time-consuming because of more extensive regulations. The process of bringing radical innovations to the market is expected to be more difficult and less routinized than bringing incremental innovations to the market (Stewart, 2010). Therefore, we expect that the learning benefits associated with the number of previous patents will be less beneficial in the creation of radical innovations when compared to incremental innovations. Therefore, it is expected that the moderating effect of the number of previous patents is stronger in incremental innovations compared to radical innovations.
Furthermore, we expect that a firm's ability to successfully exploit external knowledge has been positively linked to innovation performance in R&D alliances. Since the process of bringing radical innovations to the market is expected to be more difficult and less
routinized than bringing incremental innovations to the market. Therefore, it is expected that the number of previous patents positively moderates the relation between R&D alliance frequency and total, incremental and radical innovation performance (Stewart, 2010). Furthermore this effect is expected to be larger in incremental innovations compared to radical innovations.
H10 | The number of previous patents has a positive moderating effect on the creation of incremental innovations in R&D alliances.
H11 | The number of previous patents has a positive effect moderating on the creation of radical innovations in R&D alliances.
H12 | The moderating effect of the number of previous patents is larger for the creation of incremental innovations in R&D alliances than for radical innovations in R&D alliances.
3. Method
The current chapter describes the empirical part of this study. Firstly, an overview is provided of our research sample. Furthermore, the process of data collection is described for each the independent, dependent, moderating and controlling variables.
3.1 Sample
Firstly, we identified all US-based organizations operating within the pharmaceutical industry as of 1995. These firms were all categorized within SIC class 2834: Pharmaceutical Preparations. All companies established their primary engagement in manufacturing, fabricating, or processing drugs in pharmaceutical preparations which are targeted to dental, medical, veterinary professions as well as to the public (United States Department Of Labor, 2017). Furthermore, from this dataset, we've obtained all pharmaceutical organizations that engaged in at least one dyadic R&D alliances between 1985 and 2014. Our final data set incorporated a sample size of 89 firms and 147 dyadic R&D alliances.
The pharmaceutical industry seems to be appropriate for the current study for several reasons. Firstly, this industry is primarily driven by the generation of knowledge (Bierly & Chakrabarti, 1996). Bierly and Vhakrabarti (1996) explain that a pharmaceutical firm's ability to benefit from external sources of knowledge is a key determinant of its competitiveness. Secondly, the pharmaceutical industry is often identified as a high-velocity industry (Chi et al. 2010). Consequently, firms operating in this industry are known to be frequently looking for external knowledge through the formation of alliances. The U.S. based pharmaceutical industry, in particular, is well known for its high activity in alliances formation with the goal of enhancing innovation performance (Whittaker & Bower, 1994). Thirdly, data on organizational innovation performance is fairly easy to obtain by using the number of patents the organization in question has issued (Bierly & Chakrabarti, 1996). Furthermore, previous studies have shown that patents issued in the pharmaceutical industry are more accurately represented in online databases, compared to other industries, since accurate patenting in the pharmaceutical industry gets heavily enforced by the FDA (Bierly & Chakrabarti, 1996).
3.1.1 Data collection: Alliance data
The primary database used to retrieve data on R&D alliances is the Securities Database Corporation (SDC) Platinum from Thomson Reuters. This database is known to be a well-established and comprehensive database for retrieving detailed data on alliances and covers data on all kinds of alliances from 1990 onwards (Sampson, 2004). We focus on
dyadic inter-firm alliances since alliances in the pharmaceutical industry are almost exclusively between two individual organizations (Lin et al., 2012). Through the data presented in this database we are well able to distinguish between various alliance functions (e.g., Research & Development, Sales & Marketing, Supply & Manufacturing, Licensing & Distribution). Hence, it is possible to gain specific data on dyadic alliances that were formed for Research & Development purposes. Additionally, information on the firm age, firm size and R&D intensity of the focal firm was collected through Datastream. This database provides financial, statistical and market information and covers 99% of the world's total market capitalization with annual data.
3.1.2 Data collection: Drug Approval Data
In addition to the SDC Platinum database, we use the Food and Drug Administration database to collect data on all drug approval files from the 89 companies. The FDA database allows us to distinguish between radical and incremental innovations because they classify all drug approvals into one of ten categories (e.g., 1. New molecular entity, 2. New active ingredient, 3. New dosage form, 4. New combination, 5. New formulation or other differences, 6. New indication or claim, same applicant (no longer used), 8. Previously marketed but without an approved NDA, 9. Rx to OTC, 10. New indication or claim). Only the first category, new molecular entity, describe drug approvals that encompass at least one molecular component that can be seen as extremely novel to the market. Radical innovations incorporate large changes from existing products and technologies and constitute the basis for products and services that are completely new to the market (Tushman & Anderson, 1986). Hence, drug approvals, which are registered within category one, can be seen as radical innovations. The other nine categories concern minor changes and modifications to existing products or technologies and can, therefore, be regarded as incremental innovations. This reasoning is in line with previous research on incremental and radical innovation in the pharmaceutical industry (Kim & Song, 2007; Lin et al. 2012).
3.2 Measurements
3.2.1 Dependent Variables
Incremental Innovation Performance (IIP).
Building on the previous arguments Incremental Innovative Performance will be operationalized using the total number of category 2-10 drugs that the focal firm got approved by the FDA within 7 years after the start of the alliance. These categories include; New active ingredient, New dosage form, New combination, New formulation or other differences, New indication or claim, same applicant (no longer used), Previously marketed but without an approved NDA, Rx to OTC, and New indication or claim. Furthermore, in line with previous studies, we use a time span of eight years after the start of the alliance for the following reasons (Lavie & Miller, 2008; Lin et al. 2012). This time span aims to include all drug approvals that were granted during the period that the R&D alliance was active, in addition to the period after the alliance where the obtained knowledge might still have affected the drug development. Alliance termination dates are known to be frequently missing in databases because alliance termination is rarely publically announced (Lavie & Miller, 2008). Furthermore, based on previous studies, we assume that the average time span of an R&D alliance is three years (Lavie & Miller, 2008; Lin et al. 2012). Moreover, Argote (1999) indicates that knowledge in the knowledge intensive industries loses significant value within approximately five years. Therefore, we include all drug approvals within a time span of eight years after the start of the alliance.
A high number of category 2-10 drug approvals indicates a higher incremental innovation performance, since the focal firm was able to bring more incremental innovations to the market.
Radical Innovation Performance (RIP).
Furthermore, Radical Innovation Performance will be operationalized using the total number of category 1 drugs that the focal firm got approved by the FDA within 7 years after the start of the alliance. This category includes drug approvals that contain a new molecular entity. In coherence with the way incremental innovation performance is measured we count the number of drug approvals within a time span of eight years after the start of the alliance (Lavie & Miller, 2008; Lin et al. 2012). This time span aims to include all drug approvals that might have been influenced by the knowledge exchanged in the R&D alliance. This includes the period that the R&D alliance was active, in addition to the period after the alliance in which the obtained knowledge might still be relevant. A high number of category 1 drug
approvals indicates a higher radical innovation performance since the focal firm was able to bring more radical innovations to the market.
3.2.2. Independent Variables R&D Alliance Frequency (RDF)
R&D Alliance frequency is operationalized using the number of R&D alliances that the focal firm has engaged in during a 5-year period before the start of the R&D alliance in question.
3.2.3. Moderating Variables R&D intensity (RDI)
In line with previous research (Levitas & McFayden, 2009; Miller, 2006; Chen & Miller, 2007) R&D intensity was operationalized by dividing the total R&D expenditure for the year of the start of the alliance by the firm's total revenue in that same year. This measure reflects firm-specific research and development investments adjusted for firm size.
Knowledge distance (KD)
Knowledge distance, sometimes referred to as technological distance, is the degree of knowledge and information diversity between two organizations. In line with previous studies (Samson, 2007; Lin et al., 2012) this concept is operationalized by the extent to which the collaborating firms have innovations in the same categories. In our study, this measure is used to captures how dispersed the drug approvals of both participating firms are across FDA categories. The measure varies from 0 to 1 where a 0 indicates that there is no difference in the knowledge categories of the firms. Furthermore, a score of 1 represents the largest knowledge distance possible, which indicates that the firms don't have any drug approvals in the same FDA categories. This measure by Samson (2007) is constructed by creating multidimensional vectors for the focal firm (Fi) and for the alliance partner (Fj). Fi = (Fi...Fi), where Fi signifies the number of drug approvals approved by firm i in FDA category s. Furthermore, both knowledge portfolio's are compared using the formula presented below.
KD = 1 - 𝐹𝑖𝐹′𝑗 √(𝐹𝑖𝐹′𝑖) (𝐹𝑗 𝐹′𝑗)
Number of previous drug approvals (NPDA)
The number of previous drug approvals is measured by adding all drug approvals in FDA category 1 -10 that are assigned to the focal firm, starting from the founding date of the company until the date the alliance was announced. This measure is in line with measures used in previous studies.
3.2.4. Control Variables
When assessing the moderating effect of absorptive capacity on innovation performance in R&D alliances, it is important to control for firm age, firm size as well as the number of previous R&D alliances. We detail each of the control variables below.
When studying the moderating effect of absorptive capacity on the relation between R&D alliance frequency and innovation performance we have to consider the effect of firm age. Previous studies have suggested that the probability of innovation changes with firm age (Huergo & Jaumandreu, 2004). Entrant firms, tend to present a significantly higher amount of innovations compared to more mature firms. Therefore, we control for firm age in the current study.
There has been a lot of debate on the relation between firm size and (innovation) performance (Ahuja, Lampert & Tandon, 2008; Cohen, 1995). Previous research has suggested that firms with a smaller size are often less likely to bring innovations to the market than firms with a larger size (Huergo & Jaumandreu, 2004; Chaney et al., 1991). Additionally, the number of new products by unit of sales and proportion of innovative sales are inversely related to firm age (Hansen, 1992. Therefore, we argue that it is important to include firm size as a control variable. In line with previous studies, the variable firm size will be measured through the number of employees the focal firm possesses at the start of the alliance.
4. Results
We will start by discussing the descriptive statistics and correlation analysis. Subsequently, we will present the results from the Negative Binominal Regression analysis. The direct relationships between the number of R&D alliances in the past five years and innovation performance are discussed in advance to the moderation effects respectively.
4.1 Descriptive statistics
The descriptive analysis started with the investigation of the normal distribution of the interval/ratio variables. The regression analysis does not necessarily demand a normally distributed variable, but the regression analysis is sensitive to outliers or influence points (Stevens, 1984). These outliers may skew the distribution of the variable and may distort the interferences that are made based on the data. In addition, the linearity of the relationship, homoscedastic and normal distribution of the residuals may be improved when all interval/ratio variables in the model are normally distributed. Cohen, Cohen, West, and Aiken (2003) argue that when the relationship between the independent and dependent variable is made linear, the homoscedasticity and normal distribution assumptions of the residuals in the regression analysis are usually also satisfied. To investigate the interval/ratio variables on their normal distribution this study used the following measures. We assessed the degree to which the variables were normally distributed via both numeric and visual means. Based on the Shapiro-Wilk test statistic we've got an impression of the degree of normality. However, it needs to be noted that the Shapiro-Wilk test is sensitive to small deviations from the normal distribution, even in large datasets. In addition, the Shapiro-Wilk test does not inform on why the variable is not normally distributed. Therefore we've used a combination of the Shapiro-Wilk test and the following visual normality tests, Histograms, Normal Q-Q Plots and boxplots, to estimate the degree of normal distribution. It was found that there was a lack of symmetry (skewness) and pointiness (kurtosis) for the variables RIV, RDI, and KD. Hence, these variables were transformed using the natural log (ln).
The descriptive statistics including the mean, minimum, maximum, standard deviations, and bivariate correlations can be found in Table 1. Between 1995 and 2006 our sample contained 147 separate R&D alliances between 89 pharmaceutical firms. The age of the focal firm’s in our sample ranged from 1 to 118 years (M = 46.18, SD = 43.96), and was non-normally distributed with a skewness of .55 (SE = .20) and a kurtosis of -1.40 (SE = 0.40). This indicates that the maximum frequencies for firm age could be found at the two extremes of the range of the variate, meaning that our sample included mostly firms that were
relatively young and mature. Noteworthy is to acknowledge that the mean of incremental innovation performance (M = 3317.31) is higher than the mean of radical innovation performance (M = 1112.15), which confirms the notion that incremental innovation is more frequent than radical innovation (Dosi, 1988). Additionally, we’ve included figure 1, which shows the frequency of alliances observations for our sample.
4.2 Correlation Analysis
When looking at the Pearson correlation matrix (Table 1) we can observe the relations between de variables in this study. It appears that R&D alliance frequency is positively related to incremental innovation performance (r = .73, p < 0.01) and radical innovation performance (r = .56, p < 0.01). This correlation is stronger for incremental innovations, compared to radical innovations. Additionally a negative correlation is, found between R&D intensity and both incremental innovation performance (r = -.63, p < 0.01) and radical innovation performance (r = -.62, p < 0.05). Furthermore, we have to acknowledge that the correlation between Firm Size and Firm Age and the different dependent variables is quite high. However, when testing for multicollinearity, the Variation Inflation Factor (VIF) scores indicated there was no multicollinearity. The results of this test are further discussed in the following section.
4.3 Regression Analysis
This study used an OLS regression to investigate possible multicollinearity and serial-correlation concerns. We've studied the Durbin-Watson statistics to test for autoserial-correlation in the residuals, and the Variation Inflation Factor (VIF) scores to test for multicollinearity. The Durbin-Watson had a score of 1.99, which suggests that there is no autocorrelation present. The highest individual Variance Inflation Factor (VIF) score among all the variables was 4.62, with a mean VIF score of 1.76. Since, a score of 10 has been used as a rule of thumb to indicate excessive or serious multicollinearity (Luo & Deng, 2009), we've assessed that there was no reason to assume that there was a multicollinearity problem in the model.
This study performed a Negative Binominal regression analysis due to the nature and distribution of the dependent variable. The dependent variables counted the number of incremental (IIP) and radical innovations (RIP), which suggests that count-variables were used. The dependent variable IIP had a mean of 4712.20 and a variance of 17880486.90. Additionally, the dependent variable RIP had a mean of 1183.07 and a variance of 1970948.37, which suggests that there is overdispersion in the data. Therefore, a Negative
Binominal regression is more appropriate than a Poisson regression.
Furthermore, based on the regression we can observe that the moderators have to be tested in separate models to be able to observe their individual effects. In this study two different models are estimated. The first model investigates the explanatory variables of incremental innovation performance, and the second model investigates the explanatory variables of radical innovation performance. The results of the first Negative Binomial Regression analysis are presented in Table 2. The results of the second Negative Binomial Regression analysis are presented in Table 3. Furthermore, we provide explanatory insights for the total innovation performance of the organizations in our study.
4.3.1 Model Explanation
The results of the first Negative Binomial Regression analysis are presented in Table 2, and provide explanatory insights for the incremental innovation performance of the organizations in our study. Model 1 is the baseline model and includes only the independent variable of R&D alliance frequency (Δχ2=69.64). Model 2 adds R&D intensity, and Model 3 also incorporates the interaction terms of R&D alliance frequency and R&D intensity. The likelihood ratio test statistics indicate that by adding these variables in Model 2 (Δχ2=148.85) and Model 3 (Δχ2=151.30) the overall model-fit significantly increases as compared with Model 1. Furthermore, Model 4 incorporates knowledge distance and Model 5 also incorporates the interaction terms of the number of R&D frequency and knowledge distance. Based on the likelihood ratio test statistics, the addition of these variables in Model 4 (Δχ2=72.64) and Model 5 (Δχ2=77.39) significantly increases the overall model fit as compared with Model 1, however to a lesser extent compared to Model 2 and Model 3. Furthermore, Model 6 incorporates the number of previous drug approvals and Model 7 also incorporates the interaction terms of R&D frequency and the number of previous drug approvals. Based on the likelihood ratio test statistics, the addition of these variables in Model 6 (Δχ2=115.15) and Model 7 (Δχ2=146.17) significantly increases the overall model fit as compared with Model 1. Model 8 is considered to be the full model and includes the independent variable, the three moderating variables, as well as their interaction terms.
The results of the second Negative Binomial Regression analysis are presented in Table 3, and provide explanatory insights for the radical innovation performance of the organizations in our study. Model 9 is the baseline model and includes only the independent variable of R&D alliances frequency (Δχ2=69.64). Model 10 adds R&D intensity, and Model 11 also incorporates the interaction terms of R&D alliances frequency and R&D intensity.
The likelihood ratio test statistics indicate that by adding these variables in Model 10 (Δχ2=148.85) and Model 11 (Δχ2=151.30) the overall model-fit significantly increases as compared with Model 9. Model 12 incorporates knowledge distance and the squared term of knowledge distance and Model 13 also incorporates the interaction terms of R&D alliances frequency and knowledge distance and the interaction terms of R&D alliance frequency and the squared term of knowledge distance. Based on the likelihood ratio test statistics, the addition of these variables in Model 12 (Δχ2=72.64) and Model 13 (Δχ2=77.39) significantly increases the overall model fit as compared with Model 9, however to a lesser extent compared to Model 10 and Model 11. Furthermore, Model 14 incorporates the number of previous drug approvals and Model 15 adds the interaction terms of R&D alliance frequency and the number of previous drug approvals. Based on the likelihood ratio test statistics, the addition of these variables in Model 14 (Δχ2=115.15) and Model 15 (Δχ2=146.17) significantly increases the overall model fit as compared with Model 9. Model 16 includes the independent variable, the three moderating variables, and their interaction terms, and is therefore considered to be the full model.
4.3.2 Negative Binomial Regression analysis
In model 1 the R&D alliance frequency had a significantly positive effect on incremental innovation performance (β = 0.08, p < 0.01), which supported our first hypothesis. Hypothesis 1 claims that R&D frequency is positively related to incremental innovation performance. In model 9, R&D alliance frequency also had a significantly positive effect on radical innovation performance (β = 0.08, p < 0.01), which provided support for Hypothesis 2, which claims that R&D frequency is positively related to radical innovation performance. The β coefficient in the incremental innovation model (Model 1) shows to be a little bit stronger than the β coefficient of R&D alliance frequency in model 9. Therefore support for Hypothesis 3 had been found, which claimed that R&D frequency is more positively related to incremental innovation performance, compared to radical innovation performance.
4.3.2.1 R&D intensity and innovation performance in R&D alliances
Hypothesis 4 predicts that a firm's R&D intensity has a positive moderating effect on the creation of incremental innovations in R&D alliances. To verify this hypothesis, we first tested the direct effect of R&D intensity on incremental innovation performance by using Model 2. The model showed a significant negative coefficient of R&D intensity (β = -0.47, p
< 0.01), indicating that there is a negative relationship between R&D intensity and incremental innovation performance. Model 3 introduces the interaction term of R&D alliance frequency and R&D intensity to examine the moderating effect of R&D intensity on the relationship between R&D alliances frequency and incremental innovation performance. Model 3 estimated that the coefficient of the interaction term was negative, however, this coefficient was not statistically significant (β = - 0.02, p = 0.11). Therefore, Hypothesis 4 is rejected.
Furthermore, Hypothesis 5 claims that a firm's R&D intensity has a positive moderating effect on the creation of radical innovation performance in R&D alliances. To verify this hypothesis, we first tested the direct effect of R&D intensity on radical innovation performance by using Model 10. The model showed a significant negative coefficient of R&D intensity (β = -0.46, p < 0.01), indicating that there is a negative relationship between R&D intensity and radical innovation performance. Model 11 introduces the interaction term of R&D alliance frequency and R&D intensity to examine the moderating effect of R&D intensity on the relationship between R&D alliances frequency and radical innovation performance. Model 11 estimated that the coefficient of the interaction term is negative, however, this interaction term was not significant (β = - 0.01, p = 0.29). Therefore, Hypothesis 5 is rejected.
Hypothesis 6 predicts that the moderating effect of R&D intensity is stronger in the creation of incremental than in the creation of radical innovations in R&D alliances. The β coefficient in the incremental innovation model (Model 2) shows to be a little bit stronger than the β coefficient of R&D alliance frequency (Model 9), however both coefficients were not statistically significant. Therefore no support is found for Hypothesis 6.
4.3.2.2 Knowledge Distance and innovation performance in R&D alliances
Hypothesis 7 indicates that knowledge distance has an inverted-U curvilinear moderating effect on the creation of incremental innovation performance in R&D. To confirm this hypothesis, we first tested the main effect of knowledge distance on incremental innovation performance by using Model 4. The negative coefficient of the squared term indicates that there is an inverted-U relationship between knowledge distance and incremental innovation performance (β = - 0.11, p < 0.05). Model 5 further introduces the interaction terms of R&D alliance frequency and knowledge distance to examine the moderating effect of knowledge distance on the relationship between R&D alliance frequency and incremental innovation performance. In Model 5, the estimated coefficient of
the interaction term is negative, however not significant (β = - 0.21, p = 0.13). Therefore, no support has been found for Hypothesis 7.
Hypothesis 8 indicates that knowledge distance has an inverted-U curvilinear moderating effect on the creation of radical innovation performance in R&D. To confirm this hypothesis, we first tested the main effect of knowledge distance on incremental innovation performance by using Model 12. The negative coefficient of the squared term indicates that there is an inverted-U relationship between knowledge distance and incremental innovation performance (β = - 0.09, p = 0.05). Model 13 further introduces the interaction terms of R&D alliance frequency and knowledge distance to examine the moderating effect of knowledge distance on the relationship between R&D alliance frequency and incremental innovation performance. In Model 13, the estimated coefficient of the interaction term is negative, however again not significant (β = - 0.11, p = 0.20). Therefore Hypothesis 8 was not supported.
Hypothesis 9 predicts that the moderating effect of knowledge distance is stronger in the creation of incremental than in the creation of radical innovations in R&D alliances. The β coefficient in the incremental innovation model (Model 5) shows to be a little bit stronger than the β coefficient of R&D alliance frequency (Model 13). However, both coefficients were not statistically significant. Therefore no support is found for Hypothesis 9.
4.3.2.3 The number of previous drug approvals and innovation performance in R&D alliances
Hypothesis 10 predicts that the number of previous drug approvals has a positive moderating effect on the creation of incremental innovations in R&D alliances. To verify this hypothesis, we first tested the direct effect of the number of previous drug approvals on incremental innovation performance by using model 6. The model showed a significant positive coefficient of the number of previous drug approvals (β = 1.80*10-4, p < 0.01), indicating that there is a positive relationship between the number of previous drug approvals and incremental innovation performance. Model 7 introduces the interaction term of R&D alliance frequency and the number of previous drug approvals to examine the moderating effect of the number of previous drug approvals on the relationship between R&D alliances frequency and incremental innovation performance. Model 7 estimated that the coefficient of the interaction term that was significantly negative (β = -2.07*10-5, p < 0.01). Therefore, Hypothesis 10 is rejected.
Furthermore, Hypothesis 11 claims that the number of previous drug approvals has a positive moderating effect on the creation of radical innovation performance in R&D alliances. To verify this hypothesis, we first tested the direct effect of the number of previous drug approvals on radical innovation performance by using Model 10. The model showed a significant positive coefficient of the number of previous drug approvals (β = 1.63*10-4, p < 0.01), indicating that there is a positive relationship between the number of previous drug approvals and radical innovation performance. Model 11 introduces the interaction term of R&D alliance frequency and the number of previous drug approvals to examine the moderating effect of the number of previous drug approvals on the relationship between R&D alliances frequency and radical innovation performance. Model 11 estimated that the coefficient of the interaction term is significant and negative (β = -1.89*10-5, p < 0.01). Therefore, Hypothesis 11 is rejected.
Hypothesis 12 predicts that the moderating effect of the number of previous drug approvals is stronger in the creation of incremental than in the creation of radical innovations in R&D alliances. The β coefficient in the incremental innovation model (Model 7) shows to be a little bit stronger than the β coefficient of R&D alliance frequency (Model 11). Therefore, support is received for Hypothesis 12.
Table 1
Descriptive information and Pearson correlation
Variables Mean SD Minimum Maximum n 1. 2. 3. 4. 5. 6. 7. 8. 9.
1. R&D alliance frequency 11.90 9.77 1 36 147 1.00
2. Incremental Innovation Performance 3317.31 3309.41 2 11686 147 .73** 1.00 3.Radical Innovation Performance 1112.15 1388.22 1 6298 147 .56** .68** 1.00 4. R&D Intensity 2.81 10.50 0.03 100.57 147 -.44** -.62* -.58** 1.00 5. Knowledge Distance 0.18 0.16 0.01 0.71 147 .06 .10 .02 -.17* 1.00 6. Knowledge Distance 2 0.06 0.10 0.01 0.50 147 .08 .10 .10 -.14 .69** 1.00
7. Previous Drug Approvals 3114.78 4078.46 2 15687 147 .53** .72** .60** -.48* .14 .16* 1.00
8. Firm Size 23.28 27.77 0.01 71.82 147 .72** .90** .67** -.59** .13 .14 .79** 1.00
9. Firm Age 46.18 43.96 1 118 147 .69** .79** .64** -.57** .05 .09 .75** .82** 1.00
Note. n = 147 observations ⁎⁎⁎ p < .001 ⁎⁎ p < .01 ⁎⁎⁎ p < .05
