Complementarities of Internal R&D and Alliances with Different Partner Types in the Bio-Pharma Industry

1

University of Groningen

Faculty of Economics & Business

MASTER THESIS

Master in Business Administration, Strategic Innovation Management

Complementarities of Internal R&D and Alliances with

Different Partner Types in the Bio-Pharma Industry

Gabriel Marin

S3459136

g.d.marin@student.rug.nl

Supervisor: A.A. Oleksiak, Msc.

Co-assessor: Dr. F. Noseleit

2

Abstract

Nowadays many firms invest internally in R&D for an increased innovation performance. Furthermore within many innovation projects companies also seek external knowledge. Our paper represents an empirical study that studies the influence of internal R&D towards innovation together with the interaction effect between intra-firm and inter-firm efforts i.e. alliances. We replicate the study of Noseleit and de Faria (2013) as a focal paper, by looking at the bio-pharmaceutical industry since this is knowledge intensive and driven towards innovation. Surprisingly we do not find support for internal R&D investment leading to increased innovation output. Our findings also do not support the theory provided by the focal paper i.e intra-industry and related-industry partners enhance the efficiency of R&D efforts towards innovation or unrelated-industry partner having a negative moderation effect on intra-firm efforts. However as a contribution to innovation and alliance literature our study represents a foundation for further research that can include further replication studies following our research and also case studies.

3

Contents

Introduction ... 4

Theory and Hypotheses ... 7

Absorptive Capacity ... 7

Internal R&D Investment Leading to Innovation ... 7

The Moderating Effect of Intra-Industry Alliances ... 9

The Moderating Effect of Related-Industry Alliances ... 12

The Moderating Effect of Unrelated-Industry Alliances ... 13

Methodology ... 16 Data Collection ... 17 Measures ... 19 Data Analysis ... 21 Results ... 24 Descriptive Statistics ... 24 Regression Results ... 28 Discussion... 30 Theoretical Implications ... 30 Managerial Implications... 32

Limitations and Future Research ... 33

Conclusion ... 35

References ... 38

4

Introduction

objective of this study is to analyse the effect of various types of alliance partners on R&D efficiency towards innovation performance. We perform the replication study in the bio-pharmaceutical industry. The interaction between intra-firm efforts (R&D efficiency) and inter-firm collaborations (alliances) represents the key aspect of this paper as introduced below.

5

et al. 2005; Hagedoorn 2002). Consequently we have chosen absorptive capacity as a theoretical lens for our paper, since it represents the mechanism that catalyses intra-firm and inter-firm efforts for the efficiency of R&D towards innovation,

The above leads to the research gap which relies mainly on the focal paper by Noseleit and de Faria (2013) since we are conducting a replication study. They argue that literature does not take into account the indirect effect of external knowledge towards internal R&D relationship with innovation output (Noseleit and de Faria 2013). The effect of partner diversity (complementarity of resources) on the relationship between internal R&D efforts towards innovation represents uncharted areas within strategic alliances (Noseleit and de Faria 2013). Hence we are going to study the research question of our focal paper.

How does the type of alliance partner (from the same industry, related industries, or unrelated industries) influence the relationship between an organization’s internal R&D investments and innovation performance? (Noseleit and de Faria 2013).

Furthermore Noseleit and de Faria (2013) paper looked at the electronic and equipment industry in the US. Their results show that intra-industry alliances and related-industry alliances have a positive influence with a stronger effect for related-industry partners than intra-industry while unrelated-partners have a negative effect on the relationship between R&D and innovation output (Noseleit and de Faria 2013). Limitations of the focal paper include a clear key factor for our study – replicative studies including data from other industries and other countries can aid to generalize the findings of this paper (Noseleit and de Faria 2013).

6

also clearly highlights that the planning of the research process is known to be significantly different in biotech organizations and pharmaceutical firms (Oliver and Montgomery 2000; Luo and Deng (2009). Biotech organizations generally make more investments in R&D, allow their scientists more independence, and pay closer attention to the science and technology community (Luo and Deng 2009; Oliver and Montgomery, 2000). Moreover, within the pharma industry, firms are broadly involved in strategic alliances because of the technological challenges, opportunities and costs of coming up with new products for the market (Hall and Bagschi-Sen, 2002; McKelvey et al. 2003; Sabidussi, 2017). Consequently we can argue that this type of study is relevant within the pharma industry as a heavily innovative and R&D focused area.

7

Theory and Hypotheses

In this section we review literature on absorptive capacity, intra-firm R&D and innovation in relation to alliances (inter-firm collaborations). Absorptive capacity is used as a theoretical lens in order to explain the moderation effects of different types of alliance partners on the relationship between R&D and innovation and build our hypotheses.

Absorptive Capacity

Absorptive capacity represents the ‘ability of a firm to recognize the value of new, external information, assimilate it and apply it to commercial ends – critical to its innovative capabilities’ (Cohen and Levinthal 1990 p128). Moreover Zahra and George (2002) show four dimensions of absorptive capacity – Acquisition, Assimilation of new/external knowledge; Transformation and Exploitation i.e. transforming and exploiting the knowledge acquired/assimilated (Zahra and George 2002). The above is relevant to our study since absorptive capacity as defined by Cohen and Levinthal (1990) is critical for innovation. Subsequently organizations increase their internal R&D investments to develop absorptive capacity (Cohen and Levinthal 1990). This happens since R&D generates a capacity to assimilate and exploit new knowledge (Cohen and Levinthal 1990). Tsai (2001) and Mowery and Oxley (1995) also argue that internal R&D spending represents a key factor for a firm developing absorptive capacity.

Internal R&D Investment Leading to Innovation

8

and defines problems and then pro-actively develops new knowledge in order to solve them (Nonaka 1994). This proves that robust R&D investments are needed in order to develop this new knowledge that can solve these problems and lead to innovation.

To go into more detail regarding the abovementioned process (R&D investment triggering innovation) Rosenberg (2010) highlights the importance of internal R&D by arguing that an organization requires a significant internal research capability in order to recognize, understand and apply internal knowledge that has been located on the shelf. This can be turned into innovation through knowledge recombination (Grigoriou and Rothaermel 2017). In order to transform knowledge placed on the shelf, efficient knowledge exchange is needed (Jansen et al. 2006). By taking an absorptive capacity approach, literature highlights that connectedness (density of linkages; also known as an attribute exhibited by the socialization capabilities of the firm) represents an organizational mechanism which increases the efficiency of knowledge exchange within an organization (Jansen et al. 2006). This facilitates knowledge exchange by developing trust and cooperation within the firm (Jansen et al. 2006). In this way knowledge is transformed and exploited (Zahra and George 2002). This increases the absorptive capacity of the firm. Additionally continuous investments in R&D are needed since R&D effectiveness is path dependent and thus, lack of investment in internal R&D at a specific time may exclude future options in a particular technology (Cohen and Levinthal 1989) and reduce innovation output. Furthermore literature indicates a positive direct effect of R&D upon innovation output (Helfat 1997). This shows that firms which do not continuously invest in internal R&D will render less innovation output than organizations that do invest in this area.

9

capacity is utilized as a key factor for increasing innovation performance (Zahra and George 2002; Cohen and Levinthal 1990). Here, when faced with a new technological paradigm, focal organizations can choose internal development in order to obtain the new knowledge needed (Grigoriou and Rothaermel 2017). In this case they should have a robust potential for knowledge recombination or already a significantly high level of coordination costs for generating knowledge internally. Within an organization, R&D investment can create knowledge and it will also lead to more internal resources; this in turn will boost absorptive capacity and innovation. However these resources need to be managed efficiently for an increased absorptive capacity. Since Cohen and Levinthal (1990) argue that absorptive capacity can most efficiently be developed through the analysis of the cognitive structure that underlies learning, it can be useful to look at training. Here R&D investment can require tools to manage new knowledge such as training sessions that can have a positive aspect on social capabilities like connectedness between the employees of an organization and in turn on the assimilation, transformation and exploitation (absorptive capacity) of new knowledge (Perlines and Araque 2015, Jansen et al 2006).

This shows how R&D can lead to an increased absorptive capacity through connectedness and training. In turn absorptive capacity represents a critical factor for innovation. Thus we expect that R&D investment leads to increased innovation output through absorptive capacity at the firm level.

Hypothesis 1 (H1): There is a positive relationship between internal R&D investment and firm innovation performance.

The Moderating Effect of Intra-Industry Alliances

10

organizations (Kogut 1988). Our argumentation relies on the fact that the boundaries of the focal firm do not change since our theoretical lens is absorptive capacity at the firm level as abovementioned.

Following up from H1, besides internal R&D investment, firms are also seeking alternatives to in-house R&D i.e. R&D alliances (Lin et al. 2012). Literature shows that successful knowledge renewal relies on firms developing abilities in both internal and external knowledge development and external knowledge sourcing (Helfat et al. 2007; Grigoriou and Rothaermel 2017). Furthermore extensive literature shows that innovation active organizations cannot rely only on internal R&D sourcing but also require knowledge from outside their boundaries (Cohen and Levinthal 1990; Lin et al. 2012; Zahra and George 2002). Moreover the access to external know-how may influence positively the efficiency of internal R&D (through increased absorptive capacity), at least when the organization shows an alacrity to take on external ideas and knowledge, overcoming the ‘not invented here’ syndrome for innovation (Allen 1986; Cassiman and Veugelers, 2004). As in H1 development, we are going to use connectedness as an organizational mechanism that explains how the absorptive capacity increases within the organization for a more efficient R&D (Jansen et al. 2006) (ACAP will be higher than in H1). Additionally common languages between organizations will also be needed between partners as explained below.

11

2006). Moreover connectedness within the focal firm can motivate employees to assist each other in a two-way interaction that enables the efficient interpretation and understanding of external know-how at the firm level (Jansen et al 2006). This sknow-hows know-how the assimilation process takes place. Jansen et al (2006) also argues that connectedness creates trust and cooperation and also increases the knowledge exchange inside the focal organization enabling the organization to transform and exploit external knowledge (Zahra and George 2002). Thus we can derive that sharing the same ‘languages’ with a partner in an alliance can increase the value of connectedness within an organization. This will also enable assimilation, transformation and exploitation i.e. an increased absorptive capacity of a focal organization since it can understand the external knowledge. Furthermore, as abovementioned in H1 development, training within the focal firm can also increase connectedness (Perlines and Araque 2015, Jansen et al 2006). However in this case connectedness will be taken to a higher level than in Hypothesis 1 since external knowledge comes into the picture. This shows that the absorptive capacity of the organization shall be higher than in Hypothesis 1 since external knowledge can be assimilated, transformed and exploited through common languages and increased connectedness. As organizational absorptive capacity will be higher than in H1, internal R&D is also going to be more efficient.

On a similar note, Noseleit and de Faria (2013) findings show a positive (limited) effect of intra-industry alliances on the relationship between R&D investment and Innovation Output.

12

Hypothesis 2 (H2): Intra-Industry alliance partners represent a positive moderator on the relationship between R&D Investment and firm innovation performance.

The Moderating Effect of Related-Industry Alliances

13

Hypothesis 2. This happens since redundant knowledge is no longer an issue. Here we have shown how increased socialization capabilities (connectedness) and common languages lead to higher absorptive capacity than in Hypothesis 2.

Noseleit and de Faria (2013) main findings highlight that companies that engage in alliances with related partners benefit from more efficient usage of internal R&D and are also more likely (than intra-industry) to benefit from an additional alliance premium.

Consequently we can argue that due to the reduced amount of ‘redundant knowledge’ and automatically the increased amount of relevant know-how, R&D efficiency towards innovation increases in comparison with H2. This is enabled through higher absorptive capacity triggered by highly effective socialization capabilities (connectedness).

Hypothesis 3 (H3): Related-Industry alliance partners represent a positive moderator (stronger than intra-industry alliances) on the relationship between R&D Investment and firm innovation performance.

The Moderating Effect of Unrelated-Industry Alliances

14

assimilation, transformation and exploitation (Jansen et al. 2006). This happens since a connected network within the organization will now result in low motivation to understand and interpret the external knowledge since this is ‘too distant’ (Jansen et al. 2006). In turn this happens because low levels of knowledge exchange will lead to low trust and cooperation between employees (Jansen et al. 2006). Moreover, by using the same tool as in the previous hypotheses training will not be able to increase connectedness due to the nature of the know-how (Perlines and Araque 2015, Jansen et al 2006). As in H3 development, Cohen and Levinthal (1990) argue that the technological opportunity (amount of relevant know-how) influences the ability of a firm to apply knowledge. Thus no overlap of knowledge between partners and high redundancy of external know-how for the focal firm will lower absorptive capacity.

Noseleit and de Faria (2013) also discover a negative effect of partners in unrelated industries towards the relationship between internal R&D efficiency and innovation output.

Consequently we can argue that the efficiency of internal R&D towards innovation is hampered by unrelated-industry alliance due to low absorptive capacity through reduced connectedness, lack of common languages between partner organizations. The following hypothesis adds a negative spin to our baseline hypothesis (H1).

Hypothesis 4 (H4): Unrelated-Industry Alliances represent a negative moderator on the relationship between R&D investment and firm innovation performance.

15

Figure 1- Conceptual Model

16

Methodology

Within our paper we mainly aim to examine the influence of various industry alliance partners (3 types: intra-industry; related-industry; unrelated-industry) on the internal R&D efficiency towards innovation performance (in the focal firm) in the bio-pharmaceutical industry.

Looking at different types of replication studies, literature shows that researchers might vary the way of measuring specific variables or utilize a different statistical method for data analysis (Tsang and Kwan 1999). Six different types of replication have been identified (Table 1) (Tsang and Kwan 1999). Here we can state that our thesis includes generalization and extension since we have selected our sample from a different population (pharma industry) as abovementioned and also employed different measurement of the moderator variables i.e. we look at alliances as dyads (for the focal firm) while Noseleit and de Faria (2013) considered at both dyadic and multi-party alliances. However a similar data analysis has been employed. This will be developed in the measurement of variables section.

17

Table 1 – Types of Replication

Same Measurement and Analysis Different Measurement and/or Analysis Same Data Set Checking of Analysis Reanalysis of Data

Same Population Exact Replication Conceptual Extension

Different Population Empirical Generalization Generalization and Extension (Tsang and Kwan, 1999 p766)

Data Collection

18

Focusing on only one industry (pharma) shall enable our study to avoid issues of heterogeneity across industries in a similar way with Noseleit and de Faria (2013). They have looked at the electronic and electric equipment industry.

Following Noseleit and de Faria (2002) we utilize the Securities Data Corporation’s (SDC) Platinum database (existing database) for the research process in order to recognize technology alliances. This source provides information about the strategic alliances that were publicly announced throughout the last decades. NBER patent data project database is utilized in order to get measures of innovation performance. COMPUSTAT database and annual reports are included in the methodology in order to obtain general information about the organizations i.e. sales, employment, R&D spending (Noseleit and de Faria 2013).

19

The above described sample of organizations will be analysed by utilizing STATA software for undertaking statistical tests.

Measures

Dependent Variable – Innovation Output

Literature shows that innovation output is generally measured by utilizing patents (Sorensen and Stuart, 2000; Hall et al., 2000; Archibugi and Planta 1996; Noseleit and de Faria 2013). Furthermore the tendency to patent within the pharma industry is high, making patent data feasible indicators of technological activity (Hall and Bagchi-Sen, 2002; Porkolab, 2002). Other studies also state that patents issued are a specifically suitable measure of innovation since the creation of patents represents the primary objective of R&D alliances within knowledge-intensive industries such as the pharma industry (Luo and Deng 2009; Sabidussi 2017; Arundel and Kabla 1998; Campbell 2005). On the other hand Griliches (1990) argue that the weighting process (citation-weighed count of patents) takes into account for differences in the value of patented innovations. Noseleit and de Faria (2013) use weighed patents (NBER database). Thus innovation output is measured by citation-weighed count of patents of the focal organization. To go into more detail we use the linear weighting scheme proposed by Trajtenberg (1990) and applied in our focal paper (Noseleit and de Faria 2013). Here each patent 𝑖 is weighted by the number of citations received, denoted as 𝐶𝑖 below

(Trajtenberg 1990). As presented below WPC represents the citation weighed count of patents in a given year 𝑡 where the number of patents granted during this year is represented by 𝑛𝑡 (Trajtenberg

1990). Additionally, we have utilized ln(1+𝑊𝑃𝐶𝑡) as our dependent variable since the formula below

can have the value of 0. This represents a continuous variable since it can have an infinite number of values.

𝑊𝑃𝐶𝑡 = ∑(1 + 𝐶𝑖) 𝑛𝑡

𝑖=1

20 Independent Variable

R&D Investment

Following Noseleit and de Faria (2013) yearly figures of internal R&D Investment are taken into account. COMPUSTAT database is used for obtaining data (Compustat 2018). Furthermore as data for some firms are missing from COMPUSTAT, we use annual reports to fill this gap. Logarithm (ln) is used to avoid skewed data and replicate the focal paper. Since natural logarithm is applied, this represents a continuous variable.

Moderator Variables

For all three moderator variables we use R&D collaboration by looking at the ‘Research & Development Agreement Flag’ (Yes) in SDC database and also make sure that they are not Joint Ventures by using the Joint Venture Flag (No). Here we exclude Joint Ventures following the argumentation presented within the previous chapter (Theory and Background) i.e. Joint Ventures overlap the boundaries of organizations (Kogut 1988).

Intra-industry Alliances

These are comprised by R&D collaborations including focal firms that share the same primary 4-digit SIC code for the bio-pharma industry (2834 – Pharmaceutical & 2836, 8731 – Biotechnology) (Sabidussi et al 2017) (COMPUSTAT Global Data 2002). According to Noseleit and de Faria (2013) these firms have the highest probability to represent direct competitors of the focal organization and a significant overlap in know-how. Since the number of partners can only be an integer number and no logarithm has been applied this represents a discrete variable.

Related-industry Alliances

21

Here there should be a lower probability to represent a direct competitor and possess a smaller overlap in expertise (Noseleit and de Faria 2013). In a similar way with intra-industry alliances this represents a discrete variable.

Unrelated-industry Alliances

Unrelated alliances represent all other R&D collaborations that include partners from different industries (Noseleit and de Faria 2013) i.e all four digits or the last three are different. Similarly as intra-industry alliances this is a discrete variable.

Control Variables:

Employment – Following Noseleit and de Faria (2013) we control the size of the focal firm size by using employment. This is needed since large organizations may be able to produce more innovation than small firms (Luo and Deng, 2009). This data is found in COMPUSTAT Global database (Compustat 2018). Furthermore logarithm of employment (ln – natural logarithm) is also applied in order to replicate the focal paper and to better deal with skewed data since we are including firms of various sizes and implicitly various numbers of employees. As we apply logarithm this represents a continuous variable.

Sales - This control variable is also relevant since organizations with higher sales may be able to produce more innovation output. It has also been used by the focal study (Noseleit and de Faria 2013). COMPUSTAT will also be utilized as a database for this variable (Compustat 2018). Logarithm has been applied for the same reasons as above (employment variable). Similarly, as employment, this is a discrete variable.

Data Analysis

22

types of partners. A focal organization in a specific year (time variation) shall be considered in both of them. Subsequently within the first equation 𝑦𝑖𝑡+1 considers the citation weighted count of

patents calculated as ln (1+citation weighed patents). 𝜏𝑖 – represent firm-fixed effects controlling for

variations in firm innovation performance because of time-invariant characteristics (Noseleit and de Faria 2013). This also eliminates unobserved heterogeneity which is constant over time, for example differences in ownership structure, organizational culture and propensity to innovate (Noseleit and de Faria 2013). 𝜆𝑡 - year effects take into consideration aggregate fluctuations in innovation

performance due to business cycles, policy changes and technological shocks (Noseleit and de Faria 2013). The coefficient 𝛽1 highlights the effect of internal R&D investment towards innovation

performance. 𝛾𝑋𝑖𝑡 represents a vector of additional control variables for company specific

characteristics which vary in time i.e. sales and employment. Finally 𝜀𝑖𝑡 represents the error term

that shows the degree of precision for our prediction.

𝑦𝑖𝑡+1= 𝜏𝑖+ 𝜆𝑡+ 𝛽1𝑅&𝐷𝑖𝑡+ 𝛾𝑋𝑖𝑡+ 𝜀𝑖𝑡 (1)

We have to mention that the above equation differs from the first equation of Noseleit and de Faria (2013) since this is testing for our Hypothesis 1, which is not comprised within the research question of the focal paper.

Subsequently our second equation is identical to Noseleit and de Faria (2013) since this represents the core of our study including the moderation effects of alliance partners. Here 𝛽6, 𝛽7 and 𝛽8

(coefficients of the interaction terms) highlight the effect of different alliance partners towards the efficiency of R&D towards innovation performance.

𝑦𝑖𝑡+1= 𝜏𝑖+ 𝜆𝑡+ 𝛽1𝑅&𝐷𝑖𝑡+ 𝛽2𝐼𝑛𝑡𝑖𝑡+ 𝛽3𝑅𝑒𝑙𝑖𝑡+ 𝛽4𝑈𝑛𝑟𝑒𝑙𝑖𝑡 + 𝛽5(𝐼𝑛𝑡𝑖𝑡𝑅&𝐷𝑖𝑡) + 𝛽6(𝑅𝑒𝑙𝑖𝑡𝑅&𝐷𝑖𝑡)

+ 𝛽7(𝑈𝑛𝑟𝑒𝑙𝑖𝑡𝑅&𝐷𝑖𝑡) + 𝛾𝑋𝑖𝑡+ 𝜀𝑖𝑡 (2)

23

the cluster-robustness check (cluster ID) that we use in Stata relaxes the standard requirement that the observations should be independent (Stata Manuals 2018). In other words the correlations are independent across clusters but not automatically within groups. This influences the standard errors but not the coefficients (Stata Manuals 2018). Furthermore literature shows that this check is appropriate when cluster are not ‘few’ (Cameron and Miller 2015). They mention that there is no set definition of a limit number of clusters under which this type of robustness check is not effective (Cameron and Miller 2015). Thus we can argue that it is feasible to apply the robustness check.

24

Results

This chapter will look at the results of the statistical analysis beginning with descriptive statistics and going further with regression results for the hypotheses presented earlier in our paper (Theory and Background).

Descriptive Statistics

The sample within our regression analysis included 30 firms with 149 observations between 2000 – 2004 (employment data is missing – thus 1 observation dropped). However first differences correlations include only 119 observations (30 observations are dropped). Below we are going to argue why this phenomenon happens.

By replicating Noseleit and de Faria (2013), Table 2 shows the main statistics of the variables and correlations between them. Since the first set of correlation looks at both within and between variation the problem of multicollinearity comes into place. The between variation shows how firms are systematically different from one another. Within variation looks at how firms’ behaviours vary between observations within firm groups. Thus we needed to look at correlations between first differences which take into account mainly the within variation. That is why some observations are excluded (‘t+1 – t’ means that we have x_2004-x_2003; x2003-x_2002; x2002-x_2001; x2001-x2000) This means that 30 observations shall be excluded and in this case the correlations will be drawn from 4x30=120 observations from which 1 is dropped due to missing employment data leaving 119 observations (as aforementioned). Here the issue of multicollinearity does not exist anymore i.e. all correlations are lower than the 0.70 threshold (Table 2).

25

Table 2 - Descriptives and Correlations

Furthermore we have conducted a Variation Inflation Factor (VIF) test which shows that multicollinearity does not represent an issue. In turn here we tested both for the original variables (Appendix 2) and the first difference variables (Appendix 3). Both tests rendered values lower than 10 – which is the limit for multicollinearity value (Grewal et al. 2004).

Considering our independent variable R&D investment, we observe that the average firm possesses $1.77b R&D spend. Looking at moderators the 30 firms included in our sample engaged in a total of 518 alliances divided into 182 intra-industry alliances (35%), 174 related-industry alliances (34%) and 158 unrelated industry alliances (31%) (Appendix 4). This shows that intra-industry alliances are the most common closely followed by related-industry. Unrelated industry is the least common type of alliance within our sample. However, here we can also look at the number of firms that participated each year in a particular type of alliance (Table 3). This yields similar results to the above analysis.

Mean Std. Dev.

Min Max Correlation (levels) Correlations (first differences)

26

Here we have followed Noseleit and de Faria (2013). Nevertheless this can be compared to the focal paper only by taking into account the different measurement of the moderator that we have employed.

Table 3 – Alliance participation across time

No. of firms in

Year alliance

Intra-Industry Related-Industry Unrelated-Industry

2000 19 21 22

2001 19 19 21

2002 20 23 20

2003 23 23 20

2004 23 23 20

27

Figure 2 - Distribution of citation-weighed-patents (5 year time window)

28

Regression Results

Table 4 – Regression results

Variable Model 1 Model 2 Model 3 Model 4 Model 5

Internal R&D (ln) -0.119 (0.190) -0.102 (0.182) -0.0002 (0.218) Intra-industry alliance intensity 0.010

(0.120)

0.015 (0.123)

0.322 (0.272) Related-industry alliance intensity 0.055

(0.053)

0.052 (0.054)

0.305 (0.257) Unrelated-industry alliance intensity 0.012

(0.064)

0.013 (0.066)

0.234 (0.321) Internal R&D (ln) *Intra-industry

alliance intensity

-0.042 (0.039) Internal R&D (ln) *Related-Industry

alliance intensity

-0.035 (0.034) Internal R&D (ln)

*Unrelated-Industry alliance intensity

-0.025 (0.041) Sales (ln) 0.0841 (0.346) 0.160 (0.354) 0.033 (0.356) 0.099 (0.402) 0.135 (0.374) Employment (ln) 0.0629 (0.067) 0.0688 (0.065) 0.066 (0.062) 0.071 (0.061) 0.068 (0.059) No of Observations 149 149 149 149 149 No of Firms 30 30 30 149 30 Constant 2.79 2.87 2.94 2.99 1.62 F (Prob>F) 2.37 (0.05) 2.14 (0.07) 3.13 (0.009) 2.90 (0.01) 3.26 (0.004) R-squared (within) 0.03 0.03 0.04 0.046 0.076 Significance levels:*p<=0.1, **p<=0.05, ***p<=0.01 Robust Standard errors in parentheses Standardized coefficients are reported

29

Model 1 represents the baseline model where we have included only the control variables. These two do not have significant coefficients. This means that there is no significant impact of sales and employment on innovation performance. Looking at Model 2 – this is testing Hypothesis 1. Surprisingly the model shows an insignificant effect (p≥0.1) of the internal R&D investment towards citation-weighted patents. Additionally the R2 within value is 3%. This means that a precise prediction cannot be derived from this model since it explains very little of the variability of the response data around its mean. Thus we can conclude that surprisingly our study does not show support for Hypothesis 1. On the other hand we also cannot contradict Hypothesis 1. Models 3 and 4 show no significant results for the variables involved.

30

Discussion

In this section we are going to compare our results with the focal paper of Noseleit and de Faria (2013) and other literature, draw theoretical and managerial implications. Furthermore we are going to acknowledge limitations of our study and make recommendations for future research.

Theoretical Implications

Existing literature shows that R&D has a positive effect on innovation through absorptive capacity (Lin et al. 2012, Cohen and Levinthal 1990). Previous research also encourages studies to look at contributions of absorptive capacity (Zahra and George 2002) and it is also highlighted that organizations are more and more relying on know-how from other firms in order to develop their own capabilities (Lane and Lubatkin 1998). Literature within the alliance portfolio area has called for a more in depth understanding of the alliance portfolio (Duysters and Lokshin 2011; Nooteboom et al. 2007) and particularly R&D alliances. Duyster and Lokshin (2011) state that further studies should encompass various dimensions of portfolio complexity. Scholars looked at aspects like the international nature of the portfolio and the inverted U-shaped direct effect of cognitive distance of external know-how towards innovation (Duysters and Lokshin 2011). Furthermore previous research shows a necessity to know how organizations use their portfolio to get access to various types of knowledge (George et al 2001).

31

further research in this area. Hence from our results we can derive that R&D investment is leading to less citation weighed patents. However, besides the fact that it is not significant, it has to be mentioned that the effect is also quite reduced (𝛽 = −0.119) (Table 4). Thus by following Online Statistics Education: An Interactive Multimedia Course (2018) developed by Rice University (world top 100 University) (Times Higher Education 2018) we can argue that the support for the above statement (R&D leading to less citation-weighted patents) is weak and the data are not conclusive. It is a similar case with the influence of alliances on innovation via R&D as follows. Surprisingly all three types of alliances have a positive but not significant and reduced effect on R&D efficiency with unrelated alliances (𝛽 = −0.025) having the ‘strongest’ effect out of the three moderators followed by related alliances (𝛽 = −0.035) and intra-industry alliances (𝛽 = −0.042) (Table 4). Although the positive effect of intra-industry and related-industry are in concordance with the focal paper, the unrelated-industry industry partners had a negative effect in the theory provided by Noseleit and de Faria (2013). Additionally it is relevant to mention that absorptive capacity has been used as a theoretical lens in order to build our hypotheses. The fact that we have not obtained support for the hypotheses cannot imply that other studies should not use this tool for further research since absorptive capacity cannot be measured.

32

Gittelman 2006). Additionally by looking at the documents describing patent data from NBER – the database does not differentiate between citations added by assessors and citations added by inventors (NBER 2018). Thus we can argue that the above literature suggests that 66% of the citations we have included in our formula for the dependent variable may not reflect knowledge flows and value of patents, thus leading to insignificant results. Consequently we can assume that this may represent a drawback to the manner in which the innovation is measured in our study Further research can look into this issue and studies involving inventors could highlight more thoroughly the actual value of patent citations.

Thus we can conclude that our study contributes to innovation literature and absorptive capacity in the context of intra (R&D) and inter-firm efforts (alliances) by highlighting a contradiction to previous literature. However, as abovementioned, the support for this statement needs to be considered with caution and by taking into account several limitations. We will analyse these in the limitations chapter.

Managerial Implications

Following the theoretical discussion we consider relevant to compare characteristics of the electronic/electronic and bio-pharmaceutical industries in order to draw managerial implications for the latter industry.

33

considered an industry with a significant focus on technology. Here we can observe that bio-pharma is also part of the complex product industry.

Although both industries are focused on knowledge we can derive from the above literature that the bio-pharmaceutical industry is more focused on R&D and innovation in comparison with the electrical/electronics industry. Our descriptive statistics also show that the average firm in our sample invests $1.77b in R&D. This means that our results are even more surprising since the high internal investments in R&D are deemed not beneficial for the innovation performance of the company.

Thus managers in the pharma industry can take notice of our findings keeping in mind the limitations of our study and differences between bio-pharmaceutical and electronic/electric equipment industries. The results should be used cautiously when making decisions regarding internal R&D investment and types of alliance partners chosen for technology alliances.

Limitations and Future Research

The fact that the results could not be replicated implies the need for testing our results in turn by future research.

34

and employment data for firms. This may result in a truncation in the patent data since some of these companies may have been granted patents in other countries like France for Sanofi-Aventis, Japan for Eisai.

Following Zahra and George (2002) we have measured our moderators as alliance dyads. They state that alliance dyad characteristics are important for absorptive capacity (Zahra and George 2002). Here we used dyadic alliances as they are and split multi-party alliances into dyads. Meanwhile Noseleit and de Faria (2013) considered both dyadic and multi-party alliances (Noseleit and de Faria 2013). For example, in case of a multi-party alliance, a majority of partner-firms are required to be mainly active within the same SIC code as the focal organization (Noseleit and de Faria 2013). This highlighted the category of intra-industry alliances (the other alliances were considered in a similar way) (Noseleit and de Faria 2013). Thus we have conducted generalization and extension as aforementioned in the methodology chapter (Tsang and Kwan 1999). Furthermore Tsang and Kwan (1999) argue that if the results of the replication do not support the findings of the focal paper, it is challenging to say if the lack of support stems from the instability of the findings or from the imprecision of the replication study. Here we can come up with two different explanations as reasons for the results (not significant) of our thesis. First of all the lack of support can be triggered by the instability of the findings. By looking at our results future research can include more control variables (that would have an effect on the dependent variable) since our models show quite a limited power of prediction. Furthermore as the p values are not significant, further research can increase our sample in order to come up with more significant results. On the other hand by taking into account the imprecision of the replication study as a cause – it may be the case that different measurement of the moderator variables can be employed in the future.

35

and analysis; different measurement and/or analysis in comparison with our study and the focal study.

36

Conclusion

In our study we have researched the influence of internal R&D on innovation performance. More than this we have looked at the effect of different types of alliance partners (intra-industry, related-industry, unrelated-industry) on the relationship between R&D and innovation. Surprisingly our findings do not show statistically significant support for a positive relationship between R&D and innovation or for a moderator effect of alliance partners on this relationship. We consider these results surprising since they differ from our focal paper and previous literature, which shows that R&D leads to higher innovation performance (Lin et al 2012). As a contribution to innovation literature we suggest that R&D may actually have a small negative effect on innovation. However the above need to be further tested and we cannot contradict previous research. Following our discussion we can also recommend researchers to take into account whether or not citations are relevant for measuring innovation.

Thus future research can use our results cautiously, by keeping in mind the sample used and the aforementioned limitations of our study. Besides increasing the sample and the time period (panel data) within the pharma industry, future studies can also take this theory further and conduct replication studies for other industries. Our findings represent a foundation for studies that take into account the interaction between intra-firm and inter-firm efforts. By using our sample future studies can include case studies in the bio-pharmaceutical industry in order to study the influence of internal R&D towards innovation. Managers should take into account our results by considering the differences between the industries and also the fact that our study is not without its limitations.

37

Acknowledgements

38

References

Ahuja G. and Lampert C Morris (2001). Entrepreneurship in the large corporation: A longitudinal study of how established firm create breakthrough inventions. Strategic Management Journal. 22 (6/7) 521-543.

Alcacer, J. and Gittelman, M (2006). Patent citations as a measure of knowledge flows: The influence of examiner citations. The Review of Economics and Statistics. 88 (4), 774-779.

Arundel A. and Kabla I. (1998). What percentage of innovations are patented? Empirical estimates for European firms. Research Policy. 27 (1998), 127-141.

Alfresco (2005). Top 50 Pharmexec. Last accessed 7th June 2018 at http://alfresco.ubm- us.net/alfresco_images/pharma/2014/08/22/a946f829-9686-4128-a6eb-e011ed78f0dd/article-169778.pdf

Allen, T. (1986) Organizational Structure, Information Technology and R&D Productivity, IEEE Transactions on Engineering Management, 33, 212-217.

Archibugi and Planta (1996). Measuring Technological Change through Patents and Innovation

Surveys. Technovation. 16 (9), 451-468. Available at:

http://www.danielearchibugi.org/downloads/papers/2017/11/Archi-Pianta_Technovation.pdf

Bania, N., Eberts, R. W. and Fogarty M. S., (1993). Universities and the start-up of new companies: can we generalize from route 128 and Silicon Valley. The Review of Economics and Statistics. 75 (4). 761-766.

Bos, B. Faems, D., Noseleit, F. (2017). Alliance Concentration in Multinational Companies: Examining Alliance Portfolios. Strategic Management Journal, 38, 2298-2309.

39

Cameron A. C. and Miller D. L. (2015). A practitioner’s guide to cluster-robust inference. Journal of Human Resources. 50 (2), 1-49.

Campbell, J. J. (2005). Understanding Pharma. A Primer on How Pharmaceutical Companies Really Work. Raleigh, NC: Pharmaceutical Institute, Inc.

Caloghirou et al (2004). Internal Capabilities and external knowledge sources: complements or substitutes for innovative performance? Technovation. 24, 29-39.

Camison, C., Fores, B., (2010). Knowledge absorptive capacity: new insights for its conceptualization and measurement. Journal of Business Research 63, 707–715.

Cassiman and Veugelers (2002). R&D cooperation and spillovers: some empirical evidence from Belgium. American Economic Review. 92 (4) 1169-1184.

Cassiman and Veugelers (2006). In search of Complementarity in the Innovation Strategy: Internal R&D and External Knowledge Acquisition.

Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35, 128–153.

Cohen, W. M., & Levinthal, D. A. (1989). Innovation and learning: The two faces of R&D. The Economic Journal, 99, 569–596.

Cohen, W. M., Nelson, R. R., & Walsh, J. P. (2000). Protecting their intellectual assets: Appropriability conditions and why US manufacturing firms patent (or not). NBER working paper 2000; WP no. 7552.

40

COMPUSTAT Global Data (2002). Last accessed 23rd May at:

http://fin.ncue.edu.tw/compustat/manual/globdata/Part3h.pdf

Cooper L., C. and Finkelstein S. (2012). Advances in Mergers and Acquisition. United Kingdom, Emerald.

Deck A. C., Fatas E., and Rosenblat, Tanya (2015). Replication in experimental economics, United Kingdom, Emerald.

Doz, Y., and Hamel, G. (1998). Alliance advantage: The art of creating value through partnering. Boston: Harvard Business School Press.

Duysters, G., and Lokshin B., (2011). Determinants of alliance portfolio complexity and its effect on innovative performance of companies. Journal of Product Innovation Management. 28 (4).

Enkel and Heil (2014). Preparing for distant collaboration: Antecedents to potential absorptive

capacity in cross-industry innovation. Available at:

https://doi.org/10.1016/j.technovation.2014.01.010

Faems D., Looy V. B., Debackere, K. (2005). Interorganizational Collaboration and Innovation: Toward a Portfolio Approach. 22, 238-250.

Faems D., Janssens M., Madhok A., Looy V.B (2008). Toward an integrative perspective on alliance governance: connecting contract design, trust dynamics. Academy of Management Journal. 6 (51). 1053-1078.

Fleming L. (2001). Recombinant uncertainty in technological search. Management Science. 47 (1), 117-132.

41

Grant R. (1996). Toward a knowledge-based theory of the firm. Strategic Management Journal. 17 (Winter Special Issue) 109-122.

Grewal R., Cote J.A., Baumgartner H., (2004). Multicollinearity and measurement error in structural equation models: Implications for theory testing. Marketing Science. 23 (4). 519-529.

Griliches Z. (1990). Patent statistics as economic indicator: A survey. Journal of Economic Literature, 28, 1661-1707.

Grigoriou K. and Rothaermel F., T. (2017). Organizing for knowledge generation: internal knowledge networks and the contingent effect of external sourcing. Strategic Management Journal, 39, 395-414.

Gulati R. and Singh H., (1998). The Architecture of Cooperation: Managing Coordination Costs and Appreciation Concerns in Strategic Alliances. Administrative Science Quarterly, 43 (4), 781-814.

Hagedoorn J. (2012). Inter-firm R&D partnerships: an overview of major trends and patterns since 1960. Research Policy. 31, 477-492.

Hall, L. A., and S. Bagschi-Sen. 2002. “A Study of R&D, Innovation, and Business Performance in the Canadian Biotechnology Industry.” Technovation 22: 231–244.

Hall, B., Jaffe, A., Trajtenberg, M., (2000). Market Value and Patent Citations: A First Look. NBER, Working Paper 7741.

Hall, B. H. (2004). Exploring the patent explosion. The Journal of Technology Transfer, 30, 35–48.

42

Helfat C., E., (1997). Know-how and asset complementarity and dynamic capability accumulation: The case of R&D. Strategic Management Journal. 18 (5), 339-360.

Helfat C., E., et al. (2007). Stylized facts, empirical research and theory development in management. Strategic Organization. 5 (2), 185-192.

Jaffe, A. B., M. Trajtenberg, and M. S., Fogarty, (2000). “Knowledge Spillovers and Patent Citations: Evidence from a Survey of Inventors,” The American Economic Review 90 (2), 215–218.

Jansen, J. J., Van Den Bosch, F. A., & Volberda, H. W. (2006). Managing potential and realized absorptive capacity: how do organizational antecedents matter? Academy of Management Journal, 48(6), 999-1015.

Katila R. (2002). New product search over time: Past ideas in their prime. Academy of Management Journal.

Keil, T., Maula, M., Schildt, H., & Zahra, S. (2008). The effect of governance modes and relatedness of external business development activities on innovative performance. Strategic Management Journal, 29, 895–907.

Kim L., (1998). Crisis construction and organizational learning: Capability building in catching-up at Hyundai Motor. 9 (4). 506-521.

Kogut, B. (1988). Joint ventures: Theoretical and empirical perspectives. Strategic Management Journal, 9: 319 –332.

Lane, P. J., & Lubatkin, M. 1998. Relative absorptive capacity and interorganizational learning. Strategic Management Journal, 19: 461–477.

43

Lin J., Z., Yang, H. and Arya, B., (2009). Alliance Partners and Firm Performance: Resource Complementarity and Status Association. 30, 921-940.

Lin et al. (2012). The alliance innovation performance of R&D alliances—the absorptive capacity perspective. Technovation, 32, 282 – 292.

Luo X and Deng, L. (2009). Do Birds of a Feather Flock Higher? The Effects of Partner Similarity on Innovation in Strategic Alliances in Knowledge-Intensive Industries. Journal of Management Studies 46, 1005–1030.

McKelvey, M., H. Alm, and M. Riccaboni. 2003. “Does Co-location Matter for Formal Knowledge Collaboration in the Swedish Biotechnology–pharmaceutical Sector?” Research Policy 32: 483–501.

Morbey G., K. and Reithner R., M., (1990). How R&D affects sales growth, productivity and profitability. Research-Technology Management. 33 (3), 11-14.

Mowery C., D., and Oxley E.J., (1995) Inward technology transfers and competitiveness: the role of national innovation systems. Cambridge Journal of Economics. 19, 67-93.

NBER (2018). Last accessed 4th June 2018 at: http://admin.nber.org.proxy-ub.rug.nl/patents/

Nerkar and Paruchuri (2005). Evolution of R&D capabilities: The role of knowledge networks within a firm. Management Science, 51 (5) 771-785.

Nonaka, I. (1994). A dynamic theory of organisational knowledge creation. Organisation Science 5, 14–37.

Noseleit, F. and de Faria, P. (2013). Complementarities of internal R&D and alliances with different partner types. Journal of Business Research. 66 (10) 2000-2006.

44

Oliver, A. and Montgomery, K. (2000). ‘Creating a hybrid organizational form from parental blueprints: the emergence and evolution of knowledge firms’. Human Relations, 53, 33–56.

Online Statistics Education: An Interactive Multimedia Course of Study (2018). Last accessed 22nd June at http://onlinestatbook.com/.

Perlines F. H. and Araque Y., B. (2015). Linking Training to Organizational Performance: An Absorptive Capacity-Based View. Case Study Method in Spanish Family Businesses. Journal of Promotion Management 21 (4).

Porkolab, L. (2002) “Evaluating R&D Projects and Portfolios.” Drug Discovery Today 7 (4): 230–231.

Rigby D. and Zook C. (2002). Open-market innovation. Harvard Business Review.

Rosenberg, N. (1990). Why do firms do basic research (with their own money)? Res Policy 19, 165-174.

Sampson, R. (2007). R&D alliances and firm performance: The impact of technological diversity and alliance organization on innovation. Academy of Management Journal, 50: 364-386.

Sabidussi A, Lokshin B, Duysters G. (2017). Complementarity in alliance portfolios and firm innovation. Industry and Innovation. https://doi.org/10.1080/13662716.2017.1329084

Song M., X., Thieme J., R., and Xie, J., (1998). The impact of cross-functional joint involvement across product development stages: An exploratory study.

Sørensen, J. B., & Stuart, T. E. (2000). Aging, obsolescence, and organizational innovation. Administrative science quarterly, 45(1), 81-112.

Stata Manuals (2018). Last accessed 7th June at:

45

The Times Higher Education (2018). Last accessed 20th June 2018 at:

https://www.timeshighereducation.com/world-university-rankings

Trajtenberg, M. (1990). A penny for your quotes: Patent citations and the value of innovations. The Rand Journal of Economics, 21, 172–187.

Tsang E., W., K., and Kwan, K.M (1999). Replication and Development in Organizational Science. A Critical Realist Perspective. The Academy of Management Review. 24 (4), 759-780.

Tsai W., (2001). Knowledge transfer in intraorganizational networks: effects of network position and absorptive capacity on business unit innovation and performance. The Academy of Management

Journal, 44 (5). 996-1004.

Vaduseva G. and Anand J. (2011). Unpacking absorptive capacity: A study of knowledge utilization from alliance portfolios. Academy of Management Journal. 54 (3), 611-623.

Vassolo, R. S., Anand, J., & Folta, T. B. (2004). Non-additivity in portfolios of exploration activities: A real options-based analysis of equity alliances in biotechnology. Strategic Management Journal, 25,

1045–1061. Wuyts et al. (2005). Empirical Tests Of Optimal Cognitive Distance. Journal of Economic Behavior ad

Organization. 58, 2, 277-302.

Zahra A., S. and George, G. (2002). Absorptive capacity: A review, reconceptualization, and

46

Appendix

Appendix 1 - List of Firms with patent Info in NBER

ID Firm Name Country CUSIP code

(COMPUSTAT – 9 digits; NBER/SDC – 6 digits) CUSIP code match between databases Independent Variable (R&D) Control Variables (Sales/Employment) Moderators (Alliances) Included in Sample 1 Abbott Lab US 002824100 V V V V V 2 Akzo Nobel NV NL 010199305 V V V V V 3 Allergan Inc. US 018490102 V V V V V 4 Altana AG DE 02143N103 V V V V V 5 Amgen Inc US 031162100 V V V V V 6 AstraZeneca GB 046353108 V V V V V 7 Baxter International US 071813109 V V V V V 8 Bayer AG DE 072730302 V V V V V 9 Biogen Inc US 09062X103 V V V V V 10 Bristol-Myers Squibb US 110122108 V V V V V 11 Eisai Co Ltd JP 282579309 V V V V V

12 Eli Lilly & Co. US 532457108 V V V V V

13 Forest Laboratories US 345838106 V V V V V 14 Genentech Inc. US 368710406 V V V V V 15 Genzyme US 372917104 V V V V V 16 GlaxoSmithKline Plc GB 37733W105 V V V V V 17 Ivax Corp US 465823102 V V V V V 18 Johnson & Johnson US 478160104 V V V V V 19 King Pharmaceuticals Inc US 495582108 V V V V V

20 Merck & Co. US 58933Y105 V V V V V

21 Mylan Labs US 628530 (SDC/NBER)

/N59465109 (NBER/Compustat)

X V V V X

22 Novartis CH 66987V109 V V V V V

23 Pfizer Inc. US 717081103 V V V V V

24 Purdue Pharma US 74616F (SDC/NBER) V X X V X

47

Appendix 2 – Variance Inflation Factor (levels)

Observations (149)

Variable

VIF

Tolerance – 1/VIF

Ln(R&D)

3.36

0.30

Intra-Industry Alliances

1.77

0.56

Related-Industry Alliances

1.90

0.52

Unrelated-Industry Alliances

1.78

0.45

Ln(Sales)

5.27

0.19

Ln(Employment)

2.81

0.36

Mean VIF

2.81

Appendix 3 – Variance Inflation Factor (first differences)

Observations (119)

Variable

VIF

Tolerance – 1/VIF

48

Appendix 4 – R&D Alliances