• No results found

THE EFFECT OF ACQUISITIONS ON INNOVATION IN THE PHARMACEUTICAL INDUSTRY

N/A
N/A
Protected

Academic year: 2021

Share "THE EFFECT OF ACQUISITIONS ON INNOVATION IN THE PHARMACEUTICAL INDUSTRY"

Copied!
69
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

THE EFFECT OF ACQUISITIONS

ON INNOVATION IN THE

PHARMACEUTICAL INDUSTRY

5 January 2016

International Business & Management

Author: L. de Jong

(2)

Master thesis

MSc International Economics & Business and MSc International Business & Management

The effect of acquisitions on innovation

in the pharmaceutical industry

05/01/2016

By: L. de Jong

s1887556

l.de.jong.13@student.rug.nl

Supervisor: dr. P. Rao Sahib Co-assessor: dr. R. Ortega Argiles

Faculty of Economics and Business University of Groningen

Duisenberg Building, Nettelbosje 2, 9747 AE Groningen, The Netherlands P.O. Box 800, 9700 AV Groningen, The Netherlands

(3)

PREFACE

This thesis is titled ‘The effect of acquisitions on innovation in the pharmaceutical industry’. It has been written as part of the master degrees International Economics & Business and International Business & Management at the University of Groningen, and has been written between September 2015 and January 2016.

My interest as an economist in the pharmaceutical industry, and the healthcare sector as a whole, stems from two things. First, even though the Dutch healthcare sector is increasingly privatized, no one is willing to accept the consequences of the free market. There is an equity argument in healthcare that needs to be taken into account, which constantly battles with the cost-driven mind-set. Second, I’m intrigued by the paradoxicality in acquisitions. Most mergers and acquisitions are known to fail. Apparently acquisitions provide benefits for at least one of the two parties involved. I am hoping to find in this thesis that, opposed to financial results, acquisitions could generate more innovation.

Above all, I would like to thank dr. Padma Rao Sahib for her critical eye, correcting both minor spelling mistakes and major econometric issues. For all the hours spent giving feedback and everlasting willingness to help and answer any questions I might have.

(4)

ABSTRACT

The pharmaceutical industry is currently in the fourth acquisition wave since 1989. Firms acquire targets with promising new drugs after the patents on many blockbuster drugs have expired.

This thesis examines how acquisitions influence innovation in the pharmaceutical industry. Organizational learning theory shows that two factors can enhance the effectiveness of knowledge transfer between firms. These are the similarity between acquirer and target on a firm level and industry level, and the knowledge base of the acquirer. The analysis is conducted using data on 445 acquisitions in the pharmaceutical industry, completed between 2006 and 2012.

The relative size between target and acquirer has a negative effect on post-acquisition innovation, for acquirers with annual sales less than $25 billion (95% of firms). When the acquirer and target are active in identical industries, this negatively affects the post-acquisition innovation. Previous post-acquisition experience of the acquirer, domestic and cross-border, also affects innovation. Firms with more experience with cross-border acquisitions have produced more patents post-acquisition. Total acquisition experience however, has a negative effect on innovation.

(5)

TABLE OF CONTENTS

!

1

!

INTRODUCTION 1

!

1.1

!

Prior empirical research 2

!

1.2

!

Contribution to the literature 3

!

2

!

BACKGROUND 5

!

2.1

!

The R&D process of pharmaceutical firms and their firm specific advantages 5

!

2.2

!

Acquiring knowledge from horizontal mergers 6

!

2.3

!

Acquiring knowledge from acquisitions 7

!

3

!

THEORY AND HYPOTHESES 9

!

3.1

!

Theoretical perspective 9

!

3.2

!

Conceptual model and hypotheses 12

!

4

!

METHODOLOGY 19

!

4.1

!

Data collection 19

!

4.2

!

Sample 21

!

4.3

!

Measures 22

!

4.4

!

Model specification 28

!

5

!

RESULTS 29

!

5.1

!

Descriptive statistics 29

!

5.2

!

Regression results 31

!

6

!

CONCLUSION 39

!

7

!

DISCUSSION 41

!

7.1

!

Limitations 42

!

7.2

!

Suggestions for future research 43

!

8

!

REFERENCES 45

!

9

!

APPENDICES 51

!

9.1

!

Tables 51

!

(6)

LIST OF FIGURES

Figure 1: The R&D process 5!

Figure 2: Conceptual model 12!

Figure 3: Frequency of number of acquisitions 20!

Figure 4: Distribution of firm size 20!

Figure 5: Timeline showing focal acquisition in time 22!

Figure 6: The distribution of the number of patents 31!

Figure 7: The distribution of relative acquisition size 31! Figure 8: Conditional marginal effects of industry similarity 33! Figure 9: Marginal effects of industry similarity on drug patents 35! Figure 10: Model specifications compared to observed distribution 59! Figure 11: The number of patents plotted against relative acquisition size 59! Figure 12: Distribution of acquisitions by sales of the acquirer 60! Figure 13: Relation between size and acquisition frequency 60!

Figure 14: Marginal effects when sales ≤ μ + σ 61!

Figure 15: Marginal effects when sales ≤ μ 61!

Figure 16: Number of acquisitions per year 62!

Figure 17: Sum of total deal value of all acquisitions per year 62! Figure 18: Complementarity between internally and externally sourced R&D 62!

LIST OF EQUATIONS

( 1 ) The Kogut & Singh index to calculate cultural distance 26!

(7)

LIST OF TABLES

Table 1: Sample selection 21!

Table 2: Summary of variables 27!

Table 3: Descriptive statistics and intercorrelations (n=445) 30! Table 4: Regression results for dependent variable patents 32! Table 5: Regression results for dependent variable drug patents 35! Table 6: Descriptive statistics and intercorrelations with drug patents variable (n=445) 51! Table 7: Conditional marginal effects industry similarity on patents 52! Table 8: Conditional marginal effects industry similarity on drug patents 52! Table 9: Correlations between sales, r&d, acquisition frequency and patents 52! Table 10: Regression analysis with clustered standard errors 53! Table 11: Regression results using split sample at mean + standard deviation 54! Table 12: Marginal effects industry similarity (sales ≤ μ + σ (25,000)) 54! Table 13: Regression results using split sample at mean 55! Table 14: Marginal effects industry similarity (sales ≤ μ (10,000)) 55!

Table 15: Regression results with year dummies 56!

Table 16: Robustness check acquisition experience 57!

(8)

1 INTRODUCTION

There is an increasing recognition among researchers that pharmaceutical research and development (R&D) is facing a productivity crisis. This crisis is characterised by stagnation in the numbers of new drug approvals in the face of increasing R&D costs. The need for a renewed increase in R&D productivity is enlarged by upcoming patent expiry of blockbuster drugs and increasing costs of drug development (Khanna, 2012). A drug is called a ‘blockbuster’ if it generates more than $1 billion annually (Malik, 2009). To illustrate, Pfizer’s blockbuster drug ‘Lipitor’ is one of the drugs of which the patent has expired. Lipitor is a drug to lower the levels of cholesterol in the blood. Lipitor lost exclusivity in Canada, Spain, Brazil and Mexico in 2010, and in the U.S. in November 2011. Pfizer reports a drop in sales of $1.1 billion in 2011 (-11%), as compared to 2010 (Pfizer, 2011).

In response to declines in internal productivity, firms start to acquire other firms with a promising pipeline of new medicines (Higgins & Rodriguez, 2006). Indeed, this is what happened between 2006 and 2012 (The Economist, 2015). Internal R&D efforts have decreased and firms have increased their reliance on externally sourced R&D (Rafols et al., 2014).

However, although large mergers increase shareholder value in the short run (Cha & Lorriman, 2014), the effects on innovation and product development in the long run are less promising (Comanor & Scherer, 2013; Ornaghi, 2009).

Continuing with the example of Pfizer, there are two acquisitions made by Pfizer around the period of the loss of exclusivity on Lipitor that are interesting. First, Pfizer has acquired Wyeth in 2009 for $68 billion. Pfizer stated that this acquisition was undertaken to boost sales and reduce concerns about the loss of Lipitor (Reuters, 2009). Second, there is the acquisition of FoldRX for approximately $200 million. FoldRX is focused on developing treatments for rare diseases. This acquisition was clearly undertaken to benefit from the expected new drugs that follow from FoldRX’s R&D (Pfizer, 2010).

Although the large merger with Wyeth may increase sales in the short term because of extra existing products, smaller acquisitions, such as FoldRX, may yield more innovation in the future because of complementary R&D efforts.

(9)

1.1 Prior empirical research

Empirical research on the effects of acquisitions on innovation in the pharmaceutical sector is inconclusive. Innovation, in this context, means the development of new products. Hitt et al. (1991) finds a negative effect of acquisitions on R&D intensity and patent intensity. Ornaghi (2009 p.78) also finds that “Mergers do not produce important innovation advances or significant increases in research productivity”. However, when firms possess existing research capabilities in the underlying technologies, international acquisitions have a positive effect on the number of subsequent patent applications (Penner-Hahn & Shaver, 2005). Prabhu et al. (2005) also finds a positive effect, specifically when acquisitions are driven by the pre-acquisition knowledge of the acquirer and its similarity with the targets’ knowledge. This result is support for the knowledge-based view, which suggests that acquisitions can help innovation by adding to the knowledge base.

Moreover, within technological acquisitions, the size of the acquired knowledge base is found to influence innovation performance (Ahuja & Katila, 2001). Interestingly, Ahuja and Katila (2001) find that the absolute size of the knowledge base has a positive effect on innovation, but the relative size of the knowledge base has a negative effect on innovation. In other words, the smaller the knowledge base of the target, relative to that of the acquirer, the more innovation the acquisition yields.

A straightforward negative effect as found by Hitt et al. (1991) and Ornaghi (2009) cannot explain why the pharmaceutical industry continues to make acquisitions (Cha & Lorriman, 2014); why they do so very frequently; and why some seem to be successful (Almor et al., 2009). There appear to be circumstances and/or moderating effects, that, when present, can reverse the effect of acquisitions on innovation.

There is very little research on the relative size between target and acquirer to confirm these results. The absolute size of the acquirer or target has often been included in empirical studies as a control variable, and is found to have a positive effect (Ahuja & Katila, 2001; Penner-Hahn & Shaver, 2005; Prabhu et al., 2005). This is straightforward, as larger firms generally have greater R&D bases and generate more patents. As Ahuja and Katila (2001) have shown and as the example of Pfizer above illustrates, there is reason to assume that the relative size may be an interesting factor in the relation between acquisitions and innovation.

(10)

1.2 Contribution to the literature

The contribution of this thesis to the literature is to construct and test a framework that tries to capture all factors influencing the effect of acquisitions on innovation. The framework is based on organizational learning research, using insights of the resource-based view (Barney, 1991) and the absorptive capacity on knowledge transfer from R&D (Cohen & Levinthal, 1990). There is a specific focus on the effect of relative acquisition size on innovation. Data is used from a relatively recent period, in which there were no significant changes in acquisition activity. The sample is heterogeneous as it includes both very large pharmaceutical firms, and small to medium-sized firms.

Findings include the negative effect of relative acquisition size on innovation for firms with less than $25 billion annual sales (95% of firms). A negative effect has been found of both the acquirer and target being active in identical industries. Finally, a negative effect of acquisition experience on innovation was found.

(11)
(12)

2 BACKGROUND

This section gives background information on innovation in the pharmaceutical industry. It briefly describes some of the strategies firms have used to increase research productivity and gives a first hint as to why these strategies may or may not be effective.

2.1 The R&D process of pharmaceutical firms and their firm specific advantages

The R&D process of a pharmaceutical firm can be divided into two phases: the discovery phase and the development phase (Paul et al., 2010). These are shown in Figure 1.

Figure 1: The R&D process

Often, somewhere before the pre-clinical phase, the drug is patented (Dunn, 2011). After the pre-clinical phase, the development phase starts. This includes all types of clinical trials. These clinical trials account for 53% of total R&D expenditure (Paul et al., 2010). Phases 2 and 3, where the drug is tested on sick patients and on large groups, have the highest failure rates (Sams-Dodd, 2013). The development phase is therefore very capital intensive and high risk.

(13)

By increasing the internal R&D efforts, firms can increase the number of drugs to bring to the market. As seen in Figure 1, to achieve this, the R&D funds must be invested in the first stages of the process, where the drug needs to be discovered. However, pharmaceutical firms have allocated their R&D funds mainly to the later stages of product development (Kessel, 2011; Pammolli et al., 2011). Thus, large internal R&D bases do not result in higher R&D productivity (Graves & Langowitz, 1993).

For example, GlaxoSmithKline has average annual sales of $46.1 billion. Whilst having spent $6.2 billion on average per year on R&D, GlaxoSmithKline has only been granted one patent between 2007 and 2014. This patent was not related to drug development.

2.2 Acquiring knowledge from horizontal mergers

The second way to develop more drugs is to engage in horizontal mergers. Horizontal mergers occur between firms in the same industry, often competitors. Taking over a firm in the same industry in another country, geographic expansion, or with a different product, is also seen as a horizontal merger.

This strategy has been adopted by many big pharmaceutical firms between 1994 and 2000 (Wang et al., 2015). However, horizontal mergers do not produce important drug innovations or significant increases in R&D productivity (Comanor & Scherer, 2013; Graves & Langowitz, 1993; Ornaghi, 2009). For example, the merger of Pfizer and Wyeth in 2009 was made in order to benefit from Wyeth’s new-product launches. However, in 2014, Forbes recalls “In 2007, Pfizer’s R&D spend was $7.8 billion and Wyeth’s was close to $5 billion. By 2013, Pfizer had slashed its R&D budget to $6.55 billion, almost half of what the individual companies [together] had invested in 2007” (Forbes, 2014). It is unlikely that the portfolio of Wyeth can offset such a reduction of R&D inputs. When large companies merge, they are likely to have significant overlaps in their product portfolios. Having to divest these overlaps would eliminate some of the benefits of the merger (Frantz, 2006).

(14)

between 2000 and 2010, but the margins are significantly lower (Behnke & Hueltenschmid, 2010; Kessel, 2011). Although moving into generic drugs boosts sales, this does not mean that R&D productivity or innovation is increased. On the contrary, generic drugs benefit from the R&D conducted for the development of the patented drug that came before the generic drug.

Many firms also engage in geographic expansion into emerging markets, such as the fast-growing Brazil, Russia, India and China (BRIC countries). The growth in demand coincides with a fierce growth in competition which lowers profit margins (Khanna, 2012). Given the low operational costs in emerging countries, resulting in low cost drugs, it is difficult to capture market share in these markets with expensive brand-name drugs (Hasenclever & Paranhos, 2009). The global market for brand-name drugs is still concentrated in the U.S., Japan and Western Europe, accounting for eighty percent of market share (Behnke & Hueltenschmid, 2010).

2.3 Acquiring knowledge from acquisitions

A third strategy is diversification through acquisitions. Diversification means offering a new product in a new market (Ansoff, 1957). There is an increased interest shown by pharmaceutical companies in biotech firms (Malik, 2009; Wang et al., 2015). Where pharmaceutical firms produce medicines with a chemical basis, biotechnology companies use live organisms, such as bacteria or enzymes, to manufacture their drugs (Malik, 2009).

(15)
(16)

3 THEORY AND HYPOTHESES

In the previous section, the most important strategies that pharmaceutical firms undertake in order to produce new drugs have been discussed. They can increase or reorganize their internal R&D efforts, engage in product or geographical expansion through horizontal mergers, or diversify through acquisitions. However, there is a difference between increasing sales by bringing already developed drugs to the market or discovering new drugs and thus increasing your research productivity.

The success of acquisitions in generating actual innovation (thus discovering new drugs) depends on the success of the integration of both firms post-acquisition. Acquisition synergies will only form when firms are fully integrated. Then the employees will be able to learn from each other, gather new insights and develop new drugs. The theoretical perspective that explains the process of learning from R&D is the organizational learning theory. Insights of the resource-based view (Barney, 1991) and the absorptive capacity on knowledge transfer (Cohen & Levinthal, 1990) are used to explain the process of gaining knowledge from acquisitions. These will be explained below.

3.1 Theoretical perspective

3.1.1 Resource-based view

The resource-based view (RBV) explains firm performance by arguing that a firm’s resources can be a source of sustainable competitive advantage. It is based on the insights of transaction cost economics (Coase, 1937) and growth theory of the firm (Penrose, 1959). In contrast to previous theories of competitive advantage (Porter, 1985), this view does not assume firm homogeneity and mobile resources, but firm heterogeneity and immobile resources (Barney, 1991).

Firm resources include all assets, capabilities, knowledge, etc. controlled by a firm that enable the firm to conceive and implement strategies that improve its efficiency and effectiveness. These can be physical capital resources, human capital resources, or organizational capital resources (Barney, 1991; Wernerfelt, 1984).

(17)

perfectly duplicate the resource (imperfectly imitable); or achieve the same goal with a different resource (non-substitutable).

When the condition of imperfect imitability is uncertain, the firm can use so-called isolating mechanisms. Isolating mechanisms ensure that the competition cannot duplicate the asset after the transaction. Among many, two isolating mechanisms in particular are relevant for acquisitions in the pharmaceutical industry.

First, patents protect inimitability. They make duplication, without a license or consent, illegal. Second, there is causal ambiguity. Causal ambiguity means that no one really knows what makes the asset a competitive resource. It is the inability of the firm to fully understand the causes of efficiency differences between the firm and its competitors, which limits the imitation by competition (Rumelt, 1997). When a great amount of tacit knowledge transfer is involved, it is unlikely that even employees know the exact source of the synergy between acquirer and target. Even if an employee would move to a competitor, the competitive advantage cannot be imitated.

3.1.2 Knowledge-based view

The knowledge-based view is a specific aspect of the resource-based view that considers knowledge as the most strategically significant resource of a firm. This perspective is therefore insightful for the transfer of knowledge in the post-acquisition integration process.

(18)

limited common goals, which are often temporary. When long-term goals are pursued, acquisitions are preferred (Child, Faulkner, & Tallman, 2005).

In practice, collaborations are often first shaped as alliances. Once the alliance has achieved its temporary goals, the long-term strategy of the alliance is assessed. The alliance is then either discontinued or acquired by one of the participating firms. In the pharmaceutical industry, innovations are increasingly the result of combined capabilities with other firms or (academic) institutions, rather than produced by individual firms (James, 2002).

Examples of such alliances include the partnership between AstraZeneca and Imperial College that collaborated on potential new cancer drugs, or the collaboration between Merck and Vaccine and Gene Therapy Institute of Florida, that aimed to discover and validate drugs for HIV treatment (Khanna, 2012). This process is consistent with the firm specific advantages identified in section 2.1, Figure 1. Collaborations between large pharmaceutical firms and smaller research organizations allow both firms to use their competitive advantage.

3.1.3 Organizational learning

For acquisitions to lead to innovation, it is important that the R&D knowledge is transferred properly. Knowledge is embedded within the firm, carried by the organizational culture, its routines and its individual employees (Grant, 1996; Kogut & Zander, 1992). Organizational learning theory says that successful knowledge transfer depends on at least two factors. The absorptive capacity of the acquirer and the similarity or ‘compatibility’ of the knowledge to be transferred (Argote & Ingram, 2000).

R&D is a firm-specific asset that is the result of a cumulative pattern of activity. Moreover, additional R&D efforts increasingly benefit from an existing R&D knowledge base (Dierickx & Cool, 1989). It is therefore an iterative process. Cohen and Levinthal (1990) have also found the importance of having an existing knowledge base. They described it as “the ability of a firm to recognize the value of new, external information, assimilate it, and apply it to commercial ends”, and have labelled it a firm's absorptive capacity. For the integration process to be successful, the firms must have sufficient absorptive capacity. It thus reflects the ability of the firm to embed the acquired knowledge into its own R&D system and translate it into innovations.

(19)

in. Knowledge is transferred between individuals (Singh, 2008), as well as organizations. Therefore, the cultural differences in case of cross-border acquisitions are added to the model. This reasoning is captured in the conceptual model (Figure 2), which is described in detail in section 3.2.

3.2 Conceptual model and hypotheses

Figure 2: Conceptual model

The conceptual model is shown in Figure 2. The effect that acquisitions have on innovation is influenced by two factors as described by organizational learning theory. These factors are the similarity between target and acquirer and the level of absorptive capacity, and more specifically the knowledge base that the acquirer has developed. The similarity is further divided into country level similarity, industry level similarity and firm level similarity. On the country level, cultural differences between the countries of the target and acquirer are examined. The industry level similarity closely resembles the factors that Argote and Ingram (2000) have identified, including the source of the knowledge, the nature of task, and any industry networks the firm may be active in. On the firm level, relative size is examined. This gives insight in how easily the target can be integrated without too many disruptions to organizational routines.

(20)

3.2.1 Firm level similarity: relative size

The similarity between acquirer and target on firm level can be approximated by examining the relative size between the two firms. There are two ways in which the relative size of both firms influences the knowledge transfer. First, large amounts of newly acquired knowledge may prove disruptive of existing organizational routines. Small amounts of knowledge only need minor adjustments to be absorbed and are not likely to be disruptive (Ahuja & Katila, 2001). This is one of the reasons why large horizontal mergers often fail to generate more R&D productivity (Comanor & Scherer, 2013; Patricia M. Danzon et al., 2007; Ornaghi, 2009).

Second, group dynamics play an important role in the integration process. Increased interaction between scientists of both firms leads to more productive relationships where knowledge is more easily transferred (Shibayama et al., 2008). Scientists will interact more easily when individuals are embedded in a larger group, than if two groups must merge. In the case of two distinct groups from the acquired and acquiring firm, there is an increased chance on social conflict as individuals identify themselves as ‘us vs. them’ (Vaara et al., 2012). Following social identity theory, group identification is stronger when groups are very distinct and group members experience the other group as a threat, which for the employees of the acquired firm, could well be the case. Second, when in-group members are aware of the out-group, this reinforces homogeneity within the group (Ashforth & Mael, 1989). As the relative size of the acquired firm increases, the groups become more distinct and aware of each other. This increases intergroup competition (Ashforth & Mael, 1989).

Hypothesis 1: Relative acquisition size has a negative effect on post-acquisition innovation.

3.2.2 Industry level similarity

(21)

when businesses share centralized functions (such as finance or marketing). Both arguments will be explained more in depth.

3.2.2.1 ECONOMIES OF SCALE IN R&D

From the organizational learning perspective in section 3.1.3 it is expected that greater similarity between acquirer and target has a positive effect on innovation. As Argote and Ingram (2000) explain, similarity in the source of the knowledge, the nature of task, and any industry networks the firm may be active, helps knowledge transfer in the post-acquisition integration process. Similar knowledge is easily absorbed into the knowledge base. On the firm level, acquiring similar firms may allow the acquirer to benefit from economies of scale.

Also, recalling the resource-based view, a resource is valuable when it enables a firm to conceive or implement strategies that improve its efficiency and effectiveness (Barney, 1991). For R&D to be valuable, it needs to be substantive, as R&D in the pharmaceutical industry is characterized by economies of scale (Graves & Langowitz, 1993). For example, more R&D expenditures may allow a firm to employ significantly more specialized resources (Cockburn & Henderson, 2001).

3.2.2.2 ECONOMIES OF SCOPE IN R&D

However, from past horizontal mergers as described in section 2.2, it is known that innovation mainly benefits when the target is different and there are diversification and scope economies.

Scope economies result from the use of a diverse portfolio of research projects. An important aspect of diversification is the distinction between related and unrelated diversification. Unrelated diversification benefits from scope economies resulting from shared centralized functions such as finance or marketing. Related diversification has the potential to create additional synergies from interrelationships between research projects (Panzar & Willig, 1981). These synergies could be the source of innovation.

(22)

This is why biotech acquisitions are increasingly popular, because there are some similarities, but is also allows for significant economies of scope (Malik, 2009). As seen in the description of the R&D process in Figure 1, section 2.1, the strengths of biotech firms are target identification and screening. Pharmaceutical firms have a firm specific advantage in later stages of development. When both firms cooperate, they can create synergies. Pharmaceutical firms produce chemical medicines, whereas biotech firms produce biological medicines. The approaches to target identification of both firms are sufficiently similar. Scope economies can therefore reasonably be expected. Hence to achieve scope economies in R&D, there must be some level of similarity between target and acquirer, but not too much. There is an inverted u-shaped relation between similarity and innovation.

3.2.2.3 AN INVERTED U-SHAPED RELATION

The inverted u-shaped relation between similarity and innovation results from the balance between economies of scale and economies of scope. As the acquirer and target are too different, the knowledge transfer will be too difficult and the acquisition may lead to organizational efficiencies but not to innovation. As the acquirer and target are almost identical, adding more of the same knowledge will have little effect. This is caused by the difference between scale economies for drug discovery and drug development. Whereas drug discovery is characterized both by scale and scope economies, drug development only benefits from scope economies (Cockburn & Henderson, 2001).

The idea that drug development only benefits from scope economies and not scale economies seems odd. Drug development includes many stages of clinical testing, increasing in size as the drug continues further down the process (Paul et al., 2010). This implies that possible scale economies are present in these stages. However, the benefit of allocated specialized resources diminishes as large clinical trials are no longer conducted in-house, but are carried out at hospitals by physicians.

As the R&D base grows the organization becomes more inert, requiring more formalization and bureaucratic controls. This suppresses creative processes (Hagedoorn & Wang, 2012). Furthermore, good scientists are often promoted to supervisory roles in recognition of their research, but this lowers their R&D productivity (Cockburn & Henderson, 2001).

(23)

innovation should thrive when acquirer and target are similar, but not identical, the following hypothesis is stated:

Hypothesis 2: There is an inverted u-shaped relation between industry similarity of acquirer and target and post-acquisition innovation.

3.2.3 Country level similarity: cultural distance

Cultural differences can increase the potential knowledge transfer, as the complementarity and thus value of skills and routines is higher for organizations that are culturally different (Morosini et al., 1998). However, organizational cultural differences can weaken the ability to identify, transfer and implement potentially useful knowledge (Vaara et al., 2012). As the difference between cultures increases, the probability of social conflict occurring increases, which directly harms knowledge transfer. It erodes the social cohesion and trust between employees that is needed for successful knowledge transfer (Vaara et al., 2012). Differences in organizational practices, conflict resolution strategies, management styles and human resource practices further increase the difficulty of knowledge transfer (Dikova & Rao Sahib, 2013).

Aside from organizational cultural differences, cultural differences can also manifest on the national level. Greater national cultural distance is associated with a higher cost of transaction due to information costs and the difficulty of transferring competencies and skills (Shenkar, 2001). In that sense, national cultural differences operate in similar ways as organizational cultural differences, leading to issues such as miscommunication, lack of legitimacy, adaptation costs, and unfamiliarity hazards (Bauer & Matzler, 2014). National cultural distance is only present in cross-border acquisitions.

Hypothesis 3: For cross-border acquisitions, the cultural distance between target and acquirer has a negative effect on post-acquisition innovation.

(24)

contributing factor to this decrease in cultural gap (Dikova & Rao Sahib, 2013; Shenkar, 2001).

3.2.4 Absorptive capacity: acquisition experience

Acquisition experience allows acquirers to deal more successfully and efficiently with the integration and management challenges of a particular acquisition. For example, the acquirer can have developed routines on how to efficiently use its resources, how to solve administrative problems, or has implemented dedicated integration teams that facilitate the integration process (Dikova & Rao Sahib, 2013). These skills are developed both in domestic and cross-border acquisitions.

But for cross-border acquisitions, one other aspect is important. Firms that have experience in cross-border acquisitions have had the opportunity to develop a cultural sensitivity in resolving organizational incompatibilities. This cultural sensitivity can contribute to overcome the cultural distance (Dikova & Rao Sahib, 2013; Shenkar, 2001).

Hypothesis 4a: Acquisition experience has a positive effect on post-acquisition innovation.

Hypothesis 4b: Cross-border acquisition experience has a positive effect on post-acquisition innovation.

(25)
(26)

4 METHODOLOGY

4.1 Data collection

4.1.1 Acquisition data

Bureau van Dijk has two databases; Orbis, which includes financial data, and Zephyr, which includes data on mergers and acquisitions. Zephyr has information on over 1.2 million worldwide deals. It includes mergers and acquisitions, initial private offerings, private equity deals, venture capital deals, and rumours thereof.

Included in the sample are all acquisitions undertaken between 01/01/2006 and 31/12/2012 by firms that are characterized under the NAICS 2012 industry classification (North American Industry Classification System), as 325412: “Pharmaceutical Preparation Manufacturing”, which are all producers of pharmaceuticals. Acquisitions are defined as deals where the acquirer owns less than fifty percent of the target’s voting shares before the takeover, and the acquirer increases its ownership to fifty percent or more as a result of the takeover. The sample therefore includes full acquisitions and partial acquisitions. These partial acquisitions can, for example, be equity alliances that are now fully acquired.

The period 2006 to 2012 is chosen for two reasons. First, research observing the stagnant research productivity in the sector, and the upcoming expiry of major blockbuster patents, has been written in the second half of the 2000s, using data from 1980/1990 up to 2003 (Cohen, 2005; Pammolli et al., 2011). It therefore includes data on the first two merger waves (Grabowski & Kyle, 2012), but is silent on any activity in the industry after 2003.

The second reason is of a practical nature. Orbis includes firm-level financial data as of 2006 up to today. The acquisition sample is limited at 2012 to allow a window of two years (2013 and 2014) in which firms can integrate the acquisition. After twelve months to two years, firms are expected to have completed integration and the first patents are filed (Grimpe, 2007). This assumption is further explained in section 4.3.1, when the dependent variable is discussed.

(27)

The selection yields 445 acquisitions made by 165 firms. This implies that several firms have made multiple acquisitions in the specified time period. The frequency of the number of acquisitions is shown in Figure 3. Figure 4 shows the distribution of firm size in the sample. Roughly seventy percent of firms are relatively small with sales lower than $5 billion.

Figure 3: Frequency of number of acquisitions Figure 4: Distribution of firm size

4.1.2 Patent data

Bureau van Dijk’s Orbis company information database includes data on patent applications for the acquirers. All information on the patents listed in Orbis is sourced from the EPO Worldwide Patent Statistical Database (PATSTAT). PATSTAT is one of the most extensive global patent databases available (European Patent Office, 2014). Patent applications were included when they fulfil the following conditions;

1) They were filed between 01/01/2006 and 31/12/2014;

2) The current owner includes a firm in the pharmaceutical industry with a NAICS classification of 325412 “Pharmaceutical Preparation Manufacturing”;

3) They were classified under the International Patent Classification (IPC); 4) They were granted.

(28)

4.1.3 Financial data

To this selection, financial data was added of both the acquirer and the target, which was derived from Orbis, which includes data on over 175 million private companies worldwide, and Standard & Poor's COMPUSTAT database. As summarized in Table 1, Orbis had data for 310 of the 445 acquisitions. Using the data from Compustat, the sample increased to 403. Manual search in financial reports and press releases on acquisitions brings the sample to 445.

4.1.4 Cultural distance data

For each acquisition, data on the acquirer and target nations’ cultural differences were added using the Dimension Data Matrix obtained from the website of Geert Hofstede (Hofstede, 2010), which includes scores on six cultural dimensions for 66 countries. If the acquisition is domestic, the cultural distance is zero. If there was no country code registered for either of the firms, the location of the headquarters was used, or a more specific location derived from news articles that elaborated on the location of the acquired operations. For 36 targets and 6 acquirers, the country identifier was added manually in this way. For two targets, the aggregate cultural distance score of ‘Arab countries’ was used, as these acquisitions covered the operations in Lebanon, Jordan, Syria, Libya and Yemen.

4.2 Sample

Table 1: SAMPLE SELECTION

Step Database Variables Observations left (1) Zephyr Year; Deal value; Market capitalization; (Cross-border)

Acquisition experience; Industry similarity; Cross-border dummy

2,852

(2) PATSTAT Patents; Drug Patents 999 (-1,853)

(3) Orbis Sales; R&D expenditure; R&D intensity 310 (-689)

(4) Compustat Sales; Market capitalization; R&D expenditure; R&D intensity

403 (+93)

(5) Manual search Deal value; Market capitalization; Sales; R&D expenditure 445 (+42)

(6) Hofstede Dimension Data Matrix

Cultural distance 445 (+0)

(29)

added from a manual search in company annual reports and deal announcements. The sample includes 272 domestic and 173 cross-border acquisitions. This distinction may give insight in whether acquisition experience as such is important in the integration process, or specifically the skills that are required to overcome cultural distance after the acquisition.

4.3 Measures

All measures explained below can be seen in terms of the timeframe of Figure 5. Each acquisition is seen as a focal acquisition. From the focal year, patents are measured up to two years following the acquisition date. Acquisition experience is measured up to five years prior to the acquisition.

Figure 5: Timeline showing focal acquisition in time

4.3.1 Dependent variable 4.3.1.1 PATENTS

The dependent variable employed to measure post-acquisition innovation is the count of all patents granted to a firm in the two years after the focal acquisition. Ideally, post-acquisition innovation is measured using newly introduced products (Grimpe, 2007; Hagedoorn & Cloodt, 2003; Prabhu et al., 2005). However, the average cycle time of clinical trials is six to eight years (Pammolli et al., 2011), and the total cycle time of both drug discovery and subsequent drug development (including clinical trials) takes 13.5 years on average (Paul et al., 2010). These large time lags are undesirable in this thesis. Therefore the choice has been made to use the count of granted patents.

(30)

2001; Jeon et al., 2015; Ornaghi, 2009). Two years after acquisition is chosen for three reasons.

First, the average acquisition frequency in the sample is one acquisition every 2.6 years (445 acquisitions made by 165 firms in 7 years). Limiting the time frame to two years increases the chance that the patents are indeed a consequence of the focal acquisition, and the results are not confounded by another acquisition made in this period.

The second reason concerns the timing of patent application in the drug discovery process. On average, the discovery phase takes 5.5 years, and the subsequent development phase (which includes all clinical trials) another 8 years (Paul et al., 2010). Somewhere in those first 5.5 years, the patent application is filed for. Generally, firms have the incentive to file the patent application as late as possible, to benefit from the effective patent life as long as possible. Patents are valid for twenty years, but the effective patent life during which a firm can benefit from market exclusivity is shorter. Competitors have often introduced a different, but better medicine in the mean time, which reduces the effects of the temporary monopoly (Cohen, 2005). Research has shown that a patent application may be filed up to 4.7 years prior to beginning phase I testing, without sacrificing market exclusivity and still protect the intellectual property (Dunn, 2011). This means that on average, the patent application is filed for in the first year. Taking two years allows for a slightly longer discovery phase than average.

Third, using two years does not limit the sample further, as data from 2013 and 2014 can be used for the acquisitions made in 2011 and 2012.

4.3.1.2 VALIDITY OF THE DEPENDENT VARIABLE

The validity of patents as a measure of innovation can be compromised by two factors; a type I and type II error. A type I error is an incorrect rejection of a true null hypothesis, and is also called a ‘false positive’. In this case, it occurs when not all patents lead to innovation. When the acquirer has applied for new patents, but these are not the result of the acquisition, they will nevertheless be included in the dependent variable. Hence when a type I error occurs, an effect is detected that is not present.

(31)

Previous research has shown that in the pharmaceutical industry, the validity of patent count as a measure of innovation is relatively high (Henderson & Cockburn, 1996; Penner-Hahn & Shaver, 2005).

The argument that not all patented inventions result in innovative new drugs (type I error) rests on the incentives for companies to delay patent filings as long as possible (to lengthen the effective patent life during marketing) and to file multiple patents around the same fundamental technology to provide maximal protection against competitors (Cohen, 2005). Nevertheless, there is a high correlation between patents and new products in the pharmaceutical industry (Comanor & Scherer, 1969; Hagedoorn & Cloodt, 2003). According to Hagedoorn & Cloodt (2003, p. 1368), “… patents can be an appropriate indicator in the context of many high-tech sectors”, which is the case when “[…] an industry or its average company is characterized by a high R&D intensity, high patenting intensity and a high ratio for new product introduction.” The pharmaceutical industry is characterized as such.

The concern that not all innovations are patented (type II error) is resolved by the nature of the pharmaceutical industry, in which patents give reasonable protection of proprietary knowledge (Penner-Hahn & Shaver, 2005). The pharmaceutical market is characterized by high R&D costs, and a low cost of imitation. The average costs of launching one new drug, including discovery and development phases, amounts to $1,778 million (Paul et al., 2010). Because of the low cost of imitation, the inventor will only have a short period in which he can benefit from the first mover advantage and charge a price that may be high enough to recoup the high level of R&D investment. However, this period is far too short for the full investment to be recovered, and thus the innovation will not take place. The costs of R&D cannot be recovered by sufficient sales unless the knowledge is protected within the firm by patents, and competition cannot copy the drug. Thus, intellectual property rights are present in the market for pharmaceuticals to resolve the market failure of sunk R&D costs (Hall, 2007). Empirical research also shows that patents are among the most important means to appropriate returns to innovation in the pharmaceutical industry (Arora et al., 2008; Hall, 2007).

4.3.1.3 PATENTS CLASSIFIED UNDER A61K

(32)

To increase the construct validity of this variable, a second dependent variable is added. This variable uses granted patents classified in International Patent Classification (IPC) under A61K. Patents classified under IPC as A61K include “Preparations for medical, dental, or toilet purposes” (World Intellectual Property Organization, 2015). A61K8/* is excluded as these patents are related to the development of cosmetics (such as skin cream, shampoos or make-up). More specifically, A61K includes all ‘drug or other biological compositions’ that are capable of curing or limiting diseases, influence physiological body functions (for example growth promotors and birth controls), and diagnosing by means of in vivo testing. In vivo testing means that testing is done within the body, such as using an X-ray or skin patches, as opposed to in vitro testing where for example blood or urine is taken out of the body for examination. These patents are therefore strictly related to the development of new pharmaceutical drugs.

4.3.2 Independent variables 4.3.2.1 RELATIVE SIZE

Relative size is measured as the ratio between the deal value of the acquisition and the market capitalization of the acquirer. The deal value represents the price paid for all the shares of the target. Market capitalization of the acquirer is the total market value of all of a company's outstanding shares, and can therefore be compared to the deal value of the target. Market capitalization rather than sales is used, as it includes the value of acquired knowledge in the intangible assets. In 49 observations the deal value or market capitalization was unknown. In those cases the ratio of sales was used.

4.3.2.2 INDUSTRY SIMILARITY

Two dummy variables are added that measure the similarity between the acquiring and target firms. These variables are constructed based on the NAICS Industry Classification.

The first dummy measures whether the acquirer and target operate in an identical industry, using all six digits of the NAICS Industry Classification. Thus, it will be coded 1 when the industry classification of the acquired firm exactly corresponds to the industry classification of the acquiring firm (i.e. both firms are classified under 325412 “Pharmaceutical Preparation Manufacturing”), 0 otherwise.

(33)

four digits of the NAICS Industry Classification are 3254, which stand for all firms in “Pharmaceutical and Medicine Manufacturing”. This includes 325411 “Medicinal and Botanical Manufacturing”, 325412 “Pharmaceutical Preparation Manufacturing”, 325413 “In-Vitro Diagnostic Substance Manufacturing” and 325414 “Biological Product (except Diagnostic) Manufacturing”.

4.3.2.3 CROSS-BORDER ACQUISITION DUMMY

A cross-border dummy is added and coded 1 when the acquisition is cross-border, and 0 when the acquisition is domestic. An interaction variable between the cultural distance index and the cross-border dummy is entered in the analysis to moderate the effect between relative size of the acquisition and post-acquisition innovation. Cultural distance only enters the equation when the acquisition is cross-border.

4.3.2.4 CULTURAL DISTANCE

The Kogut & Singh index (1988) will be used to measure the national cultural distance between the countries of the acquiring and acquired firm (Ahuja & Katila, 2001; Prabhu et al., 2005). This index is calculated as follows:

𝐶𝐷𝑗 = 𝐼𝑖𝑗− 𝐼𝑖𝑘 2

/ 𝑉𝑖 6

𝑖=1 / 6 ( 1 )

where Iij (Iik) is the score of dimension i (i = 1,…,6) for country j(k) and Vi is the variance of dimension i. The dimensions included are power distance, individualism vs. collectivism, uncertainty avoidance, masculinity vs. femininity, long-term orientation vs. short-term orientation and indulgence vs. restraint. The index will be zero for domestic acquisitions and will be positive for cross-border acquisitions.

4.3.2.5 ACQUISITION EXPERIENCE

(34)

4.3.2.6 SHARE OF CROSS-BORDER ACQUISITION EXPERIENCE

The share of cross-border acquisition experience is added to account for any experience the firm has built up in past cross-border acquisitions. The share of cross-border acquisition experience is calculated as the count of cross-border acquisitions divided by the total count of acquisitions completed by the acquirer in the three years before the focal acquisition. As argued in section 3.2.4, firms that have experience in cross-border acquisitions have had the opportunity to develop a cultural sensitivity in resolving organizational incompatibilities. This experience contributes to their absorptive capacity.

Table 2: SUMMARY OF VARIABLES

Name Type Measure

Patents Dependent The count of all patents granted to a firm in the 2 years after the focal acquisition.

Drug patents Dependent The count of the patents classified as A61K*, excluding A61K8/* granted to a firm in the 2 years after the focal acquisition.

Relative size Independent The ratio of deal value of the acquisition and market capitalization of the acquirer (resp. sales ratio).

Acquisition experience Independent The count of all acquisitions completed by the acquirer in the 3 years before the focal acquisition.

Share of cross-border acquisition experience

Independent The count of cross-border acquisitions divided by total acquisitions completed by the acquirer in the 3 years before the focal acquisition. Industry similarity

dummy

Independent A dummy coded 1 when the industry is similar, but not identical, and 0 otherwise.

Identical industry dummy

Independent A dummy coded 1 when the industry is identical, and 0 otherwise. Cross-border dummy Independent A dummy coded 1 when the acquisition is cross-border, and 0 when

the acquisition is domestic. Cultural distance

(interaction)

Independent The interaction variable of the cross-border dummy and the cultural distance score developed by Kogut & Singh (1988)

R&D expenditure Control The level of R&D expenditure of the acquirer

R&D intensity Control The ratio of R&D expenditures to sales of the acquirer

4.3.3 Control variables

(35)

2009; Prabhu et al., 2005). These two variables control for any effects of increased internal R&D efforts.

4.4 Model specification

4.4.1 Distribution regression analysis

A standard ordinary least squares (OLS) regression will be performed to get a first impression of the data. Furthermore, a Poisson regression model is used to model the discrete dependent variable. This methodology is common in patent output studies (Ahuja & Katila, 2001; Graves & Langowitz, 1993; Henderson & Cockburn, 1996; Penner-Hahn & Shaver, 2005; Prabhu et al., 2005). The Poisson distribution assumes that the mean and variance are the same. It might be the case that the dependent variable is not characterized by a Poisson distribution, and exhibits a greater variance than might be expected in a Poisson distribution. When a test of overdispersion shows that the variance is greater than the mean, using a Poisson regression model is inappropriate.

This overdispersion in the variance of the dependent variable may have several causes, the most common of which is a missing explanatory variable. Should overdispersion occur, a negative binomial regression is used to overcome bias in the standard errors. The negative binomial regression method relaxes the assumption of the Poisson distribution by allowing the mean and variance to be different.

4.4.2 Empirical model

log𝑒 𝑃

0≤𝑡<2 = 𝛽𝑜+ 𝛽1𝑅𝑆𝑡+ 𝛽2−3≤𝑡<0𝐴𝐸𝑖+ 𝛽3𝐶𝐵𝐴𝐸𝑖+ 𝛽4𝐼𝐼𝑡+ 𝛽5𝐼𝑆𝑡 + 𝛽6𝐶𝐵𝑡+ 𝛽7(𝐶𝐵 × 𝐶𝐷)𝑡+ 𝐶𝑖+ 𝜀

( 2 )

(36)

5 RESULTS

5.1 Descriptive statistics

Table 3 shows the means, standard deviations and intercorrelations for all variables. There are some significant correlations that result from the research design. The interaction variable of the cross-border dummy and cultural distance is correlated with the cross-border dummy (correlation coefficient 0.684). Also, the share of cross-border acquisition experience is correlated with total acquisition experience (0.470), the cross-border dummy (0.454) and the interaction variable of the cross-border dummy and cultural distance (0.343). These correlations make economic sense.

There is a positive and significant correlation between R&D expenditure and acquisition experience. R&D expenditure and acquisition experience are also positively correlated (0.371 and 0.483, respectively) with the dependent variable patents.

A table with the correlations of the regression using the dependent variable drug patents is included in the appendix. There are only minor differences between the correlations of patents and drug patents.

The distribution of the dependent variable patents is shown in Figure 6. The distribution is highly skewed. The majority of acquirers have added less than 150 patents in the two years following the acquisition, but there are some acquirers with many patents. There are 91 acquisitions made by 41 firms after which 0 patents have been granted. The distribution of the variable drug patents, being very similar to patents, can be found in the appendix.

The distribution of the independent variable relative size is shown in Figure 7. Relative size has a similar distribution to patents, with most targets being smaller than fifty percent of the acquirer size. Some acquisitions involve targets that are larger than the acquirer. There is one acquisition with the relative size zero. The sales ratio was used and the target sales for that year were zero.

(37)

THE EFFECT OF ACQUISITIONS ON INNOVATION IN THE PHARMACEUTICAL INDUSTRY

Table 3: DESCRIPTIVE STATISTICS AND INTERCORRELATIONS (N=445)

Variables Mean S.D. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (1) Patents 233.166 608.691 1.000 (2) Relative size 0.140 0.373 0.165 1.000 (0.000) (3) Identical industry 0.407 0.492 -0.135 0.026 1.000 (0.004) (0.585) (4) Similar industry 0.070 0.255 -0.029 -0.058 -0.227 1.000 (0.536) (0.219) (0.000) (5) Cross-border dummy 0.389 0.488 0.066 -0.084 0.015 0.017 1.000 (0.165) (0.075) (0.747) (0.718)

(6) Cultural Distance (interaction) 0.653 1.199 -0.017 -0.064 0.046 0.007 0.684 1.000 (0.715) (0.175) (0.334) (0.879) (0.000)

(7) Acquisition experience (3yr) 2.261 2.778 0.483 -0.010 -0.081 0.009 0.264 0.124 1.000 (0.000) (0.829) (0.088) (0.845) (0.000) (0.009)

(8) Share of cross-border acquisition experience (3yr)

(38)

Figure 6: The distribution of the number of patents Figure 7: The distribution of relative acquisition size

5.2 Regression results

5.2.1 Model fit

Table 4 shows the regression results for the dependent variable patents, using an OLS, Poisson and negative binomial regression, respectively. The goodness of fit of the Poisson regression is tested using a Pearson chi-squared test. It tests the null hypothesis that the distribution of a sample is equal to a particular theoretical distribution. In this case, the hypothesis states that the variance is equal to the mean, and the data thus have a Poisson distribution. The results from the Pearson chi-squared test reject the hypothesis that the variance is equal to the mean. Instead, the variance is greater than the mean and the data show overdispersion. Hence, the Poisson model is not appropriate for this data. Figure 10 (appendix) compares the OLS, Poisson and negative binomial model specifications to the observed distribution of the dependent variable.

(39)

Table 4: REGRESSION RESULTS FOR DEPENDENT VARIABLE PATENTS

Variables (1) OLS (2) Poisson (3) Neg. Bin. (4) ln alpha

Relative size 290.6*** 0.303*** 0.0552 (65.77) (0.00417) (0.236) Identical industry -119.1** -0.537*** -0.549*** (51.59) (0.00768) (0.206) Similar industry -114.1 -0.531*** 0.508 (98.66) (0.0145) (0.388) Cross-border dummy 1.114 0.0196*** -0.0128 (55.14) (0.00521) (0.248) Cultural Distance (interaction) -27.27 -0.0630*** -0.0220 (21.98) (0.00312) (0.0801) Acquisition experience (3yr) 88.33*** 0.170*** 0.125***

(11.14) (0.000857) (0.0333) Share of cross-border

acquisition experience (3yr)

-38.91 0.280*** 0.325 (76.74) (0.00985) (0.294) R&D expenditure 0.0478*** 0.000179*** 0.000416*** (0.0113) (9.80e-07) (5.62e-05) R&D intensity -4.579 -0.000849 -0.0228 (10.75) (0.00172) (0.0390) Constant 8.364 4.426*** 3.961*** 1.340*** (43.70) (0.00707) (0.177) (0.0618) Observations 445 445 445 445 R-squared 0.307

Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors in parentheses. Column (1) shows the results for the OLS regression. Column (2) shows the Poisson regression. Column (3) shows the negative binomial regression. Column (4) shows the goodness of fit of the negative binomial model.

5.2.2 Results for dependent variable patents

Hypothesis 1 stated that relative acquisition size has a negative effect on post-acquisition innovation. The results in Table 4 do not support this hypothesis. In all model specifications, relative size positively affects post-acquisition innovation. In the negative binomial regression, the coefficient for relative size is insignificant (p>0.1).

The interaction between cultural distance and the cross-border was hypothesized to have a negative effect on post-acquisition innovation. The sign of the coefficient is negative, but the results are insignificant (p>0.1). No conclusions can be drawn for hypothesis 3.

Hypothesis 4a stated a positive effect between acquisition experience and post-acquisition innovation. The results support this hypothesis. The sign is positive and significant on the 1% level.

(40)

The control variable R&D expenditure shows expected results. The coefficient is positive and significant (0.000416, p<0.01). More R&D expenditure thus leads to more patents. The coefficients for R&D intensity are insignificant (p>0.1).

Hypothesis 2 predicts an inverted u-shaped relation between industry similarity and post-acquisition innovation. Table 4 shows a negative and significant coefficient for identical industries (-0.549, p<0.01). If the acquirer and target industries are identical, there is less innovation than when the acquirer and target come from different industries. This is an argument for the economies of scope that can be obtained from diversification. The coefficient for similar, but not identical, industries is positive but insignificant (p>0.1). It can therefore not be concluded that there must be some level of similarity between both firms for post-acquisition innovation to increase.

The marginal effects are analysed to be able to conclude on the inverted u-shaped relation. The results are shown in Figure 8. The predicted number of patents for firms with identical industries is -61, holding all other variables at their means. The predicted number of patents for firms with similar, but not identical, industries is 74, holding all other variables at their means. So on average, two firms in identical industries, generate 61 fewer patents than firms not in identical industries. In contrast, acquisitions with firms from similar industries have a higher innovation of 74 more patents than average. The details are listed in Table 7 in the appendix.

(41)

The marginal effect of being in an identical industry is significant at the 10% level, but the effect for similar industries is insignificant (p>0.1). The results are therefore inconclusive and hypothesis 2 cannot be rejected nor supported. The results do confirm a negative effect on post-acquisition innovation of both firms being in identical industries.

5.2.3 Results for dependent variable drug patents

Table 5 shows the results when the dependent variable has been restricted to include only drug patents classified as A61K. The first three columns show the results for the OLS, Poisson, and the negative binomial regression, respectively. Column 4 shows the likelihood ratio test for the negative binomial model. This is similar to the results using all patents, and is significant (p<0.01). The negative binomial regression model is therefore appropriate.

Acquisition experience shows a negative significant effect on the number of drug patents (-0.0803, p<0.1). Firms that have more acquisition experience generate fewer patents after an acquisition.

For all patents, the effect of being in an identical industry was negative and significant. The effect of being in a similar industry was insignificant. For drug patents, the effect of being in an identical industry is still negative (-0.381) but significant at the 10% level. The effect of being in a similar industry however is now also negative and significant (-1.119, p<0.01).

(42)

Figure 9: Marginal effects of industry similarity on drug patents

Table 5: REGRESSION RESULTS FOR DEPENDENT VARIABLE DRUG PATENTS

Variables (1) OLS (2) Poisson (3) Neg. Bin. (4) ln alpha

Relative size 52.77* 0.267*** -0.135 (29.52) (0.0102) (0.272) Identical industry -23.37 -0.170*** -0.381* (23.15) (0.00955) (0.215) Similar industry -85.91* -0.730*** -1.119*** (44.27) (0.0234) (0.411) Cross-border dummy 28.82 0.189*** 0.162 (24.74) (0.00842) (0.274) Cultural Distance (interaction) -2.654 0.0523*** 0.0679 (9.863) (0.00398) (0.0849) Acquisition experience (3yr) -20.86*** -0.149*** -0.0803* (4.998) (0.00210) (0.0431) Share of cross-border

acquisition experience (3yr)

27.76 0.169*** 0.460 (34.44) (0.0129) (0.312) R&D expenditure 0.0774*** 0.000397*** 0.000562*** (0.00508) (1.54e-06) (5.96e-05) R&D intensity -5.016 -0.0352*** -0.0397 (4.826) (0.00315) (0.0521) Constant 53.95*** 3.973*** 3.441*** 1.450*** (19.61) (0.00946) (0.193) (0.0682) Observations 445 445 445 445 R-squared 0.369

(43)

5.2.4 Sensitivity analysis

Additional analyses have been performed to test the sensitivity of the results in Table 4 and Table 5. These analyses concern possible effects due to firm size, having multiple acquisitions by the same firm in the sample and year effects. Robustness checks have been performed to test the robustness of the acquisition experience measure.

5.2.4.1 SIZE EFFECTS

The results may be driven by a small number of very large firms. The pharmaceutical industry is characterized by some very large firms. The distribution is shown in Figure 12 in the appendix. These firms have a large R&D base and more frequently acquire other firms (Table 9 and Figure 13 in the appendix).

There are two options to analyse this effect. First, clustered standard errors can be used when the sample is clustered on the firm level. Second, the sample can be split into two separate subsamples. The results for the regressions using clustered standard errors are presented in Table 10 (appendix). There are only minor differences between the analysis with regular and clustered standard errors. Clustering appears not to be appropriate because the observations for different firms are not homogeneous across firms. So a single firm is not representative of the population.

The other option is to stratify the sample into two subgroups, namely one with small firms and one with large firms. The results of this analysis are shown in Table 11 and Table 13 in the appendix. The analysis has been performed twice. The first analysis uses the mean of sales plus one standard deviation as a cut off point, which is approximately $25 billion. Eight firms (5%) have sales more than $25 billion. These are Merck & Co., Pfizer, Novartis, Johnson & Johnson, GlaxoSmithKline, AstraZeneca, Basf AG and Abbott Laboratories. The second analysis uses the mean of sales (roughly $10 billion).

From Table 11 it can be concluded that for the eight very large firms only the R&D intensity has an effect on the innovation after an acquisition has been made (p<0.05). However, for the other 95% of firms (sales lower than $25 billion), there is a negative significant effect of relative acquisition size on post-acquisition innovation (-0.643, p<0.1). This confirms hypothesis 1 when Merck & Co., Pfizer, Novartis, Johnson & Johnson, GlaxoSmithKline, AstraZeneca, Basf AG and Abbott Laboratories are treated as outliers and excluded from the sample.

Referenties

GERELATEERDE DOCUMENTEN

RA: Rheumatoid arthritis; ESR: Erythrocyte sedimentation rate; CRP: C-reactive protein; DAS28: 28-joint Disease Activity Score; TJC28: 28-tender joint count; SJC28: 28-swollen

Want het uiterlijke heeft zijn begrip en betekenis niet meer in zich en op zichzelf, zoals bij de klassieke kunstuiting, maar in het gevoel (‘Gemut’).. En het gevoel vindt zijn

This special issue is based on selected papers and extended papers of the Fifth International Conference on eHealth, Telemedicine, and Social Medicine, held on February 2013, in

The aim of this part is to create a 3D surface from camera acquired images of the breast phantom and localize markers placed on the phantom to later assist in

How does the organizational cultural distance between two firms influence the number of layoffs after M&amp;A’s and is this relation moderated by the hostility

Future research, exploring the moderating influence of the target firm’s alliance management capabilities on the relationship between acquisitions with or without

Neither, is the change in product development from before to after the acquisition influenced by organizational context factors prior to the acquisition, such as the

For the target firms coming from country with stronger investor protection than the acquirers, greater national cultural distance between the acquirer and target