University of Amsterdam BSc Economics & Business Economics & Finance
The effect of mergers and acquisitions on the
innovation of U.S. pharmaceutical firms.
Author: David Meijer
Student number: 10610820
Supervisor: Drs. P. V. Trietsch Number of ECs: 12
Abstract
The aim of this paper is to investigate the effect of mergers and acquisitions on innovation. The research is done with a panel data set consisting of U.S.
pharmaceutical firms in the period from 1995 till 2014. Innovation is measured by R&D intensity, which is defined by R&D expenditure divided by net sales. M&A activity is recorded as multiple dummy variables over time to control for the possible lagged effect of M&A. This study finds that in the short run mergers and acquisitions have a significant positive impact on innovation input but in the long run this effect turned out to be negative (insignificant).
Table of Contents Abstract 2 1. Introduction 4 2. Theoretical Background 5 2.1 Innovation 5 2.1.1 Innovation defined 5 2.1.2 Innovation measured 6
2.2 Mergers and Acquisitions 7
2.3 Relationship between M&A and innovation 8
2.3.1 Negative relationship 8 2.3.2 Positive relationship 9 2.3.3 Lagged effect 11 2.4 Control Variables 11 3. Empirical Research 14 3.1 Data 14 3.1.1 Pharmaceutical industry 14 3.1.2 Dataset 14 3.1.3 Variables 15 3.2 Summary Statistics 16
3.2.1 Robustness and Multicollinearity 17
3.3 Methodology 18
3.3.1 Hausman Test 18
3.3.2 Random-effects model 18
3.3.3 Tests for model propriety 18
3.3.4 Regression equation 19
4. Results 21
4.1 Random-effects model 21
5. Limitations and Drawbacks 22
5.1 Firm Characteristics 23 5.2 Unbalanced Data 23 5.3 Simultaneous Causality 24 6. Conclusion 24 References 25 Appendices 29
1. Introduction
“Acquisitions and innovation are substitutes” stated Hitt et al. back in 1991. They argue that acquisitions negatively influence innovation, determined by the willingness to create new products and processes. But both acquisitions and innovation are indispensable in the current economic environment. Especially in the pharmaceutical sector innovation is of great importance. In 2013 the world leading pharmaceutical firms had R&D expenditures over $112 billion (GlobalData, 2014).
Combined with the technological upswing firms are pushed to innovate to get
an advantage with respect to competitors. Despite the negative link found by Hitt et al., supplemented by Stahl (2010) and Park & Sonenshine (2012), other scientists say that these innovations can be encouraged by mergers and acquisitions. They found that M&A activity positively influences the innovation of a firm, because of compiled synergies. They state that mergers and acquisitions lead to an increase in knowledge causing R&D investments to be more efficient and useful (Prabhu, Chandy and Ellis, 2005).
There is a lot of research done about this relationship but the outcomes stay
divers. Stiebale (2013) investigated the relationship between cross-border
acquisitions and innovation input of the acquirer. Innovation input is measured by research and development expenditure. He used data from German firms in a timeframe from 2002 till 2007 and found a positive association between these two parameters. It is interesting to investigate if this association is still positive in a more longitudinal and recent period.
This paper will focus on the pharmaceutical market in the U.S., because of the
high R&D expenditures and the importance of innovation. Based on the research of Stiebale (2013), this paper has been set up and the following research question has been defined:
“Do mergers and acquisitions influence the innovation input of U.S.
pharmaceutical firms?
In this thesis an answer to this question is formed by the following structure.
Section 2 describes the theoretical background of innovation and mergers and acquisitions. First, the definition and measurement of innovation will be described, followed by the possible relationship between these two variables and ending with a description of the influencing control variables. Per control variable a hypothesis is
formed to declare their influence within the model.
In Section 3 the methodology will be discussed. Panel data of 600
observations, including 64 mergers or acquisitions, from 34 U.S. pharmaceutical firms is obtained over a period from 1995 till 2014 from two databases, namely
Thomson ONE and Wharton Research Database Services (WRDS).There are two
possible models for panel data analysis: the fixed-effects model and the random-effects model. The Hausman test is done to find out which model is suitable and found that the random-effects model is the most appropriate model to consider the effect from M&A on innovation. Besides the Hausman test, there are other tests done to control for proper outcomes.
Thereafter, in Section 4, the results of the research will be exposed and the
limitations and drawbacks, including biases, will be discussed in Section 5. Finally Section 6 contains a conclusion, which will summarize the most valuable findings.
2. Theoretical Background
In this section the definition and measurement instruments of innovation are
explained. Subsequently an overview of previous studies will review the interaction between mergers and acquisitions and innovation of firms. This paragraph ends with specifications of the control variables.
2.1 Innovation
2.1.1 Innovation defined
The main incentive for innovation is the growth in value that a firm can attain if it invests in research and development. This can be broken down into two economic forces (Gilbert, 2015). The first force is the profit that can be gained from a new product of process, depending on the influence of the innovation and the degree to
which it is untouchable for imitators.The second is called the
“escape-the-competition effect”, which is a reduction in “escape-the-competition when a firm can differentiate its product by innovation. This preemption of competition is a strong force that deters rivals from entry of the market (Gilbert, 2015). This reduction in competition can also be achieved by acquiring a target firm with the same technological essence.
by the perception of a new market and/or new service opportunity for a technology-based invention which leads to development, production and marketing tasks striving
for the commercial success of the invention.”Schumpeter (1939) describes
innovation more in a general way as “doing things differently in the realm of
economic life”, while Drucker (1985) labels innovation as “the specific instrument of entrepreneurship.” It is difficult to find a perfect definition for this specific instrument.
2.1.2 Innovation measured
There are four ways to measure innovation. First, R&D expenditures are often used as an indicator of innovation input. R&D affects future output being part of the technological performance of a firm regarding finding new ideas and processes, which will hopefully finally cause new patents and products.
Not all innovations are patented and not all R&D activity leads to patents.
Nonetheless, the second measure, the number of patents, indicating the innovative performance of a firm, are, according to Hagedoorn & Cloodt (2003), generally accepted as one of the most direct indicators of innovation output. The process of a new idea becoming a patent is associated with a difficulty: it takes time. There is a difference between patents pending and patents granted. So it is important to make a distinction between these two indicators of patents count.
Third, next to the number of patents, more and more researchers are using
citations of patents as indicator of innovation. The number of patents does not say anything about the quality of the patents. The number of patent citations does represent the quality of innovation. However, specialists are doubting about this measure because of the objectively opinion people may have when assigning the citations to a patent (Hagedoorn & Cloodt, 2003).
Fourth, patent and citation count can be seen as insufficient, because some
patents do not have an effect on innovation. Therefore, some researchers are using new products as measurement of innovation. They say that this is the most
appropriate measurement for marketing insights (Prabhu, Chandy, & Ellis, 2005).
Hagedoorn & Cloodt (2003) did research about the systematic disparity
between these four indicators and their research suggest that there is no systematic disparity amongst them.
2.2 Mergers and Acquisitions
Acquisitions are a crucial part of the business process of deploying resources into more productive and efficient uses, according to Ahuja & Katila (2001). Kamien & Zang (1990) pose that, except from efficiency gains, another main motive for M&A activity is the increase of market power. According to Stiebale (2013), besides this implementation of efficiency gains, acquisitions are undertaken to get access to target firm’s assets and so the resource base of both firms will change. Zhao (2009) even states that technological innovation, having much influence on the economic activity, is the most important reason for M&A activity. So mergers and acquisitions might influence innovation. In this paper mergers and acquisition are used
interchangeably.
From an economic view, the support of M&A activity on innovation is built on
opportunities to achieve economies of scale and scope in R&D (Cassiman et al., 2005). This innovation by M&A can be accomplished by two channels. The first channel is acquiring an innovative company with certain potential innovations in the current process. The acquiring firm benefits from these potential findings and attains patents after the acquisition and increases hereby innovation output of the unified firm. The other channel is the one where the acquirer collaborates with the target firm to generate innovations (Sevilir & Tian, 2012). M&A’s can give rise to the research performance of the firms, with clear positive effects on consumer welfare (Ornaghi, 2009).
Acquisitions have influence on innovation input as well as on innovation output
(Hitt et al., 1991). They influence quality and quantity of the inventions generated after a merger or acquisition (Valentini, 2012). The outcomes of the studies done about the relationship between M&A and innovation are divers and will be reviewed in the next paragraph.
2.3 Relationship between M&A and innovation
Summary of relevant literature
Name Period Firms Dependent Variable Findings
Hitt et al. (1991) 1970-1986 U.S. firms in 29 industries (191
acquisitions completed) R&D intensity + patent intensity (divided by sales) Negative link Stahl (2010) 1974-2005 562 U.S. public firms Patent count Negative link Ahuja & Katila (2001) 1980-1991 Global, 72 leading firms in
chemical industry
Granted patents No significant link Valentini (2012) 1988-1996 U.S. firms in ‘medical devices’
industry
Patent count Positive link Prabhu, Chandy and
Ellis (2005)*
1988-1997 Random 47 companies of 185 pharmaceutical firms
Number of Phase1 products Positive link Ornaghi (2009)* 1988-2004 Pharmaceutical firms with stock
market value > $1 billion
R&D expenditure + patents Both negative link Park & Sonenshine
(2012) 1989-2008 78 U.S. firms, 47 merged R&D intensity + patents granted Negative link Sevilir & Tian (2012) 1990-2006 381 U.S. public firms in
biotechnology and
pharmaceutical sectors (105314 observations)
Filed patents Positive link Cefis & Marsilic (2015) 1994-2002 Dutch Manufacturing firms,
SMEs (13901 observations) R&D expenditure + percentage turnover of new products/services
Both positive link Hagedoorn & Cloodt
(2003)
1997-1998 1194 international companies (aerospace and defense, computer and office machinery , pharmaceuticals electronics and communications)
R&D expenditure, patents, patent citations, new products
All indicators are useful
Heffernan & Fu (2003)* 2002-2004 1100 British Financial firms Financial, product and process
innovation Positive link Stiebale (2013)* 2002-2007 German firms with up to €500
million annual sales (389 cross-border acquisitions)
R&D expenditure Positive link
*Academic paper is listed on the Tinbergen Journal List Figure 2.1
The table (Figure 2.1) above represents the most relevant previous researches about the effect of mergers and acquisitions on the innovation input and output of a firm.
2.3.1 Negative relationship
Acquisitions have, according to Hitt and his colleagues (1991), a negative influence on innovation, determined by “managerial willingness to allocate resources and champion activities that lead to the development of new products, technologies, and processes consistent with marketplace opportunities.” Ornaghi (2009) states that “cultural disparity” might lead to inefficiency and the disruption of the routines of both firms will reduce innovation outcomes. His research, based on the pharmaceutical sector from 1988 till 2004, confirms this negative impact. He shows the effect of M&A in the same year, the year before and two years before on the R&D intensity. The coefficients of the three dummy variables are all found to be negative at a one percent significant level. Conforming to this ‘cultural disparity’ argument, DeMan &
Duysters (2005) support that the time, energy, costs and resources of the acquisition process, embracing preparation, negotiation and integration, could have been used to improve efficiency and establish innovation.
All these activities required after an acquisition trying to harmonize the
structure distract managers from undertaking action to innovate as well. So
managers’ stimulus for developing new product and process concepts will decrease after an acquisition (Hitt et al., 1991). Ahuja & Novelli (2014) say that firms need high leverage to finance acquisitions, making managers more risk-averse. The
subsequent interest expenses and repayments relating to the debt will cut the funds for innovation(Prabhu, Chandy, & Ellis, 2005). Therewith, according to Ahuja & Novelli (2014) again, acquisitions can cause managers to overestimate their capability to manage the merged business, hindering the chance of successful innovations.
Besides, Ernst & Vitt (2000) came up with “scientist separation”. They found
that influential employees, including scientists, may leave the firm after an
acquisition. The cutback in number of scientific personnel then reduces the actual knowledge of the merged firm, making innovation less likely (Ornaghi, 2009).
2.3.2 Positive relationship
At the same time, the achievement of synergies, which will further be characterized by “synergy-creation”, can be seen as something positive. Acquiring firms attain access to target firms’ assets and absorb knowledge from the target firm increasing their knowledge base (Henderson & Cockbum, 1996). Furthermore, a merger or acquisition increases the budget for R&D projects and spreads the risk, so bigger projects can be started. Next to the share of risk, the fixed costs of innovation can be distributed across more projects and outcomes (Park & Sonenshine, 2012). In this way firms potentially increase innovation output by economies of scale and scope in R&D projects (De Man & Duysters, 2005).
Also Ahuja & Katila (2001) state that synergy-creation lead to higher
innovative performances. They analyzed the parameters that influence
post-acquisition innovation in the chemical sector and found that the size of the merged knowledge base increases post-acquisition innovation. Ornaghi (2009) confirms that consolidating the competences of two firms lead to synergy-creation. Discoveries made in one section can encourage the research activity of scientists in another
section and create together new ideas to end up with innovations. Likewise, as Cefis & Marsilic (2015) notify, if merging companies have supplementary technologies and are driven by improving value, innovations can be made, which would otherwise not have been achieved (Cassiman & Veugelers, 2007). Thus, the synergy-creation effect might bring an increase in the efficiency of R&D which might boost the incentives to innovate (Stiebale, 2013).
Hagedoorn & Duysters (2002) focused on a high-tech industry to investigate
the effect of acquisitions on the technological performance of the combined firms. This performance was measured by the number of patents, accepted as innovation output. They prove that acquisitions are advantageous in innovation activities. But with a small sidestep, namely there have to be both strategic and organizational fit of both companies. They determined ‘fit’ using Standard Industrial Classification codes, patent classification codes, R&D intensity and firm size (Zhao, 2009). Valentini (2012) found the same positive effect of M&A on the patenting outcome by a research of US firms in the medical devices industry from 1988 till 1996.
Acquisitions introduce the “escape-the-competition effect”. Although some
investigators, like Stiebale (2013), state that the reduction of rivals has an ambiguous effect on the innovation, because this effect is build upon the market environment and the extent of R&D spillovers. Gilbert (2005), Kamien & Zang (1990) and Ornaghi (2009) think that the previously found outflows of rivals can stimulate investments into R&D in the new firm. However, Ornaghi (2009) states that there is a negative side on this reduction of rivals. Next to the fact that the consumer price will rise, an important result is that there will be less incentive to innovate when there are less competitors. Patents withhold rivals the ability to exploit the findings of the innovation for use of own interest. Sevilir & Tian (2012) found a strong positive link between the volume of M&A activity and the number of patents the firm gains following its activity. They even found out that this effect is bigger for larger firms.
The most related study to this paper is the research of Stiebale in 2013. He
investigated the relationship between cross-border acquisitions and innovation of the acquiring company. He used data from German firms with up to €500 million annual turnover in a timespan from 2002 till 2007, including 389 cross-border acquisitions. He uses R&D intensity as measure for innovation, being the dependent variable. The data showed that M&A active firms have higher R&D intensity then firms that do not involve M&A. By a non-linear equation system, he exposes that cross-border
acquisitions let the average R&D rate increase with 1.5 percentage point. It is interesting to investigate this results in a more longitudinal and recent period.
2.3.3 Lagged effect
Mergers and acquisitions demand a sophisticated procedure. Before a merger or acquisition is executed, much time is required in advance to prepare and negotiate about the take-over or collaboration. When the firms decided to go along with each other, the integration process starts. This integration process differs among firms and depends on the cultural and organizational structure. The ‘cultural disparity’ among firms are an explanation for the lagged effect of mergers and acquisitions on
innovation. It might take years before the value of the firms will grow. Otherwise, when the structure, culture and goals of the firms are nearly identical, the effect of M&A might be immediately visible. It is important to take this possibly lagged effect into account and control for this effect in the regression.
2.4 Control Variables
Conceptual model
In order to investigate the effect of mergers and acquisitions on the innovation of a firm we have to take other variables into account. This section points at plausible parameters which could have an effect on innovation. The conceptual model (Figure 2.2) defines the relationship between the dependent and explanatory variables.
Firm Size. The impact of firm size on the level of innovation is not unanimous.
Schumpeter stated back in 1939 that relatively large firms are more appropriate for innovation. Large companies have strong market power and can afford considerable investments in R&D (Cefis & Marsilic, 2015). R&D activities are associated with extensive fixed costs that are solely regained by large sales. Besides, large firms have easier access to external finance, resulting in a simpler way of hiring R&D employees and investing in R&D projects. Therefore, large firms, which are more likely to face economies of scale and scope, also benefit from these economies in the innovation process. Moreover, advantage of experience can contribute to a higher level of innovation. Hitt et al. (1991) defined this positive relation between size and innovation to be reasonably strong. Likewise, Lerner (2002) found that patenting activity of U.S. investment banks is positively linked to their size.
In contrary, Scherer & Ross (1990) propose that smaller companies will be more likely to enjoy innovation by R&D processes. Aron & Lazear (1990) argue by a theoretical developed model that young firms, including start-ups, are more likely to innovate. Smaller companies will easier decide and adjust to new ideas. Aghion et al. (2005) state that large companies with great market power do not have the same level as inducement as smaller firms trying to grow big. Culbertson & Mueller (1985) and Lunn (1986) find uncertain evidence of a positive relation between innovation and firm size.
So the influence of firm size on innovation is discussable. In this paper firm
size will be measured by net sales as well as the number of employees. There is no consentient impact of net sales on innovation input. But the effect of number of employees is clearly positive.
Hypothesis 1: Higher level of employees is expected to result in higher innovation input
Employee Productivity. The productivity of human capital, sales per employee, in a
employee influences the process of a research. Ahuja & Katila (2001) expose that productive personnel will use the expenses on R&D in a more efficient way.
Hypothesis 2: Higher employee productivity is expected to result in higher innovation output.
Net Profit. It is plausible that firms in a well earned period will raise their expenses
on R&D and vice versa will short their R&D budget in a period of bad earnings. If a firm has more profit, it has more budget to finance R&D processes (Sougiannis, 1994). The allocation and the height of budget available for R&D are important for innovation input.
McCutchen (1993) investigated the incentives for R&D expenses in the
pharmaceutical industry. His research found that the U.S. Tax Reform Act of 1986 causes a 1.6% increase of R&D expenses in the three investigated years. This indicates that tax might have impact on the level of R&D expenses. It is beneficial for firms which are running high profits to spent a certain amount on R&D, which
otherwise should have been lost by taxes.
Firms with high profit will furthermore be less hurt by capital problems that may
occur after acquiring another firm.
Hypothesis 3: Higher net profit is expected to result in higher innovation input.
Net Debt. Smith & Warner (1979) found that if the amount of leverage is higher of a
firm, there is a higher change that managers become more and more risk-averse. Higher debt means that the involvement of debt holders is higher. Debt holders in general do not like risky investments. This risk aversion will lead to less investments in the innovation process. Conform to this finding the research of Baysinger &
Hoskisson (1989) discovers a negative link between leverage and investment in R&D, which is an indicator for innovation input.
Hypothesis 4: Higher net debt is expected to result in lower innovation input.
Market Share. The market share of a firm is an indicator of competitiveness of the
market. Theoretical models deliver that higher competition in a market leads to lower innovation, because of post-entry rents that reward innovation (Aghion et al., 2005). Meanwhile Aghion et al. (2014) found, in contradiction to his earlier research, that a higher competition level, introduces a significant increase in R&D expenditure.
Aghion et al. (2005) found that the relationship of competition and innovation has an inverted U shape. This indicates that there is an optimal level of R&D expenditures and accepts both results of his researches. So the effect of market share on innovation is inconclusive.
3. Empirical Research
This section contains a description of the compiled data, covering the specification of the used variables. Subsequently the research method will by declared and
analyzed.
3.1 Data
3.1.1 Pharmaceutical industry
This paper describes the pharmaceutical industry. It is more informative to compare the innovation of firms in the same market, because of generally same firm and environment characteristics. Moreover, only M&A activity between U.S. companies is selected, so target firms as well as acquirer firms were U.S. pharmaceutical firms. Besides, this industry was an important contributor to the international wave of mergers and acquisitions last decade. The pharmaceutical sector is one of the sectors with the highest research and development intensity. Innovation is with no doubt the most valuable aspect for competition in the market (Ornaghi, 2009).
The innovation process of pharmaceutical companies has two stages:
discovery and development. The discovery stage outlines discovering new components, called new chemical entities (NCEs). When a new component is detected, companies receive a patent, which gives them the right of employing all future economic benefits and withholds rivals to exploit this. This stage is followed by testing the efficacy and safety of the newly detected NCEs (Ornaghi, 2009).
3.1.2 Dataset
To generate the dataset used in this paper, two datasets had to be combined. The data of mergers and acquisitions has been extracted from the Thomson ONE database. This delivers a broad range of financial content, including over 400,000 M&A deals worldwide. The top 25 American pharmaceutical firms, ranked by the
R&D expenses, has been selected as sample in this paper, extended with nine random M&A-active firms, which gives a total of 34 firms. The size and market power of the firms is divers. The list of the selected firms is shown in Appendix 1 at the end of this paper.
Stiebale (2013) investigated the effect within a timespan of 2002 till 2007. It is
interesting to investigate the effect over a more longitudinal and recent period. So the mergers and acquisitions, which has been selected in this research, occurred within a timeframe of 1995 till 2014. Because the post-M&A performance has to be taken into account, the M&A activity is merely recorded till 2012. The data across time of the firms gives a total sample of 600 observations, including 64 mergers or
acquisitions between U.S. pharmaceutical firms.
The obtained M&A data has been merged with data from the Wharton
Research Data Services (WRDS) database. WRDS provides access to the database of Bureau van Dijk, which includes OSIRIS. OSIRIS, which has very detailed financial content of companies around the world, provided the other data for the selected companies needed for this research. Not all top U.S. pharmaceutical companies had M&A activity in this selection of time. However, the data for the other variables is included, which creates automatically a control group with no M&A activity.
3.1.3 Variables
The dependent variable is this paper is innovation. This paper selects R&D expenditure as parameter for innovation. There is multiple evidence for a strong positive relationship between R&D input and innovation output, defined as number of patents. Studies from Ahuja & Katila (2001), Hagedoorn & Duysters (2002) and Hitt et al. (1991) find correlations above zero, 0.9, 0.5 and 0.2 respectively.
R&D expenditure, as a single indicator, does not reflect innovation in an
appropriate manner. It is preferable to investigate the relative R&D expenditure of a firm. Therefore, the R&D intensity is a valuable parameter to indicate innovation. This is defined as R&D investment divided by net sales and is measured yearly.
As independent variable, the activity of a merger or acquisition is chosen. This
is defined by a dummy variable that equals one if a merger or acquisition took place that year and equals zero if no merger or acquistion took place. After a merger or acquisition occured, it might take some time to see the accompanying effect. To control for this lagged effect, two more M&A dummies are implemented, namely for
the previous year (t-1) and for the year before that year (t-2).
In addition, the control variables are measured yearly as well. Firm size is
defined by two indicators, namely net sales and number of employees. Further, the employee productivity, defined as the sales per employee, is formed by the number of employees divided by the net sales. The market share per firm is an approximation generated by the total firm revenues divided by the total market revenues.
3.2 Summary Statistics
Summary Statistics
Variable Mean Median Std. Dev. Min Max p-value J-B
RDInt 2.84 0.19 13.12 0.018 138.17 0.0000 MA 0.16 0 0.37 0 1 0.0000 MAt1 0.16 0 0.36 0 1 0.0000 MAt2 0.15 0 0.36 0 1 0.0000 NS 3183485 211875 7388224 0 4.8e+07 0.0000 E 16263.1 1292.5 30230.8 5 128100 0.0000 EP 0.28 0.0037 0.80 0.00028 4.32 0.0000 NP 573507.3 4697 1591428 -2071900 1.29e+07 0.0000 ND 181859.9 -13348.5 1642578 -7540000 1.35e+07 0.0000 MS 0.03 0.002 0.08 0 0.54 0.0000 Figure 3.1
The table (Figure 3.1) above analyzes the descriptive statistics of the dependent and explanatory variables used in this paper. It is interesting to know if the raw data comes from a normal distribution. To test for this normality, the Jarque-Bera (J-B) skewness-kurtosis test is done. The p-value takes into account both the skewness and the kurtosis and has been adjusted for the fact that the sample size is relatively small. The null hypothesis is that the model is normally distributed. The given p-values of the variables are all 0.0000, which declares rejection of the null hypothesis and therefore the data of this research is non-normally distributed.
Despite the high R&D expenses in the pharmaceutical market, the values of
the R&D to net sales ratios are very divers with a minimum of 0.018 (Depomed Inc. in 2014) to a maximum of 138.17 (Amag Pharmaceuticals in 1996) caused by
or a crisis, which affected all the firms individually. The net profit and net debt are also widely spread. The firms have a MA mean of 0.16, meaning that at sixteen percent of the time observations a merger took place. The R&D intensity has a mean of 2.84. This shows on average that the spending on R&D is 2.84 times higher than net sales, which could be true in the pharmaceutical sector where innovation is extremely important. The employee productivity has a maximum of 4.32, which indicates that the sales were more than four times lower than the number of employees. Also these unlike outcomes might be a result of internal problems.
3.2.1 Robustness and Multicollinearity
The summary statistics (Figure 3.1) shows two measures for firm size, namely net sales and the number of employees. To account for the robustness of the model, it is useful to check whether one variable is more appropriate to use than the other. Both parameters have a link with the employee productivity. Therefore, the correlations are calculated between the parameter and the employee productivity. The correlation between the number of employees and the productivity is -0.64 and the correlation of the net sales with employee productivity is 0.037. Hence, because it is better to use the variable with the lowest correlation, net sales is chosen as parameter for firm size.
Nevertheless, it is necessary to check for multicollinearity in the model. To
investigate this, the variance inflation factors (VIFs) are generated. VIFs define the intensity of multicollinearity in a model. It tells us the extent to which the variance, the square of the standard error, of the coefficient of interest is increased due to
collinearity. If the VIF has a value of one, there is no correlation between that variable and the other explanatory variables. Some collinearity is tolerated, but there is a limit. Therefore, a rule of thumb is set up, which states that coefficients with a factor equal or higher than four warrant further investigation. In this case the VIF of net sales is the highest of all variables with a value of 4.37. Net sales is the only variable with a VIF above four, meaning that the model has to be adjusted for this. That is why an interaction variable between net sales and employee productivity is introduced.
3.3 Methodology
This empirical analysis is based on cross-sectional time-series data, called panel data, in which the activities of companies are observed across time. Panel data describes the change over time and provides a more accurate and simple
interpretation of model variables, which is better at taking the complexity of behavior into account (Hsiao, 2007). Some data is missing in this dataset, making this panel data an unbalanced panel.
3.3.1 Hausman Test
For the companies, the data of the variables is available for several years, which makes panel data applicable. To decide between fixed effects or random effects the Hausman test is done. The Hausman test tests whether there is significant difference between the fixed and random effect estimators. It tests whether the unique errors (ui) are correlated with the regressors, with no correlation as the null hypothesis. It is
Chi-square distributed with degrees of freedom equal to the number of variables for the time-varying regressors. If the p-value of the Hausman test is insignificant, the use of the random effects model is appropriate. Otherwise, when the p-value is significant, the null hypothesis will be rejected and it is best to use the fixed effects model (Heij et al., 2004).
3.3.2 Random-effects model
In this research the outcome of the Hausman test gives a Chi-square value of 0.09 with a p-value of 0.7583, which is larger than the significance level (0.05). This means that the null hypothesis is not rejected and shows that the use of the random effects model is appropriate. An advantage of this model is that time invariant
variables can be included. The random-effects model is used when differences across firms have influence on the dependent variable, which is in line with economic theory and the ‘cultural disparity’ principle mentioned in the theoretical background review (Section 2.3.1).
3.3.3 Tests for model propriety
To test if all the coefficients in the model are different from zero, a Wald test is done. The Chi2 statistic gives a value of 275.89 with a p-value of 0.0000. The null
hypothesis that states that the coefficients are jointly zero is rejected. This proves that all the coefficients are jointly unequal to zero and that the model is significant at a one percent significance level.
To control for possible time-effects it is required to check if all dummies for all
years are equal to zero. The p-value of Chi2-statistic is 0.8998, which is larger than the significance level (5%), indicating that there is no need for time-fixed effects in this model.
Cross-sectional dependence can lead to bias in the results of the tests. The
Wooldridge test controls for this dependence. It tests if there is serial correlation between the residuals, with the null hypothesis that there is no correlation. The p-value of this test is 0.0670, which is larger than the significance level of five percent. The null hypothesis can not be rejected, meaning there is no serial correlation, also called autocorrelation, in this model. This indicates that future observations are not affected by past values. Unfortunately, this is in contrast with the economic theory. One can suppose that the R&D investments of previous year are affected R&D expenses in the current year.
3.3.4 Regression equation
When filling in all the variables, the regression equation gets the following form:
!"#$%&%= ()+ (+,-&%+ (.,-&(%0+)+ (2,-&(%0.)+ (345&% + (789&%+ (:;9&% +
(<;"&% + (=,5&%+ >&%+ ?&%
Equation 3.1
where, RDIntit is the R&D intensity of firm i at time t, measured by the R&D expenses
divided by the net sales. MAit is a dummy variable that equals one if a merger or
acquisition occurred for firm i at time t and otherwise equals zero. MAi(t-1) and MAi(t-2)
are the dummies which control for the lagged effect.
FSit is the size of firm i at time t. Firm size can be measured by the net sales
(NSit) and by the number of employees (Eit). The effect of net sales on innovation is
inconclusive, so there is no expectation about β4 in the first case, but Hypothesis 1
suppose a positive relationship between the number of employees and innovation, stating β4>0 in the latter case.
divided by the net sales, of firm i at time t. Hypothesis 2 states that employee productivity has a positive link to innovation, causing β5>0.
NPit, and NDit denote the net profit and net debt of firm i at time t, respectively.
Hypothesis 3 says that net profit has a positive influence on innovation and Hypothesis 4 indicates a negative link between net debt and innovation, so the coefficients will be: β6>0 and β7<0.
MSit is an approximation of the market share of firm i at time t. Results of
studies done about the association between the market share and innovation are different, so the effect (β8) is undetermined.
With the decision to use net sales as parameter for the size of a firm accompanying the implemented interaction variable NSEPit, the final regression looks as follows:
!"#$%&%= ()+ (+,-&%+ (.,-&(%0+)+ (2,-&(%0.)+
(3;5&% + (789&% + (:;589&%+ (<;9&% + (=;"&% + (@,5&%+ >&%+ ?&%
4. Results
Section 4 will present the outcomes of the panel data regression. The results of the random-effects model will be showed and analyzed, including clarification of the coefficients of the used variables.
4.1 Random-effects model
Random-effects model
RDIntensity Coef. Std. Err. z P>|z|
MA 17.92 6.57 2.73*** 0.006 MAt1 -8.92 6.66 -1.34 0.180 MAt2 -1.45 6.90 -0.21 0.833 NS 0.0002 0.0003 0.62 0.535 EP 149.91 10.36 14.47*** 0.000 NSEP 0.007 0.11 0.64 0.524 NP -0.0004 0.0002 -2.49** 0.013 ND 0.00006 0.00005 1.21 0.225 MS -21661.2 37761.6 -0.57 0.566 _cons 1.72 6.83 0.25 0.801
*significance at 10%, **significance at 5%, ***significance at 1%
Figure 4.1
The table (Figure 4.1) above represents the results from the random-effects model. The coefficient of MA is 17.92. This implies a positive impact of a merger on the R&D intensity. The p-value of this coefficient is 0.006, which is below the significance level of five percent. This indicates that the effect of M&A in the same year is statistically significant to explain R&D intensity. The two dummy variables for the lagged effect have both a negative coefficient. This predict that the immediate effect of a merger or acquisition is positive for the R&D intensity, but the year after and the second year after the effect is negative. However, these two dummy variables are insignificant, so this conclusion is not robust.
Net sales, as indicator of the size of a firm, has a positive (β=0.0002) relation
with R&D intensity. Researchers found either positive or negative links between the net sales and R&D expenses, but in this case this coefficient is not significant and a conclusion is inscrutable.
divided by the net sales, has a positive effect on R&D intensity (β=149,91). This coefficient is significant at a one percent level and proofs that higher employee productivity will lead to higher R&D intensity, which is in line with Hypothesis 2.
Net profit has a negative effect on the R&D ratio. Hypothesis 3, which stated
that higher net profit will cause higher R&D ratio is in contrast with this outcome. The coefficient has a p-value of 0.013, which is lower than the 5%-significance level. So this significant result of net profit negatively influencing innovation is not in line with the economic theory reviewed in Section 2.4.
Hypothesis 4 states that higher debt introduces a lower R&D ratio. The
coefficient found by this regression is negligible positive, with a value of 0.00006, and shows the contrary. But this outcome is insignificant and so can not explain the association between net debt and innovation.
Researchers found divers outcomes on the relationship of market share and
innovation. This regression shows a coefficient outcome of -21662.2. But luckily this outcome, with a p-value of 0.566, is not a significant explanation of the large negative relationship. Because of the immense value, one can consider to eliminate the
market share parameter. If the regression is run without market share, the other values, the coefficients and the p-values of the existing variables, do not alter remarkable. Only the constant moves from 1.72 to 0.83., but these are both
insignificant. The signs of the coefficients stay the same and number of significant parameters is remained. The exact outcomes for this reduced model can be found in the Appendix 2 at the end of this paper.
5. Limitations and Drawbacks
The primary difficulty in this research is that the endogeneity problem is present, which means that the explanatory variables correlate with the error term. This correlation makes it hard to separate the precise effect of a variable on the R&D intensity. The variations in the explanatory variables are affiliated with the variation in R&D intensity, not only by the coefficient (β) but also indirectly by the changes in the error term (ε) (Heij et al., 2004).
5.1 Firm Characteristics
It is difficult to indicate the exact effect of acquisitions on the following innovation because the characteristics, the organizational culture and structure of the
companies, are divers and may influence the process of research and development and the process of undertaking a merger of acquisition. This ‘cultural disparity’ among companies is not taken into account, which may lead to a bias in the model. Moreover, the reason behind a merger or acquisition is behind the scope of this research. The merger may for example not be implemented with sight on innovation but a defensive move to anticipate for shocks. Furthermore, this research does not control for the size and transaction value of a take-over. These aspects have influence on the relationship between M&A and innovation.
This bias created by characteristics distinction represents the omitted variable bias in the research. The omission of relevant variables leads to biased estimates and a reduction in variance. For example, not all R&D projects succeed, which means that there is a fraction of R&D expenses that leads into nothing. Therefore, it might be valuable to add a variable to regulate for this. A proper measure might be the return on R&D expense, defined by the current year profit or patents granted divided by previous year R&D expenditure. Likewise, one can think of other variables which may influence innovation, like liquidity, internal knowledge, diversification, research productivity, growth potential and the age of the CEO. However, this extension of explanatory variables is beyond the scope of this paper.
5.2 Unbalanced Data
The panel data was appointed as unbalanced, implying that there is data missing. For example, in the first three years (1995-1997) a part of the R&D expenses were not available for King Pharmaceuticals and the net sales were not noted for United Therapeutics. This missing values can be seen as sample selection bias. If this research includes a larger dataset with more observations and more M&A activity, the results of this research will be more robust.
Besides, the results of this study represents that the R&D intensity is very variable and even sometimes above hundred. This is because of the varying values of R&D expenditures and net sales. Because of the merge of the two databases there could be a bias in the obtaining data process.
5.3 Simultaneous Causality
Furthermore, like Zhao (2009) also finds, is that innovation is not only influenced by mergers, it encourages mergers too. Imagine a firm with deficient innovation efforts, a merger is a suitable option to solve this problem. The merger increases the internal knowledge and subsequently generates innovation. This bias, called the
simultaneous causality bias, causes the regression to pick up both effects making the estimators biased and inconsistent.
6. Conclusion
This paper is written with the purpose to investigate the effect of mergers and acquisitions on innovation. Studies done about this association have diverse outcomes. Stiebale (2013) investigated this relationship among German firms and found a positive link. This study aims at a more longitudinal and recent period, 1995-2014, for pharmaceutical firms in the U.S., because of the innovative character of the pharmaceutical sector.
From the Thomson ONE database and the WRDS database panel data have
been formed. Innovation is measured by the intensity of R&D, defined by R&D expense divided by net sales and is the dependent variable in the used model. Explanatory variables affecting innovation were, next to the dummy variables for M&A activity, firm size, employee productivity, net profit, net debt, and market share. By a random effects model, the regression is done and the results are generated.
The coefficient of the M&A indicator is found to be positive and significant with
a one percent significance level. In short, a significant positive relationship is found between M&A activity in the same year and the innovation input (R&D intensity) of a firm. It shows that if M&A activity occurred at any time in a year, the R&D intensity at the end of that year is higher. So M&A activity has in the short run a positive impact. But in the long run a negative impact on innovation is noticed. The lagged effect, one and two years after the merger of acquisition, is namely found to be negative. But this effect, described by two dummy variables, is insignificant. Employee productivity has a one percent significant positive effect on R&D intensity, which is in line with
Hypothesis 2. Net profit has a five percent significant negative effect and net debt an insignificant positive effect on R&D intensity. These two outcomes are both in
be insignificant as well.
At last, the validity is discussed in section 5, exposing biases in this model.
The missing data causes sample selection bias and the omitting of variables causes omitted variable bias. Moreover, the effect of simultaneous causality is clarified. The outcomes and explained biases can be a good start for further research into
innovation.
References
Aghion, P., Bechtold, S., Cassar, L., & Herz, H. (2014, February 20). The Causal Effects of Competition on Innovation: Experimental Evidence.
Aghion, P., Bloom, N., Blundell, R., R., G., & Howitt, P. (2005). Competition and innovation: An inverted-U relationship . The Quarterly Journal of Economics ,
120, 701-728.
Ahuja, G., & Katila, R. (2001). Technological Acquisitions and the Innovation
Performance of Acquiring Firms: A Longitudinal Study. Strategic Management
Journal , 22 (3), 197-220.
Ahuja, G., & Novelli, E. (2014). Mergers and Acquisitions and Innovation. In M. Dodgson, D. M. Gann, & N. Phillips, The Oxford Handbook of Innovation
Management. Oxford: Oxford University Press.
Aron, D. J., & Lazear, E. P. (1990). The introduction of new products. American
Economic Review Papers and Proceedings , 80, 421.
Baldwin, J. R. (1997). The importance of research and development for innovation in small and large Canadian manufacturing firms. Statistics Canada Analytical
Studies , 107.
Baysinger, B., & Hoskisson, R. E. (1989). Diversification strategy and R&D intensity in multi-product firms. Academy of Management Journal , 32, 310-323.
Cassiman, B., & Veugelers, R. (2007). Are external technology acquisition strategies substitutes or complements? The case of embodied and disembodied
Cassiman, B., Colombo, M., Garrone, P., & Veugelers, R. (2005). The impact of M&A on the R&D process. An empirical analysis of the role of technological and market relatedness. Research Policy , 34 (2), 195-220.
Cefis, E., & Marsilic, O. (2015). Crossing the Innovation Threshold through Mergers and Acquisitions. Research Policy , 44 (3), 698-710.
Culbertson, J. D., & Mueller, W. F. (1985). The influence of market structure in technological performance in the food-manufacturing industries. Review of
Industrial Organization , 2, 40-54.
De Man, A., & Duysters, G. (2005). Collaboration and innovation: a review of the effects of mergers, acquisitions and alliances on innovation. Technovation , 25 (12), 1377-1387.
DeMan, A. G., & Duysters, G. (2005). Collaboration and innovation: a review of the effects of mergers, acquisitions and alliances on innovation. Technovation , 25 (12), 1377-1387.
Drucker, P. F. (1985). In Innovation and Entrepreneurship: practice and principles (p. 30). New York: HarperBusiness.
Ernst, H., & Vitt, J. (2000). The Influence of Corporate Acquisitions on the Behaviour of Key Inventors. R&D Management , 30 (2), 105-120.
Garcia, R., & Calantone, R. (2002). A critical look at technological innovation typology and innovativeness terminology: a literature review. The Journal of
Product Innovation Management , 19, 110-132.
Gilbert, R. J. (2015). Competition and Innovation.
GlobalData. (2014, 12 17). Top 30 Pharma Companies Spent $112 Billion on
Research and Development in 2013, says GlobalData. Opgeroepen op 06 10,
2016, van GlobalData: https://healthcare.globaldata.com/media-center/press- releases/pharmaceuticals/top-30-pharma-companies-spent-112-billion-on-research-and-development-in-2013-says-globaldata
Hagedoorn, J., & Cloodt, M. (2003). Measuring Innovative Performance: Is There an Advantage in Using Multiple Indicators? Research Policy , 32, 1365–1379.
Hagedoorn, J., & Duysters, G. (2002). The Effect of Mergers and Acquisitions on the Technological Performance of Companies in a High-tech Environment.
Technology Analysis & Strategic Management , 14 (1), 67-85.
Heffernan, S., Fu, X., & Fu, X. (2013). Financial innovation in the UK. Applied
Economics , 45 (24), 3400-3411.
Heij, C., de Boer, P., Franses, P. H., Kloek, T., & van Dijk, H. (2004). In C. Heij, P. de Boer, P. H. Franses, T. Kloek, & H. van Dijk, Econometrics Methods with
Applications in Business and Economics. New York: Oxford.
Henderson, R., & Cockbum, I. (1996). Scale, Scope and Spillovers: the determinants of research productivity in drug discovery. RAND Journal of Economics , 27 (1), 32-59.
Hitt, M. A., Hoskisson, R. E., Ireland, R. D., & Harrison, J. S. (1991). Effects of Acquisitions on R&D Inputs and Outputs. The Academy of Management
Journal , 34 (3), 693-706.
Hsiao, C. (2007). Panel data analysis—advantages and challenges. TEST , 16 (1), 1-22.
Kamien, M. I., & Zang, I. (1990). The Limits of Monopolization Through Acquisition. Quarterly Journal of Economics , 105 (2), 465-499.
Lunn, J. (1986). An empirical analysis of process and product patenting: A
simultaneous equation framework. Journal of Industrial Economics , 34, 319-330.
McCutchen, W. (1993). Estimating the Impact of the R&D Tax Credit on Strategic Groups in the Pharmaceutical Industry. Research Policy , 22, 337-351. Morgan, E. J. (2001). Innovation and Merger Decisions in the Pharmaceutical
Industry. Review of Industrial Organization , 19, 181–197.
Ornaghi, C. (2009). Mergers and innovation in big pharma. International Journal of
Park, W. G., & Sonenshine, R. (2012). Impact of Horizontal Mergers on Research & Development and Patenting: Evidence from Merger Challenges in the U.S. J
Ind Compet Trade , 12, 143-167.
Prabhu, J. C., Chandy, R. K., & Ellis, M. E. (2005). The Impact of Acquisitions on Innovation: Poison Pill, Placebo, or Tonic? Journal of Marketing , 69 (1), 114-130.
Scherer, F. M., & Ross, D. (1990). In Industrial Market Structure and Economic
Performance (3 ed.). Boston: Houghton-Mifflin.
Schumpeter, J. A. (1939). In Business Cycles. A Theoretical, Historical and Statistical
Analysis of the Capitalist Process (p. 80). New York: McGraw-Hill Book
Company.
Sevilir, M., & Tian, X. (2012, May 24). Acquiring Innovation.
Smith, C. L., & Warner, J. B. (1979). Bankruptcy, secured debt, and optimal capital structure: A comment. Journal of Finance , 34, 347-351.
Sougiannis, T. (1994). The accounting based valuation of corporate R&D. The
Accounting Review , 69, 44-68.
Stiebale, J. (2013). The impact of cross-border mergers and acquisitions on the acquirers' R&D - Firm-level evidence. International Journal of Industrial
Organization , 31, 307-321.
Stock, J. H., & Watson, M. (2015). In Introduction to Econometrics (Vol. 3, pp. 396-419). Pearson.
Valentini, G. (2012). Measuring the Effect of M&A on Patenting Quantity and Quality. Strategic Management Journal , 33 (3), 336-346.
Zhao, X. (2009). Technological Innovation and Acquisitions. Management Science,
Appendices 1. List of Firms
1. ABBOTT LABORATORIES
2. ACCESS PHARMACEUTICALS INC 3. AKORN INC
4. ALEXION PHARMACEUTICALS INC 5. ALLERGAN INC
6. ALPHARMA INC
7. AMAG PHARMACEUTICALS, INCN 8. AMGEN INCORPORATED
9. BARR PHARMACEUTICALS, INCN 10. BIOGEN INCN
11. BIOMARIN PHARMACEUTICAL INC 12. BRISTOL-MYERS SQUIBB COMPANY 13. CELGENE CORP
14. CEPHALON INC 15. DEPOMED INC
16. ELI LILLY AND COMPANY
17. ENDO PHARMACEUTICALS SOLUTIONS INCN 18. FOREST LABORATORIES INC
19. GILEAD SCIENCES INC 20. IVAX CORP
21. JOHNSON & JOHNSON
22. KING PHARMACEUTICALS INC 23. MEDICIS PHARMACEUTICAL CORP 24. MERCK & CON, INCN
25. MGI PHARMA INC
26. OSI PHARMACEUTICALS INC
27. PAR PHARMACEUTICAL COMPANIES, INCN 28. PFIZER INC
29. SALIX PHARMACEUTICALS, LTDN 30. SCIELE PHARMA, INCN
31. UNITED THERAPEUTICS CORP
32. VALEANT PHARMACEUTICALS INTERNATIONAL 33. VERTEX PHARMACEUTICALS INCORPORATED 34. ZILA INC
2. Random-effects model excluding Market Share parameter Random-effects model excluding MS parameter RDIntensity Coef. Std. Err. z P>|z|
MA 16.867 5.979 2.82*** 0.005 MAt1 -9.164 6.440 -1.42 0.155 MAt2 -0.694 6.582 -0.11 0.916 NS 0.000035 0.00015 0.24 0.813 EP 150.37 9.876 15.22*** 0.000 NSEP 0.0434 0.097 0.45 0.653 NP -0.00045 0.00014 -3.29*** 0.001 ND 0.000061 0.000051 1.20 0.231 _cons 0.832 6.862 0.12 0.903
*significance at 10%, **significance at 5%, ***significance at 1%
