The effect of mergers and acquisitions on research and development output elasticity

The Effect of Mergers and Acquisitions on

Research and Development Output Elasticity

Student: Tom van der Laan - S11019867 Master Thesis – MSc Business Economics: Finance University of Amsterdam, The Netherlands

Supervisor: Tolga Caskurlu Assistant Professor of Finance University of Amsterdam, The Netherlands Date – December 14, 2016 ABSTRACT

This research examines the effect of mergers and acquisitions (M&A) on research and development output elasticity, using the research quotient (RQ) as a method of measurement. I performed a quasi-natural experiment with a control sample of failed M&A bids and a control sample of firms without M&A activity. Using a sample of 545 US deals by 419 unique firms from 1990-2010, this research finds that there is no evidence that a firm’s M&A activity negatively affects R&D output elasticity in the first three years after M&A activity. Furthermore, the results suggest that the level of R&D output elasticity of the acquirer does not affect abnormal returns of the acquirer conducting M&As. Finally, firms that conduct M&As increasingly underinvest in R&D when compared to firms without M&As. Keywords: Mergers and Acquisitions, Research Quotient, R&D Output Elasticity, TFP, Innovation output.

Statement of Originality

This document is written by Student Tom van der Laan who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

CONTENT

I. INTRODUCTION ... - 4 - II. RELATED LITERATURE ... - 6 - 2.1 M&A activity and firm performance ... - 6 - 2.2 Innovation, firm value and M&A ... - 6 - 2.3 Constructing the Research Quotient ... - 7 - 2.3 Firm characteristics and stock returns ... - 9 - III. SAMPLE SELECTION AND METHODOLOGY ... - 11 - Hypotheses ... - 11 - M&A Deal Sample ... - 13 - Control Samples ... - 15 - Descriptive statistics ... - 15 - Methodology hypothesis I ... - 16 - Methodology hypothesis II ... - 17 - Methodology hypothesis III ... - 20 - IV. ANALYSES ... - 21 - 4.1 Do M&A negatively affect firm-specific R&D output elasticity? ... - 21 - 4.2 Does the level of RQ of the acquirer affect abnormal stock returns? ... - 22 - 4.3 Does M&A activity drive underinvestment in R&D? ... - 24 - 4.4 Sensitivity Analysis ... - 25 - V. CONCLUSION ... - 26 - 5.1 Results & Discussion ... - 26 - 5.2 Shortcomings of this research ... - 27 - 5.3 Ideas for future research ... - 28 - VI. REFERENCES ... - 29 - VII. TABLES ... - 35 - Appendix I: Variable definitions ... - 43 - Appendix II: Equal Weighted Cumulative Return ... - 44 -

I. INTRODUCTION

The process of mergers and acquisitions (M&A) is one of the centrepieces of corporate finance. Worldwide, companies spent $3.8 trillion on M&A in 2015, making it the best year ever. Deals can make or break companies and have the potential to change entire industries. This makes M&A an interesting and delicate activity where much can be gained or lost. It is not surprising that this phenomenon has drawn the attention of academics. In the last decades, many researchers have tried to explain why M&A takes place and what the effects of M&A are on the companies involved and their economic environment.

This study focusses on whether M&A activity affects the efficiency of firm-specific research and development (R&D) spending. Using the alternative research quotient (RQ), which is the management measure of the total factor productivity (TFP), I examine whether US firms increase their ability to profit from R&D spending after M&A activity. Furthermore, I examine whether RQ affects the firm’s stock price when performing M&A activity, and whether M&As cause underinvestment in R&D.

Existing literature does not investigate the effects of M&A activity on R&D output elasticity. The existence of any effect gives investors and policy makers more understanding of firm dynamics in an M&A setting. Furthermore, this paper adds more insight to the existing literature on M&A, effects on R&D productivity, and the understanding of the rationale behind pursuing M&A strategies. This paper contributes to the literature in two ways. First, by using an industry wide sample, I utilise a more representative sample than research that only focusses on individual cases or plant level data, such as Ornaghi (2009), Stiebale et al. (2011) or Blonigen et al. (2015). Using an industry wide sample reduces the selection bias that arises when measuring the effects of M&A on the firm’s innovation process using patent-based measures, as fewer than 50% of the Compustat firms that conduct R&D files for patents. Second, I consider whether RQ affects abnormal returns around M&As. This gives more insight behind the relation of expectations of the firms’ value and the ability of a firm to profit from R&D investments.

When researching the effects of M&A on firm performance or the innovation process of firms, there are some endogeneity problems to address. The first source of endogeneity is caused by latent innovation outcomes (Ornaghi, 2009). Latent innovation outcome is the innovation output of a firm that researchers could have observed without an M&A transaction. The second source of endogeneity is caused by the simultaneousness of the decision to perform M&A activity and the expected changes in

innovation and market power. When firms expect a negative exogenous shock in their market, they can try to defend their position or operations by pursuing M&A activities. The third source of endogeneity comes from the characteristics of the acquiring firm (Sorescu et al., 2007). Certain firm characteristics affect inputs and outputs of the innovation process and may affect the innovation process when performing M&A activity.

To address these sources of endogeneity, I performed a quasi-natural experiment. I used two control samples, one sample of firms that experienced a failed M&A transaction, and one sample of firms that did not pursue any M&A activity. By using the two samples, the first two sources of endogeneity are addressed. To control for endogeneity from firm characteristics, I included several control variables in my regressions and used panel data to control for other unobserved firm characteristics.

This study is based on financial information of publicly traded stocks from the US. The sample includes 545 deals in the period 1 January 1990 and 31 December 2010. The source of the sample is the M&A database of Thomson One. In conclusion, I found no significant average treatment effect of M&A activity on a firms RQ, no significant effect of RQ on abnormal returns, and a positive relationship between M&A activity and underinvestment in R&D. I found no explicit explanation for these findings. The results of this research should be interpreted with certain reservations. First, the RQ measurement is a model of endogenous growth, and the predictions of these models are hard to test directly due to the scarcity of directly observable measures of technological innovation (Kogan, 2010). TFP models are based on residue, and other forces that are not directly related to innovation can also be measured, such as resource allocation (Hsieh & Klenow, 2009). This research was performed under the assumption the RQ model is valid, and therefore, all results are conditioned on that assumption. Third, for testing the first hypothesis, I used a [y-3; y+3] timeframe for practical reasons and data availability. One could argue that adjustments in R&D spending and R&D output elasticity only become measurable in the long term. The effects of M&A activity on long-term R&D output elasticity can be researched more exhaustively.

This paper is structured in four remaining sections. Section 2 summarizes previous research around M&A activity and provides the academic framework surrounding synergies, firm performance, R&D spending, and other ‘classic’ M&A characteristics such as leverage and return on assets (ROA). Section 3 explains the methodology and the method of data collection and introduces the hypotheses. Section 4 analyses the data, gives descriptive statistics of the samples, provides the results of the

analyses, and explains the sensitivity analysis. Section 5 draws conclusions from the findings of this research and discusses the results.

II. RELATED LITERATURE

2.1 M&A activity and firm performance

The existence and effects of M&A activity is one of the fundamental questions in corporate finance. Per Blonigen & Pierce (2015), it is difficult to fully attribute the effects of M&A activity on firm performance to certain drivers. The authors suggest a dynamic trade-off exists between increased market power and efficiency gains, resulting from M&A activity.

Earlier research on M&A activity uses stock prices to measure the effect of an M&A on firm value (Ravenscraft & Scherer, 1987). Yet, this early research fails to relate equity value gains directly to gains in corporate performance (Caves, 1989). Healy et al. (1992) successfully links an increase in pre-tax operating returns with an increase in corporate performance, but fails to address the underlying drivers of this effect. Overall, the evidence is mixed. Humphery-Jenner and Powell (2011) and Carline et al. (2007) both find a positive take-over effect. Humphery-Jenner and Powell (2011) find this positive take-over effect when looking at stock prices, while Carline et al. (2007) find a positive effect when researching operating cash-flows. On the other hand, Heron and Lie (2002) and Dutta and Jog (2008) find no significant effect.

The last decade of academic research focusses more on identifying the underlying drivers causing positive effects in firm value or firm performance following M&A activity. Underlying drivers that have been identified include operational synergies, innovation synergies, tax savings, the transfer of employees and an increase in (imputed) monopoly rents (Andrade et al., 2001). Per Gugler et al. (2003), this research can be grouped into three categories: two categories focus on increasing shareholder value through cost and market value strategies, and the third group focusses on managerial goals, such as growth of the firm or other ‘quasi-irrational behaviour’.

2.2 Innovation, firm value and M&A

Technological change or ‘innovation’ as driver for economic growth was identified by Schumpeter (1912) and later by Solow (1957), which research measures technology as a residual of the production function. Innovation is best viewed as the output of the innovation process of a company. The relationship between the research inputs and

outputs for a firm is complex. The outcomes can be measured by the amounts of patent grants, patent citations, or increased firm value (Bloom et al., 2002; Czarnitzki et al., 2006; Hall, 2000). The input of the innovation process is mostly measured by the number of funds invested in R&D. R&D expenses include variable costs to fund projects and fixed costs for equipment and buildings. Per Ornaghi (2009), M&A activity can affect the optimal R&D expenses of a firm through different channels. First, M&A activity can reduce fixed costs by selling poor performing assets (Healy et al., 1992), combining R&D departments, or reducing other factors that become obsolete after combining two firms. Second, an M&A can create knowledge synergies, which implies cross-fertilization of ideas. Knowledge synergies are not related to any changes to R&D input and increase research performance. For example, Hoberg and Phillips (2010) find evidence for product market synergies and Bena and Li (2015) find evidence that M&A activity can increase post-merger patent applications for firms with a certain pre-merger technological linkage.

This paper tries to answer the question whether M&A activity affects the efficiency of R&D spending of firm level revenues, and whether the elasticity between innovation inputs and outputs changes after M&A activity. To research this, I use the RQ model, introduced by Knott (2008) as alternative measurement of innovation (Cooper et al., 2015). RQ is the firm-specific output elasticity of R&D and represents the percentage increase in revenues from a 1% increase in R&D spending, keeping all other inputs and elasticities in the production function of the firm constant. According to Knott (2008), ‘a firm can have high RQ by either generate a large number of innovations and being reasonably effective exploiting them, or by generating a smaller number of innovations and being extremely effective exploiting them.’ RQ confirms with the most common means to measure returns to R&D at industry and economy levels (Hall et al., 2010).

2.3 Constructing the Research Quotient

The firm specific RQ is calculated as follows (Knott, 2008):

Yi,t = #_$, &_$,'( )*_$,' +_$,',-. /_$,',-0 1_$,'2 ɛ i,t

Where Yi,t is the firm output, Ai is a firm fixed effect, Ki,t is capital, Li,t is labor, Ri, t-1 is

lagged R&D expenses and Si, t-1 is lagged spill overs, Di, t is advertising and ɛ i,t is an error

term.1 _{This model is estimated by using a random coefficients model that allows for}

heterogeneity in the output elasticity of R&D (Knott, 2008). Random coefficient models are models where the coefficient has two components. The first component includes the

direct effect of the explanatory variable, while the second component includes a proxy that corrects for the effects of omitted variables. The random coefficient model is estimated using the following regression:

ln 8_$,' = 9:+ 9:$ + 9-+ 9-$ <= &$,' + ( 9?+ 9?$ )ln )$,' + 9A+ 9A$ <= +$,',- +

9B+ 9B$ <= /$,',- + 9C+ 9C$ <= 1$,' + ɛ i,t

When using data from the Compustat database, 8_$,' is firm revenues, &_$,' is net property, plant and equipment, )_$,' is measured in full-time equivalent employees in units of 1000, 1_$,' is advertising (mln $) and +_$,' is R&D expenses in mln $. These data items are also used to calculate the additional input firm-specific spill overs (/_$,'), which is the sum of the differences in knowledge between local firm i and rival firm j for all firms in the specific SIC industry that have more R&D expenses than the local firm:

/_$,' = _$,'+_D,' - +_$,' ∀ +_D,' ≥ +_$,'

Per Knott (2008), this construction mimics the spill over construct in endogenous growth models and represents in essence the likelihood that superior knowledge is discovered in a random encounter with a rival firm. The RQ model is based on the TFP model (Solow, 1957; Syverson, 2011), and it captures technological change in the production function as an endogenous source. It is important to distinguish between endogenous and exogenous models. Both models capture the source of technological change in the firm’s production function. But, endogenous models assume that these changes in technology are done by profit-maximizing entities, instead of resulting from public and government knowledge. Research shows that RQ is a useful proxy for R&D productivity (Lentz & Mortensen, 2015) and that it matches predictions of firm value and growth (Knott & Vieregger, 2016). RQ also conforms with the most common methods to measure returns to R&D at the industry level and economy-wide (Hall et al., 2010). The RQ measurement addresses some of the shortcomings of the patent-based measurements that are used to capture innovation output benefits. Cooper et al. (2015) argue that the flaws of patent-based measures include identification issues. First, they stress out that that the decision to patent is endogenous to firm innovation policy, and that a non-monotonic relationship between a patent’s value and citations exists (Abrams et al. 2013). Furthermore, they argue that firms who patent are different from those who do not patent, and finally that the economic value of patents is not uniform (Sherer

& Harhoff, 2000). However, the relationship between patent value and forward citations is still the subject of discussion, as Harhoff et al. (1999), Hall et al. (2005), and Kogan et al. (2011) find a strong and positive association between forward citations and patent value.2 Related research focusses on the effect of M&A activity on R&D intensity. R&D intensity is the ratio between R&D expenses and sales. Szücs (2014) found M&A activity increases sales and operations for acquirers and reduces R&D spending, ultimately leading to a lower R&D intensity. Ornaghi (2009) confirms these findings, but also finds that M&A activity decreases the innovative efforts of pharmaceuticals in patent grants and patent citations. Stiebale et al. (2011) find a similar result for 304 small- and medium-sized German firms when measuring product and process innovations for cross-border mergers and acquisitions. Furthermore, Hitt et al. (1991) find that acquisitive growth is negatively related to firm innovation in terms of inputs (R&D intensity) and outputs (patent granted). Lastly, Desyllas and Hughes (2010) find that R&D intensity of the acquirer drops in the first year following a merger or acquisition, but increases again in years thereafter. 2.3 Firm characteristics and stock returns Certain firm characteristics are widely studied as determinants of M&A activity on firm performances. Examples of widely studied characteristics include the domestic or cross-border M&A (Kaplan & Weisbach,1992; Martynova & Renneboog, 2006: Servaes, 1991); whether the deal can be categorized as a focused or diversifying merger and/or acquisition (Shleifer & Vishny, 1989); the cash reserves of the acquirer (Gao, 2011; Jensen, 1986; Himmelberg & Petersen,1994); the acquirer’s leverage (Ghosh, 2012; Heron & Lie, 2002; Jensen, 1986); and whether the payment occurs in cash or stocks (Carline et al., 2007; Ghosh, 2001; Linn & Switser, 2001).

The relationship between R&D investments and stock price fluctuations is also widely studied. Griliches (1991) and Pakes (1985) find that R&D investments and stock price fluctuations are related. Schmookler (1966) argues that these fluctuations arise on the demand side (aggregate demand, population, etcetera), while Rosenberg (1974) finds that these fluctuations represent a shift in technological opportunity. Griliches et al. (1991) use patent citation data to account for the large variance in patent value and

2 _{A forward citation for a given patent is a citation in another (newer) patent document back to that given}

find that R&D spending and patent citations explain only a fraction of the variance of market value.

Regarding M&A activity and firm value, on average the shareholders of the acquirer experience negative or zero abnormal stock returns in the short run, due to agency problems, empire building of managers (Jensen, 1986), or information asymmetry (Andrade et al., 2001). Campa and Hernando (2004) summarize earlier research and give a complete overview of Cumulative Average Abnormal Returns (CAAR) findings for different event windows, research periods, industries, and country coverage. Most event studies reporting negative CAARs find effects that vary between -4.09% (Morck et al., 1988) and -0.07 (Mitchell & Stafford, 2000). Most research finding positive effects reports CAARs that are close to 0%.

Furthermore, empirical evidence on whether the acquirer’s leverage should be treated as firm determinant is mixed. One of the hypothesis behind the acquirer’s leverage as a firm determinant, is the monitoring role that is executed by banks. When control is strict, it becomes more likely that bad acquisitions are prevented. Also, high leverage makes it harder for the acquirer to finance acquisitions by issuing more debt. Clark and Ofek (1994) find no significant relation between post-merger firm performance, while Harford (1999) does find a positive effect. The relation between R&D intensity and leverage is also widely studied. Several hypotheses have been formulated for the relationship between R&D intensity and leverage. First, the bankruptcy costs hypothesis states that R&D-intensive firms have problems in posting collateral when issuing debt, and therefore internal funds are used when highly investing in R&D (Himmelberg and Petersen, 1994). Second, the order pecking theory (Myers and Majluf, 1984) suggests that for highly innovative firms the degree of asymmetric innovation is higher, which makes it harder to finance R&D activity with debt and leads to a lower leverage ratio. Lastly, Ghosh (2012) finds that the leverage ratio is lower for Indian firms with a high R&D intensity.

III. SAMPLE SELECTION AND METHODOLOGY

Hypotheses

The main objective of this study is to research whether the acquirer’s R&D output elasticity is affected by M&A activity. First, I focus on identifying an average treatment effect of M&As on the firm’s R&D output elasticity. Second, I review whether different levels of RQ affects abnormal stock returns. Finally, I research whether M&A activity drives underinvestment in R&D. Hypothesis I: Mergers and acquisitions negatively affect firm-specific R&D output elasticity. In the literature, certain hypotheses attempt to explain the relationship between M&A activity and innovation performance of a firm. Negative effects are attributed to the distraction of managerial energy and time when integrating a newly acquired company (Hitt et al., 1991); the lack of motivation for the acquired firm’s researchers (Calderini et al., 2003); or to organizational and market differences (Cassiman et al., 2005). Also, smaller firms can rely on governance advantages. Smaller firms have fewer employees and, thus, decision makers are more involved with the customers and the technology of the firm. The positive effects of M&A on the innovation process of the firm are also attributed to economies of scale. When acquiring companies become larger, they establish a larger and more stable stream of internal capital (Himmelberg & Petersen, 1994), which can be used for risky R&D projects. The acquiring companies can absorb risks over a broader portfolio of projects, benefit from overcoming indivisibilities in R&D projects (Calderini & Garrone, 2003), or allow further specialization of managers and scientists (Cassiman et al., 2005). Also, firms can benefit from carrying out multiple R&D projects under the same roof to stimulate the cross-fertilization of ideas (Henderson & Cockburn, 1996) or carrying out multiple R&D projects increases the likelihood that R&D projects that are failed for a certain application can be given another application elsewhere in the firm. Knott and Vieregger (2016) find that a puzzle of irrational behaviour exists where the level of R&D expenses increases with scale, while R&D productivity decreases with scale. This hypothesis tests whether M&A and increased economies of scale affect the efficiency to profit from R&D expenses.

Hypothesis II: The acquirer’s level of RQ affects abnormal stock returns when

performing M&A activity.

Cooper et al. (2015) found a positive relationship between a firm’s RQ and contemporaneous and subsequent firm value, measured in market-to-book value of the firm. Therefore, the firm’s RQ is relevant to the stock price, which is compounded into the stock price of a firm per the efficient market hypothesis (Fama, 1970). Also, an informational paper made available by Wharton (see Appendix II) shows much higher equally weighted cumulative returns for firms that have RQ values above two, compared to the S&P500 and firms with lower RQ values. In a successful M&A, the target firm is incorporated into the acquirer, leading to a larger scale of operations and the acquirer’s access to the knowledge and technology of the target firm. The change of control over the target’s assets and technology by the acquirer could lead to a RQ synergy, which is caused when new investments in R&D contribute more to the firm’s revenues than they would have contributed to the pre-merger individual firms. Newly formed firm combinations where the acquirer has a higher pre-merger RQ, should profit more from new investments and/or technology than firm-pair combinations with an acquirer with a lower RQ and less ability to profit from its investments in R&D. Differences in pre-merger RQ should cause different revenues expectations when performing M&As, which leads to differences in abnormal stock returns.

Hypothesis III: M&A activity drives post-merger underinvestment in R&D.

Healy et al. (1992) find an increase in cash flow operating returns in the five years following a merger, which is caused by an increase in asset productivity. Later, Gugler (2003) finds that also increased market power could increase the cash flow operating returns. Healy et al. (1992) do not find evidence for decreased capital expenditures or R&D expenses. Szücs (2014), Ornaghi (2009), Hitt et al. (1991), Desyllas and Hughes (2010), and Stiebale et al. (2011) find a lower post-M&A R&D intensity, resulting from increasing sales and/or lower R&D expenditures. These findings suggest that firms spend relatively less on R&D as they grow. Economically. this does not make sense, as larger firms have a steadier cash flow and should be able to finance more or riskier R&D projects (Himmelberg & Petersen, 1994). Also, riskier and/or costlier projects should have higher returns to justify increased risks. If M&A activity negatively affects R&D expenses and R&D intensity, it is also plausible that it drives underinvestment in R&D.

M&A Deal Sample

This study focuses on US M&A deals for which the acquiring firm is available in COMPUSTAT North American Annual Database (Compustat), the Center for Research in Security Prices (CRSP) and for which a research quotient is available in the Wharton Research Data Services (WRDS) Research Quotient database, for the period from 1 January 1990 to 31 December 2010. The M&A deal data is extracted from the Thomson ONE database. I use a dataset of US firms because it is easier to relate and interpret the results. Most research around M&A activity has a strong focus on the US due to good data availability.

I chose the period of 1990-2010 to make sure enough M&A deals are available with at least three years of available financial data before a merger or acquisition, and at least three years of data after the year of the merger or acquisition. This range of data is necessary for forming a weakly balanced panel data, to perform a differences-in-differences, and to test the robustness of my findings. The financial data is retrieved from Compustat. Using a combination of date and historical ‘point-in-time’ CUSIP (known as NCUSIP) from CRSP, I match deal data with financial data. I only include deals in the sample when the acquirer firm is covered by Compustat.

The RQ data is available via Wharton WRDS. This database includes RQ values that are calculated using Compustat firm data, using a rolling 7-year window from 1965-2010 and with a minimum of 6-years of data within a 7-year window. The M&A deal sample is computed using the selection requirements that also used by Bena and Li (2015);

1. KEEP: Deals that are tagged as merger or as an acquisition.

2. KEEP: Acquirer did not hold more than 40% of the shares of the target before the deal-announcement and not less than 90% after the completion of the deal.

3. KEEP: Deals with total assets of minimal US$ 1 million in the year of the deal for the acquirer (all monetary figures in 2010 US dollars).

4. KEEP: Deals with a minimum deal value of US$ 1 million in the year of completion of the M&A.

5. EXCLUDE: All acquirers that are active in the financial sector, or have a Standard Industrial Classification (SIC) code between 6000-6999.

When applying these filters to form the final sample, I have 2,757 deals where the acquirer is covered by Compustat and CRSP. This sample size is comparable with the samples used in earlier research, such as research by Bena and Li (2015) who have a

sample size of 2,621 with similar parameters. For 960 deals out of the 2,757 deals in my sample, the RQ of the acquirer is available. RQ is missing for firms that do not report R&D expenses, or when no R&D expenses are reported for at least 6 years in a rolling 7-year window.

All firm characteristics are measured as of the end of the fiscal year before the date of completion. Following Bena and Li (2015), the measurement window is [y-3; y+3]. Year y-1 is the fiscal year with the largest overlap with the year before the M&A is completed. Year y corresponds to the fiscal year in which the M&A transaction is completed and year y+1 is the fiscal year with the most overlap with the first fiscal year after the year of completion. Using a measurement window from y-3 to y+3, instead of fiscal years, makes it more practical to perform the statistical analyses to measure the effect of M&A activity. After excluding deals for which data is not available for the event window [y-3; y+3], and excluding deals of firms that have overlapping event windows because of multiple acquisitions or mergers within 7 years, I have a final sample of 545 deals from 419 unique acquiring firms. Table I shows how the deal sample is constructed.

Table I

Breakdown of Successful M&A Deals sample selection This table contains the steps that where performed to form the M&A Deal sample. The table also shows the number of deals that were kept or excluded at every step.

# of Deals Keep/Exclude Description

978,644 All deals in Thompson ONE

302,701 Keep Deals with US target

263,946 Keep Deals with US acquirer

160,368 Exclude Financials (SIC 6000-6999)

142,250 Keep Date announced between 1 January 1990 to 31 December 2010

112,115 Keep Date effective/unconditional between 1 January 1990 to 31 December 2010

102,354 Keep Percent of shares acquired in transaction between 40 to high

100,083 Keep Percent of shares owned after transaction between 90 to high

99,854 Keep Deal attitude friendly, hostile, neutral

41,376 Keep Acquirer total assets (millions of dollars) last 12 months between 1 to high

21,019 Keep Deal value (millions of dollars) between 1 to high

19,968 Keep Acquirer matched with record in CRSP database (NCUSIP – CUSIP6)

2,757 Keep Acquirer matched with record in Compustat database 960 Keep RQ of acquirer is available for at least 1 year 545 Keep When RQ of acquirer is available for the period y-3 to y+3. Y is the fiscal year of the completion of the deal Control Samples

To identify the effect of the M&A activity on R&D elasticity, I construct two control samples that function as benchmarks. The first control sample includes firms without records in Thomson ONE of any M&A activity between 1 January 1990 and 31 December 2015. This sample is formed using the complete Compustat database. The second control sample contains firms with at least one failed and/or withdrawn bid recorded in Thomson ONE. Tables II and III present a breakdown of how the control samples are constructed. [INSERT TABLE II HERE] [INSERT TABLE III HERE] The M&A deal sample is also used to test whether RQ affects firm value. I matched 432 out of the 545 firms of the M&A deal sample with usable returns. Descriptive statistics

In Table IV, Panel A contains the descriptive statistics for the M&A sample, whereas Panel B and Panel C contain the descriptive statistics for the two control samples. Overall, the mean RQ in the M&A deal sample (10.62) is slightly larger than for the No M&A control sample (9.61) and the Failed M&A sample (9.76). However, due to the large and overlapping standard errors in all samples, the mean RQ values are statistically equal. The mean RQ values are lower than the mean RQ calculated by Cooper et al. (2015), as are the standard deviations. These differences can arise from the different time frames and/or sample selection, as Cooper et al. (2015) also uses the WRDS Research Quotient database. Overall, the mean RQ seems to decline after 1990. Figure 1 shows the density histograms of the M&A sample. Panel B shows the RQ in fiscal year 2001, which distribution can be used to compare with histogram of Cooper et al (2015).

Figure 1 – This Figure shows the density distribution of Research Quotient of the Deal sample. Panel A shows the density histogram with RQ values of all fiscal years (1990-2010) and Panel B only for the fiscal year 2001. All RQ values are winsorized at 5%. Panel A. Histogram of all RQ-observations for 1990-2010 Panel B. Histogram of RQ in fiscal year 2001

Furthermore, the ratio of mean R&D expenses/total assets and mean R&D intensity is slightly lower in the M&A sample, which indicates that that the firms that perform M&A activity spend less on R&D for every dollar of assets on the balance sheet. The average book value of total assets in the M&A deal sample is US$ 3.150 billion, which is comparable to the average book value reported by Cooper et al. (2015). The average book value of total assets of the No M&A control sample is US$ 2.014 billion, which indicates that the firms in the M&A sample have larger assets on average. The average book value of total assets in the failed merger M&A control sample is US$ 3.837 billion, which is larger than the US$ 3.150 billion of the M&A deal data. Also, on average the firms in the M&A deal sample yearly spend US$ 81.257 million on R&D, slightly higher than the No M&A control group (US$ 52.587 million), but lower than the failed merger M&A control sample (US$ 257.534 million). Lastly, the ROA of firms in the M&A deal sample is comparable to that of the failed merger sample, but much higher than the ROA of the No M&A sample (7.70% against -3.31% and 9.64%). The M&A sample is comparable with the sample used by Bena and Li (2015) in terms of economic values, but differs for the Book-to-Market ratio, the ROA, and the total assets. [INSERT TABLE IV HERE] Methodology hypothesis I

The methodology to test the first hypothesis is inspired by Bena and Li (2015). I perform a quasi-experiment with a Differences-in-Differences regression on panel data.

The panel data contains firms that are labelled as acquirer in a M&A deal sample from Thomson ONE. The control samples are used to create the baseline. First, I run the following regression:

RQί,t = α + β1 Afteri,t + β2 Treatmenti,t + β3 Afteri,t * β4 Treatmenti,t + β5 Controlsi, t + Firm FEi

+ ɛit RQi,t. is the dependent variable and stands for the firm specific output elasticity of R&D, also known as the Research Quotient. It measures how a 1% increase in R&D affects the firm’s revenue in percentage terms, while holding all other inputs and their elasticities constant.

Afteri,t is equal to one for the post-merger period [y+1; y+3], and zero for any other year.

Treatmenti,t is equal to one for firms that are in the treatment group (successful M&A

deal) and zero for the control deals with withdrawn bids. In this model, Treatmenti,t is

the independent variable.

Firm Fixed Effectsi is included to control for time-invariant differences among firms.

The variable ɛit is a random error term, with the assumption ɛit ~ N(0). It is not possible

to include time fixed effects, as I reorganised the data into a [y-3; y+3] format.

Controlsi,t includes several firm characteristics identified in literature as related to firm performances. The characteristics include leverage (e.g. Ghosh, 2012; Heron & Lie, 2002; Jensen, 1986), total assets, revenues, book-to-market, capital expenditures to control for firm size (e.g. Heron & Lie, 2002; Powell & Stark, 2005), cash (e.g. Gao, 2011; Harford, 1999; Himmelberg & Petersen, 1994; Jensen, 1986; Martynova et al., 2006) and return on assets (e.g. Cooper et al., 2015; Lentz & Mortensen, 2008). See Appendix I for more details about the variables used. Methodology hypothesis II To test the second hypothesis, I performed a market model event study using Eventus. By performing an event study, I analysed whether the firms abnormal stock returns are affected by performing M&A activity for different levels of RQ. Abnormal returns (AR) are used to control from exogenous factors like macro-economic and industry-specific developments. AR and average abnormal returns (AAR) are calculated following Campbell et al. (1993) and MacKinlay (1997), using a CRSP Equally Weighted Market

index, market-adjusted returns, an estimation period that is equal to 170 days [t-200; t-

30], and an event windows of 20 days before and 7 days after the Merger Announcement Day [t-20; t+7]. Information about M&A activity can be anticipated in the stock price before the announcement date, and therefore, the event window should begin before any M&A activity is announced. I used Ordinary Least Squares (OLS) as estimation method and a minimum estimation period length of 90 days. For the event study, I calculate the AR of the firm’s securities as follows: ARίτ = Rίτ – E(Rίτ|Xτ)

Where ARίτ is the abnormal return and Rίτ the actual stock return from prices and

dividends. Each daily stock return is calculated from the previous day with a non-missing price to the current day. E(Rίτ|Xτ) is the normal return for time τ, and Xτ the

conditioning information for the normal return model. When the event date is not an existing trading date, I take the previous trading date available before the event date. Second, for any firm ί, the market model is: (1) Rίτ = αί + βίRmt + είτ Where: E(είτ = 0) var(είτ) = σ2ει

where Rίτ and Rmt are the period-t returns on firm ί and the market portfolio, and είτ is

the zero mean disturbance term. αί, βί and σ2εί are the parameters of the market model,

where αί is the return on the security more than what would be predicted by the Capital

Asset Pricing Model, and βί explains the tendency of the stock to follow swings in the

market. Given this market model, the abnormal returns can be computed. Let ARίτ, τ = T1

+ 1,…,T2 be the sample of L2 abnormal returns for firm ί in the event window. When

using the market model approach, the abnormal return is: (2) ARίτ = Rίτ – Gί – 9ίRmτ where: Gί = Hι - 9ί Hm (IJK – MJ)(INK – MN) (INK – MN)? OP Q R OS T P

9ί = (IJK – MJ)(INK – MN) (INK – MN)? OP Q R OS T P U-K V U: W - where: Hι = _X-- U-_{K V U[ W -}+YZ Hm = -X- +\Z U-K V U: W -

Rίτ and Rmτ are the period-t returns on security i and the market. The Gί, 9 coefficients

are the OLS estimates of the intercept and the slope of the market model regression. The abnormal return is the disturbance term of the market model. Under the null hypothesis, conditional on the event window market returns, abnormal returns will be jointly normally distributed with a zero-conditional mean and conditional variance σ2_{(ARίτ), where} σ2_{(ARίτ) = σ2εί +} -X-[1 + (Rmτ- μm)2/σ2m] The abnormal return observations can be aggregated for the event window and across observations of the event. Given N events, the sample aggregated abnormal returns for period τ is: #+τ = -] #+YZ ] ίV- The average abnormal returns (AAR) can be aggregated over the event window for each security ι. For any interval: _##+(τ1,τ2) = K?_{KV K-}#+Z

To test whether the calculated AARs and CAARs are reliable, I both perform a parametric and a non-parametric test. The Patell test (Patell, 1976) is used as parametric test and the Wilcoxon signed-rank test as non-parametric test (Wilcoxon, 1945).

To test whether M&A activity affects ARs for different levels of RQ differently, I first divide the sample into four subsample portfolios. The first subsample includes firms that are in the lowest 25th percentile based on RQ, and firms in the fourth subsample are in the highest 25th percentile based on RQ. To further validate my

results, I regress RQ and controls on ARs. In this model, RQi,t is the independent variable

while ARi,t is the dependent variable:

AR

i,t

=

α + β1 RQi,t + β2 Controlsi,t + ɛi,t

Where: ARi,t is the Daily Abnormal Return of acquirer i for time t. RQi,t is the Research Quotient of firm i for time t.

Controlsi,t includes control variables to control for omitted variable bias. The controls

include Leverage, Total Assets, Revenues, Book-to-Market ratio, Cash on Balance Sheet and Return on Assets. Appendix I shows how these control variables are constructed.

Methodology hypothesis III

The third hypothesis is tested by looking at RQ-Difί,t as a dependent variable. RQ-Difί,t is

the difference between calculated firm-specific optimal R&D expenses and historical R&D expenses. The calculated firm-specific optimal R&D expenses are retrieved from the WRDS Research Quotient database and are the result of the optimisation of a model that links R&D with market values (Knott & Vieregger, 2016). Furthermore, the same Successful M&A deal sample, No M&A control sample and Failed M&A control sample are used as for testing hypothesis I.

To test the third hypothesis, the following model is regressed:

RD-Difί,t = α + β1 Afteri,t + β2 Treatmenti,t + β3 Afteri,t * β4 Treatmenti,t + β5 Controlsi,t +

Firm FEi + ɛit

This model is comparable with the differences-in-differences approach that is used for the analysis of the first hypothesis. The variables Afteri,t , Treatmenti,t and Controlsi,t are the same variables as used in the regression model for the analysis of the first hypothesis.

IV. ANALYSES

This section presents the results of the empirical research. In the first section the results of the main hypothesis are explained. Subsequent sections explain and discuss the results of the second and third analysis.

4.1 Do M&A negatively affect firm-specific R&D output elasticity?

Table V presents the results of the differences-in-differences (DiD) panel data regression for hypothesis I. Panel A shows the results of the Successful M&A Activity sample versus the No M&A Activity sample (control sample 1). For all three regressions, the After coefficient is negative and significant at 1%. This means that RQ is declining over the years for both samples, i.e. the mean RQ has some sort of negative trend. No effect is found for the M&A Activity * After (the DiD regressor) in the first two regressions. This means that Successful M&A Activity does not affect the RQ of the acquirer in the first three years. The second regression also includes control variables leverage, cash and return on assets. Of these variables, only leverage affects RQ with -0.016 and is significant at 5%. This implies that leverage negatively affects the RQ of acquirers and should therefore be treated as a firm characteristic. However, the negative effect of leverage contradicts with findings of Clark and Ofek (1994) which find no significant relation between post-merger firm performance and Harford (1999) who finds a positive effect.

The third differencs-in-differences regression also includes total assets, revenues, capital expenditures and the book-to-market ratio. Of these extra control variables, only book-to-market ratio affects RQ with -0.968, but is significant at 10%. Cooper et al. (2015) found that RQ affects market-to-book ratio with 0.537 at 1% significance, thus it looks like RQ and market-to-book have a (simultaneous) relationship. The negative result implies that firms with higher market-to-book have, on average, lower RQ values. In the third regression, the coefficient of the M&A Activity * After (DiD) regressor is 1.046. This implies that Successful M&A Activity positively affects the acquirers RQ, when compared to firms that did not conduct M&A activity. This result implies that M&A activity enables firms to profit more from investments in R&D, compared to firms that do not perform M&A activity.

Figure 2 - Average RQ of successful M&A activity (red line) and control sample 1 with no M&A activity (left, blue line) and control sample 2 with failed M&A activity (right, blue line).

Panel A further shows that there is no need to control for total assets, total revenues, total cash on balance sheet, or capital expenditures, as all these controls are statistically equal to zero. Therefore, these controls should be excluded to avoid bad control problems.

Panel B shows the results when controlled with the failed M&A activity sample (control sample 2). Again, the coefficient of the After regressor is negative for all three regressions at 1% significance. When looking at the M&A Activity * After (the DiD regressor), it does not show a significant effect in all three regressions (p-value (1) = 0.956, p-value (2) = 0.677 and p-value (3) = 0.491 respectively). This means that successful M&A activity does not affect the RQ of the acquirer in the first three years, when controlled with the failed M&A activity control sample. Even the third regression does not show a significant coefficient (+ 0.270 with a standard deviation of 0.154). This contradicts the findings of the third regression of Panel A. The results in Panel B imply that firms that do not pursue M&A activity are different from firms that do perform M&A activity in terms of RQ. The control variables do not ease away the differences with respect to RQ between the successful M&A sample and the control samples. As a result, omitted variable bias leads to non-uniform results in Panel A and Panel B. Furthermore, there is no need to include control variables in Panel B, except for ROA, which is significant at 5% in the second regression. In Panel B, the book-to-market regressor is not significant, which contradicts with the results of Panel A. Overall, it is impossible to conclude whether successful M&A activity significantly affects the acquirer’s RQ. 4.2 Does the level of RQ of the acquirer affect abnormal stock returns? Table VI includes the AARs for the four subsample portfolios with for the event window [-20, +7]. On the day of the M&A announcement [Day = 0], AAR is positive for the first

three subsample portfolios, when performing the Patell Z test, a parametric test. For the first portfolio, the effect is +1.01%*; for the second subsample portfolio, the effect is +0.49%*; for the third subsample portfolio, the effect is +1.06***; and for the fourth subsample portfolio, the effect is -0.41%**. The symbols * and *** denote statistical significance at 0.10 and 0.01 using a two-tail test. When performing the non-parametric Rank test, only the AAR of the third sub sample portfolio is significant at 1% significance.

Overall, the results indicate that M&A activity positively affects abnormal returns of firms in the first 0th _{- 75}th _{RQ percentile, while it negatively affects abnormal} returns for firms that are in the top 25th _{RQ percentile.} [INSERT TABLE VI HERE] Graph 1 shows the daily AAR for every portfolio. The daily AAR peaks at [Day = 0] for the first and third portfolio. Surprisingly, portfolios two and three peak at [Day = 1]. Graph 2 - Average Abnormal Returns (%) of four subsample portfolios for the event window [-20, +7]. Portfolio 1 includes the first 25th _{percentile of firms in terms of RQ on the event date [Day = 0] and Portfolio} 4 the 75th _{– 100}th _{percentile in terms of RQ on the event date.} Table VII shows the CAAR of for alternative event windows. Event windows [-20, +7], [-5,0] and [0, +5] have a mean cumulative abnormal return of +1.41% (p-value = 0.018), +0.90% (p-value = 0.031), and +0.65% (p-value = 0.026), respectively, and show a significant effect of M&A activity on 5% significance. Event windows [-2, +2] and [+1, +30] show a mean cumulative abnormal return of +1.43% (p-value = 0.000) and -3.01% -1,5 -1 -0,5 0 0,5 1 1,5 Av er age Ab n or m al R et u rn (% ) Event window [-20, +7]

Daily AAR in % for every portfolio

(p-value = 0.001). Surprisingly, the effect of M&A activity on CAAR in the shorter event windows is mainly positive, while the CAAR for event window [+1, +30] is negative. Only event window [-14, +14] shows no significant mean cumulative abnormal return when testing with the Patell Z test. [INSERT TABLE VII HERE] Table VIII shows the results of the regression of RQ on AR and suggest that RQ affects ARi,(0) with -0.001 at 10% significance. Due to the small effect of RQ and the low R2, the

model is not very reliable and/or exhaustive. Table IX shows the results of the regression where RQ is regressed on ARi,t for every day in the event window [-20, +7].

The results suggest there is no significant effect of RQ on ARi,t (p-value = 0.664 for the

first regression and p-value = 0.818 for the second regression when controls are included). R2 _{is small for both regressions (R}2 _{= 0.0000 and R}2 _{= 0.0003 respectively),}

and, therefore, I consider the findings inconclusive.3 [INSERT TABLE VIII HERE] [INSERT TABLE IX HERE] 4.3 Does M&A activity drive underinvestment in R&D? Table X shows the results of the third hypothesis. Panel A includes the coefficients of the model when regressed against control sample 1. The results suggest that M&A drives underinvestment in R&D. The coefficient of M&A Activity * After is 18.390, 17.890 and 18.420 respectively. All coefficients are significant at 1% (p-values 0.000, p-value = 0.001, and p-value = 0.001). This means that, in comparison with firms without M&A activity, M&A activity increases the difference between the calculated optimal R&D spending and the actual R&D spending of the firm. When including control variables, cash is significant at 1% and has a negative coefficient of -0.228 (p-value = 0.007), while the control variable total assets has a positive coefficient of 0.001 with p-value = 0.000. All other control variables are not significant and could be removed from the regression model.

Panel B of Table X shows the results of the analysis when regressed against control sample 2 (failed M&A activity). I do not find a significant effect for M&A Activity * After. Without control variables, the coefficient of M&A Activity * After is -5.918 with

3 _{When using CARs as dependent variable and event window [-1, 1], similar results are found. Calculations}

p-value = 0.224. When including control variables, the coefficient is -5.442 with p-value = 0.229 and -2.063 with p-value = 0.925 respectively. Overall, the results of Panel B suggest that M&A Activity does not drive underinvestment when compared with firms with a Failed M&A Activity history. [INSERT TABLE X HERE] 4.4 Sensitivity Analysis

I perform a sensitivity analysis to test whether the results of the analysis of the first hypothesis are robust. A differences-in-differences approach is used, but instead of using all years [y-3; y+3], I used different time periods of the before and after the M&A activity period. The statistical software that I have used for the analyses (STATA) automatically calculates an average of all years before the year of the M&A announcement (when After = 0), and for all years after the M&A activity event (when After = 1). By using different time periods, I test whether and how individual years influence the average treatment effect. The results of the sensitivity analyses for the first hypothesis appear in Table XI.

[INSERT TABLE XI HERE]

V. CONCLUSION

5.1 Results & Discussion

The main objective was to research whether the acquirer’s R&D output elasticity is affected by M&A activity. The first hypothesis states: Mergers and acquisitions negatively affect firm-specific R&D output elasticity.

This hypothesis should be rejected as I did not find a significant negative effect of successful M&A activity on R&D output elasticity. When testing the successful M&A activity sample versus the no M&A sample, I find a significant positive effect of 1.046. When controlled with the failed M&A activity sample no significant effect is found. This means that either the coefficients suffer from omitted variable bias or the sample size of the failed M&A activity sample has insufficient observations. If the results of the successful M&A activity sample versus the no M&A activity sample are robust, I should conclude that firms that successfully perform M&A activity, can grow their operations without a negative effect on their ability to profit from R&D in the first three years after M&A activity. This contradicts with the underlying arguments that smaller firms have governance advantages and are in general more innovative, but is in line with the findings of Knott & Vieregger (2016), who find that R&D spending and R&D productivity both increase with firm size. Overall, RQ is declining on average for all firms in my samples between 1 January 1990 and 31 December 2010.

The second hypothesis states: The acquirer’s level of RQ affects abnormal stock returns when performing M&A activity. This hypothesis should be rejected. First, AR does not increase with RQ. I found a negative CAAR for acquirers in the highest 25th _RQ

percentile, while acquirers in the 0th _{– 75}th _{RQ percentile have a positive CAAR for the}

event window [-20, +7]. Furthermore, I found that successful M&A activity affects cumulated AAR with +1.41% (p-value = 0.018) for the event window [-20, +7]. This contradicts with earlier research that finds positive abnormal returns for shareholders of the target firm and negative abnormal returns for the acquirer firm (Andrade et al., 2001), but is in line with a positive effect reported by Goergen and Renneboog (2004) or Floreani and Rigamonti (2001), who find CAARs of 0.39% and +3.65%, respectively, for comparable event windows and time periods. A possible explanation for the contradicting results with Andrade et al. (2001) could be caused by sample selection bias. This selection bias could be introduced by only including firms for which the RQ is available in the WRDS Research Quotient database.

When looking at whether the AARs for the event day [Day = 0] are affected by RQ, I found that RQ affects ARi,(0) with -0.001 at 10% significance. Due to the low R2 of

are similar. In conclusion, both the level of RQ of acquirer and RQ itself do not affect the ARs of the acquirer.

The third hypothesis states: M&A activity drives post-merger underinvestment in R&D. This hypothesis should be partly accepted. First, I found a positive average treatment effect of US$ 18.39 million, US$ 17.89 million and US$ 18.42 million respectively at 1% significance when controlling with the firms that did not perform any M&A activity. The positive coefficients imply that the difference between historical R&D expenses and optimal calculated R&D expenses becomes larger after performing M&A, and, thus, drives underinvestment in R&D. In contrast, I did not find significant results when controlling with the failed M&A sample. 5.2 Shortcomings of this research

Due to practical reasons and data availability, this research has certain shortcomings. First, the control samples in the first and third hypotheses consisted of random periods which coincided with the timeframes in my deal sample and for which at least 7 years of data was available. I did not use fiscal years in the analysis, and therefore, no time-fixed effects are included. This could introduce omitted variable bias. Second, the control sample of failed or withdrawn M&A deals is relatively small (N=32), even compared to a failed merger sample used by Bena and Li (2015), which used N=60 firms. Third, this research could also contain some selection bias, as I could only include firms which reported R&D spending and where RQ was available. Selection bias could also come from only using firms I could match to the M&A deals with Compustat data. Fourth, this research does not account for the endogeneity of the merger pair formation. The decision to pursue M&A activity is not an exogenous process, but can be related to the firm specific characteristics. This endogeneity of the merger pair formation could have been solved by using some propensity score method (Bertrand & Zitouna, 2008; Heyman et al., 2007).

Finally, the conclusions of this research cannot be used for interpretations related to whether firm’s innovation power is affected, even when Knott (2008) named the RQ of firms an alternative innovation measurement. Whether inventions or results from R&D expenses should be considered as real innovation is a subjective matter and subject to local and international patent law. In general, the literature states that other measurements should be used to link innovation to M&A, like forward patent-citations or market value of patents based on stock price reactions (Bloom et al., 2002; Hall et al., 2005).

5.3 Ideas for future research

Future research could focus on differences between large firms and small firms that conduct M&A activity. Smaller firms tend to be more innovative (Cohen, 2010) and, therefore, a different effect could exist when performing M&A activity. Also, focussing on the type of R&D projects could give new insights. Per the model developed by Rosen (1991), larger firms tend to execute R&D projects that enhance existing technology, instead of a focussing on radical innovation that could improve their price-cost ratios (Mansfield et al., 1981). Also, more focus on underlying drivers could influence outputs, such as market or technology proximity between the merger pair, corporate governance measures, or a measure that includes the amount of specialization of researchers and

VI. REFERENCES

Abrams, D.S., Akcigit, U., Popadak, J. (2013), ‘Patent Value and Citations: Creative Destruction or Strategic Disruption?’, NBER Working Paper, No. 19647, Available at: http://scholarship.law.upenn.edu/cgi/viewcontent.cgi?article=1497&context=faculty_s cholarship [Accessed: 4 August, 2016] Andrade, G., M. Mitchell, and E. Stafford (2001), ‘New evidence and perspectives on mergers’, Journal of Economic Perspectives, Vol. 15, pp. 103–120.

Bena, J., Li, K. (2014) ‘Corporate Innovations and Mergers and Acquisitions’, Journal of Finance, Vol. 69, pp. 1923-1960.

Bertrand, O., Zitouna, H. (2008), ‘Domestic versus cross-border acquisitions: which impact on the target firms’ performance?’, Applied Economics, Vol. 40(17), pp. 2221-2238.

Bloom, N., Van Reenen, J. (2002), ‘Patents, Real Options and Firm Performance’, Economic Journal, Vol 478, pp. 97-116.

Blonigen, B.A., Pierce, J.R. (2015), ‘The Effect of Mergers and Acquisitions on Market Power and Efficiency’, Working Paper, Available at:

http://pages.uoregon.edu/bruceb/blonigen_pierce_aug2015.pdf [Accessed: 4 August 2016]

Calderini, M., Garrone, P., Scellato, G. (2003), ‘Mergers and acquisitions and innovations strategies’ in: Calderini, M., Garrone, P., Sobrero, M. (Eds.), Market Structure, Corporate Governance and Innovation. Edward Elgar, Cheltenham.

Calderini, M., Garrone, P., Scellato, G. (2003), ‘The effect of M&As on the innovation performance of acquired companies’ in: Calderini, M., Garrone, P., Sobrero, M. (Eds.), Market Structure, Corporate Governance and Innovation. Edward Elgar, Cheltenham.

Campbell, C., Wasley, C. (1993), ‘Measuring Security Price Performance Using Daily NASDAQ Returns’, Journal of Financial Economics, Vol. 33, pp. 73-92

Campa, J.M., Hernando, I. (2004), ‘Shareholder Value Creation in European M&As’, European Financial Management, vol 10(1).

Carline, N.F., Linn, S.C., Yadav, P.K. (2007), ‘Operating Performance changes associated with Corporate Mergers and the Role of Corporate Governance’, Journal of Banking & Finance, Vol. 33, pp. 1829-1841.

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 Policiy, Vol. 2, p. 34. Caves, R.E. (1989), ‘Mergers, takeovers, and economic efficiency: Foresight vs. Hindsight’, International Journal of Industrial Organization, Vol. 7, pp. 151-174. Clark, K., Ofek, E. (1994), ‘Mergers as a Means of Restructuring Distressed Firms: An Empirical Investigation’, The Journal of Financial and Quantitative Analysis, Vol. 29(4), pp. 541-565. Cohen, W. (2010), ‘Fifty Years of Emperical Studies of Innovative Activity and Performance’ in B.H. Hall and N. Rosenberg (eds.), Handbook of Economics of Innovation, Amsterdam, North Holland Elsevier. Cooper, M., Knott, A.M., Yang, W. (2015), ‘Measuring Innovation’, Working Paper, Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2631655 [Accessed: 9 June, 2016] Czarnitzki, D., Hall, B.H., Oriani, R. (2006), ‘Market Valuation of US and European Intellectual Property’, in D. Bosworth and E. Webster (eds.), The Management Of Intellectual Property, Cheltenham, UK. Desyllas, P., Hughes, A. (2010), ‘Do high technology acquirers become more innovative?’, Research Policy, Vol. 39, pp. 1105-1121. Dutta, S., Jog, V. (2009), ‘The long-term performance of acquiring firms: A re-examination of an anomaly’ Journal of Banking & Finance, Vol. 33, pp. 1400-1412. Fama, E.F. (1970), ‘Efficient Capital Markets: A Review of Theory and Empirical Work’, The Journal of Finance, Vol. 25(2), pp. 383-417. Floreani, A., Rigamonti, S. (2001), ‘Mergers and shareholders wealth in the insurance industry’, Working Paper, Available at: https://papers.ssrn.com/sol3/papers .cfm?abstract_id=267554 [Accessed: 13 November, 2016] Floreani, A. and Rigamonti, S. (2001). Mergers and shareholders’ wealth in the insurance industry, Working Paper, Universita Cattolica del S. Cuore. Gao, N. (2011), ‘The Adverse Selection Effect of Corporate Cash Reserve: Evidence from the Acquisitions Solely Financed by Stock’, Journal of Corporate Finance, Vol. 17(4), pp. 789-808. Ghosh, A. (2001), ‘Does Operating Performance Really Improve Following Corporate Acquisitions?’, Journal of Corporate finance, Vol. 7(2), pp. 151-178.

Ghosh, S. (2012), ‘Does R&D intensity influence leverage? Evidence from Indian firm-level data’, Journal of International Entrepreneurship, Vol. 10(2), pp. 158-175. Goergen, M., Renneboog, L. (2004), ‘Shareholder Wealth Effects of European Domestic and Cross-Border Takeover Bids’, European Financial Management, Vol. 10(1), pp. 9-45. Greenwood, J., Jovanovic, B. (1999), ‘The Information-Technology Revolution and the Stock Market’, American Economic Review, Vol. 89(2), pp. 116-122. Griliches, Z. (1981), ‘Market value, R&D and patents’, Economic Letters, Vol. 7(2), pp. 183-187. Griliches, Z., Hall, B., Pakes, A. (1991), ‘Patents and Market Value Revisited: Is There ad Second (Technological Opportunity) Factor?’, Economics, Innovation and New Technology, Vol. 1, pp. 1983-2201 Available at: http://www.nber.org/papers/w2624.pdf [Accessed: 18 October, 2016]

Gugler, K., Mueller, D.C., Yurtoglu, B.B., Zulehner, C. (2003), ‘The effects of mergers: an international comparison’, International Journal of Industrial Organization, Vol. 21, pp. 625-653. Hall, B.H. (2000), ‘Innovation and Market Value’, in Ray Barrell, Geoff Mason and Mary O’Mahoney (eds.), Productivity, Innovation and Economic Performance, Cambridge: Cambridge University Press, pp. 177-198. Hall, B.H., Adam, J., Trajtenberg, M. (2005), ‘Market Value and Patent Citations’, RAND Journal of Economics, Vol. 36(1), pp. 16-38. Harhoff, D., Narin, F., Scherer, F.M., Vopel, K. (1999), ‘Citation Frequency And The Value of Patented Inventions’, The Review of Economics and Statistics, Vol. 81(3), pp. 511-515.

Healy, P.J., Palepu, K.G., Ruback, R.S. (1992), ‘Does corporate performance improve after mergers?’, Journal of Financial Economics, Vol. 31, pp. 917-949. Henderson, R., Cockburn, I. (1996), ‘Scale, scope and spillovers: the determinants of research productivity in drug discovery’, RAND Journal of Economics, Vol. 27 (1), pp. 32-59. Heron, R., Lie, E. (2002). ‘Operating performance and the method of payment in takeovers’, Journal of Financial and Quantitative Analysis, Vol. 37, pp. 137-156. Heyman, F., Sjöholm, F., Tingvall, P.G. (2007), ‘Is there really a foreign ownership wage premium? Evidence from matched employer-employee data’, Journal of International Economics, Vol. 72(2), pp. 355-376.