Acquiring Innovation

Merih Sevilir Kelley School of Business Indiana University Bloomington, IN 47405 [email protected] (812) 855-2698

Xuan Tian Kelley School of Business Indiana University Bloomington, IN 47405 [email protected] (812) 855-3420

* We are grateful for comments and suggestions from Nandini Gupta, Gordon Phillips, Amit Seru, seminar participant at Indiana University, and conference participants at the 2011 SFS Finance Cavalcade conference. We remain responsible for all errors and omissions.

Acquiring Innovation

ABSTRACT We find a positive association between Mergers and Acquisitions (M&A) activity of a firm and its subsequent innovation outcome measured by the number and the novelty of the patents the firm obtains. Post-acquisition patent outcome of the acquirer firm is positively related to the research and development (R&D) intensity of the target firm as well as the target firm’s patent outcome prior to the acquisition. The positive association between M&A activity and innovation appears at least as significant as that between R&D and innovation, and is more pronounced for mature firms and non VC-backed firms. Acquisitions involving target firms with greater R&D intensity and greater patenting activity generate significantly positive announcement returns. In addition, innovativeness of the target firm is positively related to acquirer abnormal returns at announcement as well as acquirer’s long-term stock price performance after deal completion. Overall, our evidence uncovers a previously unrecognized role of M&A and suggests that acquiring innovation is an important motive for undertaking M&A. Our evidence also shows that acquisitions involving more innovative targets lead to greater value-creation both in the short run and in the longer term after the acquisition.

Keywords: Mergers and Acquisitions, Innovation, Research and Development, Firm Value JEL classification: G34, O31, O32, G24

1. Introduction Mergers and Acquisitions (M&A) represent one of the most significant and controversial corporate activity. An extensive body of research establishes that firms spend considerable amount of resources to acquire other firms in order to promote growth, improve efficiency, achieve risk reduction, and secure competitive advantage with respect to competitors in the product market. Although we have a detailed understanding of these traditional motives of M&A, there has been relatively little research on the effect of M&A activity on the innovation output of a firm. In this paper, we study the relation between M&A activity and innovation output of a firm, and present evidence of a strong positive association between M&A volume of a firm and the innovativeness of the firm measured both by the number and the novelty of the patents the firm obtains subsequent to its M&A activity. Hence, our paper uncovers a novel and bright role of M&A activity as an important contributor to innovation-the main growth engine in modern firms. The innovation motive of acquisitions is consistent with anecdotal accounts and the view of practitioners in innovation intensive industries. Based on a survey of 381 executives in the pharmaceutical and biotechnology sectors, Marks&Clerk, an international intellectual property group, reports that “82% of executives predict that big pharma will be unable to innovate sufficiently from within to replenish dwindling drug pipelines, leading to an increase in acquisitions”. 1 Established firms can increase their innovation output through M&A by acquiring firms with significant technological know-how or firms with already existing patents. Aghion and Tirole (1994) analyzes the optimal organization of innovation and shows that it may be more efficient to pursue innovative ideas by establishing specialized independent research units rather than undertaking them internally within the firm since agents exerting research effort will have stronger incentives to exert effort when they are organized as independent units and own the property rights of the innovative ideas they generate. Along similar lines, a large body of research on the theory of the firm and internal capital markets argues that large and multidivisional firms may fail to motivate employees to exert effort towards risky projects since they suffer from agency problems and informational asymmetries arising from having multiple

1

“Patent cliff to drive industry consolidation”, NGP (Next Generation Pharmaceutical), February 14, 2011. 1

divisions competing for corporate resources. 2 Hence, it may be a more efficient strategy for large and mature firms to source innovation by acquiring small innovative firms rather than undertaking it internally within the firm. Consistent with these views, Seru (2010) presents evidence that target firms in diversifying mergers exhibit a significant reduction in the number and the novelty of innovation they conduct subsequent to the merger. In addition, the deterioration in the research productivity of the target firm is stronger for acquirers with a more active internal capital market. The paper also documents that acquirers respond to the postmerger drop in the innovation productivity of the target firm by shifting R&D activity outside firm boundaries by establishing strategic alliances and joint ventures with other firms. Closely related to this evidence that such firms move R&D activity outside firm boundaries as a way to deal with weak incentives to conduct innovation inside the firm, our paper focuses on whether firms rely on external acquisitions to enhance their innovation productivity by acquiring target firms with already existing patents or innovative technologies close to be patented. Put differently, we examine the role of acquisitions to source innovation from outside firm boundaries. We find a strong positive relation between the volume of M&A transactions of a firm and both the number and novelty of the patents it obtains subsequent to its M&A activity. Acquirers with greater M&A volume generate more patents in the contemporaneous as well as subsequent two years of their M&A activity than acquirers with lower M&A volume, and these patents generate more citations than those obtained by acquirers with lower M&A volume. Specifically, our univariate comparisons show that firms with high M&A volumes obtain more than 5 patents in each of the first three years following the M&A activity while firms with low M&A activity obtain fewer than 2 patents, and the difference between the two subsamples is statistically significant at the 1% level. The difference in the patent outcomes between high and low M&A activity firms is very similar to that between high and low R&D intensity firms. Similarly, our baseline regressions show that increasing M&A volume by 10% is associated with a 1.01% increase in the number of filed patents in the first year following the M&A activity, while increasing R&D intensity by 10% is associated with a 0.89% increase in the number of filed patents in the first year following the R&D expenditure. This suggests that the positive 2

See, among others, Brusco and Panunzi (2005), Rajan, Servaes and Zingales (2000), Rotemberg and Saloner (1994) and Scharfstein and Stein (2000). 2

association between M&A activity and innovation appears to be as strong as the positive association between R&D intensity (which is widely considered as the most important input for innovation) and innovation. Interestingly, the positive association between innovation outcome and M&A volume appears to be stronger for older and larger firms, suggesting that the role of M&A for increasing innovation output is more pronounced for mature and older firms. This result is consistent with the predictions of Aghion and Tirole (1994) that it may be more efficient for large firms to source innovation from outside smaller firms than to undertake it internally. While we find a positive association between acquisitions and the subsequent patent output of the acquirers, we do not have direct evidence showing that the increase in the innovativeness of the acquirers is due to the technologies innovated or possessed by the target firms. In an attempt to identify a more precise channel through which acquisitions contribute to post-acquisition innovation output of the acquirer, we investigate the relation between the preacquisition innovativeness of the target firm and the post-acquisition innovation outcome of the acquirer. Interestingly, we find that the target firm’s prior R&D intensity and patent activity are positively associated with the acquirer’s contemporaneous as well as subsequent innovation output. A 10% increase in the number of patents obtained by the target firm in the three years prior to the acquisition is associated with a 2.23% increase in the number of patents the acquirer obtains in the year following the acquisition. This evidence is consistent with the notion that acquirers increase their innovation output by acquiring innovative firms with greater R&D and patenting intensity. Although we do not observe directly whether the acquirers obtain patents for innovative technologies and products possessed by the target firms they acquire, to the extent that target firms with existing patents are likely to have innovations at the time they are acquired, this result suggests that buying innovative firms enhances the innovation output of acquirers. We next turn our attention to value implications of innovation driven acquisitions. We find that acquisitions involving target firms with higher innovation output generate positive acquirer abnormal returns at deal announcement. Similarly, such acquisitions involving a more innovative target firm lead to greater acquirer abnormal announcement returns than acquisitions involving a less innovative target firm. We also find that the pre-acquisition innovation output of the target firm is positively related to the stock price performance of the combined firm over the 3

five-year horizon after the deal, suggesting that acquisitions with an innovation focus perform well in the long-run as well. Our paper is related to the incomplete contract theory literature which focuses on firm boundaries when parties cannot write complete contracts on the allocation of property rights of innovative ideas (see, Grossman and Hart,1986; Hart and Moore, 1990; Aghion and Tirole, 1994). Our paper is also related to the literature which suggests that an alternative way for mature firms to organize innovation activities is to establish strategic alliances with specialized and small firms where specialized research intensive firms own the property rights of future innovations and mature firms later buy the innovation from the specialized units (see, e.g., Robinson, 2008; Robinson and Stuart, 2007a, 2007b; Fulghieri and Sevilir, 2009, Seru 2010). It is plausible to expect that established firms’ desire to acquire innovation from smaller and technology intensive firms affects such firms’ incentives to conduct innovation to the extent that there is an active buyer market for new technologies and products they innovate. Consistent with this view, Phillips and Zhdanov (2011) finds that the incentives of small firms to invest in R&D increase in the probability that they will be taken over by larger firms and in the liquidity of the acquisition market in their industries. Similarly, the paper provides evidence that larger firms may find it more desirable to acquire smaller innovative firms relative to conducting R&D inside firm boundaries. Our paper complements the evidence in Phillips and Zhdanov (2011) by showing that there is a strong association between the M&A activity of an acquirer and its subsequent innovation outcome measured by the number and the significance of the patents the acquirer obtains. Finally, our paper also contributes to the literature on the drivers and determinants of corporate innovation. There is an emerging body of work examining how to enhance firm innovation. Holmstrom (1989), in a simple principle-agent model, shows that innovation activities may mix poorly with routine activities in an organization. Manso (2010) shows that managerial contracts that provide tolerance for failure in the short run and reward for success in the long run are best suited for motivating innovation. The model in Ferreira, Manso, and Silva (2010) argues that private instead of public ownership spurs innovation. Empirical evidence using U.S. data shows that laws (Fan and White, 2003; and Acharya and Subramanian, 2009), venture capital financing (Kortum and Lerner, 2000), leveraged buyouts (Lerner, Sorensen, and Stromberg, 2010), corporate governance (Sapra, Subramanian, and Subramanian, 2009; and 4

Chemmanur and Tian, 2010), capital structure (Atanassov, Nanda, and Seru, 2007), product market competition (Aghion et al., 2005), investors’ attitude towards failure (Tian and Wang, 2010), corporate organizational structure (Seru, 2010), and institutional ownership (Aghion, Van Reenen, and Zingales, 2009) all affect corporate innovation. This paper is organized as follows. In Section 2, we describe our sample selection procedure and present descriptive statistics. Section 3 discusses the main results on M&A and innovation, robustness tests, and cross-sectional analysis. Section 4 investigates the relation between the pre-acquisition innovativeness of the target and the post-acquisition innovation output of the acquirer. Section 5 examines whether innovation-driven M&A activity has significant implications for firm performance in the short-run around announcement dates as well as in the long-run after deal completion. We conclude in Section 6. 2. Sample Selection, Variable Construction, and Descriptive Statistics 2.1. Sample Selection Our sample consists of all U.S. public firms recorded on the Compustat Industrial Annual Files from 1990 to 2006. Firm-year patent and citation information is retrieved from the 2006 edition of the National Bureau of Economics Research (NBER) Patent Citation database (see Hall, Jaffe, and Trajtenberg (2001) for details). M&A transaction data are obtained from the Securities Data Company (SDC) Mergers and Acquisitions database. To calculate the control variables, we collect financial statement items from Compustat Industrial Annual Files, institutional holdings data from Thomson’s CDA/Spectrum database (form 13f), venture capital (VC) backing information from the Thomson Venture Economics database, and stock return data from the Center for Research in Security Prices (CRSP) database. The final sample used to investigate the relation between M&A activity and innovation productivity consists of 105,314 firm-year observations between 1990 and 2006.

2.2. Variable Construction 2.2.1. Measuring Innovation We use the 2006 edition of the NBER Patent Citation database for information on firms’ innovation productivity. The database provides detailed information of every patent granted by the United States Patent and Trade Office (USPTO) from 1976 to 2006, including patent assignee names, the number of citations received by each patent, a patent’s application year, and 5

a patent’s grant year. Based on the information retrieved from the NBER Patent Citation database, we construct two measures of a firm’s innovation productivity. The first measure is a firm’s number of patent applications filed in a year that are eventually granted. We use a patent’s application year instead of its grant year since the application year is argued to better capture the actual time of innovation (Griliches, Pakes, and Hall (1988)). Since a simple count of patents cannot distinguish more impactful innovations from incremental technological discoveries, in order to assess a patent’s significance and impact, we construct a second measure of innovation productivity by counting the number of non-self citations received by each patent that the firm obtains subsequent to its M&A activity. Controlling for firm size, the number of patents captures innovation productivity while citations-per-patent captures the importance and novelty of innovation output. Following the existing innovation literature, we adjust the two measures of innovation to address the truncation problems associated with the NBER Patent Citation database. The first truncation problem arises as the patents appear in the database only after they are granted. In fact, we observe a gradual decrease in the number of patent applications that are eventually granted as we approach the last few years in the sample period (e.g., 2005 and 2006). This is because the lag between patent’s application year and patent’s grant year is significant (about two years on average) and many patent applications filed during these years were still under review and had not been granted by 2006. Following Hall, Jaffe, and Trajtenberg (2001, 2005), we correct for this truncation bias in patent counts using the “weight factors” computed from the application-grant empirical distribution. The second type of truncation problem is due to the possibility that a patent can keep receiving citations over a long period of time, but we observe at best the citations received up to 2006. Following Hall, Jaffe, and Trajtenberg (2001, 2005), the truncation in citation counts is corrected by estimating the shape of the citation-lag distribution. The distribution of patent grants in the pooling sample is right skewed, with the 75th percentile of the distribution at zero. 3 Due to the right-skewed distributions of patent counts and citations per patent, we then use the natural logarithm of the weight-factor adjusted patent counts and the natural logarithm of the citation-lag adjusted citations per patent, Ln(Pat) and 3

Firm-year observations with zero patents represent roughly 83.7% of our sample. This percentage is comparable to 84% reported in Atanassov, Nanda, and Seru (2009), and 86% reported in Chemmanur and Tian (2010). The first study includes the universe of Compustat firms between 1974 and 2000, and the second study includes S&P 1500 firms between 1990 and 2006. 6

Ln(Cites/Pat), as the main innovation measures in our analysis. To avoid losing firm-year observations with zero patents or citations per patent, we add one to the actual values when calculating the natural logarithm.

2.2.2. Measuring M&A Volumes To construct our M&A volume variable, we include mergers and acquisitions of public and private companies as well as acquisitions of assets. We exclude minority acquisitions to avoid transactions related with the firm’s strategic alliances and corporate venture capital investment. We then construct the yearly M&A volume variable, MA/Assets, by aggregating the transaction values of M&As undertaken by the firm in a year and then normalize it by the firm’s total book assets as of the end of the year. It is important to note that SDC does not report transaction values for almost 40% of M&A deals in our sample, especially for those transactions in which the target firm is a private firm or a subsidiary of a public firm. We consider these transaction values to be zero. This leads us to underestimate the actual acquisition volumes for our sample firms.

2.2.3. Measuring Control Variables Following the innovation literature, we control for a vector of firm and industry characteristics that may affect a firm’s innovation productivity. In the baseline regressions, our control variables include R&D intensity (measured by R&D expenditure over total assets), capital expenditure intensity (measured by capital expenditure over total assets), firm size (measured by the natural logarithm of sales), profitability (measured by ROA), asset tangibility (measured by net PPE scaled by total assets), leverage, product market competition (measured by the Herfindahl index based on sales), and growth opportunities (measured by Tobin’s Q). To capture possible non-linear relationship between product market competition and innovation (Aghion et al. (2005)), we also include the squared Herfindahl index in the baseline regressions. Since Aghion, Van Reenen, and Zingales (2009) find that institutional ownership affects firm innovation productivity; we also include institutional ownership in the baseline regressions. For each firm i over its fiscal year t, four quarterly institutional holdings observations are then weighted equally to compile an annual measure of the institutional holdings, Inst. ownership. 7

In the cross-sectional analysis, we split our sample based on firm age and VC-backing status. We construct a firm age variable, Age, which equals the number of years since the firm’s IPO date. We obtain a firm’s VC financing history from the Thomson Venture Economics database and construct a dummy variable, VC backed, which equals one if the firm obtained VC financing before its IPO and zero otherwise. Detailed variable definitions are described in the Appendix.

2.3. Descriptive Statistics To minimize the effect of outliers, we winsorize all variables at the top and bottom 1% of each variable’s fundamental distribution. Panel A of Table 1 provides summary statistics of the variables used in this study. On average, a firm in our final sample has 2.1 granted patents per year and each patent receives 1.4 non-self citations. An average firm spends 2.9% of its book assets value on M&A transactions and makes 0.5 bids each year. Panel A also reports the descriptive statistics of the control variables. In our sample, an average firm spends 5.6% and 6.1% of its book value of assets on R&D and on CAPEX, respectively, has sales of $1.58 billion, ROA of -3.6%, PPE ratio of 52.6%, leverage of 27.5%, Tobin’s Q of 2.6, and institutional ownership of 23.5%. A valid concern is that patents exist in only a couple of industries which would make our findings relevant for only a few industries. However, as reported in Panel B of Table 1 that shows the number of patents and citations per patent in each industry of Fama-French (1997) 48industry groups, firms with patents are spread broadly across industries in our sample. Industries in Panel B are sorted based on the average number of patents generated in a year. Firms in the Aircraft, Defense, and Chemicals industries generate the largest number of patents, while firms in Candy & Soda, Insurance and Real estate industries generate the lowest number of patents. Panel B also reports the industry distribution of other key variables, i.e., MA/Assets, RD/Assets, and CapEx/Assets. We, however, do not observe an obvious monotonic relation between these variables and the number of patents an industry generates. Panel C of Table 1 reports univariate comparisons. We first split the sample based on M&A volumes. We classify firm-year observations with positive M&A expenditures as a sample with high M&A volumes and firm-year observations with zero M&A expenditures as a sample with low M&A volumes. We then report the number of patents generated in the M&A year (year 8

0) and three years subsequent to the M&A year (year 1 to year 3). Our univariate comparisons show that firms with high M&A volumes generate 5 patents in the years following the M&A activity while firms with low M&A volumes generate fewer than 2 patents, and the differences between the two subsamples are statistically significant at the 1% level. We conduct the same comparisons based on R&D/Assets and CapEx/Assets. We find a similar positive association between R&D and capital expenditures, and innovation.

3. M&A Activity and Innovation In this section, we first discuss our empirical design and report the baseline regression results in Section 3.1. We then conduct a set of robustness tests and report the results in Section 3.2. We explore cross-sectional variation on the effect of M&A on innovation in Section 3.3. In Section 3.4, we provide evidence suggesting a positive relation between the R&D and patent activity of the target firm and the post-acquisition innovation output of the acquirer.

3.1. Baseline Specification To examine the effect of M&A on a firm’s innovation output, we estimate the following empirical model in our baseline OLS regressions: Ln( Innovationi ,t + n ) = α + β × Ln( MA / Assets) i ,t + δZ i ,t + Yeart + Firmi + u i ,t

(1)

where i indexes firms, t indexes time, and n is equal to zero, one, and two. Ln(Innovation) is the dependent variable and can be one of the following two measures: the natural logarithm of the number of patents filed by the firm, Ln(Pat), and the natural logarithm of the number of non-self citations per patent, Ln(Cites/Pat). Z is a vector of firm and industry characteristics that may affect a firm’s innovation productivity discussed in Section 2.3. Year captures fiscal year fixed effects. We control for time invariant unobservable firm characteristics by including firm fixed effects, Firm. Heteroskedasticity-robust standard errors are clustered at the firm level as suggested by Petersen (2009). 4

4

Besides the pooled OLS regressions reported throughout the paper, we use a Tobit model that takes into consideration the non-negative nature of patent and citation data. We also run a Poisson regression when the dependent variable is the number of patents to take care of the discrete nature of patent counts. The baseline results are robust in the above alternative models. 9

The reasons we include firm fixed effects in the baseline regressions are twofold. First, our paper’s objective is to see whether a firm’s M&A activity is related to its innovation outcome. Including firm fixed effects allows us to directly examine if and how the variation of M&A volumes within a firm explains contemporaneous as well as subsequent variation in its innovation output. In other words, we can interpret β, the coefficient estimate of Ln(MA/Assets), as the impact of a firm’s change in M&A volumes on contemporaneous or subsequent change in the firm’s innovation output. Second, as most empirical studies involve an endogeneity concern, it is possible that an unobservable variable omitted from our empirical model affects both M&A activity and innovation outcomes, rendering our findings spurious. For example, better firms may choose to grow through external acquisitions and at the same time, they may also actively engage in long-term innovative projects which result in greater innovation output. In this case, firm quality might be unobservable and correlated with both M&A volumes and innovation, which could bias our coefficient estimate of Ln(MA/Assets). Including firm fixed effects helps alleviate the endogeneity concern to the extent that the omitted firm characteristics that are correlated with M&A volumes and innovation are constant over time. Table 2 reports the OLS regression results estimating equation (1). In columns (1) – (3), the dependent variable is Ln(Pat). In column (1), we estimate the contemporaneous effect of a firm’s M&A volume on its number of filed patents (and eventually granted). The coefficient estimate of Ln(MA/Assets) is positive and significant at the 1% level, suggesting that firms with higher M&A volumes obtain a larger number of patents. In columns (2) and (3), we replace the dependent variable with the natural logarithm of the number of patents filed in one and two years subsequent to the M&A activity, respectively. The coefficient estimates of Ln(MA/Assets) continue to be positive and significant at the 1% level. The economic relation between M&A volume and innovation outcomes is meaningful. For example, the magnitude of Ln(MA/Assets) in column (2) suggests that increasing M&A volumes by 10% increases the number of patents filed the year following the M&A activity by 1.01%. The coefficient estimates of Ln(RD/Assets), perhaps referred to as the most important innovation input, are positive and significant in columns (1) and (2), but become statistically insignificant in column (3). In terms of economic magnitude, a 10% increase in R&D expenditures translates into a 0.89% increase in the number of patents filed in the first year following the R&D expenditure, based on the estimation reported 10

in column (2). Overall, these results suggest that M&A activity appears to be at least as important as R&D in contributing to innovation. In columns (4) – (6), we estimate equation (1) with the dependent variables replaced with Ln(Cites/Pat). The coefficient estimates of Ln(MA/Assets) are positive and significant at the 10% and 1% levels in columns (4) and (5), respectively. Similarly, the coefficient estimates of Ln(RD/Assets) are positive and significant in column (4). In terms of economic magnitude, column (4) suggests that a 10% increase in a firm’s M&A volume increases the number of citations per patent in the M&A year by 0.39% while a 10% increase in the R&D intensity is associated with a 0.78% increase in the number of citations per patent. However, in column (5), while the coefficient estimate Ln(MA/Assets) continues to be significant at the 1% level, that of Ln(RD/Assets) is insignificant. Once again, these results suggest that the association between M&A activity and innovation is at least as significant as that between R&D and innovation. We control for a comprehensive set of firm characteristics that may affect firm innovation output. Our results show that larger firms (firms with larger sales), firms with larger capital expenditures and higher leverage are more innovative. We find some evidence consistent with Aghion, Van Reenen and Zingales (2009) who document a positive relation between institutional ownership and innovation. The coefficient estimate of Inst. Ownership is positive and significant in column (1), but it either loses its significance or becomes negative in the subsequent years. Firm profitability, product market competition measured by the Herfindahl index, and growth opportunities measured by Tobin’s Q do not appear to significantly affect innovation output.

3.2. Robustness We conduct a comprehensive set of robustness checks in this section. First, we reestimate equation (1) and replace the main variable of interest with Ln(Number of MAs), defined as the natural logarithm of the number of M&As a firm undertakes in a year. We report the regression results in Table 3. In columns (1) – (3) where the dependent variable is Ln(Pat), the coefficient estimates of Ln(Number of MAs) are positive and significant at the 1% and 5% levels, consistent with our baseline evidence. The magnitude of the coefficient estimate in column (2) suggests that increasing the number of M&A deals by 10% increases the number of patents filed in the first year following the M&A deals the firm undertakes by 0.33%. The coefficient 11

estimates of Ln(RD/Assets) are positive and significant at the 1% and 5% levels in columns (1) and (2), respectively, but become statistically insignificant in column (3). In columns (4) – (6), the dependent variable is replaced with Ln(Cites/Pat), a measure of the quality and the impact of the patents obtained by the firm. The coefficient estimate of Ln(Number of MAs) is positive in all three columns and significant at the 5% level in column (4). Overall, our results, once again, suggest that M&A activity is positively related to the firm’s contemporaneous as well as subsequent innovation productivity and is at least as important as R&D in explaining both the number and novelty of the patents the firm obtains. Second, to address the concern that our results may be driven by the large number of firm-year observations with zero patents and citations per patent, we focus on a subsample of firms that have at least one patent in the pooling sample. 5 We estimate equation (1) in this subsample and report regression results in Table 4. The coefficient estimates of Ln(MA/Assets) are positive and significant in columns (1) – (3) where the dependent variable is the number of patents. Not surprisingly, the magnitudes of the coefficient estimates of Ln(RD/Assets) are larger than those estimated from the full sample. This is because innovation is more relevant in this subsample. We observe a similar finding in columns (4) – (6) where the number of non-self citations per patent is the dependent variable. Third, another valid concern is that a firm’s innovation productivity exhibits persistence. To address this concern, we control a firm’s lag innovation productivity and include Ln(Innovationt-3,

t-1)

in the baseline regression, defined as the natural logarithm of a firm’s

average innovation variable from year t-3 to year t-1. In columns (1) – (3) where Ln(Pat) is the dependent variable, Ln(Patt-3, t-1) is included; while in columns (4) – (6) where Ln(Cites/Pat) is the dependent variable, Ln(Cites/Patt-3,

t-1)

is included in the regressions. We report the

regression results in Table 5. The coefficient estimates of Ln(MA/Assets) continue to be positive and significant in all columns except in column (4) after controlling for the firm’s lag patenting activity, suggesting that our baseline results are not driven by the persistence of a firm’s innovation productivity.

5

While firm-year observations with non-zero patent count for only 16.3% of the sample, the sample size drops to 37,665 (about 35.8% of the full sample) firm-year observations for firms that have at least one patent in the whole sample period. 12

In the baseline model, we include firm fixed effects in the regressions and mainly rely on the time-series variation within a firm to study the effect of M&A activity on a firm’s innovation productivity. As a robustness check, we explore the effect of M&A activity on innovation in the cross section in an untabulated analysis. We replace firm fixed effects with the 2-digit SIC industry fixed effects in equation (1). The coefficient estimates of Ln(MA/Assets) turn out to be positive and significant at the 1% level in all columns. For example, the coefficient estimate of Ln(MA/Assets) is 0.111 (p-value=0.002) in column (2) of Table 2 where one-year ahead Ln(Pat) is the dependent variable, and 0.116 (p-value