Top Management Human Capital, Inventor Mobility, and … Management Human Capital... · States...

JOURNAL OF FINANCIAL AND QUANTITATIVE ANALYSIS Vol. 00, No. 00, 2019, pp. 1–40COPYRIGHT 2018, MICHAEL G. FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON, SEATTLE, WA 98195doi:10.1017/S0022109018001497

Top Management Human Capital, InventorMobility, and Corporate Innovation

Thomas J. Chemmanur , Lei Kong, Karthik Krishnan,and Qianqian Yu*

AbstractUsing panel data on top management characteristics and a management quality factor con-structed using common factor analysis on individual management quality measures, weanalyze the relation between top firm management quality and corporate innovation inputand output. We show that top management quality is an important determinant of corporateinnovation, with individual aspects of management quality affecting innovation in youngerand older firms differently. Further, firms with higher top management quality engage inmore risky (“explorative”) innovation strategies. Finally, hiring more and higher-qualityinventors is an important channel through which firms with higher management qualityachieve greater innovation output.

I. IntroductionThe effectiveness of a firm’s top management team in investing in and man-

aging innovative projects may determine the long-term success of the firm. In-deed, prior literature suggests that firms’ investments in research and development(R&D) and their innovative output (measured by patents and citations) may havea positive impact on the long-term financial health of the firm. Given this, thereis surprisingly little analysis of how the human capital or “quality” of a firm’s top

*Chemmanur (corresponding author), [email protected], Boston College Carroll School ofManagement; Kong, [email protected], University of Alabama Culverhouse College of Business;Krishnan, [email protected], Northeastern University D’Amore–McKim School of Busi-ness; and Yu, [email protected], Lehigh University College of Business and Economics. We thankPaul Malatesta (the editor) and an anonymous referee for several valuable comments that helped togreatly improve this paper. We are grateful for comments and suggestions from Kee H. Chung, MarciaMillon Cornett, Michael Ewens, Chinmoy Ghosh, Joseph Golec, Shantaram Hegde, Sahn-Wook Huh,Brad Jordan, Andrew Karolyi, Mark Liu, Gustavo Manso, Alan Marcus, Kristina Minnick, KartikRaman, Bob Taggart, Hassan Tehranian, Brian Wolfe, and Zhaoxia Xu; seminar participants at BostonCollege, Bentley University, the University of Kentucky, SUNY Buffalo, the University of Calgary,the University of Connecticut, and Hong Kong Polytechnic University; and conference participants atthe 2015 PBC-RFS conference on “Entrepreneurial Finance and Innovation around the World,” the2015 Financial Management Association Annual Meeting, and the 2016 American Finance Associa-tion Meetings in San Francisco.

1

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam

bridge.org/core . Boston College , on 01 Jul 2019 at 00:49:45 , subject to the Cambridge Core term

s of use, available at https://ww

w.cam

bridge.org/core/terms .

https://orcid.org/0000-0003-3302-8977

mailto:[email protected]




https://doi.org/10.1017/S0022109018001497

https://www.cambridge.org/core

https://www.cambridge.org/core/terms

2 Journal of Financial and Quantitative Analysis

management team may affect its innovation input and output. We aim to fill thisgap in the literature.

One strand in the theoretical literature suggests that higher quality manage-ment teams may invest in a larger number of value-enhancing long-term projects(e.g., Chemmanur and Jiao (2012)). Given that innovative projects are among suchlong-run value-enhancing projects (e.g., Hirshleifer, Hsu, and Li (2013), Griliches(1990)), we expect that higher-quality top management teams will invest in moreinnovative projects and will have a greater extent of innovative output, on average.They may accomplish this by having better foresight into the potential value ofinnovative investment opportunities and by more effectively managing innovativeresources such as physical assets (e.g., research equipment) and human capital(e.g., scientists and inventors). Further, they may provide an environment thatfosters greater failure tolerance in the sense of Manso (2011). Given this, firmswith higher-quality management teams may attract inventors with greater skills towork for them.

These arguments lead to several testable predictions. First, firms with higher-quality top management teams may invest more in R&D. Second, firms withhigher quality top management teams may have a greater extent of innovationproductivity (measured by the number of patents) and higher-quality innovation(measured by total citations and citations per patent). Third, better managementof innovative assets by higher-quality management teams may be reflected in agreater extent of innovative efficiency (e.g., patents per R&D dollar) of the firmsunder their management. Fourth, the effect of management team quality on inno-vative output may be stronger for firms facing financial constraints and for firmsin competitive industries. Since such firms are at a disadvantage relative to otherfirms, the “leg-up” provided by higher-quality top management teams may en-hance their future prospects.

It is also likely that different aspects of top management team quality mayhave significantly different effects on the innovation output of firms at differentinnovation stages: For example, younger firms may have a larger fraction of theirinnovation projects at the early innovation stage, whereas older firms may havea larger fraction of their innovation projects at the commercialization stage, sothat the individual aspects of the top management team that are most important toefficiently manage innovation in these two types of firms may be different. Thetop management quality of a firm may also affect its innovation strategy. On theone hand, it is possible that, given their greater human capital, firms with highertop management quality will pursue more risky innovation strategies involvingnew technologies and are more likely to push their knowledge boundaries out-ward. However, it is also equally possible that such firms may engage in less risky(more conservative) innovation strategies. Finally, an important channel throughwhich firms with higher-quality top management teams achieve greater innova-tion output may be by hiring a greater number of inventors (for a given level ofR&D expenditures) and by hiring higher-quality inventors (as measured by theirprior track record of citations per patent).

An empirical analysis of the relationship between top management qual-ity and corporate innovation faces two challenges. First, measuring the humancapital of a firm’s top management team (which we refer to as top management

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497



Chemmanur, Kong, Krishnan, and Yu 3

quality) involves subjective notions of what constitutes a higher-quality manage-ment team. Second, potential endogeneity may confound empirical findings onthe relation between management quality and innovation. In particular, there maybe an endogenous matching between higher-quality top management teams andhigher-quality firms.

We overcome the first challenge above by constructing a management teamquality factor from seven measures capturing various individual aspects of topmanagement quality following the existing literature (see, e.g., Chemmanur andPaeglis (2005), Chemmanur, Paeglis, and Simonyan (2011)): management teamsize, the fraction of managers with MBAs, the fraction of managers with Ph.D.s,the fraction of members with prior work experience in the top management team,the average number of prior board positions that each manager has served on, andthe average employment- and education-based connections of each manager onthe management team.1 These measures are adjusted for firm size. We create ourindex of top management quality using common factor analysis to capture the co-movement of the aforementioned seven individual measures of top managementquality, thereby extracting a single “management quality factor.”2 To test the va-lidity of this “management quality factor” as a measure of the quality of a firm’smanagement team, we analyze the relation between our management quality fac-tor and the average compensation of senior managers included in our managementquality measure, and we observe a strong and significantly positive correlationbetween these two variables, thus establishing the validity of our managementquality factor.

We overcome the second challenge related to endogeneity by using an in-strumental variable (IV) analysis. We instrument for top management quality (asmeasured by our management quality factor) using a plausibly exogenous shockto the supply of top executives available for hire by a firm, namely, the number ofacquisitions in the industry and state of the sample firm 5 years prior, weightedby an index measuring the enforceability of noncompete clauses in that state

1Starting with the pioneering work of Becker (1964) and Ben-Porath (1967), the labor economicsliterature has focused on the human capital of workers. The Becker view is that human capital increasesa worker’s productivity in all tasks, although possibly differentially in specific tasks, organizations, andsituations. In the Becker view, although the role of human capital in the production process may bequite complex, it is representable by a unidimensional measure, such as a worker’s stock of knowledgeor skills, which becomes directly part of the production function. When analyzing the human capitalof the members of a firm’s top management team, our view is that managerial human capital is mul-tidimensional, consisting of many different aspects, which we capture using the individual measureswe mention here and collapse into one factor using common factor analysis. Thus, our view of humancapital is closer to the view of the social psychologist Howard Gardner (see, e.g., Gardner (1983) andAcemoglu and Autor (2011) for a review). An advantage of such a multidimensional approach is thatit allows us to capture differences in not only the quantity but also the quality of the human capital oftop management teams across firms.

2Chemmanur and Paeglis (2005) and Chemmanur et al. (2011) use a broadly similar approach tostudy the effects of the management quality of private firms on the initial public offering (IPO) char-acteristics of these firms and on the anti-takeover provisions in their corporate charters, respectively.Chemmanur, Paeglis, and Simonyan (2009) also use a similar approach to analyze the relationshipbetween the management quality and financial policies of firms making seasoned equity offerings(SEOs). Unlike the current paper, the previously mentioned papers make use of cross-sectional datahand-collected from IPO and SEO prospectuses.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




(aggregated to the national level). We discuss the rationale for this instrumentin more detail in Section VI.

We analyze the relationship between top management quality and corporateinnovation using a panel data set of 4,389 firms covering the period 1999–2009.We obtain biographical data on the top managers of firms from the BoardExdatabase, patent and citation information from the patent data set created byKogan, Papanikolaou, Seru, and Stoffman (2017) based on data from the UnitedStates Patent and Trademark Office (USPTO), and inventor information associ-ated with each patent from the U.S. Patent Inventor Database (1975–2010); seeLi, Lai, D’Amour, Doolin, Sun, Torvik, Yu, and Fleming (2014) for a detaileddescription of the latter database.

Our empirical results may be summarized as follows: First, we find thathigher-quality management teams invest more in R&D, showing that they de-vote a larger amount of resources to innovative activities (i.e., have larger inputinto corporate innovation). Second, firms with higher-quality management teamshave a greater extent of innovation output (measured by the number of patents)and higher-quality innovation output (measured by total citations and citationsper patent). These effects are also economically significant. For instance, a 1 in-terquartile range increase in top management quality increases firm patents by12.8%. Third, we find that firms with higher-quality management teams producemore patents and citations per R&D dollar, that is, have greater innovative ef-ficiency (see, e.g., Hirshleifer et al. (2013)). The aforementioned results on therelation between management quality and corporate innovation are confirmed byour IV analysis making use of the instrument mentioned previously. Finally, therelation between top management team quality and innovation is stronger for firmsin financially constrained industries and those in more competitive industries.

We also analyze how the seven individual measures discussed previouslythat capture various individual aspects of top management quality affect corpo-rate innovation. We find that most of these individual measures have a positiveand significant effect on the quantity and quality of corporate innovation. We thenanalyze how these individual measures for top management team quality affect theinnovation output in younger firms (likely to have a larger fraction of innovativeprojects at the “early innovation” stage) versus older firms (likely to have a largerfraction of innovative projects at the commercialization stage). We find that man-agement team size, the fraction of managers with MBA degrees, and education-based connections positively and significantly affect corporate innovation only forolder firms. Prior managerial work experience positively and significantly affectscorporate innovation only for younger firms. Two individual measures, namely,the fraction of top managers with Ph.D. degrees and employment-based connec-tions, positively and significantly affect corporate innovation in both younger andolder firms; however, the magnitude of the effect of the Ph.D. measure is greaterfor corporate innovation in younger firms.

We also examine the relation between the nature of the innovation strategiesundertaken by firms and their top management quality. In particular, we analyzewhether firms with higher-quality top management teams engage more in risky“explorative” innovation strategies (where they venture into the development ofnewer technologies or pursue innovations in areas that are less familiar to the firm)

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




or in more conservative “exploitative” innovative strategies (where they may pur-sue innovations using more conventional technologies or in areas that are morefamiliar to the firm). We find support for the notion that firms with higher topmanagement quality engage in riskier innovation strategies, innovations involv-ing new technologies, and those pushing forward the knowledge boundaries ofthe firm. Consistent with this, we find that the relation between top managementquality and the fraction of firm patents using new knowledge is positive and sig-nificant (and is stronger than the relation between top management quality and thefraction of firm patents using existing knowledge). Further, top management qual-ity is positively and significantly related to innovation diversity (see, e.g., Brav,Jiang, Ma, and Tian (2018)) and to the number of firm patents that belong to themost highly cited (top 10%) group of patents (see, e.g., Balsmeier, Fleming, andManso (2017)).

Finally, we investigate an important channel through which higher-qualitymanagement teams may foster greater innovation in their firms. We argue thathigher-quality management teams may provide more resources to R&D, manageR&D resources better, and provide a more failure-tolerant climate for inventorsto succeed in. This, in turn, may make firms with higher top management qualitymore attractive to higher-quality inventors. Thus, one way that higher-quality topmanagement teams may enhance innovation is by hiring more and higher-qualityinventors to work for the firm. We find evidence consistent with this hypothesis:Firms with higher-quality management teams experience greater net inflows ofinventors (controlling for R&D expenditures), particularly of higher-quality in-ventors (defined as those inventors with a record of past citations per patent abovethat of the median inventor in our sample). We also find that the average citationsper patent of incoming inventors entering firms with higher-quality managementteams are higher than the average citations per patent of outgoing inventors fromsuch firms.

Our paper contributes to several strands in the literature. First, wecontribute to the recent literature that has analyzed the determinants ofinnovation in firms theoretically as well as empirically (e.g., Manso (2011),Marx, Strumsky, and Fleming (2009)) and their impact on firm performance (e.g.,Hirshleifer et al. (2013), Gu (2005), Eberhart, Maxwell, and Siddique (2004),Lanjouw and Schankerman (2004), Lerner (1994), and Griliches (1990)). The ex-isting literature has focused, however, on firm characteristics other than top man-agement team quality and its effects on innovation. Some of these characteristicsare managerial compensation (e.g., Lerner and Wulf (2007), Ederer and Manso(2013)), public versus private status (see e.g., Bernstein (2015) for an empiricalanalysis and Ferreira, Manso, and Silva (2014) and Spiegel and Tookes (2016)for theoretical analyses), private equity or venture capital involvement (e.g.,Lerner, Sorensen, and Stromberg (2011), Tian and Wang (2014), and Chemmanur,Loutskina, and Tian (2014)), anti-takeover provisions (e.g., Atanassov (2013),Chemmanur and Tian (2018), and Sapra, Subramanian, and Subramanian (2014)),institutional ownership (e.g., Aghion, Van Reenen, and Zingales (2013)), CEOoverconfidence (Hirshleifer, Low, and Teoh (2012)), and conglomerate structure(e.g., Seru (2014)). In a contemporaneous paper, Custodio, Ferreira, and Matos(2018) analyze how the general versus firm-specific human capital of CEOs

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




affects corporate innovation. In contrast to the last paper just mentioned, our paperdoes not focus on CEO characteristics: rather, we analyze the relation between topmanagement team quality and corporate innovation, thus moving the literature ina new direction.

Second, we contribute to the literature linking top management quality tofirm performance, investments, and financing. For instance, Bertrand and Schoar(2003) study the effect of top managers on a firm’s financial and investmentpolicies.3 They find that manager fixed effects explain some of the heterogene-ity in the investment, financial, and organizational practices of firms. Ewensand Rhodes-Kropf (2015) use a similar fixed-effects methodology to investigatewhether individual venture capitalists have repeatable investment skill and theextent to which their skill is impacted by the venture capital firms for whichthey work. They accomplish this by tracking the performance of individual ven-ture capitalists’ investments over time as they move between firms. Chemmanur,Kong, Krishnan, and Yu (2017) relate measures of top management quality similarto ours to stock return, operating performance, and firm valuation. Unlike the lastpaper just mentioned, we focus on the relation between top management qualityand corporate innovation output. Further, we add to the above literature by ana-lyzing the nature of the innovative strategies adopted by firms with higher- versuslower top management team quality.4 We also provide new evidence suggestingthat higher-quality managers enhance their firms’ innovation output by attractinghigher-quality inventors to join the firm.5

The rest of the paper is organized as follows: Section II discusses the un-derlying theory and develops testable hypotheses. Section III describes our dataand sample selection procedures, explains the construction of our measures oftop management quality and other measures, and validates our management qual-ity measure using managerial compensation data. Section IV presents our em-pirical tests and results. Section V presents our analysis of the inventor mobil-ity channel. Section VI presents our IV analysis. Section VII concludes. TheSupplementary Material, available at https://www2.bc.edu/thomas-chemmanur/,develops testable hypotheses regarding the effect of various individual aspects ofmanagement team quality on corporate innovation and presents additional robust-ness tests not included in the main text due to space limitations.

3Our paper is also indirectly related to the literature analyzing the determinants of CEO quality andhow it affects firm performance (see, e.g., Adams, Almeida, and Ferreira (2005), Malmendier and Tate(2005)). See also Kaplan, Klebanov, and Sorensen (2012), who study the individual characteristics ofCEO candidates for companies involved in buyout and venture capital transactions and relate them tothe subsequent performance of their companies.

4In more distantly related research, Bloom and Van Reenen (2007) use an innovative survey toolto collect data on management practices from various countries and show that measures of managerialpractice are strongly associated with firm-level productivity, profitability, Tobin’s Q, and survival rates.

5Two important papers in the economics literature that have implications for our paper are those bySah and Stiglitz (1986), (1991), which imply that firms with larger management teams are more likelyto reject bad projects. The organizational behavior literature on the effect of managerial discretion onfirm performance is also indirectly related to our paper: see Finkelstein and Hambrick (1996) for areview.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://www2.bc.edu/thomas-chemmanur/





































https://doi.org/10.1017/S0022109018001497




II. Theory and Hypothesis DevelopmentIn this section, we briefly discuss the underlying theory and develop hypothe-

ses for our empirical tests. Our theoretical motivation depends on arguments madeby Chemmanur and Jiao (2012) and Stein (1988). Chemmanur and Jiao (2012)study a setting in which managers with greater talent or ability are able to gen-erate greater long-run cash flows by undertaking long-term projects. However,since their talent is private information, and since short-term projects come tofruition earlier, myopia or short-termism induced by the stock market (e.g., dueto the possibility of rivals appearing and successfully taking over the firm in theabsence of favorable signals of project success in the short run) impose pres-sures on them to undertake short-term rather than long-term projects (see alsoStein (1989)). However, Chemmanur and Jiao (2012) show that more capablemanagers also have an incentive to undertake long-term (innovative) rather thanshort-term projects, since they are able to create greater long-term value by do-ing so. In such a setting, the equity market values firms undertaking long-termprojects at a discount, since they are not able to observe true managerial abil-ity; however, firms with higher perceived management team quality suffer only asmaller valuation discount if they undertake long-term projects. In summary, man-agers’ choice between long-term and short-term projects is driven by the trade-offbetween the pressure induced by a myopic stock market versus the ability (and de-sire) of more able managers to create greater value in the long run by undertakinglong-term projects.6 Given that innovative projects are long-term projects charac-terized by short-term failures and experimentation (which increases the gestationtime of these projects), managers with greater perceived ability will undertakea greater proportion of long-term (innovative) rather than short-term projects inequilibrium.7

The aforementioned theoretical framework provides us with our first set oftestable implications. First, top management teams with higher (perceived) qual-ity are likely to devote a greater amount of resources to innovation activities.Thus, firms with higher quality management teams will be characterized by alarger input of resources into innovation activities (i.e., R&D expenditures) (Hy-pothesis 1). Further, we expect such firms to be characterized by greater inno-vation output and higher quality of innovation output, after controlling for R&D

6Formally, in Chemmanur and Jiao (2012), the objective function of the manager is a weightedaverage of the short-run and long-run stock price. Thus, while talented (higher-ability) managers willsuffer a discount in the firm’s short-run stock valuation if they take a greater proportion of long-term projects (since their equity will be priced in a pooling equilibrium with firms with less talentedmanagers), more talented managers have an incentive to undertake a greater proportion of long-termprojects since these projects will allow them to create greater long-run value, yielding a higher long-run stock price.

7Note that this theoretical framework does not require the assumption that long-term projects arealways better than short-term projects. In industries where short-term projects are as valuable as long-term projects (e.g., in less innovative industries), there is no problem of equity undervaluation (sinceall managers will prefer to invest in short-term projects in such industries), so that the relation betweenmanagement quality and innovation will be weaker in such industries.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




expenditures (Hypothesis 2).8 Such firms will also be characterized by greater in-novative efficiency (i.e., greater innovation output and higher-quality innovationoutput per dollar of R&D capital investment) (Hypothesis 3).9

So far, we have discussed top management team quality without providingany detail about how we measure top management quality. In order to developtestable hypotheses about how various individual aspects of top management qual-ity are related to corporate innovation, we now briefly discuss the approach thatwe take to measure top management quality. We first measure various individualaspects of the top management team quality of a firm using seven measures pre-viously used in the literature. These seven individual measures of managementteam quality are management team size (TEAM SIZE), the fraction of managerswith MBA degrees (MBA), the fraction of managers with Ph.D. degrees (PHD),the fraction of managers with prior work experience (WORK EXP), the averagenumber of prior board positions that each manager has served on (BOARD EXP),and the average employment- and education-based connections of each managerin the management team (EMP CONN and EDU CONN). Then, given that truetop management quality itself is fundamentally unobservable, we make use of thecomovement between these seven individual measures (correlated with the un-derlying unobservable true management quality) to extract a management qualityfactor (MQF) using common factor analysis. The details of how these seven in-dividual aspects of top management team quality are measured are discussed inSection III.B.1. Due to space limitations, we confine our discussion of how each ofthese aspects (measures) of top management quality may individually be relatedto corporate innovation and the differential impact of these measures on innova-tion in younger versus older firms to Section 1 of our Supplementary Material.As is discussed in detail in the Supplementary Material, we expect each of theseven individual measures of top management team quality discussed above to bepositively related to the quantity and quality of the innovation output of a firm.This is the next hypothesis that we test here (Hypothesis 4).

We now delve deeper into the possible differences in the innovation strate-gies adopted by firms with higher- versus lower top management quality. Onepossibility is that firms with higher top management quality pursue more riskyinnovation strategies, those involving new technologies, and those that are likelyto push the knowledge boundaries of the firm outwards (Hypothesis 5A). In thiscase, we would expect firms with higher top management quality to engage inmore explorative innovation strategies (in the sense of Balsmeier et al. (2017),

8Since innovation is only one of the many activities undertaken by firms, it is not a priori obviousthat higher top management quality (as captured by our measures) will be associated with greaterinnovation output, higher-quality innovation output, or greater innovation efficiency. For example, wewill not find such associations empirically if higher-quality top management teams invest more inshort-term projects (see, e.g., Hirshleifer and Thakor (1992), who argue that managerial concern forreputation results in conservatism in project choice) or are better at managing the commercializationof innovations but not at managing the development of these innovations.

9As in practice, here we are assuming that higher-quality managers are in short supply, so that thereis considerable variation in management quality across firms. Clearly, even when higher-quality man-agement teams command a larger amount in aggregate compensation, hiring a higher-quality manage-ment team will be optimal for firms as long as the incremental surplus (shareholder value) generatedby hiring such managers after accounting for their higher compensation is positive.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




Brav et al. (2018)), so that they venture into the development of newer technolo-gies or pursue innovation in areas less familiar to the firm. Given that such anexplorative strategy is more risky, under this scenario, we would expect firmswith higher management quality to be associated with a larger number of highlysuccessful and a larger number of quite unsuccessful innovations (as measured bycitations per patent) compared to firms with lower management quality. In otherwords, in this case, firms with higher management quality will have a larger num-ber of patents in the two tails of the patent quality distribution. Further, in this case,we would also expect top management quality to have a greater effect on non-self-citations than on self-citations. Finally, we would expect top management qualityto be positively associated with greater innovation diversity. Alternatively, firmswith higher management quality may engage more in less risky (conservative)innovation strategies (Hypothesis 5B). In this case, we would expect firms withhigher top management quality to pursue more exploitative innovation strategies(in the sense of Balsmeier et al. (2017), Brav et al. (2018)), those using moreconventional technologies, and those in areas that are more familiar to the firm.Further, in this case, we would expect firms with higher top management qualityto be associated with more moderately successful innovations (again measured bycitations per patent) compared to those achieved by firms with lower managementquality.10 Finally, in the latter case, we would also expect top management qual-ity to have a greater effect on self-citations than on non-self-citations and to benegatively related to innovation diversity.

We now analyze whether the relationship between top management qualityand corporate innovation productivity is stronger for firms in some industries thanthose in others. First, consider firms in financially constrained industries. Giventheir financial constraints, such firms will have only a limited amount of resourcesto devote to innovation. If the relation between top management quality and inno-vation is partly driven by more effective resource management on the part of firmswith higher management quality, we would expect the relationship between man-agement quality and innovation to be stronger for firms in financially constrainedindustries (Hypothesis 6).11 Next, consider firms in more competitive versus thosein less competitive industries. Scientists and engineers (inventors) in more com-petitive industries are likely to have greater outside employment opportunities,so that talented inventors are likely to be in limited supply in these industries.Therefore, if firms with higher-quality top management teams are able to attract agreater proportion of these talented inventors in limited supply, we would expectthe relationship between top management quality and innovation productivity tobe stronger in more competitive industries (Hypothesis 7).

Finally, we analyze an important channel through which firms withhigher top management quality may be able to generate greater innovation

10It is difficult to predict from a priori theoretical considerations which of the above two scenarioswill be realized in practice. We will therefore leave this question to be resolved empirically.

11While firms in financially unconstrained industries may be able to partially compensate for nothaving higher-quality management teams by devoting more resources to innovative activities (e.g., bybuying higher-quality innovation equipment), firms in financially constrained industries will be lessable to do so, so that the relationship between management quality and innovation will be stronger forthe latter category of firms.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




productivity (i.e., greater innovation output for a given amount of resources de-voted to R&D expenditures). Consistent with our conjecture that higher-qualitytop management teams may be able to manage their innovative activities moreefficiently, we hypothesize that such firms are able to hire more inventors for agiven amount of R&D expenditures (Hypothesis 8). We further conjecture thatsuch firms are likely to hire higher-quality inventors, who are more innovative (asmeasured by these inventors’ prior track record of citations per patent).12 This isthe final hypothesis that we test here (Hypothesis 9).

III. Data and Sample Selection

A. Sample SelectionOur sample is derived from multiple data sources. Our primary data source

for the biographical information of senior managers is the BoardEx database. TheBoardEx database contains data on college education, graduate education, pastemployment history (including beginning and ending dates of various roles), cur-rent employment status (including primary employment and outside roles), andsocial activities (including memberships, positions held in various foundationsand charitable groups, etc.). The main information we are making use of in thispaper is education, employment history, and demographic information. We col-lect firm-year patent and citation information from the patent data set created byKogan et al. (KPSS) (2017). We collect the inventor information associated witheach patent from the U.S. Patent Inventor Database (1975–2010); see Li et al.(2014). To calculate control variables, we collect financial statement items fromCompustat and stock price information from the Center for Research in SecurityPrices (CRSP). To construct the instrumental variables as we described earlier, wecollect information on mergers and acquisitions from the Securities Data Corpo-ration (SDC) Mergers & Acquisitions database.

The unique company-level identification code in BoardEx is “Company ID,”which is unique to BoardEx and cannot be used to merge with other databasessuch as Compustat and CRSP. We link the BoardEx database to Compustat andCRSP in the following way. BoardEx provides the Central Index Key (CIK), theInternational Securities Identification Number (ISIN), and the company name.The “Company ID” in BoardEx is matched with the PERMNO in CRSP by ei-ther CIK or Committee on Uniform Securities Identification Procedures (CUSIP)(which is derived from the ISIN). After matching by CIK or CUSIP, we check theaccuracy of the matches by comparing the company names from BoardEx withthe company names from CRSP and Compustat.

The KPSS patent data set provides detailed data for all patents that aregranted by the USPTO over 1926–2011. We use the KPSS patent data rather than

12One way in which firms with higher-quality management teams may be able to attract higher-quality inventors is by promoting a more failure-tolerant work environment (in the sense of Manso(2011)). Manso (2011) has argued that an important variable in encouraging innovation is failuretolerance. While Manso (2011) does not distinguish between firms with higher- and lower-qualitymanagement teams, if we add the additional assumption that higher-quality managers are also morefailure tolerant, then it will be the case that firms that have higher-quality top management teams willalso have a more failure-tolerant work environment (more conducive to innovative activities).

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




National Bureau of Economic Research (NBER) patent data because the KPSSpatent data enable us to identify comprehensive patent portfolios of firms thatfiled patent applications up to 2009, which are granted up to 2011. The NBERpatent data contain patents that have been granted up to 2006, and most of themhave application dates up to 2004. Because our BoardEx sample starts from 1999,using the KPSS patent data increases our sample size significantly.13 The KPSSpatent data provide the PERMNO for the assignees of each patent. We use thisto merge the patent data with BoardEx as well as Compustat and CRSP. In thebase case analysis, we assign 0 patents to firm-years in the BoardEx sample with-out any patenting activity. The final BoardEx–KPSS Patent–Compustat–CRSPmerged file leaves us with 6,504 unique firms.

Using the BoardEx employment history file, we identify all the managersin each matched company for each year from 1999 to 2009. We obtain the sam-ple of senior managers from BoardEx, whom we define as managers with a ti-tle of vice president (VP) or higher. The senior managers in our sample can bebroadly categorized into seven groups: CEOs, presidents, chairmen, other chiefofficers (chief financial officer (CFO), chief information officer (CIO), etc.), di-vision heads, VPs, and others. We exclude all firm-years that have at least oneof the following characteristics: i) there is only one manager in the managementteam (because it is unlikely that large firms covered by BoardEx have only onesenior manager); ii) there is no CEO for a firm in a given year; iii) there are morethan 30 senior managers in the management team (suggesting that perhaps certaintitles are misleading and we are overclassifying senior managers); iv) financialand utility firms, defined by Standard Industrial Classification (SIC) codes from6000 to 6999 and from 4901 to 4999, respectively; and v) firm-years with missingvalues for the relevant variables that we need to use. After these exclusions, weare left with 30,432 firm-year observations for 4,389 firms.

We then obtain the demographic and education information for each seniormanager from the BoardEx database. To obtain education-based connections,we classify all graduate degrees into four different categories: business school(MBAs included), medical school, law school, and other graduate (see, e.g.,Cohen, Frazzini, and Malloy (2008)).

B. Measuring Management Quality

1. Measures of Seven Individual Aspects of Top Management Quality

We measure the quality of a firm’s top management team along two impor-tant dimensions. The first is based on management team resources, which refersto the human and knowledge resources (including both education and relevantwork experience) available to top firm management. The second is based on con-nections available to firm management, which capture their ability to reach out tomanagers in other firms, thus enabling them to obtain not only valuable informa-tion from other firm managers about potential innovation opportunities but alsobetter terms when dealing with these firms as customers or suppliers, as well as tohire more and higher-quality inventors.

13Although BoardEx data start from 1997, data prior to 1999 are sparse (e.g., see Engelberg, Gao,and Parsons (2013)).

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




The management team resources available to the firm depend in part upon thenumber of people in its management team. Therefore, our first measure of teamresources is the size of the firm’s top management team (TEAM SIZE), measuredby the number of officers with the rank of VP or higher. Management team re-sources also depend upon the knowledge and education of its members. Thus, oursecond measure of management team resources is the percentage of the manage-ment team with an MBA degree (MBA). A larger top management team and ahigher percentage of the team with MBA degrees imply better top managementquality, which will allow the management team to accomplish the tasks discussedpreviously. Our third measure of management team resources is the fraction of thetop management team with a Ph.D. degree (PHD). In some innovative firms (e.g.,technology or biotech firms), some of the top management team may have a Ph.D.degree, which may help them in choosing the appropriate innovation strategy fortheir inventors to work on (even if they themselves are not personally involvedin developing the innovation) as well as in hiring the “right” inventors (scientistsand engineers) who may develop innovations for their firms. Our fourth measureof management team resources is relevant work experience, which we measure intwo ways. First, we look at the percentage of the management team that has servedas VP or higher in other firms prior to joining the current firm (WORK EXP).Second, we look at the number of outside board positions that each manager haspreviously served on, averaged across all members of the top management team(BOARD EXP). Clearly, prior board experience may also be a useful asset whenmanaging a firm because managers may have acquired experience in solving im-portant problems when serving as board members in the firms for which they havepreviously worked.14

We measure top management connections in two ways. First, we look at con-nections built up by the members of a firm’s top management team based on theirwork experience so far (EMP CONN). For each manager, the measure of totalemployment-based connections is calculated as the number of senior managers ordirectors that each senior manager in the management team has worked with priorto joining the current firm. If individuals have worked together in the same com-pany previously during an overlapping time period, they are defined as connected.In summary, the variable EMP CONN is defined as the number of employment-based connections of the top management team divided by TEAM SIZE. Second,we look at the connections built up by members of the firm’s top managementteam during their graduate education, which may often last throughout their en-tire career (EDU CONN). For each manager, the total education-based connec-tion is calculated as the number of senior managers or directors that each seniormanager in the top management team has been in graduate school with. If indi-viduals study in the same educational institution, have degrees in the same edu-cation category (graduate school), and graduate within 1 year of each other, theyare defined as connected. In summary, the variable EDU CONN is defined as thenumber of education-based connections of the top management team divided byTEAM SIZE. In addition, we create the variable AVG TENURE as the average

14Three of the measures discussed above (TEAM SIZE, MBA, and WORK EXP) are similar tothose used by Chemmanur and Paeglis (2005) and Chemmanur et al. (2011).

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




number of years that each senior manager has worked in a firm, and we use it asa control variable.

Table 1 provides summary statistics on our seven individual measures ofmanagement team quality. For the median firm in our sample, there are seven se-nior managers in the management team; 20% of the senior management team hasan MBA degree; 10% of the senior management team has prior work experienceas a senior manager at another firm; 0% of the senior management team of themedian firm has sat on the boards of other firms; and 0% of the senior manage-ment team has a Ph.D. degree. The median level of EMP CONN is 15.4, and thatof EDU CONN is 0. The median number of years that each manager has workedin a firm is 5.2 years.

TABLE 1Summary Statistics

Table 1 reports summary statistics for our sample of public firms between 1999 and 2009. ADJ_PATENT is the truncation-adjusted number of patents filed by a firm in a given year; ADJ_CITE is the total adjusted number of citations receivedby the patents filed by a firm in a given year; ADJ_CPP is the adjusted number of citations per patent filed by a firmin a given year. TEAM_SIZE is the number of managers (vice president (VP) or higher) in a firm’s management team;MBA is the fraction of the managers who have MBA degrees; PHD is the fraction of the managers who have Ph.D. de-grees; WORK_EXP is the fraction of top managers who have experience working as a VP or higher in other companies;BOARD_EXP is the average number of board positions that each manager has held on; EMP_CONN is the averagenumber of connections that each manager has through prior employment; EDU_CONN is the average number of grad-uate connections that each manager has through education (if two managers graduated from the same university withthe same degree within 1 year of each other, those two are defined as connected); TOT_ASSETS is a firm’s total as-sets; MB is Tobin’s Q , defined as the market value of assets divided by the book value of assets, where the marketvalue of assets is computed as the book value of assets plus the market value of common stock less the book valueof common stock; ROA is defined as operating income before depreciation divided by total assets; CAPEX/ASSETS isdefined as capital expenditures divided by total assets; R&D/ASSETS is defined as research and development expensesdivided by total assets; RET is a firm’s annual stock return; the Herfindahl–Hirschman index (HHI) for an industry (de-fined at the 2-digit Standard Industrial Classification (SIC) code level) in a given year is defined by the following formula:∑No. of firms in the same 2-digit industry

i=1 ((firm salesi )2/(industry sales)2); AVG_TENURE is the average number of years that eachmanager has worked as a VP or higher in this firm.

1st 3rdVariable No. of Obs. Mean Std. Dev. Min Quartile Median Quartile Max

ADJ_PATENT 30,432 0.862 2.308 0.000 0.000 0.000 0.309 10.439ADJ_CITE 30,432 0.034 0.108 0.000 0.000 0.000 0.000 0.495ADJ_CPP 30,432 0.002 0.006 0.000 0.000 0.000 0.000 0.025TEAM_SIZE 30,432 7.751 4.469 2.000 5.000 7.000 10.000 30.000MBA 30,432 0.230 0.195 0.000 0.000 0.200 0.333 1.000PHD 30,432 0.074 0.146 0.000 0.000 0.000 0.100 1.000WORK_EXP 30,432 0.141 0.172 0.000 0.000 0.100 0.235 1.000BOARD_EXP 30,432 0.071 0.149 0.000 0.000 0.000 0.100 2.667EMP_CONN 30,432 17.949 11.108 1.000 10.000 15.400 23.000 98.667EDU_CONN 30,432 1.417 3.296 0.000 0.000 0.000 1.200 55.500TOT_ASSET ($millions) 30,432 2,887.979 12,991.060 0.306 83.810 322.609 1,335.178 304,594.000MB 30,432 2.307 2.904 0.166 1.169 1.606 2.499 137.183ROA 30,432 0.035 0.279 −2.041 0.014 0.102 0.163 0.424CAPEX/ASSETS 30,432 0.065 0.088 0.000 0.017 0.035 0.074 0.542R&D/ASSETS 30,432 0.077 0.153 0.000 0.000 0.010 0.092 0.995RET 30,432 0.173 0.767 −0.884 −0.289 0.033 0.404 3.458HHI 30,432 0.063 0.060 0.020 0.032 0.041 0.072 1.000AVG_TENURE 30,432 5.833 3.319 1.000 3.571 5.200 7.333 36.500

2. Common Factor Analysis on Individual Measures of Management Quality

Each of the seven measures described above is likely to have its unique lim-itations as a measure of the underlying management quality, and is therefore un-likely to be a comprehensive measure of management quality by itself. Therefore,we use common factor analysis to capture the variation common to our sevenobservable measures of management quality. More precisely, the aim of our fac-tor analysis is to account for, or explain, the matrix of covariances between our

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




individual measures of management quality using as few factors as possible. Next,we rotate the initial factors so that each individual measure of management qual-ity has substantial loadings on as few factors as possible. This methodology isconsistent with the implementation of common factor analysis in the literature.15

All the seven management quality measures above are aggregated to the levelof the management team and are likely to be correlated with firm size. Therefore,to make sure that these measures are independent of firm size, we use firm size-and industry-adjusted variables in our common factor analysis. Specifically, weconduct the following regression for each of the seven proxies of managementquality:

MEASUREi ,t = α[ln(FIRM SIZE)i ,t ] +β[ln(FIRM SIZE)i ,t ]2(1)

+ IND FE+YR FE+ εi ,t ,

where i indexes the firm, and t indexes the year of the observation. Industryfixed effects defined at the 2-digit SIC code level (IND FE) and year fixed ef-fects (YR FE) are included. We use the residuals from the above regression as thefirm size- and industry-adjusted measures of the management quality.

Table 2 presents the results of the common factor analysis. The common fac-tor analysis leads to seven factors. Panel A reports the eigenvalues of each factor.Factors with higher eigenvalues account for a greater proportion of the variance ofthe observed variables. Only the first factor has an eigenvalue that is larger than 1.Further, we find that this first factor explains 80% of the variation of our individ-ual management quality proxies (the eigenvalue of the first factor as a proportionof the sum of the eigenvalues of all seven factors). This suggests that the first fac-tor is the most important one, providing us with a distinct (unique) measure ofmanagement quality. We term this factor the management quality factor (MQF).16

The first column in Panel B of Table 2 reports the loadings on the first fac-tor for each of the individual management quality measures. The loadings indicatethat each of the seven individual management quality measures loads positively onthe first factor. Consistent with this, the second column of Panel B shows positivecorrelations between the first factor and each of our seven management qualitymeasures. The last column of Panel B reports the communality of each individualmanagement quality measure with the common factor, which measures the pro-portion of the variance of each proxy that is accounted for by the common factor.

15We adopt common factor analysis rather than principal component analysis as our method ofchoice for identifying a single management quality factor. The aim of common factor analysis is toaccount for or to “explain” the matrix of covariances between our seven individual management qualityproxies using the minimum number of factors. In contrast, the aim of principal component analysis isto break down the covariance matrix into a set of orthogonal components equal to the number of theindividual proxies. Given that our objective here is to identify a factor that embodies the underlyingunobservable construct, namely, “management quality,” we believe that the former method is moreappropriate here.

16In a robustness test that we describe in the Supplementary Material, we address the concern thatour results are driven by the team size measure (TEAM SIZE) alone, but not the other individualmanagement quality measures. We therefore recalculate the management quality factor by excludingTEAM SIZE from the common factor analysis. We show that our results are similar when we use themanagement quality factor derived from this alternative model.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 2Extraction of the Management Quality Factor using Common Factor Analysis

Table 2 reports statistics related to our common factor analysis. FACTOR1–FACTOR7 are the common factors obtained byusing common factor analysis on the firm size- and industry-adjusted TEAM_SIZE, MBA, PHD, WORK_EXP, BOARD_EXP,EMP_CONN, and EDU_CONN. TEAM_SIZE is the number of managers (VP or higher) in a firm’s management team;MBA is the fraction of the managers who have MBA degrees; PHD is the fraction of the managers have Ph.D. de-grees; WORK_EXP is the fraction of top managers who have experience working as a VP or higher in other companies;BOARD_EXP is the average number of board positions that each manager has held; EMP_CONN is the average numberof connections that each manager has through prior employment (if two managers worked in the same previous companyduring overlapping time periods, either as managers or directors, those two are defined as connected); EDU_CONN isthe average number of graduate connections that each manager has through education (if two managers graduate fromthe same university with the same degree within 1 year of each other, those two are defined as connected). Panel Areports the eigenvalues for the seven factors to mimic the correlation matrix of the original variables. Panel B reports theloadings on the first factor and the correlation with the first factor as well as the communality of the original variables.Panel C reports the descriptive statistics of the first factor.

Panel A. Eigenvalues

FACTOR 1 FACTOR 2 FACTOR 3 FACTOR 4 FACTOR 5 FACTOR 6 FACTOR 7

1.172 0.577 0.216 0.042 −0.022 −0.195 −0.301

Panel B. Summary for Factor Analysis

Variable Loadings on First Factor Correlation with First Factor Communality

TEAM_SIZE 0.513 0.841 0.264MBA 0.141 0.078 0.020PHD 0.053 0.038 0.003WORK_EXP 0.368 0.233 0.135BOARD_EXP 0.288 0.094 0.083EMP_CONN 0.803 0.906 0.645EDU_CONN 0.149 0.086 0.022

Panel C. Summary Statistics of the First Factor

No. of Obs. Mean Std. Dev. Minimum 1st Quartile Median 3rd Quartile Maximum

30,432 0.013 0.758 −1.992 −0.472 −0.040 0.447 2.315

Communality is bounded between 0 and 1, and higher values indicate that a largerproportion of the variation in the measure is captured by the common factor.

3. Validation of Our Management Quality Factor

This section attempts to validate our management quality factor (MQF) byusing top management team compensation. If MQF indeed captures the true qual-ity of a firm’s top management team, one immediate implication is that manage-ment teams with a higher MQF will be paid more than those with a lower MQF.To test this implication, we make use of the compensation data in the BoardExdatabase. BoardEx provides the annual compensation information for a firm’s se-nior managers, which includes salary, bonus, the value of shares awarded, thevalue of the long-term incentive plan (LTIP) awarded, and the value of optionsawarded at the manager-firm-year level. However, the coverage of the compen-sation data is smaller than the coverage of individual measures that are used tocalculate MQF because the compensation information of many managers is miss-ing in BoardEx. We therefore focus on only the managers with compensationinformation available in BoardEx.

For each firm, we construct three different measures of compensa-tion: AVG TOT COMP, AVG CASH COMP, and EQUITY/TOT COMP. AVGCASH COMP is defined as the average amount of cash compensation for amanagement team, where cash compensation includes base cash salary andbonus. AVG TOT COMP is defined as the average amount of total compensation

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




for a management team, where total compensation, in addition to cash com-pensation, also includes the equity compensation, consisting of the value ofshares awarded, the value of LTIP awarded, and the value of options awarded.EQUITY/TOT COMP is defined as the fraction of equity compensation out oftotal compensation. We test the following model:

ln(COMPENSATION)i ,t = α+βMQFi ,t + γ Z i ,t + IND FE(2)+YR FE+ εi ,t ,

where ln(COMPENSATION)i ,t is the natural logarithm of the previous three com-pensation measures for the management team in firm i in year t . We take logsdue to the right skewed distribution. Z is a set of control variables, which is de-scribed in detail in Section III.F. We report the results of the above regressions inTable 3. Columns 1–3 present regression results with the average total compensa-tion, the average cash compensation, and the fraction of equity compensation outof total compensation as dependent variables, respectively. Consistent with our

TABLE 3The Effect of the Management Quality Factor on Management Team Compensation

Table 3 reports the ordinary least squares (OLS) regression results of various executive compensation measures on ourmanagement quality factor (MQF). For each firm-year, ln(AVG_TOT_COMP) is the natural logarithm of the amount of to-tal compensation divided by the number of managers. ln(AVG_CASH_COMP) is the natural logarithm of the amount ofcash compensation divided by the number of managers. EQUITY/TOT_COMP is defined as the fraction of equity com-pensation out of total compensation. Total compensation includes cash compensation and equity compensation. Cashcompensation consists of base cash salary and bonus. Equity compensation consists of the value of shares awarded,the value of the long-term incentive plan (LTIP) awarded, and the value of options awarded. ln(ASSETS) is the naturallogarithm of a firm’s total assets; MB is Tobin’s Q , defined as the market value of assets divided by the book value ofassets, where the market value of assets is computed as the book value of assets plus the market value of commonstock less the book value of common stock; ROA is defined as operating income before depreciation divided by totalassets; CAPEX/ASSETS is defined as capital expenditures divided by total assets; R&D/ASSETS is defined as researchand development expenses divided by total assets; RET is a firm’s annual stock return; HHI is the industry Herfindahl–Hirschman index; and AVG_TENURE is the average number of years that each manager has worked as VP or higher inthis firm. Constants, year fixed effects, and 2-digit SIC industry fixed effects are included in all regressions. All standarderrors are adjusted for clustering at the firm level and are reported in parentheses below the coefficient estimates. *, **,and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

ln(AVG_TOT_COMP) ln(AVG_CASH_COMP) EQUITY/TOT_COMP

Variable 1 2 3

MQF 0.063*** 0.011 0.027***(0.015) (0.011) (0.004)

ln(ASSETS) 0.472*** 0.254*** 0.064***(0.009) (0.006) (0.003)

MB 0.478*** 0.119*** 0.090***(0.029) (0.019) (0.008)

ROA −0.047 0.131** −0.001(0.075) (0.062) (0.026)

CAPEX/ASSETS 0.049 −0.325** 0.051(0.210) (0.153) (0.059)

R&D/ASSETS 0.366*** 0.001 0.079*(0.130) (0.084) (0.045)

RET 0.150*** 0.063*** 0.018***(0.013) (0.008) (0.004)

HHI 1.035*** −0.613* 0.522***(0.393) (0.336) (0.162)

AVG_TENURE −0.019*** 0.005 −0.009***(0.005) (0.004) (0.002)

No. of obs. 12,240 12,232 12,370Adj. R 2 0.530 0.462 0.193

Industry fixed effects Yes Yes YesYear fixed effects Yes Yes Yes

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




expectation, the coefficients on MQF are positive and statistically significant atthe 5% level for two out of three specifications. The economic magnitude of theeffect of MQF is significant as well. For instance, a one interquartile range in-crease in MQF is associated with a 5.8% increase in the average total compen-sation. The positive and significant relationship between MQF and managementteam compensation suggests that MQF is a valid measure of management quality.

C. Measures of the Quantity and Quality of Corporate InnovationFollowing the existing literature (e.g., Kogan et al. (2017), Seru (2014)), we

use patent-based metrics to capture firm innovativeness. Although we also useR&D expenditures as a measure of investments in innovative activity, patent-based measures are widely used proxies of innovation output. We obtain patentdata from the database created by KPSS, which provides detailed information onmore than 6 million patents granted by the USPTO from 1926 to 2011. The KPSSdata set has matched assignees in the patent data set with CRSP PERMNOs if theassignee is a public corporation or subsidiary of a public corporation.

Patent data are subject to two types of truncation problems. First, patents arerecorded in the data set only after they are granted, and the lag between patentapplications and patent grants is significant (approximately 2 years on average).As we approach the last few years for which patent data are available, we observea smaller number of patent applications that are eventually granted. Many patentapplications filed during these years were still under review and had not beengranted by 2011. We partially mitigate this bias by restricting our analyses to 2years before the patent data end (i.e., in 2009). Further, following Hall, Jaffe, andTrajtenberg (2001), we correct this bias by dividing each patent for each firm-year by the mean number of patents for all firms for that year in the same 3-digit technology class as the patent. The second type of truncation problem stemsfrom citation counts. Patents tend to receive citations over a long period of time,so the citation counts of more recent patents are significantly downward biased.Following Hall et al. (2001) and Seru (2014), this bias is accounted for by scalingthe citations of a given patent by the total number of citations received by allpatents in that year in the same 3-digit technology class as the patent. Note thatthis methodology gives us class-adjusted measures of patents and citations, whichadjust for trends in innovative activity in particular industries.

We construct three measures for a firm’s annual innovation output based onthe patent application year.17 The first measure, ln(PATENT), is the natural loga-rithm of 1 plus the class-adjusted patent count for a firm in a given year. Specifi-cally, this variable counts the total number of (class-adjusted) patent applicationsfiled that year that were eventually granted. However, a simple count of patentsmay not distinguish breakthrough innovations from incremental technological dis-coveries. Therefore, we consider two additional measures. The second measure,ln(CITE), is the natural logarithm of 1 plus the class-adjusted total number of cita-tions received by a firm’s patents filed in a given year. The third measure, ln(CPP),

17Consistent with the innovation literature (e.g., Griliches, Pakes, and Hall (1988)), the applicationyear is more relevant for our purposes than the grant year since it is closer to the time of the actualinnovation.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




is the natural logarithm of 1 plus the number of citations per patent, constructedby taking natural logarithm of 1 plus the total number of class-adjusted citationsa firm receives on all the patents it applies for in a given year and normalizing itby 1 plus the total number of class-adjusted patents applied for in that year. Wetake the natural logarithm because the distributions of patents and citations areright skewed. To avoid losing observations with 0 patents or 0 citations, we add 1to the actual values. In untabulated analyses, we construct an additional measureof the quality of corporate innovation using the logged market value of patentsthat were filed by a firm in a given year following Kogan et al. (2017), and weobtain consistent results. Table 1 reports the summary statistics of our innovationmeasures. The median R&D to assets ratio in our sample is 1%. Further, an aver-age (median) firm in our sample has 0.862 (0) class-adjusted patents. An average(median) firm in our sample has 0.034 (0) class-adjusted citations.

D. Measures of Innovation StrategiesWe consider a suite of measures to describe firms’ innovation strategies. The

first set of measures captures the extent to which a firm’s new patents use newversus existing knowledge. Following Brav et al. (2018), we construct measuresof explorative and exploitative patents (EXPLORE and EXPLOIT), respectively.EXPLORE (EXPLOIT) is the fraction of explorative (exploitative) patents out ofall the patents filed by a firm in a given year. A patent is explorative if at least 80%of its citations do not refer to existing knowledge; a patent is exploitative if at least80% of its citations refer to existing knowledge. Existing knowledge includes afirm’s previous patent portfolio and all the patents that were cited by the firm’spatents filed over the past 5 years. Firms venturing into new knowledge and tech-nologies are likely to generate more explorative innovations, while firms focusingon existing knowledge and technologies are likely to generate more exploitativeinnovations.

Second, following Balsmeier et al. (2017), we categorize our sample pool ofpatents into three groups based on their citation distribution among all patents filedin the same technology class and year: TOP 10, which measures very successfulinnovations and is defined as patents receiving the number of citations in the top10% among all patents in the same 3-digit technology class and application year;NO CITE, which measures very unsuccessful innovations and is defined as thosereceiving 0 citations through the end of our sample period; and M CITE, whichmeasures moderately successful innovations and is defined as those receiving atleast one citation but not in the top 10%. Firms pursuing innovations in areas thatare newer and less familiar to them are expected to have a greater number of verysuccessful and very unsuccessful innovations, whereas firms pursuing innovationsthat are more familiar to them are expected to have a greater number of moderatelysuccessful innovations.

Third, we consider the number of times a firm’s patents cite patents ownedby the same firm and the number of times a firm’s patents cite patents owned byother firms (Sørensen and Stuart (2000), Faleye, Hoitash, and Hoitash (2011)).S CITE is the natural logarithm of 1 plus the number of self-citations (i.e., thenumber of times that a firm’s patent portfolio in a given year cites other patentsowned by the same firm). NS CITE is the natural logarithm of 1 plus the number

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




of non-self-citations (i.e., the number of times that a firm’s patent portfolio ina given year cites patents owned by other firms). Firms pursuing innovations inareas that are more familiar to them are likely to have more self-citations, whereasfirms pursuing innovations in areas that are newer and less familiar to them arelikely to have more non-self-citations.

Finally, we consider the innovation diversity of a firm’s patent portfolio. Wemeasure innovation diversity (DIVERSITY or 3YR DIVERSITY) as 1 minus theHerfindahl index of the number of patents filed by a firm across the 3-digit tech-nology classes in a given year or in the following 3 years. Firms adopting moreexplorative innovation strategies are likely to have greater innovation diversity(i.e., higher diversification across different technology classes).

E. Measures of Inventor MobilityTo identify inventor mobility, we collect the inventor information of each

patent from the U.S. Patent Inventor Database (1975–2010) (provided by Li et al.(2014)) from the Harvard Business School Dataverse. The U.S. Patent InventorDatabase includes inventor names, inventor addresses, assignee names, and ap-plication and grant dates for each patent. More importantly, it identifies uniqueinventors over time so that we can potentially track the moves of each inventor.Following Marx et al. (2009), we identify mobile inventors as changing employ-ers if they have ever filed two successive patent applications that are assigned todifferent entities. Because we need at least two patents to detect a move, inventorswho have filed a single patent throughout their career are necessarily excludedfrom our analysis.

For a given firm, an inventor’s move-in year is the year when the inven-tor filed his or her first patent in this firm; the inventor’s move-out year is thatwhen the inventor filed his or her first patent in the subsequent firm. For an in-ventor’s very last employer, we assume that this inventor stayed with that firmand did not move out.18 For example, in the inventor database, an inventor namedChristopher L. Holderness filed two patent applications through 2010. He filedpatent applications with Corning Inc. in 1999 and then with Dell Inc. in 2003.Thus, for Corning, Mr. Holderness’s move-in year is 1999, and his move-out yearis 2003, and for Dell, Mr. Holderness’s move-in year is 2003, and he has stayedwith Dell since 2003. Once we identify each mobile inventor’s move-in and move-out year, we aggregate the number of mobile inventors who move in and move outat the firm-year level to obtain the total inflows and outflows of mobile inventorsfor a given firm in a year. We define the difference between the natural logarithmof 1 plus the inflow and the natural logarithm of 1 plus the outflow as the netinflow of mobile inventors (NET INFLOW).

To examine the moves of inventors with different innovative ability, we clas-sify mobile inventors into two groups, namely, high-quality and low-quality in-ventors. For each inventor, we look at the average quality of his or her historicalpatents (i.e., the citations per patent for all the patents the inventor has filed prior

18As a robustness check, we redefine the dates that the inventor moved out of the last employer as1 or 2 years after the inventor filed his or her last patent in that firm. Our results remain qualitativelysimilar with this alternative definition.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




to the current year). If an inventor’s historical citations per patent are higher thanthe sample median, the inventor is considered as a high-quality inventor; other-wise, he or she is a low-quality inventor. We aggregate the number of high-quality(low-quality) inventors at the firm-year level to get the annual inflow and outflowof the high-quality (low-quality) inventors for a firm.

F. Other VariablesFollowing the existing literature, we obtain firms’ financial information from

Compustat and stock price data from CRSP and control for a number of firm char-acteristics that could affect firms’ innovation output. We compute all variables forfirm i over its fiscal year t . The controls include ln(ASSETS), which is the naturallogarithm of book value of total assets; MB, which is the Tobin’s Q, defined as themarket value of assets divided by the book value of assets, where the market valueof assets is computed as the book value of assets plus the market value of commonstock less the book value of common stock; ROA, which is defined as operatingincome before depreciation divided by total assets; CAPEX/ASSETS, which isdefined as capital expenditures over total assets; R&D/ASSETS, which is definedas research and development expenses divided by total assets; HHI, which is thevalue of Herfindahl–Hirschman index in a firm’s industry (defined at the 2-digitSIC code level) in a given year; RET, which is a firm’s prior-12-month annualcompounded stock return; and AVG TENURE, which is the average number ofyears that each manager has worked for a firm. To minimize the effect of outliers,we winsorize all independent variables at the 1st and 99th percentiles. Table 1provides summary statistics for the control variables described above. The me-dian firm size in our sample is $323 million, suggesting that our sample consistsof mainly mid-size and large firms. The median firm in our sample has an ROA of10.2%, a CAPEX-to-assets ratio of 3.5%, a Tobin’s Q of 1.6, and an annual stockreturn of 3.3%.

IV. Empirical Tests and Results

A. The Effect of Management Quality on R&D ExpendituresWe empirically test whether firms with higher-quality management teams are

likely to devote a greater amount of resources to innovative activities (Hypothe-sis 1) by estimating the following regression:

(3) R&D/ASSETSi ,t+n = α+βMQFi ,t + γZi ,t + IND FE+YR FE+ εi ,t ,

where i indexes firm; t indexes time; and n equals 1, 2, or 3. Z is a vector of controlvariables including ln(ASSETS), MB, ROA, CAPEX/ASSETS, RET, HHI, andAVG TENURE. We include year fixed effects (YR FE) and 2-digit SIC industryfixed effects (IND FE).19 In all regressions throughout the paper, standard errorsare clustered at the firm level.20

19Our results are insensitive to defining industry fixed effects at the 3- or 4-digit SIC code level.20We also examine the same regressions controlling for industry, year, and state fixed effects;

industry × year fixed effects; and industry × state × year fixed effects, with similar results. We re-port the regression results controlling for industry × state × year fixed effects in Table IA-9 in theSupplementary Material.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




Table 4 reports the results of equation (3). Columns 1–3 use R&D-to-assetsratios of 1, 2, and 3 years ahead of the year in which management quality ismeasured as dependent variables. We find that the coefficients of MQF in allthree specifications are positive and significant, both statistically and economi-cally. For instance, the coefficient in column 1 suggests that a one interquartilerange increase in MQF is associated with an increase of 0.74 percentage pointsin R&D/ASSETS for next year, which is equivalent to 70% of the sample me-dian. These results suggest that firms with higher-quality management teams areassociated with greater innovation input, supporting Hypothesis 1.

TABLE 4The Effect of Management Quality on R&D Expenditures

Table 4 reports the OLS regression results of the ratio of research and development (R&D) expenditures to total assetson our management quality factor (MQF). Control variables are defined as in Table 3. Constants, year fixed effects, and2-digit SIC industry fixed effects are included in all regressions. All standard errors are adjusted for clustering at the firmlevel and are reported in parentheses below the coefficient estimates. *, **, and *** indicate statistical significance at the10%, 5%, and 1% levels, respectively.

R&D/ASSETSt+1 R&D/ASSETSt+2 R&D/ASSETSt+3

Variable 1 2 3

MQF 0.008*** 0.008*** 0.007***(0.001) (0.001) (0.001)

ln(ASSETS) −0.004*** −0.005*** −0.005***(0.001) (0.001) (0.001)

MB 0.057*** 0.043*** 0.039***(0.002) (0.002) (0.002)

ROA −0.216*** −0.198*** −0.194***(0.010) (0.010) (0.010)

CAPEX/ASSETS −0.052*** −0.052*** −0.047***(0.011) (0.011) (0.012)

RET −0.000 −0.006*** −0.004***(0.001) (0.001) (0.001)

HHI 0.029* 0.014 0.007(0.016) (0.017) (0.019)

AVG_TENURE −0.001*** −0.001*** −0.001***(0.000) (0.000) (0.000)

No. of obs. 27,688 23,741 19,887Adj. R 2 0.559 0.493 0.472


B. The Effect of Management Quality on Corporate InnovationHere we study the relation between overall management team quality and

corporate innovation output (Hypothesis 2), both quantity (measured by the num-ber of patents) and quality (measured by total citations and citations per patent).We estimate the following model:

INNOV OUTPUTi ,t+n = α+βMQFi ,t + γ Z i ,t(4)+ IND FE+YR FE+ εi ,t ,

where the dependent variables (INNOV OUTPUT) include ln(PATENT),ln(CITE), and ln(CPP). Since the innovation process takes time, we examine theeffect of a firm’s management quality on its innovation as of 1, 2, and 3 years after

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




the year in which MQF is measured. Z is the set of control variables similar tothose in the previous section, but it now also includes R&D/ASSETS.21

Table 5 reports the results of the aforementioned tests. Across all the specifi-cations, the coefficients of MQF are positive and significant, both statistically andeconomically.22 For example, column 1 of Panel A suggests that a one interquar-tile range increase in MQF is associated with a 12.8% increase in the next year’sadjusted number of patents. These results support Hypothesis 2; that is, firms withhigher-quality management teams are associated with greater and higher-qualityinnovation output.23

In untabulated analyses, we further test whether firms with higher-qualitymanagement teams are able to use R&D resources more efficiently in producinginnovation output. Following Hirshleifer et al. (2013), we construct two measuresof innovative efficiency: the number of patents per dollar of R&D expendituresand the number of citations per dollar of R&D expenditures. Due to space limi-tations, we report the regression results of innovative efficiency measures on ourmanagement quality factor in Table IA-1 of the Supplementary Material. We findthat our management quality factor is positively and significantly associated withboth measures of innovation efficiency from 1 year up to 3 years from now. Thus,our evidence indicates that firms with higher-quality management teams are betterat getting more “bang for the buck,” that is, using R&D resources more efficientlyin generating innovation output, supporting Hypothesis 3.

C. Individual Measures of Top Management Quality and InnovationIn this section, we study the relation between various individual aspects

of top management team quality and corporate innovation. Table 6 reports ourregression results of innovation output on each individual management qual-ity measure. Specifically, we use the values of TEAM SIZE, MBA, PHD,WORK EXP, BOARD EXP, EMP CONN, and EDU CONN as independent vari-ables in columns 1–7 across all panels in the table.24 We find that most of our indi-vidual measures of management quality are positively and significantly associatedwith all three innovation output measures. These effects are economically signif-icant as well. For instance, a 1-interquartile-range increase in MBA is associatedwith a 4% increase in the number of patents.

Further, we empirically investigate whether our management quality factor(MQF) and individual management quality measures have differential effects oninnovation in younger firms (0–2 years after IPO) versus older firms (3 years or

21In untabulated analyses, we estimate our regressions using the negative binomial maximum-likelihood estimation technique using the raw number of patents and the total number of citationsreceived by these patents as dependent variables and obtain qualitatively similar results.

22In a robustness check, we conduct regressions (reported in Table IA-7 of the SupplementaryMaterial) using the sample of firms that have filed at least one patent application throughout oursample period of 1999–2009, and we find similar results.

23In analyses reported in Table IA-2 of the Supplementary Material, we also use the logged marketvalue of patents that were filed by a firm in a given year as the dependent variable, following Koganet al. (2017), and we obtain consistent results.

24Our results are similar when we use size-adjusted individual measures of management quality.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 5The Effect of Management Quality on the Quantity and Quality of Corporate Innovation

Table 5 reports the OLS regression results of the quantity and quality of corporate innovation output on our managementquality factor (MQF). Panels A, B, and C report regression results with the number of patents, the total number of citations,and the number of citations per patent as dependent variables, respectively. ln(PATENT) is the natural logarithm of 1 plusthe truncation-adjusted number of patents filed by a firm in a given year; ln(CITE) is the natural logarithm of 1 plus thetotal adjusted number of citations filed by a firm in a given year; ln(CPP) is the natural logarithm of 1 plus the adjustednumber of citations per patent. Control variables are defined as in Table 3. Constants, year fixed effects, and 2-digit SICindustry fixed effects are included in all regressions. All standard errors are adjusted for clustering at the firm level andare reported in parentheses below the coefficient estimates. *, **, and *** indicate statistical significance at the 10%, 5%,and 1% levels, respectively.

Panel A. The Effect of MQF on the Number of Patents

ln(PATENT)t+1 ln(PATENT)t+2 ln(PATENT)t+3Variable 1 2 3

MQF 0.128*** 0.129*** 0.128***(0.011) (0.012) (0.013)

ln(ASSETS) 0.146*** 0.142*** 0.138***(0.006) (0.006) (0.006)

MB 0.113*** 0.121*** 0.122***(0.011) (0.012) (0.012)

ROA −0.010 0.002 0.005(0.021) (0.022) (0.023)

CAPEX/ASSETS −0.043 −0.021 −0.008(0.064) (0.066) (0.070)

R&D/ASSETS 0.234*** 0.190*** 0.157***(0.045) (0.045) (0.045)

RET −0.031*** −0.026*** −0.016***(0.005) (0.005) (0.005)

HHI 0.109 0.067 −0.026(0.183) (0.191) (0.182)

AVG_TENURE 0.002 0.002 0.002(0.002) (0.002) (0.002)

No. of obs. 27,688 24,519 21,250Adj. R 2 0.390 0.384 0.375


Panel B. The Effect of MQF on the Total Number of Citations

ln(CITE)t+1 ln(CITE)t+2 ln(CITE)t+3Variable 1 2 3

MQF 0.016*** 0.016*** 0.016***(0.002) (0.002) (0.002)

ln(ASSETS) 0.018*** 0.017*** 0.017***(0.001) (0.001) (0.001)

MB 0.013*** 0.013*** 0.013***(0.002) (0.002) (0.002)

ROA −0.010*** −0.010*** −0.008**(0.003) (0.003) (0.003)

CAPEX/ASSETS −0.006 −0.007 −0.009(0.010) (0.010) (0.010)

R&D/ASSETS −0.000 −0.005 −0.005(0.006) (0.006) (0.006)

RET −0.003*** −0.003*** −0.003***(0.001) (0.001) (0.001)

HHI −0.021 −0.034 −0.035(0.032) (0.033) (0.031)

AVG_TENURE 0.000 0.000 0.000(0.000) (0.000) (0.000)

No. of obs. 27,688 24,519 21,250Adj. R 2 0.256 0.251 0.241


(continued on next page)

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 5 (continued)The Effect of Management Quality on the Quantity and Quality of Corporate Innovation

Panel C. The Effect of MQF on the Number of Citations per Patent

ln(CPP)t+1 ln(CPP)t+2 ln(CPP)t+3Variable 1 2 3

MQF 0.001*** 0.001*** 0.001***(0.000) (0.000) (0.000)

ln(ASSETS) 0.001*** 0.001*** 0.001***(0.000) (0.000) (0.000)

MB 0.000*** 0.001*** 0.000***(0.000) (0.000) (0.000)

ROA −0.000 −0.000 −0.000(0.000) (0.000) (0.000)

CAPEX/ASSETS −0.000 −0.000 −0.001(0.000) (0.000) (0.001)

R&D/ASSETS 0.001** 0.000 0.000(0.000) (0.000) (0.000)

RET −0.000* −0.000 −0.000*(0.000) (0.000) (0.000)

HHI −0.005** −0.005** −0.004(0.002) (0.002) (0.002)

AVG_TENURE 0.000 0.000 0.000(0.000) (0.000) (0.000)

No. of obs. 27,688 24,519 21,250Adj. R 2 0.140 0.139 0.136


greater after IPO).25 We perform our baseline regressions (as in equation (4)) inyounger and older firms and test whether the effects of MQF and each individualmeasure on innovation are significantly different across these two kinds of firms.We report the results of these tests in Table 7.26 In columns 1 and 2, we first ana-lyze the relation between MQF and corporate innovation in younger versus olderfirms. We find that MQF is positively and significantly associated with the num-ber of patents for both groups and that the difference of the effects of MQF onthe number of patents is not significant. When we analyze individual measuresof management quality, we observe some interesting heterogeneity in the relationbetween these individual measures and corporate innovation for younger firmsversus older firms. We find that management team size, the fraction of managerswith MBA degrees, and education-based connections positively and significantlyaffect corporate innovation only for older firms.27 Prior managerial work expe-rience positively and significantly affects corporate innovation only for youngerfirms. Two individual measures, namely, the fraction of top managers with Ph.D.degrees and employment-based connections, positively and significantly affectcorporate innovation in both younger and older firms; however, the magnitude

25We collect the IPO date for each firm from the SDC Platinum New Issues Database. We cross-check the IPO date with Compustat and drop observations for which the two data sources provideinconsistent IPO dates.

26We obtain consistent results using citations and citations per patent across different time horizonsas dependent variables.

27Note that the differences in the regression coefficients on education-based connections are notsignificantly different for younger versus older firms.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 6The Effect of Individual Management Quality Measures on Corporate Innovation

Table 6 reports the OLS regression results of corporate innovation on individual management quality measures. Panels A,B, and C report regression results with the number of patents, the total number of citations, and the number of citationsper patent as dependent variables, respectively. ln(PATENT) is the natural logarithm of 1 plus the truncation-adjustednumber of patents filed by a firm in a given year; ln(CITE) is the natural logarithm of 1 plus the total adjusted numberof citations filed by a firm in a given year; ln(CPP) is the natural logarithm of 1 plus the adjusted number of citationsper patent. Individual management quality measures are defined in detail in Table 2. Control variables are the same asin Table 3 in all regressions and coefficient estimates on controls are not reported to save space. Constants, year fixedeffects, and 2-digit SIC industry fixed effects are included in all regressions. All standard errors are adjusted for clusteringat the firm level and are reported in parentheses below the coefficient estimates. The coefficients and standard errors inPanel C are multiplied by 100 for ease of reading. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%levels, respectively.

Panel A. The Effect of Individual Management Quality Measures on the Number of Patents

ln(PATENT)t+1

Variable 1 2 3 4 5 6 7

TEAM_SIZE 0.029***(0.003)

MBA 0.120***(0.033)

PHD 0.356***(0.057)

WORK_EXP 0.035(0.035)

BOARD_EXP −0.002(0.026)

EMP_CONN 0.013***(0.001)

EDU_CONN 0.012***(0.002)

No. of obs. 27,688 27,688 27,688 27,688 27,688 27,688 27,688Adj. R2 0.389 0.370 0.374 0.369 0.369 0.391 0.372

Controls Yes Yes Yes Yes Yes Yes YesIndustry fixed effects Yes Yes Yes Yes Yes Yes YesYear fixed effects Yes Yes Yes Yes Yes Yes Yes

Panel B. The Effect of Individual Management Quality Measures on the Total Number of Citations

ln(CITE)t+1

Variable 1 2 3 4 5 6 7

TEAM_SIZE 0.004***(0.000)

MBA 0.013***(0.005)

PHD 0.021***(0.007)

WORK_EXP −0.006(0.005)

BOARD_EXP −0.002(0.003)

EMP_CONN 0.002***(0.000)

EDU_CONN 0.001***(0.000)

No. of obs. 27,688 27,688 27,688 27,688 27,688 27,688 27,688Adj. R2 0.261 0.240 0.240 0.239 0.239 0.259 0.242


(continued on next page)

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 6 (continued)The Effect of Individual Management Quality Measures on Corporate Innovation

Panel C. The Effect of Individual Management Quality Measures on the Number of Citations per Patent

ln(CPP)t+1

Variable 1 2 3 4 5 6 7

TEAM_SIZE 0.013***(0.002)

MBA 0.059**(0.027)

PHD 0.112***(0.042)

WORK_EXP −0.004(0.028)

BOARD_EXP −0.024(0.018)

EMP_CONN 0.005***(0.001)

EDU_CONN 0.005***(0.002)

No. of obs. 27,688 27,688 27,688 27,688 27,688 27,688 27,688Adj. R2 0.140 0.136 0.136 0.135 0.135 0.139 0.136


of the effect of the Ph.D. measure is greater for corporate innovation in youngerfirms. These results are broadly consistent with Hypothesis 4.

D. Management Quality and Corporate Innovation StrategiesIn this section, we investigate the effect of management quality on firms’

innovation strategies. As we describe in Section III.D, we adopt a suite ofmeasures to capture the extent to which firms engage in explorative versus ex-ploitative innovation strategies.

First, we study the relation between management quality and the fraction ofexplorative patents (using at least 80% new knowledge) and that of exploitativepatents (using at least 80% existing knowledge) out of all the patents filed by afirm in a given year (i.e., EXPLORE and EXPLOIT, respectively). Table 8 re-ports the results of these tests. We find that the coefficients of MQF are positiveand significant across all specifications, suggesting that firms with higher man-agement quality are associated with a greater fraction of explorative patents aswell as a greater fraction of exploitative patents. Further, we find that the coeffi-cients on MQF in the odd columns are significantly greater than those in the evencolumns, suggesting that firms with higher-quality management teams are likelyto engage more in explorative innovation strategies relative to exploitative innova-tion strategies, thus producing more innovations venturing into the developmentof new knowledge and technologies.

We use several other measures to describe a firm’s innovation strategy andreport these results in Table 9. In Panel A, following Balsmeier et al. (2017),we study the relation between management quality and a firm’s innovation strat-egy by looking at the very successful (TOP 10), very unsuccessful (NO CITE),

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497



Chem

manur,K

ong,Krishnan,and

Yu27

TABLE 7The Effect of Management Quality Factor and Individual Measures of Management Quality on Innovation for Younger versus Older Firms

Table 7 reports the OLS regression results of corporate innovation on our management quality factor (MQF) and individual measures of management quality for younger (0–2 years after initial public offering(IPO)) versus older firms (3 years or greater after IPO). ln(PATENT) is the natural logarithm of 1 plus the truncation-adjusted number of patents filed by a firm in a given year. Individual management qualitymeasures are defined in detail in Table 2. Control variables are the same as in Table 3 and coefficients are not reported to save space. Constants, year fixed effects and 2-digit SIC industry fixed effects areincluded in all regressions. All standard errors are adjusted for clustering at the firm level and are reported in parentheses below the coefficient estimates. *, **, and *** indicate statistical significance at the10%, 5%, and 1% levels, respectively.

Panel A. The Effect of Management Quality Measures on the Number of Patents for Younger Firms versus Older Firms

ln(PATENT)t+1

Younger Older Younger Older Younger Older Younger Older Younger Older Younger Older Younger Older Younger Older

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

MQF 0.067** 0.084***(0.027) (0.017)

TEAM_SIZE 0.006 0.018***(0.007) (0.004)

MBA −0.049 0.064***(0.049) (0.022)

PHD 0.659*** 0.490***(0.135) (0.082)

WORK_EXP 0.249*** 0.059(0.091) (0.052)

BOARD_EXP 0.096 −0.013(0.095) (0.041)

EMP_CONN 0.009*** 0.008***(0.003) (0.002)

EDU_CONN 0.008 0.008***(0.005) (0.003)

No. of obs. 1,923 11,212 1,923 11,212 1,923 11,212 1,923 11,212 1,923 11,212 1,923 11,212 1,923 11,212 1,923 11,212Adj. R2 0.285 0.342 0.281 0.340 0.281 0.332 0.315 0.346 0.287 0.332 0.281 0.332 0.289 0.340 0.282 0.335

Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes YesIndustry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes YesYear fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Panel B. The Differences of the Effects of Management Quality Measures on the Number of Patents

Difference Difference Difference Difference Difference Difference Difference Difference1–2 3–4 5–6 7–8 9–10 11–12 13–14 15–16

Difference −0.017 −0.012* −0.113** 0.169* 0.190** 0.110 0.000 −0.001

https://doi.org/10.1017/S0022109018001497Downloaded from https://www.cambridge.org/core. Boston College, on 01 Jul 2019 at 00:49:45, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.

https://doi.org/10.1017/S0022109018001497




TABLE 8The Effect of Management Quality on Explorative versus Exploitative Innovations

Panel A of Table 8 reports the OLS regression results of a firm’s explorative innovations and exploitative innovations onour management quality factor (MQF). EXPLORE is the fraction of explorative patents out of all the patents filed by a firmin a given year; EXPLOIT is the fraction of exploitative patents out of all the patents filed by a firm in a given year. A patentis explorative if at least 80% of its citations do not refer to existing knowledge, which includes a firm’s previous patentportfolio and all the patents that were cited by a firm’s patents filed over the past 5 years. A patent is exploitative if atleast 80% of its citations refer to existing knowledge, which includes a firm’s previous patent portfolio and all the patentsthat have been cited by a firm’s patents filed over the past 5 years. Panel B reports the difference between the coefficientestimates of MQF using EXPLORE and EXPLOIT as dependent variables and tests their statistical differences. Controlvariables are defined as in Table 3. Constants, year fixed effects and 2-digit SIC industry fixed effects are included in allregressions. All standard errors are adjusted for clustering at the firm level and are reported in parentheses below thecoefficient estimates. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A. The Effect of MQF on Explorative and Exploitative Patents

EXPLOREt+1 EXPLOITt+1 EXPLOREt+2 EXPLOITt+2 EXPLOREt+3 EXPLOITt+3

Variable 1 2 3 4 5 6

MQF 0.027*** 0.016*** 0.026*** 0.016*** 0.028*** 0.016***(0.004) (0.002) (0.004) (0.002) (0.004) (0.002)

ln(ASSETS) 0.037*** 0.013*** 0.036*** 0.013*** 0.034*** 0.014***(0.002) (0.001) (0.002) (0.001) (0.002) (0.001)

MB 0.025*** 0.012*** 0.025*** 0.014*** 0.024*** 0.016***(0.005) (0.002) (0.005) (0.002) (0.005) (0.003)

ROA 0.022** −0.001 0.015 0.005 0.016 0.008(0.011) (0.006) (0.011) (0.006) (0.011) (0.006)

CAPEX/ASSETS 0.007 −0.017 0.034 −0.010 0.021 0.002(0.026) (0.013) (0.027) (0.014) (0.028) (0.015)

R&D/ASSETS 0.225*** 0.074*** 0.168*** 0.077*** 0.140*** 0.095***(0.022) (0.012) (0.021) (0.012) (0.021) (0.014)

RET −0.002 −0.002 −0.005 −0.004*** 0.004 −0.002(0.003) (0.001) (0.003) (0.001) (0.003) (0.001)

HHI 0.177* 0.051 0.114 0.054 −0.050 0.078*(0.096) (0.036) (0.100) (0.038) (0.130) (0.041)

AVG_TENURE −0.001 0.001* −0.000 0.001 0.000 0.000(0.001) (0.000) (0.001) (0.000) (0.001) (0.000)

No. of obs. 27,688 27,688 24,519 24,519 21,250 21,250Adj. R 2 0.214 0.117 0.207 0.116 0.202 0.121

Industry fixed effects Yes Yes Yes Yes Yes YesYear fixed effects Yes Yes Yes Yes Yes Yes

Panel B. The Differences of the Effects of MQF on Explorative versus on Exploitative Patents

Difference Difference Difference1–2 3–4 5–6

Difference 0.011*** 0.010*** 0.013***

and moderately successful innovations (M CITE), based on their citation distri-bution among all patents filed in the same technology class and year. We find thatmanagement quality is positively and significantly associated with the number ofpatents in all three categories. More interestingly, management quality has a morepronounced effect on successful patents than on unsuccessful patents or averagepatents; that is, firms with higher-quality management teams are better at motivat-ing the development of patents that are highly cited afterward.28 In Panel B, westudy the relation between management quality and non-self-citations (i.e., thenumber of times a firm’s patents cite other firms’ patents) and self-citations (i.e.,the number of times a firm’s patents cite its own patents). We also test whether theeffect of management quality on non-self-citations is statistically different from

28We obtain qualitatively similar results using all three categories of innovations across other timehorizons as dependent variables (Table IA-3 of our Supplementary Material).

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 9The Effect of Management Quality on Other Measures of Innovation Strategies

Panel A of Table 9 reports the OLS regression results of the highly successful (TOP_10), unsuccessful (NO_CITE) andmoderately successful innovations (M_CITE) on our management quality factor (MQF). Panel B reports and comparesthe regression results of self-citations and non-self-citations on MQF. Panel C reports the regression results of innovationdiversity onMQF. TOP_10 is the natural logarithm of 1 plus a firm’s number of patents that received cites within the top 10%among all patents in the same 3-digit patent class and application year; NO_CITE is the natural logarithm of 1 plus thenumber of patents that received no citation; M_CITE is the natural logarithm of 1 plus the number of patents that receivedat least 1 citation but below the top 10% among all patents. NS_CITE is the natural logarithm of 1 plus the number ofnon-self-citations. S_CITE is the natural logarithm of 1 plus the number of self-citations. DIVERSITY (3YR_DIVERSITY) is 1minus the Herfindahl index of the number of patents filed by a firm across the 3-digit technology classes in a given year (inthe following 3 years). Control variables are the same as in Table 3 in all regressions and coefficients are not reported tosave space. Constants, year fixed effects and 2-digit SIC industry fixed effects are included in all regressions. All standarderrors are adjusted for clustering at the firm level and are reported in parentheses below the coefficient estimates. *, **,and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A. The Effect of MQF on the Highly Successful, Unsuccessful, and Moderately Successful Innovations

Difference Difference DifferenceTOP_10t+1 NO_CITEt+1 M_CITEt+1 1–2 1–3 2–3


MQF 0.303*** 0.147*** 0.144*** 0.156*** 0.159*** 0.003(0.038) (0.018) (0.020) (0.023) (0.022) (0.008)

No. of obs. 15,251 15,251 15,251Adj. R2 0.406 0.384 0.428

Controls, industry and year Yes Yes Yesfixed effects

Panel B. The Effect of MQF on Non-Self-Citations and Self-Citations

NS_CITEt+1 S_CITEt+1 NS_CITEt+2 S_CITEt+2 NS_CITEt+3 S_CITEt+3


MQF 0.316*** 0.253*** 0.313*** 0.251*** 0.304*** 0.244***(0.028) (0.023) (0.030) (0.024) (0.031) (0.026)

No. of obs. 27,688 27,688 24,519 24,519 21,250 21,250Adj. R2 0.389 0.331 0.382 0.328 0.374 0.322

Controls, industry and year Yes Yes Yes Yes Yes Yesfixed effects

Difference Difference Difference

1–2 3–4 5–6

Difference 0.062*** 0.062*** 0.060***

Panel C. The Effect of MQF on Innovation Diversity

DIVERSITYt+1 DIVERSITYt+2 DIVERSITYt+3 3YR_DIVERSITYt+1

Variable 1 2 3 4

MQF 0.036*** 0.039*** 0.036*** 0.041***(0.006) (0.007) (0.007) (0.007)

No. of obs. 8,295 7,016 5,750 8,851Adj. R2 0.355 0.359 0.358 0.328

Controls, industry and year Yes Yes Yes Yesfixed effects

that on self-citations and report these test results at the bottom of Panel B. Wefind that firms with higher-quality management teams are associated with a sig-nificantly greater number of non-self-citations than self-citations, suggesting thatsuch firms tend to pursue innovations in areas that are newer and less familiarto them. In Panel C, we study the relation between management quality and in-novation diversity, which captures the level of diversification of a firm’s patentportfolio over different technology classes. We find that firms with higher-qualitymanagement teams are associated with a greater level of innovation diversity forthe following 3 years. Collectively, our empirical results presented in this sectionsupport the argument that firms with higher-quality management teams are more

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




likely to engage in explorative innovation strategies, those involving new tech-nologies, and those that are likely to push the knowledge boundaries of the firmoutward (Hypothesis 5A).

E. Management Quality and Corporate Innovation across IndustriesIn untabulated analyses, we investigate whether the positive and significant

relationship established earlier between our MQF and corporate innovation isstronger in some industries than in others. We find that this relationship is strongerin financially constrained industries (as measured by the level of external financialdependence; see, e.g., Rajan and Zingales (1998)) and in more competitive indus-tries (as measured by the Herfindahl–Hirschman index in the firm’s industry).Due to space limitations, we report these results in Table IA-4 of our Supplemen-tary Material. These results suggest that firms with higher-quality managementteams are able to select better projects, use resources more efficiently, and gen-erate greater innovation output in adverse financing environments and in morecompetitive industries. These empirical findings lend support for Hypothesis 6and Hypothesis 7.

V. Analysis of the Inventor Mobility ChannelOur evidence thus far shows that top management quality is positively related

to corporate innovation. In this section, we analyze an important channel throughwhich this may occur. As argued before, higher-quality management teams mayprovide more R&D resources, manage R&D resources better, and provide a morerisk-tolerant climate for inventors to succeed in. This, in turn, may make firmswith higher-quality management teams more attractive to higher-quality inven-tors. Thus, one way that higher-quality management teams may enhance innova-tion is by hiring more and higher-quality inventors to work for the firm. In thissection, we test these hypotheses.

A. Management Quality and the Net Inflow of InventorsTo assess the relation between management quality and the net inflow of in-

ventors who move into a firm in a given year (Hypothesis 8), we test the followingmodel:

NET INFLOWi ,t+n = α+βMQFi ,t + γZi ,t + IND FE(5)+YR FE+ST FE+ εi ,t ,

where i indexes firm; t indexes time; and n equals 1, 2, or 3. Z is a vector ofcontrol variables used in prior tests. As before, we include year fixed effects and2-digit SIC industry fixed effects. Further, since location may impact an inventor’sdecision of moving into or out of a firm, we include state fixed effects (ST FE)for the state of the firm’s headquarters in all regressions in this section.

Table 10 reports the results of these tests. The coefficients of MQF are pos-itive and statistically and economically significant across all specifications. Forinstance, column 1 suggests that a one interquartile range increase in MQF is as-sociated with a 0.05 increase in NET INFLOW. The economic magnitude of suchan effect is significant, given that the sample mean of NET INFLOW is 0.21.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 10The Effect of Management Quality on the Net Inflow of Inventors

Table 10 reports the OLS regression results of the net inflow of inventors for a firm in a given year on our managementquality factor (MQF). NET_INFLOW is defined as the difference between the inflow and outflow of inventors, where theinflow and outflow of inventors are defined as the natural logarithm of 1 plus the total number of inventors who move inand who move out aggregated at the firm-year level, respectively. Control variables are defined as in Table 3. Constants;year fixed effects; 2-digit SIC industry fixed effects; and state fixed effects are included in all regressions. All standarderrors are adjusted for clustering at the firm level and are reported in parentheses below the coefficient estimates. *, **,and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

NET_INFLOWt+1 NET_INFLOWt+2 NET_INFLOWt+3

Variable 1 2 3

MQF 0.050*** 0.049*** 0.054***(0.007) (0.007) (0.007)

ln(ASSETS) 0.073*** 0.067*** 0.060***(0.003) (0.003) (0.003)

MB 0.076*** 0.080*** 0.076***(0.008) (0.008) (0.008)

ROA 0.051*** 0.053*** 0.050***(0.016) (0.015) (0.016)

CAPEX/ASSETS 0.123** 0.104** 0.070(0.050) (0.048) (0.050)

R&D/ASSETS 0.356*** 0.257*** 0.196***(0.039) (0.035) (0.036)

RET −0.025*** −0.016*** −0.007*(0.005) (0.004) (0.004)

HHI 0.344** 0.273* 0.136(0.146) (0.140) (0.156)

AVG_TENURE −0.004*** −0.003*** −0.002*(0.001) (0.001) (0.001)

No. of obs. 25,945 23,096 20,119Adj. R 2 0.255 0.244 0.234

Industry, year, and state fixed effects Yes Yes Yes

These findings support Hypothesis 8; that is, an important channel through whichhigher-quality management teams enhance corporate innovation is by hiring moreinventors.

In an untabulated analysis, we include the net inflow of inventors in the ordi-nary least squares (OLS) regressions for a firm’s innovation output. Therefore, weregress our corporate innovation output measures on the net inflow of inventorsand our MQF while including the same of set of control variables as in Table 5,year fixed effects, industry fixed effects, and state fixed effects. This test allowsus to check whether MQF still has a direct (residual) effect on innovation aftercontrolling for the net inflow of inventors. We find that whereas the coefficientson the net inflow of inventors all remain positive and significant, the coefficientson MQF are still positive and significant at the 1% level. However, the magni-tudes of the coefficients on MQF are reduced quite substantially once we includethe net inflow of inventors (e.g., the coefficient on MQF is decreased by 28%when ln(PATENT)t+1 is the dependent variable, and the decrease is statisticallysignificant at the 1% level), suggesting that the effect of management quality oninnovation is at least partly channeled through the net inflow of inventors.

B. Management Quality and High- and Low-Quality InventorsIn this section, we move on to test whether firms with higher-quality manage-

ment teams are better at attracting higher-quality inventors. We therefore conduct

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




regressions as in equation (5), using the net inflow of high-quality inventors andthat of low-quality inventors as dependent variables, respectively, and test whetherthe coefficients of our management quality factor in these regressions are signifi-cantly different from each other.

Panel A of Table 11 reports the regression results usingNET INFLOW HIGHi ,t and NET INFLOW LOWi ,t as dependent variablescalculated at 1, 2, and 3 years subsequent to the current year. We find that thecoefficients on MQF are positive using both dependent variables, suggesting thatmanagement quality has positive impacts on the net inflow of both high-qualityand low-quality inventors. More importantly, the effect of MQF on the netinflow of high-quality inventors is economically 10 times larger than on the netinflow of low-quality inventors across all time horizons. We test the statistical

TABLE 11The Effect of Management Quality on the Net Inflow of High-Quality and Low-Quality

Inventors

Panel A of Table 11 reports the OLS regression results of the net inflow of high-quality inventors and the net inflow of low-quality inventors for a firm in a given year on our management quality factor (MQF). NET_INFLOW_HIGH is the differencebetween the natural logarithm of 1 plus the number of high-quality inventors who move into the firm and the naturallogarithm of 1 plus the number of high-quality inventors who move out of the firm in a given year; NET_INFLOW_LOWis the difference between the natural logarithm of 1 plus the number of low-quality inventors who move into the firm andthe natural logarithm of 1 plus the number of low-quality inventors who move out of the firm in a given year. Panel Breports the difference between the coefficient estimates of MQF using NET_INFLOW_HIGH and NET_INFLOW_LOW asdependent variables and tests their statistical differences. Control variables are defined as in Table 3. Constants; yearfixed effects; 2-digit SIC industry fixed effects; and state fixed effects are included in all regressions. All standard errorsare adjusted for clustering at the firm level and are reported in parentheses below the coefficient estimates. *, **, and ***indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A. The Effects of MQF on the Net Inflow of High-Quality and Low-Quality Inventors

NET_INFLOW_ NET_INFLOW_ NET_INFLOW_ NET_INFLOW_ NET_INFLOW_ NET_INFLOW_HIGHt+1 LOWt+1 HIGHt+2 LOWt+2 HIGHt+3 LOWt+3


MQF 0.054*** 0.004*** 0.053*** 0.004*** 0.056*** 0.004***(0.007) (0.001) (0.007) (0.001) (0.007) (0.001)

ln(ASSETS) 0.078*** 0.004*** 0.071*** 0.004*** 0.064*** 0.003***(0.003) (0.000) (0.003) (0.000) (0.003) (0.000)

MB 0.079*** 0.002*** 0.084*** 0.003*** 0.080*** 0.003***(0.009) (0.001) (0.009) (0.001) (0.008) (0.001)

ROA 0.050*** 0.001 0.053*** 0.001 0.051*** 0.002*(0.016) (0.001) (0.016) (0.001) (0.016) (0.001)

CAPEX/ASSETS 0.123** 0.004 0.112** 0.003 0.082 −0.003(0.051) (0.004) (0.050) (0.004) (0.051) (0.004)

R&D/ASSETS 0.356*** 0.013*** 0.264*** 0.008*** 0.206*** 0.008**(0.040) (0.004) (0.036) (0.003) (0.038) (0.003)

RET −0.025*** −0.001** −0.018*** −0.000 −0.008* −0.001**(0.005) (0.000) (0.005) (0.000) (0.004) (0.000)

HHI 0.393*** −0.001 0.313** 0.001 0.161 0.001(0.151) (0.011) (0.145) (0.010) (0.162) (0.010)

AVG_TENURE −0.004*** 0.000 −0.003*** 0.000 −0.002* 0.000(0.001) (0.000) (0.001) (0.000) (0.001) (0.000)

No. of obs. 25,945 25,945 23,096 23,096 20,119 20,119Adj. R 2 0.263 0.063 0.253 0.061 0.241 0.060

Industry, year, and Yes Yes Yes Yes Yes Yesstate fixed effects

Panel B. Differences of the Effects of MQF on the Net Inflow of High-Quality and Low-Quality Inventors

Difference Difference Difference1–2 3–4 5–6

Difference 0.050*** 0.049*** 0.052***

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




significance of the difference between the coefficients on MQF for high-qualityversus low-quality inventors and report these test results in Panel B. We find thatall these differences are statistically significant at the 1% level.

In untabulated analyses, we also investigate whether management qualityis positively and significantly associated with the change in the average inven-tor quality of a firm. The quality of each inventor is measured as the citationsper patent for the patents the inventor has filed prior to the current year. The netchange in the average inventor quality of a firm in a given year is defined as thedifference between the average quality of all incoming inventors joining the firmand that of all outgoing inventors leaving the firm in a given year. We find that ourmanagement quality factor is indeed positively and significantly associated withthe net change in the average inventor quality.29 Collectively, our empirical find-ings provide strong evidence that firms with higher-quality management teams areable to hire a greater number of high-quality inventors than low-quality inventors,consistent with Hypothesis 9.

To further analyze whether hiring high-quality inventors is indeed a chan-nel through which firms with higher-quality management teams spur innovation,we include the net inflow of high-quality inventors in our OLS regressions forinnovation output. If the relation between management quality and innovation ischanneled through the hiring of high-quality inventors, we would expect the co-efficients of the net inflow of high-quality inventors to be positive and significant,whereas the magnitudes of the coefficients on MQF should become significantlylower.

Table 12 reports the results of these regressions. Panels A, B, and C use thenumber of patents, the total number of citations, and citations per patent, respec-tively, as dependent variables. The same set of control variables and fixed effectsas in Table 11 is included in all the regressions in Table 12. Consistent with ourhypothesis, we find that the coefficients of the net inflow of high-quality inventorsare positive and significant at the 1% level in all the regressions in these three pan-els. Further, the coefficients of our management quality factor (MQF) have muchsmaller magnitudes compared to the coefficients from regressions in which the netinflow of higher-quality inventors is not included (i.e., compared to our Table 5results), suggesting that the effect of MQF on corporate innovation is partiallymediated through hiring higher-quality inventors. For instance, the coefficient onMQF with the subsequent year adjusted patents as the dependent variable is 0.096when we control for the inflow of high-quality inventors, whereas it is 0.146 with-out this control (in Panel A of Table 5), reflecting a 34% decline. More interest-ingly, we find that the coefficients of MQF are smaller in the regressions wherethe net inflow of high-quality inventors is included compared to the case wherethe net inflow of all inventors is included.30 Collectively, these results support ourconjecture that an important channel through which higher-quality managementteams enhance innovation is by hiring higher-quality inventors.

29Due to space limitations, we report these results in Table IA-5 of our Supplementary Material.30In untabulated analysis, we find that the differences between the coefficients on MQF in these

two sets of regressions are statistically significant.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 12The Effect of the Net Inflow of High-Quality Inventors on Corporate Innovation

Table 12 reports the OLS regression results of corporate innovation on the net inflow of high-quality inventors and ourmanagement quality factor (MQF). Panels A, B, and C report regression results with the number of patents, the totalnumber of citations, and the number of citations per patent as dependent variables, respectively. ln(PATENT) is thenatural logarithm of 1 plus the truncation-adjusted number of patents filed by a firm in a given year; ln(CITE) is the naturallogarithm of 1 plus the total adjusted number of citations filed by a firm in a given year; ln(CPP) is the natural logarithm of1 plus the adjusted number of citations per patent. NET_INFLOW_HIGH is the difference between the natural logarithmof 1 plus the number of high-quality inventors who move into the firm and the natural logarithm of 1 plus the numberof high-quality inventors who move out of the firm in a given year. Control variables are the same as in Table 3 in allregressions and coefficients are not reported to save space. Constants; year fixed effects; 2-digit SIC industry fixedeffects; and state fixed effects are included in all regressions. All standard errors are adjusted for clustering at the firmlevel and are reported in parentheses below the coefficient estimates. *, **, and *** indicate statistical significance at the10%, 5%, and 1% levels, respectively.

Panel A. MQF, the Net Inflow of High-Quality Inventors, and the Number of Patents

ln(PATENT)t+1 ln(PATENT)t+2 ln(PATENT)t+3Variable 1 2 3

NET_INFLOW_HIGH 0.572*** 0.526*** 0.466***(0.013) (0.014) (0.014)

MQF 0.096*** 0.093*** 0.090***(0.008) (0.009) (0.010)

No. of obs. 25,945 23,096 20,119Adj. R 2 0.578 0.559 0.528

Controls, industry, year, and state fixed effects Yes Yes Yes

Panel B. MQF, the Net Inflow of High-Quality Inventors, and the Total Number of Citations

ln(CITE)t+1 ln(CITE)t+2 ln(CITE)t+3Variable 1 2 3

NET_INFLOW_HIGH 0.063*** 0.058*** 0.051***(0.003) (0.003) (0.003)

MQF 0.014*** 0.013*** 0.013***(0.001) (0.002) (0.002)

No. of obs. 25,945 23,096 20,119Adj. R 2 0.372 0.360 0.334


Panel C. MQF, the Net Inflow of High-Quality Inventors, and the Number of Citations per Patent

ln(CPP)t+1 ln(CPP)t+2 ln(CPP)t+3Variable 1 2 3

NET_INFLOW_HIGH 0.002*** 0.002*** 0.002***(0.000) (0.000) (0.000)

MQF 0.000*** 0.000*** 0.000***(0.000) (0.000) (0.000)

No. of obs. 25,945 23,096 20,119Adj. R 2 0.181 0.184 0.184


VI. Instrumental Variable AnalysisIn this section, we address the potential endogeneity concern discussed in

the Introduction using an IV analysis. The instrument we use is the number ofacquisitions in the industry and state of the sample firm 5 years prior weightedby an index measuring the enforceability index of noncompete clauses in thatstate aggregated to the national level. Our instrument is motivated by the follow-ing facts. First, potential managers available for hire by a firm often come fromestablished firms in the same industry and may leave such firms as a result of ac-quisitions. In other words, there is a strong correlation between the movement ofmanagers across firms and the number of acquisitions in the industry that the firm

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




belongs to.31 Second, the enforceability of noncompete clauses, which are com-monly used in the employment contracts for top management teams to prohibitthem from joining or founding a rival company within 1 or 2 years of leaving,affects the mobility of managers across firms.32,33,34

Specifically, the instrumental variable for the management quality factor(MQF) of the top management team in firm i in industry j in year t is computedas follows:

(6) INSTRUMENT j ,t =

∑s

ACQ j ,s,t−5×ENFORCE INDEXs,t ,

where j , s, and t index industry, state, and year, respectively. ACQ j ,s,t−5 is thenumber of acquisitions made by established (public) companies in industry j instate s in year t–5. The information on mergers and acquisitions required to con-struct this variable is collected from the SDC Mergers and Acquisitions Database.The 5-year lag allows for the expiration of retention contracts that work as “goldenhandcuffs” for managers, and thus ACQ j ,s,t−5 works as a measure of the supply ofmanagers from state s in industry j in year t .

ENFORCE INDEXs,t is the index measuring the enforceability of non-compete agreements across different U.S. states based on Garmaise (2011),which ranges from 0 to 1.35 Higher (lower) values of ENFORCE INDEXindicate weaker (greater) enforceability of noncompete clauses and thusgreater (weaker) mobility of managers. The multiplication term, ACQ j ,s,t−5×

ENFORCE INDEXs,t , therefore proxies for the supply of managers who are ableto move across firms and available for hire from state s in industry j in year t . Wethen aggregate this variable at the industry-year level across states and use this as

31In an earlier (working paper) version of their article, Ewens and Marx (2018) show that thenumber of acquisitions in an industry is strongly correlated with the movement of top managers acrossfirms in that industry.

32Because these noncompete clauses become operational only when top managers leave their priorfirms, the enforceability of these noncompete clauses can be thought of as a measure of the frictionfacing top managers when they attempt to join the current firm.

33Bishara, Martin, and Thomas (2015) analyze an extensive sample of CEO employment contractsand show that 80% of these contracts contain noncompete clauses, often with a broad geographicscope. A growing body of work (e.g., Garmaise (2011), Marx et al. (2009)) shows that higher enforce-ability of these noncompete clauses constrains employees’ mobility (including that of managers).

34One possible argument against our IV satisfying the exclusion restriction is that the noncompeteclauses in a state may affect the movement of inventors to the sample firm (as well as that of topmanagers), thereby affecting the extent of innovation by the firm. This argument, however, is in factnot valid because our IV is not driven by the enforceability of noncompete clauses in a state alone.The number of acquisitions in the sample firm’s industry and state is unlikely to affect the movementof inventors across firms. We have empirically verified that the movement of inventors across firms isnot correlated with our IV.

35Garmaise (2011) develops an index to measure the enforceability of noncompete clauses byconsidering 12 questions analyzed by Malsberger (2004), which is the central resource describingnoncompetition law in the 50 U.S. states and the District of Columbia, and assigning 1 point to eachjurisdiction for each question if the jurisdiction’s enforcement of that dimension of noncompetitionlaw exceeds a certain threshold, with possible totals therefore ranging from 0 to 12. Higher valuesof Garmaise’s (2011) index indicate higher enforceability of the noncompete agreements in this stateand thus less mobility of the managers from this state. For example, Garmaise’s index (2011) is equalto 0 for California and is equal to 9 for Florida after 1997. The ENFORCE INDEX used here isconstructed as the difference between 12 and the value of Garmaise’s (2011) index scaled by 12, andthus it potentially ranges from 0 to 1.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




an instrument for top management quality in a firm in industry j in year t . We ag-gregate our instrumental variable across all 50 U.S. states to take into account thefact that the market for top managers is likely to be a nationwide market. Further,given that noncompete clauses have a broad geographic scope (i.e., they applyeven outside the state of the managers’ original employer), our IV is likely to bea good proxy for the supply of top managers available to a sample firm from allover the United States. We expect our instrument to be positively correlated withour top management quality factor (MQF).

To instrument for the management quality (MQF) of firm i in industry j inyear t , we therefore run the following first-stage regression:

MQFi ,t = α+βINSTRUMENT j ,t + δACQ j ,t−5(7)+γ Z i ,t + IND FE+YR FE+ST FE+ εi ,t .

In both the first and second stages of our IV regressions, we include the totalnumber of acquisitions in the sample firm’s industry 5 years prior (ACQ j ,t−5) tocontrol for the effect of an industry-wide shock (e.g., merger waves) on innova-tion. We also include fixed effects for the state in which the firm is headquarteredto alleviate the concern that the relation between top management quality and in-novation may be driven by other state-level factors. We expect the instrument tobe positively and significantly related to our MQF, thus satisfying the relevancecondition required for a valid instrument. However, even though we control for thedirect effect of merger waves on innovation, it may be argued that the requirementthat the acquisitions in the industry in previous years be correlated with futureinnovation only through the supply of managers may not hold. In this latter sce-nario, the exclusion restriction for a valid instrument will not be satisfied. Giventhis, the results of our IV analysis should be viewed as suggestive and interpretedwith caution.

Column 1 of Table 13 reports the results of the first stage of our IV analysis.The coefficient of the instrument is positive (as predicted) and is statistically sig-nificant at the 1% level. The first-stage F-statistic is 40.09, which is significant atthe 1% level. These findings confirm that the relevance condition for the instru-ment is satisfied. Columns 2–4 report the second-stage results of our IV (2-stageleast squares (2SLS)) regressions using 1-, 2-, and 3-year-ahead patent countsas our dependent variables.36 In summary, even after controlling for the poten-tial endogeneity between MQF and innovation using our IV analysis, our MQFstill has a positive and significant impact on firms’ patent counts, total numberof citations, and citations per patent in all the specifications. We also conduct IVanalyses for innovative efficiency as well as for our innovation strategy measuresusing the same instrumental variable as described here. In untabulated results, wefind that the coefficients of MQF in these 2-stage regressions are also positive andsignificant.

36We report our second-stage results using the total number of citations and the number of citationsper patent as dependent variables, respectively, in Table IA-6 of our Supplementary Material andobtain similar results.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




TABLE 13The Effect of Management Quality on Corporate Innovation: Instrumental Variable Analysis

Table 13 reports the instrumental variable (IV) regression results of corporate innovation on the management quality factor(MQF). The instrument is described in Section VI. Column 1 reports the first-stage regression result, regressing MQF onthe instrument and other controls. Columns 2–4 report the second-stage regression results using the number of patentsfiled by a firm in a given year as the dependent variable. ln(PATENT) is the natural logarithm of 1 plus the truncation-adjusted number of patents filed by a firm in a given year. ACQt−5 is the total number of acquisitions in the sample firm’sindustry 5 years prior. Other control variables are defined as in Table 3. Constants; year fixed effects, 2-digit SIC industryfixed effects, and state fixed effects are included in all regressions. All standard errors are adjusted for clustering at thefirm level and are reported in parentheses below the coefficient estimates. *, **, and *** indicate statistical significanceat the 10%, 5%, and 1% levels, respectively.

MQF ln(PATENT)t+1 ln(PATENT)t+2 ln(PATENT)t+3Variable 1 2 3 4

INSTRUMENT 0.011***(0.002)

MQF 0.610*** 0.362*** 0.239(0.149) (0.139) (0.161)

ACQt−5 −0.013*** 0.000 0.001** 0.002**(0.002) (0.000) (0.000) (0.001)

ln(ASSETS) 0.069*** 0.111*** 0.125*** 0.128***(0.008) (0.012) (0.010) (0.011)

MB 0.167*** 0.031 0.076*** 0.097***(0.017) (0.028) (0.026) (0.031)

ROA −0.195*** 0.079** 0.045 0.021(0.032) (0.037) (0.033) (0.035)

CAPEX/ASSETS −0.729*** 0.325** 0.155 0.089(0.095) (0.136) (0.120) (0.124)

R&D/ASSETS 0.213*** 0.052 0.082 0.072(0.058) (0.059) (0.051) (0.051)

RET −0.049*** −0.005 −0.013 −0.010(0.007) (0.009) (0.009) (0.009)

AVG_TENURE −0.036*** 0.020*** 0.011** 0.006(0.003) (0.006) (0.005) (0.006)

HHI −0.100 0.230 0.194 0.155(0.274) (0.240) (0.220) (0.197)

No. of obs. 25,945 25,945 23,096 20,119Adj. R 2 0.107

Industry, year, and state fixed effects Yes Yes Yes Yes

VII. ConclusionIn this paper, we use panel data on top management characteristics and a

management quality factor constructed using common factor analysis on indi-vidual measures of management quality (management team size, the fraction ofmanagers with MBA degrees, the fraction of managers with Ph.D. degrees, thefraction of members with prior work experience on the top management team,the average number of prior board positions that each manager has held, and theaverage employment- and education-based connections of each manager in themanagement team) to analyze the relation between firm top management qualityand corporate innovation input and output. We show that top management qual-ity is an important determinant of corporate innovation, with different individualaspects of management quality affecting innovation in younger and older firmsdifferently. We further show that firms with higher top management quality en-gage in more risky (“explorative”) innovation strategies. Finally, we show thathiring more and higher-quality inventors is an important channel through whichfirms with higher top management quality achieve greater innovation output.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




Supplementary MaterialSupplementary Material for this article is available on the authors’ Web sites:

https://www2.bc.edu/thomas-chemmanur/.

ReferencesAcemoglu, D., and D. Autor. “Lectures in Labor Economics.” Available at http://economics.mit.edu/

files/4689 (2011).Adams, R. B.; H. Almeida; and D. Ferreira. “Powerful CEOs and Their Impact on Corporate Perfor-

mance.” Review of Financial Studies, 18 (2005), 1403–1432.Aghion, P.; J. Van Reenen; and L. Zingales. “Innovation and Institutional Ownership.” American Eco-

nomic Review, 103 (2013), 277–304.Atanassov, J. “Do Hostile Takeovers Stifle Innovation? Evidence from Antitakeover Legislation and

Corporate Patenting.” Journal of Finance, 68 (2013), 1097–1131.Balsmeier, B.; L. Fleming; and G. Manso. “Independent Boards and Innovation.” Journal of Financial

Economics, 123 (2017), 536–557.Becker, G. S. Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Edu-

cation. Chicago, IL: University of Chicago Press (1964).Ben-Porath, Y. “The Production of Human Capital and the Life Cycle of Earnings.” Journal of Political

Economy, 75 (1967), 352–365.Bernstein, S. “Does Going Public Affect Innovation?” Journal of Finance, 70 (2015), 1365–1403.Bertrand, M., and A. Schoar. “Managing with Style: The Effect of Managers on Firm Policies.” Quar-

terly Journal of Economics, 118 (2003), 1169–1208.Bishara, N. D.; K. J. Martin; and R. S. Thomas. “An Empirical Analysis of Noncompetition Clauses

and Other Restrictive Postemployment Covenants.” Vanderbilt Law Review, 68 (2015), 1–51.Bloom, N., and J. Van Reenen. “Measuring and Explaining Management Practices across Firms and

Countries.” Quarterly Journal of Economics, 122 (2007), 1351–1408.Brav, A.; W. Jiang; S. Ma; and X. Tian. “How Does Hedge Fund Activism Reshape Corporate Inno-

vation?” Journal of Financial Economics, 130 (2018), 237–264.Chemmanur, T. J., and Y. Jiao. “Dual Class IPOs: A Theoretical Analysis.” Journal of Banking and

Finance, 36 (2012), 305–319.Chemmanur, T. J.; L. Kong; K. Krishnan; and Q. Yu. “Human Capital, Management Quality, and Firm

Performance.” Working Paper, Boston College (2017).Chemmanur, T. J.; E. Loutskina; and X. Tian. “Corporate Venture Capital, Value Creation, and Inno-

vation.” Review of Financial Studies, 27 (2014), 2434–2473.Chemmanur, T. J., and I. Paeglis. “Management Quality, Certification, and Initial Public Offerings.”

Journal of Financial Economics, 76 (2005), 331–368.Chemmanur, T. J.; I. Paeglis; and K. Simonyan. “Management Quality, Financial and Investment Poli-

cies, and Asymmetric Information.” Journal of Financial and Quantitative Analysis, 44 (2009),1045–1079.

Chemmanur, T. J.; I. Paeglis; and K. Simonyan. “Management Quality and Antitakeover Provisions.”Journal of Law and Economics, 54 (2011), 651–692.

Chemmanur, T. J., and X. Tian. “Do Antitakeover Provisions Spur Corporate Innovation? A Re-gression Discontinuity Analysis.” Journal of Financial and Quantitative Analysis, 53 (2018),1163–1194.

Cohen, L.; A. Frazzini; and C. Malloy. “The Small World of Investing: Board Connections and MutualFund Returns.” Journal of Political Economy, 116 (2008), 951–979.

Custodio, C.; M. A. Ferreira; and P. P. Matos. “Do General Managerial Skills Spur Innovation?” Man-agement Science, 65 (2019), 459–476.

Eberhart, A. C.; W. F. Maxwell; and A. R. Siddique. “An Examination of Long-Term Abnormal StockReturns and Operating Performance Following R&D Increases.” Journal of Finance, 59 (2004),623–650.

Ederer, F., and G. Manso. “Is Pay for Performance Detrimental to Innovation?” Management Science,59 (2013), 1496–1513.

Engelberg, J.; P. Gao; and C. A. Parsons. “The Price of a CEO’s Rolodex.” Review of Financial Studies,26 (2013), 79–114.

Ewens, M., and M. Marx. “Founder Replacement and Startup Performance.” Review of FinancialStudies, 31 (2018), 1532–1565.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam







































http://economics.mit.edu/files/4689



































https://doi.org/10.1017/S0022109018001497




Ewens, M., and M. Rhodes-Kropf. “Is a VC Partnership Greater Than the Sum of Its Partners?” Jour-nal of Finance, 70 (2015), 1081–1113.

Faleye, O.; R. Hoitash; and U. Hoitash. “The Costs of Intense Board Monitoring.” Journal of FinancialEconomics, 101 (2011), 160–181.

Ferreira, D.; G. Manso; and A. C. Silva. “Incentives to Innovate and the Decision to Go Public orPrivate.” Review of Financial Studies, 27 (2014), 256–300.

Finkelstein, S., and D. C. Hambrick. Top Executives and Their Effects on Organizations. Minneapolis/St. Paul, MN: West (1996).

Gardner, H. “Frame of Mind: The Theory of Multiple Intelligences.” New York, NY: Basic Books(1983).

Garmaise, M. J. “Ties That Truly Bind: Noncompetition Agreements, Executive Compensation, andFirm Investment.” Journal of Law, Economics, and Organization, 27 (2011), 376–425.

Griliches, Z. “Patent Statistics as Economic Indicators: A Survey.” Journal of Economic Literature,28 (1990), 1661–1707.

Griliches, Z.; A. Pakes; and B. H. Hall. “The Value of Patents as Indicators of Inventive Activity.”NBER Working Paper No. 2083 (1988).

Gu, F. “Innovation, Future Earnings, and Market Efficiency.” Journal of Accounting, Auditing andFinance, 20 (2005), 385–418.

Hall, B. H.; A. B. Jaffe; and M. Trajtenberg. “The NBER Patent Citation Data File: Lessons, Insightsand Methodological Tools.” NBER Working Paper No. 8498 (2001).

Hirshleifer, D.; P.-H. Hsu; and D. Li. “Innovative Efficiency and Stock Returns.” Journal of FinancialEconomics, 107 (2013), 632–654.

Hirshleifer, D.; A. Low; and S. H. Teoh. “Are Overconfident CEOs Better Innovators?” Journal ofFinance, 67 (2012), 1457–1498.

Hirshleifer, D., and A. V. Thakor. “Managerial Conservatism, Project Choice, and Debt.” Review ofFinancial Studies, 5 (1992), 437–470.

Kaplan, S. N.; M. M. Klebanov; and M. Sorensen. “Which CEO Characteristics and Abilities Matter?”Journal of Finance, 67 (2012), 973–1007.

Kogan, L.; D. Papanikolaou; A. Seru; and N. Stoffman. “Technological Innovation, Resource Alloca-tion, and Growth.” Quarterly Journal of Economics, 132 (2017), 665–712.

Lanjouw, J. O., and M. Schankerman. “Patent Quality and Research Productivity: Measuring Innova-tion with Multiple Indicators.” Economic Journal, 114 (2004), 441–465.

Lerner, J. “The Importance of Patent Scope: An Empirical Analysis.” RAND Journal of Economics,25 (1994), 319–333.

Lerner, J.; M. Sorensen; and P. Stromberg. “Private Equity and Long-Run Investment: The Case ofInnovation.” Journal of Finance, 66 (2011), 445–477.

Lerner, J., and J. Wulf. “Innovation and Incentives: Evidence from Corporate R&D.” Review of Eco-nomics and Statistics, 89 (2007), 634–644.

Li, G.-C.; R. Lai; A. D’Amour; D. M. Doolin; Y. Sun; V. I. Torvik; A. Z. Yu; and L. Fleming. “Dis-ambiguation and Co-Authorship Networks of the U.S. Patent Inventor Database (1975–2010).”Research Policy, 43 (2014), 941–955.

Malmendier, U., and G. Tate. “CEO Overconfidence and Corporate Investment.” Journal of Finance,60 (2005), 2661–2700.

Malsberger, B. M. “Covenants Not to Compete: A State-by-State Survey.” Washington, DC: BNA(2004).

Manso, G. “Motivating Innovation.” Journal of Finance, 66 (2011), 1823–1860.Marx, M.; D. Strumsky; and L. Fleming. “Mobility, Skills, and the Michigan Non-Compete Experi-

ment.” Management Science, 55 (2009), 875–889.Rajan, R. G., and L. Zingales. “Financial Dependence and Growth.” American Economic Review, 88

(1998), 559–586.Sah, R. K., and J. E. Stiglitz. “The Architecture of Economic Systems: Hierarchies and Polyarchies.”

American Economic Review, 76 (1986), 716–727.Sah, R. K., and J. E. Stiglitz. “The Quality of Managers in Centralized versus Decentralized Organi-

zations.” Quarterly Journal of Economics, 106 (1991), 289–295.Sapra, H.; A. Subramanian; and K. V. Subramanian. “Corporate Governance and Innovation: Theory

and Evidence.” Journal of Financial and Quantitative Analysis, 49 (2014), 957–1003.Seru, A. “Firm Boundaries Matter: Evidence from Conglomerates and R&D Activity.” Journal of

Financial Economics, 111 (2014), 381–405.Sørensen, J. B., and T. E. Stuart. “Aging, Obsolescence, and Organizational Innovation.” Administra-

tive Science Quarterly, 45 (2000), 81–112.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497




Spiegel, M. I., and H. Tookes. “Identifying an IPO’s Impact on Rival Firms.” Working Paper, YaleUniversity (2016).

Stein, J. C. “Takeover Threats and Managerial Myopia.” Journal of Political Economy, 96 (1988),61–80.

Stein, J. C. “Efficient Capital Markets, Inefficient Firms: A Model of Myopic Corporate Behavior.”Quarterly Journal of Economics, 104 (1989), 655–669.

Tian, X., and T. Y. Wang. “Tolerance for Failure and Corporate Innovation.” Review of FinancialStudies, 27 (2014), 211–255.

https://doi.org/10.1017/S0022109018001497D

ownloaded from

https://ww

w.cam



w.cam


https://doi.org/10.1017/S0022109018001497



Top Management Human Capital, Inventor Mobility, and … Management Human Capital... · States...

Documents

Transcript of Top Management Human Capital, Inventor Mobility, and … Management Human Capital... · States...