«John R. Graham, Co-Supervisor David T. Robinson, Co-Supervisor Manuel Adelino Alon P. Brav Manju Puri Aaron K. Chatterji Dissertation submitted in ...»
I perform the analysis on the sample of patent purchases conducted by CVC parent ﬁrms and their control ﬁrms, and the unit of observation is a patent transaction.6 The dependent variable is calculated as the citation growth from the n-year (n “ 1, 2, 3) period before the patent transaction to the same length after the transaction.7 This variable intends to capture whether the purchased patents better ﬁt the buyer than the seller, thereby signaling a more eﬃcient transaction. As in Panel A, the key variable of interest is the diﬀerence-in-diﬀerences term IpCV CP arentqi ˆ IpP ostqi,t, indicating whether the patent buyer i is within ﬁve years after launching its CVC division. If ﬁrms could capitalize the information learned from CVC by conducting more eﬃcient 6 To be clear, some patents are transacted under one “deal,” and I necessarily treat each of them as one individual observation.
7 For example, when n “ 3, ∆Citationr´3, `3s is calculated as total citations received by the transacted patent from one year to three years after the transaction minus total citations received from three years to one year before the transaction.
54 patent purchases, one would expect a positive coeﬃcient to be associated with the diﬀerence-in-diﬀerences term.
In Panel B of Table 5.3, I report the citation growth around patent transactions.
The positive coeﬃcient in column (1), 0.200, means that, after benchmarked by patent transactions conducted by their matched control ﬁrms and pre-CVC transactions, patents purchased by CVC parent ﬁrms receive on average 0.2 more citations during the ﬁrst year under the new owner than the last year under the old owner. Column (2) uses a two-year horizon to calculate citation increases, and the economic magnitude increases to 0.607. Column (3) shows an ampliﬁed result due to a three-year horizon.
It is worth discussing the economic interpretation behind this spike in citations after CVC ﬁrms’ patent transactions. In principle, a spike in citations indicates that the underlying patent becomes increasingly visible and popular, plausibly because it better ﬁts the overall innovation proﬁle of the new owner or is commercialized more successfully after the transaction. Speciﬁcally in this context, such a particularly strong increase in citations is consistent with the interpretation that CVC parent ﬁrms acquire innovation that is in turn better commercialized and made visible to the industry.
5.3 Human Capital Renewal and Information Acquisition
Evidence thus far suggests that CVC parent ﬁrms devote eﬀort to integrating and using information acquired from the entrepreneurial sector. Identifying, processing, and integrating new information is diﬃcult; how do CVC parents accomplish this task? I identify one important channel that CVC parents actively manage: human capital renewal. Indeed, inventors, usually highly educated scientists and engineers, are key in absorbing, processing, and using information to produce innovation. Recent studies also ﬁnd that ﬁrms actively reallocate innovative human resources to spur innovation and adjust the scope of innovation (Lacetera, Cockburn, and Henderson, 55 2004; Bernstein, 2015; Brav, Jiang, Ma, and Tian, 2016). In this section, I explore the role of inventors in facilitating knowledge gathering and use.
I rely on the Harvard Business School patenting database for inventor-level information.8 This database includes unique inventor identiﬁers that are constructed based on a reﬁned disambiguation algorithm employing multiple characteristics (Lai, D’Amour, and Fleming, 2009). After matching inventors to employer ﬁrms, I track the employment history and annual patenting activities of each inventor.9 Using a criterion similar to that in Bernstein (2015) and Brav et al. (2016), I identify the number of inventors who leave the company and the number of inventors who are newly hired in each year.
8 Available at: http://dvn.iq.harvard.edu/dvn/dv/patent.
9 One limitation of this analysis is that we detect inventor mobility conditional on new patent ﬁlings; the observed mobility is thus associated with inventors who patent more frequently. But at any rate, these people should be those who are economically more important to the ﬁrm. See Bernstein (2015) for a detailed discussion of the limitations associated with this database.
56 Table 5.4: Inventor Adjustment and Information Acquisition This table studies the role of inventor adjustment in information acquisition for CVC parent ﬁrms. The Harvard Business School Patent Database provides inventor-level information, which allows me to identify inventor mobility and characteristics of the inventor team for each patent. In Panel A, the analysis is based on the following standard
diﬀerence-in-diﬀerences (DiD) framework:
yi,t “ αF E ` β ¨ IpCV CP arentqi ˆ IpP ostqi,t ` β 1 ¨ IpCV CP arentqi ` β 2 ¨ IpP ostqi,t ` γ ˆ Xi,t ` εi,t.
The sample consists of CVCs and their propensity score-matched ﬁrms. The dependent variables yi,t are the logarithm of inventor leavers (columns (1) and (2)), the logarithm of newly hired inventors (columns (3) and (4)), and the proportion of patents mainly contributed by new inventors (columns (5) and (6)). A patent is considered as mainly contributed by new inventors if as least half of the inventor team are have three or fewer years’ experience in the ﬁrm in the patenting year.
IpCV CP arentqi is a dummy variable indicating whether ﬁrm i is a CVC parent ﬁrm or a matched control ﬁrm. IpP ostqi,t indicates whether the ﬁrm-year observation is within the rt ` 1, t ` 5s window after (pseudo-) CVC initiations. Panel B studies the characteristics of patents produced by new inventors. The sample consists all the innovation produced by CVC parents and matched control ﬁrms from ﬁve years before the event to ﬁve years after the event. INew Inventor’s Pat 57 equals one if new inventors contribute at least half of the patent. New Cite Ratio and Explorativeness are deﬁned in the Appendix. All speciﬁcations include industry-by-year ﬁxed eﬀects αindustryˆt to absorb time-variant industrial technological trends. Firm-level control variables include ROA, size (logarithm of total assets), leverage, and R&D ratio (R&D expenditures scaled by total assets). T-statistics are shown in parentheses and standard errors are clustered by ﬁrm. *, **, *** denote statistical signiﬁcance at the 10%, 5%, and 1% levels, respectively.
Panel A: Inventor Mobility during CVC Operation (1) (2) (3) (4) (5) (6) lnp1 ` Leaversq lnp1 ` N ewHiresq New Inventors’ Pat (%)
Observations 132,407 132,407 43,236 R-squared 0.151 0.124 0.010 Controls Yes Yes Yes Yes Yes – Industry ˆ Year FE I start by examining the intensity of human resource adjustment around the years of initiating CVC investment. The analysis is performed on the same ﬁrm-year panel of CVC ﬁrms and their propensity score-matched controls. In Table 5.4 Panel A, I study the number of inventors leaving the ﬁrm (columns (1) and (2)) and the number of inventors newly hired by the ﬁrm (columns (3) and (4)). The coeﬃcient, 0.119 in column (1), can be interpreted as showing that CVC parent ﬁrms have 11.9% more inventors leaving the ﬁrm (leavers) than in the period before CVC investment. The vacancies created by leavers are ﬁlled by newly hired inventors; the 0.110 estimated in column (3) means that CVC parents hire about 11% more new inventors compared to the years before CVC investment, benchmarked by their industry peers.
In columns (5) and (6), I examine the proportion of patents mainly contributed by inventors new to the ﬁrm. A patent is considered as “mainly contributed by new inventors” if at least half of the patent’s inventor team have three or fewer years of patenting experience in the ﬁrm as of the patent application year. The positive coeﬃcient of 17.1% in column (5) means that CVC parent ﬁrms rely more heavily on new inventors when operating a CVC, consistent with the proposition that ﬁrms hire new inventors to process new information and produce innovation.
Table 5.4 Panel B presents new inventors’ intensity of incorporating new knowledge.
The patent-level sample consists of all the patents produced by CVC parent ﬁrms and their matched control ﬁrms from ﬁve years before the event to ﬁve years after it.
Beyond the standard terms IpCV CP arentsqi and IpP ostqi,t, I introduce an indicator variable INew Inventor’s Pat that equals one if new inventors contribute at least half of the patent and zero otherwise. The unconditional eﬀect of INew Inventor’s Pat is positive, meaning that patents produced by ﬁrms’ new inventors typically incorporate more knowledge new to the ﬁrm. Meanwhile, the interaction term IpCV CP arentsq ˆ IpP ostq is associated with higher New Cite Ratio and Explorativeness, consistent with Table E.7. A key result in this table is the positive coeﬃcient in front of the 60 triple diﬀerence INew Inventors’ Pat ˆ IpCV CP arentq ˆ IpP ostq, which implies that new inventors in CVC parent ﬁrms concentrate more heavily on processing and integrating new information and innovation knowledge. In column (3), I focus on the sample of all patents produced by CVC parent ﬁrms during the ﬁve-year window after CVC initiation (that is, IpCV CP arentsq “ IpP ostq “ 1) and ﬁnd that newly hired inventors are more likely to use knowledge acquired from CVC portfolio companies in their new innovation.
61 6 CVC Terminations: Staying Power and Investment Dynamics In a frictionless world, CVC parents would want to keep investing in CVC to acquire information from entrepreneurs. With frictions, however, CVC could become less appealing as the parent ﬁrm’s internal innovation recovers. For example, a capacityconstrained ﬁrm will allocate less resources to information acquisition yet more to innovation production once the internal innovation becomes more premising (Nelson, 1982; Jovanovic and Rob, 1989). In addition, the cannibalization concern (Arrow, 1962) will disincentivize innovative incumbents to search for newer ideas that will replace the existing ones, and this eﬀect could be particularly large, with high adjustment cost and organizational complexity. Overall, as ﬁrms assimilate information into their innovation decisions and begin to have an upward innovation trajectory, the beneﬁt of keeping a standalone CVC unit shrinks. In this scenario, CVC investment may fade out as internal innovation recovers and ﬁrms devote more resources to this regained innovation path. This section examines this implication of the information acquisition hypothesis by focusing on the termination stage of the CVC life cycle.
62 The analysis provides further opportunities to distinguish the important strategic motivation behind CVC investment. Under alternative CVC rationales, CVC remains advantageous in organizing innovation due to its superior ability to obtain asset complementarity (Hellmann, 2002), motivate entrepreneurs (Aghion and Tirole, 1994; Chemmanur, Loutskina, and Tian, 2014), and obtain competitive advantages (Mathews, 2006; Fulghieri and Sevilir, 2009). Even though these studies focus primarily on static trade-oﬀs and do not concern intertemporal dynamics, they implicitly imply that ﬁrms might invest persistently in CVCs for long periods of time.
6.1 The Staying Power of Corporate Venture Capital
I start by examining the staying power of Corporate Venture Capital. To do so, it is necessary to deﬁne the date of terminating each CVC unit, which is not widely disclosed. When this termination date is not available, I deﬁne it as the date of the CVC’s last investment in a portfolio company. As a result, the staying power analysis could underestimate the duration of CVCs, particularly toward the end of the sample. To mitigate bias, I categorize a CVC unit as “active” if its last investment happened after 2012 (as of March 2015) and VentureXpert codes its investment status as “Actively seeking new investments,” and I exclude those active CVCs from the analysis. The duration of a CVC cycle is calculated as the period between the initiation and termination of the division.
Table 6.1 tabulates the duration of CVC divisions.
The median duration of a CVC is four years, and a signiﬁcant portion (46%) of CVCs actively invest for three years or less,1 lending support to the argument that the beneﬁt from CVC investment shrinks as information is assimilated. However, 27% of ﬁrms operate CVCs for a long period (more than 10 years). To understand why this is so, I report the median number of 1 They certainly could interact with their portfolio companies for longer periods of time after terminating incremental investment.
63 total and longest consecutive years that a CVC is put into hibernation, deﬁned as a year when no incremental investment was made. When the CVC duration is short, the years between initiation and termination are mostly active. As their duration increases, an increasing proportion of years are under hibernation. When I examine these hibernation periods, I ﬁnd a pattern of consecutive hibernating years—for example, CVCs with eight-year durations have a median of four years of consecutive hibernation. In other words, these CVCs typically have a lengthy pause in their CVC experience, bridging two shorter active periods of investment.
64 Table 6.1: This table documents the staying power of Corporate Venture Capital by summarizing the durations of CVCs and investment characteristics sorted by duration. When the date of CVC termination is not available, I deﬁne it as the date of last CVC investment on portfolio companies. I categorize a CVC as “active” if its last investment happened after 2012 (as of March 2015) and VentureXpert categorizes the CVC’s investment status as “Actively seeking new investments.” Duration is calculated as the period between the initiation and termination of CVC investment. Hibernation (Hiber ) is calculated as the number of years that are between CVC initiation and termination yet without any investment in entrepreneurial companies.
Consecutive hibernation years are calculated as years of the CVC’s longest consecutive hibernation. An investment deal is deﬁned as a “success” if the entrepreneurial company was acquired or went public (I exclude cases when the company has neither gone public or been acquired but is still alive).