«John R. Graham, Co-Supervisor David T. Robinson, Co-Supervisor Manuel Adelino Alon P. Brav Manju Puri Aaron K. Chatterji Dissertation submitted in ...»
The CVC life cycle ends with the termination stage, when CVC parents stop making incremental investment in startups, typically when internal innovation begins to recover. The median duration of the life cycle is about four years. When CVC divisions last more than four years, ﬁrms typically hibernate CVC activities during 7 Interestingly, CVC investment appears to have a “reverse home bias”—even though CVCs are less likely to invest in geographically distant companies, they are also less likely to invest in companies in their own geographic regions, from which they may acquire information through local innovation spillover (Peri, 2005; Matray, 2014).
6 years when internal innovation remains productive. This evidence is consistent with the information acquisition rationale, which predicts decreased CVC activity when the marginal beneﬁt shrinks after information is assimilated into parent ﬁrms.
Interestingly, if innovation again deteriorates at the parent ﬁrm, the CVC life cycle begins anew.
All told, this paper presents the CVC life cycle in the course of examining CVC’s role of acquiring information in the process of innovation. Essentially, CVC serves as a transitory information-acquiring step in regaining an upward innovation trajectory, typically after a ﬁrm experiences a deterioration in internal innovation.
Centered around this overview, Section 2 discusses how to understand the information acquisition rationale of CVC from various aspects and its implications on corporate and entrepreneurial ﬁnance. Section 3 describes how the data are constructed. Section 4 through 6 cover each stage of the CVC life cycle. Section 7 concludes.
7 2 Discussion and Literature
The life-cycle pattern lends support to the information acquisition view of Corporate Venture Capital. How does this rationale help us understand existing evidence on CVC activities? What trade-oﬀs do entrepreneurs face when accepting CVC investment and sharing their knowledge? How does CVC and its role of acquiring information ﬁt into the innovation process? What other forces might shape CVC behaviors? This section discusses the implications of CVC and its information acquisition rationale on corporate and entrepreneurial ﬁnance in light of these questions.
2.1 Reconciling Existing CVC Evidence
An emerging literature, at the intersection of economics, ﬁnance, and strategy, attempts to understand the function and inﬂuence of CVC (Dushnitsky, 2006; Maula, 2007; Lerner, 2012; Chemmanur et al., 2014; Chemmanur and Fulghieri, 2014). This paper adds to the literature in three ways. First, it asks an under-explored question on CVC and ﬁnancing innovation. It deviates from the existing framework that takes 8 CVC as given1 by seeking the economic rationale behind CVC investment and its role in the innovation process. The empirical analysis in this paper complements earlier work surveying corporate venture capitalists about their motivation.
Second, in the course of seeking the CVC rationale, this paper makes two empirical contributions to the literature. Unlike previous approaches, which typcally focus on a single phase in CVC activities, I characterize the full CVC life-cycle dynamics from initiation through operation to termination. In addition, by collecting a larger and longer sample of CVCs accompanied by detailed innovation, investment, and entrepreneurship records, I can better control for the inﬂuence of anecdotal superstar CVC cases (such as Google Venture and Intel Capital), speciﬁc industries, and speciﬁc time periods, thus oﬀering many ﬁndings that could be masked.
Third, by studying the information acquisition step of corporate innovation through the lens of CVC,2 this paper helps to explain CVCs’ active participation in the due diligence of startups (Henderson and Leleux, 2002), obtaining (sometimes non-voting) board seats (Maula et al., 2001; Bottazzi, Da Rin, and Hellmann, 2004), and creating communication platforms for inventors in both the parent ﬁrm and entrepreneurial ventures (Dushnitsky and Lenox, 2005b). This information acquisition rationale is also consistent with the ﬂexibility and lower adjustment cost (Lerner, 2012) of CVC, which enables ﬁrms to respond quickly, to change course easily (abandon a project with lower sunk costs), and to leverage outside funding sources.3 1 See, e.g., Siegel, Siegel, and MacMillan (1988); Gompers and Lerner (2000); Bottazzi, Da Rin, and Hellmann (2004); Dushnitsky and Lenox (2006); Benson and Ziedonis (2010); Chemmanur, Loutskina, and Tian (2014); Dimitrova (2013); Ceccagnoli, Higgins, and Kang (2015); Wadhwa, Phelps, and Kotha (2015).
2 Closer to this paper, Dushnitsky and Lenox (2005a) and Basu et al. (2011) indirectly support the information acquisition rationale of CVC by studying environmental variables aﬀecting the eﬃciency of information acquisition with analyses tilted toward industry-level factors.
3 Indeed, this learning process often does not involve later asset consolidation (Dimitrova, 2013).
92.2 Entrepreneurs’ Costs and Beneﬁts
The life-cycle CVC investment pattern and the information acquisition rationale lean heavily on the willingness of entrepreneurs to channel their knowledge to CVC investors. But what do entrepreneurs gain from the relation with CVC? Early research ﬁnds that CVC investors can better nurture innovative entrepreneurs by providing technical support and by tolerating riskier projects. Consistent with this argument, Gompers and Lerner (2000) and Chemmanur, Loutskina, and Tian (2014) show that entrepreneurial companies backed up by CVCs have a higher possibility of making a successful exit and become more innovative after the investment. Second, building relations with established corporations through CVC investment increases the potential for future business. Ceccagnoli et al. (2015) show that an ex ante CVC relation increases the possibility of ex post technology licensing. Last but not least, the acceptance of CVC equity investment aligns the interests of the startup with the CVC investor, working as insurance against competitive behavior from incumbent CVC parent ﬁrms (Mathews, 2006).
Admittedly, the eﬃciency of information acquisition factors in how entrepreneurs trade oﬀ the costs and beneﬁts of receiving CVC investment. Hellmann (2002) and Hellmann (1998) discuss this problem and analyze why entrepreneurs might prefer CVC over alternative entrepreneurial investment—a more formal test of the model will rely on detailed startup ﬁnancing data, which is beyond the scope of this paper.
2.3 CVC and the Innovation Process
This paper highlights the information acquisition step in the innovation process (Nelson, 1982), complementing existing studies that typically assume that the information structure is predetermined (Aghion and Tirole, 1994; Robinson, 2008; Bena and Li, 2014; Seru, 2014). This is in the same spirit as the recent endeavor of incorporating 10 information acquisition in ﬁnancial economics, which studies how economic agents search, process, and use information to guide information-sensitive decisions.4 The framework also creates the ability not only to study CVC alone but also to explicitly identify the process of integrating CVC-acquired knowledge into R&D and acquisition decisions. Ideally, these analyses can be viewed as stepping stones toward understanding the whole system of ﬁnancing and organizing innovation, in which diﬀerent organizational structures interact with each other in the dynamic innovation process.
2.4 Alternative CVC Rationales
A nice feature of this life-cycle ﬁnding is that we can analyze several alternative CVC rationales by extending the framework. Early research on CVC and strategic investment proposed many economic forces that could shape CVC activities (Allen and Phillips, 2000; Hellmann, 2002; Mathews, 2006; Fee et al., 2006; Fulghieri and Sevilir, 2009). Results suggest that such factors as corporate governance, ﬁnancial constraint, and purely outsourcing innovation seem to be secondary to the intention to acquire valuable technological knowledge from startups. Exhausting all potential economic forces that could aﬀect CVC, though deﬁnitely of interest, is beyond the scope of this paper. However, the extensions provided give an overview of those forces surrounding the central theme of information acquisition.
4 See Dow and Gorton (1997); Chen, Goldstein, and Jiang (2007); Van Nieuwerburgh and Veldkamp (2010); Bond, Edmans, and Goldstein (2012); Yang (2013).
I exploit a hand-collected sample of Corporate Venture Capital units aﬃliated with US-based public ﬁrms. I start with a list of CVCs identiﬁed by the VentureXpert Venture Capital Firms database (accessed through Thomson Reuters SDC Platinum), which is standard in VC studies (Chemmanur, Loutskina, and Tian, 2014). Each CVC on the list is manually matched to its unique corporate parent in Compustat by checking multiple sources (Factiva, Google, etc.). I remove VC divisions operated by ﬁnancial ﬁrms, which diﬀer from CVC arms of industrial ﬁrms (Hellmann, Lindsey, and Puri, 2008). From VentureXpert I obtain the investment history of each CVC, including basic information about the startup companies it invests in and the timing and features of each CVC deal.
12 Table 3.1: This table provides descriptive statistics on Corporate Venture Capital activities by year (Panel A) and by industry (Panel B). CVCs are identiﬁed from the VentureXpert Venture Capital Firm Database, accessed through Thomson Reuters SDC Platinum, and are hand-matched to their unique corporate parent ﬁrms. CVC parent ﬁrms in the sample are US-based public non-ﬁnancial ﬁrms. Panel A reports the annual number of CVC initiations and investment (deals) between 1980 and 2006. Panel B reports the industry distribution of CVC activities, where the industries are deﬁned by the Fama-French 48 Industry Classiﬁcation.
Panel A: CVC Activities by Year
14 Steel Works Etc. 3 15 Transportation 3 9 Machinery 5 15 Wholesale 10 87 Electrical Equipment 9 44 Retail 14 79 Automobiles and Trucks 6 42 Restaurants, Hotels, Motels 4 13 Aircraft 2 7 The main sample consists of 381 CVC ﬁrms initiated between 1980 and 2006.1 Table 3.1 summarizes this CVC sample by tabulating the time-series dynamic and the industry composition. Panel A presents the number of CVC division initiations and investment deals by year. CVC activities are heavily concentrated in the ﬁrst half of the 1980s and the second half of the 1990s. This is consistent with existing studies on “CVC waves” (Gompers and Lerner, 2000; Dushnitsky, 2006) and is revisited and reﬁned in Section 4.3. Panel B summarizes the industry distribution of CVC parent ﬁrms, where industries are deﬁned by the Fama-French 48 Industry Classiﬁcation.
The Business Services industry (including IT) was the most active sector in CVC investment, with 90 ﬁrms investing in 821 venture companies. Electronic Equipment ﬁrms initiated 46 CVC divisions that invested in 921 companies. Pharmaceutical ﬁrms launched 28 CVCs and invested in 254 deals. Other active sectors include Computers and Communications.
The CVC sample is augmented with Compustat for ﬁnancial statement data and with CRSP for stock market performance. Variable constructions are described in the Appendix. All data items are pre-winsorized at the 1% and 99% levels. SDC Platinum provides organizational information on M&As and strategic alliances. For corporate governance data, I extract institutional shareholding information from the WRDS Thomson Reuters 13(f) data and obtain G-index data from Andrew Metrick’s data library.2 Innovation is a crucial data component of this paper for three reasons. First, because innovation knowledge generated in the entrepreneurial sector has the potential to create great value for CVC parent ﬁrms (Scherer, 1965; Acs and Audretsch, 1988;
Kortum and Lerner, 2000; Macmillan et al., 2008), it is an important part of the 1 I focus on CVCs initiated no later than 2006 to allow for investment behaviors to realize (after
2006) and to ensure the quality of the innovation database.
2 Accessed using http://faculty.som.yale.edu/andrewmetrick/data.html.
15 information set that CVC units intend to acquire. Second, comprehensive innovation data create a valuable setting to measure informational relationships (Bena and Li, 2014) and knowledge ﬂows (Gomes-Casseres et al., 2006; Gonzalez-Uribe, 2013).
Third, the quality of detailed innovation data maintained and updated by the United States Patent and Trademark Oﬃce (USPTO) is superior to most alternative data sources on corporate activities.
I obtain basic innovation data from the NBER Patent Data Project and from Bhaven Sampat’s patent and citation data.3 The combined database provides detailed patent-level records on more than 3 million patents granted by USPTO between 1976 and 2012. I link this database to Compustat using the bridge ﬁle provided by NBER. Beyond the standard database, I also introduce several data sets and cleaning procedures that are relatively new to the literature (detailed in related sections and the Appendix): I link the USPTO database to entrepreneurial companies in VentureXpert using a fuzzy matching method based on company name, basic identity information, and innovation proﬁles, similar to Gonzalez-Uribe (2013) and Bernstein, Giroud, and Townsend (2014); I also introduce the Harvard Business School inventor-level database in order to examine how ﬁrms adjust their innovative human capital as a speciﬁc channel to facilitate information acquisition and integration; and last, I introduce the Google Patent Assignment and Reassignment database, which tracks all transactions of each patent.
The combined innovation data provide three layers of innovation information that are helpful for the analysis. First, I employ two main variables to measure basic corporate innovation performance. I measure innovation quantity by calculating the number of patent applications, which are eventually granted, ﬁled by a ﬁrm in 3 For more information on the NBER Patent Data Project, please refer to Hall, Jaﬀe, and Trajtenberg (2001). The data used in this paper were downloaded from https://sites.google.
com/site/patentdataproject/. Sampat’s data can be accessed using http://thedata.harvard.