«John R. Graham, Co-Supervisor David T. Robinson, Co-Supervisor Manuel Adelino Alon P. Brav Manju Puri Aaron K. Chatterji Dissertation submitted in ...»
The rationale of information acquisition for CVC investment is convincing only if CVC parents can use newly gathered information to improve their operations. Several economic frictions could hinder CVCs from gathering and integrating information from startups, challenging the information acquisition rationale. Hellmann (2002) theoretically shows that entrepreneurs could intentionally avoid CVC investment to protect their innovation. Dushnitsky and Lenox (2005b) and Kim, Gopal, and Hoberg (2013) argue that the absorptive ability (Cohen and Levinthal, 1990) of CVC parent ﬁrms imposes a limit on the knowledge transferred through the relationship. Gompers and Lerner (2000) suggest that the eﬃciency of CVC is constrained by the incentive problem embedded in its organizational and compensation structure. In addition, 46 high adjustment costs of R&D investment (Hall, Griliches, and Hausman, 1986; Lach and Schankerman, 1989) can decrease the speed and intensity of the integration of new knowledge acquired through CVC.
Showing how information is incorporated into corporate decisions can be challenging due to the invisible nature of information. In this subsection, I undertake two empirical settings to study how information acquired through CVC inﬂuences the parent ﬁrm. First, following the literature that uses patent citations as a measure of knowledge spillover (Gomes-Casseres et al., 2006), I study how CVC parent ﬁrms internalize acquired information into organic R&D by tracking patent citations made to their portfolio companies. I then switch to another setting where I look at the eﬃciency of corporate decisions in which the acquired information could be crucial.
5.2.1 Internal Research and Development
I identify the speciﬁc information ﬂow from portfolio companies that is further incorporated into parents’ internal innovative activities. Empirically, I follow the economic literature on knowledge spillover (Jaﬀe and Trajtenberg, 2002),2 and estimate whether CVC parent ﬁrm i makes new citations to startup company j’s patents or
knowledge after the CVC invests in the startup, using the following model:
To control for observed characteristics of CVC parents that could inﬂuence their behaviors in citing entrepreneurial companies, I construct a tighter control group for those ﬁrms. I use a propensity score matching method and match each CVC parent ﬁrm i that launches its CVC unit with two non-CVC ﬁrms from its CVC launch year and 2-digit SIC industry that has the closest propensity score estimated using ﬁrm 2 Alcacer and Gittelman (2006) and Gomes-Casseres et al. (2006), among others, discuss the advantages and potential pitfalls in using this approach.
47 size (the logarithm of total assets), market-to-book ratio, ∆Innovation, and patent stock,3 similar to the sample construction strategy in Bena and Li (2014). The CVC launching year for a CVC parent ﬁrm is also the “pseudo-CVC” year for its matched ﬁrms.
Observations are at the i-j-t level. The full set of i-j pairs then denotes the potential information ﬂow that could happen between a CVC parent ﬁrm (or a matched ﬁrm) and a startup, captured by patent citations. IpCV CP arentq is a dummy variable indicating whether ﬁrm i is a CVC parent or a matched control ﬁrm.
IpP ortf olioq indicates whether company j is in the CVC portfolio of ﬁrm i. For each i-j pair, two observations are constructed, one for the ﬁve-year window before ﬁrm i invests in company j, and one for the ﬁve-year window after the investment.4 IpP ostq indicates whether the observation is within the ﬁve-year post-investment window.
The dependent variable, Citeijt, indicates whether ﬁrm i makes new citations to company j’s innovation knowledge during the corresponding time period.
The key variable of interest, IpCV CP arentq ˆ IpP ostq ˆ IpP ortf olioq, captures the incremental intensity of integrating a portfolio company’s innovation knowledge into organic innovation after a CVC invests in the company. Table 5.2 column (1) shows the regression results. The coeﬃcient of 0.159, means that the citing probability increases by 15.9% after establishing the link through CVC investment.
3 Patent stock is constructed as the total number of patents applied for by the ﬁrm up to year t ´ 1.
4 A matched control ﬁrm is assumed to have the same investment history as the CVC parent ﬁrm to which it is matched to.
48 Table 5.2: Direct Information Acquisition from Portfolio Companies This table studies the direct information acquisition of CVC parent ﬁrms from their portfolio companies by investigating how investing in an entrepreneurial company aﬀects the CVC parent ﬁrm’s possibility of innovating based on the entrepreneurial company’s innovation. I ﬁrst identify all the patents applied by a CVC parent ﬁrm (or a matched control ﬁrm) i, and all the patents cited by those patents. I then identify all the patents applied by an entrepreneurial company j. These data further allow me to determine whether ﬁrm i makes a new citation, which it never cited before, to a patent that is possessed by company j. The analysis is performed based on the
Citeijt “ α ` β ¨ IpCV CP arentq ˆ IpP ostq ˆ IpP ortf olioq ` ΦrIpCV CP arentq, IpP ostq, IpP ortf olioqs ` εijt.
The sample is at the i-j-t level. The full set of i-j pairs then denotes the potential information ﬂow that could happen between a CVC parent ﬁrm (or a matched ﬁrm) and a startup, captured by patent citations. IpCV CP arentq is a dummy variable indicating whether ﬁrm i is a CVC parent or a matched control ﬁrm. IpP ortf olioq indicates whether company j is in the CVC portfolio of ﬁrm i. For each i-j pair, two observations are constructed, one for the ﬁve-year window before ﬁrm i invests in company j, and one for the ﬁve-year window after the investment. IpP ostq indicates whether the observation is within the ﬁve-year post-investment window. The dependent variable, Citeijt, indicates whether ﬁrm i makes new citations to company j’s innovation knowledge during the corresponding time period. The key variable of interest, IpCV CP arentq ˆ IpP ostq ˆ IpP ortf olioq, captures the incremental intensity of integrating a portfolio company’s innovation knowledge into organic innovation after a CVC invests in the company. Column (1) reports the result. Column (3) performs an analysis similar to that in column (1) except that it estimates the probability that a CVC parent ﬁrm cites not only patents owned by the startup but also patents previously cited by the startup. In other words, the potential citation now covers a broader technological area that the startup works in. Columns (2) and (4) separately estimate the intensity of citing knowledge possessed by companies that either exit successfully (acquired or publicly listed) or fail at last. All speciﬁcations include ﬁxed eﬀects imposing analysis across ﬁrms in the same industry and same year of (pseudo-) launching their CVC programs to absorb time-variant industrial technological trends. T-statistics are shown in parentheses and standard errors are clustered by ﬁrm. *, **, *** denote statistical signiﬁcance at the 10%, 5%, and 1% levels, respectively.
Observations 1,406,734 1,406,734 R-squared 0.01 0.02 Yes Yes Industry ˆ CVC Year FE I further explore the depth of information acquisition from portfolio companies.
Speciﬁcally, in column (3), I perform an analysis similar to that in column (1) except that I look at the probability that a CVC parent ﬁrm cites not only patents owned by the startup but also patents previously cited by the startup. In other words, the potential citation now covers the broader technological area that the startup works in. Column (3) extends the message conveyed in column (1)—CVC parent ﬁrms not only cite the portfolio company’s own patents, but also beneﬁt from the knowledge indirectly carried by portfolio companies, reaching to the broader knowledge behind.
Does information acquisition concentrate only on successful investment? I explore this question by modifying model (5.2) and separately estimate the intensity of citing knowledge possessed by companies that either exit successfully (acquired or publicly listed) or fail at last. The result is reported in columns (2) and (4), and it appears that CVC parents acquire knowledge from both successful and failed ventures.
5.2.2 Using Information through External Acquisitions
Here I explore an alternative channel through which ﬁrms could beneﬁt from CVCacquired information—acquiring external innovation. Acquiring innovation has become an important component of corporate innovation (Bena and Li, 2014; Seru, 2014), and identifying promising acquisition targets (companies or innovation) requires a valuable information set, such as great understanding of markets and technological trends. Under the information acquisition hypothesis, CVC-acquired information allows parent ﬁrms to form more precise expectations of acquisition deals, thereby improving eﬃciencies when making such decisions.5 I ﬁrst study how eﬃciently CVC parent ﬁrms conduct acquisitions of companies.
Following the literature, acquisition eﬃciency is measured using three-day, ﬁve-day, 5 Those acquisitions are not necessarily limited to their CVC portfolio companies and can reach a broader domain using the general innovation and industry knowledge they learn from CVC experience.
51 and seven-day cumulative abnormal returns (CAR) of an acquisition deal centered on the acquisition announcement day. The analysis is performed on a cross section of M&A deals conducted by CVCs and their matched control ﬁrms between ﬁve years before and ﬁve years after (pseudo-) CVC initiations, and the unit of observation is an acquisition deal. The key variable of interest is the diﬀerence-in-diﬀerences variable IpCV CP arentqi ˆ IpP ostqi,t, indicating whether the acquirer i is within ﬁve years after launching its CVC division. If ﬁrms could conduct more eﬃcient external acquisitions based on the information gathered from CVC investment, one would expect the abnormal announcement returns to be higher for these deals.
Table 5.3: Integration of CVC-Acquired Information through External Acquisitions This table studies the eﬃciency of acquiring companies or innovation around the start of CVC
investment. 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 acquisition deals (Panel A) and patent purchases (Panel B) conducted by CVCs and their matched control ﬁrms during ﬁve years before CVC initiations and ﬁve years after CVC initiations, and the unit of observation is an acquisition deal (Panel A) and a patent purchase (Panel B). The sample consists of CVCs and their propensity score-matched ﬁrms. The dependent variables yi,t are cumulative abnormal returns (CARs) for acquisition of companies (Panel A) and annual citation growth for purchases of patents (Panel B). IpCV CP arentqi is a dummy variable indicating whether ﬁrm i is a CVC parent 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. The model includes industry-by-year ﬁxed eﬀects αindustryˆt. 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.
Table 5.3 Panel A presents the result.
Columns (1) to (3) examine three-day, ﬁve-day, and seven-day CAR (in basis points, bps), respectively. The positive and signiﬁcant coeﬃcients across all three columns conﬁrm that ﬁrms conduct more successful external acquisitions as they internalize the information acquired through their CVC investment. Quantitatively, compared to their industry peers, acquisitions made by CVC parent ﬁrms experience a 65 bps improvement in the three-day abnormal return from one day before the announcement to one day after the announcement, and a greater than 130 bps increase in abnormal return during the r´3, 3s window.
To study how CVC-acquired information is capitalized through acquisitions of 53 innovation, I compile a detailed data set on ﬁrms’ acquisition of patents (either “company and patents” or “patents only”). The database on patent transactions is based on USPTO patent assignment ﬁles, hosted by Google Patents. This database provides useful information for identifying patent transactions: the assignment date;
the participating parties, including the assignee—the “buyer” in a transaction— and the assignor—the “seller” in a transaction; and comments on the reason for the assignment. To gather additional information on the original assignee and patent technology classes, I merge the raw assignment data with the USPTO patent databases, and with the HBS inventor database. I then follow a procedure, based on Serrano (2010) and Akcigit, Celik, and Greenwood (2013), in which I separate patent transactions from all patent reassignment records, that is, I remove reassignments associated with cases such as a patent transfer from the employee inventor to the employer ﬁrm, or a patent transfer between diﬀerent subsidiaries of a ﬁrm. A more detailed description of the data and methodology is provided in the Appendix.