Variables

In the following section, we will explain which variables we use for our analysis. The variables below are deemed to be good measures or proxies of the theoretical concepts. We will argue for the aptness of the variables employed by bridging them with the theory, which was set out in section 4.

The variables are summarized in Table 2 (inspiration drawn from the configuration of Quintana-García & Benavides-Velasco, 2008).

5.2.1. Dependent variable

Internal vs. external CVC unit

Our dependent variable is a dummy variable, where 0 signifies an internal CVC unit and 1 signifies an external CVC unit (i.e. a wholly-owned subsidiary, as previously described). This way of

operationalizing this specific variable has first been introduced as a control variable by Dushnitsky and Shaver (2009) and was employed by recent studies in the field (e.g. Lee et al., 2018).

The variable is called subsidiary in the analysis.

Investment Data

idinvestor idorganization

year_inv_max

firmname idinvestor year_inv_max num_investments_tot equity_est_firmname_tot comp_age_avg_mean

num_coinvestors_round_mean num_corpinv_round_mean same_sic_proportion_mean same_nation_proportion_mean

Transcode Table Patent Data

firmname idinvestor firmnation year_inv_min year_inv_max organization idorganization sic_4

organization_source subsidiary

idorganization year_inv_max cum_patents cum_fc cum_fsc cum_bc cum_bsc

cum_distinct_upsc sd_tot_uspc

5.2.2. Independent variables

Value of innovations

To measure the value of innovations, we use the cumulative number of forward citations assigned to the patents of the parent company at the maximum year of investment. Forward citations is an often-used proxy for value of innovations since the number of times other patents cite a patent is a good observable and quantifiable metric for how important the patent is to firms, and hence, how valuable the patent is (Hall et al., 2005). Specifically, we employ the total number of forward citations assigned to the entire granted patent stock of the parent company up until the end of the investment period. We use the granted patents (instead of applied-for patents) since the measure is forward-looking, and since there is no certainty that applied-for patents will be granted (we have that information as the database only includes applied-for patents that are later granted).

The variable is called cum_fc_g in the analysis.

Firm specificity

To measure firm specificity, we employ the share of backward self-citations in total backward citations of the patents of the parent company at the maximum year of investment. Self-citations reflect “the cumulative nature of innovation”, i.e. how much a company relies on its own prior innovations when developing new innovations (Hall et al., 2005, p. 32). According to Hall et al.

(2005), self-citations offer an insight into how much of the knowledge spillovers the firm can internalize (rather than spilling over to other companies), i.e. how specific the patents are to that company. Specifically, we employ a share measure (as also seen in e.g. Hall et al., 2005), which is the total number of backward self-citations divided by the total number of backward citations. As Hall et al. (2005) also point out, it is necessary to control for the size of the patent portfolio when using this variable, as the share of self-citations might increase when the portfolio is larger. The reason is that the larger the portfolio, the larger the chance that the company will cite one of its own patents (as there are more patents to cite from, i.e. a “mechanical” effect) but not necessarily

because they are firm-specific. Still, the share of backwards self-citations is deemed a meaningful measure of firm specificity. For the self-citations, we have used applied-for patent data, as the variable is backward-looking, and thus the applied-for patents (which were at a later point granted) offer the best possible representation of the available data. The applied-for patents will also be used

for the other patent-related measures (except forward citations, as discussed above) for the same reason.

This variable is called share_bsc_app_cum in the analysis.

Absorptive capacity

To measure absorptive capacity, we employ the company’s patent stock. According to Hall, Jaffe and Trajtenberg (2001), patents proxy knowledge capital and the success of the innovation process and, consequently, the company’s ability to absorb new knowledge. Patent stock is widely used as a proxy for absorptive capacity (Cohen & Levinthal, 1990; Dushnitsky & Lenox, 2006; Hall et al., 2001), which supports its aptness. Specifically, we use the cumulative number of applied-for patents (that are later granted) in the maximum year of investment.

This variable is called cum_patents_app in the analysis.

Technological diversifications

To measure technological diversification, we use the standard deviation of the cumulative number of patents per USPC main class of patents assigned to a given company at the end of the investment period. To put it more plainly, this measure displays the dispersion of patents in different

technology classes. Some other papers use a simply count of technology classes (see e.g. Breschi et al., 2003). However, this simplified approach does not take the magnitude of patents in each

technology class into account, which is why the approach undertaken here adds nuance: it gives a measure of the magnitude and spread in different technology classes of the patents of a given company. The cumulative number of distinct USPC classes of applied-for (and later granted) patents (cum_dist_uspc_app) in the maximum year of investment is instead employed as a control variable.

This variable is called sd_tot_uspc_app in the analysis.

5.2.3. Control variables

We have employed a number of control variables in order to account for other factors influencing our dependent variable. Specifically, we have employed the following variables: a) the cumulative

number of distinct USPC classes in the maximum year of investment as mentioned above, b) the total number of investments performed by the CVC unit, c) an equity estimate of the investments performed by the CVC unit, d) a proportion of the target investment companies that have the same industry (SIC) code as the parent company, e) a proportion of the target investment companies that are from the same nation as the parent company, f) the average age of the target investment

companies, g) the average number of co-investors in any given round of investment, h) the average number of corporate co-investors (other CVC investors) per investment round.

Table 2: Summary of variables

Variables Type Measurement method Source of the

data subsidiary Dependent Dummy variable, 0 signifies internal, 1 signifies

external

Manual search by authors cum_fc_g Independent Cumulative number of forward citations of the

company’s granted patents in the maximum year of investment

PatentsView

share_bsc_app_cum Independent The share of backward self-citations to total number of backward citations of the company’s applied-for (later granted) patents in the maximum year of investment

PatentsView

cum_patents_app Independent The cumulative number of applied-for (later granted) patents of the company in the maximum year of investment

PatentsView

sd_tot_uspc_app Independent The standard deviation of the dispersion of all applied-for (later granted) patents in cumulative distinct USPC main classes of the company in the maximum year of investment

PatentsView

cum_dist_uspc_app Control The cumulative number of distinct USPC classes of total applied-for (later granted) patents in the maximum year of investment

PatentsView

num_investments_tot Control The total number of investments performed by the CVC unit

Thomson One equity_est_firmname_tot Control The estimated total equity investment of the CVC

unit in USD million

Thomson One same_sic_proportion_mean Control The proportion of the target investment

companies with the same SIC code as the parent company

Thomson One

same_nation_proportion_mean Control The proportion of the target investment companies that are from the same nation as the parent company

Thomson One

comp_age_avg_mean Control The average age of target investment companies Thomson One num_coinvestors_round_mean Control The average number of co-investors per round of

investment, in which the unit participates

Thomson One num_corpinv_round_mean Control The average number of corporate co-investors per

round of investment, in which the unit participates

Thomson One