What Changes After the Corporate Cheque Arrives
This chapter follows start-ups after their first corporate venture capital investment and asks what, if anything, changes relative to comparable companies financed by other VC. The useful judgment is not a single “innovation up” or “innovation down” label. It is whether the investment is associated with a different mix of new patenting, subsequent use of technology, and visibility—and how confidently the evidence supports that interpretation.
The design tries to isolate the investment event
The baseline comparison from the previous chapter showed that CVC-backed start-ups are not randomly selected. They are more likely to have patents, larger patent portfolios, and stronger funding signals before the deal. To examine what happens afterward, the OECD study uses a difference-in-differences event-study design. The basic idea is simple: compare the change in outcomes for start-ups receiving CVC with the change for similar start-ups receiving other VC, before and after their first funding event.
The details of the comparison determine what the result means. For the CVC group, the analysis focuses on start-ups that receive CVC in their first funding event. It excludes companies that receive CVC only later, because a later corporate investment would arrive after a different development history. The comparison group consists of start-ups whose first funding event is other VC, and the two groups are aligned by the year of that first funding event. A company receiving its first cheque in 2010 is compared with companies entering funded development around the same time, rather than with firms funded in a different market cycle.
This arrangement does not erase selection. It addresses one especially important form of it: the possibility that CVC-backed and other VC-backed start-ups were already on different innovation trajectories before their funding event. The identifying assumption is that, without the difference between CVC and other VC, the groups would have followed parallel paths. That assumption can never be observed directly, but the study tests whether the groups show different trends before treatment.
The reported placebo tests do not find significant pre-treatment differences in the event-study coefficients. The p-value is 0.81 for patent applications and 0.93 for patent citations. These results are consistent with the parallel-trends assumption in the periods the study can observe. They do not prove that every unmeasured difference has disappeared. They make the timing of the post-funding divergence more informative: the pattern begins after the investment rather than simply reflecting a long-running gap that was already widening.
The study also uses an estimator designed for staggered treatment timing and heterogeneous effects. CVC investments do not all happen in the same year, and a corporate investor may affect one start-up differently from another. A method that treats every investment as if it occurred simultaneously, or assumes that the effect is identical across periods, can combine different situations into a misleading average. The chosen approach is intended to handle those differences more appropriately.
Even with these safeguards, the result should be described as an estimated effect under an observational design, not as a laboratory experiment. The comparison is stronger than a raw before-and-after correlation because it uses a relevant control group, funding cohorts, and pre-trend checks. It still depends on the quality of the comparison and on factors that the data do not measure. That is why the magnitude and meaning of each outcome need to be examined separately.
A simple way to see the logic is to imagine two lines on a chart. One line tracks patenting by start-ups whose first funding includes CVC; the other tracks patenting by start-ups whose first funding comes from other VC. The relevant quantity is not the height of either line at the funding date. It is the change in the first line after funding minus the change in the second line over the comparable period. A pre-existing level difference can remain; the design is asking whether the lines bend differently after the event.
The same-year restriction is important because innovation and finance move with the wider economy. A start-up funded during a technology boom may have more opportunities, more patenting incentives, and more available capital than one entering a recession. Aligning first funding years makes the comparison less vulnerable to that calendar effect. It does not control for every market shock, but it avoids treating a change in the financing environment as if it were caused by the identity of one investor.
One investment, two output patterns
The headline result is a divergence. Relative to other VC-financed start-ups, CVC-financed start-ups show a negative average post-treatment effect on log patent applications and a positive average post-treatment effect on log patent citations. The reported coefficients are -0.033 for applications and 0.203 for citations, both statistically different from zero in the study’s main table.
The use of log-transformed outcomes is practical because patenting and citations are unevenly distributed: a small number of firms can have very large counts. It also means the coefficients should not be casually restated as exact percentage changes in every start-up’s patents or citations. They are average effects on the transformed measures over the post-treatment periods in the model. The safe interpretation is directional and comparative: filing intensity is lower for the CVC group relative to the other-VC group, while citation flow is higher.
The application result means that CVC-backed start-ups file fewer patents relative to the comparison group after the funding event. It does not mean that every start-up stops patenting, that patenting becomes unimportant, or that all other forms of innovation decline. The report offers a useful check on the most extreme interpretation. Among start-ups that had filed at least one patent in the three years before funding, 31% of the CVC-backed group filed no patent afterward, compared with 39% of the other VC-backed group. The CVC group’s relative reduction therefore appears to come from filing fewer patents, not from a larger share abandoning patenting altogether.
That distinction matters because a count of filings measures the intensity of one observable activity. A start-up may file fewer patents because it concentrates resources on a smaller number of projects, shifts effort toward product development, uses other forms of protection, or spends more time commercializing an existing technology. Those possibilities are not equivalent, and the patent count alone cannot choose among them. The result is evidence of a change in patenting intensity, not a complete diagnosis of the company’s innovation process.
Citations move in the other direction. CVC-financed start-ups receive more citations on their patents relative to other VC-financed start-ups. A citation is not a direct measure of revenue, product adoption, or social value, but it can indicate that later patenting activity is making use of an earlier invention. The contrast with the filing result therefore suggests a change in composition: fewer new filings alongside more subsequent recognition or use of at least some of the start-up’s technology.
The most informative part of the citation result concerns patents filed before the CVC investment. Citations to those earlier patents rise after the funding event. The CVC investor could not have improved the technical quality of an invention that had already been filed, so the increase cannot be explained simply as better post-investment patent production. The authors treat faster diffusion, greater visibility, or use inside an investor’s technological ecosystem as plausible explanations.
That is a stronger clue about mechanism than a citation increase limited to new patents would be. If the corporate partner gives the start-up access to distribution, customers, technical collaborators, or a platform, the start-up’s existing ideas may be encountered by more organizations. The investment can then increase the reach of technology without increasing the number of inventions produced each year. The start-up is not necessarily doing less valuable work; it may be moving from broad exploration toward adoption and integration.
The clue is not a proof. Citation growth to patents filed after investment also rises, although its interpretation is less clear. Those new patents could be better because the start-up gains useful complementary assets. They could be cited more because the corporate ecosystem makes them visible. Or both processes could occur at once. The OECD analysis therefore calls diffusion a suggestive interpretation rather than a fully identified causal mechanism.
The pre-investment patent result also prevents a narrow “corporate coaching” explanation. If an earlier invention receives more subsequent citations, the post-investment relationship may be changing who encounters the invention rather than improving the invention itself. That distinction is commercially important. A corporation can create value by connecting a technology to customers, suppliers, developers, or other firms, even when it does not make the underlying idea technically better. But a wider audience is not the same as a broader set of independent research projects. Diffusion and exploration can be complements in some relationships and substitutes in others.
The two paths can coexist because invention and diffusion require different resources. A start-up in an early exploratory phase may benefit from money that supports many experiments. A start-up with a promising technology may benefit more from a corporation that helps the technology travel through a market, supply chain, or technical ecosystem. The first situation might produce more filings; the second might produce fewer filings but more use. Neither is automatically the better outcome without knowing the start-up’s objective and the opportunity cost of the shift.
The measurement boundary is part of the finding
It is tempting to translate the result into a simple story: corporate investors suppress innovation, or corporate investors improve quality. The evidence supports neither universal statement. It identifies a difference between two measured dimensions after funding. Patent applications capture the creation of new patent families, while citations capture later references to the start-up’s patent stock. The first is closer to an output count; the second combines influence, visibility, and technological use. They are related but not interchangeable.
Nor do they cover all innovation. A start-up can improve a product without patenting the improvement, learn from customers, develop a process, build organizational capability, or keep a technology secret. Conversely, citations can reflect legal or strategic patenting practices as well as genuine technical importance. The OECD study uses patent data because they provide a consistent cross-country measure, but it acknowledges that patents capture only part of innovation. A higher citation count should therefore be read as evidence of greater citation impact or likely diffusion, not as a complete measure of economic or social benefit.
This boundary changes what a founder or investor should ask when reading the result. If the goal is radical technical exploration, the relevant follow-up measures might include the breadth of the research pipeline, the number of independent projects, or the ability to keep pursuing technologies outside the parent’s immediate needs. If the goal is to get a promising invention adopted, citation reach and commercial access may be closer to the intended outcome. The OECD data can illuminate the patenting and citation parts of that decision; they cannot decide the goal for the parties.
It also changes what absence means. A decline in patent filings is evidence that fewer patents were filed relative to the comparison group. It is not evidence that no useful product work occurred. Likewise, a rise in citations is evidence of more references in the observed patent data. It is not evidence that every user benefited or that the start-up’s total social value increased. Keeping those statements bounded is not a technicality: it prevents a measurable proxy from silently becoming the definition of innovation.
The timing evidence does make one interpretation less plausible: the filing decline is not simply the continuation of an obvious pre-existing divergence, because the pre-treatment placebo tests do not show significant differences in trends. But timing alone does not reveal the mechanism. It cannot tell us whether start-ups reallocated engineers to sales, adapted their research to a platform, received better complementary knowledge, encountered a change in patenting strategy, or faced a different competitive environment after funding. Those explanations require additional evidence.
This is also why the finding should not be read as a verdict on every CVC programme. The main comparison pools relationships across investors, sectors, countries, and years. The average can hide important differences. A corporate partner with distribution power may affect diffusion differently from one whose main contribution is technical expertise. A start-up in a digital platform ecosystem may face different incentives from one developing a physical technology. The overall result tells us that the modern CVC relationship is not associated with a uniform increase in patenting; it does not tell us that every individual relationship follows the average path.
The most careful conclusion is therefore two-part. After CVC funding, the start-ups in this analysis file fewer patents relative to comparable other-VC start-ups, while their patents receive more citations. The pattern is consistent with a shift from innovation intensity toward impact, visibility, or diffusion, but the study does not establish which mechanism dominates or whether aggregate innovation rises. Selection explains why the CVC group looked stronger before the deal; the event study suggests that the investment is associated with a different trajectory afterward. Both facts must be held together before the relationship can be judged.