The Evidence Check: Selection Is Not Treatment
The evidence on corporate venture capital now gives us two facts that are easy to confuse. CVC-backed start-ups look stronger before investment, and after investment they show fewer patent filings relative to other VC-backed start-ups but more patent citations. This checkpoint connects those facts into a usable way of judging a CVC innovation claim: separate who was selected, what changed after funding, which dimension of innovation changed, and how far the evidence permits an explanation.
A stronger claim has three layers
Start with the statement people often make: “CVC-backed start-ups innovate more.” It sounds like one empirical claim, but it contains at least three different claims.
The first is a selection claim. Before the corporate investment, CVC-backed start-ups in the OECD sample were more likely to have patents, larger patent portfolios, and stronger funding signals than start-ups financed by other VC. Thirty-one percent of CVC-financed start-ups filed at least one patent, compared with 15% of other VC-financed start-ups. That difference tells us about the companies that corporate investors chose. It is evidence that CVC is attracted to visible technological promise, not evidence that CVC created the promise.
The second is a treatment or trajectory claim. After the first funding event, the OECD event study compares CVC-backed start-ups with other VC-backed start-ups whose first funding occurred in the same year. The reported post-investment pattern is negative for log patent applications and positive for log patent citations. The pre-treatment placebo tests do not show significant differences in the measured trends, which makes the timing of the divergence more informative. The result is stronger than a simple comparison of two groups, but it remains an estimate from an observational design with assumptions and unmeasured factors.
The third is an interpretation claim. Why would filings fall while citations rise? The report treats faster technology diffusion, greater visibility, or use within a corporate ecosystem as plausible explanations. Those are reasonable mechanisms because citations to patents filed before the investment also increase. But they remain interpretations of the pattern, not a complete causal demonstration. The data do not decide whether a start-up redirected people from exploration to commercialization, received better complementary knowledge, changed its patenting strategy, or became more visible to organizations that could use its existing technology.
Keeping these layers separate prevents two opposite errors. The first error gives CVC credit for every advantage already present in the start-up. The second treats the filing decline as proof that corporate investment destroys innovation. The evidence supports a more precise statement: CVC selects technologically stronger start-ups, and the selected start-ups later show a different mix of measured patenting and citation outcomes relative to other VC-backed firms.
The layers can coexist without cancelling one another. A start-up can be selected because it already has promising technology, receive useful resources from a corporate investor, and still file fewer patents afterward because its priorities change. Alternatively, the investment can fail to improve its independent research while increasing the reach of its existing inventions. Selection is not a reason to dismiss a post-investment effect; it is a reason to estimate that effect against a credible comparison rather than attributing the whole observed level to CVC.
This is also the point at which evidence strength needs to be stated precisely. The pre-deal patent and funding differences are descriptive and controlled associations. The post-deal filing and citation differences come from a more demanding event-study design with pre-trend checks. The diffusion and ecosystem-alignment accounts are mechanism interpretations. A careful argument does not put all three on the same evidentiary footing.