Checkpoint: The Post-Deal Innovation Ledger
This checkpoint connects three ideas already established: innovative targets and acquirors are selected into acquisitions, patent measures are only windows onto innovation, and a credible counterfactual is needed to isolate a deal effect. The judgment supported here is narrower than a verdict about motive: after a completed acquisition, the target’s measured innovation falls substantially, and the acquiror does not show a matching patenting increase in the OECD evidence. The remaining question is why that pattern appears.
The target-side ledger has two kinds of loss
The target result is not one number. It has an extensive margin—whether the acquired start-up continues to file at all—and an intensive margin—how much and what kind of work it produces if it continues. Keeping those margins separate makes the post-deal change easier to understand and prevents a large fall in aggregate filings from hiding the experience of the minority that remains active.
The descriptive comparison is stark. Excluding the acquisition year, acquired start-ups average 2.4 patents in the period before acquisition and 1.1 afterward. Their citation-weighted patent measure falls from 93.5 to 8.5. Among the start-ups that file at some point, the mean number of patents falls from 5.3 before acquisition to 2.3 after, while the citation-weighted measure falls from 205.5 to 18.3. These figures describe accumulated periods, not a single year’s causal effect, and their observation counts change because acquisitions near the end of the data do not have a full post-deal window. They nonetheless show the shape of the problem: most visible target patenting is concentrated before control changes hands.
The table is useful as a ledger entry, but it cannot by itself tell us what the acquisition changed. It adds each firm’s activity across a pre-deal period and a post-deal period, so it mixes transactions with different dates and different amounts of time available for observation. A start-up acquired near the beginning of the window can contribute several post-deal years; one acquired near the end may contribute none. The change in the number of firms observed in the two panels is therefore part of the measurement story, not evidence that the missing firms had zero innovation. The table also reports a different object from the event study: it describes the average activity of the acquired group, whereas the event study measures how the gap between treated and comparison firms changes relative to the year before acquisition. The two pieces of evidence point in the same direction, but their numbers should not be subtracted or interpreted as interchangeable estimates.
The extensive margin explains much of that shape. Approximately half of acquired start-ups filed patents before acquisition, while approximately 10% continue filing afterward. A target that stops filing is different from one that files fewer patents, and both are different from one that continues filing at the same rate but shifts its work into less visible forms. The post-deal outcome must therefore be described as a reduction in observed patent activity, with a large component coming from firms that cease patenting in the observed data.
The two margins also change how the average should be interpreted. A fall in the number of active patenters can pull down the average flow even if the firms that remain active are unaffected.