From Curve to Policy: How to Calibrate Support Without Overpromising

For PIPE and similar SME-oriented R&D programs, the policy problem is not simply deciding whether to spend more. It is deciding whether a proposed amount fits the project and the firm that must use it. This chapter turns the study’s evidence into a conditional policy model: move beyond uniform awards, treat the 59% peak as a local hypothesis rather than a quota, and evaluate support over a long enough horizon and broad enough set of outcomes to learn whether the money is doing what policymakers value.

Why one amount for every firm is an incomplete rule

Uniform funding has an understandable administrative appeal. It is easy to explain, easy to budget, and apparently neutral. But the PIPE evidence points to a problem with treating equal amounts as equal treatment. Firms differ in financial constraints, technical capabilities, organizational readiness, strategic orientation, sector, region, and ability to use external knowledge. Projects also differ in the complementary inputs they require. The same nominal grant can be enough to remove a bottleneck for one applicant and insufficient—or difficult to deploy—for another.

The policy implication is not that program administrators should replace one rigid rule with a new rigid rule. It is that funding level should be considered in relation to a firm-project bundle. The project determines what resources are required; the firm determines how quickly and effectively those resources can be absorbed; the local context shapes what skills, equipment, knowledge networks, and follow-on opportunities are available. A grant amount detached from those conditions is an incomplete description of the intervention.

Consider two hypothetical PIPE applicants. The first has a research team in place and needs a specialized instrument to complete a project. The second has a promising proposal but must recruit expertise and build routines for managing research before the same instrument would be useful. An equal award could unlock the first project and leave the second project with a different bottleneck. Conversely, a larger award could help the second applicant build capacity over time, even if its patent response within three years is weaker. These examples are illustrations of the decision problem, not separate findings from the study. They show why headcount alone cannot be a sufficient allocation rule.

The study’s results support this conditional framing. Restricting the analysis to manufacturing and knowledge-intensive business services produces similar inverted-U patterns with larger magnitudes, while the pooled analysis controls for industry, region, and firm size. The differences indicate that the response is not independent of context. They do not provide a fully estimated best dose for every sector or firm-size category. A larger response in a selected group is a reason to investigate its project and capability conditions, not a reason to assign that group a new universal percentage.

Headcount is particularly easy to overinterpret. The paper includes firm-size categories in its model, but a size coefficient is not the same as a separate dose–response curve for each size class.