Innovation and evidence
United States
Federal instruments and state-level experience need to be examined separately. A programme's result does not sum up a country.
Can early support unlock an innovation?
A technology without a prototype may struggle to attract capital. The challenge is to support additional development, rather than simply fund what would happen anyway.
Sabrina T. Howell's study, published in 2017, examined ranked applications for research and development grants from the Department of Energy. The evaluation is quasi-experimental and concerns a specific sample of the Small Business Innovation Research programme, known as SBIR. [1]
Select a technology proposal
Fund the initial development stage
Examine subsequent investment and outcomes
What the evidence shows
In the sample studied, early support approximately doubled the probability of subsequently receiving venture capital (risk capital), with positive effects on patents and revenue. The author identifies prototype funding, rather than just a quality signal, as a relevant mechanism. [1]
Independent evaluation
How can we tell whether the support made a difference?
- Design
Howell uses the ranking of applications to compare firms just above and just below the funding threshold in a regression discontinuity design. As the number of awards was determined by the budget after ranking, firms close to the threshold form a local quasi-experiment: they had similar scores, but only one side received Phase I funding. [1][2]
- Sample
The dataset covers 7,436 small high-technology firms and more than US$ 884 million in awards between 1983 and 2013. Causal identification focuses on applications to the offices of Energy Efficiency and Renewable Energy and of Fossil Energy; the estimated effect applies chiefly to firms close to the threshold, rather than every possible applicant. [1][2]
- Observed effect
For Phase I grants of around US$ 150 thousand, the probability of receiving venture capital over the long term rose by approximately 9 percentage points, from about 10% to 19%. The study also found positive effects on patents and revenue, with larger effects among firms facing greater financial constraints. [1]
- Scope of validity
The National Academies considers this design among the strongest causal evidence available for the DOE, but stresses its local nature: it does not cover firms that did not apply or those far above or below the threshold. The programme itself has objectives beyond business growth, including meeting the DOE's R&D needs; Phase I results also do not justify assuming the same effect for Phase II or for any subsidy. [2][1]
Implication for OpenGov3: a strong evaluation does not turn a programme into a universal rule. It sharpens the question: for which group, at which decision margin and against which alternative does additional support actually change the outcome?
Costs and delivery
Public expenditure and project selection are involved. The result studied does not provide a universal fiscal return or a subsidy amount suited to Brazil. This is a historical case, not information about current calls or eligibility. [1]
Limits of the comparison
The observed effect does not validate every subsidy or selected firm. Adaptation would need to examine additionality, selection, competition and what would happen without the support.
How can policies be compared before they are expanded?
A list of well-intentioned programmes does not show which alternative delivers more consistent results within the budget.
The Washington State Institute for Public Policy (WSIPP) brings studies together, estimates costs and benefits for Washington and examines the risks in those estimates. The process distinguishes evidence on outcomes from its conversion into monetary values. [3]
Bring together high-quality studies
Estimate local benefits and costs
Examine uncertainty and update the comparison
What the evidence shows
The documentation publishes methods, components and estimates that can be revised. Values are specific to Washington and may change with data and assumptions. This does not show that using the tool alone causes better government. [3]
Costs and delivery
The comparison includes programme costs and benefits for taxpayers and other people. Total benefit is not the same as savings available in the government's budget. The page does not aggregate the institute's operating costs. [3]
Limits of the comparison
Not every value fits into a monetary calculation. Rights, distribution, delivery capacity and unmeasured effects must accompany the comparison, without importing dollar-denominated results as though they were Brazilian.
OpenGov3's interpretation
What it can teach Brazil
Before expanding support, define the problem, the alternative and the additional outcome expected. Then distinguish benefits to society from revenue or savings the government can actually realise. A useful evaluation must allow the preferred instrument to be reconsidered.
What to monitor when adapting it
- Additional investment, not just announced funding
- Outcomes for supported firms and the comparison group
- Costs per participant and distributional effects
- Uncertainty, updates and review criteria
These are criteria for analysis, not commitments already made or estimated outcomes for Brazil. A well-documented practice is not, by itself, causal proof of an entire country's success.
Sources consulted
Consulted on 10/21/2026. Each result retains its observed period; institutional descriptions and impact evaluations have different scopes.
- Sabrina T. Howell: Financing Innovation, 2017 (opens in a new tab)
American Economic Review. Quasi-experimental evaluation of Department of Energy SBIR Phase I grants, with a local comparison around the funding threshold.
- National Academies: Review of the SBIR and STTR Programs at the Department of Energy, 2020 (opens in a new tab)
Independent consensus report. Uses Howell as an important causal reference and makes explicit the limits to generalisation, competing programme objectives and outcomes beyond commercialisation.
- WSIPP: Benefit-Cost Results (opens in a new tab)
Method for synthesising evidence, costs, benefits and risk, with estimates specific to Washington.