Levels of Evidence in Grant Funding

Levels of evidence rank how strongly research supports an intervention. Federal education regulations define four: strong, moderate, promising, and demonstrates a rationale. Each names a study design and the population match it requires. All four count as evidence-based unless a notice says otherwise.

Current figures — verified 2026-08-11

Item	Value	Source
Single-study sample for strong or moderate evidence	at least 350 students or other individuals across more than one site	34 CFR 77.1
Intervention-report route to moderate evidence	“positive effect” with at least 20 individuals from more than one site, or “potentially positive effect” with at least 350	34 CFR 77.1
Handbook versions named in the evidence definitions	versions 2.1, 3.0, 4.0, 4.1, and 5.0	34 CFR 77.1
Current What Works Clearinghouse standards	Procedures and Standards Handbook, Version 5.0 (August 2022, revised December 2022)	WWC Handbooks
Evaluation budget rule of thumb	10 percent of total program cost	National Institute of Justice

These figures change. Verify against the linked source before relying on them. Report an outdated figure

Key takeaways

• Evidence tiers describe research about an intervention, not your organization.
• The tier attaches to a named project component, not the whole project.
• A quasi-experimental design cannot reach the strong tier.
• The lowest tier still requires cited research behind each component.
• Citing evidence and generating evidence are separate obligations with separate costs.

What does “evidence-based” mean in a grant application?

“Evidence-based” means a proposed project component is supported by research at one of several defined levels — not that the applicant has a track record, and not that the idea sounds sensible. The Department of Education’s regulations define the term as a component “supported by one or more of strong evidence, moderate evidence, promising evidence, or evidence that demonstrates a rationale” (34 CFR 77.1).

Two features of that definition do most of the work. First, all four levels qualify as evidence-based, so a notice that simply says “evidence-based” without naming a tier is setting a low bar. Second, the unit is the project component, defined in the same regulation as “an activity, strategy, intervention, process, product, practice, or policy included in a project.” A proposal earns a tier for a specific named practice, and a notice usually requires that practice to be central to the design.

Levels of evidence sit inside a broader federal shift toward evidence use. The Foundations for Evidence-Based Policymaking Act (Pub. L. 115-435) requires agencies to publish learning agendas and evaluation plans, and the government-wide evaluation standards in OMB M-20-12 name relevance and utility, rigor, independence and objectivity, transparency, and ethics. Within evidence, evaluation, and data, tiers are the applicant-facing edge of that system.

Other agencies use different words for the same idea. Health, labor, justice, and national service programs commonly run three-level schemes — often preliminary, moderate, and strong. Tier language from one agency does not transfer to another without checking that agency’s own definitions.

What are the four levels of evidence?

The four levels of evidence are strong, moderate, promising, and demonstrates a rationale, in descending order of rigor. Moving down the ladder relaxes three things in sequence: the required study design, the sample size and number of sites, and how closely the studied population and setting must match the one you propose to serve.

The table below states what each level of evidence requires and what a proposal must put on the page to claim it. All definitions come from 34 CFR 77.1; sample thresholds appear in the current figures above.

Level of evidence	Study design that satisfies it	Population and setting match	What the proposal must show
Strong evidence	Experimental study meeting standards without reservations	Sample overlaps populations and settings	Design, effect direction, significance, sites, sample
Moderate evidence	Experimental or quasi-experimental design meeting standards with or without reservations	Sample overlaps populations or settings	Same, plus multi-site sample
Promising evidence	Single experimental, quasi-experimental, or correlational study with controls for selection bias	No overlap requirement	Favorable significant effect, no overriding negatives
Demonstrates a rationale	No study of the intervention required	Not applicable	Logic model plus cited research behind each component

Two structural facts follow from the table. A single strong study must be multi-site, which is why many well-regarded single-school or single-clinic studies top out at promising. And each level requires the absence of “overriding statistically significant and negative effects” — a study with a favorable headline result and a significant harm elsewhere does not qualify.

What study design does each level of evidence require?

Study design sets a ceiling on the level of evidence a citation can reach, and the ceiling is written into the regulation. Randomized controlled trials, regression discontinuity designs, and single-case designs can meet What Works Clearinghouse standards without reservations. A quasi-experimental design, in the regulation’s words, “can meet WWC standards with reservations, but cannot meet WWC standards without reservations.”

That single sentence produces the most common tier error in proposals. A matched-comparison study, however carefully executed, cannot support a claim of strong evidence. It can support moderate evidence if the sample is multi-site and large enough.

Underneath the tiers sits a review apparatus. The What Works Clearinghouse reviews education studies against a published handbook and rates each finding as meeting standards without reservations, meeting standards with reservations, or not meeting standards. The handbook governs technical questions that decide ratings — attrition in randomized trials, and baseline equivalence between groups in quasi-experimental designs (WWC Handbooks).

Four designs cover most citations a grant applicant will make:

• Randomized controlled trial — assignment by lottery, which balances observed and unobserved characteristics. Reaches the highest level of evidence.
• Regression discontinuity design — assignment by a cutoff on a measured variable, with that variable controlled in the analysis. Also reaches the highest level.
• Quasi-experimental design — a comparison group matched on observed characteristics. Capped at moderate.
• Correlational study with statistical controls for selection bias — regression adjustment on covariates. Explicitly named in the regulation as sufficient for promising evidence, and the cheapest route onto the ladder.

How does tiered-evidence grantmaking work?

Tiered-evidence grantmaking sizes the award to the strength of the evidence behind the applicant’s approach. The Government Accountability Office describes the model plainly: it “allows federal agencies to award smaller amounts of grant funding to test promising ideas, and larger amounts to replicate practices with strong evidence of success” (GAO-16-818).

The canonical architecture has three funding tiers. A development tier makes small awards to applicants with a rationale or preliminary findings, and expects the grant to produce promising or moderate evidence. A validation tier makes mid-size awards to applicants with moderate evidence, and expects confirmation. A scale-up tier makes the largest awards to applicants with strong evidence, and expects evidence that the effect holds at scale.

The venture-capital logic is deliberate: many small bets on untested ideas, a few large bets on proven ones. It also means the evaluation obligation runs inversely to intuition. A development-tier grantee has the weakest evidence going in and often the heaviest evidence-building requirement coming out, because generating a defensible study is the point of the award.

For applicants, tiered-evidence grantmaking changes the reading of a funding notice. The tier you can honestly claim determines which competition you are eligible for, how large an award you can request, and what kind of study you must fund. Working that out belongs early in reading a Notice of Funding Opportunity, not after the budget is built.

What is an evidence clearinghouse and how do you use one?

An evidence clearinghouse is a federally maintained repository that reviews studies against published standards and publishes ratings, so an applicant can cite a vetted finding rather than argue methodology from scratch. Department of Education guidance directs applicants to “explore the broadest possible range of relevant evidence” and names several (Using Evidence to Strengthen Education Investments).

Five federal clearinghouses cover most program areas:

• What Works Clearinghouse — education interventions, with intervention reports, practice guides, and single-study reviews.
• Clearinghouse for Labor Evaluation and Research — labor, workforce, and employment research, maintained by the Department of Labor’s Chief Evaluation Office.
• CrimeSolutions — criminal justice, juvenile justice, and victim services programs and practices.
• Evidence-Based Practices Resource Center — behavioral health, substance use, and mental health interventions.
• AmeriCorps Evidence Exchange — research and evaluation reports on national service, civic engagement, and volunteering.

Using a clearinghouse well means citing at the level of specificity the regulation demands. Naming the intervention is not enough. State the design, the sample size, the number of sites, the direction and statistical significance of the effect, and how the studied population and setting compare with yours. An evidence clearinghouse entry supplies most of those elements; the population-and-setting comparison is yours to make and is where reviewers most often find the claim overstated.

What if your intervention has no evidence base?

An intervention with no study behind it can still qualify, at the level called demonstrates a rationale. The regulation defines it as a case where “there is a key project component included in the project’s logic model that is supported by citations of high-quality research or evaluation findings that suggest that the project component is likely to significantly improve relevant outcomes” (34 CFR 77.1).

Read that definition closely and the tier is not a free pass. It requires two artifacts: a logic model containing the component, and citations of research supporting the causal link the model asserts. The research supports the mechanism, not the packaged program. A tutoring model with no evaluation of its own can cite research on tutoring dosage, tutor training, and attendance — provided the logic model actually routes through those links.

Satisfying the rationale tier honestly takes four steps. Name the key project component precisely enough that a reviewer could recognize it in practice. Draw the logic model so each arrow is a claim you can defend. Attach a citation to each arrow that carries weight, using outcome findings rather than advocacy literature or program marketing materials. Then say plainly what the evidence does not establish.

The selection criteria reward the same discipline. The Department of Education’s criteria ask reviewers to weigh “the quality of the logic model or other conceptual framework underlying the proposed project, including how inputs are related to outcomes” (34 CFR 75.210).

How do evidence levels differ from your own evaluation?

Evidence levels describe research that already exists about an intervention. An evaluation is the study you will run about your project. A proposal frequently owes both, and confusing them produces either an unfunded evaluation or an unsupported claim.

The two obligations move in opposite directions. The stronger the evidence you cite, the less the funder needs you to prove the intervention works, and the more the evaluation shifts toward implementation fidelity and replication at scale. The weaker the evidence you cite, the more the funder expects the grant to generate evidence, and the more the study becomes a deliverable with its own design, sample, and analysis plan.

Money follows that logic. The National Institute of Justice publishes a working rule of thumb for evaluation as a share of total program cost (see current figures above), while noting that a process evaluation may take “only a few months” and that “a large-scale outcome evaluation may require years and a substantial financial outlay” (NIJ). Any cap or floor stated in the funding notice overrides the heuristic. The share required by a multi-site experimental design with primary data collection is materially higher, and pricing that design is part of writing an evaluation plan.

Independence is the other axis. Where a notice calls for an independent evaluation, the regulation means one “designed and carried out independently of, but in coordination with, the entities that develop or implement the project component.” A subcontractor who takes design direction from the program team may not satisfy that standard, a point echoed in the federal evaluation policy of the Administration for Children and Families (ACF Evaluation Policy).

What goes wrong with evidence claims in proposals?

Evidence claims fail in predictable ways, and most failures are arithmetic or category errors rather than shortages of research. Reviewers work from the regulatory definitions, so a claim that does not track the definition’s elements reads as unsupported even when the underlying study is good.

Six failure modes account for most lost points:

• Tier arithmetic errors. Claiming strong evidence on a quasi-experimental design, or moderate on a single-site study, contradicts the definitions outright.
• Asserted rather than demonstrated. Naming a study without stating design, sample, sites, effect direction, and significance leaves the reviewer unable to verify the tier.
• Population and setting mismatch ignored. Citing an urban elementary study for a rural secondary program without acknowledging the gap invites a downgrade.
• Rationale tier treated as no evidence. A logic model with uncited arrows, or citations to advocacy pieces and vendor materials, fails the tier it is claiming.
• Evidence unconnected to the design. A cited intervention that never reappears in the activities, budget, or measures reads as decoration.
• Tier language borrowed across agencies. Education tier vocabulary in a health or labor proposal signals that the applicant did not read that agency’s definitions.

One honest limit is worth stating in the proposal itself. A level of evidence is a claim about a body of research, not a forecast about your project. Strong evidence means an intervention worked somewhere, measured carefully, under conditions someone documented. Whether it works in your setting depends on implementation, dosage, staffing, and the population you actually reach — which is what your own measurement and outputs, outcomes, and impact framework exists to find out.

Frequently asked questions

Does “evidence-based” always mean listed in a clearinghouse?

No. Federal education regulations treat all four levels — including one requiring no study of the intervention at all — as evidence-based, and department guidance directs applicants toward the widest relevant range of evidence rather than a single registry. A clearinghouse listing makes a tier easier to substantiate; it is not the only route.

Can a quasi-experimental design ever produce strong evidence?

No. The regulation states that a quasi-experimental design study can meet What Works Clearinghouse standards with reservations but cannot meet them without reservations. Since strong evidence requires a study meeting standards without reservations, a quasi-experimental design caps at moderate evidence regardless of how well it was executed.

Is a pre-post comparison enough to claim an evidence level?

A single-group pre-post comparison has no comparison condition, so it supports none of the levels above rationale. Pre-post change is legitimate as a performance measure and for continuous improvement. Presenting it as evidence of effect invites maturation, selection, and regression-to-the-mean objections.

What counts as high-quality research for the rationale tier?

Research reporting outcome findings from studies with a defensible design — experimental, quasi-experimental, or correlational with controls — carries weight. Descriptive statistics, opinion pieces, conference presentations without methods, and materials produced to market the intervention generally do not, even when they are cited in volume.

Do evidence tiers apply outside education programs?

The four-level structure is defined in Department of Education regulations and applies across its programs. Other agencies run their own schemes, often with three levels and different sample and design requirements. The concept transfers; the definitions do not. Read the definitions section of the notice you are answering.

Who decides whether your evidence claim holds?

Peer reviewers score the claim against the criteria in the notice, and program staff can request substantiation before award. Where the notice ties a tier to competitive preference points, the reviewer is applying the regulatory definition directly, which is why proposals should quote the elements of that definition back in the citation.

Related topics

• Evidence, Evaluation, and Data — the hub covering logic models, evaluation design, measurement, and data
• Writing an Evaluation Plan
• Performance Measurement and Indicators
• How to Build a Logic Model
• Theory of Change vs Logic Model

Sources

1. U.S. Department of Education, 34 CFR 77.1, “Definitions that apply to all Department programs” (eCFR). https://www.ecfr.gov/current/title-34/subtitle-A/part-77/section-77.1 (accessed 2026-08-11)
2. U.S. Department of Education, 34 CFR 75.210, “General selection criteria” (eCFR). https://www.ecfr.gov/current/title-34/section-75.210 (accessed 2026-08-11)
3. U.S. Department of Education, Using Evidence to Strengthen Education Investments, non-regulatory guidance. https://www2.ed.gov/fund/grant/about/discretionary/2023-non-regulatory-guidance-evidence.pdf (accessed 2026-08-11)
4. Institute of Education Sciences, What Works Clearinghouse. https://ies.ed.gov/ncee/wwc/ (accessed 2026-08-11)
5. Institute of Education Sciences, WWC Handbooks and Reviewer Resources. https://ies.ed.gov/ncee/wwc/handbooks (accessed 2026-08-11)
6. U.S. Government Accountability Office, Tiered Evidence Grants: Opportunities Exist to Share Lessons from Early Implementation and Inform Future Federal Efforts, GAO-16-818. https://www.gao.gov/products/gao-16-818 (accessed 2026-08-11)
7. U.S. Congress, H.R.4174, Foundations for Evidence-Based Policymaking Act of 2018, Pub. L. 115-435. https://www.congress.gov/bill/115th-congress/house-bill/4174 (accessed 2026-08-11)
8. Office of Management and Budget, M-20-12, Phase 4 Implementation of the Foundations for Evidence-Based Policymaking Act of 2018. https://www.whitehouse.gov/wp-content/uploads/2020/03/M-20-12.pdf (accessed 2026-08-11)
9. U.S. Department of Labor, Clearinghouse for Labor Evaluation and Research (CLEAR). https://clear.dol.gov/ (accessed 2026-08-11)
10. U.S. Department of Justice, Office of Justice Programs, CrimeSolutions. https://crimesolutions.ojp.gov/ (accessed 2026-08-11)
11. Substance Abuse and Mental Health Services Administration, Evidence-Based Practices Resource Center. https://www.samhsa.gov/resource-search/ebp (accessed 2026-08-11)
12. AmeriCorps, Evidence Exchange. https://americorps.gov/about/our-impact/evidence-exchange (accessed 2026-08-11)
13. National Institute of Justice, “Plan for Program Evaluation from the Start.” https://nij.ojp.gov/topics/articles/plan-program-evaluation-start (accessed 2026-08-11)
14. U.S. Department of Health and Human Services, Administration for Children and Families, ACF Evaluation Policy. https://www.acf.hhs.gov/opre/report/acf-evaluation-policy (accessed 2026-08-11)