How many distinct NIH-funded projects in NIH RePORTER will have a “Project Start Date” from 28 June 2026 through 11 August 2026, inclusive?

• My most likely prediction for the number of distinct NIH-funded projects with a start date from 28 June 2026 through 11 August 2026 is 2,700.00, with a 50% chance of falling between 1,500.00 and 4,100.00.
• The forecast reflects a severe administrative and budgetary slowdown at the NIH, where FY 2026 spending was 30% lower than historical averages as of late April 2026.
• Systemic delays caused by new political vetting processes and AI-based screening have resulted in a significant backlog, with 45% of researchers reporting delays in grant start dates.
• The total volume of funding opportunities has collapsed, with targeted notices of funding (NOFOs) dropping from 756 in 2024 to just 14 by mid-2026.
• A 24% year-over-year decline in funded research projects in 2025 sets a lower baseline for expected new project starts in the summer of 2026.
• While July 1 is a major standard start date for NIH grants, the observation date of August 12 provides very little time for late-July and August award records to be updated in the RePORTER database.

Key figures

Historical context

Historically, the NIH RePORTER database tracks tens of thousands of projects annually, with the NIH budget growing from $32.3 billion in FY 2016 to $48.9 billion in FY 2024. Typical award cycles (Cycle I) see applications submitted in February, reviewed in May or June, with start dates clustered around July 1 and September. Between 2021 and 2024, success rates for grant applications hovered between 16% and 20%. However, the 2025–2026 period has seen unprecedented deviations from these patterns. In early 2025, OMB memo M-25-13 initiated a temporary pause on federal grants, and although many were reinstated, roughly 2,600 grants totaling $1.4 billion remained suspended as of January 2026. This fiscal environment has led to historic lows in success rates (13% overall, with some paylines as low as 4%) and a significant shift away from the consistent, predictable funding cycles that characterized the previous decade.

Tailwinds

• The NIH budget for FY 2026 remains at $47.2 billion, ensuring that the agency has the funds to eventually obligate for projects.
• The July 1 standard start date typically creates a large cluster of project records, which may provide a seasonal boost despite overall delays.
• Fiscal year-end pressure (Sept 30) often forces agencies to accelerate the obligation of remaining funds in late summer (July/August).
• A segment of ‘non-competing continuation’ grants and certain renewals may proceed despite the slowdown in new ‘Type 1’ awards.

Headwinds

• A 30% lower spending rate as of April 2026 compared to the 2021–2024 average, indicating severe disbursement bottlenecks.
• A 24% reduction in the total number of funded research projects in 2025 compared to 2024.
• Collapse in targeted funding opportunities, with only 14 NOFOs published by mid-2026 compared to 756 in 2024.
• New political vetting and AI-screening requirements that have reportedly delayed 45% of grant starts according to researcher surveys.
• Implementation of the ‘Unified Funding Strategy’ requiring full upfront payment, which has constrained the total volume of new grants.

Detailed reasoning

My analysis indicates a period of severe administrative and fiscal disruption at the NIH during the summer of 2026. The forecast is primarily driven by three converging factors: a sharp decline in funding volume, significant spending delays, and new administrative hurdles that have introduced unprecedented backlogs in the grant award process.

Historically, the National Institutes of Health (NIH) funds between 50,000 and 55,000 research awards annually. Under normal conditions, the 45-day window from June 28 to August 11—which includes the major July 1 start date for Cycle I awards—would typically see a significant cluster of several thousand new project starts. However, reports from early-to-mid 2026 show that the NIH funded 24% fewer projects in 2025 than in 2024, and the number of active ‘Notices of Funding Opportunities’ (NOFOs) plummeted from 756 in 2024 to just 14 by mid-2026. This indicates a much smaller pool of potential awards for the current window.

Furthermore, spending execution has slowed dramatically. As of April 22, 2026, the NIH had spent only $8 billion of its $47.2 billion budget for Fiscal Year 2026. This spending level is approximately 30% lower than the historical average for the 2021–2024 period. The slow disbursement of funds suggests that the administrative machinery for issuing awards is severely bottlenecked.

Policy changes have compounded these delays. A White House-mandated ‘Unified Funding Strategy’ now requires full upfront payment for multi-year projects, which has fundamentally altered traditional grant-making timelines and reduced the volume of new awards. Additionally, new administrative requirements, including political pre-issuance reviews and AI-based keyword screening for grant topics, have induced months-long backlogs. Survey data from 2026 indicated that 45% of researchers experienced delays in their grant start dates.

My most likely prediction of 2,700 projects accounts for this systemic slowdown. It assumes that while the standard July cycle will still occur, it will be significantly smaller and more delayed than usual. The 50% highest density interval (1,500 to 4,100) reflects the high uncertainty regarding how many awards will actually be processed and visible in the RePORTER database by the August 12 observation date. The lower end of this range (1,500) represents a scenario where administrative freezes persist, while the upper end (4,100) allows for a moderate catch-up in award issuance as the agency approaches the end of the fiscal year. The overall distribution remains well below the historical pro-rata baseline, reflecting the ‘hypercompetitive’ and ‘opaque’ environment described by researchers in mid-2026.

Key uncertainties

• The speed at which NIH RePORTER updates its records; delays in database entry could significantly undercount projects by the August 12 observation date.
• The potential for a ‘catch-up’ burst of award issuances in late July as NIH officials attempt to meet end-of-year spending targets.
• The exact impact of the OMB proposed rule (May 2026) on the specific 45-day window, as the rule’s formal effective date is Oct 1, 2026.
• The degree to which multi-year ‘upfront funding’ requirements will further cannibalize the number of distinct new project starts.

Conclusion

• The forecasted number of NIH project starts is significantly depressed relative to historical norms, driven by a 30% reduction in year-to-date spending and a 24% decline in the volume of funded research projects compared to the previous year.
• Decision-makers should anticipate that the summer peak for grant starts will be muted, with administrative backlogs and new vetting procedures likely pushing many project starts beyond the August 11 window or the August 12 observation date.
• The high level of uncertainty, reflected in the wide 50% HDI, stems from the unpredictable pace at which NIH may attempt to clear backlogs and obligate remaining FY 2026 funds before the September 30 fiscal year-end.

The question asks how many distinct NIH RePORTER project/application records will show Project Start Dates between 2026-06-28 and 2026-08-11 when RePORTER is observed on 2026-08-12. Forecasters agree the resolution is an API/database count (count distinct ApplId/Application_IDs returned by a Project Start Date query) and emphasize replicating the query exactly (paginate results, keep NIH-funded/administered records, do not filter by activity code, notice date, fiscal year, or budget start date). A major shared concern is RePORTER visibility lag: weekly refreshes and the 7–10 day typical posting delay mean many Aug 1–11 starts may not appear by Aug 12, so the observed count could be materially depressed. Teams also agree FY2026 shows early-year award-processing disruptions, administrative screening, and staffing constraints that shift timing into summer and widen uncertainty, though enacted funding overall is near FY2025 rather than a full collapse. Given those factors and lacking verified historical same-window API counts, forecasters anchor the distribution in the low‑to‑mid thousands (typical central estimates ~5–10k; one mode cited near 5.5–6.5k), allow a meaningful right tail for catch‑up or broader inclusion, and treat probabilities above ~40,100 as essentially negligible. The highest‑value next step is to run exact historical and current NIH API/ExPORTER queries for the same window to ground the forecast and quantify the Aug. 12 “mature-window” vs. last‑10‑day visibility effects.

Forecast and Reasoning

Forecast: Probability distribution:

• 1.00% chance of value below 1000.0
• 5.00% chance of value below 1600.0
• 10.00% chance of value below 2300.0
• 20.00% chance of value below 3200.0
• 30.00% chance of value below 3800.0
• 40.00% chance of value below 4300.0
• 50.00% chance of value below 4800.0
• 60.00% chance of value below 5400.0
• 70.00% chance of value below 6100.0
• 80.00% chance of value below 7100.0
• 90.00% chance of value below 8900.0
• 95.00% chance of value below 11000.0
• 99.00% chance of value below 16000.0

Forecast Reasoning

We are counting distinct RePORTER project records whose displayed “Project Start Date” falls in a 45-day window (28 Jun–11 Aug 2026), as observed on 12 Aug 2026. The key modeling point (all forecasters noted) is that this field generally reflects the start of a project period (often the initial start; at minimum it is not the annually-recurring budget start for non-competing continuations), so the count is dominated by new project periods (new awards and any actions that reset the project-period start), not the full stock of active NIH awards.

Magnitude/seasonality: NIH competing-award volumes are on the order of ~11,000/year in recent NIH Data Book figures (FY2025) and total “project-period starts” across all mechanisms plausibly somewhat higher. The queried window includes the major July 1 start-date cluster (and often Aug 1 as well), so it likely captures an outsized share of annual starts (roughly ~25–40% rather than 45/365).

Uncertainties that widen the distribution: (i) FY2026 timing disruptions could either depress starts or cause catch-up bunching into July/August; (ii) inclusion of all activity codes (not just RPGs) can add volume; (iii) the “as observed on 12 Aug 2026” criterion introduces possible reporting lag that would bias the observed count downward versus eventual totals.

Synthesis across analysts: Forecasters 3 and 2 anchor the center around ~5,000–5,500 with cited competing-award baserates and seasonality; Forecasters 1 and 4 center closer to ~3,500 and emphasize potential FY2026 slowdowns. I anchor near Forecaster 3’s empirically grounded center but keep a somewhat wider upper tail than F3/F4 (to allow for catch-up bunching and broader mechanisms), while not adopting Forecaster 1’s extremely large tail (which would be more consistent with mistakenly counting all continuing awards).

An amateur estimate would commonly (a) assume starts are uniform through the year (multiplying an annual count by 45/365), understating the July/August clustering, or (b) confuse “Project Start Date” with budget-period starts and overcount into the tens of thousands. My forecast explicitly conditions on the RePORTER field definition, incorporates known seasonality around July 1, and adds uncertainty for FY2026 timing/reporting-lag effects. I’m moderately confident this is better calibrated than an amateur approach, though residual risk remains if RePORTER’s displayed start-date behavior for renewals differs from expectations.

Forecast rationale (numeric):

— Iteration 1 — Across the forecasts, the main reasoning pattern is that NIH project starts are highly seasonal, with July 1 and August 1 treated as the key anchor dates. Because the window from 28 June to 11 August spans both of those dates, it is expected to capture a disproportionately large share of annual NIH starts rather than a random slice of the year.

Main factors driving the estimates

• Seasonality / clustering: NIH awards and project starts are concentrated around summer transition dates, especially July 1 and August 1.
• Counting distinct records: The task is interpreted as counting distinct RePORTER projects/awards, not just parent grants. Some rationales also assume subprojects count separately, which raises the total.
• Reporting lag: Because the cutoff is near the update date, some projects starting late in the window may not yet appear in RePORTER, pushing the observed count somewhat lower than the true count.
• Field semantics risk: A major source of uncertainty is whether “Project Start Date” is being interpreted narrowly as actual project starts or more broadly in a way that could capture additional records.

Areas of consensus

• The count should be well above a trivial level because the window includes major NIH start dates.
• A several-thousand estimate is the central expectation.
• There is meaningful uncertainty from RePORTER update timing, record structure, and how the field is defined.

Main disagreements

• Central magnitude: Estimates vary from the mid-4,000s to the mid-7,000s.
• Interpretation breadth: Narrow readings keep the count in the low-to-mid thousands, while broader interpretations could push the total much higher.
• Tail risk: Some reasoning assigns substantial probability to very large counts if the query is effectively capturing a broader date field or additional record types.

Overall synthesis

The collective reasoning points to a mid-thousands estimate with wide uncertainty, driven primarily by NIH’s summer start-date concentration, adjusted downward for RePORTER lag, and widened substantially to account for ambiguity in how “Project Start Date” is recorded and what entity types are included.

— Iteration 2 — Across the forecasts, the dominant view is that the count will land in the mid-thousands, with a plausible central range of roughly 5,000 to 9,000 distinct NIH RePORTER projects.

Main reasoning patterns

• Definition matters a lot: The forecasts hinge on interpreting “Project Start Date” as a narrower field than “Budget Start Date.” This generally means counting only new/competing project starts, not all NIH award activity.
• Strong seasonal clustering: The window includes July 1 and approaches August 1, both major NIH start-date clusters. This pushes the count above a simple uniform-day share of annual activity.
• Visibility lag in RePORTER: Because the observation date is 12 Aug 2026, some projects starting in the final days of the window may not yet appear due to 7–10 day reporting lag and weekly refresh timing. That especially affects the August 1 cohort.
• Annual scale anchors the estimate: The forecasts use NIH’s large annual volume of project records as a baseline, then scale down to the 45-day window and adjust upward for start-date clustering.
• Wide uncertainty remains: Tails are kept broad because of possible ambiguity around what counts as a distinct project, potential duplicates/subprojects, and the possibility of data-definition mistakes.

Areas of consensus

• The count should be well above a few hundred and far below total annual NIH project volume.
• The distribution is right-skewed, with a meaningful chance of a higher-than-central outcome if August 1 awards are already visible and counted broadly.
• A few-thousand to high-thousands result is the most likely outcome.

Main disagreement

• The forecasts differ mainly on how much of the August 1 cohort is visible by 12 Aug and how broadly to interpret “distinct project.”
• That drives the spread from a lower several-thousand estimate to a higher estimate closer to nine thousand.

— Iteration 3 — Overall, the forecasts converge on a mid-thousands count of NIH RePORTER records with a Project Start Date in the 45-day window, driven by three shared ideas:

• Seasonality / academic-cycle concentration: The window captures the heavy July 1 start period, when many NIH awards, especially R-series, fellowships, training grants, and K awards, begin.
• Strict field interpretation matters: Most reasoning assumes “Project Start Date” means the original project start date, not budget-period starts or continuation years. That keeps the count focused on newly initiated projects rather than all awards.
• RePORTER lag will suppress the most recent starts: Because RePORTER often lags by about 7–10 days, starts near the end of the window may not yet appear by the August 12 observation date, trimming the observed count.

Areas of agreement

• The count should be substantially above a few hundred and likely in the thousands.
• The window is likely front-loaded around early July.
• There is meaningful uncertainty from publication lag and from how one counts special cases like supplements, subprojects, or nonstandard award records.

Main differences in emphasis

• The models differ mainly in the annual base rate of NIH starts they assume and how strongly they weight the July concentration.
• One forecast is more conservative around ~5,300, while another is higher at ~7,300; both still treat the central estimate as several thousand.
• Wider upper tails are retained in case the query is interpreted more loosely, especially if budget start dates or broader award records are inadvertently included, which could push counts much higher.

Bottom line

The collective reasoning points to a large but bounded cluster of NIH project starts in this period, with the best estimate in the low-to-mid thousands, and the largest uncertainty coming from data-lag effects and definition ambiguity.

Summary of Rationales

1. Definition of “Project Start Date” The rationales uniformly emphasize that the NIH RePORTER “Project Start Date” strictly refers to the original inception date of a grant. Consequently, querying this specific field exclusively captures brand-new (Type 1) awards and their associated sub-projects. It filters out the vast majority of active NIH grants, such as non-competing continuations (Type 5 awards), which retain their historical project start dates from prior years.

2. Historical Clustering in the Target Window The June 28 through August 11 window naturally captures a high volume of new project starts. Start dates administratively cluster on the first of the month, meaning this timeframe encompasses the massive July 1 cohort (driven by Cycle III awards and academic training/fellowship grants) and a secondary August 1 cohort.

3. FY 2026 Fiscal and Administrative Anomalies Forecasters point to several unique downward pressures specific to Fiscal Year 2026 that will suppress the number of visible awards:

• Reduced Total Pool: Due to a delayed federal budget, the NIH heavily relied on multi-year forward funding (paying out multi-year grants entirely from the current budget). This strategy drained available capital and structurally reduced the absolute number of distinct new Type 1 awards available to be issued.
• Processing Bottlenecks: A government shutdown, delayed advisory councils, new compliance checks, and severe staffing shortages created massive administrative backlogs. While delayed grants are being forced into the summer months, the overwhelmed system is struggling to finalize Notices of Award (NoAs) on time.

4. The Reporting Lag and Observation Date Cutoff The specific observation date of Wednesday, August 12, serves as the most critical limiting factor. NIH RePORTER runs weekly updates on Sunday nights, meaning the system will only reflect data up to Sunday, August 9. Furthermore, the database enforces a 7-to-10-day delay after a grant’s start date before it becomes publicly visible.

• July Starts: Projects starting in July will generally have cleared the technical lag and be visible, provided they did not succumb to the severe administrative processing delays.
• August Starts: Projects with an August 1 start date will fall precisely within the 7-to-10-day blackout window on the August 9 refresh. As a result, the vast majority of August starts will be truncated and completely hidden on the observation date.

In conclusion, while the target window typically sees a large influx of new grants, the combination of a reduced FY 2026 grant pool, administrative backlogs, and the rigid reporting lag that hides August starts will severely depress the final observed count.

SUMMARY

Question: How many distinct NIH-funded projects in NIH RePORTER will have a “Project Start Date” from 28 June 2026 through 11 August 2026, inclusive? Final Prediction: Probability distribution:

• 10.00% chance of value below 1990.0
• 20.00% chance of value below 2605.0
• 40.00% chance of value below 3630.0
• 60.00% chance of value below 4850.0
• 80.00% chance of value below 7375.0
• 90.00% chance of value below 10150.0

Total Cost: extra_metadata_in_explanation is disabled Time Spent: extra_metadata_in_explanation is disabled LLMs: extra_metadata_in_explanation is disabled Bot Name: extra_metadata_in_explanation is disabled

Report 1 Summary

Forecasts

Forecaster 1: Probability distribution:

• 10.00% chance of value below 1580.0
• 20.00% chance of value below 2110.0
• 40.00% chance of value below 3060.0
• 60.00% chance of value below 4200.0
• 80.00% chance of value below 6750.0
• 90.00% chance of value below 9300.0

Forecaster 2: Probability distribution:

• 10.00% chance of value below 2400.0
• 20.00% chance of value below 3100.0
• 40.00% chance of value below 4200.0
• 60.00% chance of value below 5500.0
• 80.00% chance of value below 8000.0
• 90.00% chance of value below 11000.0

Research Summary

The research examines how many distinct NIH-funded projects in NIH RePORTER will have “Project Start Date” between June 28 and August 11, 2026, and concludes that answering this is difficult because FY2026 features an unprecedented disruption to NIH grantmaking. Key empirical points the research reports: by March–June 2026 NIH had seen a ~66% drop in competitive grants awarded (vs FY21–24 average), a ~54% reduction in award dollars, only $5.8 billion of an estimated $38 billion extramural budget obligated (~15%) as of June 22, 2026, and nearly a 20% workforce loss. Policy changes include a shift to funding full multi-year project periods up front (reducing the number of new awards), a fall in R01-equivalent success rates (reported from 24% to 17%), a collapse in Funding Opportunity Notices (14 NOFOs by mid-March 2026 vs. 756 in 2024), and late OMB apportionment approval (March 16, 2026) that delayed spending.

The research also summarizes relevant base rates and data limitations: NIH historically operates three annual grant cycles with the June 28–Aug 11 window aligning with “Cycle III” starts and fiscal-year-end dynamics; NIH RePORTER updates weekly and newly funded projects typically appear 7–10 days after Budget Start Date, but the researcher could not find historical counts of projects with start dates in comparable 6-week windows or prediction-market forecasts for this question. Given the extraordinary 2026 disruptions, the research identifies major uncertainties (pipeline volume, administrative capacity, possible late‑FY acceleration) and recommends directly querying NIH RePORTER for comparable historical windows, monitoring RePORTER for June–July 2026 starts, and adjusting expectations downward from historical baselines while tracking policy and budget developments.

Sources used:

• NIH RePORTER: https://reporter.nih.gov/
• NIH Data Book: https://report.nih.gov/nihdatabook/
• Association of American Universities: https://www.aau.edu/
• Congressional Research Service reports: https://crsreports.congress.gov/
• News and analysis outlets referenced in the research (examples): STAT News https://www.statnews.com/, Nature https://www.nature.com/, The Hill https://thehill.com/

RESEARCH

Report 1 Research

Research Summary: NIH RePORTER Project Start Dates (June 28 - August 11, 2026)

Question Overview

This question asks how many distinct NIH-funded projects will have a “Project Start Date” falling between June 28, 2026 and August 11, 2026 (inclusive), as observed in NIH RePORTER on August 12, 2026. Each unique application/award identifier counts once, across all NIH institutes/centers and activity codes.

Critical Context: Unprecedented NIH Funding Disruption in 2026

The 2026 fiscal year has seen extraordinary disruption to NIH grantmaking that makes historical patterns unreliable:

Funding Collapse (As of March 2026)

• 66% fewer competitive grants awarded by end of February FY26 vs. FY21-24 average [24]
• 54% reduction in monetary value of awards [24]
• Only $5.8 billion obligated of $38 billion estimated extramural funding as of June 22, 2026 (~15% of total) [13]
• NIH lost nearly 20% of its workforce over the past year [24]

Policy Changes Affecting Award Timing

• Multi-year funding shift: NIH now funds awards fully upfront for entire project periods (4-5 years), dramatically reducing the number of new grants [18][20]
• Success rate collapse: Dropped from 24% (FY24) to 17% (FY25) for R01-equivalent awards [5]
• Funding opportunity notices plummeted: Only 14 NOFOs published by mid-March 2026 vs. 756 in 2024 [24]
• Budget delays: OMB only approved NIH apportionments on March 16, 2026, causing months of delays [24]

Proposed Further Cuts

• Trump administration proposed 13% cut for FY2027 ($41.3 billion vs. $47.2 billion in FY26) [4]
• Proposed closure of three NIH institutes/centers [3][4]
• Ongoing political uncertainty with new grant regulations threatening to politicize peer review [9][10]

Relevant Base Rates and Reference Classes

Historical Grant Cycle Timing

NIH operates three annual grant cycles with typical earliest possible project start dates [17]:

• Cycle I: September or December (applications due Feb 5/Mar 5)
• Cycle II: April (applications due June 5/July 5)
• Cycle III: July (applications due Oct 5/Nov 5)

Key insight: The question period (June 28 - August 11) coincides with Cycle III project start timing and the end of fiscal year 2026 (which ends September 30).

NIH RePORTER Data Characteristics

• Newly funded projects appear 7-10 days after Budget Start Date [15]
• Data updates occur weekly (Sunday nights, available Mondays) [15]
• “Project Start Date” is distinct from Budget Start Date, Award Notice Date, and Fiscal Year fields [14]

Historical Award Patterns (Pre-Crisis)

• Monthly NIH extramural awards ranged from 324 to 64,434 depending on month [25]
• September typically shows highest volume as fiscal year-end approaches [25]
• The 6-week window in question is relatively short compared to full fiscal year

Recent Performance (FY2025)

• R01-equivalent awards dropped from 7,720 (FY24) to 5,885 (FY25) [5]
• In FY25, 2,700 fewer scientific projects funded overall (15% decline) [8]
• Average grant size nearly doubled from $472K (early 2025) to over $830K (year-end) due to multi-year upfront funding [2]

Challenges for Forecasting

Lack of Specific Historical Data

I could not locate:

• Historical counts of projects with start dates in comparable 6-week late June/early August windows
• Specific prediction markets for this question on Metaculus, Manifold, Polymarket, or Kalshi
• Weekly/daily granularity of project start date distributions from prior years

Unprecedented Uncertainty

The 2026 funding environment is historically anomalous:

• Traditional grant cycle patterns may not hold given administrative chaos [2][8]
• Multi-year funding means fewer total projects but larger individual awards [18][20]
• Only 15% of extramural budget obligated as of June 22 [13] - just 6 days before the question period begins

Timing Considerations

• Federal fiscal year ends Sept 30: Agencies often accelerate spending in August-September
• Question period spans 6 weeks: Shorter window means fewer projects than quarterly or annual counts
• Current date is June 28: The forecast window begins immediately, with resolution in just 6 weeks

Key Uncertainties

1. Will NIH accelerate late-FY spending? Historically yes, but current crisis makes this uncertain
2. How many projects are “in pipeline” with June 28-Aug 11 start dates already set?
3. Multi-year funding impact: Fewer new starts but potentially different timing patterns
4. Administrative capacity: 20% workforce loss may slow processing regardless of budget

Recommendation for Forecasting Approach

Given the lack of specific base rates and prediction markets, I recommend:

1. Query NIH RePORTER directly (if possible) for historical project start date counts in comparable windows (e.g., June 28 - Aug 11 in 2023, 2024, 2025)
2. Monitor recent trends: Check RePORTER for June 2026 starts to gauge current pace
3. Adjust downward from historical baseline: Given 54-66% reductions in awards through February and only 15% budget execution by late June
4. Consider fiscal year-end dynamics: Some acceleration possible in July-August, but dampened by crisis
5. Track news closely: Any policy changes or budget releases in next 6 weeks could dramatically shift outcomes

Data Sources Summary

• NIH RePORTER (official database): [11][12][14][15][16]
• Association of American Universities analysis: [24]
• Congressional Research Service report: [18]
• NIH Data Book: [25]
• Multiple news sources documenting 2026 crisis: [1][2][5][7][8][9][10]

Bottom line: This is an extremely difficult forecasting environment due to unprecedented disruption and lack of directly comparable historical data for the specific question window. Traditional base rates from 2023-2024 are likely unreliable for 2026.

FORECASTS

R1: Forecaster 1 Reasoning

I accept the Round 3 critic’s main correction: I was probably giving too much weight to “award processing recovered” evidence. Total NIH award counts are a weak proxy here because the target is Project Start Date, not Budget Start Date or award notice date. RePORTER explicitly separates Budget Start Date—the funding start for a particular fiscal year—from Project Start Date—the start date of the project; subprojects inherit the parent award’s start date, and competitive renewal extends the project end date rather than resetting the start date. (api.reporter.nih.gov)

I also accept that the biggest remaining missing input is an exact historical NIH RePORTER query. The API directly supports project_start_date, supports date_added, and returns a total project count, so comparable windows in 2023–2025 are in principle the right reference class. (api.reporter.nih.gov) Without that exact count, I should avoid overconfidence and let database-semantics uncertainty keep tails wide.

I still defend the core structure of the forecast. The window is not arbitrary: NIH’s standard review cycle lists July as the earliest project start date for Cycle III, so July 1 starts and backdated July starts remain a real upside channel. (grants.nih.gov) Also, RePORTER usually posts grants 7–10 days after Budget Start Date and refreshes weekly, so most July 1 starts should be visible by August 12, while some August 1–11 or late-notice awards can leak past the snapshot. (report.nih.gov)

The downward pressure is still substantial. FY2025 already showed a much smaller competing-award denominator than total award counts: 55,394 all extramural grant awards versus 8,161 new-or-renewal competing RPG awards, and Project Start Date is narrower still for renewals/continuations. (grants.nih.gov) AAMC’s March analysis showed FY2026 new awards far below recent averages, though I agree this should not be mechanically extrapolated as a continuing near-freeze, since AAMC’s later tracking is through June 20 and explicitly excludes subprojects. (aamc.org)

Net update: I move down from my Round 2 median of 3,750 to about 3,500, and trim the high tail modestly. I keep a fat upper tail for subprojects, backlog clearing, and July 1 backdating, but I no longer want 5% mass above ~14k. The low tail remains bounded by the market floor and by the expectation that at least some July-cycle starts will post.

Updated prediction:
Percentile 0.1: 1002
Percentile 1: 1070
Percentile 5: 1280
Percentile 10: 1580
Percentile 20: 2110
Percentile 30: 2600
Percentile 40: 3060
Percentile 50: 3500
Percentile 60: 4200
Percentile 70: 5150
Percentile 80: 6750
Percentile 90: 9300
Percentile 95: 12400
Percentile 99: 20000
Percentile 99.9: 32000

R1: Forecaster 2 Reasoning

(a) Time left: From the prompt date, 28 June 2026, about 45 days remain until the outcome is observed on 12 August 2026; the project-start-date window itself ends on 11 August 2026.

(b) Outside view: The best reference class is NIH RePORTER project records with new/current project-period start dates in comparable late-June-to-early-August windows. NIH’s overall extramural award counts are much larger than the target quantity: cumulative annual grants/OT awards were 64,434 in FY2022, 65,639 in FY2023, 64,277 in FY2024, and 58,736 in FY2025, and July–August award-month increments alone were roughly 20,000–22,000 in recent years. But the target is Project Start Date, not award notice month or budget-period start, so the more relevant base rate is competing/new project-period starts: NIH RPG competing awards were 11,052 in FY2023, 10,265 in FY2024, and 8,161 in FY2025, with Type 1 new RPG awards at 8,869 in FY2024 and 6,996 in FY2025. A normal 45-day summer window containing the big July 1 start date would therefore plausibly land in the several-thousand range, not the 20,000+ all-award-month range. (report.nih.gov)

(c) Inside view: FY2026 is not normal. NIH’s own monthly awards data show FY2026 cumulative awards through May at 23,447, below FY2022–FY2024 by about a quarter but close to FY2025’s 24,480; however, this includes noncompeting continuations and does not directly tell us new project-period starts. AAU found NIH had issued 66% fewer competitive awards through February FY2026 than the FY2021–FY2024 average for the same period, and CRS notes multiyear funding reduces the number of grants awarded in a given year and contributed to lower success rates. RePORTER timing also matters: new grants generally appear 7–10 days after Budget Start Date and after the weekly refresh, so some late-window August starts may not be visible by 12 August. These factors push the forecast below the normal-summer base rate and widen the right tail for a delayed-award catch-up. (report.nih.gov)

(d) Outcome if nothing changed: If current conditions persist without a new freeze or major catch-up, I would expect roughly 4,000–6,000 qualifying project records.

(e) Outcome if the current trend continued: If competitive/new starts stay depressed while most effort goes to continuations and selected multiyear awards, I would expect roughly 2,500–4,500.

(f) Expectations of experts and markets: I found no dedicated prediction market. Expert/public analyses point downward versus history: AAU emphasizes the FY2026 competitive-award slowdown, while CRS says FY2026 awards and obligations were lagging prior years and that multiyear funding reduces grant counts and success rates. NIH’s monthly data, though, show all-award activity had partially recovered by May, so the extreme-low story is less persuasive for total RePORTER records than for new R01/RPG awards alone. (aau.edu)

(g) Unexpected low scenario: A further administrative/policy hold on new awards, very slow processing after July council decisions, or prioritization of only noncompeting continuations causes most July/August starts to miss RePORTER by 12 August, producing an outcome near the lower bound, around 1,000–2,000.

(h) Unexpected high scenario: NIH pushes through a large backlog of delayed Cycle III and late-FY awards, backdating many project starts to 1 July or 1 August and getting them into RePORTER by 12 August, producing a count above 10,000 and, in an extreme catch-up, into the high teens.

Rationale: I anchor on the historical new-start base rate rather than the much larger all-award-month counts. Recent NIH award-month data show 20,000+ July–August all-award increments in recent years, but that mostly includes annual continuations; the Project Start Date field is closer to new/competing project periods. Recent annual competing RPG counts around 8,000–11,000 imply a comparable summer window usually captures only several thousand records, concentrated around July 1 and August 1.

The 2026 evidence then moves the center down. The competitive-award pipeline was sharply disrupted early in FY2026; multiyear funding means fewer discrete awards for a given budget; and RePORTER lag makes the tail end of the window less likely to be fully visible by resolution. Offsetting that, NIH’s all-award count through May 2026 was not catastrophically below FY2025, and late-FY catch-up behavior is common. My median is therefore below the “normal” several-thousand-to-low-five-figure base rate but not near the floor.

Percentile 0.1: 1050
Percentile 1: 1300
Percentile 5: 1900
Percentile 10: 2400
Percentile 20: 3100
Percentile 30: 3700
Percentile 40: 4200
Percentile 50: 4800
Percentile 60: 5500
Percentile 70: 6500
Percentile 80: 8000
Percentile 90: 11000
Percentile 95: 14000
Percentile 99: 22000
Percentile 99.9: 36000

Frontier Forecast — Post 538

Modal: 900-5800 (29.2%) • frontier aggregate • 10m58s

Interpret Summary

• Reading: strict
• Type: strict
• Window: Projects with Project Start Date between 2026-06-28 and 2026-08-11 inclusive, as recorded in NIH RePORTER on 2026-08-12

Resolution sources/checks:

• https://reporter.nih.gov/: https://reporter.nih.gov/ - Resolver URL extracted mechanically from resolution criteria

Edge cases:

• Whether projects whose start dates are backfilled or revised after initial entry (but before 12 August 2026) are included — the resolution date snapshot governs, so any data visible on 12 August counts.
• Whether ‘distinct project’ means unique application ID (appl_id) vs. unique core project number (project_num/serial_number) — RePORTER has both; the resolution criteria say ‘unique application/award identifier’, which likely means appl_id…
• Whether projects from non-standard NIH funding mechanisms (e.g., contracts vs. grants) are included — criteria say ‘all activity codes’, which should include contracts, but contracts may be indexed differently in RePORTER.

Frontier Views (3/4)

• frontier_1 - p: ?

  • failed: :timeout

• frontier_2 - Modal: 10700-15600 (20.0%)

  • The question asks for the count of distinct NIH projects with a Project Start Date in a 45-day window (28 Jun – 11 Aug 2026), observed on 12 Aug 2026. Several forces govern the count and create wide uncertainty: 1.

• frontier_3 - Modal: 5800-10700 (42.0%)

  • The resolution criteria state that the count relies exclusively on the field labeled ‘Project Start Date’ in NIH RePORTER as observed on August 12, 2026. The NIH ExPORTER data dictionary clarifies that ‘Project Start Date’ is the original start date of the project (e.g., when the Type 1 new award began).

• frontier_4 - Modal: 900-5800 (82.0%)

  • 45-day late-summer window under FY2026 conditions of 55-66% reduced award pace yields expected volume of 1.5k-5k, placing overwhelming probability in bin_0.

Frontier Lane Failures

Aggregate used 3/4 frontier lanes; failed/excluded lane(s): frontier.

Adjudication

• Status

  • 3 non-blocking adjudication note(s); no lane required discounting.

• Guidance

  • All three lanes did not obtain the direct RePORTER snapshot for 12 Aug 2026 (the explicit resolver). Each relied on background sources (data-dictionary, AAU/AAMC reports, seasonality charts) to adjust historical anchors.

• Revision

  • Frontier revision skipped: no_selected_adjudicator_reviews.

Final Distribution (discrete bins)

Sources Checked

• Resolver: Resolver URL extracted mechanically from resolution criteria - https://reporter.nih.gov/
• frontier_2: grant-witness.us funding curves
• frontier_2: AAMC Tracking NIH Awards FY2026; AAU
• frontier_2: grants.nih.gov NIH Awards by Month (Apr 2026)
• frontier_2: NIH RePORTER / API data elements documentation
• frontier_3: NIH RePORTER API / ExPORTER Data Dictionary

wooley_rhino • run 4b0eeac9 • $2.2 • 10m58s

NIH funding flows and the public record are anchored to the federal fiscal calendar and fixed budget‑period conventions.

Public project visibility is constrained by a short, stable loading lag and by accrual patterns that concentrate additions later in the fiscal year.

Late‑June visibility through the reporting frontier has been markedly stronger than the recent multi‑year baseline, pushing the expected window total upward.

Broader fiscal pressures earlier in the year reduce the attainable ceiling for new starts, but June catch‑up activity has materially offset that downside risk.

If the late‑June pace persists into July and early August, the count will sit near the higher summer regime that anchors the central estimate.

If awarding activity reverts to the early‑FY slowdown observed last year, totals will migrate toward the lower analogue and trim the upper tail.

The aggregate is sensitive to definitional choices—counting by unique application identifier versus core‑project aggregation and inclusion of subprojects changes totals by several percent.

The dominant residual uncertainties are the exact database refresh frontier on August 12 and last‑minute administrative award timing, which can shift the final count by several hundred records.

Forecast: median = 3482 distinct NIH-funded projects (NIH RePORTER) with Project Start Date from 28 Jun 2026–11 Aug 2026 (inclusive), observed Aug 12, 2026.

• It’s about “Project Start Date,” not when funding is renewed. The main reason the count isn’t huge: most NIH awards are continuations whose project started earlier, so they don’t qualify.
• The window hits the annual NIH “start-cycle” months. Late June through early August includes especially common start dates (notably around the first of the month), which boosts the number that falls in this exact range.
• Expect uncertainty from what’s actually visible by Aug 12. RePORTER updates with a lag, so some starts near the back of the window may not yet appear—pushing the result down.
• FY2026 funding looks roughly stable, not collapsing. That supports a mid-to-high count, though prior-year disruptions and shifts toward multiyear arrangements keep some suppression risk.
• Overall distribution is right-skewed. The “usual” case is moderate, but there’s a meaningful chance of a higher count if the system returns more qualifying records than expected (e.g., how subprojects are counted).