FEMA reported 15 percent of its incident management workforce available to deploy at the start of the 2026 hurricane season — as of 2026-08-08 no source states how many people that workforce contains, and no rule requires a staffing decision to clear a readiness figure

What is happening?

The Federal Emergency Management Agency recorded over 4,300 separations in fiscal year 2025, about 17 percent of a workforce that averaged 25,134 employees that year, and a 55 percent increase over the 2,791 separations of fiscal year 2024. In the same year the agency suspended the numerical workforce targets it had set for fiscal years 2023 through 2025. Both facts were published by the Government Accountability Office on 2026-08-04 in GAO-26-108427.

The agency still publishes a readiness percentage. Only 15 percent of its incident management workforce was available to deploy on 2025-07-04, when the Texas floods began, and the same 15 percent at the start of the 2026 hurricane season. An earlier reading was lower still — 4 percent in November 2024, after FEMA fielded 13,500 people for Hurricanes Helene and Milton, the largest deployment in its history. The two readings were taken with the workforce committed to very different degrees, and no source opened here states which condition is the worse one or what either figure should have been.

What is missing is the figure those percentages were meant to be measured against. Executive Order 14356 requires an annual staffing plan, and the Department of Homeland Security directed FEMA to build it from the bottom up. The program and regional offices did that analysis and recommended holding at about 23,000 employees. The plan FEMA submitted to DHS in December 2025 proposed about 11,383 employees, roughly a 50 percent cut. GAO found no indication that the smaller figure rested on any analysis of organizational demands, and FEMA officials told GAO they did not have information about why the plan was not based on mission needs. As of 2026-04-18 the agency had about 20,968 personnel, carrying 1,079 open disaster declarations as of 2026-06-02.

Whose problem is this?

The body that suspended the targets is the same body those targets would have held to account. Nothing outside it required the targets to exist.

Where does this problem end?

The boundary here is not whether disasters are getting worse but whether a staffing decision has to clear a published figure before it takes effect.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state: no source opened here states how many people FEMA needs. GAO names no headcount in any of its five recommendations, the FEMA Review Council names none, and no bill names one. FEMA itself had targets for FY2023 through FY2025 and suspended them in 2025, and no source publishes what those target values were, so even the withdrawn figure is unavailable. The absence is structural rather than a gap in this round of research.

How big is it?

Between 17.3 million and 85.1 million people, with 56.5 million as the central reading. The basis is residents of counties where a federal major disaster declaration was actually issued while the workforce shrank, so the count covers places where the federal response capacity was the operative capacity rather than a hypothetical one.

The low bound, 17,327,963, is the population of the 149 counties designated for Individual Assistance in fiscal year 2026 to date, the window that falls entirely after the reductions. The central figure, 56,492,560, covers the 360 Individual Assistance counties across fiscal years 2025 and 2026 together. The high bound, 85,117,186, covers all 969 counties designated under any major disaster declaration across those two fiscal years. Designations come from the FEMA OpenFEMA declarations dataset pulled on 2026-08-07, and residents from the Census Bureau county population estimates for vintage 2024.

The gap between the bounds is about a factor of five and it is not noise. It is the difference between a county where FEMA was paying households directly and a county where FEMA had a declared operation of any kind. The figures also understate, by roughly 3 million people, because 38 Puerto Rico municipios and 3 United States Virgin Islands districts designated in fiscal year 2025 have no matching row in the Census county file used, and 48 designated areas across the two years are unmatched in total.

Everyone counted here is someone a past declaration already covered. The readiness figure at issue is prospective, so the people it is actually about have not been counted yet.

Under what conditions does it arise?

1. A readiness percentage can outlive the target it was measured against. FEMA retains the ability to report how many of its disaster staff are free on a given day, and reported 4 percent and 15 percent to GAO. What it lost in 2025 is the number those percentages were supposed to be compared to. A figure with no target can be published every season without ever being wrong.

2. No rule requires a staffing decision to clear a readiness figure first. Executive Order 14356 requires an annual staffing plan and DHS directed a bottom-up method, but nothing in the process required the resulting number to be reconciled with the number the plan actually carried. The analyzed figure of about 23,000 stayed in the regional offices and the unanalyzed figure of 11,383 went to DHS.

3. The document that would anchor a target was withdrawn and not replaced. FEMA rescinded its 2022 through 2026 strategic plan in May 2025 and had released no revised plan as of July 2026. GAO states that without a strategic plan there is no overall strategic direction on which to base workforce planning, so the agency cannot effectively determine its future workforce needs.

4. The intake valve and the outflow were controlled by different mechanisms on different clocks. A government-wide hiring freeze ran from January 2025, was extended in October 2025, and was partially lifted on 2026-05-18 for 347 priority positions. Separations were not paused at any point. The lift arrived after the 2026 hurricane season had already begun at 15 percent availability.

5. Two federal bodies pointed in opposite directions within three months. The FEMA Review Council approved a final report on 2026-05-07 recommending that the workforce be redistributed beyond Washington while staff are reduced. GAO reported on 2026-08-04 that reductions had exacerbated long-standing workforce problems. Neither document reconciles with the other and neither cites the numbers of the other.

What has been tried?

DHS agreed with the recommendations and then attached a condition to two of them, indicating that FEMA would wait until an Administrator was confirmed before starting. Four Senior Officials Performing Duties of the Administrator have held the position since January 2025, and the nominee named on 2026-05-11 was not confirmed as of July 2026.

What was found?

Why is it still unsolved?

Measurement absent — the numerator exists and the denominator does not.

FEMA does produce readiness figures and did report them. The 4 percent of November 2024 and the 15 percent of July 2025 are real measurements of a real quantity. What was removed in 2025 is the target force structure those percentages were supposed to be read against, and what has never existed is any rule requiring a workforce decision to be tested against a readiness figure before it takes effect. A readiness percentage with no target is a thermometer with no fever line. It can be reported at the start of every hurricane season, in good faith, without ever obliging anyone to act.

The second movement is what happens to measurement that has no standing. Measurement was in fact performed here. FEMA program and regional offices carried out the bottom-up analysis that DHS asked for and concluded that the current level of about 23,000 should be held. The annual staffing plan that reached DHS in December 2025 proposed about 11,383 instead. GAO reports conflicting information about the reason for that reduction and no indication that it was based on an analysis of organizational demands, and FEMA officials told GAO that they did not have information about why their own staffing plan departed from mission needs. Nothing required the two numbers to be reconciled or the second to explain its distance from the first, so the unanalyzed number travelled and the analyzed one did not. No source opened here states what became of the 11,383 figure.

The third movement is that the correction arrives after the season and is gated on an event with no date. What GAO recommends is not a headcount but a hook — that Congress require significant workforce decisions to be informed by the results of workforce planning, and require FEMA to report those results before each hurricane season begins. Neither requirement exists in law as of 2026-08-08. Meanwhile the hiring freeze lifted on 2026-05-18, after the season had begun at 15 percent availability with a Surge Capacity Force anticipated at about 240 against 600 the year before, and 134 hires had been made against 2,545 separations in the preceding seven and a half months. DHS agreed with all three recommendations to FEMA and then said the agency would wait for a confirmed Administrator before starting two of them. There has been no confirmed Administrator since January 2025.

Nobody hid a number here. The one number that would have made a staffing decision answerable was suspended by the agency that would have been answerable to it, and no rule anywhere required it to exist. The Post-Katrina Emergency Management Reform Act, at 6 U.S.C. 316, does bar the Secretary from substantially or significantly reducing the capability of FEMA to perform its missions. GAO reproduces that provision and then frames its conclusion prospectively, saying the agency is at risk of not meeting its statutory requirements if it makes workforce decisions that are not based on an understanding of its current workforce and mission demands. The body reviewing the conduct reaches no finding on whether the bar was crossed.

What observation would mean it is solved?

Candidates — (a) FEMA publishes a strategic plan and a workforce analysis carrying a numerical target force structure by cadre (b) Congress enacts both matters, so that workforce decisions must be informed by planning and the results must be reported before each hurricane season (c) incident management availability at the start of a season rises above a stated threshold.

(a) alone is weaker than it looks. FEMA had exactly that from FY2023 through FY2025, together with a July 2024 goal that every incident management cadre reach its target by the end of FY2026, and suspended it in 2025 without publishing a rationale or the values. A target that can be withdrawn without notice is a report rather than a constraint.

(b) alone produces a document, not a deployable person. A requirement to report the results of a workforce planning process does not require the process to arrive at any particular number. The one legislative vehicle carrying structural reform has sat reported out of committee since September 2025 with no Senate companion as of 2026-05-15.

(c) alone is uninterpretable. No source states what the threshold should be, and no source publishes how many people the incident management workforce contains, so the percentage cannot be converted into a headcount or compared across years with different total staffing. The figure also moves with disaster load as much as with headcount. It fell to 4 percent in the year FEMA fielded its largest deployment ever. The honest composite is (a) and (b) together with a published denominator, because (c) cannot be read without one.

What is it connected to?

Fills with researchrisk-based pricing under the National Flood Insurance Program and the proposed transfer of policies to private insurers, Public Assistance cost thresholds and the proposed up-front lump-sum formula grant, restructuring of the Hazard Mitigation Grant Program, Senior Executive Service retention across the federal government, and state and local emergency management capacity as the review council recommends shifting responsibility toward it. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• How many people FEMA needs. Not GAO, not FEMA, not the FEMA Review Council, and not any bill opened here. The targets that existed for FY2023 through FY2025 were suspended in 2025 and no source publishes what their values were, so even the withdrawn number is unavailable.
• The denominator of the 15 percent. Every source repeats 4 percent and 15 percent as the share of the incident management workforce available to deploy. None states how many people that workforce contains, so the figure cannot be turned into a headcount or judged adequate against anything.
• Who set the 11,383 figure, on what basis, and what became of it. GAO reports conflicting information about the reason and records that FEMA officials did not have information about why the plan departed from mission-needs analysis. Whether the plan was approved, rejected, withdrawn or is still pending is nowhere on the record opened here.
• Whether the statutory bar was crossed. GAO quotes 6 U.S.C. 316 and then states only that FEMA is at risk of not meeting its statutory requirements. No source opened here reaches a compliance finding in either direction.
• Which of the two federal positions governs. The review council approved a recommendation to reduce staff on 2026-05-07 and GAO reported on 2026-08-04 that reductions had made the problem worse. No document reconciles them or says which prevails.
• What the low availability cost. No source links the 15 percent figure, the reduced call centre staffing after the July 2025 Texas floods, or the smaller Surge Capacity Force to a measured outcome for disaster survivors. There is no response-time series, no unmet-need count, and no comparison against a season at higher availability. The GAO title itself says reductions may impact mission readiness, and the conditional is carried through the report.
• When the remediation starts. DHS conditioned two of the three recommendations on a confirmed Administrator. No source states when a confirmation vote is expected, what happens to the recommendations if none occurs, or who acts in the interim.
• How many residents the declarations cover. The federal declarations dataset names designated counties and attaches no population to them, so the figures in this document required an external Census join. The Census county file has no rows for the Puerto Rico municipios or the Virgin Islands districts that were designated, and the municipio estimates file for the same vintage returned a not-found response this round.

See the evidence

One federal audit report carries this document. GAO-26-108427 was opened directly in three forms — the product page, the full report and the highlights sheet — and supplies nearly every figure above. The Congressional Research Service insight was opened for the review council and the legislative status, because the FEMA press release carrying the final report returned a 403. Two federal datasets were retrieved directly and joined locally to produce the population range, and that join is the work of this round rather than a published figure. Three press accounts were opened as corroboration only. Where sources disagree the disagreement is left visible rather than resolved. Three different totals for separations circulate over three overlapping and non-additive windows, so this document uses the fiscal-year framing of GAO and names the window on every figure. One press account is unreliable on dates, placing the strategic plan rescission a year early, and is used here only for the corroboration of 11,383 against 23,000 and for the count of acting administrators. One press account describes the 58 Senior Executive Service departures as roughly half the corps, a denominator no opened source supplies, so that characterisation is not carried here. The per-region separation percentages printed on the highlights map were not legible at the available resolution and only the stated range of 4 to 14 percent is used, which also leaves one disagreement standing — the same press account describes the regional offices as losing 9 percent or more, a floor that cannot hold if the range of GAO opens at 4 percent, and neither source acknowledges the other, so the wider range is the figure carried above. This is a Path A output (research-based definition), so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 17,327,963–85,117,186 As of FY2025-FY2026 to date, declarations pulled 2026-08-07, population vintage 2024

Sensitivity The width is about a factor of five and it is a definitional difference rather than uncertainty. The low bound counts residents of counties where FEMA was paying households directly in the fiscal year that began after the reductions had already happened. The high bound counts residents of every county where any federal major disaster declaration was live across the two fiscal years. What the figure fails to count runs in one direction. 48 designated areas have no row in the Census county file used, including 38 Puerto Rico municipios and 3 United States Virgin Islands districts, so the count understates by roughly 3 million people, on the order of 4 percent of the high bound. Emergency declarations and Fire Management Assistance Grants are excluded entirely although staff deploy on those events too. In the opposite direction the figure overstates exposure, because a resident of a designated county is not necessarily someone who needed federal assistance, and someone who needed it is not necessarily someone the low availability harmed. No source opened here supports a harm rate. The count also measures declarations rather than need, so a change to Public Assistance cost thresholds would move it without any disaster changing, and everyone counted is covered by a declaration that has already been issued while the readiness figure at issue is prospective.

Regional breakdown The declarations dataset supports a county-level count, but the two fiscal years are aggregated to produce the range and no source opened here publishes a regional figure for the quantity actually at issue, which is deployable workforce against a target. GAO reports separations of 4 to 14 percent across the ten regions without naming the value for any individual region, and the per-region map on the highlights sheet was not legible at the available resolution. Attaching residents to regions would therefore report a disaster-exposure geography as if it were a workforce-shortfall geography.