Agencies Publish Their Data Blackout Plans Before an Appropriations Lapse, Not After

When a spending deadline approaches, most of the commentary that reaches a trading desk is about what markets will do. That question has a small sample behind it and no clean way to separate the deadline from everything else moving that week. It is not a question a position can be sized against.

There is a second question with a much firmer answer. A lapse in appropriations does something specific and documented to the flow of federal economic data. Releases stop. Collection stops. Schedules are rewritten afterward. And the details — which agency, how many staff, which release survives — are published in advance, in documents that exist before the deadline arrives. For anyone whose process depends on a dated release, that is the part of the event that can be prepared for rather than guessed at.

The Prohibition That Does the Work

Two sections of Title 31 explain why a statistical agency goes quiet rather than running on goodwill.

31 U.S.C. 1341(a)(1) states that an officer or employee may not make or authorize an expenditure or obligation exceeding an amount available in an appropriation or fund for the expenditure or obligation, and may not involve either government in a contract or obligation for the payment of money before an appropriation is made unless authorized by law. Payroll for a survey field representative is an obligation. Without an appropriation, it cannot be incurred.

31 U.S.C. 1342 closes the obvious workaround. An officer or employee may not accept voluntary services for either government or employ personal services exceeding that authorized by law except for emergencies involving the safety of human life or the protection of property. The same section then defines that exception narrowly: it does not include ongoing, regular functions of government the suspension of which would not imminently threaten the safety of human life or the protection of property.

Read those together and the outcome for a data release is not ambiguous. An economist cannot finish the index on unpaid time as a favor. A price collector cannot volunteer to visit stores. The prohibition is on accepting the service, not merely on paying for it. This is why a lapse produces a clean stop rather than a slow degradation.

Excepted, Exempt, and Why the Distinction Is Worth Learning

Agency lapse plans use two words that look interchangeable and are not.

• Exempt positions are not financed by the annual appropriation at all. They keep working because the money keeping them working never lapsed.
• Excepted positions are financed by the lapsed appropriation but are retained anyway, because the function is expressly authorized by law, necessarily implied by law, tied to presidential constitutional duties, or necessary to protect life and property.

The Commerce Department's published lapse plan puts numbers on this. Against 42,984 employees on board, it identifies 8,273 excepted — of which 1,810 are financed by non-annual appropriations, 176 are necessarily implied by law, 197 relate to presidential constitutional duties, and 6,090 relate to life and property protection — leaving roughly 34,711 furloughed.

Run a data release against those four categories and it fits none of them. That is the whole explanation for why the statistical calendar is one of the first things to disappear and one of the last to be restored.

The One Release With a Statute Behind It

The Labor Department's contingency plan is unusually blunt about the Bureau of Labor Statistics. Of 2,055 employees on board, 1 is retained during a lapse and 2,054 are furloughed. The plan states that Economic data that are scheduled to be released during the lapse will not be released, with the exception of the September CPI, and that All active data collection activities for BLS surveys will cease.

The exception is worth reading closely, because it explains the general rule. The plan provides that BLS will intermittently except the minimum staff necessary to produce and release the September Consumer Price Index if the Administration determines that failing to release it would significantly interfere with the Social Security Administration's ability to meet statutory deadlines for benefit payments.

That deadline is real and dated. Under 42 U.S.C. 415(i), the cost-of-living determination must be published in the Federal Register within 45 days after the close of the relevant quarter, with notification to the congressional committees of jurisdiction within 30 days. A benefit calculation with a statutory clock on it reaches back through the appropriations machinery and pulls one release out of the blackout.

The generalizable rule is not comfortable but it is clear. The releases that survive a lapse are the ones another statute has attached a date to — not the ones carrying the most open interest. Market importance is not one of the categories in 31 U.S.C. 1342.

The other two bureaus in the Commerce plan get similarly plain treatment. For the Bureau of Economic Analysis, Most Bureau activities will cease, with only minimally necessary work to secure physical, data and IT resources continuing. For the Census Bureau, Economic Indicators: U.S. monthly economic indicators would not be available, and American Community Survey collection is discontinued.

Count What Is Already on the Calendar

The exposure is computable because the calendars are published roughly a year ahead. Counting dated entries on the BLS Schedule of Selected Releases for 2026 gives 11 releases in October, 9 in November, and 14 in December — 34 for the quarter. Two releases falling on the same date count as two, which is why October's 11 entries sit on 10 distinct dates.

Source: U.S. Bureau of Labor Statistics, Schedule of Selected Releases 2026, monthly calendars for October, November and December 2026.

The same exercise runs on the Bureau of Economic Analysis calendar, which lists the advance estimate of third-quarter 2026 GDP on October 29, 2026 and the second estimate on November 25, 2026. The point of counting is not the total. It is that the entire inventory of what a blackout would remove is knowable before anyone knows whether a blackout happens.

A Funded Agency Can Still Print a Broken Release

Appropriations are enacted in separate bills, so a lapse can be partial. That produces a failure mode that an agency-by-agency checklist misses entirely: a fully funded bureau publishing on time with a hole in the middle of its release.

BLS has documented exactly this situation on its own site. In a lapse where the Department of Labor was funded and other departments were not, BLS noted that if Gross Domestic Product data from the Bureau of Economic Analysis were not available, the productivity release would have only a limited set of data available for the quarterly estimates. Productivity is output per hour. The hours are a Labor Department number; the output is a Commerce Department number. Fund one and not the other and the ratio cannot be computed.

The practical consequence is that the relevant map is a dependency graph, not an agency list. A release survives only if every input to it survives.

Part of the Gap Never Fills In

The most common assumption about a data blackout is that everything arrives late and then normal service resumes. Roughly half of that is right.

Some releases are simply rescheduled, and BLS publishes the revised dates. Following the most recent lapses, the September Employment Situation moved from October 3 to November 20, 2025, and the September Producer Price Index moved from October 16 to November 25, 2025. Those are delays. The information still arrives.

The other half is permanent. BLS states that it did not collect CPI data from October 1, 2025 through November 12, 2025, and that Current Population Survey household data for the October 2025 reference period were not collected and will not be collected retroactively. There was no October 2025 Employment Situation release at all. Census, which conducts the Consumer Expenditure surveys for BLS, did not collect that data in October or November 2025.

Prices observed in a particular month cannot be observed later. That has a long tail. BLS has published that missing October 2025 price data affected the April 2026 rent and owners' equivalent rent indexes, and that the missing expenditure data affects the 2025 annual Consumer Expenditure release, final revisions of the 2025 Chained CPI-U, and the 2027 CPI-U and CPI-W indexes.

For anyone running a model on these series, that distinction matters more than the headline outage. A delayed print is a scheduling problem. A never-collected month is a permanent discontinuity in a series that a backtest will happily interpolate straight through.

The Plumbing That Keeps Running

Markets do not close. What changes is which official processes are available.

• Exchange and rate data continue. The Federal Reserve Board's expenses are met by semiannual assessments levied on the Reserve Banks under 12 U.S.C. 243, and the Board states it is not funded by congressional appropriations. The H.15 selected interest rates release is posted at 4:15 pm each business day.
• Treasury borrowing continues. The Bureau of the Fiscal Service lapse plan retains 2,363 of 2,612 staff, citing an appropriation under 31 U.S.C. 3129 available for borrowing, debt servicing and forecasting responsibilities.
• Filing infrastructure stays up; filing review does not. The SEC's operations plan notes that EDGAR is operated pursuant to a contract and thus will remain fully functional as long as funding for the contractor remains available. Against roughly 3,988 employees, about 405 are excepted and about 180 exempt on carryover funds. Market Watch activities, money market fund surveillance and monitoring of broker-dealers reported in financial distress continue, along with emergency enforcement. What stops is review — the plan states the agency will not review or accelerate the effectiveness of registration statements, approve applications for registration, or issue no-action letters.

The tape runs, the auctions run, and the pipe carrying corporate filings stays open. The scarce good is official information and official sign-off. A setup that depends on a scheduled macro print, a registration going effective, or a rule change being approved is exposed in a way that a setup depending only on price and volume is not.

How the Gap Shows Up Inside a Position

None of this predicts a direction, and it should not be stretched into one. It changes three things that sit upstream of direction.

The date stops being a date. An event-driven plan is normally anchored to a known timestamp — size into it, or stand aside through it. When the timestamp itself becomes conditional, both halves of that plan need a fallback, including the case where the release lands on a rescheduled date with far less attention on it.

The first print back is not a normal print. A release that resumes after a gap can carry more than one reference period, can be missing a reference period entirely, or can arrive with a methodology note attached. Any rule calibrated on the usual surprise distribution is being applied to a differently-shaped object.

Cadence assumptions break quietly. A monthly series that skips a month is not the same object as a monthly series. Anything that differences, seasonally adjusts, or rolls a window over that series inherits the gap without warning.

The political outcome of any given deadline is not forecastable with useful precision, and the price reaction to it is a thin, contested sample. The data outage is neither. It is written down in advance, statute by statute and plan by plan, and it is the part of the event that a process can be built around.

What Would Invalidate This

This frame is narrow on purpose, and several conditions take it off the table.

• An enacted appropriation removes it entirely. If funding is in place, none of the mechanics above engage. A full-year appropriation covering one department also removes that department from the picture even while others lapse.
• The plans are revised. Every staffing figure quoted here comes from a specific published version of an agency plan. Those documents are updated, and the numbers in the version current before any given deadline are the ones that govern. Treat the figures here as the shape of the answer, not as this year's answer.
• The CPI carve-out is conditional, not guaranteed. The Labor plan makes it contingent on an administrative determination about interference with statutory benefit deadlines. It is not a standing promise that any particular September CPI prints.
• The release counts are from a selected calendar. The 34 figure counts dated entries on the BLS Schedule of Selected Releases, which is not the complete universe of BLS outputs. It understates rather than overstates.
• Short lapses may cost nothing. A lapse that falls entirely between scheduled release dates can leave the data calendar untouched. The exposure is a function of the overlap between the gap and the calendar, not of the gap alone.
• It says nothing about price. This is a frame about information availability. A strategy that does not consume scheduled federal data is largely unaffected, and should be treated that way rather than dressed in a macro narrative.

Concrete Framework

1. Pull the two calendars before the deadline, not after. Download the BLS Schedule of Selected Releases and the BEA release schedule for the affected months. Mark every dated entry your process consumes. That marked list, not the count, is the exposure.
2. Read the relevant agency lapse plan rather than coverage of it. The Labor, Commerce, Treasury and SEC plans are published documents and state staffing, what ceases, and what is excepted. Note the version date on the one you read.
3. Sort every input into the three tiers. Outside the annual appropriation; appropriated with a statutory carve-out; appropriated with none. Anything in the third tier should be assumed absent for the duration.
4. Trace one level of dependency for each surviving release. Ask what upstream number it requires and which department funds that number. Productivity needing GDP is the template for this failure.
5. Separate delayed from destroyed in your own notes. For each affected series, write down whether the data can be collected later. Rent, prices and household survey responses generally cannot. That determines whether the model needs a patched value or a flagged discontinuity.
6. Decide the fallback for a moving date in advance. Write the rule for what happens if the print does not arrive, and the separate rule for what happens when it arrives on a rescheduled date. Both should be written while the calendar is still normal.
7. Size the first post-gap release as an outlier, not as a regular event. Multi-period content, missing periods and methodology notes all argue for treating it as a different distribution until proven otherwise.
8. Flag every affected observation in stored series. A gap that is not marked in the data will be silently smoothed by the next backtest that touches it. Marking it is cheap now and impossible to reconstruct later.
9. Note which non-appropriated sources you can lean on. H.15, Treasury auction results and exchange data continue. If a decision must be made during a blackout, it will be made on those inputs, so know in advance whether they are sufficient.