Financial Services Furnished Without Payment

FROM: Office of Employment Statistics, Division of Current Employment TO: All CES Staff RE: First Friday Protocol — March 2026

Embargo holds until 8:30 AM EST. Pre-cleared personnel report to Room 3124 no later than 7:45. Cell phones secured. No external communication until public release. Remind your teams: the number is the number. We do not editorialize. We do not contextualize. We present.

Key Deliverables:

• Total nonfarm payroll (seasonally adjusted)
• Unemployment rate (U-3)
• Labor force participation rate
• Average hourly earnings

Questions re: seasonal adjustment methodology should be directed to Seasonal Factors, not Communications.

The first thing to understand about being dead in a federal building is that nobody has adjusted the model, which is to say there is no field for it, no checkbox, no variable in the regression that would allow the seasonal adjustment algorithm to account for the continued presence of a person who stopped drawing salary in October 1991, who stopped drawing breath slightly before that, in August, on a Tuesday, face down in a plate of chicken piccata at a restaurant on K Street that no longer exists and whose coordinates are now occupied by a Sweetgreen, there is no survey instrument that captures this category of being, a person who remains in the building because leaving requires knowing where to go and the dead it turns out do not know where to go any more than the living do, they simply have more time to not know it in.

I have been here for four hundred and fourteen first Fridays.

Not continuously. I am sometimes near the vending machines that replaced the vending machines that replaced the vending machines I used in 1974, and once for an entire fiscal year I was in a storage room on the second floor reading decommissioned procedural manuals from the Eisenhower era, which were wrong about nearly everything in ways that felt familiar. But on first Fridays I am in Room 3124, because on first Fridays the number comes out.

Anya Kowalski — Personal Notes (Do Not Circulate)

Running diagnostics on birth/death model AGAIN. Something off in the small-business imputation. The model assumes new firms are born at a rate derived from previous quarters, but the denominator keeps shifting because gig classifications changed in Q3 and nobody updated the crosswalk. Asked Jeff about it. Jeff said the crosswalk is “in process.” The crosswalk has been “in process” since I started here.

Sometimes I think the birth/death model is doing most of the creative work and we’re just watching.

Headphones crackling again. Static on the left channel. IT says nothing’s wrong with them.

Anya Kowalski runs the seasonal adjustment models and she is twenty-eight years old and she does not know I am standing behind her, which is accurate in the sense that I am standing behind her and she does not perceive me, but inaccurate in the deeper sense that she is the only person in the building who almost does, because when I stand close enough to read her screen there is a loosening in the way she types, her fingers slow and the rhythm changes, she pauses mid-keystroke and looks at the ceiling and then writes something in her personal notes that is closer to the truth than anything in the official methodology, and I do not claim this is my influence because that would be the kind of claim I spent my career trying to eliminate from economic analysis, the post hoc fallacy, the narrative imposed on correlation, but I will say that when I am near her the static in her left earbud gets louder and she takes the earbuds out and in the silence that follows she does her best work.

Today she is looking at a field called FINANCIAL SERVICES FURNISHED WITHOUT PAYMENT.

Gross Domestic Product: Methodology Note, Section 12.4.c

Financial services furnished without payment (FISIM): When a depositor places funds in an interest-bearing account and receives a rate of return below the reference rate, the difference constitutes a financial service provided by the depositor to the institution. This imputed service is counted as both output (of the banking sector) and consumption (by the depositor). The depositor receives a financial service by receiving less money.

She has highlighted this passage on her second monitor. She is not supposed to be looking at GDP methodology; her job is employment, not output. But Anya Kowalski has the kind of mind that follows a thread from one spreadsheet to the next, and she has arrived at the sentence that I arrived at in 1987 and spent the last four years of my life trying to say something about, which is that the entire system rests on an accounting identity in which receiving less is classified as receiving a service.

In 1987 I wrote a paper about this. Seventeen pages. The referee said I had confused a methodological convention with a philosophical claim, which was true in the sense that I had done exactly that, and which was also the point.

FROM: Jeff Halloran, Branch Chief TO: Anya Kowalski RE: re: re: re: Birth/Death Model Crosswalk

Anya,

Per our conversation — I understand your concerns re: the crosswalk but want to reiterate that the current methodology has been peer-reviewed and approved at the division level. Your job is to run the models, not to redesign them. I say this supportively.

Also: please do not CC the Commissioner’s office on methodology questions. That’s not the channel.

Jeff

Sent from my iPhone

Jeff Halloran has been branch chief for eleven years and he is not a bad person, this is important to say because the system does not require bad people, it requires people who understand their role in the production of the number and who execute that role with professional detachment, Jeff Halloran runs his branch, he meets his deadlines, he does not ask what the number means because meaning is not his deliverable, his deliverable is the number.

I knew men like Jeff Halloran. I was, in certain respects, a man like Jeff Halloran. In 1968 I helped redesign the Current Population Survey and we made a decision — it was a Tuesday, I remember the coffee was terrible, the kind of memory that persists when the important ones have faded — we made a decision to exclude unpaid domestic labor from the labor force definition. A woman who cooked and cleaned and raised children and managed a household budget and negotiated with landlords was, by our definition, not in the labor force. Not employed. Not unemployed. Not even not-in-the-labor-force in a way that registered as a gap, because the category was designed to contain her so completely that her absence from the count looked like the count working correctly.

I did not notice this at the time. I noticed it in 1983 when a colleague’s wife said something at a dinner party. By then the survey had been running for fifteen years on the architecture I helped build. The field was the problem. The field had always been the problem. But the field was also the foundation, and you cannot remove a foundation while people are living in the building.

CURRENT EMPLOYMENT STATISTICS — PRELIMINARY RELEASE (EMBARGOED) Bureau of Labor Statistics March 6, 2026

Total nonfarm payroll employment increased by 187,000 in February.

The unemployment rate was unchanged at 4.1 percent.

Notable job gains occurred in health care (+52,000), government (+33,000), and financial activities (+28,000).

Note: Figures are seasonally adjusted. The birth/death model contributed an estimated +112,000 to the total. See Technical Note for methodology.

One hundred and twelve thousand of the one hundred and eighty-seven thousand jobs in this report are imputed. They are not observed. No employer reported them. No worker confirmed them. They are the model’s estimate of jobs that should exist at firms too new or too small to appear in the survey, minus jobs lost at firms that died before the survey reached them. The birth/death model is, in this sense, a ghost story: it counts the workers that the measurement cannot see by assuming they are there in the proportions that previous quarters suggest they should be.

And they might be there. Or they might not. The model will be revised in the benchmark, eleven months from now, when the unemployment insurance tax records arrive and the actual count is known, and the revision will be plus or minus three hundred thousand, and nobody will lead the evening news with the revision because the revision is not the number, the number was the number, the number moved markets and justified policy decisions and was, by the time of its correction, historical fact in the only sense that matters, which is that people acted on it.

Anya Kowalski — Personal Notes (Do Not Circulate)

112K from birth/death. That’s 60% of the headline. Sixty percent imputed. And the imputation model uses a crosswalk that hasn’t been updated for the gig reclassification. So the 112K assumes a firm-creation rate derived from an economy that categorized workers differently than the economy we’re measuring.

We are measuring 2026 with a ruler calibrated in 2019.

I keep thinking about FISIM. The depositor receives a financial service by receiving less money. What if the whole report works like that. What if the country receives an employment statistic by not counting the people who don’t fit the fields.

Static in the left channel again. Loud this time. Almost like a voice. Not a voice. Obviously not a voice. But a pressure. Like someone leaning over my shoulder.

She is so close and she will not get there, not because she lacks intelligence but because the thing she is approaching is the thing the institution was designed to make unapproachable, not by conspiracy, not by anyone’s intention, but by four decades of methodological decisions each of which made sense in isolation and together constitute an architecture of not-seeing so complete that seeing through it feels like madness and is, in institutional terms, a career-ending event.

I got close too. In 1989 I wrote a memo to the Commissioner and the memo was received politely and filed and I understood that filing was the final form of reading, that the institution had read my memo by putting it in a folder where it would never be read again, and this was not suppression, it was something worse, it was the system functioning as designed, because a system that could absorb criticism without changing was not broken, it was robust.

FROM: Communications Division TO: All Embargoed Personnel RE: Release Protocol

Embargo lifts in 3 minutes. Bloomberg terminals are set. CNBC is live. The Secretary has been briefed.

Remember: we present the number. The number speaks for itself.

The number does not speak for itself. Nothing has ever spoken for itself. Speaking requires a mouth and a body and a position in the world from which to speak and the number has none of these, it is produced by an institution and released by a protocol and received by markets and it speaks in the voice of no one in particular which is the most authoritative voice there is.

Anya Kowalski is looking at her screen. In two minutes the number will go out and algorithms will parse the headline figure in milliseconds, the ten-year Treasury will move, someone on television will say the labor market remains resilient or the labor market is softening and both statements will derive from the same number and both will be correct, because the number is capacious enough to contain any interpretation that does not question the number itself.

She closes her personal notes. She does not send them to Jeff Halloran. She does not CC the Commissioner. She saves the file to a folder on her desktop called METHODOLOGY QUESTIONS and the folder has forty-seven documents in it and none of them have been sent to anyone because Anya Kowalski has learned, in three years, what it took me twenty to learn, which is that the institution does not refuse criticism, it absorbs it, and absorption is more effective than refusal because refusal implies that the criticism matters enough to oppose.

MARKET REACTION — 8:31 AM EST

S&P futures +0.3%. Ten-year yield down 2 bps. Dollar steady. Gold flat.

Consensus: In line with expectations. No revision to rate path.

CNBC chyron: JOBS REPORT SOLID — ECONOMY ON TRACK

The static in her left earbud stops. I have moved away from her. Not because I chose to but because the number is coming out and when the number comes out I go to the window, this is my ritual, and I watch the city receive the number, which is to say I watch nothing happen, because the number’s reception is invisible, it happens in servers and trading floors and none of it is visible from the third floor of a federal building on Massachusetts Avenue, but I watch anyway, I watch the traffic and the pedestrians and I think about the ones the number does not count, the ones working off the books, the ones whose labor produces no tax record, the ones who are not in the labor force by a definition I helped write, and I count them, silently, one by one, which is not measurement, which produces no data, which feeds no model, which is just an old dead man at a window counting the uncounted because counting without an instrument is the only kind of counting that does not, in the act of counting, determine in advance what can be counted.

Forty-seven documents in that folder. She will write the forty-eighth next month. She will not send it.