For most of 2025, the government's scoreboard said the economy added 584,000 jobs. The final count was 181,000. Next week, the Federal Reserve will set interest rates on statistics built the same way, by a chairman who calls the inflation number a scientific wild guess. There is one ledger in the world that has never been revised.

» Kevin Warsh's proposed inflation measure now reads a full point below core PCE, 2.28% against 3.34%. The chairman wants a shorter ruler the week before a live rate decision.

» The statistics are disappearing. GDP was halved on revision, nearly a fifth of CPI quotes went uncollected last July, and the USDA has stopped counting household hunger.

» The bitcoin ledger has never been revised, adjusted, or restated. Anyone can audit the full supply from a laptop.

The Broken Scoreboard

The numbers that run the monetary system are provisional, politically pressured, and quietly rewritten. One ledger in the world has never been revised.

On August 28, buried under coverage of the Federal Reserve chair's first Jackson Hole speech, the Bureau of Labor Statistics published its preliminary benchmark revision: the jobs count through March 2026 will come down by another 79,000, against the 183,000 upward revision economists had expected. It is the seventh downward benchmark in eight years. The original prints moved markets, but the correction barely made the wires.

The 2025 payroll data has now been through the full revision cycle, initial prints, a preliminary benchmark, and a final benchmark. Through 2025, the establishment survey told the country that employers had added 584,000 jobs. The corrections came in two stages. The preliminary benchmark published last September erased 911,000 positions from the twelve months through March 2025, overstatement that had accumulated mostly during 2024. The monthly prints were then marked down through the rest of the year, and when the final benchmark landed, calendar 2025 had produced 181,000.

For most of a year, economic policy, electoral politics, and market pricing all ran on a labor market roughly three times stronger than the one that existed. The figure moved rate expectations, Treasury yields, and the dollar in real time, and was then withdrawn and replaced.

It would be convenient to treat this as one agency's bad run. The record describes something wider: the operating condition of every statistic the monetary system runs on. The prints are provisional estimates, assembled from models, adjusted by assumption, published under structural pressure to err in one direction, and revised after the consequences have been absorbed. The timing sharpens the point. The Federal Reserve meets September 15 and 16 to decide whether to raise interest rates, and it will make that decision on data its own chairman has publicly described as unworthy of the task.

Seven of the Last Eight

Errors of this size do not happen by accident. They are produced, and the production process is public. Three mechanisms account for most of it.

The first is the birth-death model, which imputes jobs from business formation the survey cannot yet observe. It performs acceptably in stable expansions and fails at exactly the moments that matter, when the economy turns. In recent years it has contributed hundreds of thousands of model-generated jobs that later benchmark revisions quietly strip out.

The second is seasonal adjustment, which assumes the labor market still follows patterns that broke after 2020; the Dallas Fed has shown the adjustment factors now compound distortions rather than correct them.

The third is the benchmark revision itself, the annual true-up against unemployment insurance tax records that restates history a year late. Seven of the last eight have been downward, with 2022 the single exception in the other direction. That is what a tilted instrument looks like.

In August 2025, hours after a weak report, the president fired the BLS commissioner and called the numbers rigged. William Beach, a former commissioner, once noted that trading desks in Dusseldorf stop trading in the minutes before the 8:30 release because of how much rides on it.

This past July, payrolls printed at minus 23,000 and were revised to plus 21,000 within four weeks, a swing that moved September hike odds by fourteen percentage points before being withdrawn. A statistic that reprices monetary policy, gets restated within weeks, and gets its publisher fired has become an instrument of policy in its own right.

The Ruler Shrinks

The consumer price index carries a longer pedigree of adjustment. Each change was defensible in isolation. All of them lowered the number.

The habit began in the 1970s, when Arthur Burns's Fed, battered by food and energy shocks, popularized stripping those categories to produce core inflation, a measure defined by removing whatever was rising fastest. In 1983, the BLS replaced the actual cost of homeownership, house prices and mortgage rates, with owners' equivalent rent, an imputed figure no household has ever paid. Shelter, roughly a third of the index, has been measured by survey estimate rather than market price ever since.

In 1996, the Boskin Commission declared that the CPI overstated inflation by roughly 1.1 percentage points a year, and the remedies that followed did the rest: geometric weighting that assumes consumers substitute away from whatever gets more expensive, hedonic adjustments that reclassify price increases as quality improvements, and the chained index introduced in 2002.

The incentive sits in plain sight. Social Security cost-of-living adjustments, federal tax brackets, TIPS payments, and union contracts all index to CPI, so every downward methodological tweak lowers the government's obligations automatically. The Boskin-era fixes were understood at the time to be worth hundreds of billions of dollars in reduced outlays and higher bracket revenue over the following decade, achieved without a single vote to cut benefits or raise taxes.

The newest chapter arrived this spring. At his April confirmation hearing, Kevin Warsh dismissed the Fed's preferred gauge as offering only a rough take, proposed building an inflation measure from a billion prices collected from the private sector, and endorsed trimmed-average measures that discard whichever prices moved most in a given month. He has since called core PCE a scientific wild guess and stood up a data task force to examine the quality, freshness, and accuracy of federal statistics from first principles.

Some of that scrutiny is overdue. The direction of the proposed correction still deserves attention. When Warsh made the case, the Dallas Fed's trimmed mean sat at 2.3% while core PCE printed 3.0%. The gap has since widened past a full percentage point, with the latest prints at 2.28% and 3.34% respectively: the proposed ruler now reads more than a point below the one in use. Loretta Mester, hardly an inflation dove, has warned that trimmed measures carry a downward bias, and Evercore's Krishna Guha asked the operative question: whether the chairman was shifting the goalposts or making a reasonable economic argument.

A referee who proposes a new yardstick that happens to read lower is making policy through measurement. Five years above target, the institutional answer has been a different ruler rather than a different course. Every debasing government in history has eventually shortened the ruler. None has ever lengthened it.

When the Counting Stops

Worse, the data has started to disappear. Fourth-quarter 2025 GDP was halved on revision, from 1.4% to 0.7%, in a release that arrived late because of the government shutdown. Survey response rates across the statistical agencies have been collapsing for a decade. In July 2025, nearly 20% of the price quotes that make up the CPI went uncollected for lack of staff and budget. The USDA has discontinued its annual food insecurity survey, so the country will soon stop measuring household hunger entirely.

Three Precedents

The pattern extends well beyond Washington and well beyond the present decade, and it is not confined to governments.

Greece entered the euro on reported deficits near 4% of GDP; the 2009 admission that the true figure was above 12%, later restated beyond 15%, detonated a continental crisis, and the swaps that kept the liabilities off the books had been structured for precisely that purpose.

Argentina's government intervened at its statistics agency in 2007, replaced the statisticians, fined private economists for publishing honest estimates, and shaved billions off its inflation-indexed debt payments before the IMF issued its first-ever censure of a member country over data quality in 2013.

The private sector's entry came from London. From 1986 until 2023, the interest rate beneath hundreds of trillions of dollars in mortgages, corporate loans, and derivatives was Libor, a number compiled each morning from a poll of major banks asked, on their honor, what it would cost them to borrow. The banks' own traders were positioned against the answer, and for years they nudged submissions to benefit their books, then lowballed them during the 2008 crisis to make their institutions look solvent. The affair produced some nine billion dollars in fines, prison sentences, and the abolition of the benchmark itself.

Each number was asserted by an interested party, and the people who relied on it had no way to verify it independently. That combination, consequential and unverifiable, is the whole mechanism. Numbers that matter and cannot be checked drift toward the interests of their publishers, on every continent that has tried it.

Seventeen Years Without a Revision

One financial record in the world works differently.

The bitcoin ledger has never published a preliminary figure. It has never issued a benchmark revision, never been seasonally adjusted, and never had its publisher fired, because it has no publisher. Roughly every ten minutes the network settles a new block, and every full node independently recomputes the entire history of every transaction back to January 2009 and rejects anything that breaks the rules.

The issuance schedule is written in code that has executed identically for seventeen years. Anyone can audit the complete outstanding supply from a laptop this afternoon, without permission, and arrive at exactly the same answer as everyone else who checks.

Everything else in finance is a claim about reality, published by an institution with a stake in the claim. The gap between claim and reality is where payroll revisions, Boskin commissions, Greek deficit swaps, and Libor submissions did their work. Bitcoin's ledger removes the gap: the record is the reality, re-verified every ten minutes by thousands of independent parties who share nothing except the mathematics.

Next week, the Federal Open Market Committee will decide whether to raise interest rates. It will weigh payroll data that will be restated, GDP that was halved on revision, and an inflation framework its own chairman calls a scientific wild guess, and the decision will be received as sober technocracy because the process wears the costume of measurement. Bitcoin settled another block while you read this. The scoreboard is broken. The ledger is not.

Closing Note

Onramp provides bitcoin financial services built on multi-institution custody. To learn more about our products for individuals and institutions, schedule a consultation to chat with us about your situation and needs.

Until next week,

Brian Cubellis