How to Split a Bond-Yield Move Into Real Rate and Inflation — the TIPS Breakeven Check Nobody Runs on "Oil Is Pushing Yields Up" Headlines — Field Guide #39

Every time oil spikes or a central banker turns hawkish, the same sentence appears within the hour: yields rose, which is bad for bitcoin. It is a sentence with two numbers hidden inside it, and they push on bitcoin in opposite directions. A Treasury yield is the sum of a real interest rate and an inflation compensation, and the US Treasury publishes both halves every business day, for free, in two CSV files. Splitting a yield move into its real and inflation components takes about four minutes and tells you whether the bond market just raised the cost of holding a zero-yield asset, or just raised its estimate of how much the dollar will be debased. Those are not the same news for bitcoin. This guide shows how to run the split, works it against the week of 27 August to 1 September 2026 — in which one day was almost entirely real rates and another almost entirely inflation — and marks the four ways the method misleads.

Why the split matters for a zero-yield asset

Bitcoin pays nothing. The cost of holding it, relative to the safest alternative, is the yield you give up — and the yield that matters is the real one, because inflation erodes the Treasury's coupon as surely as it erodes a dollar under a mattress. When the real yield rises, every asset that produces no cash flow becomes relatively more expensive to hold: gold, growth equities, bitcoin. That is the channel through which 2022 happened, and it is the channel a hawkish Fed operates directly. The inflation-compensation half, the breakeven, is the market's estimate of average inflation over the bond's life. A rising breakeven with a flat real yield is the bond market saying the dollar will buy less without saying holding cash pays more. On paper that is the environment a fixed-supply asset was designed for. In practice bitcoin's response to it is inconsistent, which is precisely why you want to know which kind of day you are looking at before you write the sentence.

The 27 August to 1 September 2026 window contains one of each. On Friday 28 August the ten-year nominal yield rose 6bp after Kevin Warsh's Jackson Hole keynote; on Tuesday 1 September it rose 4bp after US-Iran strikes sent Brent above $95 and Japan's ten-year through 3%. Read as nominal moves they are the same story twice. Read as components they are opposites, and we get to that below.

Where the two series live

The US Treasury publishes daily par yield curve rates (nominal) and daily real yield curve rates (derived from Treasury Inflation-Protected Securities) on its Resource Center. Both are downloadable as CSV by year. The nominal file has fourteen tenors from 1-month to 30-year; the real file has five: 5-, 7-, 10-, 20- and 30-year. The URLs are stable and the files are updated each business day in the late afternoon Eastern time, which is the first trap: the row for today does not exist until after the close, so anything labelled "today's real yield" during the US session is somebody's live quote, not the Treasury's number. The ten-year guide in this series covers reading the nominal file; this one adds the second file and the subtraction.

• • Nominal: home.treasury.gov/resource-center/data-chart-center/interest-rates/daily-treasury-rates.csv/2026/all?type=daily_treasury_yield_curve&field_tdr_date_value=2026&page&_format=csv
  • Real: the same URL with type=daily_treasury_real_yield_curve
  • Change the year twice to pull history; the 2023–2026 files together are 917 nominal rows and 917 real rows, one per business day.

Check 1 — Subtract, per tenor, per day

Breakeven inflation at a tenor is nominal minus real at that tenor. Do it for each day you care about, then take day-on-day differences of all three. The bookkeeping has to close: the change in nominal must equal the change in real plus the change in breakeven, to the basis point, because the third number is defined as the difference of the first two. If it does not close you have mixed dates or tenors. Here is the pull for the ten-year, four rows, one per business day from the close before Warsh to Tuesday, plus the window total:

Friday was a real-rate day: the real yield rose 8bp and the breakeven fell 2bp. A Fed chair saying policy may not yet be restrictive raises the expected path of real policy rates and, if anything, lowers expected inflation, because the market believes he means it. Tuesday was an inflation day: the real yield did not move and the breakeven rose 4bp. That is what an oil shock does to the curve when the central bank is not expected to fully offset it. The window as a whole reads as a real-rate move (+10 of +12) only because Friday was the biggest day in it. Two sentences that both begin "yields rose" were describing different events.

Check 2 — Run it at two tenors, because oil lives at the front

Inflation shocks from commodities show up more at short maturities than long ones, because the market assumes the central bank eventually wins. The real file's shortest tenor is five years, so the five-year is the front of this particular curve. Same four days:

Tuesday's breakeven rose 6bp at five years, 4bp at ten and 3bp at thirty: the oil shock is priced most heavily where it will be felt first and fades along the curve. Friday's real-rate move was +11bp at five years, +8bp at ten and +4bp at thirty: a policy shock is also a front-loaded shock, but on the other component. If you only pull the ten-year you will see both days as "about the same size" and miss that they were different in kind and different in shape. Pull the five and the thirty every time; it is two more columns in files you already have open.

Check 3 — Compare against the asset's own closes, in the same clock

The point of the exercise is to attribute an asset move, so the asset's closes have to sit next to the yields on the same calendar. Treasury CMT rates are end-of-day New York; the nearest bitcoin equivalent is the daily UTC candle, which closes at 20:00 ET, three and a half hours after the bond market. That is close enough for daily attribution and not close enough for intraday claims. Bitstamp daily closes for the same window: 27 August $80,278.98; 28 August $77,835.18 (−3.04%); 31 August $78,571.17 (+1.12%); 1 September $77,397.43 (−1.49%). Bitcoin fell on the real-rate day and fell on the inflation day; it fell twice as far on the real-rate day. Gold, via the PAXG token's Binance daily candle: 28 August −2.72%, 1 September −2.54%. Gold fell about the same amount on both. That is a real difference between the two assets' sensitivities inside one week, and you cannot see it without the decomposition. The event-attribution guide covers the discipline of not over-reading a single pair of days.

Check 4 — Four ways the split misleads

Trap one: the breakeven is not a pure inflation forecast. It contains an inflation risk premium (investors pay for protection) and, in the other direction, a liquidity discount (TIPS trade less than nominals, so their yields are a little higher and breakevens a little lower than the "true" number). In calm markets these roughly offset; in stressed ones the liquidity term can swing breakevens by tens of basis points with no change in anyone's inflation view. March 2020 is the canonical example. The test is whether the move persists over a week; a one-day breakeven collapse in a liquidation is usually the liquidity term.

Trap two: neither series isolates term premium. The real yield itself contains the expected path of real short rates plus a premium for holding duration. When the long end sells off on fiscal fear — a record Japanese budget request, a US buyback that does not work — the real thirty-year rises without any change in expected Fed policy. The long-end guide covers how to see that; the short version is that a real-yield rise concentrated at thirty years and absent at five is term premium, and a real-yield rise concentrated at five and fading at thirty is policy. Friday's +11/+8/+4 was policy.

Trap three: the real curve starts at five years. There is no two-year real yield in the Treasury's file, so you cannot decompose the tenor that moves most on a Fed meeting. Two-year breakevens exist as market quotes from dealers and data vendors, but they are not a primary source you can re-pull, and this desk does not print numbers it cannot reproduce. Say "the five-year" when you mean the five-year.

Trap four: "highest since" needs the series named beside it. Tuesday's 4.79% ten-year was described in two morning pieces as the highest since 2023. On the Treasury's daily par series the last close at or above 4.79% was 13 January 2025 — twenty months, not three years — and the 2023 high on the same series was 4.98% on 19 October 2023. The 2023 comparison came from an intraday feed. Intraday feeds and daily par series are both legitimate; a superlative built on one and checked against the other is not.

Check 5 — Turn it into a marker, not a narrative

The decomposition gives you a claim you can be wrong about. If you think this selloff is a real-rate event, the falsifiable version is "the ten-year real yield closes above its 2026 high of 2.47% before the FOMC"; if you think it is an inflation event, it is "the ten-year breakeven closes above 2.40% before the FOMC" (Tuesday: 2.35%). Pick one, write down the settlement source and date, and grade it in public. This desk's marker-design guide is the method; today's Markers board carries the first of those two as I1.

The five checks, in order

• • Pull both files (nominal and real par curves) for the years you need; confirm row counts match and that today's row exists before quoting it.
  • Subtract per tenor per day; confirm each day's nominal change equals real change plus breakeven change, to the basis point.
  • Run it at five, ten and thirty years; a front-loaded real move is policy, a front-loaded breakeven move is a commodity shock, a long-end real move with a quiet front is term premium.
  • Put the asset's daily closes beside it on a stated clock, and compare the size of the asset's move on real-rate days to inflation days rather than to "yields rose" days.
  • Name the four traps in print if any applies: liquidity premium in stressed markets, term premium at the long end, no real two-year, and "highest since" claims built on a different series than the one you checked.

Field Guide #39 joins the Reading Room under the Fed and rates cluster, next to the ten-year and long-end guides it extends. It was written on the morning the method was used for today's lead story, and the tables above are the same pull.

Method: bitcoin prices are Bitstamp BTC/USD daily candles (UTC) and Binance BTCUSDT; open interest, funding and long/short ratios are Binance USDT-M and COIN-M public endpoints; Treasury yields are the US Treasury's daily par nominal and real constant-maturity series; ETF flows are Farside Investors' completed daily table; gold is the PAX Gold (PAXG) daily candle on Binance as a 24-hour proxy and CME gold futures via Yahoo Finance; equities, oil and the dollar index are Yahoo Finance quotes. Every figure labelled with today's date was snapshotted at 06:11 UTC on Wednesday 2 September 2026 unless a different clock is printed beside it. Windowed pulls carry their row counts.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Cryptocurrencies are volatile and you can lose money. Nothing here is a recommendation to buy or sell any security, token or exchange-traded fund. Do your own research and consult a licensed financial advisor before making investment decisions.