Breakeven inflation: what the spread is, and what it is not

A breakeven inflation rate — BEI, or breakeven — is the difference between the yield on a nominal bond and the yield on an inflation-linked bond of the same maturity from the same issuer. It is a single number in basis points, computed from two prices, and it answers one question: what rate of inflation makes the two bonds produce the same return over the period. Nothing more.

The name is the problem. “Breakeven” sounds like a threshold someone expects to be crossed; “inflation” sounds like a forecast of prices. Put together, the phrase invites the reading that a 2.65 per cent 5-year breakeven is a prediction of 2.65 per cent inflation, and that reading is wrong in a way that survives because the number is so often quoted without its two legs attached. A breakeven is a price of a hedge. It contains expectations about inflation, and it also contains an inflation risk premium, a liquidity premium, and whatever technical pressure the two bonds happen to be under — and no arithmetic on the two prices separates those four components from one another.

What follows is the computation and its limits: how the spread is derived and why the exact form differs from the subtraction most notes use, a five-tenor table worked to a tenth of a basis point, a forward breakeven built from four inputs where the naive shortcut fails by 268 basis points, the four things the number contains and cannot attribute, and the reporting standard that keeps a breakeven from becoming a forecast by accident.

The spread, and the two ways to compute it

Take a nominal bond and an inflation-linked bond of the same maturity, from the same issuer, both trading at a real price on the same day. The nominal yield is the return in currency. The real yield is the return in purchasing power, because the linker’s principal and coupons are scaled by an index. The difference is the rate of indexation that equates them, and the exact form of that statement is a ratio rather than a subtraction:

(1 + nominal) ÷ (1 + real) − 1.

The subtraction most notes use — nominal minus real, in percentage points — is the first-order approximation of that expression, and it is a good one at low inflation. It is not exact, and at the levels in Table 1 it overstates the breakeven by between 3.8 and 4.7 basis points, because the approximation drops the cross term: real yield multiplied by the breakeven. At 1.44 per cent real and 2.65 per cent breakeven, the cross term is 3.8 basis points — wider than the bid-offer on a liquid linker, and wider than most of the moves a monthly note is written to interpret.

That is not a pedantic distinction. A breakeven quoted to two decimal places in percentage points has a resolution of one basis point, and the choice of method moves the answer by four of them. A note that says “the 5-year breakeven is 2.65 per cent” is quoting the approximation; the exact figure is 2.6124 per cent. Both are defensible if the method is stated. Neither is defensible as an unqualified number, because a reader comparing two notes that used different methods is comparing methods rather than markets.

Table 1. An illustrative nominal and real curve at five tenors, the breakeven computed both ways, and the size of the approximation error. Source: constructed for this piece; illustrative, not a market observation. The exact breakeven is (1 + nominal) ÷ (1 + real) − 1.

Two patterns in the table are worth reading before any interpretation. The overstatement tracks the product of the real yield and the breakeven, which is why it runs from 3.8 basis points at the 5-year to 4.7 at the 30-year — both inputs are larger further out. The two-year is the exception that proves the rule: its real yield is the highest in the table, so its cross term of 4.1 basis points exceeds the five-year’s even though its breakeven is the smallest. And the breakeven curve is upward sloping — 251.9 at two years to 278.3 at thirty — for reasons that the arithmetic cannot identify. A rising breakeven curve is consistent with expectations of rising inflation, with a rise in the inflation risk premium with maturity, with a liquidity premium on the linkers that is smaller at the long end where issuance is larger, and with the interaction of all three. It is not evidence for any one of them.

The forward breakeven, and the shortcut that fails

A spot breakeven is a statement about a period ending at the bond’s maturity. The quantity most analysis actually wants is a forward: the breakeven over a future window, priced today. It is computed like any other forward, from four inputs — the nominal and real yields at the start and the end of the window — and it is worth doing carefully, because the obvious shortcut produces a number that is wrong by an order of magnitude.

The shortcut is to subtract the near breakeven from the far one. On Table 1 that gives 267.9 − 261.2 = 6.7 basis points for the five years starting in five years, which is absurd on its face: it would mean the curve prices almost no inflation for half a decade, in a market where every spot breakeven is above 250. The error is that a breakeven is a rate, not a price, and rates do not add across horizons — the compounding is multiplicative and the weights differ at each horizon.

The correct route builds two forwards and differences them at the end. The 5-year nominal forward starting in five years is the rate at which a 5-year nominal investment rolled for another five compounds to the 10-year nominal yield: ((1.0424)^10 ÷ (1.0409)^5)^(1/5) − 1 = 4.3902 per cent. The 5-year real forward is the same computation on the real curve: ((1.0152)^10 ÷ (1.0144)^5)^(1/5) − 1 = 1.6001 per cent. The forward breakeven is then the exact ratio of those two forwards: (1.043902 ÷ 1.016001) − 1 = 2.7462 per cent, or 274.6 basis points.

Table 2. A 5-year breakeven starting in five years, built from four inputs and set against the shortcut. Each line is a computation from Table 1; the final row is the answer. Source: constructed from Table 1.

The gap between 274.6 and 6.7 basis points is not a rounding question. It is the difference between a quantity that can be defended line by line and one that cannot be defended at all, and it appears whenever a rate is differenced across horizons rather than compounded. The same error, in smaller form, appears in the subtraction of a spot breakeven from a forward one, and in the subtraction of breakevens quoted on different indexation bases.

The forward is also where the measurement error concentrates. A 2 basis point error in the 10-year real yield — a plausible size for a quote read off a screen in a thin market — moves the 5-year real forward from 1.6001 to 1.6401 per cent, and the forward breakeven from 274.6 to 270.6 basis points. Two basis points of input becomes four of output, because the forward is a ratio of two compounded quantities and the error is amplified by the ten-year exponent before the fifth root takes it back. A spot breakeven dilutes measurement error; a forward breakeven concentrates it.

What the number contains

Here is the honest core of the matter, and it is the part that makes the number usable only if it is stated: a breakeven contains at least four things, and the two prices cannot be made to reveal their proportions.

Expected inflation. The component everyone quotes and no one can isolate. It is the market’s central expectation of indexation over the bond’s life, in the sense of the rate at which a holder is indifferent — which is itself a statement about a distribution, not a point.

An inflation risk premium. A holder of a nominal bond bears the risk that inflation turns out higher than priced; a holder of a linker does not. The nominal yield therefore carries a premium for that risk and the linker does not, so the breakeven contains it. Its sign and size are estimated, not observed, and the estimates move with the model.

A liquidity premium, in the opposite direction. Linkers are typically less liquid than the nominal benchmark of the same maturity: fewer issues, wider bid-offer, a smaller holder base. Illiquidity raises the yield a holder demands, so a less liquid linker has a higher real yield and therefore a lower breakeven. This component can offset the inflation risk premium in whole or in part, and its size changes with issuance and with the balance-sheet cost of holding the bond.

Technicals. Index inclusion, pension demand for inflation protection, the supply calendar, and the behaviour of the two bonds’ own repo markets. A large linker auction in the week a breakeven is measured moves the number, and the move is a fact about the auction rather than about prices.

There are three further measurement facts. The indexation lag: a linker’s cash flows are scaled by an index published with a delay of around three months, so the real yield refers to a lagged index and the breakeven is not a clean forward expectation over the bond’s remaining life — the effect is small at the long end and material at the short. The deflation floor: most linkers pay at least par at maturity, which embeds an option that is worthless at high inflation and valuable near zero, so a breakeven computed near zero is biased upward by the value of that floor. And the pairing itself: a breakeven requires a nominal and a real bond of the same maturity and the same issuer, and where the two do not exist the curve is interpolated, which imports every assumption in the interpolation into the spread.

None of that makes the breakeven unusable. It makes it a price rather than a measurement of expectations, and the correct sentence is that the curve prices 274.6 basis points of indexation for the five years beginning in five years. Whether that is high or low relative to what inflation will be is a question the number cannot answer, because the answer requires knowing the risk premium, the liquidity premium and the technicals — and those are the three things the number was supposed to tell you.

How a breakeven is misread, in one example

A worked misreading is worth more than a warning. Suppose the 10-year breakeven rises 8 basis points over a month while the nominal 10-year yield is unchanged. The common reading is that the market now expects more inflation over ten years. The arithmetic permits at least three other readings that fit the same two prices.

The real yield fell 8 basis points while the nominal did not move. That is consistent with a rise in demand for inflation protection — pension funds matching liabilities, an index fund rebalancing, a large buyer hedging — which pushes the linker’s price up and its yield down without any change in what anyone expects prices to do. It is equally consistent with a fall in the liquidity premium on the linker, if a new issue has deepened the market, and with a rise in the inflation risk premium, if nominal holders have become more worried about the upside. All four stories — expectations, demand, liquidity, risk premium — produce the same 8 basis points in the same direction, and the breakeven contains no information that distinguishes them.

What does distinguish them is the other data: a fall in the real yield accompanied by heavy linker inflows reads differently from the same fall accompanied by no flow data and a rise in surveyed expectations. The breakeven is the price at which those arguments start, not the argument. A note that reports the 8 basis points, both legs, the date and the pair has given the reader something checkable. A note that reports the 8 basis points and adds a sentence about what the market now expects has given the reader a forecast with a price attached to it, and the price does not support the forecast.

What a breakeven read costs

The computation is small. Four engine calls: breakeven_inflation for the per-tenor BEIs and the 5y5y forward, rates_math for the forwards built from the two curves, validate to cross-check the nominal quote against a second observation, and a second breakeven_inflation on the earlier curve so that a change is measured rather than remembered. At $0.006125 on Base that is $0.02450.

Three drafting calls at DeepSeek’s measured $0.00700 is $0.02100, so one note is $0.04550. Eight notes a month — two a week — is 32 engine calls ($0.19600) and 24 model calls ($0.16800): $0.36400, or $0.04550 a note and $4.37 over twelve months at the same pace. At the long-prompt rate of $0.00955 the model line becomes $0.22920 and the month $0.42520, 17 per cent more; on Qwen at a flat $0.00875 the model line is $0.21000 and the month $0.40600. The document parts that render the note carry no published per-call price and sit outside all three totals, which understates the work rather than overstating it.

A reporting standard

Four rules, each of which removes one way for a breakeven to become a forecast.

Quote the pair, not the spread. Both legs, both yields, the tenors, the issuer and the date. A breakeven reported alone cannot be checked, and a reader who cannot check it has to trust it, which is the position a measurement exists to avoid.

Name the index and the lag. Which price index the linker references, and the delay between the index and the cash flow. Two breakevens on different indices are not comparable, and a breakeven on a lagged index is not a forward expectation over the stated window.

State the method. The exact ratio, or the subtraction, and say which. Four basis points at these levels is the difference between methods, and it is the only part of the number the reader can audit without the source data.

Build the forward from the legs. Never difference two breakevens across horizons. If a forward is wanted, compound the nominal and real forwards separately and take the ratio at the end, and report the four inputs beside the answer so the arithmetic can be repeated. If a spot breakeven is wanted, quote it to a tenth of a basis point rather than a whole one — the method difference is smaller than that.

The decision that follows is not about whether the number is high or low. It is about what to publish. Publish the breakeven as a price: the pair, the two yields, the index, the lag, the method, the date, and — where a forward is wanted — the four compounding inputs and the result. Then stop. The reader who wants an expectation of inflation has to bring a survey, a model or a view, and the value of the breakeven is that it tells them what the market charges for the hedge while they do. That is a smaller claim than the one usually attached to the number, and it is the only one the arithmetic carries.

Method note

The curves in Table 1, the changes in Table 2 and the eight-basis-point example are constructed for this piece and labelled illustrative; they are not an observation of any market and should not be cited as one. Each breakeven is computed to four decimal places as (1 + nominal) ÷ (1 + real) − 1 and reported in basis points; the subtraction column is the nominal yield minus the real yield, and the overstatement column is the difference between the two, which is the cross term. The 5-year nominal forward of 4.3902 per cent, the real forward of 1.6001 per cent, the forward breakeven of 2.7462 per cent and the error-propagation case at 1.6401 per cent and 270.6 basis points are computations from the four yields named in the table, with annual compounding, and every figure has been recomputed from those inputs rather than carried forward from a previous draft. Engine prices are the published $0.006125 per basic operation on Base, from catalog.json, read 31 October 2026; the analyst services are pending review and return no 402 quote, so that price is published rather than measured. Model prices are measured from the live 402 handshake on 31 October 2026: $0.00700 for deepseek-v4-flash on Base at a short prompt and max_tokens: 1000, $0.00955 at about 3,000 tokens, and $0.00875 for qwen3.8-flash at both lengths. The monthly counts are planning assumptions. Nothing here is a forecast of inflation, a market call, or a recommendation to buy, sell, hold or hedge any instrument.

The pillar is Finance and markets on x402 — three engines, four jobs, one price. The role that computes this is macro_analyst, whose breakeven_inflation operation returns the BEIs, the 5y5y forward and an expectations label, and whose second refusal — no central bank’s intentions reported as fact — is the same boundary in a different dimension. The compute is bought at x402.press/services/deepseek/ with a second model at x402.press/services/qwen/; the measured handshake is reviewed at DeepSeek: per-call chat completions over x402, and the flat-rate alternative is set beside it at Inference endpoints compared. The breakeven is the rate that belongs in a discount rate and a terminal growth assumption, so the practical continuation is DCF model structure checklist, and the inputs it lands on are the ones Financial statement red flags tests in the accounts.

Filed under Jobs · Tags: breakeven, inflation, bonds, macro