What is the yield after inflation, and where does the data say so?

A 10-year Treasury yielding 4.65% pays 4.65% in dollars. What those dollars will buy in ten years is a different question, and there is a separate endpoint whose entire job is to answer it.

The real curve

GET /ust/real-yield-rates has the identical signature to the nominal curve endpoint — from, to, filter[year], page[limit], page[offset], fmt — and the identical long-format row shape of date, tenor, rate.

What differs is coverage. Where the nominal curve publishes 14 tenors, the real curve publishes five: 5Y, 7Y, 10Y, 20Y, 30Y. In 2026 that gives meta.total of 710 rows to 27 July, which is 142 × 5. There is no real 3-month or real 2-year point, because the underlying instruments are Treasury Inflation-Protected Securities and none are issued at those maturities. You cannot build a full real curve; you can build the long half of one.

For 27 July 2026:

A real yield is the return over and above realised inflation. TIPS principal is indexed to the consumer price index, so their quoted yield is already net of whatever inflation turns out to be.

The breakeven, computed

Subtract the real yield from the nominal yield at the same tenor on the same date and you get the breakeven inflation rate — the average annual inflation over that horizon at which holding the nominal bond and the inflation-linked bond would produce the same outcome.

For 27 July 2026:

• 10-year: 4.65 − 2.44 = 2.21 percentage points, or 221 basis points
• 5-year: 4.40 − 2.22 = 2.18 pp, or 218 bps
• 30-year: 5.12 − 2.95 = 2.17 pp, or 217 bps
• 20-year: 5.15 − 2.76 = 2.39 pp, or 239 bps

Two observations follow immediately. First, the 5-, 10- and 30-year breakevens sit within four basis points of each other, while the 20-year sits 18 to 22 basis points above all of them. That is a feature of the shape of both underlying curves at 20 years, not a statement about inflation being different in year 20.

Second, and more important: a breakeven is not a forecast. It is a price difference between two instruments. It embeds an inflation risk premium, a liquidity difference between TIPS and nominal Treasuries, and the particular indexation lag TIPS use. Describing what the number is composed of is the job here; interpreting it as anyone's prediction would be a claim this data cannot support.

The arithmetic trap

A breakeven requires the same tenor on the same date, from two endpoints. Both conditions are easy to break. The real curve has no 2-year or 3-year point, so a loop over the nominal tenors will produce nulls or, worse, silently match 2Y against nothing. And both endpoints skip the same holidays but a join on date must still be an inner join — if one series published on a day the other did not, an outer join yields a breakeven computed against a stale leg.

This is the real-vs-nominal lesson from Markets Foundations, made mechanical: the concept is a subtraction, and the difficulty is entirely in lining up the two operands.

Try it now

1. Here are /ust/yield-rates and /ust/real-yield-rates, each read on its newest date (in code you would filter to that date yourself, because from and to will not). Check the two dates match, pair the rows on tenor, and compute the five breakevens in basis points. Which nominal rows found no partner?

2. A full year of breakevens is a loop over every date; its two ends fit on a page. Here are both curves on the first date of 2026 (the nominal 10Y is data[11], the real 10Y data[2]). Compute the 10-year breakeven there and on the newest date from step 1, then split the change into the nominal leg's move and the real leg's move. Say which of the two drove it.

3. List the nominal tenors that have no real counterpart. That list — 1M through 3Y — is the part of the curve where the breakeven question simply cannot be asked from this data.