How to read the Treasury yield curve

You are reading par yields, which is not the same curve

The daily Treasury curve is a set of constant maturity yields: the yield a hypothetical security priced at par would carry at each tenor, interpolated from bid-side quotes on the most recently auctioned issues. Three consequences follow. It reflects on-the-run securities, which trade at a liquidity premium over otherwise identical off-the-run bonds. It is interpolated, so the ten-year point on a day with no ten-year trading near par is a construction. And it carries coupon effects, so it is not the zero-coupon curve you want if you are discounting a single cashflow.

For most reading that does not matter. For pricing, it does: bootstrapping to zero rates and then to forwards is a separate step, and a forward rate implied from the par curve is where the market actually expresses its view on future policy. If someone tells you the curve is predicting three cuts, they mean the forwards, not the yields printed on the page.

The spread everyone quotes, and what it is worth

Ten-year minus two-year is the standard headline, and the standard claim is that it inverts before recessions. The record on direction is genuinely good. The record on timing is not: the lag between inversion and downturn has ranged from well under a year to more than two, and the signal has fired without a recession following. A leading indicator with a lead time that varies by 18 months is a regime description, not a trade.

The three-month to ten-year spread has a better statistical record than the two-year version and gets a fraction of the attention, largely because the two-year is what the media settled on. Both are saying the same underlying thing: short rates are high relative to where the market thinks they are going, which normally means policy is tight and the market expects that to end.

Our position: the curve tells you what is priced, and its value is as a benchmark you are disagreeing with, not as an oracle. If you think two-year yields are wrong, the curve is what you have to be wrong about, and that is a much more productive use of it than counting months to a recession.

The three questions the curve actually answers well

First, the policy path: forward rates bootstrapped from the curve give you the market-implied path of short rates, which is the reference any macro view should be stated against. Second, the real rate: subtracting the TIPS real yield from the nominal at the same tenor gives breakeven inflation, the compensation demanded for inflation over that horizon. Note that the subtraction is an approximation, and the exact breakeven divides the gross nominal by the gross real rather than subtracting; on top of that the number contains a liquidity and inflation risk premium, so it is not a clean expectation. Third, the price of being wrong: duration and convexity turn a change in yield into a change in the value of a specific bond, which is the only one of the three that is arithmetic rather than interpretation.

• Pull the daily par yields from the Treasury feed for the tenors you care about.
• Take the spread you want, and check it against the three-month to ten-year version too.
• Subtract the matching TIPS real yield to read inflation compensation.
• Run duration and convexity to price what a move would cost the position you hold.

Getting the numbers without inheriting someone else’s

Treasury publishes the daily par yield curve rates itself, including an XML feed, which is the same source the JMM engine collects two-year, ten-year, and thirty-year yields and the ten-year minus two-year spread from. Use it rather than a chart from a news article, for one specific reason: business-day publication means a “current” figure quoted on a Monday holiday is Friday’s, and a stale curve compared against a live price manufactures a spread that does not exist.

Bills are their own trap and deserve separate handling. A Treasury bill has three legitimate yield numbers, bank discount, coupon equivalent, and effective annual, and they are not small differences. Quoting the discount rate as though it were a yield understates what the bill actually returns.