How to Read a Treasury Yield History Across a Benchmark Break

Use the US Treasury’s 30-year series to distinguish a missing benchmark from a new curve methodology and identify when a daily yield change can be calculated.

In this article

The US Treasury reported a 20-year par yield of 4.75% on February 8, 2006, and a 30-year yield of 4.51% on February 9. Joining those two cells produces a 24-basis-point decline. It is also a switch in maturity. Treasury had suspended its 30-year constant-maturity series in 2002 and resumed it on February 9, 2006. The first valid one-day change after the restart uses February 9 and 10, when the 30-year yield moved from 4.51% to 4.55%, a four-basis-point increase. Treasury historical daily par yield observations · Treasury’s explanation of series breaks

Three observations that expose a bad join

A constant-maturity yield is read from a fitted curve at a fixed tenor. It need not equal the yield at which a particular bond traded. That matters when a historical data file contains several long maturities in adjacent columns. A blank 30-year observation is easy to fill with a plausible 20-year number. At two decimal places, the exported column can look indistinguishable from genuine observations.

Treasury date20-year CMT30-year CMTChange from prior trading day at the 30-year tenor
February 8, 20064.75%UnavailableUnavailable
February 9, 20064.72%4.51%Unavailable
February 10, 20064.76%4.55%+4 basis points

On February 9 itself, the two tenors differed by 21 basis points. The 20-year yield had moved down three basis points from February 8. Those two facts account for the apparent 24-basis-point drop obtained by switching columns. They also show why using the nearest maturity for “just one day” can distort the largest move in a risk report. Keep the observed 20-year rate if the mandate calls for it, with its original name. Federal Reserve H.15 release dated February 13, 2006

For a holdings review, mark February 9 as the start of an observable 30-year level. For a one-day risk measure, mark February 10 as the first observable change after the gap. For a total-return calculation the yield change alone is insufficient. Bond prices, coupon income and the defined holding or index methodology also matter. One source table gives three different eligibility decisions depending on the question.

A method change calls for a different treatment

On December 6, 2021, Treasury changed the method used to estimate its official par curve. It moved from a quasi-cubic Hermite spline to a monotone convex spline. A 30-year observation on each side of that date still has the same maturity. The relevant flag is the curve construction method. Treating every computed observation after December 6 as missing would discard a series the issuer continued to publish. Treating its calculation as identical across the boundary would conceal a real methodological change. Treasury yield curve methodology

The right action depends on the tolerance of the analysis. A chart of long-run yields can keep both published levels with the method change annotated. A daily model trained to detect very small residuals needs more. It should test overlapping estimates under both methods where available, then check whether the differences matter relative to its alert threshold. A change on the transition date includes market movement. One pair of closing observations cannot isolate a pure methodology effect.

Define the history a provider must deliver

Ask for a short extract covering both episodes. A usable return should preserve the blank February 8 cell, identify the source and tenor of every substitute, calculate the first valid same-tenor change and retain the December 2021 method note. A visually continuous chart is insufficient evidence for any of those decisions.

For an analyst reviewing one curve, the Treasury table and a spreadsheet can resolve the case. For a team maintaining decisions across many benchmarks, AllMind Data Rooms can hold methodology notices next to the firm’s approved mapping and surface the cited documents during review. The dated observations still need a source with the required history. Apply the final decision to the affected models and report windows, then record when a genuine series resumed. The same discipline is useful when a fund’s reported performance starts before its publicly listed share class.