A redemption yield is a number in search of a comparison. 3.58% is high or low only against what the curve says a gilt of that maturity should yield. That comparison is benchmark divergence, and it is where a screen for value in the gilt book starts. The 1 September gilt report prices all 104 gilts — 70 conventional, 34 index-linked — each measured against a fitted benchmark curve, with coverage at 104/104. Most sit close to the line: 62 of the 104 are within ±10 basis points of fair, 89 within ±25. Five are not — and the interesting part is that only one of the five is a real signal.
What the column measures
The engine fits a smooth curve through each segment — conventional gilts on a nominal curve, index-linked on a real-yield curve, ranked separately, because “fair” in nominal space is meaningless in real space — and interpolates it at every gilt’s exact maturity. Divergence is the difference between the gilt’s quoted gross redemption yield and that fitted yield, in basis points:
- Positive divergence — the gilt yields more than the line: it looks cheap. Rank 1 is the cheapest gilt to its own curve (as the column header states), and on 1 September that was the 3¾% 2027, 378.5 bps above its benchmark.
- Negative divergence — the gilt yields less than the line: it looks rich. The conventional ranking ran 1–70, and the bottom of it — Rank 70, the 0 3/8% 2026 at −50.3 bps — was the only negative flag of the day.
Divergence is computed from the unrounded yields, so it will not always tie exactly to the rounded yield pair in the table below — treat the basis-point figure as the precise one.
The five flags, 1 September
| Gilt | Matures | Clean | Gross* | Benchmark | Divergence |
|---|---|---|---|---|---|
| 3¾% Treasury Gilt 2027 | 2027-03-07 | 99.80 | 8.07% | 4.08% | +378.5 |
| 4 3/8% Treasury Gilt 2028 | 2028-03-07 | 99.84 | 6.10% | 4.32% | +163.5 |
| 4 3/8% Treasury Gilt 2030 | 2030-03-07 | 99.08 | 5.42% | 4.53% | +76.9 |
| 4 1/8% Treasury Gilt 2031 | 2031-03-07 | 97.61 | 5.31% | 4.62% | +56.7 |
| 0 3/8% Treasury Gilt 2026 | 2026-10-22 | 99.57 | 3.61% | 4.04% | −50.3 |
* Gross redemption yield as displayed in the report (annually compounded).
A tidy pattern jumps out before any analytics: the four positive flags all mature on 7 March. That is not a coincidence of value — it is a coincidence of the calendar.
Four of five are the calendar, not the market
All four pay their next coupon on 7 September 2026, and on 1 September all four sat inside their ex-dividend period — the business days before the coupon when a buyer no longer receives it. The signature is visible in the report rows: accrued interest of −0.03 to −0.04 with 179 days accrued, where a normal cum-dividend gilt shows positive accrued. Twelve gilts shared the 7 September coupon date, and every one of them ran positive that day, from +9 bps on the 1¾% 2037 up to +378.5 on the 3¾% 2027. Two more of the cohort — the 4 5/8% 2032 at +43.1 and the 4 1/8% 2033 at +40.3 — sat just under the 50bp flag.
Why? In the ex-dividend period the quoted clean price no longer embeds the right to the imminent coupon, so a redemption yield computed against the full coupon cashflow runs hot — the buyer pays clean and forgoes the coupon, yet the yield math assumes they collect it. How hot depends on how soon the coupon lands and how short the bond is: the effect is brutal on a half-year gilt (nearly 380 bps of phantom cheapness) and fades to single digits by 2037. Four of the five flags on 1 September are that phantom. An ex-dividend gilt is not cheap; it is mis-stated by settlement convention, and a divergence column that does not know about the calendar dutifully reports it as the cheapest thing in the book.
The one that is real: 0 3/8% Treasury Gilt 2026
The fifth flag is on the other side of the line and survives the calendar test. The 0 3/8% 2026 matures on 22 October — 50 days after the report — and yields 3.58% (3.61% annualised) against a benchmark of 4.04%: rich by 50.3 bps. It is a coupon-less stub: 0.38% running yield, modified duration 0.13, DV01 of £0.0013 per £100 nominal. It carries almost no rate risk and behaves like cash with a tax angle — its tax-equivalent yield at the 40% band is 5.07% against a 3.61% gross, because its return is almost entirely capital gain.
Two readings, both pointing to “check, don’t trade”. Either the stub is genuinely bid — near-cash paper prized by tax-aware buyers, which is why its gross yield sits below the line — or the fitted curve at 0.14 years is the least reliable place in the book to draw a line, anchored by only the shortest gilts. The column cannot tell you which; it can only tell you this is the most expensive gilt in the book, which is exactly the right question to be asking about paper you might otherwise treat as a parking spot.
The healthy middle
Flags are the outliers; the rest of the book is the validation. Positive and negative divergence are roughly balanced, and 85% of the book sits within a quarter of a percent of fair. The index-linked side is the tightest of all: no flags on 1 September, its widest names the 0 1/8% Index-linked 2029 at +30.1 and the 0 1/8% Index-linked 2073 at −25.6, with the segment’s median around +3 bps against −3 bps for conventionals. A book where nearly every name clears within a few ticks of its line is pricing as one liquid market — divergence is an outlier detector, and it is doing its job when the outlier list is short and explainable.
Reading the column for real
- Learn the sign and the rank before the flags. Positive is cheap, negative is rich, Rank 1 is the gilt furthest above its curve — and the index-linked half of the book is measured against a different, real-yield curve.
- Filter the calendar first. A gilt with negative accrued interest and a coupon inside the next week is not cheap — it is ex-dividend. The report’s own rows carry the accrued figure that gives it away.
- Treat divergence as a screen, not a verdict. It generates candidates; disqualifying them is the analyst’s job. Maturity structure, tax treatment and liquidity all explain flags the curve fit cannot.
- Watch persistence. One-day spikes are usually mechanics — a coupon, a new issue settling, a suspension. A divergence that survives day-over-day comparison (the report carries prior-day divergence where available) is the one worth acting on.
The divergence column is the newest lens in the gilt analytics engine, sitting alongside gross and post-tax yields, duration and DV01, and the pipeline that publishes the daily archive. Like every other column it is only useful read against the calendar and the curve that produced it. On 1 September it flagged five gilts; four were the 7 September coupon in disguise, and the fifth was a 50-day stub asking a question worth answering.
If you’re building fixed-income tooling and want value screens that survive contact with settlement conventions, discuss your project.
