The 30-year is not high. The last 19 years were low.
The long bond at 5.16% is being read as a 2007 high. Against its own 49-year history it sits below the median — and one thing about this week is genuinely new.
The 30-year Treasury yield reached 5.17% on 23 July, a level last seen in July 2007. That comparison is accurate and it is being read the wrong way round. Against the full 49-year history of the series, 5.16% sits below the median. What ended in May was not a period of normal rates — it was the longest stretch of unusually cheap long money the series has ever recorded.
Nineteen years is the anomaly, not this week
The 30-year constant maturity series starts in February 1977 and has 12,355 daily observations. On 5,731 of them — 46.4% — the yield was lower than it is now. The current reading is close to the middle of its own distribution.
The reason it feels extreme is that the comparison window everyone is using is short. The stretch running from 18 July 2007 to 18 May 2026 held the yield below today’s level for 4,711 consecutive trading days, about 18.8 years. The second-longest such stretch in the whole series lasted 1.7 years. Nothing else comes close.
Decade medians make the point without any interpretation:
| Decade | Median 30-year yield |
|---|---|
| 1970s | 8.48% |
| 1980s | 10.57% |
| 1990s | 6.90% |
| 2000s | 4.97% |
| 2010s | 3.04% |
| 2020s | 3.88% |
Today’s 5.16% is above the 2000s median and far below every decade before it. Only against the 2010s does it look like a departure.
The window decides the alarm
Ranked against its own trailing five years, the same number sits at the 99.8th percentile. Ranked against the full history, the 46.4th. One reading, two answers, and the difference is entirely the length of the lookback.
This is why the indicator pages publish a five-year percentile and say so explicitly. A five-year rank is the right tool for asking whether something has changed recently. It is the wrong tool for asking whether a level is historically unusual, and the two questions get confused constantly.
For the record, the series has run from 0.99% on 9 March 2020 to 15.21% on 26 October 1981. Measured against that, neither 5.16% nor the decade that preceded it is remarkable. The 2020 reading is the outlier in this series, not this week.
One thing about this week is genuinely new
Initial jobless claims for the week ending 18 July came in at 187,000. Across the 2,579 weeks since February 1977 for which both series exist, claims have been below 190,000 exactly once — that week. Going further back, the last time claims printed lower than 187,000 at all was 6 September 1969, at 182,000.
So the combination is real even though neither half is. A 49-year low in claims alongside a long yield that is merely back to its own middle has not happened before in the overlapping record. It is a genuinely unusual pairing, and it is a different claim from “yields are high” — which they are not.
The curve is not 2007 either
The other half of the 2007 comparison does not survive contact with the data. The 10-year minus 2-year spread was +0.34 on 27 July. During June and July 2007, the last time the 30-year held above today’s level, that spread ranged from −0.04 to +0.24 — flat to slightly inverted. During the 85-day run above this level in 2004, it ranged from +1.72 to +2.34.
Today’s curve sits between those two, and it has been positive for 690 days; the last inversion ended on 5 September 2024. A long yield at 2007 levels with a curve that looks nothing like 2007 is a different configuration, not a repeat.
What this means for trade
The 30-year is the price of long money in the world’s reserve currency, and it sets the floor under the cost of financing inventory, vessels and trade credit everywhere else. A sustained move higher raises the carrying cost of goods in transit, which is a real channel into trade volumes.
The TradeTurbulence Stress Index reads 48.9, in its Normal band. Nothing in freight, energy or credit conditions is registering strain from this yet. That is a finding rather than a reassurance: the transmission from long rates into trade costs runs over quarters, and the index would not be expected to show it this early.
What to watch
- The 10-year/2-year spread going back through zero. It sits at +0.34 today, the 60th percentile of its trailing five years. An inversion from here, with the long end at these levels, would be the configuration that actually rhymes with 2007 — which the current curve does not.
- Claims returning above 220,000. The unusual half of this week is the labour market, not the bond market. If claims normalise while long yields hold, the pairing that makes this week unique disappears and only the ordinary half is left.
- The stress index leaving its Normal band. At 48.9 it is showing nothing. A move above 60 would be the first sign that higher long rates are reaching freight and credit rather than staying in the bond market.
Every figure above is computed from three public FRED series — DGS30, T10Y2Y and ICSA — indexed by observation date and listed with links in the front matter of this piece. The 30-year series is continuous through the 2002–2006 suspension of 30-year issuance: the constant maturity yield was published throughout, at about 250 observations a year, so comparisons across that period are valid.