Selloff in the long-end bond market exacerbates elevated U.S. interest rates and tail risk.

Interest Rate Weekly · 2026-09-28 · 2026-10-04

This week's main theme: Global long-end bond selling continued to deepen, U.S. Treasury yields hit a new high for this century, Euro area inflation rose to a three-year high, and the Fed's relatively dovish remarks plus weak employment data failed to cap yields. There was divergence within the U.S. yield curve: the short end fell slightly while the long end continued to rise.

I. Review of Last Week

U.S. storyline. On 9/30 headlines said U.S. Treasury yields touched highs for this century and the U.S. economy remained resilient; on 10/1 headlines said yields rose to their highest since 2002. Contemporary commentary described the U.S. economy as 'running hot' with strong consumption growth and high inflation, and the bond market had already priced that in. A 10/2 report noted that employment data disappointed and the Fed's language leaned dovish, yet yields did not fall. On policy, market discussion around the Fed was clearly divided: one camp thinks the Fed may remain on the sidelines for an extended period and that the chance of another rate hike this year is far less certain than before; the other camp thinks that amid a 'real investment boom' the Fed needs to hike to keep policy rates near neutral. U.S. 30-year mortgage rates rose to 7.28%, and other reports said homeowners holding mortgages below 4% are reluctant to move while institutional investors are reducing holdings of mortgage-backed securities.

Curve data matched the news. U.S. 2-year Treasury yield 4.78%, 5-day -9bp; 5-year 5.01%, 5-day -2bp; 10-year 5.24%, 5-day +6bp; 30-year 5.61%, 5-day +14bp. Over the week the short end fell while the long end rose, with the 2s10s spread at +46bp. Over 20 days, all tenors rose: 2-year +39bp, 5-year +47bp, 10-year +45bp, 30-year +34bp; 60-day increases were 55 to 70bp. The 1-year z-score was between 2.4 and 3.1, and the 10-year and 30-year were 3.01 and 3.06 respectively, at extreme positions in the past year’s distribution. This fits the narrative that 'weak employment plus a somewhat dovish Fed pressure the short end, while supply, debt, and inflation concerns push the long end up,' but curve data only shows outcomes and cannot prove each causal link.

Global storyline. On 10/2 headlines placed France at the center of the global bond sell-off, and other reports said high government debt added fuel to the global bond sell-off. News that tankers were returning to the Strait could have eased energy price pressure, but headlines showed the sell-off continued to deepen. In the Euro area, inflation rose to 3.8%, a three-year high; inflation in major economies such as Germany and France also hit multi-year highs, and the ECB had previously warned about the inflation risk from energy prices. In the U.K., articles discussed whether the Bank of England would hike in Q4, house prices unexpectedly fell, and buyers were preparing for higher rates. Commentators in India noted that core inflation was moderate and growth was weak, giving the central bank limited reasons to hike. From 10/1 some asset managers began advising clients to reconsider bonds; both 'give bonds another chance' and 'do not hold bonds' voices appeared, indicating significant disagreement.

Canada storyline. The main Canada-related news this week concerned real estate: rents rose about 20%, which the Bank of Canada could not explain; after removing inflation, house prices returned to 2016 levels and buying remains difficult. There was no direct news in the sample about the Bank of Canada’s policy path this week, so no inference is made. Curve data: Canada 2-year 3.27%, 5-day -13bp; 5-year 3.62%, 5-day -7bp; 10-year 3.94%, 5-day -2bp; 30-year 4.28%, 5-day +3bp. 20-day increases were 13 to 26bp, smaller than the U.S. 34 to 47bp, but z-scores were similarly high in the 2.1 to 2.4 range. Canada’s 2s10s is +67bp, steeper than the U.S.

Credit. U.S. high-yield OAS was 3.24%, widening 59bp over 20 days; investment-grade OAS was 0.86%, widening 5bp over 20 days. These two series have only about 3.1 years of history and are described only: high-yield spreads clearly widened during the rate shock period while investment-grade barely moved.

II. Views on Future Rates

Method scope. The directional judgments in this report come only from momentum signals that have been backtested historically; the test criterion is an improvement in hit rate relative to the base rate. For instruments whose signals did not pass the tests, direction is treated as a random walk and no directional judgment is made; only static factors such as carry are discussed. Carry is the expected return if yields remain unchanged; it is a tailwind but does not constitute a directional prediction.

U.S. short and mid maturities: directional signals lean upward. Momentum signals for the U.S. 2-year, 5-year, and 10-year all passed tests, and directions are for yields to rise. Hit rates are 56.2%, 54.0%, and 52.7% respectively, base rates are 50.5%, 50.7%, and 50.5%, with sample sizes all around 12,000. The 2-year has the largest edge, about 5.7 percentage points; the 5-year about 3.3 points; the 10-year only about 2.2 points, a marginal statistical edge and not a strong conclusion. These signals describe a small tilt in win probability and are not certainties. This week the 2-year already fell 9bp, and news about the Fed’s path presents both 'on-hold' and 'need to hike' narratives, so there is tension between the signals and recent price moves and a pullback may occur.

U.S. long end: the 30-year momentum signal did not pass the test, with a hit rate of 52.8% versus a base rate of 51.2%, and is treated as a random walk. Direction is unpredictable and no judgment is made. Only static factors can be discussed: 30-year carry + roll annualized is about 6.17%, ranking first in the sample, but its reliability metric is only 0.077; going long shows a 53% win rate, so it is only marginal. The 5.61% yield level is attractive on its own, but the z-score of 3.06 indicates yields are far from the one-year mean, and neither mean reversion nor trend continuation can be statistically confirmed.

Canada: momentum signals for the 2-year, 5-year, 10-year, and 30-year all failed the tests, with hit rates between 49.8% and 51.5%, most below base rates. Directions for all four tenors are unpredictable and no judgment is made. On static factors, Canada 10-year carry annualized is about 4.58% and 30-year about 5.07%; carry ordering tests passed for both, but that only indicates the carry ranking is reliable and says nothing about which way yields will move.

Carry versus price loss. In a rising yield environment, price losses can easily outrun carry; carry alone cannot be used as a buy signal. Roughly estimated, a 10-year instrument with duration about 8 losing 45bp in 20 days corresponds to a price loss of about 3.6%, while carry in the same period is only about 0.5%. This week the 20-day and 60-day rises far exceeded what carry could cover. In the carry ranking, U.S. 2-year has the highest reliability (correlation 0.326, long win rate 68%), annualized carry about 5.46%; short-end duration is small so price moves erode carry less, making it the most solid static case. Even so, the U.S. 2-year directional signal leans upward, opposite to carry, so both aspects need to be considered together. 10-year carry is about 5.7%, reliability 0.113, win rate 55%; 5-year about 5.39%, reliability 0.178, win rate 59%.

Cross-country divergence. U.S. 2-year yields are about 151bp higher than Canada’s, and 10-year about 130bp higher. Momentum signals for the U.S. short and mid tenors passed tests and lean upward, while Canada has no directional signals, so there is no tested basis for predicting which way the U.S.-Canada spread will move. The only confirmed fact is the current state: yields in both countries rose over 20 days, with the U.S. moving more; global bond market linkage is strong. Euro area inflation, France’s debt concerns, and U.K. rate-hike expectations all point to a global re-pricing of term premium, and Canada cannot be isolated from that, but this is background description and does not constitute a directional judgment for Canada.

Curve shape. U.S. 2s10s is +46bp and Canada’s is +67bp; both curves are upward sloping. This week the U.S. curve steepened as the short end fell and the long end rose. Future curve shape has not been independently tested. If one extrapolates from the upward bias in U.S. 2-year, 5-year, and 10-year yields, the curve might flatten, but the sizes of the edges differ and are limited, so this inference is weak and for reference only.

III. Risk Warnings

The statistical edges of directional signals are small. The U.S. 10-year hit rate is only about 2 percentage points above the base rate, and the 2-year only about 6 points; the probability of an individual judgment being wrong remains high. Random-walk instruments (U.S. 30-year and all Canadian tenors) have no usable directional guidance; any view based on their direction lacks support from the data used in this report.

Current yields are at extreme positions of the past year’s distribution: U.S. 10-year and 30-year z-scores exceed 3. In extreme states yields may quickly mean-revert or trends may continue; both directions carry tail risk. There are many news-side variables: Fed path divergence (on-hold versus hikes), oil prices and energy supply, Euro area inflation, fiscal concerns in high-debt economies like France—any of these could change yield dynamics.

Carry is a static metric and does not account for yield changes. In a rising yield environment, price losses can overrun carry. High-yield spreads have visibly widened; credit data has only about 3.1 years of history and can only be described, not extrapolated.

This report is for research purposes and does not constitute investment advice.