U.S. short- and medium-term interest rates rise; long-end risks are hard to predict

Interest Rate Weekly · 2026-10-05 to 2026-10-11

The main theme this week was a marginal cooling of long-end selling after auction yields reached the highest levels in more than 20 years. U.S. Treasury yields across maturities fell 3 to 4 bp over 5 days; the 20-day and 60-day measures still rose noticeably. Central bank rate hikes and fiscal premia remain the global backdrop for bond markets.

1. Review of Last Week

U.S. front. The Treasury issued about 7280 billion USD in government debt this week, and the stop-out yields on the 3-, 10-, and 30-year auctions were the highest in more than 20 years. In the October 7 10-year auction, the stop-out yield was 5.30%, the highest since 2000. The auction generated demand, but the sell-off continued on the day and was characterized in some commentary as a "bond-market bloodbath." By the weekend, judgments such as "contagious selling is fading" began to appear, consistent with the curve data: 2-, 5-, 10-, and 30-year yields fell 3 bp, 4 bp, 4 bp, and 3 bp over 5 days, respectively. The retracement was limited; 20-day moves were still up 17 bp, 24 bp, 28 bp, and 25 bp respectively, and 60-day moves were up 64 bp, 74 bp, 67 bp, and 51 bp. The 10-year closed at 5.24% and the 30-year at 5.60%; one-year z-values across maturities ranged from 2.26 to 2.66, at the high end of the past year’s distribution.

Policy and transmission: The Fed minutes were interpreted as showing little urgency for an October rate hike, while some forums and commentary still saw further hikes ahead. Goldman Sachs downgraded its view of the U.S. economy because rates are expected to remain high. Several reports noted that hikes may not suppress the economy, and could instead exacerbate K-shaped divergence. The spillovers from the sell-off have appeared in rate-sensitive assets: mortgage rates were once reported at 7.5%, with other reports saying they reached a three-year high; small-cap stocks were under pressure and REITs faced a dilemma in the M&A market. On energy, reports said a war with Iran pushed up gasoline prices, making inflation a political issue; U.S. and euro-area inflation data will be released next week. Credit-market reaction was muted: U.S. high-yield OAS was 3.15%, the 20-day OAS widened 45 bp; investment-grade OAS was 0.82%, the 20-day widened only 2 bp. Credit data has only 3.1 years of history, so here we merely describe it.

Canada. In the collected news this week there were no domestic Canadian items, so the Canadian move cannot be explained by news and must rely on curve data. Canadian 2-, 5-, 10-, and 30-year yields were 3.24%, 3.60%, 3.93%, and 4.29%, with 5-day changes between -3 bp and +1 bp, nearly flat. The 20-day measures were up 8 to 12 bp and the 60-day up 34 to 41 bp, with increases noticeably smaller than in the U.S. Over 60 days the Canadian 10-year rose 37 bp versus the U.S. 10-year’s 67 bp. One-year z-values across Canadian maturities were 1.88 to 2.28, also elevated but lower than the U.S. range of 2.26 to 2.66.

Global. The U.K. stood out: borrowing costs hit a 28-year high, and Bank of England Governor Bailey warned that rates may rise even if the economy weakens, saying that an unbalanced fiscal position will face bond-market turbulence. BOE official Mann said U.K. inflation has become entrenched and there is a risk it could reach 4% by year-end. The U.K. will also introduce a U.S.-style monthly inflation measure in early next year. In Europe, the ECB resumed hiking as euro-area yields surged, continued quantitative tightening, and some analysis flagged France as a risk point in the global government-bond market. The Bank of Japan hesitated on hikes and instead accelerated balance-sheet reduction to support the falling yen. After an RBI hike, local bond yields in India are expected to hover around 7.25%. China’s bond market has largely stayed outside this round of yield increases. "The bond sell-off pushed up the dollar" also emerged as a theme this week, indicating this rise is a story of both dollar moves and spread changes. Another group of reports focused on debt accumulation from AI investment and the hunt for yield, arguing that bond markets still price in loose conditions.

2. Views on Future Interest Rates

Methodology caveat first: directional judgments come only from momentum signals that have been validated historically. In this issue, U.S. 2-, 5-, and 10-year signals are labeled as proven, direction upward; the U.S. 30-year and all four Canadian maturities are labeled as approximate random walks, direction unpredictable, and thus no directional judgment is offered. The proven signals’ edge is modest; details follow.

U.S. short and belly. The 2-year upward signal has a hit rate of 56.2% versus a base rate of 50.5%, sample size near 12,000; the 5-year is 54.0% versus 50.6%; the 10-year is 52.7% versus 50.5%. The edge declines with maturity: the 2-year is the strongest, the 10-year quite thin. Accordingly, U.S. 2- and 5-year yields are tilted toward further increases over the next 3 weeks; the 10-year is similarly biased but with weaker confidence. Hit rates slightly above 50% imply only a probabilistic tilt, and short-term reversals remain possible. The 5-day retracement of 3 to 4 bp is a breather after prior gains and does not invalidate the signals.

U.S. long end. The 30-year direction hit rate is 52.8% versus 51.2%, statistically indistinguishable from a random walk, so no directional judgment is made. Only static factors can be stated: annualized carry plus roll is about 6.14%, but the reliability correlation coefficient is only 0.077 and the win rate 53%, marked as marginal. The 10-year carry is about 5.68%, correlation 0.113, win rate 55%. The 2-year carry is about 5.37%, correlation 0.326, win rate 68%, the most reliable in this set.

Carry only holds when yields are unchanged; it is a tailwind, not a directional forecast. Current yields are in an upward channel, and price losses can fully offset carry. Roughly: three-week carry is about an annualized 5.8%, the 2-year three-week carry about 0.3%, the 30-year about 0.36%. The 2-year’s 20-day rise of 17 bp, using duration ~1.9 years, implies a price loss of about 0.3%, roughly equal to three-week carry. The 30-year’s 20-day rise of 25 bp, using duration ~15–16 years, implies a price loss near 4%, about ten times the three-week carry. For long-duration instruments, a few additional bp of yield rise will swamp carry, so carry rankings alone should not be used as a buy signal.

Canada. All four Canadian maturities are approximate random walks, with hit rates between 49.8% and 51.5%, indistinguishable from the base rate, so no directional judgments are made. Statics: Canadian 30-year carry is about 5.12% and 10-year about 4.59%; reliability correlation coefficients are only 0.109 and 0.139. These carries also only realize if yields do not move and lack directional signal support.

Cross-country divergence. The U.S. 10-year exceeds Canada by 131 bp and the 2-year by 156 bp. Over 60 days the U.S. rises (2-year +64 bp, 10-year +67 bp) are about 1.8 times Canada’s (+36 bp, +37 bp), so divergence has recently widened. The U.S. short end has a proven upward tilt while Canada shows no signal, therefore the USD-CAD spread direction cannot be judged and only the current state can be described. Global central bank hikes and fiscal premia are the common backdrop; U.K. and European news move in the same direction as the U.S., but pricing speeds differ across markets. China remains outside so far, with no data supporting convergence or continued divergence.

Curve shape. U.S. 2s10s is +44 bp; over 60 days the 2-year rose 64 bp and the 10-year 67 bp, an approximately parallel upward shift. Over 20 days the 10-year rose 28 bp versus the 2-year’s 17 bp, steepening the curve slightly by about 11 bp. 10s30s is about +36 bp. Canada’s 2s10s is +69 bp; over 60 days the 2-year rose 36 bp and the 10-year 37 bp, also an approximately parallel move, and 10s30s about +36 bp. Curve shape itself provides no validated directional signal. Because U.S. 2-, 5-, and 10-year yields are biased upward, no judgment is made on whether 2s10s will steepen or flatten; the U.S. 30-year direction is unpredictable, so 10s30s is similarly indeterminate.

3. Risk Warning

Momentum signals’ hit rates exceed the base rate by only 2 to 6 percentage points, so the chance of a single incorrect call remains material. The U.S. 2- and 5-year "upward" labels are probabilistic tilts, not certainties.

Carry can be overwhelmed by price losses in an upward environment, especially for long-duration instruments. Carry rankings and win rates are historical statistics, not forward guarantees; reliability correlation coefficients mostly range from 0.08 to 0.33.

Event risk is concentrated next week: U.S. and euro-area inflation, U.K. GDP, and developments in the U.S. banking sector could all change short-term yield paths. Auction demand, fiscal and policy uncertainty, and energy-price moves from geopolitical factors can amplify volatility. Credit spread data has only 3.1 years of history and is insufficient for inference.

Canada lacked supporting news this week and its curve data shows no directional signal, so any directional view on Canada is unsupported.

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