Weekly Review, Sep 21–25, 2026: Real Yields Break Higher, Equities Ride It Out

TLDR: A hot growth print and a hawkish Fed voice drove a real-rate selloff. The 10Y rose 16bp to 5.17% and 2s10s steepened 11bp to 36bp, with the 10Y real yield doing 15bp of it. Wednesday's flash PMI beat and Governor Barr's call for more hikes did most of the damage; a weak 5-year auction added the last leg. Rates vol spiked while equity vol did not: MOVE +15.4 to 96.0, VIX flat. Credit cracked at the bottom while AI-led tech carried the S&P +1.21%.

Rates and the curve

This was a real-rate selloff, and Wednesday did most of the work. The 10Y real yield rose 15bp against a 1bp move in the 10Y breakeven (2.34%), so nearly all of the 16bp rise in the 10Y was real yield and term premium, not inflation compensation. The Sep 23 sequence (10Y, market marks) shows the auction was the last leg, not the cause:

• 9:45 ET flash PMI beat, then Barr's "further policy adjustments are likely to be needed" (Yahoo Finance): 4.97% to 5.08% by midday
• 1pm 5-year auction stopped at 5.033%, tailing 3.1bp on a 2.21 bid-to-cover (InvestingLive): 5.11% at the close

The front end did not anchor all week. The 2Y was up 11bp through Thursday at 4.87%, then gave back 6bp on Friday as oil fell and October hike odds slipped, finishing +5bp at 4.81%. Funding stayed orderly into quarter-end, with SOFR at 3.90% and EFFR at 3.88% on Sep 25.

Fed repricing

October went from a lean to the base case, then Friday took some back. Barr's remarks and the PMI pushed futures-implied odds of an Oct 28 hike to 4.00–4.25% from 58% to 74% by Thursday; they closed the week at 64%. The bigger shift was further out: hikes priced during 2027 rose to about 62bp from 42bp, which is why the belly and long end, not the 2Y, led the selloff.

What decides October is data, not speeches: August core PCE on Sep 30 (consensus +0.3% m/m per TradingKey) and September payrolls on Oct 2 (consensus +90K per BBH). The front end carries that event risk; duration carries the term premium that built this week.

Credit

Spreads widened with a clear quality gradient. IG OAS +4bp to 81bp is noise; HY +25bp to 293bp and CCC +45bp to 1,128bp are not. In cash, the IG corporate ETF fell 1.42% against 0.85% for high yield: IG lost more on duration, while the spread repricing sat at the bottom of the capital stack.

Equities

Index calm concealed a rates-driven rotation. The Nasdaq rose 2.06% while the Russell 2000 fell 0.80%, and sector dispersion ran 7.4pp from Technology to Utilities.

The bottom of the table is duration, not earnings: Utilities, Real Estate and Financials sold off with the long end, and Energy followed crude. Mortgage REITs (AGNC −2.53%, Annaly −2.09%) lagged the MBS ETF (−1.04%), a book-value read rather than a pass-through price move. At the top, Meta rose nearly 13% on its Muse AI agent and Connect event, pulling the AI complex with it (TradingKey).

FX, commodities and geopolitics

Oil was a geopolitics trade that cut both ways. Brent spiked Thursday after Houthi missiles targeted Saudi Arabia, then fell Friday on hopes of a US-Iran path to reopening the Strait of Hormuz (EnergyNow).

• November WTI −3.17%, December Brent −1.32% (headline front-month WTI shows a larger drop because of the contract roll)
• Dollar index +0.82% to 101.04; sterling the weakest major at −1.10%
• Gold −2.16% to $4,320.50, as rising real yields usually imply
• Natural gas +12.1%, the week's outlier

Cross-asset relative performance

Two of the three residuals say the same thing: rates, not equities, carried the shock. Against a two-factor model (10Y change and S&P return, 120 sessions), assets with a usable fit (R² of at least 0.40) show:

• The 2Y lagged its implied move by 12.3bp (−1.78σ): +5.0bp against +17.3bp implied, with most of the gap opening on Friday's rally
• MOVE ran hot even after the selloff: +6.7pt residual (+1.07σ), consistent with a supply shock on top of the repricing; its fit is the weakest shown
• VIX should have fallen and didn't: +1.75pt residual (+1.04σ)

Dollar, crude and gold fits are too weak to support a residual read, and 2s10s, IG and HY OAS failed the fit outright.

The stock-bond correlation is +0.43 over 60 sessions versus a one-year norm of +0.29. Bonds are hedging equities less than usual. That did not bite in an up week for stocks, but it is the live risk if yields keep rising and the equity bid fails.

Economic data

Every print beat, and the PMI was the one that moved rates. The flash composite at 58.4 is the highest since July 2021, with input costs at a near four-year high (S&P Global): growth plus cost pressure is exactly what the front end priced on Wednesday. Claims and new home sales confirm no demand crack (consensus from Trading Economics and the Reuters poll).

Week ahead

Sep 30 is the event. With breakevens flat this week, an upside core PCE print is what turns a real-rate selloff into an inflation-premium selloff, and it would pull the 2Y back up after Friday's rally. Payrolls on Oct 2 then settle October.

Key takeaways

• Real yields, not inflation, set the week: the 10Y real yield rose 15bp against a 1bp breakeven move. A hot core PCE on Sep 30 would add an inflation leg on top.
• October is the base case, not a lock: hike odds sit at 64% after touching 74%. Core PCE and payrolls decide it, and the 2Y carries that risk more than the long end.
• Friday's 2Y rally is the fragile part of the steepener: if the data confirm, the front end catches up and the curve flattens from higher yields, which is worse for duration and for mortgage REITs already lagging MBS.
• Watch rates vol against equity vol: MOVE near 96, VIX under 15 and a positive stock-bond correlation mean bonds are not hedging; a failed equity bid would hit both.

Note on data availability: this report does not carry University of Michigan sentiment (last observation Apr 2026) or a consensus for the flash composite PMI.