Grant Ellison

The 10-year has given back 24bps in four weeks. The 30-year fixed has given back 4. The spread took the rest and is now the widest since September 2025. The bond rally is reaching a mortgage quote at about one-sixth strength. №12: scndordereffects.substack.co…

Cold enough in the Northeast this morning that the garage took twenty minutes to warm up. There’s a version of this job where you read the heating-oil basis and a version where you just notice the furnace ran all night. The second one is usually earlier.

Housing data has a sequencing problem that almost nobody respects. Pending contracts lead closings by 30 to 45 days. Starts lead completions by a year. Permits lead starts by whatever the builder decides. Every series is telling you about a different month.

The useful question about a benchmark revision isn’t the size, it’s whether it’s unusual. Last year’s was 0.1% of total nonfarm. The ten year average absolute revision is 0.2%. So the revision everyone argued about was half the typical year.

Oklahoma’s drywall workforce is 89% foreign born. Texas is 80%. Iowa’s roofers are 80%. The map doesn’t look like the politics of it. Plenty of the heaviest reliance sits in states nobody files under immigrant labor.

Undocumented workers are estimated to fill up to half of US residential construction trade jobs, and up to two thirds in the border states. By trade it ranges from 50% of drywall installers down to 7% of electricians. Licensure sorts the exposure almost perfectly.

Q3 Employment Cost Index this morning, which almost nobody trades and the Fed reads closely. Wages are the part of inflation that doesn’t round trip. Energy spikes unwind. A pay scale that resets does not.

CPI normally sits above PCE because it carries a much heavier shelter weight. The two have crossed, and the reason is that shelter decelerated for a year. Then shelter turned back up in September. But now we need to see where it goes from here.

Core PCE is 3.1%, trimmed mean is 2.3%, and the gap says a minority of categories is pushing headline core higher. The catch is that the Dallas Fed trims asymmetrically on the assumption prices skew negative. When they skew positive the measure understates. They flagged exactly this in April.

PMMS 7.24%, a second straight decline and down 16bps in two weeks. The survey-week 10yr average fell about 12bps to get there, so the spread widened again to roughly 2.24. Fourth week running. Borrowers keep getting most of the move and never all of it.

Personal spending rose 0.5% in September. PCE prices rose 0.5%. Real consumption came in at exactly 0.0%. Income only rose 0.2%, so people covered the difference out of savings and got nothing additional for it.

Last year’s BLS benchmark revision took 79k off total nonfarm payrolls, and most news coverage stopped there. Private employment was revised down 178k. Government went the other way by about 99k. The headline netted two moves in opposite directions.

Fed holds at 3.75 to 4.00, but the vote was 10-2 and both dissents were for a hike. September’s hike was unanimous. Six weeks later two members wanted to go again immediately. The statement says wait, the vote says December is closer than 80% (per Kalshi, 83% CME).

The rate went 7.50% to 7.41% and purchase apps fell 7.3%. Then 7.41% to 7.37% and they rose 3.8%. Four basis points did what nine couldn't. Either the two-week cumulative move is what registers, or last week's collapse was noise and this is the correction.

Q3 rental vacancy at 10am, two hours before anyone is thinking about anything except the 2pm Fed announcement. It’s the number that decides whether the rental market is tight or loose, and the answer depends entirely on which benchmark period you compare it to.

Purchase applications rose 3.8% last week, breaking a seven-week losing streak, with the MBA rate down 4bps to 7.37%. I said last week that the level was doing the work and a few basis points wouldn't register. One week later it did. I was wrong.

279k units are authorized and not started, up 9.8% YoY, and the multifamily piece of that is up 13.9%. Developers are paying to keep entitlements alive and refusing to fund them. That is a cost-of-capital decision showing up in the construction stats.

Two national house price indices for August, published the same morning. Case-Shiller 20-city: 2.7% YoY, up from 2.5%. FHFA: 2.4% YoY, down from 2.5%. One accelerating, one decelerating. FHFA only sees conforming mortgages; Case-Shiller sees cash too.

The thing that makes a labor shortage bite in housing is sequence. You can’t skip drywall. If one trade loses half its workforce the whole critical path stalls, even when framing and electrical are fully staffed. Permits keep getting pulled and nothing comes out of the ground.

Buried in new home sales: 115k homes for sale that haven’t been started yet, an all-time high. Under construction is 22% below the cycle peak. Builders are selling houses that don’t exist and declining to pour foundations until someone signs.