Mortgage spread hit 2.19 this week. A year ago it was 2.20. Every basis point of narrowing since last October is gone, and it went while the 10yr was rallying. PMMS didn’t move at all. The Tape №10: scndordereffects.substack.co…

Two hours of rolling tonight, then deep dive on the permits data. The useful thing about getting choked out by a smaller guy is that it resets your confidence interval before you go read something you already have an opinion about.

September existing home sales rose 1.5%. MBA purchase applications have fallen six straight weeks and sit 16% below last year. Both are true. Closings in September are lock decisions from July and August. The applications series is telling you about December.

Shelter CPI went 3.0% to 3.2% YoY in September, and 0.3% to 0.4% MoM. That’s the component everyone has been counting on to carry disinflation. Almost nobody covered it because gasoline was up 9.1%.

Multifamily permits +1.0% YoY. Multifamily units authorized but not started @ +13.9%. Multifamily starts -25.9%. Developers are still getting entitlements and banking them instead of funding them. All three of those clear Census significance. The single-family versions of the same numbers don’t, so take those with a grain of salt.

Existing home sales rose 1.5% last week and got read as resilience. Those were July and August locks closing. Totally different rate environment. Pending sales just printed an all-time low, 66.0 against a prior record of 70.2 in 2024. Contracts signed in September close in October and November.

Housing starts 1.221mm, a 79k miss, and it’ll get written up as a collapse. Census puts a 90% confidence interval of plus or minus 11.5 points on that change. It contains zero. The one move in this release that clears significance is multifamily, down 25.9% YoY.

Fed blackout started Saturday and runs through the 29th. Ten days where nobody at the Fed can walk anything back or leak a correction to the usual reporter. Whatever December is priced at when they walk in on the 27th, it got there on data alone.

The “housing shortage” gets quoted as about 4 million units. Published estimates run from roughly zero to 10 million. Every one of them has to assume what headship rate would have prevailed in a different supply environment, and that's the assumption that carries by far the most weight.

September housing starts tomorrow at 8:30am, the first look on whether builders kept breaking ground at a 7.4% mortgage. Permits are the one I’ll be watching. Starts tell you what got financed months ago. Permits tell you what they decided last month.

2s10s closed Friday at 35bps. It was 51bps on Oct 7. Sixteen basis points of flattening in seven sessions, with both ends falling. The front can’t rally much with a hike live into the 27th, so the long end did most of the move.

Week ahead: September starts and permits Tuesday, claims and PMMS Thursday. Thin on paper. It’s also a full week of Fed blackout before the 27th. Nobody gets to clean up a bad print, so whatever December is priced at going into that meeting is what they walk in with.

Rolling tomorrow for the first time in a couple of weeks. What jiu jitsu does that nothing else I’ve found does is hand you an immediate, unarguable score. You got swept or you didn’t. No benchmark period to pick.

The missing middle in one number. Last month the US permitted roughly 48k units in 2-to-4 unit buildings, annualized, and started about 13k. Single family: 880k permits, 918k starts. Apartments: 475k permits, 344k starts. The building type that would actually fill the gap barely exists as a category now.

Fed blackout starts today and runs through the 29th. So the setup into the 27th is whatever the data did this week, with nobody at the Fed able to nudge it either way for ten days. That’s a cleaner read on what the market actually thinks than we usually get.

New-home months of supply is 8.5. A year ago it was also 8.5. In between, builders cut the median price about 6%, ran incentives near 12% of the sale price, and posted their fastest sales month of 2026. A full year of discounting and draining netted to zero on that measure.

October in the Northeast is the two weeks a year when the basement gym loses to going outside, and then suddenly it doesn’t. Lost that argument with myself at 5am. Back downstairs.

A renter who gets priced out of buying doesn’t create a new household. He stays in the one he already has. Move-outs to buy a home at the big apartment REITs ran about 7% in 2024 against a long-run average near 16 to 17%. Retention is close to record highs. New-lease rent growth over that stretch: under 1%.

Bernie Sanders: "STOP MUSK! STOP OLIGARCHY!" Elon Musk:

Miami has the longest days on market of the 30 metros I track at 85, and the steepest inventory drawdown at -12.3%. Inventory falling fastest while homes sit longest is what delisting looks like in the data. Sellers pull the listing and wait.