ALNRI analysis — market rents vs official CPI shelter
Web-verified 2026-06-08. Structured + edges + sources: macro-rent-cpi-divergence.json. The deep-dive on the CPI-shelter strand of macro-official-data-integrity. The lag is documented BLS fact; the "convenient direction" is the analytical (contested) layer.
The series being compared
| Series | What it tracks | Lead/lag vs CPI |
|---|---|---|
| ALNRI — Apartment List National Rent Index | new-lease median rents | leads CPI rent ~16 months |
| ZORI — Zillow Observed Rent Index | asking rents, repeat-rent, quality-controlled | leads |
| NTRI — BLS New Tenant Rent Index (R-CPI-NTR) | official CPI microdata, turnover units only, dated to when the price changed | leads CPI rent ~4 quarters |
| CPI rent — CUUR0000SEHA + OER | what all renters pay this month (full lease stock) | the official, lagging series |
Shelter (rent + OER) is ~⅓ of CPI — so a one-year lag in this component drives the whole headline at turning points.
The divergence — STRONG (independent series + BLS's own NTRI)
- Dec 2021: ALNRI +18% YoY while official CPI rent showed ~3% — a ~15pp gap.
- Zillow ZORI ~15% (late 2021) vs CPI rent ~5.5%.
- ALNRI YoY peaked Nov 2021, leading the rent-CPI peak by ~16 months.
- BLS's own NTRI — same microdata, turnover units, correctly dated — leads official CPI rent by ~4 quarters. BLS built it precisely because the headline shelter series lags.
- Mechanism: official CPI shelter measures what all renters pay; most leases reset only annually, so a market-rent shock enters the index slowly, smeared over ~12 months.
The two-phase bias — lag FACT; "conveniently directional" CONTESTED
- 2021–22 (understatement): market rents +15–18%, official CPI shelter only ~3–5.5%, while the funds rate was still ~0% → the lag helped justify staying loose too long ("transitory").
- 2023–25 (overstatement): market rents cooled (ALNRI/ZORI flat-to-negative) but official CPI shelter stayed elevated for ~a year → lagged shelter was a large part of the 2024 "inflation is resurging" narrative that kept policy restrictive.
The most-weighted CPI component is structurally ~1 year stale, so it biases the headline in whichever direction is 12 months out of date — and in 2021–25 that error ran in the policy-convenient direction at each phase. Convenient or coincidental is the contested part; the lag and its sign are fact.
When policy cited which shelter series — the chronology (added 2026-06-16, #53)
Tracking which framing the Fed reached for as the stance changed. Speech-page rows are verified primaries; FOMC-presser rows are flagged for quote re-verification.
| Date | Who | Shelter framing | Series leaned on |
|---|---|---|---|
| 2021-08-27 | Powell (Jackson Hole) | "transitory", durable goods — surging market rents absent | lagged-low official (by omission) ✔︎ verified |
| 2022-03-24 | Gov. Waller | earliest explicit lag: CPI rents "slow to reflect market conditions"; CoreLogic +12% / RealPage +15%; CPI rent would "double in 2022" | leading market rents, to argue inflation would rise ✔︎ verified |
| 2022-11-02 | Powell (presser) | new-lease series "very pro-cyclical … now coming down faster" | both; leading already falling — quote unverified |
| 2022-11-30 | Powell (Brookings) | "the market rate on new leases is a timelier indicator"; new-lease ~20% "falling sharply since mid-year" | both; leading → forecast lagged ✔︎ verified |
| 2023-08-25 | Powell (Jackson Hole) | slowing new-lease rents "in the pipeline … over the coming year" | leading as forward signal ✔︎ verified |
| 2024-04 / 05 | Powell + minutes | "lack of further progress"; shelter sluggish → higher-for-longer | lagged-high official — even though market rents had normalized in late 2023 — quote unverified |
| 2025–26 | Powell (pressers) | disinflation "slow and bumpy", then "come down steadily" | lagged official, now cooperating |
The asymmetry (the thesis). 2021 leaned on lagged-low shelter to support "transitory" (downplay); 2022 used the leading new-lease data because it predicted disinflation (useful); but by 2024 — with leading rents already cooled — Powell leaned on the lagged, still-high official measure to justify higher-for-longer. The Fed invoked whichever shelter framing fit the stance. Notably, Powell never named the BLS New Tenant Rent Index by label — though the Fed's own Reserve Banks (Cleveland WP 22-38R / EC 2024-17, Richmond, Boston, Minneapolis) flagged the lag in real time. The cleanest evidence is that gap between Reserve-Bank research and FOMC rhetoric. (Anchor public quotes on the four verified speech pages; confirm presser wording before relying on exact words.)
Why it matters beyond the print
Shelter feeds COLA (CPI-W), TIPS, and bracket indexation (macro-official-data-integrity) — a smoothed series also avoided the spike that would have forced larger COLA/TIPS payouts at the 2021–22 peak. And CRE/multifamily valuations and private-credit marks depend on actual rents (macro-cre-privatecredit), not the lagged CPI proxy — so the same lag that flatters the inflation print can mask stress in apartment/CRE collateral.
Posture
Read ALNRI / ZORI / NTRI for the real shelter signal; treat official CPI shelter (CUUR0000SEHA / OER) as a ~1-year-stale, policy-relevant lagging print — the single clearest, BLS-confirmed example of "doubt the headline."
← Research index · structured data: macro-rent-cpi-divergence.json · macro-rent-cpi-divergence.md