← Results index

A new U.S Residential Housing Index | RecessionAlert — exploratory proxy campaign

2026-09-03T12:00:00Z Research completed SHA-pinned USHMI page discloses eight-component housing mainline/diffusion and a 10s/1s AND-combo without formulas, weights, native series, or on-row assets. Labeled PERMIT/HOUST YoY diffusion and DGS10−DGS1 proxies were executed with causal lag>=1 on SPY/IEF month-end overlays; disposition exploratory_proxy_campaign under Other/Research with explicit NO-GO for native fidelity.
1617 recessionalert research_campaign proxy ushmi housing yield_curve fidelity_mismatch data_limited exploratory_proxy_campaign no_go_native

Source report: research/findings/specs/recessionalert-ushmi-1617.json

Objective

Independently explore the RecessionAlert inventory row for the U.S. Housing Market Index (USHMI) from the checked-in offline snapshot without live-site requests, credentials, private account data, or crawl-database mutation.

Source and provenance

Page https://recessionalert.com/ushmi/ bound to SHA256 sha256:fafa56a40a0858e39982d1c768ac848a44c8d0e38f20dad163d06343fbfd5ae1, fetched 2026-08-22T23:46:20Z (published 2019-04-15). Linked assets on row: 0. Fourteen 2019/04/*.png chart embeds referenced in HTML remain uncrawled. Superseded crawl revision sha256:da2e61856bd4f308cfc90d5f07e9e0219cebb75c94fdf11691873581eb832e7d was not used.

Disclosed vs inferred

Disclosed (incomplete): USHMI mainline + diffusion narrative; eight housing components; pairing with 10-year minus 1-year Treasury inversion as an AND-style stress filter.

Inferred / proxy: Equal-weight YoY diffusion on FRED PERMIT and HOUST; normalized mainline sum; DGS10−DGS1 month-end inversion gate; SPY/IEF overlay with lag>=1.

Untestable offline: Native component set beyond permits/starts, weights, exact thresholds, chart PNG bytes, product series.

Data spans

SeriesStartEndn
PERMIT1960-012026-06798
HOUST1959-012026-06810
DGS10 / DGS11962-012026-07/08~16k daily
SPY month-end overlay1993-012026-07~403
IEF2002-072026-07—

Missing offline: completions, new/existing sales, months supply, NAHB HMI.

Trial summary (causal lag≥1 unless noted)

variant_idsharpecagrmaxddrisk_off_share
diffusion625_yoy_lag20.8490.103-0.3960.27
mainline_summation_lag10.7470.075-0.3430.56
housing_and_curve_lag10.7050.094-0.5080.12
diffusion50_yoy_lag10.5710.058-0.3840.49
diffusion50_yoy_lag0_control0.7090.071-0.2130.49

Full ledger: research/findings/campaigns/1617_trial_ledger.json. Module: research/research_campaign/recessionalert_1617.py.

Disposition

Limitations

Native parity impossible offline; two-of-eight components only; FRED month-end ≠ first-release Census vintages; lag0 is control_only.

Source artifact: research/results/1617.md