RecessionAlert leading U.S.-stock indicator — exploratory proxy campaign

Objective

Replicate the distinct RecessionAlert row MODELS: A leading indicator for U.S stocks | RecessionAlert from the checked-in canonical snapshot without live requests, database mutation, credentials, private account data, or invented native rules. Because the source omits observations and executable thresholds, the required best-effort path tests clearly labeled public proxies separately from source fidelity.

Result

The May 1, 2024 article discloses two international breadth concepts: the net percentage of 39 OECD countries with rising leading economic indices (LEIs), and the net percentage of 39 central banks easing rates. It reports 7- and 10-month maximum leads to the NYSE annual percentage change, respectively, says the aggregate has a 0.57 R-squared to NYSE, points to a monthly Global Economic Report / CRB Monthly charts location, and describes dips as buying opportunities.

The source snapshot contains no source-linked observations, exact netting or aggregation formula, weights, thresholded state mapping, missing-input behavior, release clock, vintage/revision policy, or executable buy/sell/hold/exit rule. The lead and R-squared values remain source claims, not independently recomputed metrics. Native disposition remains insufficient_evidence; no official or playable Other-indicators entry is registered.

As a separate exploratory exercise, eight inferred INDPRO and UNRATE variants were run against transparent SPY risk-on and IEF/BIL risk-off sleeves. The 1993-01 through 2026-06 common window is the longest primary SPY-aligned comparison available locally. Lookback, threshold, family, and lag neighbors are recorded in the trial ledger; underperforming and otherwise negative outcomes are retained. These proxies do not recover the source's 39-entity identity.

Evidence

• Full source SHA256 sha256:e5a324ca2890226d7f62553757130605a5335c6b98bd88e4a67d22b2ca76ea01 matches the manifest, read-only SQLite body digest, and inventory prefix. The selected body is HTTP 200, 83,175 bytes, fetched 2026-08-22T23:48:28Z.
• Native span is one dated article snapshot. The approximately 843-day article-to-crawl interval is archival snapshot delay, not economic publication lag. Point-in-time component availability remains unknown.
• StocksLEI.gif is referenced but absent from the six-row canonical asset table; raw HTML/media/workbook bytes are not published.
• research/campaigns/recessionalert-leading-us-stocks/state-space.json records eight inferred variants, public proxy provenance, metrics, and preserved negative outcomes.
• research/campaigns/recessionalert-leading-us-stocks/diagnostics.json records timing, four public context windows per variant, eight proxy event rows, combinations, and failure modes.
• research/campaigns/recessionalert-leading-us-stocks/lab-series.json and docs/site-data/research/recessionalert-leading-us-stocks.json provide the derived monthly signal/equity data for the interactive Research Library page.
• Any future native overlay must use documented point-in-time input availability and at least one available-bar execution lag; same-close/lookahead is a research-only sensitivity. Unknown inputs remain unavailable.

Files and validation

• research/findings/1341.md — durable public-safe native finding with rule, provenance, span, lag, causal policy, gaps, and web exception.
• research/findings/specs/recessionalert-leading-us-stocks.json — machine-readable source-fidelity spec, unchanged.
• research/research_campaign/recessionalert_1341.py — executable read-only proxy campaign and artifact writers.
• research/campaigns/recessionalert-leading-us-stocks/ — campaign manifest, eight-variant state-space, diagnostics, specs, synthesis, and lab series.
• research/experiments/1341.ipynb and research/experiments/1341-best-effort.ipynb — native evidence and executed proxy notebook respectively.
• tests/test_recessionalert_leading_us_stocks.py and tests/test_recessionalert_1341_campaign.py — focused offline source-fidelity and proxy-campaign assertions.
• docs/research/recessionalert-leading-us-stocks.html, docs/site-data/research/recessionalert-leading-us-stocks.json, and docs/experiments/1341-best-effort-analysis.html — canonical public research surfaces.

Validation includes the complete focused proxy test module, read-only database assertion, campaign manifest/payload validation, notebook conversion, Research Library generation, Results generation, and static bundle check. No live-site request or tracked-crawl write is used.

Limitations

The proxy campaign cannot validate the native source's claimed relationship or promote a signal until a separately authorized, redistributable, source-linked series and complete rule specification supply observations, transformations, thresholds, release timing, revisions, and execution semantics. INDPRO/UNRATE and SPY/IEF/BIL are inferred public substitutes only. Keep this work under Other / Research; do not present it as an official RecessionAlert indicator or trading recommendation.