# Historical Cohort - Methodology

## Overview

This dataset is an independently reconstructed historical membership record, built entirely from primary and public-record evidence (SEC EDGAR filings, publicly archived index-provider announcements, and independently hardened point-in-time fund-holdings snapshots). It is not copied or redistributed from any official S&P Dow Jones Indices data product.

## Source hierarchy

1. **Hardened point-in-time anchors.** For a fixed set of as-of dates from 2004-09-30 through 2025-09-30, the actual constituent roster is independently extracted from a publicly filed, SEC EDGAR-hosted holdings disclosure (an S&P 500-tracking fund's own regulatory filing as of that date), cryptographically hashed, and treated as ground truth for that date.
2. **Constituent event ledger.** Between anchors, every addition, removal, merger, rename, share-class change, and corporate separation is independently evidenced against a public, dated source (an SEC filing or a publicly archived announcement) before being added to the ledger.
3. **Identity continuity.** Where a security's printed name changes without a change in the underlying registrant (a rename, a share-class relabeling), Persistfolio's identity layer links the before/after labels to one stable security ID using corroborated identity evidence (shared SEC CIK where available), never a name-similarity guess.

## Reconstruction and reconciliation

Each anchor-to-anchor interval's event ledger is chronologically replayed forward from the opening anchor's real roster and reconciled against the next real, independently hardened anchor. A reconciled interval means the replayed roster exactly matches the real anchor roster at its close, accounting for corroborated identity continuity. See `reconciliation_summary.csv` for every interval's raw and canonical constituent counts and reconciliation status.

As of this release, the whole-history forensic QA gate status is: **PASS**.

## Boundary and uncertainty semantics

- `membership_start_basis` / `membership_end_basis` and `first_roster_business_date_basis` in `constituent_events.csv` disclose whether a date is a directly evidenced exact date or a bound derived from a later/earlier anchor when the exact date could not be independently evidenced.
- `effective_date_status` is `UNKNOWN` on the narrow set of events where a primary source stated an effective date was "to be announced" and no later source with the exact date was found; the date is left blank rather than estimated.
- `confidence` and `review_status` on every row disclose Persistfolio's own internal confidence and review classification for that row, rather than presenting uniform, invented precision.

## Holdings-artifact exclusions

A small number of anchor-roster listings reflect merger-consideration or custodial holdings artifacts rather than genuine index membership (e.g. a temporary tracking-stock or escrow holding retained in a fund's regulatory filing after an acquisition closed). Where independently documented, these are excluded from the canonical roster count and disclosed in `reconciliation_summary.csv`'s `holdings_artifact_exclusions_applied` column; the raw, unedited anchor filing itself is never altered.

## Known, disclosed limitation: one identity-key collision

Persistfolio's forensic review has identified exactly one case in the covered history where two genuinely distinct, simultaneously active securities share one Persistfolio stable security ID, because their registrant identities could not be distinguished using corroborated CIK evidence alone (a 2007-2019 dual-class media holding-company pair). This is a disclosed, narrow Release 1 imprecision affecting one identity group, not a general defect in the identity-resolution methodology. Date ranges for both securities' membership periods are independently correct; only the shared security ID is imprecise for that one pair.

## Unresolved CIK coverage

Not every historical security in this release has a resolved SEC CIK. `issuer_identity.csv`'s `cik_source` column discloses `UNRESOLVED` explicitly wherever Persistfolio's already-corroborated identity sources (a hand-curated issuer registry, a corroborated identity alias, or an SEC current-title cross-reference) did not resolve one, rather than ever presenting an invented or merely plausible CIK.

## Independent / non-affiliation disclaimer

Persistfolio is independent and is not affiliated with, sponsored by, endorsed by, or authorized by S&P Global or S&P Dow Jones Indices. S&P 500® is a trademark of S&P Dow Jones Indices LLC and/or its affiliates. This product is an independent research reconstruction and is not an official S&P dataset.