Add related entities disclosed on 527 Schedule R

Form 8871 Schedule R is the political-organisation analogue of Form 990 Schedule R: a 527 names the organisations it is connected to. ?md_irs527_shape profiles the file; this loads it.

Usage

md_ingest_irs527_related(con = NULL, path = NULL, version = "v2024-11-15")

Arguments

con

A connection from md_db().

path

Path to 2024-11-POL-ORGS-SCHED-R.csv.

version

Pinned irs527_related version.

Value

Row counts, invisibly.

Every row lands with ein_b NULL, and that is the source's shape

The file gives (527 EIN) -> (ENTITY_NAME, address). There is no counterparty EIN anywhere in it. Resolving a name to an EIN is a matching problem and belongs to npmatch, so these rows record the counterparty by name and leave ein_b open. PLAN.md §4 made ein_b nullable for exactly this source.

The gap is not small and pretending otherwise would mislead a consumer: of 10,597 distinct counterparty names, 1,564 (14.8%) exactly match a BMF legal name. The rest are sponsoring corporations, unions and trade associations that are largely outside the exempt universe altogether. A consumer treating related_name as resolvable should expect roughly one in seven to resolve.

The whole tail maps to related_entity, deliberately

ENTITY_RELATIONSHIP is free text with 528 distinct values. Connected (54,660) and Affiliated (5,801) are 97.4% of rows; the tail runs Sponsor (103), Partner (73), Connected Organization (56), and on down to single-occurrence strings.

Every one of them maps to relation = 'related_entity' (§4a). A "connected organisation" is the body that established, administers or funds the political organisation - a support and control tie, not loose coordination. That places it with Schedule R rather than in the coalition band, even though the counterparty is usually a corporation or union rather than a nonprofit.

The alternative - reading sub-types out of the free text - would manufacture a taxonomy the filers were never asked to use, and Common candidate-CA 33rd Assembly Dist. is not a relation type. relation_raw keeps the original string, so a later mapping can be derived without re-ingesting. That is the §4 contract for relation_raw and this is the source it was written for.

Note what this source is not: two organisations that both appear as connected to the same political organisation are a coalition, and that is a derived co-membership over a shared counterparty, never a stored edge. §4a.

None (42 rows) is kept rather than dropped: the filer still named an entity on Schedule R and merely declined to characterise the tie. The disclosure is the fact; the label is commentary.

What is dropped

171 rows name the filer itself as its own related entity. Those are self-loops and carry no information, so they are removed and counted.

Rows are then deduplicated to one per (ein_a, related_name, relation_raw) with n_obs recording how many filings asserted it - each amendment is a fresh row in the source, so the raw count overstates distinct relations.

What must not happen

The counterparty name must never become a name variant on the 527's EIN. §10 is explicit that affiliates are not DBAs, and this source most invites the error: "ASSOCIATION OF SAFETY GROUPS" is a different body from "ASGM PAC", and aliasing them manufactures precisely the false match the database exists to prevent. This function writes to ein_relation and nothing else.