Skip to contents

Standardize the 990-PF recipient foundation-status field (SIGOCPYRFSTA)

Source: R/recipient-type.R
pf_recipient_type.Rd

Form 990-PF Part XIV Table 3a, column (c) asks the filer for the "foundation status of recipient". The IRS supplies no controlled vocabulary, so the field is free text: across 155,000 sample grants it carries ~1,400 distinct spellings of what are really a handful of concepts. This module collapses that free text onto a five-level standard, which is what decides whether a grant is even eligible to appear in the IRS Business Master File.

Why a rule cascade and not a lookup table

A lookup can only cover the spellings already seen. The rules below key off the concept — an IRC section reference, the words "public charity", the words "operating foundation" — so unseen spellings land in the same bucket. Two normalizations do most of the work:

  • section collapsing — 501(c)(3), 501C 3, 501 C 3, and 501c3 all become the single token 501C3; likewise 509A1, 170B1A, 4942J3. Without this, each spacing variant needs its own pattern.

  • de-spacing — a second, space-free copy of the string is matched in parallel, so PUB CHAR, PUBLIC CH, and OTHERPUBLICCHARITY are all caught by one PUBCHAR / PUBLICCH pattern.

Standard levels (recipient_type)

  • public_charity — nonprofit, public charity (PC, 501(c)(3) with no foundation qualifier, 509(a)(1)-(3), church, school, hospital)

  • nonprofit_other — nonprofit, other subsection or exempt-but- unspecified

  • foundation_pf — private foundation, private operating foundation, exempt operating foundation

  • individual — a natural person

  • other_unknown — blank, N/A, government, foreign, and text that answers a different question ("PAID", "CASH", "VARIOUS", "ACTIVE")

Detail levels (recipient_type_detail)

A finer split carried alongside the standard, because other_unknown mixes cases that behave very differently: government and foreign can never be in the BMF, while blank and unrecognized are usually organizations the filer simply did not code. Only the detail level is precise enough to drive the out-of-sample decision — see pf_oos_details().