Skip to content

PUF donor uprating overstates mean-reverting distribution series: non_sch_d at 2.39x the SOI concept #487

Description

@MaxGhenis

Found adjudicating Build N attempt 2 (#462 thread). Measured on the sha-pinned puf_2024.h5 donor via puf_tax_unit_donor_from_arrays:

donor (E01100 lineage, uprated to 2024) SOI Pub 1304 Table 1.4 TY2023
weighted carriers 4.67M 3.21M (1.45×)
weighted mass $24.31B $10.16B (2.39×)
weighted mean $5,206 $3,165 (1.64×)
weighted median $1,450

The eCPS-era pipeline produced $13.69B / 4.01M units from the same PUF lineage (recorded in #462's motivation), so the current ETL's 2024-level uprating roughly doubles what the retired pipeline carried. Capital-gain distributions are a mean-reverting series (TY2015 — the PUF base year — was a historically heavy distribution year); uprating its level by a generic growth factor overstates the modern concept on both margins.

Interim: #486 pins the column's level to the ledger-fed aged Table 1.4 dollar fact at release time (the SS-components repair precedent), with the factor recorded in support_value_repairs. That is honest but uniform — it preserves the donor's (too-rich) shape.

Root fix: variable-specific donor uprating for distribution-class series — age E01100-class amounts by their own SOI chain (the #474 tips mechanism generalized) either in the irs-soi-puf ETL or at populace donor load, with the chain facts in the ledger. Candidates beyond non_sch_d: the other Table 1.4 rows the #464 blanket watches, and the 13 tail-concentrated sparse columns from the N attempt-2 battery (those also implicate #481's weighted-leaf-draw loss).

Relates #462, #464, #481, #486.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions