QBI v3 evidence: SCF employer-structure + SOI wage/UBIA prior resources

[MaxGhenis]

Summary

The v3 evidence base: estimated, versioned resources replacing the last invented QBI assumptions, built from the full public SCF 2022 and six SOI by-industry tables (all pre-verified by fetch; provenance and input digests recorded in the resources).

• qbi_employer_structure_v1.json — from the SCF actively-managed-business records (both detail businesses, five implicates pooled with X42001/5): P(headcount > 1 | net-income band × legal form × coarse industry bin), banded size distributions, thin-cell collapse rules (n≥30, documented fallback hierarchy), and empirical profit-margin quantiles from whole-business receipts/net income (X3131/X3132) — the replacement for the invented Beta margins. Carries its judgment_calls and an external_anchor block comparing against JCT's zero-employee splits.
• qbi_wage_capital_priors_v1.json — SOI wage shares (salaries+wages, + cost of labor where published, + officer comp for S-corps) over receipts by industry per legal form (sole prop TY2023 at up to 154-industry detail; partnership TY2023; S-corp TY2022 at ~85 industries), and UBIA intensity from gross depreciable assets over receipts (partnerships, S-corps; sole-prop depreciation-flow ratio marked proxy: true). Per-cell source table/row provenance.
• Builder module + tool (qbi_v3_evidence.py), configurable input paths, synthetic-fixture tests, byte-identical rebuild check.

The finding that governs wiring

SCF owner-only share by headcount = 35.5% vs JCT's 84.2% zero-W-2-employee firms. The gap is definitional and structural (headcount counts unpaid family; SCF's module misses gig-scale Schedule C entirely; detail covers only the two largest businesses). Wiring doctrine for the v3 simulation change: SCF conditionals supply the shape (how employer presence and margins move with business income, form, industry), JCT supplies the level the aggregate must hit. The resource documents this; the wiring PR must implement it rather than consuming SCF levels raw.

Verification (supervisor, independent)

Full workspace suite with restricted artifacts: 3,271 passed, 0 failed; ruff clean; both resources declared; wage-share ranges sanity-reviewed (S-corp high tail reflects officer comp in the wage base, as expected).

Parallel sibling to #536 (both based on repeal-validation-298); rebases trivially in either order. The v3 wiring change (consuming these in the simulation, with the shape-vs-level rule) is the next and final build step.

Refs #530

🤖 Generated with Claude Code