[MaxGhenis]

Summary

TAXSIM input has no student status, and the emulator does not default anyone to being a student (#532, where a 19-year-old dependent is a non-student). PE-US's is_full_time_student reads is_in_k12_school, which PE-US imputes for ages 5 to 17. A spouse aged 17 or younger on a joint return was therefore a full-time student, and IRC 21(d)(2) deemed them to have earned income for the child and dependent care credit. TAXSIM does not.

26 U.S.C. 21(d)(2) (LII): "In the case of a spouse who is a student or a qualifying individual described in subsection (b)(1)(C), for purposes of paragraph (1), such spouse shall be deemed for each month during which such spouse is a full-time student at an educational institution, or is such a qualifying individual, to be gainfully employed and to have earned income of not less than— (A) $250 if subsection (c)(1) applies for the taxable year, or (B) $500 if subsection (c)(2) applies for the taxable year."

This PR pins is_full_time_student to False for every person, through _pin_input in the runner (so it holds in the marginal-rate branch) and as a situation input on the single-household path. The README's input section states the convention.

Evidence

Expected values are output from the bundled taxsimtest-osx.exe (build cd2026081819, sha256 4a17af9c…). The bundled taxsim35-osx.exe (cdate-2025Aug23) gives the same v24 on the 2021–2024 rows (it does not run 2025).

Row	Year	Spouse age, wages	Children, care	taxsimtest v24	Emulator before	After
1	2024	17, $0	1, $3,000	0	600	0
2	2024	17, $2,000	2, $7,000	400	1,200	400
3	2024	40, $0 (control)	1, $3,000	0	0	0
4	2023	17 (both spouses), $0	2, $6,000	0	1,200	0
5	2025	16, $1,000	1, $3,000	200	600	200
6	2021	17, $0	1, $3,000	0	1,500	0

fiitax matches taxsimtest on all six rows after the change (it differed on rows 1, 2, 4, 5 and 6 before).

Found by Tax-Calculator's PE-TAXSIM validation (PSLmodels/Tax-Calculator taxcalc/validation/pe_taxsim): the five c22–c24 units with a CDCC difference all have a spouse aged 17 with $0–$3,000 of wages. With this change the emulator's v24 matches Tax-Calculator on all five.

One of the five moves away from taxsimtest: c22 unit 1101 (2022, joint, ages 19 and 17, spouse wages $2,000, primary self-employment income $326,000, child care $6,000) goes from $600 to $400, while taxsimtest and TAXSIM-35 give $600. Their $600 does not depend on the spouse's age: both binaries still give $600 with the spouse aged 40, and $400 once the self-employment income is removed. The old match came from the deemed earnings offsetting a separate difference tied to self-employment income.

Scope

Every PE-US tax formula that reads is_full_time_student for a dependent gives the same result below age 19 (the qualifying-child age test, Indiana's dependent child), and college status already defaults to false. Minnesota's K-12 credit reads is_in_k12_school, which is unchanged. So outputs change only for a married couple filing jointly with a spouse aged 5 to 17, since IRC 21(d)(2) deems earnings only for a spouse. Single heads aged 16 and 17 do not change. The federal CDCC moves, and with it state taxes that start from it; the Hawaii CDCC floor and Oregon WFHDC joint-return test read the same flag. Independent review ran 2,600 records (10 household types, 52 state codes, 2021–2025) through main and this branch: only the 359 married-minor records changed, and against taxsimtest v24 and fiitax now match on all 520 minor-filer records (208 before). One Missouri siitax value moved about $4 further from taxsimtest; that was not traced.

Invariants (tested)

For a joint return with wages only, children under 13, and a spouse aged 5 to 17 (Hypothesis, years 2021–2025):

1. No deemed earnings: v24 ≤ 50% of the lower earner's wages.
2. Age neutrality: v24 is the same as for the otherwise identical spouse aged 40.
3. Path agreement: the runner and the single-household path give the same v24 and fiitax.
4. Every person's is_full_time_student is False in both paths, and the pin is a registered input (so it survives the MTR branch).

On main the probe rows fail on both paths, and Hypothesis finds a counterexample (2025, spouse aged 8 with no wages, v24 = 1,200).

Tests

• New tests/test_student_status.py: the six probe rows on both paths, a macOS check that the constants match the bundled binary, the no-student check, and the property test. 15 pass locally (policyengine-us 2.58.0, Python 3.13).
• tests/test_mtr_branch_overrides.py: is_full_time_student added to EXPECTED_OVERRIDES; 8 pass.
• tests/test_mappers.py: the expected situations now carry the new per-person input.
• ruff format --check . passes. CI passes on Ubuntu, macOS and Windows at e5229fe2.

Plan

• Pin the input on both paths, with README note and changelog fragment.
• Probe and property tests, failing on main.
• Full suite (CI, three platforms).
• Independent review (Opus 5.5; approved at e5229fe2). It found the test_mappers fixtures, fixed in e5229fe2, and the c22 unit 1101 note above.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
policyengine-taxsim	Ready	Preview	Oct 22, 2026 2:45pm UTC

[MaxGhenis]

Merging at e5229fe: CI passes on Ubuntu, macOS and Windows (plus lint, changelog, S-corp validation and dashboard). Independent review by an Opus 5.5 reviewer approved this head after the test_mappers fixtures were added; its should-fix (the c22 unit 1101 note) and scope wording are in the body. Expected values are taxsimtest output (cd2026081819).