Index Washington WFTC maximum qualifying income to the current tax year

[DTrim99]

Summary

wa_working_families_tax_credit_maximum_qualifying_income computed the income ceiling from the federal EITC parameters frozen at 2022-06-09, returning the 2022 dollar limits for every year ($49,620 for married-filing-jointly with one child). Filers whose income fell between the 2022 and current-year federal EITC limits were wrongly denied the credit.

Evidence

Washington DOR, WFTC tax year 2023 (workingfamiliescredit.wa.gov) — maximum qualifying income:

Children	Single/HoH	MFJ
1	$46,560	$53,120

These match the 2023 federal EITC limits exactly (2022's MFJ-one-child figure was $49,622), so Washington indexes the ceiling to the current year.

RCW 82.08.0206(2)(d): "'Internal revenue code' means the United States internal revenue code of 1986, as amended, as of June 9, 2022, or such subsequent date as the department may provide by rule…" This pins the IRC conformity version, not the tax-year amounts; that Code still contains the EITC inflation-adjustment provisions (26 U.S.C. §32(j)), which index the limits to each year. The federal EITC structure is unchanged 2022–2024 (the 2021 ARPA expansion had expired by June 9, 2022), so the current-period parameters carry only the annual indexing.

Change

Use the current-period federal EITC parameters for the maximum-AGI ceiling. On the issue household (WA joint 2023, one child, AGI $50,047.89) the credit goes $0 → $403.29 (max $625 less 11.5% × ($50,047 − $48,120)), matching WA DOR and the reported expectation.

Testing

Updated the WFTC tests whose incomes were positioned against the frozen 2022 ceilings — Cases 4/5/6b in the baseline suite and three analytics-coverage edge cases — to the current-year ceilings (each recomputed from the statutory phase-out against the DOR/federal limit), and added a 2023 regression test (Case 16). Full WA income-tax suite + partners edge cases pass (64 tests).

Found via PolicyEngine-TAXSIM discrepancy #1155, reported by @feenberg.

Fixes #9350.

Note

This reverses a deliberate freeze (eitc("2022-06-09"), added with a detailed comment). See the design question flagged to @PavelMakarchuk below.

🤖 Generated with Claude Code

[DTrim99]

@PavelMakarchuk flagging a design question here, since this reverses a deliberate choice.

The WFTC code froze the federal EITC at eitc("2022-06-09") — with a detailed comment reading RCW 82.08.0206(2)(d) ("IRC as in effect June 9, 2022") as pinning the dollar values to 2022. This PR reads it instead as pinning only the IRC conformity version, with the amounts still indexing per IRC §32(j), so the maximum-qualifying-income ceiling tracks the current year.

The deciding evidence is that Washington DOR publishes the current-year federal EITC limits — $53,120 for MFJ/one-child in 2023 (vs the frozen $49,622), which is exactly what a filer at AGI $50,047 needs to receive the $403 credit they're due (taxsim #1155). So DOR administers with indexed limits.

Two things I'd value your read on:

1. Whether you agree the freeze should apply to rules/structure (age window, investment-income limit, ARPA exclusion) but not the inflation-indexed amounts. This PR only changes the income ceiling; the same question applies to the frozen investment-income limit in wa_working_families_tax_credit (2022 $10,300 vs 2023 $11,000), which I left alone.
2. Whether keying the ceiling to the current-period federal EITC is the right mechanism, or whether we'd want a cleaner "2022 rules + current-year indexing" construct that stays robust if the federal EITC structure changes after 2024.

Happy to adjust if you'd prefer a narrower or different approach.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (724e795) to head (b2390e9).
⚠️ Report is 29 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9351   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            1         1           
  Lines           35        24   -11     
  Branches         2         1    -1     
=========================================
- Hits            35        24   -11     

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[hua7450]

@DTrim99 @PavelMakarchuk

Verified the current-year indexing claim against all four years of DOR's official application instructions. Every year's published AGI ceiling is exactly that year's federal EITC maximum AGI — including 2022 itself, so the frozen values were only ever correct for 2022:

Tax year	Instructions (Section C eligibility)	AGI ceiling, 1 child (Single / MFJ)	Investment income limit
2022	2022 instructions	$43,492 / $49,622	"less than $10,300"
2023	2023 instructions	$46,560 / $53,120	"less than $11,000"
2024	2024 instructions	$49,084 / $56,004	"less than $11,600"
2025	2025 instructions	$50,434 / $57,554	"less than $11,950"

So the fix to wa_working_families_tax_credit_maximum_qualifying_income is right, and the legal reading holds: RCW 82.08.0206(2)(d) pins the version of the IRC, and that version's own inflation provision (26 U.S.C. §32(j)) indexes the dollar amounts each year.

Requested change: the last column above shows the investment income limit indexes the same way — each year's instructions state the current-year federal §32(i) cap, never the frozen 2022 $10,300. But the model still reads it from the frozen snapshot in two places:

• wa_working_families_tax_credit.py (frozen_eitc.phase_out.max_investment_income, used by both the baseline and state-only paths)
• wa_working_families_tax_credit_age_expansion_eligible.py (same expression)

That wrongly denies the WFTC to filers with investment income between $10,300 and the year's limit ($11,000 in 2023, rising to $11,950 in 2025). The federal parameter gov.irs.credits.eitc.phase_out.max_investment_income is already populated and IRS-cited through 2026, so the fix is the same one-line pattern as this PR: read it from parameters(period) instead of the frozen snapshot.

Precedent: this is the same correction the repo already made for Indiana's static conformity — #8875 and #8902 switched in_eitc to the current-year federal EITC, and the comment there states the principle directly: post-snapshot federal changes "are inflation indexing only. The current-year federal EITC therefore satisfies the freeze." Oklahoma is the counterexample that justifies keeping the pattern: OTC's Form 511 packet reprints the 2020 EITC tables, so OK's eitc("2020-01-01") freeze matches agency practice and should stay.

The remaining frozen reads here (childless age band 25–64, disabled-dependent age waiver, separate-filer rule) are structural rules rather than indexed dollars — the instructions confirm 25–64 in all four years — so those should stay on the 2022-06-09 snapshot.

[PavelMakarchuk]

PR 9351 — Full Review Report

Title: Index Washington WFTC maximum qualifying income to the current tax year
Author: DTrim99 | Fixes: #9350 (TAXSIM #1155, #986)
PR head: b2390e95eea8654d8d2305102af3fc0c2e730d10 | Merge-base: e2593814059788e69c9cf4380154ebd9b960f7da
Branch status: BEHIND=0, AHEAD=1 — current. CI: 34/34 checks pass.
Mode: full review (context, regulatory, code, tests, verification-planner, consolidator)

Verdict

Severity: COMMENT — 0 CRITICAL, 5 SHOULD ADDRESS, 7 SUGGESTIONS.

All three Phase 4 roles found zero critical issues. Every changed or added expected
value (5 partner lines, 3 repositioned partner inputs, main-file 50.02 / 345.02 /
990.83 / 403.29) was independently re-derived by both the tests and regulatory roles
and matches WA DOR-published limits within stated margins. The regulatory evidence
strongly supports the PR's legal reading (see Design question below).

The review recommends COMMENT rather than APPROVE for a specific reason: this PR
contains two decisions that belong to the maintainer, not the review — (1) the
PR body explicitly routes the freeze-vs-index design question to @PavelMakarchuk,
and the original freeze was a deliberate, commented design decision; the review
presents its evidence-based verdict below but intentionally does not exercise
that decision on the maintainer's behalf; and (2) the PR edits a partner API
contract test file, which under CLAUDE.md requires the maintainer to consciously
run the three-question partner gate before merge — something no reviewer or
subagent can do for him.

---

Design question: freeze vs index

The PR body flags this to the maintainer: the previous behavior — computing the WA
WFTC income ceiling from federal EITC parameters frozen at
parameters.gov.irs.credits.eitc("2022-06-09") — was a deliberate design
choice, and this PR reverses it to parameters(period).gov.irs.credits.eitc
(current-year indexed limits).

The regulatory role's verdict: INDEX is correct. Evidence:

1. Statutory text. RCW 82.08.0206(2)(d) (fetched from app.leg.wa.gov,
   2026-08-27) defines the conformity clause verbatim:

   "'Internal revenue code' means the United States internal revenue code of
   1986, as amended, as of June 9, 2022, or such subsequent date as the
   department may provide by rule consistent with the purpose of this section."

   This pins a version of the Code, not a set of dollar amounts. RCW
   82.08.0206(2)(e)(i) defines the ceiling by reference to a Code provision
   ("The maximum federally adjusted gross income under Title 26 U.S.C. Sec. 32"),
   not a fixed dollar figure.

2. The pinned Code self-indexes. 26 U.S.C. 32(j) — inside the Code as of
   June 9, 2022 — provides that "each of the dollar amounts in subsections (b)(2)
   and (i)(1) shall be increased by ... the cost-of-living adjustment ... for
   the calendar year in which the taxable year begins." So the frozen Code,
   applied to tax year 2023, produces the 2023 dollar limits. A static conformity
   date freezes rules (it would block a post-2022 congressional restructuring of
   the EITC), not the operation of the indexing formula those rules contain — the
   standard static-conformity result seen in other fixed-date states.

3. Agency practice (decisive). WA DOR's per-year application pages
   (workingfamiliescredit.wa.gov, fetched 2026-08-27) publish current-year
   indexed limits for every tax year 2022–2025 — e.g. MFJ/1-child: $49,622
   (2022), $53,120 (2023), $56,004 (2024), $57,554 (2025). Each row equals that
   year's indexed federal EITC completed-phase-out limit (Rev. Proc. 2022-38,
   2023-34, 2024-40). If the statute froze the dollar amounts, all three
   post-2022 rows would read $49,622. DOR's 2025 page also publishes an indexed
   investment-income cap ($11,600), not the frozen 2022 $10,300. WAC 458-20-285's
   worked example applies the 2022 thresholds to 2022-tax-year filers — each tax
   year uses its own year's thresholds.

4. No confounding federal change. No post-2022 federal law (including OBBBA)
   restructured the EITC amounts/phase-outs, so PolicyEngine's current-period
   federal EITC parameters coincide with "the 2022-06-09 Code's 32(j)-indexed
   amounts" for every year the model covers. parameters(period) is therefore
   the legally correct value for 2023+ (and inert for 2022 itself).

Why the freeze existed (git history). git log -S 'eitc("2022-06-09")' traces
it to commit 23a4f57083 ("Fix federal-state tax conformity for EITC, deductions,
and ALDs," issue #8243, 2026-06-07, D. Hansell) — a deliberate static-conformity
sweep (paired with the Indiana SEA 243 snapshot), with the date kept as a Python
literal because core lacks date-valued parameters. It was a considered legal
reading, but one that conflated freezing the Code version with freezing the
dollar amounts. The DOR tables show that reading does not match administration,
and TAXSIM #1155 (@feenberg) independently caught the discrepancy.

Bottom line: the review's evidence supports indexing. The final call is the
maintainer's, since the original freeze was deliberate and the PR body directed
the question to him. Note also that accepting INDEX logically commits the model to
finding A1 below (the investment-income cap embodies the opposite reading).

---

PARTNER CONTRACT TEST — maintainer gate required

The PR edits
policyengine_us/tests/policy/baseline/partners/analytics_coverage/edge_cases/state/wa/working_families_tax_credit.yaml
— an API partner contract file under tests/policy/baseline/partners/**.

What changes:

• 5 expected-value lines: wa_working_families_tax_credit 223.25→230.21 and
  50.06→50.42 (both up); 3 adjusted_gross_income lines tracking repositioned
  inputs; a third credit case stays 0.0 at a higher income.
• 3 inputs are REPOSITIONED (employment_income_before_lsr):
  15,000→18,000; 43,490→51,590; 43,495→51,600. Repositioned inputs change what
  the cases test, which is more than value churn — though the code role verified
  each case's semantic contract is preserved at the new indexed ceiling:

Case	Before	After	Role preserved?
wa_wftc_childless_mid_phase_out	15,000 mid-band under frozen ceiling 16,480.26; credit 223.25	18,000 mid-band under 2026 ceiling 19,539.74; credit 230.21	Yes — repositioning was required: 15,000 would now sit below the band
wa_wftc_1_child_min_amount_near_ceiling	43,490 ($0.45 below frozen ceiling); 50.06	51,590 ($3.38 below 2026 ceiling 51,593.38); 50.42	Yes — still just-below-ceiling probe
wa_wftc_1_child_above_income_ceiling	43,495 → 0.0	51,600 ($6.62 above ceiling) → 0.0	Yes — still the cliff probe

Structure: all 7 cases retained; no cases renamed/added/removed, no keys
added/removed. The tests role re-derived all partner expected values from the
snapshot parameter tree — every one reproduces (including the two unchanged
values 674.96 and 1,360.14).

The gate: the PR body notes the partner edge cases were updated and pass, but
does NOT explicitly acknowledge the partner-facing API impact or the CLAUDE.md
three-question gate (confirm the edit / team notified / partner notified). The
review cannot verify notification. Per CLAUDE.md this is a partner-facing API
change the maintainer must consciously accept before merge — partner values that
downstream consumers see change (223.25→230.21, 50.06→50.42, three AGI lines).

---

Findings

CRITICAL

None.

SHOULD ADDRESS

A1. Investment-income cap still frozen at 2022-06-09 while the AGI ceiling now
indexes — internal inconsistency. (regulatory + code, converged; verification
planner settled reachability)
policyengine_us/variables/gov/states/wa/tax/income/credits/wa_working_families_tax_credit.py:34,67-70
and wa_working_families_tax_credit_age_expansion_eligible.py:49-53 still gate on
frozen_eitc.phase_out.max_investment_income = $10,300 (2022) for all years. The
PR's own statutory logic — 32(j) also indexes 32(i)(1), base year 2020 — implies
the cap should be current-year ($11,000 in 2023, $11,600 in 2024, $11,950 in
2025), and DOR's 2025 page publishes the indexed figure. Reachable behavioral
gap: a 2023+ filer with investment income between $10,300 and the current-year
cap now clears the newly indexed AGI ceiling but is still wrongly denied — the
same class of error this PR fixes. Small population, but the model now embodies
two contradictory readings of the same clause in adjacent files, and after this
PR Washington is the only conformity implementation in the tree mixing frozen and
current-period reads of the same federal credit within one program (Oklahoma
freezes every component consistently). The comment at
wa_working_families_tax_credit.py:29-33 still describes the freeze as pinning
values, contradicting the new comment at
wa_working_families_tax_credit_maximum_qualifying_income.py:20-28. Recommend
extending the fix to max_investment_income (this PR or a tracked follow-up
issue) and reconciling the comments. The frozen reads of structural values
(age.min/min_student/max, separate_filer) are harmless — identical at both
instants.

A2. Partner contract gate not acknowledged. See the PARTNER section above —
the maintainer must run the CLAUDE.md three-question gate before merge; the PR
body does not acknowledge it, and the review cannot verify team/partner
notification.

A3. Stale frozen-2022 arithmetic in partner-file comments. (tests + code)
In the same partner file, two cases whose expected values did not change still
document frozen ceilings: wa_wftc_joint_1_child_max_amount ("phase-out starts
at (20,130 + 6,130 + 3,733/0.1598) − 5,000 = 44,620" — post-PR the 2026 start is
53,863.38) and wa_wftc_3_children_max_amount ("... = 54,190" — post-PR
65,243.57). The header's opening line still says the credit is "pinned to the
federal EITC rules as in effect 2022-06-09" while lines 5-8 say the ceiling
indexes — true of structural rules only, not dollar limits; the file now
documents two contradictory ceiling regimes. Outputs remain correct (income
$20,000 is below either start). Any fix touches the partner file, so it must go
through the same three-question gate (comment-only edits still touch the
contract file).

A4. No direct pin of the ceiling at a DOR-published figure. (tests)
The PR's statutory argument rests on DOR publishing $53,120 (2023 MFJ/1-child),
and the model derives it exactly (21,560 + 6,560 + 3,995/0.1598 = 53,120.00;
3,995/0.1598 = 25,000 exact) — a clean, margin-free anchor. Case 16 pins it only
indirectly through credit arithmetic. Adding
wa_working_families_tax_credit_maximum_qualifying_income: 53_120 to Case 16's
output block would make the regression self-documenting and catch a future
parameter re-index silently shifting the credit. (The only existing direct
assertion of this variable is the 2029 cash-assistance case, which exercises the
other branch.)

A5. No 2-child near-ceiling coverage. (tests)
Limit-dependent values are pinned for 0-child (2026), 1-child (2023/2024/2026),
and 3-child (2025), but the only 2-child case sits at $20,000 earnings — far
below the band, so its output is ceiling-independent. The 2-child bracket has
distinct max (7,152 in 2025) and rate (0.2106) rows; a 2-child in-band or
near-ceiling case (2025 single 2-child ceiling = 57,310.72) would close the
bracket sweep.

SUGGESTIONS

S1. Document the divergence/maintenance condition. (regulatory + code,
merged) Current-period parameters equal the frozen-Code-indexed amounts only
while Congress leaves the EITC structure untouched; the variable comment's
justification ("unchanged from 2022 through 2024") is time-bounded while the
formula applies to all years (tests reach 2026). Broaden the comment at
wa_working_families_tax_credit_maximum_qualifying_income.py:26-28 to state the
rule: if a future federal law changes EITC amounts/structure (not mere indexing),
the WFTC must revert to computing the 2022-06-09 rules with 32(j) COLAs rather
than reading current-period parameters — unless DOR advances the conformity date
by rule under (2)(d).

S2. DOR rounding. DOR publishes IRS-rounded limits ($46,560) while the model
derives unrounded ceilings ($46,560.08; 2026 HoH/1 $51,593.38). Cent-level,
immaterial, inside margins — but if exact DOR/IRS agreement is ever wanted, round
to the published Rev. Proc. figures.

S3. Reference fields. The variable's reference still points only to the
ESSB 6346 PDF (p. 59); consider adding RCW 82.08.0206 and a DOR per-year limits
page, which now carry the indexing rationale stated in the comment.

S4. Fractional-cent test inputs. New Case 16 uses
employment_income: 20_412.535 / 29_635.352 — reproduces the taxsim #1155
household exactly (defensible for a regression case); a rounded pair in the same
band would be marginally more readable. Cosmetic.

S5. Exact-ceiling boundary / $50 floor never strictly binds. (tests, merged)
The 2026 cliff is bracketed ($51,590 → 50.42; $51,600 → 0) but no case sits
exactly AT the limit. Because eligibility is <= and rate = (max − 50)/band,
income exactly at the ceiling yields exactly $50.00 — one case (natural place:
the exact 2023 MFJ ceiling $53,120.00) would pin both the <= semantics and the
floor. Relatedly, the max_(p.min_amount, ...) floor binds only at the
ceiling; both current "minimum" cases (50.42, 50.02) are above it, and Case 4's
comment / the partner case name min_amount_near_ceiling slightly overclaim.

S6. Childless MFJ and MFS near-ceiling untested. The 0-child joint-bonus row
($7,280 in 2026) is never exercised — all childless cases are single, all joint
cases have children. A childless joint couple near the 2026 joint childless
ceiling ($26,819.74) would cover it; MFS near-ceiling likewise.

S7. 2022 inertness undemonstrated. For 2022, parameters(period) and the
frozen call resolve identically (no mid-2022 parameter entries), so the change is
inert for 2022 by construction — but no existing 2022 case exercises the
phase-out band, so no test would notice if 2022 did shift. A 2022 in-band case
would document this cheaply. Low priority.

---

Behavioral impact (quantified)

Real, positive for WA households, 2023 onward, on two margins:

• Extensive: filers with max(earned income, AGI) between the frozen-2022 and
  current-year ceilings gain eligibility, $0 → credit. The 2025 single/1-child
  band is ~$6.9k wide ($43,492 → $50,434). The issue #9350 household goes
  $0 → $403.29 (Case 16, 2023). Newly eligible filers land in the phase-out, so
  gains run from the $50 statutory minimum up to a few hundred dollars.
• Intensive: every filer already inside the phase-out gets a larger credit
  because the phase-out start (ceiling − $2,500/$5,000) shifts up with the
  ceiling (partner case: 223.25 → 230.21).

Population-level WFTC totals (WA state benefit spend) rise modestly for 2023+.
Takeup caveat: the formula gates on takes_up_eitc, and WFTC is
application-based with historically low takeup, so real-world aggregate gains are
smaller than full-takeup microsim deltas.

Discriminator confirmed three times over — cases that would FAIL under the frozen
formula: Case 16 (2023: frozen ceiling $49,620.45 < income $50,047.89 → $0
pre-PR, $403.29 post); partner childless_mid_phase_out (2026: $0 pre-PR →
230.21); partner 1_child_min_amount_near_ceiling (2026: $0 pre-PR → 50.42).

---

Validation summary

Phase / role	Status	Result
Context	done	4 changed files: 1 variable, 1 baseline test, 1 PARTNER contract test, 1 changelog. No parameter files changed.
Regulatory (incl. PDF/source audit)	done	0 CRITICAL. Design verdict: INDEX, with statute + 32(j) + DOR 2022–2025 tables. Sources: RCW 82.08.0206, 26 U.S.C. 32, WAC 458-20-285, workingfamiliescredit.wa.gov per-year pages. No new PDF pulls needed beyond these (ESSB 6346 governs the untouched 2029 expansion).
Code patterns	done	0 CRITICAL. Whole-tree frozen-instant-call sweep: pattern is conventional (~12 sites, OK/AR/federal snapshots); post-PR WA is the only mixed frozen/current implementation → A1. Changelog fragment valid; fixed is the right type. Clean on hard-coding, naming, aggregation, periods, references, entity mismatches.
Tests / edge cases	done	0 CRITICAL. All 11 changed/added expected values hand re-derived from the snapshot parameter tree — all match. Year sweep 2023–2026 covered; gaps → A4, A5, S5–S7.
Reference-checker	folded into regulatory + code	No reference fields changed anywhere in the diff; no #page=XX anchors added — no separate pass needed.
Phase 5 verification queue	NONE	All roles at 0 CRITICAL; both re-derivations agree with DOR. Investment-income-cap candidate judged and settled as SHOULD ADDRESS (A1), not a Phase 5 mismatch.
5E (page-anchor checks)	N/A	No new page-anchored references.

Branch status: BEHIND=0, AHEAD=1 — current against main. CI 34/34 green at
b2390e95ee.

Recommendation

COMMENT. Zero criticals; the model change is verified correct against
statute, the self-indexing mechanism in the pinned Code, and four years of DOR
administered practice — but two items require the maintainer's conscious
decision, which this review deliberately leaves to him: (1) accepting the
reversal of the deliberate 2022 freeze (the review's evidence supports indexing;
the call was explicitly routed to @PavelMakarchuk and the review does not
exercise it on his behalf), and (2) running the CLAUDE.md three-question partner
gate for the contract-test edits. A1 (the now-inconsistent frozen
investment-income cap) should be resolved in this PR or a tracked follow-up as
part of decision (1), since accepting INDEX for the ceiling logically entails it.