Count gambling winnings in Arkansas gross income and add the Arkansas gambling loss deduction

[MaxGhenis]

Summary

Arkansas taxes gambling winnings and lets itemizers deduct gambling losses up to those winnings, outside the 2% floor on miscellaneous deductions. policyengine-us did neither.

• gambling_winnings was not in the Arkansas gross income sources, so winnings were untaxed in Arkansas.
• The Arkansas itemized sources had no gambling loss deduction.

This PR adds both. It was found while fixing Arkansas itemized deductions (#9614, #9625).

The law

Winnings are gross income. Ark. Code § 26-51-404(a)(1)(E) includes "Gains or profits and income derived from any source whatever", and none of the § 26-51-404(b) exclusions covers gambling. DFA reports winnings as an addition to income on Form AR-OI, carried to AR1000F "Other income": Line 20 for TY2015-2017 and Line 22 from TY2018.

• The only exception is winnings from Arkansas electronic games of skill. Under Ark. Code § 26-51-1310(f), from Act 732 of 2007, § 8 (2007R/732.pdf#page=5), they are not income and their losses are not deductible. DFA Subject 208 applies this to winnings before July 1, 2019, when the two racetracks became casinos. The booklets print the exception through TY2018. PE has one gambling_winnings input and cannot separate these winnings; see Limitations.

Losses: Ark. Code § 26-51-424(a)(2)(B). Act 155 of 2017, § 21 (2017R/155.pdf#page=9) added "including without limitation gambling losses, which are: (i) Deductible to the extent of gambling winnings; and (ii) Not subject to the two percent (2%) limitation on miscellaneous itemized deductions."

• § 25 (#page=10) makes it effective for tax years beginning on and after 2015-01-01.
• No later Act has amended § 26-51-424(a). The 2024 code text matches Act 155 word for word.
• It is Arkansas's own rule and does not adopt 26 U.S.C. § 165(d), so the federal 90% limit that starts in 2026 does not apply.
• Ruled by Max (d1018): as built, including pre-2015 years. DFA Subject 208 stated the same cap and exemption from the 2% floor in its 2009 version (Wayback capture, PDF created 2009-01-08), and DFA's impact statement says Act 155 "clarifies" existing law with no revenue impact. This ruling retains the deduction before Act 155 § 25's 2015 statutory start; the existing 2014 regression case pins that behavior.

DFA forms by year. All anchors are PDF page indices; the original PR work reported checking them with pdftotext -f N -l N. The base URL is https://www.dfa.arkansas.gov/wp-content/uploads/.

TY	Winnings	Losses on AR3	Booklet
2015-2016	AR-OI line 3 → Line 20 (#page=15 / #page=13); electronic games of skill excluded	Line 27, "other miscellaneous deductions not subject to the 2% AGI limit" (#page=43 / #page=39); gambling not named	LongBookwTaxTables_2015.pdf, LongBookwTaxTables_2016.pdf
2017	AR-OI line 4 → Line 20 (#page=13); #page=5 says Act 155 "clarified" gambling losses	Line 27, as above (#page=39)	AR1000FandAR1000NRInstructions.pdf
2018	AR-OI line 4 → Line 22 (#page=13); electronic games of skill excluded	Line 27, as above (#page=39)	AR1000FandAR1000NRInstructions_2018.pdf
2019	AR-OI line 4 → Line 22 (#page=13)	Line 27, as above (#page=21)	2019_AR1000F_and_AR1000NR_Instructions.pdf
2020-2021	same (#page=13)	Line 27 "Gambling Losses ... cannot exceed gambling winnings" (#page=21); the form puts it under "not subject to the 2% AGI limitation"	<YYYY>_AR1000F_and_AR1000NR_Instructions.pdf
2022-2025	same (#page=14)	same (#page=22)	same pattern

Standalone forms: 2025_AR3_ItemizedDeduction.pdf (line 27 "Gambling Losses") and 2025_AR-OI_OtherIncomeLoss_and_DepreciationDifferences.pdf (line 4 "Gambling winnings").

Changes

• gov.states.ar.tax.income.gross_income.sources.{joint,individual}:
  • Add gambling_winnings to every period.
  • Add references.
  • Add a comment on the electronic games of skill exclusion.
• New ar_gambling_loss_deduction (TaxUnit) = min(Σ gambling_losses, Σ gambling_winnings) over the tax unit. On a married-filing-separately-on-the-same-return (status 4) return, AR3 is completed for the couple and the total is prorated by AGI (AR3 lines 31-35), which is what ar_itemized_deductions_indiv already does.
• gov.states.ar.tax.income.deductions.itemized.sources.{joint,individual}: add ar_gambling_loss_deduction next to, and outside, ar_misc_deduction_*, and add references.
• Changelog fragment.

Invariants

These hold for every input and are executed in policyengine_us/tests/unit/test_ar_gambling_properties.py. That file evaluates an 8 × 8 winnings × losses grid, including 0, 1, 999, 1,000, 2,200 and 250,000, in one vectorized simulation per year, for TY2015, 2017, 2018, 2019, 2020, 2022 and 2025.

1. ar_gambling_loss_deduction = min(losses, winnings).
2. 0 ≤ deduction ≤ winnings, and deduction ≤ losses.
3. Non-decreasing in both winnings and losses.
4. Independent of the year.
5. Independent of Arkansas AGI (0, 50,000, 1,000,000), i.e. no 2% floor. The itemized total equals the deduction when there are no other itemized inputs.
6. Differential: equals the federal wagering_losses_deduction, which applies the same § 165(d) cap through 2025.
7. Gambling never lowers Arkansas taxable income and raises it by at most the winnings. For a filer who itemizes either way, it raises it by exactly max(0, winnings − losses).

Original PR mutation checks, reported before this fix round:

Mutation	Property tests failing
Drop the cap	31 of 32
Apply a 2% floor	23
Leave winnings out of gross income	14

Review fix round (2026-10-07)

• Merged current main (e3a38da446) cleanly. Preserved main's itemized-deduction sources and corrected reference anchors; no conflicted files.
• Addressed the P3 at test_ar_gross_income_source_line_comments.py:128: every period must have a nonempty source list, while the existing equality check continues to require the text parser and YAML parser to agree.
• Recorded Max's d1018 ruling above and the hub's completed impact evidence below.

Tests in this round

Check	Result
Old guard at 304757c2fa, scratch copy with all four source lists emptied in both YAML files	2 passed, reproducing the vacuous pass
Strengthened guard against the same emptied scratch lists	2 failed as expected, at the new nonempty assertion
Strengthened guard against the real source files	2 passed
test_ar_gambling_properties.py	32 passed
ar_gambling_loss_deduction.yaml, including 2014	9 passed
ar_gambling_winnings_and_losses.yaml	10 passed
ar_itemized_deductions_indiv.yaml	5 passed
ar_itemized_deductions_joint.yaml	6 passed

make format passed before both commits; git diff --check passed. The scratch copies and evidence logs were not committed.

Total on the real files: 64 passed (34 Python and 30 YAML cases). The two deliberately broken scratch cases fail as expected and are excluded from this total.

Tests ran in the foreground, one process and one targeted test file at a time. Each single-file invocation is exempt from the host's heavy-lock requirement. No folder or full-suite run, partner contract test edit, or microsimulation run was performed in this round.

Impact

The hub already completed real microsimulations on the default dataset, comparing main against this PR for both 2025 and 2026. These are hub-provided results; they were not rerun in this fix round.

• gambling_winnings and gambling_losses are $0 in every record in both years.
• Zero records change in any Arkansas output.

Year	Arkansas gross income, main	Arkansas gross income, PR	Change
2025	$105.85bn	$105.85bn	$0
2026	$111.46bn	$111.46bn	$0

The gambling fix changes results only for filers who report gambling amounts. The P3 fix beyond the main merge strengthens a test and updates documentation; it changes no formulas or parameter values, so it introduces no additional model output change.

Limitations

• Arkansas electronic games of skill winnings, 2015 to June 2019. These are not income, and their losses are not deductible. PE has no separate input for them, so any it receives through gambling_winnings are taxed.
• Lottery and contest winnings. These are a separate AR-OI line (line 5). Both are taxable, so gross income is right. PE has one input and does not model any lottery-specific cap question.
• Not modeled: professional gamblers, whose losses fall under (a)(2)(A) as trade or business losses; nonresident sourcing; the Mississippi gambling tax deduction on AR3 line 6.

Interactions

• Main's merged Give Arkansas 2015-2020 income tax parameters their own values #9615, Use Arkansas's own miscellaneous and casualty loss itemized deduction rules #9625, Replace dead casetext.com references with live sources #9627 and Fix fused reference URLs repo-wide and correct Arkansas DFA page anchors #9609 changes are preserved, including the 2015 gross-income period, state casualty-loss sources, statutory references, and corrected itemized-reference anchors.
• Fix AR, MT and NM income tax gaps found in TAXSIM comparison sweep #9585 remains pending and adds partnership_s_corp_income to the same gross-income lists. Its integration must extend this PR's form-line guard with Schedule E line 18 before 2018 and line 19 thereafter.

Federal

Federal gross income leaves out gambling_winnings while wagering_losses_deduction is still deducted for itemizers, and the OBBBA 90% limit for 2026+ is not modeled. That change reaches every state that starts from federal AGI, MAGI, and PA's net-winnings rule, and is tracked separately in #9637.

axiom: TheAxiomFoundation/rulespec-us#1417 queued

Axiom encoding remains queued in that existing pe-parity issue, which documents corpus/pin and signing blockers.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (909176a) to head (c316f1d).
⚠️ Report is 281 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9636   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            4         1    -3     
  Lines           76        14   -62     
  Branches         2         0    -2     
=========================================
- Hits            76        14   -62     

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.

[MaxGhenis]

US + core hub sweep audit (2026-10-07): squash-merged #9636 as a48dc9ac45f9bd1067ef5a60615bed4033fb35b1.

Reviewed head: c316f1d831c2a9f4013dbf0b920ff1172b6a23a3.
Independent approval: GPT-6.1 Sol, /Users/maxghenis/reviews/us-hub/reviews-gen/us-9636/review-r3.md.
Latest exact-head Pull request CI: https://github.com/PolicyEngine/policyengine-us/actions/runs/37666293142 (success).
Live gates: gh pr checks exited 0; MERGEABLE; ready; no CHANGES_REQUESTED; no unresolved Max hold; no semantic area overlap since CI base.
Merge used --match-head-commit to pin the approved head.
Approval-only --admin bypass under ~/.claude/CLAUDE.md; one required approving review was the sole branch-protection block.