Montana 2021 income tax rebate variable ignores the MCA 15-30-2191 liability cap and pays in all years >= 2021 (reporting; state_income_tax unaffected)

[PavelMakarchuk]

Summary

mt_income_tax_rebate (Montana's 2021 income-tax rebate, SB 512: $1,250 single / $2,500 joint, capped at 2021 liability, paid 2023) exists and computes correct amounts, but is not wired into mt_income_tax / state_income_tax — it's a standalone variable with no effect on any tax output.

This is inconsistent with PolicyEngine's own convention for one-time state rebates, which are booked to the liability year inside state income tax: va_rebate, ma_taxpayer_refund_rebate (62F), ga_surplus_tax_rebate, me_relief_rebate, hi_act_115_rebate all reduce their state's tax in the eligible year. Montana's does not.

Evidence

• TY2021 MT joint, $60,000 wages: mt_income_tax_rebate computes $2,500, but state_income_tax = $2,518.22 — unchanged by the rebate.
• TAXSIM-35 with opt(27)=1 (its rebates-in-eligible-year mode, the convention PE uses) returns $18.22 for the same record — i.e. $2,518.22 − $2,500. The two engines agree on everything except PE's missing wiring.
• Independently discovered twice in the TAXSIM-emulator work: summing rebate variables into a comparison column produced a $1,250/record phantom for MT precisely because the variable never reaches siitax (Emit PE one-time rebates in srebate and add --net-of-rebates comparison scoring policyengine-taxsim#1070 discussion), and the opt(27) coverage matrix in TY2022 state one-time rebates: TAXSIM books them in 2022 siitax; PE uses the liability-year convention (GA/VA/NM/ME/CO/MA/ID…) policyengine-taxsim#1068 confirmed the live effect.

Suggested fix

Wire mt_income_tax_rebate into Montana's refundable credits (or however the sibling states' rebates enter), so the 2021 liability year reflects it — matching both PE's own rebate convention and TAXSIM's opt(27)=1 mode.

Integration test

- name: MT 2021 income tax rebate nets into state income tax (taxsim #1068)
  absolute_error_margin: 1
  period: 2021
  input:
    people:
      person1: {age: 45, employment_income: 60_000}
      person2: {age: 45}
    marital_units:
      marital_unit: {members: [person1, person2]}
    tax_units:
      tax_unit: {members: [person1, person2]}
    households:
      household: {members: [person1, person2], state_code: MT}
  output:
    mt_income_tax_rebate: 2_500
    # 2,518.22 tax before rebate − 2,500 rebate
    state_income_tax: 18.22

Originating: PolicyEngine/policyengine-taxsim#1068 (opt-27 coverage matrix).

[PavelMakarchuk]

Correction — the premise of this issue is wrong; rewriting it to the actual defect.

mt_income_tax_rebate is wired: it sits in the 2021 block of credits/non_refundable.yaml (since 2025-09), correctly implementing the liability-year convention (the 2022 block swaps to the property rebate; 2024 is empty). My "never netted" evidence was an artifact of testing on an older checkout plus a misread of a TY2022 phantom. On current main, TY2021 MT joint $60k → state_income_tax ≈ 0, close to TAXSIM's eligible-year mode ($18.22).

The real defect: the formula has no statutory liability cap. MCA 15-30-2191(2)(b) sets the rebate at the lesser of $1,250/$2,500 or the taxpayer's 2021 tax liability. The formula returns the flat filing-status amount unconditionally, and the credit application does not fully floor it, producing negative Montana income tax for low-liability filers:

• TY2021 MT joint, $20,000 wages: state_income_tax = −$33.81 with the rebate; +$218.59 with the rebate zeroed.
• The variable's own unit tests (period 2023, no income at all) assert the uncapped $1,250/$2,500 — they encode the missing cap.

Fix: cap the rebate at the unit's tax before non-refundable credits (mirroring the mt_files_separately composition used by mt_income_tax), and update the unit tests to supply income / expect capped values. PR incoming.

[PavelMakarchuk]

Second correction, after implementing and testing — the defect is smaller still. The −$33.81 I attributed to the rebate is actually the refundable MT EITC (3% × federal ≈ $33.81 for that couple), which legitimately produces negative net tax. Montana's credit application already floors the applied rebate at the pre-credit liability, so tax outcomes were never wrong.

What genuinely remains (confirmed by testing):

1. The variable over-reports. mt_income_tax_rebate returns the flat $1,250/$2,500 for every MT filer in every year ≥ 2021, ignoring the MCA 15-30-2191(2)(b) liability cap (e.g. that $20k couple: reports $2,500; statutory rebate is $252.40). Harmless to state_income_tax today, but wrong for any consumer of the variable — it produced phantom amounts in the TAXSIM-comparison srebate work, and it contradicts the statute.
2. The amount parameter is unscoped (pays the flat amount 2022+; only the credits-list wiring prevents application).

PR incoming with: the statutory cap in the formula (capped at the smaller of the joint/separate-column pre-credit bases — the filing election itself is post-credit and would cycle), amounts zero-scoped from 2022, and unit tests replacing the period-2023/no-income ones that encoded the uncapped values. No expected change to any state_income_tax result.