Use the Alaska and Hawaii poverty guidelines in the premium tax credit's income fraction

changelog.d/aca-fpl-coverage-year-state.fixed.md

@@ -0,0 +1 @@
+The premium tax credit's poverty line (aca_fpg) uses the state of residence in the coverage year, so Alaska and Hawaii households use their own guidelines when the state is supplied only for the coverage year. Households in Puerto Rico, Guam and the Virgin Islands, which have no poverty guideline and no Exchange, are not Marketplace enrollees, pay no Marketplace premium and are not eligible for the credit.

policyengine_us/tests/core/test_claimable_filer_ptc_invariants.py

@@ -31,8 +31,8 @@
    on a return with a head or spouse where neither can be claimed,
    `aca_ptc_slcsp` = `slcsp`,
    `aca_magi` = `medicaid_magi`, `aca_tax_family_size` = `tax_unit_size` and
-   `aca_fpg` = `tax_unit_fpg` for the prior year (these simulations supply no
-   prior-year size, so the two sizes agree). A unit with neither a head nor a
+   `aca_fpg` = the prior year's guideline for the coverage year's tax family
+   size and state group (Form 8962 lines 1 and 4). A unit with neither a head nor a
    spouse is filed by its oldest member, so a claimable minor changes it.
 3. Monotonicity, for the drawn couples in every year and the seeded
    population: marking another filer as claimable never raises the tax
@@ -50,6 +50,7 @@
 from hypothesis import strategies as st
 
 from policyengine_us import Simulation
+from policyengine_us.variables.gov.hhs.tax_unit_fpg import fpg
 from policyengine_us.tests.core.test_dependent_elsewhere_head_spouse_swap_invariance import (
     CLAIM_PATTERNS,
     TOLERANCE,
@@ -343,9 +344,14 @@ def _identities(sim, year):
     np.testing.assert_array_equal(
         size[unclaimed], _calc(sim, "tax_unit_size", year)[unclaimed]
     )
+    # Form 8962 lines 1 and 4: the prior year's guideline for the coverage
+    # year's tax family size and state of residence (26 CFR 1.36B-1(h)).
+    state_group = sim.populations["tax_unit"].household("state_group_str", str(year))
+    expected_fpg = fpg(
+        size, state_group, str(year - 1), sim.tax_benefit_system.parameters
+    )
     np.testing.assert_allclose(
-        _calc(sim, "aca_fpg", year)[unclaimed],
-        _calc(sim, "tax_unit_fpg", year - 1)[unclaimed],
+        _calc(sim, "aca_fpg", year)[unclaimed], expected_fpg[unclaimed]
     )
 
 

policyengine_us/tests/policy/baseline/gov/aca/eligibility/aca_magi_fraction.yaml

@@ -41,6 +41,194 @@
   output:
     aca_magi_fraction: 2.54
 
+- name: Coverage in 2026 in Alaska uses Alaska's 2025 guideline.
+  # 30,000 / 19,550 = 1.5345, floored to 1.53. The contiguous states'
+  # guideline (15,650) would give 1.91.
+  period: 2026
+  input:
+    people:
+      person1:
+        age: 40
+    tax_units:
+      tax_unit:
+        members: [person1]
+        aca_magi: 30_000
+    households:
+      household:
+        members: [person1]
+        state_code: AK
+  output:
+    aca_magi_fraction: 1.53
+
+- name: Coverage in 2026 in Hawaii uses Hawaii's 2025 guideline.
+  # 30,000 / 17,990 = 1.6676, floored to 1.66. The contiguous states'
+  # guideline (15,650) would give 1.91.
+  period: 2026
+  input:
+    people:
+      person1:
+        age: 40
+    tax_units:
+      tax_unit:
+        members: [person1]
+        aca_magi: 30_000
+    households:
+      household:
+        members: [person1]
+        state_code: HI
+  output:
+    aca_magi_fraction: 1.66
+
+- name: Coverage in 2026 for an Alaska family of three uses Alaska's per-person increment.
+  # 19,550 + 2 x 6,880 = 33,310; 50,000 / 33,310 = 1.5011, floored to 1.50.
+  period: 2026
+  input:
+    people:
+      parent:
+        age: 40
+      child1:
+        age: 10
+      child2:
+        age: 8
+    tax_units:
+      tax_unit:
+        members: [parent, child1, child2]
+        aca_magi: 50_000
+    households:
+      household:
+        members: [parent, child1, child2]
+        state_code: AK
+  output:
+    aca_magi_fraction: 1.5
+
+- name: A household that moves from Alaska in 2025 to Texas in 2026 uses Texas's guideline for 2026 coverage.
+  # The state of residence in the coverage year decides the guideline, not
+  # the state in the guideline year. 30,000 / 15,650 = 1.9169, floored to
+  # 1.91. Alaska's 2025 guideline (19,550) would give 1.53.
+  period: 2026
+  input:
+    people:
+      person1:
+        age: 40
+    tax_units:
+      tax_unit:
+        members: [person1]
+        aca_magi: 30_000
+    households:
+      household:
+        members: [person1]
+        state_code:
+          2025: AK
+          2026: TX
+  output:
+    aca_magi_fraction: 1.91
+
+- name: Family size is the coverage year's.
+  # aca_tax_family_size is 3 for 2026 only; 2025 counts the one listed
+  # member. Alaska's 2025 guideline for three: 19,550 + 2 x 6,880 = 33,310.
+  # 50,000 / 33,310 = 1.5011, floored to 1.50. The 2025 size of one would
+  # give 50,000 / 19,550 = 2.55.
+  period: 2026
+  input:
+    people:
+      person1:
+        age: 40
+    tax_units:
+      tax_unit:
+        members: [person1]
+        aca_magi: 50_000
+        aca_tax_family_size:
+          2026: 3
+    households:
+      household:
+        members: [person1]
+        state_code: AK
+  output:
+    aca_magi_fraction: 1.5
+
+- name: Coverage in 2025 in Alaska uses Alaska's 2024 guideline.
+  # 30,000 / 18,810 = 1.5949, floored to 1.59.
+  period: 2025
+  input:
+    people:
+      person1:
+        age: 40
+    tax_units:
+      tax_unit:
+        members: [person1]
+        aca_magi: 30_000
+    households:
+      household:
+        members: [person1]
+        state_code: AK
+  output:
+    aca_magi_fraction: 1.59
+
+- name: Households in Alaska and Texas in one simulation each use their own state's guideline.
+  # 30,000 / 19,550 = 1.53 and 30,000 / 15,650 = 1.91.
+  period: 2026
+  input:
+    people:
+      person1:
+        age: 40
+      person2:
+        age: 40
+    tax_units:
+      tax_unit1:
+        members: [person1]
+        aca_magi: 30_000
+      tax_unit2:
+        members: [person2]
+        aca_magi: 30_000
+    households:
+      household1:
+        members: [person1]
+        state_code: AK
+      household2:
+        members: [person2]
+        state_code: TX
+  output:
+    aca_magi_fraction: [1.53, 1.91]
+
+- name: Income exactly on a whole percentage of the guideline is not floored one point low.
+  # 100 x 36,621 / 15,650 = 234 exactly, so Worksheet 2 gives 234. Dividing
+  # first, in the 32-bit arithmetic aca_magi and aca_fpg are stored in, gives
+  # 36,621 / 15,650 x 100 just under 234, which floors to 2.33.
+  period: 2026
+  input:
+    people:
+      person1:
+        age: 40
+    tax_units:
+      tax_unit:
+        members: [person1]
+        aca_magi: 36_621
+    households:
+      household:
+        members: [person1]
+        state_code: TX
+  output:
+    aca_magi_fraction: 2.34
+
+- name: Negative ACA MAGI gives a fraction of 0.
+  # MAGI is floored at 0 before dividing (Form 8962 line 3 enters -0- for a
+  # negative total). From PR 10413's cases.
+  period: 2026
+  input:
+    people:
+      person1:
+        age: 40
+    tax_units:
+      tax_unit:
+        members: [person1]
+        aca_magi: -5_000
+    households:
+      household:
+        members: [person1]
+        state_code: AK
+  output:
+    aca_magi_fraction: 0
+
 - name: 2025 joint return with one claimable spouse measures income against the one-person line (the claimable spouse is the head)
   period: 2025
   input: