Leave dependents' capital income out of state lines on the filers' return

changelog.d/dependent-capital-income-state-lines.fixed.md

@@ -0,0 +1 @@
+Leave tax unit dependents' interest, dividends and capital gains out of the state lines that adjust the filers' federal adjusted gross income or taxable income (AZ, DC, DE, IA, IL, MA, MD, ME, MI, MN, MT, ND, NH, OH, PA, SC, UT, VA, VT, WA, WI and WV), out of California's young child tax credit loss test and Georgia's and Delaware's retirement exclusions, and out of the U.S. government interest that most states subtract. Also add the half of benefits to Iowa's reportable Social Security worksheet, net Vermont's eligible gains before its percentage exclusion, floor North Dakota's long-term gain exclusion at zero, and cap Virginia's National Guard subtraction per person.

policyengine_us/parameters/gov/states/de/tax/income/subtractions/exclusions/pension/income_sources.yaml

@@ -2,7 +2,9 @@ description: Delaware counts these income sources for the pension exclusion.
 values:
   2021-01-01:
     - dividend_income
-    - loss_limited_net_capital_gains_person
+    # The filer's own share; a tax unit dependent's losses are on the
+    # dependent's own return.
+    - filer_loss_limited_net_capital_gains_person
     - taxable_interest_income
     - rental_income
     - taxable_pension_income

policyengine_us/tests/core/test_dependent_capital_income_state_tax_invariants.py

@@ -0,0 +1,304 @@
+"""Property tests: a dependent's capital income stays off the filers' North
+Dakota and Vermont returns.
+
+Federal adjusted gross income leaves out a tax unit dependent's income
+(irs_gross_income), which the dependent reports on their own return (2025
+Form 8814, line 4; the model does not implement the Form 8814 election). Both
+states start from the filers' federal return:
+
+- N.D.C.C. 57-38-30.3(2) defines North Dakota taxable income as "the federal
+  taxable income of an individual", reduced by 40% of "the taxpayer's net
+  long-term capital gain" ((d)(1); Form ND-1 line 6, from the filer's
+  Schedule D lines 15 and 16, "If zero or less ... no exclusion is allowed")
+  and of qualified dividends ((d)(2); Form ND-1 line 13, from Form 1040 line
+  3a), and by interest on U.S. obligations ((a)).
+- 32 V.S.A. 5811(21) defines Vermont taxable income as the individual's
+  federal adjusted gross income, increased by interest on non-Vermont state
+  and local obligations ((A)(i)) and decreased, "to the extent such income is
+  included in federal adjusted gross income", by interest on U.S. government
+  obligations ((B)(i)) and by 40% of gains on "assets held by the taxpayer for
+  more than three years" ((B)(ii)).
+
+So for every tax unit in either state, and any capital income its dependents
+have (qualified and other dividends, long- and short-term gains and losses,
+capital gain distributions, taxable, tax-exempt and U.S. government interest,
+and Vermont-eligible long-term gains):
+
+1. nd_income_tax and vt_income_tax are the same as when the dependents have
+   none. So are federal adjusted gross income and taxable income, the base
+   both states start from.
+2. nd_qdiv_subtraction is 40% of the filers' qualified dividends, and
+   nd_ltcg_subtraction is 40% of max(0, min(line 15, line 16)) of the filers'
+   own Schedule D, with capital gain distributions on line 15.
+3. vt_percentage_capital_gains_exclusion is 40% of the filers' eligible gains
+   netted against their own Schedule D (no more than max(0, min(line 15,
+   line 16)), Schedule IN-153 line 15), up to the cap. vt_additions is the
+   filers' tax-exempt interest, and us_govt_interest is the filers' U.S.
+   government interest.
+
+The filers earn at least $80,000, so no earned income tax credit is at
+stake: Vermont's is a share of the federal credit, whose investment income
+test (eitc_relevant_investment_income) is not changed here.
+
+Amounts are whole dollars of at most $400,000, so every sum is exact in single
+precision and twins are compared to the cent.
+"""
+
+import numpy as np
+from hypothesis import HealthCheck, given, settings
+from hypothesis import strategies as st
+
+from policyengine_us import Simulation
+
+YEAR = 2025
+TOLERANCE = 0.01
+
+# Each person's capital income inputs.
+CAPITAL_INPUTS = [
+    "qualified_dividend_income",
+    "non_qualified_dividend_income",
+    "long_term_capital_gains",
+    "short_term_capital_gains",
+    "non_sch_d_capital_gains",
+    "taxable_interest_income",
+    "us_govt_interest_person",
+    "tax_exempt_interest_income",
+    "long_term_capital_gains_on_assets_eligible_for_vt_exclusion",
+]
+SAME_AS_WITHOUT_DEPENDENT_INCOME = [
+    "adjusted_gross_income",
+    "taxable_income",
+    "nd_income_tax",
+    "vt_income_tax",
+]
+UNIT_OUTPUTS = SAME_AS_WITHOUT_DEPENDENT_INCOME + [
+    "nd_qdiv_subtraction",
+    "nd_ltcg_subtraction",
+    "vt_percentage_capital_gains_exclusion",
+    "vt_additions",
+    "us_govt_interest",
+]
+
+
+def amount(low, high):
+    return st.integers(low, high).map(float)
+
+
+@st.composite
+def capital_income(draw):
+    """One person's capital income. U.S. government interest is part of
+    taxable interest, and Vermont-eligible gains are part of long-term
+    gains."""
+    long_term = draw(amount(-30_000, 300_000))
+    taxable_interest = draw(amount(0, 50_000))
+    return {
+        "qualified_dividend_income": draw(amount(0, 50_000)),
+        "non_qualified_dividend_income": draw(amount(0, 20_000)),
+        "long_term_capital_gains": long_term,
+        "short_term_capital_gains": draw(amount(-20_000, 50_000)),
+        "non_sch_d_capital_gains": draw(amount(0, 20_000)),
+        "taxable_interest_income": taxable_interest,
+        "us_govt_interest_person": draw(amount(0, int(taxable_interest))),
+        "tax_exempt_interest_income": draw(amount(0, 50_000)),
+        "long_term_capital_gains_on_assets_eligible_for_vt_exclusion": (
+            draw(amount(0, int(long_term))) if long_term > 0 else 0.0
+        ),
+    }
+
+
+@st.composite
+def tax_units(draw):
+    state = draw(st.sampled_from(["ND", "VT"]))
+    n_filers = draw(st.integers(1, 2))
+    filers = [
+        {
+            "age": draw(st.integers(30, 64)),
+            "employment_income": draw(amount(80_000, 400_000)) if i == 0 else 0.0,
+            **draw(capital_income()),
+        }
+        for i in range(n_filers)
+    ]
+    dependents = [
+        {
+            "age": draw(st.integers(1, 17)),
+            **draw(capital_income()),
+        }
+        for _ in range(draw(st.integers(1, 2)))
+    ]
+    return {"state": state, "filers": filers, "dependents": dependents}
+
+
+def without_dependent_income(unit):
+    zero = {name: 0.0 for name in CAPITAL_INPUTS}
+    return {**unit, "dependents": [{**d, **zero} for d in unit["dependents"]]}
+
+
+def build_situation(units):
+    """One tax unit per household: the head, an optional spouse and the
+    dependents, each dependent in a marital unit of their own."""
+    people, tax_units_out, marital_units, households = {}, {}, {}, {}
+    for i, unit in enumerate(units):
+        members = []
+        for j, person in enumerate(unit["filers"] + unit["dependents"]):
+            name = f"person_{i}_{j}"
+            filer = j < len(unit["filers"])
+            people[name] = {
+                "is_tax_unit_head": {YEAR: j == 0},
+                "is_tax_unit_spouse": {YEAR: j == 1 and filer},
+                "is_tax_unit_dependent": {YEAR: not filer},
+                **{key: {YEAR: value} for key, value in person.items()},
+            }
+            members.append(name)
+            if not filer:
+                marital_units[f"dependent_{i}_{j}"] = {"members": [name]}
+        tax_units_out[f"tax_unit_{i}"] = {"members": members}
+        marital_units[f"filers_{i}"] = {"members": members[: len(unit["filers"])]}
+        households[f"household_{i}"] = {
+            "members": members,
+            "state_code": {YEAR: unit["state"]},
+        }
+    return {
+        "people": people,
+        "tax_units": tax_units_out,
+        "marital_units": marital_units,
+        "households": households,
+    }
+
+
+def calculate(units):
+    simulation = Simulation(situation=build_situation(units))
+    results = {
+        variable: np.asarray(simulation.calculate(variable, YEAR))
+        for variable in UNIT_OUTPUTS
+    }
+    p = simulation.tax_benefit_system.parameters(YEAR).gov.states
+    return results, p
+
+
+def filer_total(unit, name):
+    return sum(person[name] for person in unit["filers"])
+
+
+def check(units):
+    twins = [without_dependent_income(unit) for unit in units]
+    results, p = calculate(units + twins)
+    n = len(units)
+
+    # 1. The same taxes and federal base as without the dependents' income.
+    for variable in SAME_AS_WITHOUT_DEPENDENT_INCOME:
+        np.testing.assert_allclose(
+            results[variable][:n],
+            results[variable][n:],
+            atol=TOLERANCE,
+            err_msg=variable,
+        )
+
+    nd = p.nd.tax.income.taxable_income.subtractions
+    vt = p.vt.tax.income.agi.exclusions.capital_gain.percentage
+    is_nd = np.array([unit["state"] == "ND" for unit in units])
+    is_vt = ~is_nd
+    long_term = np.array(
+        [
+            filer_total(u, "long_term_capital_gains")
+            + filer_total(u, "non_sch_d_capital_gains")
+            for u in units
+        ]
+    )
+    short_term = np.array([filer_total(u, "short_term_capital_gains") for u in units])
+    reference = {
+        # 2. North Dakota, from the filers' own amounts.
+        "nd_qdiv_subtraction": nd.qdiv_fraction
+        * np.array([filer_total(u, "qualified_dividend_income") for u in units]),
+        "nd_ltcg_subtraction": nd.ltcg_fraction
+        * np.maximum(0, np.minimum(long_term, long_term + short_term)),
+        # 3. Vermont, from the filers' own amounts.
+        # Schedule IN-153 line 15: eligible gains after netting, no more than
+        # the filers' net gain (the draws have no collectibles or section 1250
+        # gain, so Part I line 4 is max(0, min(line 15, line 16))).
+        "vt_percentage_capital_gains_exclusion": np.minimum(
+            vt.rate
+            * np.minimum(
+                [
+                    filer_total(
+                        u, "long_term_capital_gains_on_assets_eligible_for_vt_exclusion"
+                    )
+                    for u in units
+                ],
+                np.maximum(0, np.minimum(long_term, long_term + short_term)),
+            ),
+            vt.cap,
+        ),
+        "vt_additions": np.array(
+            [filer_total(u, "tax_exempt_interest_income") for u in units]
+        ),
+    }
+    for variable, expected in reference.items():
+        in_state = is_nd if variable.startswith("nd_") else is_vt
+        np.testing.assert_allclose(
+            results[variable][:n][in_state],
+            expected[in_state],
+            atol=TOLERANCE,
+            err_msg=variable,
+        )
+    np.testing.assert_allclose(
+        results["us_govt_interest"][:n],
+        [filer_total(u, "us_govt_interest_person") for u in units],
+        atol=TOLERANCE,
+    )
+
+
+# A simulation's cost is mostly fixed, so each example is a batch of tax
+# units and there are few examples.
+@settings(
+    max_examples=5,
+    deadline=None,
+    derandomize=True,
+    suppress_health_check=[HealthCheck.too_slow, HealthCheck.data_too_large],
+)
+@given(st.lists(tax_units(), min_size=10, max_size=30))
+def test_dependent_capital_income_stays_off_nd_and_vt_returns(units):
+    check(units)
+
+
+def test_reported_examples():
+    """The two households that showed the leak (TY2025): in North Dakota a
+    head with $50,000 of wages and $1,000 of qualified dividends, and in
+    Vermont a head with $150,000 of wages and a $10,000 long-term gain, each
+    with a dependent child who has dividends or a Vermont-eligible gain."""
+    zero = {name: 0.0 for name in CAPITAL_INPUTS}
+    nd = {
+        "state": "ND",
+        "filers": [
+            {
+                "age": 40,
+                "employment_income": 50_000.0,
+                **zero,
+                "qualified_dividend_income": 1_000.0,
+            }
+        ],
+        "dependents": [{"age": 10, **zero, "qualified_dividend_income": 5_000.0}],
+    }
+    vt = {
+        "state": "VT",
+        "filers": [
+            {
+                "age": 40,
+                "employment_income": 150_000.0,
+                **zero,
+                "long_term_capital_gains": 10_000.0,
+            }
+        ],
+        "dependents": [
+            {
+                "age": 10,
+                **zero,
+                "long_term_capital_gains": 100_000.0,
+                "long_term_capital_gains_on_assets_eligible_for_vt_exclusion": 100_000.0,
+            }
+        ],
+    }
+    check([nd, vt])
+    results, _ = calculate([nd, vt])
+    assert results["nd_qdiv_subtraction"][0] == 400
+    assert results["vt_percentage_capital_gains_exclusion"][1] == 0
+    np.testing.assert_allclose(results["vt_income_tax"][1], 6_623.20, atol=TOLERANCE)

policyengine_us/tests/policy/baseline/gov/irs/income/taxable_income/adjusted_gross_income/irs_gross_income/loss_limited_net_capital_gains_person.yaml

@@ -70,3 +70,37 @@
     # Total raw: 12000 (positive, so no limit applied)
     # Return person's raw amounts unchanged
     loss_limited_net_capital_gains_person: [-3000, 15000]
+
+- name: The filer-only allocation leaves a dependent's loss out
+  period: 2025
+  input:
+    people:
+      head:
+        is_tax_unit_head: true
+        age: 45
+        long_term_capital_gains: 5_000
+      child:
+        age: 15
+        short_term_capital_gains: -10_000
+  output:
+    # The pooled allocation shares the 3,000 limited loss; the filer-only one
+    # keeps the head's 5,000 gain.
+    filer_loss_limited_net_capital_gains_person: [5_000, 0]
+
+- name: The filer-only allocation limits a filers' net loss
+  period: 2025
+  input:
+    people:
+      head:
+        is_tax_unit_head: true
+        age: 45
+        long_term_capital_gains: -8_000
+      spouse:
+        is_tax_unit_spouse: true
+        age: 45
+        long_term_capital_gains: -2_000
+      child:
+        age: 15
+        long_term_capital_gains: 50_000
+  output:
+    filer_loss_limited_net_capital_gains_person: [-2_400, -600, 0]

policyengine_us/tests/policy/baseline/gov/states/al/tax/income/al_us_govt_bond_interest.yaml

@@ -6,3 +6,22 @@
     us_govt_interest_person: 5_000
   output:
     al_agi: 5_000
+
+- name: A dependent's interest, U.S. government interest included, stays off the filer's AGI
+  # Neither the child's interest nor its exemption is on the filer's return;
+  # us_govt_interest counts only the head's and spouse's.
+  period: 2025
+  input:
+    people:
+      head:
+        is_tax_unit_head: true
+        age: 40
+        employment_income: 100_000
+      child:
+        age: 10
+        taxable_interest_income: 5_000
+        us_govt_interest_person: 5_000
+    household:
+      state_code: AL
+  output:
+    al_agi: 100_000