Pay no Colorado sales tax refund for tax year 2026 (C.R.S. 39-22-2003(2))

changelog.d/co-sales-tax-refund-2026-no-surplus.fixed.md

@@ -0,0 +1 @@
+Pay no Colorado state sales tax refund for tax year 2026: C.R.S. 39-22-2003(2) allows it only if the fiscal year ending in the tax year had excess state revenues, and FY 2025-26 fell short of the Referendum C cap.

policyengine_us/parameters/gov/states/co/tax/income/credits/sales_tax_refund/in_effect.yaml

@@ -0,0 +1,25 @@
+description: Colorado pays the state sales tax refund for a tax year if this is true, which requires excess state revenues for the state fiscal year ending in that tax year.
+values:
+  2021-01-01: true
+  # FY 2025-26 revenue fell $175.9 million short of the Referendum C cap, as
+  # the State Controller certified on 2026-09-08, so no tax year 2026 refund.
+  2026-01-01: false
+  # The LCS June 2026 forecast expects the six-tier refund to apply for tax
+  # years 2027 and 2028. Until CDOR publishes the tax year 2027 table, the
+  # tax year 2025 tier amounts stand in as placeholders; they are not
+  # published 2027 amounts (the LCS September 2026 forecast projects $43 for
+  # the lowest single tier, against the $19 carried forward).
+  2027-01-01: true
+metadata:
+  unit: bool
+  period: year
+  label: Colorado sales tax refund in effect
+  reference:
+    - title: C.R.S. 39-22-2003(2)
+      href: https://colorado.public.law/statutes/crs_39-22-2003
+    - title: Colorado Legislative Council Staff, Economic and Revenue Forecast, June 2026, Table 9 (tax year 2026 refunds $0 in every tier)
+      href: https://content.leg.colorado.gov/sites/default/files/june-2026-forecast-for-posting-accessible_0.pdf#page=31
+    - title: Colorado Legislative Council Staff, Economic and Revenue Forecast, June 2026 (six-tier refund expected for tax years 2027 and 2028)
+      href: https://content.leg.colorado.gov/sites/default/files/june-2026-forecast-for-posting-accessible_0.pdf#page=30
+    - title: Colorado Legislative Council Staff, Economic and Revenue Forecast, September 2026 (State Controller certification of FY 2025-26)
+      href: https://content.leg.colorado.gov/sites/default/files/2026-09/sept2026-forecast-with-cover-for-remediation-and-posting-accessible.pdf#page=9

policyengine_us/tests/core/test_co_sales_tax_refund_invariants.py

@@ -0,0 +1,181 @@
+"""Invariants for Colorado's state sales tax refund, C.R.S. 39-22-2003.
+
+Subsection (2) allows the refund for a tax year only "if there were excess
+state revenues for the fiscal year ending in that tax year". FY 2025-26 ended
+below the Referendum C cap (State Controller certification of 2026-09-08), so
+no tax year 2026 refund is paid. Tax year 2025 refunds come from the FY 2024-25
+surplus, at the per-person tier amounts in the Legislative Council Staff March
+2026 forecast, Table 8.
+
+Hypothesis draws batches of tax units (single, joint, or head of household
+with a dependent), each with a modified AGI, filer ages and wages; a seeded
+population of 100 units adds breadth. Each batch runs as one vectorized
+simulation covering all three years. For every tax unit:
+
+1. Gate: in 2026 the refund is zero, whatever the income, filing status or
+   eligibility.
+2. Differential: in 2025 the refund equals the number of eligible filers
+   times the published tier amount for the unit's modified AGI, with the
+   tiers taken from the forecast table rather than from the parameters.
+3. Bounds: 0 <= refund <= 2 x the top tier amount, and the refund is
+   positive exactly when a filer is eligible (2025).
+The tax year 2027 amounts are placeholders (the 2025 tier amounts stand in
+until CDOR publishes the 2027 table), so no invariant pins them.
+"""
+
+import numpy as np
+from hypothesis import HealthCheck, given, settings
+from hypothesis import strategies as st
+
+from policyengine_us import Simulation
+
+TOLERANCE = 0.01  # dollars
+
+# LCS March 2026 forecast, Table 8, "Tax Year 2025 Refunds from FY 2024-25
+# TABOR Refund Obligation", single filers: "up to $52,000" $19, "$52,001 to
+# $105,000" $25, "$105,001 to $168,000" $29, "$168,001 to $233,000" $35,
+# "$233,001 to $299,000" $37, "$299,001 and up" $59.
+PUBLISHED_2025_STARTS = np.array([52_001, 105_001, 168_001, 233_001, 299_001])
+PUBLISHED_2025_AMOUNTS = np.array([19, 25, 29, 35, 37, 59])
+
+agis = st.one_of(
+    st.integers(-50_000, 0).map(float),
+    st.integers(0, 400_000).map(float),
+    st.sampled_from([52_000.0, 52_001.0, 105_000.0, 299_000.0, 299_001.0]),
+)
+ages = st.one_of(st.integers(14, 17), st.integers(18, 90))
+wages = st.one_of(st.just(0.0), st.integers(1, 100_000).map(float))
+
+
+@st.composite
+def tax_units(draw):
+    kind = draw(st.sampled_from(["single", "joint", "hoh"]))
+    return {
+        "kind": kind,
+        "agi": draw(agis),
+        "head": {"age": draw(ages), "employment_income": draw(wages)},
+        "spouse": (
+            {"age": draw(ages), "employment_income": draw(wages)}
+            if kind == "joint"
+            else None
+        ),
+    }
+
+
+SEED = 20261006
+
+
+def _seeded_units(n=100):
+    rng = np.random.default_rng(SEED)
+
+    def person():
+        return {
+            "age": int(rng.integers(14, 91)),
+            "employment_income": float(rng.integers(0, 100_001))
+            if rng.random() < 0.5
+            else 0.0,
+        }
+
+    units = []
+    for _ in range(n):
+        kind = rng.choice(["single", "joint", "hoh"])
+        units.append(
+            {
+                "kind": kind,
+                "agi": float(rng.integers(-50_000, 400_001)),
+                "head": person(),
+                "spouse": person() if kind == "joint" else None,
+            }
+        )
+    return units
+
+
+YEARS = (2025, 2026, 2027)
+
+
+def _situation(units):
+    # One simulation holds every year: each input is keyed by all of YEARS.
+    def every_year(value):
+        return {year: value for year in YEARS}
+
+    people, groups = {}, {"tax_units": {}, "households": {}}
+    for i, unit in enumerate(units):
+        head = f"head_{i}"
+        people[head] = {**unit["head"], "is_tax_unit_head": True}
+        members = [head]
+        if unit["spouse"] is not None:
+            spouse = f"spouse_{i}"
+            people[spouse] = {**unit["spouse"], "is_tax_unit_spouse": True}
+            members.append(spouse)
+        if unit["kind"] == "hoh":
+            child = f"child_{i}"
+            people[child] = {"age": 8, "is_tax_unit_dependent": True}
+            members.append(child)
+        groups["tax_units"][f"tax_unit_{i}"] = {
+            "members": members,
+            "co_modified_agi": every_year(unit["agi"]),
+            "co_income_tax_before_non_refundable_credits": every_year(0),
+        }
+        groups["households"][f"household_{i}"] = {
+            "members": members,
+            "state_code": every_year("CO"),
+        }
+    people = {
+        name: {key: every_year(value) for key, value in values.items()}
+        for name, values in people.items()
+    }
+    return {"people": people, **groups}
+
+
+def _run(units):
+    sim = Simulation(situation=_situation(units))
+    unit = sim.populations["tax_unit"].members_entity_id
+    by_year = {}
+    for year in YEARS:
+        eligible = np.asarray(
+            sim.calculate("co_sales_tax_refund_person_eligible", year), dtype=float
+        )
+        by_year[year] = {
+            "refund": np.asarray(
+                sim.calculate("co_sales_tax_refund", year), dtype=float
+            ),
+            "agi": np.asarray(sim.calculate("co_modified_agi", year), dtype=float),
+            "count": np.bincount(unit, weights=eligible, minlength=len(units)),
+        }
+    return by_year
+
+
+def _check(units):
+    by_year = _run(units)
+    published = PUBLISHED_2025_AMOUNTS[
+        np.searchsorted(PUBLISHED_2025_STARTS, by_year[2025]["agi"], side="right")
+    ]
+
+    # 1. Gate: no tax year 2026 refund.
+    np.testing.assert_array_equal(by_year[2026]["refund"], 0)
+
+    # 2. Differential against the published tax year 2025 table.
+    base = by_year[2025]
+    np.testing.assert_allclose(
+        base["refund"], base["count"] * published, atol=TOLERANCE
+    )
+
+    # 3. Bounds.
+    for run in by_year.values():
+        assert (run["refund"] >= 0).all()
+        assert (run["refund"] <= 2 * PUBLISHED_2025_AMOUNTS.max()).all()
+    np.testing.assert_array_equal(base["refund"] > 0, base["count"] > 0)
+
+
+@settings(
+    max_examples=4,
+    deadline=None,
+    suppress_health_check=[HealthCheck.too_slow, HealthCheck.data_too_large],
+)
+@given(st.lists(tax_units(), min_size=1, max_size=12))
+def test_co_sales_tax_refund_invariants(units):
+    _check(units)
+
+
+def test_co_sales_tax_refund_invariants_seeded_population():
+    _check(_seeded_units())