NICs-only salary sacrifice cap changes ordinary income tax after age 75 #2454

"""MOCK-only diagnostic for NICs reclassification leaking into ordinary tax.

Run from the worktree with its locked venv and four-thread environment::

    uv run --frozen --no-sync python .job/reproduce_nics_income_tax.py

An optional --source-root points at an extracted source tree, allowing the
same situations and installed dependencies to diagnose the reviewed head.
No survey data, uprating or microsimulation is involved.
"""

import argparse
import json
import sys
from pathlib import Path


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("--source-root", default=str(Path.cwd()))
    args = parser.parse_args()
    sys.path.insert(0, str(Path(args.source_root).resolve()))

    from policyengine_uk import Simulation
    import policyengine_uk.simulation as uk_simulation

    # Clone a loaded system, as the review's MOCK diagnostics do. Each cap
    # comparison still receives a separate simulation and parameter tree.
    base = uk_simulation.CountryTaxBenefitSystem()
    uk_simulation.CountryTaxBenefitSystem = base.clone

    cases = {
        "post75_review_witness": {
            "age": 80,
            "employment_income_before_lsr": 160_000,
            "pension_contributions_via_salary_sacrifice": 30_000,
            "employer_pension_contributions": 100_000,
        },
        "post75_personal_allowance_witness": {
            "age": 80,
            "employment_income_before_lsr": 100_000,
            "pension_contributions_via_salary_sacrifice": 30_000,
            "employer_pension_contributions": 100_000,
        },
        "basic_ras_boundary": {
            "age": 45,
            "employment_income_before_lsr": 2_000,
            "pension_contributions_via_salary_sacrifice": 201_000,
            "employer_pension_contributions": 100_000,
            "personal_pension_contributions": 3_600,
        },
    }
    variables = [
        "employment_income",
        "salary_sacrifice_returned_to_income",
        "employee_pension_contributions",
        "pension_contributions_relief",
        "net_income",
        "adjusted_net_income",
        "personal_allowance",
        "pension_contributions_for_annual_allowance",
        "pension_annual_allowance_threshold_income",
        "pension_annual_allowance_adjusted_income",
        "pension_annual_allowance",
        "personal_pension_contributions_tax",
        "income_tax",
    ]
    rows = []
    for year in [2029, 2030]:
        situation = {
            "people": {
                name: {key: {year: value} for key, value in inputs.items()}
                for name, inputs in cases.items()
            },
            "benunits": {f"b{i}": {"members": [name]} for i, name in enumerate(cases)},
            "households": {
                f"h{i}": {"members": [name], "region": {year: "LONDON"}}
                for i, name in enumerate(cases)
            },
        }
        results = {}
        for label, cap in [("cap2000", 2_000), ("unlimited", float("inf"))]:
            simulation = Simulation(situation=situation)
            simulation.apply_parameter_changes(
                {
                    "gov.hmrc.national_insurance.salary_sacrifice_pension_cap": {
                        str(year): cap
                    },
                    "gov.contrib.behavioral_responses.salary_sacrifice_broad_base_haircut_rate": {
                        str(year): 0
                    },
                }
            )
            results[label] = {
                variable: simulation.calculate(variable, year) for variable in variables
            }
        for i, (name, inputs) in enumerate(cases.items()):
            row = {"year": year, "case": name, "inputs": inputs}
            for label in results:
                row[label] = {
                    variable: float(values[i])
                    for variable, values in results[label].items()
                }
                # The only non-ordinary tax item in these situations is the
                # AA charge. Remove it explicitly to isolate this issue.
                row[label]["ordinary_income_tax"] = (
                    row[label]["income_tax"]
                    - row[label]["personal_pension_contributions_tax"]
                )
            row["cap_minus_unlimited"] = {
                variable: row["cap2000"][variable] - row["unlimited"][variable]
                for variable in [*variables, "ordinary_income_tax"]
            }
            rows.append(row)
    print(json.dumps({"dataset": "MOCK", "rows": rows}, indent=2), flush=True)
    # These assert the diagnostic is reproducing the wider defect. They are
    # deliberately outside the regression suite: the AA fix must not fix it.
    for row in rows:
        if row["case"].startswith("post75_"):
            assert row["cap_minus_unlimited"]["ordinary_income_tax"] > 0, row


if __name__ == "__main__":
    main()