Resolution: Multiple tax units explain the itemization behavior

itemization_issue_resolved.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Resolution: Household 4428 Has Multiple Tax Units\n",
    "\n",
    "The apparent bug is explained by household 4428 having 2 tax units with different itemization decisions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from policyengine_us import Microsimulation\n",
    "from policyengine_us.model_api import *\n",
    "import numpy as np\n",
    "\n",
    "DATASET = \"hf://policyengine/policyengine-us-data/enhanced_cps_2024.h5\"\n",
    "YEAR = 2026"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Setup Reform"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reform = Reform.from_dict({\n",
    "    \"gov.irs.income.bracket.rates.2\": {\"2026-01-01\": 0.15},\n",
    "    \"gov.irs.income.bracket.rates.3\": {\"2026-01-01\": 0.25},\n",
    "    \"gov.irs.deductions.standard.amount.JOINT\": {\"2026-01-01\": 48900},\n",
    "    \"gov.irs.income.exemption.amount\": {\"2026-01-01\": 0},\n",
    "    \"gov.contrib.reconciliation.ctc.in_effect\": {\"2026-01-01\": True},\n",
    "    \"gov.irs.credits.ctc.amount.base[0].amount\": {\"2026-01-01\": 2200},\n",
    "}, country_id=\"us\")\n",
    "\n",
    "sim = Microsimulation(reform=reform, dataset=DATASET)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Check Both Tax Units"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get tax unit data\n",
    "tax_unit_ids = sim.calculate(\"tax_unit_id\", period=YEAR).values\n",
    "itemizes = sim.calculate(\"tax_unit_itemizes\", period=YEAR).values\n",
    "tax_if_item = sim.calculate(\"tax_liability_if_itemizing\", period=YEAR).values\n",
    "tax_if_std = sim.calculate(\"tax_liability_if_not_itemizing\", period=YEAR).values\n",
    "\n",
    "# Find the two tax units in household 4428\n",
    "tu1_idx = np.where(tax_unit_ids == 442801)[0][0]\n",
    "tu2_idx = np.where(tax_unit_ids == 442802)[0][0]\n",
    "\n",
    "print(\"Tax Unit 442801 (parents + 4 kids):\")\n",
    "print(f\"  Itemizes: {bool(itemizes[tu1_idx])}\")\n",
    "print(f\"  Tax if itemizing: ${tax_if_item[tu1_idx]:,.2f}\")\n",
    "print(f\"  Tax if standard: ${tax_if_std[tu1_idx]:,.2f}\")\n",
    "print(f\"  Optimal: {'Same either way' if tax_if_item[tu1_idx] == tax_if_std[tu1_idx] else 'Itemize' if tax_if_item[tu1_idx] < tax_if_std[tu1_idx] else 'Standard'}\")\n",
    "\n",
    "print(\"\\nTax Unit 442802 (23-year-old):\")\n",
    "print(f\"  Itemizes: {bool(itemizes[tu2_idx])}\")\n",
    "print(f\"  Tax if itemizing: ${tax_if_item[tu2_idx]:,.2f}\")\n",
    "print(f\"  Tax if standard: ${tax_if_std[tu2_idx]:,.2f}\")\n",
    "print(f\"  Optimal: {'Standard' if tax_if_std[tu2_idx] < tax_if_item[tu2_idx] else 'Itemize'}\")\n",
    "\n",
    "print(\"\\n✓ Both tax units make optimal decisions!\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## The Real Issue\n",
    "\n",
    "The confusion arose from looking at household-level aggregates. When mapped to household level:\n",
    "- `tax_unit_itemizes` shows True (from the first tax unit)\n",
    "- But the tax liability values are summed across both tax units\n",
    "\n",
    "This creates the appearance of irrational behavior when it's actually two rational decisions."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}