Add comparative dataset analysis to paper

Makefile

@@ -1,4 +1,4 @@
-.PHONY: all format test install download upload docker documentation data clean build paper clean-paper
+.PHONY: all format test install download upload docker documentation data clean build paper clean-paper presentations
 
 all: data test
 
@@ -50,14 +50,21 @@ build:
 publish:
 	twine upload dist/*
 
-paper: paper/main.pdf
+paper: paper/woodruff_ghenis_2024_enhanced_cps.pdf
 
-paper/main.pdf: $(wildcard paper/sections/**/*.tex) $(wildcard paper/bibliography/*.bib) paper/main.tex paper/macros.tex
+paper/woodruff_ghenis_2024_enhanced_cps.pdf: $(wildcard paper/sections/**/*.tex) $(wildcard paper/bibliography/*.bib) paper/main.tex paper/macros.tex
 	cd paper && \
 	BIBINPUTS=./bibliography pdflatex main && \
 	BIBINPUTS=./bibliography bibtex main && \
-	pdflatex main && \
-	pdflatex main
+	pdflatex -jobname=woodruff_ghenis_2024_enhanced_cps main && \
+	pdflatex -jobname=woodruff_ghenis_2024_enhanced_cps main
 
 clean-paper:
-	rm -f paper/*.aux paper/*.bbl paper/*.blg paper/*.log paper/*.out paper/*.toc paper/main.pdf paper/sections/**/*.aux
+	rm -f paper/*.aux paper/*.bbl paper/*.blg paper/*.log paper/*.out paper/*.toc paper/*.pdf paper/sections/**/*.aux
+
+presentations: presentations/nta_2024_11/nta_2024_slides.pdf
+
+presentations/nta_2024_11/nta_2024_slides.pdf: presentations/nta_2024_11/main.tex
+	cd presentations/nta_2024_11 && \
+		pdflatex -jobname=nta_2024_slides main && \
+		pdflatex -jobname=nta_2024_slides main

changelog_entry.yaml

@@ -0,0 +1,3 @@
+- bump: minor
+  added:
+    - Paper sections on inequality results, weights, demographic analysis, and PolicyEngine integration.

docs/results.ipynb

@@ -24,9 +24,25 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 1,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "ename": "ValueError",
+     "evalue": "Invalid response code 404 for url https://api.github.com/repos/policyengine/irs-soi-puf/releases/tags/1.8.0.",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
+      "Cell \u001b[0;32mIn[1], line 7\u001b[0m\n\u001b[1;32m      5\u001b[0m cps \u001b[38;5;241m=\u001b[39m Microsimulation()\n\u001b[1;32m      6\u001b[0m ecps \u001b[38;5;241m=\u001b[39m Microsimulation(dataset\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124menhanced_cps_2024\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[0;32m----> 7\u001b[0m puf \u001b[38;5;241m=\u001b[39m Microsimulation(dataset\u001b[38;5;241m=\u001b[39mPUF_2024)\n",
+      "File \u001b[0;32m~/miniconda3/envs/policyengine/lib/python3.11/site-packages/policyengine_us/system.py:150\u001b[0m, in \u001b[0;36mMicrosimulation.__init__\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m    149\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m__init__\u001b[39m(\u001b[38;5;28mself\u001b[39m, \u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs):\n\u001b[0;32m--> 150\u001b[0m     \u001b[38;5;28msuper\u001b[39m()\u001b[38;5;241m.\u001b[39m\u001b[38;5;21m__init__\u001b[39m(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)\n\u001b[1;32m    152\u001b[0m     reform \u001b[38;5;241m=\u001b[39m create_structural_reforms_from_parameters(\n\u001b[1;32m    153\u001b[0m         \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mtax_benefit_system\u001b[38;5;241m.\u001b[39mparameters, year_start\n\u001b[1;32m    154\u001b[0m     )\n\u001b[1;32m    155\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m reform \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n",
+      "File \u001b[0;32m~/miniconda3/envs/policyengine/lib/python3.11/site-packages/policyengine_core/simulations/simulation.py:164\u001b[0m, in \u001b[0;36mSimulation.__init__\u001b[0;34m(self, tax_benefit_system, populations, situation, dataset, reform, trace)\u001b[0m\n\u001b[1;32m    160\u001b[0m         dataset \u001b[38;5;241m=\u001b[39m Dataset\u001b[38;5;241m.\u001b[39mfrom_file(\n\u001b[1;32m    161\u001b[0m             dataset, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdefault_input_period\n\u001b[1;32m    162\u001b[0m         )\n\u001b[1;32m    163\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(dataset, \u001b[38;5;28mtype\u001b[39m):\n\u001b[0;32m--> 164\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdataset: Dataset \u001b[38;5;241m=\u001b[39m dataset(require\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mTrue\u001b[39;00m)\n\u001b[1;32m    165\u001b[0m \u001b[38;5;28;01melif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(dataset, pd\u001b[38;5;241m.\u001b[39mDataFrame):\n\u001b[1;32m    166\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdataset \u001b[38;5;241m=\u001b[39m Dataset\u001b[38;5;241m.\u001b[39mfrom_dataframe(\n\u001b[1;32m    167\u001b[0m         dataset, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdefault_input_period\n\u001b[1;32m    168\u001b[0m     )\n",
+      "File \u001b[0;32m~/miniconda3/envs/policyengine/lib/python3.11/site-packages/policyengine_core/data/dataset.py:66\u001b[0m, in \u001b[0;36mDataset.__init__\u001b[0;34m(self, require)\u001b[0m\n\u001b[1;32m     64\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mexists \u001b[38;5;129;01mand\u001b[39;00m require:\n\u001b[1;32m     65\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39murl \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[0;32m---> 66\u001b[0m         \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdownload()\n\u001b[1;32m     67\u001b[0m     \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[1;32m     68\u001b[0m         \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mgenerate()\n",
+      "File \u001b[0;32m~/miniconda3/envs/policyengine/lib/python3.11/site-packages/policyengine_core/data/dataset.py:308\u001b[0m, in \u001b[0;36mDataset.download\u001b[0;34m(self, url)\u001b[0m\n\u001b[1;32m    306\u001b[0m response \u001b[38;5;241m=\u001b[39m requests\u001b[38;5;241m.\u001b[39mget(url, headers\u001b[38;5;241m=\u001b[39mauth_headers)\n\u001b[1;32m    307\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m response\u001b[38;5;241m.\u001b[39mstatus_code \u001b[38;5;241m!=\u001b[39m \u001b[38;5;241m200\u001b[39m:\n\u001b[0;32m--> 308\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\n\u001b[1;32m    309\u001b[0m         \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mInvalid response code \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mresponse\u001b[38;5;241m.\u001b[39mstatus_code\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m for url \u001b[39m\u001b[38;5;132;01m{\u001b[39;00murl\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    310\u001b[0m     )\n\u001b[1;32m    311\u001b[0m assets \u001b[38;5;241m=\u001b[39m response\u001b[38;5;241m.\u001b[39mjson()[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124massets\u001b[39m\u001b[38;5;124m\"\u001b[39m]\n\u001b[1;32m    312\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m asset \u001b[38;5;129;01min\u001b[39;00m assets:\n",
+      "\u001b[0;31mValueError\u001b[0m: Invalid response code 404 for url https://api.github.com/repos/policyengine/irs-soi-puf/releases/tags/1.8.0."
+     ]
+    }
+   ],
    "source": [
     "from policyengine_us import Microsimulation\n",
     "from policyengine_us_data import PUF_2024\n",
@@ -7270,8 +7286,16 @@
     "        title=\"Weight\",\n",
     "        type=\"log\",\n",
     "    ),\n",
-    ")"
+    ")\n",
+    "format_fig(fig).show()"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {
@@ -7290,7 +7314,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.14"
+   "version": "3.11.9"
   }
  },
  "nbformat": 4,

paper/.gitignore

@@ -11,6 +11,9 @@
 *.cb
 *.cb2
 .*.lb
+*.nav
+*.snm
+*.vrb
 
 ## Generated if empty string is given at "Please type another file name for output:"
 .pdf

paper/bibliography/references.bib

@@ -184,3 +184,28 @@ @techreport{woodruff2023survey
   note        = {Demonstrates superiority of machine learning approaches over traditional methods for survey enhancement through comprehensive benchmarking},
   url         = {https://github.com/policyengine/survey-enhance/blob/main/docs/paper/project_paper.pdf}
 }
+
+@techreport{khitatrakun2023race,
+  title       = {A New Approach for Estimating the Impact of Tax Policies by Race and Ethnicity},
+  author      = {Khitatrakun, Surachai and Mermin, Gordon and Page, Benjamin and Rohaly, Jeffrey},
+  institution = {Tax Policy Center},
+  year        = {2023},
+  month       = {2},
+  url         = {https://taxpolicycenter.org/sites/default/files/publication/164920/a_new_approach_for_estimating_the_impact_of_tax_policies_by_race_and_ethnicity.pdf}
+}
+
+@misc{itep2024race,
+  title  = {ITEP's Approach to Modeling Taxes by Race and Ethnicity},
+  author = {{Institute on Taxation and Economic Policy}},
+  year   = {2024},
+  url    = {https://itep.org/itep-tax-model/iteps-approach-to-modeling-taxes-by-race-and-ethnicity/}
+}
+
+@article{cbo2024race,
+  title   = {CBO Presents Initial Estimates of Taxpayers by Race and Ethnicity},
+  author  = {Heller, Rebecca and Mok, Shannon and Pearce, James},
+  journal = {Congressional Budget Office Blog},
+  year    = {2024},
+  month   = {5},
+  url     = {https://www.cbo.gov/publication/60171}
+}

paper/main.tex

@@ -6,7 +6,7 @@
 \usepackage{hyperref}
 \usepackage{booktabs}
 \usepackage{geometry}
-\usepackage{microtype}
+\usepackage[disable]{microtype}
 \usepackage{xcolor}
 
 % Set citation style in preamble

paper/sections/methodology.tex

@@ -8,6 +8,8 @@ \section{Methodology}\label{sec:methodology}
 \end{figure}
 
 \input{sections/methodology/overview}
+\input{sections/methodology/policyengine}
+\input{sections/methodology/demographic_analysis}
 \input{sections/methodology/demographic_variables}  
 \input{sections/methodology/puf_preprocessing}
 \input{sections/methodology/aging}

paper/sections/methodology/demographic_analysis.tex

@@ -0,0 +1,24 @@
+\subsection{Demographic Analysis}
+
+A key advantage of building from CPS microdata is the ability to analyze policies by demographics not available in tax returns. While organizations using tax return data as their base must develop complex methods to impute race and ethnicity, our approach provides these characteristics directly from the survey data.
+
+The Internal Revenue Service does not collect information on race or ethnicity from tax filers. Other microsimulation models have addressed this limitation through various imputation approaches:
+
+\begin{itemize}
+    \item The Congressional Budget Office statistically matches tax returns to survey records using income and limited demographic characteristics, then validates against linked Census-IRS data \citep{cbo2024race}
+    \item The Tax Policy Center creates multiple copies of each tax unit record, then uses an algorithm to reweight these copies to match aggregate race and ethnicity statistics from survey data \citep{khitatrakun2023race}
+    \item The Institute on Taxation and Economic Policy assigns each tax record probabilities of different racial and ethnic identities based on characteristics like income, marital status, state, and homeownership \citep{itep2024race}
+\end{itemize}
+
+These approaches require complex statistical methods and face inherent limitations in accuracy, particularly when analyzing subgroups or policy impacts that may vary by demographic characteristics not used in the imputation process.
+
+In contrast, our approach provides race and ethnicity variables directly from the CPS without requiring complex imputation. This offers several advantages:
+
+\begin{itemize}
+    \item Race and ethnicity are observed rather than imputed, avoiding potential biases from statistical matching
+    \item Demographic information is available at the individual level, not just for tax unit heads
+    \item The same enhancement methodology can be applied to analyze other demographic characteristics like disability status and educational attainment
+    \item Interactions between demographics can be analyzed naturally (e.g., poverty impacts by both race and age)
+\end{itemize}
+
+This capability enables more reliable analysis of how tax and benefit policies affect different demographic groups. For example, using the enhanced CPS we can directly examine how the Earned Income Tax Credit's benefits vary by race and ethnicity, or analyze the distributional effects of Child Tax Credit reforms across both income levels and demographic categories.

paper/sections/methodology/policyengine.tex

@@ -0,0 +1,80 @@
+\subsection{PolicyEngine Integration and Access}
+
+The enhanced dataset is designed to integrate seamlessly with PolicyEngine, an open-source tax-benefit microsimulation platform available both as a Python package and a web application at \url{policyengine.org}. The platform provides comprehensive tools for analyzing tax and benefit reforms through both programmatic and web interfaces.
+
+\subsubsection{Web Interface}
+
+PolicyEngine's web interface at \url{policyengine.org/us} allows users to:
+\begin{itemize}
+    \item Modify thousands of policy parameters across federal and state tax and benefit programs
+    \item Analyze reforms' impacts on:
+    \begin{itemize}
+        \item Government budgets (federal taxes, benefits, and state/local taxes)
+        \item Income distribution (gains and losses across the income spectrum)
+        \item Poverty (by age, race/ethnicity, and sex using the SPM)
+        \item Inequality (various metrics)
+        \item Labor supply (with customizable elasticities)
+    \end{itemize}
+    \item Generate natural language summaries of policy impacts using Claude 3.5 Sonnet
+    \item Calculate household-specific impacts by entering detailed information
+    \item View marginal tax rates under current law and reforms
+\end{itemize}
+
+\begin{figure}[h]
+    \centering
+    \includegraphics[width=\textwidth]{figures/policyengine_policy.png}
+    \caption{PolicyEngine's policy editor interface, showing modification of the top marginal tax rate.}
+    \label{fig:policyengine_policy}
+\end{figure}
+
+\begin{figure}[h]
+    \centering
+    \includegraphics[width=\textwidth]{figures/policyengine_results.png}
+    \caption{Example distributional analysis from PolicyEngine showing population impacts by income decile.}
+    \label{fig:policyengine_results}
+\end{figure}
+
+\subsubsection{Python Package}
+
+The Python package provides programmatic access with just a few lines of code:
+
+\begin{verbatim}
+from policyengine_us import Microsimulation
+
+# Load enhanced CPS dataset
+sim = Microsimulation(dataset="enhanced_cps_2024")
+
+# Analyze a tax reform
+reform = {
+    "gov.irs.tax_rate.single": {
+        2024: [
+            {"threshold": 400_000, "rate": 0.396}
+        ]
+    }
+}
+reformed = Microsimulation(reform=reform)
+
+# Calculate revenue impact
+baseline_revenue = sim.calculate("income_tax").sum()
+reform_revenue = reformed.calculate("income_tax").sum()
+revenue_impact = reform_revenue - baseline_revenue
+
+# Analyze impacts by group
+income_deciles = sim.calculate("income_decile")
+for decile in range(1, 11):
+    mask = income_deciles == decile
+    impact = reformed.calculate(
+        "household_net_income",
+        mask=mask
+    ).mean() - sim.calculate(
+        "household_net_income",
+        mask=mask
+    ).mean()
+    print(f"Decile {decile}: ${impact:,.0f}")
+\end{verbatim}
+
+\subsubsection{International Applications}
+
+The same enhancement methodology and software infrastructure powers PolicyEngine UK, which incorporates additional data sources including the Living Costs and Food Survey (for consumption) and Wealth and Assets Survey. This demonstrates the approach's adaptability to different national contexts and data environments.
+
+While designed for seamless integration with PolicyEngine, the enhanced dataset is also available as a standalone HDF5 file that can be used with other microsimulation frameworks, promoting broader research applications while maintaining the benefits of integrated analysis tools.