Primary Data Sources¶
PolicyEngine-US¶
- Comprehensive tax calculator for U.S. households
- Incorporates federal and state tax codes
- Accounts for credits, deductions, and phase-outs
- Version 0.700.0 or higher required
Current Population Survey (CPS)¶
- Annual Social and Economic Supplement (ASEC)
- Sample size: ~60,000 households
- Variables: Income, employment, demographics
- Years: 2019-2024
IRS Statistics of Income (SOI)¶
- Administrative tax return data
- Aggregated to preserve confidentiality
- Marginal tax rate distributions by income
- Years: 2018-2022
Variable Definitions¶
Marginal Tax Rate (MTR): The tax rate on an additional dollar of income, including federal income tax, state income tax, and payroll taxes, accounting for all applicable credits and deductions.
Wage Rate: Hourly earnings calculated as annual labor income divided by annual hours worked (imputed as 2000 for full-time workers).
Tax Rate Uncertainty: Standard deviation of possible tax rates faced by a household, measured as either:
- Historical variation in effective rates over past 5 years
- Subjective uncertainty from survey responses
- Complexity-based measure using tax form interactions
Sample Construction¶
Starting with CPS microdata, we:
- Select households with positive labor income
- Exclude self-employed (different tax treatment)
- Match to PolicyEngine-US for tax calculations
- Compute marginal rates using $1000 income increments
- Weight observations using CPS household weights
Final sample: ~45,000 households representative of U.S. population
Robustness Sample Definitions¶
Alternative samples used in robustness checks:
- Prime-age workers: Ages 25-54 only
- Full-time workers: 35+ hours per week
- Stable employment: Same job for 2+ years
- High-tax states: CA, NY, NJ, CT, MA
- Low-tax states: TX, FL, WA, NV, TN
Code and Replication¶
All analysis code is available at: https://
Requirements:
- Python 3.9+
- PolicyEngine-US
- NumPy, Pandas, SciPy
- Jupyter notebooks for exhibits
Replication instructions provided in repository README.