Financial Payouts and Stock Market Participation: Evidence from the Paycheck Protection Program

Here is the Zenodo description ready to paste: Title: Replication Package: Financial Payouts and Stock Market Participation — Evidence from the Paycheck Protection Program Description: This repository contains the complete replication package for "Financial Payouts and Stock Market Participation: Evidence from the Paycheck Protection Program." The package includes all source data files and Jupyter notebooks necessary to independently replicate the full dataset construction, variable construction, panel estimation, instrumental variables estimation, and robustness checks reported in the paper. Source Data Files The following publicly available data sources are included directly in this repository. IRS Statistics of Income county-level files for tax years 2019, 2020, 2021, and 2022, downloaded from the IRS SOI website at irs.gov/statistics/soi-tax-stats-county-data. IPUMS NHGIS American Community Survey 5-year county-level files for the periods 2015–2019, 2016–2020, 2017–2021, and 2018–2022, covering total population, median age, race, educational attainment, and median household income. SBA Paycheck Protection Program loan-level FOIA data for all program rounds, downloaded from data.sba.gov/dataset/ppp-foia. HUD zip-to-county residential crosswalk for Q1 2020, used to allocate zip-code-level PPP loan amounts to counties by residential population share. Jupyter Notebooks Three Jupyter notebooks are included covering the complete analytical pipeline. The first notebook, covers all data loading, cleaning, and variable construction steps including IRS SOI processing and AGI group construction, NHGIS demographic variable construction with column-level documentation of the NHGIS extract codes used for each year, SBA PPP loan aggregation to county-year level including fintech lender classification following Erel and Liebersohn (2020), HUD zip-to-county crosswalk merge, and construction of the final panel dataset panel_final.csv. The second notebook, covers the funding gap instrument construction, first-stage OLS regression with state fixed effects, F-statistic and partial F-statistic computation, two-stage least squares estimation for AGI groups G1 and G5, and difference-in-differences estimation with county fixed effects. The third notebook, PPP_Robustness_Checks.ipynb, covers the lagged robustness check using 2021 stock market participation as the outcome variable with the 2020 funding gap instrument, urban and rural county subsamples, high and low income county subsamples, and per capita PPP loan amount alternative specification. Key Variables The final panel dataset panel_final.csv contains 12,570 county-year observations covering 3,142 counties across four years. The primary dependent variables are smp_g1 through smp_g5, representing the share of tax filers reporting dividend income for five progressively higher AGI groups. The primary independent variable is ln_ppp, the natural log of aggregate PPP loan amount per county per year. The instrumental variable is pregap_share, the share of 2020 county PPP loan volume approved in the 13 days prior to the April 16 funding gap. Demographic controls are drawn from NHGIS and include log population, log median household income, log median age, percent White, and percent college educated. Software Requirements All code is written in Python 3.12 and runs in Google Colab or any standard Jupyter environment. Required packages are pandas, numpy, statsmodels, linearmodels, scipy, and functools, all of which are available via pip. The linearmodels package is required for IV2SLS estimation and can be installed with pip install linearmodels. Citation If you use this replication package please cite both the paper and the IPUMS NHGIS data source. The NHGIS citation is: Jonathan Schroeder, David Van Riper, Steven Manson, Katherine Knowles, Tracy Kugler, Finn Roberts, and Steven Ruggles. IPUMS National Historical Geographic Information System: Version 20.0. Minneapolis, MN: IPUMS. 2025. http://doi.org/10.18128/D050.V20.0