Code repository

The code and raw data for this project can be found at GitHub.

Replication instructions

All the figures and tables for the paper can be replicated and reproduced with R. Ensure the following packages are installed:

This can be done with this R code:

Also, install these additional items:

The small inline histograms in the summary statitics table in the appendix will only display correctly when they use a font that supports block elements, such as DejaVu Sans or Arial.

The project is structured as an RStudio project and uses RMarkdown to generate a website in the _site/ subdirectory. This is currently hosted online at, but the _site/ directory can be uploaded anywhere else too.

Open RStudio and go to Build > Build All to run all the scripts and generate everything. After building, you should have the following files:

Bonus table conversion: The majority of the project’s tables are generated with Markdown and automatically converted to .docx when the project is built. Tables created with stargazer, however, do not support Markdown and instead exported as HTML files. You need to open these HTML files directly in Word to copy/paste to a different document.

Alternatively, the script included at bin/ will find all stargazer-generated HTML tables and save them as well-formatted .docx files automatically. This script requires Python 3 and the BeautifulSoup package, and the script will only run on macOS with Microsoft Word installed.

Currently, this script tries to run when 99_wrap-up.Rmd is built, but there are checks to ensure it doesn’t try to run on Windows or Linux, or if Python 3 isn’t installed. For an additional precaution, the variable stargazer2word also must be set to TRUE for the conversion script to run. If you have macOS and Python 3 + BeautifulSoup and Word and you want to have R convert all the tables for you, manually edit 99_wrap-up.Rmd and set stargazer2word <- TRUE.