Structure of the repo

• source: contains all code.
  • source/build assembles the dataset.
  • source/analysis runs regressions and creates tables, figures and maps.
  • source/utils contains utility code that is used by multiple scripts.
  • source/slides contains code that I added after the submission (for my presentation slides, WIP)
• paper: contains all materials for compiling PDFs.
  • paper/output: contains figures, tables and maps created by code in source.
  • paper/paper.tex: compiles into the submitted PDF.
  • paper/references.bib: bibliography (I don't cite every paper in it)
  • paper/slides.tex: not part of my submission, compiles into my presentation slides (WIP).
• drive: contains data and literature.
  • drive/raw: raw datasets from which I build everything else.
  • drive/derived: my data build(s).
  • drive/literature: downloaded PDFs of papers I cite (and some more I don't cite).

Note: The branch main_submission contains the code as it was on June 27, without additions made after the deadline.

Instructions for replication

The easiest way to replicate my paper is to install the open-source build utility SCons, which I use. SCons relies on a SConstruct file in the root directory of the project which functions like a blueprint and tracks the dependencies between code, data, and output files. With SCons installed, running the entire pipeline in the correct order is a breeze:

• clone this repo to your computer.
• move the drive folder from Dropbox into the BA root directory.
• at the beginning of every R script in BA/source, change the setwd("~/GitHub/BA") to the correct file path if necessary.
• open the command line / Terminal and cd to the BA root directory.
• Enter scons into the command line. This will automatically execute the whole pipeline in the correct order.

Tables, figures, and maps are written to BA/paper/output. The submitted PDF was separately compiled from BA/paper/paper.tex.