PrefPol Replication

This repository provides Julia code to replicate the preference polarization analysis from the PrefPol project. The entry point for reproducing the figures and tables of the paper is running/running.jl, which orchestrates the entire data pipeline and visualization workflow.

Prerequisites

• Julia 1.10.10 – download from julialang.org.
• R 4.3.3 – download from CRAN.
• Required R packages: PerMallows, haven, and MICE.

  install.packages(c("PerMallows", "haven", "MICE"))

Data

Each election year is configured by a TOML file. Within each TOML file the data_file field points to the raw dataset on disk. Obtain the datasets from:

• 2006: https://www.cesop.unicamp.br/por/banco_de_dados/v/1583
• 2018: https://www.cesop.unicamp.br/por/banco_de_dados/v/4538
• 2022: https://www.cesop.unicamp.br/por/banco_de_dados/v/4680

Copy the downloaded files to a convenient location and update the data_file path in the corresponding TOML configuration.

Note: The example configuration sets the number of pseudo‑profiles to 3 for quick runs. The paper uses 1000 pseudo‑profiles.

Any extra dataset must add a preprocessing_specific function to the preprocessing_specific.jl file and a TOML configuration file at config.

Running the pipeline

1. Activate and instantiate the project

2. Execute the main script running/running.jl

I run it interactively in a text editor (emacs, vscode).

What running.jl does

The script performs the full replication workflow:

1. Resampling – PrefPol.save_all_bootstraps generates and stores pseudoprofiles (resamples) of each year.
2. Imputation – PrefPol.impute_from_f3 fills missing rankings using multiple imputation.
3. Profile generation – PrefPol.generate_profiles_for_year_streamed_from_index creates preference profiles for every year/scenario combination.
4. Global measures – PrefPol.save_or_load_measures_for_year computes global polarization metrics for each election year (Ψ, RHHI)
5. Group metrics – PrefPol.save_or_load_group_metrics_for_year calculates group metrics (C,D,G)
6. Plots – PrefPol.plot_scenario_year and PrefPol.plot_group_demographics_lines produce scenario‑specific and group figures which are saved via PrefPol.save_plot.

R integration

Certain preprocessing steps rely on R. Ensure R can find the required packages and that R is available on your system PATH before running the Julia script.

Testing

Run the unit tests with ] test in the repl.

The tests exercise the preprocessing and polarization measures modules.