Replication Data: On the Prevalence of Condorcet's Paradox\
Authors: Salvatore Barbaro and Anna Kurella
Public Choice (2025/2026)
*DOI: https://doi.org/10.1007/s11127-025-01353-7 *\
This repository contains replication scripts for the study on the prevalence of Condorcet's Paradox. The analysis uses data from the Comparative Study of Electoral Systems (CSES). The scripts are written in R.
Note: You will need to download the CSES data directly from the CSES project website. The source and DOI are provided in the paper's reference section and below (Section Data Accessibility). Quick Access: In case you wish to skip running the replication files on inference, you can use their outputs in the subfolder "OutputData", which also includes an R-Script.
The paper uses data from the CSES project, which is subject to redistribution restrictions but can be freely downloaded after registration.
- Data DOI: 10.7804/cses.imd.2020-12-08
Main.R: Main script.
InferenceBootstrap_Parties.R: Performs 10,000 bootstrap replications on each party election.InferenceBootstrap_Candidates.R: Performs 10,000 bootstrap replications on each candidate election.Inference_Noise: Performs 10,000 replications on each election with noise interval: [-1.1, 1.1].Inference_PBW.R: Performs 10,000 bootstrap replications by considering the PBW.
cabinet.R: Assesses how often the Condorcet winner (or loser) belongs to the cabinet after the respective election.
- R Version: 4.5.1 (Great Square Root)
- Required R Libraries:
voteparalleldplyrdata.tablestringrtidyrboot
The scripts have been tested on both Linux-Debian and Windows systems. The scripts for bootstrap / noise replications we have run on the MOGON (HPC at the Johannes Gutenberg University (hpc.uni-mainz.de). Be aware that the parallel functions do not work on non-Linux-OS. Adjust the scripts accordingly.
All bootstrap and random-noise replications use a fixed seed value for reproducibility:
Seed Value: 55234 (the authors' zip code).
- Main Scripts: ~16 minutes on a standard (2024) desktop machine.
- Bootstrap: ~8 hours each on a high-performance computer with 40 CPU cores.
- Noise: ~ 16 hours on a high-performance computer with 40 CPU cores.
- Weight: ~ 3 hours on a high-performance computer with 40 CPU cores.
If running the scripts on a non-UNIX or non-Linux system, consider revising the parallelization logic for optimal performance.
Detailed descriptions of the code functionality are included as comments within each script.
- Start R.
- Run
Main.R.
For advanced replication (e.g., bootstrap and Random Noise), execute the respective scripts as per the instructions in the comments.
We thank the Comparative Study of Electoral Systems (CSES) project for providing the data. This work relies heavily on their comprehensive electoral data collection. We are grateful to the Mogon / HPC group for granting processor time on MOGON.
This project is licensed under MIT.