Representation Failure — Replication Repository

Overview

This repository contains the replication code for the paper on representation failure and candidate ideology in Brazilian elections. The analysis estimates candidate ideal points (CFscores) using campaign contribution data and validates them against survey-based measures.

Directory Structure

representation_failure/
├── data/
│   ├── input/                       # Raw input data — not committed (see Data section)
│   │   ├── tse/                     # Raw TSE electoral files (candidates, parties, members)
│   │   ├── municipal/               # Municipal shapefiles, IBGE identifiers, state IDs
│   │   ├── contribution/            # Campaign finance files (campaign_fed_state.csv.gz, etc.)
│   │   ├── ideology/                # Legislative survey ideology data, Latinobarometer
│   │   ├── corruption/              # TSE cassation files and 2018 candidate list
│   │   ├── candidate/missing/       # Manually corrected missing candidate records
│   │   ├── spoils_of_victory/       # Replication data from Brollo et al. (2013)
│   │   ├── state/                   # State-level census data
│   │   └── identifiers/             # State name crosswalks
│   └── output/                      # All script-generated outputs (not committed)
│       ├── candidate/
│       │   ├── cfscore_estimation/  # Cleaned candidate files used in cfscore estimation
│       │   ├── fed_state/           # Federal and state legislative candidates by year
│       │   └── local/               # Municipal-level candidates by year
│       ├── election/                # Cleaned election, vote count, and coalition files
│       ├── ideology/                # Contribution matrices, cfscore estimates, legislative ideology
│       ├── municipal/               # Processed IBGE municipal dataset
│       └── corruption/              # Constructed corruption/valence indicators
├── scripts/                     # R scripts (numbered workflow below)
├── figs/                        # Output figures (PDF/PNG)
├── results/                     # LaTeX tables and validation outputs
├── docs/                        # Manuscript sections and appendices
├── renv.lock                    # Package lockfile for reproducibility
└── replication.Rproj            # RStudio/Positron project file

Data

All data files are not committed to this repository. They are available in the shared Dropbox folder. Download both the data/input/ and data/output/ directories from Dropbox and place them under data/ at the root of this repository before running any scripts.

If you wish to reproduce outputs from scratch, you only need data/input/. The data/output/ directory will be populated as you run the scripts in order.

Reproducing the Analysis

1. Prerequisites

• R ≥ 4.4
• renv for package management

2. Restore the R environment

Open the project (replication.Rproj) and run:

renv::restore()

This installs all packages at the exact versions recorded in renv.lock.

3. Run the scripts

Execute the scripts in the following order:

Step	Script	Description
1	scripts/candidates_wrangle.R	Clean raw TSE candidate/election data → data/output/candidate/, data/output/election/
2	scripts/censo_wrangle.R	Process IBGE census covariates → data/output/municipal/
3	scripts/contrib_matrix_pooled.R	Build candidate–donor contribution matrix → data/output/ideology/
4	scripts/candidates_cfscore_pooled.R	Estimate pooled CFscores → data/output/ideology/
5	scripts/candidates_pooled_robustness.R	Robustness checks for CFscore estimation → data/output/ideology/
6	scripts/create_state_level_ideology.R	Aggregate ideology to state level → data/output/ideology/
7	scripts/fix_defective_candidates.R	Correct defective candidate records in data/output/election/
8	scripts/generate_valence_corruption.R	Construct valence/corruption measures → data/output/corruption/
9	scripts/validate_cfscore.R	Validate CFscores against survey benchmarks
10	scripts/validate_cfscore_with_spoils.R	Validate using spoils-based measures
11	scripts/validate_contrib_matrix.R	Validate the contribution matrix
12	scripts/validate_unobserved_valence.R	Validate unobserved valence measure → data/output/corruption/
13	scripts/candidates_diagnostics.R	Diagnostics and summary statistics
14	scripts/candidate_figs.R	Generate all manuscript figures → figs/

Helper functions used across scripts are defined in scripts/functions.R and scripts/funs.R.