Parties' ideological cores vs peripheries

Replication data and scripts for BJPS article “Parties’ ideological cores and peripheries: Examining how parties balance adaptation and continuity in their manifestos”

---

1_get_marpor_manifestos.R

Purpose: Downloads and processes the MARPOR (Manifesto Project) corpus into a quasi-sentence–level tidy dataset suitable for classification.

Output:

• marpor_corpus_quasi.RDS

---

2_classify_quasi_berta.py

Purpose: Classifies quasi-sentences using a the XLM-RoBERTa-based model manifestoberta (see here for documentation: https://doi.org/10.25522/manifesto.manifestoberta.56topics.sentence.2023.1.1).

Inputs:

• marpor_corpus_quasi.RDS

Related files:

• classify_quasi.slurm — SLURM batch script used for running the classifier on high-performance computing cluster at the University of Innsbruck (see here: https://www.uibk.ac.at/zid/systeme/hpc-systeme).

Outputs:

• marpor_corpus_quasi_classified.RDS — final merged output of classified quasi-sentences

Intermediate batch files:
These files provide access to marpor_corpus_quasi_classified.RDS in three separate batches that are later merged in the analysis stage.

• marpor_quasi_classified_a.RDS — Text of quasi-sentences with MARPOR variables and unique ID (1:nrow()).
• marpor_quasi_classified_b.RDS — Classified labels (top 1–5) and probabilities with matching IDs.
• marpor_quasi_classified_c.RDS — Classified labels (top 6–10) and probabilities with matching IDs.

---

3_analyze_quasi_classifications.R

Purpose:
Merges the classified outputs and reproduces all statistical models and figures presented in the article.

Inputs:

• marpor_quasi_classified_a.RDS
• marpor_quasi_classified_b.RDS
• marpor_quasi_classified_c.RDS
• pg_missing_coded.xlsx — Manual coding of parties' incumbency status where ParlGov information is missing.

Outputs:

• Statistical models and figures as reported in the BJPS article.

---

Dependencies

The analyses were conducted using R and Python.

• R version: 4.3.3
  • Main required packages: tidyverse, readxl, manifestoR, ggeffects, fixest, knitr, tidytext, countrycode.
• Python version: 3.13
  • Main required packages: transformers, torch, pandas, pyreadr.

---

Citation

Werner, Annika & Habersack, Fabian (forthcoming). Parties’ ideological cores and peripheries: Examining how parties balance adaptation and continuity in their manifestos, British Journal of Political Science, OSF Preprint: https://doi.org/10.31235/osf.io/pjes3_v1