The party-voter representative linkage is central to modern democracies. Because political parties, once in office, steer fiscal and social policies, the social groups they claim to represent in their group appeals matter for distributing tangible public goods and specially the equitable treatment of minority groups by the government. Combining insights from political science, communication science, and computational linguistics, this project illuminates social groups’ access to power, and social and political equality. Offering a new perspective on the party-voter linkage, the project aims to contribute to scholarly understandings of parties’ supply-side politics and voters’ reactions to parties’ group appeals, as well as to improving voters’ contemporary low levels of political trust.
The EU-funded GAPREP project does the following:
- Develops four new models that together detect group appeals in party manifestos (to be collected into an Python/R package) (models are linked below and discussed in the pre-print below)
- Constructs a novel database of group appeals (PSoGA) that covers 864 manifestos of 153 political parties in Austria, Denmark, Germany, Ireland, Netherlands, Spain, Switzerland and the UK from 1970-2025 (see pre-print below)
- Examines voters’ reactions to group appeals and recognition of representation via survey analysis.
Outputs from the project:
Dolinsky, A.O., Huber, L.M., and Horne, W. Who do Parties Speak To? Introducing the PSoGA: A New Comprehensive Database of Parties’ Social Group Appeals. (OSF pre-print)
Horne, W., Dolinsky, A.O., and Huber, L.M. Using LLMs to detect group appeals in election manifestos of parties. (OSF pre-print)
OSF Project Registery: https://osf.io/p9nvg/
Computational Models for Text Analysis:
Token classification (social group detection): https://huggingface.co/rwillh11/base-mdbertav3-token-classification-groups-bilingual
Stance detection: https://huggingface.co/rwillh11/mdeberta_NLI_stance_2.0
Policy: https://huggingface.co/rwillh11/mdeberta_NLI_policy_2.1
Meaningful groups: https://huggingface.co/rwillh11/mdeberta_groups_2.0
The ManifestoVault repository the full texts of national-level general election manifestos at a natural-sentence level of 43 political parties from Germany, Ireland, and the UK from 1970-2025. Our aim is to provide clean and accurate (as much as possible) full-text versions of election manifestos to facilitate all scholarly research for which parties’ election manifestos are relevant. [link]
Alona O. Dolinsky 