Periodic Reporting for period 1 - GAPREP (O Social Groups, Where Art Thou? A cross-national and longitudinal analysis of the place of social groups at the core of the party-voter representative linkage)
Período documentado: 2023-10-01 hasta 2025-09-30
Democratic representation depends on the relationship between political parties and voters. Yet research has focused almost exclusively on whether parties' policy positions match voters' preferences, overlooking a fundamental question: whom are parties actually talking to? Political parties regularly appeal to specific social groups—women, workers, ethnic minorities, people with disabilities, young people, and others—in their campaign communications. These group appeals matter because once in office, parties shape policies affecting social groups' access to resources, power, and equitable treatment, and group appeals signal which voter constituencies parties claim to represent. Without understanding which groups parties appeal to, we cannot fully evaluate whether democracies represent all citizens fairly.
The current knowledge gap about parties’ appeals to social groups of this kind exists partly because analyzing group appeals across countries and time periods was prohibitively expensive and time-consuming. The GAPREP project tackled this challenge by developing automated text analysis methods using supervised-LLMs to detect and measure group appeals in party manifestos.
The project analyzed over 850 election manifestos from eight European democracies (Austria, Denmark, Germany, Ireland, Spain, Netherlands, Switzerland, and UK) spanning 1970-2025, creating the first comprehensive cross-national database on parties' group appeals. We also conducted survey experiments in Ireland and the UK to test whether voters actually feel represented when parties appeal to groups they belong to.
The project provides researchers worldwide with tools and data to study previously unmeasurable aspects of representation. The output sheds light on which social groups receive political attention and which remain marginalized as well as reveals voters’ evaluations of parties’ representative intentions, both are aspects that matter for democracy. Understanding which groups parties claim to represent helps citizens make informed voting decisions, enables systematic assessment of political inclusiveness, and contributes to strengthening democratic trust and legitimacy across Europe.
The current knowledge gap about parties’ appeals to social groups of this kind exists partly because analyzing group appeals across countries and time periods was prohibitively expensive and time-consuming. The GAPREP project tackled this challenge by developing automated text analysis methods using supervised-LLMs to detect and measure group appeals in party manifestos.
The project analyzed over 850 election manifestos from eight European democracies (Austria, Denmark, Germany, Ireland, Spain, Netherlands, Switzerland, and UK) spanning 1970-2025, creating the first comprehensive cross-national database on parties' group appeals. We also conducted survey experiments in Ireland and the UK to test whether voters actually feel represented when parties appeal to groups they belong to.
The project provides researchers worldwide with tools and data to study previously unmeasurable aspects of representation. The output sheds light on which social groups receive political attention and which remain marginalized as well as reveals voters’ evaluations of parties’ representative intentions, both are aspects that matter for democracy. Understanding which groups parties claim to represent helps citizens make informed voting decisions, enables systematic assessment of political inclusiveness, and contributes to strengthening democratic trust and legitimacy across Europe.
Software Development: The project created the GroupAppeals Python package that enables researchers to analyze group appeals in election manifestos. The package functions are based on four automated models trained and fine-tuned by the project team: the first extracts social groups mentioned in the text; the second determines the stance parties take towards the mentioned social groups; the third detects whether a group-related policy is mentioned in the text; and the fourth classifies the specific group mentions into meaningful social group categories. The package supports analysis of party manifestos in eight languages and will be freely available with comprehensive documentation and usage examples by the end of 2025.
Data Creation: The project produced the Parties' Social Group Appeals (PSoGA) database containing two datasets that cover content from 864 election manifestos of 153 parties across eight countries (1970-2025), measuring appeals to 43 different social groups. This represents nearly 885,000 analyzed sentences. In addition, The ManifestoVault repository provides full-text manifestos in structured format, containing 293,000 sentences from 43 parties in Germany, Ireland, and the UK. The repository is publicly available on DataverseNL. The PsoGA database will be publicly available in early 2026.
Research Findings: Analysis identified over 315,000 social group mentions across 28% of all manifesto sentences, with 76% expressing support for mentioned groups. Social democratic and green parties make more group appeals than conservative parties. Workers, women, youth, and families receive the most attention across the political spectrum. Pilot survey study in Ireland showed that voters are able to recognize parties’ claims of representation of the groups they belong to as well as the groups they do not, and that voters do feel represented by group appeals.
Collaboration and Dissemination: The project established partnerships with universities in the US, Germany, and the UK, resulting in multiple co-authored research papers. Findings were presented at 15+ international conferences and invited talks.
Data Creation: The project produced the Parties' Social Group Appeals (PSoGA) database containing two datasets that cover content from 864 election manifestos of 153 parties across eight countries (1970-2025), measuring appeals to 43 different social groups. This represents nearly 885,000 analyzed sentences. In addition, The ManifestoVault repository provides full-text manifestos in structured format, containing 293,000 sentences from 43 parties in Germany, Ireland, and the UK. The repository is publicly available on DataverseNL. The PsoGA database will be publicly available in early 2026.
Research Findings: Analysis identified over 315,000 social group mentions across 28% of all manifesto sentences, with 76% expressing support for mentioned groups. Social democratic and green parties make more group appeals than conservative parties. Workers, women, youth, and families receive the most attention across the political spectrum. Pilot survey study in Ireland showed that voters are able to recognize parties’ claims of representation of the groups they belong to as well as the groups they do not, and that voters do feel represented by group appeals.
Collaboration and Dissemination: The project established partnerships with universities in the US, Germany, and the UK, resulting in multiple co-authored research papers. Findings were presented at 15+ international conferences and invited talks.
Methodological Innovation: The GroupAppeals Python package is the first comprehensive toolkit specifically designed for analyzing social group appeals in political communications. While existing tools focus on policy positions or social group detection alone, this specialized pacakge enables systematic identification of which groups parties address and how. The multilingual capability advances cross-national research possibilities, addressing a major limitation in the study of group appeals.
Empirical Contribution: The PSoGA database fills a critical gap. Existing datasets systematically measure party policy positions but lack information about group appeals. And the majority of existing data on group appeals is either not available to the public or is significantly limited in scope (single-country or short time-span). The PSoGA database enables researchers to examine which groups receive political attention, how this changes over time, and whether representation via group appeals is inclusive. The 55-year coverage reveals long-term trends in democratic representation across different European contexts.
Theoretical Advancement: Our experimental evidence suggests that voters are aware of parties’ signals of representation and that group appeals influence their perceptions of being represented. This advances our understanding of democratic representation, expanding it beyond the predominant focus on policy agreement as the primary source of party-voter relationships. The findings suggest representation operates through multiple dimensions and has implications for understanding political trust and democratic legitimacy.
Future Development: Continued analysis of project output, especially large-scale survey of UK voters, is currently underway. Integration with established datasets would benefit from formal partnerships with major data providers. Educational adoption is underway through university courses and conference workshops training researchers in these methods.
Broader Impact: Potential applications include monitoring representation gaps, evaluating parties' commitments to inclusion, and enhancing voter information platforms. The findings create opportunities for strengthening democratic quality and legitimacy by illuminating which groups have voice in political systems and which remain marginalized.
Empirical Contribution: The PSoGA database fills a critical gap. Existing datasets systematically measure party policy positions but lack information about group appeals. And the majority of existing data on group appeals is either not available to the public or is significantly limited in scope (single-country or short time-span). The PSoGA database enables researchers to examine which groups receive political attention, how this changes over time, and whether representation via group appeals is inclusive. The 55-year coverage reveals long-term trends in democratic representation across different European contexts.
Theoretical Advancement: Our experimental evidence suggests that voters are aware of parties’ signals of representation and that group appeals influence their perceptions of being represented. This advances our understanding of democratic representation, expanding it beyond the predominant focus on policy agreement as the primary source of party-voter relationships. The findings suggest representation operates through multiple dimensions and has implications for understanding political trust and democratic legitimacy.
Future Development: Continued analysis of project output, especially large-scale survey of UK voters, is currently underway. Integration with established datasets would benefit from formal partnerships with major data providers. Educational adoption is underway through university courses and conference workshops training researchers in these methods.
Broader Impact: Potential applications include monitoring representation gaps, evaluating parties' commitments to inclusion, and enhancing voter information platforms. The findings create opportunities for strengthening democratic quality and legitimacy by illuminating which groups have voice in political systems and which remain marginalized.