About

The Manifesto Project collects and analyses party manifestos using comparative content analysis to determine parties' policy preferences. The project looks back on a history spanning almost 50 years, which began with the work of the Manifesto Research Group (MRG, 1979-1989), and extends through the Comparative Manifesto Project (CMP, 1989-2009), and the DFG-funded project Manifesto Research on Political Representation (MARPOR, 2009-2024). In 2003, the project received the American Political Science Association’s (APSA) award for the best dataset in comparative politics. In 2016, it received the Berlin Digital Humanities Award for the Manifesto Corpus as well as the Statistical Software Award from the Scociety for Political Methodology of the APSA for the Manifesto Corpus and manifestoR. Since November 2022, the project has been carried out in cooperation between the WZB Berlin Social Science Center and the Institute for Democracy Reserach at the University of Göttingen.

The project team coordinates the generation and the analysis of the data with the support of coders from over 67 different countries. The project's website provides instant access to the project's core products, which are constantly updated and freely accessible to the research community and the general public. It also provides accessible ways to explore and visualise the data and text corpus. The data is being used by many researchers around the world. The project's website has more than 40,000 registered users, with around 5,000 of them downloading the data each year and around 10 new registrations per day.

The core products are:

Manifesto Dataset: Our main dataset covers over 1400 parties from 1945 until today in over 67 countries on five continents. It is updated once a year. It can be accessed via our R-package manifestoR and our API. All original documents can also be downloaded. More information and how to use

Manifesto Corpus: We provide a large, multilingual text corpus of over 3,200 machine-readable election programs, of which more than 2,000 are annotated at the quasi-sentence level, resulting in more than 2 million machine-readable quasi-sentences. The corpus is continuously updated and improved and is widely used in the text-as-data community and beyond. Since 2024, the documents, originally in over 40 languages, have also been available in a validated, machine-translated English version of the complete corpus. More information and how to use

manifestoR: ManifestoR is a free package for the open source statistical software R. It provides access to coded election programmes from the Manifesto Corpus and to the Manifesto Project's Main Dataset. The Manifesto Corpus contains the collected and annotated election programmes as tibble/data.frame or in the Corpus format of the package 'tm' to enable easy use of text processing and text mining functionality. More information and how to use

manifestoberta: Since 2023, we have been operating the models manifestoberta-sentence and manifestoberta-context, two large language models (based on multilingual XLM-RoBERTa large models) that have been trained on about 1.9 million annotated statements from the manifesto corpus. They demonstrate high performance in classifying political texts and can assign a variety of political discourses in different languages to the categories of the Manifesto coding scheme. The models are updated regularly and are freely available via the Hugging Face Model Hub. More information and how to use

Substantively, the Manifesto Project allows researchers, for instance, to analyse the role of parties at different stages of the political process, and to examine the quality of programmatic representation. It enables studies on the programmatic supply of parties, the relation between parties and voters, the role of parties in parliament, and the translation of party programmes into policy output.