Manifesto Project

Manifesto Project
Name	Manifesto Project
Established	2003
Discipline	Political science
Scope	Comparative politics, party politics
Main instrument	Manifesto Research on Political Representation
Datasets	Comparative Manifestos Project datasets

Manifesto Project The Manifesto Project is an academic initiative that systematically collects, codes, and analyzes political party program texts from national and regional elections. It supports comparative research on parties such as Christian Democratic Union (Germany), Labour Party (UK), Social Democratic Party of Germany, Republican Party (United States), and Liberal Party of Australia by providing standardized quantitative indicators drawn from manifestos used in elections like United Kingdom general election, 2010, United States presidential election, 2008, German federal election, 2005, French legislative election, 2007, and Swedish general election, 2010.

Overview

The project produces a longitudinal corpus covering parties and elections across countries including United Kingdom, United States, Germany, France, Spain, Italy, Sweden, Norway, Denmark, Netherlands, Belgium, Austria, Poland, Czech Republic, Hungary, Greece, Portugal, Ireland, Finland, Switzerland, Canada, Australia, New Zealand, Japan, South Korea, Brazil, Argentina, Chile, Mexico, Colombia, South Africa, Israel, Turkey, Russia, India, China, Indonesia, Philippines, Thailand, Egypt, Morocco, Tunisia, Algeria, Nigeria, Kenya, Ukraine, and Romania. It links party positions to outcomes such as coalition formation in cases like German federal election, 1998 and policy shifts following elections such as Swedish general election, 1994.

History and Development

Originating with scholars associated with institutions such as University of Mannheim, The University of Zurich, Sciences Po, Harvard University, University of California, Berkeley, London School of Economics, University of Oxford, University of Cambridge, Columbia University, Princeton University, Yale University, University of Michigan, Stanford University, and Max Planck Institute for the Study of Societies, the project evolved from earlier endeavors like the Comparative Manifestos Project and input from researchers who worked on projects connected to Data Archive, International Social Survey Programme, European Social Survey, World Values Survey, and ParlGov. Key figures and teams have engaged with datasets used in studies published in journals such as American Political Science Review, British Journal of Political Science, Comparative Political Studies, and European Journal of Political Research. Major updates to coding schemes and release versions coincided with conferences like International Political Science Association meetings and workshops at ECPR.

Methodology

The project employs content analysis protocols combining manual coding by trained coders and computerized text analysis techniques that draw from methods used in Dictionary of Political Change style approaches and machine-learning experiments inspired by work at Massachusetts Institute of Technology, Carnegie Mellon University, University of Toronto, and University of Amsterdam. Coders segment party texts and assign categories comparable to prior schemes used in studies of parties like German Green Party, French Socialist Party, and Communist Party of China. Inter-coder reliability procedures reference standards discussed in publications from American Association for Public Opinion Research and tools developed at Stanford Natural Language Processing Group. The methodology integrates deviance adjustments, rescaling procedures, and checks against external benchmarks such as expert surveys including Chapel Hill Expert Survey and datasets like MARPOR-style codings.

Dataset Content and Variables

Variables capture category frequencies reflecting emphases on topics associated with parties such as Conservative Party (UK), Socialist Party (France), and Fidesz. Standard variables include percentages of quasi-sentences allocated to issue categories, left–right position estimates, and measures of salience for items related to policy areas debated in events like Maastricht Treaty referendum, 1992 and Treaty of Lisbon referendum, 2008. Metadata records party names, election dates, country identifiers, incumbency status in contexts like Italian general election, 2001, and document type (manifesto, program, manifesto supplement). Later releases added sentence-level annotations and links to full texts, facilitating replication of studies on topics such as manifesto change prior to events like European Parliament election, 2009.

Usage and Applications

Researchers employ the data to study electoral competition exemplified by comparisons between Rassemblement National and The Republicans (France), policy convergence theories tested against cases like German federal election, 2009, coalition bargaining studies involving parties such as Free Democratic Party (Germany) and Christian Social Union in Bavaria, and analyses of party transformation observed in Scottish National Party and Sinn Féin. The dataset is used in quantitative modeling with software from StataCorp, R (programming language), Python (programming language), and network analysis packages developed at University of California, Santa Cruz. It supports teaching in graduate courses at institutions like Princeton University and Yale University.

Criticisms and Limitations

Scholars have critiqued the project for potential coder bias raised in debates involving researchers from University of Essex, Queen Mary University of London, and University of Warwick, for issues of construct validity compared with expert surveys like Chapel Hill Expert Survey, and for challenges comparing cross-national variations illustrated by cases such as Brazilian general election, 2018 and Indian general election, 2014. Methodological limitations noted by commentators at European University Institute and Central European University include sensitivity to translation choices, temporal coverage gaps for post-communist parties like Communist Party of Czechoslovakia, and the coarse granularity of some categories when applied to niche parties like Pirate Party (Sweden) or Jobbik.

Access and Licensing

Data releases are distributed under academic data-sharing terms used by repositories such as Harvard Dataverse, ICPSR, GESIS, and institutional archives at University of Mannheim and Sciences Po. Users typically cite release versions and original coders in publications appearing in outlets like Journal of Politics and Party Politics. Licensing requires adherence to citation norms and, in some cases, registration agreements used by archives such as UK Data Service.

Category:Comparative politics datasets