Since 1979, the Manifesto Research Group (MRG) has been collecting and coding election programmes with the aim of estimating the policy preferences of political parties. Detailed descriptions of this project can be found in Budge et al. (1987), Laver and Budge (1992), Klingeman et al. (1994). During the first phase of the project, while the classification scheme was being developed, each group member was responsible for his or her own data collection. The second phase of the project started in 1989. In the context of its ‘Comparative Manifestos Project’ (CMP), the Social Science Research Centre – Berlin provided resources for updating and expanding the MRG data set to 2,347 programmes of 632 different parties in fifty-two countries.1 Coders are now hired to do the content analysis according to a coding handbook. A reliability test given in the handbook is used for training coders. This paper thus sets out to review the MRG classification scheme in terms of quality control.