Understanding Party Manifestos: A Comparison of Human and Automated Coding

Analyzing political documents is essential to test a series of hypotheses regarding the priorities of political actors and institutions, but the methods used to analyze political texts are (too) diverse. Text analysis techniques based on human coding are the most widespread and provide the best information regarding the content of these documents. However, automated techniques display many advantages. This paper intends to compare the results obtained using two distinct methods of analysis, namely a method of human coding and an automated coding method, of the same body of political texts: party manifestos of all Belgian political parties between 1987 and 2007. Based on the CAP (Comparative Agendas Project) codebook, the identification and comparison of the main issues addressed in these documents will determine the similarities and differences obtained. Ultimately, this paper discusses these two methods and the lessons that can be learnt for the future analysis of policy documents.