Text Analysis: Estimating Policy Preferences from Written and Spoken Words

This chapter provides an introduction into the emerging field of quantitative text analysis. Almost every aspect of the policy-making process involves some form of verbal or written communication. This communication is increasingly made available in electronic format, which requires new tools and methods to analyze large amounts of textual data. We begin with a general discussion of the method and its place in public policy analysis, including a brief review of existing applications in political science. We then discuss typical challenges that readers encounter when working with political texts. This includes differences in file formats, the definition of “documents” for analytical purposes, word and feature selection, and the transformation of unstructured data into a document-feature matrix. We will also discuss typical pre-processing steps that are made when working with text. Finally, in the third section of the chapter, we demonstrate the application of text analysis to measure individual legislators’ policy preferences from annual budget debates in Ireland.