Supervised scaling of semi-structured interview transcripts to characterize the ideology of a social policy reform

摘要

Automated content analysis methods treat "text as data" and can therefore analyze efficiently large qualitative databases. Yet, despite their potential, these methods are rarely used to supplement qualitative analysis in small-N designs. We address this gap by replicating the qualitative findings of a case study of a social policy reform using automated content analysis. To characterize the ideology of this reform, we reanalyze the same interview data with Wordscores, using academic publications as reference texts. As expected, the reform's ideology is center/center-right, a result that we validate using content, convergent and discriminant strategies. The validation evidence suggests not only that the ideological positioning of the policy reform is credible, but also that Wordscores' scope of application is greater than expected.