My work employs a range of methods, with particular contributions in the field of text analysis. By building large open-source datasets and methods to analyze large volumes of political texts, I aim to expand the aspects of policymaking that are open to systematic study.
Primarily, I develop methods for integrating human coding of texts with computational text analysis to increase the inferential power of hand coding. Computational text analysis tools can strategically select texts for human coders, including texts that represent larger samples and outlier texts of high inferential value. Preprocessing documents can speed hand-coding by extracting key features like named entities. I show that hand-coding and text analysis tools are each more powerful when combined in an interactive workflow. For example, we can extend inferences about who is able to influence public policy by analyzing large numbers of public comments on proposed government policies. Using this iterative mix of hand-coding and computational text analysis methods, a hand-coded sample of 10,894 hand-coded public comments on proposed policies yields equally valid inferences for over 40 million comments regarding the extent to which policy changed in the direction they sought. I thus measure lobbying success—the extent to which each group got their desired changes between the draft and final policy—for millions of people and thousands of organizations. This large sample enables new analyses of the relationships between lobbying coalitions, social movements, and policy change.