Algorithmic Assessment of Agency

The Measures for Advancing Gender Equality (MAGNET) initiative aims to broaden and deepen the measurement of women’s agency, based on the development of new tools and rigorous testing and comparison of both new and existing methods for measuring agency, and promoting the adoption of these measures at scale. By increasing the availability of innovative meaningful measures of agency for a broad range of contexts, we hope our work will lead to an improved understanding of what women’s agency is, how it manifests and how it can best be measured across contexts given the research question at hand. Measuring women’s agency with women’s own narratives through open-ended interviews is considered the gold-standard. Yet the data processing relies on hand-coding presenting limitations related to budget, time, and discretion of annotators. This tool, Algorithmic Assessment of Agency, is a text-analysis measure of women’s agency, drawing on text-to-speech transcripts of women’s answers to open-ended questions. This method will adapt the Franzosi (1994, 2004) methodology for analyzing semi-structured interviews, attributing an automated agency score to women’s narratives about how they make decisions, how they live their life and how this compares to their preferences. These questions are well-suited for surveys for targeted thematic or impact evaluation surveys designed to understand individuals' agency and decision-making. We recommend the use of this tool in the design, monitoring, and evaluation of development programs aimed at improving women’s empowerment. This data study includes following files. 1. A survey document (including implementation guidelines). 2. Two files, CAPI_Choices and CAPI_Survey, along with the accompanying files, can be used to construct a CAPI program ready for survey implementation. Alternatively, users can use an Excel workbook "CAPI_.xlsx" that includes worksheets for survey and choices, along with others, for constructing a CAPI program ready for survey implementation.