Description:

Seema Jayachandran (Northwestern University): "A five-question women's agency index created using machine learning and qualitative interviews" (with M. Biradavolu and J. Cooper)

Abstract: We develop a new short survey module for measuring women's agency by combining mixed-methods data collection and machine learning. We select the best five survey questions for the module based on how strongly correlated they are with a "gold standard" measure of women's agency. For a sample of 209 women in Haryana, India, we measure agency, first, through a semi-structured in-depth interview and, second, through a large set of close-ended questions.  We use qualitative coding methods to score each woman's agency based on the interview, which we treat as her true agency. To identify the subset of close-ended questions most predictive of the "truth", we apply statistical methods similar to standard machine learning except that we specify how many survey questions are selected. The resulting 5-question index is as strongly correlated with the coded qualitative interview as is an index that uses all of the candidate questions. We also considered a second gold standard measure of agency, a real-stakes choice between money for oneself or one's husband. This lab game, however, does not measure agency cleanly in our setting. Thus, our preferred survey measure of agency is the one validated against qualitative interviews.