Research Begins with a Goal: A Framework Illustrated by Gap-Closing Estimands

Abstract

Quantitative social science is at a moment of tremendous opportunity. Computational power has been multiplying for decades, accelerating the development of new statistical and machine learning tools. But tools are only as good as the questions they are used to answer: research always begins with a goal. This dissertation argues that renewed attention to social science goals can point the way toward transparent assumptions and the use of predictive algorithms for estimation. Chapter 2 (with Rebecca Johnson and Brandon Stewart) presents a framework for quantitative research in which a precise goal---the theoretical estimand---guides all methodological choices. Our framework creates opportunities for advances that are both methodological and substantive. Chapter 3 presents a methodological advance---the gap-closing estimand---which quantifies how an intervention to a treatment (e.g. occupational class attainment) would close a gap (e.g. income by class origin). Chapter 4 applies that methodological advance to quantify the contribution of occupational segregation to a racial disparity in health. Chapter 5 concludes. Together, these chapters illustrate the fruitfulness of a framework for quantitative social science that unlocks computational tools to answer research questions that are conceptually precise. The key to this framework is clear reasoning about the research goal.