Three Essays on Causal Inference with Observational Data and the Sociology of Mental Health and Self-Harm

Abstract

The following essays focus on causal inference methods for answering questions related to the sociology of mental health and self-harm. They collectively help illustrate what I call an imperfectionist perspective on causal inference. On this view, we shift away from debating whether our causal estimators are biased and move toward carefully reasoning about the magnitude and direction of the bias. Chapter 2 (co-authored with Brandon Stewart) reviews instrumental variables (IV) analysis for causal inference and shows that an imperfectionist perspective calls into question the elevation of IV over selection-on-observables approaches. Chapter 3 applies an imperfectionist approach to studying suicide contagion with interrupted time series analysis. I provide evidence that fictional TV shows can produce imitation effects among certain subpopulations, showcasing a variety of tools to help us reason about the direction and magnitude of the bias in our treatment effect estimates. Chapter 4 analyzes a debate over the effects of digital technology use on adolescent well-being. This chapter demonstrates the limits of the imperfectionist approach: in some settings, multiple factors plausibly bias our estimates in different directions, making it difficult to reason about the bias and draw informative conclusions about the causal claims in question. By exploring this framework and its limits, this dissertation helps advance methodological practice and expand our knowledge about suicide and self-harm.