Handle with Care: A Sociologist's Guide to Causal Inference with Instrumental Variables.

This article provides a comprehensive review of instrumental variables (IV) analysis as a method for drawing causal inferences from observational data in sociology, emphasizing its assumptions, fragility, and practical challenges. It outlines key identification assumptions—relevance, unconfoundedness, exclusion restriction, monotonicity, stable-unit-treatment-value assumption (SUTVA), and positivity—and highlights that these are often unstated or misunderstood in sociological research. The authors identify three main methodological problems exacerbated by weak instruments: identification bias (asymptotic bias from assumption violations), estimation bias (finite-sample bias toward ordinary least squares estimates), and type-M error (exaggeration of effect sizes when statistically significant). Drawing on a survey of 34 IV studies in leading sociology journals, the article finds that robust diagnostics and bias analyses are rarely reported, and IV estimates frequently exceed those from selection-on-observables approaches. To address these issues, the authors propose a detailed checklist for conducting and reporting IV analyses, advocating clearer communication of assumptions, routine use of weak-instrument-robust confidence intervals, and sensitivity analyses to assess potential biases.