A0498
Title: Evaluating the validity and robustness of instrumental-variable analyses
Authors: Dean Knox - UPenn Wharton (United States) [presenting]
Luke Keele - University of Pennsylvania (United States)
Guilherme Duarte - University of Pennsylvania (United States)
Kai Cooper - University of Pennsylvania (United States)
Jonathan Mummolo - Princeton University (United States)
Kennedy Mattes - Princeton University (United States)
Abstract: Instrumental-variable (IV) designs are widely used across numerous fields to estimate causal effects when the relationship between treatment and outcome is confounded, exploiting as-if randomized encouragements that nudge units into treatment. The validity of these designs rests on several assumptions that are often regarded as difficult to test, including monotonicity, the assumption that no units defy encouragement and exclusion, and the assumption that the instrument does not directly affect the outcome. Using newly derived analytic results and recent advances in automated partial identification, a range of falsification tests and sensitivity analyses are presented that empirically evaluate the validity of these assumptions and the robustness of inferences to their violations. Replicating and extending published examples, it is shown how the techniques are flexible to the idiosyncrasies of applied settings by sharply bounding causal estimands in situations where key IV assumptions are believed to fail.
