Placebo Tests for Causal Inference

Andrew C. Eggers,Guadalupe Tuñón,Allan Dafoe
DOI: https://doi.org/10.1111/ajps.12818
IF: 4.503
2023-08-19
American Journal of Political Science
Abstract:Placebo tests are increasingly common in applied social science research, but the methodological literature has not previously offered a comprehensive account of what we learn from them. We define placebo tests as tools for assessing the plausibility of the assumptions underlying a research design relative to some departure from those assumptions. We offer a typology of tests defined by the aspect of the research design that is altered to produce it (outcome, treatment, or population) and the type of assumption that is tested (bias assumptions or distributional assumptions). Our formal framework clarifies the extra assumptions necessary for informative placebo tests; these assumptions can be strong, and in some cases similar assumptions would justify a different procedure allowing the researcher to relax the research design's assumptions rather than test them. Properly designed and interpreted, placebo tests can be an important device for assessing the credibility of empirical research designs.
political science
What problem does this paper attempt to address?