Ancillary experiments' are a new technique whereby researchers use a completed experiment conducted by others to recover causal estimates of a randomized intervention on new outcomes. The method requires pairing new outcome data with randomized treatments the researchers themselves did not oversee. Since ancillary experiments rely on interventions that have already been undertaken, oftentimes by governments, they can provide a low-cost method with which to identify the effects of large-scale and possibly ethically difficult interventions. We define this technique, identify the small but growing universe of studies that employ ancillary experiments in political science and economics, and assess the benefits and limitations of the method.
experimental methods government performance ancillary experiments downstream experiments causal inference research design