What questions about conservation have causal answers? Highlighting the problem of multiple treatment versions.

Recent years have seen increasing attention to problems of causal inference in conservation science research. However, many of the laudable advances in this area have focused on the estimation stage of the research design process: discouraging the use of estimation strategies that cannot credibly estimate causal effects, or encouraging the use of follow-up analyses to determine how sensitive correlational findings may be to unobserved confounding. In this study, I highlight a problem that has received relatively less attention, but which is no less important: carefully defining the “estimand,” the causal question a researcher wants to answer. Inattention to this issue is especially problematic when units can self-select into different “versions” of the treatment they may receive. I review select examples of this issue in the conservation science literature, with a focus on how it might interfere with the accumulation of conservation policy evidence, or how it might drive misleading criticisms of correlational studies. I end this study with a set of practical lessons for applied researchers.