Katso artikkeli

We fully endorse Rasolofoson’s (2022) main point in his comment on our essay “Statistical matching for conservation science”(Schleicher et al., 2020): scientists and practitioners using observational data for conservation impact evaluation need to pay careful attention to the process by which some units came to be exposed to the intervention and others did not (ie,“the treatment assignment mechanism”). We also appreciate his excellent illustrative examples of this main point. We build on his argument by highlighting that the treatment assignment mechanism is rarely known with certainty and that additional analyses are needed to quantify the potential effect of this uncertainty on conclusions.

Conservation science as a discipline has been somewhat slow to embrace robust impact evaluations to advance effective policy and practice (Baylis et al., 2016). However, a shift is now underway. There has been an explosion of recent studies explicitly designed to quantify the effect of a conservation intervention by comparing outcomes with unobservable counterfactual outcomes (what would have happened in the absence of the intervention)(Ferraro, 2009). There are a range of methods for estimating these counterfactual outcomes when random assignment is not possible (Schleicher et al., 2020), of which statistical matching-the focus of our original essay–is the most common (Börner et al., 2020).