Quantifying the impact of protected areas on near-global waterbird population trends, a pre-analysis plan

There is increasing interest in the effectiveness of protected areas (PAs) for supporting populations of wildlife. While there are a number of association studies showing a relationship between protected areas and abundance or trends in wild species, studies with an appropriate counterfactual (what would have happened in the absence of protection) are rare. We use the world’s largest database on waterbird counts (covering 587 species at 21,989 sites globally) to answer three questions: 1) Do PAs have a positive impact on waterbird population trends relative to a counterfactual (this includes cases where a PA has lessened, but not halted, a population decline)?; 2) are PAs performing successfully by maintaining or increasing populations? and 3) what factors contribute to PA impact and performance? We selected 15,703 waterbird populations (here defined as a site species combination), consisting of 311 species at 870 protected sites, where PA designation occurred at least 5 years after the first survey date, and 5 years before the last. We will use this to compare trends before PA designation to those afterwards. We then matched these sites to unprotected sites with similar covariates in the years before PA designation, resulting in a matching dataset of 6,451 populations pairs consisting of 39 species at 769 pairs of protected and unprotected sites. We will use this to compare trends both before and after PA designation and inside and outside of PAs. Our results will shed light on the impact of PA on hundreds of waterbird species, providing much needed evidence regarding PA effectiveness. As PA performance is a sensitive subject and it is important to develop hypotheses before knowing the results (especially for the relatively complex data analysis used in matching protected and unprotected sites), we present a pre-analysis plan. This will ensure that the final paper’s analyses are hypotheses testing, rather than generating, and avoids the risk of, or perception of, data dredging.