Causality and Resilience
Abstract
Hundreds of empirical studies report on the factors that increase resilience to negative shocks in coupledhuman-nature systems. Yet, in a set of 500 empirical studies that investigate resilience to climate shocks, I find that most studies fail to clearly define their target causal effects (estimands) or explain how their design identifies these effects. In the studies that specify the target causal effect, many use an empirical design in which the effect is not identified. Thus, the empirical literature on resilience to climate shocks is largely uninterpretable. In this study, I start with the idealized experiment: a factorial experiment in which shocks and attributes that improve resilience to the shock are randomized across units from a target population (“units” like households or watersheds). For logistical and ethical reasons, we cannot run such an experiment. However, the idealized experimental design implies that resilience scholars who use observational designs, like experimentalists who use factorial designs, can estimate multiple causal effects, some of which are more policy and scientifically relevant than others. I define these causal effects, describe their utility in terms of common scientific and policy questions, describe how identification and estimation of each causal effect requires a different observational design, and assess the climate resilience literature in light of these findings. Without this information, scientists will struggle to develop appropriate empirical designs for investigating resilience and will find it challenging to evaluate the quality of published empirical studies on resilience. I conclude by describing a set of best practices for empirical studies of resilience in a range of literatures, including climate science, economics, ecology, cognitive science (brain resilience), political science, and sociology.