Box 2.

Study 2—A set of dynamical system models to assess the effectiveness of poverty alleviation strategies in different social–ecological contexts

The goal of the study is to enhance understanding of how structural conditions, in particular the relations between poverty and state of the environment in rural contexts, influence the effectiveness of poverty alleviation strategies (Lade et al. 2017). It is an example of causal reasoning that employs a set of dynamical systems models to explore different what-if scenarios. The authors review commonly observed or assumed social–ecological relationships in local, agricultural, developing world contexts. They use this empirical knowledge and economic theory to develop a set of simple models of multi-dimensional poverty traps that incorporate different poverty–environment relationships, e.g., agricultural intensification that degrades the environment versus agricultural practices that maintain agroecological diversity. To assess the impact of these different contextual settings on the effectiveness of poverty alleviation strategies, they design a study that builds on the manipulability account of causation to assess the effects of interventions on poverty in different modeled contexts. The interventions, such as an asset input or a transformation in which farmers combine conventional agricultural production with traditional agricultural practices, are modeled through changes in initial conditions or structural changes in the model, respectively, and their effect on the emergence of non-poor system states is analyzed The authors build on economic theory that describes poverty traps systems with two equilibria, a poor and a well-being one, that are separated by a threshold (Barrett and Constas 2014). They pay much attention to extending this simple causal model by specifying different possible relations between financial, natural and cultural capital that are represented in the different model structures. Finally, they use resilience theory to inform the design of different types of interventions. Model results reveal that interventions such as asset inputs that ignore relations between agricultural production, nature and culture can, in some contexts, reinforce poverty. In such contexts, interventions that enable development while avoiding environmental and cultural degradation can help overcome traps. The causal claims are justified through systematically studying the connectivity between different factors and through contrasting the different models and interventions. In addition to manipulability, the study builds on the mechanism-based account of causation. Mechanism-based thinking directs attention to contextual factors, in this case the relations between agricultural practices and the environment, that influence what effect a particular cause has. It thus helps understand how contextual conditions may influence intervention effects, something that RCTs often do not explicitly take into account (Rodgers et al. 2020). However, the stylized nature of the models does not allow to quantify effects nor to apply them directly to a given case