Agent-Based Modeling of Resilience in Smallholder Agriculture: Toward Robust Models and Equitable Outcomes

Abstract

Smallholder farmers constitute one of the world's most vulnerable populations. Moreover, rising socioeconomic inequalities and biophysical degradation threaten to increase this vulnerability. There is therefore a pressing need to build resilience in smallholder agriculture. Socio-environmental systems (SES) modeling can support this goal, yet confronts two challenges that may limit its usefulness for informing agricultural development. First, as agricultural systems are highly heterogeneous and our ability to model them is imperfect, there is a risk that model-based recommendations inadvertently increase vulnerability. Second, there exist a range of approaches to agricultural development that prioritize distinct objectives (e.g., market integration versus social equity), and conflicts between these approaches could undermine progress toward more resilient futures. To build smallholder resilience therefore requires an integrated perspective on development as well as robust methodologies for comparing and integrating alternative development strategies. This dissertation uses agent-based modeling (ABM) to help address these challenges.

The first contribution of this dissertation is a set of methodological advances that improve the robustness of model-based policy analysis. These advances question two analytical norms within SES modeling. The first is a lack of attention to equity. For instance, by disregarding heterogeneity in outcomes, model-based recommendations may benefit the well-off at the expense of the vulnerable and thereby perpetuate inequity. Chapters two and three address this issue, first by establishing a conceptual framework for the equity-ABM interface and then by applying an agent-based model to examine equity in the effects of resilience-enhancing strategies. The second analytical norm that this dissertation questions is the use of a single, “best-fit” model to assess policy effects; due to our incomplete understanding of complex SES, multiple plausible models may exist. This common condition is known as equifinality, but it is not often considered in SES modeling or policy analysis. To attend to this challenge, chapter four develops an approach for identifying a set of diverse model calibrations and using these to achieve a more robust policy analysis. Together, these methodological advances facilitate more robust and equitable policy assessments, in agricultural systems and beyond.

The second principal contribution of this dissertation is substantive. Emerging from the modeling of smallholder resilience, I find complementarity between disparate agricultural development approaches. For instance, chapter five compares the effects of legume cover cropping (a form of ecological farm management) and microinsurance (a financial institutional support) on smallholder climate resilience. Although these approaches are traditionally promoted by distinct academic communities and development organizations, the results show that, when implemented together, they are highly complementary. Next, chapter six investigates the potential for contract farming to overcome the negative effects of large-scale land acquisitions on smallholder food security. Results suggest that preserving smallholder autonomy through contract farming can simultaneously improve smallholder food security and agricultural production, thereby better aligning the preferences of developers and smallholders. Thus, these chapters together suggest the benefits of reconciling perspectives on and approaches to agricultural development.

As a whole, this dissertation advances the application of agent-based modeling and resilience thinking in smallholder agriculture. Beyond agricultural applications, it lays the groundwork for identifying robust and equitable development strategies in SES.