Modeling economic resilience.

Summary Major ecological and climatic transformations are currently underway. They are a source of environmental instability, as extreme climatic events have become more frequent, more intense, and affecting new regions of the globe. If we cannot prevent these changes, how can human societies adapt to them? For many researchers and decision-makers, resilience is the key to success. This concept seems to contain new solutions, adapted to a turbulent and uncertain world. By definition, resilient systems are able to bounce back from unexpected shocks, learn quickly and adapt to new conditions. Despite the interest in this notion, the processes that enable a society to be resilient remain poorly understood. This thesis develops a new conceptual framework that allows, through mathematical modeling, to explore the theoretical links between economic mechanisms and resilience. This framework is based on a critical analysis of resilience in ecology - the original domain of the concept - and in economics - our field of application. We apply it to economic production systems, modeled as networks of firms and analyzed through dynamic systems theory. This thesis evaluates the ability of such multi-agent models to generate bifurcation profiles, an essential step in the mathematical analysis of resilience. We study a very general prey-predator dynamic in ecology and economics. Second, this thesis addresses a major factor that hinders resilience: the strong interdependencies between economic activities, through which production delays and interruptions propagate from one firm to another. Using realistic production networks, we show how supply delays, when embedded in particular topologies, multiply these propagation phenomena. Then, thanks to an evolutionary model, we highlight the existence of a systemic risk: cascades of incidents occur even though all agents have inventories adapted to the risk level. This phenomenon is amplified when supply chains become specialized and fragmented. These theoretical results are of general value, and can be used to guide future empirical research. This thesis also advances knowledge on very recent mathematical methods and objects, such as Boolean delay equations forming a complex network, and evolutionary dynamics on graphs. The proposed models and conceptual framework open new research perspectives on resilience, in particular on the impact of environmental feedbacks on the structural evolution of production networks.