Abstract
In this thesis, I explore approaches to modeling energy systems as complex adaptive systems using agent-based models. I focus on the oil market as a prototype for applying those approaches. In the first chapter “Why scenario models fail to model the cyclicality of the oil market?” I empirically make the argument that the process of oil extraction is characterized by the presence of significant frictions, heterogeneity, and feedback loops. Oil projects have long lead-times and geologically constrained production profiles, while oil producers have different resource access and behavioral profiles. Scenario models of the oil market often generate linear outlooks that fail to foresee the possibility of significant “tipping point” dynamics in prices, even though those “boom-and-bust” have been a regular market characteristic since OPEC was formed in 1973. I review those models and discuss why they fail to model the oil market as a continuously evolving system with changing structural features and behaviors.In the second chapter “Interpretable Economics: Using Fuzzy Logic to Model Agent Behavior in Energy Markets” I focus on exploring the use of Fuzzy Logic as a more interpretable method of modeling agent behavior, and constructing a model of energy markets with linguistic rules to describe agent behavior. I prototype a stylized model of the oil market to simulate the production and investment decisions of producers and the price negotiation process between producers and consumers. I show that the interaction between agents can lead to a realistic pattern of endogenous cyclicality due to field development lags and that the price negotiation mechanism can successfully clear the markets.In the third chapter “Imperfect OPEC: an Oil Market Model” I construct a model of the oil market with heterogeneous, rationally-bounded agents. I incorporate field development lags and empirically estimate reserve creation as a function of investment. In the model, agents learn investment and production decisions using a genetic algorithm to optimize their value functions. I find that imperfect cartelization arises naturally due to the different characteristics between- and within-OPEC and non-OPEC agents. Boom and bust patterns are endogenous to the market as producers respond imperfectly to shocks. Finally, I show that the scenarios arising from my model exhibit more stationary price paths than the energy scenario models used by energy agencies, and that cyclicality in price is a feature of the market even when averaging across different simulated shock paths. This thesis aims to contribute to the field of natural resources and environmental economics by embedding the modeling methodology firmly within a complex adaptive systems framework. The typical features that make a system complex – path-dependence, adaptive behavior, heterogeneity, and feedback effects – exist in almost all social and economic systems. The policy motivation for this thesis is that when it comes to studying energy systems, climate change, and structural transitions need to be modeled as dynamic processes spanning decades in time and multiple intertwined physical and social systems. The implications of our policy choices will span that far out as well, and We need to establish a framework for examining them through that lens.