The Santa Fe Institute has played a major role in developing the emerging field of Complexity Economics, and has recently received a large grant to explore its connections to political economy.
The approach we refer to as Complexity Economics directly and more realistically models the behavior of boundedly rational agents and the institutions and networks that shape and constrain those behaviors. This school of thought, supported by many psychologists, assumes that agents make decisions using a combination of heuristics, learning algorithms, and imperfect reasoning. Whereas most economic models are based on assumption of equilibrium and utility maximization, complexity economics is based on bounded rationality (a phrase introduced by Herb Simon in the 60’s). Complexity economics models can take many forms, including dynamical systems, statistical mechanics, and agent-based modeling. The agent-based models employed by Complexity Economics can be simple or complicated, but directly simulate the decisions and interactions of agents and model them as dynamical systems which may or may not reach an equilibrium.
When the dust settles, perhaps the most important advantage of complexity economics models is that they are much more tractable in complicated circumstances. Problems such as climate change involve heterogeneous agents, complicated institutional constraints and competing economic forces. These all operate in tandem in a future that is fundamentally uncertain. Models based on optimizing agents quickly become intractable in such circumstances. This limits the institutional realism of such models. Agent-based models, in contrast, do not assume optimal behaviors and remain tractable in much more complicated set ups. Agent-based models can therefore incorporate a much higher degree of realism, which is critical in many policy contexts. Simpler dynamical models and statistical mechanical models can also be used to derive results in simpler contexts.
Another important advantage of Complexity Economics is that it is more appropriate for situations with uncertainty. There are two senses in which the future can be unknown. Risk refers to situations where we do not know the future, but where the possible outcomes and their probabilities are known. Uncertainty, in contrast, corresponds to situations where outcomes and/or their probabilities are unknown. Mainstream economics models are by their very nature designed to deal with risk. In contrast, as many psychologists have argued, the direct models that we use to reason about the future, such as heuristics, have evolved to deal with conditions of uncertainty. Real world problems occur in circumstances of uncertainty rather than risk. Three concrete examples are: 1) the 2008 global financial crisis, 2) the policy-driven increases in economic inequality from the 1970s-2010s, and 3) the slow response to climate change.
There are several reasons for believing that complexity economics can play a key role in political economy.
First, as its name implies, political economy lives at the intersection of economics and politics, where “politics” more generally refers to issues of power. How does economic activity confer power, how is this power used, and what are the consequences of economic policies regarding the allocation of power? Complexity economics and agent-based models in particular offer a powerful method to study the interactions between economic outcomes and power. In agent-based models it becomes much easier to create models that incorporate more realistic policies and study their interaction with economic and political outcomes.
Second, inequality plays an important role in political economy. While understanding that inequality is a hot topic in contemporary macroeconomics, the methods for incorporating it into canonical models are limited. In contrast, in an agent-based model this is easily done by creating a heterogeneous synthetic population whose demographic characteristics match those of a real population. It then becomes possible to use techniques developed in microsimulation to study how policies affect synthetic populations. Unlike microsimulation models, however, agent-based models incorporate feedback between consumption and production, so that it becomes possible to study how policies will affect the economy both at the macrolevel and at an individual level.
Third, with the rise of the digital economy, increasing returns to scale have become important, giving rise to companies with strong market power such as Amazon and Google, giving such organizations enormous political power. Agent-based models have a good potential for understanding increasing returns in a quantitative manner.
Complexity economics has begun to show how it can be used to explore all of the four topics mentioned above. Over the past decades, Complexity Economics has generated insightful contributions to policy questions and demonstrated progress toward better quantitative models in a variety of topics, including financial stability and systemic risk, climate change and the transition to net-zero, drivers of economic growth, business cycles and inequality, political polarization, technological change and its potential impacts on the labor market. Agent-based models are now competing head-to-head with standard macro (DSGE) models and roughly match their empirical performance but promise much improved performance as more data is incorporated and refinements in calibration are made. Several models of production networks have been developed, ranging from “agentized”, behavioral versions of standard models, to full-fledged ABMs calibrated on administrative datasets. A dynamic disequilibrium model produced accurate real time forecasts of the impact of the COVID pandemic on the United Kingdom. A recent agent-based model of the Hungarian housing market realistically represents all four million Hungarian households, incorporating information about the characteristics of every house in Hungary. The time is ripe to build on these successes to address core questions in political economy.
Complexity Economics has the potential to influence other fields of social science. The core ideas of economics have substantially influenced other branches of social science, such as law and political science. Complexity economics potentially offers a more tractable and empirically valid alternative that could dramatically improve our understanding of political economy.
The proposed SFI working group will focus on recent successes in addressing the challenges posed above, and discuss how we can build on this success to accelerate progress in the future. Outputs will include a special issue in the Journal of Economic Behavior and Organization, as well as a set of essays to be collected in a volume that will become “The economy as an evolving complex system, Part IV”.