Episode Summary
Executive Summary: Sean Carroll interviews Doan Farmer about complexity economics: why standard equilibrium-based models often miss endogenous dynamics, crises, and heterogeneity, and why agent-based simulations better capture real economies. Farmer explains how bounded rationality, feedback loops, specialization, and networked supply chains produce oscillations and crashes, and how these tools have already improved forecasts for COVID and market risk.
Main Topics: Why complexity economics differs from mainstream economics (Priority: 5/5): Farmer contrasts equilibrium-first, rational-expectations models with agent-based simulations built around interacting, boundedly rational agents and evolving information. Endogenous dynamics and disequilibrium (Priority: 5/5): He argues that many major economic phenomena, including business cycles and financial crises, arise from within the system rather than from external shocks, so static equilibrium models miss the mechanism. Chaos, oscillation, and unstable equilibria (Priority: 4/5): Farmer connects chaos theory to economics through feedback, overshooting, and unstable equilibria that create persistent fluctuations rather than smooth convergence. Agent-based modeling and bounded rationality (Priority: 5/5): The discussion explains how simple decision rules, heuristics, learning, and network interactions can be simulated at scale to generate realistic macro behavior. COVID modeling and policy relevance (Priority: 4/5): Farmer describes a disequilibrium model built for the UK government to estimate lockdown impacts using labor, input, and demand constraints. Market ecology, innovation, and regulation (Priority: 4/5): Markets are framed as ecosystems of specialized trading strategies; regulators could simulate new financial products like invasive species to test systemic risk. Specialization, trophic levels, and technological progress (Priority: 3/5): Using ecology analogies, Farmer argues that deeper supply chains can lead to faster innovation and cheaper products over time.
Key Arguments: Mainstream economics often assumes rational expectations and equilibrium, which works only in simple settings; complex real-world economies need bounded rationality, heterogeneity, and dynamics. Agent-based models let researchers simulate millions of interacting agents, allowing richer institutional detail and emergent outcomes that equations alone often cannot capture. The 2008 financial crisis was largely an endogenous disequilibrium event driven by new financial instruments and feedback effects, not a simple external shock. Business cycles and persistent fluctuations can arise naturally from feedback and imperfect policy correction, like an unstable pole being repeatedly over- and under-corrected. Chaos matters because irregular endogenous dynamics require sensitive dependence, oscillation, and nonlinear feedback; otherwise systems tend to settle into equilibrium. COVID was an exogenous shock, and a disequilibrium model that tracked labor, inputs, and demand could reproduce how disruptions propagated through supply chains over time. Market efficiency is incomplete: markets may be informationally efficient yet still allocate capital badly and become unstable when novel instruments or leverage are introduced. Ecological thinking helps economics by emphasizing specialization, interdependence, and the effects of long supply chains on innovation and costs. Simulations are not just practical tools; they can help derive better theory by isolating mechanisms through knockout and addition experiments.
Data Points: US economic growth since the Revolutionary War: ~2% per year on average - Farmer cites long-run U.S. growth as one reason the economy is not static, even without shocks. Financial crisis frequency in the U.S. before the Federal Reserve: about every 7 years - Used to argue that crises are historically recurrent and not anomalies. Agent scale in simulations: millions of agents - Complexity models can simulate far more actors than tractable equation-based approaches. Prediction horizon for trophic-level/price improvements: 14 years ahead - Farmer says trophic-level analysis predicted which products would become cheaper or improve over time. Central banks using a housing-model variant: about 6 to 8 central banks in Europe - Evidence of growing adoption of complexity-based tools in policy settings. Workforce modeling input: Bureau of Labor Statistics occupational data - Used in the COVID model to infer which occupations could continue working under lockdown. Mainstream model dimensionality concern: more than a dozen agents becomes unsolvable - Farmer criticizes standard optimization models for becoming intractable once systems grow nonlinear and large.
Pivotal Quotes: "the economy is more like a drunk driver on a mountain road" — Doan Farmer: Explaining why endogenous feedback and bounded rationality create swerving dynamics rather than smooth convergence. "once the oscillation, once you have more than two frequencies oscillating, you almost always get chaos" — Doan Farmer: Describing why irregular endogenous business cycles require chaotic dynamics. "it takes a model to beat a model" — Doan Farmer: On why complexity economics must prove itself through empirical success against mainstream alternatives.
Implications: If Farmer is right, policymakers should rely more on agent-based, disequilibrium simulations to stress-test markets, pandemics, and supply chains. Complexity economics may shift regulation, forecasting, and macro policy toward mechanism-rich, data-driven modeling.
About Sean Carroll MindScape
Ever wanted to know how music affects your brain, what quantum mechanics really is, or how black holes work? Do you wonder why you get emotional each time you see a certain movie, or how on earth video games are designed? Then you’ve come to the right place. Each week, Sean Carroll will host conversations with some of the most interesting thinkers in the world. From neuroscientists and engineers to authors and television producers, Sean and his guests talk about the biggest ideas in science, ...