Making Sense of Chaos (with Doyne Farmer)

Episode Summary

Executive Summary: The episode argues that economics should model real-world complexity, uncertainty, and disequilibrium rather than rely on simplified equilibrium assumptions. Doan Farmer explains how agent-based, bottom-up simulations can better predict shocks like COVID-19, improve policy design on inequality and climate, and eventually replace mainstream models as data and computing power improve.

Main Topics: Complexity economics vs. orthodox economics (Priority: 5/5): Farmer contrasts mainstream equilibrium models—built on utility maximization and fixed assumptions—with complexity economics, which simulates people, firms, and institutions making decisions over time under changing conditions. COVID-19 as a test case for agent-based modeling (Priority: 5/5): The discussion highlights Farmer’s pandemic model, which tracked labor, demand, and input shocks day by day and closely matched actual GDP, unemployment, and industry outcomes. Risk, uncertainty, and heuristics (Priority: 4/5): Farmer distinguishes measurable risk from true uncertainty and argues that simple heuristics can outperform fancy optimization when probabilities and outcomes are unknown or unstable. Policy relevance for inequality and wages (Priority: 4/5): The hosts connect complexity economics to middle-out policy, especially the idea that higher wages and stronger demand can improve growth rather than destroy jobs. Data, computation, and model building (Priority: 4/5): Farmer says modern computing and richer transactional data now make large-scale simulation feasible, including synthetic populations and global supply networks. Climate transition modeling (Priority: 4/5): The conversation extends complexity methods to climate policy, arguing for testable, nudge-based strategies and predicting rapid growth in solar and wind deployment. Future of the economics profession (Priority: 3/5): Farmer expects complexity models to spread first through private industry and policy institutions, eventually pressuring academia to adapt.

Key Arguments: Mainstream economics relies on an unrealistic as-if framework that assumes utility-maximizing agents and equilibrium, which limits realism and makes complex problems hard to solve. Complexity economics models the economy as a dynamic system of interacting agents, allowing feedback loops, shocks, and emergent outcomes to be studied directly. Modern computing now makes it feasible to simulate entire economies at scale; what was impractical decades ago is increasingly tractable today. The COVID model worked because it captured disequilibrium: labor shortages, demand collapse, input disruptions, and their propagation through supply networks. In uncertainty-heavy environments, simple heuristics can outperform optimization because they avoid false precision about unknown probabilities. Policy debates such as the minimum wage are distorted by oversimplified models; complexity approaches suggest wage gains can raise demand and improve economy-wide outcomes. Better data—especially transaction-level supply network data—would dramatically improve economic modeling and policy analysis. Climate policy should focus on realistic adoption pathways and measured experimentation, not idealized optimal-path models that assume a world dictator. A faster clean-energy transition can be economically beneficial, not just costly, because falling technology prices and higher deployment reduce overall expenses. The economics field will change as complexity models prove their predictive value, especially if they outperform mainstream models across countries and policy domains.

Data Points: Years of confidence in complexity economics: 30 years - Farmer says his confidence in applying complexity theory to economics has built over the last three decades. COVID model GDP prediction: 21.5% predicted decline - Farmer’s team predicted the UK’s second-quarter 2020 GDP contraction. Actual UK Q2 2020 GDP decline: 22.1% - The transcript compares the model’s forecast to the realized economic hit. Pandemic prediction horizon: 1 year - Farmer says the team explicitly limited its COVID forecasts to a one-year window. Simulated economy scale: 8 billion people, 3.5 billion households, 200 million firms - Farmer says current computing power could support a global-scale simulation at this magnitude. Model testing countries: 38 countries - Farmer says his current model has been run across 38 countries. Amount of investment compared with mainstream economics: 1/1000th - He estimates the complexity approach has received roughly one-thousandth the cumulative investment of mainstream economics. Other institutions pursuing similar work: EASA, Bank of Canada, Bank of Hungary, Bank of Italy - Farmer cites a growing international set of teams developing comparable models. Solar and wind deployment outlook: 5 to 10 years of exponential growth - He predicts continued rapid growth before deployment begins to taper off. Climate target reference: 2-degree target - Farmer discusses policy modeling around limiting warming to roughly two degrees. Benevolent dictator climate timeline: 5 years - Farmer says a world dictator could put society on the right climate path quickly if he were in charge.

Pivotal Quotes: "The more the middle class thrives, the better the economy is for everyone, even rich people like me." — Nick Hanauer: Opening framing for the podcast’s middle-out economics perspective. "We can simulate the whole world." — Doan Farmer: Farmer describing the scale now possible with modern computational power. "The problem with bad models is that they change the economy with every forecast because it is a complex adaptive system." — Nick Hanauer: Closing reflection on how forecasts influence policy and behavior.

Implications: The episode suggests economics is moving toward more realistic, testable models that can improve wage, climate, and crisis policy. For listeners, it reinforces that better data and complexity thinking could produce win-win economic outcomes.