# Making Sense of Chaos **J. Doyne Farmer**  --- _The economy doesn't converge to equilibrium. It never stops moving._ Standard economics assumes rational agents optimising in equilibrium. The economy changes only when external shocks disturb it. But real economies exhibit endogenous motion: they never settle, even without outside disturbances. Real agents aren't rational optimisers but boundedly rational creatures using simple heuristics to navigate the [[Unknown and unknowable]]. The [[Variance]] that standard models treat as noise is often the signal itself. Business cycles, financial crises, technological revolutions: these phenomena emerge from the system's internal dynamics, not from external shocks. If your model can only generate motion by being kicked from outside, it cannot explain the thing you most want to understand. --- **Complicated and complex are not synonyms, and the difference explains why equilibrium models fail.** Complicated means many moving parts. Complex means emergent behaviour: properties that arise from interaction that no individual component exhibits. The Lorenz equations have only three variables but produce geometric objects of extraordinary intricacy. Three variables, infinite unpredictability. And standard macroeconomic models? They have only fixed-point attractors. Left alone, the model economy freezes and stays frozen. All motion requires exogenous shocks. This rules out endogenous business cycles from the outset. That inability is structural, not a matter of calibration. Chaos has two essential properties: sensitive dependence on initial conditions, and endogenous motion. Even without external shocks, a chaotic system never settles to rest. Standard macro models cannot generate cycles on their own. They can only predict how the economy relaxes back to equilibrium after a disturbance. If the economy is genuinely complex, and there is good reason to believe it is, then studying it with equilibrium tools is studying turbulence with a model that predicts still water. --- **The economy is an ecosystem of specialists, and the supply chain finding is the sharpest result in the book.** In biology, each species has its own strategy for extracting energy from the environment. Species eat each other, compete, cooperate, and collectively alter the environment in ways that feed back to every participant. To understand grass, you must think about lions, because lions control zebras, and zebras eat grass. Production networks work the same way. Roughly 65% of improvements in any given industry come from other industries, not from within the industry itself. Industries with deeper supply chains, higher trophic levels, improve faster because they have more innovating industries below them compounding their capabilities. Manufacturing improves faster than services for a structural reason: manufacturers sit higher in the supply chain and absorb compounding innovation from below. This is the structural logic behind [[Niches]]: narrow focus amplifies improvement because it concentrates supply chain benefits. Specialisation doesn't just make the economy function. It automatically accelerates the rate of change. --- **Heuristics are not second-best.** Herbert Simon introduced [[Bounded Rationality]] in 1955. The implication isn't that people are bad at decisions. In genuinely complex environments, optimisation is often impossible. The search for good-enough choices is the rational response, not a cognitive failure. The less-is-more effect is real: simple heuristics frequently outperform complicated methods, precisely because complicated methods overfit to data that isn't representative of future conditions. A model that fits the past perfectly by adding enough parameters is useless for forecasting. The test is predictive power, not explanatory elegance. Agent-based modelling takes this seriously. Instead of equations derived from rational optimisation, the models simulate heterogeneous agents using varied, imperfect heuristics, and observe what emerges. Economic forecasting lags weather forecasting by decades, and Farmer thinks the gap is structural. Weather models are built from the bottom up at the finest possible scale, run on nonlinear equations, and cover the entire world as a closed system. Economic models operate at aggregate level with representative agents, use linear approximations, and treat outside influences as exogenous shocks. Better models require better data, bottom-up construction, and accepting that the whole is genuinely different from the sum of its parts. ---