The Economy on a Computer
From a single artificial stock to entire economies: how the institutions that regulate the world’s financial systems are abandoning the equilibrium framework, and what that means for the strategies built on its failure
On February 12, 2026, the Santa Fe Institute Press published The Economy as an Evolving Complex System IV. Two volumes. Over a thousand pages. Edited by R. Maria del Rio-Chanona, Marco Pangallo, Jenna Bednar, Eric Beinhocker, Jagoda Kaszowska-Mojsa, François Lafond, Penny Mealy, Anton Pichler, and J. Doyne Farmer.
To understand the significance of this publication, you have to understand the series it belongs to.
In September 1987, the then-nascent Santa Fe Institute held a conference to look at the economy as an evolving complex system. It was organised by the physicist Philip Anderson and the economist Kenneth Arrow. The proceedings were published as The Economy as an Evolving Complex System in 1988. It was SFI’s first research programme. Brian Arthur headed it.
Volume II came in 1997, the same year the artificial stock market paper was published. Volume III came in 2005. Each volume marked a phase transition in the field: from provocation to demonstration to maturity.
Volume IV is the phase transition that matters most for anyone who manages money.
Because Volume IV is not about whether complexity economics could work. It is about where it is already working. In central banks. In regulatory agencies. In the institutions that set the rules for the global financial system. The question is no longer theoretical. It is operational.
The shift has been building for years, but it accelerated after 2008.
The financial crisis exposed the limitations of the equilibrium models that central banks had relied upon. The Dynamic Stochastic General Equilibrium models (known as DSGE) that formed the backbone of monetary policy analysis assumed a representative agent, rational expectations, and market clearing. They could not produce financial crises. They could not model contagion. They could not explain how a housing market collapse in the United States could cascade through the global banking system in a matter of weeks.
Andrew Haldane, then the Bank of England’s Executive Director for Financial Stability, was one of the first senior central bankers to say so publicly. In a series of remarkable speeches between 2009 and 2012, Haldane argued that the financial system should be understood as a complex adaptive network, not a system in equilibrium. He drew on ecology, epidemiology, and engineering to describe the structural vulnerabilities that had built up in the decade before the crash. The financial system, he argued, had become a robust-yet-fragile network: highly connected, apparently stable, but susceptible to catastrophic failure when a key node came under stress.
Haldane was not a complexity theorist. He was a central banker with responsibility for the stability of the British financial system. And he was saying, in effect, that the models his institution relied upon were structurally incapable of seeing the risks they were supposed to manage.
Haldane and his co-author Arthur Turrell published a paper in 2019 in the Journal of Evolutionary Economics that laid out the case for agent-based models as a complementary tool for macroeconomic analysis. The argument was precise: macroeconomic modelling had been under intense scrutiny since the crisis, with serious shortcomings exposed in the dominant methodology. Agent-based models, which had been extensively used across other disciplines, were particularly well-suited to answering questions where complexity, heterogeneity, networks, and adaptive behaviour played important roles.
This was not a fringe proposal. This was the chief economist of the Bank of England arguing that the central banking toolkit needed to be rebuilt using complexity science.
In 2020, Farmer and his colleagues at Oxford’s Institute for New Economic Thinking published a Bank of England working paper titled “Foundations of System-Wide Financial Stress Testing with Heterogeneous Institutions.” The paper laid out a framework for stress-testing the financial system using agent-based models that could capture the feedback loops, cascading failures, and emergent behaviours that DSGE models could not.
In 2025, the Bank of England published a comprehensive working paper surveying the use of agent-based models across central banks globally. The findings were striking: ABMs had emerged as effective complementary tools for central banks in carrying out their mandates. The paper documented their use across three categories: applied research connected to central bank mandates, technical research advancing the methodology, and direct integration into policy work.
The European Central Bank has explored agent-based approaches for macroeconomic forecasting. The US Office of Financial Research has used them for systemic risk assessment. Research groups at central banks from Hungary to Japan have developed their own implementations. The conversation has shifted from “could this work?” to “how do we deploy it?”
An agent-based model of an economy works on the same principle as the Santa Fe Artificial Stock Market, but at vastly greater scale.
Instead of a hundred agents trading a single stock, picture millions of agents: households that consume and save, firms that produce and invest, banks that lend and borrow, investors that allocate across asset classes, regulators that set capital requirements, and central banks that set interest rates. Each agent follows rules that are simple individually but interact in ways that produce complex aggregate behaviour. Households adjust their spending based on income, wealth, and expectations. Firms adjust their production based on demand and inventory. Banks adjust their lending based on capital ratios and perceived risk.
None of these agents has perfect information. None of them optimises in the way that standard economics assumes. They learn from experience. They follow heuristics. They adapt when their strategies stop working. They are, in Arthur’s language, boundedly rational, inductively probing their environment.
The models do not assume that the economy tends towards equilibrium. They do not assume that markets clear. They do not assume that a representative agent can stand in for the entire distribution of heterogeneous actors. They simply define the agents, their rules of interaction, and the institutional environment, and then let the simulation run.
What emerges is an economy that looks startlingly like the real one. Business cycles arise endogenously. Credit booms build and collapse. Asset price bubbles inflate and burst. Inequality grows through path-dependent dynamics. Cascading failures propagate through banking networks. All of these phenomena, which equilibrium models either cannot produce or must import as external shocks, arise spontaneously from the interaction of adaptive agents.
I want to be clear about what is happening, because it matters.
The institutions that regulate the global financial system are, slowly but decisively, supplementing their equilibrium-based analytical frameworks with complexity-based tools. They are not doing this because complexity economics is fashionable. They are doing it because the equilibrium models failed catastrophically in 2008, continued to miss the dynamics of the recovery, and remain unable to capture the feedback loops that generate systemic risk.
The DSGE models are not being abandoned. They remain useful for certain narrow questions about monetary policy transmission. But for the questions that matter most (how does stress propagate through the financial system? what happens when a large institution fails? how do leverage cycles build and unwind? where are the vulnerabilities that could trigger the next crisis?), the equilibrium framework has been found structurally wanting.
The replacement is not another equilibrium model with better parameters. The replacement is a fundamentally different way of thinking about what an economy is. Not a system that tends towards a stable resting point, but a complex adaptive system that is perpetually out of equilibrium, perpetually generating novelty, perpetually producing outcomes that no individual within it anticipated.
This is not a prediction about the future. This is a description of what is already happening.
Consider the implications for what we do.
For fifty years, the theoretical justification for the efficient market hypothesis has been the equilibrium framework. If markets are in equilibrium, then prices reflect all available information. If prices reflect all available information, then price movements are random. If price movements are random, then trend following cannot work, because there are no persistent patterns to exploit. The logical chain is tight. And for fifty years, it has been the intellectual foundation for dismissing trend following as either luck, data-mining, or compensation for hidden risk.
Now the institutions that regulate the world’s financial markets are abandoning the premise.
If markets are not in equilibrium (and the agent-based models show they are not), then prices do not reflect all available information. If prices do not reflect all available information, then price movements are not random. If price movements are not random, then persistent patterns can exist. Trends can exist. Momentum can exist. Fat tails can exist. Clustered volatility can exist. All of the empirical features that trend followers have been exploiting for decades can exist, because the theoretical framework that said they could not is being replaced by one that says they must.
The artificial stock market showed this for a single stock. The economy-scale agent-based models show it for entire financial systems. The dynamics are the same. The mechanism is the same. Adaptive agents form predictions, the predictions crowd, the crowding generates instability, the instability produces directional moves, and the moves persist long enough for a responsive strategy to detect and follow them.
This is not a theoretical argument any more. It is the operating assumption of the Bank of England’s financial stability division.
Farmer’s Complexity Economics programme at Oxford’s Institute for New Economic Thinking sits at the centre of this shift. The programme, which Farmer directs as Baillie Gifford Professor of Complex Systems Science, brings together physicists, mathematicians, computer scientists, and economists to build the next generation of models.
The ambition is sweeping. Farmer and his team are building models that simulate entire economies at the level of individual agents: millions of households, thousands of firms, hundreds of banks, and the regulatory apparatus that connects them. The models are calibrated to real data. They produce quantitative predictions, not just qualitative stories. And they are being used to answer policy questions that equilibrium models cannot touch: what happens if interest rates rise sharply while household leverage is elevated? How does a supply chain disruption propagate through an interconnected production network? What are the systemic consequences of a rapid transition to clean energy?
During the COVID-19 pandemic, Farmer’s team used agent-based models to forecast the economic impact of lockdowns by modelling the supply and demand shocks at the level of individual industries and occupations. The work, published in the Oxford Review of Economic Policy, demonstrated that agent-based models could produce actionable forecasts in real time, under conditions of radical uncertainty, where equilibrium models had nothing to offer.
This is the trajectory. From Brian Arthur’s thought experiments in the 1980s, to the artificial stock market in the 1990s, to Farmer’s ecology of strategies in the 2000s, to economy-scale models used by central banks in the 2020s. The ideas that were dismissed as fringe when Arthur first proposed them are now embedded in the analytical infrastructure of the institutions that govern global finance.
In Complex Adaptive Markets (forthcoming 2026), I describe the economy itself as a living system. Not as an analogy. As a description.
A living system is one that maintains itself far from equilibrium through the continuous exchange of energy and information with its environment. It adapts. It evolves. It produces emergent order that no individual component designs or controls. It is resilient to small perturbations and catastrophically vulnerable to large ones. It exhibits cycles, feedback loops, path dependence, and phase transitions.
This is what the agent-based models show. This is what the EECS IV volumes document. This is what the central banks are beginning to accept. The economy is not a machine that can be tuned to an optimal setting and left to run. It is a living system that is perpetually constructing itself, perpetually out of equilibrium, perpetually generating the dynamics that our programme harvests.
In Carved by Impossibility (forthcoming 2026), I argue that this understanding changes everything about how portfolios should be built. If the economy is a living system, then the robust portfolio is not the one optimised for a predicted future. It is the one that survives whatever the living system produces. It is carved by impossibility: structured not around what you expect to happen, but around the elimination of everything that would destroy you if you were wrong.
And in the Trend Following Manifesto with Niels Kaastrup-Larsen (forthcoming 2026), we describe the practical expression of this insight: the diversified systematic programme that treats every market as a living ecology, that deploys responsive sensors across dozens of these ecologies simultaneously, and that harvests the dynamics that the ecology produces without needing to predict which dynamics will appear or when.
Let me state plainly what has happened.
In 1987, when Arthur and Anderson and Arrow gathered at the Santa Fe Institute, the idea that the economy should be studied as a complex adaptive system was so heterodox that mainstream economists dismissed it as physics envy. In 1997, when the artificial stock market produced fat tails and bubbles from agents with simple rules, it was treated as a curiosity. In 2008, when the models that assumed equilibrium failed to see the crisis coming, there was a brief window of openness that quickly closed as the profession patched its models and returned to business as usual.
In 2026, the fourth volume in the series documents a field that has arrived. Agent-based models are in central banks. Complexity economics is taught at Oxford, the Santa Fe Institute, and a growing number of universities worldwide. The Bank of England uses complex network analysis and heterogeneous agent models as standard tools. The intellectual framework that said trend following should not work is being replaced, at the institutional level, by one that explains why it does.
The ground has shifted beneath the profession. And most of the profession has not yet noticed.
But we have noticed. Because the shift confirms what we have known empirically for decades: that markets are not in equilibrium, that prices are not random, that trends are real, that fat tails are structural, and that a diversified systematic programme built to respond to the dynamics of a complex adaptive system is not an anomaly to be explained away. It is the strategy that the science says should work.
The economy is on a computer now. And on the computer, our strategy is the one that thrives.
Next in the series: “What Complexity Economics Means for Your Trading.” The final article draws every thread together: from Arthur’s El Farol bar to Farmer’s agent-based economies, and what it all means for portfolio construction, risk management, and the future of systematic trend following.
Want the theoretical foundation for why trend following works?
The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.