Agent-Based Markets and Zero Intelligence
Can simple market agents generate price patterns without sophisticated forecasts?
Simulate simple order-submission rules and compare their price or order-book patterns with the observed market under the same measurement procedure. Such models show that complex-looking patterns can emerge without sophisticated forecasts; they do not prove the mechanism in every security.
Zero-intelligence models are illustrative benchmarks, not mandatory null models for every security.
Evidence map
| Aspect | Finding |
|---|---|
| What it is | Two results that bound what the stylized facts can be taken as evidence for. |
| Key result / formula | In the Lux-Marchesi model, agents switch between a fundamentalist strategy and two chartist moods, optimistic and pessimistic, according to realised profits and the opinions of others, and the price moves with excess demand. |
| Why it matters for backtesting | These are the right nulls for two kinds of claim. |
What it is
Lux & Marchesi (1999) showed that fat tails and volatility clustering emerge from the interaction of heterogeneous traders even when the news feeding the market is Gaussian. Farmer, Patelli & Zovko (2005) showed that much of the microstructure of a real order book is reproduced by agents with no intelligence at all, placing and cancelling orders at random.
Key result / formula
The fundamental value follows a random walk with Gaussian innovations, so the input has neither heavy tails nor clustering; yet the simulated returns show a heavy tail and a slow decay of the autocorrelation of absolute returns like the empirical ones, while the price tracks the fundamental on average. The bursts come from temporary surges in the chartist share — a critical regime the system enters and leaves on its own. Farmer, Patelli & Zovko took the opposite tack on the London Stock Exchange: a continuous double auction where limit orders, market orders and cancellations arrive as Poisson streams at rates measured from the data, with no strategy at all. With one free parameter, the model accounts for 96% of the variance of the bid-ask spread and 76% of the variance of the price diffusion rate across stocks — spread and short-horizon volatility are largely mechanical consequences of order-flow rates and the book's rules.
Why it matters for backtesting
That returns are fat-tailed and volatility clusters is not evidence of an exploitable behavioural regime, since a Gaussian fundamental produces both through interaction alone. That a stock's spread or short-horizon volatility is "anomalous" must first be compared with what a zero-intelligence book with that stock's order rates generates, before any story about informed traders is entertained. Neither model has a directional signal: they generate the texture of prices, not their sign. Their limits run the other way — Lux-Marchesi reaches the facts with many free choices, and the zero-intelligence book assumes independent arrivals, so it is silent on whatever depends on the long memory of order signs, which is where structure beyond mechanics would live, if anywhere.
Source
Lux & Marchesi, "Scaling and criticality in a stochastic multi-agent model of a financial market", Nature 397, 1999, 498-500; Farmer, Patelli & Zovko, "The predictive power of zero intelligence in financial markets", Proceedings of the National Academy of Sciences 102(6), 2005, 2254-2259. Primary source