Price Formation: How Price Emerges Without a Price-Setter

The price of a financial asset at any moment is not set by anyone. No central authority, no auctioneer, no regulator determines what a share of Apple stock costs. The price is the outcome of a decentralized process in which thousands of agents --- each acting on local information about their own inventory, their own estimate of value, and the current state of the order book they can observe --- submit orders that interact through the matching engine. Price formation is emergence in the precise sense: a global quantity (the price) arises from local interactions (order submissions) without any agent computing or intending the global outcome.

The Walrasian Auctioneer and Its Absence

Introductory economics textbooks describe price formation through the Walrasian auctioneer: a fictional entity that collects all agents’ demand and supply schedules, computes the market-clearing price where quantity demanded equals quantity supplied, and announces this price to all participants. The auctioneer is a pedagogical device. No such entity exists in any real market.

In a real continuous double auction market --- the mechanism used by every major stock exchange, futures exchange, and electronic trading venue --- orders arrive continuously, execute immediately when they cross existing orders, and prices change with every transaction. There is no aggregation step, no clearing computation, and no announcement. The price is simply the last transaction price: the price at which the most recent market order was matched against the most recent resting limit order.

This absence of a price-setter is what makes price formation a genuinely emergent phenomenon. The price is not a computed equilibrium; it is the trace left by a sequence of bilateral transactions, each determined by the local decision of one agent to submit a market order and the prior decision of another agent to place a limit order at that price.

The Continuous Double Auction

The rules of the continuous double auction are simple and deterministic:

  1. Any agent may submit a limit order (buy or sell at a specified price) or a market order (buy or sell at the best available price) at any time.
  2. When a buy order’s price meets or exceeds the lowest sell order’s price, a transaction occurs at the resting order’s price.
  3. The transaction removes the matched quantity from the book (or reduces it, if partially filled).
  4. The transaction price is recorded as the market’s current price.

A sequence of transactions traces out a price path: P(t_1), P(t_2), P(t_3), … This path is the market price. It is not a smooth function --- it jumps at every transaction and is constant between transactions. The transactions occur at irregular intervals (seconds or milliseconds in liquid markets, hours in illiquid ones).

The distinction between the transaction price and the bid-ask midpoint matters. The transaction price alternates between the bid side and the ask side as buy and sell market orders alternate, producing a mechanical “bounce” between two price levels. The midpoint of the best bid and best ask is a smoother estimate of the underlying value and is preferred for analytical purposes. The midpoint changes only when the best bid or best ask changes, which happens when limit orders are placed, cancelled, or executed.

Information Aggregation

The Efficient Market Hypothesis (Fama, 1970) asserts that market prices reflect all available information. The mechanism by which information enters prices is trading: an informed agent who knows the price is too low submits buy orders, pushing the price up; an informed agent who knows the price is too high submits sell orders, pushing the price down. The price adjusts until it reflects the information.

Albert Kyle’s 1985 model in Econometrica formalized this mechanism precisely. The model has three agent types:

One informed trader who knows the asset’s true value v and submits an optimal order to maximize profit.

Noise traders who submit random buy and sell orders for liquidity reasons unrelated to information (portfolio rebalancing, tax considerations, consumption needs).

A market maker who observes the total order flow (the sum of informed and noise orders) but cannot distinguish informed from uninformed orders. The market maker sets the price as a linear function of the net order flow: P = lambda * (x_informed + x_noise), where lambda measures the market’s “illiquidity” --- how much the price moves per unit of order flow.

Kyle proved that in equilibrium:

  • The informed trader submits an order proportional to the difference between the true value and the current price, modulated by the noise trader volume (more noise trading provides more cover for the informed trader).
  • The market maker’s pricing rule is a martingale: the expected future price, given all public information, equals the current price.
  • The informed trader’s private information is incorporated into the price gradually over the trading period, at a rate that depends on the noise trader volume.

The model establishes that price formation is information aggregation through a specific mechanism: the market maker infers information from net order flow, adjusting the price in response. More order flow in one direction is interpreted as evidence that informed traders are buying (or selling), and the price moves accordingly.

Price Impact: Permanent and Temporary Components

When a large market order arrives, it moves the price. This price impact has been studied extensively and has two components:

Temporary impact. The immediate price change caused by consuming liquidity from the book. A large buy order executes against multiple price levels, pushing the transaction price above the pre-order mid-price. This impact is mechanical: the price moves because the order consumes the standing limit orders at the best prices, and subsequent transactions occur at worse prices. After the order is complete, new limit orders refill the depleted price levels, and the price partially recovers. The temporary component reflects the cost of immediacy, not the information content of the order.

Permanent impact. The portion of the price change that persists after the book has refilled. This component reflects the information content of the order: a large buy order may signal that the buyer knows the asset is undervalued, and the market maker updates the price to reflect this inference. The permanent impact is the mechanism by which private information enters public prices.

The empirical finding (Bouchaud et al., 2018) is that approximately half of the initial price impact is permanent and half is temporary, though this ratio varies across markets and time horizons. The total impact scales approximately as the square root of order size: impact ~ sqrt(q), where q is the order quantity. This concave scaling means that the marginal impact of additional quantity decreases with order size, reflecting the increasing depth of the book at successive price levels.

The square-root impact law is one of the most robust findings in market microstructure. It holds across U.S. equities, European equities, U.S. Treasury futures, and foreign exchange (Almgren et al., 2005; Bouchaud et al., 2018). Its universality suggests that it is a property of the order book mechanism --- the way limit order depth accumulates away from the best quotes --- rather than a property of any specific market or trader population.

What Price Formation Reveals About Emergence

Price formation exemplifies a specific kind of emergence: the global quantity (price) is not computed by any agent but is the residual of many agents’ local decisions. Several features make it a canonical case:

No agent knows the price before it forms. Each agent submits an order based on their own estimate of value, their own inventory, and their own observation of the order book. The transaction price is the outcome of these submissions. Even the market maker, whose role is to set bid and ask quotes, does not determine the transaction price --- the transaction price is the quote at which a market order happens to arrive.

The price aggregates dispersed information. Kyle’s model shows that the equilibrium price incorporates the informed trader’s private information, even though the market maker cannot observe that information directly. The aggregation occurs through the mechanism of order flow: the informed trader’s order changes the net flow, the market maker infers information from the flow, and the price adjusts. This is not a voting mechanism or an averaging mechanism --- it is an inference mechanism.

The price has properties that no agent intended. The square-root impact law, the autocorrelation of order flow, the volatility clustering, and the fat-tailed return distribution are all properties of the price process that no individual order submission was designed to produce. They are consequences of the interaction mechanism operating on the statistical properties of order flow.

Further Reading