Price Discovery

Every price you see on a screen is a slightly wrong version of a number you can never observe directly: the efficient price, the market's true consensus estimate of what an asset is worth given everything currently known. The printed price wobbles around it because of the bid-ask spread, temporary supply-and-demand imbalances, and plain noise. Price discovery is the process by which fresh information gets absorbed into that efficient price and the observed price is dragged toward it. It is, quite literally, the core job a market performs.

The right mental model splits every observed price into two parts. There is a permanent component — the efficient price — which moves only when genuine new information arrives and, once moved, stays moved. And there is a transitory component — microstructure noise from spreads, inventory, and fleeting order imbalances — which mean-reverts and washes out. Price discovery is all about the permanent part.

$$p_t = m_t + s_t, \qquad m_t = m_{t-1} + w_t .$$

Here $p_t$ is the observed price, $m_t$ is the hidden efficient price, and $s_t$ is the transitory noise. The efficient price $m_t$ follows a The Random Walk: each period it takes an unpredictable, permanent step $w_t$ driven by new information. That $w_t$ — the efficient-price innovation — is the thing price discovery is trying to pin down, and its size measures how much genuinely new information hit that period.

Who discovers the price first?

Because one asset trades in many places at once (Market Fragmentation), a natural question is which venue moves the efficient price. When news hits, does the futures market react first and the ETF follow, or vice versa? Hasbrouck's information share answers this: it is the fraction of the efficient-price innovation variance $\operatorname{Var}(w_t)$ that originates in a given venue. A venue with a 70% information share is where 70% of the permanent price moves are born; the other venues mostly import that price. High information share = a price leader; low = a price follower.

Worked example

A stock trades as both an S&P futures contract and a cash ETF on the same underlying. You estimate the efficient-price innovation and decompose its variance by where each move originated. Over your sample the total innovation variance is $\operatorname{Var}(w) = 1.44$ (in squared basis points), split as:

$$\underbrace{1.08}_{\text{futures}} \;+\; \underbrace{0.36}_{\text{ETF}} \;=\; 1.44 .$$

The information shares are then

$$\text{IS}_{\text{fut}} = \frac{1.08}{1.44} = 0.75, \qquad \text{IS}_{\text{ETF}} = \frac{0.36}{1.44} = 0.25 .$$

So the futures leads price discovery with a 75% share: three-quarters of the permanent price moves originate there, and the ETF's price mostly catches up afterward. This matches reality — futures are cheaper, faster, and more leveraged, so informed traders hit them first (exactly the The Kyle Model logic of informed order flow choosing where to trade). A stat-arb desk would read this as: watch the futures, trade the laggard ETF. The information share turns "where does the truth show up first?" into a single number.

Where it misleads

Price discovery is the unifying theme behind much of microstructure. The The Glosten-Milgrom Model and The Kyle Model describe how informed trading forces quotes to update, i.e. the mechanism of discovery; Adverse Selection is the cost market makers bear for providing it; Market Impact is the permanent price move a large trade causes precisely because it looks informative. And in a fragmented market, measuring which venue does the discovering — the information share — is how you find the price leader worth watching.