The microstructure lens
M1 — Market microstructure
M0 taught you the machinery: the book, the spread, the path from click to settlement. M1 asks a different question — why does that machinery behave the way it does? — and it requires one shift in how you look at the book.
In M0, the order book was plumbing: a data structure that sorts orders and matches them. In M1, the order book is an inference engine: a public, continuously revised estimate of what the asset is worth, updated by what people do.
That sentence is the whole module. Everything below is machinery for taking it seriously.
Three questions M0 couldn’t answer
Market microstructure is the study of how the mechanism of trading affects the prices that come out of it. Three questions define the field:
- Why does the price you transact at differ from the value? M0.4 answered “the spread,” and decomposed it. M1 derives it.
- Why does trading itself move the price? Not the news — the trade. A large buy pushes the price up even when nothing about the company has changed. Why should the market care that you bought?
- Why does liquidity evaporate precisely when it’s wanted? Spreads widen and depth vanishes in exactly the moments — earnings, shocks, crises — when people most need to trade.
Notice all three are about the gap between value and price. Classical finance assumes that gap away. Microstructure is the study of the gap.
The organising split: informed vs uninformed
Everything in M1 rests on dividing order flow in two.
INFORMED FLOW UNINFORMED FLOW
─────────────────────────────────────────────────────────────
knows something the maker trades for reasons
doesn’t — earnings, a model, unrelated to value —
a client order, faster data rebalancing, a house
deposit, an index change
trades in ONE direction, direction is essentially
correlated with the truth a coin flip
the maker LOSES to them the maker WINS from them
also called: “toxic” flow also called: noise, or
liquidity traders
Two things about this split are easy to say and hard to keep hold of:
Neither wears a badge. At the instant a buy order arrives, it is a buy order. The maker sees identical bytes whether it came from a hedge fund with a signal or a dentist with a bonus. This is the fact that generates almost every result in M1 — if makers could tell them apart, they would simply quote a tight market to one and refuse the other, and there would be no spread to speak of.
“Informed” doesn’t mean insider. It means anything the maker doesn’t know yet — a better model, a faster feed, a view on where a big parent order is heading. Most informed flow is entirely legal and quite mundane.
The maker’s whole problem reduces to one line: quote to everyone, and learn from who takes you up.
What “the book is a belief” means
Take that seriously and the numbers on the screen change meaning.
what you see what it actually is
─────────────────────────────────────────────────────────────
the midpoint ≈ the market’s expected value of the asset,
CONDITIONAL on everything it has seen —
including the order flow so far
the spread ≈ the price of being wrong about that estimate
the depth ≈ how confident the market is / how much
noise there is to hide in (this is M1.4’s λ)
a trade ≈ EVIDENCE. Not just an event.
That last row is the hinge of the module. A buy order is weak evidence the asset is worth more than the maker thought — because informed traders buy when it is, and uninformed traders buy at random. So a rational maker, seeing a buy, must raise its estimate, even having learned nothing about the company.
This is why trading moves prices even with no news, and it is not a market failure. It’s Bayes’ theorem running in public, one order at a time. M1.2 writes the update down.
Three forces, three models
A real spread is set by several pressures at once. The literature’s strategy — and this module’s — is to isolate them one at a time by building a model that switches the others off.
model isolates switched off result
────────────────────────────────────────────────────────────────────
Glosten–Milgrom information maker is risk- spread = μ × the
(M1.2) / learning neutral, holds range of value;
no inventory quotes are
conditional
expectations
Ho–Stoll inventory risk nobody is quotes SKEW with
(M1.3) informed the maker’s
position
Kyle (1985) a STRATEGIC one informed depth λ; price
(M1.4) informed trader trader, who impact is LINEAR
chooses his size in order flow
Hold on to what each one throws away, because that’s what makes each tractable — and it’s the honest answer to “which model is right?” None of them. Each is a clean statement of one force. Real quotes carry all three, which is exactly the three-part decomposition you already met in M0.4: order-processing, inventory, adverse selection.
M1.5 then asks what survives contact with data, and M1.6–M1.7 ask what the firms who live in this layer actually optimise.
Order-driven and quote-driven, revisited
One recap worth making explicit, because M1 slides between them. In an order-driven market (M0.2’s book) there is no designated maker — liquidity is whatever limit orders happen to be resting, posted by anyone. In a quote-driven market (M0.5’s bond dealers) a designated dealer is obliged to show a two-sided price.
The models in M1 are written in the language of a single representative market maker setting a bid and an ask. That is literally true in a quote-driven market and a useful fiction in an order-driven one, where “the maker” stands in for the competing crowd of limit-order posters at the top of the book. The economics carry over; the institution doesn’t.
Source: Maureen O’Hara, Market Microstructure Theory, ch.1 for the framing — or stay in Larry Harris and go deeper into ch.10–14. If you read one thing before M1.2, make it Harris’s treatment of informed traders; his taxonomy of why people trade is what makes the informed/uninformed split feel like a description of real people rather than a modelling convenience.