Was the Electronic Order Book Always Going to Win? Glosten's Prophecy

In 1994, most of the world's stock trading still ran through human intermediaries. The NYSE had its specialists. Nasdaq had its dealers. London had market makers on the phone. The idea that trading would end up as an anonymous electronic queue of limit orders, with no designated market maker at all, was a plausible future but hardly a certainty.

Larry Glosten wrote a paper that year with a provocative title and an even more provocative argument: not merely that the electronic limit order book would win, but that there is a structural reason it must. It is the trading institution that nobody can profitably attack.

Reading it three decades later, when nearly every liquid market on earth is an electronic limit order book, is a slightly eerie experience.

The problem: which market design survives competition?

Imagine you run an exchange. You have chosen some way of organizing trade: a specialist system, a dealer market, an order book, whatever.

Now a competitor sets up shop next door and tries to steal your business. They do not have to take all your business. They just need to skim the profitable parts. In market structure this is called cream-skimming, and it is the mortal threat to any trading venue: a rival who takes your easy, harmless, uninformed order flow and leaves you with the toxic remainder.

That is not a hypothetical. It is what "third market" dealers, internalizers, and payment-for-order-flow arrangements have always done. So the sharp question is: is there a market design that cannot be cream-skimmed?

The key idea via analogy: the ladder priced for the worst case

To see Glosten's answer, you have to understand what a limit order book actually is, economically.

A limit order book is a ladder of prices. At the best offer, some shares are available. Slightly higher, more shares. Higher still, more. If you want to buy a small amount, you pay the best offer. If you want to buy a lot, you walk up the ladder and pay progressively worse prices. The book is, in effect, a schedule that prices every possible order size in advance.

Now, the crucial question: who puts those orders in the book, and what are they thinking?

Consider someone posting a limit order high up the ladder, far from the current price. When does that order get filled? Only when a very large buy order comes sweeping through. And who submits very large buy orders that sweep multiple levels of the book? Disproportionately, someone in a hurry with a strong opinion, which is to say, disproportionately, someone informed.

This is the paper's beautiful and slightly grim insight. Each level of the book is being picked off by a different kind of order. The orders near the top get hit by ordinary small trades, a benign mix. The orders deep in the book only get hit by the big, aggressive, likely-informed sweeps. Therefore, a rational limit order trader must price each level conditional on the kind of order that would reach it, demanding worse and worse prices further out, because the flow that reaches those levels is more and more dangerous.

The result is that the book's price schedule is not arbitrary. It is precisely the schedule at which liquidity providers break even against the adverse selection they face at each depth. Every rung of the ladder is priced exactly at cost, given the danger of that rung.

And now the punchline. If every level of the book is already priced at break-even, there is nothing left for a competitor to skim. A rival dealer who wants to steal business must offer a better price than the book at some size. But the book's price at that size is already break-even. Beating it means losing money. There is no fat in the schedule.

Compare that with a dealer market where a single spread is quoted for all sizes. That spread is necessarily an average: too wide for small, safe orders and too tight for large, dangerous ones. That average creates an opening. A competitor swoops in, offers small retail orders a slightly better price than the average spread (safely, because those orders are harmless), and leaves the incumbent dealer with only the big, toxic ones. A single spread is cream-skimmable, precisely because it is an average. A price schedule is not, because it is not.

Hence the title. The open limit order book is inevitable not because it is elegant, or fast, or cheap, but because it is the only design with no cream to skim.

Glosten also derived two further results worth knowing. The book supports a genuine spread even for tiny trades, so limit orders do earn something on small, harmless flow. And in extreme conditions, the book provides as much liquidity as it is possible for any institution to provide, because when things get bad enough that even the book empties out, no dealer would have quoted either.

Why it mattered

• It called the future, and it was right. In 1994, this was a bold theoretical claim. Today, the overwhelming majority of the world's exchanges are order-driven electronic books. Equities, futures, foreign exchange, and crypto all converged on this design. The paper is a rare case of microstructure theory making a large structural prediction that the world then went and confirmed.
• It explains cream-skimming rigorously. Every argument about internalization, payment for order flow, dark pools, and retail wholesaling is, at bottom, an argument about whether someone is skimming benign flow away from the public book. Glosten gave that argument its theoretical foundation, and the framing, that you can only skim where prices are averaged rather than schedule-priced, is exactly the right lens.
• It gave the limit order book an economic theory. Before this, the book was plumbing. After this, the book was an equilibrium object: its shape reflects the adverse selection profile of the order flow that reaches each depth. Almost all modern theory of limit order books traces back here.
• It quietly explains why deep liquidity is expensive. The book's upward-sloping cost of size, which every execution trader knows in their bones, is not a mechanical accident. It is the rational price of the fact that only scary orders reach deep into the book.

The honest limitations

• The prediction was right but incomplete. Yes, the world became an electronic limit order book. But it also became a world of internalizers, wholesalers, dark pools and payment for order flow, all of which are, in Glosten's own terms, cream-skimming the public book. The book won the core, but a very substantial fraction of retail order flow in the United States never reaches it. The theory says that should be hard. Reality found a way, largely because a wholesaler can identify the flow as retail and therefore harmless before quoting, which is exactly the information the anonymous book does not have.
• Anonymity is doing a lot of work. The model's logic depends on limit order traders not knowing who they are facing. To the extent that a venue can segment order flow by counterparty type, the whole break-even-at-each-rung argument weakens.
• Nobody has to post limit orders. The model assumes a competitive supply of them. But in stressed markets, liquidity providers simply leave, and the book is a design that offers no obligation to quote. A designated market maker with obligations, whatever its inefficiencies, at least has to show up. This is the standing argument against pure order books in a crisis.
• It is a static equilibrium. The book in reality is a fiercely dynamic queue, with cancellations, queue priority, order splitting, and speed races that this framework does not attempt to capture. The mechanism design work that followed, and the empirical order book models of Cont and others, exist to fill that gap.

The one-line takeaway

Glosten argued that the electronic limit order book wins because every rung of the book is already priced to break even against exactly the kind of order that would reach it, leaving a competitor no profitable flow to skim, whereas any market quoting one average spread for all sizes is inviting someone to steal the easy orders and leave it the dangerous ones.