Market Makers and HFT: Who Provides Liquidity

The two risks a quote is paid for

A market maker publishes a bid and an offer at the same time, and stands ready to trade either. The spread is not a fee it charges; it is the price of the two risks it takes on by quoting both sides all day. **Inventory risk** is the obvious one. Every fill changes what you hold, and if the flow is one-directional you end up long a falling stock or short a rising one. The larger the position, the more the mark moves, and the mark is usually the biggest number in a market maker's day. **Adverse selection** is the subtle one, and it is the reason the business is competitive rather than easy. Some of your counterparties are uninformed, and you earn the spread on them. Others know something, and when they trade with you the price moves against you by more than the spread you earned. Your expected profit is the spread on uninformed flow *minus* the loss on informed flow, and the second term is why a quote that looks safe on average can bleed in a session. A session quoted a penny wide, from the lab — Fills (60 tickets of 100 shares): 52 prints, 5,200 shares traded · Spread captured: $52 at a 1¢ half-spread · Informed prints: 10 sellers who were right · Inventory at the close: 1,000 shares, bought into a falling market · Session P&L: −$155 — the mark, not the quote ← Read the last line twice. The spread earned $52 and the session lost $155. Almost none of that came from bad luck in the quoting; it came from being the last buyer on a one-way tape.

Who actually provides liquidity — and what front-running is not

Liquidity providers are not one species. **Designated market makers** on NYSE and **registered market makers** on Nasdaq take on quoting obligations in exchange for incentives and, historically, advantages in the book. A much larger share of liquidity now comes from **electronic market makers** — firms with no quoting obligation beyond the rules they choose, whose edge is speed, hedging and the ability to quote a very large number of names at once. That speed is where the popular story goes wrong. An algorithm cannot see your order before it reaches the market; your order goes to a broker, and the market learns about it from the print. What fast firms do is react to *public* information more quickly than anyone else — a quote changes on one venue and they adjust on five others in microseconds. That is a **latency race**, and it is legal, and it is what keeps quotes in line across venues. **Front-running** is a different act with a real victim: trading ahead of a specific customer order you have been entrusted with, because you know it is coming. That is fraud. The two are routinely conflated in market commentary, and the difference is exactly what the learner should be able to state: racing public information is competition; trading on non-public knowledge of a customer's order is a breach of duty. • **Obligational** makers must quote — designated market makers and specialist-style roles • **Non-obligational** electronic makers quote when it pays, and withdraw when it does not — which is why liquidity vanishes in stress • **Latency competition** is reacting faster to the same public information; it narrows spreads across venues • **Front-running** is trading ahead of a known customer order; it is a duty breached, not a race won

Leaning the book: how a maker sheds inventory

A maker holding 1,000 shares it does not want has two ways out, and only one of them costs money in a hurry. It can **cross out** — sell the position at the market, paying the spread and announcing to everyone that stock is for sale. Or it can **skew** its quotes: bid a little lower and offer a little more generously, so that the flow it attracts is one-sided in the direction it needs. Skewing is the craft. If the maker is long, it moves both sides of its quote down: a worse bid, so fewer willing sellers trade with it; a better offer, so more willing buyers do. It trades a little edge on every fill for a far better chance of ending the session flat, and how far it leans is a direct function of its inventory measured against the risk limits its desk sets. The visible consequence is that quotes are never neutral. When you see a bid for 100 shares and an offer for 5,000, you are usually looking at a maker that wants to sell. **Quote sizes carry information about the quoter’s position**, and reading that — two-sided versus one-sided, symmetric versus lopsided — is a real skill that most retail interfaces hide behind a single best-bid line. • **Crossing out** pays the spread and reveals your hand to everyone • **Skewing** shifts both quotes so the flow you meet shrinks the position • **Skew size** grows with inventory and tightens with the desk’s risk limits • **Asymmetric quote sizes** are the visible trace of a maker leaning The trade-off is the whole business in miniature: a maker that skews hard manages risk and earns less per fill, while one that quotes symmetrically earns more and carries more mark risk. The lab gives you that dial and shows you the P&L split.

The obligations that come with the quote

A quote looks like a business decision — you publish two prices and hope to earn the difference — but in registered market making it is also a legal duty. Firms registered as market makers in a stock take on **continuous two-sided quoting obligations**: they must show a bid and an ask in at least the required size during regular hours, keep the quote within a defined distance of the best price, and keep it live for a minimum share of the trading day. Quote too far off the market and you are not quoting at all for the purposes of the rule; step away entirely and you may lose the registration that comes with the privileges. That obligation is the reason liquidity does not disappear the instant conditions worsen. A market maker who would rather not hold inventory during a news event still has to publish a two-sided market, and the only legitimate way to reconcile the obligation with the risk is to widen. This is why the phrase “liquidity providers stepped back” is half wrong: they did not stop quoting, they quoted further away and in smaller size, which is a real withdrawal of *usable* liquidity but not an absence of quotes. Watch the size at the touch during a shock and you will see it thin before the spreads blow out. The other side of the privilege is what the market pays for it. Registered market makers get fee rebates for adding liquidity, exemptions from some short-sale restrictions when hedging, and access to order flow, and those economics are what make the business viable at spreads measured in fractions of a cent. It is worth seeing the whole arrangement as a bargain rather than a free market: the exchange offers privileges to firms willing to accept a quoting duty, and the public gets a continuous, if sometimes distant, two-sided market in return. • Registered market makers must quote both sides continuously, in size, near the market. • The obligation is what makes widening the honest response to risk — not walking away. • Rebates, hedging exemptions and flow access are the payment for accepting that duty. • Watch size before spread in a shock: size thins first, and the spread follows. Designated market makers who manage an opening or closing auction carry stronger obligations still — they are expected to set a price when the book is one-sided, which is a duty to take the other side of an imbalance, not merely to show up.

The economics of the race

The reads above said what a maker is paid and what it is obliged to do. They did not say why the largest providers spend extraordinary sums chasing microseconds, and the answer is neither sinister nor obvious: speed has a price, and the price is the spread on the last few shares of a race that many firms enter and one wins. Start with the fee model, because it pays for much of the quoting. Exchanges charge a fee to the firm that removes liquidity and pay a **rebate** to the firm that provided it — a fraction of a cent per share, which sounds trivial until it is multiplied by a billion shares a day. For a firm that quotes continuously and trades in enormous volume, the rebate is a revenue line of its own, sitting on top of the half-spread it earns. Note the conflict that creates: the same exchange is paying your broker’s router partly on the basis of whether it added or removed liquidity, which is the disclosed tension the routing lesson is about — a structural conflict, not a fraud. Then speed. A quote on one venue is repriced and a slower venue has not caught up yet; the firm that reacts first can trade at the stale price before it disappears, and that profit is the difference between the old and new quotes multiplied by the size on offer. This is **latency arbitrage**, and it is arbitrage of *public* information — nothing is known that the tape does not already show. Firms compete for it by colocating servers next to the matching engine, buying direct data feeds instead of the consolidated one, and, at the extreme, using microwave links and specialised chips to shave microseconds off a message’s travel time. The payoff is that quotes across venues stay consistent with each other; the cost is that the race is a genuine arms race, and the losers spend the money without earning it back. Order-to-trade ratios are the statistic most often cited as evidence of something wrong, and they deserve a plain reading. A firm that posts and cancels quotes all day while executing rarely will show a ratio in the tens or hundreds to one — that is what a quoting strategy *is*, because a quote that is not taken has to be refreshed as the market moves. A high ratio is a number regulators watch, and it has sometimes prompted minimum-resting-time proposals, but the number alone does not distinguish a market maker managing inventory from a manipulator, which is why enforcement turns on intent and effect rather than on the ratio. The “phantom liquidity” complaint has a real grain. A displayed quote that vanishes the instant a slow trader reaches for it is a genuine cost to that trader, and it is the honest reason some investors would like a minimum time a displayed order must remain. The middle position is neither “the market is rigged” nor “nothing is happening”: the race to react first does make quotes more consistent and cheaper for most participants, and it also means that some of the liquidity you see is contingent on nobody using it. Finally, the transfer. The spread a fast provider earns is money paid by whoever demanded immediacy — and when you cross the spread, a small part of that payment funds the race. That is the actual flow of funds behind the fee debate, and it is why the argument about maker-taker rebates is really an argument about who should pay for a market that everyone uses. For a retail account the practical summary is short: the competition tends to tighten quoted spreads, and you still pay the spread every time you demand an immediate fill. Both truths sit together. A useful test for any claim about high-frequency trading: ask whose money moved, and whether the information used was available to everyone. If the answer is “public data, traded faster” it is competition; if it is “a specific customer’s order” it is the other thing entirely.

Sources

• Market makers, designated market makers and quoting obligationsNYSE / Nasdaq — Exchange rulebooks
• High-frequency trading and liquidity provisionSEC — Equity market structure literature review
• Trading and Exchanges: Market Microstructure for PractitionersLarry Harris, Oxford University Press (2003)