Market makers, spreads and the cost of immediacy

There is a price at which you can sell right now and a slightly better price at which you could sell if you were willing to wait. The gap between those two numbers is not a fee, a tax, or a conspiracy. It is the price of a service: someone stands ready to take the other side of your trade at a moment of your choosing, and charges for the privilege of being available.

Harold Demsetz gave that service its name in 1968: immediacy. Once you see the spread as the price of immediacy rather than as friction, most of the odd behaviour of quoted markets becomes predictable. Spreads widen before earnings, in illiquid small caps, in the first minute after the open, and in a crisis — all for the same underlying reason, which is that the cost of standing ready has gone up.

What a market maker actually does

A market maker quotes two prices at once: a bid, at which it will buy, and an ask (or offer), at which it will sell. It does not forecast direction. Its business is to buy slightly below the consensus value and sell slightly above it, many times a day, ending each day with as little inventory as possible.

Suppose a large Swiss blue chip is quoted 99.98 bid / 100.02 ask. The mid is 100.00. The quoted spread is 0.04, which on a 100 mid is 4 basis points (a basis point, bp, is one hundredth of a percentage point). The half-spread — what one side of a round trip costs you — is 2 bp. Buy and sell immediately with no price movement in between and you are down 4 bp before commissions.

The market maker's income is that 4 bp, but only on the trades where it buys from one client and sells to another at roughly the same time. Everything else — the inventory it is stuck with, the trades it does against someone who knows more — is cost.

Not all market makers are formally designated. On SIX Swiss Exchange, blue chips have plenty of voluntary two-sided quoting from proprietary firms and banks; ETFs and structured products, by contrast, typically depend on a designated market maker contracted by the issuer to post quotes within an agreed maximum spread for an agreed minimum size. That distinction matters more than most retail investors realise: in the second case, "liquidity" is a contractual obligation with defined limits, not the natural result of buyers meeting sellers. The lesson on measuring liquidity develops this point.

Where the spread comes from: three costs

Decades of theory and evidence converge on three components.

1. Order processing. Exchange fees, clearing, settlement, technology, capital, staff. This is the smallest and most predictable piece in modern equity markets, and it sets a floor: nobody quotes tighter than it costs to operate. The tick size — the minimum price increment — can bind here. In a stock trading at CHF 5 with a tick of one rappen, the spread cannot be narrower than 0.01, which is 20 bp. The security is not illiquid; the grid is simply too coarse for a tighter quote to exist.

2. Inventory risk. When a market maker buys from you, it owns something it did not want to own. Until it can lay that position off, it carries the asset's price risk on its own book.

The arithmetic is brutal. Say a firm ends the day holding CHF 2,000,000 of a stock whose daily standard deviation is 1.2%. The one-day standard deviation of its P&L on that position is 0.012 × 2,000,000 = CHF 24,000. The half-spread it earned to acquire the position, at 2 bp, was 0.0002 × 2,000,000 = CHF 400. It earned four hundred francs and took on twenty-four thousand francs of daily risk. No sane operation runs that way, which is why market makers do not sit on inventory: they hedge it, or they skew their quotes to attract the offsetting flow.

Quote skewing is the visible symptom. A firm that is long too much stock will move both bid and ask down — say to 99.95 / 99.99 — keeping the spread at 4 bp but making it more attractive for the next trader to buy from it and less attractive to sell to it. The mid has moved without any news. Ho and Stoll formalised this inventory-control behaviour in 1981.

3. Adverse selection. This is the deep one. Some of the people trading against a posted quote know something the quoter does not. The market maker cannot tell them apart from the uninformed at the moment of trade, so it charges everyone a premium that covers its losses to the informed.

The adverse selection calculation

Glosten and Milgrom's 1985 model makes this precise and is worth working through, because the result is a formula you can carry around.

Take an asset that will turn out to be worth either 110 or 90, each equally likely, so its expected value today is 100. Suppose a fraction α of incoming orders come from traders who already know which outcome it is, and the remaining (1 − α) come from uninformed traders who buy or sell for their own reasons with equal probability.

Set α = 20%. If the true value is 110, the probability the next order is a buy is 0.20 (the informed, who all buy) + 0.80 × 0.5 = 0.60. If the true value is 90, the probability of a buy is 0 + 0.80 × 0.5 = 0.40. Starting from a 50/50 prior, the probability that the value is 110 given that a buy arrives is 0.60 / (0.60 + 0.40) = 0.60.

A competitive market maker must set its ask equal to the value conditional on being hit by a buy — otherwise it loses money on average. So the ask is 0.60 × 110 + 0.40 × 90 = 66 + 36 = 102. By symmetry the bid is 98. The spread is 4 on a mid of 100: 400 bp.

Now set α = 5%. The probability of a buy given the high value is 0.05 + 0.475 = 0.525; given the low value, 0.475. The posterior after a buy is 0.525, and the ask becomes 0.525 × 110 + 0.475 × 90 = 57.75 + 42.75 = 100.50. The spread is 1, or 100 bp.

The pattern generalises: in this setup the half-spread equals α × (V_high − V_low) / 2. It is the product of how likely you are trading against someone informed and how much there is to know. That single sentence explains why spreads widen ahead of an earnings release (both terms rise), why they are wider in a thinly followed small cap than in Nestlé (the informed are a larger share of a small flow), and why they blow out in a crisis (nobody knows the range of outcomes any more).

Quoted, effective and realised spread

The quoted spread describes the top of the book. What you actually pay is the effective spread, measured against the mid at the moment your order arrives.

Return to 99.98 / 100.02, mid 100.00. You buy 5,000 shares. Only 2,000 are available at 100.02; the remaining 3,000 fill at 100.06 as your order walks up the book. Your average price is (2,000 × 100.02 + 3,000 × 100.06) / 5,000 = (200,040 + 300,180) / 5,000 = 100.044. Your effective half-spread is 0.044, or 4.4 bp — more than double the 2 bp quoted half-spread. The quote was real; it was just not deep enough for you.

Now measure what the market maker earned. Five minutes later the mid is 100.05, because your buying revealed information. The realised half-spread — the market maker's actual profit — is 100.044 − 100.05 = −0.006, a loss of 0.6 bp. The price impact of your trade was 100.05 − 100.00 = 0.05, or 5 bp.

The identity is exact: effective half-spread = realised half-spread + price impact, and here 4.4 = −0.6 + 5.0. This decomposition is the workhorse of execution analysis. It tells you what portion of your cost was payment for immediacy and what portion was the market updating its view because you traded. The lesson on slippage and impact builds directly on it.

Who is allowed to quote, and under what licence

Quoting two-sided prices professionally is a regulated activity in Switzerland, and the licence behind a quote tells you something about how much the quote is worth.

The Financial Institutions Act (FinIA / FINIG), in force since 1 January 2020, consolidated what had been a scattered set of authorisations into one ladder: portfolio managers and trustees at the entry level, then managers of collective assets, fund management companies, and securities firms (Wertpapierhäuser), alongside banks licensed under the Banking Act. The securities-firm category is the one that covers market making. It captures firms that deal in securities on a professional basis for their own account on a short-term horizon, and firms that publicly quote prices in securities on a permanent or on-request basis — which is the statute describing the business set out at the start of this lesson. Such firms are prudentially supervised by FINMA and carry capital, organisational and risk-management requirements accordingly.

That distinction has a practical edge when you read a screen. A price posted by a supervised securities firm sits behind an entity with a capital requirement and a regulator entitled to ask what its inventory limits are. A price posted by an issuer in its own structured product sits behind a contractual undertaking in a term sheet, with the limits the term sheet states and no more. Both render on your screen as a bid and an ask of similar width. They are not the same promise, and the difference becomes visible on the day you want to sell something the issuer would rather not buy back.

What to take away

• The bid-ask spread is the price of immediacy, and it decomposes into order processing costs, inventory risk, and adverse selection — of which adverse selection dominates in liquid markets and explains almost all of the interesting variation.
• In the Glosten-Milgrom framework the half-spread equals the probability of facing an informed trader multiplied by half the range of possible values, which is why spreads widen before news and in thinly covered securities.
• A market maker that skews its quotes is managing inventory, not forecasting; the mid can move without any information arriving.
• Your true cost is the effective spread measured against the arrival mid, not the quoted spread, and it splits exactly into the market maker's realised spread plus the price impact your own order caused.
• Market making in Switzerland is a licensed activity — firms quoting prices publicly on a permanent or on-request basis fall into the securities-firm category under FinIA and are supervised by FINMA — which is why a supervised firm's quote and an issuer's contractual quote in its own product are different kinds of promise despite looking identical on screen.

Sources

• Harold Demsetz, "The Cost of Transacting," Quarterly Journal of Economics, 1968 — the original treatment of the spread as the price of immediacy.
• Lawrence R. Glosten and Paul R. Milgrom, "Bid, Ask and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders," Journal of Financial Economics, 1985.
• Thomas Ho and Hans R. Stoll, "Optimal Dealer Pricing under Transactions and Return Uncertainty," Journal of Financial Economics, 1981 — the inventory-control model.
• Thomas E. Copeland and Dan Galai, "Information Effects on the Bid-Ask Spread," Journal of Finance, 1983.
• Richard Roll, "A Simple Implicit Measure of the Effective Bid-Ask Spread in an Efficient Market," Journal of Finance, 1984.
• Yakov Amihud and Haim Mendelson, "Asset Pricing and the Bid-Ask Spread," Journal of Financial Economics, 1986.
• Larry Harris, Trading and Exchanges: Market Microstructure for Practitioners, Oxford University Press, 2003.
• Swiss Federal Act on Financial Institutions (FinIA / FINIG), in force 1 January 2020, on the authorisation categories for securities firms, and FINMA's published guidance on licensing.