Liquidity Is a Story We Tell Ourselves
Liquidity feels like a property of markets. It is closer to a collective agreement — and agreements can be revoked without notice.
Open a brokerage app during an ordinary Tuesday and the market looks like a lake: wide, calm, bottomless. A quote appears, you trade against it, another quote appears in its place, and the process repeats thousands of times a second without apparent friction. It is easy to mistake this for a physical fact about the asset — a large float, an active shareholder base, genuine depth. Mostly it isn’t. What you are looking at is a service, provided continuously by a small number of firms who can withdraw it at will, and who are contractually and economically obligated to do exactly that at the moments it would help you most.
This is not a cynical observation so much as a structural one. Modern liquidity — in single stocks, in the ETFs that wrap them, in the futures that hedge both — is manufactured by market makers running inventory models, not accumulated by patient buyers waiting to be found. The distinction matters because manufactured liquidity has a kill switch. Accumulated liquidity, the old-fashioned kind built from standing limit orders left by investors who genuinely wanted to transact at a given price, degrades gradually as conditions change. Manufactured liquidity can go from full depth to a wide, hollow spread in the space of one bad print, because the thing generating it was never a resting order — it was a real-time calculation, rerun continuously, about whether making a market is still a good idea.
The machinery behind the calm
A market maker’s job, reduced to its essentials, is to quote both sides of a security, capture the spread, and hedge away the resulting inventory risk fast enough that the position doesn’t become a directional bet. The economics work only because the risk is small and short-lived on average — the firm holds inventory for seconds, sometimes fractions of a second, and relies on being able to offload it in a correlated instrument (an index future, an ETF, the underlying basket) almost immediately. Strip away the technology and this is the same function a specialist performed on the floor of the NYSE a generation ago. What has changed is the scale at which it happens and the degree to which it has been centralized in a handful of electronic firms whose models are, by design, extremely sensitive to volatility.
That sensitivity is the whole point, and also the vulnerability. A market maker who kept quoting tight, deep markets through a violent repricing event would be accepting essentially unlimited risk in exchange for a few basis points of spread — a trade no rational firm makes. So the models are built to widen automatically, and to widen fast, the moment realized volatility, order flow imbalance, or correlation breakdown crosses a threshold. This is not malfunction. It is the system working as intended. The unsettling part is that “as intended” and “just when you need it” turn out to be the same moment, because both are triggered by the identical signal: the market moving hard and fast in one direction.
ETFs complicate this further by adding a second, less visible layer of manufactured depth. An ETF’s on-screen liquidity is not really a function of how many shares trade day to day — it is a function of the authorized participant mechanism, the process by which large market makers can create or redeem shares against the underlying basket to keep the ETF’s price tethered to its net asset value. That mechanism works beautifully when the underlying basket itself is liquid and orderly. It works much less beautifully when the basket contains names that are themselves thin, foreign, or halted, because then the arbitrage that keeps the ETF’s price honest requires trading things that cannot currently be traded at a sane price. The ETF wrapper advertises the liquidity of the vehicle; it inherits the liquidity of the contents. In stress, the gap between those two becomes the story.
Liquidity is not a reserve that gets drawn down. It is a decision, remade every few milliseconds, about whether the compensation for providing it still exceeds the risk of doing so.
Why the air pocket shows up exactly when you need air
Consider what happens in a genuine dislocation — a macro surprise, a large forced seller, a correlated deleveraging across several asset classes at once. The first thing that breaks is not the market’s capacity to trade; it’s the market’s capacity to price. Market-making models depend on being able to hedge quickly and on a reasonably stable relationship between the instrument they’re quoting and the instrument they’re hedging with. When correlations shift abruptly — when the future no longer tracks the basket, when the ETF no longer tracks its NAV, when cross-asset relationships that held for years suddenly don’t — the hedge becomes unreliable, and an unreliable hedge means unbounded risk on every quote. Firms respond, rationally and near-simultaneously, by pulling back size and widening spreads. Because most electronic market makers run broadly similar risk models reacting to broadly similar signals, this withdrawal is correlated across firms rather than staggered. The result is not a gradual thinning of the order book; it is closer to a phase transition — deep, then suddenly shallow, with little in between.
Retail and institutional investors experience this as an “air pocket”: the price a screen showed a moment ago simply isn’t executable size, and the next real price might be a percentage point or more away, discovered only when a large order actually clears. Circuit breakers and trading halts exist precisely because regulators learned, from a series of incidents over the past two decades, that this kind of gap is a recurring feature of electronic markets rather than a rare accident. The halts don’t add liquidity; they buy time for it to be rebuilt, which is itself an admission that the liquidity on offer moments before was more fragile than it looked.
What durable liquidity actually requires
None of this means electronic market-making is a net negative — average spreads and average execution costs are almost certainly lower than they were in the specialist era, and that benefit is real and persistent for the vast majority of trading days. The honest framing is narrower: the liquidity visible on any given day is a function of prevailing volatility and correlation stability, provided by profit-seeking firms with the right, and the risk models, to step back the instant those conditions shift. It is a service with a cancellation clause, not a standing commitment.
For an investor, the practical implication is to size positions and choose order types on the assumption that the depth you see is not a promise you can call on when the market is failing — that’s the one moment it isn’t reliably there. For a market structure that increasingly routes retail flow through a small number of large market makers and ETF authorized participants, it means understanding that “efficient” and “resilient” describe different things, and that the recent decades of ever-tighter spreads have made markets considerably more efficient without making them proportionally more resilient. The two virtues get conflated because most of the time, in most conditions, they happen to move together. They stop moving together at exactly the moments that matter, which is the entire, uncomfortable point.