Liquidity is not a number — it is a system property

Markets are full of numbers that look like liquidity.

Daily volume. Open interest. Total value locked. Assets under management. The size of an order book at a particular moment. Each can describe activity or capacity, but none answers the question an investor actually cares about:

Can I change my position, at the size I need, when the system is under pressure?

That is a different question.

A market may process billions in a quiet session and still fail to absorb a large sale without moving sharply. An order book may look deep while most of its quotes disappear as volatility rises. A fund may hold liquid assets and offer redemptions on terms that become difficult when every holder wants the same exit.

Liquidity is not inventory sitting on a screen. It is the behavior of a system when demand for immediacy becomes one-sided.

Volume measures motion, not resilience

Volume is useful. It tells us that transactions happened. It does not tell us what would happen if the next transaction were larger, more urgent, or pointed in the same direction as everyone else’s.

Consider two markets trading the same daily notional.

In the first, activity is distributed across many independent participants, quotes remain close to the market, and no single venue or market maker is essential. In the second, most volume comes from a small group of firms trading with each other, using the same funding source and similar risk models.

The headline number can be identical. The failure mode is not.

When conditions change, the first market can lose some participants and continue functioning. In the second, one funding constraint can cause several liquidity providers to reduce risk at once. Volume observed under normal conditions says little about the capacity available under stress.

This is the same mistake engineers make when they confuse average throughput with system reliability. A service handling 10,000 requests per second in normal conditions is not necessarily resilient to a traffic spike, a database slowdown, or the loss of one dependency.

Average activity describes the happy path. Liquidity is a stress property.

The displayed market is only an interface

An order book gives the comforting impression that liquidity is stored at known prices. There are bids below, offers above, and a visible spread between them.

But a quote is not a reserve. It is a conditional intention.

The participant providing it can cancel, widen, or reduce the quote when volatility changes, funding becomes expensive, inventory limits are reached, or the information in incoming orders begins to look dangerous. The price is firm only for the quantity available at the moment the order arrives.

This means displayed depth is partly endogenous: it changes in response to the same stress that causes a trader to need it.

In calm markets, selling creates little movement, so market makers can quote tightly. Under stress, price moves increase inventory risk. Quotes widen or disappear. That makes each sale move the price further, which raises measured volatility and forces more participants to reduce exposure.

The exit does not merely become crowded. It becomes smaller as the crowd approaches it.

Software interfaces can fail in a similar way. A dashboard may report that every component is healthy while the dependencies behind it are approaching their limits. The visible interface is not the system. It is a snapshot of promises the system is currently willing to make.

Liquidity has a dependency graph

To understand a market, map the dependencies that allow a trade to happen.

A buyer needs capital. A leveraged buyer needs financing and acceptable collateral. A market maker needs inventory capacity, exchange access, hedging venues, risk limits, and settlement rails. A fund offering redemptions needs assets it can sell on terms compatible with the promise it made to investors.

Those dependencies are often shared.

Different firms may use the same prime broker. Several venues may rely on the same stablecoin, custodian, clearing member, oracle, or cloud region. Strategies that appear independent may respond to the same volatility threshold. Portfolios that hold different assets may have pledged them against the same source of leverage.

The market therefore contains hidden single points of failure.

This matters because diversification at the asset layer does not guarantee diversification at the infrastructure layer. Ten positions funded through one fragile channel are ten expressions of the same dependency. Five exchanges that settle through the same collateral asset are not five independent exits.

Architecture determines whether local stress remains local.

Leverage turns price movement into forced flow

Price declines are not automatically market failures. A functioning market can reprice an asset sharply as information changes.

Fragility appears when price movement triggers transactions that are not based on a new investment view.

A leveraged participant breaches a collateral threshold and must sell. That sale moves the market, weakening someone else’s collateral and triggering another sale. A volatility-targeting strategy reduces exposure as measured risk rises. A market maker reaches an inventory limit and stops absorbing flow. Investors request redemptions, forcing a fund to sell what it can rather than what it would prefer.

The original price move becomes an input to the system. The system then produces more of the flow that caused it.

This is a feedback loop, and feedback loops explain why liquidity can disappear nonlinearly. Conditions do not worsen in a smooth proportion to the first shock. They cross thresholds.

Before the threshold, participants rebalance voluntarily and spreads remain manageable. After it, risk constraints begin making decisions. The relevant question is no longer what an asset is worth over the next five years. It is who must transact in the next five minutes.

Time is part of liquidity

An asset is often called liquid if it can be sold. Given enough time and a large enough discount, almost anything can be sold. That definition is too weak to be useful.

Liquidity always includes a time requirement and an acceptable cost.

A position that can be exited over six months may be perfectly liquid for an unleveraged investor with a ten-year horizon. The same position is illiquid for a fund promising daily redemptions or for a borrower facing an hourly margin check.

Many failures described as asset problems are therefore duration mismatches. The asset settles later than the liability. The portfolio can be sold more slowly than investors can redeem. The collateral can be moved, but not before a liquidation engine acts. The hedge exists on another venue, but capital cannot reach it in time.

This is why the sentence “we can always sell” is operationally incomplete.

At what size? Over what period? Through which venue? At what expected slippage? Under whose risk limit? And what happens if the normal route is unavailable?

Without those constraints, liquidity is a story rather than a capability.

Measure the cost of changing state

If volume and displayed depth are insufficient, what should an operator examine?

Start with the cost of changing the portfolio’s state.

Market impact. How far does price move when the position is reduced at realistic size, rather than at the size used for a screen quote?

Concentration. How much executable capacity depends on the largest venues, counterparties, or liquidity providers?

Persistence. Does depth remain available during volatile periods, or is it present only when nobody needs it?

Funding sensitivity. Which positions become forced sellers when rates, haircuts, collateral values, or volatility thresholds move?

Settlement time. How long does capital take to move between the places where it is held and the places where it is needed?

Exit correlation. Which apparently different participants are likely to seek liquidity at the same time for the same reason?

No single metric captures the whole system. The point is to replace a static label — liquid or illiquid — with a set of testable conditions.

Stress the route, not just the price

Traditional stress tests often begin with a price shock: what happens if the asset falls 20%?

That is necessary, but incomplete. A robust test should also damage the path used to respond.

What happens if the asset falls while the main venue is unavailable? If spreads triple? If collateral haircuts rise? If the fastest settlement rail pauses? If a counterparty cuts limits? If the hedge and the asset diverge precisely when the position must be closed?

These are not exotic additions to the scenario. They are often how stress propagates. A system rarely fails because one variable changed in isolation. It fails because the response depended on components that degraded together.

Engineers call this testing under partial failure. Market participants should think the same way.

Do not ask only whether the portfolio survives a lower mark. Ask whether the operation can still observe, fund, hedge, and exit the position while the market is producing that mark.

The real liquidity question

Liquidity cannot be proven by a calm market. Calm conditions reveal which routes are convenient, not which ones are resilient.

The useful work happens before stress: mapping dependencies, matching the timing of assets and liabilities, limiting concentration, testing alternate routes, and deciding which thresholds require action while action is still optional.

This does not predict the next crisis or tell anyone what to buy or sell. It does something more durable: it separates a position that merely looks tradable from a system designed to remain operable when many participants need the same thing at once.

Volume records activity under normal conditions. Liquidity reveals what can still move when the system is under load.

Systems > Emotions.