Quant Memo
Core

Trading Costs In A Stressed Market

Bid-ask spreads and market impact don't just widen a little in a stressed market — they can multiply several times over, and a cost model calibrated on calm days will badly understate what a trade actually costs when it matters most.

Prerequisites: When Liquidity Disappears

A cost model built from a year of normal trading days will tell you that selling $10 million of a liquid large-cap stock costs a few basis points of slippage. That number is only true on days that look like the days the model was built from. In a stressed market — a sharp selloff, a liquidity crunch, a day everyone wants to sell the same thing at once — the same trade can cost five or ten times as much, not because the model was wrong, but because the market it was calibrated on no longer exists that day.

Why costs blow up, not just widen

Three things happen together in stress, and they compound rather than add. Bid-ask spreads widen because market makers get less certain about fair value and demand more compensation for holding inventory in a fast-moving market. Available depth at the best price shrinks because everyone pulls resting orders rather than risk getting run over, so the same order size that used to sit comfortably inside the spread now has to walk through several price levels. And market impact — how much your own trading moves the price against you — rises because there are fewer natural buyers on the other side of your sell order; you're increasingly trading against other people trying to do the same thing, not against patient capital. A trade that costs 5 basis points on a normal day can easily cost 30–50 basis points in a genuine liquidity crunch, and the tail is fat: it can be far worse than that for large or illiquid positions.

Transaction costs in a stressed market don't scale linearly from their calm-market level — spreads widen, depth vanishes, and market impact rises simultaneously, so a cost model built on normal days can understate stress-day costs by an order of magnitude.

A worked example

Suppose a strategy's cost model says selling a $5 million position costs 8 basis points, or $4,000, based on trailing 90-day averages. On a day when the market is selling off hard and the stock's typical top-of-book depth has dropped to a fraction of normal, the same order might have to cross several price levels to fill, realistically costing 60–80 basis points — $30,000 to $40,000 — roughly eight to ten times the model's estimate. If the strategy's expected edge on that trade was only 20 basis points, the stress-day cost alone turns a profitable trade into a clear loser before you even consider the position's actual P&L.

What this means in practice

Desks that trade through stress build in a stress multiplier — a deliberately conservative markup applied to normal-day cost estimates when volatility or spread indicators cross a threshold — rather than trusting the calm-market number. They also slow down: breaking an order into smaller pieces and giving up on immediacy is usually cheaper than paying up for a fast fill, unless the position itself is the source of risk and needs to come off regardless of cost. The decision to trade at all, at that cost, has to be weighed explicitly against the cost of not trading — sometimes the stressed-market spread is simply the price of getting risk off, and it's worth paying.

The trap is using a single point-in-time cost estimate for the whole order. Costs during stress are not stable through the life of a large trade — they often get worse as you trade, because your own selling adds to the pressure that's already there, and other participants can see size moving through the market and adjust their quotes accordingly.

Related concepts

Practice in interviews

Further reading

  • Kissell, The Science of Algorithmic Trading and Portfolio Management, ch. 4
ShareTwitterLinkedIn