Why Margin Calls Jump in a Crisis: CCP Margin Models Explained

The Call That Arrives at the Worst Possible Moment

Picture a clearinghouse — the quiet plumbing behind almost every derivatives trade. Its job is simple to describe: it steps between buyer and seller so that if one side blows up, the other side still gets paid. To do that, it holds collateral, and when markets get jumpy, it asks for more. Usually that request lands right when everyone is already short of cash. Awkward, to put it mildly.

This is the world of central counterparties, or CCPs. After the 2008 crisis, global leaders agreed that standardised derivatives should be cleared through these entities rather than left in a tangled web of one-to-one deals. The reform worked in important ways. But it also concentrated risk in a small number of nodes, and it made margin — the collateral CCPs collect — a load-bearing wall of the financial system.

Margin, in Two Flavours

Margin comes in two parts. Variation margin settles up yesterday's price moves; it flows back and forth and roughly cancels out across the market. Initial margin is the buffer. It's sized to cover potential losses if a clearing member defaults and its positions have to be closed out — from the last collateral collection to the end of that liquidation. In plain terms, it's the cushion for the messy gap between "something went wrong" and "we've cleaned it up."

That word "potential" is where things get interesting. Nobody knows the future, so the CCP estimates it with a model. And a model built to be sensitive to market risk will, by design, demand more collateral when markets get risky. That's not a bug. It's the point.

Procyclicality: When Safety Measures Amplify the Storm

Here's the tension. Margin that rises as volatility rises is doing its job — but if it rises steeply and suddenly, it can turn a market wobble into a funding scramble. Regulators call this margin procyclicality: the tendency of initial margin to react to market volatility in a way that amplifies stress rather than absorbing it. The Bank of England has described large, unexpected margin changes as a possible propagator of liquidity strain across centrally cleared markets.

March 2020 is the reference point. During the "dash for cash," volatility spiked, initial margin requirements jumped sharply, and some participants faced liquidity needs materially larger than they had planned for. Nobody had done anything wrong. The models were simply doing what they were designed to do, all at once, for everyone.

The Trilemma Nobody Can Fully Escape

Supervisors frame this as a trade-off between three things: coverage (does the margin actually cover the losses?), cost (what does it cost on average across the cycle?), and reactivity (how sharply does it move when volatility changes?). Push down reactivity and you usually weaken coverage or raise the average cost of posting collateral. You cannot optimise all three at once.

The Bank of England's expectations are blunt about this: it would be "reductive" for a CCP to look at procyclicality in isolation. A model isn't good or bad on one dimension alone. It's a set of trade-offs, and the sensible question is where you want to sit.

To limit the worst of the reactivity, UK rules require CCPs to use at least one anti-procyclicality measure. Three common options: hold a buffer of at least 25% above the calculated regulatory initial margin; give at least 25% weight to stressed observations in the lookback window; or ensure margin never falls below what a ten-year historical lookback would produce. Different routes, similar intent.

Why the Tools Didn't Fully Deliver

Many CCPs already had anti-procyclicality tools in place. Then March 2020 happened, and margin models around the world still reacted severely. The Bank of Canada's staff work offers a useful diagnosis: the problem often wasn't the stressed margin level itself, but the weight given to it. A stressed component weighted too lightly barely moves the answer. Calibrate the weight high enough, and the same tool becomes genuinely effective.

There's a blunter critique too, from within the industry. One argument is that focusing on initial margin models misses the bigger picture, because in a stressed market the bulk of the cash demanded comes from variation margin, not initial margin. What looks like a margin-model problem may really be a system-wide liquidity question — one that no single CCP can solve alone.

From SPAN to VaR: Same Risk, New Arithmetic

Margin models are also changing. For decades, exchange-traded derivatives relied on SPAN, a methodology developed in 1988 that assesses risk product by product and was prized for being conservative and legible. Since 2022, major clearinghouses have been migrating to Value-at-Risk (VaR) models, which look at a portfolio as a whole.

VaR asks a specific question: over a set period, at a given confidence level, what's the most we could expect to lose? A 99% one-day VaR of £1 million means that on 99 days out of 100, losses should stay below that. It's a useful shorthand — and a deliberately incomplete one, because it says nothing about how bad things get on the hundredth day. That gap is why Expected Shortfall exists: it averages the losses that exceed the VaR threshold, capturing the tail.

VaR models generally use capital more efficiently and handle complex, hedged portfolios better. They also come with honest costs: they're computationally heavy, harder to explain, harder to allocate across desks, and harder to predict. Under SPAN you could often see roughly where your margin was heading. Under VaR, a change in prices or volatility can shift it immediately. One industry paper's summary lands about where you'd expect: the winners are portfolios that hedge well, and the losers are the ones that don't.

Did the Collateral Move Too?

There's a second channel, easy to overlook. Margin is posted in collateral, and collateral has a price. If the assets you posted fall in value, the CCP asks you to top up — not because risk rose, but because your cushion shrank. In a stressed market, that top-up can arrive alongside an increase in required margin, and both land on the same balance sheet on the same morning.

Research from the Bank for International Settlements notes that margin models based on value-at-risk can understate risk in calm conditions and amplify it in stress, and that a collateral squeeze can force deleveraging and fire sales precisely when the rest of the system is under strain. The network can be "robust-yet-fragile": strong against small shocks, less so against large ones.

Where Crypto Exchanges Enter

Crypto venues face a version of the same problem with less of a safety net. Crypto assets are highly volatile, show no obvious stable correlation to traditional asset classes or economic variables, and respond to events — cyber incidents, social media waves — that barely register in conventional risk models. Analysis of Bitcoin pricing has found that a position could lose as much as 3.5% of its value in a single hour at the 99th percentile. That's the kind of number that makes a daily risk report feel quaint.

The practical implication is that crypto margin needs intraday monitoring, not just an end-of-day calculation. It also needs shorter lookback windows and more sensitivity to regime shifts, because correlations between crypto assets drift over time and can dislocate sharply in stress. Traditional portfolio risk methods still apply; they just need faster clocks and more scepticism about the stability of relationships.

One more complication: newer tokens often lack enough price history to build meaningful historical scenarios at all, which pushes risk teams toward proxies and benchmark-adjusted estimates — reasonable tools, but ones that carry their own uncertainty.

What to Take From All This

Margin is not a punishment. It's insurance, priced and collected in real time, and the people setting it are managing a genuine trade-off between protecting the system and not strangling it. The remaining questions are honest ones. How much procyclicality is unavoidable? Who is responsible for the liquidity strains that margin calls create — the CCP, the clearing member, the client, or the system as a whole? Rules differ by jurisdiction, and firms operating across borders should expect different answers in London, Brussels and Washington. This is education, not advice: your own collateral position depends on your portfolio, your venue and your regulator. But understanding the machinery is the first step to not being surprised by it.