How Exchanges Decide Your Margin: SPAN, VaR and Stress Tests

Start with the pizza rule

Imagine you order a pizza with friends. Nobody wants to pay the whole bill up front, so you all put down a deposit. Then someone adds extra jalapeños, someone else adds double cheese, and the delivery guy is late. The restaurant wants a little more money on the table, just in case. That, in essence, is margin.

Margin is the good-faith deposit a trader posts to show they can cover potential losses on a futures or foreign-exchange (FX) position. It is not a down payment on the asset. It is a cushion. The interesting question — and the one this article is about — is how an exchange or clearing house works out the size of that cushion.

Three broad approaches dominate: SPAN, value-at-risk (VaR) models, and stress-based or portfolio margining. They are not rivals in a boxing match. They are different tools for different jobs, and many clearing houses use more than one.

SPAN: the grandparent of margin maths

SPAN stands for the Standard Portfolio Analysis of Risk. It was designed and implemented by the Chicago Mercantile Exchange (CME) in 1988, and it was the first portfolio margin calculation methodology. It is currently used by around 50 exchanges, clearing houses, calculation service providers and market regulators worldwide, and the methodology is continuously developed to support a broad range of products on both derivatives and cash markets. It even travels with a little family of software products, including PC-SPAN and SPAN Risk Manager.

SPAN’s logic is disarmingly straightforward. Rather than asking “what might this entire book lose in a terrible week?”, it typically asks: what would happen to this portfolio under a set of specific, stylised market moves? Those moves are called scenarios. The clearing house runs the scenarios, looks at the worst plausible outcome for the account, and that becomes the margin requirement.

Think of it as a driving instructor testing you on a fixed list of tricky junctions: wet roundabout, blind bend, sudden braking. It is rigorous, it is comparable across firms, and everybody knows the test in advance. The trade-off is that a fixed list of scenarios may not capture every kind of unpleasant surprise.

VaR: asking the model what usually goes wrong

Value-at-risk takes a different angle. Instead of a fixed set of scenarios, it uses historical market data to estimate a probability distribution of potential losses, then sets margin against a chosen point on that distribution.

LCH provides a clean illustration. Its proprietary portfolio methodology, called PAIRS — the Portfolio Approach to Interest Rate Scenarios — is used to calculate margin for OTC interest rate derivatives, foreign exchange derivatives and listed rate derivatives cleared at LCH, across SwapClear, Listed Rates and ForexClear. PAIRS is an expected shortfall VaR model based on filtered historical simulation incorporating volatility scaling. It uses ten years of historical market data to simulate changes in portfolio value, and from those simulations it estimates the potential loss distribution.

That is a mouthful, so here is the chai-shop version. Instead of testing you on a fixed set of junctions, the instructor pulls up ten years of dashcam footage, notices that you drive faster when it rains, and designs the test around your actual habits. Filtered historical simulation with volatility scaling is a formal way of saying: recent market turbulence gets weighted more heavily than a quiet patch from years ago.

Even so, no model is complete. LCH applies margin add-ons covering credit risk and liquidity risk where a particular member’s inherent risk exposure is not captured within the PAIRS model. That is the honest admission every good risk framework makes: the maths has edges.

Portfolio margining: the net-risk view

Portfolio margining aligns the amount of collateral required in a customer’s account with the risk of the portfolio as a whole. It first became available on a limited basis in July 2005 and was expanded in April 2007.

The contrast with older practice is the whole point. Under longstanding strategy-based margining, requirements are set by formula for a single option position or a recognised strategy such as a spread — regardless of what other offsetting positions sit in the account, and regardless of potential market moves. Long a put and long the underlying stock? Each position is margined separately, with no recognition that one gains value when the other falls.

Portfolio margining does the opposite. It is calculated by simulating market moves up and down and accounting for offsets between and among all products held in the account that are highly correlated. Cboe gives the example that options on the S&P 500 Index (SPX) receive offset with options on the SPDR S&P 500 ETF Trust (SPY) and options on the SPDR Dow Jones Industrial Average ETF Trust (DIA). The framework, developed by The Options Clearing Corporation, is known as the TIMS system — the Theoretical Intermarket Margin System — and it prescribes how profits and losses are calculated, what price shocks to assume, and how offsets are applied. The practical effect: margin calculated on net risk is generally lower than under position- or strategy-based methods, which lets a customer make more efficient use of capital.

Lower margin is not free money, of course. It means your cushion is thinner relative to gross positions, which is exactly why the offsets must be genuinely correlated.

Where stress comes in — and where it gets messy

The shift from SPAN to VaR-based portfolio margining has been the big story in cleared energy contracts. Clearing firms are finding it difficult to decompose and explain the outputs of VaR models used to set margin requirements for cleared energy contracts. The transition from SPAN to VaR-based portfolio margining — CME made the switch in 2023, with ICE following suit in 2025 — has materially reduced margin volatility during periods of market stress.

Sit with that for a second. Lower margin volatility during stress sounds like good news, and in many ways it is: fewer violent margin calls whipsawing firms at the worst possible moment. But it also means the models are doing more work, and when a firm cannot easily explain why a number moved, trust and governance start to strain. A clearing house that cannot explain its own output is like a chef who will not tell you what is in the sauce. It may be delicious. You would still like to know.

This is the central tension in modern margining. SPAN is transparent and comparable but may miss novel shocks. VaR is statistically richer but opaque. Stress-based portfolio margining captures net risk elegantly but depends on correlations holding — and correlations have a habit of misbehaving precisely when everyone needs them most.

Why the margin framework around you matters

Margin requirements do not float in a vacuum. They sit on top of a capital and segregation architecture that determines what happens if something goes badly wrong.

In the United States, each futures commission merchant — an FCM, the firm that holds your futures account — must maintain adjusted net capital of at least $1,000,000, or a greater amount determined under CFTC Regulation 1.17(a)(1)(i). Introducing brokers not operating under a guarantee agreement must maintain at least $45,000. An FCM that is also a registered swap dealer faces a higher floor of $20,000,000, and one approved to use internal models to compute market and credit risk charges for uncleared swaps must maintain net capital of at least $100 million.

Client money is treated separately. All customer funds for trading on designated contract markets must be kept apart, or segregated, from the FCM’s own funds, including cash deposits and any securities or other property deposited to margin or guarantee futures trading. Segregated accounts must be titled for the benefit of the FCM’s customers, and acknowledgements must preclude a bank or clearinghouse from recognising a right of offset against the account for the FCM’s debts. Customer funds in segregation have a bankruptcy preference if the FCM becomes insolvent.

There are also moving parts on the swaps side. The CFTC has amended margin requirements for uncleared swaps for swap dealers and major swap participants not subject to prudential regulator margin rules. The amendments revised the definition of “margin affiliate” so that certain collective investment vehicles receiving start-up capital from a sponsor — seeded funds — will not have margin affiliates when calculating the thresholds that trigger the requirement to exchange initial margin, for up to three years after the asset manager begins investing on behalf of the fund. They also expanded eligible initial margin collateral by removing a disqualification on securities issued by certain pooled investment funds, and adopted specific percentage haircuts for money market and similar funds.

Two cautions worth stating plainly. First, rules differ by jurisdiction and change over time; what applies in one market may not apply in another, and nothing here should be read as current for your situation on any given date. Second, brokerage firms may require customers to post higher margins than the regulatory minimums. The number you see in your account is not necessarily the number the exchange calculated.

The takeaway in one breath

Margin is a cushion built from scenarios. SPAN tests a fixed set of them. VaR asks a statistical model what could go wrong, weighted toward recent history. Portfolio margining nets correlated positions to reflect genuine combined risk. Each approach trades transparency against sophistication, and each is only as good as the assumptions underneath it. Understanding which framework sits behind your margin is not trivia — it is the difference between knowing why your cushion changed and merely being told that it did.