Quick answer: counterparty credit risk asks a moving question: if the other party fails before a transaction is finally settled, how much could we lose at that future moment? Unlike a simple loan whose exposure is often close to its outstanding balance, a derivative can move from asset to liability as markets change. Banks therefore model current replacement cost, potential future exposure, legally enforceable netting, collateral, the counterparty’s credit quality, and the possibility that exposure becomes largest precisely when the counterparty becomes weakest.

The counterparty can become more dangerous at the same time the trade becomes more valuable.

Why this belongs in mathematics

Counterparty credit risk combines probability, future-state simulation, max functions, netting algebra, conditional dependence, discounting and stress testing. It also teaches a powerful modelling lesson: the amount at risk is not always visible from today’s balance sheet. A contract with almost zero market value today can produce a large positive exposure after rates, currencies or prices move.

This is why counterparty risk is not simply ordinary borrower credit risk with a different label. The exposure itself is stochastic.

1. Start with replacement cost

Suppose Bank A has an interest-rate swap with Counterparty B. If B defaults today, Bank A asks: what would it cost to replace the economic position in the market?

If the swap is worth +S$8 million to Bank A, the current positive exposure is approximately S$8 million before collateral and netting effects. If the swap is worth −S$8 million, Bank A does not normally have an S$8 million credit exposure to B from that trade, because replacing a liability does not create the same loss from B’s default.

A simplified replacement-cost expression is therefore built around a floor at zero. Under Basel’s Standardised Approach for Counterparty Credit Risk, replacement cost for a simple unmargined netting set is related to the positive part of current value after recognised collateral.

2. But today is not the dangerous horizon

A swap worth almost zero today can become strongly positive next month. An FX forward can move sharply as exchange rates change. An option can gain value as volatility or its underlying price moves. That creates potential future exposure (PFE).

Think of an exposure profile as a family of future distributions. At each future date t, the model generates possible market states, revalues the portfolio, floors negative counterparty exposures appropriately, and asks how large the positive exposure could become.

Two common concepts are:

Expected Exposure (EE_t) — the average positive exposure across modelled future states at date t.
Potential Future Exposure (PFE) — a high percentile of the positive-exposure distribution at a future horizon.

Expected exposure answers an average-style question. PFE asks for a high but plausible exposure boundary. Neither is a maximum possible loss.

3. SA-CCR turns current and future exposure into one regulatory amount

The Basel Standardised Approach for Counterparty Credit Risk uses the structure:

EAD = α × (RC + PFE)

where EAD is exposure at default, RC is replacement cost, PFE is potential future exposure, and α is set to 1.4 in the framework. The formula is intentionally more conservative than simply using today’s mark-to-market exposure.

See the Basel SA-CCR framework and its credit-risk-mitigation treatment.

4. Netting changes the unit of measurement

Suppose Bank A has two derivatives with the same counterparty:

Trade 1 is worth +S$12 million to Bank A.
Trade 2 is worth −S$9 million to Bank A.

If the two trades sit inside a legally enforceable close-out netting agreement, the current net value may be only +S$3 million rather than +S$12 million of gross positive value. On default, the contracts can be terminated and combined into one net amount, subject to the agreement and applicable law.

This is why a netting set is such an important object. The algorithm does not merely sum every positive trade and ignore every negative one. It asks which trades are legally and operationally allowed to collapse into one close-out amount.

Mathematics alone cannot create netting. Legal enforceability is part of the model boundary. Basel requires recognised netting arrangements to be enforceable in relevant jurisdictions and to provide for termination, close-out and netting following default.

5. Collateral reduces exposure—but not always to zero

Variation margin transfers collateral as market values move. If Bank A is owed S$8 million on a derivative and holds S$7.5 million of eligible cash collateral, its current unsecured exposure is much smaller than without collateral.

But collateral introduces new questions:

How often is margin called?
What is the threshold before collateral must move?
How long is the margin period of risk?
Can collateral itself fall in value?
Is the collateral highly correlated with the counterparty?
Is the collateral legally available during default?
Can it be liquidated quickly enough?

A collateralised trade can therefore retain future exposure because market values can move between the last margin exchange and the final close-out of the defaulted portfolio.

6. Wrong-way risk: when protection weakens exactly when it is needed

Wrong-way risk occurs when exposure and counterparty default risk move unfavourably together. Basel distinguishes general wrong-way risk from specific wrong-way risk.

General wrong-way risk arises when counterparty default probabilities are positively related to broad market factors. Specific wrong-way risk is more direct: the nature of a particular transaction causes the exposure to rise as that specific counterparty becomes more likely to default.

Imagine a bank takes shares issued by the counterparty itself as collateral. If the counterparty deteriorates, the collateral may lose value at the same moment default becomes more likely. The hedge and the hazard are tied together.

That is mathematically different from treating exposure and default as independent random variables. The joint distribution matters.

7. CVA converts counterparty credit quality into valuation

Credit Valuation Adjustment (CVA) adjusts a default-risk-free derivative value for the possibility that the counterparty may fail. Conceptually, a simplified unilateral CVA can be thought of as a discounted sum of future expected positive exposures multiplied by incremental default probabilities and loss given default.

That means a derivative can lose value even before the counterparty defaults. If the counterparty’s credit spread widens sharply, the market value of counterparty credit risk can worsen immediately.

The Federal Reserve’s public counterparty-risk guidance notes that during the 2007–09 financial crisis, substantial counterparty losses arose through CVA changes rather than only through completed defaults. Basel now maintains a separate CVA-risk capital framework because the credit spread of a counterparty is itself a market-sensitive source of loss.

8. A simplified expected-exposure example

Suppose a bank simulates a netting set one year ahead and obtains five equally weighted exposure outcomes after recognised netting and collateral:

S$0m, S$2m, S$5m, S$9m and S$14m.

The expected exposure in this tiny teaching sample is:

(0 + 2 + 5 + 9 + 14) / 5 = S$6 million.

If a high-percentile PFE were needed, S$14 million might be closer to the relevant tail region than S$6 million. If the counterparty’s probability of default is also highest in the scenarios producing S$14 million, using the unconditional average exposure could materially understate wrong-way risk.

9. The algorithmic pipeline

Inventory transactions. Identify derivatives, securities-financing trades and other counterparty-sensitive positions.
Assign legal netting sets. Separate trades that cannot legally offset one another.
Revalue today. Compute replacement cost using current market data.
Model future market states. Rates, FX, equities, spreads and volatilities drive future values.
Revalue in each state. Produce future positive-exposure distributions.
Apply collateral mechanics. Margin frequency, thresholds, minimum transfer amounts, haircuts and close-out delays matter.
Calculate regulatory or internal exposure metrics. RC, PFE, EE, EPE or internal-model quantities are generated according to the use case.
Overlay counterparty credit quality. PD, rating, market spread, recovery and concentration information enter.
Test wrong-way risk. Stress exposure and credit quality together rather than separately.
Calculate CVA or related valuation adjustments. Credit deterioration becomes visible before default.
Aggregate by counterparty and group. A bank must know total exposure across desks and products.
Apply limits and margin actions. Measurement without control is incomplete.
Backtest and validate. Compare modelled exposure behaviour with realised collateral calls, close-outs and market changes.

10. Archegos: the useful failure case

The 2021 default of Archegos Capital Management is a powerful public counterexample to the idea that collateral and margin automatically solve counterparty risk. The Federal Reserve reported that the default caused more than US$10 billion of losses across several large banks and highlighted weaknesses in due diligence, counterparty monitoring, margin practices and the ability to respond as the fund’s risk profile changed.

See the Federal Reserve’s Archegos supervisory letter.

The lesson is not “margin failed.” It is more precise: margin terms, transparency, concentration limits, close-out speed and risk sensitivity must remain appropriate as the counterparty evolves.

11. Creative-work lens: why stories focus on the moment everyone discovers everyone else

Films such as Too Big to Fail and Margin Call are not technical sources for counterparty-risk mathematics, but they are useful narrative lenses for one recurring feature of crises: institutions discover that the risk is not contained inside one balance sheet. Contracts create a web of claims, collateral calls, forced sales and confidence effects. The creative work supplies the human tempo; the evidence must still come from balance sheets, legal agreements, exposure data and supervisory records.

12. Failure modes

Gross/net confusion. Offsetting is assumed where legal netting is not enforceable.
Static exposure. Today’s mark-to-market is treated as tomorrow’s maximum exposure.
Margin complacency. Collateral is assumed to eliminate gap risk during close-out.
Wrong-way blindness. Exposure and default probability are simulated independently when they are economically connected.
Counterparty fragmentation. Several desks each see a moderate exposure while the group-wide exposure is large.
Opacity. A leveraged client’s concentration across prime brokers is not visible.
Stale limits. Credit limits do not tighten as market volatility and credit spreads worsen.
Model comfort. A mathematically elegant exposure model is used despite poor collateral, legal or trade data.

13. Diagnostics and falsifiers

What is the largest exposure by counterparty after legal netting—not just by trading desk?
How much does PFE change if volatility doubles?
What happens if collateral cannot be liquidated for several extra days?
Which counterparties share the same concentrated market factor?
Do exposure and credit-spread shocks become positively correlated in stress?
How much exposure is created by contracts that cannot be netted?
Does an independent revaluation reproduce the exposure profile?
Are margin terms still appropriate for the counterparty’s current leverage and concentration?

Suppose someone claims, “The counterparty is fully collateralised, so the credit exposure is zero.” A falsifier is a plausible market move during the margin period of risk that creates new positive exposure before collateral can be collected and the portfolio closed out. If such a path exists, the claim was too strong.

14. Verification and update triggers

reconcile trades, collateral and netting agreements across systems;
independently validate pricing and exposure models;
compare predicted and realised margin movements;
stress concentrated counterparties and common risk factors;
review legal enforceability when jurisdictions or agreements change;
recalibrate after volatility regimes change;
tighten limits when transparency, liquidity or margin quality deteriorates;
reassess models after major counterparty failures reveal new behaviour.

Connections across the Bukit Timah Tutor finance-and-banking algorithms lane

Market risk: VaR, Expected Shortfall and stress testing — the market factors that move derivative exposure.
Yield-curve algorithms — a key valuation input for interest-rate derivatives.
Bank capital models — where counterparty exposure becomes part of capital requirements.
A Formula Can Be Correct and Still Be the Wrong Model — the model-risk discipline underneath exposure measurement.

Research anchors

The deeper lesson

Counterparty risk is the mathematics of a promise whose value changes before the promise ends. The bank must model the contract, the market, the legal netting set, the collateral process and the counterparty at the same time. The strongest system is not the one that produces the smallest exposure number. It is the one that can explain why the exposure is small—and can show what would make it large.

Educational note: This article explains banking mathematics and public risk-management concepts. It is not financial advice, trading advice, legal advice or institution-specific regulatory guidance.

Small Group Tutorials

How Banks Calculate Counterparty Credit Risk: Exposure Profiles, Netting Sets, Collateral, CVA and Wrong-Way Risk