Wrong-Way Risk & Central Clearing

You’ve spent this whole course building one number: exposure — how much you’d lose if your counterparty vanished. You learned to net it down, post collateral against it, and price the chance it goes bad as CVA, then the whole XVA family that follows. But there’s been a quiet, load-bearing assumption hiding under all of it: that how much you’re owed and how likely they are to default are independent — two unrelated dice. This lesson breaks that assumption in the worst possible way (wrong-way risk), and then walks through the post-2008 plumbing the regulators built so that one bad counterparty can’t take the system down with it: central clearing, the default waterfall, margin, and the capital rules. This is the last content lesson — the place where everything you’ve measured meets the machinery that contains it.

Wrong-way risk: exposure and default, correlated

Analogy. Imagine you lend your umbrella to a friend who promises to return it the moment it rains. Fine — except your friend is made of sugar. The instant the rain starts (exactly when you need the umbrella most), your friend dissolves. That’s wrong-way risk: the thing that makes you need the payout is the same thing that makes the payer unable to deliver it. The exposure peaks precisely when the counterparty fails.

The definition. Wrong-way risk (WWR) is an adverse correlation between your exposure to a counterparty and that counterparty’s probability of default — they rise together. The CVA formula you built earlier multiplies exposure (EE) by default probability (PD) and quietly treats them as independent:

$\text{CVA} \approx \text{LGD} \times \sum_t \text{EE}(t) \times \text{PD}(t)$

When exposure and PD are positively correlated, that product is too small — you should be weighting the big-exposure states by the higher default probabilities that actually accompany them. Independence understates the charge; WWR is the correction. The two flavors:

• General WWR — driven by macro factors that move many things at once. The emerging-market FX forward from the pretest is the textbook case: a sovereign/currency shock simultaneously inflates the dollars you’re owed and guts the local bank’s solvency. No direct legal link between the trade and the counterparty — just a shared macro driver.
• Specific WWR — a direct, structural link between the exposure and the counterparty’s own credit. Classic examples: buying CDS protection on a company from a counterparty closely tied to that company (you’re protected against a default that would also sink your protection seller), or taking the counterparty’s own shares as collateral (the collateral evaporates exactly as the counterparty fails).

Worked example. You buy $10m of CDS protection on Company X from Dealer D, and D is X’s parent. In calm times your expected exposure to D is small — say the protection is worth $0.5m mark-to-market, PD around 1%. Naive CVA on that leg ≈ LGD × EE × PD = 0.6 × $0.5m × 1% = $3,000. But these aren’t independent. If X defaults, your protection is suddenly worth its full $10m payout and D — X’s parent — is in deep trouble, with PD spiking to, say, 40% in that state. Weight the $10m exposure by that 40%: 0.6 × $10m × 40% = $2.4m of risk in the joint bad state alone. The independent number missed it by three orders of magnitude. The correlation is the whole risk.

Right-way risk: the benign cousin

Analogy. Same umbrella, but now your friend is a duck. When it rains, the duck is delighted and perfectly capable of waddling the umbrella back to you. The thing that triggers your need (rain) makes the counterparty more reliable, not less. That’s right-way risk — the friendly mirror image.

The definition. Right-way risk (RWR) is a favorable correlation: your exposure falls as the counterparty’s default probability rises (or, equivalently, your exposure is largest exactly when they’re healthiest). It reduces the CVA charge, because the big-exposure states coincide with low PD.

Worked example. A copper-mining company hedges by selling copper forward to you at $9,000/tonne. Two states at maturity:

When copper crashes, your forward is deep in-the-money (high exposure) — but a copper crash is good for a company that pre-sold at $9,000, so their default probability is low exactly when you’re most exposed. Exposure up, PD down: right-way. Your effective CVA on this trade is lower than the independence assumption would suggest. Some structures are deliberately built to offset this way — it’s one reason commodity producers are attractive hedging counterparties.

Central clearing: a CCP steps into the middle

The definition. At the 2009 Pittsburgh G20, the world’s regulators agreed that standardised OTC derivatives must be centrally cleared. The mechanism is a central counterparty (CCP) — a clearing house that, through a legal process called novation, becomes the buyer to every seller and the seller to every buyer. Your original bilateral trade with Dealer D is torn up and replaced by two trades: you ↔ CCP and CCP ↔ D. You no longer face D’s credit at all; you face the default-remote CCP.

Three things fall out of standing in the middle:

• Novation — the CCP legally substitutes itself as counterparty to both sides, so no member faces another member directly.
• Mutualisation — if a member defaults and blows through its own resources, the loss is shared across the surviving members’ pooled contributions. Risk is socialised, not left to detonate on one unlucky bilateral counterparty.
• Multilateral netting — instead of netting only within each bilateral pair, the CCP nets across all members and all trades. If you’re +$50m to the CCP on one trade and −$30m on another, you face a single $20m net — and offsetting positions across the whole membership collapse gross notionals dramatically.

Margin at a CCP: variation and initial

The definition. A CCP defends itself with two layers of cash/collateral:

• Variation margin (VM) — collected (and paid out) daily, often intraday, to settle the current mark-to-market. If your cleared swap moved $2m against you today, you wire $2m of VM today. VM keeps the CCP’s exposure to live moves close to zero — losses don’t accumulate.
• Initial margin (IM) — a forward-looking buffer posted up front, sized to cover the potential future loss over the margin period of risk (MPoR): the window between a member defaulting and the CCP fully closing out or auctioning their portfolio. For cleared trades the MPoR is short — typically around 5 days; for uncleared bilateral trades it’s roughly 10 days (you’re slower to unwind a bespoke book). IM is segregated — held apart from the CCP’s own money so it survives the CCP’s own troubles and is returned if you don’t default.

Worked example. You clear a swap. Day 1 it moves $2m against you → you post $2m VM, settling the mark to zero. Separately, the CCP holds, say, $4m of IM against your portfolio, calibrated to a 5-day, high-confidence (e.g. 99%) potential loss. If you default, the CCP has the $4m IM cushion to absorb adverse moves while it spends those ~5 days hedging and auctioning your book — before it ever has to reach into anyone else’s money. VM keeps the running tab at zero; IM covers the messy gap on the way out.

The default waterfall

The definition. When a clearing member defaults, the CCP absorbs the losses through an ordered sequence of resources called the default waterfall. Each layer must be fully exhausted before the next is touched. The canonical order:

1. The defaulter’s own initial margin — their posted IM is consumed first.
2. The defaulter’s default-fund contribution — their own slice of the mutualised pool goes next. (Layers 1–2 = the “defaulter pays” principle.)
3. The CCP’s own “skin-in-the-game” (SITG) capital — a tranche of the clearing house’s own equity, burned before surviving members. This is deliberate: it forces the CCP to actually care about the risk it clears.
4. The mutualised default fund — the surviving members’ pooled contributions. Only now is loss socialised across the innocent.
5. Further assessments / recovery tools — if even the fund is exhausted: powers to call additional contributions from members, variation-margin gains haircutting, and other end-of-the-line recovery measures.

Worked example. Member M defaults owing $120m of close-out losses. The waterfall pays in order: M’s IM $50m (→ $70m left), then M’s default-fund contribution $10m (→ $60m), then the CCP’s SITG $15m (→ $45m), then $45m from the mutualised fund. The survivors only ever feel the final $45m — and only after M’s money and the CCP’s own capital were both wiped out first.

Uncleared margin rules (UMR) & SA-CCR

The definition. Not everything can clear — bespoke and illiquid trades stay bilateral — so regulators extended margining to them via the Uncleared Margin Rules (UMR). UMR phased in (by counterparty size) mandatory exchange of:

• Variation margin — daily settlement of the mark, just like a CCP demands.
• Initial margin — two-way IM for larger counterparties, commonly computed with the industry-standard ISDA SIMM (Standard Initial Margin Model) and calibrated to the longer ~10-day MPoR of uncleared trades.

The result: the bilateral world now looks much more like the cleared world — collateralised on both sides — which is exactly why CVA/FVA shrank across the board, cleared or not.

SA-CCR — the capital measure. Margin protects against default; capital is the regulatory cushion the bank must hold anyway. The current standard for measuring counterparty exposure for capital is SA-CCR (Standardised Approach for Counterparty Credit Risk). It computes the Exposure-at-Default (EAD) that feeds capital:

$\text{EAD} = \alpha \times (\text{RC} + \text{PFE}), \qquad \alpha = 1.4$

where RC is the replacement cost (what it’d cost to replace the trade today — essentially current exposure net of collateral), PFE is the potential future exposure add-on (a forward-looking buffer for how much exposure could grow), and the \(\alpha = 1.4\) multiplier is a conservatism/correlation scaler baked in by the regulators.

Worked example. Take a netting set with RC = $2m and a PFE add-on = $3m. Then:

$\text{EAD} = 1.4 \times (\$2\text{m} + \$3\text{m}) = 1.4 \times \$5\text{m} = \$7\text{m}.$

That $7m EAD is the number capital is charged on — and the cost of holding capital against it is exactly the KVA the previous lesson priced. Margin shrinks RC (less current exposure) and PFE, but the \(\alpha = 1.4\) floor means even a well-collateralised book carries capital weight. The loop closes: exposure → EAD → capital → KVA.

The big picture: risk transformed, not erased

Analogy. Central clearing is like replacing a tangled spaghetti of personal IOUs between everyone at a poker table with a single, well-capitalised house bank that everyone settles through. It’s a genuine improvement — no more chasing the guy who left early, no more one person’s bust taking down three others. But you didn’t make the risk disappear. You concentrated it in the house (now the house itself is too-big-to-fail), you turned “who owes whom” into “everyone must keep cash on deposit at all times” (a funding cost), and you still can’t fix the fact that some players go broke precisely when their bets go bad (wrong-way risk lives on).

The honest summary. Clearing swaps bilateral default risk for three new things: margin-funding cost (all that posted VM and IM has to be funded — hello MVA), CCP concentration risk (the hub is now a systemic single point of failure that must itself be impeccably managed), and the residual that no plumbing removes — wrong-way risk, which still lurks wherever exposure and default share a driver. Counterparty risk isn’t a problem you solve; it’s a problem you measure, price, collateralise, and contain. That’s the entire course in one sentence: exposure → netting and collateral → CVA and the XVA family → and the market structure (clearing, margin, capital) that wraps around all of it.

That’s the last piece of content in the course. You can now build exposure from a trade, net it, collateralise it, price its credit cost as CVA and its cousins across the whole XVA family, spot the wrong-way correlation that makes the naive number lie, and read the post-2008 market structure — clearing, margin, the default waterfall, UMR, and SA-CCR capital — that contains it. Next up is the final exam: one question at a time, no going back, no retries. Bring everything.