Module 9 · Advanced
Wrong-way risk — when exposure and default arrive together
A naive credit charge multiplies the counterparty's default probability by your expected exposure to them — implicitly assuming the two are independent. Wrong-way risk (WWR) is the failure of that assumption in the bad direction: your exposure is largest precisely in the states of the world where the counterparty is most likely to default. The number that matters is not expected exposure, but expected exposure conditional on default — and under WWR it can be a multiple of the unconditional figure.
The standard taxonomy splits it in two:
- General WWR— a macro correlation. The same factor drives your exposure and their credit: an emerging-market bank selling you dollar protection against its own currency's collapse, an oil producer short oil swaps, a leveraged fund whose solvency and whose positions both ride the same market.
- Specific WWR — a structural or legal link. The textbook cases: a counterparty posting its own (or an affiliate's) securities as collateral; buying put protection on a bank from that bank; and the canonical modern one — a crypto fund posting a token as collateral against a long position in a correlated token. The collateral evaporates in exactly the scenario where the exposure spikes.
The 2008 monolines are the definitive case study: banks bought CDS protection on mortgage risk from insurers whose entire balance sheet wasmortgage risk. The protection was worth the most at the precise moment the protection sellers couldn't pay. Exposure and default weren't just correlated — they were the same trade.
🎛 Wrong-way scenario lab
400 scenarios: where do the defaults land relative to your exposure?
Gray dots: scenarios where the counterparty survives. Red dots: defaults. With positive correlation, the red dots crowd the upper-left — default arrives precisely when they owe you the most.
Expected exposure
$789,379
Exposure | default (19 defs)
$3,074,530
WWR multiplier
3.89×
A naive credit charge is PD × expected exposure — implicitly assuming default and exposure are independent. The number that matters is expected exposure givendefault. Their ratio is the wrong-way multiplier: push correlation to +90% and watch it climb well above 1; flip it negative (right-way risk — the counterparty prospers exactly when they owe you) and it drops below. Basel's blunt version of this multiplier is the α = 1.4 factor on EAD. Educational tool — not investment advice.
How it's actually handled
- Regulatory capital — Basel's blunt instrument is the α = 1.4 multiplier on exposure at default, a standing tax for the correlation the models don't capture; identified specific WWR trades must instead be measured with the bad scenario hard-wired in.
- Pricing — CVA done properly correlates the exposure simulation with the default intensity, so wrong-way trades price wider.
- Margin — a prime broker charges a WWR add-on (Module 4) and, for specific WWR, acts on the collateral schedule directly: punitive haircuts on correlated collateral, or refusing it outright. No haircut fixes collateral that is the same bet as the exposure.
- Structuring — the cheapest fix is at inception: demand uncorrelated collateral, cap the correlated concentration, or route the trade to a counterparty on the other side of the factor (turning wrong-way into right-way).
Things to try
- • Correlation 0%: red dots scatter evenly and the multiplier hovers near 1 — the naive PD × EE charge is honest.
- • Correlation +90%: the defaults crowd the upper-left of the scatter and the multiplier climbs well past Basel's 1.4 — the flat α is a floor, not a truth.
- • Correlation −60% (right-way risk): defaults land where exposure is near zero, the multiplier drops below 1 — this is why banks like collecting premium from counterparties who prosper in the states where they owe.
- • Raise PD to 20% and resample: more red dots, tighter estimate — conditional statistics need defaults to condition on, which is exactly why real-world WWR measurement is hard.
Test yourself
A fund posts token X as collateral against a leveraged long in token Y, where X and Y have 0.9 correlation. Trace the default scenario end to end: where does the wrong-way risk appear, and why does no ordinary haircut on X fix it? (Y falls → the exposure spikes and the fund is impaired → but X has fallen with Y, so the collateral covering the spike has evaporated at the same time. The haircut would need to approach 100% — at which point X isn't collateral, it's decoration.)
End of the track
The full arc: the default-and-liquidate question, VM and IM on the default timeline, the scenario grids, the add-ons that patch their blind spots, cross-margining's correlation traps, digital assets, the engine, the backtest — and finally the risk that correlates against you. Natural next stops: VaR & expected shortfall, CVA, ISDA SIMM and repo & secured funding in the Fixed Income track.