Central counterparty clearing

A central counterparty (CCP) interposes itself through novation: the original bilateral contract is legally extinguished and replaced by two contracts, buyer-to-CCP and CCP-to-seller. Three consequences follow, and only the first two are the ones people remember.

What does a central counterparty change?

First, netting becomes multilateral, and the compression is dramatic. Bilateral netting can only offset what A owes B against what B owes A; multilateral netting offsets everything against everything. Worked Example 5 computes a five-member book: ₹10,590 crore gross across 15 legs compresses to ₹2,610 crore bilaterally (75.35%) and to ₹740 crore of net payments to the CCP (93.01%), a further 71.65% reduction beyond bilateral netting. Less liquidity to fund, fewer payments to make, fewer places for an operational failure to start.

Second, anonymity and fungibility become possible. Because everyone faces the same counterparty, participants need not assess each other's credit, positions become fungible and can be offset, and an exchange can run an anonymous order book at all.

Third (the part that must be taught explicitly, because it is where confident-wrong answers cluster) the CCP does not make risk disappear. It transforms it: many small diffuse bilateral exposures become one large concentrated exposure to the CCP, mutualised across surviving members and pre-funded through a layered default waterfall:

1. The defaulter's initial margin: posted in advance, sized to the estimated close-out cost of its own portfolio.
2. The defaulter's default-fund contribution.
3. The CCP's own capital ("skin in the game") deliberately placed ahead of surviving members so the CCP's incentives align with prudent margining.
4. Surviving members' mutualised default-fund contributions, pro rata.
5. Assessment rights: the power to call further contributions, up to a capped multiple.
6. Recovery tools (variation-margin gains haircutting, partial tear-up) and beyond them, resolution.

Worked Example 5 walks a ₹1,850 crore close-out loss down that waterfall to the last rupee. The number to carry: a surviving member with no exposure to the defaulter at all still loses ₹162.50 crore, because the loss ran past the defaulter's own resources and into the mutualised pool. That is what "mutualised" means, in numbers.

How do CCPs manage margin and default risk?

Two further mechanics matter. Initial margin versus variation margin: VM is the daily or intraday exchange of the actual mark-to-market change, a transfer of realised value; IM is a buffer against the potential future change over the close-out period, and is what determines how much collateral the system locks up. Procyclicality: because IM models are risk-sensitive, margin calls rise exactly when volatility rises and collateral is scarcest, an amplification channel examined after the March 2020 dash for cash and the 2022 UK gilt episode; anti-procyclicality floors, stress-period weighting and buffers exist to blunt it. The governing standard is CPMI-IOSCO's Principles for Financial Market Infrastructures (2012), whose Principle 4 requires a CCP systemically important in multiple jurisdictions to hold resources covering the default of the two largest participants under extreme but plausible conditions, the "cover 2" standard. India's CCIL is a Qualified Central Counterparty under RBI regulation and publishes its PFMI disclosures.

Why are CCPs everywhere?

CCPs are everywhere because of the G20 Pittsburgh commitment of September 2009 that standardised OTC derivatives be centrally cleared (Title VII of Dodd-Frank in the US, EMIR in the EU) plus uncleared-margin rules that make not clearing expensive. The result is a system genuinely safer against ordinary defaults and genuinely more concentrated against extraordinary ones: a trade the profession made deliberately and should describe honestly.

{
  "skill": "Assembling a defensible market-plumbing statistic (turnover, settlement volumes, offshore dollar credit) without importing somebody else's error",
  "use": "Use AI to locate the exact primary release, table and vintage that answers a plumbing question, and to summarise a long statistical release's methodology notes -- then pull the figure yourself from the named table rather than accepting the model's recollection of it.",
  "tools": ["Research/browsing assistants that can reach bis.org, sec.gov, rbi.org.in and sebi.gov.in", "Document/PDF AI over a downloaded BIS Quarterly Review article or statistical release", "pandas or a spreadsheet for every ratio, share and growth rate you quote"],
  "prompt": "I need <STATISTIC> for <PERIOD>. Do NOT state a number from memory. Instead: name the exact primary release (institution, publication, release date), the table or chart number within it, the reporting scope and any double-counting adjustment, and the units and multiplier. Then tell me which OTHER release of the same institution measures a similar-sounding but different quantity, and how the two differ, so I do not quote one as the other.",
  "verify": "Open the named table and read the figure yourself; match the reporting date, the scope (locational vs consolidated, net-gross vs net-net, obligations vs turnover), the units and the multiplier column; recompute every share, ratio and growth rate from the raw figures rather than quoting a derived number; date-check the vintage against any revision; and cite the release and table so a reviewer can re-verify. Here the model's most dangerous output is a plausible number attached to the WRONG SURVEY -- BIS turnover and BIS settlement obligations are both correct and are not the same thing (Guardrail 1-5).",
  "diy": "You must be able to say which release answers which question before you ask -- turnover from the Triennial, outstandings from the semiannual OTC statistics, obligations from the settlement survey, credit from the global liquidity indicators. That map is what sections 5.1, 5.2 and 5.5 install; without it you cannot audit the model's citation.",
  "market": "US"
}