Guide

Central counterparty clearing and variation margin explained

Harbor Capital ran $4.2 billion of USD interest-rate swaps bilaterally under ISDA CSAs with three dealer banks. When the 2-year Treasury yield jumped 85 basis points in six weeks, variation-margin calls from counterparties spiked: the desk posted $118 million of Treasuries and cash in a single week, breaching internal liquidity buffers and forcing an emergency repo line draw. Treasury asked why cleared futures on the same rate view produced smoother cash flows. The answer was central counterparty clearing: one net margin relationship with LCH instead of three bilateral VM ladders, standardized SIMM initial margin, and a published default waterfall. After migrating the IRS book to cleared swaps, peak weekly collateral movement fell from 22% of NAV to 6%, and operational errors on margin calls dropped from nine per quarter to one.

A central counterparty (CCP) inserts itself between buyer and seller through novation, becoming the buyer to every seller and seller to every buyer. It collects initial margin (IM) to cover potential future exposure and variation margin (VM) to settle daily mark-to- market P&L in cash or high-quality collateral. This guide covers novation and netting, IM models (SPAN, SIMM), VM thresholds and collateral eligibility, default waterfalls and mutualized default funds, the Harbor Capital refactor, a technique decision table versus bilateral CSA and exchange-traded futures, pitfalls, and a production checklist.

What a CCP does

Before clearing, Party A trades an interest-rate swap with Bank B. If B defaults, A must replace the hedge and may not recover VM it already posted. A CCP breaks that bilateral chain:

Novation — the trade is replaced by two contracts: A vs CCP and CCP vs B. Economically the same swap; legally the CCP is counterparty to both.
Netting — offsetting positions with the same member net to one obligation. A pay-fixed 10Y and receive-fixed 5Y with the CCP may produce a single net DV01 rather than two gross VM streams.
Margin collection — IM covers tail risk if a member defaults before the next VM cycle; VM settles daily (or intraday in stress) P&L so the CCP never carries large uncollateralized exposure.
Default management — if a member fails, the CCP closes or auctions its portfolio using the defaulter's IM and mutualized default fund before tapping non-defaulters.

Major CCPs include LCH SwapClear and CME for rates, ICE Clear for credit and energy, and OCC for US equity options. Clearing is mandatory for many standardized OTC derivatives under post-crisis rules (US: Dodd-Frank; EU: EMIR). Bilateral trading persists for customized structures, but liquidity in plain-vanilla IRS concentrates at cleared venues.

Initial margin vs variation margin

Both margin types protect the CCP, but they answer different questions:

Variation margin (VM)

VM is mark-to-market settlement. Each day the CCP revalues open positions. If rates rose and your receive-fixed swap lost value, you owe VM to the CCP; the CCP credits the gain to your net account. VM is typically 100% of daily MTM change, subject to:

Thresholds — no VM transfer until exposure exceeds a de minimis amount (often zero for cleared members).
Minimum transfer amounts (MTA) — small daily moves accumulate until they breach MTA, reducing operational noise.
Eligible collateral — cash (USD Fed Funds), Treasuries with haircuts, sometimes agency debt. Non-cash collateral is revalued and may trigger recall if haircuts widen.
Settlement timing — T+0 or T+1 depending on product; intraday VM calls escalate in volatile sessions (March 2020, SVB week).

VM is not a buffer — it is full daily P&L settlement. A fund that is right on direction but wrong on liquidity can be forced to sell assets to meet VM even when mark-to-market will reverse next week.

Initial margin (IM)

IM covers potential future exposure from the time of default until the CCP closes out positions. It is calculated with statistical models, not yesterday's P&L:

SPAN — scenario grid used by CME and ICE for futures and options; scans price/volatility moves and takes worst loss.
ISDA SIMM — Standard Initial Margin Model for cleared and bilateral uncleared swaps under BCBS-IOSCO rules; sensitivity-based with correlation factors across rate, FX, credit, equity buckets.
House margin — CCP may add buffers above regulatory minimum when concentration or illiquidity warrants.

IM is usually segregated in a clearing account and may be portable on member default if documentation allows. VM sits in a settlement account and turns over constantly. Confusing the two is a common treasury error: IM is relatively stable week to week; VM swings with market moves.

Default waterfall and mutualization

When a clearing member defaults, the CCP follows a prescribed default waterfall:

Defaulter's IM and VM collateral on hand.
Defaulter's contribution to the default fund (mutualized pool from all members).
Non-defaulters' additional assessments (capped by rule, rarely invoked in practice for large CCPs).
CCP own capital (skin in the game, typically single-digit percent of default fund).

The waterfall is designed so defaulter resources absorb losses before mutualizing to survivors. Stress tests (CCAR-style for CCPs under CPSS-IOSCO principles) model simultaneous member failure plus market gaps. For investors, the key insight is that cleared exposure is not “risk-free” — tail mutualization exists, but it is orders of magnitude smaller than uncollateralized bilateral exposure pre-2008.

Portability allows a healthy member to transfer a defaulter's client positions without closing them, reducing market disruption. Buy-side firms negotiate direct clearing memberships or use FCM/agent models (futures commission merchant posts margin on their behalf).

Harbor Capital refactor

Harbor's bilateral IRS book had three pain points:

Asymmetric CSA terms — different thresholds, eligible collateral schedules, and dispute windows per dealer; operations reconciled nine margin calls per quarter.
Gross VM spikes — offsetting 10Y pay-fixed with one dealer and receive-fixed with another did not net bilaterally; both sides called VM on rate moves.
Funding drag — posting Treasuries with 2% haircuts across three CSAs tied up more collateral than a single cleared net line.

Migration steps:

Opened direct LCH SwapClear membership via sponsoring FCM with segregated IM account.
Compressed legacy bilateral trades through portfolio compression with dealers (tear-ups and replacements) before reposting as cleared swaps.
Centralized VM funding in a single Treasury repo sleeve against the repo market instead of three bilateral collateral pipelines.
Implemented SIMM sensitivity feeds into the firm's stress-testing engine for IM spikes under parallel and twist scenarios.

Outcomes after two rate cycles: peak weekly VM movement 22% to 6% of sleeve NAV; margin-call disputes 9/quarter to 1; all-in funding cost on posted collateral down 14 bp annually from netted repo lines. Trade-off: clearing fees and IM lock-up rose ~$8M annually, acceptable versus liquidity risk reduction.

Technique decision table

Goal	Prefer	Over
Net rate exposure across tenors and directions	CCP-cleared IRS (SIMM IM + daily VM)	Multiple bilateral swaps with gross VM
Standardized delta with daily liquidity	Exchange-traded SOFR or Treasury futures	Custom OTC swaptions without clearing
Bespoke amortizing notional or embedded options	Bilateral CSA with negotiated IM (SIMM)	Forcing non-standard terms through CCP
Credit exposure to a weak counterparty	CCP novation (dealer becomes agent)	Uncollateralized bilateral derivatives
Minimize IM footprint for short-dated macro hedge	Futures with SPAN (lower IM for liquid contracts)	10Y IRS with full SIMM sensitivities
Operational simplicity for small buy-side	FCM-cleared futures only	Direct CCP membership with full VM treasury

Common pitfalls

Treating IM as VM buffer. IM does not offset daily losses; you still owe full VM on adverse MTM. Treasury must fund both buckets.
Ignoring intraday VM. End-of-day models miss CME/LCH intraday calls during gap moves; liquidity plans must cover same-day wires.
Collateral eligibility drift. Haircuts on long-dated Treasuries widen in stress; VM posted yesterday may be recalled today. Monitor CCP haircut schedules, not just MTM.
Assuming bilateral offsets net. Only CCP netting (or a single counterparty) collapses opposing DV01. Multi-dealer hedges can double VM.
SIMM model updates. ISDA publishes SIMM version bumps that can raise IM 10–30% without market moves. Budget IM capacity with version overlays.
Dispute latency. Bilateral CSA disputes freeze VM while exposure grows; cleared venues have shorter arbitration windows but require pre-agreed valuation agents.

Production checklist

Map every derivatives position to cleared vs bilateral and document CSA terms.
Feed daily CCP statements into treasury with automated reconcile to internal MTM.
Separate IM and VM accounts in liquidity forecasts; stress both independently.
Model SIMM/SPAN IM under ±100 bp parallel and key-rate twists quarterly.
Maintain VM funding sleeve (cash + repo) sized to peak weekly move in last two years.
Track eligible collateral haircuts and concentration limits per CCP schedule.
Negotiate FCM porting rights and default-fund assessment caps before membership.
Run compression cycles annually to collapse stale bilateral trades before clearing.
Document intraday VM escalation contacts and wire cut-off times per venue.
Include CCP mutualized default-fund exposure in enterprise risk disclosures.
Reconcile DV01 between internal risk, dealer marks, and CCP sensitivity reports.
Review clearing fees vs bilateral funding drag when rates or balance sheet changes.

Key takeaways

CCPs novate trades and collect IM for tail risk plus VM for daily MTM — they are not optional plumbing for standardized derivatives.
VM is full P&L settlement; IM is a statistical buffer — treasury must fund both.
Harbor Capital cut peak weekly collateral movement from 22% to 6% of NAV by clearing IRS and netting at LCH.
Default waterfalls mutualize tail risk, but defaulter IM and default funds absorb losses first.
Choose cleared swaps for netting, futures for liquid delta, bilateral only when customization demands it.