Skip to content

Total Return Swap Synthetic Exposure

total-return-swap-synthetic-exposuresource

Use when pricing or risk-managing a share-locked equity total return swap: the total return leg with manufactured dividends filtered by the ISDA dividend period, and the funding leg on a benchmark rate plus spread.

Domain
Multi-Asset Derivatives
Version
2.0.0
Reading
6 min
Hands off to
5
Handed off from
5
License
Apache-2.0
Covers2002 ISDA Equity Derivatives Definitions2006 ISDA Definitions (day count)BCBS-IOSCO Uncleared Margin RequirementsPython Standard Library (dataclasses, datetime)

When to Use

Use this skill when engineering, pricing, risk-managing, or accounting for a share-locked equity Total Return Swap (TRS) — synthetic economic exposure to a stock, ETF or basket without holding the physical asset.

It gives you:

  • Total Return Leg — capital return plus manufactured dividends, with dividend eligibility filtered by the ISDA Dividend Period and the Record / Ex / Paid basis actually specified in the Confirmation.
  • Funding Leg — benchmark (SOFR, €STR, SONIA, EURIBOR, TONA, Fed Funds) plus prime broker spread, accrued on the period-reset notional over an ACT/360, ACT/365 or genuine 30/360 Bond Basis day count fraction.
  • Periodic net settlement signed for whichever side you are on, plus a side-signed mark-to-market and synthetic share delta ($\Delta$).
  • ISDA CSA margin requirements — initial, maintenance and variation margin reported as three separate numbers, because initial margin is segregated and cannot fund a variation margin call.

When NOT to Use

  • Fixed-notional TRS. This engine keeps quantity_shares fixed and resets the notional each period. In a fixed-notional structure the share count changes at each reset instead; the funding accrual and the delta both differ. Do not use this engine for one.
  • Path-dependent or intraday MtM. process_reset_period values one complete reset period from two prices. It does not path-simulate, does not accrue daily, and does not model intra-period collateral movements.
  • A legal or tax determination. The engine applies whatever withholding haircut you give it. Whether Section 871(m) applies, at what rate, and under which treaty is a counterparty- and transaction-specific determination — see Common Pitfalls.
  • Regulatory margin calculation of record. The initial_margin_pct default reflects the BCBS-IOSCO standardised schedule for equity (15% of notional). If your CSA prices IM with ISDA SIMM or a bilateral house grid, that number governs, not this one.
  • Non-equity underlyings. Bond and commodity TRS carry accrued-interest, coupon and roll mechanics this engine does not model.

Prerequisites

  • Python 3.10+ (standard library only).
  • The executed Confirmation — specifically: share count vs. notional reset basis, the Dividend Amount basis (Record / Ex / Paid), the dividend pass-through percentage, the day count convention, and the CSA's VM threshold and Minimum Transfer Amount.
  • Benchmark rate fixings (SOFR, €STR, SONIA) and a corporate action feed carrying ex-date, record date, payment date, gross amount and withholding rate.

Workflow

  1. Configure the contract. Instantiate TRSContractConfig with the swap ID, symbol, booked notional, initial reference price, fixed share quantity, benchmark, spread in bps, day count, margin percentages, dividend basis, VM threshold and MTA. config.validate() rejects non-positive quantities and prices, percentages outside [0, 100], and NaN/Inf. config.consistency_warnings() reports the non-fatal booking contradictions — a booked notional that disagrees with quantity_shares × initial_reference_price, or a day count that is not the market convention for the chosen benchmark. Read those warnings; they are the difference between a mis-booked trade and a correct one, and they never block pricing.
  2. Define the reset period and dividends. Build a TRSResetPeriod with the period start/end dates, start/end reference prices, the average benchmark fixing, and every DividendEvent you have — including ones you are unsure belong to the period. The engine filters them; you do not have to pre-filter. Supply record_date whenever the Confirmation specifies the Record Amount basis.
  3. Calculate the total return leg. calculate_total_return_leg() returns (capital_return, net_manufactured_dividends). Only dividends whose relevant date falls in (start_date, end_date] accrue. If the net dividend is smaller than you expected, inspect TRSSettlement.excluded_dividend_ids before assuming a bug — the dividend probably belongs to the adjacent period.
  4. Calculate the funding leg. calculate_funding_leg() accrues period_notional × (benchmark% + spread_bps/100)/100 × day_count_fraction, where period_notional = quantity_shares × start_price. A negative all-in rate produces a negative accrual — that is correct, not a sign bug, and the engine warns when it happens.
  5. Process the settlement. process_reset_period(config, reset, side, collateral_already_posted_usd=0.0) returns a TRSSettlement whose net_cashflow_usd and current_mtm_usd are signed for the side you passed. Positive is an inflow to that party. Do not read the receiver's numbers and negate them mentally — pass PAYER_TOTAL_RETURN and let the engine do it.
  6. Act on margin as three separate numbers. variation_margin_due_usd is the incremental transfer after the VM threshold, the already-posted collateral and the MTA. initial_margin_requirement_usd is a segregated requirement that stands alongside it. Never subtract one from the other to decide what to post.

Common Pitfalls

  • Netting initial margin against a variation margin call. Under the BCBS-IOSCO uncleared margin framework initial margin is exchanged on a gross basis, held so as to protect the posting party, and cannot be re-hypothecated — so it cannot fund a VM call. A $113,750 mark-to-market loss against $150,000 of posted IM is still a $113,750 variation margin call, not a zero call. Variation margin fully collateralises the MtM at a zero threshold, subject only to the Minimum Transfer Amount (capped at €500,000 across IM and VM combined in the EU implementation).
  • Computing 30/360 from a day count. actual_days / 360 is Act/360, not 30/360. The Bond Basis fraction is [360(Y2−Y1) + 30(M2−M1) + (D2−D1)] / 360 with D1 = 31 mapped to 30 and D2 = 31 mapped to 30 when D1 ∈ {30, 31}. Over 1 Jan → 1 Jul the two conventions differ by a full day of interest on the whole notional. Use day_count_fraction_for_dates(start, end, convention); the day-count-only helper raises for 30/360 rather than returning a plausible wrong number.
  • Assuming 15% Section 871(m) withholding. The statutory rate on a dividend equivalent paid to a non-US person is 30% (IRC §871(a)/§881); 15% is a treaty rate that applies only where a treaty and the counterparty's documentation support it. Separately, IRS Notice 2024-44 extended the phase-in so that for transactions issued before 1 January 2027 only delta-one transactions are §871(m) transactions — a TRS tracking its underlying one-for-one is delta-one and is in scope today. Confirm no later notice has moved that date again before relying on it.
  • Accruing every dividend in the list. A dividend accrues to the period only if its relevant date falls inside the Dividend Period. Which date is "relevant" is a Confirmation term: Record Amount, Ex Amount or Paid Amount (2002 ISDA EDD §10.1). A dividend with an ex-date in March and a payment date in April lands in different periods under the EX and PAID bases — booking the wrong basis silently shifts cash between quarters.
  • Mismatched day-count conventions. USD SOFR funding accrues ACT/360; SONIA and TONA accrue ACT/365. Booking a SONIA leg ACT/360 over-accrues interest by roughly 1.4%. The engine warns on any benchmark/day-count mismatch instead of silently correcting it, because a bespoke Confirmation may genuinely depart from convention.
  • Confusing notional reset with fixed share reset. In a share-locked TRS the share count is fixed and the notional resets to shares × start_price each period — so funding interest and margin both scale with the reference price. In a fixed-notional TRS the share count moves instead. This engine implements the former only.
  • Reading a payer's position off the receiver's numbers. The mark-to-market of the two sides is a mirror image; only the funding accrual is a shared gross figure. A model that reports one MtM for both sides tells the short synthetic party it lost money on a day the underlying fell.
  • Neglecting financing drag in backtests. Backtesting a synthetic long without the funding leg (benchmark + spread on the resetting notional) overstates net return and Sharpe — the drag compounds with the reference price, not with the trade-date notional.

Verification

Run the test suite. It covers the ISDA Bond Basis end-of-month rules, dividend period boundary inclusion, the Record/Ex/Paid bases, negative funding rates, side-signed MtM, IM/VM separation, MTA suppression, and input validation:

python -m unittest discover -s skills/total-return-swap-synthetic-exposure/scripts

Check a settlement's warnings and excluded_dividend_ids before signing off a reset — an empty warnings list is part of the expected result, not an optional extra.

Verify it, from the repository root

python -m unittest discover -s skills/total-return-swap-synthetic-exposure/scripts

Hands off to 5

Skills this document names, usually in When NOT to Use, as the owner of a case it excludes.

Options, futures, swaps

Kill switches and exposure limits

Allocation across strategies

Handed off from 5

Skills that name this one as the place a case belongs. The reverse edges of the graph.

Options, futures, swaps