# Block 027 — FX carry trades

**Status:** ● closed (Targets 1-2 closed Session 43; Targets 3-4 closed Sessions 44-45)
**Phase:** Phase 1 (FX)

## Topic scope

Carry trade construction (high-yielder funded by low-yielder). Volatility drag. Carry-to-risk ratio. Funding-currency selection. Carry unwind asymmetry.

## Plan

**Pre-read (canonical reference: `_session-state/2026-08-10-block-027-open-resume.md` `pre_reads` field — see below for detail):**

- **Menkhoff, Sarno, Schmeling, Schrimpf (2012) "Carry trades and global FX volatility"** — Journal of Finance (~25 min). The foundational academic reference for carry trade P&L: volatility drag, carry-to-risk ratio, the conditional distribution of carry trade returns. Skim the math derivations; the load-bearing substrate is the empirical decomposition (carry-to-risk ratio < 0.3 sustained = the "compressed-risk-premium" signature).
- **BIS Bulletins #90 (Aquilina et al., August 2024 yen carry unwind) + #124 (2025 state-dependent monetary policy transmission)** — re-read with focus on carry trade CONSTRUCTION (~20 min). Block 026 used these for unwind patterns; Block 027 reframes for funding-currency selection and carry asymmetry.

**Skip:**
- Detailed academic derivations of carry trade returns (the substrate is the application, not the math)
- ECB SNB-specific intervention histories (Block 025 territory)
- Hedge accounting for carry trades (Block 075 cross-asset territory)

**Pre-read compliance (2026-07-17 convention):** at session open, the user says "done" or "yes" and the mentor proceeds without per-source verification. Re-probe only triggered if the probe loop surfaces substrate as the failure cause.

**Delegation option (per 2026-07-27 S32-side convention):** if reading is too much, user can delegate pre-reads to side-session teach. Substrate → one mechanism cold-recall probe at a time → atomic Anki cards at end of each source.

**Targets (provisional, refine at next session open per 2026-07-03 protocol):**
- **Target 1 (likely):** Carry trade construction — high-yielder funded by low-yielder, the funding-currency selection problem, the carry-to-risk ratio as the sizing input
- **Target 2 (likely):** Volatility drag — the math of why carry < risk-adjusted carry, the breakeven vol regime
- **Target 3 (likely):** Carry unwind asymmetry — why positioning unwinds are faster than funding unwinds, the time-to-resolution read
- **Target 4 (likely):** PM implications — funding-currency selection rules, position sizing under carry-to-risk, interaction with cross-currency basis (Block 023) and CB reaction functions (Block 024)

**Carry-forwards from Block 026:** see Block 026 file `## Carry-forward to next block` for the six substrate items. Most load-bearing for Block 027: (1) carry direction locked, (2) carry = interface trade, (3) three-layer alignment diagnostic, (4) unwind = positioning vs funding, (5) trade construction, (6) basis is separate diagnostic.

## Carry-forward to next block

**Targets 1-2 closed (Session 43, 2026-08-12).** Targets 3-4 pending.

**Target 1 — Carry trade construction (closed Working+):**
- Direction locked: borrow low-yield, invest high-yield.
- Funding-currency selection: rate persistence + own-vol + correlation with risk-off. Sovereign risk is a cross-check, not a primary criterion. JPY case: structurally safe-haven AND carry-funding currency = correlated failure mode at the wrong time.
- Sizing input: carry-to-risk ratio, but *positioning is the separate timing input*. Four-cell matrix: (1) carry-to-risk > 0.5 + neutral positioning = take carry, full size; (2) carry-to-risk > 0.5 + crowded positioning = fade the carry, smaller size, tight stop (Aug 2024 setup); (3) carry-to-risk < 0.3 sustained + neutral positioning = compressed risk premium, wait for re-entry; (4) carry-to-risk < 0.3 + crowded positioning = unwind in progress, hold off.
- Positioning proxy: IMM COT non-commercial net short futures on CME for the funding currency.

**Target 2 — Volatility drag (closed Working+):**
- Mechanism: conditional asymmetry (downside tail fatter than upside) + path dependency (timing of loss affects realized Sharpe, not just magnitude) + forced-selling channel (trader-side in positioning unwinds, bank-side in funding unwinds).
- Realized Sharpe gap (e.g., 0.5 naive vs 0.2 realized) = the vol drag from these three channels, not "vol was bigger than expected."
- Breakeven vol regime = forward-looking kill threshold. Carry is profitable only when realized vol < breakeven; above breakeven, it's a vol bet disguised as carry.
- Forward-looking vol read: implied vol (FX options) vs historical vol. Spread widening = market pricing regime shift.
- Aug 2024 mechanism: VIX jumped, carry trade underwater on a single-day move, forced selling compounded JPY strength beyond what vol regime alone predicted.

**Carry-forwards to Targets 3-4:**
- Unwind asymmetry substrate already loaded via Probe 3 teach: positioning unwinds self-stabilize (finite forced sellers, days), funding unwinds require Fed action (infinite constraint on bank balance sheet, weeks-months).
- PM trade construction for each: fade the crowded carry (positioning, short horizon, tight stop, smaller size) vs fade the basis (funding, longer horizon, wide stop, larger size because Fed action is the structural backstop).
- Cross-currency basis as the discriminator: stable = positioning unwind, blow-out = funding unwind.

**Anki cards pushed Session 43:**
- 1786534453742 — fade the basis trade construction (USD/JPY legs, P&L source, entry trigger). Tagged `user-flagged`.
- 1786534817367 — carry-to-risk four-cell matrix (take / fade / wait / hold off). Tagged `user-flagged`.

**Targets 3-4 closed (Sessions 44 + 45):**

**Target 3 — Carry unwind asymmetry (closed Working+ Session 44):**
- Positioning unwind vs funding unwind discriminator locked: time-to-resolution read (days vs weeks-months) + cross-currency basis discriminator (stable vs blow-out).
- PM trade construction: Unwind A = long the carry (FX spot/forward, days-weeks, finite forced sellers); Unwind B = long the basis (cross-currency basis swap, weeks-months, Fed is structural backstop). Different trades on different instruments because the mechanism that resolves them is different.

**Target 4 — PM implications / vol-derived sizing (closed Working+ Session 45 via Castagna smile extension):**
- Three-instrument smile framework (ATM / RR / butterfly) as the vol-side sizing input.
- Three-channel discriminator for smile shifts with butterfly gate (customer-flow absorption / spot-trend break / vol regime shift).
- Customer-type discriminator (real-money / levered / dealer) determines how to read the smile signal.
- Dealer-BS chase dynamic = the mechanism behind negative convexity in unhedged carry. Path-dependency matters more than magnitude.
- Forward-spot smile decomposition: strip carry + butterfly → expected spot path residual = tradeable signal.
- Trade 1 (forward → spot convergence) vs Trade 2 (calendar convergence) for trading the residual.
- Cross-link carry-forward: spot-vs-option-driven COT crowding → Block 026 unwind architecture (to be drilled in Block 030).

**Anki cards pushed Session 45 (14 cards, all tagged `block-027-ext` after re-file):**
- 1786784978217-342: Q1/Q2/Q3 precision-slip atomic Basic cards (6 cards)
- 1786785906691-755: Convexity teach atomic cards (3 cards)
- 1786784978242: Q2 card refactored for butterfly discriminator gate (1 edit)
- 1786787185466-585: Discriminator table + 4 scenario cards (5 cards)

Re-file note (2026-08-15): session originally filed as "Block 028 close" but on-disk block 028 = FX intervention mechanics. Castagna work re-filed as Block 027 Target 4 extension per Block 027 carry-forward ("Castagna chapter on FX vol surface if Target 4 lands on vol-derived sizing"). All card tags corrected from `block-028` to `block-027-ext`.

**Pre-read for Block 028 (assigned at close per 2026-07-03 protocol):**
- BIS Working Paper on FX intervention effectiveness (~20 min) — sterilized/unsterilized distinction, signaling channel.
- FRB-NY FX operations doc or Neely (2008) successor (~15 min) — actual intervention mechanics.
- SNB 2015 case study (SNB press releases + BIS Quarterly Review March 2015 recap) (~15 min) — floor removal case.
- BOJ 2022 case study (BOJ press releases + BIS Bulletin recap) (~10 min) — actual intervention case.
- Total: ~60 min read, ~20 min teach-first substitute.
- Sign-off requested at next session open.