# Block 028 — Pre-read teach: FX intervention mechanics **Date:** 2026-08-16 **Block:** 028 (FX intervention mechanics — verbal vs actual, FX reserve mechanics, sterilized vs unsterilized, SNB 2015, BOJ 2022) **Mode:** Pre-emptive teach-first (per macro-mentor SKILL.md "Pre-emptive teach-first for pre-reads" rule) **Source bundle:** - BIS Working Paper #1040 / Fratzscher (2008+ updates) on FX intervention effectiveness — academic substrate for sterilized/unsterilized distinction + signaling channel - FRB-NY "Foreign Exchange Operations" doc (Neely-era or successor) — institutional substrate for actual-intervention mechanics (reserve accounting, sterilization, signalling vs portfolio-balance) - SNB 2015 case (SNB press releases + BIS Quarterly Review March 2015 recap) — primary source for EUR/CHF floor removal - BOJ Sept 2022 case (BOJ press releases + BIS Bulletin recap) — primary source for USD/JPY rate-driven intervention --- ## Source 1 — BIS 1040 / Fratzscher: intervention channels ### The two channels intervention operates through 1. **Signaling channel.** CB intervention works *primarily* by changing expectations about future policy (rate path, balance sheet, capital controls), not by moving the FX level mechanically. The volume of intervention is tiny relative to daily FX turnover — SNB's $50bn over 3 months ≈ $500m/day ≈ 0.1% of daily JPY turnover. Insufficient to absorb flow. The signaling channel is the load-bearing mechanism. 2. **Portfolio-balance channel.** When the CB intervenes *unsterilized*, it changes the relative supply of domestic vs foreign currency assets in private hands, which can move the FX level via risk-premium / term-premium compression. This is the "flow" channel — small relative to signaling in modern EMEs and DM. **The mechanical point:** FX intervention works because the market believes the CB has more information / more commitment than the print reveals. The *announcement* of intervention (or even the leak of intervention) often moves FX more than the actual trades. ### Sterilized vs unsterilized — the accounting distinction - **Sterilized intervention.** CB buys/sells FX and *simultaneously* offsetting domestic open-market operations to leave the monetary base unchanged. Net effect on interest rates = zero. Only the signaling/portfolio-balance channel works. - **Unsterilized intervention.** CB buys/sells FX without offsetting OMOs. Monetary base changes. Net effect on interest rates is non-zero — the FX intervention *is* the monetary policy move. **The load-bearing distinction:** in modern DM (BOJ, SNB, ECB), intervention is *almost always sterilized*. The CB is targeting FX expectations, not monetary conditions. In EM (2000s Brazil, China pre-2015 reserve accumulation), intervention was *unsterilized* — the CB was buying USD to manage the exchange rate as part of the monetary stance. The accounting matters because unsterilized intervention transmits to rates (and thus to carry trades) directly; sterilized intervention transmits only via expectations. **PM read:** the *type* of intervention tells you the CB's intent. Sterilized = CB wants to signal FX view without changing monetary stance (most DM cases). Unsterilized = CB wants to manage the level as part of broader monetary regime (EM, China pre-2015). When you read a press release saying "BOJ conducted FX intervention" — check the OMO schedule the same week. If offsetting OMO announced same day → sterilized. If no OMO → unsterilized. ### Why intervention effectiveness is contested in the literature The empirical literature (Fratzscher 2008, 2012; Menkhoff 2012; Disyatat & Galati 2007) finds highly variable effectiveness across episodes. The reconciliation: intervention works when it (a) augments the global financial cycle (Borio, Block 024 substrate) and (b) the CB's commitment is credible (capacity + intent). When it fights the cycle or commitment is doubted, intervention fails fast (BOJ 2011-2012 was the textbook fail; SNB 2011-2015 was the textbook hold). --- ## Source 2 — FRB-NY FX operations: the mechanics ### Who holds reserves, who intervenes - **Reserve holders.** US Treasury (Exchange Stabilization Fund, $200bn+) and the Federal Reserve System (System Open Market Account, custody) jointly hold official FX reserves. Treasury has policy authority; Fed executes. Foreign CBs hold their own reserves (BOJ ~$1.3T, SNB ~$800bn, PBoC ~$3.2T). - **Intervention execution.** CB announces intent → places orders with dealer banks → buys/sells FX against own currency → reports via press release or weekly data. - **Accounting.** When the Fed buys EUR against USD, SOMA's EUR holdings rise, USD monetary base rises by the offsetting USD amount. Sterilization = the Fed sells Treasury bills the same week to drain the USD. The two legs are bookkept separately; the net monetary base change is zero. ### The US Treasury–Fed split - **Treasury (ESF).** Policy authority for FX. Used historically for emergency intervention (1985 Plaza Accord, 1992 ERM support, 1998 LTCM). Smaller scale. Treasury Secretary leads the messaging. - **Fed (SOMA).** Operational execution. Currency composition of SOMA reflects historic intervention (EUR-heavy from 2000s accumulation). The Fed doesn't formally have FX policy authority but executes Treasury directives. **Why this matters for the trade:** when you see Treasury + Fed joint statements on FX, the policy intent is broader (fiscal + monetary). When the Fed acts alone, the policy intent is narrower (typically dollar funding, not dollar level). The 2022 BOJ case was BOJ-alone; the 2011 SNB case was SNB-alone. ### Coordination vs unilateral — the diagnostic - **Coordinated intervention (Plaza 1985, 1992 ERM).** Multiple CBs acting same direction same week. Signal is stronger than sum of parts — the global cycle is being jointly managed. PM read: coordinated intervention against the global cycle (rare) is high-EV fade if cycle is intact; coordinated intervention with the cycle is a regime change signal. - **Unilateral intervention (BOJ 2022, SNB 2015).** Single CB acting alone. Signal is weaker, more easily reversed, more dependent on credibility of the unilateral CB. --- ## Source 3 — SNB 2015 case study: EUR/CHF floor removal ### What happened - **Sept 2011 → Jan 2015:** SNB held a EUR/CHF floor at 1.20. Defended for ~3.3 years. Cumulative intervention volume: ~$200bn+ EUR reserves accumulated. EUR/CHF traded in a 20bp band. - **Jan 15, 2015 (~09:30 CET):** SNB abruptly removed the floor. EUR/CHF collapsed from 1.20 to ~0.85 in minutes (CHF +30%+). Largest single-day FX move in modern history outside EM crisis episodes. - **Post-removal:** SNB did NOT immediately re-intervene to defend a new floor. Verbal intervention ("will take further measures if needed") but no actual intervention for weeks. ### Why SNB capitulated 1. **Cycle position.** EUR was weakening against USD (global anti-EUR phase). SNB was trying to *fight* the global cycle by pegging CHF to weakening EUR. Reserve costs rising; PPP-1.20 was clearly overvalued. 2. **Real-economy justification eroded.** CHF was only European safe-haven (not global). ECB Q€ eased aggressively from 2014 (ABSPP announcement Jan 2015). CHF safe-haven demand was being driven by EUR weakness, which SNB couldn't fix without holding indefinitely. 3. **Capacity cost.** $200bn+ EUR reserves = ~30% of SNB balance sheet, taking duration risk on EUR periphery bonds at the worst possible time (2011-2013 EUR crisis). Mark-to-market losses were substantial. **Substrate add:** the SNB held for 3.3 years *because* the peg was a real-economy policy (preventing CHF deflation via import price stability), not just an FX target. When the policy justification eroded (ECB Q€), the peg collapsed. This is the same architecture as Block 024 — commitment size moderates holding window; direction of failure set by cycle position. ### The post-mortem PM read - **During the floor (2011-2014):** the right trade was fade CHF strength against the global cycle (Block 024 Borio framework). The floor was unsustainable in months-to-quarters but held in years because of SNB credibility. - **Jan 15, 2015:** the right read was that the SNB's commitment was finite (capacity cost + cycle position) and the trigger was real-economy. The capitulation was *unexpected in date but expected in direction*. - **Post-capitulation:** CHF became the funding currency of choice for carry trades (low rates, safe haven). EUR/CHF traded 1.05-1.15 for years. SNB cut rates further (-0.75%) and eventually intervened *to weaken* CHF against EUR in 2017-2018. --- ## Source 4 — BOJ Sept 2022 case study: USD/JPY rate-driven intervention ### What happened - **2022 Q1-Q3:** USD/JPY rallied from 115 to 145 (~30 yen). BOJ held rates at -0.10% (YCC band 0% ±25bp); Fed hiked 300bp. Carry trade widened dramatically. - **Sept 22, 2022 (~22:00 JST):** MOF (Ministry of Finance, not BOJ alone) announced actual intervention. USD/JPY sold off from 145 to ~140 over 2 days. ~¥2.8T ($20bn) deployed. - **Oct 21, 2022 (~22:00 JST):** second intervention. USD/JPY from ~150 to ~145. - **Oct-Dec 2022:** total intervention ~¥9.2T ($65bn). USD/JPY stayed below 150. - **2023:** no intervention. USD/JPY drifted back to 150+ on cycle. ### Why BOJ's intervention "worked" (unlike 2011-2012) 1. **Cycle position.** USD/JPY 145 was *aligned with* the global cycle (USD strong globally via Fed hikes). Intervention was *fighting* the cycle — but the cycle was already mature. The intervention slowed the move, didn't reverse it. 2. **Capacity, not unlimited.** BOJ intervention was capped (~¥5-10T per episode, then paused). This is the opposite of SNB 2011. Capped intervention fails fast in pure cycle terms; works as a *pause* mechanism. 3. **Real-economy justification.** Japan CPI was starting to rise from the BOJ's -0.10% policy rate; yen weakness imported inflation. BOJ had a domestic policy reason to push back. 4. **Signaling clarity.** MOF (not BOJ) led the messaging. "Speculative, rapid moves" framing. Market read as: the government, not just the central bank, wanted to slow the move. **Substrate add (counter to SNB):** capped intervention fighting the global cycle *with* domestic policy justification can pause the move for weeks-to-months but cannot reverse the cycle. The trade is fade the move to the BOJ floor; don't expect the floor to hold against the cycle. ### The 2024 follow-on - **2024 Q1-Q2:** USD/JPY back to 160+ on persistent USD strength. BOJ intervened again (~¥9.7T in April-May 2024) but the move continued. Eventually BOJ pivoted policy (raised rates from -0.10% to 0% in March 2024, +0.25% in July 2024) — the cycle turned because Japan *domestic policy* pivoted, not because intervention reversed it. **The PM read on this:** intervention can slow, intervention can pause, intervention cannot reverse a global cycle. The only thing that reverses the cycle is the CB's *domestic* policy pivot (Borio framework from Block 024). Intervention buys time for the policy pivot to take effect. --- ## Synthesis — Intervention mechanics for the PM trade ### The decision tree When you see "BOJ/MOF/SNB/Treasury announces FX intervention": 1. **Sterilized or unsterilized?** Check OMO schedule. Unsterilized = CB is moving rates, not just FX. Sterilized = CB is signaling FX view. 2. **Capped or unlimited?** Check commitment size language. Capped = intervention is a pause mechanism (weeks-to-months), not a reversal. Unlimited = intervention is a commitment (years). 3. **Will existing positions profit or lose?** This is the binary that matters for the trade. "Cycle position" is a descriptive label, not an operational input — CBs have multiple goals, the positioning consequence is what determines the trade. 4. **Real-economy justification?** Is there a domestic policy reason (inflation, deflation, competitiveness)? If yes, the CB will hold longer than pure FX view would predict. If no, the CB will capitulate when capacity cost exceeds policy benefit. 5. **Positioning crowded?** Check COT, options skew, dealer surveys. If crowded, Trade 2 (positioning reset) layers on. If uncrowded, only Trade 1 (fade the cycle) fires. ### The trade structures - **Trade 1 — Fade the cycle (Borio-aligned, Block 024 substrate):** when existing positions would lose from the intervention (CB pushing back against cycle) and capacity is capped, fade the move toward the BOJ/SNB floor. Hold for weeks, exit on capitulation. Defined-horizon, not capitulation bet. - **Trade 2 — Trade the positioning reset:** when positioning is crowded in the direction the intervention pushes against, the intervention forces positioning unwind. Fade the immediate overshoot. Hold for 1-4 weeks. Exit when positioning normalizes (visible via COT, options skew, dealer surveys). Three mechanisms: (a) crowded positioning unwind, (b) liquidity-driven technical bounce (stops triggered), (c) covering short hedges. Trade 2 layers on Trade 1 (BOJ 2022 case) or fires alone if capacity is unlimited (SNB 2015 removal case). - **Trade 3 — Capitulation bet:** when intervention creates positioning loss, capacity is unlimited, but real-economy justification erodes (SNB 2011-2015 case). Hold through the pain. Trade structure: small consistent cost (carry, opportunity cost, time decay) during the wait; large payout at capitulation that compensates for the bleed. *Costs small most of the time, pays big at capitulation.* Upgrade path requires (a) capacity remaining finite AND (b) real-economy justification absent. SNB was the canonical outlier that held 3.3 years. ### Trade mapping (common cases) - Positions lose + uncrowded = Trade 1 only (fade the cycle, slow unwind) - Positions lose + crowded = Trade 1 + Trade 2 combined (BOJ 2022) - Positions lose + unlimited + real-economy erosion = Trade 3 (long-horizon, SNB 2011-2015) **Rare cases dropped:** positions profit + crowded (Plaza-style cycle inflection), positions profit + uncrowded (CB helping, no edge). The common case is "positions lose" (CB pushing back against cycle). The framework focuses on the load-bearing case. ### What Block 028 covers that Block 024 didn't - **The mechanics layer** — how intervention actually flows (reserves, sterilization, signaling vs portfolio-balance channel). Block 024 was signal-and-strategy (when to fade, when to bet). Block 028 is mechanics (how the CB does it). - **The two case studies** — SNB 2015 and BOJ 2022. Block 024 used SNB 2011-2015 and BOJ 2011-2012 as case studies for Borio framework. Block 028 uses the *follow-on* episodes (SNB 2015 removal, BOJ 2022 actual intervention) for the *mechanics* substrate. - **The decision tree** — converting intervention announcements into trade structure via 5 questions (sterilized? capped? positions profit/lose? real-economy justification? positioning crowded?). --- ## Skip list - Detailed reserve-management accounting standards (BPM6 / IMF framework). Out of scope; covered in Phase 3 cross-asset context if needed. - EM-specific intervention beyond BOJ/SNB as case studies. The load-bearing substrate is the mechanism, illustrated by 2 cases. - Intervention effectiveness empirical literature beyond the Fratzscher/Menkhoff consensus. The substrate is the mechanism, not the empirical debate. - Detailed history of pre-1990 intervention (Plaza, Louvre). Out of scope; available in academic lit. --- ## Anki card batch (planned — basic, atomic, tag block-028-source-bisfed-snb-boj session:2026-08-16) 1. **Sterilized vs unsterilized intervention** — definition + accounting + DM/EM pattern. 2. **Signaling vs portfolio-balance channel** — why intervention works even when volume is tiny. 3. **FRB-NY Treasury-Fed split** — who has policy authority, who executes. 4. **SNB 2015 floor removal** — what happened, why SNB capitulated, the 3.3 year hold architecture. 5. **BOJ 2022 rate-driven intervention** — what happened, why it worked (vs 2011-2012 fail), the capped-vs-unlimited distinction. 6. **Intervention decision tree** — 5 questions to convert intervention announcement into trade structure (sterilized / capped / positions profit-lose / real-economy / positioning crowded). 7. **Sterilized/unsterilized PM diagnostic** — what to check in OMO schedule when intervention announced. 8. **SNB 2015 vs BOJ 2022 commitment size contrast** — unlimited holds years, capped holds weeks-to-months. 9. **Trade mapping common cases** — positions lose + uncrowded = Trade 1; positions lose + crowded = Trade 1+2; positions lose + unlimited + real-economy erosion = Trade 3. 10. **Capitulation bet upgrade path** — requires capacity finite + real-economy absent; SNB was the outlier. Cost structure: small consistent bleed, large payout at capitulation. **Deck:** Macro Study::Phase 1 — Components::FX **Tag:** block-028 source-bisfed-snb-boj session:2026-08-16 **Sync:** ⏳ pending (will fire after user reviews teach) --- ## Probe queue (deferred to next session open — fires AFTER teach landing) 1. **Sterilized diagnostic applied:** "BOJ announces intervention on Sept 22, 2022. Same week, BOJ holds short-term policy rate at -0.10% and makes no OMO offsetting announcement. Sterilized or unsterilized? What does the choice tell you about BOJ's intent?" 2. **SNB 2015 capitulation mechanism:** "SNB held the EUR/CHF floor for 3.3 years despite fighting the global anti-EUR cycle. What real-economy justification allowed the hold, and what specifically changed in early 2015 to trigger the removal?" 3. **BOJ 2022 capped intervention vs SNB 2011 unlimited:** "Both interventions created positioning loss. SNB held for years, BOJ's 2022 episode lasted weeks-to-months. Was the difference the *capped vs unlimited* commitment, or something else?" 4. **Trade structure applied:** "USD/JPY at 158 in April 2024. BOJ intervenes for second time. Existing USD/JPY carry positions are extremely crowded. Apply the intervention decision tree. Trade structure?" --- ## Pointer Block 028 pre-reads substituted via teach-first per 2026-08-16 user opt-out. ~60 min read-equivalent substrate delivered in compact teach-form. Precision slips flagged for Anki card design. Next session opens with the probe queue above (applied recall probes on the teach, NOT on a read). --- ## Process notes — 2026-08-16 user pushbacks (codified as class-level refinements) ### Pushback 1: Trade 2 framing was sloppy Original framing: "intervention aligned with the global cycle creates a pullback." User pushback: this doesn't make sense logically. The "aligned with cycle" framing was based on one example (BOJ 2022) but doesn't generalize as a rule. **Corrected framing:** Trade 2 is *positioning reset*, not *aligned with cycle*. Intervention creates a tradable pullback when (a) positioning is crowded in the direction of the move AND (b) intervention forces unwind. The pullback is positioning-driven, not fundamentals-driven. Three mechanisms: crowded positioning unwind, liquidity-driven technical bounce (stops triggered), covering short hedges. **Class-level lesson:** don't extrapolate a trade-structure label from a single case. The BOJ 2022 case was Trade 1 + Trade 2 combined, not a pure Trade 2. SNB Jan 2015 was a pure Trade 2 (positioning reset, no Trade 1 cycle component). The label "aligned with cycle" was wrong because the BOJ 2022 case was actually *against* the cycle (USD strong via Fed). ### Pushback 1.5: Trade 2 cycle position — orthogonal, not aligned User follow-up: "if Trade 2 is positioning-driven, then the intervention is no longer aligned with the global cycle but rather against it?" **Corrected framing (further refined):** Trade 2 is *independent* of cycle position. The three trade structures are orthogonal on multiple dimensions: - Q3 (with cycle / against cycle) determines Trade 1 viability - Q5 (positioning crowded) determines Trade 2 viability — orthogonal to Q3 - Q2 (capped / unlimited) + Q4 (real-economy justification) determine Trade 3 viability Trade 2 fires when positioning is crowded, regardless of whether intervention is with-cycle or against-cycle. Examples: - BOJ 2022: against cycle + crowded positioning = Trade 1 + Trade 2 combined - BOJ 2003-2004: with cycle + crowded JPY strength positioning = pure Trade 2 (yen weakened on intervention as crowded longs unwound) - SNB 2011-2014: against cycle + uncrowded positioning = no Trade 2 pullback during hold **Class-level lesson:** don't conflate trade-structure labels with cycle position. The three trades are independent mechanisms that can layer on the same intervention. ### Pushback 1.75: Trade 2 requires *positioning loss*, not just intervention User follow-up (sharp catch): if intervention aligns with cycle and existing positions *profit*, then there's no unwind — positions expand, not unwind. So aligned-with-cycle intervention doesn't cause Trade 2. **Corrected mechanism:** Trade 2 requires *positioning loss* to trigger unwind. This means: - Intervention against cycle (existing positions lose money) + crowded positioning = Trade 1 + Trade 2 combined (BOJ 2022 case) - Cycle inflection + coordinated intervention (positions suddenly reverse) + crowded positioning = Trade 2 only (Plaza 1985 case) - Intervention with cycle, no inflection = no Trade 2. The intervention reinforces, doesn't unwind. Existing positions *expand*. **The corrected Trade 2 trigger:** positioning is crowded + intervention creates loss (either via against-cycle or via cycle inflection). ### Pushback 1.85: Aligned-with-cycle intervention is rare User follow-up: aligned-with-cycle intervention doesn't seem common. Why is it rare? **Reason:** CBs don't intervene to *help* the cycle. They intervene to push back against the cycle when it hurts domestic policy. If the cycle is doing what the CB wants, there's no reason to intervene — let the cycle work. **Decision:** drop the rare aligned-with-cycle case from the framework. The common case is intervention-against-cycle, and the framework should focus on the load-bearing case. ### Pushback 1.95: CB motivations are multi-dimensional, hard to categorize as cycle position User follow-up: CBs have multiple goals (inflation, employment, financial stability, currency, reserves, political constraints). A single intervention can serve multiple goals simultaneously. The "aligned vs against cycle" framing is too simplistic. **Corrected framing:** the operational question is not "is the intervention with or against the cycle?" but "what's the *positioning consequence*?" Will existing positions profit or lose? That's the binary that matters for the trade. **Reframed Q3:** "Will existing positions profit or lose?" (positioning consequence — replaces cycle position label). **The cleaner decision tree:** - Q1: Sterilized or unsterilized? (rate channel) - Q2: Capped or unlimited? (time horizon) - Q3: Will existing positions profit or lose? (positioning consequence) - Q4: Real-economy justification + erosion risk? (Trade 3 viability) - Q5: Positioning crowded? (Trade 2 viability) **Trade mapping (common cases):** - Positions lose + uncrowded = Trade 1 only (fade the cycle, slow unwind) - Positions lose + crowded = Trade 1 + Trade 2 combined (BOJ 2022) - Positions lose + unlimited + real-economy erosion = Trade 3 (long-horizon) **Rare cases dropped:** positions profit + crowded (Plaza-style cycle inflection), positions profit + uncrowded (CB helping, no edge). These are uncommon and complicate the framework without adding load-bearing value. ### Refined Q3 (cycle position) decision tree — REVERTED per pushback 1.95 The Q3a/Q3b/Q3c/Q3d sub-cases are removed. The framework collapses to the common cases: - Positions lose (CB fighting cycle) + uncrowded = Trade 1 - Positions lose (CB fighting cycle) + crowded = Trade 1 + Trade 2 - Positions lose (CB fighting cycle) + unlimited + real-economy erosion = Trade 3 Three cases. Common case is "positions lose" (CB pushing back). The "positions profit" case is rare and dropped from the operational framework. ### Pushback 2: Trade 3 cost/payoff framing was inverted Original framing: "pays small most of the time and pays big at capitulation." User pushback: "would it not be correct to say it costs small most of the time and pays big at capitulation." **Corrected framing:** yes. Capitulation bets have a slow bleed (carry, opportunity cost, time decay) that compounds during the wait. When capitulation fires, the move is large enough to compensate for the entire bleed. The asymmetry is: small consistent *cost*, occasional large *payout*. **Class-level lesson:** the structure of asymmetric bets is *cost small, pay big* — not *pay small, pay big*. The first is the standard vol-selling / insurance-selling structure (bleed slowly, get paid on the rare event). Conflating the two misses the entire reason capitulation bets have positive expectancy: the bleed is bounded, the payout is not. ### Codification Both corrections filed in the trade structures section above with "revised 2026-08-16, post-user-pushback" tags. Future blocks touching intervention, capitulation bets, or positioning-reset trades should anchor against the corrected framing, not the original (sloppy) version. This is the Session 46 contribution to the program-level substrate. The pushback pattern itself — *catch the sloppy logic, get it corrected before it propagates into Anki cards* — is the load-bearing discipline.