## 1eurea2025008

## Source details

**Canonical URL:** [1eurea2025008](https://www.imf.org/-/media/files/publications/cr/2025/english/1eurea2025008.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2025/english/1eurea2025008.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2025/english/1eurea2025008.pdf.json)

---

### CCP counterparty credit risk on reverse repo
- Coverage and data:
  - Analysis covers 16 CCPs authorized to clear in the EA regardless of their domicile, using data with reference date December 11, 2024.
  - Data sources: EU SFTR trade-state tables and CCP supervisory reports; PD time series from National University of Singapore, Credit Research Initiative (NUS CRI).
  - Sample limitations:
    - Of 14 EU CCPs, only seven reported reverse repos (mostly in non-EUR currencies: USD, GBP, and other EU currencies).
    - EU CCPs have at most eight different counterparties; third-country Tier 2 CCPs have up to 54 counterparties.
    - SFTR sample covers trades reported by EU firms; for EU CCPs IMF/ESMA obtained 100 percent coverage except one EU CCP excluded for data quality.
    - For Tier 2 CCPs, after cleaning for outliers, included trades represent no more than 29 percent of Tier 2 CCPs’ reverse repo positions—making loss estimates a lower bound.
- Methodology:
  - Expected Shortfall (ES) combining historical simulations (January 2020–December 2024) over a five-day holding period to capture time-varying exposures and market dependencies across tenors and currencies.
  - For each day compute conditional loss per CCP–counterparty as difference between discounted value of collateral and cash at maturity; netting allowed across positions of same counterparty but not across counterparties.
  - Loss given default after collateral value assumed to be 1.
  - ES reported at α = 0.001.
- Key empirical findings:
  - Expected shortfall relative to total cash investment: in most cases modest, up to 15 basis points, with some CCPs experiencing no losses.
  - Losses concentrated in a few CCPs characterized by relatively smaller SITG compared to DF; sample SITG / DF ranges between 0.3 and 4.5 percent.
  - Potential severity relative to SITG: in some cases losses could consume a significant part of own resources, with one case reaching 120 percent of SITG.
  - NDL-cover-2 (sum of two largest conditional losses per day) could reach up to 13 percent of the default fund, with three CCPs exceeding 7 percent.
- Supervisory considerations:
  - Assessment compares potential losses to skin-in-the-game (SITG) and default fund (DF); NDL are normally absorbed by equity but CCP equity data unavailable so SITG used as proxy.
  - Policy recommendation: ESMA should monitor non-default loss risk against CCPs’ resources and close data gaps for Tier 2 CCPs.

### Collateral revaluation methodology (duration and convexity approximation)
- Collateral revaluation:
  - Use duration (D) and convexity (s) at cut-off date T; map each bond to a bond index by issuer country and tenor to obtain yield time series.
- Notation and formulas:
  - VL(t) = value in denomination currency of a bond on day t.
  - Δk y(t) = change in yield over holding period of length k (between t−k and t).
  - Given portfolio value in EUR at cut-off date T, V(T), change in value:
    - Δk V(t,T) = V(t) − V(t−k) = VL(T) ⋅ Δ%k V(t,T)
    - Δ%k V(t,T) = (FX(t) − FX(t−k)) / FX(t) − (D ⋅ Δk y(t,T) − s/2 ⋅ Δk y(t,T)^2)
      - FX(t) is exchange rate to EUR at time t; D is duration; s is convexity at cut-off date T.
  - Change in yield for a bond determined by interpolating changes in yield for bond indices with tenors closer to its duration.
- Excess collateral and FX treatment:
  - For counterparty i, excess collateral Ci,t:
    - Ci,t = V(T) − cash ⋅ FX(T) ⋅ (1 + (FX(t) − FX(t−k)) / FX(t))

### CCP loss enumeration, aggregation, and risk metrics
- For ncpt counterparties there are 2^ncpt combinations of defaulting counterparties.
- Defaults assumed independent conditional on each day’s market conditions; joint probability for combination u[k]:
  - P(u[k]) = ∏ PDi^{u[k]_i = 1} ⋅ ∏ (1 − PDi)^{u[k]_i = 0}
- Conditional losses for defaulting counterparties summed to obtain CCP loss for a given day.
- Loss aggregation:
  - Use matrices: C (ncpt × ndays) of excess collateral; H (ncpt × nsec) of collateral posted; S (nsec × ndays) percentage shocks Δ%k V(t,T).
  - Conditional loss per combination uses [ [C + H ⋅ S]^T ]⋅u[k] for k = 1,...,2^ncpt.
  - Expected shortfall at level α across all combinations and ndays (equal weight per day); α = 0.001 used in results.
- Alternative metric:
  - NDL-Cover-2: sum of two largest conditional losses each day; NDL-Cover-2 defined as expected shortfall at level α across days (α = 0.001).

### System-wide spillovers from NBFI liquidity distress (investment funds focus)
- Coverage and scope:
  - Stress test included around 32,000 investment funds, representing about 70 percent of AUM of the EA fund sector, covering UCITS, AIF and MMF.
  - Spillovers to/from insurance sector quantified using detailed EA insurers’ asset holdings and derivative positions shared by EIOPA.
- Time horizons and scenarios:
  - Two market scenarios combining steepening yield curves, equity market corrections, and widening spreads in sovereign and corporate funding markets.
  - Time horizons: two days and two weeks (10 business days).
  - Two-day scenario: share redemptions and asset sales excluded.
  - Two-week scenario: decline in funds’ NAV induces investor redemptions.
- Aggregate liquidity outflows:
  - Two-day scenario: liquidity outflows could total EUR 16 billion, concentrated in bond funds; comparable to margin calls faced by EA funds in March 2020.
  - Two-week scenario: combined liquidity outflows are close to EUR 700 billion, corresponding to aggregate outflow rate of 4.7 percent of AUM, concentrated in equity funds.
- Liquidity shortfalls and margin risks:
  - Two-day scenario: aggregate liquidity demand exceeds available liquidity, resulting in shortfall up to EUR 9.8 billion.
  - Repo borrowing could mitigate shortfall but effectiveness depends on counterparties’ willingness and ability; operational challenges limit many funds' access to repo market, with majority of repo counterparties located outside the EA.
  - Liquidity shortfall translates, on average, to 52 percent of variation margins due from in-scope funds to banks potentially at risk; median = 65 percent.
- Market impact amplification (two-week scenario):
  - Funds meet liquidity demands by liquidating assets, amplifying initial shocks.
  - Estimated EA government bond price impacts vary between 12 and 350 basis points depending on scenario specifics.
  - Corporate bonds: price impacts between 45 and 174 bps.
  - Bank unsecured bonds: price decline up to 110 bps; covered bank bonds: up to 54 bps.
  - Spillovers could extend to non-EA markets due to EA funds’ global investment focus.
- Impacts on institutions and instruments:
  - MMFs could encounter up to EUR 40 billion in redemptions.
  - Insurance companies facing margin calls could contribute to MMF redemptions of around EUR 9 billion.
  - If MMFs liquidate corporate paper after exhausting cash buffers, CP sales could be up to EUR 26 billion, majority issued by banks.
  - Defensive NBFI actions could amplify liquidity strains, trigger outflows from bank deposits and drop in HQLA values.

### Modeling, data and methodological observations
- Historical simulations span January 2020–December 2024 (CCP reverse-repo) and market variables time series include January 2008 - October 2024 for fund stress tests.
- Collateral revaluation uses duration and convexity approximations; changes in yields interpolated across bond indices.
- Derivatives valuation:
  - Variation margin (VM) calls computed; initial margin (IM) calls not considered.
  - Instruments treated via sensitivity analysis (equity options, CDS, bond futures, STIR) or full repricing (equity futures, equity swaps, IRS).
  - Key assumptions: CDS DV01 reference tenor = 5Y CDS; equity swap spread s = 2 percent; dividend yield set to zero for equity futures; equity swap maturities > 1 year truncated to 1 year.
- Data gaps and limitations emphasized:
  - Need more granular information on use of liquidity management tools (LMTs), heterogeneity of investor base, liquidity lines with custodian banks, instrument-level holdings across all funds, and systematic fund-to-custodian mapping.
  - SFTR trade repository quality limits full visibility, especially for Tier 2 CCP reverse-repo positions.
  - Encumbered assets excluded from liquidation; other SFTs not fully included due to data quality.
- Price-impact modeling:
  - Power-law price-impact functions calibrated with market depth depending on volatility and traded volumes.
  - Volatility estimated via GARCH(1,1) on total return series between January 1, 2020 and October 28, 2024; median used.
  - Volume estimates use MiFID transparency calculations re-scaled for two-week horizon.

### System-wide liquidity stress testing: scope, data, and methodology
- Objective: estimate impact on EA core markets and financial institutions from investment funds’ liquidity needs driven by redemptions, derivative margin calls, and repo collateral calls.
- Data sources and cut-off:
  - Granular holdings and historical fund flows from Refinitiv Lipper for UCITS; regulatory data for AIF and MMF by ESMA staff.
  - Transaction-level data from EMIR and SFTR for derivatives and repo exposures; market data from MiFID, Haver and Bloomberg.
  - Cut-off date for fund holdings: June 30, 2024. Transaction level data spans June 2023–June 2024. Derivatives and repos data collected for June 28, 2024.
- Two-step analysis:
  1. Simulate cumulative outflows/inflows over stress horizon and compare to available liquidity supply.
  2. Measure spillover effects to core markets, funds, insurance firms, and banks.
- Institutional perimeter and market shares:
  - Over 32,000 funds domiciled in the EA across nineteen countries; about 70 percent of AUM of EA fund sector.
  - Approximately 13,000 UCITS funds with AUM EUR 7.1 trillion (coverage ~50 percent).
  - Approximately 19,000 AIFs with AUM EUR 7.4 trillion (near complete coverage).
  - Approximately 400 MMFs with AUM EUR 1.6 trillion (90 percent coverage).
  - Solvency II supervisory data covers about 70 percent of EA insurance firms’ total investments (reference date 2024Q2).
- Sources of liquidity considered:
  - Withdrawing bank deposits; allowing reverse repo positions to mature; redeeming MMF and ETF shares; selling portfolio assets; borrowing via repo.
  - Committed/uncommitted liquidity lines with custodians not accounted for due to lack of information.

### Sample overview and portfolio composition (exact figures)
- Major asset holdings of in-sample funds:
  - EA equities: EUR 1.19 trillion.
  - US equities: EUR 2.12 trillion.
  - EA sovereign debt: nearly EUR 700 billion.
  - US Treasuries: EUR 281 billion.
  - UK sovereign debt: EUR 242 billion.
  - Other sovereign debt: roughly EUR 256 billion.
  - EA corporate/bank debt: approximately EUR 1.18 trillion.
  - US corporate debt: EUR 554 billion.
  - ROW corporate debt: EUR 622 billion.
  - Crossholdings of in-sample funds: about EUR 1.81 trillion (about 14 percent of total AUM in-sample).
  - ETFs: EUR 186 billion.
  - MMF shares owned by funds: EUR 117 billion (compared to aggregate MMF balance sheets in EA of EUR 1.7 trillion in 2024Q2).
  - Derivatives positions: EUR 2.01 trillion.
  - Cash: EUR 291 billion.
  - Remainder across other asset classes: approximately EUR 1.7 trillion.
- Derivatives notional / percent of total derivative notional exposure:
  - Interest rate swaps: EUR 1.5 trillion (14.9 percent).
  - Interest rate futures: EUR 1.3 trillion (13 percent).
  - Equity options: EUR 0.5 trillion (5 percent).
  - Credit swaps: EUR 443 billion (4 percent).
  - Equity swaps: EUR 383 billion (4 percent).
  - Interest rate swaptions: EUR 298 billion (3 percent).
- Repo activity (typical week around cut-off):
  - AIFs mostly engage in repo borrowing: about EUR 120 billion (typically with UK banks).
  - MMFs almost exclusively lend securities via reverse repo: outstanding amounts vary between EUR 300 and 350 billion.
  - UCITS average outstanding volume: repos ~EUR 35 billion / reverse repos ~EUR 25 billion.
  - In aggregate, 13 EA banks provide repo funding to funds in sample, accounting for 20 percent of volume; remainder to non-EA banks.

### Scenarios and shock calibration (exact shocks)
- Two horizons: two-day and two-week (10 business days).
- Calibration based on expected shortfall at 0.1 percent of market factors’ marginal distributions over holding period.
- Interest rate shocks (swap curve):
  - Two-day scenario: increase of about 20 bps for the 6 months horizon and 30 bps for longer horizons.
  - Two-week scenario: shift upward by 50 bps for the 6 months horizon, 65 bps for the 1-year horizon and 75 bps for the 5 year and longer terms.
  - Similar magnitude shocks applied to non-EUR swap curves.
- Sovereign risk premium shocks (above swap curve):
  - Two-day scenario: increase of 32 bps for low-risk sovereigns, 47 bps for medium-risk sovereigns, 91 bps for high-risk sovereigns.
  - Two-week scenario: 72 bps for low-risk, 96 bps for medium-risk, 172 bps for high-risk sovereigns.
- FX shocks are not considered.
- Shock application:
  - Shocks applied instantaneously to starting balance sheets; cumulative valuation effect quantified.
  - Short horizons chosen because liquidity distress episodes tend to be short-lived and to avoid overlap with regular portfolio rebalancing.

### Shock formulation, flow-performance, and redemption calibration
- Asset depreciation linear in risk factors:
  - ΔA_i = A_i^0 ⋅ ∑_k α_{k,i} ⋅ δ_k
- Redemption calibration uses Fama-McBeth two-step approach estimated on UCITS monthly flows between August 2019 and October 2024.
- Flow-performance sensitivities (Table 2 estimates):
  - Bond 0.79***
  - Equity 0.55***
  - Mixed 0.31***
  - Alternatives 0.52**
  - Notes:
    - For Funds-of-Funds the average coefficient of bond and equity funds applied.
    - For LDI funds the bond funds value is attributed.
    - For MMFs a flow-performance parameter of 0 is applied.
    - *** p-values <1 percent; ** p-values < 5 percent.

### Repo, collateral calls, and leverage scenarios
- Collateral calls channels: longer-term repos top-up/deleveraging and overnight repos roll-over risk.
- For roll-over risk assume all repos are rolled over; funds engaging in reverse repos (almost exclusively MMFs) do not roll-over reverse repos when conserving liquidity.
- Three leverage scenarios:
  - (i) no additional leverage,
  - (ii) additional leverage up to the highest ratio observed for that fund in last year,
  - (iii) additional leverage up to full encumbrance (upper bound).
- Repo eligible assets: assets eligible for the ECB GC Pooling or ECB GC Pooling Extended repo baskets.
- Repo haircuts: apply haircuts from the ECB’s collateral framework as applicable on June 24th, 2024.
- Synthetic leverage proxy: IM/AUM (lower bound estimate; may overestimate UCITS ability to engage in new repo).

### Asset liquidation sequencing and pecking order
- Sources of liquidity: bank deposits; maturing reverse repos; MMF shares; ETF shares; selling portfolio assets; repo borrowing.
- Asset sale strategies: vertical (pro-rata) slicing and horizontal (waterfall) slicing considered.
- Liquidation begins after exhausting cash, cash-equivalents and repo resources.
- Pecking order:
  - MMFs treated as cash-like with high pecking order.
  - ETFs treated as liquid and assigned high pecking order before securities.
  - Other investment fund shares positioned after securities.
- Encumbered assets excluded from liquidation; encumbrance high for LDI and hedge funds:
  - About 38 percent of government bonds encumbered for LDI.
  - About 35 percent of government bonds encumbered for hedge funds.

### Key quantitative results and stress outcomes
- Two-day scenario:
  - Aggregate liquidity shortfall up to EUR 9.8 billion.
  - Aggregate liquidity outflows could total EUR 16 billion, concentrated in bond funds; comparable to March 2020 margin calls.
  - Funds with insufficient liquid buffers hold EUR 1.6 trillion in AUM (11 percent of the sample).
  - UCITS account for 91 percent of the AUM exposed to shortfalls.
  - Bond UCITS contribute the most to shortfall in absolute terms: EUR 4 billion shortfall; corresponding AUM EUR 0.6 trillion (28 percent of Bond UCITS AUM).
  - Alternatives UCITS: 47 percent (in terms of AUM of funds with this strategy) face a liquidity shortfall.
  - Among AIF, LDI funds: 15 percent (in terms of AUM) face potential liquidity distress.
  - UCITS and AIF collectively redeem between EUR 1.4 and 1.8 billion in MMF shares (aggregate 0.1 percent redemption rate for EA MMF).
  - Allowing repos up to historical limit could increase funds’ leverage by 1.2 percentage points; up to full encumbrance could increase leverage by 3.8 percentage points.
  - Available eligible unencumbered collateral could cover up to EUR 7 billion.
- Two-week scenario:
  - Combined liquidity outflows close to EUR 700 billion (aggregate outflow rate 4.7 percent of AUM).
  - Aggregate liquidity shortfall limited to less than EUR 0.5 billion (or less than 0.1 percent of total liquidity demand), pertaining to funds representing less than 0.2 percent of AUM.
  - Redemptions represent 5 percent of initial system AUM overall.
  - Equity UCITS: 16 percent redemption rate on average; Alternative UCITS: 10 percent.
  - Price impacts and second-round effects:
    - EA government bonds: price impact between 12 and 350 bps, translating into yield impact up to 40 bps.
    - Corporate bonds: 45–174 bps price impacts.
    - Bank unsecured bonds: decline up to 110 bps; covered bank bonds: up to 54 bps.
  - Borrowing via repo instead of selling would limit price falls but impose additional leverage on remaining shareholders—up to an additional six percentage points in scenarios considered.
- Impacts on insurers and MMFs:
  - Insurers’ VM total: EUR 34 billion (two-week scenario), concentrated in very large Life insurers (49 percent) and very large Composites (27 percent).
  - Insurers’ VM could trigger an additional EUR 9 billion in MMF share redemptions (53 percent from composites; 44 percent from life).
  - MMFs could face up to EUR 40 billion in redemptions (around 3 percent redemption rate); 59 percent hit VNAV MMFs.
  - Funds would liquidate about two-thirds of their total EA MMF holdings, causing MMFs to liquidate around EUR 21.5 billion in CP, of which EUR 19.2 billion are issued by banks.
- Banks and deposits:
  - HQLA bond values on banks’ balance sheets could decline a further 80 bps on top of market scenario decline.
  - Deposits outflows from funds: up to EUR 70 billion assuming funds use cash before asset sales, corresponding to average 45 percent runoff rate.
  - Repo outflows from cancelled reverse repo: at most EUR 0.86 billion.
  - Potential new repo between EUR 0.93 and EUR 4.14 billion for the 13 EA banks with repo trades with EA funds (dependent on leverage assumption).
- Cross-fund redemptions and liquidation of fund shares:
  - Under waterfall liquidation, total sales of other UCITS and AIFs around 2 to 2.3 percent of funds’ inventory of these shares.
  - Under pro rata liquidation, funds redeem around 5-6 percent of aggregate inventory of other funds’ shares, causing up to an additional 0.5 percentage points of redemptions.

### Conclusions, policy implications and key recommendations
- Findings:
  - Two-day market shock: material gaps to meet aggregate demand; unmet margin calls by NBFIs could trigger CCR losses for banks.
  - Two-week horizon: defensive actions by NBFIs may amplify funding market shocks and produce second-round effects weakening solvency and liquidity of financial institutions.
  - Correlated defensive strategies can propagate stress to core financial markets and institutions, creating financial stability risks.
- Recommended actions (summarized from main recommendations):
  - CCP systemic risk analysis:
    - Regularly monitor risks in investments (non-default losses) by CCPs. — Paragraph 11; Authorities: ESMA / NCA; Timing: ST
    - Close data gaps for Tier 2 CCP regarding investment positions, including on collateral received and currency. — Paragraph 11; Authorities: ESMA; Timing: MT
    - Include risk from non-default losses in the next CCP stress test. — Paragraph 12; Authorities: ESMA; Timing: MT
  - Investment fund sector systemic risk analysis:
    - Operationalize and conduct regular macroprudential EU-wide liquidity stress tests, including UCITS, AIF and MMF. — Paragraphs 51, Box 1; Authorities: ESMA; Timing: MT
    - Strengthen system-wide financial risk monitoring on a cross-country and cross-sectoral basis and conduct system-wide stress tests, including bank and nonbank sectors. — Paragraph 52; Authorities: ESMA / ESRB / EIOPA / ECB / EC; Timing: MT
    - Ensure revised supervisory templates comprise necessary information to conduct stress tests, including:
      - Instrument level holdings;
      - Returns and flows with at least weekly frequency;
      - Use of liquidity management tools (LMTs);
      - Temporary borrowing arrangements and custodian bank;
      - Asset encumbrance, with source of encumbrance;
      - Funds redemption schedule;
      - Harmonized leverage. — Paragraphs 26, 35, 38, 41, Box 1; Authorities: ESMA / EC; Timing: ST
    - Conduct currency-specific liquidity stress tests for the most significant currencies (EUR, USD, GBP). — Paragraph 33; Authorities: ESMA; Timing: MT
    - Conduct multi-period liquidity stress test. — Paragraph 23; Authorities: ESMA; Timing: MT
    - Make proper data infrastructure and software tools available to economists and risk analysts. — Box 1; Authorities: ESMA; Timing: ST
    - Ensure trade repositories improve the quality of transaction level data collected under EMIR and SFTR. — Paragraph 6; Authorities: ESMA; Timing: MT
  - Data priorities:
    - Collect instrument-level holdings for all funds, systematic information on standing credit lines with custodians, prevalence/details of LMTs, fund-to-bank custodian mapping, and harmonized higher-frequency returns and flows by share class.
  - Governance and tools:
    - Regular system-wide stress tests require strong governance, centralized/harmonized data and ad-hoc ICT tools and capacity for teams conducting financial stability analysis.

*Italic: Source — EXECUTIVE SUMMARY and selected sections (Technical Note on NBFI systemic risk assessment — Euro Area FSAP; Appendices I–III and related paragraphs) as provided in the supplied content.*

### EXECUTIVE SUMMARY __________________________________________________________________________ 5

### EXECUTIVE SUMMARY

### CCP counterparty credit risk on reverse repo
- Analysis covers 16 CCPs authorized to clear in the EA regardless of their domicile, using data with reference date December 11, 2024.
- Reverse repos are used by CCPs to invest cash received from clearing members as margin and are characterized by lower haircuts on collateral received (government bonds) than haircuts applied in clearing business.
- Data sources include EU SFTR trade-state tables and CCP supervisory reports; PD time series are market-based and sourced from National University of Singapore, Credit Research Initiative (NUS CRI).
- Sample limitations:
  - Of 14 EU CCPs, only seven reported reverse repos (mostly in non-EUR currencies: USD, GBP, and other EU currencies).
  - EU CCPs have at most eight different counterparties; third-country Tier 2 CCPs have up to 54 counterparties.
  - SFTR sample covers trades reported by EU firms; for EU CCPs IMF/ESMA obtained 100 percent coverage except one EU CCP excluded for data quality.
  - For Tier 2 CCPs, after cleaning for outliers, included trades represent no more than 29 percent of Tier 2 CCPs’ reverse repo positions—making loss estimates a lower bound.
- Methodology:
  - Expected Shortfall (ES) set-up combining historical simulations (January 2020–December 2024) to capture time-varying exposures and market dependencies across tenors and currencies over a five-day holding period.
  - For each day, compute conditional loss per CCP–counterparty as the difference between discounted value of collateral and cash at maturity; netting allowed across positions of same counterparty but not across counterparties.
  - Loss given default after collateral value assumed to be 1.
- Key findings:
  - Most CCPs experience limited losses (up to 20 basis points) in the stress tests.
  - Losses can reach up to 120 percent of a CCP's own capital at risk in cases where skin-in-the-game is minimal.
  - Potential losses, although modest in absolute terms, can be substantial compared to CCP capital.

### System-wide spillovers from NBFI liquidity distress (investment funds focus)
- Coverage and scope:
  - Stress test included around 32,000 investment funds, representing about 70 percent of assets under management (AUM) of the EA fund sector, covering UCITS, AIF and MMF.
  - Spillovers to and from the insurance sector to the fund sector quantified based on detailed EA insurers’ assets holdings and derivative positions shared by EIOPA.
- Scenarios and time horizons:
  - Two market scenarios combining steepening yield curves, equity market corrections, and widening spreads in sovereign and corporate funding markets.
  - Time horizons: two days and two weeks.
  - Two-day scenario: share redemptions and asset sales are excluded.
  - Two-week scenario: decline in funds’ net asset values induces investor redemptions.
- Aggregate liquidity outflows:
  - Two-day scenario: liquidity outflows could total EUR 16 billion, concentrated in bond funds; comparable to the margin calls faced by EA funds in March 2020.
  - Two-week scenario: combined liquidity outflows are close to EUR 700 billion, corresponding to an aggregate outflow rate of 4.7 percent of AUM, concentrated in equity funds.
- Liquidity shortfalls and margin risks:
  - Two-day scenario: aggregate liquidity demand exceeds available liquidity, resulting in a shortfall of up to EUR 9.8 billion.
  - Repo borrowing could mitigate shortfall but effectiveness depends on counterparties’ willingness and ability to provide funding; operational challenges limit many funds' access to repo market, with majority of repo counterparties located outside the EA.
  - The liquidity shortfall translates, on average, to 52 percent of variation margins due from in-scope funds to banks potentially at risk, which could trigger counterparty credit risk (CCR) losses.
- Market impact amplification in the two-week scenario:
  - Funds predominantly meet liquidity demands by liquidating assets, amplifying initial market shocks.
  - Correlated defensive strategies by funds could propagate stress to core financial markets and institutions.
  - Estimated EA government bond price impacts vary between 12 and 350 basis points depending on scenario specifics.
  - Spillovers could extend to non-EA markets due to EA funds’ global investment focus.
- Impacts on financial institutions and instruments:
  - MMFs could encounter up to EUR 40 billion in redemptions, leading to significant asset liquidations.
  - Insurance companies facing margin calls could further contribute to MMF redemptions for around 9 billion.
  - If MMFs liquidate corporate paper (CP) after exhausting cash buffers, CP sales could be up to EUR 26 billion, with a majority issued by banks.
  - Defensive actions by NBFIs could amplify liquidity strains, triggering outflows from bank deposits and a drop in the value of high-quality liquid assets (HQLA).

### Modeling, data and methodological observations
- CCP reverse-repo loss modeling uses netting at counterparty level and assumes loss given default after collateral equal to 1.
- Historical simulations span January 2020–December 2024 to capture market dependencies and time-varying exposures over a five-day holding period.
- The analysis acknowledges significant data gaps and the need for more granular information to accurately assess funds’ liquidity demand and behavioral responses, including:
  - Use of liquidity management tools (LMTs) and redemption fees.
  - Heterogeneity of investor base.
  - Information on liquidity lines with custodian banks.
  - Instrument-level holdings across all funds.
- Operational and data constraints:
  - Many funds face operational challenges accessing repo markets; majority of repo counterparties are located outside the EA.
  - Trade repository quality under SFTR limits full visibility, especially for Tier 2 CCP reverse-repo positions.

### Policy recommendations (summarized from Table 1: Euro Area: Main Recommendations)
- CCPs systemic risk analysis:
  - Regularly monitor risks in investments (non-default losses) by CCPs. — Paragraph 11; Authorities: ESMA / NCA; Timing: ST
  - Close data gaps for Tier 2 CCP regarding investment positions, including on collateral received and currency. — Paragraph 11; Authorities: ESMA; Timing: MT
  - Include risk from non-default losses in the next CCP stress test. — Paragraph 12; Authorities: ESMA; Timing: MT
- Investment fund sector systemic risk analysis:
  - Operationalize and conduct regular macroprudential EU-wide liquidity stress tests, including UCITS, AIF and MMF. — Paragraphs 51, Box 1; Authorities: ESMA; Timing: MT
  - Strengthen system-wide financial risk monitoring on a cross-country and cross-sectoral basis and conduct system-wide stress tests, including bank and nonbank sectors. — Paragraph 52; Authorities: ESMA / ESRB / EIOPA / ECB / EC; Timing: MT
  - Ensure revised supervisory templates comprise necessary information to conduct stress tests to meet the financial stability mandate, including:
    - Instrument level holdings;
    - Returns and flows with at least weekly frequency;
    - Use of liquidity management tools (LMTs);
    - Temporary borrowing arrangements and custodian bank;
    - Asset encumbrance, with source of encumbrance;
    - Funds redemption schedule;
    - Harmonized leverage. — Paragraphs 26, 35, 38, 41, Box 1; Authorities: ESMA / EC; Timing: ST
  - Conduct currency-specific liquidity stress tests for the most significant currencies (EUR, USD, GBP). — Paragraph 33; Authorities: ESMA; Timing: MT
  - Conduct multi-period liquidity stress test. — Paragraph 23; Authorities: ESMA; Timing: MT
  - Make proper data infrastructure and software tools available to economists and risk analysts. — Box 1; Authorities: ESMA; Timing: ST
  - Ensure trade repositories improve the quality of transaction level data collected under EMIR and SFTR. — Paragraph 6; Authorities: ESMA; Timing: MT

*Source: EXECUTIVE SUMMARY, Technical Note on NBFI systemic risk assessment — Euro Area FSAP.*

### 8.      We approximate collateral revaluation using its duration and convexity. We map each

### 8.      We approximate collateral revaluation using its duration and convexity. We map each

### Collateral revaluation methodology
- Collateral (bond) revaluation is approximated using duration (D) and convexity (s) at the cut-off date T.
- Each bond used as collateral is mapped to a corresponding bond index by issuer country and tenor to obtain the relevant yield time series over the historical simulation horizon.
- Notation and approximation:
  - VL(t) = value in denomination currency of a bond on day t of the historical simulation.
  - Δk y(t) = change in yield over the holding period of length k (i.e., between t−k and t).
  - Given portfolio value in EUR at cut-off date T, V(T), the change in value over the historical simulation:
    - Δk V(t,T) = V(t) − V(t−k) = VL(T) ⋅ Δ%k V(t,T)
    - Δ%k V(t,T) = (FX(t) − FX(t−k)) / FX(t) − (D ⋅ Δk y(t,T) − s/2 ⋅ Δk y(t,T)^2)
      - FX(t) is the exchange rate to EUR at time t.
      - D is duration and s is convexity at cut-off date T.
- Change in yield for a given bond is determined by interpolating across changes in yield for bond indices with tenors closer to its duration.

### Excess collateral and FX treatment
- For each counterparty i, excess collateral Ci,t accounts for FX shock on the cash leg analogously to the collateral leg, noting FX for the cash leg may be distinct from collateral FX:
  - Ci,t = V(T) − cash ⋅ FX(T) ⋅ (1 + (FX(t) − FX(t−k)) / FX(t))

### CCP loss enumeration and probabilities
- For a CCP with ncpt counterparties, there are 2^ncpt combinations of defaulting counterparties.
- Defaults assumed independent conditional on each day’s market conditions; joint probability for a combination u[k]:
  - P(u[k]) = ∏ PDi^{u[k]_i = 1} ⋅ ∏ (1 − PDi)^{u[k]_i = 0}
  - u[k] (ncpt × 1) is the vector of default indicators (u[k]_i = 1 if counterparty i defaults).
- Conditional losses for defaulting counterparties are summed to obtain CCP loss for any given day.

### Loss aggregation over historical simulation and risk metric
- Losses over ndays days are computed across all 2^ncpt combinations and ndays days.
- Matrices used:
  - C: ncpt × ndays matrix of excess collateral for each counterparty and day (monetary amount).
  - H: ncpt × nsec matrix of collateral posted across nsec collateral (monetary amount in Euro at cutoff date).
  - S: nsec × ndays matrix of percentage shocks Δ%k V(t,T).
- Conditional loss per combination uses [ [C + H ⋅ S]^T ]⋅u[k] (for k = 1,...,2^ncpt).
- Loss estimate defined as the expected shortfall at level α across all combinations and all ndays, assuming equal weight per day.
  - For the presented results, α = 0.001.

### Alternative metric: NDL-Cover-2
- Cover-2 losses concept adapted: sum of the two largest conditional losses for each day.
- NDL-Cover-2 defined as expected shortfall at level α across all days, equal probability per day.
  - For the presented results, α = 0.001.

### Comparison to CCP resources and supervisory considerations
- Assessment compares each CCP’s potential losses to:
  - skin-in-the-game (SITG) — CCP’s own capital exposed to clearing member default losses — used as a proxy for CCP loss absorption capacity.
  - default fund (DF) — size related to Cover-2 losses in the clearing portfolio.
- Non-default losses (NDL), including losses from CCPs’ own investments, are part of NDL and not absorbed via the CCP clearing waterfall; in normal times NDL are supposed to be absorbed by equity.
- Data limitations: FSAP team could not access CCP equity or yearly profit; hence SITG used as proxy.
- Policy recommendation: ESMA should monitor this risk against CCPs’ resources to absorb NDL, and close existing data gaps for Tier 2 CCPs.

### Key empirical findings on CCP reverse repo exposures and losses
- Expected shortfall relative to total cash investment:
  - In most cases modest, up to 15 basis points (bps), with some CCPs experiencing no losses.
- Concentration:
  - Losses concentrated in a few CCPs characterized by relatively smaller SITG compared to DF.
  - In sample, ratio SITG / DF ranges between 0.3 and 4.5 percent.
- Potential severity relative to SITG:
  - In some cases losses could consume a significant part of own resources, with one case reaching 120 percent of SITG.
  - Implication: significant portion of estimated potential losses from CCR could be distributed to clearing members in recovery.
- NDL-cover-2 relative to default fund:
  - NDL-cover-2 could reach up to 13 percent of the default fund, with three CCPs exceeding 7 percent.
  - Implication: in a market downturn, NDL could contribute to spillovers from CCPs to the clearing ecosystem.
- Policy recommendation: ESMA should include losses from reverse repo and NDL more broadly as part of the regular CCP supervisory stress test.

### System-wide liquidity stress testing: scope and methodology
- Objective: estimate impact on EA core markets and financial institutions from investment funds’ liquidity needs driven by redemption shocks, derivative margin calls, and repo collateral calls.
- Data sources and cut-off:
  - Granular holdings and historical fund flows from Refinitiv Lipper for UCITS; regulatory data for AIF and MMF by ESMA staff.
  - Transaction-level data from EMIR and SFTR for derivatives and repo exposures.
  - Market data from MIFID, Haver and Bloomberg.
  - Cut-off date for fund holdings: June 30, 2024.
  - Transaction level data spans June 2023–June 2024.
  - Data on derivatives and repos collected for June 28, 2024.
- Two-step analysis:
  1. Simulate cumulative outflows and inflows over stress horizon and compare to available liquidity supply.
  2. Measure spillover effects to core markets, funds, insurance firms, and banks.
- Scope includes quantifying spillovers from insurance to MMF sector due to margin calls, using Solvency II supervisory data covering about 70 percent of EA insurance firms’ total investments (reference date 2024Q2).
- Pension funds not covered (4 percent of EA financial assets); to be covered by upcoming EIOPA stress test.

### Sample overview and portfolio composition (exact figures)
- Sample coverage:
  - Over 32,000 funds domiciled in the EA across nineteen countries; covers about 70 percent of AUM of the EA fund sector.
  - Approximately 13,000 UCITS funds with AUM totaling EUR 7.1 trillion (coverage ~50 percent).
  - Approximately 19,000 AIFs with AUM of EUR 7.4 trillion (near complete coverage).
  - Approximately 400 MMFs (UCITS and AIFs) with AUM of EUR 1.6 trillion (90 percent coverage).
- Major asset categories held by in-sample funds:
  - EA equities: EUR 1.19 trillion.
  - US equities: EUR 2.12 trillion.
  - EA sovereign debt: nearly EUR 700 billion.
  - US Treasuries: EUR 281 billion.
  - UK sovereign debt: EUR 242 billion.
  - Other sovereign debt: roughly EUR 256 billion.
  - Corporate and bank debt:
    - EA corporate/bank debt: approximately EUR 1.18 trillion.
    - US corporate debt: EUR 554 billion.
    - ROW corporate debt: EUR 622 billion.
  - Crossholdings of in-sample funds: about EUR 1.81 trillion (about 14 percent of total AUM of in-sample funds).
  - ETFs: EUR 186 billion.
  - MMF shares owned by funds: EUR 117 billion (compared to aggregate MMF balance sheets in EA of EUR 1.7 trillion in 2024Q2).
  - Derivatives positions: EUR 2.01 trillion.
  - Cash: EUR 291 billion.
  - Remainder across other asset classes: approximately EUR 1.7 trillion.
- Derivatives breakdown (notional / percent of total derivative notional exposure):
  - Interest rate swaps: EUR 1.5 trillion (14.9 percent).
  - Interest rate futures: EUR 1.3 trillion (13 percent).
  - Equity options: EUR 0.5 trillion (5 percent).
  - Credit swaps: EUR 443 billion (4 percent).
  - Equity swaps: EUR 383 billion (4 percent).
  - Interest rate swaptions: EUR 298 billion (3 percent).
- Repo activity (typical week around cut-off date):
  - AIFs mostly engage in repo borrowing: about EUR 120 billion (typically with UK banks).
  - MMFs almost exclusively lend securities via reverse repo: outstanding amounts vary between EUR 300 and 350 billion.
  - UCITS average outstanding volume: repos ~EUR 35 billion / reverse repos ~EUR 25 billion.
  - In aggregate, 13 EA banks provide repo funding to funds in sample, accounting for 20 percent of volume; remainder to non-EA banks.

### Asset segmentation and pecking order
- Holdings aggregated into 47 asset baskets; 30 flagged as eligible for General Collateral (GC) pooling and GC pooling extended.
- Sovereign bonds: four maturity buckets above one year plus a separate bucket for bonds < one-year residual maturity.
- Corporate bonds and equities aggregated by region (EA, US, ROW); EA further separated by bank bonds and corporate bond rating (HY, IG).
- Funds’ shares differentiated:
  - MMFs treated as cash-like instruments with high pecking order.
  - ETFs treated as liquid and assigned high pecking order before securities.
  - Other investment fund shares (OIF) positioned in pecking order after securities.

### Scenarios and shock calibration (exact shocks)
- Two market scenarios considered with two horizons: two-day and two-week (10 business days) horizons.
- Scenarios calibrated based on expected shortfall at 0.1 percent of market factors’ marginal distributions over holding period (two or 10 days).
- Shocks applied instantaneously to starting balance sheets, cumulative valuation effect quantified.
- Interest rate shocks (swap curve):
  - Two-day scenario: increase of about 20 bps for the 6 months horizon and 30 bps for longer horizons.
  - Two-week scenario: shift upward by 50 bps for the 6 months horizon, 65 bps for the 1-year horizon and 75 bps for the 5 year and longer terms.
  - Similar magnitude shocks applied to non-EUR swap curves.
- Sovereign risk premium shocks (above swap curve):
  - Two-day scenario: increase of 32 bps for low-risk sovereigns, 47 bps for medium-risk sovereigns, 91 bps for high-risk sovereigns.
  - Two-week scenario: 72 bps for low-risk, 96 bps for medium-risk, 172 bps for high-risk sovereigns.
- FX shocks are not considered.
- Time-horizon considerations:
  - Short horizons chosen because liquidity distress episodes tend to be short-lived and to avoid overlap with regular portfolio rebalancing.
  - Multi-period dynamics not modelled due to FSAP timeline; recommendation: ESMA should consider developing a multi-period framework to align with best practices.

*Source: 1eurea2025008 - 8.      We approximate collateral revaluation using its duration and convexity. We map each*

### 24.      The shock is applied to funds’ assets and is linear in all risk factors. Formally, the change

### 24. The shock is applied to funds’ assets and is linear in all risk factors.

### Shock formulation and asset depreciation
- Change in the value of asset i:
  - ΔA_i = A_i^0 ⋅ ∑_k α_{k,i} ⋅ δ_k
  - A_i^0 denotes the asset value at the start of the exercise.
  - α_{k,i} denotes exposure to risk factor k.
  - δ_k measures the change in value of risk factor k (the scenario).
  - Summation runs through all k factors the asset i is sensitive to.
- For funds holding other funds’ shares:
  - Apply the corresponding fund devaluation to the share held.
  - Use the median for out-of-sample funds for which balance sheet information is not available.
- For the residual asset category:
  - Apply the median depreciation across other assets.
- For derivatives:
  - Valuation impact computed with the equations described in paragraphs 29 and 30, and Appendix III.
  - Effects added to overall depreciation at fund level to account for potential hedges through derivative positions.

### Market scenarios and timing
- Two scenarios considered explicitly:
  - Two-day scenario: margin calls dominate outflows; redemptions excluded as they are supposed to be met with T+2 settlement.
  - Two-week scenario: redemptions dominate outflows.
- Spread bucketing labels:
  - ‘low’, ‘mid’, ‘high’ refer to bucketing of countries as in the macro scenario.
  - ‘IG’ = Investment Grade, ‘HY’ = High Yield, ‘ROW’ = Rest of the world.

### Redemption calibration and flow-performance
- Flow-performance calibration (flow-return relationship) links redemptions to market scenario using a Fama-McBeth two-step approach:
  - Regression: f_{i,t} = α + ∑_{k=1}^4 β_k f_{i,t−k} + ∑_{h=1}^4 γ_h return_{i,t−h} + θ_{i,t} X_{i,t} + ε_{i,t}
  - f_{i,t−k} measures lagged fund flows.
  - X_{i,t} denotes log(AUM) as a control variable.
- Calibration:
  - Estimated at four fund sub-groups: Bond funds, Equity funds, Mixed strategy funds and Alternatives.
  - Coefficients estimated from UCITS monthly flows between August 2019 and October 2024.
  - Assumption: cash withdrawn from the fund sector is temporarily deposited by investors into their bank (no modeling of flows between funds).
  - For MMF only, redemptions are endogenously determined based on other funds and insurers’ reaction to their liquidity outflows.
- Estimated flow-performance sensitivities (Table 2):
  - Bond 0.79***
  - Equity 0.55***
  - Mixed 0.31***
  - Alternatives 0.52**
  - Notes:
    - For Funds-of-Funds the average coefficient of bond and equity funds is applied.
    - For LDI funds the bond funds value is attributed.
    - For MMFs a flow-performance parameter of 0 is applied.
    - *** p-values <1 percent; ** p-values < 5 percent.

### Derivatives valuation and margin calls
- Variation margin (VM) calls:
  - Computed when market value of a netting set decreases; typically computed daily.
  - Initial margin (IM) calls are not considered.
- Valuation approaches by derivative type:
  - Equity options, bond futures, interest rate futures and CDSs: use sensitivities to changes in underlying asset prices.
    - Equity options: ΔV_{i,k} = δ_k ⋅ Δf(k), where δ_k is option delta and Δf(k) is change in underlying.
    - Bond futures: use BPV of cheapest-to-deliver bond and conversion factor to derive impact of rate changes.
    - Interest rate futures: change in prices determined by change in relevant forward rates.
    - CDSs: change in market value given by DV01.
  - Equity futures, equity swaps and IRDs: full repricing tools used.
    - IRD pricing uses ESMA tool with contract characteristics from EMIR and market information; fixed leg priced using reported fixed rate and discount factors (risk-free curve based on OIS); floating leg uses forward curve and discount factors.
    - Equity swaps: NPV of equity leg (using equity futures and discount factors) and NPV of floating leg (assumed based on an overnight rate set at maturity). Maturity taken from EMIR when < 1 year; otherwise fixed at one year. Floating rate uses the one-year overnight forward rate as of end-June 2024 and a spread of 2 percent is applied to the floating rate.
- VM calculation at fund level:
  - For each fund i with derivatives exposures, liquidity demand from margin calls equals sum of VM over all derivative positions across netting sets P_i:
    - VM_i = ∑_{p∈P_i} ∑_{k∈p} ΔV_{i,k}
  - Two-week scenario: margins received reduce liquidity demand related to margin calls.
  - Two-day scenario: assume funds cannot net incoming margin with outflows; sum across portfolios for which ΔV_{i,k} is negative:
    - VM_i = ∑ [ ∑_{k∈P_i} ΔV_{i,k} ]_{P_i^-}
- Insurance undertakings:
  - Margin calls computed based on sensitivity from de Jong et al. (2019), for two-week scenario only.
  - VM considered assuming interest rate shock as most relevant risk factor and netting across counterparties.
  - Total VM for insurers amounts to EUR 34 billion, concentrated in very large Life insurance companies (49 percent) and very large Composites (27 percent).

### Repo, collateral calls, and repo funding assumptions
- Collateral calls on repos:
  - Two channels: (i) longer-term repos where collateral value falls below threshold requiring top-up or deleveraging; (ii) overnight repos via roll-over risk.
  - Aggregate repo market has a small share of longer-term repos; collateral top-ups are limited.
  - Computation of additional liquidity needs follows approach described in paragraph 8.
  - For roll-over risk, assume all repos are rolled over (no modeling of rebalancing behavior).
  - Funds engaging in reverse repos (almost exclusively MMFs) do not roll-over reverse repos to conserve liquidity when repo terms allow.
- Repo usage scenarios and leverage ceilings:
  - Three leverage scenarios:
    - (i) no additional leverage,
    - (ii) additional leverage up to the highest ratio observed for that individual fund in the last year,
    - (iii) additional leverage up to full encumbrance (upper bound).
  - For scenarios (i) and (ii) consider funds with standing repo agreements during the last year; scenario (iii) allows establishing repo relationships.
  - Repo eligible assets: assets eligible for the ECB GC Pooling or ECB GC Pooling Extended repo baskets.
  - Repo haircuts: apply haircuts from the ECB’s collateral framework as applicable on June 24th, 2024.
  - Synthetic leverage proxy: IM/AUM.
  - Limitations: estimate for synthetic leverage is a lower bound; may overestimate UCITS ability to engage in new repo.

### Liquidity supply, asset sales, and liquidation sequencing
- Sources of liquidity considered:
  - (i) withdrawing bank deposits,
  - (ii) allowing reverse repo positions to mature without renewal,
  - (iii) redeeming shares held in MMFs,
  - (iv) redeeming shares from ETFs,
  - (v) selling portfolio assets,
  - (vi) borrowing (secured via the repo market).
  - Committed or uncommitted liquidity lines with custodian banks not accounted for due to lack of information.
- Asset sale strategies:
  - Vertical (pro-rata) slicing and horizontal (waterfall) slicing considered.
  - Asset liquidation begins after exhausting cash, cash-equivalents and repo resources.
  - Waterfall ranks follow asset segmentation in Appendix IV; baskets with same rank liquidated concurrently pro-rata.
  - Approaches exhaust liquid resources first; do not account for cash hoarding behavior or heterogeneous liquidation across funds.
- MMFs and insurers:
  - MMFs meet redemptions by using cash buffer, canceling maturing reverse repo, then liquidating assets pro rata to cover remaining needs while maintaining portfolio composition.
  - Insurers: for VM-driven outflows, use cash equivalents and MMF shares first, then sell assets to preserve long-term liability-driven allocations.

### Key quantitative results and stress outcomes
- Two-day scenario:
  - Largest absolute outflows: bond funds EUR 6 bn, mixed funds EUR 4 bn (both UCITS and AIF).
  - Collateral calls more muted: bond AIF and LDIs with less than EUR 1 bn each.
  - Alternative UCITS: liquidity outflows reach 2 percent of AUM (most affected in relative terms).
- Two-week scenario:
  - Redemptions represent 5 percent of initial system AUM overall.
  - Heterogeneity by strategy and regulatory framework:
    - Equity UCITS: 16 percent redemption rate on average (most affected).
    - Alternative UCITS: 10 percent redemption rate.
  - AIFs show relatively lower redemption rates than UCITS with comparable strategies due to less asset devaluation and longer redemption schedules (funds with redemption schedules above 10 days or close-ended funds do not experience redemptions in this set-up).
- Insurers’ VM total: EUR 34 billion (two-week scenario), concentrated in very large Life insurers (49 percent) and very large Composites (27 percent).

*Sources: Bloomberg, AIFMD, EMIR, Lipper IM, IMF, ESMA, and IMF staff calculations.*

### 38.      We exclude encumbered assets from the liquidation. Sources of encumbrance include

### 1eurea2025008 - 38–51: Encumbrance, Price-Impact Modeling, Scenarios, and Stress-Test Results

### Asset encumbrance and data limitations
- Encumbered assets excluded from liquidation; sources of encumbrance include existing securities financing transactions (SFTs) (including repos) and initial margin (IM) posted for outstanding derivatives positions.
- For repo, SFTR data provide asset-level information at ISIN level, accounting for netting, mapped to the segmentation described in paragraph 19.
- Other SFTs were not included in the encumbrance estimate due to data quality limitations.
- For derivatives, counterparties report only total IM posted; IM is assigned pro-rata to the set of eligible assets.
- LDI and hedge funds (AIFs) have the highest government bond encumbrance:
  - About 38 percent of government bonds encumbered for LDI.
  - About 35 percent of government bonds encumbered for hedge funds.
- Most other fund strategies exhibit low encumbrance ratios.
- Policy implication: information on asset encumbrance should be part of supervisory templates (see recommendation referenced in text and Table 1).

### Price-impact modeling and market-depth calibration
- Asset sales' market effects estimated using power-law price-impact functions requiring sales amount q and market depth D parameters.
- Market depths are functions of volatility and traded volumes.
- Volatility estimation:
  - GARCH(1,1) estimated for each asset basket on total return series between January 1, 2020, and October 28, 2024; the median across the time series is used.
- Volume estimation:
  - Security-level information from MIFID transparency calculations used.
  - Volumes re-scaled to reflect two-week scenario (instead of six months) and normalized by outstanding volume of each ISIN.
- Calibration caveat: basket-level averaging may underestimate market impact during episodes where a small subset of instruments (e.g., long-dated Gilts) drive distress; instrument-level holdings for all funds are needed to avoid averaging effects (collection of which is a key policy recommendation).

### Simulation assumptions (summary)
- Behavioral and scenario assumptions summarized in Tables 4 and 5.
- Four configurations of assumptions for each scenario to capture uncertainty in funds’ mitigating actions.
- Two-day scenario:
  - Horizon: 2 days; Redemptions: No (T+2 constraint).
  - Liquidity supply excludes asset sales; funds can use cash and MMF shares to meet margin calls.
  - Key scenario variants include leverage limit settings (Full, encumbrance, Historical, Not applicable), asset liquidation strategies (Waterfall up to MMF shares; Pro-rata up to MMF shares), repo rollovers and new repo availability, margin netting, and treatment of maturing repo.
- Two-week scenario:
  - Horizon: 2 weeks; Redemptions: Yes.
  - Funds can use all available assets to satisfy liquidity demands.
  - Key scenario variants include leverage limit settings (Full, encumbrance, Historical, Not applicable), asset liquidation strategies (Waterfall; Pro-rata), reverse repo rollovers if maturity < 10 days, margin netting = Yes (in scenarios), and new repo availability up to leverage limit in some scenarios.

### Two-day scenario: funds’ preparedness and immediate spillovers
- Aggregate liquidity shortfall of up to EUR 9.8 billion.
- Funds with insufficient liquid buffers (cash and MMF shares) hold EUR 1.6 trillion in AUM, or 11 percent of the sample.
- Among funds potentially missing margin calls:
  - UCITS account for 91 percent of the AUM exposed.
  - Bond UCITS contribute the most to shortfall in absolute terms: EUR 4 billion shortfall; corresponding AUM of those funds EUR 0.6 trillion (28 percent of Bond UCITS AUM).
  - Alternatives UCITS: 47 percent (in terms of AUM of funds with this strategy) face a liquidity shortfall.
  - Among AIF, LDI funds: 15 percent (in terms of AUM) face potential liquidity distress.
- Mitigating factor not captured in data: standing credit lines with custodian banks (up to 10 percent of AUM); no data available—collection recommended.
- Knock-on effects on MMFs and banks:
  - UCITS and AIF collectively redeem between EUR 1.4 and 1.8 billion in MMF shares (aggregate 0.1 percent redemption rate for EA MMF).
  - For EA banks counterparty to derivatives, variation margins due from in-scope funds potentially at risk:
    - Average (mean): 52 percent of variation margins.
    - Median: 65 percent of variation margins.
  - Missed variation margins can lead to banks liquidating funds' positions and triggering CCR losses (example cited: Archegos).

### Access to repo in short horizons
- Allowing funds to access repo could mitigate liquidity shortfalls in short horizons.
  - Funds’ leverage could increase by 1.2 percentage points when repos up to historical limit are allowed.
  - Funds’ leverage could increase by 3.8 percentage points when repos up to full encumbrance are allowed.
- Available eligible unencumbered collateral could cover up to EUR 7 billion.
- Operational hurdles and most funds not using repo currently; around 80 percent of existing repo transactions are with counterparties outside the EA.
- Funds may also borrow unsecured from banking sector in stress, contingent on bank willingness/ability.

### Two-week scenario: asset sales, market amplification, and spillovers
- Aggregate liquidity shortfall limited to less than EUR 0.5 billion (or less than 0.1 percent of total liquidity demand), pertaining to funds representing less than 0.2 percent of AUM.
- Focus shifts from fund resilience to spillovers to core markets and institutions due to asset sales.
- Price impacts (varied by asset and defensive actions):
  - EA government bonds: price impact between 12 and 350 bps, translating into a yield impact of up to 40 bps.
  - Corporate bonds: price impacts between 45 and 174 bps.
  - Bank unsecured bonds: price decline up to 110 bps.
  - Covered bank bonds: price decline up to 54 bps.
- Borrowing via repo instead of selling would limit price falls but impose additional leverage on remaining shareholders—up to an additional six percentage points in scenarios considered.
- Data gap: prevalence and details of liquidity management tools (LMTs) not available; would be highly relevant for analysis and calibration.

### Impact on financial institutions and markets (second-round effects)
- Insurers:
  - Price impact would reduce insurance asset values by an additional 58-63 bps (lower bound associated with waterfall liquidation).
  - If funds engage in repo up to available collateral, price falls limited to 10 bps.
  - VM from insurers in scenario: around EUR 34 billion, triggering an additional EUR 9 billion in MMF share redemptions (53 percent from composites; 44 percent from life insurance undertakings).
  - Additional CP sales by MMFs from insurer-driven redemptions: EUR 5.9 billion (under same liquidation assumptions).
- Banks:
  - HQLA bond values on banks’ balance sheets could decline a further 80 bps on top of market scenario decline.
  - Deposits outflows from funds:
    - Up to EUR 70 billion assuming funds use cash before asset sales, corresponding to average 45 percent runoff rate.
    - LCR framework: 25 percent weight to “operational deposits”; 100 percent to “excess operational deposits” and “non-operational deposits”.
    - Banks report above 90 percent of NBFI deposits as “excess operational deposits” or “non-operational deposits” on average.
  - Repo outflows from cancelled reverse repo: at most EUR 0.86 billion.
  - Potential new repo between EUR 0.93 and EUR 4.14 billion for the 13 EA banks with repo trades with EA funds (dependent on leverage assumption).
  - Mapping gaps: no systematic matching of funds to custodian banks—further mapping recommended to identify liquidity distribution imbalances and banks’ potential supply of secured/unsecured liquidity.
- MMFs and commercial paper (CP) market:
  - When funds liquidate assets following pecking order and do not increase leverage, EA MMFs face up to EUR 40 billion in redemptions (around 3 percent redemption rate).
  - Of those redemptions, 59 percent hit VNAV MMFs.
  - Funds would liquidate about two-thirds of their total EA MMF holdings, causing MMFs to liquidate around EUR 21.5 billion in CP, of which EUR 19.2 billion are issued by banks.
  - The secondary market for these securities is OTC with very limited depth; majority of securities term = 182 days and a large share held to maturity.
  - Small amounts of CP sales can be difficult in distress; March 2020 “dash-for-cash” cited as comparable.
  - Discounts on asset sales could trigger NAV deviations and further outflows from MMFs, including LVNAVs.
- Cross-fund redemptions and liquidation pecking order:
  - Under waterfall liquidation, total sales of other UCITS and AIFs around 2 to 2.3 percent of funds’ inventory of these shares.
  - Under pro rata liquidation, funds redeem around 5-6 percent of aggregate inventory of other funds’ shares, causing up to an additional 0.5 percentage points of redemptions (bond funds most impacted).

### Conclusions and policy implications
- Two-day market shock: material gaps to meet aggregate demand; unmet margin calls by NBFIs could trigger CCR losses for banks.
- Two-week horizon: defensive actions by NBFIs may amplify funding market shocks and produce second-round effects weakening solvency and liquidity of financial institutions.
- Correlated defensive strategies can propagate stress to core financial markets and institutions, creating financial stability risks.
- Findings support FSB efforts to enhance liquidity preparedness for margin and collateral calls for NBFIs (FSB, 2024d).
- FSAP recommendations:
  - ESMA should conduct EU-wide liquidity stress tests of the EU fund sector, including UCITS, AIF and MMF, focusing on systemic risk from direct and indirect interconnectedness and cross-border links—beyond NCAs’ supervisory stress tests.
  - Regular system-wide stress tests would strengthen cross-country and cross-sectoral financial risk monitoring and help calibrate CCR losses from banks’ exposures to NBFIs.
  - Data and resource intensity of the exercise requires strong governance and tools for regular implementation.
  - Key data gaps to address: instrument-level holdings for all funds, systematic information on standing credit lines with custodians, prevalence/details of LMTs, and fund-to-bank custodian mapping.

*Italic: Source — Extract from IMF EURO AREA chapter (paragraphs 38–51) as provided in the supplied content.*

### 52.      Taking a system-wide perspective in liquidity stress testing helps reveal intersectoral

### 52.      Taking a system-wide perspective in liquidity stress testing helps reveal intersectoral

### System-wide perspective: purpose and limitations
- Findings:
  - A system-wide view helps reveal intersectoral dependencies in terms of funding and liquidity and complements sector-specific stress tests conducted by NCAs and ESAs.
  - It could inform the calibration of risk buffers to strengthen resilience and address systemic risk.
  - Significant data gaps and hurdles in granular and comprehensive data sharing across EA/EU institutions limit the scope and depth of system-wide liquidity analysis.
- Recommendation:
  - ESMA, alongside ESRB and other ESAs, should continue to jointly develop a framework for system-wide liquidity stress testing tailored to the EA financial markets specificities and sectors heterogeneity.

### Funds data: enhancements to support stress testing and systemic risk analysis (Box 1)
- Identified data issues:
  - Ongoing ESMA work to revise and harmonize supervisory templates across fund segments is an opportunity to overcome current data limitations.
  - AIFMD requires relatively detailed data but gaps remain (e.g., granularity of asset holdings).
  - UCITS data collection is fragmented at the member-state level with different standards and limited access for ESMA.
  - Transaction-level data quality under EMIR and SFTR should be improved.
- Data and identifier recommendations:
  - All funds should be identified by LEI, in line with FSB recommendations of global LEI adoption.
  - For each fund, additional information should include: investment strategy with economically meaningful granularity; custodian bank with standardized information on committed and uncommitted temporary borrowing arrangements; shares’ primary currency; redemption schedules and fees by share class; footprint in terms of investors’ location and sector (jointly).
- Liquidity management tools and holdings:
  - Standardized data on LMTs, available to each fund including their triggers, would be critical; consider automatic notification to ESMA/NCAs when such tools are activated.
  - Fund holdings should be collected at instrument level, at least at quarterly frequency.
  - Reports should include standard identifiers (e.g., ISIN, CUSIP) and ancillary asset information: currency, asset type, country of exposure, credit quality, average traded volume (for example the CIC categories in Solvency II reports).
  - Asset exposures (short and long) should be complemented by information on encumbrance including the source by activity.
- Risk sensitivities and stress metrics:
  - Static balance-sheet and fund information should be complemented by risk sensitivities (e.g., collateral received and posted; DV01 for IRDs; CS01 for credit derivatives).
  - Include a harmonized definition of leverage, ideally with contribution of each activity (repo, derivatives, etc.).
  - Results of standard univariate stress tests (e.g., impact on NAV of equity markets down/up by 5 percent) should be reported.
  - These requirements align with recent FSB draft proposals related to NBFI leverage (FSB 2024b).
- Frequency and harmonization:
  - Fund returns and flows should be measured at higher frequency, by share class; frequency should align with redemption frequencies offered to investors.
  - High frequency data could be reported alongside holdings reports, with ESMA or NCAs able to request more timely information in periods of stress (analogous to banks’ liquidity templates).
  - Ensure definitions and identifiers are harmonized across datasets (e.g., overlap with data reported to NCBs for ECB statistical reporting).
- Data centralization and tools:
  - Data centralization and harmonization can ease data sharing across member states and EU authorities and support system-wide stress testing under the ESRB mandate.
  - Granular data require ad-hoc ICT tools and capacity for teams conducting financial stability analysis and stress testing; ESMA’s ability to leverage data depends on experts’ access to adequate tools.

### System-wide Liquidity Stress Test: scope, methodology, and shocks
- Scope and data coverage:
  - Includes liquidity demand from UCITS (using commercial Lipper data), AIF, and MMF (using AIFMD and MMF Regulation data), interaction in bond, repo and derivative markets, and margin requirements using transaction-level derivatives and repo data.
  - Incorporates additional liquidity demand on MMFs from insurance companies (data from EIOPA).
  - Data access constrained scope: excluded certain interactions due to lack of centralized, granular exportable data; channel of credit line usage not analyzed due to lack of access to AnaCredit; level of asset-holdings granularity did not allow identification of cross-holdings in all cases.
- Institutional perimeter and market shares (STeM summary):
  - 13,000 (EUR 7.1 tr) UCITS funds with holdings data in Lipper, 19,000 (EUR 7.4 tr) AIF and around 1,000 (EUR 1.5 tr) MMF.
  - EA insurance undertakings from a significant subset of EA jurisdictions.
  - Banks: 109 EA Significant institutions.
  - Market share: Funds: 50 percent of assets under management of UCITS Funds, over 90 percent of open-ended AIF, and over 90 percent of MMF.
  - Insurers: around 70 percent of EA insurers total investment.
  - Banks: Around 99 percent of the banking sector assets.
  - Data and baseline date: 2024: Q2 for stock; 2020–2024 for time series of market variables.
- Methodology and risk channels:
  - Liquidity measure based on cash and high-quality liquid assets.
  - Models of market depth integrate second-round effects from asset sales, calibrated using MiFID trading disclosures.
  - Limited incorporation of intersectoral linkages via deposits.
- Risks and buffers:
  - Risks: severe redemption shock; additional liquidity needs from margin calls on derivatives and collateral calls on repo; funding liquidity (outflows) and inability to sell assets; market liquidity leading to second-round price effects; margin calls leading to second-round liquidity effects; liquidity demand from funds to banking sector.
  - Buffers: stock of liquid assets, cash, and reverse repo positions.
- Tail shocks and calibration:
  - Market scenario calibrated based on expected shortfall at 0.1 percent of market factors marginal distributions over the holding period of the scenario (i.e., 2 or 10 days). Historical data covered the period January 2008 - October 2024.
  - Redemption shock calibrated based on historical net flows and returns data and linked to the market scenario.
  - Second round effects from price effects due to sales of assets.
  - Sensitivity analysis includes access to repo market.

### Appendix I: Stress Testing Matrix (STeM) — CCP reverse repo example
- Institutional perimeter:
  - 14 EU CCPs + 2 Tier 2 CCPs; Market share: 100 percent.
  - Data and baseline date: SFTR, CCP supervisory return, 2024 Q4.
- Methodology:
  - Type of exposures: bilateral reverse repo.
  - Conditional loss based on stressed collateral value.
  - Historical simulation over 5 years (2020-2024), Holding period of 5 days; cumulative distribution function of conditional losses computed for each CCP by sampling past 5 years market conditions with equal weights for each day.
  - Total loss for each day based on all possible combinations of conditionally independent defaults of counterparties. Probabilities of default based on National University of Singapore, Credit Research Initiative - https://nuscri.org/en/.
- Risks and buffers:
  - Risks: Counterparty credit risk.
  - Buffers: Skin in the game and default fund.
- Reporting format:
  - Output presentation: Expected shortfall at 0.1 percent; Cover 2.

### Appendix II: AIF data and balance-sheet reconstruction
- AIFMD reporting scope:
  - AIFMs report periodic information on exposures aggregated by asset type (equity instruments, debt securities, derivatives, physical assets and real estate, cash) and include breakdowns by geographical area and economic sector; leverage reported capturing direct investments and derivatives.
  - Framework applies to hedge funds, private equity, real estate, fund of funds, and ‘other’ AIFs (including equity funds, fixed income funds, commodity and infrastructure funds, mixed vehicles). LDI funds categorized as ‘Other’ AIFs.
- Strategy classification approach:
  - Funds assigned strategy labels based on asset type and geographic exposure (over 50 percent of NAV threshold); examples: Equity Europe, Bond Europe, Equity RoW, Bond RoW, Equity NA, Bond NA, Mixed Assets, FoF HF, FoF Other, RE, HF, LDI (distinguished by USD, GBP and EUR).
- Reconstruction methodology:
  - Stylized balance sheets reconstructed via top-down approach due to limited asset-level granularity.
  - Estimates net exposures (long minus short) and applies rules-based allocation; uses representative benchmark index weights for fixed income.
  - Does not aim to replicate actual fund-level portfolios but to generate consistent representation of AIF exposures.
- Allocation rules and categorizations:
  - Sovereign bond classification: EA sovereigns derived from reported EU government bond exposures with an assumed 10 percent share attributed to non-EA countries; allocated across credit ratings (AAA, AA, BBB) and maturity buckets using EA bond indices.
  - Other G10 sovereigns: For LDI funds denominated in USD or GBP entire exposure allocated to U.S. or U.K. sovereign bonds; maturity split follows fixed benchmark weights. For other funds, residual G10 exposure assigned to U.K. (non-LDIs), Japan and sovereigns using stylized proportions.
  - RoW sovereigns: net positions in non-G10 bonds and reallocated portions of EU sovereign exposure not attributed to the EA.
  - Debt securities disaggregated by issuer type (financial vs. non-financial) and credit quality (investment grade vs. high yield); EA financial institution exposures allocated using EA NAV share; EA non-financial corporate exposures divided between investment grade and high yield and disaggregated using representative Euro corporate bond indices.
  - Equity exposures allocated by region and sector using NAV shares; EA equity further disaggregated into Financials (primarily banks) and Non-financial corporates.
  - Real estate exposures derived from reported holdings in physical property and allocated by region based on NAV shares (EA, North America, RoW).
  - Remaining exposures grouped into: Cash and cash equivalents; Commercial paper; Collective investment undertakings (MMFs, ETFs, other fund structures); Residual exposures (structured finance, private equity and infrastructure, art and collectibles).
- Indices used for sovereign bond allocation:
  - A list of ICE BofA indices used (e.g., ICE BofA AAA Euro Government Index - Full market value; ICE BofA 0-1 Year US Treasury Index - Full market value; ICE BofA 0-1 Year UK Gilt Index - Full market value; ICE BofA 1-3 Year Japan Government Index - Full market value, etc.) as reported by ESMA.

*Source: International Monetary Fund.*

### Appendix III. Modelling the Impact of Shocks on Derivatives

### Appendix III. Modelling the Impact of Shocks on Derivatives

### Overall approaches
- Two modelling approaches are used to estimate the impact of market shocks on the mark-to-market (MtM) value of derivatives positions and to calculate variation margin (VM) calls:
  - Sensitivity analysis: simple to implement; considers market shocks to the underlying but does not consider changes in the risk-free rate.
  - Full repricing: more complex; models changes in all risk factors including changes to the underlying and to risk-free rates.

### Sensitivity analysis — instruments and formulas
- Equity options, credit default swaps (CDS) and bond futures are treated with sensitivity analysis. The change in MtM is calculated using sensitivities to changes in the underlying.

- Equity options:
  - Delta (훿) retrieved from commercial providers using the ISIN reported in EMIR.
  - MtM change formula: ∆푀푡푀_equit y options = 푁 ⋅ ∆푆 ⋅ 훿
    - 푁 is the notional.
    - ∆푆 is the change in stock prices stemming from the market scenario.

- Credit default swaps (single-name and index CDS):
  - Use DV01 (change in CDS value for a one-basis point change in the underlying spread) retrieved from commercial sources using 5Y CDS as reference.
  - MtM change formula: ∆푀푡푀_CDS = 푁 ⋅ ∆푆푝푟푒푎푑 ⋅ DV01
    - 푁 is the notional.
    - ∆푆푝푟푒푎푑 is the change in credit spreads from the market scenario.

- Sovereign bond futures:
  - Use basis point value (BPV) of the cheapest-to-deliver bond and conversion factor (CF) accounting for remaining maturity and coupon.
  - BPV and CF retrieved from commercial data sources.
  - MtM change formula: ∆푀푡푀_bond future = 푁 ⋅ ∆퐼푅 ⋅ BPV / CF
    - 푁 is the notional.
    - ∆퐼푅 is the change in interest rates from the market scenario.
    - BPV is the change in future price for a one basis point change in interest rates.
    - CF is the conversion factor.

- Short-term interest rate futures (STIR), typically on 3-Month reference rates (SOFR or ESTR):
  - Price definition: 푃 = 100 − (푛/12) ⋅ 푅_forward
    - 푛/12 is the maturity of the future (1/4 for 3-Month futures).
    - 푅_forward is the forward rate at the maturity of the future.
  - MtM change formula: ∆푀푡푀_STIR = 푁(푅_forward,0 − 푅_forward,shock)
    - 푅_forward,0 is the forward rate as of the reference date (end June 2024).
    - 푅_forward,shock is the new forward rate after the market scenario is applied to the interest rate curve.
    - 푁 is the notional.

### Full repricing — instruments, valuation, and assumptions
- Full repricing applied to equity futures, equity swaps and interest rate swaps; captures both underlying shocks and changes in risk-free rates.

- Equity futures:
  - Initial futures value: 퐹0 = 푆0 e^{(푟0 − 푑)푇}
    - 푆0 is the underlying stock value.
    - 푟0 is the risk-free rate over maturity 푇.
    - 푑 is dividend yield (set equal to zero).
  - MtM change: 퐹1 − 퐹0 = (1 + 훼) 푆0 e^{푟1 T} − 푆0 e^{푟0 T}
    - 훼 is the shock from the market scenario (훼 < 0).
    - 푟1 is the new risk-free rate after the shock.
    - Time horizon is short (two days or two weeks); maturity adjustment ignored (푇1 = 푇 instead of 푇1 = 푇 − 2 weeks).

- Equity swaps:
  - Modelled as two legs:
    - Equity leg (asset leg): PV_asset(t) = (푆t − 푆0)/푆0 ⋅ 푁 * 퐷
      - 퐷 is the discount factor over swap maturity.
      - 푆t is the future price of the stock (푆t = 푆 e^{rT}).
    - Floating leg: PV_float(t) = 푁 ⋅ 퐷 (푅_forward + s)
      - 푅_forward is the forward rate.
      - s is the spread, set equal to 2 percent (as in Jukonis (2022)).
  - Swap value: PV_swap(t) = PV_asset(t) − PV_float(t)
  - MtM change: ∆푀푡푀_equity swap = PV_swap(shock) − PV_swap(baseline)
  - Additional implementation details:
    - Equity swaps underlying retrieved using ISIN reported in EMIR.
    - Use maturity reported in EMIR; if maturity > 1 year, set maturity = 1 year to consider possible reset dates over contract maturity.

- Interest rate swaps (IRS):
  - Swap value decomposed into floating leg minus fixed leg:
    - IR Swap = ∑_{t=0}^{T} f_t N D_t ︸︷︷︸ floating leg − ∑_{t=0}^{T} r N D_t ︸︷︷︸ fixed leg
      - D_t are discount factors.
      - f_t are forward rates.
  - Discount factors derived from the risk-free curve based on overnight index swaps on SOFR, ESTR and SONIA for USD, EUR, and GBP swaps respectively.
  - EMIR data used to retrieve swap characteristics: notional, maturity date, fixed rate, payment frequencies, day count convention.
  - MtM change: ∆푀푡푀_IRS = PV_IRS(shock) − PV_IRS(baseline)

### Key modelling assumptions and data sources
- Sensitivities (delta, DV01, BPV, conversion factors) retrieved from commercial data providers.
- CDS DV01 reference tenor: 5Y CDS.
- Interest rate forwards reference date: end June 2024.
- Dividend yield in equity futures set to zero.
- Equity swap spread s = 2 percent.
- Equity swap maturities above one year are truncated to one year to reflect potential resets.
- Time horizon for futures valuation: two days or two weeks; maturity adjustment ignored.
- Discount factors for IRS derived from overnight index swaps on SOFR, ESTR, SONIA (currency specific).

### Repo data preparation (Appendix V summary points relevant to derivatives liquidity)
- "Trade-state" is a stock representation capturing transaction state at end of a reporting date.
- Scope: repo and repo equivalents (buy-sell-back) where one counterparty is in the funds in scope; special handling for CCP trades:
  - For EU CCP-involved trades: use CCP reports.
  - For non-EU CCPs: rely on reports from European clearing members.
  - For funds: consider trades reported by in-sample funds.
- Data preparation steps:
  - Pairing of double-sided transactions and deduplication.
  - In double reporting, quantitative fields use the average of reported values (e.g., notional amount, collateral amount).
- Outstanding repos: contracts with start date ≤ reference date and maturity > reference date.
- Cleaning procedure: outlier identification, removal of stale records, removal of records with implausible collateral-to-loan ratios.
- For details on cleaning methods, see section "Statistical methods" in ESMA (2024b).

*Source: Appendix III. Modelling the Impact of Shocks on Derivatives.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2025/english/1eurea2025008.pdf_