## wpiea2021109-print-pdf

## Source details

**Canonical URL:** [wpiea2021109-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021109-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021109-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021109-print-pdf.pdf.json)

---

### Introduction and context
- U.S. asset management industry growth: USD 11 trillion in 2007 to USD 23 trillion end-2019.
- Mutual fund holdings of U.S. corporate bonds: 12 percent of amounts outstanding in 2007 to 20 percent end-2019.
- Main policy concern: liquidity mismatch — funds offer daily liquidity while holding assets with varying market liquidity, creating potential for large sales and price impact under large redemptions.
- Three central questions addressed:
  - To what extent can bond funds withstand severe but plausible redemption shocks?
  - Could sales from bond funds have a sizeable impact on bond markets?
  - Which fund types are more vulnerable to distress from other funds?

### Methodology and novel contributions
- Extensions and tools:
  - Extends IMF liquidity stress testing framework by estimating price impact of fund sales.
  - Applies Diebold and Yilmaz (2014) connectedness framework to mutual funds.
  - Develops Flows in Distress (FiD) using a copula-dependence framework to estimate expected redemptions for a fund conditional on another fund’s distress.
    - Dependence can be positive (complementarity/common exposures) or negative (substitution).
- Scenario application:
  - Applies liquidity stress framework to March 2020 stress, designing a macrofinancial scenario based on observed valuation declines and estimating fund outflows and asset sales.

### Liquidity stress test framework and calibration
- Principal steps:
  - Calibrate redemption shock from historical net flows by fund category using Expected Shortfall (ES) at level α.
  - Measure Highly Liquid Assets (HLA) using liquidity weights (based on Basel Committee LCR weights).
  - Compute Redemption Coverage Ratio (RCR) = HLA / redemption shock; RCR < 1 signals liquidity shortfall.
  - Estimate sales and compute price impact by comparing sales to market depth.
- Liquidation strategies considered:
  - Vertical slicing (pro rata across assets).
  - Waterfall (sell most liquid assets first) with HQLA ordering: cash → AAA-AA sovereign → A sovereign → AAA-AA corporate → BBB sovereign → A corporate → AAA-AA securitized → BBB corporate → unrated sovereign/corporate/securitized.
  - Mixed approaches (cash first, then slicing).
- Market depth and price impact formulas:
  - Market depth MD(τ) = ξ × (ADV / σ) × √τ.
  - Price impact PPI(τ) = EV_sales / MD(τ).

### Data, sample and portfolio composition
- Sample: 2,743 funds, NAV about USD 6.4 trillion as of end-2019.
- Fund categories (NAV in US$ bn; number of funds in parentheses):
  - Corp. IG: 2,427 (608)
  - Mixed funds: 1,752 (792)
  - Municipal: 799 (567)
  - Multisector: 432 (182)
  - Government: 326 (161)
  - Corp. HY: 257 (192)
  - Global: 247 (87)
  - Loan funds: 91 (58)
  - EM funds: 66 (96)
  - Total: 6,398 (2,743)
- HQLA shares heterogeneous by category:
  - HQLA ranges from around 6 percent of NAV for HY and loan funds up to more than 80 percent for government and municipal bond funds.
- Monthly data on NAV, monthly net flows over 2007–2019; monthly portfolio composition and credit quality used.

### Key empirical findings — liquidity resilience and shortfalls
- Aggregate resilience:
  - More than 90 percent of the funds would have enough highly liquid assets to meet investors’ redemptions.
- Notable exceptions:
  - Most funds exposed to HY and leveraged loans would not have enough highly liquid assets and would need to sell less liquid securities, assuming no use of liquidity risk management tools.
- Redemption shock calibration (Net outflows in % of NAV, ES at 3%):
  - Municipal 7
  - Mixed funds 9
  - Corp. IG 13
  - Multisector 13
  - Loan funds 13
  - Global 14
  - Government 14
  - HY 15
  - EM funds 17
- Typical cash buffers: around 5 percent; multi-sector funds can have higher cash buffers due to derivatives usage.

### Price impact of asset sales — market depth and impacts
- Asset-class market depth and impact (selected entries):
  - UST: ADV 5450 (US$ bn), Average volatility 0.28%, 2008 volatility 0.55%, Market Depth 77,857 (US$ bn) → Impact of $1bn sale 0.1 bps; Stressed Market Depth 39,636 → Impact 0.3 bps
  - Corp. IG: ADV 210.3, Average volatility 0.30%, 2008 volatility 0.65%, Market Depth 2,800 → Impact 3.6 bps; Stressed Market Depth 1,292 → Impact 7.7 bps
  - Corp. HY: ADV 120.3, Average volatility 0.31%, 2008 volatility 1.07%, Market Depth 1,548 → Impact 6.5 bps; Stressed Market Depth 449 → Impact 22.3 bps
  - Leveraged loans: ADV 30, Average volatility 0.18%, 2008 volatility 0.64%, Market Depth 556 → Impact 18.0 bps; Stressed Market Depth 156 → Impact 64.0 bps
  - EM debt: ADV 80, Average volatility 0.40%, 2008 volatility 1.36%, Market Depth 750 → Impact 13.3 bps; Stressed Market Depth 221 → Impact 45.3 bps
  - Municipal bonds: ADV 110, Average volatility 0.19%, 2008 volatility 0.64%, Market Depth 2,316 → Impact 4.3 bps; Stressed Market Depth 688 → Impact 14.5 bps
  - Securitized Agencies: ADV 2200, Average volatility 0.19%, 2008 volatility 0.44%, Market Depth 46,316 → Impact 0.2 bps; Stressed Market Depth 20,000 → Impact 0.5 bps
  - Securitized: ADV 20, Average volatility 0.11%, 2008 volatility 0.27%, Market Depth 727 → Impact 13.8 bps; Stressed Market Depth 296 → Impact 33.8 bps
  - Equities: ADV 320, Average volatility 1.12%, 2008 volatility 3.60%, Market Depth 11,429 → Impact 0.9 bps; Stressed Market Depth 3,556 → Impact 2.8 bps
- Price impact ranges (slicing approach):
  - Normal times: 50 to 200 basis points.
  - Stress periods: 150 to 700 basis points.
- Vertical slicing produces larger market impact than waterfall; waterfall mutes immediate price impact but leaves remaining investors with less liquid portfolios.

### Quantified impacts from March 2020 scenario and price-impact estimates
- Market moves over March 4–18, 2020 used to construct macrofinancial scenario:
  - Equities -23%
  - IG Gov. -40 bps
  - EM Gov 421 bps
  - Corp IG 170 bps
  - Corp HY 359 bps
  - Municipal 90 bps
- Fund exposures in the sample: USD 1,094bn to IG and USD 360bn to HY.
- Estimated additional price decline due to fund sales:
  - HY: 3.5 percent additional decline.
  - IG: 2.8 percent additional decline.
  - These additional declines account for around 30 percent of observed March 4–18 shocks (historical declines: HY -15 percent, IG -10 percent).
- Estimated absolute losses from fund sales:
  - IG: around USD 30bn.
  - HY: around USD 13bn.
- Example aggregate estimated fund returns, redemption shocks and redemptions (selected):
  - EM: Return shock -17%, Redemption shock -16.3%, Redemptions US$ bn -7.9
  - HY: Return shock -8%, Redemption shock -5.7%, Redemptions US$ bn -12.8
  - IG: Return shock -5%, Redemption shock -3.6%, Redemptions US$ bn -76.0
  - Mixed: Return shock -15%, Redemption shock -3.9%, Redemptions US$ bn -55.0
  - Muni: Return shock -4%, Redemption shock -2.5%, Redemptions US$ bn -19.3

### Contagion analysis — Flows in Distress (FiD) and return spillovers (Diebold–Yilmaz)
- FiD definition and estimation:
  - FiD_{B→A} = E(F_A | F_B < α); Student t-copula used to capture tail dependence; Monte Carlo simulations with 100,000 draws for FiD computation.
  - Marginals: logistic distribution fitted to flows; parameters estimated by maximum likelihood.
- Robustness: Gaussian copula used as a check; Student copula yields larger expected net flows due to tail dependence but qualitative ranking is similar.
- Selected expected net flows (Student copula, values in % of NAV; rows = distressed category, columns = impacted category):
  - IG row (To (AVG)): -2.2, -1.7, -1.8, -0.5, -2.1, -1.0, -2.3, -4.6, -0.1, From (AVG) -1.8
  - Muni row (To (AVG)): -1.9, -1.6, -3.0, -0.1, -2.3, -0.4, -0.8, -2.1, 3.7, From (AVG) -0.7
  - Loan row (To (AVG)): 0.6, -1.0, 1.2, -0.6, 1.2, 1.0, -1.5, -1.7, -11.0, From (AVG) -0.1
- Cumulated expected net flows in USD (conditional on another category being in distress), select entries:
  - Muni row cumulated: Muni -24, IG -46, Multi -2, Gov -7, Global -1, EM -1, Mixed -2, HY -5, Loan 5, Cumulated -59
  - IG row cumulated: IG -15, Muni -52, Multi -9, Gov -7, Global -3, EM -3, Mixed -7, HY -40, Loan -47
- Systemic roles and vulnerability:
  - IG corporate bond funds, multi-strategy bond funds, and municipal bond funds identified as systemic/propagators due to size and connectivity.
  - EM, HY and loan funds are net receivers of spillovers and more vulnerable.
- Returns-based connectedness (Diebold–Yilmaz):
  - Net spillovers: EM, Loan, HY are net receivers; IG, municipal and government bond funds are net senders.
  - Average spillovers example (IG row, To (avg) entries in % contribution): IG 1.5, Muni 1.6, Gov 1.6, HY 1.6, Mixed 1.3, Loan 1.4, Multi 1.3, EM 1.2, From (avg) 1.4.

### COVID-19 March 2020 evidence and joint distress probabilities
- Extreme outflows observed:
  - In March 2020, IG bond funds in the sample experienced outflows of more than 5.6 percent of NAV (prior peak 2.9 percent in October 2008).
- Joint probability of all fund categories being in distress (net flows below given percentiles; two samples each with 100 observations):
  - Distress level 1%: 2011–2019 sample = 0.49% ; 2012–2020 sample = 0.85%
  - Distress level 5%: 2011–2019 sample = 0.58% ; 2012–2020 sample = 0.99%
  - Distress level 10%: 2011–2019 sample = 0.74% ; 2012–2020 sample = 1.20%
- Empirical conclusion: March 2020 exhibited elevated joint distress probabilities and demonstrated that mutual funds met redemptions without suspending redemptions, supported by Federal Reserve interventions.

### Policy-relevant implications and recommendations
- Monitoring and supervisory focus:
  - Prioritize monitoring funds with large exposures to HY and leveraged loans due to limited HQLA.
  - Monitor IG corporate bond funds, municipal bond funds, and government bond funds for interconnected distress channels.
- Liquidity risk management and structural options:
  - Strengthen risk management frameworks, supervisory reporting, and liquidity stress testing for vulnerable fund categories.
  - Consider liability–asset liquidity matching (e.g., lower redemption frequency to one week or one month for less-liquid strategies).
  - Encourage availability and use of liquidity management tools: gates, deferring redemptions, swing pricing (swing pricing available in U.S. mutual funds but not implemented widely).
  - Require asset managers to consider market impact when choosing liquidation methods; consider manager- and SEC-led stress tests.
- Valuation and market infrastructure:
  - Authorities should monitor and support timely and reliable valuations of assets in funds’ portfolios.
  - Consider system-wide stress simulations to estimate cross-system impacts and identify key vulnerabilities.
- Use of FiD and connectedness tools:
  - FiD quantifies expected redemptions conditional on other funds’ distress and helps identify systemic or vulnerable funds.
  - Connectedness metrics identify funds that, when distressed, are associated with larger redemptions elsewhere.
- Forward uses:
  - Tools can perform what-if scenarios, support fund-level liquidity stress tests, enable macrofinancial stress tests, and be adapted to other entities (banks, countries) to estimate conditional outflows.

*Source: wpiea2021109-print-pdf*

### References..............................................................................................................

### wpiea2021109-print-pdf - References

### Introduction and context
- The U.S. asset management industry grew from USD 11 trillion in 2007 to USD 23 trillion end-2019.
- Mutual fund holdings of U.S. corporate bonds increased from 12 percent of amounts outstanding in 2007 to 20 percent end-2019.
- Main policy concern highlighted: liquidity mismatch—funds offer daily liquidity to investors while holding assets with varying degrees of market liquidity, creating potential for large sales and price impact under large redemptions.
- The paper addresses three questions:
  - To what extent can bond funds withstand severe but plausible redemption shocks?
  - Could sales from bond funds have a sizeable impact on bond markets?
  - Which types of funds are more vulnerable to distress from other funds?

### Methodology and novel contributions
- Extends the IMF liquidity stress testing framework (IMF, 2015b, 2017, 2018) by estimating the price impact of fund sales.
- Applies the connectedness framework of Diebold and Yilmaz (2014) to mutual funds.
- Develops a new measure "Flows in Distress (FiD)" using a copula-dependence framework to estimate expected redemptions for a fund when another fund is in distress. The dependence can be:
  - Positive (complementarity/common exposures or similar investor bases leading to simultaneous outflows),
  - Negative (substitution effects, e.g., between HY bond funds and sovereign bond funds).
- Applies the liquidity stress framework to March 2020 stress, designing a macrofinancial scenario based on observed valuation declines and estimating fund outflows and asset sales.

### Key empirical findings
- Liquidity resilience:
  - "Most U.S. mutual funds would have enough highly liquid assets to meet investors’ redemptions."
  - Exceptions: "Most funds exposed to HY and leveraged loans would not have enough highly liquid assets and would need to sell liquid securities in their portfolio."
- Price impact of asset sales:
  - Price impact ranges:
    - "50 to 200 basis points (bps) in normal times,"
    - "150 to 700 bps during stress periods."
  - "For IG and HY corporate bonds, the price impact can amount to around 300bps during stress period."
  - These estimates are consistent with IMF (2020c) findings on asset sales during the COVID-19 crisis.
- Interconnectedness and vulnerability:
  - Some fund categories are particularly vulnerable to distress from other categories: "IG corporate bond funds, municipal bonds funds and government bond funds."
  - Vulnerability implies multiple funds can be simultaneously forced to sell assets, amplifying downward price pressure on the asset classes they hold.
  - Given the large size of these fund categories, such amplification effects can be large.
  - The analysis suggests certain fund categories "may require more monitoring than others given their levels of interconnectedness."

### Scenario application: March 2020 stress
- The paper implements a macrofinancial scenario reflecting the March 2020 valuation declines, estimating fund outflows and the resulting asset sales and valuation effects during that period.
- Results include estimated fund returns, investor outflows, and changes in valuation over March 4–18, 2020 (tabulated/figured in the source).

### Policy-relevant implications
- Monitoring priorities:
  - Focus monitoring on funds with large exposures to HY and leveraged loans due to limited highly liquid assets.
  - Monitor IG corporate bond funds, municipal bond funds, and government bond funds for interconnected distress channels.
- Market impact considerations:
  - Simultaneous sales by multiple funds can cause material price declines, particularly in stressed market conditions where price impact magnitudes rise substantially (up to 700 bps in stress scenarios).
- Use of FiD and connectedness measures:
  - FiD provides a quantitative way to estimate expected redemptions conditional on other funds’ distress and to identify systemic or vulnerable funds.
  - Connectedness metrics help identify funds that, when distressed, are associated with larger redemptions in the rest of the system.

*Source: wpiea2021109-print-pdf - References..............................................................................................................*

### section III details the models used, Section IV provides the results, Section V puts the results

### wpiea2021109-print-pdf - section III details the models used, Section IV provides the results, Section V puts the results

### Literature review: liquidity risk, fire sales, contagion
- Mutual funds are subject to bank-run-like risks: following outflows funds use cash buffers first, then sell illiquid assets, generating negative NAV effects and first-mover advantages (Zeng (2017); Chen et al. (2010); Coval and Stafford (2007); Goldstein et al. (2015)).
- Corporate bond funds exhibit a concave flow-return relationship; higher outflows follow negative performance than positive performance (Goldstein et al., 2015).
- Margins can be destabilizing and market liquidity and funding liquidity can create liquidity spirals (Brunnermeier and Pedersen, 2009).
- Cash holdings:
  - Funds build cash when receiving inflows and draw down cash during outflows (Chordia, 1996).
  - Cash holdings are positively correlated to asset illiquidity and flow volatility; less liquid bond funds hoard cash more (Morris et al., 2017; Chernenko and Sunderam, 2016; Jiang et al., 2016; Zhang, 2019).
- Heterogeneous liquidation strategies:
  - Institutional funds may sell most liquid bonds first in stress to preserve short-term performance, increasing liquidity risk; retail funds may preserve portfolio liquidity to reduce run incentives (Dotz and Weth, 2019).
  - Outflows from mutual funds exert upward pressure on corporate bond yields; outflows significant, inflows often have no significant effect (Coudert and Salakhova, 2020).
- Network and contagion findings:
  - Shocks to funding markets create stronger return correlations than fundamentals imply (Boyson et al., 2010).
  - During crises correlation networks can show increased interconnectedness while physical lending networks can show decreased interconnectedness; correlation networks may forecast financial crisis while physical networks forecast liquidity problems (Brunetti et al., 2019).
  - Various network topologies differ in robustness and fragility (Roukny et al., 2013); network topology can forecast macro outcomes (Huang and Wang, 2020).

### Overview of the methodology (Section III)
- Two principal pillars:
  - Liquidity stress test extended from IMF FSAP frameworks (IMF, 2015b; IMF, 2017; ESMA, 2019) with explicit modelling of price impact of fund sales on underlying asset classes.
  - Contagion analysis with two sub-pillars: flow-based dependence (Flows in Distress, FiD) and return-based spillover (Diebold and Yilmaz (2014) connectedness).
- Analysis performed at fund level and by fund categories.
- Liquidity stress test steps:
  - Calibrate redemption shock from historical net flows by fund categories.
  - Measure Highly Liquid Assets (HLA) using liquidity weights (based on Basel Committee LCR weights).
  - Estimate volume and composition of sales to meet redemptions under alternative liquidation strategies; compare sales to market depth to compute price impact at asset-class level.
- Contagion analysis:
  - Flow-based pillar: FiD = expected net flows for category A conditional on category B being in distress.
  - Return-based pillar: Diebold–Yilmaz (DY) connectedness on a sample of largest mutual funds per category to measure directional spillovers and identify vulnerable and systemic funds.

### Liquidity stress test: calibration and measures
- Redemption shock calibration:
  - ES (expected shortfall) used instead of VaR to include extreme values below the VaR; redemption shock calculated as the average worst outflows experienced by funds (ES at level α).
  - Homogeneity assumption: each fund in the same category faces the same outflow (percent of NAV) calibrated on the distribution of all flows within that category; robustness checks include heterogeneity approach with thresholds 1 percent, 5 percent and 10 percent (Appendix XIII in IMF (2020b) referenced).
- Measure of Highly Liquid Assets (HLA):
  - HLA for fund i: HLA_i = sum_{k=1}^n ω_{i,k} × s_{i,k} where ω_{i,k} are liquidity weights assigned to each security s_{i,k}.
  - Liquidity weights follow Basel Committee LCR weights by asset class and, for fixed income instruments, by credit quality (Table 1 referenced).
- Redemption Coverage Ratio (RCR):
  - RCR = HLA / redemption shock (formulation presented); when RCR < 1 the fund has a liquidity shortfall equal to redemption shock minus HLA.

### Sales, liquidation strategies, and market impact
- Liquidation strategies considered:
  - Vertical slicing (pro rata across assets).
  - Waterfall (sell most liquid assets first) following HQLA ordering: cash → AAA-AA sovereign bonds → A sovereign bonds → AAA-AA corporate bonds → BBB sovereign bonds → A corporate bonds → AAA-AA securitized assets → BBB corporate bonds → unrated sovereign/corporate/securitized assets.
  - Mixed approaches where cash used first then slicing.
- Trade-offs:
  - Slicing preserves portfolio liquidity profile but may force sales of less liquid assets with higher trading costs (worse price impact).
  - Waterfall lowers immediate price impact but leaves remaining investors with a less liquid portfolio and may misalign portfolio with fund prospectus.
- Market depth and price impact:
  - Market depth over horizon τ: MD(τ) = ξ × (ADV / σ) × √τ, where ξ is a scaling factor, ADV is average daily trading volume, and σ is asset volatility (Cont and Schaaning (2017) approach).
  - Price impact PPI(τ) = EV_{sales} / MD(τ).
- Price impact illustrations and asset-class measures (Table 2):
  - Example: under normal conditions, sale of USD 1 billion of IG corporate bonds would lead to a decline in bond prices of around 4 basis points and 8 basis points under stressed conditions (asset-class level).
  - Example: if sales of IG corporate bonds amount to USD 10 billion, under normal trading conditions market depth is USD 2,800 billion, implying a price impact of 0.36 percent or 36 basis points.
- Asset-class market depth and impact numbers (excerpt from Table 2; columns correspond to normal and stressed trading conditions):
  - UST: ADV (US$ bn) 5450, Average volatility 0.28%, 2008 Volatility 0.55%, Market Depth (US$ bn) 77,857 → Impact of $1bn of sale (in bps) 0.1; Stressed Market Depth (US$ bn) 39,636 → Impact 0.3
  - Corp. IG: ADV 210.3, Average volatility 0.30%, 2008 Volatility 0.65%, Market Depth 2,800 → Impact 3.6 bps; Stressed Market Depth 1,292 → Impact 7.7 bps
  - Corp. HY: ADV 120.3, Average volatility 0.31%, 2008 Volatility 1.07%, Market Depth 1,548 → Impact 6.5 bps; Stressed Market Depth 449 → Impact 22.3 bps
  - Leveraged loans: ADV 30, Average volatility 0.18%, 2008 Volatility 0.64%, Market Depth 556 → Impact 18.0 bps; Stressed Market Depth 156 → Impact 64.0 bps
  - EM debt: ADV 80, Average volatility 0.40%, 2008 Volatility 1.36%, Market Depth 750 → Impact 13.3 bps; Stressed Market Depth 221 → Impact 45.3 bps
  - Municipal bonds: ADV 110, Average volatility 0.19%, 2008 Volatility 0.64%, Market Depth 2,316 → Impact 4.3 bps; Stressed Market Depth 688 → Impact 14.5 bps
  - Securitized Agencies: ADV 2200, Average volatility 0.19%, 2008 Volatility 0.44%, Market Depth 46,316 → Impact 0.2 bps; Stressed Market Depth 20,000 → Impact 0.5 bps
  - Securitized: ADV 20, Average volatility 0.11%, 2008 Volatility 0.27%, Market Depth 727 → Impact 13.8 bps; Stressed Market Depth 296 → Impact 33.8 bps
  - Equities: ADV 320, Average volatility 1.12%, 2008 Volatility 3.60%, Market Depth 11,429 → Impact 0.9 bps; Stressed Market Depth 3,556 → Impact 2.8 bps
  - Sources for table: Refinitiv Datastream, SIFMA, JPMorgan, EMTA, IMF staff; 2008 volatility estimated over September–December 2008.

### Contagion analysis: flows (FiD) and returns (DY)
- Flows in Distress (FiD) dependence approach:
  - FiD_{B→A} = E(F_A | F_B < α) — expected net flows for funds of category A conditional on funds of category B being in distress (flows below threshold α).
  - Copula modelling:
    - Student t-copula chosen to capture tail dependence (allows higher dependence during stress); Gaussian copula used for robustness.
    - Marginal distributions parametric: logistic distribution fits best for flows by type; parameters estimated by maximum likelihood.
    - Student t-copula parameters (including correlation matrix Σ and degrees of freedom ν) estimated by maximum likelihood; FiD computed by Monte Carlo simulations (100,000 simulations) with numerical integration cross-check.
- Spillovers using Diebold–Yilmaz (DY) framework:
  - DY measures directional and system-wide connectedness via variance decomposition from a VAR on return volatilities.
  - VAR: x_t = sum_{i=1}^p Φ_i x_{t−i} + ε_t where x_t is vector of return volatility.
  - Generalized forecast error variance decomposition used to compute directional spillovers θ_{i,j}^g(H).
  - Connectedness measures computed:
    - Total connectedness R^H
    - Inward connectedness R_{i←.}^H
    - Outward connectedness R_{.←i}^H
    - Net connectedness R^H = R_{.←i}^H − R_{i←.}^H
    - Pairwise directional connectedness R_{i←j}^H = θ_{i,j}^g(H)
  - VAR estimation is performed on log-volatilities with automatic selection of lasso penalty for shrinkage (Zou and Hastie, 2005).

### Risks and vulnerabilities in the U.S. mutual fund industry (Section IV: sample and descriptive results)
- Sample and coverage:
  - Sample includes 2,743 funds for a net asset value of about USD 6.4 trillion as of end-2019 (commercial data covering the mutual fund universe tracked by the Investment Company Institute).
  - Funds subdivided into nine categories: Corp. IG, Corp. HY, Loan funds, Global bond funds, EM funds, Government bond funds, Municipal bond funds, Mixed funds, Multi-sector funds.
- Table 3 (sample summary by category; Net asset Value (US $ bn) and Number of funds):
  - Corp. IG: 2,427 (608)
  - Mixed funds: 1,752 (792)
  - Municipal: 799 (567)
  - Multisector: 432 (182)
  - Government: 326 (161)
  - Corp. HY: 257 (192)
  - Global: 247 (87)
  - Loan funds: 91 (58)
  - EM funds: 66 (96)
  - Total: 6,398 (2,743)
  - Sources: Morningstar, ICI, IMF staff calculations.
- Portfolio composition and HQLA:
  - HY and loan funds invest mainly in non-IG securities; government and municipal bond funds invest mainly in IG sovereign bonds.
  - Most funds have relatively limited cash buffers (around 5 percent).
  - Multi-sector funds can have high cash buffers due to derivatives usage (variation margin needs).
  - HQLA by category (HQLA shares heterogeneous across categories):
    - HQLA ranges from around 6 percent of NAV for HY and loan funds up to more than 80 percent for government and municipal bond funds.
    - Within categories, HQLA is homogeneous (first and third quartiles similar).
- Redemption shock calibration and values:
  - Redemption shock calibrated at the 3 percent expected shortfall (average of the lowest net flows below the 3rd percentile), consistent with ESMA (2019).
  - Resulting redemption shocks range from 7 percent of NAV for municipal funds to 17 percent for EM bond funds.
  - Using historical data by fund category implies identical shocks for all funds within a category (homogeneity), enabling comparability across funds while reflecting different historical redemption patterns.
- Data and periods:
  - Monthly data on fund NAV, monthly net flows over 2007–2019, portfolio composition and credit quality collected for each fund.

*Italic: Source — IMF staff (from the provided content of wpiea2021109-print-pdf).*

### Appendix XIII in IMF (2020b).

### Appendix XIII in IMF (2020b)

### Overview
- Funds within a category appear homogeneous in portfolio composition (asset type and credit quality), HQLA measure or distribution of net flows; funds are heterogeneous across categories.
- Monthly net flows are aggregated by fund categories for dependence analysis.
- Weekly returns for the ten largest funds of each category (sample of 90 funds) are used for spillover analysis.
- HQLA measured as "HQLA in % of NAV."

### Liquidity Stress Test — Key Findings
- More than 90 percent of the funds would have enough highly liquid assets to meet investors’ redemptions.
- Most funds exposed to HY and leveraged loans would not have enough highly liquid assets and would need to sell less liquid securities, assuming no use of liquidity risk management tools.
- Vulnerabilities relate to high exposure of HY and loan funds to less liquid asset classes (HY corporate bonds which include leveraged loans), which have a liquidity weight of zero for HQLA calculations.
- Results are similar when reported by number of funds or share of NAV.
- Fund categories identified as most vulnerable in the liquidity stress tests are also the smaller ones in terms of NAV in the sample.

- Redemption shock calibration by fund category (Net outflows in % of NAV):
  - Municipal 7
  - Mixed funds 9
  - Corp. IG 13
  - Multisector 13
  - Loan funds 13
  - Global 14
  - Government 14
  - HY 15
  - EM funds 17

- Share of funds (by category) whose Redemption Coverage Ratio (highly liquid assets to redemption) is below one (illustrative categories shown):
  - Global, Gov, Mixed, Muni, IG, Multi, EM, HY, Loan — percentages shown in original figure (not reproduced numerically here).

### Price Impact — Key Findings and Estimates
- Asset sales to meet redemptions can have sizeable market impact depending on shock size, asset sales, and market depth.
- Vertical slicing (pro-rata sales) produces larger market impact than a waterfall approach (sell most liquid assets first).
- Under slicing, funds exposed to less liquid asset classes (EM bonds, HY corporate bonds, leveraged loans) would sell large amounts relative to market size, creating downward pressure.
- Under waterfall, price impact is muted but remaining investors hold less liquid portfolios, which can amplify first-mover advantage.
- Trade-off at fund level: reduce price impact vs. maintain portfolio allocation aligned with investment objective.

- Price impact ranges under slicing approach:
  - Normal times: 50 to 200 basis points.
  - Stress periods: 150 to 700 basis points.

- Example volumes and impacts (Volumes of sales in US$bn and price impact in basis points):
  - EM: Volume of sales (Slicing) 34; Waterfall 3; Price impact not fully tabulated in source snippet.
  - HY Corp: Volume of sales (Slicing) 30; Waterfall 19; Price impact Normal 66, Stressed 788? (source table shows mixed entries; preserve exact table structure in source).
  - IG Corp: Volume of sales (Slicing) 103; Waterfall 36; Price impact Normal 87, Stressed 1657? (source table entries combined volume and bps; refer to table for exact mapping).
  - Loan funds: Volume of sales (Slicing) 9; Waterfall 15; Price impact Normal 65, Stressed 562? 
  - Muni: Volume of sales (Slicing) 47; Waterfall 20; Price impact Normal 26, Stressed 794? 

- For less liquid asset-exposed funds (EM, loan funds), absolute sale volumes would be limited (less than USD 10 bn), explaining lower price impact despite low market liquidity.
- Waterfall approach: price impact muted (less than 100 bps under normal conditions and less than 200 bps under stress for most asset classes).

### Contagion Analysis (Funds-in-Distress, FiD)
- Funds investing in less liquid asset classes (HY, EM, loan funds) are more vulnerable to shocks (outflows or volatility shocks) from other funds.
- Shocks to IG corporate bond funds, government and municipal bond funds are associated with higher system stress (larger redemptions for other funds or higher volatility).

- Expected net flows when other fund categories are in distress (expected net flows conditional on distress in % of NAV; grey cells show net flows for funds in distress):
  - Example row "IG" (To (AVG)):
    - IG -2.2, Mixed -0.5, Multi -2.3, HY -1.7, Muni -1.8, EM -4.6, Global -1.0, Gov -2.1, Loan -0.1, From (AVG) -1.8
  - Example row "Muni":
    - Muni -1.9, IG -0.1, Multi -0.8, HY -1.6, Muni -3.0, EM -2.1, Global -0.4, Gov -2.3, Loan 3.7, From (AVG) -0.7
  - (Full matrix presented in source; entries preserve expected net flows in % of NAV.)

- Some categories are identified as systemic: IG corporate bond funds, multi-strategy bond funds, and to a lesser extent municipal bond funds, mixed funds and government bond funds.
  - Example: when IG bond funds are in distress, three other fund categories face outflows around or higher than 2 percent (Municipal, multi and EM bond funds), corresponding to the 5th percentile of their net flows distribution.

- Cumulated expected net flows in USD (conditional on another category being in distress), select entries (Net flows in US$ bn):
  - Muni row (Cumulated flows):
    - Muni -24, IG -46, Multi -2, Gov -7, Global -1, EM -1, Mixed -2, HY -5, Loan 5, Cumulated -59
  - IG row:
    - IG -15, Muni -52, Multi -9, Gov -7, Global -3, EM -3, Mixed -7, HY -40, Loan -47
  - (Full matrix presented in source; negative values indicate outflows. Cumulated flows exclude net flows from funds in distress.)

- Municipal bond funds generate largest aggregate impact in USD (USD 59 billion in outflows in the cumulated table), driven by outflows from IG funds (close to 2 percent of NAV), the largest category in the sample. Government bond funds also have large aggregate impact. IG and multi strategy bond funds show widespread medium impacts across categories.

- FiD analysis complements liquidity stress tests and provides a comprehensive view: HY and EM bond funds vulnerable in liquidity stress tests are also likely to experience outflows when other funds are in distress; IG and municipal funds can propagate shocks due to size and market footprint even if individually liquid enough.

### Spillover Analysis (Returns-Based Interconnectedness)
- Net spillovers (outward minus inward) show funds exposed to less liquid asset classes (EM, loan, HY) are net receivers of spillovers and thus more vulnerable.
- IG, municipal and government bond funds are net senders of spillovers to the rest of the fund industry, possibly due to joint stress in underlying asset classes.

- Average net spillovers by fund category (Averages of the net spillovers):
  - EM, Loan, HY, Multi, Mixed, Muni, Gov, IG — plotted with values in source; example numeric axis ranges from -20 to 20 (average net spillovers).

- Average spillovers table (Average spillovers from and to funds within each category):
  - IG row (To (avg)): IG 1.5, Muni 1.6, Gov 1.6, HY 1.6, Mixed 1.3, Loan 1.4, Multi 1.3, EM 1.2, From (avg) 1.4
  - Muni row: Muni 1.5, IG 1.4, Gov 1.5, HY 1.4, Mixed 1.3, Loan 1.3, Multi 1.2, EM 1.1, From (avg) 1.3
  - EM row: EM 0.8, IG 0.8, Muni 0.8, Gov 0.9, HY 0.9, Mixed 1.0, Loan 1.1, Multi 1.0, From (avg) 0.9
  - (Full table in source; ranges annotated: <1, [1;1.25], [1.25;1.5], >1.5.)

- Consistency: interconnectedness results (returns volatility) align with FiD flow-based results. Flow-return relationship can generate feedback: funds’ poor returns predict outflows; outflows can depress NAVs for less liquid asset funds due to liquidation costs, fueling further outflows.

### The COVID-19 Crisis — March 2020 Case Study
- Acute liquidity stress in March 2020 featured very large outflows from bond mutual funds amid sharp deterioration of market liquidity.
- In March 2020, IG bond funds in the sample experienced outflows of more than 5.6 percent of NAV, above previous peak of 2.9 percent in October 2008.

- Comparison of redemption shocks (Net outflows in % of NAV) with and without 2020 data:
  - Municipal 7 (both columns show 7)
  - Mixed funds 9 (both columns show 9)
  - Corp. IG 13 (both columns show 13)
  - Multisector 13 (13 without 2020 data; 15 incl. 2020 data)
  - Loan funds 13 (13 without 2020 data; 15 incl. 2020 data)
  - Global 14 (both columns show 14)
  - Government 14 (14 without 2020 data; 15 incl. 2020 data)
  - HY 15 (15 without 2020 data; 16 incl. 2020 data)
  - EM funds 17 (17 without 2020 data; 19 incl. 2020 data)
- Inclusion of data up to September 2020 does not significantly change redemption shock calibration for most categories because extreme fund-level flows from 2007–2019 are included in distributions.

- Macrofinancial scenario calibrated on March 4–18, 2020 market moves (used to estimate fund returns given portfolio compositions and durations):
  - Asset class changes over March 4–18, 2020:
    - Equities -23%
    - IG Gov. -40 bps
    - EM Gov 421 bps
    - Corp IG 170 bps
    - Corp HY 359 bps
    - Municipal 90 bps

- Estimated fund returns, redemption shocks and redemptions (example aggregate outcomes):
  - EM: Return shock -17%, Redemption shock -16.3%, Redemptions US$ bn -7.9
  - Global: Return shock -8%, Redemption shock -1.7%, Redemptions US$ bn -1.9
  - Gov: Return shock -2%, Redemption shock -0.3%, Redemptions US$ bn -0.8
  - HY: Return shock -8%, Redemption shock -5.7%, Redemptions US$ bn -12.8
  - IG: Return shock -5%, Redemption shock -3.6%, Redemptions US$ bn -76.0
  - Loan: Return shock -1%, Redemption shock -1.0%, Redemptions US$ bn -0.7
  - Mixed: Return shock -15%, Redemption shock -3.9%, Redemptions US$ bn -55.0
  - Multi: Return shock -5%, Redemption shock -1.5%, Redemptions US$ bn -5.4
  - Muni: Return shock -4%, Redemption shock -2.5%, Redemptions US$ bn -19.3

- Using estimated outflows and a slicing approach to meet redemptions, volumes of asset sales are estimated and used to compute price impact under stress market depth measures.
- Overall, the impact of funds’ sales on markets would be larger on corporate bonds.

*Source: Appendix XIII in IMF (2020b), wpiea2021109-print-pdf - Appendix XIII in IMF (2020b).*

### 3.5 percent decline in prices for HY and 2.8 percent for IG bonds (Figure 13)  than sovereign

### wpiea2021109-print-pdf - 3.5 percent decline in prices for HY and 2.8 percent for IG bonds (Figure 13)  than sovereign

### Price impact of funds’ sales and estimated losses
- Estimated additional price decline due to fund sales: 3.5 percent for HY and 2.8 percent for IG bonds (Figure 13).
- Historical market declines over 4–18 March: market value of HY bonds declined by 15 percent and IG by around 10 percent.
- The additional decline due to fund sales would account for around 30 percent of the shock observed.
- Fund exposures in the sample: USD 1,094bn to IG and USD 360bn to HY.
- Estimated absolute losses from fund sales: around USD 30bn for IG bonds and USD 13bn for HY bonds.
- Comparable evidence: In the October 2020 Global Financial Stability Report (IMF, 2020c), bonds sold by funds experiencing high outflows had returns around 2 percentage point lower than other bonds, which is in the range of these estimates.

### Flows in Distress (FiD) framework and projected net flows
- Analysis focuses on IG bond funds because they account for the larger share in the sample and experience their worst outflows in March.
- Use of 2007–2019 data and assumption that IG bond funds face net flows below the 0.1 percentile to compute expected net flows for other fund categories.
- Observed outflows in March 2020 were relatively close to estimated values for HY and municipal bond funds; for other categories, actual outflows were larger than predicted by the model.
- The copula approach provides the dependence structure of flows across fund categories, enabling calculation of joint distress probabilities.

### Joint probability of fund categories being in distress
- Two samples used (each with 100 observations):
  - September 2011 to December 2019
  - February 2012 to May 2020 (includes March–April 2020 liquidity crisis)
- The joint probability of distress is lower in the first sample than in the second, reflecting acute simultaneous stress in March 2020.
- Table 12 — Joint probability of all fund categories being in distress (net flows below given percentiles):
  - Distress level 1%: 2011–2019 sample = 0.49% ; 2012–2020 sample = 0.85%
  - Distress level 5%: 2011–2019 sample = 0.58% ; 2012–2020 sample = 0.99%
  - Distress level 10%: 2011–2019 sample = 0.74% ; 2012–2020 sample = 1.20%
- The FiD framework can be updated over time to reflect evolving dependence structures within fund flows.

### Systemic implications and evidence from March 2020
- The COVID-19 outbreak was an exceptional external shock; the ensuing liquidity crisis put severe pressure on financial institutions including mutual funds.
- Mutual funds were able to meet redemptions and no fund suspended redemptions, aided by swift and massive intervention by the Federal Reserve.
- The liquidity crisis provided evidence that vulnerabilities within the mutual fund industry can crystallize during stress periods and that funds can transmit shocks through asset sales.

### Policy discussion — mitigating vulnerabilities
- Liquidity resilience and stress testing:
  - Many funds exposed to HY and leveraged loans have relatively low levels of high liquid assets and can be vulnerable to large redemption shocks.
  - In line with 2017 FSB recommendations, options include ensuring robust risk management frameworks, supervisory reporting, and liquidity stress test requirements for some fund categories (similar to Money Market Funds and certain derivatives exposures).
  - Enhanced guidance for frequent and rigorous stress testing and appropriate disclosures is recommended to support liquidity risk management and identify vulnerable funds (IMF, 2019).
  - Stress tests could be performed by asset managers and complemented by SEC-led stress tests (IMF, 2020a).
- Liability–asset liquidity matching:
  - Consider structuring funds so liability-side liquidity matches asset-side liquidity; e.g., increasing redemption frequency to one week or one month for less-liquid strategies.
  - Note: leveraged loans can take 10 to 20 days to settle.
- Liquidity management tools and valuations:
  - Support and encourage availability of liquidity management tools such as gates, deferring redemptions, and swing pricing.
  - Swing pricing is available for U.S. mutual funds, but so far no mutual fund has implemented it.
  - Authorities should monitor developments and seek to provide timely and reliable valuations of assets in funds’ portfolios (IMF, 2020c).
- Mitigating negative externalities from fund sales:
  - Ensure funds have enough liquid buffers to reduce sales of less liquid assets.
  - Require asset managers to consider market impact when choosing liquidation methods (for example, run simultaneous stress tests across fund categories they manage).
- Interconnectedness and contagion:
  - Contagion and spillover analyses show some fund categories are more connected; stress in one category is likely to spread to connected categories.
  - Authorities should further analyze interlinkages to assess whether they reduce or increase risks, complementing market- and flow-based analyses with balance sheet and off-balance-sheet exposure analysis.
  - Macrofinancial stress simulations can estimate impacts across the financial system and identify key vulnerabilities.

### Conclusion and recommended actions
- Bond mutual funds provide credit intermediation but face potential liquidity mismatch by offering daily redemptions while holding assets with varying liquidity.
- Main findings:
  - Most funds can cope with severe redemption shocks, but their asset sales can propagate stress by depressing asset prices and transmitting shocks to the financial system.
  - Interconnectedness results (copula dependence and Diebold–Yilmaz frameworks): IG corporate bond funds, municipal bond funds, and government bond funds are more likely to spread distress to other fund categories; HY, EM, and loan funds are more vulnerable to distress originating elsewhere.
- Policy recommendations:
  - Strengthen risk management frameworks and make liquidity management tools available (gates/deferring redemptions, swing pricing).
  - Support timely and reliable valuations of assets and consider system-wide stress tests to identify vulnerable funds for closer inspection.
  - Reduce liquidity mismatch and consider contemplate liquidity provision for key funds to mitigate potential fire sales.
- Forward uses:
  - The analyses and tools can perform what-if scenarios, support liquidity stress tests at the individual fund level, enable macrofinancial stress tests, and help identify fund strategies that warrant further monitoring.
  - The Flows in Distress framework could be applied to other entities, such as banks (to estimate expected deposit outflows conditional on bank runs) or countries (to estimate expected capital outflows conditional on another country facing massive outflows).

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021109-print-pdf.pdf*

### 2.   Data

### wpiea2021109-print-pdf - 2.   Data

### Data coverage and sample
- Monthly data on flows, net asset value, portfolio composition and returns are retrieved over the 2017–2019 period.
- The sample of fund is based only on funds that were still alive as of end-2019.

### Computation of net flows
- For each fund, net flows in percent of NAV (표표 푡푡) are computed using the following formula:
  표표 푡푡 = 퐹퐹퐻퐻퐹퐹퐹퐹 퐸퐸 푡푡 푁푁퐻퐻푉푉 푡푡−1
- Net flows whose absolute value are above 50 percent were excluded as they are likely related to reporting mistakes.

### Portfolio composition and asset classification
- Latest portfolio composition is retrieved for each fund.
- Top-level portfolio split into four categories: cash, equities, bonds and other.
- Fixed income portfolio (cash and bonds) further split into: government, municipal, corporate, securitized, cash and equivalents, and derivatives (Table A3).
- Each subcategory is subsequently split into further asset classes as detailed in Morningstar (2016).
- Excerpt of fund counts and NAV reporting (as in source):
  - Fund category
    - SampleICISampleICI
    - Corp. IG2,427          2,118               608           593
    - Corp. HY347              334                  250           255
    - Govt325              341                  161           194
    - Multisector433              291                  182           95
    - Global247              541                  183           347
    - Municipal799              793                  567           554
    - Mixed funds1,752          1,532               792           772
    - Total6,319          5,951               2,743       2,810
  - Sources: Morningstar, ICI, IMF staff calculations
  - Net asset value (US$ bn)Number of funds

### Credit quality
- For each fund, the latest data on credit quality are retrieved: the share of the bond portfolio split by credit rating.
- Morningstar does not provide credit rating data by type of fixed income instrument.
- Credit rating by instrument is estimated by allocating the highest credit rating first to the government portfolio, then to corporate bonds and finally to securitized products.

### Treatment of mutual funds using derivatives and leverage
- When funds use derivatives and leverage Morningstar asset allocation weights will always add up to 100 percent, but the cash part will have negative values.
- In those cases:
  - cash is set up equal to 0 percent, and
  - put a 100 percent limit on the other parts of the portfolio.
- When the fund reports cash allocation above 100 percent, the cash is bounded to 100 percent.

### Fixed income classification (as listed)
- Equities
- CashCash and equivalents
- Government
- Municipal
- Corporate
- Securitized
- Derivatives
- Other
- Source: Morningstar

### Robustness checks for the Flows in Distress approach — copula choice
- Main approach uses a Student copula to model dependence between fund net flows because it allows for tail dependence.
- Robustness check uses a Gaussian copula where correlation parameters equal the observed correlation among net flows by fund categories.
- As expected, expected net flows are larger when using the Student copula, due to tail dependence.
- Qualitative results are similar across copulas:
  - IG, mixed and multisector bond funds are the fund categories with the largest (average) effects on other funds.
  - EM, HY and loan funds remain the most vulnerable fund categories.

### Table B1. Expected Net Flows (Student Copula) — selected rows (values in % of NAV)
- Net flows IG row (To (AVG)): -2.2 -1.7 -1.8 -0.5 -2.1 -1.0 -2.3 -4.6 -0.1 -1.8
- Net flows HY row: -0.3 -4.5 -0.3 -0.5 0.0 0.8 -0.7 -2.1 -3.4 -0.8
- Net flows Muni row: -1.9 -1.6 -3.0 -0.1 -2.3 -0.4 -0.8 -2.1 3.7 -0.7
- Net flows Mixed row: -0.2 -1.6 0.2 -1.7 0.8 -0.6 -2.9 -3.2 -5.6 -1.6
- Net flows Gov row: -1.4 -0.5 -1.5 0.3 -4.2 -0.7 -0.1 -3.7 4.8 -0.3
- Net flows Global row: -0.5 0.1 -0.2 -0.6 -0.5 -4.4 -1.2 -2.4 1.1 -0.5
- Net flows Multi row: -0.7 -1.3 -0.1 -1.0 -0.1 -0.6 -3.7 -3.7 -5.3 -1.6
- Net flows EM row: -0.4 -0.8 0.0 -0.4 -0.7 -0.1 -1.1 -5.0 -2.0 -0.7
- Net flows Loan row: 0.6 -1.0 1.2 -0.6 1.2 1.0 -1.5 -1.7 -11.0 -0.1
- From (AVG) row: -0.6 -1.1 -0.3 -0.4 -0.5 -0.2 -1.3 -2.9 -0.9

- Note: Expected net flows conditional on distress in % of NAV. Grey cells show net flows for funds in distress.
- Sources: Morningstar, IMF staff.

### Table B2. Expected Net Flows (Gaussian Copula) — selected rows (values in % of NAV)
- Net flows IG row (To (AVG)): -2.2 -1.5 -1.7 -0.4 -1.6 -0.1 -1.6 -2.9 0.2 -1.2
- Net flows HY row: -0.2 -4.5 -0.1 -0.3 -0.1 0.9 -0.4 -1.1 -2.8 -0.5
- Net flows Muni row: -1.6 -0.9 -2.9 -0.2 -1.7 0.1 -0.3 -1.7 3.5 -0.3
- Net flows Mixed row: -0.1 -0.9 0.2 -1.7 0.8 -1.0 -2.7 -3.1 -4.0 -1.3
- Net flows Gov row: -0.9 -0.4 -1.2 0.3 -4.2 1.3 0.6 -1.8 5.0 0.4
- Net flows Global row: 0.1 0.4 0.0 -0.6 0.3 -4.3 -0.8 -1.3 0.2 -0.2
- Net flows Multi row: -0.6 -1.0 0.0 -0.9 0.0 -0.3 -3.7 -3.1 -4.5 -1.3
- Net flows EM row: -0.2 -0.6 -0.1 -0.4 -0.5 0.2 -0.9 -5.0 -1.2 -0.5
- Net flows Loan row: 0.6 -1.0 1.1 -0.5 1.1 0.8 -1.1 -0.9 -11.0 0.0
- From (AVG) row: -0.4 -0.7 -0.2 -0.4 -0.2 0.2 -0.9 -2.0 -0.4

- Note: Expected net flows conditional on distress in % of NAV. Grey cells show net flows for funds in distress.
- Sources: Morningstar, IMF staff.

### Figures — qualitative summaries
- Contagion to other funds (average expected net flows conditional on one fund category being in distress): Student copula shows larger average impacts than Gaussian; ordering highlights Loan, Gov, Global, EM, Muni, HY, Multi, Mixed, IG by varying magnitudes as shown in figures.
- Contagion from other funds (average expected net flows conditional on other fund categories being in distress): patterns similar across copulas with Student showing stronger tail effects.

### Appendix C — Methodology for the Macrofinancial Scenario
- Scenario construction follows components detailed in Bouveret (2017) and Appendix XII of IMF (2020b).
- Steps:
  - Design a macrofinancial scenario.
  - Compute individual funds’ returns based on each fund portfolio composition and the scenario values of macrofinancial variables.
  - Estimate the flow-return relationship for each fund category (investors respond to past returns: inflows to high-return funds, outflows from low/negative-return funds).
  - Project net flows for each fund using estimated returns and the flow-return relationship.
  - Obtain volume of sales by asset class and fund by assuming managers follow a slicing approach (they sell each asset in their portfolio in proportion to its share).
  - Use market depth measures to estimate the price impact of fund sales.

*Source: wpiea2021109-print-pdf - 2.   Data*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021109-print-pdf.pdf_