## textrevised

## Source details

**Canonical URL:** [textrevised](https://www.imf.org/-/media/files/publications/gfsr/2024/october/english/textrevised.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfsr/2024/october/english/textrevised.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfsr/2024/october/english/textrevised.pdf.json)

---

### Financial system transformation and NBFI growth
- Strengthened supervision, crisis management preparedness, enhanced data collection, and a macroprudential approach have raised resilience to recent shocks.
- Structure changes:
  - Growing participation of nonbanks in financial intermediation (NBFI).
  - Pension funds, insurance companies, and sovereign wealth funds grew from 50 percent of global GDP to close to 90 percent over the past two decades, with much deployed to nonbanks.
  - New digital technologies are transforming distribution, credit evaluation, trading, and market making.
- Persistent recurrent vulnerabilities:
  - Liquidity mismatches in open-ended mutual funds.
  - Highly leveraged trading strategies used by hedge funds.
  - Opaque interconnectedness within the broad NBFI sector.
- Key statistics and dates:
  - Report reflects information available as of September 30, 2024.
  - Editorial changes noted on 11/21/24: three post-release corrections.

### Market conditions, volatility episodes, and Growth‑at‑Risk (GaR)
- Near-term risks and aggregate assessment:
  - IMF’s one-year-ahead GaR measure around the "40th historical percentile".
  - Over the next year, there is a 5 percent probability that global real growth will fall below 1.2 percent.
  - Baseline growth forecast: 3.2 percent (World Economic Outlook).
- Early August 2024 turmoil:
  - Nikkei index declined by "12 percent" on "August 5".
  - VIX intraday: 16 → more than 65 → 37 on August 5, 2024.
  - S&P 500: –3 percent (August 5, 2024).
  - STOXX Europe 600: –2 percent (August 5, 2024).
  - VIX front futures expire on August 21 (2024).
- Financial conditions and tail-risk scenario:
  - If financial conditions tighten by 2.5 standard deviations and persist one quarter, year-ahead GaR could worsen to its lowest historical quintile.
  - Medium-term (four years ahead) GaR "remains at its worst quintile currently."
- Yield-curve and QT facts:
  - Group of Ten central banks reduced balance sheets from a peak of $28 trillion in March 2022 to $21.5 trillion.
  - US Federal Reserve policy-rate expectation: cut by almost 150 basis points by end-2025 (markets priced relative to April 2024 GFSR).

### Nonbank financial intermediation (NBFI), leverage, and data gaps
- NBFI contribution to stress:
  - Rapid unwinding of leveraged positions can generate liquidity imbalances and amplify stress.
  - NBFI growth includes open-ended bond funds, hedge funds, private credit with increased leverage.
- Data and transparency gaps:
  - Inadequate data hinder assessment of nonbank leverage and concentrated positions.
  - Recommendation: enhance reporting requirements for NBFIs and strengthen policies to mitigate vulnerabilities and amplification mechanisms from nonbank leverage.
- Fund and leverage metrics:
  - Hedge funds industry size: $7 trillion.
  - Leveraged bond funds and repo usage: leveraged funds small share but deleveraging by even a small set could have outsized effects.
  - Nonbank financial intermediaries’ assets under management in median emerging market: 25 percent of GDP (current) vs less than 10 percent in 2003.

### Artificial intelligence (AI) — adoption, risks, and market‑structure implications
- Adoption and capabilities:
  - ML integrated in finance ~20 years; GenAI and foundation models in infancy but rapid gearing-up.
  - AI-driven ETFs: less than $1 billion AUM; AI-driven ETF AUM chart values cited (2018–24): 0.003, 0.001, 0.001, 0.002, 0.002 (billions of US dollars).
  - Trailing six‑month model counts and open-source shares tracked; training compute measured in floating-point operations (exa = 10^18).
- Potential benefits:
  - Efficiency, productivity, refined portfolio frameworks, improved return forecasting, SupTech/RegTech uses.
- Principal risks (four categories preserved):
  - Increased market speed and volatility under stress.
  - Opacity and monitoring challenges.
  - Operational concentration on a few AI service providers.
  - Increased cyber and market-manipulation risks.
- Market‑structure concerns:
  - AI could increase correlation, turnover, and speed of price incorporation; could concentrate advantages via high fixed costs of training notable models.
  - Simulation and outreach findings: potential for tacit algorithmic collusion, herding, and higher procyclicality.
- Empirical signals and simulations:
  - ML‑GaR predictive gains: ML‑GaR improves out-of-sample accuracy by up to 7 percent; adding REU further improves performance by 5 to 13 percent.
  - ML‑GaR variable importance: REU contributes at least as much as the financial conditions index for downside GDP-risk prediction.

### Commercial real estate (CRE), corporate sector, and private credit vulnerabilities
- CRE corrections and exposures:
  - Global CRE prices: –12 percent year over year.
  - US office sector: –23 percent year over year.
  - European office sector: –16 percent year over year.
  - US maturing CRE debt between 2024 and 2025: nearly $1 trillion with a funding gap of almost $300 billion.
  - CMBS office delinquencies above 8 percent, up 3 percentage points year over year.
- Bank stress-test adverse scenario (50 percent office value loss):
  - Aggregate Tier 1 capital ratio: US banks 12.3 → 11.3 percent.
  - Aggregate Tier 1 capital ratio: European banks 17 → 13.3 percent.
  - 4 percent of banks in sample (US and European) representing 1 percent of assets would have Tier 1 dip below 7 percent.
- Corporate debt and refinancing:
  - Fixed-rate debt accounts for close to 50 percent of global corporate debt coming due in 2025, coupons between 3.5 and 4 percent versus current refinancing yield of 5.5 percent.
  - Refinancing 2024 and 2025 bonds at higher rates would bring ICRs down by an average of 12 percent if monetary policy does not ease.
  - Average remaining life of high-yield debt reached 4.6 years in Q4 2023.
- Private credit:
  - Growth beyond midsized borrowers; signs of stress: declining ICRs, more PIK coupons, frequent restructurings under broader default measures, opaque valuations.
  - Risk pathways: stale valuations → deferred losses → spike in defaults → frozen fundraising → runs on semiliquid funds → spillovers.

### Treasury and bond market structure, dealer inventories, and market liquidity risks
- Treasury market dynamics:
  - Primary dealers increasingly warehousing longer-dated securities; household/hedge fund Treasury holdings rise slowed since April 2024.
  - Bloated dealer inventories risk: may constrain absorption of sales and worsen sell-offs.
- Bills share annotation preserved: "Steady and predictable15–20% recommended bills".
- QT and US Treasury free float:
  - QT increased free float of Treasuries; Treasury issuance shifting toward shorter-term debt.
  - Net supply of Treasury bonds relative to GDP projected to remain elevated, potentially pushing up term premiums.

### Emerging markets, fiscal buffers, and capital‑flows‑at‑risk
- Emerging market resilience and vulnerabilities:
  - EMs remain resilient since April 2024; near-term pressure may moderate with advanced-economy easing.
  - Slowing growth in China and geopolitics increase risks; portfolio flows may become more volatile.
- Capital‑flows‑at‑risk:
  - 5 percent probability that EM outflows could reach 2.4 percent of GDP over the next three quarters (marginal increase since April 2024 GFSR).
- Fiscal buffers and sovereign spreads:
  - Fiscal-buffer classification: "Small or worsening" beyond deficit of 2 percent; "Borderline" between –2 to 2 percent; "Large or improving" exceeding 2 percent.
  - Of 80 sampled sovereigns, 17 (21 percent) have average ratings at CCC+ or worse vs 4 (5 percent) in December 2019.
- Frontier market specifics:
  - Roughly $4 billion maturing remainder of 2024; roughly $13 billion in 2025; roughly $14 billion in 2026.
  - Roughly 60 percent of maturing bonds issued by countries with prevailing yields close to or above 10 percent.
- Sustainable debt and climate finance:
  - Global sustainable debt issuance rebounded in H1 2024; EMs account for just 13 percent of year-to-date issuance.
  - International adaptation finance to developing countries is 10 to 18 times below estimated needs.

### China: property sector, AMCs, and systemic considerations
- Inflation and housing metrics:
  - One-year-ahead expected CPI inflation nearly halved to 1.3 percent; probability one-year-ahead inflation falls below 0.3 percent increased.
  - Primary home prices down 7 percent from peaks; secondary home prices down 13 percent from peaks; primary market sales 40 percent lower than prepandemic peak.
- Central bank action:
  - On August 31, secondary market operations resulted in net liquidity injection of 100 billion yuan.
- AMCs and NPLs:
  - Banks’ reported NPL ratios less than 1 percent for mortgages; direct exposures to developers less than 6 percent of total bank loans.
  - Since 2012, reported cumulative NPLs amounted to less than 3 trillion yuan in 2023.
  - Write-offs and disposals topped 3 trillion yuan each year since 2020; cumulative disposals totaled 22 trillion.
  - Four national AMCs hold 80 percent market share in primary NPA market; two national AMCs have equity-to-asset ratios below 5 percent.

### Macroprudential framework, crisis preparedness, and supervisory priorities
- Macroprudential recommendations:
  - Tighten appropriate macroprudential tools to contain excessive risk taking in nonbank sector while avoiding destabilizing broad tightening.
  - Strengthen financial regulation and supervision to address tail of weak banks; promote full, timely, and consistent implementation of Basel III.
  - Require banks to test access to central bank instruments periodically and prepare frameworks for emergency liquidity assistance with solvency/viability, collateralization, and haircut principles.
- Crisis preparedness and resolution:
  - Progress on recovery and resolution frameworks for weak/failing banks is critical to avoid undermining financial stability or risking public funds.
  - Supervisors should collect detailed CRE exposure information, conduct stress tests (including smaller banks), review valuation assumptions, and ensure provisions are adequate.
- NBFI-specific supervisory steps:
  - Implement Financial Stability Board recommendations on liquidity mismatches in open-ended funds and enhance stress testing for nonbanks.
  - Require mapping and disclosure of AI‑relevant information for large traders and NBFIs; enhance cross-border supervisory cooperation and data sharing.

### Policy recommendations — enumerated and specific
- Central banks:
  - Communicate clearly; avoid overreacting to single data points.
  - Where growth and inflation momentum slow, gradually ease toward neutral stance; where inflation remains above target, remain restrictive.
  - Monitor liquidity conditions, uneven reserve distribution, and remain ready to address market stresses; be transparent about liquidity removal steps.
- Fiscal authorities:
  - Sustain gradual, well‑communicated fiscal adjustments to rebuild buffers and stabilize debt; protect the vulnerable and public investment.
  - Frontier and low‑income sovereigns: strengthen creditor communications, multilateral cooperation, and use IMF Integrated Policy Framework where appropriate.
- NBFI and private credit:
  - Enhance reporting and data collection for NBFIs and private credit.
  - Improve liquidity preparedness, adopt FSB standards, and implement more intrusive supervisory approaches for private credit given exponential growth and retail participation.
- AI and market resilience:
  - Calibrate circuit breakers and review margining practices for potential rapid AI-driven moves.
  - Map interdependencies of data, models, and tech infrastructure; identify critical AI third‑party service providers and strengthen cyber resilience.
  - Require human-in-the-loop for significant decisions; adopt balanced regulation that preserves innovation.
- Cross-border and emerging market measures:
  - Build international reserve buffers, maintain exchange rate flexibility, and ensure bank/nonbank assessment of cross-border spillovers.
  - Scale up adaptation finance: align public/private interests, improve tracking, provide taxonomies, and integrate adaptation across asset classes.

*Source: Preface; Foreword; Executive Summary; Chapters 1–3, Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability (information in this unit reflects data and analysis as of September 30, 2024).*

### Preface                                                                                                                 

### Preface

### Key messages on financial system transformation and resilience
- Strengthened supervision and regulation, better crisis management preparedness and resolution processes, enhanced data collection, and a macroprudential approach to financial sector oversight have raised the financial sector’s resilience to recent shocks.
- The structure of the global financial system is undergoing substantial transformations, notably with growing participation of nonbanks in financial intermediation (NBFI).
- Drivers of NBFI growth highlighted:
  - Constraints imposed by postcrisis regulations on banks’ leverage have encouraged diversification and the transfer of risk to other financial intermediaries.
  - Pension funds, insurance companies, and sovereign wealth funds have grown from 50 percent of global GDP to close to 90 percent over the past two decades, with much of their assets deployed to nonbanks.
  - New digital technologies are revolutionizing distribution of financial services, credit evaluation, and trading and market making.

### Benefits and uneven distribution of market-based finance and NBFI growth
- Positive implications:
  - Alternative sources of financing for firms through market-based finance, private equity, private credit, hedge funds, and high-frequency market making and trading.
  - Potential for better capital allocation and greater market efficiency.
  - Broader set of intermediaries with different risk profiles, time horizons, and expertise can reduce overreliance on banks, increase competition, provide diversification, and enable risk transfer away from the banking system.
- Uneven benefits:
  - Many advanced, emerging market, and developing economies remain bank-centric and stand to benefit from further development of NBFI and market-based finance.

### Risks and policy imperatives
- Recurrent vulnerabilities identified in past GFSRs that remain relevant:
  - Liquidity mismatches in open-ended mutual funds.
  - Highly leveraged trading strategies used by hedge funds.
  - Opaque interconnectedness within the broad NBFI sector.
- Market liquidity in core markets (government and corporate bonds) can be impaired as nonbanks become central to intermediation—episodes of stress have required central bank intervention.
- International standard setters are making progress to enhance NBFI resilience; urgency and consistent national implementation are paramount.

### Uncertainty, geopolitical risk, and amplification channels
- Enhanced relevance of resilience-building given elevated economic and policy uncertainty and rising geopolitical risks.
- Reference to Chapter 2 findings: economic uncertainty increases downside risks to future growth, asset prices, and bank lending, and can trigger cross-border spillover effects through trade and financial linkages.

### “Future of finance” considerations
- Policymakers need to think through technological innovation’s impact on the future of finance:
  - Innovation can increase efficiency and competition while disrupting traditional bank-provided services.
  - Novel lending modalities for private credit are expected to continue growing.

### Editorial and publication details (selected)
- This GFSR reflects information available as of September 30, 2024.
- The report benefited from comments and suggestions from staff in other IMF departments and from Executive Directors following their discussions of the GFSR on October 8, 2024; however, the analysis and policy considerations are those of the contributing staff and should not be attributed to the IMF, its Executive Directors, or their national authorities.
- Editor’s Note (11/21/24): three changes made after initial release:
  1. In the Preface, a contributor’s name was transposed.
  2. The sources for Figure 3.8 were corrected.
  3. The storyboard for Figure 3.11, panel 3, was corrected.

*Source: Preface, Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability (information in this unit reflects data and analysis as of September 30, 2024).*

### FOREWORD

### FOREWORD

### Nonbank Financial Intermediation (NBFI) and systemic risks
- Artificial intelligence may support further growth in NBFI (see Chapter 3 of the GFSR), and digital banks are growing in systemic importance.
- Traditionally, prudential regulation of nonbanks tends to be either absent or less strict because they do not take deposits from retail investors and largely do not have recourse to central bank backstops.
- With the growth in the relative size of NBFI and its close linkage with the banking sector, more substantive externalities may be generated, potentially requiring novel policy approaches.
- Market stress episodes illustrate nonlinear amplification channels: when volatility spiked in August 2024, many leveraged investors reached risk limits and received increased margin calls, forcing rapid position closures that likely exacerbated the sell-off.
- Examples of nonbank failures and spillovers noted: liquidity mismatches in bond funds; significant increases in margin calls in derivatives markets; failure of highly leveraged nonbanks such as Archegos generating substantial losses for banks.

### Policy recommendations to manage NBFI and systemic risk
- Expand data collection:
  - Collect more comprehensive data on NBFIs to better evaluate risks to global financial stability and map interlinks of the sector.
  - Obtain information on the use of leverage and asset holdings to develop more effective policies while avoiding stifling financial innovation.
- Increase transparency:
  - Address opacity of nonbanks; require more information to investors and the public given growing spillover potential and rising retail investor participation.
  - Strengthen conduct requirements, including public disclosure, to support market discipline and price discovery.
- Design appropriate liquidity facilities and backstops:
  - Recognize trade-offs central banks face between short-term financial stability support and moral hazard.
  - Develop central bank support mechanisms that minimize moral hazard and encourage nonbanks to internalize liquidity risks.
  - Prepare communication plans to avoid perceptions of policy incoherence (for example, purchasing assets to restore financial stability while tightening monetary policy to fight inflation).
- Improve the financial “plumbing”:
  - Ensure payments and settlements systems work effectively and securely, including interoperability across systems and cross-border integration.
  - Integrate new technologies, including artificial intelligence, to enhance efficiency and security.
- Enhance resilience of central counterparties (CCPs):
  - Ensure CCPs have enough resources to cover potential losses, business continuity plans, and recovery and resolution plans to restore stability or wind down operations.
  - Review margining requirements to protect CCPs while considering the broader-system impact of margin and collateral calls during stress.
- Undertake a systemic approach to NBFI resilience:
  - Move beyond institution- and sector-specific prudential frameworks toward system-wide and cross-sectoral perspectives.
  - Improve coordination among relevant authorities to ensure governance structures, monitoring mechanisms, and tools to address systemic risks are in place.
  - Sharpen existing tools and potentially develop new ones to address potential systemic risk.

### Financial market conditions, volatility, and uncertainty
- Since the April 2024 Global Financial Stability Report:
  - Global economic activity has moderated, and inflation has continued to slow.
  - Monetary easing is under way among major central banks; financial conditions have remained accommodative; emerging markets have remained resilient; asset price volatility has stayed relatively low, on net.
- Near-term financial stability risks:
  - IMF’s one-year-ahead growth-at-risk measure remains contained at around the 40th historical percentile.
  - Accommodative conditions can facilitate buildup of vulnerabilities: lofty asset valuations, global rise in private and government debt, and increased use of leverage by nonbank financial institutions.
- Interaction of elevated uncertainty and vulnerabilities raises probability of adverse shocks and sudden volatility surges:
  - The market turmoil in early August 2024 showed stock market volatility spikes in both Japan and the United States and significant declines in global asset prices.
  - Chapter 2 quantifies that if global real economic uncertainty jumps by an amount equivalent to its rise during the global financial crisis, the downside outcome (the 10th percentile) of one-year-ahead global real GDP growth worsens by 1.2 percentage points (Figure ES.4).
  - The adverse effect is stronger when macrofinancial vulnerabilities are elevated or when market volatility is more disconnected from uncertainty.
  - Uncertainty can trigger cross-border spillovers through trade and financial linkages.

### Vulnerabilities and imbalances highlighted
- Sovereign debt:
  - High levels and rapid growth of sovereign debt remain a global challenge, with many jurisdictions failing to achieve longer-term debt-stabilizing primary balances.
  - In many advanced economies, increasingly large shares of government debt issuances will need to be absorbed by price-sensitive buyers amid ongoing quantitative tightening by their central banks, potentially increasing bond market volatility.
  - Emerging markets and frontier economies with weak and worsening fiscal buffers have seen their sovereign bond and credit default swap spreads increase more than those of other jurisdictions.
- Emerging markets:
  - Emerging markets have continued to demonstrate resilience since April 2024; central banks have focused on domestic conditions and used exchange rate adjustments to mitigate external headwinds.
  - With major advanced economies set to ease monetary policy, near-term pressure on emerging markets could moderate.
  - Elevated uncertainty regarding trade policies and geopolitics and a slowing growth outlook in China could make preserving financial stability in emerging markets more challenging.
  - Portfolio flows may become more volatile and access to international funding may be more difficult, especially for frontier economies.
  - Interest rate spillovers from advanced economies to emerging markets have increased over the past decade; changes in the 10-year US term premium have explained an increasing share of changes in the term premiums of 10-year emerging market bonds (Figure ES.5).
- Sustainable debt and climate finance:
  - Global issuance of sustainable debt has rebounded in 2024.
  - Emerging markets account for just 13 percent of year-to-date issuance (Figure ES.6), and the share of emerging-market sustainable debt denominated in local currencies is small.
  - Underinvestment in climate finance could delay mitigation and adaptation efforts and could challenge financial stability in the future.
- Corporate sector and private credit:
  - Even as global interest rates decline, many firms will find debt servicing challenging in coming years.
  - Defaults have steadily risen as weaker firms have struggled.
  - Some midsized companies borrowing at high interest rates in private credit markets are resorting to payment-in-kind methods, deferring interest payments and accumulating more debt.
  - Trade restrictions and geopolitical events are likely to affect corporations through higher costs and disruptions.

*Source: FOREWORD, textrevised - FOREWORD (Global Financial Stability Report), October 2024.*

### ExECuTIvE SuMMARY

### ExECuTIvE SuMMARY

### Key findings on vulnerabilities and market dynamics
- Corporate bonds have continued to trade within tight spreads by historical standards, producing pricing misalignments that indicate an increased risk of an abrupt repricing of credit risk.
- Although stability risks from residential real estate appear contained in most countries, pressures on the commercial real estate (CRE) sector remain acute, with misalignment in prices and fundamentals pointing to further corrections—especially in the office sector.
- The global banking sector has remained resilient, with ample capital and liquidity buffers; however, net interest margin and bank profitability could be negatively impacted by interest rate cuts, and a relatively large tail of weaker institutions faces business-model challenges.
- Near-term financial stability risks have remained contained: the IMF’s Growth-at-Risk (GaR) model places near-term risks at around the "40th historical percentile".
- The severe, albeit short-lived, market turmoil in early August illustrated how quickly volatility can spike: the Nikkei index declined by "12 percent" on "August 5".

### Nonbank financial intermediaries (NBFIs), leverage, and AI
- The rapid unwinding of leveraged positions in NBFIs can generate liquidity imbalances that increase volatility and amplify stress.
- The growth of open-ended bond funds, hedge funds, and private credit has been accompanied by increased use of leverage among several NBFI segments.
- Data gaps hinder authorities’ ability to assess vulnerabilities associated with nonbank leverage and to identify large and concentrated positions.
- Adoption of AI and machine learning in trading and investment is rising: the share of applications related to AI and machine learning in patent filings in asset management has risen impressively in recent years (see Figure ES.9).
- Widespread adoption of AI in capital markets could worsen financial fragilities by:
  - increasing volatility during market stress,
  - creating more opacity and challenges in monitoring AI use in capital markets,
  - increasing reliance on a few key AI service providers and associated operational risks,
  - growing risks of cyber and market manipulation.

### Commercial real estate (CRE) and banks’ exposures
- Misalignment in CRE prices and fundamentals points to further corrections, particularly in the office sector; funding withdrawals could push down prices and create adverse feedback loops that put more financial institutions under pressure.
- Both banks with outsized concentrations in CRE and nonbank investors such as real estate investment trusts may experience strains.
- Nonperforming loan ratios have increased for some forms of lending (consumer credit cards, automobile loans, and CRE), though overall bank asset quality has not deteriorated significantly.

### Market events and amplification channels
- The early August market turmoil showed how nonbank intermediation can transmit and amplify strains through rapid unwinding of leveraged positions and resultant liquidity imbalances.
- Algorithmic traders and hedge funds that unwind leveraged positions can exacerbate price declines, and recent AI and machine-learning advancements suggest algorithms may play a larger role in future episodes of turbulence.

### Executive Directors’ assessments and policy priorities
- Executive Directors broadly agreed with staff’s assessment of the global economic outlook, risks, and policy priorities, welcoming continued growth resilience amid recurring shocks.
- Directors highlighted that monetary policy has managed to bring disinflation with so-far limited cost to output and employment, raising the likelihood of a smooth landing—but noted the recovery remains uneven and underwhelming owing to weak productivity growth.
- Directors warned of risks from potentially more persistent underlying inflation, increased geopolitical conflicts, intensifying protectionist policies, and the widening disconnect between subdued financial market volatility and elevated economic and geopolitical uncertainty.
- Directors emphasized:
  - careful calibration of monetary policy, remaining data dependent and clearly communicating policy decisions;
  - keeping policy rates in restrictive territory where core inflation persists above target;
  - moving to a more neutral stance where inflation is unambiguously abating, long-term inflation expectations remain anchored, and output gaps are closing;
  - standing ready to mitigate disruptive impacts of foreign exchange volatility and capital flows, including leveraging the IMF’s Integrated Policy Framework where appropriate.
- Directors called for sustained, gradual, and well-communicated fiscal adjustments to stabilize debt and build buffers, calibrated to country-specific conditions and protecting the most vulnerable and public investment.
- Directors urged advancing structural reforms to boost productivity, competition, human capital, and labor force participation, and to accelerate the green transition while enhancing social acceptability of reforms.
- Directors stressed stronger multilateral cooperation to facilitate debt restructuring, mitigate geoeconomic fragmentation risks, and accelerate the green transition consistent with World Trade Organization rules.

### Chapter 1 at a glance — core messages
- Since April 2024, near-term financial stability risks have remained contained; global activity has moderated, inflation has slowed, emerging markets have been resilient, financial conditions have been accommodative, and market volatility has remained low overall.
- Accommodative financial conditions facilitate further buildup of vulnerabilities: lofty asset valuations in equity and corporate credit markets; rising global debt levels; increased leverage in NBFIs; and fragilities in corporate and CRE sectors.
- These imbalances could amplify future downside risks, made more probable by elevated economic and geopolitical uncertainty and the disconnect between uncertainty and low volatility.
- The early August market turmoil demonstrated how quickly volatility can surge, force the unwinding of leveraged trades, and trigger feedback loops between asset prices and deleveraging.
- Certain types of NBFIs amplified the early August turmoil and warrant more active supervisory engagement.
- The banking system remains sound overall, though a weak tail of banks faces exposures to troubled sectors like CRE and business-model challenges.
- Emerging markets have generally shown resilience, but risks include slowing growth in China, fragilities in China’s financial system, constrained funding access for frontier markets and countries with weaker fiscal buffers, and underinvestment in climate finance delaying mitigation and adaptation.

### Policy recommendations (enumerated)
- For central banks:
  - Communicate clearly that the path of monetary policy should not react excessively to any individual data point to reduce uncertainty.
  - Where growth and inflation momentum are set to continue, gradually ease monetary policy toward a more neutral stance.
  - Where inflation remains stubbornly above targets, push back against overly optimistic investor expectations for monetary policy easing that would further stretch asset prices.
- For fiscal authorities:
  - Focus fiscal adjustments primarily on credibly rebuilding buffers to keep financing costs reasonable and anchor medium-term inflation expectations, given sovereign debt in many countries is substantially above prepandemic levels.
  - For sovereign borrowers in frontier economies and low-income countries, strengthen efforts to contain debt-vulnerability risks through creditor communications, multilateral cooperation, and international community support.
- To address NBFI and nonbank leverage risks:
  - Enhance reporting requirements for NBFIs and strengthen policies that mitigate vulnerabilities and amplification mechanisms stemming from nonbank leverage.
  - Improve NBFIs’ liquidity preparedness, implement the Financial Stability Board’s agreed-upon standards, and enhance stress testing for nonbanks.
- Macroprudential and banking resilience:
  - Strengthen the macroprudential policy framework to contain excessive risk taking in the nonbank sector and ensure adequate capital and liquidity buffers in banking systems to support credit provision through stress periods.
  - Tighten macroprudential tools to increase resilience against a range of shocks while avoiding broad tightening of financial conditions.
- CRE and corporate sector monitoring:
  - Conduct stress-testing exercises that incorporate scenarios involving trade restrictions, geopolitical events, and significant declines in CRE prices.
  - Enhance reporting requirements for private credit to improve monitoring and management of risks.
- Crisis preparedness and resolution:
  - Ensure supervisors are equipped to intervene early and banks are prepared to access central bank liquidity.
  - Progress on adopting and implementing recovery and resolution frameworks is critical to address weak or failing banks without undermining financial stability or risking public funds.
  - Full, timely, and consistent implementation of international standards (including Basel III) remains important to enhance prudential frameworks.

*Source: Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability (Executive Summary).*

### CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARkETS NAvIGATE uNCERTAINTY

### CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARkETS NAvIGATE uNCERTAINTY

### Executive overview
- With postpandemic supply chain disruptions and commodity price pressures having largely dissipated and labor markets coming into better balance, inflation has continued to move toward central banks’ targets, and most have begun to ease monetary policy.
- Markets are pricing in multiple cuts in policy rates among major central banks over the remainder of this year and during the next.
- The slowing growth outlook in China and fragilities in its financial system are key downside risks to the global economy.
- Financial flows may become more volatile and access to international funding may be more difficult, especially for frontier economies.

### Monetary policy outlook and expectations
- The Federal Reserve is expected to cut its policy rate by almost 150 basis points by the end of 2025, more than was expected at the time of the April 2024 Global Financial Stability Report.
- Since the April 2024 GFSR, the European Central Bank, Bank of England, the Federal Reserve, and Riksbank have cut policy rates; the Bank of Japan raised its policy rate in July.
- In emerging markets, policy paths have generally been revised downward, with some central banks pausing cutting cycles or raising rates to ensure convergence of inflation to target.

### Financial market volatility versus economic and geopolitical uncertainty
- Financial market volatility has compressed despite elevated economic policy uncertainty and geopolitical risks; the wedge between volatility and uncertainty is currently quite large.
- Inflation uncertainty remains elevated, with analysts forecasting upside risks—especially a 2 percent or higher core inflation in the year ahead—for both the euro area and the United States.
- Option-implied monetary policy scenarios indicate substantial likelihoods of both shallow cuts and deep adjustments, especially in the United States.

### Yield curve developments and term premiums
- Long-term interest rates in most advanced economies and many emerging markets have changed little, on net, since the April 2024 GFSR; in some major emerging markets, long-term rates have seen upward pressure from rising term premiums.
- The US yield curve has disinverted after a historically long period of inversion; steepening reflects both falling expected short-term rates and rising term premiums.
- The inflation risk premium component of the term premium has displayed continued persistence.
- Historical associations:
  - Bear steepening episodes have tended to favor risk assets more than bull steepening episodes, because bear steepening is typically associated with expected strong growth momentum.
  - Recent US steepening is unique in featuring a decline in the expected policy rate path (bull steepening characteristic) coupled with a rise in term premium (bear steepening characteristic) of broadly comparable magnitude.

### Quantitative tightening (QT) and bond market implications
- Group of Ten central banks have reduced their balance sheets from a peak of $28 trillion in March 2022 to $21.5 trillion.
- Ongoing QT has proceeded in an orderly fashion so far, reflecting calibrated pace and scope of balance sheet reduction aimed at maintaining smooth functioning of government bond and short-term funding markets.
- Key tail risks from QT:
  - QT may drain bank reserves too much, risking funding squeezes similar to US repo market turmoil in September 2019.
  - Simultaneous QT across many central banks raises the odds of spillovers through interconnected funding markets.
  - QT could shift the buyer base of government bonds toward more price-sensitive investors, introducing more volatility—especially in noncore bond markets where reduced central bank holdings are offset by households and the hedge fund sector.
- In the United States:
  - QT has increased the share of free float Treasury securities (portion of outstanding securities net of the Federal Reserve’s holdings), which could exert upward pressure on Treasury yields and volatility over time.
  - The Department of the Treasury has increasingly issued more shorter-term debt to meet funding needs, which might lower borrowing costs in the near term but could expose the Treasury to higher future financing cost.
  - Larger issuances have led hedge funds and dealers to keep more securities on balance sheet.

### Market structure and investor base shifts
- In the euro area, ECB’s reduced holdings of core issuer bonds like Germany have been offset by more holdings by domestic banks and foreign investors; for noncore issuers, reduced ECB holdings are offset by “other domestic investors” (including households and hedge funds), increasing price sensitivity in those markets.
- Net supply of Treasury bonds relative to GDP is projected to remain elevated, which may push up term premiums.

*Italicized source: IMF, Global Financial Stability Report: STEAdYING ThE COURSE: UNCERTAINTY, ARTIFICIAL INTELLIGENCE, AND FINANCIAL STABILITY (October 2024).*

### 1. Bills Share and Treasury Free Float Since the Mid-1970s

### 1. Bills Share and Treasury Free Float Since the Mid-1970s

### Treasury market dynamics
- Chart annotation: "Steady and predictable15–20% recommended bills" (preserves source phrasing).
- Primary dealers are increasingly warehousing longer-dated securities.
- Panel note: In panel 2, “Hedge funds” reflects the percentage share of Treasury securities held by households and nonpro¡t organizations—by and large composed of hedge funds—relative to outstanding marketable Treasury securities. “Primary dealer” reflects the percentage share of primary dealer positions in Treasury coupon securities relative to the corresponding outstanding issued amount.

### Dealer inventory and hedge fund holdings
- Time series coverage in panel 2: Aug. 2015 through Aug. 2024.
- Sources cited for data: Bloomberg Finance L.P.; EUROPACE AG/Haver Analytics; Federal Reserve Bank of New York; Kuttner (2006); and IMF staff calculations.
- Note from text: Since the April 2024 Global Financial Stability Report, the rise in household Treasury holdings (primarily driven by hedge funds) has slowed, consistent with the increased warehousing by primary dealers shown in the latest Federal Reserve Board flow of funds statistics.

### Risks from bloated dealer inventories
- Bloated dealer inventory presents a medium-term risk:
  - In adverse market conditions where investors are selling Treasury securities (for example, if hedge funds were to unwind the Treasury basis trades as described in the April 2024 Global Financial Stability Report) primary dealers with larger Treasury inventories are more likely to face internal balance sheet constraints that could prevent them from absorbing the sales, worsening the sell-off.
- Intermediaries (brokers or primary dealers) may buy securities to facilitate trading and liquidity, whereas end users (pension funds, insurance companies, mutual funds, corporations, individual investors) are typically the ultimate holders.

### Equity markets: rally, volatility, and valuation vulnerabilities
- Rally context:
  - The rally in global equity markets continued since the April 2024 GFSR, briefly interrupted by a severe but transitory sell-off in early August (gray vertical lines mark the sell-off peak: August 5).
- Recent country performance highlights: Since April, Canada, China, and the United States have experienced the largest equity gains; Japan’s performance driven by information technology stocks.
- Volatility spike: Implied volatility for equities spiked during the early August sell-off (Figure 1.7, panel 2).
- Corporate spreads: Investment-grade and high-yield corporate bond spreads widened in Europe and the United States after a long period of decompression (Figure 1.7, panels 3 and 4).
- Nonbank financial intermediation: NBFIs like momentum-following and commodity trading advisor hedge funds and algorithmic and quantitative traders reportedly contributed to the sell-off by cutting positions to stop losses.

### Concentration, correlation, and required earnings growth
- Concentration and correlation:
  - Since January, the share of the Magnificent 7 (M7) increased from 20 to 30 percent of the overall S&P 500 index (market capitalization).
  - Rolling six-month correlation estimates indicate the correlation between M7 and the S&P has increased from around 40 percent to just above 65 percent since May.
  - Correlation of average pairwise M7 has increased from 10 to 50 percent over the same period.
  - Since 2023, there have been 69 days on which fewer than 150 stocks have moved in the same direction as the index.
- Valuation pressure and required earnings growth:
  - The S&P 500 is trading above its historical upper quartile in terms of forward price-to-earnings ratio since 1990.
  - For the ratio to return to its historical 10-year average by 2026, earnings per share on of the S&P and Nasdaq would need to post compounded annual growth rates of close to 25 and 30 percent, respectively—far higher than current market expectations.
  - MSCI World and MSCI Advanced Economy indices require higher growth rates than current expectations to return to historical valuations; emerging market indices and the Russell 2000 require less than current expectations.

### Emerging markets and fixed income
- Emerging market sovereign spreads:
  - US dollar emerging market sovereign spreads remain tight relative to spreads on US investment-grade corporations (Figure 1.9, panel 1).
- Local currency government yields:
  - Spreads of local 10-year government bonds as a share of 10-year Treasury yields and 10-year Treasury yields are tracked (Figure 1.9, panel 2).
- Equity performance:
  - Emerging market equities have performed positively year to date in many cases, though valuations remain lower than historical averages (Figure 1.9, panel 3).
- Risks ahead: Uncertainty from monetary policy in advanced economies—especially the United States—and policies of newly elected governments could challenge emerging market assets.

### Crypto assets and correlations
- Market capitalization: Total market capitalization of crypto assets at $2.2 trillion remains below its historical peak in November 2021.
- Recent dynamics:
  - The crypto rally earlier this year has started to fade after approvals of spot Bitcoin and Ethereum exchange-traded products in January and May 2024, respectively.
  - Crypto valuations have been driven recently by high rolling correlation between Bitcoin and other asset classes, such as equities (S&P 500) and gold, rather than idiosyncratic developments within crypto.
- Risks identified: Widespread adoption of crypto assets could undermine the effectiveness of monetary policy, circumvent capital flow measures, exacerbate fiscal risks, divert financing from the real economy, and threaten global financial stability. Growing interlinkages between crypto and broader financial markets may increase contagion risks.

### Macrofinancial stability: Growth-at-Risk (GaR) and financial conditions
- Financial conditions:
  - Financial conditions have marginally tightened in many regions following the early August market turmoil, but still-elevated equity and corporate bond valuations have kept financial conditions in advanced economies relatively easy by historical standards.
  - In China, price indicators show looser financial conditions due to monetary easing and narrower corporate credit spreads, while quantity indicators such as credit growth continue to weaken.
- GaR assessment:
  - Over the next year, there is a 5 percent probability that global real growth will fall below 1.2 percent.
  - This is appreciably lower than the baseline forecast for growth of 3.2 percent in the World Economic Outlook.
  - GaR is around the 40th historical percentile, indicating near-term risk is contained owing to still accommodative financial conditions and moderate credit growth.
  - The forecast distribution of growth is skewed slightly more to the left than in the April 2024 Global Financial Stability Report, in line with the World Economic Outlook’s assessment that the balance of risk to the global outlook is tilted to the downside.

*International Monetary Fund | October 2024 — Chapter 1, "Steadying the Course: Financial Markets Navigate Uncertainty" (figures and notes as presented in the source).*

### 2. Key Drivers of Financial Conditions Indices

### 2. Key Drivers of Financial Conditions Indices

### Financial conditions and Growth-at-Risk (GaR)
- The IMF’s Financial Conditions Index is designed to capture the pricing of risk. It incorporates various pricing indicators, including real house prices, but does not include balance sheet or credit growth metrics.
- Scenario: If financial conditions were to tighten by 2.5 standard deviations—broadly corresponding to the average of the intraday increases of the Chicago Board Options Exchange Volatility Index level on August 5 relative to its level at the open—and remain at that restrictive level for one quarter, the year-ahead GaR could worsen to its lowest historical quintile.
- Near-term vs. medium-term GaR:
  - Near-term (year-ahead) GaR could move to its lowest historical quintile under an abrupt tightening scenario described above.
  - Medium-term (four years ahead) GaR has been at around historically elevated levels since 2023 and "remains at its worst quintile currently."
- Methodological notes and calibration details:
  - The scenario used in panel 1 calibrates the response of global financial conditions to a spike in Chicago Board Options Exchange Volatility Index, as seen on August 5, 2024 (EDT).
  - The level of Chicago Board Options Exchange Volatility Index used to calibrate the FCI tightening is computed as the average of intraday VIX, recorded at five-minute intervals, from the start of business on August 5 to end of business on that day.
  - Response functions of financial conditions are computed via linear regression methods applied separately over the full sample and the post–COVID-19 sample.
    - The full sample regression starts in 1991:Q1.
    - The post–COVID-19 sample estimation begins in 2021:Q1.
    - A 75 percent weight is applied to the post–COVID-19 regression estimates.

### Emerging markets: resilience, differentiation, and risks
- Post-pandemic actions:
  - Emerging markets deployed proactive monetary policy and in certain cases measures related to foreign exchange to strengthen resilience to external headwinds.
- Overall assessment:
  - The aggregate heat map for emerging market assets shows that market stress has remained largely moderate in interest rates, foreign exchange, and other assets.
  - Market turmoil in advanced economies in early August has not changed this assessment.
- Outlook and risks:
  - As advanced economy central banks cut interest rates while global growth remains resilient, the dollar could weaken and investor sentiment on emerging market assets could turn more positive, spurring renewed portfolio inflows.
  - Global uncertainty is likely to remain elevated owing to geopolitical developments as well as uncertain future policies of newly elected governments.
  - Divergence across emerging markets may become more pronounced; some countries will likely face further external headwinds and idiosyncratic risks that led to recent depreciations of some emerging market currencies.
  - Financial conditions for frontier markets remain challenging, with many countries grappling with higher borrowing costs and financial instability and still not having access to funding through international markets despite sovereign spreads that are moderating lower.

### Global monetary policy synchronization and spillovers to emerging markets
- Interest differentials and carry trades:
  - Positive interest rate differentials in emerging markets vis-à-vis advanced economies narrowed since the April 2024 GFSR, putting pressure on emerging market currencies.
  - Increased volatility, including the rapid appreciation in early August in the Japanese yen, made carry trades less attractive on a risk-adjusted basis.
- Drivers of 2024 currency moves:
  - An IMF staff model finds that while the carry factor dominated currency moves in 2023, in 2024 an idiosyncratic factor (a proxy for domestic policy risks and uncertainty in global markets) has played an important role alongside the strength of the US dollar, notably for Latin American currencies and the South African rand.
  - High-yield sovereigns and commodity exporters have generally been more susceptible to larger FX swings.
- Central bank responses:
  - Several EM central banks became more cautious and slowed or paused their rate cut cycles; some conducted foreign exchange interventions to smooth currency volatility.
  - The Fed rate cut in September and the subsequent weakening of the US dollar have eased some pressures faced by EM central banks, and markets continue to expect easing across emerging markets broadly.
- Synchronization effects:
  - Market participants expect monetary policy cycles in emerging markets to be more synchronized with those in the United States after two years of decoupling.
  - Greater policy alignment should stabilize interest rate differentials between advanced economies and emerging markets, but may increase sensitivity of EM bond yields to advanced economy yields through:
    - More synchronized expected policy paths.
    - Larger spillovers from the term premium component that captures uncertainty in interest rates.
  - Increases in term premiums in most emerging markets have primarily driven recent changes in yields, likely resulting from larger spillovers from higher US term premiums.

### Portfolio flows and capital-flows-at-risk
- Recent flows and issuance:
  - Portfolio flows to emerging markets have been positive on net in recent months.
  - Several countries, notably Egypt and Türkiye, experienced large inflows into local currency bonds; flows into Indian markets have benefited from India’s inclusion in global bond indices.
  - Equity flows have been under pressure in some countries.
  - Year-to-date international issuance of sovereign bonds has risen to its highest level since 2021.
  - Dedicated emerging market bond and equity funds domiciled in the United States have experienced cumulative outflows since March 2022.
- Capital-flows-at-risk metric:
  - The IMF’s capital-flows-at-risk measure indicates that there is a 5 percent probability that emerging market outflows could reach 2.4 percent of GDP over the next three quarters, a marginal increase in outflow risk since the April 2024 GFSR.
  - Rising market volatility, as seen during the early August shock, would materially increase outflows risks if sustained over a longer period.
- Investor base and stabilization:
  - Changes in the investor base have mitigated portfolio outflow risks to some extent: long-term domestic investors like insurers and pension funds have absorbed increasing shares of emerging market bonds, likely serving as a stabilization force.
  - Foreign investors appear more cautious, with portfolio inflow cycles becoming shorter and smaller on average.

### Fiscal buffers, debt dynamics, and policy implications for emerging markets
- Fiscal trajectories and market perceptions:
  - After post-pandemic progress, momentum on fiscal consolidation has waned.
  - Market analysts’ consensus expectations regarding the budget balance for the aggregate government in 15 major emerging markets over the next three years have become more pessimistic and are firmly in deficit territory, with 11 of these countries set to underperform analysts’ forecasts for fiscal year 2024.
- Debt servicing challenge:
  - Still-high global interest rates, larger financial spillovers from advanced economies, and weaker prospects for longer-term growth are making it more difficult to service existing debt for some sovereigns.
  - Some sovereigns risk entering a “debt begets more debt” dynamic unless primary balances improve.
- Fiscal buffer concept and estimation (method summary from source):
  - The 2024 fiscal buffer is estimated by subtracting the long-term debt-stabilizing primary balance from the expected 2024 primary balance.
  - Long-term debt-stabilizing primary balance computed using a simplified formula with assumptions on long-term nominal growth and an effective steady-state long-term interest rate; gross debt is based on prevailing gross government debt as of end-2023.

*Italic: Source — textrevised - 2. Key Drivers of Financial Conditions Indices (PDF chapter/section).*

### 1. Portfolio Flow Tracking: Local Currency Bonds and Equities

### 1. Portfolio Flow Tracking: Local Currency Bonds and Equities

### Portfolio flows and capital-flow-at-risk
- Capital-flows-at-risk worsened modestly; risks appear slightly higher than average.
- “Portfolio flows at risk” is defined as the 5th percentile of the three-quarters-ahead nonresident portfolio flows’ probability density.
- Panel sample notes:
  - Panel 3 includes monthly data on 16 countries for equity flows and 20 countries for bond flows.
  - “EM ex China” includes an unbalanced sample of 20 emerging markets.
  - Daily data on Chinese equity flows ceased being available as of August 11.
- US-domiciled dedicated emerging market bond funds have seen large outflows since 2022.

### Portfolio inflow cycles and fund flows
- Portfolio inflow cycles have become shorter in recent years.
- Cumulative equity and local currency bond inflows before outflows are tracked (monthly data).
- Flows in US-domiciled emerging market mutual funds and exchange-traded funds show pronounced outflows across Jan. 2023 to Oct. 2023 and into 2024.

### Data sources and notes
- Sources: Bloomberg Finance L.P.; EPFR; EUROPACE AG/Haver Analytics; IMF, World Economic Outlook database; national sources; and IMF staff calculations.
- Note: Inflow episodes are reset at the first monthly occurrence of outflows. AUM = assets under management; EM ex China = emerging markets excluding China.

### Emerging message
- Portfolio flow cycles are shorter and capital-flow-at-risk has modestly worsened, with US-domiciled EM bond funds experiencing sustained outflows since 2022.

### _Sources: IMF staff calculations; Bloomberg Finance L.P.; EPFR; EUROPACE AG/Haver Analytics; IMF, World Economic Outlook database; national sources._

---

### Emerging Market Fiscal Buffers and Sovereign Financial Costs

- With fiscal consolidation delayed, deficits have remained above prepandemic levels.
- Panel 1 covers 15 major EM sovereigns whose fiscal balance trajectories are tracked by broad analysts and for which expectations for a two-year forward-looking horizon are available.
- Panel 2 and 3 sample: 16 major EM sovereigns with outstanding external debt denominated in US dollars, with eight in each category.
- Panels 4 and 5 sample: 80 EM sovereigns continuously rated by at least one of three international rating agencies from December 2008 to August 2024 (excludes withdrawn ratings; includes defaults).
- Fiscal-buffer classification:
  - “Small or worsening” sovereigns identified as those with fiscal buffers beyond a deficit of 2 percent.
  - “Borderline” sovereigns have fiscal buffers ranging from –2 to 2 percent and expect widening fiscal year 2024 primary deficits.
  - “Large or improving” sovereigns have large fiscal buffers (exceeding 2 percent), and borderline sovereigns expected to experience narrowing fiscal year 2024 primary deficits.
- A “reasonable rate” is within the –2 to 2 percent range, as the average five-year standard deviation of sample sovereigns’ primary balances is about 2 percent of GDP (based on expectations from fiscal year 2020 to fiscal year 2024).

### Sovereign risk and ratings dynamics
- Emerging markets with worse fiscal buffers generally have higher credit default spreads (Figure references).
- Pricing of sovereign credit risk: spreads are diverging between “large or improving” and “small and worsening” buffers.
- Ratings drift defined as total net change (as a percentage of total) in credit ratings by the three agencies over the preceding six months; positive drift is green, negative drift is red.
- Historical default and ratings context:
  - B-rated sovereigns have a cumulative default rate of up to 17 percent over a period of five years (Fitch, Moody’s, S&P historical studies).
  - The Moody’s study indicates that ratings of defaulted sovereigns, on average, tend to be in the B-rating range one year before a default event.
  - Of the 80 sovereigns sampled, 17 (21 percent) have average ratings at CCC+ or worse, compared with 4 (5 percent) in December 2019.
- Market behavior:
  - Markets often front-run ratings actions; hard-currency spreads for 12 out of a sampled 19 defaulting sovereigns exceeded 10 percent before a downgrade to CCC or worse.
  - Some sovereigns’ spreads have exceeded 10 percent before downgrade to CCC or lower.
- Concern: sequential downgrades and “cliff effects” can be extreme and may push sovereigns closer to losing market access; recent defaults have tended to spend longer durations in CCC or worse bands.

### Policy implication
- The continued struggle for market confidence underscores the importance for emerging market sovereigns to maintain sufficient fiscal buffers and flexibility, especially during periods of strong growth, to mitigate effects of unexpected shocks.

### _Sources: Bloomberg Finance L.P.; Fitch Ratings; JPMorgan; Moody’s Investor Services; S&P Global; and IMF staff calculations._

---

### Frontier Market Developments

- Frontier sovereign spreads tightened further in the second quarter, approaching long-term average levels.
- Progress on debt restructuring has improved investor sentiment; recent restructurings:
  - Eurobond restructurings in Suriname, Zambia and Ghana completed in December, June, and October, respectively.
  - An agreement in principle was reached with creditors in Sri Lanka in September.
- Market access and issuance:
  - Frontier economies continued to issue international debt in the second quarter, although yields remained high.
  - Just 14 percent of frontier economies have sovereign spreads above 1,000 basis points—a lower share than a year ago.
  - Roughly a fifth of frontier economies still have yields close to 10 percent or higher.
- Upcoming maturities and exposure:
  - Significant amounts of frontier debt are coming due: roughly $4 billion in the remainder of 2024, roughly $13 billion in 2025, and roughly $14 billion in 2026.
  - Roughly 60 percent of maturing bonds were issued by countries with prevailing yields close to or above 10 percent.
- Maturity and refinancing risks:
  - Decline in weighted average maturity of frontier debt issuance indicates reliance on shorter-term debt and greater exposure to monetary policy expectations.
- Fiscal metrics:
  - Debt-to-GDP ratios for both emerging market and frontier economies remain well above historical average levels.
  - Under IMF staff projections, these debt levels are not expected to come down meaningfully in the medium term.
  - Interest repayment burdens for frontier economies are projected to ease somewhat but remain relatively high in the medium term.

### Notes and definitions
- Frontier market classification: 43 countries included in the JPMorgan Next Generation Market index or, if not included, low-income countries with international bond issuance.
- Panel 1 shows 25th and 75th percentiles of the JPMorgan Next Generation Market Index.
- Panel 3 shows weighted average maturity of international debt issuance by frontier sovereigns.
- EMBIG = JPMorgan Emerging Market Bond Index Global; WAM = weighted average maturity.

### _Sources: Bloomberg Finance L.P.; EUROPACE AG/Haver Analytics; IMF, World Economic Outlook database; JPMorgan; and IMF staff calculations._

---

### Sustainable Debt Issuance and Climate Finance in Emerging Markets

- Global issuance of sustainable debt rebounded in the first half of 2024.
- Green bonds remained the largest component, accounting for roughly half of sustainable debt issuance and exceeding the amount issued in the first six months of past years.
- The share of issuance by emerging markets has somewhat declined recently.
- Emerging market sustainable debt:
  - Sustainable debt continues to account for a relatively small portion of total debt issuance in emerging markets.
  - The share of offshore issuance of sustainable debt in total issuance of sustainable debt is somewhat higher, indicating emerging-market sustainable issuance is relatively larger in offshore markets than in total debt issuance.
- Country-level data: panel 2 shows country-level sustainable bond issuance in emerging markets as a share of total domestic and offshore issuance (cumulative since 2015); data labels use ISO country codes.
- Climate finance allocation and gaps:
  - Different estimates suggest about 75 to 90 percent of climate finance flows are directed toward mitigation efforts (CPI 2023; OECD 2023; UNEP 2023).
  - International adaptation finance flows to developing countries are 10 to 18 times below estimated needs, and the gap is widening (UNEP 2023).
- Investor behavior:
  - A significant portion of private sector capital providers are unfamiliar with the adaptation investment thesis; perceived risk for adaptation remains prohibitively high.
  - Among private sector investors, mitigation is typically seen as an opportunity, whereas adaptation is often deprioritized.
- Policy concern:
  - Underinvestment in climate change mitigation and adaptation in emerging market and developing economies could lead to global risks to financial stability through greater exposure to systemic climate-related financial risks, including contagion effects along value chains.

### _Sources: Bloomberg Finance L.P.; BloombergNEF; and IMF staff calculations._

### CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARkETS NAvIGATE uNCERTAINTY

### CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARKETS NAvIGATE uNCERTAINTY

### China: slowing growth, deflationary pressures, and market signals
- One-year-ahead expected consumer price index inflation has nearly halved from a year ago, to 1.3 percent.
- The probability that one-year-ahead inflation will fall below its current level of 0.3 percent has increased.
- Home price and housing market indicators:
  - Primary home prices down 7 percent from their peaks.
  - Secondary home prices down 13 percent from their peaks.
  - Primary market sales 40 percent lower than their prepandemic peak.
- Recent policy and market responses:
  - Authorities unveiled a series of monetary and regulatory stimulus measures at the end of September aimed at bolstering the domestic economy and stabilizing the property sector and consumer sentiment.
  - The announcement initially triggered a strong appreciation in stock prices, partially retraced as investors awaited details on potential fiscal stimulus.
- Government bond and fixed-income dynamics:
  - Both the 2- and 10-year central government bond yields have fallen to near record lows.
  - Term premiums, especially for longer-term rates, have compressed, indicating a weaker economic outlook and flight to safety.
  - Outperformance of defensive and high-dividend sectors (utilities, energies) points to low appetite for risk.
  - Foreign investors have increased holdings of renminbi-denominated bonds, particularly negotiable certificates of deposit in the interbank market.
  - A sudden rise in benchmark bond yields could trigger a sharp repricing in broader fixed income markets, redemptions from investment funds, and significant market volatility.
- Short-term interventions and liquidity action:
  - On August 31, the central bank announced it conducted secondary market transactions in August by buying short-term central government bonds and selling long-term central government bonds, resulting in a net liquidity injection of 100 billion yuan.
- Retail and institutional investor behavior:
  - Retail-focused wealth management products and mutual funds have displayed strong appetite for fixed income assets.
  - Local government financing vehicle (LGFV) debt led declines in other bond yields following fiscal support for financially weak regions.

### Banks, nonperforming assets, and asset management companies (AMCs)
- Reported bank asset quality and NPLs:
  - Low mortgage defaults with NPL ratios less than 1 percent.
  - Manageable direct exposures to developers at less than 6 percent of total bank loans.
  - Since 2012, cumulative reported NPLs amounted to less than 3 trillion yuan in 2023.
- Write-offs and disposals:
  - Banks have been proactive in addressing NPAs, with write-offs and disposals topping 3 trillion yuan each year since 2020.
  - Write-off and other disposal totaled 22 trillion (cumulative) which have effectively lowered banks’ headline NPL ratios by 1.5 percentage points.
  - Disposals have been done mainly through transferring NPAs to state-owned AMCs in the primary market; the secondary market—non-AMC buyers of NPAs—remains nascent.
- AMC market structure and risks:
  - The four national AMCs established in 1999 hold 80 percent market share in the primary NPA market.
  - More than 50 regional AMCs have been established since 2015; regional AMCs target smaller banks in their regions.
  - A fifth national AMC, established in 2020, remains small, holding less than 0.2 percent of total AMC assets.
  - AMCs have expanded into conglomerates offering lending, trust, insurance, brokerage, and real estate services.
- AMC fundamentals and vulnerabilities:
  - Fundamentals of national AMCs have weakened since 2018; profitability has suffered.
  - Capital levels, as proxied by equity-to-asset ratios, have dropped to distressed levels of below 5 percent at two of the national AMCs.
  - Regional AMCs appear more resilient on limited disclosure measures but are unlikely in the near term to fill gaps left by national peers.
  - AMCs are intertwined with the financial system through investments, lending (receivables), and reliance on bank and market financing; a credit event at a large AMC would hamper a source of NPA disposal for banks.
  - Distress in one national AMC generated ripple effects requiring a $6.6 billion state-led bailout in 2021.
  - Authorities have strengthened regulations on AMCs, including centralizing supervision of local AMCs under the National Financial Regulatory Administration.
- Composition and exposures:
  - Disclosures from listed AMCs show the bulk of NPA acquisitions in 2023 originated from the property market, small and medium enterprises, and LGFV-related sectors.
  - AMCs’ balance sheets include investments, receivables, trading securities, cash, long-term debt, short-term debt, and other current liabilities (composition detailed in source charts).

### Corporate credit: valuations, refinancing risks, and vulnerabilities
- Corporate bond valuation and investor behavior:
  - Misalignment in corporate bond valuation, relative to a model accounting for macro fundamentals, has remained at levels similar to those in April 2024.
  - Degree of overvaluation among US issuers is elevated by historical standards.
  - Japanese investors have reportedly preferred US investment-grade corporate bonds to Treasury securities because yields on the former more than compensate for foreign exchange hedging costs.
  - Foreign-exchange-hedged yield of US investment-grade corporate bonds has been attractive to Japanese investors (chart series shown in source).
- Credit instruments and market structure:
  - Syndicated lending has recently regained share from private credit direct lenders.
  - Collateralized loan obligation issuance volumes in the United States and euro area for the second quarter of 2024 were 60–90 percent higher than the average volumes between the first quarter of 2022 and the first quarter of 2024.
- Defaults, insolvency risk, and debt structure:
  - Loan and bond defaults have steadily risen as weaker firms have struggled.
  - Forward-looking metrics indicate around one-quarter of firms are vulnerable to insolvency (distance to insolvency measures).
  - Bankruptcy cases have continuously risen in recent months, with cases exceeding prepandemic levels in Europe and Japan.
  - Among investment-grade firms, the amount of debts issued by “fallen angels” is now roughly equal to the amount of “rising stars,” a reversal from the position reported in April 2024.
- Maturity, leverage, and refinancing risks:
  - Average debt maturity has decreased as borrowers have issued less long-term debt, increasing refinancing risks by concentrating repayment obligations over a short period.
  - A debt-to-GDP ratio adjusted for remaining years to maturity shows corporate leverage has been rising and is now at about its highest levels since the period after the global financial crisis.
  - The risk is more pronounced in the high-yield segment, where the maturity of debt has dropped much more steeply.
  - The average remaining life of high-yield debt reached 4.6 years in the fourth quarter of 2023.
- Firm-level liquidity pressures:
  - The share of firms with cash-to-interest-expense ratios below 1.5 has been increasing, especially among smaller firms.

*Global Financial Stability Report: STEAdYING ThE COuRSE: uNCERTAINTY, ARTIFICIAL INTELLIGENCE, ANd FINANCIAL STABILITY (October 2024)*

### 6.7 years in the third quarter of 2009, when leverage based on the

### textrevised - 6.7 years in the third quarter of 2009, when leverage based on the

### Corporate debt sustainability
- Fixed-rate debt accounts for close to 50 percent of global corporate debt coming due in 2025, with existing coupons between 3.5 and 4 percent, significantly lower than the current refinancing yield of 5.5 percent.
- Should monetary policy not ease, refinancing 2024 and 2025 bonds at higher interest rates would bring ICRs down by an average of 12 percent.
- Refinancing costs remain elevated specifically for emerging market corporations, especially for foreign currency bonds, putting pressure on debt sustainability.
- Issuance in emerging markets has remained much slower than before the current monetary tightening cycle.
- An easing in monetary policy, or a shift toward issuing in local rather than foreign currencies, would help firms in emerging markets with debt sustainability.
- Trade restrictions or geopolitical events that raise input costs would compress margins and further deteriorate ICRs:
  - In a scenario where input costs increase by 10 percent, the weak tail of firms (ICRs less than one) would increase by an additional 3 to 6 percentage points, depending on the region, with impact especially large in emerging markets.
  - The scenario calibration reflects higher marginal financing costs (by 150 basis points) and potential upward pressures on input costs from factors like recalibration of international trade policies or supply chain disruptions.
- About 90 percent of firms in advanced economies have little or no meaningful market power (April 2019 World Economic Outlook), implying higher input costs would likely adversely affect profit margins for most firms.

### Private credit: expansion and rising vulnerabilities
- Private credit continues to grow and has expanded beyond lending to midsized corporate borrowers, intensifying competition with banks in syndicated loan markets.
- Favorable outlooks have pushed up stock prices of specialized asset managers, which have outperformed bank stocks and the broader equity market.
- Signs are mounting that high interest rates are pressing private credit borrowers:
  - ICRs have continued to decline because of borrowers’ high leverage, the floating-rate nature of loans, and the slowdown of economic activity.
  - Defaults narrowly defined (missed payments) are relatively rare due to flexibility in private credit vehicles, but defaults under broader measures—including restructurings or breaches of covenants—are becoming frequent.
  - A significant share of borrowers are facing cash flow pressures, evidenced by an ever-growing share of payment-in-kind (PIK) coupons.
- Opaqueness of the private credit industry makes it challenging to assess risks and quantify deterioration of private credit loans.
- Downside scenario risks:
  - Stale and uncertain valuations could lead to deferred realization of losses followed by a spike in defaults.
  - Crisis of confidence could be triggered by an outsized share of defaults in a group of funds, leading to frozen fundraising, runs on semiliquid funds, withdrawal of leverage and liquidity from banks and other investors, simultaneous reductions in exposures, and spillovers to other markets and the broad economy.

### Residential real estate: modest declines, contained stability risks
- On an annual basis:
  - Real home prices in emerging markets have declined by 1.6 percent.
  - Real home prices in advanced economies have declined by 0.3 percent.
- Global real house prices remain 5 percent above the prepandemic average, keeping affordability stretched globally.
- Supply-side constraints (rising construction costs, shortages of construction materials) have partly dampened pass-through of elevated interest rates on demand by lowering affordability; price elasticity of new housing supply varies across countries.
- Country-specific developments:
  - Norway and Canada have recorded significant declines; Korea, South Africa, Sweden, the euro area, and the United Kingdom have undergone annual declines; China’s property remains weak despite recent government support.
  - US house prices have increased 2 percent year over year, supported by brisk absorption of housing inventories and lower mortgage rates boosting refinancing and mortgage origination.
- Risks and projections:
  - There is room for house prices to decline further in jurisdictions with high household leverage, overvalued markets, or where substantial easing in monetary policy is less likely.
  - Risks to financial stability are contained: further increases in mortgage rates are not projected to raise household debt-servicing expenses significantly (see “Scenario 1”); a limited number of risky and complex financial instruments are tied to housing; household and bank balance sheets are sound overall.

### Commercial real estate: acute pressures and repricing
- Global CRE prices have fallen by 12 percent year over year, weighed down by still-high interest rates and poor investor sentiment.
- Sector-specific declines:
  - US office sector is experiencing a 23 percent decline.
  - European office sector is experiencing a 16 percent decline.
- Risks:
  - CRE is at risk of further correction, especially if financial institutions active in lending to this market (including real estate investment trusts, commercial mortgage-backed securities, and some banks) come under strains.
  - Funding could be withdrawn discreetly, pushing down prices and creating adverse feedback loops that put more institutions under pressure.
  - Price declines appear to be stabilizing for CRE owned by institutional investors, but vulnerabilities remain acute.

*Italicized source attribution as provided in the original content.*

### 1. Changes in Private CRE Valuations

### 1. Changes in Private CRE Valuations

### Key findings on markets and valuations
- Transaction volumes were just over $130 billion in 2023, a 37 percent decrease from the previous year.
- Offices historically accounted for 40 percent of cross-border CRE investments between 2010 and 2023; this share has declined by close to 10 percentage points since 2022.
- The spread of prime property yields over long-term government bond yields has eased in some regions, while in Europe the excess spread is rebounding and nearing its 25-year historical average.
- In the United States, market agencies project rates of capitalization will peak in 2024.

### Regional and sector differences in demand and vacancy
- US metro areas have higher vacancy rates than other global cities and are projected to have negative net absorption rates (occupancy is outpaced by newly vacant space).
- Technological transformations (artificial intelligence and cloud computing) are expected to boost demand for data centers and similar CRE, especially in Asia-Pacific.
- Postpandemic shifts to remote working and changing international trade patterns are producing diverging country and regional performance across CRE sectors.

### Funding conditions and investor behavior
- Sources of CRE funding have shifted: tight bank lending standards and subdued investor sentiment are expected to further restrict CRE financing, leading to project delays or cancellations and reducing supply.
- Equity investments by institutional investors have declined significantly as they favor debt instead.
- Debt funds have significantly outperformed equity investments in European real estate since the end of 2022 (per MSCI’s Europe Quarterly Private Real Estate Debt Fund Index).
- Presence of CRE debt premiums (spread between 10-year fixed-rate CRE and corporate A to Baa rates) may be contributing to property value corrections.

### Maturing CRE debt and funding gaps
- In the United States nearly $1 trillion in CRE debt will mature between 2024 and 2025, with a funding gap of almost $300 billion.
- Globally, about 40 percent of loans held by banks, 25 percent by commercial mortgage-backed securities, and 20 percent by investor-driven lenders (like debt funds) are maturing over this period.
- CMBS lenders have the largest exposure to loans maturing in 2024, accounting for nearly 30 percent of the balance.
- Delinquencies of CMBSs specializing in office properties are above 8 percent, up 3 percentage points from the previous year.

### Risks for banks and stress-test results
- Banks with global footprints have the greatest exposure to vulnerable loans on central business district (CBD) offices: this segment accounted for 26 percent of their total CRE loan originations over the past three years; the same share is 4 percent for national and regional and local banks.
- A review of 398 banks (Asia, Europe, United States) shows many have high ratios of CRE loans to Tier 1 capital, particularly in the United States.
- In an adverse scenario where CRE office exposures lose 50 percent of their value:
  - Aggregate Tier 1 capital ratio of US banks would decrease from 12.3 to 11.3 percent.
  - Aggregate Tier 1 capital ratio of European banks would decrease from 17 to 13.3 percent.
  - 4 percent of the banks in the sample (US and European banks)—representing 1 percent of assets—would have Tier 1 capital ratios dip below 7 percent.
- Among banks that report CRE office exposures, about 25 percent of sample US banks and almost 50 percent of sample European banks report CRE office exposures to Tier 1 capital greater than 50 percent.

### Nonbank investors and REITs
- Real estate investment trusts (REITs) that depend on bank funding for liquidity have elevated expected frequencies of default in Canada and the United States.
- Funding conditions are also affecting alternative investors in CRE markets, with elevated expected default frequency for some REITs (panel 5 in source material).

### Downside price risk and scenario projections
- CRE price-at-risk model estimates with 5 percent probability that real prices could decline over the next three years by about:
  - 20 percent in North America.
  - 19 percent in Europe.
- The CRE price-at-risk approach follows Deghi, Mok, and Tsuruga (2021) and accounts for supply, demand, and financing factors; prolonged high interest rates and tighter financing heighten downside risks.

### Implications for supervisors and investors (policy-relevant observations)
- Rate cuts alone might not resolve all challenges facing CRE investors due to postpandemic remote work and reshaped demand, especially for CBD offices.
- Banks could lend more conservatively toward CBD office and other vulnerable CRE segments, posing challenges to refinancing of large loan volumes coming due.
- The lack of granular CRE disclosures complicates risk assessments; investors appear to penalize banks that forgo detailed information—US banks with high CRE concentrations that disclose office exposures tend to outperform those not disclosing.
- Supervisory attention is paramount to assess the effect of a downturn on banks’ safety and the soundness of business models, especially for weak institutions.
- Attention is needed on the integration of anti–money laundering and combating the financing of terrorism measures within the broader financial stability framework.

*Source: IMF staff, “1. Changes in Private CRE Valuations,” GLOBAL FINANCIAL STABILITY REPORT: STEADYING THE COURSE: UNCERTAINTY, ARTIFICIAL INTELLIGENCE, AND FINANCIAL STABILITY (October 2024).*

### 1. US Bond Fund Assets Under Management

### 1. US Bond Fund Assets Under Management

### Key findings on bond funds and leverage
- Peak outflows tend to be higher for institutional mutual funds and ETFs.
- Fund flows-at-risk are particularly high for emerging market bond funds, especially ETFs.
- Leveraged bond funds take on significant leverage through repurchase agreements.
- Leveraged bond funds tend to experience larger peak outflows compared with their nonleveraged peers.
- Leveraged funds currently constitute a small share of the bond fund sector, although there are differences across jurisdictions.
- Regulators should be aware that deleveraging by even a small set of funds could have an outsized effect on the broader financial system.

### Fund-level metrics and vulnerabilities (methodology notes)
- “Fund flows-at-risk” are the 5th percentile of flows, based on historical flow data (in 5 percent of cases, outflows would have been larger).
- In panels 2–5, analysis is based on Lipper data covering US-domiciled bond funds; panels show median values and interquartile ranges across funds within each category.
- The median flow, the fund flows-at-risk, and the median and peak repo usage are first computed for each individual fund, based on monthly data spanning 2014–24, before the distribution across funds is computed.
- Peak repo usage refers to the 95th percentile of a fund’s monthly data on repo usage.
- In all panels, the analysis of MF flows covers open-ended MFs only, and “MFs” is used as shorthand for open-ended MFs.
- Definitions and abbreviations: EM = emerging market; ETF = exchange-traded fund; HC = hard currency; HY = high yield; IG = investment grade; LC = local currency; MF = mutual fund.

### Hedge funds, carry trades, and market amplification
- Hedge funds are a $7 trillion industry very much connected to the rest of the financial markets.
- Hedge funds with momentum and macro strategies participated heavily in carry trades, building substantial short positions in yen.
- These short yen positions were often matched with long positions in US equity futures and in currencies of emerging markets.
- After the Bank of Japan’s monetary policy decision and worse-than-expected US labor market data, the interest rate differential between Japan and the United States rapidly narrowed, equities declined, and the yen appreciated.
- Many hedge funds reportedly reached risk limits and received increased margin calls, forcing them to rapidly close positions and erasing the year’s returns for many hedge funds.
- The rapid unwinding of crowded and concentrated positions could exacerbate price movements across global indices and propagate stress throughout the financial system.
- Limited transparency in the hedge fund industry makes it difficult for investors and supervisory authorities to gauge leverage in real time and what might trigger another bout of hedge fund deleveraging.

### Evidence (selected chart annotations)
- Noncommercial net positions on yen futures are shown in billions of US dollars.
- Ratio of long to short futures and options in US equities is reported for futures and options on E-mini contracts for the S&P 500, the Dow Jones, and the Nasdaq.
- Returns of commodity trading advisor hedge funds are presented as an index; the index in panel 3 is calculated as the weighted average return of 10 selected mutual funds managed by some of the largest commodity trading advisor hedge funds globally.
- Key dates annotated in the hedge fund returns series include July 2, 2024 and Aug. 6, 2024.

### Illiquid investments by pensions and insurers and maturity mismatches
- The share of defined-contribution pensions and unit-linked insurance products has risen globally in recent years.
- Providers of defined-contribution plans typically offer clients frequent opportunities to enter or exit investment options, which may exacerbate liquidity mismatches between liabilities and illiquid underlying assets.
- Australian superannuation funds are required to allow clients to switch between different investment options generally within three business days, even though these funds hold, on average, illiquid exposures exceeding 20 percent of their total assets.
- Selected large private defined-contribution pension and superannuation funds have increased allocations to illiquid private equity and credit in recent years.
- A sample of 26 selected defined-contribution private pension and superannuation funds has $1.4 trillion in assets under management (domiciled in Australia, Canada, Germany, Mexico, Sweden, Switzerland, the United Kingdom, and the United States).
- Illiquid level 3 assets in five of the largest Australian superannuation funds, with assets under management exceeding $0.5 trillion, are estimated to account for almost one-quarter of total assets.
- US insurers’ illiquid investments are calculated as the sum of miscellaneous assets, mortgages, and real estate investments as of the end of 2022, according to the American Council of Life Insurers’ latest published factbook.
- Illiquid investments of EU insurers are calculated as the sum of investments in real estate funds, alternative funds, private equity funds, infrastructure funds, real estate structured notes, real estate collateralized securities, mortgages, loans, and property as defined by the European Insurance and Occupational Pensions Authority.
- European unit-linked insurance products have not increased the shares of illiquid investments in their portfolios to the same extent as US annuities; exposures are materially smaller than those of European general account insurers.

### Policy recommendations (selected)
- Central banks should be data dependent while communicating clearly that the path of policy rates should not react excessively to any individual data point; where growth and inflation momentum slow, central banks should gradually ease policy toward a more neutral stance.
- Central banks should monitor a broad spectrum of indicators encompassing liquidity conditions and funding rates in money markets, remain attuned to uneven distribution of liquidity and reserves across banks, and stand ready to address market stresses; policymakers should clearly communicate objectives and steps for removing liquidity.
- Emerging market central banks should continue to ensure inflation targets are met and preserve resilience against external pressures; countries should integrate policies where applicable using the IMF’s Integrated Policy Framework.
- The use of foreign exchange interventions may be appropriate as conditions warrant, provided intervention does not impair macro policy credibility or substitute for necessary adjustment; capital flows management measures may be an option in imminent crises as part of a broader package but should not substitute for warranted macroeconomic adjustments.
- Fiscal adjustments should focus on credibly rebuilding buffers to keep external financing costs reasonable and to help anchor medium-term inflation expectations; countries with less fiscal space should ensure credibility of fiscal plans to prevent cliff effects in ratings.
- Authorities should enhance creditor coordination, use the Group of Twenty Common Framework when applicable, and promote wider use of enhanced collective-action clauses and majority voting provisions in syndicated loans to facilitate preemptive orderly restructurings.
- China needs accommodative macroeconomic policies and structural reforms to bolster near-term activity; property sector policies should prioritize completion of presold unfinished housing and timely restructuring of troubled developers; additional easing of monetary policy (especially lower interest rates) and reorientation of public expenditures toward households could bolster near-term recovery.
- Additional regulatory measures in China to enhance management of liquidity and maturity risk and to close regulatory and data gaps in the nonbank financial sector could help contain systemic risks.
- To scale up adaptation finance, align public and private sector interests, improve tracking and measurement of adaptation finance flows, provide investment guidance and adaptation taxonomies, and integrate adaptation considerations across asset classes; the Resilience and Sustainability Trust has integrated adaptation support in its 18 programs since establishment in 2022.
- Authorities should collect detailed information on CRE exposures, conduct stress-testing exercises (including smaller banks with material CRE exposure), review banks’ CRE valuation assumptions, and ensure provisions are adequate.
- Given private credit’s exponential growth and increasing retail participation, authorities may consider enhancing reporting requirements and adopting a more intrusive supervisory and regulatory approach to monitor credit, liquidity, leverage, valuations, and interconnectedness.

*Source: IMF staff calculations and analysis as presented in the chapter excerpt.*

### CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARkETS NAvIGATE uNCERTAINTY

### CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARkETS NAvIGATE uNCERTAINTY

### Macroprudential framework and banking resilience
- The buildup of debt amid elevated uncertainty underscores the need to strengthen the macroprudential policy framework to contain excessive risk taking in the nonbank financial sector and to ensure that capital and liquidity buffers in banking systems are adequate to support the provision of credit through periods of stress.
- Policy recommendations:
  - Tighten appropriate macroprudential tools to increase resilience against a range of shocks and forestall further increases in pockets of elevated vulnerabilities, while avoiding a destabilizing tightening of financial conditions.
  - Enhance financial sector regulation and supervision to address the tail of weak banks and the risk of contagion to healthy institutions.
  - Promote full, timely, and consistent implementation of Basel III and other international standards.
  - Ensure that financial institutions are prepared to access central bank liquidity and intervene early to address liquidity stress.
  - Require all banks to test their access to central bank instruments periodically.
  - Central banks should set up frameworks for emergency liquidity assistance in normal times and be ready to provide liquidity against a broad universe of assets while abiding by principles concerning solvency and viability, collateralization, and appropriate haircuts.
  - Make further progress on adopting and implementing frameworks for recovery and resolution to address weak or failing banks proactively without undermining financial stability or risking public funds.

### Nonbank financial intermediation (NBFI), leverage, and data gaps
- Market turmoil in early August highlighted how leveraged NBFIs can amplify stress through rapid unwinding of leveraged positions that generate liquidity imbalances even in the absence of defaults.
- Key vulnerabilities and actions:
  - Inadequate data hinder authorities’ ability to assess vulnerabilities associated with nonbank leverage and to identify large and concentrated positions.
  - It is crucial to enhance reporting requirements for nonbank institutions and to strengthen policies that mitigate vulnerabilities and amplification mechanisms stemming from nonbank leverage, where judged to pose a threat to financial stability.
  - The growth of bond funds highlights the need to reduce systemic risks by ensuring the effectiveness of tools for managing liquidity.
  - Timely and consistent implementation of the Financial Stability Board’s revised recommendations to address structural vulnerabilities from liquidity mismatches in open-ended funds is crucial.
  - Data gaps often hinder monitoring of interconnectedness risks posed by pension funds and insurers; supervisors should fill data gaps and cooperate, including across borders.
  - In jurisdictions where defined-contribution pensions and unit-linked insurance products are material, supervisors should closely monitor the share of illiquid investments held by these products.
  - Conduct liquidity stress tests that consider crises related to liquidity availability across major asset classes and ensure compatibility between asset liquidity and notice periods for client switches between investment products.

### Synthetic risk transfers (SRTs): scale, drivers, and risks
- Observations on market size and issuance:
  - Globally, more than $1.1 trillion in assets have been synthetically securitized since 2016, of which almost two-thirds were in Europe.
  - Industry estimates expect issuance of SRTs to remain above $200 billion in Europe and to more than triple in the United States to surpass $50 billion in 2024.
- Investor returns and motivations:
  - Investors purchase SRTs to access loan categories not easily accessible through public markets or direct lending and to earn attractive returns of 8–12 percent compared with those from other asset classes, as well as to meet mandates to allocate capital in private credit.
- Structural features and potential systemic risks:
  - SRTs move credit risks from banks to investors through a financial guarantee or credit-linked notes while keeping loans on banks’ balance sheets; through this credit protection, banks can claim capital relief and reduce regulatory capital charges.
  - Stylized example (preserve exact figures shown in panel):
    - Assumption: Tier 1 Capital Requirement of 10.5 percent
    - Without SRT: $87.50 (20% RWA); $11.00 (0% RWA); $1.50 (1,250% RWA)
    - With SRT: Tier capital requirement: $10.5; Loan portfolio $100 (100% RWA); Reference pool $100; Tier 1 capital requirement: $3.80
  - Risks identified:
    - Elevated interconnectedness and potential negative feedback loops during stress, including anecdotal evidence that banks may provide leverage for credit funds to buy credit-linked notes issued by other banks.
    - Opaqueness of the market: only a fraction of deals are public and there is no centralized repository for SRT data.
    - SRTs may mask banks’ degree of resilience by increasing regulatory capital ratios while overall capital levels remain unchanged.
    - Possible overreliance on SRTs if banks cannot build capital organically due to weaker fundamentals and profitability.
    - Exposure to business challenges if SRT market liquidity dries up.
    - Signs of increased concerns regarding deterioration of asset quality in SRT reference pools (credit risk indices point to worsening, particularly for SMEs).
    - Risk of financial innovation leading to securitization of riskier asset pools and reduced clarity about ultimate risk holders.
    - Cross-sector regulatory arbitrage may reduce capital buffers in the broad financial system even if bank-level charges fall.
- Supervisory actions:
  - Financial sector supervisors need to closely monitor SRTs, ensure necessary transparency regarding SRTs, and assess their impact on banks’ regulatory capital.

### Tokenization of real-world assets and interconnectedness with crypto markets
- Tokenization involves creating a digital representation of real-world assets on a blockchain and has been used for money market funds’ shares, repos, and Treasuries.
- Benefits observed or expected:
  - Potential immediate trade settlement, lower costs related to ownership, fractional use of safe and liquid collateral, and timely receipt of asset yields or coupons.
  - For repos, blockchain-based transactions can provide immediacy and cost-efficiency to manage intraday liquidity and mitigate costly intraday central bank repos.
- Market participation and performance:
  - Traditional finance institutions have entered the space; examples cited include BlackRock and Franklin Templeton launching tokenized Treasury funds (BlackRock USD Institutional Digital Liquidity Fund and Franklin OnChain US Government Money Fund).
  - Performance example through end of July 2024: Franklin OnChain US Government Money Fund’s average annual return was 5.3 percent, compared with 5.0 for Federated Hermes’ Treasury Obligations Fund.
- Financial stability considerations:
  - Current concerns are limited given small scale, but medium-term risks arise from deepening nexus between crypto and traditional markets.
  - Potential channels for volatility or shocks include investor uncertainty about token valuation or redemption, shocks occurring when underlying real-world assets cannot be traded (for example, over a weekend), technology risks, and increased use of leverage via tokenization.
  - Growth of tokenized safe and liquid assets could interact with the rise of stablecoins.
- Supervisory actions:
  - Supervisors should continue to monitor risks related to interconnectedness within crypto markets and between those markets and traditional capital markets.

### Carry trades, monetary policy divergences, and market volatility
- Carry trades involve borrowing in low-cost currencies and investing in higher-return assets to earn the “carry” spread.
- Recent developments and effects:
  - Over the past several years, the relatively low interest rates in Japan vis-à-vis other advanced economies made the yen a preferred funding currency for carry trades.
  - Many investors used the yen to invest in Brazilian and Mexican government bonds, Indian equities and corporate bonds, and US technology stocks in artificial intelligence.
  - Carry trades accumulate during periods of sustained low volatility and can unwind rapidly when volatility surges, potentially destabilizing markets.
  - Worse-than-expected US labor market data following the Bank of Japan’s monetary policy decision in July meant carry trades were no longer profitable, and their unwinding led to spikes in stock and currency volatility in early August.
  - The interest rate differential between the dollar and yen narrowed, and the yen appreciated in a speed-up of trends that began in July.
  - High-yielding currencies targeted by carry trades depreciated; the Nikkei experienced a collapse (panel references in source indicate currency and equity impacts).
- Measurement challenges:
  - It is difficult to estimate the overall size of carry trade positions; one upper-bound guide is the amount of Japanese yen borrowed by nonresidents.

*Source: CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARkETS NAvIGATE uNCERTAINTY (textrevised - CHAPTER 1 STEAdYING ThE COuRSE: FINANCIAL MARkETS NAvIGATE uNCERTAINTY).*

### 12.4 percent on August 5, its largest one-day move

### 12.4 percent on August 5, its largest one-day move

### Carry trade unwind and the VIX surge (August 5, 2024)
- VIX recorded a 12.4 percent move on August 5, its largest one-day move since 1987.
- The Chicago Board Options Exchange Volatility Index (VIX) surged from 16 to more than 65, before lowering to 37 by the end of the day.
- Other major indices moved sharply: S&P 500 (–3 percent) and STOXX Europe 600 (–2 percent).
- Drivers identified:
  - Unwind of carry trades by nonbank financial intermediaries (including hedge funds).
  - Momentum and algorithmic traders guided by trend-following algorithms.
  - Broker–dealers selling stocks to hedge risks created by selling large amounts of options to clients.
- Episode characteristics and interpretation:
  - The period of high volatility was short-lived; risk assets regained most losses in subsequent days.
  - Traders judged declines not justified by macroeconomic fundamentals.
  - The massive VIX move largely reflected issues related to the index’s construction (greater role of out-of-the-money options) rather than widespread market transactions.
  - Front VIX futures saw much smaller moves on August 5 than the index itself.
  - Front VIX futures expire on August 21, and much lower futures prices relative to the VIX indicated traders did not view the high VIX level to be sustained for more than two weeks.
  - In past episodes (COVID-19 pandemic and the 2015 China devaluation), the VIX and VIX futures surged in tandem.

### Potential systemic and regulatory implications
- The episode highlights the potentially destabilizing role of leveraged strategies such as carry trades in global markets.
- Emphasis on need for more regulatory scrutiny, especially regarding nonbank financial intermediaries.
- Reminder that a disconnect between heightened uncertainty and low market volatility can abruptly close, with adverse consequences for asset prices.
- Uncertainty remains about the potential scale of future unwinding of carry trade positions.

### Nonbank financial intermediaries and local bond markets
- Total assets under management in nonbank financial intermediaries have grown since 2002 to reach 25 percent of GDP in the median emerging market from less than 10 percent as recently as 2003.
- Nonbank financial intermediaries—especially pension and insurance funds—are playing an increasingly important role in markets for local currency government bonds (LCGBs).
- Effects of a larger domestic institutional investor base:
  - Provides governments with a stable source of funding as other asset classes are less developed.
  - Reduced reliance on foreign capital, mitigating risks of capital outflows.
  - Enhanced market depth and liquidity; reduced volatility.
  - Countries with increased share of LCGB holdings by pension and insurance companies have experienced less volatility in term premiums; those with declining shares have seen term premiums rise.
  - A sizable domestic investor base with a long-term horizon could mitigate a sovereign–bank nexus.
- Risks from concentration:
  - Funds may become overly concentrated in LCGBs, leaving them vulnerable to large losses if interest rates rise precipitously, the yield curve steepens sharply, or inflation surges.
  - Unexpected increases in redemptions from these funds could drive a sudden rise in domestic funding costs.
- Instruments and market development:
  - Long-dated and inflation-linked LCGBs are already a large fraction of outstanding LCGBs in several countries and could present opportunities to extend the domestic yield curve.

### Key statistics and dates preserved from the source
- VIX: 16 → more than 65 → 37 (intraday on August 5, 2024).
- S&P 500: –3 percent (August 5, 2024).
- STOXX Europe 600: –2 percent (August 5, 2024).
- Front VIX futures expire on August 21 (2024).
- Nonbank financial intermediaries’ assets under management in the median emerging market: 25 percent of GDP (current) and less than 10 percent of GDP (2003).
- Local currency government bond holdings by pension and insurance funds assessed from December 2021 to December 2023.

_Italic: International Monetary Fund, Global Financial Stability Report: Steadying the Course: Financial Markets Navigate Uncertainty (October 2024)._

### Introduction

### Introduction

### Elevated macroeconomic uncertainty since the COVID-19 pandemic
- Uncertainty about economic outcomes and policies "spiked during the COVID-19 pandemic and has remained high since then" compared with earlier years amid "inflation shocks, escalating geopolitical tensions, rapidly emerging new technologies, and increasing climate-related risks."
- Different measures of macroeconomic uncertainty remain volatile but "on average have stayed elevated since the pandemic." Some measures, such as global economy policy uncertainty of Baker, Bloom, and Davis (2016), "declined in the first quarter of 2024 but rose again in the second quarter amid electoral uncertainty in some major economies" (Online Annex Figure 2.1.1).
- Chapter prepared by Rafael Barbosa, Yuhua Cai, Mario Catalán (co-lead), Andrea Deghi (co-lead), Li Lin, Tatsushi Okuda, Mustafa Yenice, and Aleksandr Zotov, under the guidance of Mahvash Qureshi. Ian Dew-Becker and Stefano Giglio served as external advisors.

### Channels through which macroeconomic uncertainty affects macrofinancial stability
- Macrofinancial stability (defined in terms of systemic risk measured by downside tail risks to future real GDP growth) can be affected through three key channels:
  - Market channel: "It can exacerbate downside market tail risks in the event of an adverse shock"—raising downside market tail risk and amplifying realized negative asset returns.
  - Real channel: "It can delay private sector consumption and investment decisions, slowing economic activity and raising credit risks for financial institutions," potentially triggering adverse macrofinancial feedback loops.
  - Credit channel: "It can reduce the supply of domestic credit by financial institutions by exacerbating challenges in determining the creditworthiness of new borrowers."
- These three channels "can interact and mutually reinforce each other, amplifying the effect of macroeconomic uncertainty on macrofinancial stability."
- Macrofinancial vulnerabilities (for example, "high levels of public debt relative to GDP," "high leverage in the corporate sector," or "overly compressed credit spreads") can magnify investor reactions and market corrections when uncertainty is high.
- Financial variables may not fully span macroeconomic uncertainty; financial indicators may explain about "80 percent of the variation in commonly used measures of macroeconomic uncertainty for advanced economies like the United States, and 40 to 50 percent of the variation in those for major emerging markets such as Brazil."

### Cross-border spillovers and market contagion
- "The effect of macroeconomic uncertainty can spill over across borders" via forced asset sales by regionally hit investors, causing asset-price declines and international financial contagion.
- By reducing domestic consumption and investment, macroeconomic uncertainty can weaken import demand and "raise downside risks to economic activity in trading partner countries."
- Bond and stock market volatility tend to be positively correlated across major economies, and this correlation "seems to have increased since the pandemic," suggesting faster spread of asset-market stress (Online Annex 2.1).

### Measures and sources of macroeconomic uncertainty
- Sources of macroeconomic uncertainty considered:
  - Real-sector innovations (output, product prices, factor costs, firms’ profitability).
  - Domestic policies (monetary, fiscal, trade, regulatory).
  - Geopolitical tensions (conflicts, cross-border barriers).
- Specific measures used:
  - Real sector uncertainty: the real economic uncertainty index (REU) of Jurado, Ludvigson, and Ng (2015) and Ludvigson, Ma, and Ng (2021); dispersion in real GDP forecasts from Consensus Economics.
  - Domestic policy uncertainty: the text-based economic policy uncertainty index of Baker, Bloom, and Davis (2016); the world uncertainty index of Ahir, Bloom, and Furceri (2022).
  - Geopolitical uncertainty: the text-based geopolitical risk index of Caldara and Iacoviello (2022).
- Correlations among measures:
  - Different measures "tend to be positively but not strongly correlated," with the REU exhibiting the strongest correlation with other measures.
  - Measures of macroeconomic uncertainty correlate with financial uncertainty measures (for example, Ludvigson, Ma, and Ng (2021) financial uncertainty or the VIX) in a generally positive but modest manner.
  - Episodes of synchronized spikes across measures occur in major crises (global financial crisis, COVID-19 pandemic); other events may be reflected only in particular measures (for example, the dot-com bubble captured mainly by financial uncertainty measures; US–China trade tensions largely captured by the economic policy uncertainty index).
- Macro-market disconnects can occur when "macroeconomic uncertainty is high and financial market volatility (realized and implied) is low." Factors driving disconnects include investor expectations of policy protection, low-quality political signals, divergence in investor opinions, strong equity performance, and hedging strategies keeping implied volatility low.

### Empirical approach and extensions to growth-at-risk (GaR)
- The chapter studies risks to macrofinancial stability using an augmented growth-at-risk (GaR) framework:
  - Uses panel data from "a sample of 43 advanced and emerging market economies since 1990 (or the earliest year for which data are available)."
  - Empirical questions:
    - Does macroeconomic uncertainty help predict downside risks to output?
    - How does macroeconomic uncertainty interact with macrofinancial vulnerabilities to affect downside risks to output?
    - Does the effect of macroeconomic uncertainty spill over across borders to affect downside risks to economic activity in a country’s major financial and trading partners?
  - Extensions to GaR:
    - Augments the GaR model with measures of macroeconomic uncertainty—three types: (1) accuracy and dispersion of forecasts for key macroeconomic variables, (2) domestic policies, and (3) geopolitical tensions.
    - Implements the augmented GaR using machine learning tools in addition to standard panel quantile regressions to exploit advantages in prediction and improve forecasting of downside tail risks to future GDP growth.
- Downside risks to future GDP growth are "typically captured by the 5th or 10th percentile of the distribution."
- Financial conditions in GaR are proxied by a composite indicator of risky asset prices (equity and corporate bond returns, real house price growth, etc.) and measures of financial uncertainty (such as the VIX).
- Machine learning models are used because they "can accommodate many predictors and complex, nonlinear relations between variables."

### Technology, fintech, and AI considerations
- Recent innovations and social media can aggravate uncertainty and its effects on market tail risks by increasing investor and depositor attention to surprises and accelerating stress episodes.
- Fintech has made transactions "faster and easier," exacerbating funding and market liquidity risks.
- AI penetration into finance—particularly institutional investors’ use of AI-based algorithmic trading strategies—may "further rais[e] market volatility risks because of a potential increase in herding behavior among investors using similar AI models" (see Chapter 3).
- While these innovations have benefits, they create systemic complexities that can accelerate shock transmission and amplify the effect of macroeconomic uncertainty on financial stability.

*Source: Chapter 2, Introduction, Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability (textrevised - Introduction).*

### 1. Correlations between Selected Measures of Macroeconomic and

### 1. Correlations between Selected Measures of Macroeconomic and

### Correlations and macro‑market disconnect
- Panel evidence (1990–2024): correlations between macroeconomic and financial uncertainty measures vary over time and across episodes; a “macro-market disconnect” can arise when macroeconomic uncertainty is high while implied (or realized) market volatility is low.
- Examples noted: the 2016 US presidential election and Brexit—high macroeconomic uncertainty with limited short-term stock market volatility.
- Panel 3 (1997–2024): monthly ratios of the global economic policy uncertainty index to realized S&P 500 volatility, to the VIX, and to MSCI World Index volatility are normalized to have a mean of one over 1997:M1–2024:M2.
- Panel 4 (1990–2024): scatterplot of standardized country‑specific economic policy uncertainty vs MSCI country‑specific realized stock market volatility; the fourth quadrant (above average macroeconomic uncertainty and below average stock market volatility) identifies macro‑market disconnect observations.

*Sources: See Online Annex 2.1; and IMF staff calculations.*

### Macroeconomic uncertainty and downside risk to output
- Methodology:
  - An augmented GaR model (based on Adrian, Boyarchenko, and Giannone 2019) is estimated using panel quantile regressions to assess the full distribution of future GDP growth, focusing on the left tail (the 10th percentile) as downside tail risk.
  - Controls: current GDP growth, financial conditions, and country fixed effects. Robustness exercises address endogeneity concerns (see Online Annex 2.3).
- Key quantitative findings:
  - A one-standard-deviation increase in measures of macroeconomic uncertainty reduces one-quarter-ahead real GDP growth (annualized) by 0.5 to 2.0 percentage points (Figure 2.4, panel 1).
  - Measures based on real outcomes (REU and GDP forecast dispersion) have the largest quantitative effect.
  - The impact persists up to about seven quarters after the shock.
  - In cumulative terms, an increase in the REU equivalent to that observed on average across countries during the global financial crisis translates into a decline in one-year-ahead GaR of about 1.2 percentage points.
  - Context: the annual output decline in the bottom 10th percentile of the historical GDP growth distribution for the full sample across advanced and emerging market economies is 1.2 percent.
- Asymmetry:
  - Increased macroeconomic uncertainty affects downside (“bad”) tail risks more strongly than upside (“good”) tail risks.
  - Macroeconomic uncertainty has a negligible effect on the median of the future real GDP growth distribution but a large statistically significant effect on lower and upper quantiles (Figure 2.4, panel 3).
  - Examples of “good” uncertainty: 1990s US dot‑com bubble, mobile phone revolution in Finland, postcrisis reforms in Korea, German reunification. Examples of “bad” uncertainty: onset of the global financial crisis, the COVID‑19 pandemic.

### Machine learning (ML) enhancements to Growth‑at‑Risk (ML‑GaR)
- ML approaches applied: panel quantile random forest and panel quantile neural network incorporated into the GaR framework (ML‑GaR).
- Predictive gains:
  - ML‑GaR improves out-of-sample prediction accuracy relative to the standard benchmark GaR model by up to 7 percent at different horizons (Figure 2.5, panels 1 and 2, green bars).
  - Adding the real economic uncertainty index (REU) as a predictor further improves out-of-sample forecast performance of ML‑GaR models by 5   to 13 percent relative to standard GaR models that exclude uncertainty (Figure 2.5, panels 1 and 2, red bars).
- Variable importance:
  - ML‑GaR models show the REU contributes at least as much as the financial conditions index to predicting downside risk to real GDP growth.
  - For one- and four-quarter‑ahead forecasts across advanced and emerging market economies, the REU on average contributes more to predictions than the financial conditions index (Figure 2.5, panels 3 and 4).
  - Contributions measured as average absolute Shapley values; contributions may vary across countries and over time.

### Transmission channels and role of vulnerabilities
- Market and credit channels:
  - Increases in macroeconomic uncertainty are associated with a greater likelihood of large negative realizations of stock market returns and spikes in sovereign bond spreads.
  - Macroeconomic uncertainty influences tail risks to future bank lending, especially where banking exposure to sovereign debt is high.
- Interaction with macrofinancial vulnerabilities:
  - High real economic uncertainty combined with excessive domestic credit (credit‑to‑private‑sector‑to‑GDP gap measured as deviation from long‑term trend) reduces one-quarter‑ahead downside tail risk to GDP growth (10th percentile) by 0.6 percentage points (Figure 2.6, panel 1).
  - High public debt levels (deviation of public‑debt‑to‑GDP ratio from long‑term trend) significantly increase downside risks to GDP growth, particularly when real economic uncertainty is high.
- Intertemporal trade‑off of easing financial conditions:
  - A one-standard-deviation easing in financial conditions affects the term structure of GaR differently amid high vs low real economic uncertainty (Figure 2.6, panel 2) and amid high vs low macro‑market disconnect (Figure 2.6, panel 3).
  - High (low) macro‑market disconnect is defined as the ratio of real economic uncertainty to realized market volatility being above (below) the mean.

### Policy implications and macroprudential role
- Macroprudential policies can help mitigate the intertemporal trade‑off posed by easy financial conditions and maintain financial stability.
- Evidence:
  - The impact of a one-standard-deviation easing in financial conditions during periods of macroprudential tightening is materially different from periods without macroprudential tightening (Figure 2.6, panel 4).
  - The macroprudential tightening regime refers to quarters in the preceding year with net macroprudential tightening (methodology similar to Chapter 2 of the April 2021 Global Financial Stability Report).
- Overall recommendation from analysis:
  - Monitoring and addressing macroeconomic uncertainty is critical for downside risk assessment.
  - Incorporating measures of macroeconomic uncertainty (REU, GDP forecast dispersion, other indices) into forecasting and stress‑testing (including ML‑enhanced GaR models) improves early detection of downside risks.
  - Macroprudential policy can be an effective tool to offset the adverse intertemporal effects of loose financial conditions, especially when macro‑market disconnect is present.

*Sources: See Online Annex 2.1; and IMF staff calculations.*

### CHAPTER 2 MACROFINANCIAL STABILITY AMId hIGh GLOBAL ECONOMIC uNCERTAINTY

### CHAPTER 2 MACROFINANCIAL STABILITY AMId hIGh GLOBAL ECONOMIC uNCERTAINTY

### Macroeconomic Uncertainty and the Intertemporal Trade-off
- High macroeconomic uncertainty amplifies macrofinancial stability risks associated with loose financial conditions through the GaR framework.
- Short-term effect of easing financial conditions:
  - Rising asset valuations and compression of credit spreads and stock market volatility reduce downside tail risks to GDP growth.
- Medium-term effect of easing financial conditions:
  - Encourages buildup of debt vulnerabilities that exacerbate downside tail risks to GDP growth.
- A large macro-market disconnect (high real economic uncertainty relative to realized stock market volatility) increases the risk of sudden jumps in financial market volatility and market crashes when adverse shocks occur.
- Empirical findings:
  - Under high macroeconomic uncertainty, looser financial conditions exacerbate downside tail risks to GDP growth (GaR model augmented with the REU).
  - The impact of looser financial conditions is more pronounced when there is a large macro-market disconnect.

### Role and Effectiveness of Macroprudential Policy
- Macroprudential policies can mitigate the intertemporal trade-off:
  - A net tightening of macroprudential policies can offset the rise in medium-term downside risks associated with easy financial conditions.
  - When the macro-market disconnect is large, a loosening of financial conditions coupled with a net tightening of macroprudential policies is associated with a reduction in downside risks to GDP growth of 0.3 to 0.6 percentage points in the medium to long terms, compared with a scenario in which no macroprudential measures are put in place.
- Recommendations:
  - Policymakers may need to be more proactive in deploying policies aimed at preserving financial stability when macroeconomic uncertainty is high relative to market volatility.
  - Credible policy frameworks may help reduce macroeconomic uncertainty and its impact on downside risks to output.

### Cross-Border Spillovers of Macroeconomic Uncertainty
- Foreign macroeconomic uncertainty, constructed as trade- or financial-exposure-weighted averages of partners’ REU, raises domestic downside tail risks through real, market, and credit channels.
- Quantitative findings:
  - The 10th percentile of the one-quarter-ahead distribution of GDP growth declines by 1.7 percentage points following a one-standard-deviation increase in the trade-weighted foreign REU.
  - The effect from trade-weighted foreign uncertainty is less persistent than a similar increase in domestic macroeconomic uncertainty and peters out in about three quarters.
  - Similar spillovers arise when foreign REU is weighted by banking exposures or portfolio exposures.
  - For portfolio-weighted foreign uncertainty the effect is notably more persistent, suggesting nonbank financial intermediaries can play an important role in transmitting macroeconomic uncertainty across borders.
- Mitigants:
  - Building adequate international reserve buffers.
  - Greater exchange rate flexibility.
  - Prudential authorities should ensure banks and nonbanks assess vulnerabilities to cross-border spillovers.

### Asset Markets, Banking, and Machine Learning Insights
- Macrofinancial channels and asset-market effects:
  - A one-standard-deviation increase in the REU is, on average, associated with an increase of 150 basis points in upside tail risks to sovereign bond spreads (90th percentile) in emerging market economies at a six-month horizon.
  - For the average advanced economy, a similar REU shock increases upside tail risks by about 25 basis points.
  - The impact on sovereign bond spreads is more pronounced when fiscal vulnerabilities such as public debt service and banks’ exposure to public debt are high.
- Machine learning and systemic risk assessment:
  - Machine learning models can improve forecasting capacity of frameworks like GaR.
  - Practical challenges of machine learning: significant data and technological requirements, weak signal-to-noise ratios, and issues with transparency and interpretability.
  - The chapter addresses these shortcomings using cross-validation methods, overfit mitigation, numerical simulations, and variable importance analysis.

### Conclusion: Policy Priorities and Recommendations
- Core findings:
  - High macroeconomic uncertainty increases downside risks to future real GDP growth, stock and bond market returns, and bank lending.
  - Macrofinancial vulnerabilities (high public and private debt ratios) can interact with high uncertainty to amplify adverse effects on output.
  - High macroeconomic uncertainty worsens the intertemporal trade-off posed by easing financial conditions, especially when accompanied by low financial market volatility (macro-market disconnect).
  - Macroeconomic uncertainty spills over internationally via trade and financial linkages.
- Policy actions:
  - Reduce policy uncertainty by enhancing credibility of monetary and fiscal frameworks (for example, adoption of policy rules supported by strong institutions), improved transparency, and well-designed policy communication.
  - Maintain a stable financial regulatory framework with clear communication, robust calibration, phase-in periods, and practical supervisory discretion to avoid generating unnecessary policy uncertainty.
  - Deploy adequate macroprudential and fiscal policies to contain financial stability risks amid high uncertainty; remain vigilant and proactive, especially when financial conditions are loose.
  - Fiscal policies should prioritize debt sustainability to contain adverse effects of elevated public debt on borrowing costs.
  - Ensure adequate international reserves and exchange rate flexibility to cushion foreign uncertainty shocks.
  - Build adequate safety nets and buffers to mitigate risks from geopolitical uncertainty; pursue diplomacy and multilateral cooperation where possible.
  - Enhance systemic risk monitoring by explicitly considering macroeconomic uncertainty and exploiting AI tools (machine learning and natural language models) while addressing their conceptual and practical challenges.

*Source: CHAPTER 2 MACROFINANCIAL STABILITY AMId hIGh GLOBAL ECONOMIC uNCERTAINTY (textrevised - CHAPTER 2 MACROFINANCIAL STABILITY AMId hIGh GLOBAL ECONOMIC uNCERTAINTY, October 2024).*

### Chapter 2 of the April 2022 Global Financial Stability

### Chapter 2 — Market and Credit Channels, and Policy Credibility (April 2022 Global Financial Stability Report, Chapter 2)

### Market and Credit Channels: key findings
- Macroeconomic uncertainty raises upside tail risks of sovereign bond spreads and downside tail risks of stock market returns.
- A one-standard-deviation increase in the REU can raise downside tail risks to stock market returns (the 10th percentile of the distribution of stock market returns) by about 30 percentage points, one year after the shock in advanced and emerging market economies (Figure 2.1.1, panel 3).
- For sovereign spreads:
  - Panel estimates use monthly data for 20 advanced and 9 emerging market economies from 1990:M1 to 2023:M12.
  - Effects are reported at horizons h = 1, 3, 6, and 12 months for upside tail risks (the 90th percentile of changes to sovereign bond spreads).
  - The difference between low and high fiscal vulnerabilities (25th vs 75th percentiles) is not statistically significant for advanced economies in the analysis.
- Credit channel / bank lending:
  - A one-standard-deviation increase in the REU is associated with a decline of about 1 percentage point (annualized) in the 10th percentile of the distribution of one-quarter-ahead real credit growth (Figure 2.1.2, panel 1).
  - This effect persists through about seven quarters, although it becomes smaller over time.
  - Estimates are based on panel quantile regressions for a sample of 18 advanced and 13 emerging market economies using data from 2001 to 2023. The model controls for lagged credit growth dynamics, output growth, financial conditions, and financial vulnerabilities (position in the credit cycle; banking sector fundamentals such as capital adequacy, asset quality, profitability, and exposure to sovereign risk), and includes country and time fixed effects.
- Bank-level text-based uncertainty measure:
  - Constructed by calculating the percentage of sentences including words related to uncertainty in earnings call transcripts of specific banks, using definitions from the February 2024 update of the Loughran-McDonald Master Dictionary; these percentages are averaged across banks in a country to form a country-level indicator.
  - The bank-level text-based measure generally exhibits a low degree of correlation with other measures of uncertainty.
  - Although the bank-level text-based measure has a somewhat smaller impact on future tail risks to bank lending than the REU, it remains statistically significant when all other measures of uncertainty are included in the regression.
- Amplification by financial vulnerabilities:
  - Existing financial vulnerabilities amplify the effects of higher uncertainty on downside risks to bank lending.
  - Extending the model to include interaction terms shows countries with higher bank exposure to sovereign risk exhibit a greater likelihood of a sharp decline in future bank loan growth when macroeconomic uncertainty rises.
  - Example: a one-standard-deviation increase in the REU is associated with an increase in the one-year-ahead downside risk to lending (10th percentile of the real credit growth distribution) of about 1 percentage point when domestic banks’ exposure to sovereign risk is high (one standard deviation above the mean) compared to at the mean level.
- Methodological notes:
  - Panel quantile regressions used for market and credit analyses include country fixed effects and relevant controls; confidence intervals reported (90 percent for bank-lending panels; 95 percent for some sovereign/stock panels).
  - Panel for stock market returns uses monthly data for 21 advanced economies and 19 emerging markets from 1990:M1 to 2023:M12.

### Monetary policy frameworks and uncertainty
- Theory and evidence:
  - Early proponents argued rules reduce policy uncertainty and associated inefficiencies; enhanced monetary policy credibility can help stabilize inflation expectations.
  - Empirical evidence cited indicates that enhanced monetary policy credibility can anchor inflation expectations to target levels and that policy rules can reduce uncertainty.
- Empirical findings in this chapter:
  - Countries where inflation expectations deviate more from the policy (inflation) targets experience higher levels of economic policy uncertainty (Figure 2.2.1, panel 1).
  - Increased macroeconomic uncertainty (real or policy related) has a larger effect on downside risk to one-quarter-ahead GDP growth when policy targets were missed by wider margins over the preceding three years (weaker monetary policy frameworks) (Figure 2.2.1, panel 2).
  - Panel 1: high/low deviation regimes are defined by the sample median of the deviation of inflation expectations from the policy inflation target (absolute value summed over preceding three years); “high deviation” proxies less effective monetary policy frameworks.
  - Panel 2: shows the effect of a one-standard-deviation increase in measures of real and economic policy uncertainty on the 10th percentile of the distribution of one-quarter-ahead real GDP growth without (baseline) and with (interaction term) sound monetary policy frameworks; whiskers indicate 90 percent confidence intervals (significance).
- Conclusion:
  - Credible monetary policy frameworks tend to help reduce economic policy uncertainty and its impact on downside risks to output.

### Fiscal policy frameworks and uncertainty
- Fiscal rules and uncertainty:
  - Fiscal rules can reduce fiscal policy uncertainty, fiscal procyclicality, and market volatility, and enhance fiscal sustainability.
  - Studies cited document that more stringent fiscal rules can reduce overall macroeconomic volatility (and hence real economic uncertainty).
- Additional points:
  - Discretionary fiscal policy is prone to deficit bias; fiscal rules can act as a commitment device to limit discretionary deficits.
  - Evidence also indicates increased uncertainty can impair the effectiveness of fiscal policy, implying policy responses could become more uncertain and magnify macroeconomic uncertainty.

### Figures, panels, and indices referenced
- REU = real economic uncertainty index.
- Key panels and samples:
  - Figure 2.1.1 Panel 1: monthly data for 20 advanced and 9 emerging market economies from 1990:M1 to 2023:M12; shows effects of a one-standard-deviation increase in REU on upside tail risks to sovereign spreads at horizons 1, 3, 6, and 12 months.
  - Figure 2.1.1 Panel 3: monthly data for 21 advanced economies and 19 emerging markets from 1990:M1 to 2023:M12; shows effects of a one-standard-deviation increase in REU on the 10th percentile of stock market returns at horizons t+1, 3, 6, and 12 months.
  - Figure 2.1.2 Panels 1–3: panel quantile regressions using country-level data for both advanced and emerging market economies; effects reported as average quarterly rate (annualized) for horizons up to 12 quarters; confidence intervals shown (90 percent).
  - Figure 2.2.1 Panels 1–2: show average economic policy uncertainty by soundness of monetary policy framework and effect of uncertainty on GaR conditional on soundness; GaR = growth-at-risk.

*_Italicized source attribution line_*: International Monetary Fund, Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability — Chapter 2 (April 2022).

### CHAPTER 2 MACROFINANCIAL STABILITY AMId hIGh GLOBAL ECONOMIC uNCERTAINTY

### CHAPTER 2 MACROFINANCIAL STABILITY AMId hIGh GLOBAL ECONOMIC uNCERTAINTY

### Major findings on AI and capital markets
- Generative artificial intelligence and related breakthroughs have the potential to dramatically increase the efficiency of capital markets—trading, investment, and asset allocation—through artificial intelligence–assisted process automation and analysis of complex unstructured data; evidence suggests these effects are already beginning to be felt.
- Financial institutions have been actively using machine learning and other AI-related computation methods for approximately 20 years; GenAI and related applications are in their infancy but adoption signals indicate rapid gearing-up.
- Recent surveys report a vast majority of respondents expect a significant expansion of the use of GenAI-driven models (IIF and Ernst & Young 2023), and more than half of investment managers said that they planned to use GenAI in the future (Mercer Investments 2024).
- AI-powered ETFs—where AI is used to construct and adjust an ETF’s portfolio—still account for a very small share of the market, with less than $1 billion in assets under management.
- The unit cost of training AI models has dropped dramatically, but “notable” models have simultaneously become much more complex, leading to much higher overall costs; high fixed costs may exacerbate market concentration (notable models defined as models in the running for the top 10 largest training compute, expressed in terms of required floating-point operations (FLOP) (Epoch AI 2024)).
- Development of foundation models has predominantly been based in the United States.

### Potential benefits identified
- Efficiency improvements and higher productivity (Boukherouaa and others 2021).
- Refined portfolio investing frameworks (Park and others 2023).
- Improved return forecasting (Chen, Kelly, and Xiu 2023).
- Quantification of crash risks (Swinkels and Hoogteijling 2022).
- SupTech and RegTech applications for supervisors and regulated institutions.
- GenAI increases efficiency across tasks: helping analysts write code, improving customer-facing activities, generating new investment ideas, and improving the predictive power of textual analysis when used as inputs into analytical models.
- Potential to lower barriers to entry for quantitative investors into less liquid asset classes that require extensive analysis of legal documents.

### Four broad categories of potential risks
- Increased market speed and volatility under stress, especially if AI trading strategies become highly correlated or respond similarly to shocks.
- Opacity and monitoring challenges as extreme behaviour of AI systems becomes increasingly difficult to anticipate and as AI activities migrate to nonbank financial intermediaries (NBFIs).
- Increased operational risks due to reliance on a few key third-party AI service providers.
- Increased cyber and market manipulation risks, particularly through fraud and social media disinformation.

### Market-structure and transmission concerns
- AI could cause large changes in market structure through greater and more powerful use of algorithmic trading, novel trading and investment strategies, increased turnover, higher asset correlations, and faster incorporation of information into prices.
- Migration of market-making and investment activities to hedge funds, proprietary trading firms, and other NBFIs could create uncertainty about interaction effects of AI models used by different investors and traders.
- High fixed costs of infrastructure and talent and potential data monopolies could concentrate advantages among a few players with superior nonpublic data and large volumes of trading and client data.
- For emerging market and developing economies, GenAI is widely viewed as a tool to enable technological leapfrogging and increase financial development and inclusion, but differential speeds of adoption could create fragmentation risks.

### Current adoption patterns and outlook
- Mainstream GenAI use dates back only a few years, while ML methods have been integrated into investment processes for approximately 20 years.
- GenAI adoption so far is largely “evolutionary”—extensions of existing analytical methods and investment strategies—rather than “revolutionary” autonomous strategies.
- Most market participants surveyed by IMF outreach are uncomfortable with fully autonomous AI executing trades without human oversight and favor having a “human in the loop.”
- GenAI is likely to increase the speed of market reactions to new information through real-time processing of unstructured data such as textual central bank announcements.

### Analytical and data observations
- Labor market data, patent filings, and investor outreach suggest institutions are rapidly preparing for significant integration of AI technologies.
- The chapter draws on IMF staff market outreach and analytical work leveraging novel data sources to compensate for limited direct data on AI adoption in capital markets.

### Policy recommendations and suggested supervisory responses
- Undertake the calibration of circuit breakers and a review of margining practices in light of potentially rapid AI-driven price moves.
- Enhance monitoring and data collection of the activity of large traders, including nonbank financial intermediaries.
- Address dependency on data, models, and technological infrastructure by requesting a risk mapping from regulated entities (that is, data on the internal and external interconnections and interdependencies that are necessary to deliver the institutions’ critical services).
- Adopt a coordinated approach for the definition of critical AI third-party service providers and continue to strive for resilience in capital markets by enhancing cyberattack protocols.
- Adopt measures that ensure continued market integrity, efficiency, and resilience of over-the-counter markets when AI use proliferates.

*Source: GLOBAL FINANCIAL STABILITY REPORT: STEAdYING ThE COuRSE: uNCERTAINTY, ARTIFICIAL INTELLIGENCE, ANd FINANCIAL STABILITY — Chapter material (October 2024) contained in the supplied PDF text.*

### 2. National Affiliation of Foundation Models and Access Type

### 2. National Affiliation of Foundation Models and Access Type

### Overview of figures and training-cost context
- Trailing six-month count of published models and trailing six-month percentage of open-source models (right scale) are depicted with axis markers: 0, 100, 20, 40, 60, 80 and timepoints/labels: 2011 13 15 17 19 21 23 Feb. 2021 Nov. 2021 Aug. 2022 May 2023 Feb. 2024.
- Training compute is measured using “oating-point operations.
- Estimated costs for panel 1 were derived from Epoch AI’s data sets and the methodology to estimate costs of training notable arti•cial intelligence models and graphics processing unit price-performance data.
- Exa denotes a factor of 10 18.
- The blue and red lines in panel 1 are best-•t lines.
- Sources cited for the model counts and affiliations: Epoch AI; Stanford University’s Ecosystem Graphs; and IMF staff calculations.

### Cost of compute and training notable AI models
- Cost measures shown in inflation-adjusted 2023 US dollars per exa floating point operation and in millions of inflation-adjusted 2023 US dollars (right scale).
- Panel framing: "1. Cost of a Single Precision Processor and the Cost of Training Notable AI Models."

### Key findings on AI adoption in capital markets
- Penetration of advanced AI applications in investment management is described as "relatively modest."
- IMF staff conducted qualitative outreach to major industry players directly involved in AI-related strategy to assess adoption and transformation in capital markets.
- Participants highlighted accelerating pace of AI adoption driven mainly by proliferation of GenAI tools.
- Prospects for creating value through AI appear most promising in publicly traded liquid asset classes; equities, government bonds, and listed derivatives cited as high-potential because of real-time data and transparency.
- IMF outreach results:
  - Equities and derivatives are most likely areas for AI adoption in the investment process, followed by fixed income and foreign exchange.
  - Some participants expect AI advances to benefit less-liquid markets such as private credit and some emerging market segments.
- Use cases reported by market participants:
  - Incorporation of alternative data sets and development of forward-looking indicators.
  - Market analysis including sentiment analysis of social media and regulatory filings.
  - Buy-side uses: productivity enhancement, exploration of new asset classes, extraction of signals to support investment decisions, portfolio optimization and allocation, back-office automation.
  - Sell-side uses: risk assessment, pricing and forecasting, customer service improvements, and trading automation.
  - Market infrastructure and academia: democratization of code writing and prototyping, information extraction and summarization.

### Adoption patterns, timelines, and autonomy expectations
- Participants expect greater integration of sophisticated AI in investment and trading decisions within a three- to five-year horizon.
- One prevalent near-term pattern: AI generates signals that feed existing analytical systems where human traders typically make final trading decisions; fully autonomous AI trading agents are not yet widely implemented.
- Consensus expectation: a "human in the loop" approach will persist in the near term (three to five years), especially for large capital allocation decisions; complete autonomy is not anticipated soon.
- Some participants anticipate future developments such as agent-to-agent trading and complete AI-driven trading workflows.

### Evidence from specific market segments and instruments
- Robo-advisors:
  - Figure 3.3 panel 1 shows rapid growth in robo-advisor assets under management (AUM) with forecasts for 2025, 2026, and 2027 shown as light blue bars.
  - Axis and scale references: Billions of US dollars; share of US equity market cap in percent (right scale). Years shown: 2017 18 19 20 21 22 23 24 :Q3 25F 26F 27F.
- AI-driven ETFs:
  - Figure 3.3 panel 2 indicates AI-driven ETF assets under management have grown but remain tiny relative to market size.
  - Numeric small values shown for AI-driven ETF AUM: 0.003, 0.001, 0.001, 0.002, 0.002 for years 2018–24 (chart sequence).
  - Chart axes include 0, 1.0, 0.2, 0.4, 0.6, 0.8 and AUM in Billions of US dollars; share of US equity market cap in percent (right scale).
- Asset-class opportunity breakdown (Figure 3.4):
  - Allocation shares listed: Publicly traded assets vs privately traded assets and a breakdown that includes the following percentages in a chart: 57%, 14%, 11%, 18%.
  - Liquidity and AI value creation opportunities show a strong correlation between market liquidity and current AI adoption.
- Algorithmic trading and autonomy:
  - Evidence is mixed on the use of sophisticated AI for autonomous trading; survey data in one energy market suggests more autonomous algorithms may still rely on simpler methods.
  - Dutch Authority for the Financial Markets finding: on the Euronext exchange, “trading firms tell the Authority for the Financial Markets that machine learning is implicitly or explicitly used in 80 to 100 percent of their trading algorithms.” The Authority notes "explicit" use cases may include applications that are not autonomous, such as signal generators.

### Innovation signals: patents and labor markets
- Patent filings:
  - Filings referencing AI/ML in the context of high-frequency or algorithmic trading have increased (Figure 3.6, panel 1; time series spanning 2009–23 shown).
  - Filings also show a surge in patents related to asset allocation or portfolio management (Figure 3.6, panel 2; time series spanning 2009–23 shown).
  - Recent filings focus on improving operational efficiency of brokerage/trading platforms and developing systems for low-latency, high-throughput trading signals; asset management filings emphasize ML techniques for cash flow and liquidity management, automated asset-class rebalancing, improved valuation and forecasting, tailored capital requirements, interpretation of unstructured data, and novel access to alternative asset classes (emissions trading, digital assets, cryptographically signed transactions).
- Labor market and skills:
  - Only a small share of workers currently claim AI skills, but the talent pool within financial services appears to be growing.
  - ML, natural language processing, and deep learning are among the top 30 competencies listed in quantitative researcher and analyst profiles in the US financial industry.
  - Front office job-posting statistics: job postings containing AI terminology for front office roles rose from a monthly average of 4.5 percent of front office roles in 2019 to 4.9 percent in 2023 and peaked at 6.6 percent in 2022.
  - Competition to attract AI talent is cited as a major challenge that could limit acceleration of AI developments.
  - AI talent concentration within the US financial services industry exceeds that of the broader economy.

### Implications for emerging markets and synthetic data
- Potential benefits for emerging markets and developing economies:
  - Improved access to financial services, credit scoring, loan origination, robo-advising, and portfolio construction.
  - GenAI-enabled parsing of fragmented and unstructured data could reduce investment barriers and improve liquidity of some emerging market assets, potentially attracting larger capital flows.
- Use of synthetic (AI-generated) data:
  - Synthetic data may help train investment models where real data are scarce.
  - Two caveats when relying on synthetic data:
    - Unintended over- or under-representation of certain values in real-world data distributions, undermining extreme-event performance of AI systems.
    - Potential biases perpetuated by synthetic data when generation processes fail to account for specificities and requirements of second-order applications.
- Market fragmentation risk between advanced economies and emerging markets is seen as limited by some participants; AI could support greater financial inclusion in many cases, though automation risks for lower-skilled jobs were noted.

### Market-structure implications
- AI adoption may alter market structure and dynamics by amplifying existing trends and potentially producing revolutionary changes in certain areas.
- Expected near- to medium-term effects include improved efficiency and productivity, cost savings in designing trading algorithms, better processing of unstructured data, and more compressed bid-ask spreads.
- Supervisory uses:
  - Financial supervisors are beginning to use AI-driven SupTech tools to monitor markets and institutions, detect anomalies in large data sets, and assist with regulatory compliance checks.
  - Banks use RegTech tools to enhance anti-money laundering/know your customer processes, automate tasks to improve accuracy in client data, monitor transactions, and detect fraud.

*Italic: Source — IMF staff analysis and outreach, Chapter 3, "Advances in Artificial Intelligence: Implications for Capital Market Activities," Global Financial Stability Report, October 2024.*

### 1. Concentration of LinkedIn Members in the Financial Services Industry

### 1. Concentration of LinkedIn Members in the Financial Services Industry with AI Skills

### AI Talent and Job Postings in Financial Services
- Average monthly percent of total job postings containing AI skills (figures shown for US and US financial services; exact plotted values not all provided in text).
- AI talent concentration measured as percent of total members (average percent of aggregated country profiles; exact plotted values not all provided in text).
- Observation: demand for these skills, particularly in the financial services sector, has increased in recent years and appears to outpace job postings of the broader US economy.
- Sources: Indeed Hiring Lab; LinkedIn Economic Graph; and IMF staff calculations.
- Note: Financial services include entities that make financial transactions (creation, liquidation, or change in ownership of financial assets) and/or that facilitate financial transactions across 30 advanced economies and 14 emerging markets.

### Larger Role for Nonbank Financial Institutions (NBFIs) and Algorithmic Trading
- NBFIs now hold over half of all financial market assets globally.
- In the United States:
  - Algorithmic trading constitutes about 70 percent of equities trading.
  - Algorithmic trading constitutes more than half of futures trading.
- Other jurisdictions lag in share of algorithmic equities trading but could catch up briskly.
- Increasing returns to scale have resulted in markets with relatively high algorithmic trading activity tending to see concentration among a limited number of players.
- High fixed costs for internal development or deployment of sophisticated AI favor larger trading firms.
- Smaller players may resort to third-party cloud and AI software service providers, amplifying outsourcing, market concentration, and vendor lock-in risks.

### Empirical and Literature Findings on Algorithmic Trading Effects
- Algorithmic trading is largely assessed to have a positive impact on market liquidity and efficiency, but can increase short-term volatility:
  - Research references: Hendershott, Jones, and Menkveld 2011; Hendershott and Riordan 2012; Boehmer, Fong, and Wu 2021.
- Algorithmic trading can increase volatility following macroeconomic news and disincentivize informed traders, potentially harming market efficiency (Scholtus, van Dijk, and Frijns 2014; Yadav 2015).
- In the US Treasury market, digitalization improved liquidity on aggregate but may have increased the occurrence of rare extreme bouts of illiquidity (Bouveret and others 2015).
- Market liquidity is affected by the presence of high-frequency traders (Adrian, Fleming, and Vogt 2017).
- Decomposition of high-frequency US stock returns shows idiosyncratic jumps in individual stock returns are less frequent over time; idiosyncratic jumps are more frequent when liquidity conditions are poor.
- High-frequency traders often use order cancellations, but order cancellation rates drop significantly as implied volatility increases; hidden order rates increase under stressed liquidity conditions.
- Algorithmic trading strategies often include algorithmic risk limits (position, price, volume, and other limits), which can be triggered under stress and contribute to cascading liquidity evaporation.

### Concentration and Dependence on Third-Party Providers
- IT infrastructure remains strongly concentrated; the AI software services market is becoming more concentrated.
- Cloud service disruptions (hours per year) show longer outage time for IT infrastructure in 2023 than in 2022:
  - 2023 (205.3 hours)
  - 2022 (133.5 hours)
- Note: Infrastructure represented by aggregated revenue for Infrastructure-as-a-Service; AI software represented by AI and predictive analytics Platforms-as-a-Service. Disruption events include critical events as defined by Parametrix.

### How AI Could Change Algorithmic Trading and Market Dynamics
- GenAI could lower barriers to entry for algorithmic trading by facilitating coding, testing, and automation in less technologically sophisticated trading venues.
- GenAI can aid processing of complex text-based data (for example, bond indentures) to enable more standardized risk analysis and pricing tools to support liquidity. Examples cited: Overbond; BondGPT.
- Potential new dynamics:
  - AI-driven strategies could drive higher and more procyclical trading volumes.
    - Evidence: AI-powered ETFs have experienced significantly higher turnover than other active or passive ETFs.
  - Markets could react faster to textual news; intraday market data suggest that after the introduction of large language models, the initial market reaction following the release of the Federal Open Market Committee minutes (up to 45 seconds) tends to reflect its eventual impact more accurately than before.
  - AI algorithms could collude or manipulate markets; theoretical models are investigating potential interactions and manipulation risks.

### Empirical Signals on AI-driven ETF Turnover and Market Reaction
- AI-powered ETFs have experienced significantly higher turnover than other active or passive ETFs; three sample AI-driven ETFs increased portfolio turnover during the March 2020 market turmoil, providing evidence for procyclicality.
- Market reaction to Federal Open Market Committee minutes:
  - Comparison before and after LLMs indicates faster and more accurate initial market adjustment (proportion of price change at the 15-minute point; both subsamples contain 15 qualifying datapoints).
- Simulation evidence (Fan, Pelger, and Yu forthcoming) shows market inefficiency (price gap between market price and fundamental value) can increase with underinformed AI traders; panel displays price gap as number of underinformed agents increases.

### Table of Potential Positive and Negative Scenarios (as in Table 3.1)
- Market liquidity
  - Negative Scenario: AI magnifies existing risks related to algorithmic trading by facilitating growth and democratizing algorithmic trading activity across assets and geographies, exacerbating risks of sudden liquidity withdrawal under stressed conditions.
  - Positive Scenario: AI increases stability of algorithmic trading under stressed conditions; AI-driven algorithms could operate in a wider set of market conditions with lower flash-crash risk and reduced liquidity-withdrawal under stress.
- Leverage
  - Negative Scenario: AI-driven strategies boost short-term leverage as arbitrage opportunities are exploited more efficiently, leaving remaining opportunities requiring higher leverage to deliver similar returns.
  - Positive Scenario: AI improves management of leverage and related risks by facilitating more frequent and automated management of leveraged positions based on more inputs, mitigating operational lags.
- Interconnectedness
  - Negative Scenario: AI increases interconnectedness by proliferating algorithmic trading across asset classes, geographic regions, and trading venues, leading to higher correlations and facilitating spillovers and transmission of stress.
  - Positive Scenario: Market access, efficiency, and liquidity improve for some market segments, including emerging markets.

*Source: IMF staff, Chapter 3, "Advances in Artificial Intelligence: Implications for Capital Market Activities," Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability (October 2024).*

### CHAPTER 3 AdvANCES IN ARTIFICIAL INTELLIGENCE: IMPLICATIONS FOR CAPITAL MARkET ACTIvITIES

### CHAPTER 3 AdvANCES IN ARTIFICIAL INTELLIGENCE: IMPLICATIONS FOR CAPITAL MARkET ACTIvITIES

### AI-driven market dynamics and model interactions
- Simulated markets with informed and uninformed reinforcement learning agents can produce a variety of outcomes, including tacit algorithmic collusion and “winner takes all” scenarios.
- In a simulation from Fan, Pelger, and Yu (forthcoming):
  - One informed reinforcement learning agent holds one-eighth of the total market buying power.
  - Remaining buying power is evenly split among varying numbers of uninformed reinforcement learning agents.
  - When the number of uninformed agents increases, it becomes harder for the informed agent to manipulate the price, moving the equilibrium price closer to the fundamental value.
  - The scenario with two uninformed agents most likely generates self-perpetuating trends that take the form of local price bubbles.
- Manipulation is more likely if one algorithm has an information or latency advantage and when the market has fewer players (Figure 3.11, panel 4).

### Financial stability implications identified by market participants
- Primary concerns from IMF outreach participants (market infrastructure providers, asset managers, academia) include:
  - Herding and market concentration.
  - Vendor concentration leading to overdependence on a limited number of AI model providers and data vendors.
  - Market fragility and manipulation (including deepfakes and misinformation).
  - Cyber risks and large-scale data poisoning.
  - Model explainability and model hallucination.
  - High costs of fine-tuning sophisticated models creating an unlevel playing field favoring large firms.
  - Job replacement, unauthorized data use, reputational risks, and potential EMDE fragmentation.
- Specific fragility pathways:
  - Fast-paced AI decision making may lead to drying up of market liquidity, excess volatility, and flash crashes if guardrails are ineffective due to poor design, system complexity, or malicious intervention.
  - Lack of model explainability and hallucinations could erode trust in markets.
  - Acquisition competition for data scientists and AI professionals could concentrate talent and capabilities.
  - Wider AI adoption could exacerbate spillovers from advanced economies to emerging market and developing economies, particularly if AI models are more sensitive to price fluctuations and managed against a basket of various asset classes.

### Summarized financial stability challenges (current and prospective)
- Increased market speed and volatility under stress:
  - AI-enhanced algorithmic trading can enhance liquidity and lead to more prompt price adjustments and thinner margins, but may incentivize increased leverage and amplify falling asset prices, volatility, and deleveraging in stress periods.
  - AI models may herd and produce similar decisions, creating procyclical risks and potential self-fulfilling fire-sales during adverse shocks.
  - Novel adverse events (for example, the COVID-19 pandemic in 2020) may drive outcomes difficult to comprehend, with models potentially shutting down and requiring humans to process voluminous trades.
- More opacity and monitoring challenges:
  - Continued migration of trading and investment activity to NBFIs may raise systemic opacity because NBFIs often face lighter explainability and transparency requirements than banks.
  - AI-generated portfolios spanning asset classes, regions, and venues could create correlations and interconnectedness that undermine holistic regulatory monitoring.
  - Emergent risks may arise from complex interactions between autonomous AI agents not visible at the institutional or regulatory level.
- Increased operational risks from dependence on few third-party AI providers:
  - AI models and IT services currently reside with a handful of key providers with dominant computational power and large language models; failure of these providers may lead to market stress akin to the failure of key financial market utilities.
- Increased market manipulation and cyber risks:
  - Fraud, disinformation, and deepfakes will likely become more sophisticated and could be used to manipulate financial markets and asset prices.
  - Compromised data integrity and confidentiality could lead AI models to produce suboptimal trading and investment decisions.

### Regulatory and supervisory expectations and developments
- Market participants expect regulatory authorities to:
  - Provide clarity and guidance on model risk management.
  - Emphasize stress testing for extreme scenarios.
  - Provide transparency and clearer disclosures.
  - Offer guidance on industry-specific regulatory structures and accountability frameworks.
  - Issue guidelines on AI use in consumer-facing applications.
  - Focus on balanced regulation that ensures responsible AI use without stifling innovation while protecting consumers.
  - Prioritize guidelines and best practices over strict rulemaking given the rapidly evolving nature of AI in financial markets.
  - Address bias in AI models and enhance supervisor preparedness through continuous upskilling and integration of AI/ML into supervision and market surveillance.
- International and national initiatives:
  - Existing regulatory and supervisory frameworks for capital markets are largely technology-neutral and applicable to AI systems, with ongoing work to adapt prudential frameworks to AI-specific risks.
  - Standard-setters and authorities (FSB, BIS, BCBS, NIST, IOSCO referenced) are issuing or revisiting guidance on:
    - Financial stability, market integrity, investor protection.
    - Managing third-party risk and cyber incidents.
    - Data governance, operational and cyber risk management.
  - Recommended supervisory actions include updating skills and tools to monitor complex strategies and processing more granular data in real time, and proactively assessing whether extant frameworks adapt to novel AI forms.

### Key areas of regulatory focus (as presented)
- Governance of AI use
- Outsourcing
- Data quality and bias
- Reporting requirements
- Need for a human-in-the-loop
- Transparency and explainability

*Source: CHAPTER 3 AdvANCES IN ARTIFICIAL INTELLIGENCE: IMPLICATIONS FOR CAPITAL MARkET ACTIvITIES, Global Financial Stability Report, October 2024*

### CHAPTER 3 AdvANCES IN ARTIFICIAL INTELLIGENCE: IMPLICATIONS FOR CAPITAL MARkET ACTIvITIES

### CHAPTER 3 AdvANCES IN ARTIFICIAL INTELLIGENCE: IMPLICATIONS FOR CAPITAL MARkET ACTIvITIES

### Regulatory and supervisory landscape
- IOSCO is conducting a two-year project to assess risks and challenges associated with the use of AI, with potential policy guidance expected by the first quarter of 2025 (IOSCO 2024).
- An IMF review of actions taken by 26 authorities in large capital markets finds:
  - Governments have begun to formulate comprehensive AI strategies and act in areas of data protection, governance, and cybercrime.
  - Some jurisdictions are considering dedicated AI legislation to ensure robust governance.
  - Supervisory authorities have focused primarily on clarification and outreach rather than on enforcement.
- Adoption of SupTech tools (periodic measures of supervisory technology adoption):
  - In 2023, 79 percent of advanced economies and 54 percent of emerging market and developing economies had adopted SupTech tools.
  - In 2022, adoption was 50 percent for advanced economies and 31 percent for emerging market and developing economies.

### Best practices and supervisory use cases
- Engagement and outreach:
  - Establish public/private forums to develop overarching principles (Office of the Superintendent of Financial Institutions 2023).
  - Partner with industry to build a risk framework (Monetary Authority of Singapore 2024).
  - Conduct surveys on applicability of existing frameworks (US National Archives 2021; Institute for Workplace Equality 2022; Bank of England 2024).
- Specific supervisory measures:
  - Request notification by banks prior to adoption of certain technologies or arrangements with third parties (BCBS 2024).
  - Periodic upskilling and upgrading to identify AI-specific issues such as models designed to “game the regulation” and detect algorithmic coordination.
  - Use of SupTech and GenAI to automate data quality checks, combine multiple data sources without unique identifiers, detect anomalies in trading patterns (changes in prices, volume, volatility), identify misleading information, and perform real-time monitoring of market transactions.
- SupTech adoption is uneven between advanced economies and emerging market and developing economies (Cambridge SupTech Lab 2023).

### Policy recommendations (high level)
- Regulatory and supervisory frameworks should follow a balanced approach, allowing participants to reap benefits of AI while acknowledging risks (IMF and World Bank 2018).
- Across sectors, supervisors should:
  - Continue to strive for cyber resilience.
  - Address dependency on data, models, and third-party service providers by requesting risk mapping.
  - Strengthen regulation and oversight of nonbank financial institutions (NBFIs) to require identification and disclosure of AI-relevant information.
- Specific capital markets actions:
  - Strengthen monitoring and resilience in over-the-counter markets and existing measures to address volatility.
  - Implement regulatory reporting, outreach, or survey approaches to support continued structural assessment of AI developments and risks.
- Coordinated approach to regulation and supervision of AI service providers:
  - Map relationships between critical AI service providers and essential IT infrastructure providers.
  - Ensure comparable and interoperable regulatory approaches for critical service providers (FSB 2023b).
  - Ensure the definition of critical service providers captures systemic use of common AI models (Bank of England 2024).

### Addressing increased market speed and volatility under stress
- Authorities and trading venues should determine whether to design new or modify existing volatility response mechanisms for AI-driven trading crash events.
- Circuit breaker considerations:
  - Existing circuit breakers may need re-parameterization in light of changing market structures.
  - Poorly designed circuit breakers may exacerbate volatility and interfere with market efficiency and price discovery (Vereckey 2023).
  - Testing algorithms in controlled environments is recommended to assess behavior in extreme circumstances.
- Margining and buffers:
  - Financial sector authorities, trading venues, and central counterparties should review margining requirements and other buffers in light of potentially rapid AI-driven price moves.
  - International work is needed to:
    1. Foster market participants’ preparedness for large variation margin calls during market stress.
    2. Identify good practices for variation margin collection and distribution by the central counterparty.
    3. Understand central counterparty margin models’ responsiveness to volatility and market stresses.
    4. Review initial margin levels in non-stress times and the effectiveness of tools to reduce procyclicality of margin models (BCBS, CPMI, and IOSCO 2022).

### Addressing increased opacity and monitoring challenges
- Financial sector authorities should require regular mapping of interdependencies between data, models, and technological infrastructure supporting AI models.
  - Models may feed on shared or interdependent data sources, share common architectures, and rely on a small number of providers for software, data, and cloud services.
  - Data sets may not cover a complete financial cycle, undermining model reliability (Gensler and Bailey 2020).
- Regulatory frameworks require assessing cumulative effects of models, but do not mandate joint assessment of data dependencies; an updated view of interdependencies will enable proactive risk management and a more resilient ecosystem.
- NBFI oversight:
  - Require NBFIs to identify themselves and disclose AI-relevant information.
  - Monitor large traders by unique identification and reporting to registered broker-dealers to enable supervisory monitoring (IOSCO 2011).
  - Enhance risk management and strengthen liquidity buffers to mitigate asset mispricing or liquidity runs (see Chapter 2 of the April 2023 Global Financial Stability Report; FSB, n.d.).

### Addressing operational concentration risk from third-party AI service providers
- Map relationships between critical AI service providers and essential IT infrastructure providers; failure or disruption of critical third parties may affect financial stability and confidence (Federal Reserve System, FDIC, and Department of the Treasury 2023; European Securities and Markets Authority 2023).
- Comparable and interoperable regulatory approaches facilitate compliance and coordination across the financial sector.
- Ensure protocols to avoid, protect against, respond to, and recover from attacks across AI system lifecycle stages: design, development/procurement, deployment, and operations (National Cyber Security Centre 2023).

### Over-the-counter market monitoring and resilience needs
- Authorities should be prepared to adopt measures to ensure continued market integrity, efficiency, and resilience of over-the-counter markets as AI use proliferates:
  - Collect and disseminate more detailed information on over-the-counter transactions.
  - Require market participants to account for liquidity shifts in their risk management frameworks.
  - Establish or expand incentives for market-makers to enhance liquidity.
  - Improve incentives for central clearing and establish margin requirements for non-centrally cleared derivatives.
  - Backstop measures in event of shocks could include central bank liquidity provision to market-making banks or indirect support for non-bank dealers by easing market funding conditions (CGFS 2014).

### IMF outreach findings and market intelligence
- IMF staff conducted extensive outreach across stakeholders to gather market intelligence on AI use in capital market activities:
  - Engaged with bank/dealers, AI vendors, asset managers, academia, rating agencies, and market infrastructure firms.
  - Received detailed responses from 27 stakeholders directly involved in AI topics and business.
  - Conducted regulatory outreach with 10 capital market supervisors of advanced and emerging markets.
- Definitions adopted by IMF staff for outreach:
  - “AI/ML models”: well-established predictive analytics, including neural networks, clustering algorithms, natural language processing, decision trees, and so on.
  - “Sophisticated AI models”: latest innovations such as deep learning, reinforcement learning, and large language models, including generative AI capable of generating text, codes, images, and other content.
- Participants in IMF market intelligence outreach (percent):
  - Asset managers: 20
  - Bank/dealers: 28
  - Market infrastructure: 12
  - Academia: 12
  - Others: 28
  - Note: “Others” includes nonprofit financial organizations, artificial intelligence finance conferences, artificial intelligence vendors, and rating agencies. This figure does not include regulatory outreach.

### Cybersecurity, manipulation, and AI incident observations
- Cyberattacks have increased, with the finance and insurance sector share rising as well (2014–2023 trends shown).
- IMF prior work: one measure of potential maximum annual financial firm losses from cyber incidents increased from $300 million to $2.2 billion since 2017 (Chapter 3 of the April 2024 Global Financial Stability Report).
- Generative AI can be used by bad actors to manipulate markets or conduct cyberattacks:
  - Deepfakes can impersonate key individuals, leading to fraudulent transactions, manipulated stock prices, or erosion of trust that triggers selloffs or deposit runs.
  - Critical financial market or information technology infrastructure can be targeted with significant market disruptions.
- AI incident tracking:
  - Several databases track AI incidents (AI Incident Database; OECD’s AI Incidents Monitor; AI, Algorithmic, and Automation Incidents and Controversies Repository).
  - Despite better AI preparedness, most AI incidents have occurred in advanced economies, even when accounting for differences in GDP.
  - IMF AI Preparedness Index incorporates four macro-structural indicators relevant for AI adoption: digital infrastructure, innovation and economic integration, human capital and labor market policies, and regulation and ethics.
- Examples of recent incidents:
  - A 2024 case: a finance worker at a multinational firm in Hong Kong SAR was reportedly tricked by AI-generated deepfake video and audio, allegedly leading to a $25 million payout to fraudsters.
  - A 2023 ransomware attack on the Industrial and Commercial Bank of China reportedly affected US Treasury market conditions (not known to be related to AI).
  - Fake or genuine social media activity has amplified news and contributed to panic; reports suggest First Republic Bank was targeted by an online manipulation campaign.

*Italic: IMF — Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability, October 2024.*

### Box 3.2 (continued)

### Box 3.2 (continued)

### Glossary — AI-related concepts (as used in the chapter)
- Algorithmic trading (AT)  
  Trading in financial instruments whereby an algorithm independently executes trading decisions.26 Algorithmic trading can be used for trade execution, market-making, or in other proprietary trading strategies. Algorithmic trading strategies vary in complexity and latency; in its simplest guise, algorithmic trading may involve the use of basic trading rules or instruction to feed portions of an order into the market at preset intervals to minimize market impact cost. More complex applications can involve multi-asset trading strategies based on advanced machine learning models. Reinforcement learning allows algorithms to learn dynamically from evolving trading patterns, as well as the actions of other algorithms.

- Artificial intelligence (AI)27  
  The theory and development of computer systems able to perform tasks that traditionally have required human intelligence. The definition of AI is very broad and would include many simple applications that would generally not be described as AI in the public discourse. For example, simple linear regression would fall under this broad definition, even though most would not classify this as AI. The focus of this chapter is on more sophisticated AI, which includes not only generative AI but also more complex nongenerative applications such as clustering algorithms, neural networks, gradient-boosted decision trees, support vector machines, etc.

- Deep learning  
  A form of machine learning that uses algorithms that work in “layers” inspired by the structure and function of the brain. Deep learning algorithms, whose structures are called artificial neural networks, can be used for supervised, unsupervised, or reinforcement learning (itself a form of machine learning).

- FinTech28  
  Technology-enabled innovation in financial services that could result in new business models, applications, processes, or products with an associated material effect on the provision of financial services.

- Foundation models29  
  An umbrella term referring to a diversity of models that are usually trained by applying deep learning to massive quantities of data, such as text and images. Because the expertise, time, and computing power involved in training foundation models from scratch are typically prohibitive for most nonspecialist firms, these models are usually pretrained and shared with end users for further refinement.

- Generative AI  
  AI that generates new content, such as text, images, and videos, often based on user prompts. Generative AI is powered by foundation models, such as large language models.

- High-frequency trading (HFT)  
  HFT is frequently equated to algorithmic trading. However, whereas HFT is a type of algorithmic trading, not all forms of algorithmic trading can be described as high frequency. Algorithmic trading predates HFT and has been extensively used as a tool to determine some or all aspects of trade execution like timing, price, quantity, and venue. Many intermediaries use algorithmic trading for their own proprietary trading or offer it to their clients. It has also become a standard feature in many buy-side firms, mainly with the purpose of devising execution strategies that minimize price impact or to rebalance large portfolios of securities as market conditions change. A number of common features and trading characteristics related to HFT are identified by IOSCO.30

- Large language models (LLMs)  
  Large language models are AI systems designed to learn grammar, syntax, and semantics of one or more languages to generate coherent and context-relevant language.

- Machine learning (ML)  
  A method of designing a sequence of actions to solve a problem that optimizes automatically through experience and with limited or no human intervention.

- Market-making31  
  The provision of liquidity for clients in financial instruments, whereby a trader sets firm bid-offer quotes and thereby provides liquidity for a specific product or a particular product class. This is designed to avoid temporary imbalances between supply and demand for certain products.

- Proprietary trading  
  Describes a trading unit which is separate from the rest of an organization’s trading activities and is not involved in client business. It generates profits exclusively from taking positions. This trading unit has no client contact and is not involved in the broker market.

- RegTech32  
  Any range of applications of FinTech for regulatory and compliance requirements and reporting by regulated financial institutions. This can also refer to firms that offer such applications.

- Reinforcement learning (RL)  
  A type of machine learning paradigm where an agent learns to make decisions by taking actions in an environment to achieve some goal. The learning process is driven by the feedback the agent receives from the environment in the form of rewards or penalties.

- Robo-advisors or automated advice  
  Applications that combine digital interfaces and algorithms, and can also include machine learning, to provide services ranging from automated financial recommendations to contract brokering to portfolio management to their clients, without or with very limited human intervention. Such advisors may be standalone firms and platforms or can be in-house applications of incumbent financial institutions.

- SupTech  
  Any application of FinTech used by regulatory, supervisory, and oversight authorities.

- Synthetic data  
  Artificially generated information that is designed to mimic real-world data in terms of statistical properties and structure. Unlike real data, which are directly collected from real-world events or interactions, synthetic data are created through algorithms and simulation models.

### Citations and notes appearing in the glossary
- Footnotes and institutional sources cited inline: 26, 27, 28, 29, 30, 31, 32 (as presented in the glossary).

*International Monetary Fund | Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial Stability — Box 3.2 (continued)*

---


_Source: https://www.imf.org/-/media/files/publications/gfsr/2024/october/english/textrevised.pdf_