Stablecoin Systemic Risk: What IMF Research Reveals About Run Dynamics and Fire Sales

Stablecoins now represent over $320 billion in circulating supply, with Tether and Circle alone holding more U.S. Treasury bills than many sovereign nations. At this scale, stablecoins are no longer just crypto infrastructure: they are participants in the global safe-asset market. A wave of recent research from the IMF, BIS, and NBER has begun quantifying what happens when these instruments experience stress. The findings reveal feedback loops between stablecoin reserves, bond markets, and redemption dynamics that regulators and protocol designers can no longer ignore.

This article examines four key papers published between 2025 and 2026 that collectively map the systemic risk landscape for stablecoins. The research matters for anyone building or using stablecoin infrastructure, including Bitcoin Layer 2 protocols that support dollar-denominated assets.

How Big Is the Problem?

The stablecoin market has grown from roughly $5 billion in early 2020 to over $320 billion by mid-2026. USDT accounts for approximately $190 billion of that total, with USDC at roughly $78 billion. Together, the two largest fiat-backed stablecoins represent about 85% of the market. Combined trading volumes for USDT and USDC reached $23 trillion in 2024, exceeding the GDP of most countries.

The reserve portfolios behind these stablecoins now hold substantial positions in U.S. Treasury markets. As of December 2025, Tether reported $141 billion in direct and indirect Treasury exposure. Circle holds approximately 80% of USDC reserves in Treasury allocations through the Circle Reserve Fund, a registered 2a-7 government money market fund managed by BlackRock. Combined, the two issuers hold roughly $108 billion in T-bills alone, representing approximately 1.7% of the total U.S. T-bill market. This concentration makes stablecoin flows a measurable force in sovereign debt pricing.

IMF "Stablecoin Shocks": Measuring Treasury Yield Transmission

The first paper to quantify this channel is "Stablecoin Shocks" (IMF Working Paper WP/26/044), published in March 2026 by Cerutti, Firat, Hengge, and Sagawa. The authors construct novel measures of stablecoin demand shocks by combining a daily narrative dataset of stablecoin-specific news events with changes in the combined market capitalization of USDC and USDT. They use heteroskedasticity-based identification within both event-study and structural vector autoregression (SVAR-IV) frameworks.

Key Quantitative Findings

The central result: a $3.5 billion inflow into stablecoins (roughly a 1% increase in combined USDC and USDT market capitalization) lowers 3-month T-bill yields by approximately 2 to 2.5 basis points. The effect on 1-month T-bill yields is slightly smaller at about 1.9 basis points, with the trough reached around week 24. Effects attenuate at longer maturities: 1-year and 10-year Treasuries show limited impact.

The paper also identifies cross-asset spillovers. Stablecoin demand shocks trigger a depreciation of the U.S. dollar and boost the Bloomberg Galaxy Crypto Index by roughly 1.5% at peak. The S&P 500 shows a more muted response of about 0.25%. Payment provider stocks benefit from stablecoin inflows, while banks (including community and small banks) show no priced disintermediation risk, suggesting that stablecoin growth has not yet begun displacing bank deposits in investor expectations.

Scale matters: The 2 to 2.5 basis point impact per $3.5 billion inflow may seem small, but stablecoin flows are not small. In 2024, combined USDT and USDC trading volumes reached $23 trillion. During stress events, redemption volumes can reach $7 billion in 48 hours. At those magnitudes, the yield impact becomes macroeconomically significant.

BIS Working Paper 1270: Supply Scarcity Amplifies the Effect

The BIS independently corroborated and extended these findings in "Stablecoins and Safe Asset Prices" (Working Paper 1270), authored by Rashad Ahmed and Inaki Aldasoro. Originally published in May 2025 and revised in February 2026, the paper uses daily data from 2021 to 2025 with an instrumental variable approach.

The BIS estimates are slightly larger: a 2-standard-deviation stablecoin inflow (approximately $3.5 billion) reduces 3-month T-bill yields by 2.5 to 3.5 basis points within 10 days. The critical contribution is the finding that effects are state-dependent. During periods of ample bill supply, the yield impact is statistically insignificant. But during bill scarcity (debt ceiling standoffs, growth in the Fed's reverse repo facility), the effect jumps to 5 to 8 basis points.

The Asymmetry Problem

The most consequential finding for systemic risk analysis is the asymmetry between inflows and outflows. The BIS paper shows that outflows raise Treasury yields by 2 to 3 times as much as inflows lower them. A stablecoin redemption event that triggers $3.5 billion in outflows can push 3-month yields up by 6 to 8 basis points. This asymmetry means that depeg events and mass redemptions hit bond markets harder than the equivalent inflow benefits them.

The paper attributes approximately 70% of the estimated yield impact to USDT, 19% to USDC, and 11% to other stablecoins. This concentration mirrors the market share distribution and underscores the systemic importance of Tether's reserve management decisions.

The Redemption-Fire Sale Feedback Loop

While the IMF and BIS papers measure the transmission from stablecoin flows to bond prices, a separate IMF working paper models the internal dynamics of a stablecoin run. "From Par to Pressure: Liquidity, Redemptions, and Fire Sales with a Systemic Stablecoin" (WP/26/005), published in January 2026 by Marco Gross and Richard Senner, formalizes the feedback loop that makes stablecoin runs self-reinforcing.

The mechanism works in stages. First, a negative shock (regulatory action, counterparty failure, reserve transparency concern) triggers an initial wave of redemptions. The issuer must liquidate reserve assets, typically short-term Treasuries, to meet these redemptions. If volumes are large enough, the asset sales depress bond prices. This reduces the mark-to-market value of the issuer's remaining reserves, eroding the collateral ratio. The declining collateral ratio triggers further redemptions from holders monitoring reserve adequacy. The loop repeats.

Design Dials

The paper identifies several design parameters (which the authors call "dials") that govern the severity of this feedback loop. These include capital buffers above the 1:1 reserve requirement, the composition and duration of reserve assets, and redemption gates or fees that slow outflows during stress. The model shows that robust prudential design can substantially stabilize stablecoins and their surrounding market environment, but the specific calibration involves tradeoffs.

Higher capital buffers reduce run probability but impose costs on issuers that reduce yield competitiveness. Shorter-duration reserves (overnight repos versus 6-month T-bills) reduce fire sale losses but lower returns. Redemption gates reduce fire sale intensity but may erode confidence if holders interpret gates as a signal of distress. Each dial addresses a different part of the feedback loop, and the paper argues that a combination of all three is necessary for resilience.

The analogy to money market funds is deliberate: The authors draw explicit parallels to the 2008 Reserve Primary Fund break-the-buck event. The same fire sale dynamics apply, but stablecoins face additional complications: they operate 24/7, lack access to central bank liquidity facilities, and their secondary market pricing is visible in real time, which accelerates information cascades.

The Arbitrage Paradox: When Price Stability Amplifies Run Risk

Perhaps the most counterintuitive finding comes from an NBER working paper (No. 33882, May 2025) by Yiming Ma, Yao Zeng, and Anthony Lee Zhang, titled "Stablecoin Runs and the Centralization of Arbitrage." The paper demonstrates that efficient stablecoin arbitrage in secondary markets, which keeps prices close to $1.00 during normal times, paradoxically increases the risk of a catastrophic run.

The logic: when arbitrage is efficient, individual sellers face low price impact. Selling $10 million worth of USDT on a liquid secondary market barely moves the price. This low price impact makes selling more attractive during stress, because each holder knows they can exit without taking a large haircut, as long as they act quickly enough. The result is a classic coordination failure: each holder rationally runs, and the aggregate effect is a stampede.

Concentrated Arbitrage as a Stabilizer

The paper finds that concentrated arbitrage (fewer, larger market-making entities) actually reduces run risk by increasing price impact. When selling stablecoins moves the price significantly, holders have less incentive to panic-sell, because early sellers absorb losses that discourage others from following. This finding has direct implications for depeg risk management.

Tether's redemption structure provides a real-world example of this dynamic. The paper notes that Tether permits only approximately 6 agents per month to redeem stablecoins for cash, with a minimum redemption threshold of $100,000 and a 0.1% fee (minimum $1,000). This extreme concentration of primary-market access functions as a de facto gate, channeling most selling pressure through secondary markets where price impact provides a natural brake on runs.

The policy implication is provocative: regulations designed to improve stablecoin price stability through more efficient redemption mechanisms may have the unintended consequence of increasing systemic run risk. The peg mechanism itself becomes a vector for instability.

Historical Stress Tests: What Actually Happened

Theory aside, the stablecoin market has already experienced two major stress events that partially validated these models.

Terra/UST Collapse (May 2022)

UST was the third-largest stablecoin by market capitalization at approximately $17.5 billion before its collapse. Unlike fiat-backed stablecoins, UST relied on an algorithmic mechanism tied to the LUNA token. Roughly 75% of UST was deposited in Anchor Protocol earning a 19.5% yield. On May 7, two large withdrawals from Anchor triggered the initial depeg. By May 9, UST had fallen to $0.60. LUNA's supply hyperinflated approximately 80x between May 10 and 12, a textbook death spiral that erased roughly $45 billion in combined market capitalization.

The contagion to fiat-backed stablecoins was immediate. Tether processed approximately $7 billion in redemptions within 48 hours and $10 to $13 billion over one week. USDT briefly fell to $0.97 on major exchanges. Tether's market capitalization declined from $83 billion to roughly $73 billion in the immediate aftermath, eventually reaching $65 billion over two months.

USDC/SVB Depeg (March 2023)

When Silicon Valley Bank was closed on March 10, 2023, Circle disclosed that $3.3 billion (approximately 8% of USDC reserves) was held at the bank. USDC fell to between $0.86 and $0.87 on secondary markets. Hourly trading volume peaked at nearly $2 billion on March 11. Primary-market redemptions were suspended over the weekend, which amplified secondary-market selling pressure. The peg was restored on March 13 after the Federal Reserve, Treasury, and FDIC announced a depositor backstop for SVB.

The contagion pattern confirmed the feedback mechanisms described in the IMF papers. DAI depegged as approximately $1 billion in USDC was deposited into its Peg Stability Module per day on March 10 and 11. USDP fell to roughly $0.91. Meanwhile, USDT and BUSD traded marginally above $1 as capital rotated toward stablecoins perceived as unexposed to U.S. banking system risk.

Design Levers for Stablecoin Resilience

The combined research points to several mechanisms that can reduce stablecoin systemic risk. The April 2026 IMF paper "Making Stablecoins Stable" (WP/26/074) by Li, Mancini-Griffoli, Miccoli, Tan, and Zhang provides a theoretical framework for evaluating these levers.

Capital Buffers

Excess reserves above the 1:1 peg requirement create a loss-absorption layer that prevents the feedback loop from activating during moderate stress. Tether currently maintains approximately $6.3 billion in excess reserves as of year-end 2025. The "From Par to Pressure" model shows that capital buffers reduce both run frequency and fire sale intensity, but the required buffer size depends on reserve composition and redemption mechanics. The "Making Stablecoins Stable" paper finds that without regulatory requirements, issuers rationally underinvest in buffers because the costs (lower yield competitiveness) are borne privately while the benefits (reduced systemic risk) are distributed across the financial system.

Reserve Composition and Duration

The duration of reserve assets determines fire sale severity. Overnight repos and very short-term T-bills can be liquidated with minimal price impact. Longer-duration instruments offer higher yields but create larger mark-to-market losses during rapid liquidation. Circle's reserve fund holds Treasuries with a weighted-average maturity under 60 days plus overnight repos collateralized by Treasuries. This short-duration profile limits fire sale risk but compresses the yield available for reserve management.

The BIS paper's finding that outflows raise yields 2 to 3 times more than inflows lower them creates an additional incentive for short-duration reserves: the asymmetric market impact means that even moderate-duration holdings can generate outsized losses during stressed liquidation.

Redemption Gates and Fees

Redemption gates limit the rate at which holders can convert stablecoins back to fiat, preventing the kind of instantaneous mass redemption that overwhelms reserve liquidity. The NBER paper's finding about concentrated arbitrage suggests that Tether's restrictive redemption structure (approximately 6 authorized agents per month, $100,000 minimum) may actually enhance stability by channeling selling pressure through price-sensitive secondary markets.

However, gates create their own risks. The SEC's experience with money market funds is instructive: the SEC introduced redemption gates for prime money market funds after 2008 but rolled back the gate authority in 2023, retaining only liquidity fee provisions. The concern was that gates could trigger preemptive runs as investors rush to redeem before gates activate, a phenomenon known as "gate risk."

Regulatory Responses: The GENIUS Act Framework

The U.S. GENIUS Act, signed into law in 2025, addresses several of the risk vectors identified in the research. The law requires 1:1 reserve backing with specified permitted assets (U.S. dollars, T-bills, short-term repos, money market funds). Reserves must be segregated from operational funds, and rehypothecation is explicitly forbidden. Regulators are tasked with establishing additional capital and liquidity requirements through rulemaking.

The European MiCA regulation takes a different approach, requiring e-money token issuers to hold reserves in credit institutions and limiting investment to low-risk financial instruments. Both frameworks attempt to constrain the reserve composition dial, but neither directly addresses the arbitrage paradox identified in the NBER paper. More efficient redemption mechanisms mandated by regulation could inadvertently increase run risk by reducing price impact in secondary markets.

The IMF's December 2025 departmental paper "Understanding Stablecoins" goes further, calling for harmonized legal definitions, strict reserve and redemption standards, granular disclosure requirements, and cross-border supervisory coordination. The paper notes that approximately 97% of outstanding stablecoin tokens reference the U.S. dollar, making this primarily a dollar-system regulatory challenge. For a detailed breakdown of how these frameworks compare, see our analysis of global stablecoin regulation.

What This Means for Stablecoin Design on Bitcoin

The research has direct implications for how stablecoins should be designed and managed on Bitcoin Layer 2 networks. The fire sale feedback loop applies regardless of which blockchain a stablecoin operates on: the underlying risk is in the reserve portfolio, not the settlement layer. But the settlement layer does affect several key parameters.

Bitcoin-native stablecoin designs face different tradeoff surfaces than their Ethereum or Tron counterparts. On-chain transparency is structurally different in UTXO-based systems. Redemption mechanics depend on the Layer 2's settlement guarantees. And the absence of complex DeFi composability on most Bitcoin Layer 2s limits the contagion pathways that amplified the UST and USDC stress events (no equivalents to Anchor Protocol or DAI's Peg Stability Module absorbing concentrated stablecoin deposits).

Spark, as a Bitcoin Layer 2 supporting stablecoins like USDB, engages directly with these risk considerations. USDB's reserve backing through Brale reflects design choices informed by exactly these dynamics: fiat-backed reserves, regulated issuance infrastructure, and a separation between the settlement layer (Spark) and the reserve management layer (the issuer). This separation means that a stablecoin run on USDB would not compromise Spark's Bitcoin-denominated operations, an important isolation property that the IMF's feedback loop models highlight as critical.

For a deeper look at how yield-bearing stablecoins introduce additional complexity to these risk models, including the interaction between yield distribution and redemption incentives, see our dedicated analysis.

Open Questions and Future Research

Several important questions remain unresolved in the current literature.

• How do stablecoin fire sales interact with Treasury market microstructure during periods when primary dealer balance sheets are already constrained?
• The BIS paper projects that a $2 trillion stablecoin market (Standard Chartered's 2028 forecast) could see 2-standard-deviation flows suppress T-bill yields by 6 to 8 basis points. What happens at $5 trillion?
• Should stablecoin issuers have access to central bank liquidity facilities (discount window, repo) to prevent fire sales, or would access create moral hazard?
• How do cross-chain stablecoin bridges affect contagion dynamics when a depeg occurs on one chain but not another?
• The NBER arbitrage paradox suggests that perfectly liquid stablecoins are inherently less stable. Is there an optimal level of redemption friction?

These questions will likely be addressed as the stablecoin market continues to grow and regulatory frameworks mature. The research published so far provides a rigorous foundation, but the models have not yet been tested against a stress event at the current $320 billion market scale.

Key Takeaways

The IMF, BIS, and NBER research collectively establishes several findings that anyone involved in stablecoin infrastructure should understand.

• Stablecoin flows measurably affect Treasury yields, with outflows having 2 to 3 times the impact of equivalent inflows.
• Redemption-fire sale feedback loops are self-reinforcing and require multiple design interventions (capital buffers, reserve composition, gates) to mitigate.
• Efficient secondary-market arbitrage improves normal-time price stability but paradoxically increases catastrophic run risk.
• Reserve duration, not just reserve adequacy, determines fire sale severity.
• The 24/7 operation and real-time pricing of stablecoins compress the time available for intervention compared to traditional money market stress events.

For developers building on Spark or other Bitcoin Layer 2 protocols that support stablecoins, understanding these dynamics is not optional. The choice of which stablecoins to integrate, how to handle depeg scenarios programmatically, and how to communicate reserve risk to end users are all informed by this research. Our comparison of peg mechanisms provides additional context for evaluating specific stablecoin designs.

This article is for educational purposes only. It does not constitute financial or investment advice. Bitcoin and Layer 2 protocols involve technical and financial risk. Always do your own research and understand the tradeoffs before using any protocol.