Tracing Stablecoin Contagion during the USDC Depeg after the Silicon Valley Bank Collapse

Abstract: The March 2023 collapse of Silicon Valley Bank (SVB) disrupted the core premise of stablecoins, which are digital tokens designed to maintain a fixed value against the U.S. dollar and serve as on-chain substitutes for dollar liquidity. The event triggered a sharp depeg of USDC, creating a rare exogenous shock to the stablecoin ecosystem. While price deviations during this crisis are well documented, the underlying behavioral reorganization of on-chain activity remains less understood. Here, we analyze high-granularity transaction data to measure the shock's effects on network activities, volumes, and prices, reconstructing the contagion pathway from market-wide synchronization down to account-level reallocation. By extracting phase dynamics, we first show that transaction activity across major stablecoins became strongly synchronized during the crisis window, indicating a collective market-level response. We then uncover a bifurcated contagion pathway. While USDT, WBTC, and WETH reacted primarily as liquidity absorption channels with larger trade volumes, only USDC-related assets exhibited immediate price responses alongside surging transaction counts. This reflects the dominant role of USDC-related assets in this incident and their immediate behavioral connection to user panic, driving a mass reallocation from single-coin to multi-coin portfolios. Finally, governed by persistent intraday time-zone rhythms and balance-size heterogeneity, these findings provide a comprehensive empirical framework for understanding systemic risk and flight-to-quality mechanisms in fractional-reserve digital asset networks.

• The paper demonstrates a bifurcated contagion mechanism during the March 2023 Silicon Valley Bank collapse, revealing that the depeg of USDC and secondary stress in USDC-linked assets, (including DAI and FRAX), propagated through both broad user activation and large-value liquidity transitions across eight stablecoins, through ERC-20 activity data, weighing directional state transitions, and wealth-groups, revealing how contagion played out at the granular level.
• The study uses high granularity Ethereum ERC-20 transfer data and optimized by the Hilbert-Huang phase extraction and various models like autoregressive distributed lag, and the Hilbert’s linear regression to examine market-level synchronization, contagion modes, and significant intraday transaction changes across their sample period of March-April 2023 post the depeg event.
• Markedly, this analysis takes an in-depth system risk analysis of the broad-and-sharpen liquidity transitions in the crypto subsectors in identifying speculative user-trends dominantly from the SVB cryptoreceivables to USDT, and provides a methodological scaffolding for understanding, and monitoring the in-depth behavioral, macroeconomic effects of such international contagion events on macro-resistant stablecoin reliance.

Overview

This paper reconstructs how the March 2023 Silicon Valley Bank (SVB) collapse propagated through the on-chain stablecoin ecosystem following the depeg of USDC. Circle disclosed that approximately USD 3.3 billion—nearly 8% of its cash reserves—was held at SVB, triggering a sharp secondary-market depeg of USDC and secondary stress in USDC-linked assets such as DAI and FRAX. While prior work has documented price, volatility, and abnormal-return effects of this episode, the authors argue that the behavioral reorganization of on-chain activity beneath the price layer remained poorly understood. Using high-granularity Ethereum ERC-20 transfer data for eight assets (USDC, DAI, FRAX, USDP, USDT, BUSD, WBTC, WETH), they combine phase synchronization analysis, autoregressive distributed lag (ARDL) models, account-state transitions, intraday activity profiling, and wealth-group segmentation to trace contagion from market-wide synchronization down to account-level reallocation.

Data and methodology

The study covers March 1 to April 30, 2023, with March 9–13 defined as the main depeg window and matched three-day pre-depeg (March 6–8) and post-depeg (March 14–16) comparison windows. Data are drawn from a full Ethereum archive node (Erigon), including decoded ERC-20 Transfer events, contract-creation records used to distinguish externally owned accounts (EOAs) from contract accounts, and daily per-account balance snapshots reconstructed at last-block-of-day block heights. Daily prices come from CoinGecko.

For each asset and time period, the authors construct a weighted directed transaction graph whose nodes are active accounts and whose edge weights are aggregate transferred amounts. Four observables are derived: transaction count $N_{a,t}$, active-node count $n_{a,t}$, transaction volume $Q_{a,t}$, and average degree $\bar{k}_{a,t}$.

Synchronization is measured via a Hilbert–Huang transform (HHT)-based phase extraction applied to hourly transaction counts over March 2023 (744 hourly observations per asset). Intrinsic mode functions with dominant periods of at least four hours are retained, Hilbert-transformed to obtain instantaneous phases, and combined into a Kuramoto-style order parameter $r(t)$; leave-one-out variants test robustness against single-asset dominance.

Asset-level propagation is tested with ARDL models on first-differenced series, fixing both autoregressive and distributed-lag orders at three daily lags. Specifications regress transaction count, volume, and average degree on price, and volume on node count and transaction count. Account-level analysis classifies accounts as single- versus multi-asset holders, tracks daily asset-focus states (USDC, USDT, WETH, Inactive) for a persistent cohort of 1,359 USDC-active accounts observed between March 6–16, and regresses directional state-transition flows on the USDC price.

Market-level synchronization

The order parameter $r(t)$ remains relatively high throughout the observation window, reflecting a persistent common intraday rhythm in stablecoin transaction activity. During the March 9–13 depeg window, $r(t)$ exhibits a pronounced local increase, reaching one of its highest values in the sample. Leave-one-out robustness checks show that the synchronization rise persists when individual assets are excluded, indicating a collective market-level response rather than the behavior of one dominant coin. This establishes that the SVB shock produced a coordinated change in activity cycles across the broader stablecoin ecosystem before any asset-level decomposition—an early-warning signal that operates independently of peg deviations.

Bifurcated asset-level propagation

The central empirical finding is a bifurcated contagion mechanism. For USDC-related stablecoins (USDC, DAI, FRAX, USDP), the depeg translated into broad participation: sharp increases in transaction counts and active-node counts, with ARDL results showing sustained price sensitivity of transaction counts from lag $L_0$ through $L_3$ for USDC and DAI, and through $L_2$ for FRAX and USDP. Notably, USDC's average degree shows no significant price sensitivity: because transaction counts and node counts scaled nearly proportionally during the panic, local network density was preserved even as total activity surged—the network scaled up its baseline topology rather than rewiring densely. DAI and USDP show contemporaneous average-degree sensitivity at $n_{a,t}$0, and FRAX at $n_{a,t}$1 and $n_{a,t}$2, indicating weaker connectivity adjustment alongside their participation response.

By contrast, USDT, WBTC, and WETH reacted primarily as liquidity absorption channels. USDT's transaction count is not price-sensitive, but its average degree and transaction volume are significant from $n_{a,t}$3 to $n_{a,t}$4, and its volume responds to active-node count at $n_{a,t}$5–$n_{a,t}$6 but to transaction count only at $n_{a,t}$7—evidence of value-flow adjustment rather than broad user activation. WBTC and WETH show transaction-volume spikes without comparable increases in counts or nodes; WETH's volume becomes price-sensitive only at $n_{a,t}$8 and is associated with node count but not transaction count, consistent with larger transfers by fewer participants. BUSD serves as a control: it exhibits no price sensitivity in any price-related specification while retaining ordinary internal volume–activity coupling, remaining outside the main propagation channels.

The implication is that contagion was heterogeneous in both magnitude and mechanism: confidence shocks propagate through distinct participation channels (broad user activation in shocked and collateral-linked stablecoins) and liquidity channels (large-value transfers into alternative assets).

Account-level reallocation

At the address level, the ratio of single-coin accounts falls during the depeg window while the ratio of multi-coin accounts rises, and both ratios are price-dependent at lag $n_{a,t}$9—indicating that diversification from concentrated toward cross-asset positioning emerged with a short delay after the price deviation rather than instantaneously.

Tracking the 1,359-account persistent USDC-active cohort across daily asset-focus states reveals the direction of movement. The dominant price-linked transition is USDC→USDT, significant at $Q_{a,t}$0 (coefficient −931.27, $Q_{a,t}$1), providing direct account-level support for USDT as the immediate alternative liquidity destination. Critically, USDC→WETH and USDC→Inactive flows are not price-dependent: the immediate response was not an exit into volatile wrapped assets or general withdrawal, but a stablecoin-to-stablecoin flight. On the USDT side, the USDT→USDC flow is price-dependent at $Q_{a,t}$2 and $Q_{a,t}$3, which the authors interpret as re-entry, arbitrage, or purchase of discounted USDC rather than retention; the USDT→WETH transition is significant only at $Q_{a,t}$4, suggesting a two-step process in which a subset of accounts moved from stablecoin liquidity into wrapped crypto exposure with a longer delay.

Intraday rhythms and wealth heterogeneity

Under normal conditions, transaction activity follows a recurring intraday rhythm: onset around 7:00–8:00 UTC, consistent with European daytime hours, strengthening through midday-to-afternoon UTC overlapping U.S. East Coast morning trading. During the depeg, USDC and DAI show elevated activity across much of the day rather than a narrow band—the crisis partially overwhelmed normal market-hour timing into a broad all-day activation—with USDT showing a milder version of the same disruption. The authors caution that blockchain addresses provide no geolocation, so this constitutes timing-based evidence of market-hour structure, not direct identification of user locations.

Within the same 1,359-account cohort, segmented into bottom-, middle-, and top-20% wealth groups by reference-period USDC-denominated balances, responses diverge systematically. Lower- and medium-wealth accounts exhibit larger relative daily balance fluctuations around the depeg and upward cumulative median balance adjustment afterward, whereas top-wealth accounts (spanning roughly $Q_{a,t}$5–$Q_{a,t}$6 USDC) show muted relative changes and flat-to-declining cumulative positions. The authors acknowledge that percentage changes are naturally larger for smaller initial balances, so part of the relative-fluctuation effect may be mechanical; nevertheless, the divergence in cumulative adjustment indicates the depeg was not experienced uniformly across the account population.

Limitations

The paper is explicit about several constraints. The ARDL results identify contemporaneous and lagged associations, not causal effects. The account-level analysis captures only Ethereum ERC-20 activity and omits centralized-exchange trades, off-chain redemptions, and other chains—a material omission given that much USDC redemption pressure occurred off-chain. The intraday geographic interpretation rests on timing signatures rather than identified locations. The wealth-heterogeneity result depends on reference-period balance classification within a single cohort and is partly sensitive to the mechanical properties of relative changes. Open questions include extending the framework to cross-chain flows, pool-level DeFi liquidity, exchange data, and whether synchronization-based indicators can function as real-time early-warning signals with explicit stablecoin–stablecoin and stablecoin–DeFi dependency modeling.

Conclusion

The paper demonstrates that the SVB-induced USDC depeg was a multiscale behavioral event: synchronized transaction activity across stablecoin-related assets, bifurcated propagation through broad-participation channels (USDC, DAI, FRAX, USDP) and large-value-transfer channels (USDT, WBTC, WETH), account-level diversification from single- to multi-asset exposure dominated by short-lag USDC→USDT movement, crisis-broadened intraday activation, and wealth-dependent adjustment intensity. Its main implication is methodological: stablecoin contagion should be monitored as a behavioral-network process in which transaction synchronization, network observables, and user-flow transitions complement peg deviations as systemic-risk indicators.