Researchers introduce the Aggregated Systemic Risk Index (ASRI), a composite measure designed to flag system-wide danger coming from interactions between decentralized finance (DeFi) and traditional financial institutions. The index combines four sub-measures with fixed weights: Stablecoin Concentration Risk (30%), DeFi Liquidity Risk (25%), Contagion Risk (25%), and Regulatory Opacity Risk (20%). The team builds ASRI using data from DeFi Llama, the U.S. Federal Reserve’s FRED database, and on‑chain analytics.

Each sub-index focuses on a different channel of risk. Stablecoin Concentration looks at reserve mix and peg stability (for example, total value locked ratios, treasury holdings, a concentration index, and peg volatility). DeFi Liquidity captures protocol concentration, leverage changes and smart contract vulnerabilities, including flash‑loan exposure. Contagion Risk measures links to traditional finance, tokenized real‑world asset growth, bridges and changing correlations. Regulatory Opacity scores transparency, custody concentration, and regulatory arbitrage. These pieces are combined into a single ASRI score meant to reflect how fragile the whole system is.

The authors test ASRI against four real crises: the Terra/Luna collapse (May 2022), the Celsius/Three Arrows Capital contagion (June 2022), the FTX bankruptcy (November 2022), and the Silicon Valley Bank (SVB) episode (March 2023). An event‑study finds statistically significant abnormal signals for all four crises (t‑statistics between 5.47 and 32.64, all p < 0.01). Using a threshold rule, ASRI flagged three of the four events ahead of time with an average lead of 30 days; the Terra/Luna event was not caught by that threshold, which the authors attribute to limits of market‑based indicators for algorithmic stablecoins. A walk‑forward (out‑of‑sample) test detected all four crises with an average lead time of 18 days, which the paper uses to argue against look‑ahead bias. The paper also reports that each sub‑index is statistically stable over time and that a three‑state Hidden Markov Model (a statistical method that groups observations into low, moderate, and elevated risk states) finds very persistent regimes (>97% persistence). Structural stability tests also pass (Chow p = 0.993).