Methodology
The taxonomy is derived through a three-pillar approach that triangulates evidence from the academic literature, first-principles theory, and real-world crisis episodes.
Three Pillars
Systematic Literature Review
2,433 papers retrieved from OpenAlex using Boolean keyword queries across 14 candidate channels.
Theoretical Derivation
First-principles analysis of mechanism distinctiveness, novelty relative to traditional finance, and cross-channel interactions.
Crisis Case Evidence
25 crisis episodes (2009–2026) mapped to channels, weighted by log-transformed loss estimates.
Data Collection
The literature corpus was assembled from OpenAlex, an open scholarly database. For each of the 14 candidate systemic risk channels, a Boolean keyword query was constructed targeting titles, abstracts, and concepts. Results were sorted by citation count (descending) with a 200-paper cap per channel, yielding 2,433 papers in total after deduplication across channels.
The 14 candidate channels were identified through a combination of prior literature surveys, the authors' domain expertise, and iterative refinement during the review process.
Search Keywords
The exact Boolean keyword queries passed to the OpenAlex API, as defined in references/search_queries.json and executed by scripts/openalex_search.py:
Search Parameters
- API
- OpenAlex (
api.openalex.org/works) - Publication years
- 2009–2026
- Minimum citations
- 5
- Sort order
- cited_by_count descending
- Max results per channel
- 200
- Language
- English
Per-Channel Queries
- systemic risk contagion cryptocurrency network
- DeFi interconnectedness systemic risk
- crypto exchange network contagion
- financial network topology cryptocurrency
- liquidity spiral cryptocurrency
- AMM liquidity crisis DeFi
- exchange run crypto
- fire sale digital assets
- stablecoin de-peg systemic risk
- algorithmic stablecoin collapse
- stablecoin bank run
- USDT USDC de-pegging contagion
- DeFi composability risk
- smart contract exploit cascade
- money legos systemic risk
- DeFi protocol dependency
- on-chain liquidation cascade
- DeFi leverage systemic risk
- cascading liquidation cryptocurrency
- margin call crypto
- counterparty risk cryptocurrency exchange
- FTX contagion systemic
- centralized exchange failure
- crypto concentration risk
- proof of reserves cryptocurrency
- CeFi opacity systemic risk
- information asymmetry crypto exchange
- fiat crypto gateway systemic
- banking channel cryptocurrency
- on-ramp off-ramp fragility
- Silvergate SVB crypto
- oracle manipulation DeFi
- price feed contagion blockchain
- oracle dependency systemic risk
- Chainlink oracle attack
- DAO governance attack
- crypto governance systemic risk
- hard fork contagion
- protocol governance failure
- regulatory shock cryptocurrency
- crypto regulation financial stability
- compliance contagion digital finance
- cross-chain bridge exploit
- bridge hack systemic risk
- Wormhole Ronin bridge contagion
- mining concentration systemic risk
- validator centralization blockchain
- consensus layer risk
- tokenized assets systemic risk
- real world asset DeFi bridge
- tokenization financial stability
Broad Queries (Cross-Channel)
- systemic risk digital finance
- systemic risk cryptocurrency
- DeFi financial stability
- stablecoin systemic risk
- crypto contagion
- tokenization systemic risk
- CBDC financial stability
references/search_queries.json. The search was executed by scripts/openalex_search.py using scripts/openalex_client.py (rate-limited at 10 req/sec via the OpenAlex polite pool). Results were deduplicated across channels by OpenAlex work ID. Per-channel raw results are stored in output/data/openalex_raw_{channel_id}.json; the merged corpus is in output/data/openalex_merged.json.
Composite Scoring
Each candidate channel was scored on three normalized dimensions, combined with the following weights:
- Literature volume (35%): paper count per channel, normalized to [0, 1]
- Citation impact (35%): mean of top-10 cited papers per channel, normalized to [0, 1]
- Crisis evidence (30%): count of associated crisis events, weighted by log-transformed aggregate losses, normalized to [0, 1]
Channel Selection
All 14 candidate channels ranked by composite score, with disposition decisions based on qualitative assessment:
| Rank | Channel | Lit | Cit | Crisis | Composite | Decision |
|---|---|---|---|---|---|---|
| 1 | Cross-Chain Bridge Vulnerability | 0.575 | 1.000 | 0.517 | 0.706 | Merged into Composability Risk |
| 2 | Liquidity Spirals | 0.756 | 0.307 | 0.996 | 0.671 | Retained |
| 3 | Gateway Risk (Fiat-Crypto Bridging) | 1.000 | 0.325 | 0.680 | 0.668 | Retained |
| 4 | Composability Risk | 0.575 | 0.578 | 0.740 | 0.626 | Retained |
| 5 | Counterparty Concentration | 0.575 | 0.211 | 1.000 | 0.575 | Retained |
| 6 | Validator/Miner Concentration | 0.575 | 0.964 | 0.076 | 0.561 | Merged into Counterparty Conc. |
| 7 | Information Asymmetry and Opacity | 0.575 | 0.177 | 0.618 | 0.449 | Retained |
| 8 | Liquidation Cascades | 0.575 | 0.220 | 0.343 | 0.381 | Retained |
| 9 | Network Contagion | 0.575 | 0.164 | 0.406 | 0.380 | Retained (foundational theory) |
| 10 | Governance Failure | 0.575 | 0.232 | 0.210 | 0.345 | Split: Composability / Counterparty |
| 11 | Stablecoin Runs | 0.626 | 0.031 | 0.245 | 0.304 | Retained (foundational theory) |
| 12 | Oracle Manipulation | 0.598 | 0.131 | 0.131 | 0.295 | Merged into Composability Risk |
| 13 | Regulatory Contagion | 0.575 | 0.179 | 0.073 | 0.286 | Distributed: Gateway / Counterparty |
| 14 | Real-World Asset Transmission | 0.575 | 0.220 | 0.000 | 0.278 | Deferred (insufficient crisis evidence) |
Sensitivity Analysis
Four alternative weight schemes were tested to assess the robustness of the channel selection:
| Scheme | Literature | Citations | Crisis |
|---|---|---|---|
| Primary | 0.35 | 0.35 | 0.30 |
| Equal | 0.333 | 0.333 | 0.333 |
| Crisis-dominant | 0.25 | 0.25 | 0.50 |
| Literature-dominant | 0.50 | 0.25 | 0.25 |
All four schemes produce the same qualitative retention decisions, confirming selection stability. While individual rank orderings shift under alternative weights, the set of eight retained channels remains unchanged.
Known Limitations
- Off-topic papers in citation impact: Broad Boolean keyword matching retrieves papers only tangentially related to each channel. The top-10 citation metric may therefore reflect the prominence of a research area rather than the specific systemic risk mechanism.
- 200-paper cap compresses literature volume: Channels with large literatures (e.g., gateway risk at 348 papers before capping) lose information when truncated, compressing variance in the literature volume sub-score.
- Temporal citation bias: Sorting by citation count favors older, heavily cited works and may underweight recent contributions.
- Single-coder classification: Channel assignment and crisis mapping were performed by a single researcher, introducing potential classification bias.