GCRI · 全球挑战
系统性风险与级联风险
Systemic risk is the risk that a shock in one part of a connected system spreads, amplifies and disrupts the whole. A drought travels through water, energy, food prices, budgets and health. A financial failure halts payments that hospitals and utilities rely on. The Global Centre for Risk and Innovation (GCRI) traces where shocks cross systems, where they amplify and where they can be stopped. Part of global challenges.
有记录的系统性风险和级联风险。
- 01机构全球风险与创新中心(Global Centre for Risk and Innovation,GCRI)
- 02Key ideasFour features make a risk systemic
- 03Terms and measuresWhat systemic risk means in practice, and what can be measured
- 04方法及其局限专家用于关联风险的模型,以及各模型在何处失效
- 05Data and its statusThe evidence on connected risk is thinner than the evidence on single hazards
- 06GCRI’s partWhat GCRI does here, how it reaches Reports, and how you can help
- 07How GCRI approaches itMethods for connected risk
使风险具有系统性的有四个特征。
Systemic risk is not a separate hazard. It is a property of how systems are connected, which is why it is often invisible to assessments made one sector at a time.
Tight connections
由电力、水、货物、数据和资金流动联结的系统会迅速传导压力。
反馈回路
价格飙升、囤积、恐慌和被迫抛售会使冲击在扩散过程中不断放大。
共同故障点
众多服务背后的某一项资产、供应商或假设一旦失效,可能使这些服务同时失效。
Thresholds and tipping points
系统在一定限度内可以吸收压力,随后会突然改变状态。
What systemic risk means in practice, and what can be measured.
Two definitions anchor the work. The International Risk Governance Center describes systemic risks as risks to the functioning of a system that arise from the interactions of its parts. Their effects reach well beyond the point of origin. The UNDRR Global Assessment Report of 2019 adds that such risk is often embedded in a system without being recognised as a risk, until a shock reveals it.
| 概念 | 描述内容 | How it is measured |
|---|---|---|
| 相互依存 | A two-way state relation between systems; Rinaldi and colleagues distinguish physical, cyber, geographic and logical links | Dependency matrix of services and assets; share of a service’s inputs that come from each other system |
| 级联 | 一处故障引发其他地方的故障,可能在数小时、数周或数季内依次发生 | Fraction of nodes or services failed after propagation; time from first failure to last; time to recover in days |
| 放大效应 | A feedback that makes the shock larger as it spreads: fire sales, hoarding, price spikes, loss of confidence | 模型运行中总损失与初始损失之比,并附其范围报告 |
| 集中度 | Many functions resting on one asset, supplier, route or assumption | 供应商或路线份额;赫芬达尔-赫希曼集中度指数;单点故障数量 |
| 临界转变 | 越过阈值后状态之间的突然转变 | 长序列中方差和一阶自相关的上升,应视为可能的预警,而非预测 |
专家用于关联风险的模型,以及各模型在何处失效。
| 方法 | What it answers | 失效之处 |
|---|---|---|
| 网络传染模型,包括 DebtRank(Battiston 等,2012) | 哪些机构或资产会使冲击沿风险暴露网络传导得最远 | 风险暴露网络很少完整或公开;缺失的联系会改变谁看起来处于中心 |
| Interdependent network percolation (Buldyrev and others, 2010) | Why coupled networks can collapse abruptly where a single network would degrade gradually | Stylised: real operators reroute, shed load and repair, which the theory leaves out |
| 投入产出模型与失能模型 | 一个部门的产出损失如何减少向其采购的各部门的产出 | Fixed technical coefficients allow no substitution; results overstate losses over longer horizons |
| 基于主体的模型 | 囤积、恐慌或更换供应商等行为如何影响扩散 | 参数众多而验证数据很少;结果属于探索性分析,并连同其敏感性一并发布 |
| 情景故事线与复合压力情景 | 一个在物理和经济上自洽的不利序列是什么样子,逐步说明 | 不附带概率;其价值取决于各系统的专家是否检验每一步 |
Haldane and May showed in 2011 how homogeneity and tight coupling made a banking system fragile in ways no single bank’s risk model could see. The same lesson holds for water, power and food systems: the property that matters belongs to the network. Every model result here carries its assumptions, its range and the label Modelled or Scenario.
关于关联风险的证据比关于单一危害的证据更为薄弱。
| 证据 | 它提供什么 | 如实说明现状 |
|---|---|---|
| 国家供给使用表,以及国家间投入产出表(OECD ICIO) | 用于传播模型的行业间和国家间流量 | Published with a lag of several years; sector detail too coarse for many infrastructure questions |
| 灾害损失数据库(EM-DAT、DesInventar Sendai) | Recorded direct losses by event and place | 间接和级联损失很少被记录;小型事件存在漏报 |
| Infrastructure dependency data held by operators | 哪些资产支撑哪些服务 | Mostly restricted for security reasons; public studies work with permitted or synthetic topology |
| 监管金融风险暴露数据 | 机构间的双边风险暴露 | Held by supervisors; public work relies on aggregates and reconstructed networks |
凡数据不足以支持某项论断之处,页面会说明该论断尚不能成立。缺口就如实显示为缺口,绝不以看似合理的数字填补。
GCRI 在此做什么、如何接入 Reports,以及您如何提供帮助。
GCRI is a research, development and policy institution. On systemic risk it develops and tests methods, helps national teams map their own dependencies on their own data, and publishes what review supports. It does not regulate, supervise, fund or approve projects, and every decision on a system stays with the authority or operator that runs it. Work is organised through the Systemic Risk Management platform and the Systems Science and Modelling platform.
Findings are designed to reach the Global Systems Risk and Resilience series and the Scenario and Stress-Test series. Under the proposed Nexus Standards, each dependency record and indicator series would carry a record identifier (NS-3) and each model a model card (NS-12). A reader could then trace a result back to its inputs.
- 一条有记录的级联链,每一步都附有来源,包括归因不确定的环节。
- 对某种传播方法的批评,以及该方法失效的案例。
- 针对某一城市或行业的获准使用的依赖关系数据,经秘书处按规定条款提供。
互联风险的方法。
- 01
Map the system
Dependencies across sectors and borders with network science and contagion.
- 02
Simulate propagation
复杂系统与基于智能体的模型 ,说明行为如何影响扩散,并明确陈述假设。
- 03
Stress-test
Compound scenarios using scenario design and stress testing, labelled as scenario evidence.
- 04
Find points of control
Interventions that cut or buffer transmission, with their trade-offs and limits.
Where this page’s references lead.
阅读本页时所对照的机构和文件。状态和版本会变化;每个链接都指向发布机构。
- 《IRGC 系统性风险治理指南》(EPFL 国际风险治理中心,2018 年)
- 《2019 年减少灾害风险全球评估报告》系统性风险专辑(UNDRR,2019 年)
- OECD Recommendation of the Council on the Governance of Critical Risks (OECD, 2014)
- Haldane and May, Systemic risk in banking ecosystems, Nature 469 (2011)
- Battiston and others, DebtRank: too central to fail?, Scientific Reports 2 (2012)
- Buldyrev 等,《相互依赖网络中的灾难性级联失效》,Nature 464(2010)
- Rinaldi、Peerenboom 和 Kelly,《关键基础设施相互依存关系的识别、理解与分析》,IEEE Control Systems Magazine 21(2001 年)
- Scheffer and others, Early-warning signals for critical transitions, Nature 461 (2009)
- OECD 国家间投入产出表和增加值贸易数据库
- NS-3、NXS-GRX(规范数据、来源追溯和风险)记录 UNI,为拟议标准
- NS-12,软件、模型与 AI/ML 制品,拟议标准
问题与解答。
How is systemic risk different from ordinary risk?
Can systemic risk be measured with one number?
谁使用这项工作?
How do countries take part?
Can systemic risk be eliminated?
从本页贡献证据。
Reports 是联合体的季刊。您对该主题的了解可以成为下一期中的证据、案例或更正,署名归您,并在属于您本人作品时以 CC BY 4.0 发布。
- 电子邮件网关 GW-GCRI从本页发送证据、更正、案例、方法、审查、翻译或概况投稿。它会带上本页标签送达,由负责的编辑接收。
- ZenodoOpen depositDeposit it and submit to the communityPublish your own record on Zenodo with your ORCID iD, then submit it to the Reports community there. Curators review it; the DOI stays yours.
- 报告GCRI查看所有贡献方式贡献类型、各自的评审方式、您获得的署名、许可协议,以及绝不接受的内容。
切勿发送个人数据、敏感基础设施细节或您无权分享的任何内容。贡献在使用前会经过审查;发送贡献不赋予署名权或编辑权,赞助也绝不能换取研究结论。
法律文件。
本网站上的地图、数据和图表仅作为背景和一般信息展示。它们不是建议、预测或预警;信息的局限见免责声明和数据说明。