Why do failures cascade through power grids and finance?

A cascading failure spreads because the parts of a network rely on each other. When one element fails, its load, its function or its obligations pass to other elements, which may then fail in turn. Research on interdependent networks adds that failures can also cross between separate networks, such as a power grid and the communication system that controls it. [1] [2]

In a power grid, a failed element shifts its load to nearby elements. Those elements can be pushed beyond their capacity, become overloaded and pass their load on to others. This is common in high-voltage systems, where a single point of failure on a fully loaded or slightly overloaded system can cause a sudden surge across the nodes. The surge can push already overloaded nodes into failure and take down the whole system very quickly. [1]

The August 2003 outage in North America shows the pattern. It began with a high-voltage line failing in Ohio. Because utilities share power through the grid, the outage spread through southern Canada and the northeastern United States, leaving as many as 50 million people without power, some for days, with about $6 billion in losses. [3]

A review of cascade mechanisms lists several origins: load redistribution that causes overloads, direct dependencies in which a node fails when something it relies on fails, and threshold rules in which a node fails if too few of its neighbours still work. The review notes that overload models resemble dependency models, since a node that fails from redistributed load effectively depends on the node that failed first. [2]

Interdependent network models separate connectivity links within a network from dependency links between networks. A node failing in one network disables its partner in the other. Pieces cut off from the main connected cluster also shut down, and this feeds failure back and forth between the networks until the system settles. In such a model, when the initial damage reaches a critical level, both networks collapse. [2] [4]

A study in Nature found that a failure of a very small fraction of nodes in one network can fragment a system of several interdependent networks. It also found, against what holds for a single network, that a broader degree distribution makes interdependent networks more vulnerable to random failure. The authors cite the Italian blackout of 28 September 2003, in which power station shutdowns disrupted the Internet network, which in turn caused more power stations to fail. [5]

The collapse can be abrupt. In isolated networks the breakdown is a continuous transition, while in interdependent systems it is a hybrid transition. Near the transition point the system is metastable, so very small extra damage can destroy a system that otherwise looks stable. Spatial embedding of networks also reduces their resilience to failures. [2] [6]

Finance relies on a web of electrical, technical and financial networks, each exposed to its own threats but dependent on the others. The US Office of Financial Research describes three channels by which operational incidents can threaten stability: disrupting a firm that provides critical services, reducing confidence in firms and markets, and damaging the integrity of key data. It also notes that one firm's default can set off defaults at other firms through the network of exposures. [3]

Shared holdings add another path. Where multiple banks hold the same set of assets, coupling between institutions rises, which increases the likelihood of systemic failure. [4]