Systemic risk along a chain

How a single failure hits several apparently independent positions at once.

This lesson has had no expert review. It was written for this platform and against the evidence it cites; nobody has gone through it independently.

Learning objectives

  • You can report direct, indirect and dependency exposure for a portfolio separately.
  • You can describe a cascade as a sequence rather than as an event.

Check your prior knowledge

Answer these for yourself before reading on. Wherever you hesitate is where this lesson pays off.

Core concept

Fact

A cascade is a sequence, not an event

It begins with a price move. That lowers collateral values, triggers liquidations whose sales lower the same price further — and with it the next round. Each step on its own is intended behavior. The risk is not in a faulty step but in the steps feeding one another.

Interpretation

Three kinds of exposure, to be reported separately

Direct is where the capital sits. Indirect is what that protocol itself uses. Dependency exposure is what several positions share. A listing showing only the first can present a portfolio as broadly spread when it hangs from a single point — and that is the normal case, not the exception.

Risk

Correlation arises under stress, not in calm

In calm markets, positions with shared dependencies often behave unremarkably differently. Exactly when diversification is supposed to count, they move together — because the shared dependency is also the transmission path. A correlation measured historically over calm periods therefore systematically fails to describe the scenario it is needed for.

Definitions

Cascade
A sequence in which each step worsens the condition for the next.
Contagion
Transmission of a failure to parties with no direct contractual relationship.
Dependency chain in the glossary
The sequence of components a position rests on.

Model

  1. The price of a widely used collateral falls

  2. Oracles pass the value to every lending market

  3. Positions near the threshold become liquidatable simultaneously

  4. Liquidation sales hit the same, already thin market

  5. The price falls further — back to step 2

A cascade in five steps — No step contains a fault; the problem is the feedback.

Formulas

Exposure per shared dependency

Exposure(dependency) = Σ position value across all positions that use it
dependency
stablecoin, oracle provider, bridge or chain
position value
capital deployed in the respective position

Limit: The sum is as complete as the underlying mapping. A dependency not established shows as zero and so looks like a checked absence — which is why every such listing needs its own “not established” row.

Worked example

One portfolio, two listings

By protocol
3 protocols, 1/3 each — looks spread
Shared stablecoin
3 of 3 positions
Shared oracle provider
2 of 3 positions
Shared chain
2 of 3 positions

Computed by dependency: stablecoin 100 % of the portfolio, oracle provider 67 %, chain 67 %.

The same positions, two very different pictures.

Reading: The first listing is not wrong — it answers a different question. For “what hits me simultaneously”, only the second is usable.

Retrieval

Why does a historically measured correlation systematically misdescribe the stress scenario?

Exercise on real data

Pick three markets from different protocols and build the second listing: by dependency rather than by protocol, with a row for what you could not establish.

Check three markets' shared dependencies →

Work through the case study and compare your answer with the written-out reasoning.

Case study: dependency chain →

Application

Design a stress scenario for a portfolio of three positions sharing one stablecoin. Which steps do you describe, and which number is the statement at the end?

Related case studies

Institutional reading

Bank
Would the simultaneously affected share be reportable as concentration risk?
Insurance
How does aggregation through shared dependencies affect the calculation?
Asset management
Which position would be reduced first to lower the shared dependency?

Metrics in this lesson

Key takeaways

Evidence