The Market Is Pricing a Variable. What's Actually Happening Is a Cascade.
Why correlation-based risk models are blind to the crisis unfolding right now — and what that means for every CFO and board in the world.
Early this week, I received a message from a colleague — a Head of Analysis at a major energy intelligence firm, someone I regard as one of the sharpest minds working at the intersection of geopolitics and global supply chains. He was in New York, coming out of a series of high-level meetings, and what he wrote stopped me cold.
I'm sharing it here — fully anonymized — not because it's alarming (though it is), but because it perfectly illustrates a fundamental failure in how organizations model and price risk. A failure that will cost some of them dearly.
"By my firm's math the world is short several million barrels of oil today, Asia has lost a large chunk of its natural gas supply, and about 30% of fertilizer trade is stopped. China has stopped exporting gasoline, diesel, and jet fuel. Asian refineries are running at 60% of normal and some will shut down very soon. Chemicals value chains are already shutting down — so basic materials for everything get more expensive and more scarce."
He went on: Oil at $90/bbl signals concern, but only with the embedded assumption that the Strait of Hormuz reopens soon. If it stays closed for a month, he said, oil goes well over $100/bbl. And then he asked the question that no model can answer: "Who do we negotiate with in Iran? Who can you trust? Why would they trust us?"
That last question is not an economic question. It's a causal intervention question — and standard risk models have no apparatus to address it at all.
What the Market Is Actually Saying
When the energy sector prices a scenario, it's doing correlation math. It's pattern-matching against historical disruption events and assigning a probability-weighted outcome. At $90/bbl, the market is essentially saying: this disruption is real, but it resolves within a normal historical timeframe.
That assumption might be right. But it rests on a foundation of independent variables that are no longer independent.
Here's what my colleague described, translated into causal language:
The Strait closure is not a single event. It is a causal trigger activating multiple downstream chains simultaneously — energy supply, natural gas flows, refinery capacity, chemical feedstocks, fertilizer trade, and liquid propane for cooking and heating across Asia and Africa. These are not separate risks that happen to coincide. They are causally connected. Each one amplifies the others.
"The world is short several million barrels of oil today" is not a supply-demand imbalance. It is a causal cascade in motion.
The Problem With Correlation-Based Risk Pricing
Standard enterprise risk models — whether in financial services, energy, manufacturing, or strategic planning — are built on correlation. They identify historical patterns, assign probabilities, and price risk accordingly. This works reasonably well in stable environments where the underlying causal structure doesn't change.
But we are not in a stable environment. We are in an environment where multiple independent causal roots are converging simultaneously. And that is precisely what correlation models are structurally blind to.
Fiduciari’s Grey Swan framework addresses this directly: what markets call tail risks are not rare statistical events. They are what happens when multiple causal roots converge faster than standard models can detect. The rarity isn't in the event — it's in the model's inability to see the convergence coming.
The $90/bbl assumption is a perfect example. It prices one variable — the duration of the closure — while treating the downstream cascade as either independent or recoverable. But fertilizer trade doesn't restart the moment the Strait reopens. Chemical value chains don't spin back up overnight. Refineries running at 60% capacity create their own downstream constraints. The causal damage accumulates even after the triggering event resolves. Time lag knows no master.
The Intervention Point Problem
My colleague's most penetrating observation wasn't about oil prices. It was this: "And how do we reopen the Straight for shippers? Who do we negotiate with in Iran? Who can you trust? Why would they trust us?"
This is the question that exposes the deepest limitation of correlative risk analysis. Correlation models can tell you what historically follows from a disruption. They cannot tell you where in a causal system you can actually intervene to change the outcome.
Identifying leverage points — where intervention produces the desired effect — requires understanding causal structure, not statistical patterns. In a world where the key negotiating parties distrust each other and the institutional architecture for resolution has been eroded, the historical resolution patterns are simply not applicable. The causal environment has changed.
This is what my colleague meant when he said the US risks becoming a pariah nation for "recklessly blowing up the modern economy." He's not making a political point. He's making a causal point: the interventions that created this situation may have permanently altered the causal pathways through which resolution normally occurs.
What This Means for CFOs and Boards
If you are a CFO or board member reading this, the practical implication is not "buy oil futures" or "hedge your energy exposure." It's a harder question: does your organization's risk framework have the capacity to see this kind of cascade coming — and to identify where you can actually intervene?
Most don't. Not because the people are unsophisticated — my colleague's firm is full of brilliant analysts who are reasoning causally in real time — but because the official models they produce are still correlation-based. There is a systematic gap between what the smartest people in the room know intuitively and what the models say officially. That gap is where organizations get blindsided.
This is precisely the problem that causal AI is designed to solve. Not as a replacement for human judgment — the kind of geopolitical intuition my colleague demonstrated is irreplaceable — but as a formal apparatus that can represent causal structure, identify convergent risks before they become cascades, and map the intervention points where action actually changes outcomes.
The fiduciary question is becoming unavoidable: if your risk models are structurally incapable of detecting causal cascades of this kind, and one occurs, what is the board's accountability exposure? Delaware courts are already moving in a direction that treats decision process quality — not just outcomes — as a fiduciary matter. The SEC is examining AI and implicitly HI decision governance in its 2026 examination priorities. The regulatory and legal environment is catching up to what thoughtful risk professionals already know.
The question is no longer whether causal AI is a competitive advantage. It's whether operating without it represents a fiduciary liability.
A Final Note
My colleague ended his message with something that struck me as both sobering and honest: "We still have the capacity to produce the materials that have been disrupted. So maybe there's a ton of money to be made, but the human suffering could be very significant."
That tension — between the opportunity that disruption creates and the human cost it carries — is exactly why the quality of organizational decision-making matters so much right now. Organizations that can reason causally will navigate this environment with clarity. Those that can't will be navigating with a map that was drawn for a world that no longer exists.
To work, the GPS has to know where you actually are.
—
Mark Stouse is the CEO of Proof Causal.ai, which builds causal AI solutions for enterprise decision-making and governance.