Earlier this week, I shared a curious finding:
Across twelve years of ArcelorMittal data, we found that higher groundwater extraction tended to accompany lower profits. The correlation? -0.26.
It raised an obvious question: Is water trying to tell us something?
In today’s audio post, I went deeper into the story behind that analysis—both technically and personally.
Here’s the outline for those who prefer reading to listening:
The Question Behind the Correlation
Water is often treated as a passive resource in financial models. But what if it’s an active signal?
The observed correlation between water use and profit wasn’t random—it was statistically significant. But as any analyst knows, correlation isn’t causation. So we asked: What causes what?
From Coincidence to Causal Hypothesis
We explored whether:
Water overuse drives profit down,
Lower profits cause increased water use,
Or whether a third factor—like climate, inefficiencies, or regulation—is at play.
Surface-level models stop at “what moves together.”
Causal thinking digs into why they move together.
Steelmaking as a System
Water isn’t just a cost input—it’s an operational lifeline.
From cooling furnaces to controlling emissions, its role is fundamental. And steel factories exist within broader social-ecological systems:
Local water laws
Regional politics
Climate volatility
Community pressure
When a resource becomes volatile, sustainability becomes financial.
What We Do at Wangari
Instead of stopping at correlation, we apply causal inference tools:
Causal graphs to map hypotheses
Temporal and counterfactual analysis to test directionality
Controls for confounders like commodity prices, input costs, and environmental conditions
The goal isn’t to prove one arrow—it’s to understand the system so decision-makers can act more wisely.
Why We Didn’t Sell to the Banks
Some major investment banks wanted to buy our correlation work early on.
We walked away—not because it wasn’t valuable, but because it wasn’t deep enough for what we were building.
At the time, I’ll admit, I still had some limiting beliefs about money—beliefs rooted in early life, not in strategy. But moving through those opened the door to something more fulfilling.
Instead of becoming correlation calculators, we’re building systems that let water—and energy, and land—speak causally.
Who This Is For
We’re still exploring that question.
Industrials? Hedge funds? Water agencies? Agri coops?
We’ve had promising conversations in all these sectors.
What we know is this:
The more complex the system, the more causal clarity is needed.
And the more volatile the resource, the more financial the sustainability risk becomes.
Coming Friday: The Technical Walkthrough
This Friday, we’ll walk you through:
How we modeled the ArcelorMittal case
What tools and code we used
What we found—and didn’t find—when testing for causality
It’s our first public look at how we apply causal inference to sustainability data in practice. And we’re excited to share it.
Until then, here’s one thing to try:
Next time you build a model, ask yourself not just what moves—but why.
Because sometimes, even water has something to say.
Share this post