Decisions Are the Pinch Point

I saw a billboard the other day promoting eating vegetables “amid cost crisis.” The implication: people eat badly because healthy food is expensive.

Is that true? Not really. Rice, beans, frozen vegetables, eggs, bananas — these are among the cheapest foods per calorie and per nutrient. The real objection isn’t cost. It’s something like: “But that’s not a satisfying meal.” OK — but now we’re being honest. You’re not choosing the burger because vegetables are expensive. You’re choosing the burger because you prefer the burger. When you adjust for taste, texture, convenience, and immediate satisfaction — yeah, the unhealthy option wins. That’s the whole point. That’s why you’re eating it.

The mechanism isn’t cost. It’s a decision.

The Shape of the Problem

Pick your favourite explanation for why low-SES populations have worse health outcomes. Poverty. Low education. Poor health literacy. Food deserts. Intergenerational trauma. “Social determinants of health” — the term people reach for when they want to gesture at systemic factors without having to be specific about any of them.

I’m not saying those factors don’t exist. They do. But notice: none of them directly cause disease. They influence the thing that does.

The shape looks like an hourglass on its side:

Many causes on the left. Many outcomes on the right. Everything passes through the same pinch point in the middle.

The bold lines trace one example: poverty and food environment shaping someone’s food decisions, producing obesity and chronic disease. The faded lines are paths that could activate but didn’t — because the decision went a particular way.

The point isn’t that poverty doesn’t matter. Poverty matters because it influences decisions. It’s not the terminal cause. The decision is.

So What?

Once you name the mechanism, you can do something about it.

If the problem is really “healthy food is too expensive,” the solution is subsidies or price controls. We’ve tried those. They don’t work well, because cost was never the real barrier.

If the problem is “people are making bad food decisions” — now you’re somewhere useful. Why are the decisions bad? What information is missing? What framing are they operating under? What would need to change for them to choose differently?

“Social determinants” is the term people use to avoid asking these questions. It lets you point at systemic factors and stop there, as if the system directly causes the disease without passing through any individual agency. It’s comfortable because it removes blame. But it also removes the lever. If nobody’s making a decision, where do you intervene?

This works at the personal level too. Self-improvement that doesn’t change your decisions is theatre. You can read about nutrition, understand the biochemistry, follow the right accounts — but if none of that changes what you actually do, nothing has improved. The only test that matters: are your decisions different?

Not Just You

Yes, there are factors outside your control. Where you grew up, what your parents taught you, what’s available at the shop on your corner. Those are real inputs. I’m not pretending otherwise.

But they’re inputs to a decision. And the decision is the pinch point.

It’s also not just your own decisions. A kid’s nutrition is their parents’ decisions. Your diet is partly what your partner cooks. Your exercise habits track whether your friends are active or sedentary. The pinch point includes the decisions of people close to you.

People are unhealthy because of decisions. Not exclusively. Not without context. But mechanistically — yes. That’s where it happens.

Once you accept that, the conversation gets productive: how do we help people make better decisions? Better information, better framing, better defaults. Make the good choices easier and the bad choices harder.

That’s the only framing that actually leads somewhere.

Appendix: The Causal Structure

If you’ve done epidemiology or causal inference, you already see it: the hourglass diagram is a structural claim. Here it is formally.

The wrong DAG

The “social determinants” framing implicitly draws this causal graph:

Poverty causes obesity. Done. No mechanism specified, no point of intervention identified, no individual agency in the model. This is how most public health literature treats the relationship — an arrow straight from social circumstance to disease.

The right DAG

The correct graph has a mediator:

The dashed line is the direct effect of poverty on obesity — the part that doesn’t pass through decisions. My claim is that this arrow is negligible. Poverty doesn’t give you diabetes. Poverty influences the decisions that give you diabetes.

In causal inference terms: decisions are a mediator — and not a partial one. The total effect of poverty on health runs through decisions. Nearly all of it is indirect effect.

Why this matters for intervention

Judea Pearl’s do-operator makes this concrete. do(good decisions) — intervene directly on the decision node — and you block the causal path from poverty to health outcomes. Poverty still exists, but its path to disease is cut.

“Social determinants” focuses intervention on the left side of the graph — reduce poverty, improve education, fix food deserts. All indirect. They work only if they change decisions. Target the mediator directly.

Colliders and conditioning

One subtlety. If decisions are caused by multiple upstream factors (poverty, education, genetics, culture), then decisions are a collider — a node with multiple parents. Conditioning on a collider (studying only people who made the same decision) creates spurious associations between its causes. This is why studies that “control for behaviour” can produce misleading results — you’re conditioning on the collider, which opens spurious paths between the upstream variables.

The right approach isn’t to control for decisions. It’s to understand how upstream factors flow through decisions and intervene at the decision point.

The formal claim

Stated precisely: for most modifiable health outcomes, the direct causal effect of socioeconomic variables (bypassing behavioural decisions) is approximately zero. The observed association between poverty and disease is almost entirely mediated by decisions — the person’s own, and those of the people around them.

This isn’t a controversial claim in the data. It’s controversial in framing. The data shows that behaviour mediates most of the SES-health gradient. The controversy is whether you’re allowed to say so.