#52 - Single-variate vs. Multivariate Models

Why the world runs on many variables, and why we stubbornly insist on using one

Last time we established that IQ is real (post 51 - IQ is Real), measurable, predictive, and normally distributed. But we also hinted at something important: IQ is not the only variable that matters. Today, we formalize that idea.

Everything we deal with in the real world is multivariate. Your health, your income, your happiness, the success of a business, the stability of a country... each is the product of many interacting factors, not one.

This is not a particularly controversial claim when stated plainly. And yet, most of public discourse, policy debate, and even academic research operates as if single variables can explain the world. This is the single-variable trap, and almost everyone falls into it.

The Single-Variable Trap

Humans are wired for simple stories. We covered this in our discussion of simplification (post 45 - Simplification). Our brains naturally compress complexity into manageable narratives. That instinct serves us well in daily life, but it becomes a serious liability when we’re trying to understand complex systems.

The single-variable trap works like this: you identify one factor that is genuinely correlated with an outcome, and you treat it as the explanation. Everything gets filtered through that one lens. “It’s all about education.” “It’s all about income inequality.” “It’s all about IQ.” Each of these captures something real, but someone who reduces all of life’s complexity down to any single factor has found a real signal and mistaken it for the whole picture.

What Multivariate Actually Means

In statistics, a multivariate model is one that considers multiple independent variables simultaneously to explain an outcome. The power of this approach is that it allows you to see how much each variable contributes after accounting for the others.

This is critical. Many variables that look powerfully explanatory on their own shrink dramatically once you control for other factors. And some variables that seem unimportant on their own turn out to matter a great deal once other noise is removed.

A practical example: imagine you’re trying to predict a student’s college GPA. SAT scores alone show a decent correlation. High school GPA alone does too. But if you look at both together, along with conscientiousness, family stability, and whether the student is working a part-time job, you get a much richer and more accurate picture. Each variable adds something, and none of them alone captures the full story. This is how reality actually works.

Why We Resist Multivariate Thinking

If multivariate thinking is more accurate, why don’t we default to it?

Simply put: it’s harder. Holding multiple factors in your head and reasoning about their relative contributions is genuinely more cognitively demanding than latching onto one compelling explanation. Our brains find a pattern (post 37 - Pattern Recognition & Intelligence) and want to stop there.

Second, it makes for worse stories. “This problem is caused by a complex interaction of seven factors, three of which we can partially influence” doesn’t grab attention nearly as much as “eggs cause cancer!” The incentives of media and politics actively punish multivariate explanations.

Lastly, it undermines moral clarity. Single-variable explanations give us built-in heroes and villains. Multivariate explanations distribute responsibility, which makes righteous indignation harder to sustain. And for many people, the indignation is the point.

This tendency is rampant across the political spectrum. The Left gravitates toward systemic single-variable explanations (racism, patriarchy, income inequality). The Right has its own versions (culture alone, individual responsibility alone, government as the singular cause). Each captures something real. None alone explains the complexity of what we observe. The truth, as is almost always the case, requires you to hold multiple factors in tension. Anyone telling you otherwise is either simplifying (post 45 - Simplification) for convenience, selling you a narrative (post 20 - What is a Narrative), or genuinely doesn’t understand the problem.

Thinking Multivariately

Multivariate thinking isn’t natural, but it can be practiced. When you encounter a compelling single-variable explanation, ask: “What else could be contributing?” Force yourself to name at least two or three other plausible factors before accepting any single one as dominant.

When someone presents data showing a correlation between one variable and an outcome, ask: “What happens when you control for other factors?” This is the difference between a naïve correlation and a meaningful one. We’ll explore this more when we cover soon with our post on correlation and causality.

And when a policy proposal is built around a single lever, be skeptical. Real-world outcomes almost never hinge on a single intervention.

Connecting the Dots

Over the last few installments we’ve built a toolkit: normal distributions (post 50 - Normal Distributions) show us how traits are actually distributed, IQ shows us that cognitive ability is one such measurable and predictive trait, and now we see that no single trait, however real, can explain an outcome on its own.

Next time, we’ll look at what happens when people take the single-variable error to its logical extreme: the curse of monocausality.