Separate Correlation From Causation

Use to understand what truly needs to be addressed to solve a problem.

In efforts to understand why certain events happen, we must go looking for instigating factors. It's only logical that we try to find a previous event directly responsible for causing the event we're looking at. This is what we should spend our time trying to fix, but it turns out that we might be spending all of our time on the wrong issue. We're fooled into confusing correlation for causation. One of this mental model's shining examples follows.

Say you're looking at a graph that shows two data comparisons - one axis shows the total number of sunglasses sold over a period of time, and the other shows the total sales of ice cream. During the summer months, you note that sales of both items increase and that they tend to go down after summer is over.

Looking at this graph, you might come to the conclusion that sales of ice cream directly impact sales of sunglasses. People are buying more sunglasses because they're buying more ice cream - or the other way around. No matter the direction, it appears that one is causing the other.

Why might this be the case? Is it because there are stores that sell both ice cream and sunglasses? Is there something about buying a sundae or root beer float that triggers one to grab a pair of Ray-Bans immediately after? Do sunglasses press on a facial nerve that triggers thirst?

These theories sound ridiculous, don't they? That's because they are.

When you first read the example, you probably figured out that sales of ice cream and sunglasses increased due to the arrival of summer. Since there are more hot and sunny days in summer, people are more inclined to buy cold treats like ice cream and protective eyewear like sunglasses. People don't buy sunglasses as a direct result of ice cream purchases - they buy both when the summer heat hits them. Just because two things occur simultaneously doesn't mean there is a relationship between them.

Even though that's a pretty broad example, it reflects a logical error that lots of people make - sometimes about matters even more elementary and basic than ice cream and sunglasses. That error is believing that since two events have similar patterns or related behaviors that one must be causing the other to happen. This is the mistake of believing that correlation implies causation. In fact, they are entirely separate concepts.

Correlation is a statistical term. It shows that two individual elements or variables share similar traits or trends - "ice cream and sunglasses sales both increased." That's all there is to correlation: two things behave similarly in this way or that way. Correlation does not describe why or how the relationship between two items is the way it is; it doesn't give a reason. It just says, "These two things are generally doing the same thing at the same time."

Causation, on the other hand, is an effort to establish the reason things happen - also referred to as "cause and effect." The message of causation is: "This thing changed, which in turn caused this other thing to change." In our super-basic example, the thing that actually caused the increase in sunglasses revenue was the arrival of summer, which was also responsible for the increase in ice cream sales. There was a causal relationship between summer and sunglasses and summer and ice cream, but there's only a correlative relationship between sunglasses and ice cream.

To believe that the increase in ice cream sales caused the increase in sunglasses sales is a logical mistake. This is countered by the phrase correlation doesn't imply causation - just because two events are similar doesn't mean one is causing the other one to happen. There may be another underlying factor that's causing both things to happen.

This error in thinking usually happens when there's a lack of information at our disposal - or, perhaps more frequently, when we don't take the time to observe all the information we should. Jumping to conclusions is always a temptation when we feel under pressure to come up with a definitive answer. In order to avoid that fallacy, one should identify as many potential factors as one can: research, study trends, gather more data, and make reasonable, unhurried judgments.

In a lot of cases, correlations are nothing more than flukes or chance, yet we rapidly jump to causal thoughts. When evaluating cause and effect, the default mental model should always be to separate correlation from causation and not assume a causal relationship unless you can definitively say so.

There's one more wrinkle when it comes to discussing cause and effect. It's a bit more complex than we're led to believe as children, when we're taught that pushing on a toy truck will make it move.

As we gain more life experience, causal factors become a little more complex. There are more conditions, underlying motives, and elements that affect events. Sometimes it's hard to point to a singular cause, because it's hard to say that it acted alone or wasn't the product of multiple mini-causes.

This process involves looking past the immediate reason things happen (the proximate cause) and searching for certain greater, more fundamental basis that things happen (the root cause). The proximate cause is to the root cause as correlation is to plain causation. Solving for the former (proximate cause; correlation) won't rid you of your troubles.

For example, say somebody's driver's license gets suspended. Let's call him Hal. Traffic court has been waiting for Hal to respond to a series of speeding violations, but he's never complied. A warrant for Hal's arrest is issued; the police go over to his home, bust down his door, and throw him in jail for a long weekend.

At this point, we can ask the question, why is Hal in jail? Well, he's there because police were acting on an arrest warrant that said he needed to answer for multiple speeding violations. This is the proximate cause: the most recent, basic actions that led to Hal being thrown into the slammer.

But the proximate cause doesn't explain the deeper issues that have led to Hal's being in jail. You could say the arrest warrant was issued because Hal's a lead-foot who needs to lighten up on the accelerator pedal. So you could consider Hal's need for speed to be the root cause.

But is it?

One can keep going down a rabbit hole to find out why Hal is this way, and you could continue to consider each new level a more root cause. If he's going to change his ways, simply telling him to stop speeding so much might not be effective. What's causing him to speed? Maybe his parents never taught him restraint in certain situations; they just let him dart around the house and make a mess of things, and that recklessness followed him into adulthood. At this point, Hal has a deeper root cause - some have called this level arriving at the ultimate cause. Unless Hal deals with the emotional basis for his speeding habit, there's a great chance he'll re-offend. If he blows them off and just blames "the man," he hasn't learned anything.

This is the proximate/root cause portion of this mental model in a nutshell. It's a more critical and profound way of discovering the real answers and explanations for events. Quality thinking means going past the proximate cause - which is usually just a physical sequence of cues - and understanding the factors, thinking or emotional patterns, or environmental elements that set the groundwork for something happening.

It might help to imagine each set of actions as motivated by something psychological. One way of putting this discovery plan into action is the "five whys" method, which is simply asking "why" five times to establish a deeper root cause:

Why is Hal in jail? Because there was an arrest warrant out for him (proximate cause).

Why? Because he hadn't responded in court to his multiple speeding violations.

Why? Because he exceeded the speed limit nine times and got caught.

Why? Because he has a "need" or impulse to go super-fast on the highway.

Why? Because he never had a set of boundaries as a child and thought he could do whatever he wanted without consequences.

Differentiating between proximate and root causes makes one keep going in the discovery process - whereas, left to one's own instincts, they might just stop asking once they identify the immediate cause or even when they see a vague correlation. By going deeper you'll get a better understanding of why things happen and be better positioned to deal with problems.