Home » The Danger of Mixing Up Causality and Correlation: Ionica Smeets at TEDxDelft (Transcript)

The Danger of Mixing Up Causality and Correlation: Ionica Smeets at TEDxDelft (Transcript)

Welcome. I came here today to warn you about the dangers of ice-cream.

You may not be aware of this, but these innocent looking cones full of sweetness are one of the major causes of drownings. And I’ve got the numbers to prove it.

So, if you plot a graph of the number of ice-creams that are sold, and you compare it with the number of drownings, you can see there is clearly an upwards trend, and I think it’s very safe to conclude from this that we should ban ice-cream because it’s very dangerous.

Since you’re all smart people, you’ve probably figured out there’s something wrong with my example. What’s really happening here is, of course, that there is an underlying factor, which is nice weather, you might have guessed it. And if the weather is nice, more people will go out swimming, and unfortunately drown, and at the same time, more people will buy ice-cream. And it’s not the ice-cream that’s causing the drownings. And here it’s really easy to see that there is something wrong, but jumping to an incorrect conclusion about causality when you see a correlation is the most often made logical mistake. And today my goal is to make sure that in the future you can recognize this mistake. And I really hope you can avoid making it in the future for yourselves.

And I’ll do this by just giving some famous examples. And the first one is really rather innocent. The fact is that married men live longer than single men. If you look at the statistics, you see that this is really happening. And women’s magazines, they like to conclude from this that marriage is very healthy for men, because it makes them live longer.

Well, a friend of mine, he likes to joke that marriage mainly makes life seem longer, but — that’s because his wife is — But so, can anyone guess what’s going on here? Because there is a causal relation, but it’s the other way around. The fact is that men who are healthy, and rich, and well educated, and have a much higher life expectancy, these are the men that are much more likely to find a wife — that’s the way women are — and the guys who have a very low life expectancy, so they’re unhealthy and poor, they are not as likely to get married. So it’s the high life expectancy that is causing the marriage, not the other way around.

Well, and this, of course, you know, it’s not so serious, no one will get married just because he read this. So let’s move to a more serious example.

It was also more serious research. In Nature there was a study in 1999 that showed that young kids who sleep with the lights on, that they have a much higher probability of becoming short-sighted later in life. But the researchers, they were smart, and they wrote very careful that they had found a correlation, and they didn’t know how the causal relation might work, but just to be sure they advised all parents to turn off the lights at night. And in the popular media this became that bed lamps were night abuse, children’s abuse, and that it was very bad if parents used lamps in the bedroom. And many parents were worried.

I can imagine, if this would have happened when my son was sleeping with the lights on, I would had felt really bad. But luckily, the article had to be corrected the next week, and maybe some of you can guess, and if there are biologists in the audience, they know Short-sightedness is genetic. And so, it’s parents who are short-sighted. And those are the parents who like to leave the light on in the bedroom, and they also are the parents who have short-sighted kids.

So again, a simple mistake, easy to make. Then, what is I think the worst example I know — I know many of them, I see at least one of these in the newspapers every week. But this is a classic one: in the ’70s, researchers found that there is a very strong link between kids who do well in school, get good grades, and kids who have a high self-esteem. And they concluded from this that it’s very important to make sure that young kids are, you know, raised to be confident and proud of themselves, because if their self-esteem is high, the good results will follow. And this forethought was told to parents, especially in the US, for generations, that just make sure that your kid is proud and confident, then all will turn out well.

And many years later, someone did another study just to see in which direction the cause was working. And they found that it was in the opposite direction. So the good grades were causing the self-esteem, and self-esteem wasn’t causing good grades. And it was even worse. So kids who are raised just to have high self-confidence, and not excel at anything — it can be sports or music, doesn’t have to be school — the kids who are just proud of themselves, and then fail at everything, in the end they will have a very low self-esteem, and not be able to make anything of their life.

So this was a very serious correlation mistake. And what I want for today is for you to remember that the next time someone wants to prove that there is a causal relation between something and something else — it can be anything, it can be vaccines and autism, it can be female bankers and the financial crisis, and if they point out to you that there is a very strong relation, remember that it’s not enough to have a correlation. It gives a very good hint of what might be happening, but before you can conclude that one thing causes something else, you need to know why it does and how it does.

So, when in doubt, just remember the ice-cream. Thank you very much.