Correlation vs. Causation
Correlation can cause bad decisions
January 1, 2021
I suspect that many of you, perhaps all of you, have heard something about correlation versus causation, e.g., “Correlation doesn’t mean causation.” And that’s true. What I hope to impress upon you in this missive is that this fact has much wider application than you might think, in sometimes subtle ways. Three examples follow.
First, consider the now-discredited belief that MMR (measles, mumps, rubella) vaccinations cause autism.(1) Why might someone believe this to be true? Answer: some children start showing autistic characteristics after they were vaccinated. This is an example of “Post hoc ergo propter hoc” or, more simply, the post hoc fallacy. Just because one thing follows another doesn’t mean that there is causation. But nor does it mean that there isn’t causation—further investigation is needed. For example, what is the rate of autism in children who don’t get vaccinated? Turns out it is about the same as the rate for vaccinated children. Vaccinations and the first signs of autism arise at about the same age, but there is no causative relation between the two.(2) The consequences of the belief that vaccinations cause autism: many more cases of measles, a potentially deadly disease.
Second, it is widely believed that an authoritarian parenting lifestyle leads to children with more troubles: they (the children) have lower social skills, are more likely to be engaged in bullying, and are less resourceful, to name a few.(3) The seemingly obvious conclusion: Parents, don’t behave in an authoritarian manner.
Now imagine for the moment that you have a child who is rather unruly. It seems likely that you might be rather strict with him. This is not a question of punishment per se; you just want him to have the skills to succeed in the world. A researcher will see a correlation between your parenting style and the behavior of your child.
What is really going on is what Judith Rich Harris calls, “child-to-parent” effects.(4) In this case, there is causation, but it runs contrary to the way ‘round, from child to parent, not parent to child. Of course, the causation can run both ways; the point is that the seemingly obvious interpretation of a causation might be wrong, or at least more nuanced than at first glance. Maybe authoritarian parenting is called for in some circumstances.
Third, consider the positive relation between health and lifetime income. It is easy to imagine the causal arrow pointing both directions here—better health means that you can work hard or longer (and hence higher income), and higher income means that you can afford better health care.
Yet there may be an underlying factor that drives both good health and lifetime income—your parents. By “your parents,” I mean the genes they pass on to you as well as environmental influences while you were growing up. (There is an ongoing debate about the relative importance of these two influences, but they’re both important.) In this case, there is a third factor that drives the correlation between the two variables. That is, they are correlated, but not because one causes the other. A possible danger: You worker harder, believing that it will make your healthier. In the absence of causation, you may suffer from worse health, not better.
These three examples differ from each other. The first involves a correlation without any causation. The second involves causation running in the opposite direction from what might seem obvious at first. And the third involves causation arising from a “push” from an underlying, third factor.
Mistaking correlation for causation has not been widely identified as a cognitive bias, but perhaps it should be. An immense amount of doubtful science has been based on false connections between the two, over the centuries and even today. Yet somehow its powers to persuade persist, not just within the halls of science but across nearly every field of human endeavor.
I suspect this is the case because in large part the human mind, for both good and ill, is a pattern-seeking device.(5) Hence the provision of even a whiff of a pattern at the center of a hypothesis and its subsequent conclusion often is enough to create a perception of credibility. Papered over with enough layers of jargon and dubious data, many such correlation-causation claims have found a home in the world. It’s the job of true scientists, not to mention responsible adults in general, to see through these false patterns on their way to discovering truth.
By the way, and on something of a lighter note, I have noticed a strong causal (?) correlation between people who read this newsletter and an increase in critical thinking. My data are informal admittedly, but I’ve noticed a definite pattern. Maybe, however, you were on that path already.
(1) The data in the study that purported to show the vaccination-autism link were carefully selected (shall we say). The doctor behind the study has been barred from practicing medicine.
(2) If anything, children with autism are less likely to be vaccinated. https://time.com/5213493/autism-vaccination-rates/
(3) For a quick summary, see https://www.parentingscience.com/authoritarian-parenting.html#:~:text=Emotional%20problems,self%2Desteem%2C%20and%20depression%3F&text=Other%20research%20in%20China%20suggests,(Muhtadie%20et%20al%202013).
(4) Harris, Judith Rich, The Nurture Assumption: Why Children Turn Out the Way They Do, Revised and updated, Free Press, 2009. I also enjoyed her other book, No Two Alike: Human Nature and Human Individuality,” W.W. Norton & Company, 2006.
(5) See Chapter 4 of my book, The Foolish Corner: Avoiding Mind Traps in Personal Finance Decisions, Stuart Charles Group, 2017, available on Amazon.