If you don’t read my blog, you will die.

That caught your attention, didn’t it?

And on the face of it, it’s true and backed up by facts and statistics:

  1. About 150,000 people died worldwide yesterday.
  2. Of these, none had read my blog yesterday.
  3. The few people who did read my blog yesterday are still alive today.

Conclusion: Reading my blog keeps you alive.

“Damn. We should have read Andreas Moser’s blog.”

Or does it? You immediately sense that a few things are fishy here, even if you believe that I know each one of my readers personally and know that they are alive: Of the people who died yesterday, many had no access to the internet. If they also live in countries with higher mortality rates, their chance of death is of course higher than those of the bulk of my readers who are in the USA and Europe (see the flag counter on the right hand side). This increased mortality risk might have less to do with lack of wifi, but with lack of food or the danger of diseases or with a war. On the other hand, of the people who read my blog, some might be couch potatoes who never venture outside. They don’t have the risk of car accidents, of being shot or struck by lightning. But again, the reason for their longevity is not reading my blog, the reason is staying inside.

You see that just because two events (reading my blog and staying alive) happen at the same time or to the same people, there is not necessarily a causal connection, as long as there are other possible more causal connections (like the country where you live or the dangers that you are exposed to, in this example).

Correlation is not causation!

In fact, correlation does not even imply causation.

With a silly example like the one I have chosen, this is obviously clear. But keep this sentence in mind when you read or hear about the latest “new study” that tries to show a link between something and cancer or something and crime or something and divorce. This is especially true for the social sciences, because human interactions are so complex that they will rarely be open to monocausal explanations.

A few more examples that we come across regularly:

  • If a study suggests that married people are happier, this is presented as proof that marriage is a way to happiness. However, it could simply mean that happy people find partners more easily than grumpy ones and therefore get married more often.
  • Children who watch TV or play computers too much, have psychological problems. Parents use this as an argument to cut down on your computer use. In fact, it might be interpreted the other way round: Children who already have psychological problems don’t want to interact with other children and thus prefer to sit in front of a screen.
  • Babies who have been breast-fed will have a higher IQ. Proponents of breast-feeding will cite this again and again and probably even petition for tax relief for breast-feeders. However, if we look at the group of breast-feeding mothers, these might be the mothers who generally take more time for their babies, have a closer emotional bond, will be more supportive and will also read books to the child. All of which might have a greater impact on the IQ than where the milk comes from.

To be clear: What these studies pretend to show, could be true. But it could also be false. Even the opposite could be true. – Without the context of many other studies, they just don’t tell us anything of value.

More critical thinking, please!


About Andreas Moser

Travelling the world and writing about it. I have degrees in law and philosophy, but I'd much rather be a writer, a spy or a hobo.
This entry was posted in Economics and tagged , , . Bookmark the permalink.

10 Responses to If you don’t read my blog, you will die.

  1. John Erickson says:

    Then again, if you stuck with your original concept, you could have a long, prosperous career in politics, especially here in the US. Based upon recent events, it’s not what you know, but how long and strong you are willing to stand BEHIND what you don’t know! (For evidence, I submit John Boehner and Donald Trump.)
    Or, to paraphrase an old Latin statement, “I think, therefore I cannot be a politician.” :D

  2. Sonia M. says:

    It amazes me how often folks get the two confused. Sometimes makes it very difficult to talk rationally.

  3. Great blog and important too! I would suggest an extension as for a correlation to become a causation it is necessary that you have a theory that explains why A should be the cause for B. While the requirement of a theory will not increase a correlations probability to become a causation, it will root out the most obvious nonsense, simply by asking:
    Why should married people be happier than unmarried people?
    Why should children who watch TV too much have psychological problems?
    And why should breast feeding influence intelligence (and not the other way round?)…

    By the way, your syllogism is not correct (the middleterm is not identical) and you make a fallacy in your conclusion (affirming the consequence, because there are still people alive that did not read your blog), however, this does not affect the argument.

  4. oros says:

    Loved this. Just wish you’d credited the comic http://xkcd.com/552/ :)

  5. Maybe I’ve read too much David Hume, but if correlation doesn’t even *imply* causation, how do we ever infer causation? Sure, correlation does not prove causation, but how do we ever come up with theories of causation without making inferences based on observed correlations? There doesn’t seem to me to be anything logically wrong with taking the correlation between rotted meat and flies as implying that rotted meat causes flies. It’s not until further observation shows that the correlation breaks down that the causation is invalidated.

    I just don’t understand what it means to deny that correlation even implies causation. If no or negative correlation is found, doesn’t that imply a lack of causation?

    • GradStudent says:

      Causation is shown through controlled experiments, rather than data mining or nonexperimental studies. The difference between the rotten meat/flies example and the way most people look at correlations is the next step – the further observation that shows the correlation breaks down. Most correlations thrown about aren’t easily fit into that paradigm on an individual scale – it’s hard for an individual to collect enough data about breastmilk and IQ, for instance.

      Additionally, you can have causation without correlation, as correlation tests for a linear relationship only. There are several examples of this on Wikipedia: http://en.wikipedia.org/wiki/File:Correlation_examples2.svg. The bottom row shows many examples of a causal relationship that produces a correlation of zero.

  6. Pingback: How people found my blog | The Happy Hermit – Andreas Moser's Blog

  7. Pingback: Chocolate makes smart | The Happy Hermit

  8. Pingback: Schokolade macht schlau | Der reisende Reporter

Please leave your comments, questions, suggestions:

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s