We always seem to be on a search for cause.
In college my statistics professor’s favorite expression was “correlation does not imply causation.” In case you’re not familiar with the phrase, I’ll borrow the explanation I learned in school:
When male college students wake up with a headache, a large percentage of the time they are still wearing their shoes. Therefore, sleeping with shoes on is correlated with waking up with a headache. However, it would be incorrect to conclude that sleeping with shoes on causes headaches. The more likely explanation is that both are caused by going to bed drunk. Correlation is not the same as cause.
While the example is vivid, it’s not particularly practical in distinguishing between correlation and cause. It’s unlikely anyone thought shoes caused the headaches. However, we tend to confuse correlation and cause frequently. In fact, most of the research studies I read are designed to test correlation even though the popular press report these correlations as cause.
For example, many studies have shown that eating breakfast is correlated with success in elementary school. This has turned into the popular wisdom that eating a good breakfast causes students to be better learners. However, this isn’t really the case; students who don’t eat breakfast are also more likely to be absent or tardy which in turns causes their poor performance. As it turns out, breakfast only helps undernourished children perform better.
Here’s a useful explanation from Khan Academy:
As humans, we gravitate towards order and explanations of the unknown. This means we are constantly on a search for cause, even if one doesn’t exist. Erik Hollnagel explains in his book Barriers And Accident Prevention:
Whenever an accident happens there is a natural concern to find out in detail exactly what happened and to determine the causes of it. This trait of human nature is so strong that we try to find causes even when they do not exist, such as in the case of misleading or spurious correlations. One very good reason is that we have created a way of living that depends heavily on the use of technology, and that technological systems are built to function in a deterministic, hence reliable manner. A second reason is that our whole understanding of the world is based on the assumption of specific relations between causes and effects, as amply illustrated by the Laws of Physics. A third reason is that most humans find it very uncomfortable when they do not know what to expect, i.e., when things happen in an unpredictable manner. This creates a sense of being out of control, something that is never desirable since – from an evolutionary perspective – it means that the chances of survival are reduced.
So, be wary of your own biases. Two events occurring frequently together does not mean one caused the other, even if it seems to makes perfect sense. Especially if it involves headaches.
Another excellent example is that having a college degree is a common trait in a successful career. So thus in order to be successful you must have a college degree.
It is also a common problem in technology. One of the most common complaints I would get from users is “When I click the button the program doesn’t work. The button must be broken.” I wish it were EVER that simple.
The button is not broke, it would probably be fine if the underlying problem were fixed. The button not functioning correctly is being caused by the same problem. The button click issue correlates to the underlying problem but it is not the cause.
Dean Spitzer wrote, “In the absence of data, anyone’s biased opinion is as good as anyone else’s.”
Our personal frame of reference has to be reoriented from time to time to prevent tunnel vision or jumping to self-serving conclusions.
We don’t, usually, get the benefit of feedback from others – since how we perceive ourselves is rarely how others view us. While a hungover male college student may point at his shoes rather than his drinking decisions for the cause of his headache, his college professor, as well as most of his friends won’t agree and won’t provide constructive feedback that could illuminate the student.
I am always happy when students remember what their college professors say, in this case about correlation and causation. This is one of the more challenging topics to teach. It is easy to have the class remember to repeat “correlation does not imply causation”, but remembering in the practical situation is more difficult. I find the “vivid” examples help us remember, in the class and in research. When professors write empirical papers (looking at data) one of issues we are often and correctly challenged on is whether we have found a correlation, or truly the correlation.
An amusing version of this was the point of an XKCD comic strip.
If you are unfamiliar with XKCD, it is an online comic that comes with the warning
“This comic occasionally contains strong language (which may be unsuitable for children), unusual humor (which may be unsuitable for adults), and advanced mathematics (which may be unsuitable for liberal-arts majors).”
The increased focus on marketing ROI and data-driven decisions are requiring marketers to be more like scientists that build theories and do tests with increased discipline.
I especially enjoyed this post because just a week ago I wrote on this topic – http://theadaptivemarketer.com/2013/11/09/does-your-marketing-formula-pass-the-correlation-test/ I would love to hear your thoughts on it.
Jonathan, I just discovered your site a few moments ago and it is delightfully intelligent and refreshing. Thank you for restoring my faith that intelligent and witty people still roam the planet!