Causation

By Bradley Wright

Sociologists spend a lot of time talking and thinking about causality. (We probably spend even more time with office politics, but that’s not very interesting, so I won’t write about that).

Now, I have a sneaking suspicion that philosophers spend a lot of time defining exactly what is the essence of causality, and they probably trace it back to Romans and Greeks and people like that. Rather than go into this type of philosophical analysis, I will simply focus on how we might test for its existence.

If “A” is a cause of “B”, what does that mean? In this case, “A” and “B” could be just about anything—characteristics of people, interactions, groups, societies.

First off, we would expect some level of association between “A” and “B”. By this we mean that as levels in “A” change, we would expect usually to see some change in “B”. Some associations are positive, meaning that “A” and “B” move in the same direction. So, as “A” increases, “B” does also. (Or, conversely, if “A” decreases, so does “B”). Other associations are negative, meaning that as “A” increases, “B” decreases or the reverse.

Second, we should see changes in “A” occur before changes in “B”. Since very few sociologists can afford time-travel machines (though I think that I saw a colleague with a flux capacitor in their office), we are stuck with the linear progression of time. That means that changes in a cause have to happen before the resulting consequences in the effect. Sometimes this time difference is miniscule, so that changes in “A” and “B” seem to happen almost simultaneously, but there is still some ordering. At the very least, if “A” causes “B”, then changes in “B” can not happen before corresponding changes in “A”.

Third, there should be no spurious correlation. A spurious correlation means that some other variable causes both “A” and “B” such that they correlate with each other, and maybe “A” comes before “B”, but in fact there is not causal connection between them. (For a fuller explanation, read this previous post). This is where things get a little tricky. Researchers can measure if two variables are associated, and he or she can measure which came first, but how can you know that there is no secret variable out there that makes the correlation spurious? Who knows, given what “A” and “B” are, there could be dozens if not hundreds possible spurious correlates. How can a researcher rule out all of them? They can’t. The researcher can measure and rule out any obvious spurious correlates, but ultimately it’s an act of faith (or, as it’s called in sociology, “theory”) that a correlation between “A” and “B” is not spurious.

Finally, we like to know how “A” causes “B”. There can be a causal relationship between the two even if we don’t know how they affect each other, but knowing “how” makes us more confident the causal connection. Basically, sociologists sleep better at night if they know the causal mechanism.

So far I’ve discussed this in rather abstract terms, and you’re probably wondering if I had intended to put you to sleep at your computer. (Sociologists sometimes forget that regular human beings don’t get excited talking about vague “A”s and “B”s).

Here’s an example.

Suppose that a friend told you that they had a bag of magic M&Ms. Now, I realize that for some people, any bag of candy is magic, but these are special M&Ms, according to your friend. If you eat a green one, you will instantly become amazingly physically attractive (if you’re not already). You’ll be so handsome or beautiful, that you’ll end up on lists like this, this, and this. (Okay, the last one was just to see if you were paying attention.)

You are intrigued, but you want to find out if it’s true. Does eating green M&Ms make you attractive? Or, to put into boring sociological notation, does “A” cause “B”? To test this, you give a bag of the magic M&Ms to your friends, and then you take notes.

First you notice whether the friends who ate green M&Ms are more attractive than those who didn’t. If so, this would be a positive association—more green M&Ms = more good looks.

Then you would look to see which came first. Perhaps beautiful people just happen to eat more green M&Ms; if so, they “B” comes before “A”, and we don’t think “A” is a cause of “B”.

Can you think of any spurious correlation between green M&Ms and attractiveness? At this particular moment, I can’t (but then again, I may just be thinking about how thin this example is getting).

Finally, you wonder how green M&Ms would change a person so dramatically. You might send them off to the lab and have them analyzed.

Once you’ve answered all these questions, you can decide for yourself if there is hope that green M&Ms will make you so good looking. Then again, maybe you should have some anyway… just in case.

Comments

[Chelsea T.]

That's a good analogy for understanding the definition of causation. Of course, it's easier to understand when dealing with any sort of sweets.

[Kaila C.]

Referencing this term back to candy really caught my attention which aided me in understanding what causation is. I now realize that it is the thought that one thing always leads to another, but that thought can really never be proven completely true. It is based more off of what you choose to believe. Thanks : )

[Kristie]

Wow. This was really helpful to me in understanding what causation was. The M&M illustration makes it easier to understand. If only green M&Ms did make people more attractive.

[Aditya M.]

You have explained the concepts of causation and spurious correlation very clearly. Sometimes abstract concepts do need a sugar coating. I understand the definition of spurious correlation better. We should not assume a relationship between two variables without first investigating the existence of other variables that may cause the relationship. I am looking for a bag of M&M's anyway.

[Justin]

Refering the definition of causation to candy made it more easy for me to understand. How you you explained it, I could relate and understand what causation is.

[Makenzie]

This was a great way to describe causation. With a green M&M causing someone to look better makes understanding causation a lot sweeter. Knowing what is the cause of something and where it came from is a good understanding and something to experiment with as you said.

[Madison M.]

This article really helped me understand causation better. Using the green M&M's helped me actually visualize what causation is. Too bad your story wasn't true.

[Eric Irish]

I agree with you, although I believe that it's possible that A can change without a single chang in B, also that it is possible for B to be the first one to change if at all. Also that their could be more then just one extra variable involved, each day new problems occur that have multiple variables and causes.

[Aditya Voruganti]

I completely agree with your stance on the spurious correlations. There are an infinite possibilities of things that could affect the results of a causation test. I think that in order to truly test to see if there are any spurious correlation, one would have to isolate A and B and try to see if the causation (or lack there of) continues in this isolated area. Then slowly one should add normal parts of the environment around the variables being tested to see if they start to react differently. For example: in the M&M experiment, have the subject eat green M&Ms without eating anything else for the day. See if the subject becomes more attractive, then the next day feed them M&Ms and let them eat. If a different result occurs, then it could be assumed that what the subject eats, coupled with the M&Ms effects attractiveness.

[Darius France]

That was very intresting the way you used the comparison of the green m&m's. It also helped me to better understand causality in a whole new light. (And were i come from we could use some of those magic green m&m's!!)

[Sophia]

This article really helped me to understand what causation is and how it works. At first, I ddn't understand the A, B explaination, but when you related it to candy, it made a lot more sense. Thanks for this article!