Marriages: Causation or Correlation?

Submitted: Feb. 2, 2008 @ 10:38 pm by Donna Daines

As I'm scanning the news for my first column I come across the article, "A Good Fight May Keep You And Your Marriage Healthy." Not exactly the Life Sciences, but here it is under the health section on Google News, and it is just the topic I want to take on.

The article talks about a study tracking 192 couples over 17 years and comparing the mortality rates of self-identified couples broken into three groups: those who express anger, those who don't, and mixed couples. Ultimately the study finds a link between expressing anger in marriage and mortality rate – those who self-identified as being expressive were shown to live longer.

What caught my interest was not the content of the study itself – which is seemingly trivial at first inspection, and even more so under careful scrutiny – but the media attention that such a study was receiving, the misleading headlines, and the false conclusions that were being drawn in light of the findings.

Not so long ago
To understand what I mean we need to backup. Picture an empty chair in the back of a small lecture room in the basement of the Widstoe building on BYU's Provo campus. Next to you in the chair is a tired young father who has fallen asleep despite a very interesting lecturer giving a very interesting lecture at the front of the class. The lecturer's name is Duane Jefferey's and he's taught at BYU for ages. In fact, he's going to retire this year, but you don't know that yet. The teachers talking about two words that begin with the letter "C." And if you haven't guessed it yet, the tired father was me.

Correlation vs Causation
Luckily I woke up at this point so I can explain an important principle that I will remember for the rest of my life. These two words have everything to do with avoiding the sorts of spurious conclusions that news writers would like us to make. Simply put, correlation is the observation that two things tend to happen in proximity to each other, whether it's in the same individual or the same point in time. Correlation says (to borrow from Dr. Seuss) "thing one" and "thing two" hang out a lot. On the other hand, causation is a much stronger conclusion; it means that "thing one" is responsible for "thing two" because by some means or other "thing one" leads to "thing two."

So when do I have causation? When you can show that "thing one" not only correlates with "thing two" but there is a means by which it brings "thing two" about. In the sciences we call this means a mechanism. The primary motive of basic research is to identify these important mechanisms.

Back to the Study
So what does this mean for our interpretation of the results? Well, statistics and assumptions aside (in other words, whether the researchers did their job or not), we can not conclude much from the study except that in the group of people studied, lower mortality rate correlated with couples who self-identified as being expressive of anger. We can not conclude that expressing anger in marriage will lead to a longer life. This would be causation.

So what do we say to the authors that want us to make unfounded conclusions like yelling at your spouse will make you live longer, or to ABC producer's whose upcoming show leads viewers to believe that autism is caused by vaccinations? We say show me the mechanism; show me the way by which "thing one" leads to "two."

What good are the social sciences then?
Much research goes into understanding how environment, behavior, and other aspects of a person's life affect their health. Most research of this nature falls into the general category of the social sciences. If such studies never uncover a mechanism, one might wonder if they are any good. The answer is yes, they are good and useful. Identifying a correlation is often an indication of the direction to look. "Thing one" might not cause "thing two," but it's possible. It's also possible that "thing two" causes "thing one," as funny or backward as that may seem sometimes. So correlation is often a good guide post for basic research; it points us in the direction we should go to explore and find the all important mechanism.

What's it mean to me?
If you get nothing else from my column, go away with this thought: Don't believe everything someone is trying to conclude if all they give you as evidience is correlation. When you're doubtful, say, "Show me the mechanism." If there isn't a mechanism, there isn't causation, and without causation their conclusions are likely spurious.