Margarine causes divorce!

My last post was about how easy it is to become confused about nutrition.  In that post, I cautioned that we have to careful about believing headlines, and also to be careful when an article says something is associated with something else.  I thought I would delve into that content a bit deeper.  See the headline above (the title of this post.)  Certainly sounds crazy and hard to believe, but follow along and see how reporting on a study, not unlike the typical reporting on the latest nutrition-related study, can be misleading.

Margarine Linked to Divorce, recent study reveals

Researchers at Big University of Maine (BUM) were out to discover the reason for the 18% decline in divorce rate that has occurred over the past 10 years and made a startling discovery.  Trying to find any cause for this decline, they turned to dietary habits and examined USDA data on per capita margarine consumption in the United States.  What they found was a strong association between the use of margarine and the divorce rate in Maine. The decline in margarine consumption they found correlated very strongly with the decline in divorce.
When asked about their groundbreaking research the authors of the study commented; "This study appears indicate that if we limit our consumption of margarine it will lead to better marital relationships and thus a reduction in the divorce rate."  If these findings prove to be correct, it indeed seems possible to reduce divorce.  Speculating on the reason for this relationship the authors say they believe it has something to do with the reduction in fat consumption that leads to better overall health and a corresponding improvement in marital relations. "We think this is a big step in reducing the divorce rate.  If people will limit their fat intake, we can see divorce rates continue to decline.  There's no reason this couldn't be applied across the whole country."

This is a real association using actual data.  I got it from a great web site titled "Spurious Correlations."  The creator of the site, Tyler Vigen, emphasizes that these correlations are not meant to imply causation.  The difference is key to understanding reporting on any research, but particularly nutrition research.

If there is a correlation between two things, it means that as one changes, the other changes.  In other words, if A and B are correlated, then as A goes up, B tends to go up, or as A goes up, B tends to go down.  (One of these is a positive correlation, the other is a negative correlation.) Notice that I didn't say that A causes B.  Correlation does not mean that something causes something else, it just means that when one changes, the other one seems to change.  They are probably related in some way, but you cannot conclude that one of them causes the other.

So for the example with divorce rates and consumption of margarine, you can see by the graph that they really are correlated.  They are strongly correlated.  Over the 10 years of data provided, as one changes the other changes in step.  But you should never conclude from this that margarine consumption causes divorce, or even that divorce causes margarine consumption.  So when you read a story or hear about something being correlated with (or associated with) something else, DO NOT THINK THAT ONE CAUSES THE OTHER.

So why would margarine consumption be correlated to divorce rates?  Maybe both of them are related to something else.  One possibility might be that over those 10 years the income increased or declined influencing both the chances of divorce as well as the food choices that people buy.

There is value to studies that look at correlation using existing data.  They are great clues to what may be going on.  For example, let's think about the relationship between cholesterol and heart disease.  Early in the study of heart disease it was observed in people with atherosclerosis (where plaque builds up in people's arteries) that where this exists, there is a lot of cholesterol there.  So that's a great observation that there is something going on between the disease and cholesterol.  That does not mean that cholesterol causes atherosclerosis. Maybe it means that atherosclerosis causes the cholesterol to be there!

An excellent analogy for this that I've heard is that if we wanted to prevent fires we could go to a bunch of fires and make observations to see what else is there and try to figure out how to prevent them.  What we likely would find is that there is a very strong correlation between house fires and the existence of fire trucks.  We could conclude that fire trucks cause fires.  A better conclusion though would be that fires bring fire trucks.  The same is probably more true about Cholesterol.

Be very wary when you hear about studies citing one thing is correlated with (or associated with) another.  Don't believe it when it is then claimed that to prevent some disease, that the correlation means you should reduce (or increase) the consumption of something else.