In statistics, confounding is when things are mixed up to the point of confusion, such that it is difficult to determine what causes what. Why is this a concept you should care about? Consider the following textbook example, which is similar to some of the headline-grabbing claims you may come across in newspapers or on the internet:
“In New York City, researchers have documented a positive association between ice cream consumption and burglary. On days when New Yorkers eat lots of ice cream, the burglary rate tends to be high. On the other hand, on days when people refrain from ice cream consumption, these rates are much lower.”
Given the above findings, what sort of conclusions would you draw? A lousy newspaper may report the above as follows:
- “Study Shows Eating Ice Cream Causes Burglary.”
Another media outlet may spin the same research in a different way:
- “Is Burglary Causing Us to Eat More Ice Cream?”
But in both cases, we have to be careful never to jump to any immediate conclusions because it is possible that a confounding variable exists. That is, a third variable may be confounding (mixing up/confusing) the relationship between ice cream sales and burglary.
In the above example the confounding variable happens to be temperature. This is because when it is hot people eat a lot of ice cream. They also spend a lot of time outdoors and this increases the opportunities for burglary.
So while ice cream sales (X) and burglary (Y) appear to move together, this does not mean X causes Y or vice-versa. A third variable, temperature, is the culprit.
Researchers usually make an effort to identify and eliminate the effects of confounding variables but this isn’t always successful. So be sure to maintain a bit of a skepticism whenever you come across studies that claim some factor, X, causes another factor, Y.
Statistics and Data Interpretation for Social Work by James A. Rosenthal
Wikipedia Article on Confounding byWikipedia