Correlation Not Equal Causation

Apparently there is a new study being trumpeted by social conservatives to help them begin a new attack on raunchy song lyrics.  The Rand study to be published in the journal Pediatrics did phone interviews with a bunch of teenagers, asking them about their sexual habits and what songs they listen to.  They found a correlation that teenagers tend to listen to more sexually-degrading music also are more likely to subsequently begin having sex (or having more sex).  The articles reporting on this study have headlines like this:

Study: Raunchy Music = Earlier Sex

The implication is that listening to sexually-charged* music causes more early sex.  But in fact, they don't know that.   Before it starts to rain here in Phoenix, the sky goes dark and the winds pick up.  Does this mean that darkening skies and increasing winds cause rain?  Or are darkening skies and winds merely a leading edge symptom of a broader phenomena that also includes rain, which we might call "a thunderstorm moving through town."

Does interest in sexually degrading lyrics actually cause teenagers to have sex when they might not have otherwise?  Or is this interest in such music merely a leading indicator, a symptom on the leading edge of a larger phenomena that one might label "adolescence" or "hormone overload."  As an alternative hypothesis to explain the data, one could argue that listening to this music is merely an early low-risk form of sexual experimentation, like sneaking a peek at the Playboy magazines at the local 7-11, which then gets followed up by (but doesn't cause) physical sex. 

I don't know the answer.   Though few would describe me as a puritan, I certainly won't let some of that crappy music in the house (I do check what the kids are downloading on iTunes).  On the other hand, "Don't Fear the Reaper" was one of my favorite songs for years and I never felt the slightest urge to kill myself. 

What I do know is that you absolutely have to beware of the media when they report studies and statistics, and correlation=causation is their absolute favorite mistake to make.

* the article makes a second mistake, in that the study authors found a difference between the correlation of teen sex with sexually degrading music vs. just music with sexual content.  The media also misses this distinction.

Postscript:  This article about professional wrestling leading to teenage violence seems to make the same mistakes


  1. Kyle Bennett:

    The 4 modes of causality: If A and B are correlated (by time, physical proximity, common attributes, or some such), then either A caused B, B caused A, Some C caused both, or coincidence (Meaning that they are co-incident, they are correlated, but no causal relationship exists). Studies like this (or the distorted reporting of them) just assume the first case, with A being the thing that happened first, or was observed first, or that simply stands out for some superficial reason. Any observation that stops at identifying the correlation and takes that in itself to be evidence of causality, and for of one of the modes of causality in particular, is meaningless at best, deliberately deceptive at worst.

  2. gc:

    Here's an interesting use of statistics I found in an article on youth violence:

    "Among youth ages 10 to 14 years, homicide and suicide are the third and fourth leading causes of death, respectively; among 15 to 19 year olds, they are second and third"

    Shocking? No. I'm guessing the #1 cause is auto acciddents. That doesn't leave many other leading causes of death. What else is going to kill teenagers - heart disease, cancer? Not likely.