web analytics

April 18, 2014



Correlation is NOT Causation

I have many pet peeves, from the oh-so common misuse of scientific terms (as any number of fellow Star Wars fans could point out to you in great detail, a parsec is a unit of distance, not time) to the need to include onions in almost every dish (I’m looking at my fiancee, Sondra, for this one).  One in particular that bothers me (and more to the point, has a rather negative influence on society in general) is the tendency to mix up correlation and causation.

In a nutshell, correlation is simply when two events are close together in time and space, whereas causation is when one event leads to the second event occurring.  Events can be correlated without having a causal link; just because you got hit by a car after losing your lucky penny, doesn’t mean that losing the penny caused you to get hit by a car.  There’s an tendency of the human brain to assume that all events, particularly bad events, are caused by something we or someone else did.  This is good in a way; if smoking really does cause cancer, for example, it’s good to know in order so you can avoid predictable problems in the future.

They don't call them 'cancer sticks' for nothing

They don’t call them ‘cancer sticks’ for nothing

The problem is that humans, as we so often do, tend to take it too far, seeing causes in places where there are none (or where the causes are subtle and difficult to discern, or where the causes are well outside of our control, etc).  A complete list of such situations could fill many pages on its own, but a short list can include:

  • Any good luck charm, particularly when you’re trying to influence events in which you are not participating.  As much as you may think the outcome of the game you’re watching hinges on whether youare wearing your lucky underpants, trust me when I say that it doesn’t.
  • Whenever some form of media is blamed for someone’s (usually a teen’s) bad behavior.  Cracked made a good point about video games and violence in a larger article about ways the news media distorts the news.  Pornography or other sexual explicit material is another area where this comes into play.
  • An awful lot of scientific studies, particularly in the social sciences (psychology, sociology, etc.).  It’s hard to prove definitively that certain personal traits cause other traits or particular behaviors; at best, you can sometimes point out that two traits are often seen together, but that doesn’t tell you much about which one causes the other, or even if one is in fact a cause.  (To cite just two examples I’ve seen floating around, liberals are more intelligent and conservatives are more generous, if these studies are to be taken at face value.)

There’s many more examples that could be mentioned, but you’ve probably gotten the point: we humans have a tendency to ascribe as causes earlier events that just ‘seem right’, without proof that they actually caused the later events.

What Can We Do?

One of the best ways to protect yourself from falling into this fallacy of assuming correlations are causes is to be aware of the ways that one event that seems to cause another can be related in another way.  Let’s consider an example from the realm of the political and financial: On January 20, 2009, the day Barack Obama was inaugurated, the Dow closed at 7949; then, on January 20, 2010, the Dow closed at 10,603, an increase of over 33%.  A particularly smug Democrat might say, “Look, Obama boosted the stock market by a third in his first year.”

It seems like a logical conclusion; event A (Obama’s inauguration) causes event B (a rise in the Dow).  That’s how these things work, one thing causing another, right?  But there’s several other options besides the straightforward ‘A causes B’; that’s see what some of them are, as well as how they could explain our observation about Obama and the Dow:

1) A causes C, which in turn causes B: There’s frequently a series to events that occur.  One thing triggers another, causing another, and so on.  There might be an intervening event (or multiple events) that connects our ’cause’ and our ‘effect’.  In our example, perhaps Obama’s election caused such a rush on gun manufacturers, talk radio hosts, and bottled water makers that the resulting upswing in sales/listeners had a ripple effect through the economy, and the rising tide lifted all boats.

2) C causes both A and B: It’s always possible that there is a third event that leads to both of the events listed in your supposed causation.  So, the bad economic climate of 2008 might have cleared the way for Obama’s victory, and created circumstances that allowed the Dow to grow so robustly in 2009 (by knocking it down fifty percent or so from its high, giving it lots of room to grow).

3) A was one of (many) contributing factors to B: With most things in life, there’s not a single cause.  There’s frequently numerous events contributing to the final outcome.  So, the Dow rising might be attributable in part to Obama’s actions, but also to the actions of the Fed, of foreign bankers, of the media and of individual investors (among others).  It’s a group effort, even if one part of the group gets most of the credit.

4) C causes B, and A decreases the effects: This is probably the most counter-intuitive, but it can happen; rather than A causing B, A serves to decrease B.  In the video game/violence correlation mentioned earlier, maybe playing violence video games actually decreases the amount of violence we see in the real world, by giving those with a tendency for violence an outlet for their urges that doesn’t blow a hole in someone’s head.  (There’s a term in psychology called ‘catharsis’, where primal urges toward lust and violence are expressed in a socially acceptable way.)  Similarly, maybe Obama and his policies have actually slowed the recovery we’d see with a more hands off economic approach.  (A Republican retort to the smug Democrat might be, “Maybe, but we’d see an even better recovery if McCain was in office.”)

5) Event A and Event B have no direct relationship: One of the hardest relationships for us to accept is that, well, there simply isn’t a relationship.  Event A simply happened first and seems like it should be the cause (or at least, a cause) of Event B.  Maybe it’s sheer coincidence that one year after Obama was inaugurated, the Dow has risen significantly.

That covers many of the major relationships between events other than direct causation; there are more, of course, but these suggestions, alone or in combination, should help provide plenty of alternatives to consider.  Just remember that simply because A came before B (no matter what A and B are), doesn’t mean that A caused B.  Look for alternate explanations, and you’ll go a long way toward inoculating yourself against false arguments.

Comments

  1. This is stats/logical reasoning all over again from college! Ouch lol
    .-= Evan´s last blog ..What is Keyman Insurance? Is Your Business Protected From Death? =-.

  2. Great rightup on spurious relationships! This is one of my pet peeves too!!!

    I’ve caught my son making this mistake a few time lately. I’m in the process of schooling him on the differences. Hard to do with a 9 year old ;)
    .-= Money Reasons´s last blog ..Asset Class Diversification For The Middle Class =-.

  3. Captain Rick says:

    What? Smoking causes cancer? When did they decide that? hehe

    • @Captain Rick: I suppose we could argue about just how ‘decided’ the link between smoking and cancer is right now. Still, it seemed an apt metaphor.

  4. Chuck G says:

    I myself tend to be guilty of this at times. I guess people, at some point, just need to make some kind of meaning as to why something happens. It was David Hume who said that causality is based on experience, and experience is similarly based on the assumption that the future models the past, which in turn can only be based on experience. I don’t think this is harmful, as long as it is not overdone. Your example about Obama’s win seem too farfetched to me, maybe people just want something to talk about.

Speak Your Mind

*

CommentLuv badge