Correlation is Not Causation
Alright people, here’s another tip drilled into me writing for the President’s Daily Brief: correlation is not causation. Honestly I didn’t…
Alright people, here’s another tip drilled into me writing for the President’s Daily Brief: correlation is not causation. Honestly I didn’t need PDB training for this…but I see so many people make this mistake, all the time. And I make it too sometimes. It’s an easy trap to fall into.
So, here are three rhetorical techniques often used when making this mistake, illustrated with examples related to my dog. Then, I’ll explain how I prefer to express potentially worrisome correlations where we don’t have enough data.
Also if you haven’t seen it, check out these spurious correlations by Tyler Vigen. And check out this picture of my dog. Then read on below!
1. B Increased When A Did — A Caused B!
This is a classic. We’ve all seen the chart with a line showing an increase in one thing, say “adoption of puppies,” at the same time as an increase in another thing, say “spending on pet toys.”
Yes it’s possible that people adopted more puppies, causing them to buy more dog toys.
But, it’s also possible people adopted more cats, causing them to buy more cat toys, and dog toy sales were flat.
Or, perhaps less likely, maybe people just bought more pet toys first. And then they adopted the puppies to have something to do with them.
The point is, we don’t what what caused what, or if there was an intervening variable.
2. B Happened After A Changed — A Caused B!
Here’s another classic. We’ve also all seen the chart where a trend, say “purchases of cinnamon Chow Chows” is flat, and then it starts going up after a certain date. And marked at that date is some key event, say “release of Star Wars.”
Sure, is it possible, even likely, that people bought more dogs that look like Wookies or Ewoks after Star Wars? Of course!
But, it’s also possible people just bought more of these dogs anyway. And started naming them with Star Wars-themed names after the fact.
3. A is Part of Group B and Did C —All B’s Always do C!
This one is also so common. It’s not exactly a correlation equals causation argument, but I don’t like it so I’m including it. For example, I could say that because my dog always barks at the doorbell, therefore ALL dogs bark at the doorbell.
Ok yes it’s true, lots of dogs bark at the doorbell.
But some don’t. It’s not always true!
Ok But What About Worrisome Correlations?
Yes indeed, I know sometimes correlation DOES equal causation, but we just don’t have hard evidence of causation yet. Like, smoking causing cancer.
So, what if we see A correlated with B, but can’t prove that A causes B. But, it’s possibly really really bad if A indeed is causing B? How should we express that?
This is totally ok! Say “we need to dig into the issue more.” Or “the risk that A is causing B is so high we need to take action just in case.”
Just don’t say “A is causing B” as a statement of fact. This way, you’re expressing your opinion, not making a statement you can’t back up.