Ghost Variables: Why Ice Cream Doesn't Make People Drown

The Setup

Every summer, something strange happens.

As ice cream sales go up, so do drowning deaths.

The data is real. The correlation is real. So... should we ban ice cream? Add drowning warnings to Popsicles? Make people pass a swimming test before buying a soft serve?

Obviously not. But why not — that's the actually interesting question.

There's a hidden player in this story that nobody talks about. A ghost variable. And once you learn to see them, you'll spot them everywhere.

What's a Ghost Variable?

A ghost variable (also called a confounding variable or lurking variable) is a third factor that secretly causes both things you're looking at.

It's not one thing causing another. It's both things being caused by something you didn't notice.

The ice cream / drowning connection isn't weird at all once you see the ghost: hot weather.

When it's hot:

• People buy more ice cream ✓
• People swim more → more drowning accidents ✓

Heat causes both. Ice cream and drowning are connected only because they share a common cause. Remove the summer heat, and the correlation disappears.

That's a ghost variable. It haunts your data, making things look connected when they're actually just both responding to something invisible.

Real-Life Ghost Variable Hall of Fame

Nicolas Cage movies and pool drownings

This is a real dataset someone made. Years when Nicolas Cage released more films also had more pool drownings. Obviously Nicolas Cage isn't killing people. The ghost? Probably summer blockbuster seasons, economic cycles, random noise in small datasets — classic ghost stuff.

Shoe size and reading ability in children

Older kids have bigger feet. Older kids also read better. So... big feet make you smarter? No. The ghost is age. Age explains both.

Countries with more TVs per person have lower birth rates

Richer countries have more TVs. Richer countries have lower birth rates. The ghost is wealth — it affects both TV ownership and family planning decisions. Sending TVs to poor countries won't lower birth rates.

(This one actually influenced bad development policy for a while. Ghost variables have consequences.)

Your TikTok feed and your mood

You might notice that on days when you use TikTok more, you feel worse. Is TikTok causing your bad mood? Maybe. But the ghost variable could be: having a bad day makes you seek distraction and also makes you feel bad. The causation might be backwards — or there's a ghost.

Why This Matters More Than You Think

This isn't just a statistics puzzle. Humans make serious decisions based on correlations all the time.

• "Kids who play violent video games are more aggressive" → Ghost variable: kids who are already more aggressive are drawn to violent games
• "Students who eat breakfast do better in school" → Ghost variable: stable home environment → both regular breakfast and academic support
• "People who go to church live longer" → Ghost variable: strong social community affects both religious attendance and health outcomes

Every time someone says "X causes Y," the first question should be: is there a Z causing both?

How to Hunt the Ghost

When you see a surprising correlation, run this checklist:

1. Who are these people?

What else do they have in common? Is there a hidden trait that explains both things?

2. What's the timing?

Does one thing consistently happen before the other? (Correlation has no direction. Causation does.)

3. Does it make mechanical sense?

For A to cause B, there needs to be an actual mechanism — a way it could physically happen. Ice cream → drowning has no mechanism. Temperature → both does.

4. What happens if you control for Z?

If you separate the data into "summer" and "winter" groups, does the ice cream / drowning correlation disappear? If yes, you found your ghost.

The Challenge

Ghost hunting mission:

Think of a correlation you've heard stated as a cause:

• "Social media causes depression"
• "Good grades lead to success"
• "Being popular in school leads to worse outcomes as an adult"

Pick one. Try to identify at least two possible ghost variables that could explain the connection without one thing causing the other.

Then flip it: what would you need to prove it's actually causation, not just correlation?

Share your ghost hunt. The best ones are always the most counterintuitive.

The world is full of things that move together for completely unrelated reasons. The ones who see the ghost are the ones who don't get fooled.