WS MoreOrLess: 'Spurious Correlations'

Summary

Podcast Summary: More or Less – "Spurious Correlations"

BBC Radio 4, World Service Edition – June 9, 2014
Host: Tim Harford
Featured Guest: Tyler Vigen (Harvard Law student, creator of Spurious Correlations)
Producer/Contributor: Ben Crichton

---

Overview of the Episode

This episode of "More or Less" dives into the pitfalls of confusing correlation with causation, using the viral phenomenon of "Spurious Correlations" as a humorous and enlightening case study. Tim Harford and his team, with help from guest Tyler Vigen, explore how unrelated data trends can line up perfectly, tricking us into seeing connections where none really exist. The second segment applies probability theory—via a listener’s question— to the office tradition of World Cup sweepstakes, debunking myths about strategy and luck.

---

Key Discussion Points & Insights

1. Spurious Correlations: When Two Trends Align by Chance

• Tim Harford introduces the episode with a tongue-in-cheek claim: margarine causes divorce, supported by a convincing-looking chart he found online (00:18).
• Audio Illustration: With the help of Ben Crichton, Harford uses music to represent how two unrelated statistics (US margarine consumption and the divorce rate in Maine) can move together– but this does not mean one causes the other (01:18-02:28).
  • Ben Crichton: “If it was a graph of something increasing steadily over time ... it’s going to sound like this.” (01:24)
• Tyler Vigen (02:37) explains the joke behind his website [SpuriousCorrelations.com], which uses computer algorithms to mine for data pairs with similar trends.
  • He highlights combinations such as "the age of Miss America winners" correlating with "murders by steam, hot vapors and hot objects," and "non-commercial space launches" aligning with "US sociology doctorates" (03:16-03:53).

2. Why Correlation is Not Enough

• Harford and Vigen emphasize that just because data sets show similar trends, it doesn't mean there's a meaningful or causal connection (04:06-04:45).
  • Tyler Vigen [04:10]: “It’s really important for us to be critical of whether that correlation exists, whether we can show that there’s actually statistical significance, and whether there’s a causal mechanism.”
• Golden Rule Reminder: Correlation does not equal causation.
  • The BBC cheekily reinforces this: [04:58]: “Staff are reminded that correlation does not equal causation and that sometimes correlation isn’t even really correlation.”

3. Listener Question: The Office Sweepstake and Probability

• Transition to Next Topic: [05:24] – Tim teases the World Cup office sweepstake as a context for exploring probability.
• Listener Andrew from Surrey asks whether it’s smarter to wait before joining the random team draw in an office sweepstake to improve his odds (06:14-06:36).
• Harford’s Analysis:
  • It makes no difference when you enter—the good teams are as likely to be picked first as last (06:49-06:53).
    • Tim Harford: “Brazil is as likely to be picked first as last.”
  • If you delay joining and some favorites are already taken, your odds worsen—so any illusion of strategy is a fallacy unless you’re willing to skip the sweepstake entirely (06:53-07:24).
• Additional Layer: If everyone tries to delay entry for better odds, it becomes a circular situation—almost a game-theory problem (07:24-08:24).
• Real-World Adaptation: Radio Current Affairs sweepstakes organizer, Richard Fenton Smith, explains their solution: first, everyone buys a ticket, then there’s a random order draw to keep things fair (08:24-08:41).

---

Notable Quotes & Memorable Moments

• Tim Harford [00:18]: “We’re starting a campaign to ban margarine ... We’ve seen convincing evidence that margarine consumption causes divorce. It’s all in a cool chart on a website. ... It’s totally, totally convincing.”
• Tyler Vigen [02:45]: “It allows people to be their own scientists for a few minutes ... and then realize that we have no basis for that in reality.”
• Ben Crichton [03:36]: “Correlates with murders by steam, hot vapors and hot objects.”
• Tyler Vigen [04:10]: “It’s not that hard to find overlapping lines that kind of curve or vary together. And just because two things vary together on a linear scale doesn’t mean that they’re actually correlated or that they’re actually causing each other.”
• BBC Announcement [04:58]: “Staff are reminded that correlation does not equal causation and that sometimes correlation isn’t even really correlation.”
• Tim Harford [06:53]: “Brazil is as likely to be picked first as last.”
• Richard Fenton Smith [08:24]: “This year I decided to take all the money first and then have a draw ... then they will pick their team.”

---

Timestamps for Important Segments

• 00:18 – Intro to margarine/divorce “correlation” joke
• 01:18-02:37 – Audio dramatization of how two trends can align
• 02:37-04:45 – Interview with Tyler Vigen; explanation of Spurious Correlations website and statistical thinking
• 04:58 – BBC tongue-in-cheek notice about correlation and causation
• 05:24-08:41 – World Cup sweepstake probability analysis and office example

---

Episode Tone and Style

• Playful and witty, with tongue-in-cheek humor directed at statistical missteps
• Logical, clear explanations for lay listeners, mixed with musical and comedic interludes
• Cautious advocacy for skepticism and critical thinking about the meaning behind big data and headline stats

---

This episode is a lively yet practical reminder: always question apparent patterns and supposed ‘facts’ in statistical reports, and remember—just because two things happen together, doesn’t mean one caused the other.

Transcript

[BBC Announcer]

This is the short edition of More or Less, first broadcast on the BBC World Service.

[BBC Podcast Promoter]

Thank you for downloading from the BBC. For details of our complete range of podcasts and our terms of use, go to bbcworldservice.com podcasts

[Tim Harford]

hello and welcome to More or Less on the BBC World Service. We're your weekly numerical haggis stuffed full of tasty statistics. And I'm Tim Harford here at More or Less. We're starting a campaign to ban margarine. That's partly because butter is delicious, but it's mostly because of the side effects of margarine. You see, we've seen convincing evidence that margarine consumption causes divorce. It's all in a cool chart on a website. Let me just bring it up. Yeah, I'm looking at it now. It's totally, totally convincing. All I need is some way to convey to you just how convincing it is.

[Ben Crichton]

Ben, could you come in for a minute?

Ok.

Right. Ben Crichton is more or less his violist in residence, and Ben is going to play some swooping, elegant glissando notes. You can do that, right, Ben?

[Tim Harford]

Yes.

[Ben Crichton]

Hey, Ben, what's the difference between a viola and a chainsaw?

I don't know, Tim. You tell me.

If you have to, you can get a tune out of a chainsaw. Ok, let's just get to the glissando, shall we? They'll mimic what these graphs look like. For example, if it was a graph of something increasing steadily over time, a graph that starts low on the left and ends high on the right. It's going to sound like this. Okay, so here's what the graph for per capita consumption of Margarine in the US from 2000 to 2009 sounds like.

[BBC Announcer]

Like.

[Ben Crichton]

Falling over time. And here's what the graph for the divorce rate in the US state of Maine over the same period sounds like. And put them together and you'll be able to hear. They correlate perfectly. Thank you, Ben. Now, you've probably seen graphs like this too, in your favourite newspaper or website.

[Tim Harford]

And you'll agree that when two things

[Ben Crichton]

line up perfectly like this, it's time to take action. Bane to save marriages.

[Tim Harford]

Who's with me?

[Tyler Vegan]

My name's Tyler Vegan. I'm a student here at Harvard Law

[Tim Harford]

School, and Tyler runs the website which published this graph.

[Ben Crichton]

So I'll let him explain its importance.

[Tyler Vegan]

What's kind of fun about it is it allows people to be their own scientists for a few minutes because they get to come up with a hypothesis like Maybe when there's more margarine in the house, it's more likely to cause divorce, or there's a link between some of the molecules and margarine or something. And we get to think about that for a second and then realize that we have no basis for that in reality, that there's just nothing that can confirm that for us and we kind of can reject our own hypothesis outright.

[Ben Crichton]

It all seemed too good to be true, didn't it? Tyler's site is called Spurious Correlations and

[Tim Harford]

he set up a computer program to

[Ben Crichton]

mine data sets and find things which seem to be correlated.

[Tim Harford]

He graphs them and he publishes the

[Ben Crichton]

graphs on his website. So you can see that the age

[Tim Harford]

of the winners of the Miss America contest.

[Ben Crichton]

Correlates with murders by steam, hot vapors and hot objects.

[Tim Harford]

You can also see that worldwide non commercial space launches correlate with, of course,

[Ben Crichton]

sociology doctorates award in the United States. As Tyler says, if you think about these for a minute, or even if

[Tim Harford]

you think about them for a second,

[Ben Crichton]

the idea that there's any causal relationship seems ridiculous.

[Tyler Vegan]

I've seen a lot of headlines, especially sensationalist ones, where it might say scientists find connection between X and Y. And in a lot of those situations there might be a correlation or even a mathematically insignificant correlation between two things. But it's really important for us to be critical of whether that correlation exists, whether we can show that there's actually statistical significance, and whether there's a causal mechanism. It's not that hard to find overlapping lines that kind of curve or vary together. And just because two things vary together on a linear scale doesn't mean that they're actually correlated or that they're actually causing each other.

[Ben Crichton]

Tyler has given us a humorous way to remind us of the golden rule of statistical thinking. This is certainly something we take very seriously here at the BBC. In fact, if I just open the studio door, you might be able to

[BBC Announcer]

hear Staff are reminded that correlation does not equal causation and that sometimes correlation isn't even really correlation. And remember, Alan Yentov's homemade scones are on special in the canteen today.

[Ben Crichton]

For lots of other spurious correlations, check out Tyler's [e-mail] or just search for spurious correlations.

[Tim Harford]

You're listening to, more or less. If you've ever felt a sense of excitement building recently, you'll know we're coming up to a major event which happens only once every four years. I'm talking about the office sweepstake for the World Cup. Who cares about the football when there's a chance to turn a pound one steak into pound 32, which you're then forced to spend on drinks for your colleagues. For those of you not aware of the peculiar institution of the office sweepstakes,

[Ben Crichton]

the basic idea is that you pay

[Tim Harford]

an entry fee into a pot, you draw a team at random, and if your team wins, you scoop the entire lot. Naturally, the thoughts of more or less listeners are turning to sweepstakes strategy, and listener Andrew from Surrey contacted us with this.

[Andrew from Surrey]

I wonder if there's any advantage to be had in not picking first. When the sweepstake was announced, there was a rush to enter, but I held back. Surely those people who entered the sweepstake at the start stand a greater chance of drawing one of the 28 weak teams, because they far outnumber the four favourites and so it's therefore optimal to defer entering. So my question is, does it make sense to wait until a predetermined point

[Tim Harford]

and then enter time to peel the probability onion? The outer layer is Andrew's question. Is there a benefit from entering later rather than sooner? And the answer is no.

[Ben Crichton]

Brazil is as likely to be picked first as last.

[Tim Harford]

What's lured Andrew in? It's the idea that the first team probably won't be any good, which is true. But if the first team out of the hat is Brazil and you've been hanging back, your chance of winning has just taken a hell of a beating. Or to put it more simply, the teams you want are just as likely to come out of the hat first as last. If you're determined to play the sweepstake, Andrew, it doesn't matter when you buy a ticket. But let's look at the next layer of the probability onion. Does Andrew have to play? If he's happy to miss out entirely, then the calculus changes. He should wait and see what happens in the early rounds. If a few poor teams come out of the hat and the favourites are still in there, then Andrew's odds have improved and he should consider stepping up and buying a ticket. If some of the favourites are taken early, Andrew can just wait four more years for a more opportune contest. Now, the decision rule for exactly how long to linger and when to buy in is mathematically complex, so we leave that as an exercise for the loyal listener. The final layer of the probability onion, though, is what happens when not only Andrew but everyone hangs back to await more favourable odds. Why does anyone ever buy the first ticket in a sweepstake? Well, let's leave the probability onion behind us. Having chopped enough to make a delicious bowl of French probability onion soup. And ask the man who's had to grapple with this in our very own office, BBC Radio current affairs sweepstakes supremo Richard Fenton Smith.

[Richard Fenton Smith]

This year I decided to take all the money first and then have a draw. And then it dawned on me, perhaps we should have a draw for the draw. So we're going to have a draw whereby people pick a name out the hat for the order of the draw and then they will pick their team.

[Tim Harford]

Sensible statistical precaution or sad indictment of the likelihood of BBC journalists complaining?

[Ben Crichton]

You be the judge.

[Tim Harford]

And that's all we have time for today. Please keep your questions and your comments coming. We're at more or lessbc.co.uk.com and as always, there's further information and a downloadable edition of the show available at bbcworldservice.com moreorless we'll be back next week. Until then, goodbye.

[BBC Podcast Promoter]

There are dozens of different podcasts now available from the BBC, including news, documentaries, science, business, arts and sports. The details of them all include go to bbcworldservice.com podcasts.