Predicting the Presidency

Summary

More or Less — "Predicting the Presidency"

Host: Tim Harford (BBC Radio 4)
Date: October 8, 2012

---

Episode Overview

This episode of "More or Less" explores the use of statistics in political prediction, focusing on Nate Silver's influential election forecasts and the Bayesian methods underpinning them. The show also examines a quirky listener question about the impressive longevity of US presidents' mothers compared to those of British prime ministers. The tone is engaging, accessible, and lightly humorous, with a focus on demystifying statistical thinking for the general audience.

---

Key Discussion Points and Insights

1. Nate Silver and Political Prediction

(00:13–05:52)

• Nate Silver's Rise to Prominence

  • Tim Harford introduces Nate Silver, a statistician famed for his baseball predictions who turned to politics and gained national attention in 2008 by accurately predicting 49 out of 50 state results and all Senate races in the US presidential election.
  • "This young, bespectacled baseball nerd was suddenly the most talked about statistician in America." (Tim Harford, 02:14)

• The Role of Thomas Bayes

  • Silver’s predictive success is linked to Bayesian statistics, specifically Bayes' theorem, which helps update predictions as new data comes in.
  • "What Bayes theorem is...it's just a very simple mathematical formula that tells you what do you do when you encounter new information.” (Nate Silver, 02:42)

• Bayesian Framework in Forecasting

  • When new polls emerge, instead of discarding previous predictions, Silver combines old and new data to refine the probabilities systematically.
  • "Bayesians like Silver don't throw out their old forecasts...the new data is used to adjust rather than replace what went before." (Tim Harford, 03:33)
  • Silver emphasizes skepticism and systematic processing: "You use all the information, but tend to use it more skeptically...design a set of rules ahead of time for how you process these polls." (Nate Silver, 03:43)

• Accounting for Uncertainty and State-Level Analysis

  • Silver underscores the importance of analyzing elections at the state level and communicating the odds and uncertainty for each candidate in every state.
  • "That's probably the most challenging...Obama's ahead by, say, four or five points. But what's that translate to in terms of odds? What odds would you have to have to place a bet on Mitt Romney?" (Nate Silver, 04:38)
  • He notes the similarity between his statistical odds and those offered by bookmakers: "You couldn't really make money using our predictions right now because the consensus is where we happen to be as well." (Nate Silver, 05:07)

• Added Value of Statistical Sites vs. Bookmakers

  • Tim asks about the usefulness of 538 if the odds closely match those at bookmakers.
  • Silver responds: "Frankly people probably should consult bookies more often when they're looking at the risk of political events...If you're a gambler, you have to be more precise than [the media]." (Nate Silver, 05:26)
  • He warns that media coverage oversimplifies probabilities, swinging from uncertainty to unwarranted certainty.

• Practical Example

  • As of the recording date (October 4, 2012), Silver's model gives Obama an 86% chance of re-election. (05:52)

---

2. The Mystery of Presidential Mothers’ Longevity

(06:14–08:56)

• Listener’s Statistical Query

  • Listener Mike Shearing notes many US presidential candidates had mothers living to very old ages and asks if this is statistically significant.
  • "Are these figures statistically unusual or significant? How often should I get out?" (Mike Shearing, 06:14)

• Expert Analysis

  • Dr. George Leeson, a gerontologist at Oxford, examines age-at-death data for presidential and prime ministerial mothers.
  • Both groups’ mothers show the expected normal, bell-curve distribution, but the US cohort skews older:
    • "The US curve has moved further to the right...we’ve got more people in the distribution who are reaching 100, in their 90s in the US simply because the curve has all moved to the right." (Dr. George Leeson, 07:02)

• Is It Statistically Significant?

  • Despite small numbers (~20 in each group), Leeson affirms the difference is statistically significant.
  • "Absolutely, yes. It's so significant." (Dr. George Leeson, 07:44)
  • Life expectancy varies by cohort and birthplace; interestingly, UK mothers of prime ministers outlive their contemporaries by a greater margin than their US counterparts, but the absolute ages of US mothers are higher.
  • "The US mothers are living much longer than the comparable cohorts of UK Prime Ministers mothers." (Dr. George Leeson, 08:56)

• Conclusion

  • "There is an advantage to being a leader's mother, but the advantage is much greater in the US than in the UK. Who knew?" (Tim Harford, 08:56)

---

Notable Quotes & Memorable Moments

• Nate Silver on Bayesian Updating:
  "What Bayes theorem is...it gives it a way for you to weigh new information against your previous conception of a race." (Nate Silver, 02:42)
• On Interpreting Polls:
  "You're always going to look at prediction as a work in progress." (Nate Silver, 03:21)
• Media vs. Bookie Odds:
  "If you're a gambler, you have to be more precise than [the media]. There's a big difference between 2 to 1 odds and 5 to 1 odds." (Nate Silver, 05:26)
• Presidential Mothers' Longevity:
  "The US curve has moved further to the right...we’ve got more people in the distribution who are reaching 100." (Dr. George Leeson, 07:02)

---

Timestamps for Key Segments

• 00:13 — Introduction to Nate Silver & Thomas Bayes
• 02:42 — Nate Silver explains Bayes' theorem
• 03:43 — Handling polling data in political prediction
• 04:38 — State-level uncertainty and betting odds
• 05:26 — Discussion of bookmakers vs. statistical models
• 05:52 — Example of odds: Obama’s 86% likelihood in 2012
• 06:14 — Listener email: Presidential mothers and old age
• 07:02 — Dr. George Leeson on statistical analysis of longevity
• 08:56 — Conclusion on the “leader’s mother effect”

---

Summary

This episode delves into how modern political forecasting, epitomized by Nate Silver, relies on Bayesian statistics to treat prediction as a constant process of refinement, integrating prior beliefs and new poll data. The segment on mothers’ longevity adds a lighter, quirky touch, yet illustrates solid statistical reasoning. Listeners come away with a richer understanding of how to interpret probabilities and the ongoing nature of “knowing” in the face of new evidence—whether predicting elections or comparing presidential parents.

Transcript

[BBC Announcer]

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

[Tim Harford]

hello and welcome to More or Less on the BBC World Service, your weekly dose of mathematical medicine. I'm Tim Harford. This week's prescription is a little different. Today we're meeting two stars of the statistical world. One is a 34 year old new Yorker called Nate Silver, who's just published a book, the Signal and the the Art and Science of Prediction. The other was an English Presbyterian minister who died in 1761. His name was the Reverend Thomas Bayes. Nate Silver says he can predict the result of the forthcoming US Presidential election. And Thomas Bayes, well, Thomas Bayes helps him do it. Let's go back to 2008. Nate Silver had already achieved some success forecasting baseball results. But as Senators McCain and Obama slugged it out for the presidency, Silver switched from predicting batting averages to votes.

[Narrator/Reporter]

He managed to enrage half of Chicago last year by correctly predicting the White Sox World Series winners of 2005 would finish with 72 wins and 90 losses. They finished with 72 wins and 90 losses. He has moved up in the world now. And now John McCain's strategists and supporters are going to be mad at him. Or they will be after they see this. Our third story in the countdown, why a man who has elevated statistical predictions in sports to an art form now says it's statistically more probable that Barack Obama will win the 2008 election.

[Tim Harford]

Nate Silver was right, my friends.

[Barack Obama]

We have. We have come to the end of a long journey. The American people have spoken and they have spoken clearly. A little while ago, I had the honor of calling Senator Barack Obama. To congratulate him, please. To congratulate him on being elected the next president of the country that we both love.

[Tim Harford]

In fact, Silver wasn't just right, he was spectacularly right. He correctly predicted which way 49 of the 50 states would go. And he got every US Senate race right, too. This young, bespectacled baseball nerd was suddenly the most talked about statistician in America. That was then. Now, as you might have noticed, Barack Obama is in another presidential race. And of course, just like last time, the TV networks are awash with polling data. This time, Nate Silver's blog, FiveThirtyEight, is being very closely watched. But one of the things you notice about Silver's predictions is that they keep changing. And this is where Thomas Bayes comes in.

[Nate Silver]

What Bayes theorem is and it's become very fashionable nowadays. If you read mathematics or statistics journals, it's just a very simple mathematical formula that tells you what do you do when you encounter new information. And it gives it a way for you to weigh new information against your previous conception of a race.

[Tim Harford]

So Bayes Theorem says that you start with what you think you know, then adjust those beliefs as new evidence emerges. It also provides a precise mathematical method for doing this. Imagine all the polls broadly agree on the election result and you think you know who's going to win. Then a new poll comes along which suggests a wildly different result.

[Nate Silver]

How much should that poll change your view of the odds? And that's literally what Bayes Theorem is trying to do. And it says, among other things, that you're always going to look at prediction as a work in progress.

[Tim Harford]

Bayesians like Silver don't throw out their old forecasts. When the new poll numbers come in, the new data is used to adjust rather than replace what went before.

[Nate Silver]

These polls are reported, and people tend to treat each of them as being a highly newsworthy event unto itself. My view is that first of all, you use all the information, but tend to use it more skeptically. And I also use it in a way that's systematic. So you design a set of rules ahead of time for how you process these polls, and it spits out essentially a set of odds that updates each day as new polls come in. But it does ensure that you're looking at the data in the hole and looking for kind of what the consensus will tell you and not. And not becoming overly fixated with any one survey because they will bounce up and down quite a bit.

[Tim Harford]

And of course, you always try and break things down by state, which clearly the US election is determined state by state. It's not determined by the overall poll. And you're trying to give a spread of the level of uncertainty as well. How likely is it that each of the candidates will actually take each of these states?

[Nate Silver]

Yeah, that's probably the most challenging is you can see that Obama's ahead by say, four or five points. But what's that translate to in terms of odds? What odds would you have to have to place a bet on Mitt Romney? And we give you that number every day. Interestingly, I guess the numbers we have are often very, very similar to the ones that if you go and want to make a bet at a bookie in the UK or in Vegas, are the ones on offer there. Where right now Romney's about a 5 to 1 underdog.

[BBC Announcer]

It looks like.

[Nate Silver]

And so, you know, you couldn't really make money using our predictions right now because the consensus is where we happen to be as well.

[Tim Harford]

If I could just go to a bookie and find out the odds and that's pretty much the same as I'm getting from your site. What's the added value of a site like 538?

[Nate Silver]

Well, you know, I think frankly people probably should consult bookies more often when they're looking at the risk of political events. Sometimes if you look at the portrayal in the media, they tend to to shift from saying oh, it's too close to call to saying oh, one candidate's going to win for sure. And if you're a gambler, you have to be more precise than that. There's a big difference between 2 to 1 odds and 5 to 1 odds, even though you have one candidate who's a favorite in either case.

[Tim Harford]

It's Thursday the 4th of October as I record this the day after the first presidential television debate. And Nate Silver's blog currently gives Barack Obama An 86% chance of of retaining the presidency. We'll see. Now, sticking with US presidents or at least their mothers, here's an intriguing email from Mike Shearing, a listener in Malaysia.

[Mike Shearing]

It's come to my attention that of 24 Republican or Democrat US presidential candidates since the war, three of them had mothers who lived to 100, six more who made it to 90 and five more who have made 80 plus mothers of British prime ministers seem to fare much worse. Are these figures statistically unusual or significant? How often should I get out?

[Tim Harford]

Frankly Mike, you should get out more. But thanks for your question. We collated the age of death for every mum of a post war US president or losing presidential candidate, except those who are still alive of course. And for the mothers of every post war British Prime Minister and losing opposition leader. Then we sent the whole lot to Dr. George Leeson, a gerontologist at Oxford University in the UK.

[Dr. George Leeson]

What we would expect to find with ages at death is that we would have what we call a normal distribution. This is a nice bell shaped curve. So we would have most people congregated around the middle observations and then we'd have some small numbers usually out at the extremes. And we do have that for both of these groups of mothers. However, if you like, the US curve has moved further to the right so that as your listener pointed out, we've got more people in the distribution who are reaching 100 in their 90s in the US simply because the curve has all moved to the right.

[Tim Harford]

So we're fairly sure that the difference in the average is real?

[Dr. George Leeson]

Absolutely, yes. It's so significant.

[Tim Harford]

Now, might it still be a coincidence because we're still only talking about 40 people? I suppose about 20 in each?

[Dr. George Leeson]

We are, yes. And it is a small number of observations, but we can begin to look at other things where we might want to get a more clear picture of what is actually happening. Because, of course, there are a number of things that affect your age at death, what we call life expectancy. And one of them, of course, is the time in which you are born. For every decade going back to the 1840s, when we have our first mother born, to the 1940s, when we have our last mother born, for every decade, the life expectancy of that cohort of women in the UK is higher than the life expectancy for that cohort in the United States. The UK mothers of Prime Ministers are outliving their peers quite significantly. The US mothers are also outliving their peers. However, the US mothers are living much longer than the comparable cohorts of UK Prime Ministers mothers.

[Tim Harford]

Dr. George Leeson. So there is an advantage to being a leader's mother, but the advantage is much greater in the US than in the uk. Who knew? Well, that's all we've got time for this week. Next week, Yan Wong will be back with more number y. Goodness. But for now, from me, goodbye.

[BBC Announcer]

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