Nate Silver - Statistician

Download Subtitles

Transcript

:00:02. > :00:12.world's busiest ports. You're watching BBC News. Now, it's time

:00:12. > :00:25.

:00:25. > :00:28.for Hardtalk. Welcome to HARDtalk. He is seen as a political

:00:29. > :00:31.astrologer with an uncanny ability to get it right. His prediction of

:00:31. > :00:35.the US presidential election result last year was spot on when most

:00:35. > :00:38.pundits were saying the race was too close to call. No wonder then

:00:38. > :00:43.that Nate Silver is being credited with re-shaping the art and science

:00:43. > :00:45.of political forecasting. But has he robbed electoral campaigns of

:00:45. > :00:55.their substance by reducing them to mere statistics and number-

:00:55. > :01:24.

:01:24. > :01:31.Nate Silver welcome. When it comes to elections, what is the point of

:01:31. > :01:37.prediction. Why not wait and see why people vote? People like to

:01:37. > :01:41.prepare for the future. The word for custom means planning under

:01:41. > :01:47.conditions of uncertainty. Trying to plan who the next president

:01:47. > :01:51.might be, for your personal life or for investing. Good information

:01:51. > :01:57.instead of the bad information you may get from the media. Even if it

:01:57. > :02:03.is a prediction as late as before an election result, what is some

:02:03. > :02:08.point of that? People cannot really make plans. There is a lot of horse

:02:08. > :02:13.race coverage in the States. Some people follow it like they may

:02:13. > :02:17.follow sport. The coverage is often poor when exaggerate how dynamic

:02:17. > :02:21.the horse race is. They claim the candidate behind welcome back and

:02:21. > :02:27.often they will not. They are not looking what the voters are telling

:02:27. > :02:31.them. They are superimposing the pundits's viewpoint on the

:02:31. > :02:38.information. We try to do this and we want to be accurate and truthful

:02:38. > :02:42.and strive to be more scientific with projections. Your predictions

:02:42. > :02:46.consist of three parts, dynamic modelling, data analysis and human

:02:46. > :02:52.judgement. Most people would say you have proved yourself on the

:02:52. > :02:56.first two. How good is your human judgement? She would have to talk

:02:56. > :03:03.to my friends and partners. What you think about your human

:03:03. > :03:08.judgement, due are famous for these computer models? A lot of people

:03:08. > :03:11.are very enamoured with this idea you can make a prediction based on

:03:11. > :03:17.your gut feeling. The evidence shows and a lot of feelings that

:03:17. > :03:26.gut feelings are not very accurate most of the time. Special when it

:03:26. > :03:31.comes to estimating probabilities, we do not really know how the maths

:03:31. > :03:35.works very well. Elgar often leads us astray. You do except it human

:03:35. > :03:44.judgement forms an important part of forecasting and predictions. It

:03:44. > :03:48.is messy? Has people analyse more different factors, it is messy.

:03:48. > :03:53.Whether that judgement, especially if it comes from a pundit that is

:03:53. > :03:58.out of touch with what voters think, and they are going to cocktail

:03:58. > :04:03.parties at Georgetown and serving other A-League friends, the chatter

:04:03. > :04:07.they here, may not be very useful. But if they hear from the campaigns

:04:07. > :04:11.noise and to listen to put that as providing valuable insight and

:04:11. > :04:15.information. Sometimes, sticking with simple things like poles that

:04:15. > :04:22.give us the voters the chance to express themselves to wreck there

:04:22. > :04:26.is a better approach. You process statistics and make models and

:04:26. > :04:30.predictions, when you hear what somebody li?I ? somebody liitish

:04:31. > :04:37.pollsters said he found that you guv, he applauds you and says you

:04:37. > :04:41.are brilliant at taking a lot of data, understanding it, he says

:04:41. > :04:46.without opinion polls and surveys they would be no Nate Silver. You

:04:46. > :04:49.did well because the polls did well. If you have taken every just of

:04:49. > :04:55.those polls you would have reached the same conclusion? I could have

:04:55. > :05:00.done that. Anybody could. But you get all this attention? They is

:05:00. > :05:05.because in politics the one-eyed man is the king in the land of the

:05:05. > :05:08.blind. The coverage is so poor a lot of the time. You have to sell

:05:08. > :05:15.newspapers and television programmes to say it will be a

:05:15. > :05:19.close race. You want to appear unbiased. Obviously, people

:05:19. > :05:24.involved in politics directly were trying to spin a narrative of how

:05:24. > :05:28.the party may win, a lot of inaccurate information. When he

:05:28. > :05:35.sees the did well because the polls were good and you manage to key in

:05:35. > :05:41.is that a key point that anybody could have done it. I said before

:05:41. > :05:47.the election, I would get too much attention of Barack Obama 1 and too

:05:47. > :05:51.much blame if Mitt Romney won. I look at probabilities, there is a

:05:51. > :05:55.margin of error with polls. We look at how accurate polling has been.

:05:55. > :06:01.You have many Poles from many different states which can increase

:06:01. > :06:09.the accuracy. That can influence whether you place a bet. We do not

:06:09. > :06:15.know what is going to happen, it is about Wayne information and data

:06:15. > :06:21.within the context of uncertainty. Did you generate any date yourself?

:06:21. > :06:28.Were each generate data with the models. Do you actually bring in

:06:28. > :06:33.fresh data yourself? I am not a pollster. You're likely lead on the

:06:33. > :06:40.work of opinion polls? The use economic data and a judgement as

:06:40. > :06:46.well in a review of the empirical scientist. I am very widely read.

:06:46. > :06:50.We do not rely on spin from campaigns. We do not rely on

:06:50. > :06:54.conventional wisdom which is not worth much in Washington DC. We

:06:54. > :06:59.rely on poles which reflect what individual people say. Polls that

:06:59. > :07:06.go out and interview action will voters and what they think about

:07:07. > :07:11.the election. As Michael Gerson a prominent member of the Republican

:07:11. > :07:15.establishment, an aide to George bush said, elections are not a

:07:15. > :07:21.mathematical equation? These are the people that thought Mitt Romney

:07:21. > :07:27.would win. That may be the case, there. Is your approach trivialises

:07:27. > :07:37.politics. It's trivialises their ability to spin and confuse people.

:07:37. > :07:39.

:07:39. > :07:46.They have an interest in a site like mine which takes the view from

:07:46. > :07:54.data and best practices. They were proven wrong last November. Michael

:07:54. > :07:58.Gerson makes the point. Why should we be listening to him? He is a

:07:58. > :08:03.senior member of the bush establishment. Why is that a

:08:03. > :08:08.credential? Used to measure public opinion. The measure what people

:08:08. > :08:16.think and how voters may behave. build models based on polling data

:08:16. > :08:23.and economic data. To discuss what forms public opinion. As the

:08:23. > :08:31.election came to a close, we talked to Nate Silver's statistical model,

:08:31. > :08:36.there was not much about social mobility and debt. He said your

:08:36. > :08:40.approach was devoid of substance and trivialises politics.

:08:40. > :08:47.approach is to write a better version of the horse race coverage

:08:47. > :08:50.we see in the United States media every day. It should not be

:08:50. > :08:57.difficult to look at the polls and take an average of them and see who

:08:57. > :09:02.is head. And why do not we trust parties since to talk about the war

:09:02. > :09:06.if they cannot get basic facts right. Adding up one converses the

:09:06. > :09:12.other. I'm not sure why we should listen to sum up like that on

:09:12. > :09:16.complex issues. Because he is a Republican? Because he is wrong. I

:09:16. > :09:23.think he's deluded about what reality is. His biases have

:09:23. > :09:28.overcome his ability to see the polls. I am biased? Are you vised?

:09:28. > :09:34.They everyone is biased. They historian Kim Stanley is accusing

:09:34. > :09:42.you of predicting a win and making him look like a winner, and it is

:09:42. > :09:52.your goal to create, convincing Obama will win because he is a

:09:52. > :09:54.winner. I am telling you what Tim Stanley says. If you take two

:09:54. > :09:59.people are not credible and take the arguments and ignore lots of

:09:59. > :10:09.other people in the field that think highly of my research. It

:10:09. > :10:13.makes for better TV. I was many same that. Looking at data from

:10:13. > :10:19.other interviews to create better ratings for this show what type of

:10:19. > :10:25.journalistic values are those. putting up to that Michael Gerson

:10:25. > :10:30.is being partisan. I think he is wrong. He would say he has a fair

:10:31. > :10:35.point to make. He is saying what you do makes elections devoid of

:10:35. > :10:40.elections by reducing them to a mathematical equation. We have gone

:10:40. > :10:43.through that. Do you worry that even if you if you are not being a

:10:43. > :10:48.vet the partisan that he might be used as a political tall? A thing

:10:48. > :10:54.the media coverage influences the way people can vote and behave. We

:10:54. > :10:58.are trying to use simple models that take poll's as direct measures

:10:58. > :11:03.of public opinion and take them back to the public domain. I do not

:11:03. > :11:09.like the business of politics, I do not live in Washington. The goal is

:11:09. > :11:15.to be to inform the public. Does it worry you that Barack Obama said at

:11:15. > :11:19.a dinner in March this year there are some one very special in my

:11:19. > :11:24.life who is missing who always has my back and stands with me no

:11:24. > :11:34.matter what the matter how dark things seem, my rock, my foundation,

:11:34. > :11:35.

:11:35. > :11:38.thank you Nate Silver. That was a joke. So he just said it as a joke.

:11:38. > :11:45.That is the number that is the dinner where he makes a number of

:11:46. > :11:55.jokes. His is on the record that you are a democrat. Do you support

:11:55. > :12:03.Barack Obama. I'm am on the centre- left in the US. The idea that

:12:03. > :12:07.people in politics should not have political views is ridiculous. I

:12:07. > :12:09.say that is my subject of a vantage-point, but I'm looking at

:12:09. > :12:15.an object of world we all share. Can I make accurate judgements

:12:15. > :12:18.about that world. By making verifiable predictions and claims

:12:18. > :12:23.that are not the biggest to be inscrutable, you can test on

:12:23. > :12:27.reality. People in the Republican Party for Mitt Romney, came close

:12:27. > :12:30.up to reality when they saw they were looking at the world and a

:12:30. > :12:35.jaded and biased way. You are an interesting phenomenon because you

:12:35. > :12:40.said I try not to talk to campaigns because they are mostly noise. You

:12:40. > :12:45.talk about politics being a game with a lot of and vested interests.

:12:45. > :12:50.ECGD not like to get into political debate and live in Washington. You

:12:50. > :12:57.cannot remain above the fray? think most journalists would not

:12:57. > :13:01.like to be the story. Part of the reason is that my block became

:13:01. > :13:07.symbolic of the polls. When you become invested with symbolic power

:13:07. > :13:12.it is not always your choice. It is a distraction. There are other

:13:12. > :13:19.sites that take a similar approach to what they do. They had largely

:13:19. > :13:22.similar results. All of them by election day, had Barack Obama

:13:22. > :13:26.ahead in the electoral college. When you have any type of signs or

:13:26. > :13:31.academic inquiry we have consensus view., different assumptions that

:13:31. > :13:37.you the same result, that is richer than when your model is based on

:13:37. > :13:42.one parameter. That might be changed. You can no longer remain

:13:43. > :13:47.above the political fray? I think I can. I do not really care about

:13:47. > :13:54.politics. There are things I worry about sports and economics. I have

:13:54. > :14:00.other ways to make a living. I have political views but as compared to

:14:00. > :14:04.people most people, I am a more detached from that world. The study

:14:04. > :14:08.politics at the University of Chicago, then you joined KPMG, you

:14:08. > :14:13.talk to online gambling and made $0.5 million from that. You said

:14:13. > :14:18.Poker gave you better training than anything I can think of about how

:14:18. > :14:24.to weigh new information and what may be less important information.

:14:24. > :14:34.The went into looking at baseball and looking at how players too. His

:14:34. > :14:50.

:14:50. > :14:56.politics another game like poker or You will get lucky sometimes as

:14:56. > :15:01.well. What is the best decision to make under an environment of

:15:01. > :15:08.uncertainty. That perspective is helpful. Knowing what we can and

:15:08. > :15:18.cannot control. Light politics where people take hyperbolic fuse.

:15:18. > :15:18.

:15:18. > :15:26.Manipulating public opinion. -- the views. You want $0.5 million. Why

:15:26. > :15:32.did you stop? That was at the peak. We lost some money after that.

:15:32. > :15:36.Poker was a bubble economy. In the meagre 2000, it was being on

:15:36. > :15:42.television. A lot of players who were not very good were depositing

:15:42. > :15:48.money. I took their money. Eventually the players got better

:15:48. > :15:53.off the bat once gave up all went broke. Anyway, you say the

:15:53. > :16:00.principles for piker, baseball, games and politics all were. Can

:16:00. > :16:06.your model predict any event? Could it have predicted the Arab spring

:16:06. > :16:11.for instance? We look at individual feels. Where there is some degree

:16:11. > :16:18.of predictability. Especially when the competition is not handling it

:16:18. > :16:23.in a very intelligent way. Sports, in the US, ten years ago baseball

:16:23. > :16:27.teams were not very sophisticated. Looking for cases like that.

:16:27. > :16:32.Predicting global and international events is something of a fool's

:16:32. > :16:42.errand. The best that democracies can do is prepare themselves

:16:42. > :16:42.

:16:43. > :16:47.overwriting of outcomes. Your success depends on... In the detail

:16:47. > :16:57.modelling Euan Dunn, worked on the US because there is so much data

:16:57. > :16:57.

:16:57. > :17:01.available. -- you haven't done. would be much more cautious about

:17:01. > :17:07.making predictions about because in Egypt. People may not be

:17:07. > :17:12.comfortable telling their honest opinion. In some countries, where

:17:12. > :17:17.the votes are counted fairly. what about the financial crisis for

:17:17. > :17:21.which there is a great deal of data and there was - could that have

:17:21. > :17:29.been predicted? There were a lot of people who understood the housing

:17:29. > :17:33.bubble well in advance. I should stay that the housing market was

:17:33. > :17:43.the relevant in the room. Seen the prices go up in the US and Western

:17:43. > :17:46.Europe, in the way that you always had a crash to follow. That was an

:17:46. > :17:52.obvious truth. One of the lessons of the book is that you can build

:17:52. > :17:56.are these very complicated and fancy models but if you lack of

:17:56. > :18:03.judgement it still lead you astray. You're talking about the book, the

:18:03. > :18:09.signal and the noise. Amaze all that noise and data, all that data

:18:09. > :18:15.we are overwhelmed with, looking at the financial crisis, do you thank

:18:16. > :18:21.you would be able to, however you do it, to say yes, the euro will

:18:21. > :18:25.survive or not it will crash? of the findings of the book is that

:18:25. > :18:30.the more sure of themselves the pound it is the less likely the

:18:30. > :18:35.outcome. I do not have a strong opinion on whether the eurozone

:18:35. > :18:40.will remain intact or not. I think it teaches us lessons

:18:40. > :18:43.retrospectively about maybe what happens with unknown unknowns. We

:18:44. > :18:50.take a big complex system that is working more less well and you

:18:50. > :18:53.change it, you can anticipate a problem like a common currency but

:18:53. > :18:59.the centralised control of public finances that will create

:18:59. > :19:04.difficulties. I do not think people anticipated how that would player.

:19:04. > :19:10.How funny you predict about how something is going to happen - the

:19:10. > :19:14.British elections in 2015 - the data is too premature? Politics and

:19:14. > :19:18.economics are closely tied together. Economists have been unable to

:19:18. > :19:23.predict recessions or recoveries more than six months in advance. To

:19:23. > :19:28.know the state of British economy in 2015 you would know more about

:19:28. > :19:31.how the politics would turn out all the elections were turnout. What

:19:31. > :19:36.about climate change signs which may spread dish -- predictions

:19:36. > :19:44.about what is going to happen in 70 years' time? Are you saying that is

:19:44. > :19:49.really based close? The reason why in the climate signed predictions

:19:50. > :19:54.but Test worthy is that they are based on theory which has been

:19:54. > :19:59.tested on the bases of relationships that have been

:19:59. > :20:04.understood for years. And you see a basic warming trend that correlates

:20:04. > :20:11.with carbon dioxide emissions. Climate change science can forecast

:20:11. > :20:18.what is going to happen to the earth in decades to come? If you

:20:18. > :20:22.Trapmore Hague in the atmosphere -- if you trap more heat in the

:20:22. > :20:26.atmosphere then you can predict something. But I cannot predict

:20:27. > :20:33.what is going to happen in Ireland in the summer of Myatt they know

:20:33. > :20:35.2058. The basic notion that the plan and is getting warmer and when

:20:35. > :20:42.things get that the economic consequences is on solid ground.

:20:42. > :20:47.There are those people, like the professor of statistics in

:20:47. > :20:53.Pennsylvania, says that there is a lot we need to know about the

:20:53. > :20:57.climate. These are her simplified representations of complex systems.

:20:57. > :21:04.These are just approximations. course, I agree with that entirely

:21:05. > :21:09.but the question is what evidence heavily absorbed so far? Could you

:21:09. > :21:19.discount that there is a big freeze that might happen a new Ice Age?

:21:19. > :21:19.

:21:20. > :21:24.think it is unlikely. But possible? Do you want to face a bet? I do not

:21:24. > :21:28.know if we're going to be around. I want to ask about the nature of

:21:28. > :21:33.predictions. The odds of a big freeze are hundreds to one against.

:21:33. > :21:38.When it comes to you, or something like climate change science based

:21:38. > :21:43.on predictions and forecast, and you accept what the professor has

:21:43. > :21:46.said, does one go to policy makers and politicians and public and say,

:21:46. > :21:52.we will base their policies on these predictions and commit

:21:52. > :21:56.billions and billions of dollars... A lot of policy positions have been

:21:56. > :22:04.based on predictions that were on a much less solid ground and then

:22:04. > :22:10.that, for example the invasion of Iraq. If, the decision to enter the

:22:10. > :22:13.and have a common currency. And there much more systematic

:22:13. > :22:21.uncertainties than in climate change. We have to make decisions

:22:21. > :22:25.all the time. With the information available. A statistician from the

:22:25. > :22:28.Copenhagen looked at climate change and accepts that it is happening

:22:28. > :22:33.undeniably but says the narrow focus of reducing carbon emissions

:22:33. > :22:38.that some governments have could leave future generations with major

:22:38. > :22:42.costs without major cards to temperatures. Does he have a point

:22:42. > :22:47.of as my this is not about the signs are predictions. This is

:22:47. > :22:53.about what we do as or alleviation technique. If we have that debate

:22:53. > :22:59.over here in Europe, in the US you have people still engaged in a

:22:59. > :23:04.scientific debate. If you read the chapter in my book, I'm not someone

:23:04. > :23:11.who says everything is always right on the climate science and be back.

:23:11. > :23:15.It is still at a premature debate in the US. You say in your book,

:23:15. > :23:20.you hope we might get more insight into planning our futures and

:23:20. > :23:26.become less likely to repeat our mistakes. That is right.Therefore,

:23:26. > :23:31.there is a purpose to predictions and forecasting. That is the whole

:23:31. > :23:35.point stop on his swing, in my view, forecasting is not the most

:23:35. > :23:38.critical thing in the world. But some of those same lessons you can

:23:39. > :23:42.learn from those fields are instructive when it comes to

:23:42. > :23:45.economic planning or obviously winning comes to anticipating where

:23:45. > :23:54.the climate might go and what we can do to change that and mitigate

:23:54. > :24:00.the effects. Wycombe and a lot from games, sports and politics. -- we