Are we still hanging on the wisdom of the Nate Silver? The headline suggests he knows he has a credibility problem. (This time he really means it.) But let's look beyond the headline. There's a lot of stuff crunched over there — that's the confidence-building methodology — and here's how it ends:
There’s still some uncertainty about the outcome.... Although the polls haven’t systematically underestimated nationalist and right-wing parties, they also haven’t been all that accurate in pinning down their support, having come in both high and low in different elections. In cases since 2012 where the right-wing party polled at 25 percent or more, the polls missed the party’s actual support by an average of 3.6 percent of the vote. That translates to a true margin of error (or 95 percent confidence interval) of about plus or minus 9 percentage points. And because any vote that Le Pen gets is one that Macron won’t get, the margin of error for the gap between Le Pen and Macron is twice as large, or about 18 percentage points.That's excitingly put and it sounds strong. As I said, confidence-building. But we're all skeptical now, aren't we?
An 18-point margin of error is huge! But it still isn’t enough when you’re 26 points behind, as Le Pen is against Macron.... She could beat her polls by as much as Trump and Brexit combined and still lose to Macron by almost 20 points.
This was the final prediction, published on America's Election Day, at Nate Silver's FiveThirtyEight:
Posts
67 comments:
Heh.. ouch that graphic leaves a mark. I think a 'nuff said is in order.
I don't get the fascination with LePen among conservatives. She's a socialist in every policy that matters. The pro immigrant left likes to paint her as 'far right' only because they want to associate right and conservative with racist. If she is racist and not just anti immigrant, then she's a lefty racist.
It rains on a lot of days with a 28% chance of rain.
I don't think conservatives particularly like LePen, except that she is despised by all of the usual suspects, and we can identify with that. She is a stick in the eye for the right people. We don't have to live with her policies. Hillary should be really proud about how her time as SoS and her role as the farce that launched a thousand refugee ships.
Stop taking polls.
At least in the U.S., conservatives like to lie to pollsters. I know I do.
The only poll that matters is the election. All others are meant to influence the election.
About how here time as SoS has destabilized Europe.
Speaking of polling, Hillary would still lose. This time by a lot more! LOL.
The pollsters asked for second choices and those went overwhelmingly to Macron. She did best with Fillon voters but Fillon immediately endorsed Macron. I spoke to 2 Fillon voters who will vote Macron.
As I understand it, from the conservative point of view, we have already won, as the entrenched elites in France take a poke in the nose, either way. The Democrats are the party of entrenched elites, which is why the billionaires like them so much. If you have billions of dollars, what you really want is a government that is for sale. Democrats are more than happy to oblige. How did Harry Reid get so rich in a life of "public service"? How did the Clintons get to be worth hundreds of millions of dollars never holding a job outside of government that paid much at all.
"Left Bank of the Charles said...
It rains on a lot of days with a 28% chance of rain."
Sometimes it pours.
Jamais Le Pen.
The globalists have overplayed their hand. They are as scared of this France Firster as they were of America Firster DJT.
Greg said...
I don't get the fascination with LePen among conservatives. She's a socialist in every policy that matters. The pro immigrant left likes to paint her as 'far right' only because they want to associate right and conservative with racist. If she is racist and not just anti immigrant, then she's a lefty racist.
You are correct, you don't get it.
Any of it.
Nate Silver writes the press releases for elite opinion. That's his job. That's what gets him invited to all the best parties, the ones with movie stars.
"She's a socialist in every policy that matters."
She is but that's true of just about every French pol, isn't it? Fillon is the closest thing to an economic conservative they have. The French won't give up on their social welfare state until it comes crashing down, no matter how many of their more ambitious young people leave for more dynamic economies like the UK. They're left with an aging native population, a pool of unmotivated and unemployed or underemployed young people and a hell of a refugee problem - and those refugees will be an additional drain on the public coffers.
Still, if that's the route they chose for themselves, so be it. I'm not a leftist who wails and sobs about how democracy is dead when elections don't go my way.
It's a true thing that the food in Britain has gotten a lot better as young French entrepreneurs leave France and open restaurants in the UK. Economic liberalism is dead in France. Still, any poke in the eye for the EU leadership that fancies themselves "post democratic" is a good thing.
I'd say this is an "asymmetrical" election.
If LePen wins, it's another poke in the eye to the global elites following Brexit and Trump. So that's good.
If LePen loses, well, it's still France. No need to worry about it too much. I assume most of its citizens want the atheist, socialist, bisexual, multi-cultural value system in the ascent. The problem, of course, is when Muslims start to plant their flag...,,,
"I don't get the fascination with LePen among conservatives. She's a socialist in every policy that matters."
If your country is being invaded by people who would violently overthrow your way of life, and some of whom would happily behead you, concerns about economic policy become secondary.
People used to say France is wonderful if only there were no Frenchmen there.
We are going to find out but the old "Nature abhors a vacuum" thing is also true and it is filling that beautiful country with 7th century Medievalists. It may be that cities, like those here ruled by Democrats, become hellholes and the countryside is the last refuge of modernity. The French countryside is filling with English retirees. There are now towns in south eastern France that are 100% English speaking.
Who killed Seth Rich?
The stock market is going to be up big this morning based on the futures because there isn't a socialist in the runoff. The big worry was that both a socialist and Le Pen would be in the runoff. Money, money!
Greg: I don't get the fascination with LePen among conservatives. She's a socialist in every policy that matters.
I don't get the insistence on interpreting people's views via no longer useful categories. I know we use "liberal" and "conservative" around here as convenient shorthand for "people whose views I don't like", but they're leftover categories from the post-War 20th century American political landscape, which no longer fit very well.
Once you stop assuming that any American "conservative" is some variant of a WSJ (or National Review or George Mason University) economic conservative, it becomes obvious why some American "conservatives" might have positive views of Le Pen. Same thing applies to lots of people who get slotted as "lefties", too.
James Kahn said...
I was just about to make the same point. As Ann Coulter so aptly pointed out in the lead up to our election, if you don't control immigration you're not going to have a country, then nothing else matters.
Survival instincts don't lend themselves to nuance and compromise.
AprilApple said...
"Who killed Seth Rich?"
A voice cries out in the wilderness, but sadly on deaf ears, liberal ears.
I always have doubts when foreigners try to devine the voting preference of a country.
Not that I care who wins. Neither are terribly good for us.
"Heh.. ouch that graphic leaves a mark. I think a 'nuff said is in order."
-- I want a time lapse GIF of the NYT's election predictor that updated through the night, going from something like 80% Hillary Clinton to a Trump win. I followed that and CNN, and you could sort of track exactly as CNN's heart broke as it ticked ever rightward.
I think the reason that "conservatives" like Le Pen is that there is a pent up desire to see misery visited on a large scale to a Muslim population resident in a Western country. She seems the most likely person to do it.
As to why this could be the case, I suppose that years of terror, Middle-East fatigue, and PC pro-Muslim propaganda have taken their toll.
I get the sense that I would like LePen less if I knew more about her, but that's true about just about every French politician. At any rate, the markets are way up. I don't interpret that as a good sign for her chances. The people who put their money where their mouths are are not betting on her.
No one loses more with a LePen victory than Angela Merkel.
"Marine Le Pen Is In A Much Deeper Hole Than Trump Ever Was."
One might say, ahem, Nate Silver's "holier-than-thou" attitude toward Le Pen compared to Trump is sexist.
Don't know what you mean by Silver's credibility problem. He did very well on the Trump election. Unlike pretty much every other statistician, who kept insisting that the election was "over" and Clinton had won, Silver kept insisting that Trump had a pretty good chance right up to election day. Even when the liberal media pilloried him for it.
As for Le Pen's chances, it's pretty simple. If nothing changes, she loses handily. But ISIS may decide to hack the election. A couple of terrorist attacks before the election and who knows what will happen. Imagine if Macron makes fun of Le Pen's paranoia in a debate, and then there's a major truck attack the next day.
So go ahead, statistician, tell me the probability of that.
I talked about this a lot concerning the Clinton email scandal. Liberals refused to believe it could be a problem. Conservatives were sure she was toast - except there might be a fix in. No one had a clue what the FBI would do, except everyone said they did, and the FBI had a pretty broad range of choices. No statistician can plan for that, but a smart one (like Silver) has to include a buffer for it.
I suppose it depends on how suicidal the French are at this point. For a country that can be exceedingly pedantic about their culture, it amazes me that they also insist on importing foreigners that hate French culture, want French culture to die, and are not being the least bit sly about it. It appears the highest priority is the short-term survival of the welfare state status quo, everything else be damned.
To channel my inner Laslo, France is a man standing on a stool, head in a noose, getting "serviced" by a beautiful woman down below. Len Pen is the only one who recommends taking off the noose. Everyone else does not want to change anything. The only possible exception is the Communist who likes the noose just fine, but this the woman is too bourgeoise and wants to start a prostitution collective which will probably involve a mean homely woman coming by once a month to spit on you.
Excellent article here:
https://www.city-journal.org/html/french-coming-apart-15125.html
"and we can identify with that. She is a stick in the eye for the right people. We don't have to live with her policies. "
Sickening to think some people believe that. As if the destruction of NATO and the breakup of the EU will not affect the US.
Yeah, but ... Silver was right. Other forecasters had Hillary at 90, 95, even 98%. Silver was the only one warning that there was a good chance of a Trump victory. And if a 25% chance result doesn't happen, you know, 25% of the time, you're doing your statistics wrong.
Moreover, Silver identified *why* Trump had a puncher's chance: large numbers of undecided voters, sparse polling in swing states, unjustified assumptions about Clinton's "blue wall". He actually pegged the popular vote pretty accurately: Clinton +2%. No one could have anticipated how badly Clinton mismanaged her campaign. Which is why, despite the lead, he gave Trump a much better chance than anyone else.
@Greg:
I don't get the fascination with LePen among conservatives. She's a socialist in every policy that matters.
I think that's a ridiculous characterization. Le Pen believes in private enterprise with taxation to pay for government social services. In other words, what every developed nation does, including the US, to varying degrees. If you want to find a country trying to implement socialism, look at Venezuela, not France.
@MPH:
Sickening to think some people believe that. As if the destruction of NATO and the breakup of the EU will not affect the US.
Sure it will affect the US...positively. NATO has outlived its purpose, and the EU is a German-dominated organization meant to check US and Russian power.
MikeR is the only commenter today with an ounce of sense.
Otherwise, the comments today are overrun with Putinist trolls.
This was the final prediction, published on America's Election Day, at Nate Silver's FiveThirtyEight
No, that was not the final prediction. Because that is not a prediction. It is a statement of probability.
If you ask me the probability of getting a 6 on a single roll of a standard die, I'll tell you the probability is one-in-six. If you roll the die and it comes up 6, my statement of the probability is still correct.
The fact that a statement of probability does not tell you what you actually want to know does not make the statement of probability incorrect.
Otherwise, the comments today are overrun with Putinist trolls.
Ha. Pretty astounding that in 2017 people are still compelled by this kind of Russophobia. Because of course when you look around at the problems the US is facing today or the problems faced by the wider western world, the obvious culprit is Putin.
When it's Le Pen vs. The Press, she loses.
But when it's Le Pen vs. The System, she wins.
Macron's slogan is "neither left nor right" which means, in practice, "left." Economically, the French electorate is of the left, overwhelmingly, meaning they value the welfare state above all else. Le Pen is certainly in accord with this. The choices with regard to economics were left, lefter, and communist. The one thing that sets Le Pen apart from the pack is her recognition that third-world immigration is a death knell for France and the rest of Europe. This is important to a portion of the electorate, but not a majority. So France will pretty much continue on the same path it has trod the last few decades, with their fate ultimately in the hands of the immigrants who hope to do to the nation what they have already done to large swaths of cities. They hope to make the entire country a no-go zone for infidels.
Mike: He actually pegged the popular vote pretty accurately.
This may be in part to the fact that the Clinton campaign put a lot of resources in GOTV efforts in places like New Orleans, where she was definitely going to lose, and Chicago, where she was definitely going to win. One of the fears was Trump would win the popular vote but Clinton would win the Electoral College and they wanted to avoid the headaches that entailed. Of course, this led them to ignore states that they thought were in the bag.
Nationwide popular vote polling is of dubious usefulness when the actual goal is the Electoral College and the election is close. This goes triple if there is any inkling of a realignment. This goes quadruple when one candidate does something eccentric.
Problem is, Trump was never in a hole.
Getting between your opponent and the center of the electorate is generally a surefire way to win an election, but . . .
Does Macron have any previously unknown scandals? Besides boinking his high school teacher, I mean? (They did get married.)
Will ISIS stage another big attack between now and June 20? Or maybe a couple of the Muslim "no go" zones will go on a rampaging riot, as they seem to love to do? In other words will something happen that makes the French say "enough is enough" (trop c'est trop)?
Will women decide that it's a woman's turn, and have France join Britain and Germany in having a female PM?
28.6% meant there was a real chance. Play enough backgammon or any other game that is all about probabilities, and it becomes very apparent that 1 in 4 or 1 in 3 will happen enough to frustrate (or thrill) the hell out of anybody. For all the criticism thrown at Nate Silver, why were his models wrong, if they were wrong? Every bookmaker out there had similar odds, also, so what to make of them? Were they just following his assumptions when making their betting lines?
I don't have a problem with Silver being challenged, but it would be good to see something beyond he only gave Trump a 28.6% chance in his final model, so how good was the model? I will say it was clear that he blew it when predicting Trump had no chance in the primary. If anybody is going to question Silver, that seems like a good place to start. It was clear that Trump offered positions that no other candidate was willing to touch, yet a sizable minority of the party agreed with, such as birtherism. In a crowded field, that is a winning formula. I knew at the time his chances would depend largely upon how quickly other candidates who had little chance dropped out of the race. Scott Walker alluded to this when he urged others to seriously consider cutting bait, as he knew that the longer others stayed in the race, the more it helped Trump.
Percentage chance of winning is not the same thing as percentage lead in the polls. Lots of things that have a 30% chance of happening do in fact happen - three out of ten, normally.
I think part of the reason Nate Silver was so pilloried was the paradox of rising expectations. Given Silver's successes in 2008 and 2012, the resulting good press and high-dollar offers, and the general innumeracy of much of today's press and pundit class, it is no wonder they essentially believe that anytime he does not get something 100% right, he's a failure.
MPH,
Someday, soon, being called a "Putinist troll" will be seen as a badge of honor. I'm overjoyed to have gotten in on the ground floor!
[And really (going into Sarcastro-block-quotes mode]: the sooner we recognize Russia as a competitor rather than an enemy the better off we'll be.]
I also think that where Silver and the polls tend to break down is in measuring intensity. There were people -- and I'm proud to say I was one of them -- who would have crawled five miles on two broken legs through broken glass to vote against Hillary, and people who would have done the same to vote for Trump vice against Hillary. And these groups badly outnumbered the people who have gone way out of their way to vote for Hillary. Pollsters say no, their models capture intensity. But look at Silver's map. Trump captured all of the pale red states and over half of the light blue ones. If intensity of support doesn't explain that, then what does?
Reminds me of the theme song from The Wire:
"Got to keep the devil way down in the hole!"
He actually pegged the popular vote pretty accurately.
That and $2.50 will get you a cup of coffee at Starbucks.
Silver's genius was to call races for the Democrats when the Democrats won. Calling one for them when the Democrats lost it was definitely a brand error. It's like they say about day traders, a rising market makes everybody look like a genius.
@Kirk Parker:
[And really (going into Sarcastro-block-quotes mode]: the sooner we recognize Russia as a competitor rather than an enemy the better off we'll be.]
I think this formulation is exactly correct. Russia poses no significant threat to our prosperity as a nation, but because the US defines its interests in such an absurdly broad way (essentially desiring a global hegemonic position), we are constantly bogged down in trying to control events that we have very little control of. But in the interventionist mind, when an intervention fails, it is because of a lack of will or a lack "getting the job done," not because of any inherent limits on what American power (especially American military power) can achieve.
Big Mike, there were also a lot of people who would crawl on broken glass to vote against Trump. One thing I often hear is "shy Trump voter" when someone tries to explain the result. I think Scott Adams has alluded to this phenomenon a number of times. Silver did try to incorporate some leans towards Trump in his model, and he gave a higher likelihood for Trump than other models out there. I would have looked at how the races down ballot were shaping up, also. When I saw how close Ron Johnson was in his race in comparison to Trump in Wisconsin, it certainly made me wonder if Trump was being undercounted in polls there. Perhaps a shy Trump voter in Wisconsin would be more likely to voice public support for Johnson than for Trump?
"One thing I often hear is "shy Trump voter" when someone tries to explain the result."
-- Well, when you see people at rallies being physically attacked and someone tries to pull a gun on Trump, you get that effect. I'm not even a Trump fan, and I can understand how some who are might decide, "Screw that pollster, of course I'm going to tell him I'm undecided or Voting For Clinton. Did you see what they did to those people in California?"
God Bless the Secret Ballot.
I'm not disagreeing with you, Matthew, but with modeling, one has to look more at numbers than at anecdotes. Trump's numbers versus Johnson's is one possible way to see this in action. Or maybe people were motivated by Johnson and ended up voting up ballot for Trump. A lot was spent on the GOP held seats in the Senate by people who were not going to spend it on Trump, so I suppose that's another possible explanation. Did Trump provide coattails for the other battleground states' Senate races (PA, OH, NC, besides WI), or did the Senate races pull up Trump to victory in these battleground states?
I think Macron will win in two weeks, but the election is going to be closer than most pundits think right now.
I have been laughing my ass off on all the stories about how so-called mainstream parties were shut out of the 2nd round in France. This is at best a mistaken interpretation, or at worst an outright lie. The center-right mainstream in France was shut out of the 2nd round, but the Left played a different game that had a two-pronged strategy. First, they found an attractive candidate with a skimpy record and had him run as an "independent" or "third way" candidate. Second, the present government arranged to knee-cap the center-right's candidate, Fillon, so that he would fail to make the second round at all.
It was a truly brilliant strategy for hanging onto power, but it is going to be a pyrrhic victory in the end since Macron is all but certain to end up in the same state Hollande is in today, and Le Pen's party will grow to be the center-right opposition in the next election, which she is likely to win in a landslide.
And, I will say this again about the pre-election national polls:
Yes, they got the national vote about right when combined together, but that isn't good enough. Anyone with access to the national raw data should have been able to look at the state subset data and pointed out that Clinton ran a large risk of losing the electoral college- like say a 50% chance of doing so. Not one analyst did that. This tells me that one of two things has to be true- none of the national polls, nor their combinations, had data that actually predicted Trump would win enough states to win the electoral college, or the data was right there in front of them and they refused to report that possibility. Being surprised on election night, you have to blame either the polling data, or you have to blame the people doing the data analysis. Pick one.
71% to 29% hur hur Nate Silver is pretty dumb amirite?
A 29% chance of being President is a pretty high probability. The chance of dying in one round of Russian roulette is only 18%, and most people find that probability large enough that they care not to play.
I have no idea how a numerate person could look at Silver's prediction, see that Trump won the election, and then say Silver was doing it wrong. But there are not a lot of numerate people.
I have no idea how a numerate person could look at Silver's prediction, see that Trump won the election, and then say Silver was doing it wrong.
Well, he did seem to think that each state was independent of the others, like so many coin tosses. If a segment of the population moved toward Trump, and, as it happened ,that segment of the population was significantly represented across many states, even states Hillary won, states like for instance Pennsylvania, (Upstate) New York, to Wisconsin, then sure, a numerate person could say Silver was doing it wrong. I am such a person. I pointed this out *before* the election in these very comment sections.
It's kind of like the mortgage crisis, where they thought that regions were independent of each other and couldn't all go south at once, then they applied a national, uniform forcing on all of the real-estate markets by pumping money into them without worrying about the ability of the borrowers to repay, or whether they were overpaying, or even if they had a freaking job.
Another case would be the hockey stick, that Michael Mann built on the assumption that the climate each year was independent of the years on either side. This assumption was incorrect, and without this assumption, just about any autocorrelated data you put into his algorithm produced hockey sticks.
It is easy to fuck this stuff up, especially when you have something riding on the outcome, where that is the time when you should be most careful.
@tin in vermont: If a segment of the population moved toward Trump, and, as it happened ,that segment of the population was significantly represented across many states
I find your collectivism a little disturbing, honestly. Humans are not cattle. But let's go with it and assume that they are, then how did you implement your model mathematically and what odds did you model?
I am such a person. I pointed this out *before* the election in these very comment sections.
Can you send me a link to your contemporaneous calculations where you got a different result?
My magic 8 ball correctly called Trump and Brexit, but I am very confident that was just a coincidence, since the magic 8 ball is not numerate and didn't do any actual work to arrive at that conclusion.
tim in vermont said...
Well, he did seem to think that each state was independent of the others, like so many coin tosses.
Why did it seem to you that he thought that? If you look at his methodology, he explicitly states that the states are not independent:
The error from state to state is correlated. If Trump significantly beats his polls in Ohio, he’ll probably do so in Pennsylvania also.
I'm not claiming that he accounted for this correlation correctly. I'm just pointing out that the way it seemed to you was not the way it was.
"I'm just pointing out that the way it seemed to you was not the way it was."
That's right--it accounts for why he had Trump at 30% or so vs others who had him at less than 5%. But he still erred in his assessment of the precision of the state polling in the battleground states. Any understanding or accounting of that should have pushed him toward 50-50.
Maybe he erred, maybe he didn't. Who knows that answer for sure? Saying 30% was too low is too easy. Is there a mathematical model that can explain why it's too low?
Silver says, See I was right: I said Trump had a 30% chance of winning, and he won! I said Hillary was only a little bit ahead in Wisconsin and Michigan and Pennsylvania, and I was right: Trump pulled ahead and won them!
But what if Hillary wasn't a little bit ahead in those (and other) states. What if the polls were wrong and Trump was ahead, but we didn't know it?
Maybe Trump wasn't "in a hole" after all? What if the polls systematically over-weighted Hillary's support and under-weighted Trump's? Not necessarily intentionally, but because the pollsters didn't understand what was going on?
In fact, I submit that that's probably what happened. And when the pollsters are now telling us how negatively the people are rating Trump's performance as president, are they making the same mistakes? If they are mistakes. Or are they intentional?
There's no way to check the accuracy of an opinion poll unless it's about a forthcoming election. What -- other than blind faith -- would make you believe ANYTHING that pollsters tell you about support for, or opposition to, Trump's program today?
" Is there a mathematical model that can explain why it's too low? "
Yes, as I said, there were good reasons to think that the state polls were noisier than believed. Just allow for a larger margin of error in the state polling and you get probabilities shifting toward 50-50. As for evidence, all I can tell you as that I put sizable money down on Trump because I didn't believe the probabilities I was seeing in the betting markets.
Money won is twice as sweet as money earned, as Paul Newman's character said in Color of Money. Margin of error works both ways. It's just fattening the tails, ie expanding the range of outcomes. You seem to be talking about skewing the curve more towards Trump, which is a bit different. How to get there, that's the question.
Post a Comment