The critique of Nate Silver’s pure reason
What interests me is the way people talk about math as if it were divinely prophetic. They seem to subscribe to a religion that simply apes the terminology of science. To listen to many of Silver’s defenders, questioning his methodology is akin to rejecting evolution or the laws of thermodynamics, as if only his model is sanctified by the god Reason.
I wonder: What kind of scholarship do we have to look forward to when, in the words of Krugman, “facts really do have a well-known liberal bias” and a difference of opinion over poll-weighting foretells the end of science?
Don’t get me wrong; I do understand that math can be ironclad. We know the decay rates of isotopes, how fast things will fall in a vacuum, what compounded interest rates will yield, and all that.
But I like to think that people are different, more open to reason, and that the soul — particularly when multiplied into the complexity of a society — is not so easily number-crunched.
Blowback
Note from Hot Air management: This section is for comments from Hot Air's community of registered readers. Please don't assume that Hot Air management agrees with or otherwise endorses any particular comment just because we let it stand. A reminder: Anyone who fails to comply with our terms of use may lose their posting privilege.
Trackbacks/Pings
Trackback URL
Comments
What are the chances Jonah would have written this if Silver had correctly predicted a Romney win?
Pablo Honey on November 7, 2012 at 11:25 AM
For the love of….
GO AND LOOK AT NATE SILVER’S MAP. It looks exactly like what happened last night. The only swing state he gave to Romney was North Carolina.
Having read Hot Air very often these past months, I was kind of perplexed by all the unskewing of the polls. “These people are in for a tough election night.” I thought. But, now that it’s happened accept the facts:
Math matters. Ignore the polls at your peril.
agirlacamera on November 7, 2012 at 11:25 AM
Jonah has been saying this for a while now actually. Read Tyranny of Cliches – it’s in there too.
gwelf on November 7, 2012 at 11:29 AM
What a pantload.
tommyhawk on November 7, 2012 at 11:30 AM
Ron Paul was right – the status quo won, and the status quo would have won if Romney had been elected as well. I’m sorry you had to realize that.
MoreLiberty on November 7, 2012 at 11:30 AM
Indeed. You may recall all the talk a while back about ‘epistemic closure’ on the right–the conservative penchant for self-selecting facts and rejecting facts that don’t conform to ideology. This past few weeks Hot Air became an echo chamber of just this sort. All the polls were wrong. Silver was wrong. Romney in a landslide.
A selection:
Reality matters.
lostmotherland on November 7, 2012 at 11:33 AM
Maybe Jonah will have a more thought provoking piece after a good night’s sleep.
faraway on November 7, 2012 at 11:34 AM
Nate silver is a fucking dope. The RCP poll of polls correctly predicts the actual out come as well. All nate did was feed that into his super model and get a result that was slightly more wrongt. Awesome work!
sheikh of thornton on November 7, 2012 at 11:34 AM
Nate Silver is a dope. The RCP poll of polls also predicted the correct result. All Nate did was feed that into his model and pump out a slightly less good prediction.
Well done Nate!
sheikh of thornton on November 7, 2012 at 11:35 AM
while i don’t remember posting anything similar, i’ll admit to thinking the same thing. i lost quite a few bets last night. i was very, very wrong.
and to think, i even supported T Paw way back in the primary season!
Steven McGregor on November 7, 2012 at 11:37 AM
Nate Silver nailed it.
Most people use reason to reinforce their gut feelings, rather than to objectively assess things.
When the early voting numbers came in for Colorado yesterday, most posters here desperately wanted to believe they reinforced the landslide narrative. When I questioned the math of how these numbers were so great, I was ridiculed as if I was an Obama supporter deluding myself.
Truth is, those of us who thought we knew better than Silver’s model were all deluding ourselves.
Chameleon on November 7, 2012 at 11:41 AM
LOL. I guess the Republicans are still inclined to deny hard truths rather than confront them.
Alpha_Male on November 7, 2012 at 11:42 AM
If I had correctly predicted the electoral map last night by casting oracle bones and divining from pig entrails, would you also sign onto the accuracy of those measures?
Good Solid B-Plus on November 7, 2012 at 11:43 AM
What, unlike the liberal ass holes of America who thinks things are fine and Obambam is doing a great job!
Yikes, nation of stupid.
sheikh of thornton on November 7, 2012 at 11:46 AM
Conservatives reject facts? Please explain to me how Democrats are going to prevent a fiscal collapse due to skyrocketing debt by taxing the rich bringing in 80 billion more per year?
gwelf on November 7, 2012 at 11:48 AM
Lest we all forget, the contempt for Nate Silver came from his declaring the race all but over two, three, four, five months ago.
As the first debate showed, that’s a lot of sabermetric show-boating nonsense. Things happen during campaigns. That’s why they have them.
All Nate did was feed the final polls that showed Obama winning into his machine and declared that Obama would win. This is more than many Republicans were willing to do, but that’s to be expected.
HitNRun on November 7, 2012 at 11:49 AM
Silver’s model is basically just the poll of polls. He accepted the large D+ samples; most of us didn’t.
It turns out the GOP base isn’t as big we thought it was.
BadgerHawk on November 7, 2012 at 11:49 AM
Nates model completely depended on the MSM assisting in the election. Something he knew he could count on. It’s an engineered outcome of carefully controlled information.
Keep the voters dumb and you can control the outcome of any election. Not exactly rocket science.
HotAirian on November 7, 2012 at 11:54 AM
Well that’s the problem right there, isn’t it? Why weren’t many Republicans willing to do this?
ernesto on November 7, 2012 at 11:54 AM
The thing is, you never had any reason to believe it was as big as you thought it was. You just simply made it up out of thin air. This makes Silver more deserving of trust and attention than pundits who ignore reality, like Ed Morrisey at HA.
ernesto on November 7, 2012 at 11:56 AM
Nate Silver’s model, whatever it is, works a hell of a lot better than the analysis seen around here.
Yeah he gets hype because he is a kos kid telling the left what they want to hear. They don’t have a clue what he is talking about as far as his model goes. Most liberals think he is talking about a chick.
Glenn Reynolds has said all election season Don’t get cocky, kid. I never did. If you just read the numbers around here, I’m sure you are shocked.
Moesart on November 7, 2012 at 11:57 AM
Easy. By hiding the increases in the form of price increases. QE4-QE20 here we come. Plus fuel taxes, health taxes, etc.
faraway on November 7, 2012 at 11:57 AM
Because they were like every losing party ever?
FL was a tie, OH was within 2, VA within 3. One more state wins it. It would be pretty stupid of them to accept defeat in advance.
Troll harder.
HitNRun on November 7, 2012 at 11:57 AM
I know a hardcore conservative who voted Obama. I don’t blame Romney, but we really could use better candidates.
Darth Executor on November 7, 2012 at 11:58 AM
Goldberg is perpetually confused. Most people don’t have souls, and are in fact quite easily number-crunched automatons.
Darth Executor on November 7, 2012 at 12:00 PM
There were plenty of reasonable facts you could draw on to discount Silver’s model and the poll sampling. The dismal state of the economy, an incumbent who consistently polled below 50%, 2008 was a very large record turnout and unlikely to be repeated, the 2010 elections (midterm elections are directly comparable to presidential ones but 2010 indicated something). Many polls showed Obama getting bigger D+ margins than he did in 2008 – and he didn’t.
gwelf on November 7, 2012 at 12:02 PM
Thank you LML
You guys shewed me over and over on this blog just for adding an independent view to the mix. You are stuck in the 20th century, and you would not hear any different no matter what…
From your campaign to your thinking…..#losing
Can.I.be.in.the.middle on November 7, 2012 at 12:04 PM
ALL the fundamentals – all of them – favored Romney, except for Demographics.
This election was a tipping point. We finally reached the point where the Dem coalition consists of enough people to win elections no matter how bad the rest of the fundmentals end up being.
BadgerHawk on November 7, 2012 at 12:08 PM
oh…and for those wondering what the heck is Nate doing…it’s all about data mining (collected through mobile and social networks), pattern behavior at a local level, the genius is being able to find the true “signals” vs “the noise”…the rise of Analytics!
Check out IBM’s article about it
https://www-304.ibm.com/connections/blogs/smarteranalytics/entry/distilling_the_signal_from_the_noise_with_nate_silver11?lang=en_us
Can.I.be.in.the.middle on November 7, 2012 at 12:10 PM
Exactly.
The final RCP poll of polls was Obama +2.2. As of right now, that looks to be almost exactly right. RCP’s individual battleground state polls were also correct. So remind me, what is it exactly that Nate Silver did that’s so amazing?
Yes, we were wrong about the turnout models. We thought Democrat turnout couldn’t possibly reach 2008 levels, but they did. But Nate Silver is nothing special; he gives us no insights that we couldn’t already glean from the polls themselves.
Good Solid B-Plus on November 7, 2012 at 12:15 PM
So after filtering out all the “noise,” Silver ended up being no more precise than the RCP Poll of Polls.
As someone who has been in the baseball analytics community since Clinton was the President, I know more about Nate Silver than just about anybody here. He’s a smart guy, but he’s always overrated his methods. Back when he was a forecaster in the sabermetrics realm, he was no more accurate than any of his other Baseball Prospectus colleagues (or readers like me, for that matter). He’s seen as an oracle by people who aren’t intelligent enough to deal with the raw material by themselves.
If Tom Tango starts doing political analysis, that’ll be a real get.
Good Solid B-Plus on November 7, 2012 at 12:20 PM
Nate Silver is great. Thanks to him, I made a bundle at intrade
Flapjackmaka on November 7, 2012 at 12:21 PM
We had a lot of reasons to believe that Obama’s turnout wouldn’t equal 2008, and we were right. However, Romney’s turnout also didn’t equal McCain’s, which wasn’t something any of us expected.
Good Solid B-Plus on November 7, 2012 at 12:27 PM
That’s the best summary I’ve seen so far.
Can RINOs turn out the base? Has the base shrunk?
faraway on November 7, 2012 at 12:36 PM
Nothing Silver did is impressive. Running a Monte Carlo simulation of the RCP polls gave virtually the exact same result (can be done with a simple free App by anyone).
When polls are wrong, Silver is wrong. When polls are right, Silver is right.
Norwegian on November 7, 2012 at 12:36 PM
Good commentary.
I predicted Walker’s recall win (I bet someone he’d win by 7-8 points) and was pretty accurate about the 2010 election results, but I certainly don’t consider myself to be a genius or expert at that sort of thing. I thought 0dumba wouldn’t do nearly as well in 2012 as he did in 2008, and he didn’t, but like you, I erroneously believed that Mitt would better McCain’s result.
What so many Silver fanbois do is ignore the times he’s been wildly wrong, which I am aware has happened more than once.
Anti-Control on November 7, 2012 at 12:42 PM
Ugh…So much bad information here.
First, Nate Silver is using state polls much like RealClearPolitics. He is running state polls through a monte carlo simulation to obtain probabilities, and giving different weights to those polls. He is not getting a “signal” through the “noise” of social media.
Nate Silver’s model is not so much unique as it is well presented.
However, Silver did beat RCP’s poll aggregation this time around. He called all the states with Florida at 50-50. RCP had Florida in Romney’s camp. Also, the final RCP poll of polls had Obama +.7. The final result looks like Obama +2.2. Silver projected Obama +2.5.
There were many conflicting data points in this election. The most prominent was the disconnect between the state polls and the national polls. The national polls had Romney ahead by about 1 point until the end. The state polls suggested that Obama would win by about 2 points. Then there were outliers like Gallup that were suggesting a Romney wave.
In the end, the state polls were the most accurate, which isn’t all that surprising given the volume and focus of all those polls.
Chameleon on November 7, 2012 at 12:47 PM
Actually, the final RCP of polls was actually .7. Nate is predicting an Obama popular vote win of 2-3 points.
And RCP had Romney up by 1.5 in Florida (he lost to Obama there).
Anyway, if you want to follow RCP instead of Silver, that’s perfectly fine. Follow some sort of rational model, is my point.
According to the survey yesterday 94% of Hot Air readers thought Romney would win. Which is you what you would expect from low, low, low information voters.
agirlacamera on November 7, 2012 at 12:48 PM
Much of the criticism of Silver’s model was legitimate.
There were good reasons to question the party sampling being used in the polls used in his models.
2008 was a historic turnout year and it’s not ‘living in a fantasy world’ to have thought that the models should have looked like a different year.
That said, it turns out the models were right – I’m still having trouble figuring that one out since Obama lost 5 of his 2008 voters for every 1 McCain voter that Romney lost. I guess the votes they lost were in states that weren’t being polled.
JadeNYU on November 7, 2012 at 12:51 PM
There were like 10 poll aggregators aside from Nate Silver and RCP. There was Sam Wang, that guy at Davidson, electoralvote.com etc, etc.
I can’t remember all of them. But ALL of them predicted an Obama victory.
agirlacamera on November 7, 2012 at 12:53 PM
What a revealing comment, which is what you’d expect from someone with low, low, low rational argumentation skills.
Anti-Control on November 7, 2012 at 12:53 PM
It may be harsh, but surely there’s no excuse for that kind of wrongness. I can imagine low-information voters who pay no attention to politics having no clue about the state of the race, sure. But HA readers are just as misinformed if a whopping 94% are expecting a Romney victory even as of virtually all swing state polls have Obama in the lead. So my point is that Math matters, and that you would be smart to take Nate Silver seriously from now on.
agirlacamera on November 7, 2012 at 1:04 PM
What do you expect? Partisans always engage in wishful thinking prior to elections. All high information posters know this.
Go back and read the Kerry-will-win-in-a-landslide threads from Halloween ’04 on DailyKos.
The fact of the matter is that people use reason primarily to reinforce their own opinions or wishes. We all do, and that includes you. Ironically, you are using your reason to puff yourself up as a rational thinker (what you want to believe), while making a dubious assertion that wishful-thinking among partisans is related to knowledge of election issues.
Chameleon on November 7, 2012 at 1:22 PM
It’s only harsh because you are harsh.
Plenty of high-information voters looked at the same data as Nate Silver did, and came to a different conclusion about it’s meaning. So what? The way you tied their error to a needless and stupid insult reveals more about you and your critical thinking skills than it says anything about them.
Anti-Control on November 7, 2012 at 1:28 PM
No, I think that is a false equivalency. “Unskewing”, propped up by conservative media as a whole, is pretty unique to this election. Nor was Bush as consistently ahead in state polls in 2004. There was some hope about ground game or getting undecideds, and some cherrypicking of polls etc. But nothing on the scale I have seen at Hot Air and in the conservative echo chamber.
agirlacamera on November 7, 2012 at 1:39 PM
Again, there were about 10 poll aggregators. They ALL predicted an Obama victory.
Anyway, I won’t keep insisting since I think clearly the lesson HAS been learned and Hot Air will definitely start paying attention to polls.
agirlacamera on November 7, 2012 at 1:43 PM
What some people are forgetting is that the backlash against polls formulated in September, when Obama was nationally plus 5-7, the swing states were in double digits, and MSM pundits were declaring the election authoritatively over.
This was nonsense and we were right to call it such. Things happen in campaigns, like debates. Romney had a 1-5 point lead in RCP before the hurricane.
The takeaway should be the polls are not biased (though individual polls often are) but they can change and quickly, so getting the vapors over their state is pretty stupid.
HitNRun on November 7, 2012 at 1:54 PM
nate silver is an obnoxious nerd, but I trust nerds to do the number crunching and and give us the cold mathematical representation of reality from which we can more easily take conclusions and decisions!
soul can give motivation, but we should know when despite out best efforts, that motivation was just not enough to change reality.
nathor on November 7, 2012 at 2:21 PM
Honestly, could you miss the mark any more than you have right here?
First of all, HA readers are pretty much primae facie *not* low-info voters. Anyone who regularly reads a political blogis more informed than about 90% of the electorate. And yes, that also applies to people at HuffPo (the political part of the site, at least, not the part that only does Katy Perry/Selena Gomez news), DU, and Kos.
Secondly, we weren’t paying attention to polls? Ed Morissey does more analysis of poll cross-tabs than just about any other pundit on the internet. That’s what all of the snark form liberals is missing. You act like Republicans wanted to predict the election by using Tarot Cards or a Ouija Board. Of course math matters. Of course polls matter. We didn’t buy the polls, because we were certain that turnout wouldn’t match 2008, so D+6 or D+7 was unrealistic for sampling. We ended up being both correct and incorrect; turnout wasn’t nearly as high as 2008 for Obama, yet turnout was also depressed for Romney, so D+6 ended up being accurate.
It’s the same mistake Silver made at BP, that complexity by itself is an indicator of higher quality. PECOTA ended up being outdone by simpler projection systems like CHONE and Marcel, which used much simpler modes of statistical analysis. Marcel in particular was the simplest type of regression analysis possible, yet it often provided a better predictor than the over-complicated systems like PECOTA.
The issue I have is when people treat Reason/Math like a deity and Silver as a unique prophet, called upon to interpret divine information for the masses.
Good Solid B-Plus on November 7, 2012 at 2:32 PM
You’re cherry-picking right now.
Bush was ahead 2.1 points in RCP national polls on election day, Obama was only ahead by .7. Moreover, Romney led in national polls in the weeks leading up to the election, while Kerry never led. Gallup had Romney up 5-7 points in the weeks leading to the election, based in part on a huge 9000-voter survey checking party identification.
You say, “pay attention to the polls”, but then ignore Romney’s strength relative to Kerry’s in the major national polls as soon as those polls are inconvenient to your argument.
In other words, you do exactly what you gripe about others here doing in order to support your dubious opinion that Obama’s victory was easier to see coming than Bush’s.
But nothing on the scale I have seen at Hot Air and in the conservative echo chamber.
What scale? Is your analysis here based on math, or just your gut feeling?
Let’s take a look at DailyKOS diaries from 11-1-2004.
Hey look! A prediction poll! What percentage of DailyKOS readers thought Kerry would win the day before the election?
97.4% think Kerry wins with over 270 electoral votes.
65.4% think Kerry wins with over 300 electoral votes.
http://www.dailykos.com/story/2004/11/01/67841/-Electoral-Vote-projection-thread-poll
Wow! Seems like they were even more confident than HotAir readers!
Now that the data has contradicted your false assumptions, will you acknowledge your mistake, or will you dig in your heels and start cherry-picking data to try and resuscitate your false narrative?
Here are some other funny Kos diaries from 11-1-04:
Election Prediction: Kerry with 51% of Vote; 316 EVs
Stephen Yellin predicts Dems retaking WH, Senate and House
GREAT News: I called 50 voters in Ohio, 0 for Bush!!!!!!
Kerry’s horoscope for tomorrow is great!
What the Heck EV Prediction: Kerry 360
Republicans Know. (Lots of speculation that they know they’ve lost)
1960 / 2004 parallels
The Stars say Kerry Wins
John Stewart Wins it for Kerry!
Larry Kudlow Throwing in the Towel on Bush
Scientific prediction (predicts K – 323 EV)
The case for the Kerry landslide
Gallup numbers are even better than they look!
FanFREAKINGtastic poll info for our side!!!
I could go on and on….
There are many dozens of these if you’d care to look…
Chameleon on November 7, 2012 at 3:22 PM