Nate Silver’s Election Odds Model is totally screwed up.
Despite obvious momentum in favor of Trump, Silver projects Trump has a better chance of winning today than a month from now. This is of course ridiculous.
Let’s investigate three charts as of 2016-09-16 11:40 PM Central.
- Silver’s Nowcast: Who would win the election if it were held today
- Silver’s Polls-Only Forecast: What polls alone tell us about Nov. 8
- Silver’s Polls-Plus: What polls, the economy and historical data tell us about Nov. 8
Nate Silver Nowcast Results
Nate Silver Polls Only Results
Nate Silver Polls Plus Results
Recap of Trump Odds
- Nowcast: 42.4%
- Polls Plus: 40.7%
- Polls Only: 39.8%
Mathematical Idiocy
Despite the fact that momentum has clearly shifted in favor of Trump, Silver has Trump’s odds of winning today higher now than in the future.
This negative time preference is ridiculous.
Earlier this week, but I failed to capture it, Silver had the the Nowcast and Polls only forecasts 0.1 percentage points apart.
Are the odds that Hillary keels over, is indicted, or hospitalized by November 8 greater than 0.1%?
Any actuaries or doctors care to chime in?
Meaning of Time
Not only does Silver fail to understand time, he fails to understand momentum, and human psychology as well.
His previous 89.2% Hillary forecast on August 14 looks totally laughable. But it was nearly as laughable then, as now. And I said so at the time.
Given the human emotions in play, Hillary’s obvious (even back then) medical difficulties, the extreme dislike of both candidates etc, Silver never should have had an 89% confidence level on Hillary that far in advance.
Time and time again, Silver has proven he does not understand time or human psychology. He should have held off giving election odds until the final month.
Mike “Mish” Shedlock
Why do you bother giving this guy so much attention if you think he’s so bad?
Analysis?
I see a definite trend: Nate Silver turns out wrong 100% of the time this election.
No kidding.
This post should have been one sentence — the opening one,
“Nate Silver’s Election Odds Model is totally screwed up.”
— and then ‘fin’, no more on his guy.
Have to agree. Silver appears to be able to prognosticate well if the future in not in doubt. He, and others in his business, don’t do well when an excess of ‘uncertainty’ is floating around in the ether. I don’t know why the uncertainty epidemic has pummeled us so badly this year. It just plopped down from the sky like a giant turd and backed up the full works, flooding everything. Lots of giant mud pies everywhere and they don’t appear to be going anywhere. His skills at edumacated guesswork have been affected.They will probably remain as such until we get back to normal and everything gets fixed again.
“Why do you bother giving this guy so much attention if you think he’s so bad?”
For the same reason he exposes how totally clueless central bankers are.
Every morning when I turn on my satellite for CNN and BBC they are at it again. I mean mocking Trump. It seems there is no thing Donald can or could do right. For the amount of bad will alone, I would vote Donald if there would be a chance.
There is no such thing as bias free media.
Here is the psychology angle Mish mentions, but for some reason doesn’t discuss. Every media entity is attacking Trump 24×7, ridiculing anyone who would support him. If you are a Trump supporter, you are not only a racist, you also torture and eat cute puppies.
So how open will people be when pollsters call? I’d say most are keeping their Trump support quiet, so as not to sound like a social outcast. Much like Brexit, there will be a big surprise once the votes are counted.
99.99% shit-show, 0.01%news. I don’t believe any of it. Meanwhile, what is really happening out there? If you don’t dig deep for it, you won’t find anything useful.
I think the movements in the polls is screwing with the models. Seriously, if Trump wins Ohio, Florida, and Iowa, he will win Virginia. And, in my opinion, if he wins Iowa, he will also win Colorado. It would require a major realignment of voting patterns for Clinton to win Virginia but lose those other states. Sure, the state has drifted left for over 15 years, but it hasn’t drifted that far.
I would revisit this in a week or two after you get new data for all the states that are in any doubt.
And, let me say this- Kaine makes no difference in who wins VA, though I am absolutely certain Clinton chose him precisely for that reason. Clinton will win or lose VA on her own merits.
Rubio lost to Trump in the primary despite his popularity in his home state. My bet is Trump will take VA. Kaine is a non factor.
I believe, and I’m not sure, that Nate’s election odds always referred to “if the race were held today”. If that isn’t what he meant then he is a moron.
Nate is no longer the Leftist boy wonder for 2 reasons: 1. He epically blew the primaries, exposing himself for the globalist shill hackjob that he is 2. His model is much kinder to Trump than most of the others. Leftists have switched over to Sam Wang as their new guru.
The thing is, Hilary just had a brutal 2 weeks and is still up by nearly 3 points in the polling averages. Trump’s numbers with college educated whites are horrendous, he needs to win 55% of college educated whites and right now he is looking to get more like 47%.
One reason I maintain some hope for Trump is that Obama was only up by 1 against Romney in 2012 in the polling averages and won by 4. If Trump could copy that 3 point swing that the polls aren’t capturing, he could just about it off.
A lot is hanging on the first debate. If Trump could score a clear win like Bush did over Gore in 2000, he has a chance.
Trump should be leading in the up-to-date polling averages. The RCP polling average still includes data from pre-9/11. As of today, the two polls, Fox and USC tracking, that were taken (mostly) post 9/11 have Trump leading. The poll averages will start favoring Trump sometime next week. This is likely the final inflection point in this campaign.
As long as there is no paper trail and no possibility of a recount, then the probability of fraud needs to be factored into USA electoral predictions.
Mrs. Clinton’s supporters I talk to says she has an overwhelming “electoral advantage”. What are they talking about?
Lots of careers depend on her winning so the staus quo remains intact. They constitute an army of people who would otherwise have to start doing actual work if she loses. They are looking after their own self interest. Some of them might actually have to get real jobs, as opposed to flunky work.Thus, her ‘electoral advantage’. A lot of people are frightened of a Trump presidency for this reason.
Nate Silver is 1000% correct. Mish is totally wrong.
Mish did not factor in poll rigging as suggested by Trump.
Agree. Most people are honest as long as it serves their purpose. They lie only enough to bring what they do to the ‘full’ line. Pollsters like to see the glass completely full. That’s how they get paid.
Remember the quote, “When it gets serious, you have to lie” by Jean-Claude Juncker?
You are spot on.
The only thing that with a 89.2% probability is that Nate (((Silver))) is wrong.
These types love making money by providing “services” that add NO ECONOMIC VALUE WHATSOEVER and add nothing to society as a whole.
They love to yap, yap, yap…..go pick up a hammer an nails and make something useful.
“This negative time preference is ridiculous.”
You might say the same of interest rates…
I don’t think Silver’s methodology is the problem. I think he is rooting against Trump (I am too) and his results intentionally reflect this bias in order to assist Hillary.
For the last several election cycles Nate Silver’s polls have been within 2 points or less of the actual results.
I wonder if Nate Silver has learned anything from his primary forecast. LOL.
http://fivethirtyeight.com/features/how-i-acted-like-a-pundit-and-screwed-up-on-donald-trump/
mish,
time for that article ‘investment thenes of a Trump prez’
Trump already gave us a hint, last week he said HR Block would have a very hard time
Are you kidding? Investing after Trump wins will go back to being what the textbooks describe. Fundamental analysis will have meaning again. Technical analysis will work over the short term. It’s pretty bogus over the long term, but it takes all kinds to make an investor community and if they want to believe in magic and pixies, OK by me.
Crooks will go back to being only a part of the group, as opposed to the majority group today except for those who somehow believe markets are honest today and don’t believe in crooks. Or law enforcement who chooses to ignore most crooks. Your interest expense will be my interest income and I will spend it. Debt issuance will have to take returns that cover real interest expense into consideration. As a result, project will create jobs as opposed to today, which concerns itself with the finance of paper flipping and executive comp schemes.
Do you really think an election can change the way EVERYBODY in power does thing?
Don’t you think undoing the damage, or doing a different kind of damage might take 12-20 years?
If this thing ever reverts back to fundamentals, a smoking heap of rubble is all that will remain.
Too far gone now. It has to run it’s course into creative self destruction. Then, who knows what rises from the ashes.
That said, it is unwise to fall in love with politicians,,,,,they’ll break your heart every time.
Nate Silver had better be careful, or he will become a laughingstock like Dennis Gartman.
The media cites that goofball all the time.
You’ve got to get rid of the anachronism called the electoral college. Then you can have an chance for an accurate forecast.
“You’ve got to get rid of the anachronism called the electoral college.”
Elections are an anachronism. Obama prefers being a dictator.
You would be wrong; they kept the elections, but rigged the system. And they did not tell anybody.
“Not only does Silver fail to understand time, he fails to understand momentum, and human psychology as well.”
Nate Silver’s propaganda is running out of time
Poor Nate. He must be the laughing stock of the political forecasting business. I haven’t seen him lately on any of the rubber-chicken media circuit shows delivering his gibberish.
I wonder if he’s considering a career change? How about being an economics forecaster, in particular affiliated with some Left Coast university. Boy, that would be a hand in glove fit.
His GOP primary calls should go down in the records books under “off by a country mile”.
Or the chance that she will make another comment that will result in something even better than Les Deplorables.
Maybe she coughs because of a frog in her throat – Pepe. Both Pepe and Hillary seem about to croak.
Also note how you’ve not seen any Elephants or Donkeys, only Pepe the Frog.
Someone up above alluded to the problem Silver and others have with building models and then running with the results as if they can’t be wrong (and this happens *a ton* in sports): models only work when (A) the past is indicative of the future, and (B) all relevant information is accurately captured and considered. When either or both of those don’t happen, then models look great for predicting the past but have issues projecting into the future. (See “our models don’t account for home prices decreasing” pre-2008.) The real “problem” is that we only have some 50 observations (presidential elections) in U.S. history, and perhaps 15 of them are suitable for using to predict outcomes of presidential elections. *Maybe* you can use gubernatorial elections, Congressional elections, and state elections to supplement – and maybe those elections behave similarly to Presidential elections – but to the extent they don’t, you have issues in building an accurate model.
One other item to note: many people (grossly) misunderstand statistics. If I say an outcome to a certain event will happen 95% of the time and for event it doesn’t, that doesn’t mean the model is wrong. It easily could have been one of the 5% of scenarios where the event wasn’t expected to happen (for whatever reason). If 150 times out of 1000 that event doesn’t happen, though, … well, that’s a problem even if 85% of the time the model was right.
Maybe Silver was lucky previously. Or, maybe he’s right and this cycle is the exception because of factors he can’t (or doesn’t) account for. We have no idea, and unless he gets to predicting thousands of elections every cycle there’s no way to know. Suffice it to say, though, that Silver’s models are not working well right now because this election isn’t behaving like any in history – and there’s no “good” way to make an accurate adjustment beyond spitballing assumptions and hoping they’re close.