MOBILE USERS: m.isthmus.com
Connect with Isthmus:         Newsletters 
Sunday, October 26, 2014 |  Madison, WI: 41.0° F  Fair
Collapse Photo Bar

Nate Silver's Flawed Model

Races for the Senate, U.S. House, etc. and other issues of national importance.

Nate Silver's Flawed Model

Postby Bludgeon » Thu Oct 25, 2012 6:53 pm

To quote a wise lefty:
Huckleby wrote:I love Nate Silver's blog, his analysis is always plausible and impressive, at least to me. I hunger for his each and every new release, it's a great pick-me-up in a sea of often anxiety-provoking polls, it serves as a "stiffening" shot of whiskey when courage is flagging.


To Kurt W: It's cool you like Nate Silver, I'm glad any time people get interested in electoral polls and prognosticating, but c'mon, fivethirtyeight.com is a wet dream on wheels. Especially this year. An orchard full of cherries to pick. Selective reasoning at its finest. There's no poll he won't take seriously if it casts Obama's chances in a sunny light; and no poll he won't shake a stick at (even PPP) if it makes his prospects appear less rosy. Not only that! His weighting system is as leading as may possibly be. He openly takes seriously the idea that not only will Democrats have a turnout advantage over Republicans, but that they'll have a turnout advantage over Obama's turnout in 2008! Really?

Josh Jordan wrote:I outlined yesterday why Ohio is closer than the polls seem to indicate by looking at the full results of the polls as opposed to only the topline head-to-head numbers. Romney is up by well over eight points among independents in an average of current Ohio polls, the overall sample of those same polls is more Democratic than the 2008 electorate was, and Obama’s two best recent polls are among the oldest.

But look at some of the weights applied to the individual polls in Silver’s model. The most current Public Policy Polling survey, released Saturday, has Obama up only one point, 49–48. That poll is given a weighting under Silver’s model of .95201. The PPP poll taken last weekend had Obama up five, 51–46. This poll is a week older but has a weighting of 1.15569.

The NBC/Marist Ohio poll conducted twelve days ago has a higher weighting attached to it (1.31395) than eight of the nine polls taken since. The poll from twelve days ago also, coincidentally enough, is Obama’s best recent poll in Ohio, because of a Democratic party-identification advantage of eleven points. By contrast, the Rasmussen poll from eight days later, which has a larger sample size, more recent field dates, but has an even party-identification split between Democrats and Republicans, has a weighting of .88826, lower than any other poll taken in the last nine days.

Furthermore, Silver explained on Saturday that a tie in the Gravis Marketing Ohio poll is actually a negative for Romney in his forecast because Gravis shows a Republican-leaning bias in polling. But the Gravis poll released Saturday has a nine point advantage in party identification for Democrats — almost double the Democrats’ advantage in the 2008 election. Then, regarding the PPP Ohio poll mentioned above (where Romney cut Obama’s five-point lead to one in a week), Silver notes that “Public Policy Polling has lost most of the strong Democratic lean that it had earlier in the cycle.”


Now, look at Obama's face in the debates, look at the 5,000 person crowds he's been drawing, look at small minded attacks about Big Bird, Binders, and "Romnesia." Then, let me know if you honestly believe Obama and his campaign share Nate's view of the political landscape. I would say Nate's out of his mind, but the truth is I don't believe he thinks it's an accurate representation either. My suspicion is that Silver knows he's churning out the universe's ultimate push-poll.

Peace.
Bludgeon
Forum God/Goddess
 
Posts: 1291
Joined: Tue Apr 06, 2010 2:27 am

Re: Nate Silver's Flawed Model

Postby rabble » Thu Oct 25, 2012 8:11 pm

Wow. The republicans think Nate's weighting his polls wrong.

Well there might be something to it. They certainly know a thing or two about that.
rabble
Forum God/Goddess
 
Posts: 6307
Joined: Tue Jun 10, 2008 7:50 pm

Re: Nate Silver's Flawed Model

Postby Stu Levitan » Thu Oct 25, 2012 8:23 pm

Bludgeon wrote:Now, look at Obama's face in the debates, look at the 5,000 person crowds he's been drawing, look at small minded attacks about Big Bird, Binders, and "Romnesia." .....


15,000 in Richmond VA on a Thursday afternoon. 12,000 in Cleveland on a Thursday evening. And I kinda think he did just a little bit better than 5K in Madison a few weeks ago, eh?
Stu Levitan
Forum God/Goddess
 
Posts: 3244
Joined: Sat Feb 08, 2003 8:40 pm
Location: Studio B of the historic Abernathy Building

Re: Nate Silver's Flawed Model

Postby Stebben84 » Thu Oct 25, 2012 8:32 pm

Is that confirmed :)
Stebben84
Forum God/Goddess
 
Posts: 4948
Joined: Thu Mar 02, 2006 12:59 pm

Re: Nate Silver's Flawed Model

Postby fennel » Thu Oct 25, 2012 8:50 pm

I think Nate is flattered that now even the wingnuts follow his blog.
fennel
Forum God/Goddess
 
Posts: 3194
Joined: Thu Sep 28, 2006 4:24 pm
Location: Inside the Green Zone, Madison

Re: Nate Silver's Flawed Model

Postby Detritus » Thu Oct 25, 2012 9:43 pm

Howzabout you explain exactly what it is about Silver's weighting system that is wrong? The thing you quoted doesn't do that--it just tut-tuts the weighting her provides. F'r instance, which element or elements of his weighting system are given incorrect values, and what should those values be? And if you think, as Jordan seems to suggest, that all polls should be weighted equally, what is your justification for that?
Detritus
Forum God/Goddess
 
Posts: 2435
Joined: Wed May 13, 2009 9:42 pm

Re: Nate Silver's Flawed Model

Postby Huckleby » Thu Oct 25, 2012 11:20 pm

Bludgeon wrote: c'mon, fivethirtyeight.com is a wet dream on wheels. Especially this year. An orchard full of cherries to pick. Selective reasoning at its finest. There's no poll he won't take seriously if it casts Obama's chances in a sunny light


I agree that Nate Silver is a left-leaning outlier in the world of prognostication. But I certainly do not agree that his methodology is dishonest. He backs-up his claims very convincingly. Silver has just as much of a chance to turn-out right in this game of tea-leaf reading as anybody else.

The point I was making in the other thread was not to get hopes up too high on any particular analysis. I suspect in a few days there may be stronger consensus one way or another. Then again - maybe it's a guessing game until the day after the election, or even weeks after the election, yuk.
Huckleby
Forum God/Goddess
 
Posts: 6743
Joined: Thu Oct 20, 2005 5:12 pm
Location: parents' basement

Re: Nate Silver's Flawed Model

Postby kurt_w » Fri Oct 26, 2012 4:42 am

Huckleby wrote:I agree that Nate Silver is a left-leaning outlier in the world of prognostication.


Nate Silver is solidly in the middle of the pack. See this post here, in which I report on 12 sites that do statistical analysis of polling data.

The median prediction is Obama 292.5, Romney 245.5. Nate Silver is tied with two other sites as closest to this median.
kurt_w
Forum God/Goddess
 
Posts: 5245
Joined: Tue Jul 22, 2003 3:11 pm

Re: Nate Silver's Flawed Model

Postby Bludgeon » Fri Oct 26, 2012 5:44 am

Huckleby wrote:
Bludgeon wrote: c'mon, fivethirtyeight.com is a wet dream on wheels. Especially this year. An orchard full of cherries to pick. Selective reasoning at its finest. There's no poll he won't take seriously if it casts Obama's chances in a sunny light


I agree that Nate Silver is a left-leaning outlier in the world of prognostication. But I certainly do not agree that his methodology is dishonest. He backs-up his claims very convincingly. Silver has just as much of a chance to turn-out right in this game of tea-leaf reading as anybody else.

The point I was making in the other thread was not to get hopes up too high on any particular analysis. I suspect in a few days there may be stronger consensus one way or another. Then again - maybe it's a guessing game until the day after the election, or even weeks after the election, yuk.


As far as that goes, nobody knows what's going to happen; but everybody knows it's not going to be a landslide. Romney's odds are good; Obama's odds are good. But nobody knows what the electorate is going to look like. What percentage will be: the white vote; the black vote; hispanics; the youth vote; the senior vote...

Last time Obama won 80% of the minority vote, and historically enthusiastic minorities made up 26% of the electorate. McCain won only 57% of the white vote, who made up 74% of the electors. Who here thinks minority turnout is going to be what it was in 2008? Average WaPo, Pew, Gallup, Politico, NBC, CBS, Monmouth, PPP, Rasmussen, IBD and Fox; Obama 2 weeks before the election is averaging 37.9% of the white vote. Almost half the president's margin in 2008 came from historically high levels of black turnout. I just don't see that happening in 2012.

This is not 2008. The minority vote is not going to be what it was in 2008, and the white vote is not going to be what it was in 2010. Romney is winning independents by 8%, the president is struggling in a bad economy. And there are people who voted for Obama in 2008 that are (or already have) voted against him in 2012.

Nobody thinks Obama is going to perform as well with any of these groups as he did in 2008. Take away just a few ticks from his margin with minorities, women, youths, give him just 38%-40% of the white vote and increase its share of the electorate by 1%-2%; a coin toss is what it comes down to.
Bludgeon
Forum God/Goddess
 
Posts: 1291
Joined: Tue Apr 06, 2010 2:27 am

Re: Nate Silver's Flawed Model

Postby kurt_w » Fri Oct 26, 2012 7:48 am

Bludgeon, we get it. You're a true believer.

The thing is, though, similarly convincing rationalizations can be found on left-leaning sites in which people explain why Obama's going to win by a wider than expected margin. For example:

* Whites' percentage of eligible voters has decreased by 2 points since 2008, while Latinos have increased by 2 points.

* Romney has almost no chance of winning without Ohio, and out of a dozen polls there in the past two weeks he hasn't had the lead in a single one.

* Claims that (a) "the polls don't count enough Republicans", and (b) "more independents are supporting Romney" are (inadvertently?) causing conservative analysts to double-count their own voters. What's happened is that in many polls, a certain fraction of Republican-leaning voters don't identify themselves as "Republicans". This makes "independents" look more Republican, but it also makes the reported fraction of Republicans lower. When conservative analysts then "un-skew" the polls by boosting the number of "Republicans", they don't simultaneously remove those "independent Republicans" from the count of independents. This leads to double-counting Republican poll respondents.

And so on and so forth. The bottom line is that you really need to avoid falling into the trap of accepting all the arguments that support your point of view while rejecting all the arguments that don't. Unfortunately, human beings are very prone to that kind of rationalization and self-delusion.

It's best to look at the arguments coming from a range of different sources, including those you vehemently disagree with. Which is exactly what I do here.
kurt_w
Forum God/Goddess
 
Posts: 5245
Joined: Tue Jul 22, 2003 3:11 pm

Re: Nate Silver's Flawed Model

Postby kurt_w » Fri Oct 26, 2012 8:10 am

As for the claim that Obama's support won't be as high as it was in 2008 ... well, so what?

In 2008, Obama won with a 365-vote electoral college landslide. There's plenty of room for him to do worse than in 2008, while still winning a convincing victory.

Nate Silver's projected 294 EC prediction implies that Obama will lose three-quarters of his 2008 margin above the 270-vote winning threshold. Intuitively, that seems reasonable to me (though it's the quantitative model, not my intuition, that's more important here).
kurt_w
Forum God/Goddess
 
Posts: 5245
Joined: Tue Jul 22, 2003 3:11 pm

Re: Nate Silver's Flawed Model

Postby Huckleby » Fri Oct 26, 2012 8:41 am

There is exactly one piece of information in the entire world of election data that I trust as reliable, and that is the trend shown by the RCP average of polls. I don't trust the levels - RCP average may be completely wrong about who is winning and by how much - but if one candidate is increasing, that is God's truth. I'm a first derivative true believer.

The reason I say this is that despite any bias in the polls due to over/under sampling or other faulty methodology, I trust the polls are done consistently day-to-day. So changes are meaningful, even if not the absolute numbers are not.
Huckleby
Forum God/Goddess
 
Posts: 6743
Joined: Thu Oct 20, 2005 5:12 pm
Location: parents' basement

Re: Nate Silver's Flawed Model

Postby Huckleby » Fri Oct 26, 2012 8:55 am

kurt_w wrote:
Huckleby wrote:I agree that Nate Silver is a left-leaning outlier in the world of prognostication.


Nate Silver is solidly in the middle of the pack. See this post here, in which I report on 12 sites that do statistical analysis of polling data.

The median prediction is Obama 292.5, Romney 245.5. Nate Silver is tied with two other sites as closest to this median.


I never take electoral college projections seriously. Tiny shifts in the overall popular votes blow those numbers out of the water.

Despite all of the focus on the electoral college, the national poll numbers are an excellent indicator of who is going to win the election, especially trends of poll numbers. (I haven't looked much at the poll of all swing states aggregated, but maybe that one is better than national poll.) The chance of the electoral vote going against the popular vote is quite small, estimates I've seen are around 5% this year, and in those cases the election is too tight to forcast anyway.

Put another way: ya, maybe Ohio is gonna decide the election, but the swings in Ohio tend to parallel national trends.

I notice you consider Real Clear Politics "activist/conservative". I think that designation reflects your bias, not theirs. They report data very honestly, aggregate a fair mix of lib/conserv articles, and I don't believe they conduct any polls of their own. Right-wing advertising at the site doesn't make them "activist."
Huckleby
Forum God/Goddess
 
Posts: 6743
Joined: Thu Oct 20, 2005 5:12 pm
Location: parents' basement

Re: Nate Silver's Flawed Model

Postby kurt_w » Fri Oct 26, 2012 9:55 am

Huckleby wrote:I notice you consider Real Clear Politics "activist/conservative". I think that designation reflects your bias, not theirs. They report data very honestly, aggregate a fair mix of lib/conserv articles, and I don't believe they conduct any polls of their own.


In the past, RealClearPolitics has been very erratic about which polls they include and which they exclude. During the 2008 election, they engaged in some pretty obvious manipulation, choosing to include a given pollster's work one day and exclude it another, with no explanation, and generally in ways that had the effect of benefiting Republicans.

I don't know whether they are still doing that. But it is one of the reasons I trust them less than other sites.
kurt_w
Forum God/Goddess
 
Posts: 5245
Joined: Tue Jul 22, 2003 3:11 pm

Re: Nate Silver's Flawed Model

Postby Mad Howler » Tue Oct 30, 2012 1:24 am

OK, let me understand this.
1) Silver closely calls the result in 2008.
2) Silver closely calls the result in 2010.
3) Silver is continuing to do what those gifted in math and statistics do. You know, kind of like Romney and Singer's math guys have done in a different context.

http://mediamatters.org/blog/2012/10/29 ... gut/191001

Has anyone considered that there have been nearly 21 centuries since Archimedes got run through by some ill informed soldier? Has anyone noticed that your real economy is called by math? I suppose if you do not see it already, then in ten to fifteen years you might as that is how it is being packaged. Regardless, say what you might about Silver, the way I see it math is math and this is brutal and no nonsense soldiering with a nod to Archimedes.
Mad Howler
Forum God/Goddess
 
Posts: 1393
Joined: Wed Apr 04, 2012 8:36 pm

Next

Return to National Politics & Government

Who is online

Users browsing this forum: No registered users and 1 guest

moviesmusiceats
Select a Movie
Select a Theater


commentsViewedForum
  ISTHMUS FLICKR
Created with flickr badge.

Promotions Contact us Privacy Policy Jobs Newsletters RSS
Collapse Photo Bar