Silver's flawed assessment of this race.

[Drunken Irishman]

No, we're not going to throw Silver under the bus. However, he is not above reproach. I think we have every right to question his model when it shows such a radical shift over the course of only a couple weeks. Not because it might not be accurate ... but because Silver has built up this idea that his formula was not only a genuine forecast of what could happen, but that it wasn't heavily influenced by polls, which, let's be honest, consistently show noise.

This past week, though, has proven that Silver's model is nothing more than a glorified RCP ... essentially solely tied to the radical shifts of polls.

But when he says there is no evidence of Romney's bounce fading, he's being absolutely disingenuous because the evidence we have is choppy at best.

What I mean is that there just isn't enough post-debate polling to suggest whether the bounce has faded, solidified or grown. That's not to say he's wrong... it's just that we just don't know.

We don't know because every poll we're basing this on, outside the Pew poll, has shown inconsistent results so far this year.

1. Rasmussen - Which even Silver admits has a Romney bias in its numbers. Rasmussen has consistently shown either Romney or Obama up narrowly ... with no rhyme or reason. Yesterday, Obama was up 1 ... today, Romney is up 1. But on the whole, even when Obama was polling well, it wasn't unusual for Romney to be polling higher by a couple points than Obama. That is a fact.

2. Gallup - We're comparing Gallup's RV model with their LV model and that's just not fair. Yes, Gallup shows Romney up among LV, but their RV model has consistently shown Obama in the lead ... even if it increases or decreases depending on the day. But the reality is, because Gallup for this entire election season has been operating under a RV model, and not a LV model, we don't exactly know what Obama's numbers were ... it's entirely possible, for weeks, Obama has narrowly trailed in Gallup's LV model ... even though he's led in their RV model. This makes perfect sense because, in the end, that's exactly what we see right now. So, any gain Romney has seen in Gallup has solely been by their changing of models. Had the LV model been used a couple weeks ago, it's entirely possible he was leading Obama in their polling all along.

3. IBD/TIPP Tracking - This is a poll that just suddenly popped up after the debates. It wasn't around prior to the debates and it shows Romney and Obama tied ... but did show Romney with a 4 (or 5?) point lead just a couple days ago. But again, here is a poll where we can't compare its numbers to the pre-debate because it just didn't exist ... or wasn't being reported on extensively.

4. Pew - This is the lone poll released the past few days that shows any significant Romney surge. Every daily tracking poll either showed the race tight before the debate (Rasmussen) or had a different model (Gallup).

5. No major network polls - CNN, CBS, ABC and NBC have yet to release any national numbers. That means, the extent of Romney's surge is almost universally tied to tracking polls ... specifically recent tracking polls ... that, as I've shown, are either showing minimal change or hardly any change. Romney has had a great deal of days prior to the debate where he was +1 on Rasmussen and Gallup's LV model could have shown Obama down even before the debate ... but since they were using RV, which still shows Obama up, we were under the illusion the switch supported the Romney bounce. No it didn't. I mean, it's possible it did, but it's also possible the debate bounce coincidentally coincided with the switch that shows Obama now trailing.

So, what I'm saying is that the polls we have right now can't suggest anything other than this race is tightening. But even so, races tighten, generally, the closer you get to election day. That happened in 2004, it happened in '96 and it even happened in '92 ... every leading candidate lost some support.

But without more non-tracking poll national polls out there, it's hard to state definitively whether Romney's bounce is lasting ... whether it's receding or whether Obama is now going to see a bounce. Basing your view on that bounce on polls that, as I've outlined above aren't necessarily negatively indicative of the race, is just not smart.

I would have expected Silver to see this. And while I get that he doesn't want to oversell Obama ... the worst thing he can do is undersell Obama. What if the next batch of national polls, which no one knows how they'll turn out, show Obama in the lead? Then what? Does Nate flip-flop on his model and talking point?

The reality is, and I think most people will tell you this, even Republicans, the dynamics haven't changed. What we do know is that the polls tightened, but the amount they've tightened is skewed by the lack of overall polls we have to base it on. Gallup, Rasmussen ... these are legitimate polls, but without ABC or CNN or CBS or NBC releasing their numbers ... we can't tell anything about this race.

So, Silver's assessment is flawed solely because the metrics he's using are flawed ... or at least don't support his claim.

[Mutiny In Heaven]

1. I think this is an excellent post

Don't get me wrong, I'm not an optimist by nature; I suspect that if held today, the election would be extremely close, with the popular breaking narrowly for Romney.

However, Silver's model looks like it's been imbibing, so all-over-the-place it has been.

The big question I have with regards the entire election is whether the pollsters have the LV judged correctly. This week, we've seen polls that show Obama anywhere from 2 to 6 points up among registered voters, but lagging by a couple when the LV screen is applied. It would be more or less unprecedented for a President to be defeated with such numbers.

[RBInMaine]

2. Silver sways and shifts day to freakin' day. Don't listen to all that daily chatter. Just noise.

[Baitball Blogger]

3. Background please.

Is this the Silver who just a week ago had Obama at a 70% chance of a win?

[aletier_v]

11. More like 97%

[TroyD]

12. Some people are confusing the NOW CAST

With the NOV 6 CAST.

Nate never had Obama with a 97% chance of winning on NOV 6. That was the NOW CAST. What the election would be if decided today.

[MFM008]

4. a few weeks??

Silver had Obama up 321. electoral votes and 87.1 percent of the vote chance of winning. % 51.6 vs RMoneys 47.3 of the vote on Oct 4. The morning after the debate. Stupid fking polls.

[gkhouston]

5. I think a big part of the problem is that Nate is used to doing baseball statistics,

where the input data is less subjective. Yeah, umpires make bad calls from time to time, but it's harder to cook the data without people noticing. As it stands, I don't think his current model's doing a very good job of compensating for LV screens. A lot of the recent volatility, IMO, has to do with switching from RV to LV models and the ways pollsters decide someone is a "likely" voter.

[cheriemedium59]

6. Thank you for all your hard work on this post. Excellent points!

[TroyD]

7. Just posted by Nate Silver:

Real good day for Romney in the polls. Maybe his best yet. He's now up to almost 40% in our forecast.

https://twitter.com/fivethirtyeight/status/256887679323619329

Hopefully the trend will start reversing in the days to come, but it's certainly a lesson never to take anything for granted again.

[ailsagirl]

8. Is NS a political animal as well as a statistician?

Or is he strictly a statistician?

[Maximumnegro]

9. Even Nate has admitted I believe that

his numbers from 2008 (?) were matched by other sites that use less, er, complicated methods. 538 is an accessible, visual delight but no one should take anything besides his numbers seriously. He is terrible at political analysis but he knows numbers. His model still has Obama winning that's all I care about. Anyway some things are palpable. When a candidate has indeed turned the tide we can sense it. And that's not happening for Romney. He's closed the gap which sucks but there is no true feeling of momentum, of okay this is really a problem. Especially with the SENATE still holding for goodness sakes - I think that's hilarious. So yeah all we can do is wait and do what we can to encourage turnout. But I definitely don't sense a real game change. Once again the O team analysis from earlier this week still holds: the race is essentially stagnant (which also shows how conservatively they judged Obama's lead in the previous weeks). They are on top of it.

[Drunken Irishman]

14. That's how I see it.

[TroyD]

10. 'the worst thing he can do is undersell Obama.'

How so?

[Drunken Irishman]

13. Are you really asking that?

It's pretty obvious why he can't undersell Obama. If he does, and next week the polls shift, he's going to have to shift his tune and it proves my point exactly ... Silver is nothing more than someone who looks up to the sky and tells us if it's cloudy or not. He's telling us nothing we don't already know.

[Denzil_DC]

15. I used to read 538 regularly in 2008

when it was an independent site.

If you remember, he used to have co-contributors who wrote some fantastic series of posts about their experiences visiting campaign field offices around the country etc. They were fascinating, and I used to really enjoy them.

It took the pressure off Nate having to come up with new content every day when sometimes there wasn't much new to say. Now he has to write posts even when he doesn't have much to report, so he often reaches and stretches, and he's not that good a pundit compared to many out there (Booman, for instance).

There wasn't much new data over the holiday weekend following the first presidential debate, hence the undue focus on some pretty weird outliers and plain scam polls. I can read rubbish like that anywhere on the Web. If he's not going to serve as a reliable editor of what's worth paying attention to and what isn't, then I have better places to spend my time nowadays.

His model is certainly contaminated at the moment with some rubbish polls like Gravis, and he doesn't seem to care as far as I can see. It's one thing to weight polls if they have a house effect, quite another to sometimes rely on them for a lede even if they're very suspicious or blatant outliers, as he has done.

I couldn't believe it last night when he cited that rubbish widely debunked CNN instapoll about the VP debate. Sure, he'll usually include a caveat when he mentions these less reputable polls, but why dignify them by mentioning them at all if they can't be trusted to signify anything? It's just filler.

I'll still be paying attention to him closer to the election because of his track record, but I'm afraid he's doing his reputation no good nowadays. We'll see whether he nails it this time, or even close.

[treestar]

16. Let us not let this guy panic us or for that matter comfort us

He is not god and can't tell us anything that will happen in the future - he can only talk about today. His "odds" of winning seem to be rather impossible to rely on. Only one poll really counts. Now there are so many polls, when it is close like this, there's no telling what is going to happen.