POTUS Debate II: The Empire Strikes Back at the Wrath of Electric Mittensaloo

[derspiess]

Even an RCP average, as unsophisticated as it is, is still a better choice than any one individual tracking poll. 

Yeah, and I check RCP every day.  I check 538 every couple of days.  We will see in a few short weeks if the two polls I pay particular attention to are suboptimal choices.  I know one of them was dead on in 2008.

Question for you:  What do you say to the accusations from the right that some of the polls are sampling democrats too heavily?  From what I understand, some of them were supposing even greater enthusiasm (i.e., likelihood to get out and vote) for Obama than in 2008, and I have a hard time believing he can match the enthusiasm his supporters had for him 4 years ago.

[DGuller]

Question for you:  What do you say to the accusations from the right that some of the polls are sampling democrats too heavily?  From what I understand, some of them were supposing even greater enthusiasm (i.e., likelihood to get out and vote) for Obama than in 2008, and I have a hard time believing he can match the enthusiasm his supporters had for him 4 years ago.

I'm not equipped to answer that question.  Sample weighting is the black art of polling, because you won't know how right you are until election is over.  You have to do it, because without it, you're going to have too much error.  But, if you do it poorly, you're also going to have too much error (which is what I suspect the problem with Gallup is). 

This is one reason why I put a lot of weight on 538.  I don't know exactly what Nate Silver does to account for that, but whatever it is has to mitigate the overall effect of some pollsters not getting their weights right.

[Jacob]

Oh look, Romney made money from government bailouts while outsourcing jobs and denying healthcare and pensions to American workers: http://www.thenation.com/article/170644/mitt-romneys-bailout-bonanza#

[derspiess]

The Nation?  Should I start posting anti-Obama articles from NRO??

[garbon]

The Nation?  Should I start posting anti-Obama articles from NRO??

At any rate, I'm not sure I understand this. Isn't this further ammunition that the bailout didn't really help those that it should have?

It isn't like Romney championed a bailout in order to profit from it, is it?

[MadImmortalMan]

I'm not equipped to answer that question.  Sample weighting is the black art of polling, because you won't know how right you are until election is over.  You have to do it, because without it, you're going to have too much error.  But, if you do it poorly, you're also going to have too much error (which is what I suspect the problem with Gallup is). 

This is one reason why I put a lot of weight on 538.  I don't know exactly what Nate Silver does to account for that, but whatever it is has to mitigate the overall effect of some pollsters not getting their weights right.

Silver aggregates the polls. That's why he's good. He takes all of them and does his own analysis. FWIW, Gallup has been closest to the actual election results the last several times so I wouldn't discount them easily. Pew either.

It stands to reason that since Silver is making use of an aggregate of a bunch of polls, the most accurate single poll in the group will necessarily be more accurate than 538's average. It just has to be the case mathematically.

[DGuller]

Silver aggregates the polls. That's why he's good. He takes all of them and does his own analysis.

Aggregating the polls is easy.  Figuring out how to extract the most information and the least noise out of the whole bunch is the very hard part.  That's what Silver does, and he does it pretty well.

FWIW, Gallup has been closest to the actual election results the last several times so I wouldn't discount them easily. Pew either.

I don't keep track of such stuff on my own, but my impression was that this wasn't the case.

It stands to reason that since Silver is making use of an aggregate of a bunch of polls, the most accurate single poll in the group will necessarily be more accurate than 538's average. It just has to be the case mathematically.

  I don't think that's true.  And even if that were true, that's kind of pointless to know, because there is zero predictive power in knowing that one of the points that make up the averages will get it pretty close in hindsight.

[Jacob]

The Nation?  Should I start posting anti-Obama articles from NRO??

Is that what the Nation is equivalent to? I'm not familiar with all American media, alas

[The Minsky Moment]

Anyone invested in stocks, bonds or just about financial instrument would have money from the bailouts.

[Habbaku]

The Nation?  Should I start posting anti-Obama articles from NRO??

Is that what the Nation is equivalent to? I'm not familiar with all American media, alas

The Nation does some good work on occasion, just like NRO, but they are a very biased source, yes.

[derspiess]

The Nation?  Should I start posting anti-Obama articles from NRO??

Is that what the Nation is equivalent to? I'm not familiar with all American media, alas

I would say so.  It has called itself "the flagship of the left".

[The Minsky Moment]

Oh look, Romney made money from government bailouts while outsourcing jobs and denying healthcare and pensions to American workers: http://www.thenation.com/article/170644/mitt-romneys-bailout-bonanza#

This is a pretty silly article.  Romney's investment was in the hedge fund Elliott Associates.  Elliott had a total return of just over 4% in 2011, which is not terrible but not really lucrative either.

[Jacob]

I trust your judgement in this, Minsky 

[The Minsky Moment]

I think the whole attack on Mitt's investments and his tax rate misses the point.  What matters is the policies he would push for.  The problem isn't that Mitt is taxed at 14%; the problem is that Mitt supports a tax system that taxes work (wages and ordinary income) at a much higher rate than taxes on asset returns, as a result of which people like him are taxed at much lower rates than ordinary workers.

[garbon]

I don't keep track of such stuff on my own, but my impression was that this wasn't the case.

538 posted an article yesterday that actually said the opposite for Gallup in that it wasn't good at predicting election results.