Nate Silver vs. Real Clear Politics

[cthulu2016]

RCP has a good track record and there's no point entirely dismissing their view—it is one of many, and worthwhile as such. However, I think their averaging method counts robo-call polls the same as live polls.

Robo-call polling is becoming increasingly distorted with each passing year. You cannot, legally, robocall a cell phone.

(I honestly have no idea what sort of weighting assumptions PPP makes to correct their robo-polls.)

I think Nate Silver downgrades robo-polls in his formulas. (I may be wrong, but that is my impression.)

[DemocratSinceBirth]

1. I Thought Nate Rates Polls By How Much They Diverged From Actual Results In The Past

I don't know how he rates pollsters without a polling history.

[cthulu2016]

2. Not on history, on method

He has written several times about trying to quantify the very real and skew in robo-polling and I am assuming that he downgrades his weighting of robo-polls in his election forecasts.

Wisconsin, for instance, is blue on 538 but a toss-up on RCP. I think the Romney friendly polls are robo-polls and that 538 counts them less, hence the large difference.

That's my impression from reading him over time. I cannot cite a statement to that effect.

[moondust]

3. These days

I'd be wary of any poll that doesn't include cell phones and Internet phones. I have both but I don't have a landline. Younger people are even more apt than me to be without a landline.

[JackN415]

4. About Nate Silver...

May I reprint one of my posts regarding Nate Silver's method and approach?

If Nate Silver stills consider "Gravy fraud" polling as legitimate, he loses credibility with me.

Any scientist must know and take a stand on the source of information or data. He might be political to be inclusive, playing the role of poll statistical meta-analysis, treating all pollsters with equal weight at face value.

But that is wrong at best and dishonest at worst.

It's his job to review the methodology, the data, the credibility of polling firm. Taking information without thorough investigation of the source is absolutely negligent.

Let's us consider an analogy: a medical researcher analyzes research results from other publications about certain thing, like whether red meat might cause colon cancer. The researcher cannot say that I will treat all publications on this subject with equal face value and hence my meta-analysis is this and this, etc.

It is the researcher's responsibility to determine the credibility and validity of the publication to be included in his/her meta-analysis. He/she cannot take a publication from a never-heard-of journal, from authors with dubious credential and say, I will treat this study the same as other.

Nate Silver is very wrong to include Gravy fraud poll firm.