« BACK  |  PRINT

RS

FRONT PAGE CONTRIBUTOR

The Times’s Nate Silver Punishing Pollsters?

By: Neil Stevens (Diary)  |  August 25th, 2010 at 02:30 PM  | 

RESIZE: AAA

538

So 538 moved to the New York Times this morning and in the process made Marco Rubio the favorite finally.

But seriously, my issue with Nate Silver today comes from the old site and specifically, his primary night commentary.

Nate Silver said this:

Rasmussen — which polled the McCain-Hayworth primary eight times in a race where there was some disagreement among pollsters — was not willing to do so during the final four weeks of the campaign. Our pollster ratings are always becoming more sophisticated and we’re going to be looking at appropriate ways to punish pollsters who dodge putting their necks on the line.

If I relied on Nate Silver for objective commentary on polling, this comment would make me uncomfortable. Why is this guy looking to punish public pollsters for what races they cover or not, especially when it was obvious that Hayworth’s deficit had grown insurmountable? Why should Rasmussen have spent money to keep polling Arizona to keep tabs on whether Hayworth lost by 30, 40, or 50, when the firm clearly is not a charity, primaries aren’t the firm’s usual thing, and there were much more interesting races to follow?

Seriously: Rasmussen doesn’t even poll primaries in general. I’ve lamented that many times myself. That Silver wants to penalize Rasmussen for having gone beyond the norm by publishing a few Arizona results to me suggests an ulterior motive. And given his track record as a member of the old Journolist steering committee of political media coverage, I have to think this comment marks him as part of the partisan jihad against Rasmussen Reports.

But I’m shocked to find out that Silver would weight pollsters less, and thus alter the accuracy of all his work on a site, for reasons that have nothing to do with accuracy or reliability of the results. If Rasmussen is right, but is penalized for other reasons, then the integrity of Silver’s whole operation is called into question.

Once again, Nate Silver puts partisanship over math. And as always, he knows better, but he just doesn’t care. I’m sure his buddies in Journolist II are happy with him, though.

From Unlikely Voter

Tags: , , , , , , , , ,

COMMENTS

POSTING RULES  |  LOGIN TO COMMENT  |  REGISTER TODAY
  • Death_of_the_Donkey

    so long as it is applied equally. I didn’t see where Rasmussen was in this race for the 8 polls he did, but if he had Hayworth close and then didn’t poll in the last 4 weeks, it does make his polls “less” valuable. Like I said though, if Rasmussen has Hayworth getting blown out in his “last” poll though, then I agree with you and why waste the effort. While Nate may have some personal bias against Rasmussen, I highly doubt that he would let that get in the way of the data, as I think Nate’s ego about being right is bigger than his bent against Rasmussen.

  • http://www.hakubi.us/ Neil Stevens

    Why should Rasmussen polls be weighted less in every race, every time, when the weightings are usually accuracy-based?

    He already weights for time.

    All this will do is make Rasmussen polls count less in general election races that the firm covers very well.

  • Brian_Roastbeef

    Has hated on Rasmussen for a while. He is always looking for an excuse to jump on them.

    Maybe it is because Rasmussen is telling him stuff he doesn’t want to hear? If so, then he might as well go ahead and punish Rasmussen in his numbers and report liberal bias so he can be thoroughly discredited. I see he’s already made the move to the proper website for doing so.

    Or he can retain credibility and listen to what Rasmussen and most other reputable pollsters (and a few less reputable pollsters) are telling him about what is coming this November.

  • Death_of_the_Donkey

    them, but that every given cycle is different and it becomes harder to gauge who is using the most accurate polling model when you do not get closer to election polls that may/may not back up/shoot down earlier polls. Like I said, this makes no sense if Rasmussen was showing a big McCain win earlier, but if he was showing Hayworth winning/really close, then it makes some sense to question the modeling itself (again without a more recent poll to rely on). Having said that, I think that in the end, Nate was just whining and will likely take no action, as like I said above, he is more concerned with his accuracy than ideology for calling elections.

  • http://www.hakubi.us/ Neil Stevens

    So your defense of Silver is that he’s just ranting emotionally, and doesn’t actually mean it?

  • grateful_red

    The interactive map has WI as Lean Dem at 71% Feingold

    The interactive map has MO as toss up at 91% Blunt

    The interactive map has OH as toss up at 81% Portman

    The interactive map has OH as toss up at 88% Toomey

    So it’s a toss up if it’s 80% or more likey to go republican but it’s leaning dem at 70% ?!?!?

  • azaeroprof

    The idea behind 538 is awesome and would be a great resource if put in the hands of a professional who could separate his/her bias from the model. Maybe someone like…..Scott Rasmussen!

    I started following 538 back in the early stages of the ’08 election and was intrigued by the sophistication of the model (c’mon, I’m an engineer!). But it became clearer and clearer and finally demonstrably obvious that 538 was nothing more than an advocacy site for Barack Obama. It has now become the official Statistics Modification Arm of the Obama WH.

    ‘Tis a pity, as a model that assimilates the whole body of polling data to discern reliable trends is sorely missing in American politics. The RCP average is a step above individual polling results, but way short of being scientifically reliable.

  • nvrepub

    nt

  • finaljeopardy

    Ace of Spades collects leftist comments about “the Hitler” of polling. Silver is a diehard Obama supporter who moved to New York recently from Chicago to become part of the establishment media. For him, Rasmussen is to polling as FOX is to TV. Let him poll like Kos, but watch his polls. He will reveal the Democratic strategy to you in the upcoming elections.

  • finaljeopardy

    That’s a good one to look at now. It’s up in HotAir’s Green Room. Ishmael uses a score he calls the Partisan Voting Index Multiplier Electoral Risk (PVIMER)which he explains at length here:

    I?ve aggregated what I believe are the latest predictions from The Cook Report, Sabato?s Crystal Ball, Congressional Quarterly, and the Swing State Project. I?ve assigned numbers to each of the races they?ve rated and added an enthusiasm multiplier, which I?ll explain later.

    The scale is straightforward: Toss-ups are rated ?0?, Leans rated ?1?, Likelies rated ?2?, and Solids rated ?3.? Negative numbers mean, based on these four sources, a likely GOP seat takeover; positive numbers suggest better prospects for Democrats.

  • lucretius

    but Silver suspects that Ras is avoiding accountability. Basically, in polling the only poll that can be graded somewhat objectively is the last one. Which is why many pollsters re-weight and tweak their likely voter screens; nobody wants to be the outlier who blows it.

    Many of us on the left (likely including Silver) suspect that Ras tries to drive the narrative or test candidate strength in its unnacountable early polls, and only refines its data later for the big races in order to keep up its track record. If the early and later polls are too out of whack, they open themselves up to charges of data manipulation. The easiest way to avoid such accusations is to simply stop polling late in the race.

    Maybe Ras isn’t doing this intentionally, I don’t know. But any pollster who tries to avoid accountability invites suspicion and distrust.

  • finaljeopardy

    is the one who complains that the McCain-Hayworth primary isn’t being polled enough by a firm. Unless you were trying to “drive a narrative” there would be no point to keep polling once the winner’s lead was too big. Only a hack pollster would waste money trying to prove that the gap was narrowing.