Nate Silver pooh-pooh's PPP's PA senate poll which has Sestak up by 3...

[book_worm]

because it doesn't use his magic formula--and obviously he feels PPP is a parisan organization but not Rasmussen!? I wonder what he will think of the Independent Morning Call poll showing Sestak up by 3 today! I think Nate is starting to fit in just fine with the MSM:

I’ve been fielding a lot of questions about the Public Policy Polling survey in Pennsylvania’s Senate race that gives Representative Joe Sestak, the Democrat, a 1-point lead over the Republican Pat Toomey.

I wrote last week that a comeback by Mr. Sestak, who had been trailing Mr. Toomey by around 6 or 7 points in most surveys until now, was not more likely than the chances assigned to him by our model, which had been about 5 percent.

Obviously, we have new information in the form of the Public Policy Polling survey. But a few things to keep in mind:

Usually, general elections don’t turn on a dime without good reason — and going from 7 points down (or 9 points, as Public Policy Polling had Mr. Sestak in its August survey) to 1 point ahead would be fairly unusual. It can happen, but it doesn’t happen that often.
You’ll rarely make a mistake by holding out for more data. While pollsters have been distracted by states like Delaware that are unlikely to be close on Election Day, Pennsylvania has received uncharacteristically scant polling. So far, no other nonpartisan poll confirms Mr. Sestak’s surge, and the next-most-recent survey of the state, from Rasmussen Reports, gave Mr. Toomey a 10-point lead — the largest advantage Rasmussen had given him to date. One poll does not a trend make, and even when several polls do agree on a trend, it can often reverse itself.
Public Policy Polling conducts surveys for Democratic candidates (and Daily Kos, a liberal blog) in addition to issuing surveys under its own name. Until recently, we have not found an especially large “house effect” for Public Policy Polling — that is, they’ve had plenty of surveys showing poor numbers for Democrats. But lately, such an effect has arguably become more noticeable: they are the only pollster, for instance, to show the Democrat Michael Bennet with a lead in Colorado, although several other pollsters have shown that race tightening. And their survey of Arizona’s Third Congressional District, which showed the Republican Ben Quayle trailing in a district that ordinarily leans Republican, has raised a few eyebrows, although Mr. Quayle may be an unappealing enough candidate that the result is not necessarily implausible.
The most prudent conclusion is that Mr. Toomey still holds a lead, but it is probably smaller than the one he held before. Still, even small leads can be meaningful at this time of year, and Mr. Sestak may not be helped by the fact that Democrats are performing poorly in the governor’s race in Pennsylvania as well as several competitive House races around the state.

When I reran our numbers with the Public Policy Polling survey included, Mr. Sestak’s chances were improved, but only to 10 percent.

With that said, our Senate model does not use partisan polling in its forecasting. But publications like The Hotline have indicated, and some contacts of mine have related, that not only do Democratic internal polls show the race tightening, but Republican ones do as well. For that reason, I would be inclined to take Mr. Sestak’s side of the 9:1 odds that our model is offering him.

http://fivethirtyeight.blogs.nytimes.com/

[monmouth]

1. I think Nate has gone to the dark side. Thanks for the info..n/t

[girl gone mad]

13. Nate Silver is an enemy of the people!

[tiptoe]

38. Richard Charnin (TruthIsAll) challenged Nate Silver with an open letter and 25 questions:

Here are Charnin's comments and questions for Silver:

Open Letter to Nate Silver

Twenty-five Questions for Nate Silver

Zogby/Silver:1996-2008 True vs. Recorded Vote Pollster Rankings

[eleny]

2. I guess political outcomes are more fluid than sports

:hi:

[bigdarryl]

3. He's an ASS!!! PPP polling what every you say about them are usually spot on

[Blue_In_AK]

4. This may be sacrilege here,

as I've gotten called on it before, but Nate Silver can blow it out his ass, as far as I'm concerned.

And, once again, POLLS ARE VOTER SUPPRESSION TOOLS! Ignore them and vote your conscience.

[DarthDem]

5. Nate . . . Done as a Relevant Force

Nate gets a lot of attention around here because he was right in 2008 - - but (1) before he was right, he was dead, dead wrong, indicating that McClown would win as late as Labor Day before finally reading the tea leaves correctly, and then only very cautiously until the end when it was pretty obvious that Obama was going to do VERY well; and (2) he's changed. He sold out, and there's no way that's not affecting his modeling.

He's not to be accorded any more weight than Chuckie Todd at this point. But watch for backpedaling as Election Day nears.

[Drunken Irishman]

7. A LOT of people were right...

Hell, even I was right.

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=132x7713304

The only states I got wrong was Missouri!

[Cosmocat]

23. It is interesting that you noted Chuck Todd ...

5-6 years ago when Todd was cutting his teeth on Hotline on CSPAN, I could have swore he was progressive ... He pretty much called it straight, but when he was reporting good news for the dems, he just seemed to enjoy it more ...

Today, obviously he has moved in the other direction in a VERY big way ... I have seen his progession along the way, and I think it highlights the perverse reality to the "liberal" media nonsense ... I saw it with Luke Russert, too ... First time they let him on the air in the 08 campaign he reported on the "youth vote" and simply by nature it was very good news for BO ... The very next day they put him on again and he basically apologized for not giving McCain enough credit and obediently started to note how it was very possible that it would be bad news for BO with the youth vote for some hair brained reason ... Since, he has been the happy little D basher and R cheerleader ...

I think when the career, and the big pay checks, start to get build, all of these people find a GREAT deal of pressure to confirm to the "Ds in trouble, Rs on the prowl" memes that MSMers are under orders to push ...

Either way, again, I have pointed out recent trends in statewide PA elections, and no one really has PA calibrated correctly ...

[geek tragedy]

6. He's being cautious--he can't reverse his take on a race

based on 1-2 recent polls.

[vaberella]

32. The man sat there and put all the Dem candidates at the bottom of the barrel and unlikely to win.

Then out of nowhere there are massive surges on state polls, independent polls, and so on. I think he just doesn't want to look like a punk based on his first statement.

[impik]

8. Enough with this f*&*(&^&g "model" already

[DrToast]

9. Funny how everyone thought Nate was a genius in 2008

He's just reporting the numbers, folks.

[grantcart]

21. The problem is the swing in 'likely voters'. If Democrats get motivated later than the

Republicans then a lot of people are going to be wrong.


As for General elections turning on a dime I find that surprisingly because many general elections were turning and likely to get a different result of the GE was only a few days away.

[DrToast]

22. They won't be wrong

Nate is reporting on the state of the race right now. He reports what the polls say.

[grantcart]

26. You are holding a position that is empirically 100% untrue and that is

a rare ahcievement at DU.


The reason that it cannot be true is that the pollsters are using widely different models for 'likely voters'. As Blumenthal at pollster.com
remarked

http://www.huffingtonpost.com/2010/10/05/likely-voters-how-pollsters-choose-them_n_751560.html

Needless to say, those numbers can't all be right. The enormous variation has left a lot of reporters and readers emailing to ask, just how do pollsters identify a likely voter? More important, does anyone know what the heck a likely voter is anymore

Just how bad can it get? How about 'shocking' and 'implausible'?


http://www.huffingtonpost.com/alan-abramowitz/gallups-implausible-likel_b_764345.html


It's a shocking result. According to the Gallup Poll, a generic Republican candidate currently leads a generic Democratic candidate by 17 points among likely voters in a hypothetical House matchup. A margin of that magnitude on Election Day would almost certainly result in a Republican gain of at least 80 seats in the House of Representatives and the largest GOP majority since the 1920's. But how plausible are Gallup's results?

An examination of some of the internals from the latest Gallup survey of likely voters leads to the conclusion that these results are wildly implausible . First, Gallup shows a much larger percentage of Republicans (55% Republican identifiers and leaners vs. 40% Democratic identifiers and leaners) and conservatives (51% conservative vs. 28% moderates and 18% liberals) than we've ever seen in a modern election. They also show a smaller percentage of voters under the age of 30 (7%) and a larger percentage of voters over the age of 65 (27%) than we've seen in any modern election. But that's not all. The candidate preference results for some subgroups of voters are just wildly implausible.

Gallup's latest likely voter survey shows a generic Republican leading a generic Democrat by a whopping 28 points among whites, 62% to 34%. To put those numbers in perspective, in 1994, according to national exit poll data, Republicans only won the white vote by 16 points, 58% to 42%, and that was their best showing since the advent of exit polling. Gallup is telling us that right now the Republican lead among whites who are likely to vote is 12 points larger than the GOP margin among whites in 1994.

But that's not the most implausible result in the latest Gallup likely voter survey. Among nonwhites other than blacks, a group that comprises about 13% of likely voters, a generic Republican is leading a generic Democrat by 10 points, 52% to 42% That's a group that voted Democratic by a 2-1 margin in the 2006 midterm election. Moreover, it's a group that has never given a majority of its vote to Republican candidates for Congress in any election since the advent of exit polling. According to the 2006 exit poll results, about two-thirds of these "other nonwhite" voters are Latinos. How plausible is it that at a time when the Republican Party is closely associated with stridently anti-immigrant policies that Latino voters are moving in droves toward Republican candidates? Not plausible at all, especially when Gallup's results are directly contradicted by other recent polls of Latino voters.

f

As I am sure you know Nate Silver has developed his political forecasting based on his baseball model. It is based on an endless stream of statistical information that is of the same quality. It worked well on the electoral map because there was such a high volume of polls in each state that he was working on.

Now he is trying to use the same method on Senate races that are polled infrequently and, as noted above, by pollsters that are using wildly different LV models. It is even worse at the Congressional Seat level.

Just look at the most recent articles on polls showing dramatic increases for Democrats

http://www.huffingtonpost.com/2010/10/21/polls-joe-sestak-rebound-boxer-brown-lead_n_771128.html

http://www.huffingtonpost.com/2010/10/20/sestak-and-bennet-running_n_769469.html

[DrToast]

27. You fail to understand what Nate does

He's forecasting results of elections based on the variables that have worked historically.

Every election pollsters have different likely voter models. It doesn't matter if some of them are right and some of them are wrong. His model makes the best use of this information.

It's embarrassing to see so many people shitting on his methodology basically because they don't like what it's saying.

[grantcart]

30. You haven't read the pollster.com articles that are documenting

that the polls are using widely varried methods on 'likely' voters.


I understand exactly Nate Silver's modelling and how he adapted it from his work on Baseball.

For the computer models to work well he needs a large supply of polls that cover the same population. That was fine for a large presidential race where there were daily polls in the contested states.

Up until now most of the contested states only had a few polls a month and the contested Congressional districts a few polls since the begining.

When the polling experts call the screening for LV "shocking" and "implausible" then you have garbage in/garbage out. In the Presidential election of 2008 there was a general consensus on what a LV voter was. That consensus is gone and NS is giving equal weight to all polls no matter how 'implausible' they are. Again Gallup has a poll of LV that shows Black voters preferring a Republican over a Democrat 52-42. Those kind of discrepancies were not present in 2008.

[DrToast]

28. Additionally...you changed your argument

You claimed that somehow if Democrats became motivated in the future, that would retroactively make the historical polls wrong. That's ridiculous. Polls show where the electorate is today, not in the future. Likely voter models are based on polling, too. If Democrats became more motivated in the future, then likely voter models would reflect that.

[grantcart]

35. Campaigns matter.

It has been an unusual season because the Tea Party started their campaign 6 months ago.

Democrats have started their campaign now.

The polls have showed all along that Registered voters favored Democrats in many of the contested elections but not in likely voters.

What is so ironic about your replies is that Silver has already agreed with the point I am making:


As I have warned repeatedly in the past, we believe that the uncertainty in the forecast is intrinsically quite high, stemming from the unusually large number of seats in play, and from differences of opinion among pollsters in how to calibrate their likely voter models to account for the so-called “enthusiasm gap.”

http://fivethirtyeight.blogs.nytimes.com/


Unfortunately Silver has also said that he does not think that it is possible for campaigns to have signifciant changes at the end. While that may be true for "who" people are voting I don't think that will hold for "which" people are voting. There is no polling for it but I suspect that a lot of Democrats in PA are seeing the race in DE and all of a sudden got a lot more enthusiastic about Sestak. That would account for the sharp increase in Sestak numbers that Silver discounts as possible at the end of a campaign.

[DrToast]

36. Let's review what you said

The problem is the swing in 'likely voters'. If Democrats get motivated later than the
Republicans then a lot of people are going to be wrong.

You're suggesting that present day likely voter models are flawed because they fail to capture potentially more motivated Democrats on election day. This makes no sense at all. Likely voter models are derived from current polls. Whether or not pollsters are going to be wrong with their likely voter model is completely unrelated to whether or not Democrats are motivated in the future.

Democrats can be depressed in November and likely voter models are wrong. Democrats can have record turnout and likely voter models could end up wrong. They're unrelated.

[vaberella]

33. I never did. Numbers are numbers and can be wrong.

As an Economics major---I never bet on numbers unless I have all the variables under some likely eventuality. Most numbers people don't count all the factors---messes with their numbers.

[DemocratSinceBirth]

10. In Less Than Two Weeks We Will Know If Nate Is A Sage Or A Stooge

I will continue to look at all the polls.

[Bluenorthwest]

11. Nate is too new to take seriously,

The thing that bothers me with Nate is that his track record is very short, and the sort of credibility he assumes is for those who have a very long and consistent record.

[Arkana]

12. Nate seems to be ignoring a genuine trend in the race.

I think ultimately the only way to prove who is right will be to see what happens on November 2, though.

[Aramchek]

14. Nate is rushing headlong into Political Hackery

[Dawson Leery]

15. Nate has joined (R)as.

I mean that. GOTV!

[PBS Poll-435]

16. Go back to Baseball stats, Nate. nt

[Clio the Leo]

17. I poo-poo anyone who poo-poo's Nate Silver. nt

[thewiseguy]

18. Had Nate Silver said something that we would have liked

Then he would have been touted as a genius. Oh the hypocrisy...

[DrToast]

19. Seriously, this thread is embarrassing.

[DFLforever]

20. He wasn't a genius in the MN gubernatorial primary

this year: one hour AFTER the polls had closed, he called the election for the eventual LOSER.

Obviously, his MN model sucked this year, although he was very close in 2008 with the Franken recount.

There is another poll out today confirming the Sestak lead in the PPP poll.

[vaberella]

34. For some maybe....for myself hardly. n/t

[ProSense]

24. New Quinnipiac poll, race a statistical dead heat

The race for Pennsylvania's U.S. Senate seat is now a statistical dead heat with Republican Pat Toomey getting 48 percent of likely voters to 46 percent for Democratic U.S. Rep. Joe Sestak, according to a Quinnipiac University poll released today.

link

[Unvanguard]

25. What is your complaint with his commentary, exactly?

He thinks PPP has a house effect, and provides examples of why he thinks so (and, yes, he thinks Rasmussen has a house effect too); he points out, correctly, that there are such a thing as polling outliers, and waiting for more data is wise; he also points out, also correctly, that such a drastic comeback would be a pretty are event.

So?

(Also, he's referring to the +1 PPP poll, not the more recent +3 poll from a different pollster, which in my view makes the "Sestak comeback" idea substantially more robust.)

[budkin]

29. Nate jumped the Shark when he joined the Times

[polichick]

37. sorry, self-delete

[vaberella]

31. Nate's an idiot. n/t

[leftygolfer]

39. Nate, drinking the MSM Kool-Aid

He is wrong, I think he is just in denial now. Dems hold both the House and Senate. Nov. 3rd will be so sweet!!!