« BACK  |  PRINT

RS

FRONT PAGE CONTRIBUTOR

Pennsylvania update

By: Neil Stevens (Diary)  |  May 12th, 2010 at 03:37 PM  | 

RESIZE: AAA

Pennsylvania Flag

We’re less than a week away from primary election day in Pennsylvania, so let’s take a look at the latest polling news from the House Special election to replace Jack Murtha, the Senate primary for Arlen Specter’s seat, and a surprisingly interesting House race.

Susquehanna Polling and Research came out with a pair of House polls. Their poll of the 12th district special election looks strange to me, though. Their sample of “400 Registered Likely Voters” shows a better than 2-1 advantage for Democrats, 63-29, and unsurprisingly Mark Critz leads Tim Burns in their poll 44-38 (MoE 4.9).

That partisan advantage strikes me as off. Cook Political Report gives the seat an R+1 PVI. For a comparison, Research 2000/Daily Kos gives Democrats a 53-36 (MoE 4) advantage in that seat. Susquehanna is giving Democrats a 17 point swing in registration, resulting in a 12 point swing for the Democrat.

I’m inclined to discount the poll pending confirmation.

So I’m also suspicious of this other Susquehanna poll I’ve come across, featuring trouble for incumbent Democrat Tim Holden in CD 17. Only 38% of likely primary voters want to see Tim Holden re-elected, though he leads the primary 54-27. The Republican four-way primary is a mess, with Dave Argall ahead of Frank Ryan 21-17.

But after seeing the huge discrepancy between Susquehanna and the usually reasonable DKos/R2000 poll, I don’t know whether to pay it much heed. Though oversampling Democrats wouldn’t really make Tim Holden feel better, would it?

Finally, the race we’ve all been waiting for, the primary between Joe Sestak and Arlen Specter. I said before that Specter needed to stop the bleeding to keep from losing badly. Per Real Clear Politics, Specter has regained a tie in the Morning Call/Muhlenberg College tracking poll, evening it up at 45. Anyone could yet win this, it seems.

Crossposted from Unlikely Voter

Tags: , , , , , , , , , , , , , , , , , , ,

COMMENTS

POSTING RULES  |  LOGIN TO COMMENT  |  REGISTER TODAY
  • tngal

    for chasing down the release on the poll. I found mentions of the it but they all led back to the tribune which gave little in the way of specifics. Although they’ve done polls before, it appears as if the burns- critz was simply not good polling. 400 people? almost a 5 MOE?

  • jaybo

    It is clear to see this when you see a sample size of only 400 and a higher than average margin of error that is 4.9%.

    Those are red flags that should warn anyone that the poll is not reliable.

  • IJB

    This is a Dem-leaning poll, and it *still* only gets Critz to 44%. That can’t be good news if you’re a Dem…

  • jaybo

    Their polls consistently inflate support for democrats. If you watch their tracking poll for Presidential Approval, you will see that it regularly gives The President 3-4 points more than most other polls give him.

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

    You can’t discredit every poll they put out because of a slight difference in a tracking poll.

    Besides, there’s no way of knowing who’s right on a tracking poll.

  • http://www.hakubi.us/ Neil Stevens
  • Return to Revolution

    Anyone who votes for Specter, the squishy, traitorous moderate, isn’t paying attention, so that matchup won’t tell us much about the partisan future of PA. Sestak/Toomey, however, is different – both have similar name recognition and neither sits in the middle. If PA ultimately elects Sestak in a wave year favoring republicans, there is no hope for us turning red or even remaining purple. I think Toomey will pull it off against either dem in the end.

    The PA-12 poll would be more accurate if Murtha were still here. Now that voters no longer have a “beloved” representative to blindly vote for, I wouldn’t put much stock in it given their Cook voting preference.

  • nateborn

    Dear Redstate Staff,
    Please look at Rep. Sam Rohrer. He’s PA’s Nikki Haley. Please research him, and take a sincere look at supporting him. The primary is quickly approaching.

    http://www.samrohrer.org/

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

    The Daily Kos/ R2k Poll tracks the Presidents Favorablity, which typically is higher than approval. There is a difference, and typically, with polling, wording like that matters.

    Though, most of the time, they do show slightly more favorable numbers for the democrats, so I am not discounting a house effect entirely.

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

    Very good point.

    And I suspect any house effect is in a different Likely Voter secret sauce. They probably weight 2008 more than we would.

  • jaybo

    The smaller the sample size the greater the chance that the poll is NOT a representative picture of the population you are measuring.

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

    I’m curious.

  • smoothsailing

    This is an accurate sample based on voter registratioion as cited by the Pennsylvania Dept. of State.

    The poll was commissioned by the Pittsburgh Tribune-Review, a conservative leaning paper that has endorsed Burns in the special election.

    The poll was conducted by Susquehanna Polling & Research, a Republican firm.

    The poll also shows Burns to be in a statistical deadheat with Lt.Col.(ret) Bill Russell in the Republican Primary

    “There were 256,281 Democrats compared with 118,175 Republicans as of last year?s general election, according to the Pennsylvania Department of State.”
    Source: http://tribune-democrat.com/local/x892954498/Special-election-Burns-Critz-work-to-gain-an-edge

    There are 410,000 registered voters in PA-12

  • jaybo

    “They probably weight 2008 more than we would.”

    Wouldn’t it have been easier to just say that right from the beginning?

    If you weight a poll in a way that is different from most (if not all) reputable pollsters, isn’t that an unreliable poll?

  • jaybo

    In other words you cannot draw any conclusion from it unless you have supporting data that supports its results.

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

    I said they’re weighting 2008 more than *we* would. But we’re partisan activists.

    The only reputable pollster who seems to be discounting 2008 is Rasmussen. All the rest are taking it in stride.

    Research 2000′s results are not out of line in any I’ve seen so far. I don’t know where you’re getting that other than the fact that the polls are commissioned by Daily Kos.

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

    You brought up statistics. What statistical test are you using to declare 400 an insufficiently large sample?

    The standard tests are giving it a perfectly reasonable Margin of Error, but I’m interested in what math you did to draw your conclusion.

  • Achance

    would be in big sparsely populated places like the Dakotas, Montana, Wyoming, Alaska, etc. Those places are so diverse that it takes a larger sample and they’re realy difficult and expensive to get.

  • jaybo

    On occasion you will see a poll that appear to come close to other reputable poll like Gallup but most of the time they favor democrats.

    Taken by themselves, they are not reliable.

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

    Why don’t you write a diary showing us all the times The R2000/DKos has messed up, and prove the case once and for all?

    If they’re so bad you should have lots of examples, right?

  • Richard Mullins

    I remember that it was big when I first lived in it and it seemed to have gotten larger the second time. No district in this state is as big as that even thought there are more sparsely populated districts here. PA-12 could fit many times over.

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

    You may be surprised to learn that not all people who are registered to vote actually bother to do it.

    Thus, despite the partisan registration advantage, the district went only 54-43 Gore over Bush, 51-48 Kerry over Bush, and finally 49-49 McCain over Obama.

    So some Democrats are voting Republican, but a bunch of Democrats probably aren’t voting at all.

  • jaybo

    This is the biggest problem that I have with the plethora of polling that is done today. The average reader that is not familiar with statistical analysis, simply assumes that it must be correct. But this is not necessarily the case at all. When polling politics you must also “estimate” the population subsets (republican/democrat/independent) in order to “weight” the raw data that comes from the sample. But again, the smaller the sample size the greater the possibility that you do not have an accurate representation of the population

    I am not saying that the poll is incorrect. I am saying that you cannot conclude anything with this poll alone.

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

    But you keep making claims without showing that you, yourself, have done a shred of statistical analysis on the matter.

  • smoothsailing

    Yes, we are talking about likely voters….

    F1: As you may know, a special election for United States Congress has been scheduled in conjunction with the May 18 Primary Election due to the untimely death of former Congressman John Murtha. How would you rank your chances of voting in this upcoming special election: Would you say you are definitely voting, very likely to vote, somewhat likely to vote, not likely to vote, or definitely not voting?
    3/10 %Ch
    1. Definitely voting 308 77% 68% +9
    2. Very likely voting 68 17% 21% -4
    3. Somewhat likely voting 24 06% 10% -4
    4. Not likely voting (TERMINATE)
    5. Definitely not voting (TERMINATE)

    It may surprise you to learn that although their is some intensity among Republican voters in the district to vote in the specia; election, there is an equal untensity among Democrat voters to vote in the Senate Primary between Spectewr and Sestak. Both events are the same day.

    That’s why I maintain the sample size is accurate and no pollster worth their salt would apply a filter or “special sauce” to the May 18th special election.

    It’s unfortunate for Burns,whom I support in the special election, but that’s the way it is.

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

    Yeah, there are better ways of modeling voter likelihood than that.

    You don’t come out and ASK though, sheesh.

  • smoothsailing

    on this ’til the cows come home.

    I live in the district and will vote for Burns in the special and Russell in the primary.

    We’ll see how it all shakes out on the 18th.

  • proudgop

    The thing with that congressional seat is that over 50% of registered voters in district are registered Dems so in essence whoever wins Dem vote wins the seat

    Its going to be close race I hope Burns prevails we need this big time

  • IJB

    Who won’t even endorse the GOP candidate in the Special Election?

    I sure as heck wouldn’t vote for somebody like that.

  • throwback59

    primary, what party is he going to jump to next?

  • http://www.hakubi.us/ Neil Stevens
  • http://www.suvstrategery.blogspot.com SoFiMil

    n/t

  • Richard Mullins

    Will Specter go Independent if he loses the Primary to Sestek? That seems to be a logical question to ask.

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

    My understanding is he’d have had to change his own party registration already.

  • http://www.suvstrategery.blogspot.com SoFiMil

    .

  • Richard Mullins
  • http://www.suvstrategery.blogspot.com SoFiMil

    nt

  • Richard Mullins
  • http://www.hakubi.us/ Neil Stevens
  • http://www.suvstrategery.blogspot.com SoFiMil

    Then: http://www.politico.com/news/stories/0409/21800.html

    Now: (With some expert commentary by Joe Biden)
    http://tpmlivewire.talkingpointsmemo.com/2010/05/biden-sure-ill-campaign-for-specter—-but-im-not-sure-obama-will.php

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

    I mean, there is that one thing about him that is, ahem, “Not Proven.”

  • novimir

    If the Susquehanna Polling & Research has the correct voter model, how can you argue that a perfectly weighted sample of 400 voters will give an incorrect result.

    Your argument boils down to “poll and pray” i.e. the bigger the sample size, the less accurate your polls weighting needs to be and the more the pollster is hoping that they got a random sample.

    I am with Neil Stevens. make a mathematical argument for why the Susquehanna Polling & Research model is wrong.

  • Richard Mullins

    Having him so close to Specter is sort damaging politically. Biden isn’t the smartest person to have around anyone. Having Obama isn’t much better either.

  • smoothsailing

    doesn’t do endorsements, but it has said this about Tim Burns and the special election…

    “But we certainly do hope he wins the special election,” said Russell spokesman Steve Clark.”

    http://www.pittsburghlive.com/x/pittsburghtrib/news/westmoreland/s_680706.html

    That should be enough for any reasonable Burns supporter,

  • IJB

    Sounds like just another politician to me.

  • http://moelane.com/ Moe Lane

    …Specter would have needed to change his party registration 30 days before the primary in order to be able to run as an independent.

  • streiff

    N=400 in a CD, assuming a statistically random sample, is plenty.