VA-05: I Will Make You Hurt

SurveyUSA (7/16-19, likely voters, no trend lines):

Tom Perriello (D-inc): 35

Rob Hurt (R): 58

Jeff Clark (I): 4

Undecided: 3

(MoE: ±4.1%)

Whoa — those are some dreadful numbers for Tom Perriello. But let’s first compare this poll to SUSA’s final poll of this race from 2008. In that one, SUSA’s likely voter universe was 40% Democratic and 38% Republican. This time, it’s 42% Republican and 27% Democratic. In 2008, SUSA pegged the electorate as 22% black — this time, just 13%. Furthermore, African-American voters give 27% of their votes to Hurt in this poll, a significantly higher share than the 13% given to Goode. Young voters, too, have completely flipped against Perriello; Hurt racks up a 62-30 lead among 18-to-34 year-olds after Perriello rocked Goode among those voters by a 61-34 margin two years ago.

Back in February, Public Policy Polling released a poll showing Perriello and Hurt tied. That poll, which used a slightly softer likely voter screen, had a 39D-35R sample, was 24% African-American, and had a sub-sample of 18-to-29 year-olds who favored Perriello by a slim 45-42 margin.

Which poll are you buying stock in?

100 thoughts on “VA-05: I Will Make You Hurt”

  1. Christ the crosstabs for this poll sound crazy. Its like Hurt and Perriello are running in a district deep in Utah or something.

  2. PPP has both Democratic and black turnout figures above where they were in ’08, which could be the case in ’12 but not this year. SUSA has weird crosstabs as usual that suggest Perriello may be doing better than the poll suggests. What  I take from this poll, though, is that Hurt has consolidated conservative and Tea Party (over 85% of both). It was a real worry that the Tea Party crowd were going to sit this one out after the semi-moderate Hurt won the primary.

    I would characterise this race as Lean R now and give Hurt a 10 point lead.

  3. There is no getting around it, this absolutely sucks. It is happening to such a great guy as well. I even gave him twenty bucks and I am pretty conservative when it comes to donations. I want to believe PPP but in all reality a lot has changed since Febuary, Hurt just got a lot of exposure coming out of the primary and has higher name id now so the results are not completely unbelievable. Don’t get me wrong I want so badly just to dismiss these results and it could be a bad poll but chances are that a good Congressman is going down because of a bad year. The crosstabs are crazy as hell but it is SUSA so what do you expect? What a shame.  

  4. There is an enthusiasm gap, but that doesn’t equal unenthusiastic Democrats not going to the polls. I do have VA-5 listed high on my list of seats to turn over, but if it does flip, I don’t expect a lopsided result like this.  

  5. … days before the elections we will have 60 house seats with democratic incumbent with -20% or worse in the polls against the republican challengers.

  6. As an actual resident of Virginia’s 5th district, let me just say how unlikely this is. Even in the more rural area of the district where I am located, Perriello has a fairly consistent popularity base, and if you drive a short way into towns like Bedford, its even larger.

    Hurt being up a bit wouldn’t surprise me, but in a district where the GOP is very unenthusiastic about him (and considering I actually reside in his State Senate district, that’s saying something) and the Democrats are unusually revved up about Tom because of his health care votes and such, and with no top-ballot race to drive up turn out, I honestly believe that the 5th is less likely to be flipped back to red than almost any other dangerous district.

  7. Only statewide data available (AFAIK) from ’06 and ’08 CNN exit polls


    16% African American (AA)

    36/39/26 (D/R/I)


    20% AA

    40/33/27 (D/R/I)

    VA-05 has an AA population of 24%. VA as a whole has an AA population of 20% — a proportion differential of 4%


    1) The CNN exit polls are good

    2) AA turnout in VA-05 is in proportion to the whole state

    In that case —

    AA turnout in VA-05 in ’06 was 20%

    AA turnout in VA-05 in ’08 was 24%

    3) If we assume an unenthusiastic D electorate / enthusiastic R electorate relative to ’06/’08 — it leads to:

    Assumption 4) — lower AA turnout than ’06

    Then I can believe say an AA turnout in ’10 of say 16%, which is closer to Survey USA than PPP.

    Furthermore, Survey USA did well in its noted last poll in ’08 in VA-05. So their ’08 numbers for D/R/I probably was close to the actual voting electorate. Thus, based on

    a) the difference between the ’06 and ’08 statewide electorate — means a more R sample in mid-term elections

    b) Assumption 3 means a more R sample in ’10

    It is reasonable to assume some R advantage in likely voters in VA-05. However, I don’t see enough data to justify (or refute) the presumed 27/42/31 (D/R/I) split in that district.

  8. Having gone to school in the 5th and having worked for Perriello, the partisan split SUSA employs here just doesn’t jive with the reality there. Sure, this district is conservative but there is no way the gap will be THAT wide in November. Moreover, given the distaste over half the Republican base has for Hurt (he only received 48% of the vote in the primary, making me wish VA had a run-off system), I have a hard time believing he has already consolidated conservatives in the district to the point where he would win 95% of their vote with Clark on the ballot and leave only 3% undecided.

    I think it is much more likely that at this moment, in July, at least 10% of the electorate is undecided overall and that conservatives are also more undecided than this poll would have it appear. While this doesn’t mean they won’t mostly end up voting for Hurt, it also doesn’t preclude Clark causing him trouble.

    Additionally, there is zero chance that 27% of African American voters in that district are going to vote for Hurt. And don’t sleep on the Charlottesville turnout for Perriello as folks there are absolutely thrilled with his key votes so far. One can look at his fundraising success so far and see how he’s resonating with his constituents and with the national Democratic base.

    (crossing my fingers for Tom)

  9. According to the larger press release, even if this poll were re-weighted, Hurt would have an 11 point lead. Could this be an outlier? Sure. I doubt Hurt is leading by this much.

    Still, I think Perriello is in real trouble.  

  10. You know a poll really stinks when the person who is the recipient of the poll releases a statement that says the poll stinks.

  11. Polling showing outcomes favoring Ds or Rs this early are useless. We constantly see weird weighting and crosstabs producing even weirder results. Some firms seem to have an agenda to push a narrative, while others just seem to have issues, and rightfully so, in figuring out the configuration of the electorate over 3 months out. The only people who are seeing trends and where races stand currently are the campaigns who are on the ground, and even they don’t know how they will stand by election night.

    Perriello has every reason to be ranked at high risk of losing his race, but there is no reason to think he won’t break 40%; as he is an incumbent and more than credible in terms of electability. In this environment he probably should lose, but we won’t know until election night who gets their voters to the polls and how the independents will break.

  12. SUSA even mentions in there news brief that if the partisan split is even Hurt still wins by 11. Dissecting information from the cross tabs is risky too as the margin of error on those is huge. I remember the SUSA polls of the VA-Gov race had weird break downs in demographic cross tabs. In the end though their final margin was only 2 points off. SUSA has been ranked has one of the best, if not the best in Nate Silver’s pollster ratings so I’m inclined to believe them. Without a recent poll to compare this to we can’t know if this is an outlier or not yet.

    Based on evidence outside of polling I’d consider a 10 point gap more likely this November.  

    The PPP poll is almost 6 months old. If the information was unpleasant anyone could instantly disregard it based on its age. So I’m buying SUSA stock.


  13. because it’s summer, all those college students at the University of Virginia are not there, and thus not polled (or severely underpolled)?  That would seem to explain why Hurt is doing so well with young voters, if you only look at young people who grew up there, versus those who are going there for school.

    That could also then explain the numbers SurveyUSA had in this district in August 2008, when school was similarly out, and how different the numbers became by the week before the election.

  14. One thing about VA though is that local races and congressional races is all thats on the ballot. No Senate race, no Gov race, so down ballot statewide office, no legislative races.  

  15. I actually disagree with that. While the 5th isn’t too favorable towards Obama these days or the Democratic Party, with the likely low turnout (lack of a Senate or Prez race at the top of the ballot) and the fact that Democrats within the district are largely MUCH more passionate about Perriello (one of the few conservadems to vote for the public option, final health care bill, stimulus, and cap-and-trade) than Republicans are Bob Hurt, and if there is an enthusiasm gap, chances are its more the GOP than the Democrats, particularly when you look at the fact that 60% of primary voters voted against Hurt to be their nominee.

    I think it’ll be close, but I’d rank it a lot lower on my turnover list than you implied.  

  16. SurveyUSA is specializing in predict big democratic failures. Very curious.

    If you predict 20 failures and you get right in one… Well, is not very difficult…

  17. SUSA is good but has outliers like eveyrone else.

    I got depressed when I first saw and posted here about this VA-05 SUSA poll, but now I feel calmer about it.  The Perriello campaign pushed back on it instead of staying ominously quiet, and ultimately the House race ratings by Stu Rothenberg and others don’t support Hurt up by 23.  These campaign reporters like Rothenberg and Cook are privy to private polling and other information we don’t get at SSP, and they base their ratings on that stuff, not on a SUSA poll.

    Rothenberg has this race as tossup/tilt GOP, which to me translates to a low-to-mid-single digit Hurt lead in both sides’ polling.  And Rothenberg isn’t shy about dinging incumbents in his ratings, he’s got Cao’s seat as “Dem favored” and even Djou’s seat, which many at SSP worry about, as “tossup/tilt Dem.”

  18. I’m as unhappy with this poll as anyone else, but just because a poll produces unfriendly results doesn’t mean the pollster is biased. (And if SUSA is Republican, why did FDL choose to use them?) Did you think SUSA was a Republican pollster when they showed Obama up by double digits in Virginia in October 2008?

    SUSA has a tendency to produce quite a few outliers in the summer, but their numbers become reliable towards the end, and they’re nowhere as bad as Rasmussen.

  19. SUSA has a tendency to produce quite a few outliers in the summer, but their numbers become reliable towards the end, and they’re nowhere as bad as Rasmussen.

    And you think this is casual? (I think not) Why more outliers in the summer? (Because they are not elections for see the true results) Why more outliers in the underpolled races? (Because they are not enough polls for make your outlier as a clear outlier)

    Casual outliers are randomly distributed by the time, and favore booth sides, not only to your side always.

    You are giving keys what mean this are not random outliers. Then, is that really ridiculous? I think not.

  20. perfectly fine explanations for outliers, especially given that SUSA uses IVR. In the summer people are less engaged and low-info voters are probably less likely to respond to polls, and it’s much easier to hang up on a recording than an actual person.

    Attacking every poll you don’t like as “Republican” just makes you look desperate. I would be more inclined to agree with you if this were an openly Republican pollster like Magellan or a closet Republican pollster like Rasmussen, whose shenanigans are not limited to outlier polls.

  21. what Hurt can have the same results than Goode when he was an incumbent (now is a defeated incumbent).

    Perriello should make public his own internal poll in the next weeks. And more democratic candidates should do the same.

  22. there will be a 9 point drop in Democratic turnout in that district. Particularly not with Perriello’s vibrant campaign and how loved he is by a pretty surprising mix and spectrum of Democrats, from Charlottesville college liberals, to rural populist Dems.  

  23. although either way this poll smelled of outlier. If SUSA is still showing a significant Hurt lead in October, then I’ll really be worried.

  24. That are only comments what have not influence about have always the outliers going to the republican side.

    But im not desperate, and you can believe me, I know something about numbers, and this is not right.

  25. biased towards Republicans because these are the hyper-engaged voters who are willing to sit through a robotic voice talking to them on the phone.

    I should also point out that SUSA doesn’t weight its numbers, which probably tilts it away from low-info voters even more.

  26. 27 is 75% of 36, so the sample projects Dem turnout 25% less than 2006.

    And it projects nearly a full 1/3 of Democrats won’t vote compared to 2008… 27 is 67.5% of 40.

    In contrast, African American turnout in Virginia was 20% in 2008, and 16% in 2009… which is only a 20% drop from 08 to 09.

    The sample is poppycock.

  27. Hurt’s probably no more than 5 points ahead right now, and he’s simply not the campaigner Perriello is.  

  28. An 11-point spread would still be depressing, but the more I think about it, the more this SUSA poll doesn’t pass the sniff test.  An incumbent Congressman anywhere isn’t down 23 points in July without some kind of personal or criminal scandal.  If this district were that conservative, Perriello never could’ve won in the first place.

    The biggest flaw I see in the SUSA poll is the 27% Dem and 31% indy vote shares.  Indies made up about 20% in every SUSA poll in 2008, and Dems in the 30-40 range, increasing with each poll.  There is nothing else on the ballot in Virginia this year to drive turnout, and in the first place partisans are more motivated to vote in midterms anywhere than nonpartisans, so it makes no sense that indies would be that much higher a percentage of this November’s vote than in 2008.

    But, as the article noted, reweighting for a more credible spread still puts Hurt comfortably ahead, and there’s no prize for losing narrowly.

  29. I don’t think it’s statistically or otherwise valid to simply split the difference between PPP or SUSA.

    My fallback on this is simply to trust what the campaign media are saying, mostly Stu Rothenberg.  The campaign media are fed private polling and other inside information on the state of the House and Senate races, and they base their ratings on this valuable info we SSPers never get to see.

    And for VA-05, Rothenberg rates this as tossup/tilt GOP.  That translates to me as Hurt leading by low-to-mid-single digits in both sides’ private polling.

    I rely on Rothenberg and not Charlie Cook because Cook has a stupid policy of refusing to rate any unindicted incumbent running for reelection any worse than tossup.  So he’ll rate a sure goner the same as an incumbent with a 50-50 shot of surviving.

    I don’t know what CQ or others do.

  30. has Blanche Lincoln’s race rated Lean Republican and they rated Reid’s race Lean Republican at one point before the GOP found a nutty Angle down there.  

  31. But I’m not sure. Even if Rothenberg is getting internal polling, how can he possible evaluate it? There is no downside risk to feeding Rothenberg lousy numbers. If you lose, you won’t be worth Rothenberg’s time. If you win by a lesser margin, the changes in political cycles/opponent quality make it meaningless. There are just no reputation consequences.

    And Rothenberg has done worse than Sabato and Cook, so maybe he should stop listening to campaigns internals.

  32. If you’re saying pundit race ratings treat incumbents as less endangered than public polling suggests in real time, fair enough, and maybe even Rothenberg does that to some extent.

    But even if Rothenberg’s rating is off in that way, VA-05 would go from tossup/tilt R to lean R, which still isn’t a 23-point spread.

    In all seriousness, the more I think about it, the more doubts I have about this poll.  It just doesn’t make sense that Perriello as an incumbent would be down 23 right now, even in this district.  If a poll said he’s down 10, I could believe that…it would still depress me, but it wouldn’t be crazy.  But an incumbent Congressman needs a lot worse than a bad environment and a tough district to be down 23 at this stage, usually it would take an official or personal scandal.

  33. No reason his career needs to end.  I’m sur ehe can find a non-lobbying job for a few years and run for Gov in 2013.  He’d have to be better than McAuliffe or Deeds.

  34. Rothenberg and Cook and others are not just talking to campaigns, he’s talking to party officials on the ground in the states and districts, and to the party committee staff and officials in D.C.  And he’s talking to both sides for every race!

    So it’s not possible to fool these guys, they’ll find out fast if someone is trying to pull the wool over their eyes.

  35. I don’t think SUSA has an agenda or anything a la Ras, but I do think they messed up with this poll.

  36. if Perriello loses, he’ll run for Deeds’ state senate seat next year. It’s looking like Deeds will be retiring at the end of his term. I don’t think he’s really recovered from his crushing loss last year, plus his marriage fell apart around the time of the campaign.

  37. I know Petersen is a young and up and coming state senator from the northern Virginia area who was in the Virginia House of Delegates and ran for Lieutenant Governor in 2005, finishing third.  What do you think about him running for governor?

  38. The marriage part that is.  I’d rather Periello do something more high profile, are there any appointments he might be eligible for.  I know the VA-Gov is a Republican but Periello might be qualified for something in his eyes maybe?

  39. McDonnell has a tendency to pull a Steve Beshear/Barack Obama…that is, appoint politicians from the other side to positions when it opens up seats that are ripe for takeover. If Perriello wins in 2010 and becomes entrenched, McDonnell might want to get him out of the picture so the Republicans can take back his district. Otherwise, I wouldn’t bank on it.

  40. There’s nothing “easy” about evaluating political races.

    I still believe that independent, public polling is the best barometer of political information. Every other source is incentivized to tell a narrative favorable to their side. Sure, maybe there are people who are brutally honest about their side’s chances, but having worked in politics, that’s a slim minority. Even professionals have a tendency to “drink the kool-aid” and/or outright spin people.

    And any political professional will be judged on how well they work for their candidate, not how truthful they are to Charlie Cook and Stu Rothenberg.  

  41. He’s not going anywhere but up.  If he can hold that seat during this cycle he’s got mad skills as a politician and a BRIGHT future.

  42. I jsut wodner though, when they poll VA kids, how do the screen (LV or RV)?  Is it possible students in VA but voting absentee to another state might later on inflate Periello’s numbers with this crowd?

  43. That July ’06 poll included

    29/45/23 (D/R/I)

    19% AAs

    For contrast, there was also a Nov ’06 poll, ref

    32/44/22 (D/R/I)

    17% AAs

    While the current poll’s

    27/42/29 (D/R/I)

    13% AAs

    sound low, it’s within the MoE in all respects.

    Now the ’06 contrast:

    The SUSA November ’06 poll had:

    61/35 (Goode/Weed)

    (the July ’06 poll had 59/35)

    The final result

    59/40 (Goode/Weed)

    So based on ’06 data, SUSA understated the D vote by about 5 points.

    Assuming SUSA uses the same basic methods, assuming a similar electorate to ’06, Perillo may be down by “only” 18.

    HOWEVER, ’10 is not as favorable a year as ’06.

    Conclusion: the SurveyUSA poll is credible.

  44. Numbers don’t tell the whole story, it’s over-reliance to look at the old SUSA polls and conclude that the current one is good.

    I’ve taken a step back since last night based on the simple fact that no incumbent Congressman anywhere is down 23 points in July.  If the district was that conservative, Perriello never could have won last time against a scandal-free conservative Republican incumbent, but he did.

    Take a look at all the House race results for incumbents we lost in 1994 or GOP incumbents we beat in 2006 or 2008.  What was the biggest margin in any of those wave-driven takeaways?  I bet not one challenger sniffed 60%.  I remember Chris Murphy beat Nancy Johnson by 12, 56-44, and that might be as big a margin as there was.

  45. ref

    In mid-October, an opinion poll commissioned by the Evansville Courier & Press showed Ellsworth leading Hostettler, 55% to 32%.[39]

    which is similar to the 58/35 split from SUSA.

    In the November election, Hostettler was soundly defeated, taking 39 percent of the vote to Ellsworth’s 61 percent. His defeat was the first announced that night.[citation needed] The 22-point margin was the largest margin of defeat for an incumbent in the 2006 cycle, and the second-biggest margin of a defeat in a Republican-held district.

    I want to be wrong on this. But this is looking like a a Congressman who is out of step with his district.  

  46. So I looked up 1994, at results regarding defeated Democratic incumbents that year.  The 4 worst beatings incumbent Dems suffered were, in order of margin of defeat, as follows:

    Georgia:  Charlie Norwoord beat Don Johnson 65-35.

    California:  Radanovich beat Richard Lehman 57-40.

    Arizona:  Karan English lost to J.D. Hayworth 55-41, with a Libertarian getting 4%.

    Utah:  Karen Shepard lost with 36% to a GOPer who got 46%, and indy Merrill Cook (later a short-time GOP Congressman himself) got 18%.

    Keep in mind regarding Shepard, there were other Dem incumbents who lost by 10-12 points to a Republican, but all of those Dems were in the mid-40s with no strong 3rd candidate absorbing votes.  I list Shepard because her 36% was the 2nd-worst for an incumbent that year, even though her margin of defeat was only 10 points.

    So there was, indeed, exactly ONE precedent in 1994 to support the SUSA poll on Perriello now, that being Don Johnson’s loss in GA-10.  Johnson was a freshman like Perriello now, and had won the seat 54-46 when it was open due to a sitting Democrat’s retirement.  Johnson’s actual vote totals actually dropped from over 106K in 1992 to just 51K in 1994, a drop of 55K, while the total turnout in the district dropped by only a slightly larger 59K.

    So SUSA is saying Perriello is getting Johnsonized.

  47. benefited from a friendlier district than Perriello, Perriello is just so much better than Hos as a candidate. And I would even go so far to say that Ellsworth is better than Hurt too.

  48. …was that he didn’t campaign.  He always had been a poor fundraiser, but that last time he was just tired out and didn’t really want to run again, and in fact he really didn’t.  It was bad enough that I remember having wondered in summer 2006, when I read on the Howey Report blog how non-existent a campaign Hostetler had, why Hostetler didn’t just retire so that some other Republican could run a real campaign.

    Ellsworth would have won anyway, but the 61-39 blowout happened because Hostetler ran the kind of campaign you expect from a local crank nominated as a sacrificial lamb against a safe incumbent.

    That said, see my follow-up comment about Don Johnson in 1994.  That might be a better example.

  49. So the HCR vote doesn’t hurt him in the district?  That’s contrary to everything us outsiders have been hearing.  Good news if true.

  50. it’s possible Perriello will run for governor in 2013. If he holds on to the 5th this November, he’ll decide not to run for re-election in 2012 because the district may be even more conservative after redistricting (though admittedly, Obama’s name will be on the ballot again which helped him the last time). That would set him up nicely to run a gubernatorial campaign for 2013.

  51. assuming a quality campaign.

    Nevertheless, Perrillo is swimming against the tide, given the year and the R+5 nature of the district. Hos had the R lean of the district in his favor.

    Thus I don’t see him making up all of the deficit unless Hurt pulls a Coakley. (But I assume that the SUSA poll numbers are credible, and I see others think not.)

    OTOH, w/r/t Don Johnson, were there mid-decade redistricting issues? I think there was some ’90s VRA related stuff in the area. I don’t know if it had a significant effect on Johnson’s ’94 district.

  52. About voting for “unpopular bills” (unpopular in quotes bc I’m quite convinced the US still has no clue what hcr is) revs up your own base and will get them out for you.

    Perriello is probably one of the few who can count on a much more normal turn-out model (I’d think.). Also why idiots like Rep Arcuri shot themselves in the foot bc they voted yes on the more liberal hcr but no on the final and more conservative bill.  He just went and pissed off everyone.

  53. Is because of Republican tea-baggers moving to Independent because those gosh darn Republicans are just too liberal.

  54. I don’t buy the 23-point SUSA margin, and I distinguish it from IN-08 in 2008 because Perriello is, in fact, running an aggressive campaign.  He’s had TV ads airing for awhile already, he had $1.7 million banked in his last finance report, and he’s an acknowledged to run a fantastic field campaign, in 2008 and again now.

    That doesn’t mean he might not be down now, and in reality I could believe he’d lose by 10 if the election were held today.  I don’t doubt that the SUSA poll isn’t off BY 23 points, meaning Hurt must have some sort of lead.

    But I don’t buy it that Hurt is up 23.

    And the Don Johnson example in 1994 was the exception that proves the rule, which was that most incumbents who lose still are competitive.  Regarding redistricting stuff, I looked that up, and yes redistricting that was later struck down had created a couple extra majority-black districts that hurt white Dem incumbents, but Johnson had won the seat under the same lines when it was open in 1992.  So Johnson can’t blame redistricting on his blowout margin of defeat.

  55. I’m on my blackberry so I can’t look up the numbers but Obama came within at least 2-points.  And Goode was the first GOPer to win this district (since reconstruction?  Blah I shouldn’t do anaylsis on a blackberry) so it’s not like it’s iron clad against us.

    Not to mention, a candidate with opposing viewpoints than the district can always win with some hardwork and a great campagin strategy

  56. I was actually digging around for campaign finance data for that Johnson-Norwood race to see if Johnson perhaps had fundraising problems, and in Googling for that I kept getting goddam Miami Vice links!

  57. If anything is misleading, it’s that Obama came close to winning here.  “Normal” here is a more comfortable win by the Republican presidential nominee.  Obama combined a manufactured spike in black turnout with a strong anti-Republican environment to come close here, and that’s not something Democrats can replicate hardly ever.

    And Goode was elected to the district in the 90s as a very conservative Democrat, and then switched first to independent and later formally to a full-fledged Republican, all while holding the seat.  The district itself was a traditional conservaDem Southern district, one of many that inevitably would flip to the Republicans in the post-Civil Rights era.

    Really, the R+5 PVI is a fair characterization of what a Democratic Congressional candidate should expect in a typical year.

  58. If the electorate is more Democratic than shown in SurveyUSA’s model here, the Republican still wins, though his margin of victory is less. For example, if hypothetically there were an even number of Democrats and Republicans in the likely voter model, Republican challenger Hurt wins today by 11 points, not 23.

  59. sleep-deprived stupor I am missing something, but I don’t see how that quote proves that he weights by party ID. It sounds more like he just took the party crosstabs and crunched the numbers to see what it would look like with different party ID numbers yet the same result within each party. Whereas weighting means you have an idea of what the district’s party or demographic breakdown looks like, and you adjust your numbers based on a certain number of each group. To the best of my understanding, SUSA simply polls at random and lets the chips fall where they may, aside from their likely voter screen.

    The reason I am inclined to believe this is because I pretty clearly remember this as an explanation (by Nate Silver, maybe?) for their outlier poll showing McCain up 20 in NC in September 08. If my memory is foggy or they have changed methodologies since then, feel free to correct me.

  60. in SUSA’s 3rd-to-last poll of NJ Gov 09, their party breakdown was 41-37-20 D-R-I, in the 2nd to last poll it was 44-34-20, then finally to 44-33-22. In ten days, the spread went from D+4 to D+11? Doesn’t sound like they weighted the party ID numbers.

  61. Or crazy like a fox.  He probably shouldn’t have won his last race but did.  Maybe voting for something and against plays well somehow/some way to his constiituents.

    Who knows?

  62. You at least paint a picture that the pitch fork brigade (anti-hcr) is either not as big as we’d heard or not as angry.

  63. that Perriello will likely benefit more from his unpopular votes, because he has made really hard votes for somebody in an R+5 district, and the base voters (particularly the large black populations in areas such as Franklin and Bedford Counties) will turn out much more than they typically would in a midterm with no Senate candidate, and on the opposite scale, Robert Hurt’s voting record in the State Senate has largely suppress enthusiasm by base voters contrary to national trends, and I think it paints a much more favorable picture of the race than some believe.

    You do have to remember, Virginia is one of the few states this year that has neither a Gubernatorial or Senatorial race, or actually any other major election in November, and because of that, national trends will not play as big because typically, if its just House Races, its all about turning out the base.

  64. They focus on the non-weighted result, but they do provide the weighted result, which is 11 pts.  

  65. It was a strong Democratic year, Arcuri had won an open seat 2 years earlier, and he should have won handily in 2008 against a weak opponent.  But he almost blew it, and no one saw that coming.

    His health care votes really are stupid.  You vote no on both, or yes on the 2nd regardless of your vote on the 1st…there is political logic to any of those choices.  You can establish yourself as either firmly opposed, or firmly favoring, with any of those.

    But there’s no political logic behind Arcuri’s combo of yes on the 1st and no on the 2nd.  You can take the Stephen Lynch approach and say it’s not liberal enough, but that applies only in a true blue district and has no political logic in a district like Arcuri’s.

  66. Keep in mind 42% Republican already is a little high, but only a little, and it’s very plausible in this anti-Democratic environment.  But that high self-identifying GOP vote share makes it impossible that some previous GOPers are calling themselves indies now.

    And the independents break for Hurt by 11, which is actually about the same as all the 2008 SUSA VA-05 polls, even the one on election eve showing Perriello surging and down just 3–that one had indies still favoring Goode by 9 points.  Really, all the crosstabs are roughly comparable to that last 2008 poll!  They’ve got Hurt with a little more GOP unity, 90-7 compared to 84-14 for Goode, and Perriello with a little more Dem bleeding, at 79-17 compared to 83-14 back then.  And the indy split is now 11 points rather than 9.  But those differences are small and alone would still leave Perriello down only in the mid-to-high single digits under the turnout model of that final 2008 poll.

    What’s really driving the margin in this poll is the disastrous turnout model.  There is almost no place in America today where a Democrat can win where Democratic turnout lags GOP turnout by 15 points in vote share.  We’ve pulled it off recently in upstate New York in a few districts, but that’s the last bastion of Republicans who’ll readily defect.  Suburban Philly has been such a place, but a lot of those voters are flipping their party registration formally anyway in recent years.

  67. It might turn off indies, but Arcuri might actually have a logic–Dems in Ithaca might very well think the bill is too conservative.  Most are probably biding their time, like me, for a better option, but some might be more pissed.  I’m not sure, but if you want to look at “crazy liberals” in marginal districts, didn’t Grayson vote no on the 2nd bill?  Also, Kucinich’s district is surprisingly moderate at D+8, whereas one might expect him to be in a D+25 or so district.

  68. For the exact reasons you listed, he should have lost.  

    I don’t beleive in the candidates who ride waves and take advantage of weak opponents or gaffes.  I’m a big fan of candidates and campaigns mattering.  So he definitely should have lost, and in a non-wave year he would have.

  69. It doesn’t matter what the “real” numbers are.  This poll is poppycock, and it should both embarrass and make one skeptical of other SUSA products.

    That doesn’t mean their other polls will be as ludicrous, but it does mean they released something this comically impossible.

  70. Just because your totals come out close to the end result doesn’t make the poll any better.  I know that sounds crazy but think about it.

    For example, if you’re an accountant and you get the right totals by adding up all the wrong numbers, is your accounting good?  

    If you do Walmart’s corporate budget and you hit your totals but are way off on the underlying numbers, are you good at budgeting?

    At the end of the day SUSA’s poll of VA-GOV could just as easily be coincidence as it was accurate.  

  71. There is –almost always– larger random variation in subsamples.

    In fact, the lack of random variation was part of the problem with R2K, at least per the data cited by Nate.

    Polling problems come from all subsamples going one way. And that happens one time out of 20, thus the “95% confidence margin” that is a proviso with every pol.  

  72. However if your crosstabs are all wrong you can still get the right answer.  Even if your poll is flawed.  If the subsamples net out to the rigth total, the poll is considered valid, even if the underlying data makes no sense.    

    At the end of the day since the majority of races for Congressional seat are no more than 20 points between candidates, its really hard for a poll to ever be off more than 10%.  Just based on numbers and common sense, in any competitive race if you keep them within 10% you have a close chance of getting with a 5% MOE.

    I personally am just out and out sick of polls, crosstabs and stuff.  Its like people want elections daily.  There’s a reason that we have a govt setup to have elections only every so often.

    Polls are so fallible even if people think the right model is used.  It assumes so many things its beyond insane to trust any poll ever.

  73. If people keep analyzing pollsters trying to disprove their methods, I have a feeling we’ll start considering pollsters as being accurate/reliable like we consider reality TV to be real and not scripted.

    But that’s just my opinion.  I just don’t think accurate polling is physically possible via random phone samplings using screens based on the pollsters concept of an electorate that may not materialize or even exist.

    Elections are awesome for determing who people are going to vote for.

  74. I’m not sure I agree with your analogies. Polling is about testing a race where it is at that moment. It’s almost certain that the gap will close towards election day. In the press brief SUSA admits the results are due to a highly energized Republican base. Once more D’s get engaged the numbers will close (Hurt wins by 11 when its even R-D).

    SUSA has always had weird cross tabs. Which have margins of error in the 15% range. This hasn’t prevented them from being #3 on Nate Silver’s pollster rating list. Clearly they have a long record of high quality, accurate polling.

    Like I said this can’t be dubbed an outlier until we have similar data to compare it to.    

  75. One thing I learned from the 2004 presidential election is that even in a very close race, polling is by and large reliable.  Most polls late in that cycle showed Bush barely ahead of Kerry, by 1-4 points, with a few undecideds.  On our side we all hoped and presumed undecideds would break to Kerry, that an incumbent can’t cut it that close to the bone and still win.  But Bush won by just under 3, and those polls were right.

    So ever since I’ve taken the position that if polls consistently show a particular candidate winning, even if always only by a point or two or three, then that candidate is winning.  And that’s almost always proved right.  The exceptions are few.

    Statistical sampling is a long proven science.  It’s not perfect, there are outliers, and the science itself admits the outliers.

    And setting that aside, I find the first sentence of your last comment very confusing and perhaps contradictory.  If people keep trying to discredit pollsters, then why does that make pollsters look more accurate and reliable?

  76. Is like polling people and excluding a candidate.  There’s just no way you can have a model that surveys people more skewed towards being active and then try to move that sample around to get what you think it will be and have it be accurate.

    Also, how exactly can you say it polls the race where it is today?  It doesn’t do that at all.  Nor do people close gaps in polls.  They narrow as the election closes because the earlier polls were wrong.  There’s no spinning it.  

  77. 600 out of say 200,000 registered voters per poll means that very few people are sampled more than once.

    Furthermore, I see no “adjustment” in SUSA figures as it sounds like you’re suggesting.

    As for where the race is today, that’s addressed by the phrase: “if the election were held today”

    Opinions do change over time, as they are influenced by campaigns. So the polls reflect that. No big deal.

    But if I’m hearing the totality of what you’re saying correctly, you do have a point — how can people not be influenced by a constant stream of polls hitting the front pages of a newspaper/local TV news? It’s sort of like expecting a pure jury pool where the accusation has been discussed ad nauseum.

    Nevertheless, most voters aren’t “engaged,” i.e. thinking more than a few seconds about whom they will vote for, until closer to election day.

    Thus the influence of the ad nauseum polls is fairly small, except perhaps in the post-Labor Day period in which most voters are engaged.

  78. Candidates and campaigns and current events decide elections, and those things overwhelm and smother polls and other things in the news cycle.

    There’s a lot of clutter in the 24-hour news cycle, and polls are among the clutter, not a driver of anything.  Only people like us on this site are focused on them.  People tune out the clutter because they have to, they don’t have the mental or emotional bandwidth to absorb it.

    I actually think voting itself has a remarkable psychology in how independent every voter really is.  People really do tune out their surroundings for the most part and vote based on their own personal values and priorities, no matter the noise surrounding a particular election.

    The only sense in which voters can be influenced by the noise is in the marginal turnout of infrequent voters.  Believing a race won’t be remotely close can get politically disinterested voters to stay home.  There are tens of millions of such people, so their import is not to be belittled.

    But for people who actually consistently vote, which in America is roughly 40% of adult U.S. citizens, they don’t get distracted very easily.

  79. Marginal turnout is everything in US elections — the races are close enough in a sufficient number of districts — to make the difference in control of both houses of Congress.

    In addition, there’s the effect on campaigns — good close polls tend to enhance the energy associated with the campaign, and the productivity of campaign workers. And that also influences the electorate (though that’s only a secondary effect of polls).

    But I also think it isn’t such an influence on either before Labor Day. That “paranoia” is left for political junkies.

  80. No doubt polling is one part of what can become a self-reinforcing loop of momentum, either positive or negative.

    But polling is never a trigger for that momentum.  Candidates and campaigns and current events trigger all momentum, and that momentum is what decides outcomes.

    And “marginal turnout” really is overstated, it doesn’t matter that much.  It makes a difference only in the very closest of races.  And definition, the closest of races are close enough in polling so as not to have turnout problems for either party.

    When I was doing research on defeated incumbents in 1994 earlier this week for purposes of another comment on SSP, I was startled to be reminded how NOT close some of those margins of defeat really were.  There were a LOT of incumbents defeated by solid margins, more than I would have guessed.  And those were incumbents, on average they’re going to lose by smaller margins than Democratic nominees in open Democratic-held seats.

    Of course, where turnout does matter is with coattails.  I witnessed in my own Virginia House district how coattails virtually alone defeated my Democratic Delegate, dragging down turnout in all but one of the Democratic precincts while turnout edged up in all but one of the Repbublican precincts.  Creigh Deeds cost us some Virginia House seats we didn’t have to lose.

    So I’ll give you that polling “matters” in that regard.

    But the problem is even absent all the polling, people tend to be aware when an election is very one-sided.  Was there any doubt in 1964 that LBJ was on his way to crushing Goldwater?  Was there any doubt in 1932 that FDR was going to spank Hoover?

  81. Like GOPvoter tell they give booth numbers, but they are emphasizing the outlier 23%.

    I think they mean the people know and see before this 23% against Perriello, while they want to defend from the critics giving too the this 11%, what is like another different poll (with a different composition of the sample). If you see the poll without give more attention you will get only with this 23%.

    Is obvious they are playing with the samples. The GOPvoter quote and your comment about New Jersey prove it.

    You must think about the reasons for reduce the republican weight in the polls for a rece, just for the elections.

  82. “Is obvious they are playing with the samples.”

    Do you mean that they aren’t attempting to estimate in good faith who is going to vote? Is there any evidence that SUSA is not acting in good faith?

  83. I don’t exactly understand what you’re saying but from what it sounds like you’re saying, you’re being completely unfair. SUSA’s policy is and has always been not to weight for party ID. The 11% versus 23% thing was saying, well, IF we weighted for party ID and decided that the % of Democrats will be the same as the % of Republicans, it’s 11 points…but we DON’T weight for that stuff, so it’s 23%. It’s really bizarre that you’re trying to paint this as some SUSA pro-Republican bias when they never weight for party ID numbers.

    My New Jersey example doesn’t prove that they’re playing with the samples…it proves that they don’t weight for party ID. If you don’t weight by party ID then obviously the party ID numbers will change from poll to poll…that’s one of the vagaries of polling, that if you’re polling 500 voters out of several hundred thousand or million, then you might get a slightly different sample each time. This is why polls have MoEs, and it’s why polls that aren’t weighted for party ID can tend to switch around a bit.

    Besides, if SUSA were really a Republican pollster why would they release a poll showing the Republican governor of their own state with a -29 approval rating?

    Let’s leave the conspiracy theories about pollsters who give unfriendly results to the Republicans.

  84. But I think the succession of samples what explain Sapelcovits help not believing in SurveyUSA’s “good faith”, and the explanation what they give in the release about the 23% and the 11% goes by the same way. We must think in the reasons for explain the fast pre-election evolution of the sample. They are not logical political reasons for explain this change between September and November. Only we can think what they assume the previous mistake of take an excesively pro-republican sample, and if they make the same more times, you can think bad about this procedure what would explain well mathematically the pro-republican outliers in the summer what are habitual in SurveyUSA, like Sapelcovits comment.

    Is not casual what their outliers goes always by the same way helping to the republican side. These are not statistically random outliers. The random outliers what every pollster can have go in favor of the two sides with a random distribution. And more, I can tell without very much fear of be wrong, what SurveyUSA has an excesive percentage of outlier results for be random outliers. In their last 10 polls they can have 3 or 4 pro-republican outliers.

    Well, maybe sometimes I use not the right technical words for explain this in english. This is because I learn mathematical statistic in the university in my native country (the Basque Country), in other language (spanish), as part of my technical formation.

  85. But I think you have a much higher burden of proof in accusing a pollster of acting in bad faith. SUSA believes more Republicans will vote this year than usual. We won’t know whether they’re right or wrong until the elections are over and the votes are counted. Just because you for some reason disagree with SUSA doesn’t mean they are acting in bad faith.

  86. I am not defending this poll per se, I’m just sick of people automatically assuming that a poll with a Republican house effect means that the pollster is pro-Republican.

Comments are closed.