home

How To Look At Polls

By Big Tent Democrat

Speaking for me only

Having derided polls (mostly because they over claim their value, particularly when omitting how fragile their likely voter models are) for most of my time blogging, I find my intense interest in them in this election contest rather perverse. But I think there is a good reason for my interest. The polls have produced data worth considering when taken in conjunction with the exit polls. The reason is that demographics have completely driven this contest. But what can we make of polls that show such wide divergences between each other? I think a lot. I'll explain why on the flip.

Unlike most any contest I have seen, in this year's Dem contest, if we know the demographics we pretty much know the result. This is especially so in the big contested states. Demography has been political destiny.

So when we look at the SUSA PA poll, which has Clinton up 18, and compare it to the just released PPP poll, which has Clinton now up 3 (previously PPP had Obama up 2), I think we can see why and make judgments on the data presented.

PPP writes:

Clinton has a 46-43 advantage in the state after trailing Obama by two points in PPP’s poll last week. Clinton’s improvement since last week comes almost entirely from her core demographics. Her lead among women improved from 10 points to 16, her lead among white voters went from 11 to 17, and her lead with senior citizens showed a gain from 16 to 21.

By contrast, SUSA has Clinton winning whites by 61-32, seniors by 65-29 and women by 61-33. SUSA expects whites to be 82% of the electorate (PPP does not tell us what it expects the demographic breakdown to be) and women to be 58% of the electorate (again PPP is silent on this.) The undecided in SUSA is 6 points. In PPP, it is 11. That is not the explanation for these differences. While PPP does not tell us, I suspect they have Obama doing much better with A-As than does SUSA which has Obama winning "only" 75% of the black vote.

We can project PPP's findings (I will assume an 85-15 split in the A-A vote) using the SUSA turnout model. Let's assume, using PPP's numbers, whites go 59-41 for Clinton X 0.82(SUSA's assumed white percentage of the turnout), that is 48-4%, plus 0.15 (I assume Obama winning A-As by 85-15) X 0.14 (SUSA's assumed A-A percentage of the turnout), that's 2 percent, plus 2 percent of the remaining 4 percent of the vote.

That means the PPP result using SUSA voter turnout models will show a Clinton 5 point win in PA, 52.5 to 47.5.

SUSA, contrary to PPP's prediction of Obama capturing 41% of the white vote, predicts Obama capturing about a third of the white vote. If we extrapolate the undecided whites we have Clinton winning 66% to Obama's 34%. We do the math as described above and that is 54%. I adjust Clinton's percentage of the A-A vote down to 15% from SUSA's 24%, and add another 2% to Clinton and 2% of the remaining 4% and that gives me Clinton 58%, Obama 42%.

So who will be right? I do not know, but I think PA is more like Ohio than Missouri. And PA is certainly not Wisconsin. It seems impossible to me to see how Obama can beat Clinton in Pennsylvania. The question is by how much. I think Obama's best case is a 5 point loss. Worst case? As much as 20. Demography is destiny in Pennsylvania.

< Obama's Day With His Wealthier Supporters, Changes Tune on Public Financing | New Obama Ad Spending Numbers Out >
  • The Online Magazine with Liberal coverage of crime-related political and injustice news

  • Contribute To TalkLeft


  • Display: Sort:
    And of course, demographics do NOT (5.00 / 1) (#1)
    by MarkL on Wed Apr 09, 2008 at 12:52:43 PM EST
    imply racism. John Kerry did terribly among white males, but last I checked, he was not black.

    Of course, since he found out (5.00 / 1) (#49)
    by Cream City on Wed Apr 09, 2008 at 03:22:33 PM EST
    that he had Jewish heritage, there is anti-Semism. . . .

    Fyi, Kerry and Gore pulled about the same with whites overall at about 41-42% -- and much better than Obama has been doing, at about 35%, last I saw.  So the gender difference you cite is interesting -- Gore must have done a lot better with white women than either Kerry or Obama.

    And, of course, white women are by far the largest voter bloc and Dem voter bloc by race and gender.  

    Parent

    Heh. (none / 0) (#3)
    by madamab on Wed Apr 09, 2008 at 12:57:49 PM EST
    Indeed. ;-)

    Parent
    Actually (none / 0) (#4)
    by BlacknBlue on Wed Apr 09, 2008 at 12:59:52 PM EST
    Democrats as a whole do poorly with whites because of they're associations with Blacks and other minorities. You may recall that the South was solidly Dem until the passage of civil rights legislation.

    Parent
    I think that's wrong (5.00 / 1) (#36)
    by badger on Wed Apr 09, 2008 at 02:29:37 PM EST
    LBJ won his landslide victory in 1964 after the Civil Rights Act was passed, and I can't imagine too many voters who voted at that time weren't aware of it.

    Additionally, go back a look up his losing margins in the southern states in that election. Some - MS, AL - are fairly wide (though he still hit nearly 40% or better), and some are fairly close. And those people all knew the Civil Rights Act passed, because all of the segregated facilities they lived with were disappearing.

    Why the Democrats do poorly is a lot more complicated question than simply race or minorities. I think a real analysis would show that the other problems Democrats have had the last 40 years have actually made race more of a problem - it wouldn't have had the effect it's had without the other stupid things Dems have done.


    Parent

    Voting behavior is not as simple (none / 0) (#59)
    by kenosharick on Wed Apr 09, 2008 at 04:47:44 PM EST
    as you imply. The entire country was not voting on a single issue. Besides, LBJ had an opponent that scared the crap out of most people.

    Parent
    That would be my point (none / 0) (#62)
    by badger on Wed Apr 09, 2008 at 05:19:32 PM EST
    If, as the post I was responding to implied, race were the single issue that has caused the Democratic Party to lose elections in the past 40 years or so, then it certainly would have shown up in the 1964 election. It didn't, except in parts of the south.

    Certainly Goldwater wasn't an attractive candidate, but making the same point, he did vote against the 1964 Civil Rights Act. If there was a huge white demographic that voted solely based on the Democrats' association with blacks, they would have had a candidate in Goldwater. They didn't, except in parts of the south.

    I agree that votes rarely come down to a single issue for most voters, and certainly not nationwide.


    Parent

    I suggest (none / 0) (#65)
    by swiss473 on Thu Apr 10, 2008 at 12:39:06 AM EST
    I suggest you look at the longer trends than simply an election that came months after the Civil Rights Act was passed in 1964 and before the Voting Rights Acts in 1965.

    Also, after Kennedy was shot there was no way Johnson was losing in 1964.  The sympathy vote and just the desire to not reward Kennedy's killer(s) by changing power in the WH was a very powerful effect.  Throw in a "extreme" candidate like Goldwater and Johnson's very effective Daisy ad and him winning wasn't a big surprise.

    But look at what happened AFTER that election.

    In 1966, after both the Civil Rights and Voting Rights Acts were passed, the Democratic Party lost 48 seats in the house.

    In 1968, the dems went from LBJ getting 61% to Humphrey getting 43% and would plummet even lower with 37% or so for McGovern in 1972.

    After dominating the White House from 1932 by winning 7 of 9 elections, the Democrats have lost 7 of the last 10 elections.  More immediately they lost 5 of 6 from 1968-1988.

    The Democrats share of the white vote, both male and female, nationwide has cratered  since then.  Bush won beat Kerry 58-41 among whites.  All democrats, except for Carter in '76 have pretty much polled around 35-42% of the white vote since 1964.

    Their share of the white vote since then has sunk.  In the South, outside of decent success when Southern Governors Carter and Clinton ran, Humphrey, McGovern, Mondale, Dukakis, Gore and Kerry combined won a grand total of 0 states in the South.  That's 6 candidates spread over 40 years and they won a combined 0 states and you're trying to say that there was no effect?

    Johnson said after passage that he'd thrown away the South(and much of the white vote) for the next generation and the facts have certainly proved him right.  Now, I'm not saying that race is the sole factor that has caused Democrtas to lose 7 of 10 elections since 1964 after dominating the WH the previous 36 years, but the evidence and the numbers certainly show that it's definitely one of the bigger factors, if not the biggest.

    Parent

    True (none / 0) (#26)
    by ruffian on Wed Apr 09, 2008 at 01:21:48 PM EST
    and a point overlooked in a lot of the analysis of this election, even though it has been accepted as conventional wisdom in other elections.  I'm not sure I can say this the right way, but one could make the point that Obama is at no more of a disadvantage in the GE with people who think that way than a white Dem would have been anyway.

    Parent
    Chuckle (none / 0) (#7)
    by Faust on Wed Apr 09, 2008 at 01:04:47 PM EST
    First, an excellent analysis. (5.00 / 1) (#2)
    by madamab on Wed Apr 09, 2008 at 12:57:27 PM EST
    Second, I feel that Obama's own tactics have come back to bite him.

    Had he actually run a post-racial, unifying campaign, he would have gotten a lot more of the white vote. As I pointed out earlier, he could have used progressive issues to convince white men to vote for him (like Montana Gov. Brian Schweitzer did), and he could have been more respectful to women in general and to Senator Clinton in specific, and not alienated women so much.

    But instead, he felt he had to race-bait and divide the electorate to take the AA vote away from HRC. He did it all right, but at what cost?

    It ain't over till they vote, but I'm not seeing a lot of "Hope" for Obama in the GE at this point.

    Chivalry is not dead in Hawaii (5.00 / 1) (#51)
    by Cream City on Wed Apr 09, 2008 at 03:28:20 PM EST
    thanks to Donald.  I love this post -- that you started out leaning Clinton because of your mom, because of chivalry, but then Clinton cinched your vote with her brains, experience, and hard work.

    Btw, your story of your mom makes me think of my ex-mom-in-law, a remarkable centenarian today who was widowed in WWII with an infant son -- yet even decades ago, and despite all she endured as well in the workplace, she managed to make it and buy a home of their own as a single mom in the '50s.  And then she had to send her only child to the Viet Nam War.  

    Fortunately for her and me and our kids, that man in her life made it back.  I am so sorry about your father.

    And if you weren't married, I might declare undying love for the guy who wrote this.  As I'm married as well, you'll have to settle for my respect for the guy who must be raising fine granddaughters for your mom to enjoy.  It takes a village of good men as well as great women to do so.

    Parent

    Aloha to you (none / 0) (#54)
    by madamab on Wed Apr 09, 2008 at 03:49:38 PM EST
    and thanks for your kind heart and thoughtful words.

    Parent
    So Obama wins by 19? (5.00 / 1) (#21)
    by Big Tent Democrat on Wed Apr 09, 2008 at 01:18:19 PM EST
    You are wrong on your facts I believe.

    The campus edge is not there in Penn (5.00 / 1) (#52)
    by Cream City on Wed Apr 09, 2008 at 03:38:04 PM EST
    as it was in my Wisconsin for a simple reason:  Wisconsin has same-day registration.  That always makes a major difference here in college towns.  (Not that there are as many large campuses -- there are many, but not as large -- in Wisconsin as in Penn, btw.)

    And Wisconsin has an open primary, so we saw one of those years with (widely encouraged by the GOP) a sizeable crossover, and it went for Obama.

    Why?  Above all, Wisconsin was the primary after Super Tuesday, when the need for a new Clinton campaign manager became clear, as there were not plans for a serious Wisconsin campaign.  Then the primary came only a few days after the upheaval with the replacement of campaign managers for Clinton, so the campaign here began 'way too late.  

    And then it got cut back to almost nothing -- no public event in Milwaukee, ever! -- because of the awful weather this winter here.  Btw, the totals now have been tallied for the winter here, anyway, and it has been the second-snowiest of all time in Milwaukee, officially, although THE snowiest in Madison -- those being the two Dem bases in the state -- and the snowiest ever in many parts of Milwaukee, such as mine.  More than ten feet of snow in total, and much of it was that darn week before the primary.  And don't even talk to me about how cold it was this winter, too.

    Anyway, for many reasons beyond the demographics, I don't think Wisconsin is a good predictor for Pennsylvania.  Here, weather is destiny as well.

    Parent

    willie, read this article ... (none / 0) (#57)
    by cymro on Wed Apr 09, 2008 at 04:25:28 PM EST
    No Really. Hillary Has A Decent Shot.

    ... then see if you still believe that WI is a good predictor of the PA result. This analysis seems to suggest that SUSA is on the right track.

    For further analysis of the same subject, see this:

    In Terms of Geography, Obama Appeals to Academics and Clinton Appeals to Jacksonians

    These two articles contain the most thorough explanations I have read of the relationships among demographics, geography, and the Democratic primary results.

    Parent

    Consistently (5.00 / 1) (#33)
    by Steve M on Wed Apr 09, 2008 at 02:01:45 PM EST
    In state after state, we've seen polls suggesting that Obama would only win 60-70% of the black vote, and he always does far better on Election Day.

    I don't think anyone is necessarily lying to pollsters, but the evidence does suggest a lot of people are saying they're undecided when they really aren't.  If the level of indecision suggested by the poll results were valid, those voters wouldn't break the same way in every single state, time after time.

    Agree (none / 0) (#37)
    by nell on Wed Apr 09, 2008 at 02:34:29 PM EST
    Clinton should expect no more than 10 percent of the African American vote. If, by chance, she should get more, that is great, but she (or we) should not expect that. I mentally adjusted the SUSA numbers to account for him getting 90 percent of African American voters.

    Parent
    True (none / 0) (#58)
    by Ga6thDem on Wed Apr 09, 2008 at 04:37:14 PM EST
    but PA has a different set of circumstances than the other states. Nutter supports Clinton and is actively campaigning for her. And this is all post Wright too which turned off some AA's. I guess we'll just have to wait and see what happens.

    Parent
    Thanks for posting this (4.66 / 3) (#22)
    by Lil on Wed Apr 09, 2008 at 01:18:48 PM EST
    I find it fascinating and have always trusted your analyses. I trust it because I think you try to be fair and not spin it. I appreciate that because sometimes I get dizzy reading so many conflicting views and comments. Whenever you write about polls I sit up a bit straighter.

    What would be interesting (none / 0) (#5)
    by frankly0 on Wed Apr 09, 2008 at 01:00:32 PM EST
    would be to do a projection in PA entirely based on exit polling data, and known election results, using similar states, OH most obviously, as the basis, and fitting them onto the known demographics of PA.

    My guess is that such a projection would be at least as good as any polling would ever be, given the general unreliabilities of polling. The fact is, the projections onto the demographics of PA is probably more solid when looking at previous primary results in like states than it is based on a poll, which can be very unreliable in getting correct results for the targeted demographic.

    I did that (5.00 / 2) (#8)
    by Big Tent Democrat on Wed Apr 09, 2008 at 01:04:49 PM EST
    Check my links. Clinton wins by 22 if Ohio is the model.

    Parent
    Fyi, some support for what SUSA saw (none / 0) (#48)
    by Cream City on Wed Apr 09, 2008 at 03:19:02 PM EST
    from InsiderAdvantage today, a lesser poll but also seeing a trend back to Clinton:

    Latest InsiderAdvantage Poll Pennsylvania Primary: Clinton 48%, Obama 38%

    April 9, 2008 -- A new InsiderAdvantage/Majority Opinion poll shows Sen. Hillary Clinton regaining some ground she'd previously lost to Sen. Barack Obama in the Pennsylvania Democratic presidential primary race.  [It also has undecided at 13% and notes that results when rounded off may not add up to 100%.]

    The survey was conducted April 8 among 681 likely registered voters in the April 22 Pennsylvania Democratic primary. The data have been weighted for age, race and gender with a margin of error of +/-3.6%.

    InsiderAdvantage's Matt Towery: "Sen. Clinton has made progress among both men and among all white voters. Her support among women also appears to be consolidating.  My guess is that whatever damage she might have sustained by recent gaffs and media missteps have been largely discounted by the public. The race in Pennsylvania is clearly still fluid.

    "But, at least for now, it's tending back towards the result that was originally anticipated by most -- a Clinton lead.  Her big task now is to maintain a double-digit lead and expand on it; Obama's is to force her back into a single-digit race. Clinton needs a resounding victory in Pennsylvania to relieve the pressure on her to quit the presidential race," said Towery.

    Parent

    I hope you are right (none / 0) (#12)
    by ajain on Wed Apr 09, 2008 at 01:09:33 PM EST
    But it seems to me that the Obama camp have something up their sleeve.
    I mean, if they know what you know, why would they be spending such a load of cash? Why would they be trying to go for the kill? Are they just praying for something or do they know something others don't?

    Because 5% is better than 20 (5.00 / 1) (#13)
    by Big Tent Democrat on Wed Apr 09, 2008 at 01:10:42 PM EST
    Yea (3.00 / 1) (#18)
    by ajain on Wed Apr 09, 2008 at 01:14:43 PM EST
    But if SUSA is right they are pretty screwed.

    Parent
    The "shoe-in" can't afford (none / 0) (#45)
    by andgarden on Wed Apr 09, 2008 at 03:00:26 PM EST
    to get blown out in a state like PA. Of course, one would have though that about Ohio, New Jersey, California etc. . .

    Parent
    They have nothing to lose (5.00 / 1) (#25)
    by litigatormom on Wed Apr 09, 2008 at 01:21:42 PM EST
    Everyone expects Clinton to win. But every point Obama can shave from Clinton's margin gets closer to Obama claiming that he "beat the spread" or won a "moral victory."  And every point Obama can shave from Clinton's margin diminishes her argument that she is a better GE candidate.

    I'm guessing that even if Clinton gets a 10% margin, Obama will argue that he beat expectations.  Anything below 10% and Obama will claim a moral victory.

    So to Obama this is money well-spent.  


    Parent

    I think 12% is the break even mark (none / 0) (#28)
    by MKS on Wed Apr 09, 2008 at 01:30:09 PM EST
    If Hillary wins by 11, Obama will still make the same argument....

    And, perhaps the best thing to do is see what the delegate break down would be with a 10-12 point win for Hillary....If Hillary nets just a dozen or so, or even 20, delegates, it would have been a big loss for her.....She is running out of chances.....

    Parent

    Don't agree (none / 0) (#43)
    by Marvin42 on Wed Apr 09, 2008 at 02:53:54 PM EST
    I think its a game of psychology (like price points). The magic number is probably 9-10%. For whatever reason a 9-20 point win sounds a lot bigger than a 6-8 point win.

    Parent
    He Can Try to Spin a Loss (5.00 / 3) (#46)
    by BDB on Wed Apr 09, 2008 at 03:13:30 PM EST
    But for someone whose supporters - including sitting U.S. Senators - were calling for Clinton to drop out because the race was over, losing by any margin in a state you've had six weeks to concentrate on and where you've outspent your opponent 5 or 6 to one (and set spending records) is a sign of electoral weakness, IMO.  And trying to spin it otherwise (Obama closed the gap in a big state, but didn't win, AGAIN!) starts to sound kind of pathetic.

    Parent
    I have no disagreement (none / 0) (#55)
    by Marvin42 on Wed Apr 09, 2008 at 04:01:27 PM EST
    With what you said, I am just saying for Sen Clinton to get the impact she needs from a PA win it should be in that range. A squeaker win won't play as well.

    And I say that as a supporter.

    Parent

    Enough with the race bating off topic nonsense (none / 0) (#24)
    by Big Tent Democrat on Wed Apr 09, 2008 at 01:19:42 PM EST
    Why can you commenters NOT stay on topic?

    All of you. Must you fight in every thread?

    Then do not. ask Jeralyn for an Open thread if you must.

    Parent
    Open Thread (none / 0) (#30)
    by Elijah Trotsky on Wed Apr 09, 2008 at 01:44:30 PM EST
    Hi,
    My mistake.  I am new to this.

    When you have a moment please explain what an "open Thread" is.

    Thanks,

    Elijah Trotsky

    Parent

    An open thread is (none / 0) (#32)
    by vicsan on Wed Apr 09, 2008 at 02:01:44 PM EST
    a thread where you can talk about ANY topic you wish.

    THIS thread is about polls, so that's what we're suppose to discuss in this thread.

    Parent

    Numbers are fun (none / 0) (#29)
    by AF on Wed Apr 09, 2008 at 01:37:07 PM EST
    And demographics are important, but the bottom line is that turn-out models only explain a small part of the discrepancy between the PPP and SurveyUSA results.


    Also, the MOEs are quite different (none / 0) (#53)
    by Cream City on Wed Apr 09, 2008 at 03:42:32 PM EST
    as SUSA's sample is less than half the size of PPP's, so SUSA's MOE is more than 4%, and that can affect the subgroups a lot.  With about 600 total in the SUSA sample for the state, as I recall, and the state's low population of AAs, that comes down to a really small sample of AAs.

    Parent
    If he loses by less than 7% (none / 0) (#31)
    by ruffian on Wed Apr 09, 2008 at 01:48:45 PM EST
    and the vagaries of proportional delegate allocation means that he ties or beats her in delegates...there will be the loudest clamoring yet for it to be over.

    But it won't be.  She'll stay in the race if she beats him in the vote count, no matter the margin.


    From what I've been reading (none / 0) (#34)
    by ahazydelirium on Wed Apr 09, 2008 at 02:07:48 PM EST
    the college vote will not help Obama in Pennsylvania.

    Yes, but can they vote in PA? (none / 0) (#47)
    by ahazydelirium on Wed Apr 09, 2008 at 03:13:40 PM EST
    As the article mentions, many factors work against college students in the Pennsylvania primary.

    Many of those out-of-state students are registered to vote in their home states. Sean Coit, 20, a junior at Saint Joseph's University in Philadelphia and the news editor of the school newspaper, The Hawk, is among them.

    Although he follows politics closely and is interning with a Democratic consultant in Philadelphia, he will not be voting in the Pennsylvania primary since he is registered to vote in Virginia at his parents' house.

    "A lot [of students] are voting absentee at home," said Coit of his fellow out-of-state students.

    Pennsylvania's election rules also act to suppress the college vote. The state has a closed primary, which means only registered Democrats can vote. Since college-age young people are disproportionately likely to be registered as independents, that will serve to limit the number of students who can vote.

    Additionally, as the Students for Barack Obama organization stated, only 5000 new registrants were grabbed. Such a number is not large enough to drastically (or even marginally?) alter the demographics of the voting pool.

    Parent

    i am curious (none / 0) (#39)
    by cpinva on Wed Apr 09, 2008 at 02:37:59 PM EST
    to know why PPP is using 14% as the AA portion of total voter turnout? i ask because, looking at the US Census Bureau database for 2000, AA's make up only 10% of PA's total population. by what logic do they assume they will come out to vote at a rate 40% greater than their overall representation in the PA population?

    granted, that data is nearly 8 years old, but i suspect the demographic breakdowns, for race, haven't changed significantly between then and now.

    however, it does confirm assertions i made in an earlier post, with respect to the distribution of the total AA population: it tends to be highly concentrated in the southeastern states. states that will, in all likelihood, go republican in nov., regardless of who the dem. nominee is.

    and repubs... Many more AA democrats, which may increase turnout percentages of AA voters in the democratic primary.

    Parent
    yeah, (none / 0) (#56)
    by cpinva on Wed Apr 09, 2008 at 04:03:06 PM EST
    i went back and looked at that again, comparing it to the (admittedly a bit stale) 2000 census data. you're probably correct.

    Parent
    FYI here, exit polls v demo (none / 0) (#60)
    by Pacific John on Wed Apr 09, 2008 at 05:13:11 PM EST
    Here are the CNN exit poll Dem turn-out numbers for African Americans and Hispanics versus fraction of the population (% of votes/% of population):

    State, AA, Hispanic
    Ohio, 18/12, 4/2 (too much noise here)
    Texas, 18/11.6, 30/35.7
    California, 7/6.2, 30/36
    Pennsylvania ?/11 ?/3.5

    AA's should provide around 15.5% of the Dem vote to Obama, Hispanics should provide around 2% of the Dem vote to Clinton.


    Parent

    shoot! (none / 0) (#40)
    by cpinva on Wed Apr 09, 2008 at 02:39:11 PM EST
    i meant to also add, good analysis BTD.

    And as usual, the undecideds (none / 0) (#44)
    by andgarden on Wed Apr 09, 2008 at 02:59:35 PM EST
    are considerably lower in SUSA than PPP.

    I predict a 10-15% win for Hillary, with the outside chance of a 20% win.

    We'll see at least 8% (none / 0) (#61)
    by Pacific John on Wed Apr 09, 2008 at 05:17:25 PM EST
    but if Hillary gets a last minute tail wind, the numbers will run way up. It's 8% if she's out-spent 4:1 and has hostile media.


    Parent