Forget the Polls, Get to Work

Chuck Sheketoff

Don't get depressed or distraught by the daily news about Kerry and the polls. Read "Forget the Polls" and see why.

  • brett (unverified)
    (Show?)

    That plus this: we’re gonna win, for sure. No matter what the polls say.

    Fingers in ears: Check. Eyes wide shut: Check. Loud yelling to drown out all sound: Check.

    Man, is that a convincing article.

  • (Show?)

    Jeez, Chuck, this theory about the weaknesses of polling might be enough to make me--the eternal pessimist--feel better. At least for a couple days.

    However, here's the pessimist speaking: voter registration is great, but this theory is contingent upon our side getting more voters physically to the polls than the other side does on Nov. 2. And I have to think that the Christian conservatives are working very, very hard to get every churchgoing voter concerned with social issues out. Probably just as hard as we are working to get the traditional Democratic voters (including minority and youth) out.

    I guess we'll see who won the turnout war come early November.

  • The Prof (unverified)
    (Show?)

    I'm not sure all of you will be able to get access to this, but the best recent article that I have seen that evaluates the accuracy of pre-election polls is in a recent: Political Analysis

    It's obscure social science stuff, but the basic gist is this: pre-election polls are intended to measure voter preferences at a particular time point, not project forward to a vote in November. That is why these polls are more variable, and that is why some other forecasting methods (e.g. the Iowa Political Stock Market) seem to function better as forecasters.

    Another nice source of social science forecasting is here .

  • The Prof (unverified)
    (Show?)

    ok, sorry, it looks like that is password protected. email me and i might be able to get you a PDF.

  • Betsy (unverified)
    (Show?)

    I've been saying for the last couple of weeks that traditional telephone polling as it's being done is an outmoded method that will soon fall by the wayside.

    Last week, both Jimmy Breslin and John Zogby said the same thing (here's a link to my post, which points to the Breslin column.)

    There are 169 million cell phone users, for example - and not one of those folks are being polled. Add in the fact that people don't answer their own phones any more and/or block calls from unknown sources, and you have a process that's creaking towards oblivion alongside horse-drawn carriages.

  • brett (unverified)
    (Show?)

    This is all true -- polls are unreliable, lots of people don't have landlines, some people are going to change their minds between now and November. But -- how do you explain the fact that the skew consistently favors Bush? I'm not saying his lead is 13, or 11, or even 9 as the NYT poll says, but it's clear that there is a lead.

  • (Show?)

    Leslie, I think you point to a sort of backhanded advantage in all this. So long as the polls don't show Kerry trailing by too much, it should keep the left fixated on showing up. This, along with the lesson of 2000 (every vote DOES count), may well work to the activist's advantage in helping get those couch potatoes off to the polls.

  • Marcello (unverified)
    (Show?)

    The polls done since the RNC show us that a larger number of 'publicans tell pollsters that they are likely or very likely to vote than Democrats do. They are passionate about Bush and they are going to vote come hell or high water. Democratic voters in general are passionate about voting but not quite as passionate. But there are more of them, and independents also prefer Kerry.

    What does it mean? That this election will be determined by how well Democrats GOTV for Kerry. We are going to see the mother of all GOTV efforts in November (and starting mid-October here in Oregon). Between the Democratic party, the unions and the PACs tied to the 527s, Democrats will get nagged to get their ballot in the mail every single day. Turnout will be the highest in recent history. And Kerry will win nationwide the popular vote and the electoral vote.

  • The Prof (unverified)
    (Show?)

    169 million cell phone users. Are you saying that all 169 million don't have land based phones?

    RDD dialing systems do not rely on a phone directory; they can easily include cell phone exchanges.

    Cell phones are soon going to be listed in directories so if you think a cell phone will keep you away from telemarketers, pollsters, etc, think again.

    Alternative polling technologies are already being developed (see knowledgenetworks.com).

  • Betsy (unverified)
    (Show?)

    No, but I am saying that many people - particularly younger people - use their cell phone as their primary or only number.

    In addition, a huge subset of people (yours truly included) never answer their home phone (if they have one) if the call is coming in from an unknown source - or they have calls blocked from unknown sources. Anyone who truly wants to reach me calls me on my cell phone - and I'm no spring chicken!

    I have some numbers up on my site that are admittedly unscientific - my own night's total doing phone banking for the Kerry campaign - but it's only echoing what I'm starting to see show up in some of the reporting taking place on telephone polling.

    That people in general are much harder to reach, some segments of the population just aren't represented in the listings available to pollsters, and phone polling is a 'science' that needs to be either retired or drastically overhauled to remain relevant.

  • raging red (unverified)
    (Show?)

    Fingers in ears: Check. Eyes wide shut: Check. Loud yelling to drown out all sound: Check.

    Funny, that's the same checklist that's required to support the Bush administration.

  • (Show?)

    Case in point about polling - I was just over at electoral-vote.com looking at the map (which, btw, is entirely too red) over the last few days. Four EVs jumped back on the Kerry side since yesterday and it turned out to be NH.

    I flipped back and forth between the last few maps. According to a Mason-Dixon poll (as reported on electoral-vote.com) on 9/15 it was 40-49. According to a Rasmussen poll from the same day it stood at 51-45. The previous poll was a Zogby poll on 9/3 which had it at 50-45. I didn't do any investigating into the details of the polls, but it just proves the point that polls cannot be taken at face value and definitely not as gospel.

    I'm with Jeff. It could completely work in our favor that the polls look like they do. The more comfortable they are in their lead the more they'll take it for granted. And, of course, the harder we'll work. Yup, could definitely work to our advantage. I love irony. :-)

  • The Prof (unverified)
    (Show?)

    thought folks interested in this topic might like Charlie Cook's most recent column. I am pasting it here because National Journal is a subscription only site:

    OFF TO THE RACES Bush, Kerry Polls Apart

    By Charlie Cook NationalJournal.com Tuesday, Sept. 21, 2004

    Rounding the turn into the fourth week of September, President Bush's lead over Massachusetts Democratic Sen. John Kerry seems to be somewhere in the mid-single digits. The differences between individual polls are amazing, however, ranging from Bush ahead by 16 points all the way to a statistical dead heat.

    Scaling the sample down to likely voters to reflect a "normal" turnout, less than 55 percent of the voting age population, doesn't make a lot of sense.

    According to the handy RealClearPolitics average of national public polls, the president has a 6.8-point advantage in a three-way race over Kerry and independent Ralph Nader. The average was 6.5 points in a two-way, Bush-Kerry trial heat and the president's job approval rating averaged 51.1 percent. All three averages cover national polls released over the most recent seven-day period.

    The numbers bounce around each day, as new polls are added and older ones are dropped. Privately, sources in both campaigns and the pro-Democratic 527s say the Bush lead is a bit closer but steady for the last few days.

    Obviously, with such wide variations in the polls -- from a dead heat to a Republican landslide of historic proportions -- something is amiss. That the country is so polarized would seem to put the latter in question, but events of the last couple of weeks and the feel of the race put the former in doubt as well. The averages and private estimates of top strategists from both parties make far more sense, and indeed, the consensus forecasts of many economists are usually more accurate than any one. The same is true with polls as well.

    Some of these differences are due to the sampling frame used in each poll. In most polls over the last few months, Bush has fared better in polls of likely voters than among all registered voters. Indeed, a Gallup poll showing a Bush lead of 14 points was among likely voters, and among registered voters, the margin was seven points. In the 2000 election, the final Gallup Poll of registered voters was virtually dead on, and in 1996, the likely voter number was the closer of the two. With all indications pointing to a much higher voter turnout than normal -- I'm guessing the highest in 30 years -- scaling the sample down to likely voters to reflect a "normal" turnout, less than 55 percent of the voting age population, doesn't make a lot of sense.

    There is some evidence that there has been an increase in Republican Party identification since the GOP convention. Some pollsters who had been seeing a fairly pronounced Democratic edge in party identification since the first of the year are now seeing a Republican edge. With roughly 90 percent of Republicans backing Bush and almost 90 percent of Democrats supporting Kerry, just a few points difference in the proportion of Republicans and Democrats in the sample can make a difference.

    This has renewed a long-standing debate over partisan weighting of polls. It is standard operating procedure for pollsters to "weight" (statistically adjust) just-completed surveys if the poll comes back with a sample that is abnormal in terms of gender, race, income or education, from what the Census tells us the true numbers are from the adult population. The purists believe that it is a sin to weight the sample of a poll by party, since unlike gender, race, income or education, party identification is not a known figure and could change. On the other hand, we have seen some samples that were scandalously tilted one way or the other. My favorite was a Los Angeles Times survey from early August that gave Democrats a 13-point advantage in party identification and showed them 19 points ahead in the generic congressional ballot test. That was simply laughable. Pollsters acknowledge variances from one poll to the next in gender, race, income and education, and they correct for it, but refuse to acknowledge that partisan numbers fluctuate just the same, and need to be corrected.

    My own view is that samples should be weighted by party to the average party breakdown in a combination of the polls for the last several months, linking it to a very large sample of combined surveys to reduce sampling error. While this method might be a bit sluggish if party identification is changing dramatically, it would mean that when a candidate is gaining or dropping, it is most likely because they really are, not because of a sample that is too tilted in favor of one party or the other. If Republicans are indeed gaining in party identification, it will show up after a couple of polls in the average.

    Rather than sampling, some are pointing to questionnaire design, both on the issue of presidential preference and even party identification. Some argue that polls asking the presidential preference question early in the questionnaire tend to show Bush doing better than when it is further back in the questionnaire and that Republican Party identification is higher when that question is early rather than back in the questionnaire.

    With so many moving parts, it may be distressing but is not that surprising that poll numbers can vary this much. But it certainly keeps things interesting.

    Charlie Cook, a NationalJournal.com contributing editor, is the founder and publisher of the Cook Political Report. This column, which also runs in CongressDailyAM when Congress is in session, appears each Tuesday morning. In addition, Cook writes a weekly column for National Journal magazine. His e-mail address [email protected].

  • brett (unverified)
    (Show?)

    red: good to see someone has a sense of humor. I like it.

    I guess Oregon is a swing state again:

    OREGON (Gore won in 2000 by 0.5 percentage points)

    •Bush-Cheney, 47 percent

    •Kerry-Edwards, 43 percent

    Still doesn't matter, though, right?

  • (Show?)

    Or maybe Oregon isn't a swing state, brett: A KOIN TV/Portland Tribune poll released today has Kerry up over Bush 51 percent to 44 percent. Which leads us back to Chuck's original premise--forget the polls.

  • (Show?)

    I guess Oregon is a swing state again:

    Or not. According to a KOIN/Trib poll out today, Kerry's beating Bush 51-44%. And even that probably doesn't capture the margin, because the poll has Portland at 57%-36%. But in 2000, Gore carried Portland with 69%. Either Portland's politics have shifted dramatically right in the past four years or pollsters oversampled Bushies.

  • brett (unverified)
    (Show?)

    36% of Portlanders support Bush?? Where the hell are they all, and why are my friends not among them? That seems way high to me. Better than France, though, where he's at 5%. I do think these polls are screwed up. Either that or Kerry is a worse candidate than I thought.

  • (Show?)

    In 1998 a notable Oregon Democratic pollster said George Bell didn't have a snowball's chance against Kevin Mannix. The House D's believed the pollster and decided not to spend an extra $50k they had... George Bell lost to Kevin Mannix by a mere 384 votes.

    The same pollster also said Civiletti didn't have a chance. He lost to Lokan by 100 votes.

    Some folks won't do anything without a poll saying its safe to do. That's my frustration.

    <h2>Polls can help with message if well written (that's the subject of further blogs), but they shouldn't be used to decide what or whether to do something.</h2>

connect with blueoregon