Home > Political Ranting > The Polls Must Be Crazy

The Polls Must Be Crazy

September 18th, 2004

One really has to wonder if the polls mean anything this year. Ever since the Republican convention, they’ve been going nuts. Just a few days ago Harris had Kerry out in front by one point. But now Gallup comes up and says that Bush leads by fourteen points. No way both are true, or even close to agreeing upon anything. Other polls show similar discrepancies. Kerry is at 41% or at 48%, Bush at 46% or 54%.

Considering that the margin of error is between 3% and 4%, it is highly unlikely that reality is converging at the center and all the polls are off completely to one side. It is more likely that the data is simply greatly unreliable. Perhaps, for some reason or another, the different polling organizations started using very different models, or introduced new variables at about that time. Or maybe it’s just that this election is so unpredictable that surveys simply don’t work at all.

We may have to fly this one in blind.

Categories: Political Ranting Tags: by
  1. September 18th, 2004 at 04:51 | #1

    It depends on who is doing the poll and what area they chose to do it. They can pick a nice neighborhood and it be completely Democrat, or it could be Republican.

    You can never really tell when you are asking under 1,000 people

  2. Ron
    September 18th, 2004 at 10:28 | #2

    I found these two sites to be worthwhile in explaining the highlights of polls, and… why we shouldn’t put too much absolute trust in them.

    All about polls

    A Polling Primer

    The important thing to remember these days of neck-and-neck campaigning for America’s future is to look at the poll’s margin of error. For polls comparing Bush & Kerry, for instance, if the difference between the candidates is less than twice the margin of error, than there’s some overlap, i.e., the poll indicates a dead heat and is virtually meaningless.

    Gallup has consistently been predicting Bush wins since last year, even while all the other polls show Kerry with a slight lead. This latest from Gallup, though, is really strange, as if they polled only white men waiting in line to buy NASCAR tickets, beer in one hand, a gun in the other, and a T-shirt that said “Bush is my savior”.

    What the guys want
    The president’s macho persona has given him a big lead among white males
    By Kenneth T. Walsh, US News & World Report, Sept. 13, 2004, issue Sept. 20, 2004

    I would personally be very interested to know how these polling outfits determine the margin of error. I’ve read (maybe at one of the links above) that polling today is a near science in its accuracy and portrayal of reality, but I have my doubts. For instance, I recently e-mailed Pew Research Center about whether they do/have done any polling of overseas voters, and whether they knew of anyone else who did, and though they shared my sentiments about the probable importance of the overseas votes this November, they don’t poll this bloc, have never done so, and don’t know of anyone who does. Too expensive, time-consuming and difficult, they said. This makes me immediately question the absolute validity of their polls when whole blocs are being eliminated like this. As another not-so-far-fetched example, getting the opinions of the growing number of poor people in the States, many of whom may not have phones, could likewise produce very different numbers than what Gallup and others continue showing.

    BTW, this poll (using the USA Today link you provide) showing Bush leading by 14 points was done with only 1022 adults, aged 18 and over, over a span of only 3 days, The margin of error was +/- 3%. Also, the 14% figure is from Question #2, about who people would “lean toward”, not necessarily vote for. Question #1 covers who people would actually vote for (Likely Voters: Bush 55%-42% / Registered Voters: Bush 52%-44%). So when asked who they’re “leaning toward”, among Likely Voters it’s Bush by 14%; but for Registered Voters it’s Bush by 8% (50% to 42%). The RealClearPolitics site doesn’t indicate any of these details. You have to look at the complete poll data, if possible, to get the whole story. Gallup has been around a while, with consistent polling techniques over time, one of the good things about them, and a quick look at these same questions shows Kerry trailing by similar margins in January this year and other times. I agree it’s a bit late to be trailing by so much, but looking at how the numbers have varied over time tells me not to worry too much. Also, Kerry’s known supposedly for pulling out all the stops late in the game. I for one am hoping he can still do it, over and above these polls and whatever October Surprise Bush and Rove have up their sleeves…

  3. Ron
    September 18th, 2004 at 11:56 | #3

    Just to add a few more comments to the above, I was just listening to CNN’s analyst talk about the latest Gallup poll and how it differs so much from all the other polls, and he made a good point. As discussed in the first of two links on polls I listed in my previous post, how a poll’s questions are worded AND the order in which the questions appear can greatly affect the results.

    In this Gallup poll with Bush supposedly commanding a 14-point advantage, Gallup asks about voting preferences BEFORE asking any other questions. In all the other polls, the respondent’s candidate choices are asked AFTER questions about the issues like the Iraq War, the war on terror, the economy, etc. As the CNN analyst pointed out, when people are asked who they’d vote for right off the bat, they usually go for their personal favorite, and Bush holds an advantage here. However, when people are asked the same question after being queried on and asked to think about the issues, Kerry is almost always favored.

    This thing isn’t over by a long shot.

Comments are closed.