Did the AP Fix Its Poll to Help McCain?

Yesterday, the Associated Press’ Liz “Donuts for McCain” Sidoti published a story on the new AP poll showing Obama leading McCain by only one point – 44 percent to 43.  The headline to Sidoti’s piece falsely claimed the race was “all even” (see Kevin Drum’s explanation as to why a poll showning a lead within the margin of error is not “all even”).

Drudge and numerous blogs on the Right immediately heralded the poll.

My initial thought was it must be an outlier.  However, The Atlantic‘s talented Mr. Ambinder has turned up a clue that there may be a more devious explanation for the result in this AP poll.

Marc has discovered that 44% of the AP poll’s likely voter sample is made of self-described evangelicals/born agains.  Ambinder notes: “That’s about double the weighted average that pollsters generally assume.”

Those of us on the progressive side have repeatedly noted the AP’s biased reporting against Barack Obama.  This, of course, has come while the AP has been under the stewardship of McCain ally Ron Fournier.

While a narrative was forming that the race was over and the McCain campaign was falling apart, the AP poll provided a spark of life or a boost of energy perfectly timed to keep McCain’s hopes afloat.  Could the AP have really tanked a poll to help McCain just 13 days before a presidential election?   The evidence warrants suspicion and, at least, an explanation from the AP.

(edited by DrGail)

6 Responses to “Did the AP Fix Its Poll to Help McCain?”

  1. Holy crap that’s a good catch. I just ignored the poll; the data point was just so fabulously off compared to everything else we’ve been seeing I didn’t think much about it.

  2. Angellight says:

    The enemies of America are those who hiber hate and separtism in their hearts and who pit one group against another when we all come from the one same Source which shines on us all, good and bad, and when the blood in each of us is red, the symbol of the inner man and our oneness.
    “The journey will be difficult. The road will be long. I face this challenge with profound humility, and knowledge of my own limitations. But I also face it with limitless faith in the capacity of the American people. Because if we are willing to work for it, and fight for it, and believe in it, then I am absolutely certain that generations from now, we will be able to look back and tell our children that this was the moment when we began to provide care for the sick and good jobs to the jobless; this was the moment when the rise of the oceans began to slow and our planet began to heal; this was the moment when we ended a war and secured our nation and restored our image as the last, best hope on Earth. This was the moment – this was the time – when we came together to remake this great nation so that it may always reflect our very best selves, and our highest ideals. Thank you, God Bless you, and may God Bless the United States of America.”
    Obama’s Nomination Victory Speech In St. Paul June 3, 2008.

    “Obama is a LightWorker – an Attuned Being with Powerful Luminosity and High-Vibration Integrity who will actually help usher in a New Way of Being”


  3. Forrest Milder says:

    I had noticed the same thing. (It’s great that you can read the actual date on the Internet!)

    I also noticed that 51% of the respondents were Protestant or “other-Christian”, and yet 44% were Evangelical. Even assuming that ALL of the “other Christians” were Protestant (and not, for example, Catholic), I can’t imagine that 85% of all Protestants are Evangelical.

    On the other hand, I was surprised by certain other parts of the poll which turned out to be surprisingly accurate (like the percentage of people who own their own homes, or who live in different regions of the country; these were both consistent with Census or other respected data).

    A few other items didn’t add up — the number of people who were certain/uncertain/might change their mind about their vote only added up to 571, even though about 800 were “likely voters”. What happened to the other 229 people? And is it really possible that 99% of the 800 likely voters had an opinion on what the state of the economy would be a year from now? I would have expected at least a 10% undecided for that (By comparison, most questions like “which candidate will improve the economy?” had about a 10% undecided, so why not for that question?).

  4. Forrest,

    I haven’t dug into the cross tabs of this poll, and tobe honest, I doubt I will ever get around to it, but I did want to answer your question on “What happened to the other 229 people?”

    And you otherwise seem well informed, so if my answer is something you don’t already know, I apologize and am in no way trying to insult your intelligence.

    The 229 people that you refer to basically didn’t make it through the likely voter filters of the poll. That is, whenever you are given the finalized results of a poll, before those results are given the RAW data from the interviews conducted is crunched eight ways from Sunday before it is put out in the finished form that is reported in news articles.

    What most pollsters do is they create a likely voter model. This is because not everyone who is registered to vote actually goes out there and votes. We know this, and we try to account for this because if we didn’t, polls would be even more inaccurate than they already are.

    Now, each pollster creates their likely voter model differently. Usually historical trends are taken into account for each demographic. For instance, the elderly are among the most dependable voting blocs in this country, so you will see likely voter modelswill consist of a high turn out rate among the elderly. But you also have other factors that will come into play in creating this model as well.

    Once you have the model established, you take the raw data from the sample, and you set up filters in such a way that you essentially throw out enough people so that your processed data reflects what your likely voter model would suggest.

    So, what you’ll see happen is so many first time voters will get thrown out, so many young voters, so many of this kind of voter and that kind of voter because this is essentially accounting for those people we expect not to actually vote on election day.

    Thus, this is what happened to your 229 people, they didn’t make it through the likely voter filters. Those 229 people also contribute, I believe, to what I call the Obama Effect.

    The Obama effect is a phenomenon wherein a politician outperforms his polls based on three reasons. 1) aggressive and successful GOTV effort. 2) significant support from people who are not typically polled (new registrants, cell phone only voters). 3) And this is the important one to our discussion here, those voters that do not make it through more conventional likely voter filters.

    First time registrants and young voters historically have a low turn out rate, and as a result pollsters tend to filter a lot of these folks out when they build their likely voter models. But what we saw an awful lot in the primaries was that Obama actually did what few other politicians were able to do; promise to bring out the young vote, and then actually deliver.

    So this is just some stuff to mull over. The more Iearn about this poll, the less I trust it, and I already was skeptical enough about it to not pay it any mind. Not only was the weighting ludicrous as Mac pointed out in his post, for instance, but the sample size was surprisingly small with only eight hundred interviews. By contrast, the major daily trackers have sample sizes over two thousand (the larger the sample size, the more accurate and reliable the results).

    For more information on this stuff I got two things for ya. Keep reading here, I usually do at least one or two polling analyses a day. Also go check out Nate Silver at fivethirtyeight.com who just recently wrote up a very informative article on his theory on this election season’s likely voter models which is similar to my own take (specifically that a lot of the pollsters are basing this election cycle’s likely voter models on the 2004 election).

    Thanks for stopping by and don’t be afraid to speak up again.

  5. Ace Armstrong says:

    I also was skeptical when I saw this poll, but dismissed it to the AP’s preference for McCain.
    I might also add to the comments on voter models for polling that another unique feature of this election cycle is a fuller presence of the Internet and its communication potential. Those of us who are using the Internet as a primary source of news for the elections are way ahead in information and communicating with others. A lot of the nonsense is filtered rather quickly.

  6. DrGail says:

    Macswain, Forrest, Kyle, and Ace: This discussion was very interesting: I wish I had something to add to it, but the fact is that I learned quite a bit more than I ever knew about the subject. Thanks guys!

Leave a Reply

Your email address will not be published. Required fields are marked *

Connect with Facebook