Articles and Analysis


POLL: Datamar Florida Primary

A new Datamar Inc. automated survey of likely voters in Florida (conducted 5/14 through 5/18) finds:

  • Among 413 Democrats, former Sen. John Edwards (at 25.7%) edges out Sen. Hillary Clinton (24%) and Sen. Barack Obama (18.9%) in a statewide primary.

  • Among 607 Republicans, former Mayor Rudy Giuliani edges out former Sen. Fred Thompson (27% to 22.4%) in a statewide primary; former Gov. Mitt Romney trails at 17.6%, Sen. John McCain at 11.9%.




I looked at the pdf and nothing was divulged. A "proprietary algorithm" of a "likely turnout model based on election cycles and other factors." Very strange. "Barak" Obama was misspelled--not promising from someone who supposedly deals with political names on a constant basis.

Democratic numbers are especially "eyebrow-raising."

The few demographics it did divulge were simply hilarious. Look at this.

High school
graduate or less 17.2
Some college/trade
school 32.4
Graduated college or
more 50.4
Total 100.0

18-34 8.0
35-49 19.1
50-64 31.2
65 or older 41.6
Total 100.0

This is polling malpractice if anything is. Sounds like Edwards needed a poll to show donors, even if it was utterly fraudulent.



A brief follow-up on Alex's comments. Based upon 2000 census data, in the general population in Florida, 35% of the population is 55 or older. For education status, 48.8% of the population has no education beyond high school, 28.8% went to trade school and/or attended some college, 22.4% received a bachelor degree or more.

I would suspect that the 55+ population and the college graduates would be higher among the voting populace than the general populace, but the numbers given by DataMar do seem somewhat unrealistic.

Having said that though, Alex, what proof do you have that this poll was paid for by the Edwards' campaign? That is an awfully strong charge, and I can see nothing on Edwards website or the DataMar website to support this claim.

Is there another explanation for the variance in the demographics from the polling data and the general population numbers from the census?


It was a strong charge, but I recalled an similarly anonymous group ("Environmental Policy" or something) releasing a poll c. December 14 2006, showing Edwards with a 'commanding' 36 percent in Iowa... except that it was taken from September 2006, before the landscape-changing Nov. 06 election.

Campaigns certainly spin data as best they can, but this marks a second very suspiciously-favorable to Edwards outlier, released to national media (which never do due diligence on polling internals) at a time when Edwards was slipping in other polls.

It's hardly a smoking gun. But it makes two points in a distinct pattern of anonymous outfits suddenly kinking-up Edwards' downward-sloping poll average with data that are actually very dubious.

Combine that with some very questionable demographics, and you have a problem.

I am aware that Florida is a very gray state, and a lot of those voters are Democrats. But in that case, the sample shouldn't show over half of its respondents as college graduates. It's not as ridiculous as it first seemed, but it still doesn't add up.


Jeff Baker:

After some more researching of this issue, I see no reason to make the claim that John Edwards campaign falsified data. Based upon 2004 election data from Florida, 36% of voters were between 46 and 64, and 26.2% were over 64 (62% over 46). Still under the percentage reported by DataMar, but these are general election numbers, not primary election numbers. And, as the demographic breakdowns are based upon the respondants they reached (which could be biased by the time of day that the phone calls were made, and I would expect older people to be home more frequently, and thus respond more frequently than younger people)

As DataMar notes on their results page: "where necessary responses were weighted according to age, gender, and political party."

Looking at the Republican data, we see very similar numbers for both age and education (18-34: Dems - 8%, Reps - 7.4%, 35-49: Dems - 19.1%, Reps - 20.4%, 50-64: Dems - 31.2%, Reps - 31.1%, 65+: Dems - 41.6%, Reps - 41.0%). Same for education (high school or less: Dems - 17.2%, Reps - 14.2%, Some college: Dems - 32.4%, Reps - 34.6%, Grad college or more: Dems : 50.4%, Reps 51.2%).

The age numbers do not seem too unusual compared with 2004 statistics available from the Census Bureau, particularly if we take into account differences between the primary and general electorate and consider who might have been available to answer the survey.

The education data does look weird for both Democrats and Republicans. For the entire country (data on the US Census Bureau does not break down the data by state for education level of voters), 36.6% voters in 2004 did not go beyond a high school diploma, 30.95% went to college or trade school, but did not receive a bachelor's degree, and 32.44% received at a minimum, a bachelor's degree. I have no explanation for the demographics of the DataMar Florida survey.

Contrary to Alex's suggestion on his blog, DataMar does appear to be a legitimate company. According to their webpage, DataMar has been in business since 1981. They provide results on a number of surveys going back to 2005. I'm not sure if that is when they started doing them or simply that is all that they have posted on their webpage.

Now, the question becomes, why is Alex calling out Sen. Edwards on this poll. This is not the first time he has done so. His comments above about a "recalling" a "similarly anonymous" group suggests it is a faint recollection. His blog on the DataMar polls belies this inference. He provides a direct link to his blog in December of 2006 in which he takes Edwards to task for the "Environmental Defense" survey.

So, Alex, why this vendetta agains Edwards? Whose campaign employs you?

For the record, I am not employed by any political campaign or party, although I made some political contributions to the Edwards campaign (less than $500) and plan on voting for him in February.



I don't have a vendetta per se--you raise good points--but the data are still somewhat suspect.

I don't see a reason to assume that 55+ are more represented in primaries. I agree that in general, older people vote more, but I would think that in primaries you would get a more even balance.

I didn't say Edwards himself falsified this. Campaigns produce "good" polls for media spin quite often, especially when more mainstream and better known polling shows them slipping significantly. It is a common PR tactic; single-issue groups in DC are really promiscuous about it, although they usually use heuristics embedded in the phraseology as opposed to outright bad sampling.

As you admitted, the education data is very strange; that was the deciding factor to me.

I have a hard time believing that one-third of Americans have gotten a bachelor's degree. Today, 25% of Americans graduate from college, itself a high, no? Older voters should have lower rates of college education.

The sample's skewing old, *and* skewing towards very high education, *and* the fact that this is so contrary to other polling data I've seen, *and* the fact that a company I'd never heard of will release these surprising results, while at the same time hiding behind a proprietary algorithm to mask all methodology, *and* the fact that they misspelled a leading candidate's name twice, *and* the fact that I could find no names behind the organization -- for me, that is a tough sell, and I have a huge pet peeve with junk polling getting quoted by MSM without any due diligence.

I am employed by no one except myself; however, I do have a pretty penny riding on Obama over at Intrade :-)



If you ;ook at their web page they were right on the money in predicting the Governor's race in California:

Scwartznegger 53.8
Angelides 35.8

Actual: Scwartznegger 54 Angelides 39

They had the closest prediction than any other pollster's final poll for CA governor's race.

They are not a junk pollster


Robert P.:

Alex, the poll in Iowa you are talking about was just county Chairs and Vice-Chairs (political insiders). The same poll showed the normal lead amongst likely caucus goers, but showed a HUGE lead among county leaders.


Robert P.: I know. In that case, my beef wasn't with the sampling, but with the fact that someone tried to spike an adverse (to Edwards) polling trend with information that was completely irrelevant because it was so outdated.

However, it was still reported in December in the same breath as other, up to date polls.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.