Daily KOS/Research 2000 Polling on Maine Elections and Marriage Campaign

Looks like if the SFMM people can focus on and capture a more few female democrats, it will be a lock.

The poll is fatally flawed. The sample is biased by including these statements: "A yes vote takes away the right of same-sex couples to marry. A no vote keeps the right of same-sex couples to marry." This information will not be on the ballot. By suggesting a vote will "take away a right," the poll has pushed people to a No vote.

It is interesting to note that the numbers for Mills and Jacobsen are very similar. Mills apparently did not gain much name ID from his last campaign or his years in Augusta. Goes to show that the general public is much different from political junkies and the media. This poll shows that Mills is not a front runner. He is unknown like most of the candidates.

Yep, it is going to be close.

All the more reason for the SFMM to work harder!

Get out there and WORK!

I take the results to mean that it will not be close. Despite phrasing the question in a biased way intended to push people to a No reponse, the result was 48% Yes, 46% No.

Well, perhaps, but we've got to get out there and bust our bunions to get more support.

I would take nothing lying down. At the same time, however, I do think the pro-repeal vote may, because of the nature of the issue, underpoll by several % points. Basically, I think there will be sort of a gay "Bradley" effect at work.

That certainly was the case during the past gay rights battles of years gone by. I remember during the first two referendums the pro gay right vote polled in the high 50's % and it ultimately lost - albeit narrowly.

We've always have been the underdogs.

I have to agree with Dan. Even with a biased question, they could not get a majority to agree with them.

Steven Scharf
SCSMedia@aol.com

I have to agree with Dan. Even with a biased question, they could not get a majority to agree with them.

Steven Scharf
SCSMedia@aol.com

I think an earlier poll this summer by one of the Maine polling firms had the Yes votes prevailing by 2-3% with about 4% undecided. That's not that far away from these results. Polling on this issue has got to be difficult because of the larger number of cell phones out there, and the reluctance of some of the yes proponents to give an opinion due to all the PC atmosphere out there. It does appear that there will be several million dollars chasing about 4-5% of the voters.

Pollster Patrick Murphy agrees with me that the Kos poll was badly done and not very accurate.

DId you agree with the part where the highly-disrespected Murphy said, "The poll could be biased in favor of those seeking to repeal the law, because the second part of the question wasn't put to the people taking the survey, said Murphy."

I don't. I don't think that part matters to the average voter.

But leaving that out shows how sloppy the poll was. It should not be reported as a serious poll. It is trash. The media should know this and ignore it.

The worst part of the poll was describing the question as taking away rights. That certainly biased the results.

The poll failed to follow basic polling techniques. The results are meaningless.

That depends. The poll is meaningless if the respondents knew nothing about the issue or the passing of LD 1020 or the move to not enact the law. Then the criticism is accurate -- it was done all wrong. If, at the other end of the scale, the body of respondents all knew about the question on the ballot and, in fact, had a pretty good idea of how they plan to vote, then they could say, "Oh, you mean Question One. Yeah, I'm voting ..." In that scenario, the results would be valid as the wording of the question would not have biased the results.

Since we don't know the level of awareness and/or commitment among the general population, the results are indeed suspect. But to the extent that they reflect what people already know of their voting intentions, it could be valid. But even a broken clock is right twice a day. And a digital one never, unless it's flashing 12:00:00.

My guess is that the poll is accidentally accurate.

Most people are not paying as close attention to this stuff as you suggest.

There is a science to polling. When that science is not followed, the results are no better than a guess.

The media should not report as news polls that do not follow the basic rules of survey research,

In addition to all of the problems mentioned above, there is the fact that any professional statistician should know that "margin of error" values given to poll results are based on random sampling. As in all polls, pure random sampling was not used here ( as was acknowledged, calls were adjusted by telephone exchanges and by county) precluding randomness and thereby making any assertions of "margin of error" of no greater than 4% (as they did here) bogus.
Of course with no knowledge of the actual "margin of error" it is, of course, impossible to come up with results of any known accuracy.

Unless they employed stratified random sampling, which is intended to minimize over or under sampling of a particular sub-group.

Stratified sampling by definition, regardless of the purpose, precludes a purely random sample and thereby makes it impossible to assign any quantitative value for sampling error. Stratified sampling is, of course, always done for practical reasons such as cost, time or physical constraints, but does not justify the misleading pronouncement of "margin of error".

We have always been the underdogs.

Pretty much like a fish trying to ride a bicycle would be an "underdog" compared to a human trying to ride the bicycle.
A turtle trying to race a gazelle would be an "underdog", because a gazelle is designed for speed,
or a car trying to float like a boat would be an "underdog", because a car is intended for the road,
or anything trying to excel outside of its intended role and discovering it's ill equipped.
You are the underdog, but you can only blame nature, for either creating you as such, (which I don't believe is true),
or blame nature because it won't magically accommodate you as you have changed or been changed by society so that you no longer conform with nature.

IMHO, check this source:

"Once you know the correct formulas, you can readily estimate population parameters and standard errors. And once you have the standard error, the procedures for computing other things (e.g., margin of error, confidence interval, and region of acceptance) are largely the same for stratified samples as for simple random samples."

Agatha, if you were trying to make a rational point...you missed. Is that the best you can come up with?

Thanks for the reference AJ.
The key word in the quotation is "largely", which implies a certain degree of difference. The exact difference is of course unknowable and given the subjectivity and of the word "largely" the sentence becomes vague and meaningless.

That quote says "the procedures...are largely the same," not the results. Accurate margins of error are certainly calculable from stratified random samples.

But not if you ask crappy questions -- which was the main issue with the Kos poll.

Problem is the procedures for stratified sampling make "assumptions" such as a Gaussian distribution of the "massaged" data which are not always necessarily correct. Some times the data is Binomialy distributed or follows a Poisson distribution, or some other unknown distribution in which case the end result is only an unquantifiable guess.

On top of all this there is also the problem of poll respondents lying or refusing to answer the questions, or in the case of telephone sampling, simply not being availble at the time of the calls, all of which tend to "skew" the sample data distribution away from the true population distribution thereby rendering any strict quantitative conclusions impossible.

So, then, I take it that your position is that accurate survey sampling is impossible. Well, there goes another whole industry, along with Dist's dissing of the fast food industry.

So much for economic recovery!

So, then, I take it that your position is that accurate survey sampling is impossible.

Exactly.

(Oh, brother)