A Few Words on Polling

Polling Report is a good resource to view ALL the polls. An index of polls is your best bet to get at the truth of opinion for a moment in time. Remember, a poll is only a snap-shot of a single moment in time, and therefore polls have a very limited utility. The media would make you think that polls have a maximum utility, they just don't.

As polls go, most suck, but Ipsos/Reid/Cook Political Report and anything from The Field Company is above standard, Zogby is usually pretty good, and Quinnipiac or the other regional University based institutions believe strongly in the science of polling and do a good job. The smaller the aspirations of a poll, the better. It's tough to take a sample of 3,000 people and say this is what the the whole country feels, without some very serious grains of salt added.

An indicator of a quality science is when they show their MoE, (Margin of Error), to the tenth, as in +/- 3.1. To really know if a sample is valid, and the pollster are serious about the science involved. When judging a poll not only do you need to know the questions, size of the sample, but where the sample was taken from, and how. Phone? In-person?

The right track/wrong track is still the best measure of the current national 'snap-shot', really the only poll I believe in, (from Ipsos-Cook.) The right/wrong track works because it is elegant in it’s simplicity and has a long track record of being a great indicator of the nations disposition for change. The best non-open ended, (when I say open-ended I mean polls where respondents fill in the answers instead of checking prefabricated answers…very bad), polls for opinion on individuals use a 'Temperature' scale developed by Mervyn Field in the 1960s, (the bomb-shizzle of polling.)

The place where good polls go bad, if it’s not the diversity and quality of the sample, it’s providing the question AND the answer. The question and answer shade one another, giving you a false reading. If you can’t conduct a poll where the respondents fill in their own answers in their own words, (ideal), then you have to make the provided answers clear, unbiased, and simple: 1-10, 1-being coldest, 10-being warmest, how do you feel about George W. Bush? Real simple, real elegant, devoid of words that may skew.

Gallup and Fox Dynamic often fail because of the language they use can distort results, and there have been charges that Fox Dynamic doesn't really try very hard to get a random sample, (let's call folks in Alabama, and Virginia Suburbs and call that America!) Most polling companies disregard the need for regional diversity. Also the people that answer polls aren't always what we would call 'prime indicators', for example the test to see if you are talking to a 'likely voter', is too often, asking, 'are you a likely voter'? Or the other lazy trick that polling companies use to declare ‘likely voters’ is asking if the respondent has an individual retirement account. There are a series of questions you can ask, that in combination give a good indication of ‘likely voter’, not two or three, I could tell you what these are, but then I would have to kill you, and I just don’t have the time to get you all.

Anyway good polls have open-ended questions in clear unbiased language, don’t provide the answers but instead allow the respondent to provide the answers, feature long interviews, and a real diverse sample. It's tough to pull off. These companies get paid to put out a poll; they rarely care about the science.