Some opinion polls are also hidden ways of promotion a policy point. I remember several mails that presented themselves as polls but were in fact trying to get you to question a particular political candidate or issue in favour of another.

People love polls. Think about all the magazines with reader polls. Many pool don't tell you anything more then the opinions of the 10 people who took the poll and as your story illustrates sometimes they don't even tell you that much.

As for the reliability of the poll results that's another question. Leading questions or answer options are one issue but how the poll is conducted, who the respondents are, what the sampling frame is (i.e. how representative is the group you are interviewing compared to the population you are making statements about)are all questions to ask to assess the relevance of any poll results.

Why favorite opinion poll question is "do you favor welfare" which always get a low response.

However, if you ask "do you favor giving money to the poor, ill and others unable to care for themselves" you get a high response.

What this shows is that most people's opinions are based upon a superficial response to buzz words. Recently a US poll asked which party has the majority in congress. The result was slightly above 50% correct. Given that there are only two choices, that's not saying much.

For an organization that tries to do a better job look up PIPA.

Policies not Politics
---- Daily Landscape

OK, first... what the hell were you doing on that site?  Yeah, I know you warned us, but... yeeech.

Second, and more on topic... no survey conducted on the Internet is going to be mistaken for a serious one by anybody who knows what they're talking about.  It's a decidedly unscientific self-selected sample.  Online polls are for novelty purposes only and should not be taken as indicators of anything generalizable.

My problem with opinion research has more to do with the inept way it is often covered in the media.

Even if this were a scientifically conducted poll, it would be rendered invalid based on the directed wording of the question.  It is certainly possible to word questions in a way that they are designed to elicit specific answers, but it's also possible to word them in a way that makes it much more likely to elicit the respondent's real opinion.

Any reporter attempting to write about a survey needs to see the wording of the questions being answered, and if they're asked in a leading way, the research needs to be either disregarded or at least acknowledged for what it is, which is a poll designed to produce a certain result.

I partly financed my graduate education by conducting telephone surveys (what can I say, I have a nice voice), some of which were political (if the election for Senator were held today, would you vote for....) and some commercial (how many cans or glasses of soft drink have you had in the past week?).  I also have some experience in analyzing public opinion research.  There are growing and serious problems with its reliability, especially related to the great unreported secret, the rejection rate.

When it comes to telephone surveys, this is partly the result of the over-polling of certain populations (e.g. Pennyslvania) who then just don't want to take part in any more surveys, and partly a result of technology, e.g. caller ID, which allows people to decide which phone calls to answer.  There's a lot more to it than that, but it's something the industry doesn't like to talk about because it threatens its very existence -- if you can't produce reliable surveys, what's the point of doing them at all?  (Assuming you're one of the people who really wants reliable opinion data....)

Telephone surveys are also, of course, completely useless in societies where most people don't have landlines, e.g. sub-Saharan Africa.