[Dooby]

I really hated polls for a long time until people that were a lot more experienced and paid a lot more than I was convinced me otherwise. And I started seeing how closely the polls mimicked actual results. Statistically, they work. Polls are usually inaccurate for one of several reasons: (i) the question is poor; or (ii) sample is poor; or (iii) the poll is interpreted badly.

While it is hard to believe, you really can gauge the public sentiment by randomly sampling just over 1,000 people.

Example of a poor sample is best shown by the poll that started this thread. We live in a nation of about 50% turnout of eligible voters. This sample is taken from "adults", which, if representative of the actual sample, means that about half of those people surveyed won't actually vote. There are also a number of recent events that currently play havoc with random sampling, including recent legislation and technology. Frankly, call blocking and caller ID has played havoc on pollsters. On the one hand, it makes it difficult to accurately sample tech savvy younger voters, but on the other hand it makes it difficult to sample older, more affluent voters. Additionally, the Motor-Voter and Rock-The-Vote legislation initiatives played havoc on the voter rolls. There are numerous people inadvertently registered 2, 3 or 4 times and numerous people who are ineligible to vote that are listed as registered voters (immigrants and felons).

After the 1996 Republican Primary, I turned over a list of 350+ names that were registered multiple times to the Dallas Co. Elections Dept. That was just in a single congressional district of only past republican primary voters.

Poorly worded questions are frequently found in the mass media polls, such as this one. The "Who will you vote for" questions are not so bad, but the problem often lies in issue questions like "Is the environment important to you?", which is an idiotic question. But media outlets often ask stupid questions like that because they have no vested interest in the outcome. You also run into a huge problem with negative-push polls, but that is a different issue because those "polls" aren't polls at all and are not intended to yield results.

I have personal experience with poorly interpreted polls. I cannot stress enough how often people are led astray by reading something into a poll that simply isn't there. You often see this in advocacy groups trying to argue that their position has the people's support. Frankly, if there was one thing that was the kiss of death for Al Gore's campaign during the summer of 2000, this was it-polling and trying to match what people said they wanted with what Gore said and did. It made Gore look like a fool.

An important thing to remember is that people don't know what they want. The best examples are not politically related, but in technological innovations of the last 25 years. If you described the fax machine or federal express to someone in the 70's and asked that person if they were interested is using such a service or buying such a product, they'd have said no, or maybe that they'd use it once and a while. If you polled using a blind-ID (no names but by description only) of Ted Kennedy or George Bush or Ann Richards and asked the people if they'd vote for someone like that, they'd say "no." Same principle-Nixon would have blown Kennedy away. Asking someone "Do you think America should do more to help the working poor?" is not the same thing as asking "Do you support a $300 Billion tax increase for job training and childcare tax credits for part time and migrant workers?" The Clinton health care plan in 1993 was a dramatic example of how to screw up by misinterpreting public sentiment.