News-Record.com

Alright! [Enthusiasm explaineed later]

Mark, I think I've received a call for one of these polls recently. If it's the same, it advised something like, "if you're not eighteen, hang up now." I'm over eighteen and hung up anyway. Surely bored yongsters stay on and participate. I wonder how accurate this "push-botton" methodology is

Roch:

Lots of points here, starting with the one you raise the most prominantly. Almost all surveys short of the census, particularly automated ones, have self-selection problems. If someone didn't want to put up with an automated questioner, their opinion didn't get mesaured.

Of course, I have my doubts about regular ol' voice polling as well. Back in college when I was a young pyschology major learning the nuts and bolts of social science (really!) out statistics and research professors warned us that phone polls had their limitations. Not everyone had phone lines - either by choice or because they couldn't afford them. And there were people with unlisted numbers and those new-fangle things called cell phones. I imagine with the increasing complexity of our communications environment today, the problem is worse.

An interesting point raised by the owner of this company when I talked with him today was that on live voice polling, somewhere around 90 percent of respondents say they are likely voters in the next election. Now, voter turnout never approaches 90 percent. Ergo, some sizeable part of your sample is lying. If they'll lie about being likely voters, what else will they be less than straight-forward about. I'm not sure if I buy that all the way, but it's an interesting argument.

I have a little bit of comfort with the survey because it draws its sample from the statewide voter registration database and the party affiliation breakdown tracks what we know of registered voters in North Carolina pretty well.

This particular sample skews older and whiter than the state as a whole. Older I would expect because of the voter reg. database serving as the potential sample. Whiter doesn't make sense to me. The owner of the company said it was because there are cultural differences in the willingness to take surveys or even answer the phone. Again, I'm not sure I buy the party line there, but it's an interesting thought.

One final thing that gives me a little added comfort with the poll: this survey was done on the company's own volition, but they're in the business to make money. Candidates and causes pay them for data and it doesn't make sense to poll for bogus data. (If you're in a campaign, you want to know the truth - even the ugly truth.) So if they're managing to stay in business, they must be doing something right for their clients.

It's actually and intereting company and a neat concept. I think you're right to take their results with a grain of salt...if the +/- 4.4 percent margin of error didn't do that for you already. But I think given the sample, the sample size, and the robust nature of the results in this case, they're probably on to something. The numbers may not be exactly on target, but they've probably got a pretty good case for saying that there's a sentiment trend in one direction.

Now about that enthusiasm...?