More Polling Problems

And it's not even a Sun poll this time. Both the Capital and the Sun have been shopping this Dan Nataf poll on the Governor's race in Anne Arundel County as well as the County Executive race. Of note in the poll is that the County Executive's race is basically tied with a quarter of the vote undecided.

If you can get beyond the bias of Nataf's organization, as I have previously noted, take a look at this blurb on page 26 of their press release on this poll:

The survey polled a random sample of 382 county residents who were at least 18 years old. Phone numbers were derived from a database of listed numbers as well as computer chosen, randomly assigned numbers. There is about a 5 percent statistical margin of error for the overall sample; the error rate is higher for subgroups such as "men"” or "“Independents."” The dataset was weighted by demographic variables such as party registration and gender to better represent the general population.

What this means is that, for lack of a better term, this poll is meaningless crap. The sample was not created using any professionally recognized way of creating polling data. The sample used random digit dialing, however used the phone book in order to create these numbers. They did not use lists of registered voters in created its number list. Who knows if the people who were polled are even registered to vote, much less deciding whether or not they are likely voters. The sample size is much too small to use in a competitive general election.

Nataf's crew has a history of this. Four years ago, their poll indicated that Janet Owens was going to be re-elected by 12 percent. Of course, the statistical data oversampled Democrats and (amazingly) used random-digit dialing instead of actual voter lists. I asked Nataf about this following the 2002 County Exec debate. He did not have an answer then as to why they did this, and I presume he is not particularly worried about it considering he is knowingly making the same basic mistakes four years later.

Nataf has the right to produce whetever poll he wants (doing it on the taxpayer dime is another story). But the local papers need to be responsible enough not to publicize meaningless drivel in lieu of actually useful statistical information.