What about accuracy of polls?

Question: I am not a fan of polls because they can and often do reflect information that is not valid. An example in today’s paper stated that you called 500 people and interviewed 406 of them that said they were registered voters. Since only around 50 percent of all potential voters are registered to vote how did you select 500 people and 80 percent of them said they were registered voters? How many actual people were called that did not answer the phone, refused to be interviewed, or hung up once they found out who was calling?

If these were random calls it is very unlikely that you got 406 registered voters out of 500 calls. If these were selected calls and you knew in advance that these people were registered voters then why did 94 drop out because they said they were not registered to vote? -- Wayne Lythgoe, Colbert

Answer: I've asked Jim Camden, our veteran political reporter, to respond to your question. Camden has worked with poll results for years and regularly helps us formulate questions for political polls that we commission.
Here's Camden's response to the question:
Registration in Washington state is actually much higher than the 50 percent the reader cites.
According to the most recent census figures, there are 3,282,777 registered voters and an estimated 4,555,664 persons 18 or over in Washington state, and not all people 18+ are eligible to vote (felons and non-citizens are included in the census figures, but they can't vote.) But even by just using the raw data for those over 18, Washington has a voter registration rate of at least 72 percent, with more voters being signed up every day. So getting an 80 percent success rate is in the ballpark and not at all unlikely
It's my understanding that the poll we cited Monday used a random-digit system broken down by the exchanges for the different communities and divided evenly across area codes. That means we make calls based on the proportion of the state's population in the different area codes, and the different exchanges. Each pollster has methodology that they believe best handles the statistical problems of non-answers, hangups and refusals.
There are other polls that work from a list of registered voters, or even those that work from a list of voters who have cast ballots in two of the last four, three of the last four or four of the last four elections. Those are polls that are more likely used by campaigns, political consultants and political parties. Because addresses and phone numbers change, however, there are always some persons contacted on those surveys who are not registered voters.
For people who are not fans of polls, there's always something to discount. For people who are using polls, the most important thing is to get someone reputable to do the poll. -- Gary Graham, managing editor

Useful links
About Steve Smith
About Gary Graham
About Ken Sands
About Doug Floyd
About Carla Savalli