We are often asked how to we get our samples. So let me explain…
The vast majority of our surveys are dealing with campaigns, politics and public policy. So the opinions of people who vote, and actually influence the system, matter most. Therefore, our samples are usually based on the voter registration database of a given state.
We find this method to be more accurate for political purposes than random digit dialing.
Voter registration databases have lots of information on each individual voter including gender, race, party affiliation, age, address, precinct, municipality, county, districts, and voter history. But they do not have phone numbers.
A company called Aristotle International, Inc. takes the voter databases of each available state and then appends home phone numbers to match each voter. We purchase our samples from these lists.
In targeting our surveys we can select on any of these criteria. For example, we could conduct a survey of only black voters from Cumberland County who are registered Democrats.
Most of our surveys are only targeted to likely voters. For the most recent “Dole vs. Miller” poll we created a pool of likely 2008 general election voters. That pool consisted of any voter who voted in either 2004 or 2005 or 2006 in North Carolina. Then from that list of more than 2 million voters who also had home phone numbers, Aristotle removes duplicates and randomly selects numbers for us to poll.
In addition to the sample, at the beginning of each survey we ask everyone if they are a registered voter—to make sure we are talking to the right person, and weed out non-registered phone answerers. For some surveys we may get even more specific like “…do you intend to vote in the Republican primary in 2008…” It’s not perfect, but we try our best to talk with only the people who will actually be casting ballots.