When CBS News/New York Times survey figures are released in the weeks before a national election, the numbers are not based on all registered voters in our survey, or on all those who say they will definitely vote. Instead, those numbers reflect the results for a likely electorate.
About 85% of adults say they are registered to vote, but fewer actually are registered, and even fewer vote. 90% of our registered voter respondents say they will vote, but according to the Federal Election Commission, only 68% of all registered voters actually cast their ballots in 2000.
Polling organizations deal with this problem in several ways, usually by reporting results only for "likely voters". Our solution is to create a "likely electorate." We classify our respondents into several groups, and assign a different probability of voting to each group. When we add up the number of voters for a candidate, or with a particular opinion, a respondent with a 0.75 probability of voting counts as ¾ of a voter, and so on.
Respondents who aren't registered to vote are assigned a probability of zero. Registered voters are asked about their self-evaluated probability of voting, their history of voting, the attention they have paid to the campaign, when they last registered, when they last voted and when they moved to their current address. We use the last three questions to figure out whether a respondent has registered or voted since they last moved. If they have not, we give the respondent a lower probability of voting because, depending on whether they moved between jurisdictions, they may not be eligible to vote. Before the passage of "Motor Voter" legislation, it was possible to "purge" some of these individuals, using each state's rules, just as their election boards would have. Since Motor Voter, this kind of vote purging is not permitted.
To assign probabilities of voting, we first relied on voter validation studies conducted for the University of Michigan's National Election Studies through the 1980s. In November 2000, we called back respondents to CBS News pre-election surveys to ask whether they actually voted to refine what by then were outdated probabilities. Beginning in 2000, respondents who have already voted absentee are given a probability of 1.00. The lowest weight (given to a respondent who has moved since casting a ballot or registering, and who says they are not "definitely" going to vote) is 0.05.
The exact wording of the questions we ask follows.
Some people are registered to vote and others are not. Are you
registered to vote in the precinct or election district where you now
live, or aren't you?
Attention paid to campaign
How much attention have you been able to pay to the 2004 Presidential
campaign -- a lot, some, not much, or no attention so far?
Self-evaluated likelihood of voting
How likely is it that you will vote in the 2004 election for President
this November -- would you say you will definitely vote, probably vote,
probably not vote, or definitely not vote in the election for President?
Voted in 2000 or 2002
For the following questions on registration and voting, it's important
that our information be accurate. I'd appreciate your taking some time to
think about these answers.
Did you vote in the 2000 presidential election, did something prevent
you from voting, or did you choose not to vote?
IF VOTED, ASK: Did you vote for Al Gore, George W. Bush, Pat Buchanan,
or Ralph Nader?
Did you vote for U.S. House of Representatives in the elections held
in 2002, did something prevent you from voting, or did you choose not
to vote for U.S. House of Representatives in 2002?
Has registered or voted since moving
What year was the most recent election of any kind you voted in?
What year was the last time you registered to vote?
How long have you lived at the address where you live now - less than
one year, one to two years, three to four years, or 5 years
Copyright 2004 CBS. All rights reserved.