[NickMPK]

[ QUOTE ]
[ QUOTE ]
The people who are going to show up to vote are likely voters. There is no possible distinction between the groups barring a time machine. When pollsters ask who is likely to vote, they are asking who is going to show up. The empirical evidence has borne out that asking if a voter is likely to show up effectively screens out non-voters and produces accurate results.

[/ QUOTE ]

A couple things...

"Likely voters" aren't determined simply by just asking someone if they're likely to show up. That wouldn't be accurate enough, as people on average tend to overstate their intentions. It is usually one of the questions they ask, but there are usually five or ten more questions also. But that isn't really relevant to my point anyways.

More importantly, my point was about turnout, and turnout isn't going to be determined by "likely voters". The likely voters are just that... likely voters. In 2004, voter turnout was about 55%. In 2000, it was about 51%. In 1996, it was under 50%. The difference isn't in the percentage of "likely voters", but rather the people who wouldn't be qualify as "likely voters". They might be considered more of "possible voters", "might vote", "occasional voters", etc. The likely voters are going to vote no matter what; it is the non-likely voters that influence the turnout rate.

[/ QUOTE ]

Actually, the screen for likely voters (in a general election) is typical just "How likely are you to vote" (only asking people who are already registered). Many firms use an additional screen for "definite voters", which may rely on past voting history, etc, but this is almost never used/reported by the media. And of course, there are various things that pollsters do to weight survey respondents by likelihood to vote, etc, but again, the polls you see in the media are usually not that sophisticated.

People do overstate their intentions, as you say, so the screen for "likely voters" used by most polling firms nets the vast majority of people who end up voting. It is much more probable that a "likely voter" will end up not voting than than that an "unlikely voter" will vote.