question about the polls, if someone can help me out

[rch35]

the polls that say they are comprised of likely voters, they consistently show a close race, or obama pulling ahead slightly

well

what is the definition of likely voters, do they take into account all they newly registered voters? or do they completely cut them out?

[Tansy_Gold]

1. Generally --

"Likely" voters are those who are not newly registered, have voted in previous elections, and/or tell the pollster they are planning to vote in the upcoming election.

The pollster has the ability to define "likely" however they wish. Maybe they only want people who voted in the past two presidential elections, or the last presidential and the last off-year.

"Registered" voters are, as the name implies, officially registered to vote. They may be newly registered or veterans of 10 election campaigns.

Some pollsters rely only on self-identified likely and registered voters. In other words, they ask the respondent "Are you registered to vote?" Others may use voter rolls so that they only call people who really are registered.


Pollsters tailor their sampling to be as effective as they believe it can be. With all the newly registered voters, there's no precedent for determining how they will actually vote. The pollsters can only compare their poll results to actual PAST results, and the new voters don't have any history.

Any poll that includes them, therefore, is going to be less accurate than one that includes people who have a verifiable history of actually getting out there and voting.

If the Dem advantage in newly-registered voters pans out -- meaning they actually vote -- the polls may be shown to have been skewed. Right now, however, if they're polling "likely voters," they probably ARE NOT polling the newly registered.

I've posted on this exact subject several times.



Tansy Gold

[rch35]

2. thanks very much nt

[grantcart]

3. good question and one that is debated among pollsters

The real problem however is not in identifying 'likely voters' as much as it is how much weight should they give the different parties when one party has added so many new registrations.

Part of the complication is that newly registered voters have widely different patterns of turn out. Typically voter registration drives were powered by a bounty paid by the party involved. Often times that meant that the newly registered voter wasn't particularly motivated to actually vote and the turn out on newly registered voters was not high.

That changed in 1993 when a skinny kid in Chicago registered hundreds of thousands and got them motviated to vote without any bounty.

The question for pollsters in a state like Florida, for example where the Obama campaign has identified 1.5 million registered voters who did not vote in 2004 and another 500,000 newly registered voters. How many of these will actually vote?

So the pollsters have no idea what percent of Democrats they should include in their weighted polls.

[endthewar]

4. Basically to sum it up

A pollster decides what the race "feels" like to them, then they adjust the likely voter internals to match their preconceived prejudice.