Gallup: Traditional vs Expanded? DU Statistics experts, which is a better model?

[Quixote1818]

It seems to me they realize the dynamics have probably changed and so the expanded poll numbers are probably the more accurate. Anyone have any strong opinions on this?

Here is what Gallup writes:

Gallup's likely voter scenarios show differing patterns. If turnout in this year's election follows traditional patterns by which the voting electorate skews towards those who usually vote as well as those who are interested in this year's election, the race is a close one, with Obama holding on to a two percentage point margin, 49% to 47%. If a much higher than usual proportion of new voters turn out, thus increasing the potential impact of groups of voters traditionally less likely to vote, such as young adults and minorities, Obama has a six-point lead, 51% to 45%. -- Frank Newport

http://www.gallup.com/poll/111241/Gallup-Daily-Little-Impact-From-Debate-Far.aspx

[IndyOp]

1. All 3 are informative - they make the point that Obama has been making publicly -

what happens in the election is ALL about the turnout.

If only those people who regularly vote show up then Obama really could lose (Gallup's Likely Voters - Traditional).

If the people who currently plan to vote (Gallup's Likely Voters - Expanded) really do get out and vote then Obama is in much better shape, but with determined disenfranchisement of voters and electronic vote fraud on behalf of the GOP, then Obama could still lose.

If everyone who is registered - with an especially good turnout from the youth and minorities then we can overcome anything the GOP tries in order to steal the election. (See Gallup's Registered Voters model).



It all depends on turnout.

[mvd]

3. I think both can be used but model "expanded" is closer

There will be a bigger turnout, but you can subtract a little for disenfranchisement and the fact that not all turn out.

[Traveling_Home]

2. Just Keep Working!!!!! nt

[TheCoxwain]

4. ****************Here is the difference********************

Traditional : McCain Loses Badly

Expanded: McCain Loses Really Badly

[writes3000]

5. Look at early voting for signs of which model is more on target.

The early voting all indicates that this election will be very, very different from elections past. Gallup's "traditional" model will not be accurate if these trends continue.

[RoyGBiv]

6. I'll let you know Nov. 5th ...

The point is, no one actually knows at this point which is a better model. There are opinions and theories, but they need testing, and that's what the election itself will help to do.

[Thrill]

7. How can you use traditional in an election with so many new voters?

[Essene]

8. Here's the truth. Nobody can predict turn-out so there's a ton of assumptions being made

The most cautious approach is to use Registered voters and to just simply admit the truth that turn-out is unknown.

These "likely voter" methods use very debatable weightings based on what they think will be real voter patterns. Even if their assumptions and methods nail an election in 2004, there is no definite reason to assume the same assumptions apply to 2006 or 2008.

The MAIN PROBLEM with this approach is that they confuse excitement with political leaning.

In other words... if you are trying to flesh out your sample categories with strong "likely voters" that means you are putting aside those who may simply be really angry and thus report they are not as likely to vote.

Thus, when you have a situation where a party's base is low energy and then suddenly events dramatically increase their excitement level... you can see this skew your "likely voter" results because it confuses political leaning trends with excitement to vote.

Gallup tries to justify their continued use of using their own likely voter assumptions because it better predicted a Bush victory against Kerry.

The fact is... they cannot predict turn-out... and they know it.