How useful is the concept of "likely voter" right now?

By now many have heard of the USA Today/Gallup pollthat shows McCain leading among “likely voters” by 4 even though he is behind in the same polled group of all registered voters by 3. Which means that unlikely voters (per Gallup’s methodology) are disproportionately likely to be Obama supporters… 8 times more likely.

Who are likely voters and how useful is it to attempt to identify them at this time?

Here is Gallup’s explanation.

Is it a useful exercise right now? My sense is not so much and that in this election in particular predicting who will vote will be very difficult. What do you think?

For the interested here is a discussion of the seven question likely voter model and it limitations/usefulness in the next linked installments.

Doesn’t ths kind of polling, conducted through landlines, risk undercounting young people who may be inclined to vote Obama but only have cellphones?

Gallup polls cellphones if they determine that you don’t have a landline. I don’t know about the other firms.

From the Gallup Daily Tracker:

Pollsterhas an article about Gallup starting it back in January.

A “likely voter” is, among other things, a voter unlikely to get purged.

However, it would not include the massive voter signups that the Obama campaign has been doing for the last year or so.

At least the cell phone only issue would effect the likely voter and the registered voters total similarly, if not the same.

The bigger issue is that even the seven question list to determine likely voters makes it virtually impossible for a first time voter to make the tope 60% cut. That 60% will be filled up almost if not completely with perfect scores (according to the link and following installments in my second post) and first time voters really cannot get perfect scores. The three question method is similar: how many first voters can answer " How often would you say you vote – always, nearly always, part of the time, or seldom?" with “Always”?

The model may do well in typical election cycles as younger voters are also typically less likely voters most cycles. And of course there are other groups that traditionally do not vote as often - such as Hispanic voters, for example. In an election cycle that is typical, the past performance of these groups will be at least moderately predictive of the future performance.

The problem comes in when the there is a reasonable possibility that those particular demographics won’t behave typically, when in fact, they have not been behaving typically during a primary season.

Is it more predictive to use the behavior of those demographics in past Presidential elections, or the behavior in the current election cycle so far, when the two are significantly at odds with each other.

In other words, if one believes that when all is said and done, younger voters will forget to vote that day, either out of apathy or forgetfulness or having something better to do, and that Hispanic voters will have typical turnouts in the general in contrast to their record primary turnouts, and that older Conservative voters will indeed come out in typical numbers, then the Gallup model may not be too far off. If, however one believes that those are not reasonable assumptions, and believes instead that the recent behavior of those demographics is more predictive of their behavior in this general than the behavior of those demographics in past general elections, then their model is likely very very far off.

This election will indeed turn on who comes out to vote that day.

Voters are only considered “likely” if they say they’ve voted in the past. “Likely voters” does not include newly registered voters, which Obama has in abundance. The average of all polls has shown Obama with a small but consistent lead of about 3 points, and with a fairly decisively lead on the electoral map. I think that’s probably about accurate. I hear a lot of pundits saying they think Obama should have a bigger lead, but I disagree with that. There really isn’t all that much flex in the electorate. Most people default to one side or the other no matter what. True swing voters are a pretty small group. No one’s going to run away with anything anymore. Like DSeid says, it’s going to be about turnout, and, for once, the kiddies might show up.

So here’s an example of what I’m talking about. Look at the way these polls fall out now for the major polls since 7/23.

“Likely voter” screens used: USA Today Gallup McCain +4; Rasmussen latest McCain +1 (and same yesterday). McCain +2.5 avg.

Registered voter results with no screen for “likely voter”: Pew Obama +5; CNN Obama +7; Gallup Obama +4 (was +3 yesterday); AP-Ipsos Obama +6. Obama +5.25 avg.

Notice a pattern?

For current national numbers the difference between using “likely voter screen” or not is 7 to 8 points. Consistent with the USA Today Gallup results of one polled population with and without screen on (a 7 point swing).

Now again, if the “likely voter” screens are reliable devices at this point in time for this particular cycle then those results bode well for McCain as Obama’s support is disproportionately among those who will stay home. If the screen is inaccurate this cycle or at least used at this point in the process then Obama’s lead is consistent.

In my state, even registered voters would miss some voters as you can register at the polls when you show up to vote.

Yes and no, but mostly yes. Both campaigns are about getting their list of folks out to vote. New voter registrations tend to follow Democrats, so it also is worth mentioning that if more people get registered, that favors Dems until the minority groups act as they traditionally do and not vote (that includes the coveted 18 to 25 demographic).

Getting folks out to vote will be the key.

Some interesting information relevant to this discussion buried in the most recent CBS national poll (I got to it through here and then clicking on the CBS link which downloaded a pdf).

Not the fact that it is another registered voter poll that shows Obama +6, and not the fact that Obama’s supporters are marginally less likely to say that they may change their mind, but the very sizable enthusiasm gap: “The enthusiasm gap remains: Obama’s supporters are three times as likely as McCain’s to be enthusiastic about their candidate.” And McCain supporters are three times as likely as Obama’s to be dissatisfied about their candidate.

To be precise 43% of Obama supporters are enthusiastic about their choice while only 13% of McCain supporters feel enthusiastic about their choice. 17% of McCain supporters are dissatisfied about their choice while only 6% of Obama supporters are dissatisfied with theirs.

6 out of 10 of both Democrats and Republicans are a lot of attention to the campaign, more than did in 2004 until a month before the election.

What these numbers say is that Obama supporters are likely voters if such is measured by some combination of enthusiasm vs disatisification with their choice and amount of attention being paid to the process. It leaves only the previous voting question as the cause of the selective filtering out of Obama supporters by likely voter screens.