Daily Kos/SEIU State of the Nation poll: Despite switch to likely voter model, Obama still up by 6

[ProSense]

Daily Kos/SEIU State of the Nation poll: Despite switch to likely voter model, Obama still up by 6

by David Nir

Public Policy Polling for Daily Kos & SEIU. 9/7-9. Likely voters. MoE ±3.1% (8/23-26 results):

Q: If the candidates for President this fall were Democrat Barack Obama and Republican Mitt Romney, who would you vote for?

Barack Obama: 50 (44)
Mitt Romney: 44 (44)
Undecided: 6 (6)

So where's Obama's convention bounce? You're looking at it! Allow me to explain. This week, Daily Kos and SEIU made an important change to how we conduct our national polling. Specifically, we switched from a registered voter model to a likely voter model—meaning we're now confining our pool of interviewees to those that our pollster, Public Policy Polling, thinks are "likely" to vote in the November elections. Typically, the switch from RVs to LVs is made relatively late in the cycle, because asking someone if they're likely to vote when election day is, say, a year off is a very dicey proposition. (Do you know what you're going to be doing a year from today?) So we felt that changing to likely voters after Labor Day made sense.

This means that our polling from this point forward isn't directly comparable to the surveys we conducted before the holiday. But it doesn't mean all comparisons are out of bounds, and that's where the "bounce" comes in. Nate Silver helpfully explained what switching to LVs typically means in presidential polling:

In the past six presidential election years, the shift to likely voter models has always helped the Republican candidate, but the difference has also always been small, usually amounting to a net of one or two percentage points in the margin between the two candidates.

This research suggests that had we not made the changeover and were still conducting polls among registered voters, Obama might instead have a 7- or 8-point lead. Put another way, the fact that we're now using a more pro-Republican screen nevertheless didn't help Romney. We can't know these things for sure, of course, but I think it's reasonable to conclude (particularly when viewed in light of other polling from other firms) that Obama did indeed get a small bump heading out of the conventions.

- more -

http://www.dailykos.com/story/2012/09/11/1130271/-Daily-Kos-SEIU-State-of-the-Nation-poll-Despite-switch-to-likely-voter-model-Obama-still-up-by-6

[frazzled]

1. The definition of "likely" voters seems very subjective

and varies widely from polling outfit to polling outfit.

Mostly, I ignore the polls until the last few weeks, when they all adjust everything to be truly predictive. Because that's the point at which none of them wants to be wrong. They'd be out of business if their last polls were wildly off. So while this is all interesting and fun, I take everything with a grain of salt until the fat lady sings (two metaphors!!). Ignore the polls and keep working!

[ProSense]

2. We're talking

Republicans and the media. Take nothing for granted. They can and have attempted to shape perception, depress the vote with their polls and reporting. Complacency plays into their hands. After 2004, I simply can't ignore them. It's getting easier to tell bullshit from reality.

IMO, if the polls are worth paying attention to in four weeks, they're worth paying attention to now. This isn't a year before the election, it's two months away.

[frazzled]

5. There are professionals to debunk the fishy polls

Nate Silver does a good job of it at 538. I'm certainly not qualified to judge a poll's reliability, and probably neither are 99.9% of DUers. We simply know the polls we "like" and "don't like."

My point was simply that "likely voter" models can mean anything. To the layman, beware. Yes, keep your eyes on the good polling analysis sites, but two months is an eternity in politics.

[ProSense]

8. True

Nate Silver does a good job of it at 538. I'm certainly not qualified to judge a poll's reliability, and probably neither are 99.9% of DUers. We simply know the polls we "like" and "don't like."

My point was simply that "likely voter" models can mean anything. To the layman, beware. Yes, keep your eyes on the good polling analysis sites, but two months is an eternity in politics.

...which is why comparing different polls isn't good practice, but a poll's trend is revealing. Nate is the person people following the polls turn to so that's still paying attention.

Two months can seem like an eternity, but it does fly by. We're already nearing the second week after the RNC. Mitt's pulling out of states and planning to blitz others, and this is likely the result of what the polls are showing.

[former9thward]

3. The main issue in polls is not likely vs. registered.

Its the mix of Rs, Ds, and Indies that are sampled. Depending on what mix the poll is using it can come up with all sorts of results. Who is correct? It will all depend on turnout of those three groups. Every pollster has to make an educated guess at the mix. No one knows what will be correct until after the election.

[ProSense]

4. True, and

Its the mix of Rs, Ds, and Indies that are sampled. Depending on what mix the poll is using it can come up with all sorts of results. Who is correct? It will all depend on turnout of those three groups. Every pollster has to make an educated guess at the mix. No one knows what will be correct until after the election.

...that's the case with all these polls. In fact, some of the RV polls were downright weird, like one of the main polls recently indicating Romney leading among women.

[ananda]

7. In that one they mostly interviewed pro life tea party women.

nt

[ProudToBeBlueInRhody]

6. Why would I care who someone would vote for if they aren't going to vote?

To me, the first questions in any poll should be....

1. Are you a registered voter?

2. Are you voting this year?

If the answer to number 2 is "no", the call should end.