Below is a chart of all reasonably current polls in Pennsylvania. On one axis, I've plotted Clinton's lead in each poll, and on the other, the number of undecided voters in that poll.
As you can see, there is a strong, if not overwhelming relationship between these two things: Obama tends to do better where there are more undecideds, and Clinton tends to do better where there are fewer. The relationship is statistically significant at the 95 percent level.
If we trace the regression line such that it crosses the y-axis -- meaning, there are zero undecideds -- we'd project a Clinton win by 16 points.
Is this a valid way to look at the polls? I have no idea. But we can run a gut-check of sorts. The average of these 14 polls is: Clinton 48.3, Obama 40.9, Undecided 10.0. For Clinton to win by 16 points on election day, that would imply results of Clinton 57.5, Obama 41.5, assuming that 1 percent of the vote goes to minor candidates.
In order for that to occur, Clinton would need to pick up 9.2 points from undecided voters, to Obama's 0.6. In other words, she'd have to win nearly every undecided voter. Even when elections break at the end -- they don't break that strongly. Maybe Clinton could win 2 out of 3 undecided voters (which would imply a victory margin of about 10 points), but not 9 out of 10.
So -- I wouldn't take these numbers all that seriously. At the same time, I think there is a case to be made that Clinton has a couple extra points worth of cushion in her Pennsylvania numbers versus what the polling averages currently imply.
Saturday, April 19, 2008
Are Undecideds leaning Clinton in Pennsylvania?
-- Nate at 10:00 AM
Labels: depth of support, pennsylvania, primaries
Subscribe to:
Post Comments (Atom)
15 comments
I think it is true that "Last-Minute-Deciders" tend to vote CLinton- we saw that in Ohio, Texas, Cali [I know, many Clinton votes were mail-in ballots, but nevertheless she gained ground in the last 60 hours before the election, and we saw the same in NH- I don´t think her surge there had much to do with that event where she started crying.] and New Hampshire.
So I think we will see a 9-10 point lead for Clinton on Tuesday.
Another thought: Could you put together such a graph for McCain in some toss-up GOP primaries and project how he is doing with undecided voters?
Maybe this is something we can include in our sheets... that a poll with many undecided voters loses weight or that you do not take the Democrat/Republican lead but their percentage and give McCain 60% of the undecideds and 50% against Clinton, when we should see that McCain is able to get "Last-Minute-Voter" support...
We also see, that Obama improved again in states where the primaries are over- NH is a good example again, New Jersey and California, too.
It seems that the undecided voters decide to vote for Clinton in the Primary (or stay at home) and fall back to their undecided status again, then.
Hm, but looking at the pollster.com graphs for Missouri or Georgia I don´t see that. He is actually doing better in polls with many undecided voters, like Obama, I think- Huchkabbe did better with those voters, Romney too...
What are the numbers without ARG? That and PPP look like outliers, and ARG is low-quality.
I think that you are right but I hope that you are wrong. Hoping that Obama can win Philly by a large enough margin to keep Clinton below 10 points statewide.
You've already made a gut-check in your demographically based prediction of the PA outcome -- 12 points.
One problem with both methods of estimation is that I don't think you've explicitly taken into consideration whether it's an open or closed primary. The closed primary in this case probably favors Clinton, too, since almost all of those "undecideds" are probably (long-term) registered Democrats who may have been "thinking" about Obama but many will "come home" to Clinton if they think she's a safer choice.
OTOH, if you also consider what you know about the tendency of black voters to say "don't know" when polled, then if you can separate out the white from black "undecideds" there could be more room for optimism for Obama.
One other factor that you may not have considered in your demographically based model is the "new voter" phenomenon. Although history tells us not to overstate the turnout or the impact of such new voters (they really didn't show up for Kerry, for example), if the newly registered Democrats are large enough and motivated enough in PA (stories suggest it's a huge number), then this may give Obama a couple of percentage points. Do such data exist? (In the PA case the new Dem registrants are likely to be committed in the polls and have registered "with a purpose." OTOH if a substantial percentage are African Americans many will still show up as "undecided" in the polls.
This is off topic but I would be intersted in knowing what you all think of the blip on the daily gallup towards Clinton?
Random noise. We had that some days ago for Obama, and that 10-point lead did not hold for long...
There will be no momentum until Wednesday.
Actually, If I recall correctly, Obama won California among voters who decided in the last 72 hours. If not for mail-in votes, he would have been very close.
According to the Exit Poll data at CNN 26% decided in the last 72 hours and the break was 1% in favor for Clinton 48/47
http://www.cnn.com/ELECTION/2008/primaries/results/epolls/index.html#CADEM
But in the last DAY, that was 14% of all voters and 51/43 in Clinton's favor
When analyzing past performance, you also have to take into account the reasons for the pattern, if possible. Throughout most of this campaign, Hillary has been the "default" candidate. She also had everybody sold on her experience. The two competing messages were "experience" vs "change". When people are confronted with a choice and they have to make it now, people will more frequently go with the safer choice, which in this case is Clinton with the experience message.
Over the last month, that experience message has weakened considerably. Additionally, the inevitability of Obama is beginning to sink in. I read the article about one of the recent PA polls and something like 75% of Clinton supporters believe Obama will get the nomination.
This could weaken the last minute effect for Clinton. Though Poblano's analysis of the polls suggests otherwise.
Another thing that changed last month was among the cross over voters. In Texas and Ohio, got a larger share of the cross over vote, but analysis of CNN's exit polls show she got something like 95% of the vote among those who thought highly of John McCain. This indicates that there were a fair number of strategic cross over voters who will not vote Democrat in the fall, but are trying to meddle with the Democratic primaries.
In PA, it takes more forethought to be a crossover voter since it's a closed primary. Still about 100,000 Republicans switched their registration to Democrat so they could vote in the primary. How many of those are strategic voters and how many are truly supporting one of the Democratic candidates? I suspect that those who take the effort to change their registration are more likely motivated positively towards a candidate (which would be Obama for most), than just trying to mess with the Democratic primary process.
The message that Clinton is mathematically finished going around, the closed primary, and Clinton's message of experience weakening, I think that the last minute voters might break with past patterns and vote Obama.
I could be wrong though.
Can one of the matth geeks (I say with affection) run the numbers to see if the following theory is true
Late breaking undecided generally go to the winner of the state?
The same thing happened in New Hampshire too, if you look at the polls taken in the last 4 days before the election. But in that case, the undecideds seemed to be favoring Obama by similar margins. In fact, that is one of the things that pushed Obama's final pre-election numbers so high.
What happened, of course, is that Obama lost New Hampshire. Those polls that showed the **most** undecideds were closest to the truth -- in other words, those polls that pushed the undecidededs hardest were the worst, and those that pushed the least were the best.
Why? Perhaps because undecideds are the least committed, and therefore least likely to vote. In other words, pushing undecideds screws up the pollsters' own likely-voter models.
Just a theory. But if true, Clinton's true lead may be more like 2 points than 16.
On the other hand, could pissing off Democratic party activists and MoveOn.org members cost her a few points? Or has she already lost them?
I understand that about 1 Mil MoveOn members voted to endorce Clinton. PA accounts for about 3.5% of the US population. Hard to say whether it would be over or under represented in MoveOn. So suppose there are 35,000 Clinton MoveOn supporters in PA. That is quite a lot - a bit over 2% of the turnout projects here are correct. Obviously most of these are not going to charge their vote based on Clinton's comments, but it seems possible that she could have lost up to a few thousand votes this way. This is not enough to explain shifts in the polls, but enough to make a slight difference on Tuesday.
Does it really make sense to follow the regression line all the way to zero undecideds? Isn't it more likely that some of the undecided voters will opt to simply stay home?
If you assume that the polls (ARG, SUSA, Muhl and Quinn) with the lowest percent of undecideds are showing this floor, at around 5%, then the regression line comes in at around a 12 point Clinton victory, which would match your earlier prediction based on demographics.
Post a Comment