Poll roller coaster continues, with SUSA showing Clinton holding firm with 12 hours to go

Whatever the results tomorrow, there will be a group of pollsters that will come out the clear losers of the run-up to Super Tuesday. On the one side we have groups like Zogby and Rasmussen which show Obama with higher-than-average results, on the other are Mason-Dixon, ARG and SUSA which are much much more reassuring to Clinton's chances. Depending on which institute has released the latest wave of polls, Clinton and Obama supporters are trading optimism for depression and vice-versa.

What can these discrepancies be attributed to? It is important to note that each institute is consistent in showing an Obama uptick or a Clinton uptick: Zogby has great Obama results in California but also in Missouri and New Jersey; and Mason-Dixon and SUSA have Clinton up higher than expected in most of their polls as well. This suggests that pollsters are using widely differing turnout models -- not only with the composition of the electorate but how many new voters show up and who these new voters are. Don't necessarily assume that new voters favor one candidate or the other, as both Iowa and New Hampshire Democrats voted in record numbers... to differing results. At this point there is no way of knowing which pollster we should trust, and we will have to wait for the results to come in tomorrow.

Meanwhile, here are the latest trends, mostly based on SUSA's massive release of polls. First up, of course, is the big prize of California, the state for which we could stay up very late tomorrow night:

• SUSA has Clinton up 53% to 41% in a week-end poll, which is a slight improvement for the NY Senator who led 49% to 38% last week. The key finding of the poll is that 34% of voters have already voted, and Clinton leads by 12% among those as well. The second crucial internal from the poll is that Latinos (27% of the electorate here) are giving 71% of their vote to Clinton.
• The Republican numbers are in line with other numbers and have a toss-up between McCain at 39% and Romney at 36%. A week ago, McCain had 36% and Romney had 25%, so this is a significant progression for Romney. Among those who have already voted (35%), Romney is down 5%.

I have long emphasized that early-voting was a big hurdle that Obama and Romney would have to overcome. And while there is no question that many of the people who voted Clinton early would have voted Hillary anyway, there is little doubt that some voters who are now locked in could have been susceptible to this late Obama momentum.

Next, we get three polls from Clinton and Obama's home states:

• In New York, SUSA shows Clinton staying ahead 56% to 38%, a bigger margin than in some other polls we have seen. Clinton really needs to get a massive delegate lead here to offset possible losses elsewhere. Worth noting is that Obama wins the male vote, in what is a 44% gender gap!
• In PPP's survey, Clinton has a similar lead, 51% to 32%.
• The GOP numbers have McCain up 56% to 23% in SUSA and 49% to 24% in PPP.


• In Illinois, it's Obama that is widely ahead in the SUSA poll 66% to 30%. Clinton only gets 17% of the white vote.
• Among Republicans, McCain does not tremble, 46% to 25% for Romney and 17% for Huck.

The Democratic numbers in these states are actually very important, as Illinois and New York will attribute a large numbers of delegates and both candidates want to get a lead in their home states to offset potential losses elsewhere. In New York, Clinton has a robust lead in all regions of the state, but Obama is also coming in consistently above 30% which should allow him to split delegates equally in all districts that allocate an even number of delegates! Clinton's Illinois position is a bit more precarious, as she is weak enough overall that she could let Obama get two extra delegates in even districts. And don't be surprised if Clinton gets shut out in some districts of Chicago.

Completing the Northeast picture, SUSA releases final numbers from New Jersey, Connecticut and Massachusetts. Polls from all three have shown tight races in the past week but SUSA partly disagrees:

• In New Jersey, SUSA gives Clinton a reassuring 52% to 41% lead, in line with another poll released this morning that showed Clinton holding firm in the Garden State. The gender gap is 42%.
• In Massachusetts, SUSA has Clinton surviving 56% to 39% thanks to a 62% gender gap!! SUSA does show a 7% Obama progression in a week.
• In Connecticut, however, it's Obama, 48% to 46% (he led 48% to 44% a few days ago) in a race with a small gender gap, small age gap and small geographical differences.
  


• All those states show no suspense in the GOP race, and only MA is not winner-take-all and so the margin has to be watched. In NJ, it's McCain 54-25; in Connecticut, 52-30. And in Massachusetts, Romney is up 58-30.

In many ways, Clinton would rather lose Massachusetts than Connecticut. Losing the former could be attributed to Kennedy's influence, while Connecticut is in Clinton's own backyard. And with that we move to Missouri, a swing state in both parties in which SUSA once again is kind to the New York Senator:

• Clinton is up 54% to 43%, a noticeable progression since SUSA's poll released earlier this week in what SUSA calls "Last Minute Momentum to Clinton." It had been a while Hillary's camp hadn't read those words.
• The GOP 's Missouri primary might be the single most important contest in the Republican race, as the winner will get all of the state's 58 delegates. And as in many other polls, SUSA has it too close to call, with McCain at 33%, Huckabee 31% and Romney 28%.

Both Obama and Romney really want to win this state. A Clinton victory here would allow her to show she has appeal in the Midwest and that she can attract voters in a purple state that has been leaning red. And a McCain win would virtually leave Romney no path towards keeping the delegate count close.

And with that we move South, to Georgia and Alabama which are holding very tight contests as well. In Alabama, it is the Democratic race which is key now that Edwards's withdrawal has opened up a path for Clinton to win if the electorate in the South stays as racially polarized as in SC. We got three polls from AL:

• SUSA shows Obama up 49% to 47%, CPR has Clinton up 48% to 44% and Insider Advantage shows a toss-up, with Obama at 45% and Clinton at 44%.
• Among Republicans, AL has long been a McCain-Huckabee contest. SUSA has McCain stumbling a bit in the past few days and now up 37% to 35%, CPR has it at 36-30.


• In Georgia, PPP confirms Obama's lead 53% to 37%, but it is the GOP race that is more important in this state. PPP pits it at 31-29-27 (McCain-Romney-Huckabee), with Insider Advantage at 32 McCain-31 Romney-26 Huckabee.


• The latest updates from Tennessee and Oklahoma (PPP and SUSA) respectively confirm that Clinton has a wide lead there (56-34 and 54-27) but the GOP race is closer (34-28 for McCain-Huck in TN and 37-32 between them in Oklahoma).

As always, these numbers tell us first and foremost that the delegate count will probably stay close in the Dem race. And odds are that, if one candidate emerges with a significant delegate lead, it will be because of the margins in New York and Illinois rather than because of who won Missouri, California and Connecticut by a point of two.

• General election poll
  

A quick note about RAsmussen's latest general election numbers which are as over the place as the primary polls: McCain leads Clinton 47% to 39% but ties Obama at 44%. But the electability is reversed in Romney's matchups, as Clinton crushes him 50% to 37% and Obama is much closer, 44% to 41%. Rasmussen had only released Clinton-McCain and Romney-Obama yesterday and the rest today, so it appears that those two match-ups were tested in one sample, and that Clinton-Romney and McCain-Obama were tested in another. Simple sample differences would then explain why the first two are much more favorable to Democrats than the last two. I don't know, however, why such a sample split would be a good idea.