Tuesday, October 21, 2008

Close Race in Florida

Barack Obama 48
John McCain 47

Florida has tightened up a little since PPP's last poll of the state, with what was a 49-46 lead for Obama now 48-47.

There has been some movement in John McCain's direction with both white voters and Hispanics over the last three weeks. McCain's margin has increased five points with whites, from 52-41 to 55-39. Hispanic voters, who have swung back and forth both in PPP's polls and those of most organizations conducting surveys in Florida, are supporting McCain 50-46 after going for Obama by the exact same margin in our previous poll.

Two of the major factors that could determine who wins the state are:

1) The proportion of the electorate that ends up being white voters relative to the proportion who end up being black voters. We are projecting white voters at 69-70% and blacks at 13-14% and right now McCain's solid support with whites and Obama's overwhelmingly strong standing with blacks are basically canceling each other out. But if the final turnout figures end up skewing either more or less African American, depending on the success of Obama's turnout operation, that will have a major impact on the race whichever way it goes.

2) Because the white and black votes are practically a combined wash right now, Hispanic voters have the power to tip the election one way or another in these closing weeks. Most polling has shown the race extremely close with that demographic. If their votes tip strongly to either McCain or Obama in the closing, the person that benefits will win the state.

I think those are the main two things to watch in the final weeks in Florida.

Both candidates are having more trouble winning over their party's voters in Florida than they are in most places. Obama gets just 80% of Democrats while McCain is earning only 82% of Republican voters. Independents are going heavily for Obama, 52-35.

McCain is up 49-47 with folks who voted in 2004 while Obama has a 64-26 lead with new voters.

Full results here.


Sreenu said...

This poll gives me hope for Obama.
The party ID advantage for the Dems shown in the latest registration statistics does not shown up in the poll (I am guessing there are several Dems that did not get called by this poll!)

Registration Statistics:
D 42%; R 36%; I 22%

This poll:
D 42%; R 42%; I 16%

Anonymous said...

Some random assorted comments:
- Do you weigh by race at all? Because looking at the wikipedia Florida entry, its population is about 15-16% African-American and 15-16% Hispanic. So you may be oversampling whites.
- Your breakdown of votes by party is rather close to SUSA; the main difference is that you project Obama to do significantly better with independents (though that's probably in the MOE because of small sample size), and your party IDs are even (SUSA had a 4-point R advantage.) Notably, however, the actual breakdown is around a 4-point Democratic advantage.

- Interestingly enough, it looks like a disproportionate percentage of those who didn't vote in 2004 are African-American or Hispanic.

Andrew H said...

Sreenu stole my thunder. As a resident of FL, I would never have guessed that 29% of the state considers itself to reside in a small town or rural area. Seems a bit high to me.

Andrew H said...

Regarding weighting, I would guess that the Hispanic number is semi-accurate because many Hispanics down here in South FL are probably not citizens yet, and lots are under 18.

Unknown said...

the dates on the poll are Oct 16-19, so it only catches one day of the Powell endorsement. Florida has a lot of military, of course.

Sreenu said...

As per the latest FL voter registration statistics:
White - 69%
AA - 13%
Hispanic - 12%
Others - 6%

This poll:
White - 69%
AA - 14%
Hispanic - 13%
Others - 3%

Tom Jensen said...

Sreenu stole my thunder. I was about to post the same statistics.

As for the party id thing, party registration and party id aren't always the same thing, especially in southern states. That doesn't mean I'm sure the breakdown in this poll is the one and only correct answer (fine with me if there turns out to be a six point Democratic id advantage among those who vote). But we don't weight to registration statistics because of that difference.

Anonymous said...

Thanks Tom. What's up next?

andgarden said...

Curiously, your respondents are overstating their 2004 support for Bush. [49/(49+43) * 100 = 53%]

Anonymous said...

Was there any significant difference in the horserace on the last day of polling (after the Powell endorsement)?

Web Statistics