Tuesday, October 21, 2008
How accurate are polls? In 2004, VERY
Obama (for whom I voted a few days ago - WOO!) is, as of today, in the lead in every national poll - although by remarkably divergent amounts.
He also is far enough in the lead in various state polls that Real Clear Politics considers him "safe" in states that would give him 259 electoral votes - and "leaning" in another 3 states equaling 27 electoral votes. So before you even look at the "toss-up" states (an additional 6 states with 89 electoral votes), he's up by enough to win the election. And he's leading in 5 of those 6 toss-up states!
So when I saw this excellent video today from Ohio, I got a bit concerned. Was it really true, what they said in their admonition for us not to get complacent (and to give money), that Kerry was up in the polls in 2004?
Well, not really. I mean, yes, he did lead at some point. But as Oliver Willis points out, once the GOP convention happened, Bush never *really* looked back. Yes, the race *did* tighten once October rolled around, and Kerry was ahead in a few polls (and it was a two-point game at the end), the Democrat was always playing catch up.
Check it out:
And looking at RCP's state tallies from 2004, some remarkable data comes to light.
In 2004, RCP identified 17 states as "battleground" states. (I'm leaving out Hawaii, which was never really in doubt.) With the exception of Wisconsin, the winner in each state matched the projected winner according to the polls! And Wisconsin, of course, ended up breaking for the Democrat.
Not only that, but state by state, the election results tracked very closely to the polls. In Ohio, they got the margin of 2.1% exactly right. In 9 other states, they got it right to within 1%. And in two others, they were within 2%. The states pollsters were "right" in included OH, PA, WI, IA, MN, MI, NH, ME, and NJ - aka "Northern" states - plus OR, NM and CO (good ol' Western states).
Now, what states did the pollsters get wrong (i.e. 3% or worse difference)? Florida, Missouri, West Virginia, and Arkansas - all of which broke hard for Bush - plus Nevada (a Western state), which went from a 6.3% Bush poll victory to a 2.6% Bush electoral victory. And still *none* of the states broke more than 5% from the polls.
What this suggests to me is that Bush's vaunted "ground game" in 2004 was really only effective in the South - aka his base.
Assuming polls stay more or less stable, and are anywhere close to as accurate as they were in 2004, Obama wins this election comfortably. Even if there's a significant shift of 5% nationally towards McCain in the final two weeks, Obama'll win this. And what if there's a significant shift of 5% nationally - in every state - AND an additional 3% movement in EVERY "battleground" state toward McCain (thanks to a church-based stealth ground game, a la 2004)? This would give McCain FL, OH, MO, NV, CO, IN, and NC. Well, then it would be Obama 264, McCain 261, and a virtual tie in Virginia. And I gotta say, given Warner at the top of the ticket, Webb helping out in his way, and improving demographics for the Dems over the past decade, I like those chances! Oh, plus Nancy Pfotenhauer's "real Virginia" gaffe:
He also is far enough in the lead in various state polls that Real Clear Politics considers him "safe" in states that would give him 259 electoral votes - and "leaning" in another 3 states equaling 27 electoral votes. So before you even look at the "toss-up" states (an additional 6 states with 89 electoral votes), he's up by enough to win the election. And he's leading in 5 of those 6 toss-up states!
So when I saw this excellent video today from Ohio, I got a bit concerned. Was it really true, what they said in their admonition for us not to get complacent (and to give money), that Kerry was up in the polls in 2004?
Well, not really. I mean, yes, he did lead at some point. But as Oliver Willis points out, once the GOP convention happened, Bush never *really* looked back. Yes, the race *did* tighten once October rolled around, and Kerry was ahead in a few polls (and it was a two-point game at the end), the Democrat was always playing catch up.
Check it out:
- In all October polls prior to the final polls from each pollster, Bush led in 25, Kerry led in 3, and 8 were ties.
- In the final polls, Bush led in 10, Kerry led in 2 and 2 were ties.
- Bush's biggest October lead in any poll was +8. Kerry's biggest lead in any October poll was +2.
- Bush's final margin of victory was very, very close to what the combined polls said they'd be.
- Of all October polls, Obama has led in 34, McCain has led in 0, and 0 have been ties.
- Obama's biggest October lead in any poll is +14. His smallest lead is +1.
And looking at RCP's state tallies from 2004, some remarkable data comes to light.
In 2004, RCP identified 17 states as "battleground" states. (I'm leaving out Hawaii, which was never really in doubt.) With the exception of Wisconsin, the winner in each state matched the projected winner according to the polls! And Wisconsin, of course, ended up breaking for the Democrat.
Not only that, but state by state, the election results tracked very closely to the polls. In Ohio, they got the margin of 2.1% exactly right. In 9 other states, they got it right to within 1%. And in two others, they were within 2%. The states pollsters were "right" in included OH, PA, WI, IA, MN, MI, NH, ME, and NJ - aka "Northern" states - plus OR, NM and CO (good ol' Western states).
Now, what states did the pollsters get wrong (i.e. 3% or worse difference)? Florida, Missouri, West Virginia, and Arkansas - all of which broke hard for Bush - plus Nevada (a Western state), which went from a 6.3% Bush poll victory to a 2.6% Bush electoral victory. And still *none* of the states broke more than 5% from the polls.
What this suggests to me is that Bush's vaunted "ground game" in 2004 was really only effective in the South - aka his base.
Assuming polls stay more or less stable, and are anywhere close to as accurate as they were in 2004, Obama wins this election comfortably. Even if there's a significant shift of 5% nationally towards McCain in the final two weeks, Obama'll win this. And what if there's a significant shift of 5% nationally - in every state - AND an additional 3% movement in EVERY "battleground" state toward McCain (thanks to a church-based stealth ground game, a la 2004)? This would give McCain FL, OH, MO, NV, CO, IN, and NC. Well, then it would be Obama 264, McCain 261, and a virtual tie in Virginia. And I gotta say, given Warner at the top of the ticket, Webb helping out in his way, and improving demographics for the Dems over the past decade, I like those chances! Oh, plus Nancy Pfotenhauer's "real Virginia" gaffe:
Labels: 2004, Pfotenhauer, polling
