Breaking Down Swing State Exit Polls: When Voters Decided

In this post, we are going to look through swing state exit poll data relating to when voters decided which candidate to support.

National Exits

Election day
(2004, 5%) Kerry 52, Bush 45
(2008, 4%) Obama 50, McCain 45

Last three days
(2004, 4%) Kerry 55, Bush 42
(2008, 3%) McCain 52, Obama 47

Last week
(2004, 2%) Bush 51, Kerry 48
(2008, 3%) McCain 50, Obama 48

In October
(2004, 10%) Kerry 54, Bush 44
(2008, 15%) Obama 54, McCain 43

Before October
(2004, 78%) Bush 53, Kerry 46
(2008, 74%) Obama 52, McCain 47

Analysis. Conventional wisdom (or at least just Republican postgame spin) has said that the financial crisis took a race that was veritably tied in late-September and gave it to Obama. The data says otherwise. Granted, the race was tighter prior to the Wall Street meltdown; Obama only led by by five points with voters who made a decision before October, and by 11% with voters who made a decision during October. Still, given that three-fourths of voters decided before October, McCain was likely in dire conditions anyway. In other words, even though the election may have turned out close had Lehman Brothers never collapsed, I think McCain still would have lost. Heck, he lost the voters who decided on November 4, and barely won voters who made a choice the last week. However, let me add that 50% more voters made a decision in October this year than in 2008. There is little doubt in my mind that the crisis had something to do with that, much to the McCain's detriment. Going further, just note how much better Obama did than Bush: in only one subset -- the final three days -- did McCain enjoy a lead outside the margin of error. Whereas Bush opened strongly enough to hold on to an overall 51-48 victory, Obama's campaign opened and closed strong -- a very impressive feat.

Colorado

Election day
(2004, 6%) Kerry 51, Bush 45
(2008, ?) ?

Last three days
(2004, 3%) ?
(2008, 0%) ?

Last week
(2004, 2%) ?
(2008, 3%) ?

In October
(2004, 12%) Bush 50, Kerry 49
(2008, 12%) Obama 53, McCain 45

Before October
(2004, 77%) Bush 55, Kerry 45
(2008, 82%) Obama 53, McCain 46

Analysis. Unfortunately, the small sample sizes give us no data for voters who decided late in the game. Nevertheless, we see that Obama had Colorado locked up before October. Even after the crisis broke and McCain foolishly "suspended" his campaign, he was already cooked in the state, In fact, Obama's level of support with pre-October deciding voters was just about identical to his support among voters who made a decision in October. In a state Kerry was trounced, Obama did spectacular.

Florida

Election day
(2004, 6%) Kerry 53, Bush 44
(2008, 3%) McCain 52, Obama 45

Last three days
(2004, 3%) ?
(2008, 3%) McCain 51, Obama 48

Last week
(2004, 3%) Kerry 68, Bush 29
(2008, 4%) Obama 55, McCain 42

In October
(2004, 12%) Kerry 59, Bush 40
(2008, 12%) Obama 60, McCain 38

Before October
(2004, 77%) Bush 56, Kerry 44
(2008, 76%) Obama 51, McCain 48

Analysis. This was a close one, and McCain closed here late in strong fashion. The data seems to indicate that this is a state that was won because of the economic crisis. Why? Prior to October, three-fourths of the voters had decided, and Obama held a three-point lead with them. However, October-deciders broke for Obama by over 20 points, ensuring that McCain's late charge would not be enough to come back. This did not occur in 2004, where, despite the fact that people seemed to decide in the same percentages in October and pre-October, Bush had a huge lead with the latter group -- more than enough to counter-balance the +19% Kerry advantage with October voters.

Indiana

Election day
(2004, 5%) ?
(2008, 9%) Obama 50, McCain 46

Last three days
(2004, 4%) ?
(2008, 5%) Obama 50, McCain 50

Last week
(2004, 3%) ?
(2008, 5%) Obama 50, McCain 45

In October
(2004, 11%) Kerry 56, Bush 43
(2008, 13%) Obama 55, McCain 44

Before October
(2004, 77%) Bush 65, Kerry 34
(2008, 67%) McCain 50, Obama 48

Analysis. The numbers that stand out here are the scores of people who remained undecided late in the game: in the last week of the campaign, about one in five voters were undecided, nearly double the national average. What happened here was that Obama was down prior to October, though not substantially, and as a result he wisely decided to spend big here. The economic crisis gave the Democrat a good cushion, and those voters who rarely went for a national Dem broke for him by just enough to award Obama a tiny statewide victory in one of the most unlikely of states.

Nevada

Election day
(2004, 8%) Bush 49, Kerry 48
(2008, 4%) Obama 53, McCain 47

Last three days
(2004, 3%) ?
(2008, 3%) Obama 50, McCain 44

Last week
(2004, 3%) ?
(2008, 3%) McCain 48, Obama 44

In October
(2004, 13%) Kerry 57, Bush 40
(2008, 14%) Obama 59, McCain 40

Before October
(2004, 72%) Bush 54, Kerry 45
(2008, 74%) Obama 56, McCain 42

Analysis. Nevada was an old-fashioned beat-down. Obama organized the heck out of the state, and McCain never had a chance, even before the Wall Street crisis. Needless to say that the pre-October polls showing a real horse-race were not terribly accurate.

North Carolina

Election day
(2004, 4%) Kerry 71, Bush 26
(2008, 3%) Obama 55, McCain 42

Last three days
(2004, 4%) Kerry 57, Bush 43
(2008, 2%) Obama 54, McCain 42

Last week
(2004, 3%) ?
(2008, 3%) McCain 67, Obama 32

In October
(2004, 11%) Kerry 50, Bush 48
(2008, 14%) McCain 60, Obama 39

Before October
(2004, 78%) Bush 60, Kerry 39
(2008, 77%) Obama 52, McCain 48

Analysis. Again we find another example where Obama was in strong shape before October. While his lead was fairly narrow, most of his work was done prior to October 1. McCain closed very strong in October and the last week -- the former period is interesting as the economic crisis did not deliver a new majority to Obama over October -- but Obama countered over the final days.

Ohio

Election day
(2004, 5%) Kerry 61, Bush 39
(2008, 5%) Obama 48, McCain 44

Last three days
(2004, 4%) Kerry 64, Bush 36
(2008, 3%) McCain 52, Obama 43

Last week
(2004, 3%) ?
(2008, 4%) McCain 52, Obama 48

In October
(2004, 11%) Kerry 61, Bush 36
(2008, 16%) Obama 51, McCain 47

Before October
(2004, 78%) Bush 55, Kerry 45
(2008, 70%) Obama 53, McCain 46

Analysis. Like in North Carolina, Obama actually did better pre-October than in October. Food for thought. And like in Indiana, we had more undecided voters after September than in many other states. It's likely that a lot of voters in Ohio could not commit to Obama, and many of the late deciders, didn't.

Virginia

Election day
(2004, 5%) Bush 48, Kerry 48
(2008, 5%) Obama 49, McCain 49

Last three days
(2004, 3%) ?
(2008, 5%) McCain 67, Obama 30

Last week
(2004, 3%) ?
(2008, 4%) ?

In October
(2004, 8%) Bush 52, Kerry 45
(2008, 11%) Obama 59, McCain 40

Before October
(2004, 81%) Bush 56, Kerry 44
(2008, 74%) Obama 53, McCain 46

Analysis. Obama did most of his hard work in Virginia before the meltdown, but the event certainly clinched the state for him.