Does a Lead in the Polls Mean a Win?

A poll is a (fairly) scientific attempt to take a snapshot of public sentiment at a given point in time (often performed over 2-3 days). But even accepting for a moment that a really good poll will show current sentiment, do people wind up voting the way they say they are thinking they will vote just 10-14 days prior?

Gallop does a good job digging through their historical numbers and showing that Late Upsets Rarely Happen. The story is that it is rare for a Presidential Candidate to overcome a polling deficit just a week away from the Election, but it has happened. Bush (2000) and Carter (1980) were leading in polling in late October, but did not win the popular vote on Election Day. Bush still won the electoral vote count which is all that mattered. Carter and Reagan had their only head to head debate on October 28th and as I have mentioned previously, that was the kicker for Reagan. He looked as if he could be OK handling the presidency and people wanted a change. A poll taken after the debate but before Election Day had Reagan moving into the lead, so in many ways, that should have been the poll used by Gallop in this analysis.

What history does show is that if a candidate is behind by double digits in late October, they just can't make up the difference. Early voting in over 30 states has also enabled Obama to lock in the sentiment 2 weeks out and decreases the chances McCain has for a late comeback.

Although Electoral-vote.com shows that the average of its 8 polls has an Obama 6% lead, the Gallop expanded poll has a 9% Obama lead (10% announced this afternoon for 10/27 results). The expanded poll for likely voters uses only current voting intentions, not historical turnout. This estimate is estimating that there will be a higher turnout among groups of voters traditionally less likely to vote, such as young adults and minorities. While this is reasonable, it is not certain. Gallop's traditional model has a 5% Obama edge. I am sure the Republicans are focused on this number. So if I keep harping on the fact that this is an electoral vote race, not a popular vote contest, why care about the national popular vote poll results? Assuming it is possible for an underdog to make up as much as 3-5% in the last week of a campaign, a gap of 5% or more is not insurmountable. A candidate could lose the popular vote by up to 1-2% and still win the electoral vote total. If the gap winds up being 3% or more, it is highly unlikely that those votes will be distributed in such a fashion that would give them the electoral victory.

I think the Gallop expanded poll is a pretty fair assessment of how things will play out. As of today, they have the race 53%-43% for Obama with 4% undecided. Assume that most undecideds vote for the party in power which is fairly common and you wind up with a 6% gap on Election Day. The important thing for Obama is that even in the traditional Gallop poll, he still has 50% of the vote, which is important when you know the historical trends of the undecideds.