Politics: What happens if the late deciders decide for the challenger?

The polls are close.

Some have Romney ahead and some have President Obama ahead.

The "conventional wisdom" is that if a voter is undecided they are more likely to vote for the challenger. Thus, let's assume for the purpose of discussion that 6 of 10 voters will vote for the challenger if they decide in the final weekend before the vote.

The Politico 10/22-10/25 poll has: Obama at 49 and Romney at 48.
Thus, there are 3% undecided.
If the 60/40 rule holds, then Obama will add 1.2% to his totals and Romney will get 1.8% more.
Final result: Obama 50.2% Romney 48.8%.

This scenario would probably mean a very long election night and possibly many days of counting absentee ballots and perhaps recounts in some swing states.

In the IBD 10/22-10/27 poll, Obama has 45 and Romney at 44.
Thus, 11% undecided.
If the 60/40 rule holds, then Obama adds 4.4% and Romney 6.6%.
Final result: Romney 50.6% Obama 49.4%.
Again, likely a very long night with a possible Romney popular vote win but an Obama re-election because the electoral votes are determinative.

On the other hand, if the Gallup 10/22-10/28 poll is right: Romney 51 Obama 46.
Thus, 3% undecided.
Final result: 51+1.8 = 52.8 Romney and 46+1.2 = 47.2 Obama.
A 5.6% popular vote differential would probably translate into a solid electoral college win with Romney probably right around 300.

So the key numbers:
(1) How many undecided there really are: 3% or 11%?
(2) If 3% then is the level of support for the President closer to 49 or 46?
(3) Will late deciders break toward Romney 60/40?

The three polls above have different readings on items #1 and #2. Of the two polls that agree the undecideds are only 3%, they differ on the level of support for the President.

Poll fans may note my omitting Rasmussen. Rasmussen is very close to Gallup as of 10/31 with Obama at 47% and 4% undecided. I wanted to highlight three very different polling scenarios.

Nate Silver appears to hold to the view that the polls point to about 3% undecided, the President's support at 49 to 50% (like Politico) and that the undecideds will probably split 50/50 yielding a popular vote of 50.5 to 51.5% and a comfortable electoral college win for the President. He also gives more weight to aggregating the state based polls over the national polls.

Electionprojection.com, like Nate Silver, also attempts to aggregate the polling data and come up with a forecast of what will happen and his model also predicts an Obama victory. However, he does point out the key is the turnout model. If the current polling data is adjusted to a 2004 turnout model or even a model somewhere in between 2004 and 2008, Romney wins. But if the data is taken as is, Obama wins. He puts it this way: Either the polls will be right and so will EP, or they'll be wrong and Romney will win the White House.

Jay Cost thinks the key polling number is Romney's edge on the economy and among independents.

Keep in mind, in the end, about 120 million votes will have been cast for president. A reported poll is on the order of 500 to 3000 people which forms a sample to assess possible outcomes in a state or national context. When they report a margin of error of 3%, that means a 51 to 49 lead could really be a 54 to 46 lead or a 48-52 deficit. Also, there is a 5% chance that the actual results will fall outside the stated range.

In 1984, 1988, 1992, 1996 and 2008, the polls all pointed to easy wins and the only question was what will be the final margin of victory. In 2000 and 2004, the polls were all very close and exit polling that major new outlets used on election night turned out to be problematic. Barring any last minute surprises, 2012 is looking to be very close.

November 6 may tell the tale. But if it is really, really close, there could be additional days of counting and recounting and we won't have an answer until several days later.