I have yet to see any diary on this topic in the Kos-o-sphere, or anywhere for that matter. But I'm only a part-timer, so maybe it's been discussed . . .
Let's use Obama winning the very young vote (i.e. 18 years old) by a 3-1 advantage, and likewise assume McCain's poll numbers reflect the same 3-1 margin at the other end of the voting age range. (continue below for rest of the analysis . . .)
If we use a voting age range of 18-83, that's 65 years of voters. While there are voters older than 83, the drop-off in people able to vote over the age of 80 vs. the numbers of 18-year-old newbies, should make that range a good estimate.
Then if we assume in the next months that an equal number of older age voters that favor McCain either die or become infirm as there are new 18-year-old voters that favor Obama, we get the following calculation:
(65 years) divided by (3 months or 1/4 year) = 260. So 1 out of every 260 voters will switch from very old to 18-years-old between now and the 3 months until November 4th. That's equal to 0.4 percent approximately. If Obama is getting 3 out of 4 of the 18-year-old vote, then that would add (0.4 percent) x (3/4) = 0.3 percent. So every poll we see around today will be biased in McCain's favor by 0.3 percent. With the polls as close as they are right now, the 0.3 percent could end up tipping the election to Obama.
(I couldn't find Poblano's estimated median conversion rate of electoral votes for every 1 percent of popular vote - anyone know that figure?)