Why are the national polls tightening when in fact the state polls are not?
If you were just to look at some of today's national polls you would be very scared Obama is slipping and in danger of losing on election night.
For example some 10/29/08 Polls show:
1.DailyKos poll shows Obama up 50% to 44% in a poll with a very favorable party breakdown of 35%D / 26%R and 39% other.
2.Rasmussen Reports shows Obama up only 50% to 47% in a poll with a moderately favorable party breakdown of 40% D / 32.8% R and 27.2% unaffiliated.
(I know other polls have other leads more favorable to Obama).
Now, since Obama's lead is shrinking in polls where the Democrats outnumber the Republicans you might start to get worried.
However, as everyone knows this is a state-by-state election for electoral votes and on that front McCain is making no dent in Obama's chances of getting at least 270 electoral votes (I am not so sure a 269-269 tie, where the House of Representatives decides the winner, is an absolute lock for Obama). fivethirtyeight.com gives McCain only a 3.8% chance of winning the election and the intrade.com market now only gives McCain a 14.4% chance of winning. Obama has a comfortable lead in all Kerry 2004 states and leads substantially in VA, IA, CO and NM.
So, what is up with national polls? First off, maybe national polls are something left over from a past when state polling was not so common. This year is certainly the most I have ever seen state polling at this volume.
Anyway, when I see McCain have any hint of a gain I get concerned. What is going on? Why can the national polls be closing when McCain's chances of winning are not?
Any thoughts would be appreciated.
UPDATE: I just noticed one thing that may be a flaw in the geographical distribution of the DailyKos national Poll whic is tightening. The internals indictae that the poll surveyed 29% of its sample from the South (FL, NC, SC, AL, MS, GA, VA, TN, KY, LA, AR). Those 11 states ONLY account for 127 electoral votes which is ONLY 23.6% of the 538 total. The poll should only assign a weight to a group of states the same as their impact on the electoral college.