| I think the problematic issue now is the likely voter model. It's fairly hard to find cross-tabs for polls, and when I do, I note that they have very high estimates for democratic turnout vs. republican. In other words, they assume that democrats have the edge in enthusiasm, organization, party ID, and so forth. A reasonable assumption, but difficult to translate into numbers.
In the primary, BTD of talkleft wrote that demographics is destiny. In other words, voting patterns among demographics groups were remarkably stable after February.
Now, if you look at different polls and even exit polls of past, you can say that Party ID is destiny. Rarely, a tab is so well correlated with the outcome. Take 2006:
CNN exit poll stated that 38% of voters identified as democrats and they voted by 93% to 7% for democratic candidates; 36% identified as republicans, and they voted by 8% to 91% for republican candidates. Independents formed 26% of voters and preferred democrats by 18% (57 to 39). Exit poll for 2004 showed the party ID at 37-37-26 (D-R-I).
Now, Rasmussen has a likely voter model for this election which assumes a partisan breakdown of 40-33-27 (D-R-I). Say Democrats vote D by 88-10 and vice versa for republicans. An 7 point spread in party ID gives Obama a 5.5% advantage (leaving independents out for now), whereas a 2 point spread translates to 1.5% advantage. A HUGE difference, and one completely hidden in any poll results that become public. More important, the difference is about the same size as the MOE for the poll. So, be ready for those after the election who cry fraud, who say how could every poll be wrong?? Yes, EVERY poll can be wrong, if they are all influenced by the same mistake in who is going to turn out this election.
Second to party ID, I find a good indicator of an election is the preference of independents. Take these numbers, for example, all among independents only:
2000: Gore 45% vs. Bush 47%
2004: Bush 48% vs. Kerry 49%
2006: Democrats 57% vs. Republicans 39%
2006 (VA-Sen): Webb 56% vs. Allen 44%
2006 (MO-Sen): McCaskill 51% vs. Talent 43%
Now compare these party ID's:
HOUSE: 38-36-26 (D-R-I 2006)
NOW: 40-33-27 (RAS.); 41-28-31 (CBS); 38-29-29 (ABC-WAPO)
OHIO: 40-37-23 (D-R-I 2006); NOW: 45-35-19 SUSA
VIRGINIA: 36-39-26 (D-R-I 2006); NOW: 38-36-24 SUSA.
Pennsylvania: 43-38-19 (D-R-I 2006); NOW: 49-41-10 SUSA.
All 2006 numbers are taken from CNN 2006 exit polls.
Put another way, the national party ID gap has gone from +2 to +7 (RAS), +13 (CBC), or +9 (ABC-WAPO), all in favor of D.
The state party ID gaps are gone from +3 to +12 (OHIO), -3 to +2 (VIRGINIA), and +5 to +8 (PENNSYLVANIA).
Yes, CBS national poll has a 41-28-31 R-D-I breakdown, and a 54-46 breakdown according to gender. Any wonder that they show a 13 point lead for Obama? Remember how these polls would underestimate the women turnout in primary?? Now, they overestimate it since it suits them. On the other hand, if (a big if) you trust FOX, the party ID and enthusiasm has dramatically increased for the republicans in October.
So, my point is? Guess a number for party ID gap, say 4%. It translates to about 3 points for D. Add the edge among independents (say 10%, for a roughly 30% of voters=3 points). Add the two, and Obama's margin is 6%. You too can play this game.
Seriously, like every other recent election, this one comes down to turnout and whoever takes the independents.
NOTE: with apologies, I am afraid the diary may turn into another venue for horse race discussions and heated exchanges. So, I have turned off the comments. Thanks for understanding. Comments now open. Please be good, and don't make Alegre kick me out!! |
