Some interesting crosstab data
DATE: Have voted Obama/Romney vs Plan to vote (%) Obama/Romney:
10/9: Have voted (11%) 59/39- Plan to vote(89%) 44/45
10/16: Have voted (18%) 57/38 - Plan to vote(82%) 43/43
10/23: Have voted (26%) 59/39 - Plan to vote(74%) 44/46
10/30: Have voted (25%) 56/40 - Plan to vote before election day(14%) 58/35 - Plan to vote on election day(60%) 41/50
The 'have voted' number dropped 1% probably because they changed to allow 3 options to answer
Do cell phones matter?
Cellphone: 52 Obama/36 Romney (Obama +16)
Landline: 46 Obama/48 Romney (Romney +2)
Coincidence that the only OH poll to show Romney +2 (Rasmussen) only calls landlines?
It obviously does matter, I don't think anyone besides the most strident GOP hacks try to argue it doesn't.
While I see the concerns over calling only landlines, as long as landline only polls adjust their aggragate sample to reflect the anticipated turnout share of young voters, why would that effect the poll's accuracy?
In other words, if a pollster accurately estimates (e.g.) 15% of voters will be 18-29, and in a landline only poll ensures that 15% of respondents are in that age bracket, why would the poll be inaccurate, unless one postulates there's a material difference in political leanings between 18-29 year olds with land lines vs. those with cells only (which I don't see). The same question applies for other demographic subgroups (Af. American & Hispanic voters perhaps?) than youngs which may also be more landline-free.
Of course, this all assumes the pollster (Rassy in this case) gets the estimated turnout of youngs correct in advance, or even attempts to account for the demographic slant of a landline only poll.