|
Edited on Mon Jul-14-08 12:40 PM by jobycom
The formulas pollsters use take a lot of varied demographics into account, based on historical turnout patterns, and on turnout expectations based on poll questions, so most people who are cell-only (like me) are included in other demographics as well. Unless there is some expectation that cell phone users will vote as a solid block, or as a recognizable demographic that takes prominence over any other income, gender, ethnic, or regional demographic they also fit into, then not skipping them won't affect things much.
But as the article points out, with Obama using high tech promotional methods, he may be creating such a demographic. They will still pick up that demographic by making hand calls on cell phones, but their formulas may not account for that, and since it's easier to robo-call, their sampling of land-liners will be more extensive, so more reliable, than their sampling of cell phone users.
The bigger problem, though, is the classic question of whether pollsters catch demographics changes as they happen. As with war, you are always polling last term's election, because the formulas are based in part on last term's turnout. If a new demographic pops up--in this case, a high-tech demographic that might be voting as a group independent of other groups people belong to--they may not notice it right away, and may not notice it for several cycles. Their numbers will just be off.
This cell-phone bias could work either way, too. Say they randomly sample white males between 35 and 40 in a middle income bracket--a typical Republican base--but since they aren't filtering for cell phone use or high tech criteria--most of their sample would also fit into this cell phone bracket. So they are more pro-Obama, and when plugged into the whole formula, they may make the entire graphic look more pro-Obama than they are.
The reason I say it's probably overstated is because this cell-phone group is captured in polling in other demographics, so it's not likely to be a large percentage over a long period over several polls. Rather, it will cause some polls to be wildly inconsistent with others, and we won't know which are right until it's over.
Yeah, I write too much. Sue me. :) No one reads my posts anyway.
|