The current (13 Sep 08) CNN “super poll” (combination of 4 polls) says that McCain is ahead 45% to 44% with “no margin of error.” (That last means, I assume, that they can’t calculate the error, rather than that the poll is 100% accurate.)
Anyway, is there information about why those people who were polled chose their candidate?
Sometimes they’ll ask directly. For automated polling calls, they may say something like “which issue is most important to you this election? If it’s the economy, press 1… if it’s health care, press 2…” etc. Live polling might simply ask and then try to slot your answer into a chart. Obviously it’s not perfect but if someone says that National Security is most important to them and they’re voting for McCain, you can make a reasonable assumption that their vote for McCain was swayed by his stance on national security.
This information is sometimes freely given when you look at the polling results and sometimes it’s kept and sold to people who subscribe to that polling firm’s data. For example, Rasmussen has daily polling, the basic results of which are available to anyone on their website. They also have information such as how many white female Republicans are saying they’ll vote for Obama (and how many white Republican women think that Iraq is the #1 issue) but you have to subscribe to Rasmussen to access those “cross-tabs”.
As for the CNN “margin of error”, you’re correct in your guess that they simply aren’t giving it a MOE due to each individual poll having a different MOE rather than claiming that the super-poll is dead accurate.
This week I got a live telephone pollster who asked me three questions:
Which candidate did I favor?
Would I describe my support as “definite” “leaning toward” or “maybe”?
Did I favor or oppose abortion rights?
Obviously someone out there is trying to draw a link between the two.