Which is it? The URL (and experts in your original story) implies the endorsement won't matter but the headline says it will? It's the former.
Pew, which generally does stellar work, should be a little ashamed of the question. You can not really measure an endorsement's impact by asking a more-or-less-likely question. First, marketing and political science is awash with data showing that people simply cannot accurately report what influences them -- to buy a product, vote a certain way, etc. That is why marketers spend huge money for "unobtrusive" research that measures influence by observation rather than direct report.
Secondly, respondents answer the question differently -- some attempt to say whether or not an endorsement actually sways them (and these include many who do not want to admit they could be so influenced) while others are simply giving you a favorable or unfavorable opinion of the endorser. I seriously doubt that 13% of Democrats really will be less likely to vote for Obama because of Oprah -- most of those are people who just don't like her, and use the question to register that.
It is also useless to ask generically about an endorsement, without naming the specific candidate. Oprah's endorsement might have a completely different impact if it were on Hillary's behalf than if it were for Obama. An endorsement might have limited impact if it's expected and greater if it is unexpected and therefore tells voters something they didn't know about a candidate, or if it drives part of their message (Wesley Clark picking Hillary) or if it draws publicity to a struggling campaign. As I said, Pew should be embarrassed.
There is only one way to reliably gauge an endorsement's impact (which, as Larry Sabato said in your original story, is almost always nil). That is to use a "split sample" experiment: take a random half of your 1,032 voters and ask who they are voting for, take the other half and ask the same question after they have been told that (say) Oprah supports Obama, and see if his support is any higher among the people who hear about the endorsement. This way you are not relying on voters to accurately self-report their decision process, you are observing the impact that the information has.
Which is it? The URL (and experts in your original story) implies the endorsement won't matter but the headline says it will? It's the former.
Pew, which generally does stellar work, should be a little ashamed of the question. You can not really measure an endorsement's impact by asking a more-or-less-likely question. First, marketing and political science is awash with data showing that people simply cannot accurately report what influences them -- to buy a product, vote a certain way, etc. That is why marketers spend huge money for "unobtrusive" research that measures influence by observation rather than direct report.
Secondly, respondents answer the question differently -- some attempt to say whether or not an endorsement actually sways them (and these include many who do not want to admit they could be so influenced) while others are simply giving you a favorable or unfavorable opinion of the endorser. I seriously doubt that 13% of Democrats really will be less likely to vote for Obama because of Oprah -- most of those are people who just don't like her, and use the question to register that.
It is also useless to ask generically about an endorsement, without naming the specific candidate. Oprah's endorsement might have a completely different impact if it were on Hillary's behalf than if it were for Obama. An endorsement might have limited impact if it's expected and greater if it is unexpected and therefore tells voters something they didn't know about a candidate, or if it drives part of their message (Wesley Clark picking Hillary) or if it draws publicity to a struggling campaign. As I said, Pew should be embarrassed.
There is only one way to reliably gauge an endorsement's impact (which, as Larry Sabato said in your original story, is almost always nil). That is to use a "split sample" experiment: take a random half of your 1,032 voters and ask who they are voting for, take the other half and ask the same question after they have been told that (say) Oprah supports Obama, and see if his support is any higher among the people who hear about the endorsement. This way you are not relying on voters to accurately self-report their decision process, you are observing the impact that the information has.