About the Polls

The Bradley Effect is merely a subset of the "socially desirable answer" phenomenon. People tend to tell pollsters the socially desirable answer.

Here's one simple example: When a pollster is asking if someone is a likely voter, and if they've voted in three of the past three elections, etc., in which way do you imagine people tend to lie?

Towards claiming they're reliable, conscientious voters, of course. That's the socially desirable answer. Only a smaller number of people, such as steve_in_hb, are honest enough to state the true socially undesirable answer.

In steve_in_hb's case, that answer is "Voting is for fags."

But most people, when asked by a stranger, will give the socially desirable answer.

Unlikely voters tend to lie and say they're likely voters, or very likely voters.

But there's more to it than that.

Many pollsters don't just poll elections. They do "voter contact," which means pretty much what it sounds like. They call up likely voters for a particular candidate and tell them all the great things they should know about that candidate. It's not to get information -- it's to give it, and to encourage people to vote for the candidate hiring them to do the contact.

Here's the thing; Pollsters try to keep the callers doing voter contact away from also polling on the same race. Why? Because they find that often the callers doing voter contact wind up getting better results in their polling calls in favor of the candidate whose praises they've sung in the previous voter contact calls.

Why? Well, the suspicion is that after having repeated so many wonderful things about Candidate X, they've internalized those wonderful things (even if they never heard of Candidate X before making 200 calls on his behalf) and it comes through, when they call to poll someone, who they think is the better candidate. Whether through voice inflection or the like, they give subtle cues as to whom they believe is the better candidate. They tip to the interviewee what the socially desirable answer is to the polling questions.

How much of an effect does this have? Well, at minimum, it's enough of an effect that pollsters try to keep voter-contact callers from working on the polls they've done contact on -- so there is an effect.

Someone in the field tosses a number to me. This is anecdotal, based upon limited personal experience, and not a scientifically determined number. But in the experience of this person:

About half the time, such "cross interference" of praising one candidate and then polling on him results in a 6% rise in the polling results for the "socially desirable" candidate.

Gallup and etc. probably are pretty good about not using voter contact callers on polls. In fact, most of the big polling firms probably have so much polling work they don't ever do voter contact calls at all. Their phone banks are exclusively for polls.

But that doesn't mean there's not a socially desirable answer being suggested by the callers.

For one thing, if you walk into any polling firm phone bank, many of the callers are young. Probably not a majority, but a lot of them.

For another thing, all of the callers aren't making a whole lot of money. Calling people on the phone is not a high paying gig.

And, of course -- around half of the callers are themselves minorities. Yup, half.

Put it all together, and you have all of the nation's pollsters using interviewers who almost certainly skew in political preference to Barack Obama.

No, they probably didn't do any voter contact work for Obama, and hence get "poisoned" by repeating his virtues over and over again on the polling company's dime.

But a large number of them have probably done so, for free. On their own time. Not because they were paid to call people and tell them how swell Barack Obama is. But because they simply believe it in their hearts.

And I really don't think that I need to prove that Obama's supporters tend to be very enthusiastic supporters. Black supporters among the most enthusiastic of all.

Does this have an effect on polling? Well, once again: Pollsters are on alert for workers who may have been tainted by doing voter contact. They know such callers get too many positive responses for the candidate they have learned to prefer.

But what can they do when 80% of their calling staff walks in through the door with exactly that same taint?

They can't do anything.

Combine that with the fact that black callers, for example, tend to find more support for black candidates generally when polling. Same deal, really, except more pronounced. If a guy who sounds black asks you who you're voting for, your tendency is not going to be expressing a greater desire to see John McCain elected than you actually might have.

I don't know to what extent this impacts the race. But this sort of "contamination" of callers is something polling firms watch out for. And I don't see how they can avoid it when most of their callers are pro-Obama, and a significant fraction of such zealously so.

posted by Ace at 02:54 PM

| Access Comments