Wednesday, January 09, 2008

The Bradley Effect?

I don't want to sound like a sore loser or anything, but Clinton's win in New Hampshire tonight is mighty fishy. She's won it 39-36, but eight national polls released over the last 24 hours, no more, all gave Obama a hefty lead (an average of 7-10%). How did a 10% lead become a 3% loss? Isn't a 13 point swing in one day, on the day votes were cast, a little strange?

Not if you consider the Bradley Effect. For those who don't know, the Bradley Effect is named after a former black mayor of Los Angeles who ran for governor of California against a white opponent. Leading up to the election, Bradley dominated all the polls. He ended up losing the race. Why? Because white voters, in an effort to not appear racist to pollsters, said they would vote for Bradley or that they were undecided. When the time came to actually vote - and, crucially, when no one could see which candidate they voted for - they went the other way. The Bradley Effect, then, refers to an artificial bump in pre-election polling numbers for black (or Hispanic) candidates running against white ones. From my vantage point, it appears to have done Obama in tonight.

Tell me I'm wrong. Tell me there's another sound, logical explanation for a 13 point swing in one day. Please, go ahead. I'm waiting.


zeyd said...

Recount, anyone?

Anonymous said...

I think Obama is "white" in many ways that other candidates have not been. His mother is white, he speaks like a white man, he looks kind of mixed more than black. I don't think his being black is playing a major role in the campaign, at least yet.

I just think the polls were wrong. Hillary won the womans vote BIG TIME.

Ahsan said...

how can eight out eight polls be wrong the morning of the primary?

ayla said...

I was watching Fox News (entertainment purposes only I assure you) and the anchor posed the same question to a correspondent in NH. He didn't mention the 'Bradley Effect' but he did mention the following:

- that the number of voters was unprecendented and polling techniques and outputs need to be revised to take into consideration the fact that more people from a wider range are voting now.

- many voters didn't make their minds up until once they were in the polling station and actually casting their vote.

Go Obama!!! It was awesome... last night i was watching the news Al-Jazeera or CNN and apparently he went to school in Jakarta and his step father was Indonesia. How random is that! I hope he wins.

Ahsan said...


1. the number of voters argument is wrongheaded. most polls (if done correctly, and i can't see how eight out eight polls were done incorrectly) will use a sample size of 800-1000 people, IRRESPECTIVE of the population size. in other words, whether you're taking a poll of americans (population 300 million) or brits (population 60 million?), you will see the same sample size. this is because, for large numbers (i.e. in the hundreds of thousands, or millions) you get very little additional traction past a certain threshold, which is in this case about 1000 people. so taking a poll of 10000 people and 1000 people will give you pretty much the same results, in terms of accuracy.

2. the "making up their mind in the polling booth" doesnt necessarily contradict my hypothesis. as i said, many white voters, in an effort to not appear racist, say they will vote for the black candidate OR (and this is key) that they are undecided - even though they KNOW they will vote for the white candidate. this is what i suspect happened here.

ayla said...

On the BBC their correspondent in NH said that the polls were taken latest Sunday night... so that might be another reason. BBC said it, not me.

Also, I disagree with your argument in point 1. that a poll of 1000 people out of 100,000 will be just as accurate as 10,000. Also, we do not know the exact techniques used when polling. Anyway... there's always Nevada and Michigan to go!

Ahsan said...

the time lag (sunday-tuesday) may have been an issue, what with her crying and gaining sympathy in the interim. but you have to figure her crying attracted some voters and pissed off others, so why don't we hear about the second group of voters?

as for the 1000/100000 stuff, i can only advise you to consult the stats book nearest to you. this is not a particularly controversial idea so i'll just leave it at that.