Why did WSJ poll wildly oversample black voters? Blacks make up about 13% of the general population. In this poll they represent nearly 25% of the sample (177 out of 700 polled). Doesn’t that make this poll a non-starter? http://dirtyharrysplace.com/?p=808#comment-1210
Has the discrepancy between the robo-poll and the in-person WSJ/NBC survey shown us that The Bradley Effect has now gone into effect?
The theory suggests that statistically significant numbers of white voters tell pollsters in advance of an election that they are either genuinely undecided, or likely to vote for the non-white candidate, but that those voters exhibit a different behavior when actually casting their ballots.
http://en.wikipedia.org/wiki/Bradley_effect
It’s standard practice to oversample a sub-category of interest to obtain a lower Margin Of Error, and then to re-weight for final poll reporting. The idea here is that white representation is large enough, not requiring an oversample. If one were to draw conclusions on ‘black opinion’ based only on 13% of 700, the MOE would be over 10.
As Cheney would say, “So?”
Tags: Bradley Effect, cooked books, different behavior, hillary, oversampling
