Black voter behavior in National Polling

[cthulu2016]

When you see a national poll that says what black people are up to, take it with a big grain of salt.

This isn't anything about black people, it's about polling. The same goes for any minority... an age group, an income group, a religion, etc..

A typical national poll has around 1,000 people in the sample. (800-1,200)

100-160 of the people in the sample will self-identify as black. That 10-16% of the sample will be "weighted" to equal about 13% of the poll total. (That kind of weighting is valid.)

A national poll of only 100-150 people has a LARGE margin of error. It is valid as a subset of the whole poll of 1,000 people, but you can't make fine tuned observations about things within that sample.

Example: Does the PEW poll say that Obama lost 7% of black voters to undecided? No, not really. That 7% is probably 7 to 11 actual people in the poll. Not percent... people. In a subset where the margin of error is probably 5% or 6% or 7%.

Was there some increase in "undecided" among black voters? Sure. Probably. Was it 7%? Who knows... the sample isn't really large enough to talk with precision about what 7% of 13% of the population is thinking.

7% of 13% is a little less than 1% (0.91%) and no poll cuts that fine.

[MADem]

1. I have to wonder how many Republicans are fucking with the pollster! nt

[ProSense]

2. You know, the poll is what it is, but

there is no need to try to portray the sample as anything but screwed up.

I mean, party ideology is one thing. When every demographic is off, then the poll begins to seem like an outlier.

[cthulu2016]

3. That's fine. It might be the worst poll ever. But the observation pertains

When I see people talking heatedly about what 7% of black people did or did not say in some poll what springs to mind is, "Hell the margin or error on black voters in that poll is probably more than 7%."

We just went through the thing with the AP debate poll where (because the poll was only 440 people) the non-white subset was too small to even responsibly report.

I don't think PEW should have listed those break-outs without the MOE of each sub-set right next to the numbers because the MOE for the headline number doesn't pertain to the sub-sets..

As for the quality of the PEW poll, this OP doesn't speak to it one way or another. I am not defending it here... saying a result is meaningless would be an odd mode of defense.

I do not think a lot of black people will be voting for Romney. Period.