Repartay

Yesterday, Byron York wrote an opinion piece making the argument (?) that if you eliminate blacks from poll results, Obama’s poll numbers are lower. I put a “?” next to the word “argument,” because York is really making more of an observation with a hanging and deniable implication based on the banal point that Democrats have historically received lots of support from blacks. Anyway, York saw fit to defend the column against charges of racism.

I wrote my post because of the striking numbers in the New York Times poll.  Those numbers raise a question: What if a president were wildly popular with one group, and only middlingly popular with another group and yet was often portrayed as being hugely popular with the whole group?[...]

[...]Perhaps some people find those numbers entirely uninteresting, but I think it is entirely reasonable to point them out. It is also entirely reasonable to point out that a poll result can be shaped by an extremely high number in one component of the poll result.  It’s the old joke:  Six people are in a bar.  They’re all middle class; their average net worth is about $100,000.  Bill Gates walks in.  Seven people are in a bar; their average net worth is in the billions.  A wealthy group, right?  Internal numbers are revealing.

This argument would be really clever if it weren’t so staggeringly stupid. The obvious difference between the two scenarios — one where Obama is more popular among a certain segment of the population than another, and the other where certain segments have a grossly distorting weight — is that the opinion of one person still only counts as the opinion of one person (though York may prefer it somewhere closer to 3/5ths). For York’s analogy to work properly, Obama’s poll numbers would have had to rely disproportionately on the opinion of a particular group, which indeed, would have skewed his overall popularity numbers. But of course if pollsters did that, their product would be totally worthless. Instead, they use a fancy method called “representative sampling,” so that when they release a poll, it accurately reflects the overall population, thus making it basically impossible to “portray” results as “being hugely popular with the whole group,” without the results actually demonstrating huge popularity with the whole group. The one exception — and the point York seems to be making — is that if you don’t consider blacks to be part of the whole group, then it’s true that elections polls that include blacks overstate Obama’s popularity.