The myth of the monolithic subgroup.

I come to writing about politics from the perspective of 30 years' experience in marketing research. And one thing I know from that experience (among many) is that market segments are often stable, but there's nothing to suggest that they can't be divided further, sensibly, if you introduce a new variable into the analysis. Descriptions of groups are convenient categorizations which help users chunk a whole host of information; at the same time that you identify a segment and characterize it, you not only collect a lot of traits in the characterization, but you also introduce a simplification. That's part of the modeling process; and segmentation is a modeling process.

There actually aren't very many genuinely monolithic segments out there; there will always be some variation. For instance, even if African Americans can pretty much be counted on to vote Democratic, thanks to Herbert Hoover's duplicity after the 1927 flood (cf John Barry), there will be some who won't. I knew that recent poll result which showed 0 percent support for Romney vs. Obama had to have some sampling variance, even if immeasurably small. Had the sample included Herman Cain, Alan Keyes, or Allen West I don't think the number would have been zero.

So there's this discussion (you've heard it) about Obama doing poorly with the White Middle Class voter. Heard it? Well, it turns out that it's not a general national fact, that it's limited to the South. Kevin Drum explains.

Home | | 7:54 PM
(DISCLOSURE: I work for Abt SRBI. My company does polling. My opinions should not be construed as representing those of my employer.)