All of this writing and data analysis is a lot of work! So after more than five years of posting, I've finally launched a Patreon to help pay the bills.


Some intersectional analysis on the Trump and Clinton coalitions

The demographic divides that define the 2016 election have been fairly clear for months, though some pundits, of course, persist in getting them wrong. As I tweeted out earlier tonight, the general state of affairs looks something like this:

If you're white, and/or if you are a young boomer, you're more likely to vote for Trump than Clinton. Otherwise, you're more likely to vote for Clinton than Trump. Within those two coalitions, of course, the strength of support also varies, but any demographic analysis of this election has to begin with race and age.

This, of course, is a significant simplification. A more intersectional approach wouldn't just look at how identity, as defined by a single dimension, determines one's politics - it would look at how the intersection of multiple identities do so. And when we do this, the picture gets more complicated. 

The boomer race split

Consider, for example, how the numbers look at the intersection of age and race (here, black vs. white):

A few significant trends stand out. First, of course, race is by far the most important factor predicting one's vote - this is implicit in the first chart. This tendency is so strong, in fact, that it masks a second trend that only becomes visible here: a split between black and white middle-aged voters. Support for Trump isn't simply coming from young boomers. It's specifically coming from young white boomers, and also from ageing white gen-xers. Among black Americans, meanwhile, we see a complete reversal of this trend: support for Clinton is the strongest among black Americans between 50 and 59.

This fact of black American life gets completely erased by the usual simplified way that we talk about political demographics. It is numerically correct that Trump's support comes from boomers, but only because the overwhelming majority of boomers happen to be white. Such statistics tell us just as much about racial proportions among age groups as they tell us about political preferences. It's only when we dig into the data, and do an intersectional analysis of the trends, that we get a more accurate picture of what's going on.

In what sense is Trump the candidate of white men?

Or consider, meanwhile, the common characterization of Trump as the candidate of white men. This is obviously true in the narrow sense that Trump has the highest margin of support among white men - but let's put this fact in context:

Once again, we find that the most important divide at work is race. Race is what divides preference for Trump from preference from Clinton, and race proves far more important than gender even within the two coalitions. (This is why, for instance, support for Clinton is stronger among black men than Hispanic women.)

This point is worth attending to when, for example, we have multiple white women in media characterizing Trump as the candidate of white men. As we see above, this is mostly an exercise in last-place-avoidance - but it does little to distinguish white women otherwise. Their shared whiteness with white men is what ends up mattering; being a woman knocks off a few points off of their overall preference for Trump, but it does nothing to bring them into a coalition with people of color. Here, the intersectional high ground white women can claim over white men is about as legitimate as 40-49 year old whites scolding 60-69 year old whites for liking Trump just a little more than they do.


One takeaway from all of this is that the overwhelming majority of our demographic analysis is painfully simplistic. What I've done here is more sophisticated than what we usually see, and even this was an extremely limited effort. A more ambitious analysis would look at how multiple demographic dimensions - race, age, gender, income, and so on - all combine in unique points of intersection with unique political tendencies. Doing this with two genders, four races, five age groups, and six income brackets would leave me with 240 data points to analyze, and I don't do that kind of work for free. It is, nevertheless, the bare minimum of what anything resembling an adequate intersectional demography would have to involve.

A more important point, however, is that once you start digging into the way various forms of identity intersect, the usual simplifications that popular demographic analysis deals in end up obscuring as much as they reveal. It's fine to talk about the role that boomers and white men play in supporting Trump, but not if this erases the role of middle-aged black voters in opposing Trump, or if it masks the complicity of white women.

This kind of simplification isn't the only problem with what passes for identitarian demographic analysis, but it's one of the most pernicious and egregious - and if we're going to do this, we should at least get it right.