/ About / Archive / Other media
Which voter groups give you the biggest payoff at the polls? - 1/24/19
One of the most puzzling things about Nate Silver's new method for evaluating Democratic presidential contenders - or as Libby Watson calls it, "the weird pentagon chart thing" - is his choice of "key" constituencies. Silver insists that his groups -
  • Party Loyalists
  • The Left
  • Millennials and Friends
  • Black Voters, and
  • Hispanic (with perhaps Asian) voters
- were "chosen because they represent the dividing lines in recent Democratic Party primaries"...but what does this mean? Wade into his rationale and it's just an endless parade of conjecture, holistic judgment calls, rehashed media narratives, personal intuitions, and isolated data points. Silver concedes that his approach "definitely reflects a mix of art and science," but there isn't any actual science to be found.

I've been thinking about how you could identify "key" constituencies with some minimal objectivity and rigor, and it seems to me that the main question is whether you care about potential margins or probable margins. On one hand, large groups obviously have the potential to give you larger margins than smaller groups can; that's why you can never take for granted groups that have a high absolute turnout. On the other hand, however, large groups often don't exercise this potential; instead, they split their vote, and smaller groups that overwhelmingly favor a given candidate contribute disproportionately large margins. This is the more probable outcome, which is why so many of these demographic analyses end up looking for decisive group preferences and disregard absolute turnout.

So it seems that there are two metrics one needs to look at to identify "key" constituencies: absolute turnout and absolute margins, which is what you get when you adjust the margin of preference by turnout. These numbers have changed over the years, but since Silver seems to be focusing exclusively on 2008 and 2016, I'll follow suit with an average of numbers from those years.

  1. Black voters. No surprise here: black voters are typically less than a quarter (23%) of the voting public, but since they often have decisive preferences for particular candidates they end up contributing huge margins. On average, in fact, they end up contributing a ridiculous 13.3 points to their preferred candidate's total.
  2. Women. Women gave Clinton a 14 point bump in 2016 - comparable to what she got from black voters - but only gave her a 5 point bump in 2008, putting their average at nearly 10.
  3. Boomers. Strong preferences and superior turnout mean that olds still contribute significantly bigger margins than young people. Voters over 65, for example, have contributed 7 point margins on average, compared to 5 from those younger than 30.
  4. Electability voters. As with women, this number varies dramatically from election to election. In 2008, Clinton won a paltry .2 points from voters who thought she had the best chance to win against John McCain; but in 2016, she won 12 from those who thought she was better positioned to beat Trump than Bernie Sanders.
  5. Economy voters. Voters who rank the economy as their "most important" issue typically contribute about 6 points to a candidate's margins.

  1. White voters. 2016 was a paradigm year for white voters: they made up nearly two-thirds of the voting electorate, but since they split their votes evenly between Clinton and Sanders, their marginal contribution was roughly zero. Nevertheless, their size means that even slight preferences can have big consequences. In 2008, for example, they gave Clinton nearly as many points as black voters gave Obama (10 vs. 13) even though their preference for her was much less strong than the black preference for Obama (55% vs 82%).
  2. Women. Women have an average turnout of about 57%.
  3. Liberal. This label is a bit misleading - as the leftmost option in polling that typically includes "moderates" and "conservatives", it probably encompasses respondents who would otherwise identify as leftists, progressives, and even socialists. In any case, whatever this group actually is, it typically gives you a turnout of around 54%.
  4. Suburban voters. 46% of primary voters, compared with 38% urban voters and 19% from rural areas.
  5. Economy voters. Also 46% on average, though in 2016 there were slightly more male voters (41% vs 42%).

Narrowing this down to five "key" groups would probably just be a question of strategy, though a few conclusions seem inevitable. There is certainly some overlap here with Silver's scheme (in particular, with black voters and the left); but these numbers also suggest groups that he omits (most significantly, women) and contradicts those that he includes (favoring olds rather than Millennials). Concerns about electability and the economy probably vary in importance from year to year, but their presence here suggests that Silver's omission of issue or priority defined groups is probably significant.

Thanks to Michael for helping me pull these numbers.