The giant Republican presidential field has created tough choices for the hosts and organizers of presidential debates. You can't let ALL 17 candidates on stage! They say.
There's something to that — a super-crowded stage would result in either a very long debate or very perfunctory answers, or very few questions.
So drawing a line somewhere makes sense. After all, it's not true that there are 17 candidates. In fact, over 100 people have filed to run for president, almost all of whom you've never heard of (though one may well be your next-door neighbor).
I would prefer taking the top 17 or so, and dividing them randomly into two different debates. Then maybe in a second go-around, mix them up again to maximize the number of combinations.
Instead, it looks like we'll have an "A" debate and a "B" debate. One objection to this is that it's arbitrary to include No. 10, and not No. 11 when they're so close in the polls. NPR has an article leaning on the fact that the polls' margins of error are way larger than the gaps between the candidates.
This way of discussing the polls (as the article notes in a parenthetical) is statistically misleading. First, an average of many polls — which is what the debate organizers will do — could be understood to have a far smaller margin of error than any one poll. Secondly, those margins of error basically apply in a poll where the percentages are something like 48% to 46%.
In other words, an average of polls will have some fuzziness around who's in 10th versus who's in 11th, but nothing close to what that NPR chart suggests — that No. 6 might "really" rank below No. 14, given the margin of error.
My AEI colleague Stan Veuger ran a statistical simulation to see how "off" the average of the five latest high quality polls might be. To put it simply, any given set of five polls would be 85 percent likely to agree with the first set on who the top ten is — and thus who makes the cut.
Veuger adds: "Furthermore, the top five candidates make the debates every single time, and number six through eight fail to make it just once [out of 1000 simulations]. In addition, the three weakest candidates among the 16 never make the cut. If we decide to draw the line somewhere, this is a reasonable way to do it."
Put another way, among the top 16: [Tier 1] there's a top 8 that definitely belong in the debates (Trump, Walker, Bush, Rubio, Carson, Huckabee, Paul, Cruz), [Tier 2] a borderline 4 (Kasich, Christie, Perry, Jindal), and [Tier 3] a bottom 3 (Fiorina, Santorum, Pataki).
Of course, there's also a [Tier 4] of about 100 candidates, from which Veuger, a leading Dutch blogger, provides a colorful example:
"Let's dedicate a moment to Michael Bickelmeyer from Ohio, who advocates for an orbital weapons platform called "A Gift for Children." This weapon system would, if I understand correctly, magnify outer atmosphere sunrays by having them pass through a series of magnification lens systems to produce lethal beams that can be used to combat terrorists and drug traffickers or even to eliminate entire countries."
As we said, one has to draw the line somewhere.