Rigged Math

One of my guilty pleasures this campaign season comes in the form of Bill Mitchell’s Twitter feed. He’s a radio host or something. But in my mind he’s the foremost champion of the pro-Trump magical thinking. No one seems more committed to the idea that online polls are totally legit. For instance:

Last night he offered this gem:

I should just leave it there, and let the incandescent idiocy speak for itself. But just in case some don’t see the flaws in this, I’ll explain.

It is true that online polls — when not manipulated by bots and other repeat voting techniques — can be an indication of enthusiasm by diehard supporters. I suppose that can be interesting, when put in the proper context. But the idea that fan polls are scientifically predictive is barmy.

A scientific poll that uses a representative sample can tell us about the electorate because it is representative.

I will avoid all talk of statistics, because ugh. Imagine you have a big pot of chili. If you take a spoon and scoop out a sufficient portion, you can get a good idea of what the chili tastes like and what it’s temperature is. The spoonful represents a tiny fraction of the pot, but enough to convey the information. Now, it’s possible the spoonful didn’t capture enough beans (shut up Texans), or grabbed too much liquid or meat. That’s why people average scientific polls. The more spoonfuls you take — and then average together — the more accurate the picture gets.

Online polls are very different. To mangle the metaphor horribly, they measure the intensity of different ingredients. If you cook all the beans separately and then take a big spoonful of beans — without seasoning, meat, stock etc. — you will get a good idea of what the beans taste like before you mix them in and let it all simmer. That’s useful information — about the beans — but it tells you very, very little about the final product. Bill Mitchell has been doing that all season, which is just one reason he’s full of beans.

Most Popular