How To Tokenize with Proportions

13% of American Muslims voted for Donald Trump.

That's a minuscule proportion. It is around half the proportion that Hillary Clinton got in Idaho. It is fair to say that Muslims overwhelmingly voted against Trump, just like it is fair to say that Idahoans voted overwhelmingly against Clinton.

13% also translates, roughly, into "1 in 8". And when you think of it that way, it shouldn't be that hard to find a Muslim Trump supporter. Statistically, all you'd need to do is know eight American Muslims, and one of them is probably a Trump voter. And across a population of roughly 3.3 million Muslims, that means there are roughly 412,500 American Muslims who support Trump -- a lot of people! Yet it would be clearly, obviously wrong to use those "lot of people" to try and argue against the above conclusion that "Muslims overwhelmingly voted against Trump."

In short, it is simultaneously true that "Muslims overwhelmingly dislike Trump" and "it is not hard to find Muslims who do like Trump." Likewise, we can simultaneously know that Idaho is exceptionally conservative and know that finding liberal Idahoans doesn't take any herculean effort.

When one doesn't keep those two thoughts in mind, it is very easy to mislead oneself. I've noted that 13% is also the percentage of UK Jews who planned to vote Labour last election, but that still means it should not be remotely hard to find Jews -- quite a few Jews -- who are loud-and-proud for Jeremy Corbyn. If one is a Corbyn fan, one can (accurately!) think "look at all the Jews I know who support Corbyn" and then (inaccurately) conclude that the stories of widespread Jewish consternation over Corbyn are ginned-up nonsense. Same with Black Republicans -- they're simultaneously rare and not that difficult to find, and so it is easy for conservatives to dupe themselves into thinking they have no race problem by pointing out all the Black Republicans out there.

Ditto when one sees big crowds of angry constituents in a deeply conservative or liberal representative's town hall meeting. One can see those and think "wow -- even here people are turning against [Insert Party]!" But even in the most electorally lopsided districts, there are still going to be quite a few members of the other side -- certainly enough to pack an auditorium, if they're feeling motivated.

Or take this article, "To Understand White Liberal Racism, Read These Emails." It is about angry emails sent to school administrators regarding the decision by Seattle school teachers to wear "Black Lives Matter" t-shirts. The article observes that these emails came from one of the whitest, most affluent" and "staunchly liberal neighborhoods" in the city, places "dotted with rainbow yard signs that say 'All are welcome.'"

Applying the "staunchly liberal" label to these neighborhoods is entirely justified. The (Democratic) state senator in this part of Seattle was last re-elected with 80% of the vote. That's a crushing margin! But it still means that 1 in 5 voters in the district cast their ballot for Republicans. On the one hand, that's not a lot of people. On the other, that's a lot of people! Certainly, if 1-in-5 school parents have retrogressively conservative views on  race, that'd be enough to make their voices known in a letter-writing campaign.

Now, to be clear, it is entirely possible -- plausible even -- that these emails didn't come from the 1-in-5 Republicans but from the 4-in-5 Democrats. "Democrats" are a wide tent, and there are, indeed, plenty of putative progressives who are on a hair-trigger about race issues and would be prime candidates to send out letters like these. I'm not saying that because these emails were racist, they couldn't have come from liberals. They very much could have.

What I am saying is that we can't say "because this neighborhood is staunchly liberal, these emails must have come from liberals." That's because that conclusion entails a shift from the accurate observation that this part of Seattle is overwhelmingly liberal, to the inaccurate observation that any political or social activity substantial enough to make it onto the social radar screen must be emerging from liberals. It's quite possible for conservatives in a place like Northeast Seattle to be simultaneously a marginal presence and a visible one, under the right circumstances. Ditto liberals in a place like Idaho.

More broadly, this is just a particular example of an obvious point: words with the same meaning can nonetheless communicate very different messages. When we want to erase the minority presence, we talk in percentages (20% is teensy-tiny!). When we want to elevate it, we talk in ratios (1:5 is really common!). Both are right, and in fact both connotations are right: a minority of 20% is a very small minority (as against an 80% majority), but 1:5 people is very common. Keeping both connotations in mind is good deliberative practice. Jumping from one to the other as argumentatively-necessary is very bad practice.