Health insurance follies

I was looking at my HSA web site the other day, trying to pay some bills, and looked in particular at the charges for my recent periodic endoscopy/colonoscopy (isn't being Over Fifty wonderful?). There were two charges of $1545.50, apparently one for the endoscopy and one for the colonoscopy. As I've discussed here and here, the "amount billed" is pure fiction, having little resemblance to what anybody actually pays or gets paid.

But there was something unusual about this one. For one of the two procedures, the "allowed charge" was $1128.50. For the other, it was $2257.00, larger than the "amount billed" and (I think not coincidentally) exactly twice the other allowed charge, as though somebody had entered the number into a spreadsheet twice and the software had added the two numbers. 10% of each of these was charged to me, and the other 90% to the insurance company. (It's nice to have a wealthy employer with generous bennies....) As a result, my insurance company appears to have paid the hospital an extra $1015.65 that they shouldn't have paid, and I'm being charged an extra $112.85 that I shouldn't have to pay.

I contacted the HSA and was told "we just enter whatever's in the EOB we get from the insurance company; it's not our problem." So I called the insurance company; the customer service agent didn't see the same numbers I did, but agreed that something sounded fishy. She thanked me for bringing this to her attention and suggested I wait until I get an EOB in the mail. I'm not sure they'll actually do anything about it; the extra $1015 they paid may not be worth the trouble of getting it corrected.

Brexit, or, Designing Poll Questions is Hard

In 2016, the British government posed the British people the following question:

"Should the United Kingdom remain a member of the European Union or leave the European Union?"

The wording of this question was the result of months of debate, and was determined by the Electoral Commission to be clear, unambiguous, and neutral... and yet as we can all see now, it wasn't. It compared a concrete option (staying in the EU, status quo ante) with an abstract option (leaving the EU) on which different voters could project whatever they wanted. If a lot of them projected fear, it would lose; if a lot of them projected hope, it would win.

As you know, the latter happened. People in favor of a "soft" Brexit with a free flow of goods and people between Ireland and Northern Ireland could vote for it. People in favor of a "hard" Brexit with a customs border between Ireland and Northern Ireland could vote for it. People in favor of a "no-deal" Brexit" could vote for it. People in favor of returning to a romanticized 1972, before Britain joined the EEC, could vote for it. People in favor of Northern Ireland or Scotland seceding from the UK could vote for it as a spur to that outcome. Even some people who thought EU membership basically worked for the UK but needed tweaking could vote for it as a "protest vote".

As a result, "Remain" didn't get majority support. I'm pretty sure that if you took an up-or-down referendum on any of these other options, none of them would get majority support either. This is the problem with trying to answer a multi-way question using a series of binary questions: if I get to choose the order of the binary questions, I can get almost any desired answer to win. In the first round, I put the option that I think poses the strongest competition on one side, and all the others on the other side: unless the strongest option actually has majority support, it will lose, and then I can go on to eliminate the remaining competing options one by one in the same way. By the last round, when I've winnowed the field to two thoroughly unpopular options, most of the voters may have tuned out entirely and not vote at all.

So reducing a multi-way public-policy question to a series of binary questions works poorly in practice. Why do we do it? Because we know how to handle binary questions, and multi-way questions are harder to think about. People are either male or female; there is no third option. People are either white or non-white (in which case we go on to ask other binary questions).

A voting system that allows people to cast one vote for their favorite choice works very well when there are two choices. It works less well when there are three choices, and really badly when there are more than that. It fails in a predictable way: it favors not the most popular option, but the option that differs most from the rest of the options. Witness the 2016 Republican Presidential primaries, in which lots of similar candidates who would have beaten Trump in head-to-head popularity contests split the vote, and the one candidate with no history of public service, no history of elected office, and no qualifications whatsoever in economics, military, diplomacy, or law went on to win the nomination and eventually the Presidency. And it seems entirely possible that something similar will happen to the Democrats this time around.

The world doesn't actually come in clean binary choices.