Obamacare Exchanges Were in Big Trouble Before Trump

From Megan McArdle.

"Healthcare.gov enrollment came in well below what was anticipated last month. After running very slightly ahead of last year’s numbers in December, January brought the news that about 400,000 fewer people had enrolled on the federal exchanges than did so in 2016. Those are scary numbers, not so much for the absolute size of the decline -- it’s roughly 4 percent -- but because any backwards movement is very bad news for the exchanges.

As many readers will recall, 2016 brought hefty premium hikes to make up for years of losses on the exchanges. Going into this year’s open enrollment period, there seemed to be a significant risk that we’d see what economists call “adverse selection”: People who were getting good value out of their insurance (because they used a lot of services) would keep paying the premiums, even if they grumbled a bit, but people who didn’t use a lot of health care would decide that at those prices, they might as well go uninsured. When those people exited the market, the average cost to cover each person remaining in the pool would be higher … and insurers still wouldn’t make money, forcing them to raise premiums again next year. This process is known as the “adverse selection death spiral.”

December’s numbers seemed to indicate that people were still willing to buy insurance at the new, higher prices. January’s numbers suggest that maybe they aren’t. Most worryingly, young people, who don’t worry that much about the health insurance they rarely use, tend to sign up fairly late in the game, so a rise in December and a decline in January means that the pool could end up substantially older and sicker than last year’s.

Okay, but why are liberals blaming all of this on President Donald Trump? The man was only president for a few days worth of open enrollment. Could he really have somehow caused 400,000 people to forgo health insurance?

That’s exactly what people like Charles Gaba are suggesting. The culprit: outreach advertising designed to encourage people to sign up, which Donald Trump cut. Gaba points out that many states who run their own exchanges didn’t see the same kinds of declines that the federal exchanges did. That seems almost like a natural experiment, which makes it persuasive.

Except: We’re talking about $5 million worth of advertising over a brief period, and it’s not clear how much of it actually failed to run. If HHS advertising is that effective, then they need to open a branch office on Madison Avenue. In other words, the eyeballing of the data looks okay, but the mechanism is frankly unbelievable. Moreover, if advertising, or some other federal exchange policy, explains it, then how come Connecticut, California and Maryland also saw declines?

You can argue that the problem was not the advertising, but the fact that the Trump administration was obviously against Obamacare. And I suppose that’s possible -- except that Trump won in November, so how come people only reacted in late January? It’s not as if it was exactly a secret that Republicans wanted to repeal the thing. Nor do I find it plausible that potential customers were following the play-by-play of fiddling changes to health care policy during the last 10 days of the month -- in part because the news was completely dominated by other things, but mostly because I’ve never seen evidence that these consumers were following this sort of deep-in-the-weeds wonkplay at any time in the history of the program.

Besides, there’s another mechanism that’s just as plausible: Young and healthy people tend to sign up late, and young and healthy people are the ones who are most likely to have balked at higher premiums. Say Hillary Clinton had won, and her administration ran the ads just as planned. However, suppose as well that (unbeknownst to her), Obamacare was poised for a decline. What pattern would we expect to see? Pretty much the one we did: December would still look about like last year, but January would come in lower, because the people most likely to fail to buy insurance are the young, healthy folks who buy late. We’d see some states where enrollment increased, and some where it fell, because small populations do not mirror larger ones exactly. But the overall impact would be a late-January decline.

We’ll never really know the answer. I spent a morning slicing and dicing the data looking for some relationship to premiums -- size of premium increase, ratio to local incomes, absolute size of premium. You can kind of tease a relationship out, if you look hard enough, but you have to squint pretty hard. The same is true of the state/federal distinction, however: Gaba ends up separating out various states for various reasons, which can be reasonable, but can also be a good way to fool yourself into thinking that you’ve identified an effect that isn’t there.

After long squinting, the best I’m willing to say is that smaller states tended to show larger effects one way or another: Hawaii’s enrollment increased 30 percent, Mississippi’s declined 20 percent. 

Unfortunately, that’s the kind of “information” that anyone could have predicted using nothing more than a basic familiarity with statistics. It’s another instance of the Obamacare Rorschach blot, offering policy experts as much insight into their own psychology as to what’s on the page.

What we can say, however, is that any of these theories of the decline suggests that Obamacare was already incredibly fragile. The program was either:

a) primed to decline anyway,
b) primed to decline as soon as a Republican took the White House and voters began worrying about the future of the program, or
c) so vulnerable that a small amount of advertising could make the difference between enrollment growth and a significant decline.

All three of these theories suggest that this program badly needs to be replaced with something that doesn’t begin to topple as soon as anyone looks at it funny."