More Puzzling Obamacare Numbers

The best data set we have on Obamacare enrollment is a puzzling but useful study.
Photographer: Andrew Harrer/Bloomberg

Late last month, the Los Angeles Times got a lot of buzz when it used data from an unpublished RAND Corp. report to assess the state of Affordable Care Act enrollment. That report is now available, and a lot of folks have been poring over its findings. Actually, puzzling over its findings might be a better way to put it.

Before we get into why it's so puzzling, here's the rundown on the study and what it found:

  • Based on an ongoing survey of 2,425 working-age adults (ages 18 to 64), it tracked the changes in their responses from September 2013 to March 28, 2014. The numbers presented are extrapolated from that sample to the general population.
  • There is a lot of churn in the insurance markets generally, with people moving from employer insurance to being uninsured to Medicaid to the Veterans Administration to Medicare. The RAND study was no exception:


  • According to the RAND sample, 40.7 million people were uninsured in 2013; this dropped to 31.4 million in 2014. However, 4.4 million lost their insurance during the same period. The net gain in the insured during the open enrollment period thus comes to 9.3 million.
  • Most of the gain came from employer-sponsored insurance, followed by Medicaid. Fewer than 2 million formerly uninsured people purchased individual policies, either on or off the exchanges.
  • Most people who lost insurance over the same period previously had employer-sponsored insurance. The next biggest group was "other," followed by those on Medicaid. Only 700,000 people from the individual market lost insurance entirely; most migrated to another form of insurance or bought new individual policies outside the exchanges.
  • A total of 3.9 million purchased exchange policies through March 28, 2014. About a third of them were people who had been previously uninsured.
  • For 2014, 18.2 million ended up on Medicaid. Of that number, 3.6 million were previously uninsured; the rest were people who had previously had Medicaid or migrated there from some other form of insurance.
  • The number of people with employer-sponsored insurance rose by a lot, to 116 million from 108 million. Most of the increase came from people who had been uninsured, or had "other" insurance.
  • The number of people with "other" insurance shrank by 7.2 million. Most of those people ended up either on employer-based insurance or uninsured.

OK, so why is this puzzling? Well, umm, what's all that employer-sponsored insurance doing there? Why did so many people with "other" insurance lose it? There is nothing in Obamacare that should have caused either outcome; the employer mandate hasn't even taken effect.

I mean, I could tell a story about how the exchanges make a trivial contribution to solving the problem of the uninsured, but a lot of uninsured people who are afraid of the individual mandate bite the bullet and sign up for that employer-sponsored insurance they've been declining because their share of the premium is $150 a month. Where the existence of the exchanges causes a lot of companies to dump their retirees onto the individual market in order to pick up some subsidies. Where people who already had individual policies take one look at the new premiums they have to pay and decide it's better to just sign up for their spouse's insurance, even if they have to pay the whole premium for the additional coverage. It's not a particularly flattering story for Obamacare, but it's a story you can string together from these data.

But this isn't the story I would have been telling on March 30, before I'd seen some of the RAND results. I hadn't heard much about widespread cancellation of retiree benefits, nor about hordes of people rushing to sign up for employer-based policies. I'm not saying that it couldn't have happened, especially if much of the changes were taking place at small and medium-sized businesses, where benefit changes rarely attract much publicity. I'm just saying that if this was happening, folks were awfully quiet about it.

And I could also tell a story where the midpoint of the RAND estimates are off -- 2,500 people is a nice big sample, and following them over six months means that you can actually track changes among individuals, rather than trying to guess about aggregates. For that reason, RAND's work is very valuable.

However, it is only 2,500 people. Random chance could have thrown the results off by quite a bit. That's why the error bars are quite large.

In fact, we know that at least one midpoint is quite far from the truth, because there is no way to reconcile the 3.9 million people it says bought exchange policies with the Barack Obama administration's official data showing that 7.3 million had selected exchange plans by the end of March. Even if we factor in the three days the survey missed -- which saw a titanic surge in enrollments -- and even if we assume that only 80 percent of the people who selected a plan have actually paid for it, 3.9 million is too small. Well before data collection ended, the administration had already announced that plan signups had hit 6 million. Even with attrition of 20 percent, that should still have yielded 4.8 million marketplace policies in the RAND study. And indeed, 4.8 million is within its margin of error, though at the very high end.

Eventually, some of these questions will be resolved; we will know how many exchange policies are in effect and how many newly eligible people went on Medicaid, because the administration needs to know that number in order to pay for them. Until then, the best data set we have is this puzzling but useful study.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

    (Megan McArdle writes about economics, business and public policy for Bloomberg View. Follow her on Twitter at @asymmetricinfo.)

    To contact the author on this story:
    Megan McArdle at

    To contact the editor on this story:
    Brooke Sample at

    Before it's here, it's on the Bloomberg Terminal.