Rand Corp. Survey "3.9M enrolled" Mystery Resolved (for real!)

A few days ago, the much-discussed Rand Corporation healthcare coverage survey was released.

There's a lot to absorb and discuss in it, but the one number which I found the most puzzling was their estimate of around 3.9 million private QHP enrollments via the ACA exchanges.

As I noted at the time, this puzzled me because around "mid-March" (the point at which the vast bulk of survey responses had come in), the actual number of exchange QHPs was roughly 1 million higher, around 4.9M.

The survey itself lists a +/- 1.1 million margin of error for that figure, but it still didn't make much sense to me, since the official HHS report through 3/01 (2 weeks earlier) was already 4.24 million.

At first, I figured that perhaps they were lopping off about 20% to account for the unpaid enrollments. This would make the 3.9M figure correct, but there was no mention of payments in the survey.

Then, however, a contributor going by the handle of Simple_Soap had a "Eureka!" moment. As he describes it:

Charles, I finally got around to downloading and reading the actual Rand survey. I think i had an Eureka moment when I read Table 3. The population base is 198.5M. The survey tracked the movement of the 18-64 group. There are ~ 2M in that group according to Census data. In other words children were NOT included. That would explain your 4.9 vs 3.9. Rand did not get involved in the paid/unpaid issue.

The 9.3M increase is relative to 198.5M.

Also Rand has the 2013 uninsured pop. at 40.7M when the other data sources, Census GPS, KFF, Urban Inst. etc have it at 48-50M TOTAL pop.
What do you think?

This seemed a bit outlandish to me, but I decided to look into it. Yes, I had seen the reference to "18-64" in the survey itself...

The results presented here are based on 2,425 adults between the ages of 18 and 64 who responded in both March 2014 and September 2013.

...but I just assumed that this only referred to the person who was filling out the survey for the household. I figured they weren't allowed to have minors fill out the forms or whatever, but thought that whoever did was including all of the household info, including kids.

Simple_Soap followed up by contacting Rand Senior Economist Christine Eibner, one of the co-authors of the study, and reported this:

I'm certain the 0-17 pop. was not included. Christine confirmed it. The survey tracked the changes for ~ 2,500 adults from Sept to March and then extrapolated the changes to their total population (18 -64) of 198.5M.

If you were sampled both you and your wife would be counted in the total population but not your child. There are ~ 72M children not sampled or counted.

The survey also estimated that ~ 36% of the QHP marketplace adult enrollees were previously uninsured but only 6.4% off-exchange enrollees.

Just to be certain (since this was still coming in 2nd-hand), I contacted Rand Economist Katherine Carman, the other co-author, and sure enough, she confirmed it:

The Rand study not only did not include any children, it didn't even include spouse/partner data. In other words, each person who was surveyed only gave information about themselves for the study, not other members of their households.

If I had been surveyed, there would be no information about my son or wife; our household would only count as 1 QHP enrollee, not 3.

I'll be discussing the study with Ms. Carman further as soon as I can, but for now...thus endeth the mystery.

UPDATE: Just to clarify, I'm not saying there was anything wrong with the survey, just that at least one part of it is being misinterpreted by a lot of people.