I did not get to watch much of the debate last night, but I learned this morning that at one point in the debate Bernie Sanders said that:

George, you talked about, was it 150 million people on private insurance? 50 million of those people lose their private insurance every year when they quit their jobs or they go unemployed or their employer changes their insurance policy.

Emmarie Huetteman of Kaiser Health News declared this claim “mostly false” for a piece she wrote at Politifact. Huetteman’s analysis of this claim is both strange and incorrect.

On Its Face

If you want to estimate how many people lose their employer-sponsored health plan every year, you need two figures: 1) the total number of people with employer-sponsored insurance, and 2) the percentage of those people who transition off their health plan every 12 months.

For the first figure, we can go to the latest Census health insurance figures that were released earlier this week. Those figures show that, in 2018, there were 178,350,000 people who had employer-sponsored insurance.

For the second figure, we have to be somewhat more creative because, as far as I know, there are not regular surveys that track people over time to see how often they switch off their current health plan. The most recent study I know of that did this was conducted in Michigan between 2014 and 2015. That study found that only 72 percent of people who were on an employer health plan at the beginning of the study remained on that same plan for 12 continuous months. This means 28 percent had lost their initial health plan over the year with some going into uninsurance and others going onto other health plans.

If 178,350,000 people are on employer-sponsored insurance and 28 percent of people on such insurance lose their current plan each year, then that means 49,938,000 lose their employer health insurance plan every year. This rounds up nicely to 50 million, which is the number Sanders gave.

Thus Sanders’s statement, read literally on its face, is actually correct, according to the best data we have on this subject.

Muddled Version Is Also Probably Correct

Huetteman never actually tried to assess whether the statement was literally correct, which it is. Instead she quickly detoured into whether Sanders’s statement was an accurate description of a piece I wrote back in July that uses data from the The Behavioral Risk Factor Surveillance System to claim that 50 million adults between the ages of 18 and 64 face a spell of uninsurance every year.

Huetteman is right that Sanders’s claim does not perfectly mirror the claim I made. Sanders says “people” when my post says “adults between the ages of 18 and 64.” Sanders says “lose their private insurance” when my post says “face a spell of uninsurance.” Sanders says these people face this fate “when they quit their jobs or they go unemployed or their employer changes their insurance policy,” which is only three of the approximately ten reasons why people fall into uninsurance.

But this is where we get into a bit of muddled terrain here. If we are analyzing Sanders literally, then he is correct for the reasons given above. If we are analyzing him non-literally, i.e. trying to get some sense of what he might be getting at, then we need to figure out what exactly he is trying to say here. And it seems that Huetteman believes that he is primarily trying to say that 50 million people face a spell of uninsurance every year.

This claim is also probably true though, at least based on the best data we have.

Huetteman’s only reason for believing that this is not true is that the BFFRS survey that this claim is based upon was conducted in 2014, and that the number of Medicaid expansion states has increased from 23 to 38 in the intervening 5 years. Note here that Huetteman does not offer any alternative estimates of how often people face spells of uninsurance. She just reasons that the Medicaid expansions must have brought that number down significantly.

But this reasoning does not actually get you to the conclusion that the claim is mostly false or even that it is probably false. In fact, even if you grant that Medicaid expansions cut uninsurance spells some, the claim is probably still true for two reasons that Huetteman does not consider.

1. Bernie used the word “people” but the 50 million number in my July post only counted “adults between the ages of 18 and 64.” The most recent Census release says that 4.3 million children and elderly people were uninsured for all of 2018, which means that even more than 4.3 million children and elderly people faced a spell of uninsurance during 2018. Bringing kids and old folks into the mix almost certainly offsets any reduction in adult uninsurance spells caused by the incremental Medicaid expansions.
2. Job separations have increased since 2014. Short uninsurance spells are driven in significant part by labor turnover, i.e. how many people separate from their job each year. According to JOLTS, the job separation rate increased by 10.6 percent between 2014 and 2018. This bump means that 2018 had 6 million more job separations than it would have had if the job separation rate had remained at 2014 levels, all else equal.

Of course, as a general matter, I would not advise trying to guess at an uninsurance figure using these kinds of methods. This is why, in my initial post, I just take the data we do have from the 2014 BFFRS and present it on its face. But if you are going to use that estimate as a baseline and then reason abstractly about how many people probably faced spells of uninsurance in 2018, then this is the kind of analysis you would do. Merely observing that more states expanded Medicaid is not enough in this case, and once you bring in kids and the elderly (“people”) as well as account for 10 percent higher labor turnover, you almost certainly find yourself right back at the 50 million figure Huetteman claims is mostly false.

Politifact should retract this story for multiple reasons.