• Extrapolate Medicaid crowd out estimates with caution

    Around the 1:16:30 mark in of this video of a symposium* on the ACA at the University of Michigan, Cato Institute’s Michael Cannon cited my work.

    Austin Frakt estimated that of the people who gain coverage under the Medicaid expansion 80% of them already have private health insurance. You have crowd out rates of 60% — Jonathan Gruber has clocked crowd out at 60% prior to the enactment of this law. Austin Frakt, another economist, has has estimated it as high as 80% for the Medicaid expansion.

    Actually, what Cannon is referencing is a working paper by Steve Pizer, Lisa Iezzoni, and me, which you can download here.

    In the literature, “crowd out”  is the proportion of the positive effect of Medicaid eligibility expansion on public insurance coverage that is offset by its negative effect on private coverage. Notice that this doesn’t quite answer the question Cannon is interested in: what’s the proportion of new Medicaid enrollees who already had private coverage? It doesn’t quite answer that question because that doesn’t account for the uninsured. You can get a 100% crowd out rate if only the uninsured take up Medicaid and, for every uninsured individual who enters Medicaid, a privately insured individual becomes uninsured.

    So, strictly speaking, the crowd out rate only tells you the growth in Medicaid relative to the reduction in private coverage. If there were no uninsured, the crowd out rate would be exactly what Cannon is interested in. When the size of the uninsured population is small and/or few uninsured take up Medicaid under an expansion of the program, Cannon’s interest and the crowd out rate are close.

    Cannon is correct that some work, including ours, has found crowd out rates well above 50%. However, other work finds lower rates. A lot depends on the precise estimation methodology, as well as the circumstances of the expansion considered. That’s why one must be careful extrapolating prior estimates to predict experience under the ACA’s expansion.

    A major point to consider is that the ACA includes an individual mandate, a feature not common to the circumstances surrounding prior crowd out estimates, including our own. That makes a big difference, as my coauthor Steve Pizer explains:

    Our estimates don’t apply to a world with a mandate. Without a mandate, most of the working near poor who prioritize insurance are already getting limited coverage somewhere. Medicaid expansion gets some of them a better deal, but doesn’t attract large numbers of currently uninsured. That means crowd out is high. But with a mandate, large numbers of uninsured will obtain insurance, so crowd out ceases to be a meaningful concept. Both public and private insurance rolls expand at the same time. The main thing getting crowded out is uninsurance.

    That’s not to say there won’t be any shift from private to public coverage. My coauthors and I acknowledge there will be. We just don’t know the rate from prior studies because they did not include a mandate, which has the effect of encouraging private coverage as well. And, as discussed above, we can’t infer much from crowd out rates when a Medicaid expansion attracts a lot of uninsured.

    This is corroborated by results from Massachusetts. There, the combination of a mandate and an expansion of Medicaid eligibility led to an expansion of private coverage.

    *You can read about the event here, or watch the archived webcast, which is in two parts: a policy panel and a journalism panel.


    • I think there is pretty much a consensus in health economics that Michael is using the right definition and you are using the non-standard one.

      Also, crowd out rates will be higher when there is a mandate. If you impose the mandate alone then you’ll get pretty high private coverage rates if the penalty is large enough but if you let people enroll in Medicaid instead a lot of people will switch to the free plan causing high crowd out.

    • You say

      “(In Massachusetts), the combination of a mandate and an expansion of Medicaid eligibility led to an expansion of private coverage.”

      Why point your readers to a four-year-old article — which is not even about Massachusetts — that uses six year old data? Anyone that wants to know what happened to private coverage (specifically employer sponsored coverage) in Massachusetts can get it directly (and as up to date as has been released) from the state of Massachusetts (see Figure 1 on page 4 of Massachusetts’ 2013 Annual Healthcare Market report, August 14, 2013, .

      Combining the three years shown in this three-month-old report with data released previously by the state for earlier years, “private (insurance) coverage” — combining ESI and individually purchased private insurance, counting Commonwealth Choice but not Commonwealth Care) has dropped by hundreds of thousands of people in Massachusetts (more than 5% of the population, not counting the possibly related movement of the disabled onto Medicare) since the mandate was instituted.

      (In addition, in Massachusetts, the mandate and Medicaid expansion took place almost a decade apart.)