*The following is co-authored by Austin Frakt and Garret Johnson. Garret is a research assistant for Dr. Ashish Jha at the Harvard T.H. Chan School of Public Health, where he also works with Austin on Medicare Advantage studies. *

The recent paper by Kate Baicker and Jacob Robbins estimates a spillover effect from Medicare Advantage (MA) to traditional Medicare (TM). The idea is that MA, through care management, influences patterns of care in such a way to reduce costs, which then affects the quality and costs of care for TM beneficiaries as well. (Prior posts on this kind of spillover here and here).

We wanted to convey the estimates in the paper in a different way, one we find more helpful. Here, using their results, is the answer to the question, for the marginal dollar spent on MA, how much is saved in TM?

- They find that $1200 in additional payment per MA enrollee per year yields 2.2 pct points higher MA market penetration.
- They also calculate a $252 per TM beneficiary per year in spillover “savings” for every 10 percentage points in higher MA market penetration. (The savings come through more efficient use of hospital services, in particular shorter lengths of stay, mediated by a concurrent increase in less costly outpatient service use. Therefore, it’s not really immediate savings to Medicare since diagnosis-based Medicare hospital payments aren’t sensitive to shortening lengths of stay. But, longer term, perhaps these savings could be captured through changes in rate increases.)
- Therefore, combining (1) and (2), for each $1200 per MA enrollee per year in additional payment there’s a $252 × (2.2/10) of spillover savings per TM beneficiary per year, or $55.44. Dividing, that’s $0.0462 of spillover per TM beneficiary for every $1 of payment for an MA enrollee.
- There are a lot more TM beneficiaries than MA enrollees, so we need to scale these. At the midpoint of the authors’ study window (2005), MA had a market penetration of 13%, leaving 87% in TM. Multiplying these by the figures from (3), we get $0.040 of spillover per $0.13 in MA payment. That’s $0.31 of savings per $1 of payment.
- The results of (4) is at the particular margin examined. Not every $1 of MA payment is associated with 31 cents of savings.
*Extrapolating out of sample*, today, when the MA/TM enrollment split is 31%/69%, the savings is closer to 10 cents on the dollar. (This is just repeating the calculation in step (4), but with 31%/69% instead of 13%/87%.)

There are lots of caveats:

- We mentioned it above, but it should be emphasized that these figures only pertain to the time period (1999-2011) and range of MA payment (from about 5% to 15% above TM costs) and market penetration (9%-11.9%) examined. In particular, our out of sample extrapolation is not likely to be correct. We just did it to illustrate how the results change a lot as market penetration changes.
- Related, the results above are for the marginal dollar spent on MA. What about all the other dollars? No doubt, some of them contribute toward TM savings, but some do not. In the paper, the authors estimate TM utilization as a quadratic function of MA penetration. For some lower range of penetration, MA is associated with higher per person TM costs (perhaps due to favorable selection into MA). Spillover effects probably taper off at a higher range of penetration, too.
- What the total cost or savings of MA to TM is today, we cannot tell. We don’t even know if it’s net positive or negative. That estimate would require more estimation work and a bunch of assumptions.

This is, on the one hand, interesting. Some MA spending is associated with a high degree of TM savings (31 cents on the dollar circa 1999-2011 is *HUGE*). To the extent that MA growth makes TM “cheaper” for taxpayers, then some amount of payment above TM rates to MA plans may be efficient. On the other hand, this is unsatisfying. We don’t really want to know if the marginal dollar spent on MA in, say, 2005 saved money. We want to know if, in total, MA is saving or costing money today. We just don’t know.