A quibble with Avalere over CSR takeup

9/3/15: See update at bottom for Avalere's response to questions posed here 

I may have to "never mind" this post if I'm missing something basic, but...it seems to me that Avalere Health lowballed Cost Sharing Reduction (CSR) takeup a little in an analysis released today.

Avalere's calculation is simple: 5.9 million buyers of private plans on ACA exchanges bought plans with Cost Sharing Reduction subsidies, available only to buyers with incomes under 250% of the Federal Poverty Level (FPL), and only if they selected silver plans.  8.1 million buyers were eligible for CSR, according to Avalere; hence about 2.2 million left the benefit on the table; most of them probably brought bronze plans with sky-high deductibles. The CSR takeup rate comes to just 72.8%,

Avalere says that it derived these numbers from HHS's March enrollment report,which broke out enrollees by income level in the 37 states using healthcare.gov, and a CMS update released in June, which adjusted for attrition (mostly no-pays) as of March 31.

We know where the 5.9 million figure for those who accessed CSR came from: the June update, which reported 5,850,936 enrollees with CSR. But what about the denominator, the 8.1 million? That's trickier. We know there were 10,187,197 enrollees as of March 31. But how many were CSR-eligible?

For that we have to go back to the March report, unless I'm missing something  That report broke out income for the states using healthcare.gov, which accounted for just about three quarters of all enrollees, but not for the states running their own exchanges (state-based marketplaces, or SBMs).  In the healthcare.gov states, 83% of those for whom income data was available had incomes under 251%  FPL, and so were CSR-eligible. Since a half million enrollees did not report their incomes, and so almost surely earned too much to qualify for subsidies, that comes to 78% of all enrollees on the platform. Extend that figure to all 10,187,197 enrollees (on all exchanges) remaining on March 31, and you get 7.95 million.

But those 37 states probably don't reflect the demographics of the SBM states.  Three quarters of the healthcare.gov enrollees were in states that refused the Medicaid expansion, and in those states, eligibility for subsidized private plans began at 100% FPL rather than 138% FPL, as in the expansion states. In the states on hc.gov that did expand Medicaid, just 69% of enrollees were CSR eligible.

We don't know what percentage of the 2,662,964 enrollees in SBM states (as of 3/31) were CSR-eligible. But since all of them except Idaho expanded Medicaid, the income distribution probably looks more like that of the expansion states on hc.gov than the overall total, which is dominated by the non-expansion states. Let's assume for the moment that 69% were eligible -- though the percentage may be even lower, as the states that fielded their own exchanges are pretty wealthy on average.

If 69% of 2.66 million were CSR-eligible, and 78% of the 7.52 million on hc.gov were as well, that comes to 7.70 million CSR eligibles in total. That's 75.9% --  close to the 77% of CSR eligibles whom we know selected silver plans and accessed the benefit on hc.gov as of the March 10 enrollment report. It's significantly higher than Avalere's 72.8% figure -- which, again, is perhaps based on some figure hiding from me in plain sight.

As I've asserted many times, to me the crucial figure in CSR takeup is the percentage of buyers under 201% FPL who access the benefit. That's because CSR is really strong up to that level and almost negligible from 201-250% FPL.  CSR raises the actuarial value of a silver plan from a baseline of 70% to 94% for those with incomes up to 150% FPL, to 87% for those in the 150-200% FPL range, and to just 73% for those from 201-250% FPL. In states that break out metal level selection by income, CSR takeup falls to anywhere from 50% to 64% for eligibles in the 201-250% FPL range. 

On that basis, I've calculated previously that CSR takeup for those under 201% FPL was probably a bit over 80% in the 37 states using healthcare.gov. While adding SBM states seems to depress total takeup a bit, in those states a higher percentage of CSR-eligibles are probably in the upper range of eligibility (201-250% FPL). So I think we're at status quo ante -- a bit more than 80% of buyers with incomes up to 200% FPL buy silver.

Too many low-income ACA customers buy bronze plans, which rarely have single-person deductibles below $5,000. But most don't.

P.S. An afterthought: My guess is that Avalere is assuming that a higher percentage of enrollees for whom income is unknown were CSR-eligible than I am. I'm assuming zero.

UPDATE, 9/3/15: I heard from Avalere manager Chris Sloan, who told me two things that explain the difference in our estimates. First, Avalere does not assume that the marketplace enrollees who do not report their income earn too much to qualify for subsidies; rather, they distribute those "unknowns" evenly throughout the population, leaving the percentages at each income level unchanged. Second, Avalere did not use the expansion states on healthcare.gov as a proxy for the SBM states -- they applied the overall percentage of CSR-eligibles for all healthcare.gov states to the SBMs.

While no one really knows the percentage of income non-reporters who earn too much to qualify for subsidies, Larry Levitt has suggested to me that it's reasonable to assume that almost all of them do. A small number might be non-citizens who illegally access the marketplace and bought without benefit of subsidies; a few might also be subsidy-eligible applicants who made application mistakes, such as stating that they did not intend to file a tax return, which would disqualify them for subsidies. While all too many exchange visitors may have made such errors, presumably few could afford to buy unsubsidized plans.

As to the second point, I think it's highly likely that the income distribution of buyers in SBM states would look more like the nonexpansion states on hc.gov than on the average for all states on the federal platform. In sum, I do think it likely that total CSR takeup was closer to 76% in 2015 than to 73% -- and, again, about 5 points higher for those under 201% FPL.  But perhaps I go too far in assuming functionally 100% of "income unknowns" are not eligible for CSR (or any subsidy).

UPDATE: A CMS enrollment update, released in early Sept. and showing effectuated enrollment through June 30, shows enrollment down  to 9.9 million. That would suggest about 7.3 million CSR-eligibles. The report shows 5.57 million CSR enrollees.