Why 2019 ACA enrollment drops were concentrated in Medicaid expansion states on HealthCare.gov, Take 4

Subscribe to xpostfactoid via box at top right (requires only an email address; you'll get 2-3 emails per week on average)

In December and January, I took several whacks at explaining why 2019 enrollment losses in the 39 HealthCare.gov states were concentrated in the states that had expanded Medicaid. In those 21 states, plan selections as of the end of Open Enrollment were down 7% from 2018, compared to a 1% drop in 16 states that have refused to expand Medicaid. (I excluded Virginia and Maine, which expanded Medicaid in 2019, affecting enrollment options and results.)

First hypothesis was that the heavy concentration of enrollees in non-expansion states with incomes that would have qualified them for Medicaid in expansion states is especially "sticky." People in this income group (100-138% of the Federal Poverty Level) pay just 2% of income for a benchmark silver plan that comes with the highest level of Cost Sharing Reduction (CSR), raising the actuarial value of a silver plan to 94%, which usually translates to a deductible in the $0-500 range.  About a third of enrollees in nonexpansion states are in this income category. While about 15% of enrollees in expansion states have incomes in the 138-150% FPL range, which qualifies them for the 94% AV silver, they pay 3-4% of income for the benchmark.

Second hypothesis was that silver loading effects are stronger in nonexpansion states, because all those low income enrollees raise the average actuarial value of silver plans (which varies with income). In Florida, the blended AV of all silver plan enrollees is 91.5% in 2019.  (If you're unfamiliar with silver loading, see the note at bottom.)

Recently a third possible factor occurred to me: expansion states have a higher percentage of enrollees with incomes too high to qualify for subsidies, and enrollment losses among the unsubsidized were much steeper in the market as a whole than among the subsidized. This factor did play a role, but it was partly offset by the fact that in this category losses were steeper in nonexpansion states. The higher concentration of unsubsidized enrollees in expansion states cancelled out that advantage, however, so that losses in unsubsidized enrollment took basically equal bites out of total enrollment in both groups.

Subsidized and Unsubsidized Enrollment Change, 2018-2019

Expansion vs. Non-expansion States, HealthCare.gov, as of the end of Open Enrollment 2019

State group	Subsidized enrollment 2019 vs.2018	Unsubsidized enrollment 2019 vs. 2018	Decrease in 2019 unsubsidized as % of 2018 enrollment
Expansion*	-6.0%	-12.0%	2.4%
Nonexpansion	+1.1%	-17.7%	2.4%

* Excluding Virginia and Maine
   Source: CMS state-level public use files 

This breakout led me to compare enrollment performance between the two groups at subsidized income levels as well, which wasn't yet possible when I first hypothesized that the 100-138% FPL cohort in nonexpansion states was likeliest to maintain enrollment, as CMS's public use files weren't out yet.

First, a look at the very different income distribution of enrollees in expansion vs. nonexpansion states. I've included overlapping categories 100-150%, 150-200% and 100-200% FPL to highlight the high concentration of should-be-in-Medicaid enrollees in nonexpansion states (unfortunately there's not a separate breakout of 100-138% FPL, but they're about 85% of the 100-150% FPL total in nonexpansion states.)

Income distribution of HealthCare.gov enrollees, 2018

Expansion vs. Non-expansion states

State group	100-150% FPL	150-200% FPL	100-200% FPL	200-400% FPL	Unsubsidized
Expansion*	15%	25%	40%	44%	20%
Nonexpansion	44%	20%	63%	28%	12%

* Excluding Virginia and Maine
  Totals sum to more than 100% because some enrollees at 100-400% FPL are not subsidized

  Source: CMS state-level public use files 

Now, a look at enrollment performance at different income levels in each group.

Enrollment changes by income, HealthCare.gov states

2019 enrollment as a percentage of 2018, as of the end of OE in each year

State group	Total enrollment	100-150% FPL	150-200% FPL	100-200%  FPL	200-400% FPL	Unsubsidized
Expansion*	93%	93%	91%	92%	95%	88%
Nonexpansion	99%	100%	98%	100%	102%	82%

* Excluding Virginia and Maine
   Source: CMS state-level public use files 

The enrollment performance gaps at 100-200% FPL and 200-400% FPL are comparable. I would be tempted to suggest that the gap at 100-150% FPL is due to the concentration of low income enrollees in the nonexpansion states (those at 100-138% FPL, who pay less for top-level CSR silver plans than do those at 138-150% FPL), and that the gap at 200-400% FPL is due to stronger silver loading in the non-expansion states. But the large performance gap at 150-200% FPL kind of belies that -- unless, perhaps, silver loading  at that income level is pulling more prospective enrollees into free or ultra-cheap bronze plans in the nonexpansion states. In the silver loading era, silver plan selection at 150-200% FPL has dropped from 83% in 2017 (pre silver loading) to 76% in 2019 in HealthCare.gov states. The lure is bronze that is often free at that income level, while benchmark (second cheapest) silver costs about $130/month at 200% FPL. I guess I need to test whether bronze selection in this income band has risen faster in nonexpansion states.

Some further caveats to these rather nebulous conclusions are in order. First, myriad factors affect enrollment in each state -- for example, the haphazard discounts or lack thereof generated by premium spreads (silver loading provides grist for discount-generating spreads, but many other factors also come into play); the behavior of monopoly insurers, which exist in many states and regions; insurer entries and exits; the degree of oversight (or conspicuous lack thereof) exercised by regulators; providers' pricing leverage in rural or concentrated markets; employment in each state, etc. etc. Any of these factors can swamp the effects of enrollee income distribution in any given state, and often do.

Then there's behemoth Florida, the most robust market in the country for idiosyncratic reasons, where enrollment rose 4% in 2019, and which alone accounts for 31% of 2019 enrollment in nonexpansion states on HealthCare.gov. Further, the 12 states (including D.C.) than run their own exchanges have outperformed HealthCare.gov states, notwithstanding their wealthier enrollment pools (all but one, Idaho, have expanded Medicaid): enrollment in SBEs was higher in 2019 than in 2018 -- or 2016. State governments in SBE states are committed to making the marketplace work, e.g. through advertising and enrollment assistance, extra state-based subsidies, and more active rate review. Finally, the enrollment drops recorded at the end of each Open Enrollment period from 2017 through 2019 compress considerably, and may disappear entirely, when effectuated enrollment, average monthly enrollment, or enrollment as of December in each year are compared. Retention appears to have improved across all these measures.

Update, 10:30 a.m.: Man, was this in need of a proofread, now provided. Sorry!
--
Silver loading is the byproduct of Trump's October 2017 cutoff of direct federal reimbursement to insurers for the Cost Sharing Reduction (CSR) subsidies they are required to provide to low income marketplace enrollees who select silver plans. Faced with the cutoff at the brink of open enrollment for 2018, most state insurance departments allowed or encouraged insurers to price CSR into silver premiums only. Since premium subsidies, designed so that the enrollee pays a fixed percentage of income, are set to a silver plan benchmark, inflated silver premiums create discounts for subsidized buyers in bronze and gold plans.

For enrollees whose incomes are too high to qualify for CSR, a silver plan has an actuarial value of about 70%. Prior to Trump's CSR reimbursement cutoff, silver plans were priced as if they were 70% AV. But CSR raises the AV of a silver plan to 94% for enrollees with incomes up to 150% of the Federal Poverty Level, and to 87% for those with incomes in the 150-200% FPL range. The more low income enrollees a state has, the higher silver's "real" AV, and the more expensive it should be.

Related:
Silver loading and 2019 ACA enrollment: A compendium
Why 2019 ACA enrollment drops were concentrated in Medicaid expansion states on HealthCare.gov, Take 3
Two key factors in state-by-state ACA enrollment performance