Medicaid expansion states on HealthCare.gov account for nearly all enrollment losses in 2019

In yesterday's post, I  missed the most startling point in the ACA enrollment data I'd been staring at.

It's this: In 2019, virtually all of the enrollment decline in the ACA marketplace was concentrated in states on the HealthCare.gov platform that have expanded Medicaid. That's excluding Virginia and Maine, which are expanding Medicaid in 2019.*

Charles Gaba projects virtually flat enrollment in the states that run their own exchanges, in many of which enrollment for 2019 is still open. In the 16 remaining non-expansion states using HealthCare.gov, enrollment in 2019 is down less than 1% from enrollment in 2018 (excluding late adjustments, which should roughly cancel out similar adjustments made last year). Enrollment in the 21 HealthCare.gov states that have expanded Medicaid (excluding VA and ME) is down 7% in 2019. It was also down 7% in 2018, compared to 4% the in non-expansion states.

As I noted yesterday, outsized enrollment losses in 2018 were concentrated at lower income levels, where enrollment assistance is probably most vital, and where the value of Cost Sharing Reduction (CSR) mostly outstrips the value of  bronze and gold discounts generated by silver loading** (an income breakout is not yet available for 2019). In HealthCare.gov expansion states, enrollment in the 100-150% FPL income band, where CSR is strongest, cratered in 2018, dropping 14%. At 150-200% FPL, enrollment in these states dropped 11%.

At higher income levels, as was the case generally in 2018, silver loading discounts in these states seem to have partially offset the forces driving enrollment down at lower income levels. At 300-400% FPL, enrollment in these states was up 8% in 2018 (it was up 12% in non-expansion states).

Two key factors in state-by-state ACA enrollment performance

Last week, I noted that in the 39 states using the HealthCare.gov platform, enrollment declines from 2018 to 2019 were generally steeper in states that had accepted the ACA Medicaid expansion than in states that have so far refused to expand.

Excluding Virginia and Maine, which are expanding Medicaid as of Jan. 1 2019, 13 of the 18 states with enrollment performance above the hc.gov median are nonexpansion states, while 15 of 18 below the median are expansion states (as is Kentucky, at the median).

My working assumption is that a measure of enrollment stability in non-expansion states is grounded in the one third-plus of enrollees in those states who have incomes that would qualify them for Medicaid in expansion states (100-138% of the Federal Poverty Level). This is the income level at which ACA marketplace offerings are most affordable and comprehensive.  At 100-138% FPL, the premium for a benchmark silver plan with an actuarial value of 94% (average deductible $355) costs just 2% of income.

Apparently cutting against this hypothesis is the fact that in HealthCare.gov states in 2018, enrollment dropped 7% at 100-150% FPL while remaining essentially flat at 200-400% FPL, as at the higher income levels, silver loading discounts (see note at bottom) largely offset the effects of cuts to advertising and enrollment outreach. But the drop at 100-150% FPL was much steeper in the expansion states on the platform, where subsidy eligibility begins at 138% FPL (so that the income band is effectively 138-150% FPL). In expansion states on HealthCare.gov, enrollment in this income band cratered 14% in 2018, compared to 6% in the nonexpansion states.* At 138-150% FPL, a benchmark silver plan -- still at 94% AV -- costs 3-4% of income, compared to 2% of income for those below 138% FPL.  And while silver loading reduced enrollment losses at 200-400% FPL in 2018, the resulting discounts in many cases evaporated in 2019 (though new discounts also emerged).**

The chart below sets 2019 enrollment performance beside the percentage of enrollees in a state who obtain silver plans with the highest level of CSR, which raises the actuarial value of a silver plan to 94%.  That percentage is generally much higher in nonexpansion states, where over three quarters of the enrollees in 94% AV silver would be in Medicaid if their state had accepted the expansion. The relationship is significant, I think,  but can be trumped by silver loading effects, as they are in some (not all) of the highlighted states below, as well as by other factors noted below. In states marked in red, the relationship between the percentage of enrollees at 94% AV silver and enrollment performance is most sharply out of whack; in those marked in tan, more moderately so. The apparent effects (or lack thereof) of premium changes from 2018 to 2019 in highlighted states are discussed below.

The median percentage of enrollees with 94% AV silver plans in these 37 states is 19%. The median enrollment change is Kentucky's -5.38%.  I have excluded Virginia and Maine, as they are in the midst of Medicaid expansion.

Sources: Charles Gaba, CMS state-level Public Use Files, 2018.

About that "late enrollment surge"

At the end of Week 6 of Open Enrollment for 2019 in the 39 states that use HealthCare.gov, all-state enrollment, at 4,132,432, was 88.3% of the Week 6 total for 2018.

By the end of Week 7, the final week, the 2019 total (8,454,882) was 95.8% of the 2018 total. Surprise! News stories reference a late enrollment surge. But the surge was a mirage -- generated by a slight difference in the weekly breakdown of an enrollment season that was 45 days in both years. Or rather, the surge was normal and happens every year -- this year the pace did not accelerate more than last.

In CMS's weekly enrollment snapshots, "Week 1" had four days in November 2017 and three days in November 2018. "Week 7," conversely, had six days in December 2017 and seven days this December.

Enrollment season starts slowly and ends fast. This year, enrollments averaged 123,892 per day in Week 1 and 617,493 in Week 7. More than half of all enrollment -- 51% -- happened in the final week (boosted by about 1.7 million auto-enrollments or passive renewals, added in at the end). Last year, 47% of enrollees were booked in the final (6-day) week.

A key factor in state-by-state enrollment performance on HealthCare.gov

Okay, a rather elemental discovery regarding ACA marketplace enrollment for 2019, which was down 4.2% overall in 39 states using HealthCare.gov.

Courtesy of Charles Gaba, here is the enrollment performance by state, 2019 vs. 2018, with a little coloring book intervention of mine in the far-left column. What distinguishes the highlighted states?

The invisible primary for judicial repeal of the ACA

Slapped on a Friday evening with a hyperpartisan judge's palpably ridiculous decision purporting to strike down the entire ACA, I suspect that many progressive minds restarted a two-stroke motor that's been running since Texas v. U.S. first stained the horizon.

Stroke 1: this suit is too ridiculous even for Republican judges.  It claims in effect that Republicans in Congress repealed the entire ACA in a fit of absent-mindedness when they zeroed out the mandate penalty last December.

Stroke 2: we had the same reaction to the two prior anti-ACA suits to reach the Supreme Court, one of which failed by a 5-4 and the other by a 6-3 count. Will the absurd again become Republican orthodoxy?

Will this legal nightmare recur? Jack Balkin offers crucial perspective:

Blue waivers! Synthetic silver loading, anyone?

First the bad news: Seema Verma's CMS has invited states to undermine guaranteed issue and comprehensive coverage in their ACA marketplaces, floating four "waiver concepts" that converge on one idea: subsidizing medically underwritten, lightly regulated insurance.

Now the good news: CMS has invited states to be creative. Blue states might think about taking up the offer to redesign the ACA marketplace subsidy structure. Those states, however, will want to stay within the statutory "guardrails" for ACA Section 1332 innovation waivers that recent Trump administration guidance seeks to, um, wave away: provide coverage as comprehensive and affordable to as many people as does the existing marketplace design, without increasing the federal deficit.

The financial constraint -- deliver better benefits without spending more money -- has rendered Section 1332 something of a Catch-22 for states aiming to maintain a private market in which plans offer comprehensive coverage without regard to applicants' medical history.

But Trump's October 2017 cutoff of direct  reimbursement of insurers for Cost Sharing Reduction (CSR) and the resulting "silver loading" -- pricing CSR into silver plans only, since CSR is available only with silver plans -- has created a windfall that might be leveraged if CMS plays by Section 1332 rules. (See note below if you're unfamiliar with how silver loading works.)

Silver loading has inflated subsidies as well as premiums and created discounts in bronze and gold plans benefiting some two million enrollees, boosting enrollment and partially offsetting other forms of Trump administration sabotage. CBO estimates the cost to the Treasury at $194 billion over ten years. But the benefits have been haphazard and inefficiently distributed. Via waiver, perhaps a state could make better use of its share.

Is New Jersey's unsubsidized marketplace enrollment migrating off-exchange?

Politico and Modern Healthcare have both wondered why ACA marketplace enrollment for 2019 is down in New Jersey, as in most of the 39 states using HealthCare.gov. As MH's Shelby Livingston put it, New Jersey "did everything right" to boost enrollment -- swiftly passed a state individual mandate and implemented a reinsurance program that brought 2019 premiums an average of 9% below 2018 levels. Yet 5 weeks into Open Enrollment for 2019, enrollment is down 13% compared to 2018. In all 39 HealthCare.gov states, it's down 10.4%.

One hope is that lower premiums are driving more enrollees whose incomes are too high to qualify for subsidies into the off-exchange ACA compliant market. There should be at least some movement in that direction, for two reasons.

First, benchmark silver plan premiums dropped a long way -- 15% in much of the state. Consequently, fewer people with incomes at the upper end of subsidy eligibility qualified for subsidies this year, because the unsubsidized premium cost less than the "affordability" threshold, 9.86% of income at 300-400% FPL. In 2018 in New Jersey, 37,991 enrollees as of the end of Open Enrollment had incomes in the 300-400% FPL range.

Second, this year New Jersey's Department of Banking and Insurance actively encouraged insurers to use on-exchange-only "silver loading" (see note below) to offer cheaper silver plans off-exchange only. AmeriHealth and Oscar responded. AmeriHealth's three cheapest silver are offered off-exchange only; the cheapest one, a narrow network HSA, costs a 46 year-old $108 per month less than the cheapest on-exchange silver.  Oscar's cheapest off-exchange silver plan is $39 per month less for a 46-year old than its cheapest on-exchange silver.  And in Jersey, an unusually high percentage of off-exchange enrollees do choose silver -- 65% in 2018.

As of the end of Open Enrollment for 2018, New Jersey had 62,360 unsubsidized enrollees on-exchange. Some of those enrollees, plus a significant share of the 38k in the 300-400% FPL band, may have some incentive to move off-exchange.

A closer look at off-exchange enrollment in New Jersey, however, suggests that such movement may be limited.

Silver loading improves with age

I have noted before that ACA marketplace enrollment in 2018 was stronger at the upper income levels among those eligible for subsidies.

That's because the discounts in bronze and gold plans created by silver loading (see note below) are chiefly relevant to those who earn too much to qualify for strong Cost Sharing Reduction (CSR), which is available only with silver plans. Also, the huge spike in benchmark silver premiums in 2018 rendered more people in the 300-400% FPL subsidy-eligible than in previous years.

I just stumbled on a breakout I never used that shows a corollary: enrollment as a percentage of 2017 enrollment also rose with age This is in HealthCare.gov states:

Waiving the ACA marketplace away (at healthinsurance.org)

It doesn't matter to anyone but me, but work on articles for outside publications sometimes leads to long blogging pauses. Pending [update: now posted] : a look at likely fallout from CMS's "waiver concepts" released Nov. 29 -- very broad templates for how the Trump administration would like to see states implement ACA innovation waivers under the guidance CMS issued on Oct. 24.

That guidance took an ax to the so-called "guardrails" previously constraining the ACA Section 1332 innovation waivers: statutory requirements that alternative schemes developed in waiver proposals provide coverage as comprehensive and affordable to as many people as would the default ACA design, without increasing the deficit. Trump's CMS decoupled the requirement to provide coverage as comprehensive and affordable from the requirement to cover as many people, declaring that equally comprehensive/affordable insurance had to be merely "available," not "provided."

CMS also erased 2015 guidance stipulating that changes could not harm vulnerable groups such as low income, older or high-medical-need enrollees, declaring, in effect, that such populations could be harmed "slightly" in pursuit of covering more people more cheaply.

Three of the four waiver concepts ring variations on one uber-concept: using the federal dollars currently devoted to subsidizing coverage in the ACA marketplace to subsidize ACA-noncompliant products such as short-term plans.

In a pending post [now up] at healthinsurance.org, I try to flesh out how programs conforming to the "concepts" might affect the vulnerable populations named in the superceded 2015 guidance: low income, older, and sicker enrollees. In an outtake, I question whether proposals conforming to the concepts can adhere even to the weakened guidance CMS issued in October, let alone violating the ACA statute:

Further, it's questionable whether the waiver concepts could pass muster even under the new CMS guidance. The new guidance affirms that coverage as comprehensive and affordable as that on offer without the proposed scheme has to be available, even if fewer people choose it.. While the CMS fact sheet laying out the waiver concepts states that proposals must meet the statutory guardrails, it offers no hints as to how the developed concepts might satisfy even the "availability" standard.

That argument is developed in more detail (with respect to subsidizing noncompliant plans, prior to release of the waiver concepts) here. 

UPDATE, 12/8: Joel McElvain, a former Dept. of Justice lawyer who participated in the defense of the ACA in the two cases that reached the Supreme Court, has since argued in detail that the October guidance violates the ACA's statutory language. He also, by the way, affirms the thesis advanced in this post:

There is, at a minimum, substantial doubt as to whether some of the concepts that HHS described in the discussion paper could meet even the agencies’ newly-announced interpretation of Section 1332; it is questionable whether state residents would continue to have access to the same level of comprehensive, affordable health coverage under a waiver that siphons off healthy people from the risk pool, as HHS appears to be contemplating.     

We're in Cassidy-Collinsville, Chapter 2

Last March, Peter Suderman wrote a clever column claiming that Republican changes to the ACA -- repeal of the individual mandate, creation of a parallel ACA-noncompliant individual market -- were achieving the goals of the ACA repeal/replace bills. I noted the missing piece there: defunding the ACA, in particular the Medicaid expansion and Medicaid more generally. That was the "hot-beating heart" of the failed Republican repeal bills. Instead of AHCA- light, I suggested:

In the aftermath of the 2017 assault, the ACA resembles not so much a system established by a mainstream Republican repeal-and-replace bill as it does the kind of compromise that might have emerged from a negotiation over the Cassidy-Collins bill introduced in January 2017, if negotiation over such a plan had been possible (it wasn't, because no more than a handful of Republicans were interested in the bill).