The Case for Replacing “Silver Loading”

These changes have not affected all enrollees equally. Enrollees with incomes in the 100-200% of FPL range are mostly unaffected because they are eligible for generous CSRs and so overwhelmingly enroll in silver plans even under silver loading (see Figure 2, Panel A). By contrast, subsidized enrollees with incomes above 200% of the FPL have benefited substantially. These enrollees are eligible for small or no CSRs, so they selected non-silver plans in large numbers even before silver loading, and many more have switched into non-silver plans to take advantage of the decline in net premiums for those plans (see Figure 2, Panel B). Unsubsidized enrollees who previously selected silver plans are potentially worse off, but in most states insurers have limited silver loading to on-Marketplace silver plans, so enrollees can avoid the silver load by switching to off-Marketplace plans if they wish. Regardless, there will be fewer unsubsidized enrollees going forward if the extension of the premium tax credit to people with incomes above 400% of the FPL in the American Rescue Plan Act is made permanent in the coming months.

Evidence suggests that silver loading has increased Marketplace enrollment. People with incomes in the 200-400% of FPL range, the main group that benefited from silver loading, saw higher Marketplace enrollment growth following the transition to silver loading relative to enrollees at lower income levels. More compelling, states that saw larger increases in silver premiums following implementation of silver loading saw faster enrollment growth—and that additional enrollment growth was concentrated in the 200-400% of FPL group that benefited most from silver loading. These findings are consistent with projections from microsimulation models that predicted silver loading would increase Marketplace enrollment (and empirical work in other settings showing that reducing premiums increases enrollment).

By increasing the value of the premium tax credit and increasing Marketplace enrollment, silver loading has also increased federal costs—even after netting out the federal savings from no longer making CSR payments. In brief, the increase in premium tax credits paid out on behalf of silver plan enrollees have approximately offset the savings from the lack of CSR payments,[4] while the increased premium tax credits paid on behalf of non-silver enrollees and the tax credits provided to new enrollees have represented new federal costs. Consistent with this, CBO previously projected that resuming direct CSR payments would reduce federal costs by around $10 billion per year over the 2019-2021 period.[5]

Absent policy changes, it appears likely that silver loading will continue indefinitely. While insurers sued the federal government to recover the missing CSR payments, court rulings issued in August 2020 held that insurers are entitled to compensation only to the extent that they have failed to mitigate their losses by raising silver premiums and receiving higher premium tax credits; this implies that most insurers will recover little or nothing for years after 2017. Moreover, the logic of these court decisions suggests that insurers that have the opportunity to silver load but decline to do so might be entitled to nothing. Thus, it appears that insurers will struggle to recover CSR payments through litigation, implying that their best response to the continued lack of CSR payments will be to continue silver loading in the years to come.

The transition to silver loading has, in effect, brought about an expansion of the ACA’s Marketplace subsidies, with the main benefits (lower premiums for enrollees and broader insurance coverage) and costs (higher federal costs) that such expansions typically entail. Correspondingly, policymakers who generally oppose Marketplace subsidy expansions have good reason to support resuming CSR payments and thereby returning to the pre-silver-loading status quo. By contrast, policymakers who generally support Marketplace subsidy expansions are likely to look favorably on silver loading. Nevertheless, for several reasons, policymakers in the latter group may wish to consider replacing silver loading by resuming CSR payments and using the savings to finance a direct expansion of Marketplace subsidies.

One important rationale for replacing silver loading is that a future administration could try to end silver loading administratively. In particular, a future administration could seek to force insurers to spread the cost of the missing CSR payments across premiums in all metal tiers instead of just the silver tier, an approach sometimes called “broad loading.” By reducing silver premiums, this approach would reduce federal premium tax credit costs, but it would increase premiums for people buying non-silver plans since these enrollees would now pay higher gross premiums and receive smaller tax credits. Thus, this approach would unwind the main benefits that silver loading provides to Marketplace enrollees.

Broad loading would also create some pernicious incentives. Under broad loading, insurers would incur large losses on their silver plans since premiums for those policies would no longer reflect the full cost of providing CSRs. On the other hand, they would now make large profits on their non-silver plans since premiums for those plans would be set to defray CSR costs that those plans did not incur. In this environment, insurers would have strong incentives to make their silver plan offerings unappealing to consumers in order to push enrollees out of silver plans and into non-silver plans.[6] This behavior would particularly harm people who are eligible for CSRs since CSRs are only available in silver plans.

It is not hard to imagine a future administration that was hostile to the ACA (or that simply viewed the reduction in federal spending from ending silver loading as more important than the corresponding reduction in insurance coverage and increase in enrollees’ premiums) might go down this path. Indeed, the Trump administration appears to have seriously considered taking action to end silver loading. Congress took this threat seriously enough that it barred the administration from taking any action to end silver loading for the 2021 plan year in appropriations legislation enacted in December 2019.

Now, a future administration might not be successful in moving to broad loading even if it tried. One issue is that enforcing broad loading would be challenging in practice. Under-pricing silver plans and over-pricing non-silver plans is not in any insurer’s interest, so insurers would likely try to “cheat” away from broad loading and toward silver loading to the extent they could get away with it. In practice, they would likely be at least somewhat successful since regulators likely could not perfectly distinguish surreptitious silver loading from permissible premium differences between silver and non-silver plans.

Mandating broad loading might also be illegal. While the ACA regulates how insurers can vary the premiums they charge different enrollees for the same plan, it generally does not regulate how the premiums of different plans must relate to each other, beyond a general requirement that insurers cannot set the premiums of different plans to reflect differences in risk mix.[7] Insurers might therefore reasonably argue that mandating broad loading would go beyond the federal government’s authority.

Even if silver loading were guaranteed to continue, there would still be other rationales for replacing it. Most importantly, for enrollees who are not eligible for generous CSRs, silver loading causes silver plans to be “overpriced” relative to other plans. This has caused a large shift out of silver plans among people with incomes above 200% of the FPL, most of it into bronze plans (see Figure 2, Panels B and C).

Because of these shifts, many subsidized enrollees with incomes above 200% of the FPL are likely now in the “wrong” level of coverage. To be precise, many enrollees who have shifted into bronze plans would likely prefer a silver plan if the silver plan’s premium reflected the actual cost-sharing they would pay, rather than primarily reflecting the cost-sharing faced by enrollees who are eligible for generous CSRs. Thus, unless these enrollees were mistakenly opting for plans with too little cost sharing prior to the implementation of silver loading, something there is not much evidence for, these enrollees are likely now in plans that include too much cost sharing.[8] (Similarly, many of the much smaller number of enrollees who have shifted into gold plans may now be in plans with too little cost-sharing.)

To be clear, this does not mean that silver loading has made subsidized enrollees in this income range worse off. Enrollees who previously bought silver plans had the option to stay in silver plans without paying a higher premium (since the premium tax credit offset the increase in silver premiums). The fact that they nevertheless shifted into non-silver plans indicates that they viewed these plans as better options than they had before. However, these enrollees would have been even better off if they could have benefited from the increase in the value of the premium tax credit caused by silver loading while staying in silver plans, which a replacement for silver loading could make possible.

Shifts out of the silver tier are less consequential for unsubsidized enrollees. As noted above, unsubsidized enrollees generally have the option of switching to off-Marketplace silver plans that do not incorporate a CSR load. To the extent this has occurred in practice, the cost that these enrollees incur from leaving Marketplace silver plans are mostly limited to hassle costs. Additionally, unsubsidized enrollees will become rarer if the subsidy expansions in the American Rescue Plan Act are made permanent.

There has also been a modest shift out of silver plans among enrollees with incomes below 200% of the FPL (see Figure 2, Panel A). Many of these enrollees are likely making mistakes in light of the very large differences in cost-sharing between bronze and silver plans for these enrollees (although, by making the value of CSRs partially “portable” across metal tiers for people in this income group, silver loading may also have allowed some enrollees to opt for plans with a mix of premiums and cost-sharing that better meets their needs). Regardless, if the subsidy expansions in the American Rescue Plan Act become permanent, most of these enrollees are likely to shift back into silver plans whether or not silver loading continues since these expansions guarantee a zero premium silver plan to everyone with an income below 150% of the FPL and a very low or zero premium silver option to everyone below 200% of the FPL.

A final rationale for replacing silver loading is that doing so could eliminate a modest disincentive for states to expand Medicaid. When a state expands, people with incomes between 100% and 138% of the FPL lose eligibility for Marketplace coverage and instead become eligible for Medicaid, thereby removing most people with 94% actuarial value CSRs from the Marketplace. This in turn reduces insurers’ need to silver load; the average actuarial value of Marketplace silver plans (inclusive of the value of CSRs) is 90% in non-expansion states versus 84% in expansion states.[9] The resulting reduction in silver premiums diminishes the benefits of silver loading for the higher-income enrollees who remain in the Marketplace, so Medicaid expansion has the unintended effect of making these enrollees somewhat worse off.

Even taking account of this interaction with silver loading, expanding Medicaid is an overwhelmingly attractive proposition for states in light of the benefits of Medicaid expansion for access to care, health, and financial security, as well as the fact that the federal government picks up 90% of the long-run cost. Nevertheless, eliminating this modest disincentive for expansion would likely be worthwhile.

Silver loading creates a much larger disincentive for states to adopt a Basic Health Program (BHP). As background, under the ACA, states have the option to create a BHP as a substitute for Marketplace coverage for people with incomes under 200% of the FPL. States that opt for a BHP receive per enrollee payments from the federal government equal to 95% of what the federal government would have paid on behalf of those enrollees in the Marketplace. New York and Minnesota currently operate BHPs, and their programs offer lower premiums and cost-sharing than Marketplace coverage. (This is likely because the existing BHPs offer Medicaid-like plans that pay providers less than Marketplace plans, and the states are then using the savings to offer more generous benefits to BHP enrollees.)

Creating a BHP shifts all enrollees who are eligible for generous CSRs out of the Marketplace and into BHP. This all but eliminates the need for insurers to silver load, which in turn essentially eliminates the benefits of silver loading for the higher-income enrollees who remain in the Marketplace.[10] In light of this fact, it is doubtful that it currently makes sense for states that do not already have a BHP to adopt one. To the extent that broader adoption of BHPs is desirable, this is a downside of silver loading.

If policymakers do want to resume direct CSR payments and reinvest the savings by directly expanding the ACA’s Marketplace subsidies, they have two main decisions to make about how to do so.

A first decision is whether to reinvest funds by expanding the premium tax credit or by expanding CSRs. Expanding the premium tax credit would lower premiums in all metal tiers and then allow consumers to select whether they prefer plans that offer higher premiums and less cost-sharing or prefer plans that offer lower premiums and more cost-sharing. This approach is more attractive to the extent that consumers vary in what mix of premiums and cost-sharing best suit their needs and do a good job of choosing plans in accordance with those varying needs. Additionally, this approach would also tend to increase the number of consumers who have a zero-premium plan option available to them, which may have a particularly large effect on enrollment outcomes, plausibly by eliminating the hassle costs associated with remitting a monthly premium. Broader availability of zero-premium plan options could also facilitate various types of automatic enrollment efforts.

By contrast, CSRs are only available to people enrolled in silver plans, so expanding CSRs would make silver plans more attractive without changing the attractiveness of other plans.[11] This may be desirable if policymakers are concerned that consumers do a poor job weighing tradeoffs between lower premiums and higher cost-sharing (which there is evidence for in at least some contexts) and thus wish to steer consumers toward plans with a particular mix of premiums and cost-sharing. A risk of this approach is that some enrollees might not respond to the incentives policymakers were aiming to create and, thus, would end up enrolling in non-silver plans that do not provide the additional assistance; this is the case today for a small, but not trivial, minority of people with incomes below 200% of the FPL (see Figure 2, Panel A).

Policymakers would also need to decide what types of enrollees to target with the reinvested funds. From a substantive perspective, that choice should likely be shaped by two factors. The first is where additional assistance would do the most to increase insurance coverage. Those effects could vary by income (although it is not immediately clear how since take-up of ACA-compliant coverage was incomplete at all income levels as of 2019) or by other characteristics like age; they would also likely depend on whether the broad-based expansion of the premium tax credit enacted in the American Rescue Plan Act (ARP) had been made permanent. The second is where additional assistance would do the most to relieve financial hardship; in general, this consideration argues for targeting additional assistance at people with lower incomes, where household budgets are likely to be tightest.

In practice, policymakers might see political advantages to reinvesting funds in a way that mirrored the benefits of silver loading in order to minimize the number of consumers made worse off by replacing silver loading. These political advantages might be smaller if legislation that replaced silver loading also made the ARP subsidy expansions permanent since very few consumers are likely to be made worse off by the combined package. However, that might only be true if consumers first encountered the end of silver loading at the same time they first encountered the ARP subsidy expansions. That likely will not be the case in practice. Premium setting is already underway for the 2022 plan year, so an end to silver loading will likely need to wait until the 2023 plan year, whereas some Marketplace consumers are already receiving the enhanced ARP subsidies and the rest will during open enrollment for 2022.

If policymakers decide that their priority in reinvesting the savings from ending silver loading is to replicate the consumer benefits created by silver loading as closely as possible, then there is a straightforward way to do so. Specifically, policymakers could change the benchmark premium that is used to calculate the premium tax credit to be a multiple of the second-lowest-cost silver premium (rather than the second-lowest-cost silver premium itself, as is the case today) for enrollees not enrolled in generous CSRs.

If this multiple was selected so that the benchmark premium under the policy (for the enrollees eligible for the new higher benchmark) equaled the benchmark premium that would have prevailed if silver loading had continued, then this change to the premium tax credit would almost exactly reproduce the pattern of consumer benefits created by silver loading. The exception is that this approach would reduce the net premiums of plans in all metal tiers, not just the premiums of non-silver plans.

As described in the appendix, I estimate that, on average nationwide, the second-lowest-cost silver premium is currently around 28% higher than it would have been if direct CSR payments had continued. Thus, setting the benchmark premium at 128% of the second-lowest cost silver premium would approximately replicate the benefits of silver loading. It is important to note that while this would be true on average, it would be not be true in every geographic area since the impact of silver loading varies across areas due to differences in the income mix of CSR enrollees and other market-specific factors. The percentage increase in the benchmark required to replicate the effects of silver loading could also change over time if the impact of silver loading changed over time; this could occur if people ineligible for generous CSRs continue to migrate out of silver plans, as suggested by the trends shown in Figure 2.

The cost of enhancing the premium tax credit in this way would likely modestly exceed the savings from ending silver loading because this type of subsidy expansion would benefit silver enrollees, whereas silver loading does not. In the appendix, I estimate an incremental cost of around $2.3 billion per year over the 2023-2030 period relative to current law (or $3.0 billion per year over that period if the ARP subsidy expansions were made permanent). As discussed in the appendix, this estimate is relatively crude and accounting for factors omitted from this simple estimate could reduce the estimated cost modestly.

A related—and more elegant—approach to reinvesting the savings from ending silver loading would be to change the benchmark plan to be the second-lowest-cost gold plan rather than the second-lowest cost silver plan (and, correspondingly, to modify CSRs so that they are available on gold plans rather than silver plans). This type of policy has been previously proposed by researchers at the Urban Institute.

At first blush, this “re-benchmark to gold” policy appears to be much more expensive than the policy outlined above since, on average nationwide, the second-lowest cost gold premium is about 12% higher than the second-lowest cost silver premium.[12] But for two reasons, that may not be the case.[13]

First, re-benchmarking to gold would likely reduce the average claims risk of gold enrollees. At present, it appears that gold plans are attracting much sicker enrollees than silver plans (in ways that risk adjustment is not offsetting); indeed, without a substantial difference in risk mix, it is hard to explain why gold premiums are higher than silver premiums even as the average actuarial value of a Marketplace silver plan (87% on average nationwide, inclusive of CSRs) exceeds the actuarial value of a gold plan (80%). However, if CSRs shifted from the silver tier to the gold tier, most people with generous CSRs would shift into gold plans. About 5.1 million people selected silver plans with generous CSRs during 2021 open enrollment (accounting for 77% of all silver plan selections), whereas only 1.0 million selected gold plans. Thus, shifting this group into gold plans would likely cause the risk mix in the gold tier to converge toward the current risk mix in silver, likely substantially reducing gold premiums. Second, the presence of more low-income consumers in the gold tier could intensify price competition, further reducing gold premiums.

Quantifying these effects would require further analysis, but it appears very likely that re-benchmarking to gold would be much less expensive that it looks—conceivably actually cheaper than continuing silver loading.[14] Reducing the premiums of gold plans could also directly benefit consumers who value plans with lower cost-sharing.[15] Thus, if policymakers do want to reinvest the benefits from ending silver loading in a way that would mirror the benefits of silver loading, this approach merits consideration.

I note that a disadvantage of either of these two approaches is that they (like silver loading) deliver roughly the same benefit to subsidized enrollees at all income levels, whereas targeting more of the assistance at lower income levels would likely do more to relieve financial hardship.  If policymakers wanted to target more assistance to people with lower incomes, they could consider phasing out the more generous benchmark premium for consumers with incomes above a specified level.

If policymakers concluded that they did want to provide an appropriation for direct CSR payments and then reinvest the savings that generated, they may face a scoring obstacle. Specifically, CBO has previously said that it would score a provision that appropriated CSR payments as having no effect on federal deficits, even though CBO agrees the provision would, in fact, reduce deficits. Thus, legislation that resumed direct CSR payments and reinvested the savings would be scored as increasing deficits.

Why has CBO adopted this odd scoring treatment? Under the law that governs how CBO constructs its baseline, CBO must assume that funding will be “adequate to make all payments required” under an entitlement program (that is, a program that, like the CSR program, gives some entity a legal right to a payment from the federal government).[16] Relative to that baseline, a provision that appropriates funds to make payments under an entitlement program typically has no budgetary effect, so CBO conventionally scores such a provision as causing no change in deficits. CBO has said in the past that it intends to apply that convention in scoring a provision that would appropriate CSR payments.

This scoring treatment would pose a problem if policymakers wanted to resume direct CSR payments and reinvest the savings. Legislation along these lines would likely need to be considered through reconciliation to avoid a filibuster in the Senate. But the Byrd rule, which governs Senate consideration of reconciliation bills, bars legislation that increases deficits outside the budget window.

However, this problem has a straightforward solution. In particular, CBO has said that its current scoring approach was arrived at in consultation with the House and Senate Budget Committees. Thus, the Committees could likely request that CBO adopt a scoring approach that accurately captures the real-world consequences of providing an appropriation to make direct CSR payments.

In addition to being more realistic, this alternative scoring approach would be simpler and more internally consistent. Indeed, CBO has sensibly decided that its baseline projections should reflect what is happening in reality: silver loading, not direct CSR payments.[17] CBO scores virtually all legislation against that realistic baseline. The sole exception is legislation that would resume direct CSR payments, which is, in effect, scored against a different baseline in which direct CSR payments are already occurring.

The main text reports an estimate of the percentage increase in the average benchmark premium in 2021 due to the lack of CSR payments. This portion of the appendix describes how that estimate was derived.

In a simple competitive model, the change in the benchmark premium caused by the lack of CSR payments will equal the induced change in insurers’ average costs.[18] The lack of CSR payments may change insurers’ costs in at least three ways. First, and most importantly, insurers now bear the claims spending previously covered by CSR payments. Second, silver loading may cause people ineligible for generous CSRs to leave silver plans, as suggested by the trends in Figure 2, decreasing the amount of cost-sharing faced by the typical silver enrollee and thereby increasing utilization. Third, those same enrollment shifts may change the risk mix of silver enrollees, which could also change utilization in silver plans.

To arrive at my main estimate of how much silver loading has increased the benchmark premium, I perform a calculation that accounts for the first two channels. To that end, I use the 2021 open enrollment public use file published by the Centers for Medicare and Medicaid Services (CMS) to estimate the average actuarial value of a Marketplace silver plan (inclusive of the value provided by CSRs). For states that use HealthCare.gov, I calculate the average actuarial value directly using data on how many people selected each tier of CSR. The public use file lacks the needed information for states that do not use HealthCare.gov, so I assume that the average actuarial value in other states is the same as the average for HealthCare.gov states with the same Medicaid expansion status.[19] Weighting each state by its total number of Marketplace plan selections, I obtain a national enrollment-weighted actuarial value of 86.7%. Using a similar method, I estimate that this actuarial value was 84.4% as of 2017.[20]

Under the assumption that 10% of insurers’ expenses do not scale with claims spending, the increase in the benchmark premium required to offset the direct loss of CSR payments is 21.4% (=0.9*[0.867/0.7] + 0.1). To estimate the contribution from higher utilization, I assume that the change in the average actuarial value of a silver plan since 2017 was entirely attributable to the transition to silver loading, and I treat a silver plan as if it had a constant coinsurance rate equal to one minus its actuarial value (an admittedly imperfect approximation). Applying a conventional elasticity of utilization with respect to price of -0.2 yields a utilization increase of 3.1% (=Exp[-0.2*ln[0.133/0.155]]). Combining these two effects yields an increase in the benchmark premium of 25.0% (=0.9*[0.867/0.7]*1.031 + 0.1).

All of the calculations presented above assume that the actuarial value of a silver plan without CSRs exactly equals 70% and that the actuarial value of a silver plan with CSRs exactly equals the nominal actuarial value of the relevant CSR tier (that is, 73%, 87%, or 94%). In practice, however, silver plans are permitted to have a base actuarial value as low as 66% and silver CSR variants are permitted to have an actuarial value up to 1 percentage point below the nominal actuarial value associated with that CSR tier.[21]

To account for this, I use CMS’ plan attributes public use file to obtain the actual actuarial value of the benchmark plan in each county that used HealthCare.gov in 2021. Using these data, I estimate that the base actuarial value of the benchmark plan averaged 68.3% in 2021, while the actuarial values of the corresponding 73%, 87%, and 94% actuarial value CSR variants averaged 72.9%, 87.3%, and 93.8%, respectively, where averages are weighted by county-level Marketplace plan selections. If these actuarial values are substituted for the nominal actuarial values in the calculations above, then my estimate of the average actuarial value of a Marketplace silver plan in 2021 becomes 86.4% and the corresponding estimate for 2017 is 84.1%. The estimated impacts on the benchmark premium also end up being slightly larger, primarily because of the lower base actuarial value of the benchmark plan; the estimated effect is 23.9% when ignoring utilization effects and 27.6% when incorporating those effects.

As a check on this model-based calculation, I also estimate how silver loading affected benchmark premiums by comparing the trend in benchmark premiums to the trends in lowest-cost gold and bronze premiums. The virtue of this approach is that it does not depend on a specific model of insurers’ pricing decisions. However, this approach does require that the trends in bronze and gold premiums be a reliable guide to the path that silver premiums would have taken without silver loading.

This assumption will fail if silver loading directly affects bronze and gold premiums, such as by changing the risk mix of people who select those plans. This assumption will also fail if other factors cause bronze and gold premiums to trend differently from silver premiums. Figure A1 shows that benchmark premiums were declining relative to both lowest-cost gold and bronze premiums prior to the start of silver loading (perhaps reflecting ongoing changes in risk mix across metal tiers); this suggests that this second issue is indeed a concern in this context. To cope with this problem, I assume that benchmark premiums would have continued their pre-silver-loading decline relative to lowest-cost bronze and gold premiums in the absence of the transition to silver loading; the implied paths are depicted by the dashed lines in Figure A1. While this assumption has intuitive appeal, its validity is subject to uncertainty, particularly absent a full understanding of why benchmark premiums were declining in relative terms prior to silver loading.