Function 550 - Health
Establish Caps on Federal Spending for Medicaid
CBO periodically issues a compendium of policy options (called Options for Reducing the Deficit) covering a broad range of issues, as well as separate reports that include options for changing federal tax and spending policies in particular areas. This option appears in one of those publications. The options are derived from many sources and reflect a range of possibilities. For each option, CBO presents an estimate of its effects on the budget but makes no recommendations. Inclusion or exclusion of any particular option does not imply an endorsement or rejection by CBO.
|Billions of Dollars||2023||2024||2025||2026||2027||2028||2029||2030||2031||2032||2023–
|Caps on Overall Spendinga|
|Apply Caps to All Eligibility Categories, With Growth of Caps Based on the CPI-U|
|Change in Outlays||0||-2||-48||-61||-81||-96||-111||-129||-151||-173||-193||-853|
|Change in Revenuesb||0||*||-1||-1||-2||-2||-2||-3||-3||-3||-3||-17|
|Decrease (-) in the Deficit||0||-2||-47||-60||-79||-94||-109||-126||-148||-170||-190||-836|
|Apply Caps to All Eligibility Categories, With Growth of Caps Based on the CPI-U Plus 1 Percentage Point|
|Change in Outlays||0||-2||-25||-33||-48||-57||-67||-79||-95||-110||-108||-516|
|Change in Revenuesb||0||*||-1||-1||-1||-2||-2||-2||-3||-3||-3||-15|
|Decrease (-) in the Deficit||0||-2||-24||-32||-47||-55||-65||-77||-92||-107||-105||-501|
|Caps on Spending per Enrolleec|
|Apply Caps to All Eligibility Categories, With Growth of Caps Based on the CPI-U|
|Change in Outlays||0||-2||-3||-69||-86||-104||-123||-143||-165||-188||-161||-884|
|Change in Revenuesb||0||*||*||-1||-1||-1||-2||-2||-2||-3||-3||-13|
|Decrease (-) in the Deficit||0||-2||-3||-68||-85||-103||-121||-142||-163||-185||-158||-871|
|Apply Caps to All Eligibility Categories, With Growth of Caps Based on the CPI-U Plus 1 Percentage Point|
|Change in Outlays||0||-2||-3||-38||-50||-62||-76||-90||-106||-123||-94||-550|
|Change in Revenuesb||0||*||*||-1||-1||-1||-1||-2||-2||-2||-2||-11|
|Decrease (-) in the Deficit||0||-2||-3||-37||-49||-61||-75||-88||-104||-121||-92||-539|
Medicaid is a joint federal-state program that covers acute and long-term health care for groups of low-income people, chiefly families with dependent children, elderly people (those 65 or older), nonelderly people with disabilities, and—at the discretion of individual states—other nonelderly adults whose family income is up to 138 percent of the federal poverty guidelines. The federal and state governments share in the financing and administration of Medicaid. The federal government provides the majority of Medicaid's funding; establishes the statutory, regulatory, and administrative structure of the program; and monitors states' compliance with the program's rules. As part of its responsibilities, the federal government determines which groups of people and which items and medical services states must cover if they participate in the program and which can be covered at states' discretion. For their part, the states administer the program's daily operations, determine eligibility, reimburse health care providers and health care plans, and determine which optional eligibility and service categories to adopt. The result is wide variation among states in levels of enrollment, the scope of services covered, payment rates for providers and health care plans, and spending per capita, among other aspects of how the program is implemented.
In 2021, the states received, in aggregate, $521 billion in federal funding for Medicaid and spent $260 billion of their own funds for the program. The Congressional Budget Office expects that, under current law, federal spending for Medicaid will increase faster than general inflation, in part because of continued growth in health care costs and in part because more states are expected to expand Medicaid coverage under the Affordable Care Act (ACA). (To date, 38 states and the District of Columbia have done so.)
Under current law, almost all federal funding is open-ended: If a state spends more because enrollment increases or costs per enrollee rise, larger federal payments are generated automatically. The federal government's share varies by state, by the type of cost (that is, costs for medical or administrative services), and by eligibility category. On average, the federal government pays about 65 percent of the program's costs, with a range among the states of 54 percent to the current high of 79 percent. That range mainly reflects the variation in states' per capita income and in the share of enrollees (if any) in each state that became eligible for Medicaid as a result of the optional expansion of eligibility under the ACA. The federal government currently pays 90 percent of the costs of medical services provided to those enrollees.
Under this option, the amount that states receive from the federal government to operate the program would be capped. If the combined federal and state costs of Medicaid exceeded the upper limit established by the federal government, then federal spending would not increase above the cap and states would be responsible for providing additional funds.
Key Design Choices That Would Affect Savings
There are a variety of ways to design caps that would yield significant federal Medicaid savings. If lawmakers decided to consider such caps, a number of major policy choices, with important implications, would have to be made. Those key design choices include the following:
- Whether to set overall or per-enrollee caps;
- What categories of Medicaid spending and what eligibility categories to include in the spending limits;
- Which year's spending to select as a base year so that the specified growth factor, or percentage rate of increase, could be applied to calculations of total spending;
- What growth factor to use to limit the increase in Medicaid spending over time; and
- Whether spending for the optional expansion of coverage under the ACA would also be subject to the caps (thus creating additional complexities for states that have not yet expanded coverage but might do so in the future).
Overall or per-Enrollee Spending Caps. The first key design choice would be whether to establish a cap on federal Medicaid spending across the board or to cap the amount available per enrollee.
Overall Caps. In general, overall caps would consist of a maximum amount of funding that the federal government would provide a state to operate Medicaid. Once established, and depending on the way they were scheduled to increase, the federal caps generally would not fluctuate in response to rising or falling enrollment or as a result of changes in the cost of providing services. However, the rate of growth in the caps could be set so that they accounted for population growth or allowed for automatic increases during economic downturns.
Per-Enrollee Caps. For caps on per-enrollee spending, the federal government would set an upper limit on federal payments for each Medicaid enrollee, on average. Under such a plan, the federal government would provide funds based on each person enrolled in the program, but only up to a specified amount per enrollee. As a result, each state's total federal funding would be limited to the product of the number of enrollees and the capped per-enrollee spending amount. Individual enrollees whose care proved to be more expensive than the average could still generate additional federal payments, as long as the total per capita average did not exceed the cap. Unlike an overall spending cap, such an approach would allow for additional funding if enrollment rose (for example, when a state chose to expand eligibility under the ACA or as a result of an increase in enrollment during an economic downturn). Funding would decline if Medicaid enrollment fell (for example, when a state chose to restrict enrollment or when enrollment fell as a result of an improving economy).
Spending and Eligibility Categories. A second key design choice would relate to spending and eligibility categories. Options to cap federal Medicaid spending could target all Medicaid spending and eligibility categories or limited categories of services and enrollees. For example, the caps could cover all acute care services but allow long-term services and supports to remain uncapped. The caps could also target spending for nondisabled adults and children but leave spending for the disabled and elderly uncapped. Although many possible combinations of services and enrollees could be subject to caps, in general, the more categories that were capped, the greater the potential for federal budgetary savings.
Base Year. A third key design choice for establishing caps on federal spending for Medicaid would entail selecting a particular year of Medicaid outlays as a base year, calculating that year's total spending for the service categories and eligibility groups that were included, and then increasing those amounts by the selected growth factor. The base year would not usually be the first year in which the caps took effect; rather, it would be the year from which the future spending growth would be measured and limited (as described in the next section). Thus, for overall and per-enrollee spending caps alike, the selection of the base year would be important: A higher base-year amount would lead to higher caps (and smaller federal savings) than a lower base-year amount would.
An important consideration in selecting a base year is whether to use a past or future year. Choosing a past year in which actual Medicaid expenditures were known would prevent states from increasing spending in the base year to boost their future spending limits. States could increase spending in a future base year by taking the following steps: raising payment rates for providers and health care plans; making additional onetime supplemental payments; or moving payments for claims from different periods into the base year. Those responses would increase Medicaid spending and lower federal savings.
Another consideration about the base year is whether any unique policies or economic conditions were in effect that influenced Medicaid spending and enrollment in that year. The years 2020 and 2021 provide examples: Specifically, the coronavirus pandemic and the federal policy response to that public health emergency dramatically affected Medicaid spending and enrollment. The federal government increased its own share of total Medicaid expenditures by 6.2 percentage points in states that agreedto allow all eligible people to remain enrolled in the program, regardless of changes to their economic status or personal circumstances that would otherwise cause them to lose coverage. Choosing one of those years as the base year would essentially lock in the greater spending that arose from the economic disruption and the policies implemented in response, which would not persist once the public health emergency ended. To avoid the distortionary effects of the public health emergency when selecting a past year for the base year, it would be necessary to select 2019 or an earlier year; or spending in the base-year calculation would need to be adjusted to remove the effects of the public health emergency on Medicaid.
Growth Factors. A fourth key design choice would be determining the annual growth rate of the limit on Medicaid spending. The growth factor would be one of the most important drivers of savings from Medicaid caps because the caps would essentially be limits on the degree to which the federal government allowed its payments to grow over time. The growth factor could be set to meet specific savings targets or to achieve other specific policy purposes. For example, if a growth factor was set higher than the rate of increase projected for Medicaid spending under current law, little or no budgetary savings might be anticipated, but some other policy objective could be met, such as protecting the federal government from unanticipated cost increases in the future. Alternatively, the growth factor could be set to make the increase in federal Medicaid spending—overall or per enrollee—match changing prices in the economy as measured, for example, by the consumer price index for all urban consumers (CPI-U). Or the growth factor could be set to reflect the growth in health care costs per person, perhaps as measured by the per capita increase in national health care expenditures, or at a rate that was consistent with economic growth as measured by the increase in per capita gross domestic product.
Growth factors that were tied to price indexes or to overall economic growth, however, would not generally account for increases in the average quantity or intensity of medical services of the sort that have occurred in the past. Moreover, the growth factors would not account for advances in medical technology that affect health care costs and could lead to a disconnect between the cost of care and the limit on federal payments.
In general, the smaller the growth factor relative to CBO's projected growth rate for federal Medicaid spending under current law, the greater the projected federal budgetary savings would be. Smaller growth factors would increase the possibility that federal funding would not keep pace with increases in states' costs per Medicaid enrollee or, in the case of overall caps, with increases in Medicaid enrollment. If so, the likelihood that states would not be able to maintain current services or coverage would increase.
The Optional Expansion of Medicaid. A fifth key design choice would pertain to the optional expansion of Medicaid. Since January 2014, states have been permitted to extend eligibility for Medicaid to most people whose income is below 138 percent of the federal poverty guidelines. Under the terms of the ACA, the federal government covers 90 percent of the costs for this eligibility category. Designing the federal spending caps to include the expansion of Medicaid would add complexity, particularly for states that chose to adopt the expansion after the base year.
For states that have not yet expanded coverage under the ACA, data from an earlier base year would not reflect spending for this category. Should any of those states subsequently adopt the expansion, the annual limits established by an overall spending cap would fail to account for the spending of that group of enrollees. For per-enrollee caps, the additional enrollment from the coverage expansion would generate additional federal spending, but average per capita spending for adults in the base year would not account for the higher federal payment for newly eligible people under current law. In addition, the average would not reflect any differences between expected costs related to the health status of those new enrollees and costs for people who would have been eligible before the expansion. In designing Medicaid caps, those issues could be addressed in one of several ways or there could be no special adjustments for that group.
CBO analyzed two approaches that would limit federal Medicaid spending: establishing overall spending caps and establishing per-enrollee caps. For both approaches, CBO analyzed limits on spending for all medical services to all eligibility groups. Further, to illustrate a range of savings, CBO used a pair of alternative growth factors for each type of cap: either the annual change in the CPI-U or the change in the CPI-U plus 1 percentage point (referred to here as the CPI-U plus 1). Under each alternative, states would retain their current authority concerning optional benefits, optional enrollees, and payment rates for providers and health care plans.
For all of the alternatives, CBO chose 2019 as the base year to avoid the impact of the public health emergency on Medicaid enrollment and spending in 2020 and 2021. Overall caps would take effect in October 2024; per-enrollee caps would take effect one year later. That additional year would be the minimum necessary to allow the Centers for Medicare & Medicaid Services (CMS) to complete the complex gathering of data needed to arrive at state-specific caps for each group of enrollees. For overall and per-enrollee caps alike, federal matching rates would continue as they are under current law. Medicaid's disproportionate share hospital payments (which are already capped), the Vaccines for Children program, and administrative spending would all be excluded, as would Medicaid assistance with Medicare cost sharing and premiums for those dually eligible for both programs.
For the per-enrollee spending caps, CBO assumed that separate spending limits would be set for five Medicaid eligibility groups in each state: the elderly (people age 65 or older); people with disabilities; children; nondisabled, nonelderly adults whose eligibility category existed before enactment of the ACA; and adults made eligible by the ACA (in states that have expanded coverage). States would be permitted to cross-subsidize groups, meaning that states could spend above the upper limit for an eligibility group as long as they spent less than the limit for another group by an amount sufficient to maintain total spending below the overall program limit. CBO also assumed that CMS would either create a new data source or modify an existing data source to capture the necessary spending and enrollment information for the five groups. CBO anticipates that no additional states would expand coverage under the ACA and therefore the Secretary of Health and Human Services would not need to adjust the caps to reflect estimated additional spending in any state that adopted the expansion after the base year.
Effects on the Budget
The savings to the Medicaid program under each alternative would vary widely. The most important factor affecting the amount of savings would be the rate by which Medicaid spending was permitted to grow under the alternatives.
Caps on Overall Spending. Under the specifications listed here, CBO estimates that the overall caps would reduce the deficit by $836 billion between 2024 and 2032 using the CPI-U growth factor and by $501 billion using the CPI-U plus 1 growth factor. Those net effects on the deficit reflect larger gross savings to Medicaid that are partially offset by increases in other types of mandatory spending and a reduction in revenues. The gross savings to Medicaid would be $921 billion between 2024 and 2032 using the CPI-U growth factor and $576 billion using the CPI-U plus 1 growth factor. Savings in 2032 would amount to about 17 percent of projected federal Medicaid spending using the CPI-U growth factor and 11 percent using the CPI-U plus 1 growth factor.
The gross savings from establishing caps on overall spending would be partially offset because of responses by states and individuals to the caps. With the policy specifications described above, reductions in federal Medicaid spending resulting from the overall caps would represent large reductions in state revenues. Therefore, in CBO's assessment, states would take a variety of actions to reduce the additional costs they would face, including restricting enrollment. Some states would discontinue coverage for enrollees made eligible by the ACA, and all states that would have adopted such coverage in the future would no longer choose to do so. Of those people who lost Medicaid coverage, some would gain access to subsidized health insurance coverage through the marketplaces established by the ACA because they would qualify for subsidies to buy coverage if other eligibility criteria were met. The rest would enroll in other coverage, principally through an employer, or become uninsured. CBO and the staff of the Joint Committee on Taxation (JCT) estimate that roughly 65 percent of people who lost Medicaid coverage would become uninsured under caps that were adjusted for inflation using either the CPI-U or the CPI-U plus 1 growth factor.
For the caps on overall spending, the agencies estimate—using the CPI-U growth factor—that the additional marketplace and employment-based coverage would increase federal outlays by $68 billion and decrease revenues by $17 billion from 2024 through 2032. Using the CPI-U plus 1 growth factor, the agencies estimate that the additional coverage would increase outlays by $59 billion and decrease revenues by $15 billion over the same period.
The net savings from capping overall spending would depend greatly on the growth factor. The lower CPI-U growth factor, when compared with the CPI-U plus 1 growth factor, would increase savings by an additional $334 billion. The net savings could be made larger or smaller by adding or subtracting additional percentage points from the CPI-U, and the change to net savings would be reasonably, though not perfectly, proportional to the change. That is, each additional 1 percentage-point reduction in the growth factor would increase savings to Medicaid by a similar dollar amount, but the offsets attributable to the loss of Medicaid coverage would increase by less. The reason for the difference is that states have limited flexibility to lower enrollment while complying with federal guidelines. Therefore, increasingly smaller growth factors would cause states to reach the limit of that flexibility, and no additional reductions would be anticipated.
Caps on per-Enrollee Spending. Under the policies specified above, CBO estimates that the per-enrollee caps would reduce the deficit by $871 billion between 2024 and 2032 using the CPI-U growth factor and by $539 billion using the CPI-U plus 1 growth factor. Those net effects on the deficit reflect larger gross savings to Medicaid that are partially offset by increases in other types of mandatory spending and a reduction in revenues. CBO estimates that establishing caps on per-enrollee spending would generate gross savings to Medicaid of $934 billion between 2024 and 2032 using the CPI-U growth factor and of $593 billion using the CPI-U plus 1 growth factor. The savings would represent about 20 percent and 13 percent, respectively, of projected federal Medicaid spending in 2032.
As with the caps on overall spending, the gross savings from per-enrollee caps would be partially offset. Although per-enrollee caps would provide additional federal payments for each enrollee, caps below projections of federal per-enrollee spending would create a loss of revenues to states for each enrollee relative to current law. Therefore, CBO anticipates that some states also would take action to restrict enrollment under per-enrollee caps. In addition, CBO and JCT estimate that roughly 63 percent of enrollees who lost Medicaid coverage would become uninsured using either growth factor. The remainder would instead either obtain subsidized health insurance through the marketplaces or enroll in an employment-based plan. For per-enrollee caps, the agencies estimate—using the CPI-U growth factor—that the additional coverage would increase federal outlays by $50 billion and decrease revenues by $13 billion from 2024 through 2032. Using the CPI-U plus 1 growth factor, the agencies estimate that the additional coverage would increase outlays by $43 billion and decrease revenues by $11 billion over the same period.
As with caps on overall spending, the net savings from capping per-enrollee spending would depend greatly on the growth factor. The lower CPI-U growth factor, when compared with the CPI-U plus 1 growth factor, would increase savings by an additional $331 billion. The estimated net savings could be made larger or smaller by adding or subtracting additional percentage points from the CPI-U, and the change to net savings would be reasonably, though not perfectly, proportional. That is, each additional 1 percentage-point reduction in the growth factor would increase savings to Medicaid by a similar dollar amount; the offsetting increases in outlays and reductions in revenues attributable to the loss of Medicaid coverage would increase by less.
Using the same base year, the same growth factors, and the same implementation date, CBO estimates that per-enrollee caps would save the federal government more than caps on overall spending. The per-enrollee caps would have a larger effect on the deficit because of the way federal spending would change in response to state eligibility restrictions. As explained above, CBO expects that states would respond to both the per-enrollee caps and the overall caps by seeking to offset a portion of the additional costs they would face, including by taking steps to restrict eligibility. Under per-enrollee caps, the reduction in enrollment would cause the states to receive less federal funding, and the federal government to save more, because funding would be tied directly to enrollment. By contrast, under the overall caps, the reduction in enrollment would not change the amount of federal funding that would be available to states because that funding would not be affected by changes in enrollment. However, other combinations of base-year data, growth factors, and implementation dates could result in overall caps' saving more than per-enrollee caps.
Uncertainty About the Budgetary Effects
There are two principal sources of uncertainty in the estimates of savings arising from this option. First, if projected spending growth, which averages 5.5 percent in the second half of the 10-year period, was substantially lower in the absence of the caps than CBO projects, the savings realized by capping Medicaid spending would be significantly smaller. In an extreme case, if spending growth under current law was less than the CPI-U in each year, then capping Medicaid growth by implementing either the overall caps or the per-enrollee caps would produce no savings. By contrast, if spending growth under current law was substantially higher than CBO projects, then the savings would be significantly larger, as would the pressure on states to make adjustments to their programs.
The second source of uncertainty pertains to whether and how states would choose to alter their Medicaid programs in response to the caps. Under per-enrollee caps, if a state chose to leave its Medicaid program unchanged and instead found other ways to offset the loss of federal funds, there would be little or no change in total combined federal and state Medicaid spending or enrollment. In addition, the federal government would incur few or no offsetting costs and revenue reductions associated with former Medicaid enrollees' obtaining other subsidized health insurance. By contrast, if under per-enrollee caps states made more significant reductions than expected to future Medicaid spending and enrollment, federal Medicaid savings would be larger and more former Medicaid enrollees would obtain subsidized health insurance or become uninsured, which would increase the associated offsetting costs. Under overall caps, states' changes to enrollment would have no effect on federal savings because federal payments would not adjust as enrollment changed. However, such changes would affect enrollees and health care providers.
In its distributional analysis, CBO allocates reductions in spending directly to the beneficiaries of that spending program. Most Medicaid enrollees' income is under 138 percent of the federal poverty guidelines, so the effects of reduced Medicaid spending would fall principally on households toward the bottom of the income distribution. (In 2022, the federal poverty guideline is $13,590 for single-person households in the 48 contiguous states and the District of Columbia and increases by $4,720 with each additional household member.)
Medicaid enrollees are not the only group that would be affected by a reduction in federal Medicaid spending. Medicaid payments from the federal and state governments go directly to health care providers, health care plans, and companies that sell prescription drugs. If, in response to lower federal payments, states reduced providers' payment rates, discontinued coverage for optional services, or covered fewer people, compensation throughout the health care industry would fall, affecting people across the income distribution, including some health care providers at the top of the distribution.
For the purposes of these estimates, CBO anticipates that, on average, states would replace about one-third of the lost federal funds; however, the agency does not project how individual states would respond to the change. To replace lost federal spending, states could reduce spending in other areas, increase existing taxes, or introduce new taxes. Each of the potential responses would have its own specific distributional effects, and the net effect of the option would reflect the impact of reduced Medicaid spending and the consequences of those other changes.
In addition to having the behavioral effects reflected in conventional budget estimates, such as the ones shown above, a reduction in Medicaid spending could affect the labor supply and people's saving; those effects would apply both to enrollees and to employees in the health care industry. For enrollees, a reduction in Medicaid spending could lead to poorer health outcomes and thus reduce the number of able-bodied workers and their productivity. Because many enrollees are disabled, elderly, or children, and do not or cannot work, the decrease in the labor supply would most likely be small. But, for those who do work, a loss of benefits could increase the number of hours they work to compensate for the need to spend more of their own resources on health care. Whether the combination of those two effects would increase or decrease the total number of hours that enrollees work is uncertain, but the economywide effect on hours worked would probably be small.
A reduction in benefits that caused enrollees to increase their own medical spending could cause them to cut back on other types of consumption and on saving. Because lower-income households have lower saving rates, which can be zero or even negative, the effect on such households' finances could be consequential, possibly leading to a significant increase in medical debt and bankruptcies. Economywide, the effect on saving would probably be small.
Across the health care industry, the effect of Medicaid cuts would vary widely and would depend on each provider's mix of Medicaid patients and other types of patients. The labor supply of health care workers and the amount they save could be reduced because of a decrease in their income. That decrease in income would result if there was a drop in the demand for services or a reduction in Medicaid payment rates. Reductions in Medicaid eligibility and enrollment could also lead to increased enrollment in higher-paying private plans, increasing some health care workers' income, the number of hours they work, and the amount they save. Economywide, the net effect on hours worked and saving would probably be small.
As with distributional effects, CBO does not project how individual states would respond to the changes and does not estimate the specific economic effects of each potential response. The net effect of the option would reflect the impact of reduced Medicaid spending and the consequences of those other changes.
Caps on federal Medicaid spending would represent a fundamental restructuring of Medicaid financing. In addition to their consequences for the federal budget, the limits on federal spending would have significant consequences for states. Capped federal spending would create uncertainty for states as they planned future budgets because it could be difficult to predict whether Medicaid spending would exceed the caps and thus require additional state spending. Moreover, depending on the structure of the caps, Medicaid might no longer serve as a countercyclical source of federal funds for states during economic downturns. (Under overall caps, the states might not automatically receive more federal funds if a downturn caused an increase in Medicaid enrollment.)
If the limits on federal payments were set low enough, additional costs—perhaps substantial costs—would be shifted to states. States then would need to decide whether to commit more of their own revenues to Medicaid or reduce Medicaid spending by cutting payments to health care providers and health care plans, eliminating optional services, restricting eligibility for enrollment, or (to the extent feasible) arriving at more efficient methods for delivering services. Under proposals that led to significant reductions in federal payments, many states would find it difficult to offset the reduced federal payments solely through improvements in program efficiency. If reductions in federal payments were large enough, states would probably resort to a combination of all approaches. All of those effects would be magnified beyond 2032 as the difference between the capped federal payments and the full cost of providing services to Medicaid enrollees grew wider over time.
Enrollees would be affected in various ways if states reduced providers' payment rates or payments to managed care plans or cut covered services. If states reduced payment rates, fewer providers might be willing to accept Medicaid patients, especially given that, in many cases, Medicaid's rates are already significantly below those of Medicare or private insurance for some of the same services. If states reduced payments to Medicaid managed care plans, some plans might shrink their provider networks, curtail quality assurance, or drop out of the program altogether. If states reduced covered services, some enrollees might decide either to pay out of pocket for medical services or to forgo those services entirely.