Executive Summary

Budget issues are central to the ongoing debate over health care reform. Many policymakers want to expand insurance coverage, but boosting federal spending is a challenge given today's record deficits. At the same time, some policymakers want to address our nation's long-run fiscal imbalance, which is being driven by rapidly rising health care costs and population aging. Finally, many people in the policy community remember how budget considerations helped to derail the reform proposals put forward by President Bill Clinton in the 1990s.

For all of those reasons, policymakers, analysts, and journalists are paying particularly close attention to the budget analyses of the Congressional Budget Office (CBO). CBO analyses often rely on sophisticated economic modeling and are usually framed in ways that match the specific, sometimes arcane, requirements of the congressional budget process. As a result, the cost estimates and related analyses may sometimes be challenging to understand. The unfortunate result can be confusion about what the scores mean and, equally important, what they do not mean.[1]

To avoid such confusion, it is important for lawmakers and commentators to understand fully the CBO's role in the legislative process and how it develops its scoring of proposed legislation. It is also important to understand the particular challenges that the CBO faces in scoring health legislation.

What the CBO Does

The CBO, working for the committees and Members of Congress, is a key part of the budget process. The Senate and House Budget Committees are its most important clients, given the key role that they play in guiding the congressional budget process, but the CBO works closely with every committee.[2]

The CBO is providing six basic services in the ongoing health discussion.

1. The CBO estimates the budget costs of proposed legislation. [3] These scores are often made public, either as published cost estimates or as tables that are distributed during legislative negotiations. Equally important, however, are the confidential, behind-the-scenes estimates that the CBO prepares for staff and Members while they are in the process of developing legislation.
2. The CBO provides estimates of policy impacts that are related to the budget. In its analysis of recent health insurance proposals, for example, it included estimates of how the proposals would affect coverage.[4]
3. The CBO decides whether proposed policies should be treated as part of the government and thus be recorded on the budget for congressional purposes. [5] This is a particularly important aspect of health budgeting and was a key issue during debate over President Clinton's health proposals. In that case, the CBO decided that the regulatory requirements on the private sector were so tight that much of the health care system ought to be included in the budget.

   The CBO recently released a report that outlines how it will approach this decision this time around.[6] The key takeaway, as summarized in Director Douglas W. Elmendorf's blog, is that:

   In CBO's view, the key consideration is whether a proposal would be making health insurance an essentially governmental program, tightly controlled by the federal government with little choice available to those who offer and buy health insurance--or whether the system would provide significant flexibility in terms of the types, prices, and number of private-sector sellers of insurance available to people. The former--a governmental program--belongs in the federal budget (including all premiums paid by individuals and firms to private insurers), but the latter--a largely private-sector system--does not.[7]

4. The CBO decides whether proposed legislation would impose mandates on either state or local governments or the private sector. Moreover, it estimates how large those mandates are. Mandates can sometimes be used to raise points of order against proposed legislation.
5. Upon request, the CBO often provides supplemental information that goes beyond the scope of basic cost estimates. Such information can be very useful in understanding the basis for and implications of budget scores.
6. The CBO devotes substantial resources to studies related to health policy and budget issues. The studies take many forms, including broad evaluations of policy options, examinations of the factors driving spending and coverage, and empirical analyses of past policy changes.

The CBO does not make policy recommendations. Although the CBO provides many services to Congress, it is categorically not its job to encourage or discourage particular policy actions. The CBO's role is limited to ensuring that Congress has the best possible information--on budget impacts and related factors--to make policy decisions.

This distinction is occasionally overlooked. Observers sometimes suggest, for example, that the CBO opposes a particular policy just because it concludes that the policy would have significant costs. Similarly, policy advocates sometimes suggest that the CBO should give favorable scores to policies that they favor. Such views fundamentally mistake the CBO's role, which is to provide unbiased cost information so that Congress can evaluate whether particular policies make sense.

The Basics of Budget Scoring

Scoring focuses on revenues and mandatory spending. The budget debate about health care focuses primarily on revenues and mandatory spending, the two largest components of the budget.

Revenues are the money that the government raises through taxes, most importantly those on payrolls, individual incomes, and corporate profits. Revenues are important in the health debate because of the special tax treatment that some health spending and insurance premiums receive, because tax credits are one policy option to increase coverage, and because increasing revenues may be one way to offset the budget costs of expanded coverage.[8]

Mandatory spending is any spending that occurs outside the annual appropriations process. The most important examples are the major entitlement programs such as Medicare and Medicaid. CBO cost estimates generally refer to mandatory spending as "direct spending" because the legislation that authorizes these programs also directly appropriates the money to pay for them. Most proposals to expand health coverage or reduce federal health spending involve mandatory spending.[9]

Scores may also address discretionary spending, but they ignore interest. Although health budget debates usually focus on revenues and mandatory spending, there are two other types of spending: discretionary spending and interest on the national debt.

Discretionary spending is handled through the annual appropriations process. In general, this means that two laws are required for the spending to occur: one to authorize the spending and the second to actually appropriate the money. Some health reform proposals would provide new authorizations for discretionary spending. As a result, some CBO analyses will include estimates of the amount of money that is authorized for such programs and the amount that will be spent over time. These amounts are not part of the official budget score, however, because the legislation merely authorizes the spending; it does not actually cause it to happen. Actual spending will occur only if it is so directed in a future appropriations act.

Interest on the national debt is technically a form of mandatory spending, but the congressional budget process does not consider it when evaluating legislation. Legislation might increase or decrease the size of the debt, and thus increase or decrease the amount of interest payments, but those changes fall outside the official budget process.[10]

Legislation can affect revenues and mandatory spending in various ways. When evaluating proposed legislation, budget analysts try to track all the ways that it might change revenues and mandatory spending. Some are obvious. An increase in Medicare eligibility, for example, will increase Medicare spending, and limiting the tax exclusion for employer-sponsored health insurance will increase revenue.

Other ways may be less obvious, with changes in one program leading to higher or lower spending in others. In addition, changes in spending programs sometimes lead to changes in tax revenues and vice versa. Such indirect effects are commonplace but can often cause confusion for observers who have focused only on what they expect to be the direct effects of a policy change.

Typical examples of indirect effects include the following:

• Expanding Medicare eligibility might reduce enrollment in Medicaid. The gross costs of increased spending on Medicare might thus be offset in part by reduced spending on Medicaid.
• Encouraging enrollment in the State Children's Health Insurance Program (SCHIP) might lead to greater enrollment in Medicaid. The costs of increased spending on SCHIP might thus be amplified by increased spending on Medicaid.
• Subsidies for non-employer health insurance might reduce enrollment in employer-sponsored health insurance. The costs of increased spending on subsidies might thus be offset in part by increased tax revenues.

That final linkage--the indirect connection between revenues and health spending programs--is a frequent source of confusion in health cost estimates. Revenues usually rise or fall because of changes in the tax code. In health, however, spending and regulatory policies that affect private insurance can also have revenue implications because employer-sponsored health insurance is exempt from payroll taxes and individual income taxes.

If a policy change would result in less spending on employer-sponsored health insurance (e.g., by reducing private costs or by reducing enrollment), CBO predicts that employers would respond by increasing salaries and wages. The amount of overall compensation would remain the same, but the amount of taxable compensation would increase, thus boosting tax revenues.

This effect can be large. The CBO's preliminary analysis of the first version of the Kennedy health plan, for example, found that reductions in employer-sponsored insurance, with a corresponding increase in cash compensation, would increase tax revenues by $257 billion over 10 years. New spending in the bill totaled $1.3 trillion. Because of the tax effect, however, the net budget cost of the bill was reduced to $1 trillion.[11]

Policy proposals are evaluated relative to a baseline. To estimate the budget effects of proposed legislation, the CBO compares spending under the legislation to a baseline estimate of what would happen without the legislation. In preparing those baselines, the CBO takes account of existing law, trends in enrollment, known cost drivers, and a host of other factors. Baselines can cause confusion in at least two ways.

Baselines sometimes have features that can be unreasonable in practice. For example, Congress enacted the sustainable growth rate (SGR) formula in 1997. Under this formula, physician fees in Medicare Part B were to be adjusted each year to maintain a Part B budget target. Given the high trend of spending, the CBO Medicare baseline assumes that the mechanism will lead to large reductions in physician payment rates, even though Congress has overridden those reductions in most recent years.

The CBO baseline assumption makes sense as a matter of budgeting, since it forces Congress to grapple with the budget implications of changing the SGR each year, but it also has the unfortunate side effect of creating a budget baseline that may differ from a reasonable expectation of future policy.

Baselines include projections about how the practice and financing of medicine will change in coming years. Projecting the effect of underlying trends in medicine can have significant effects on the potential budget impact of proposed policy changes. For example, suppose a new treatment is developed that would reduce the costs of Medicare treatments by $10 billion per year.

Some policymakers might expect that they could get credit for budget savings of $10 billion per year if legislation includes a provision requiring the use of that technique, but that logic is not correct from the CBO perspective. The potential budget savings depend on how often the new technique would be used, relative to what would happen without the legislation. If the CBO expects that all providers would adopt the technique on their own--and so includes the procedure in developing the baseline--then it would not give the proposed legislation any credit for potential budget savings.

Key Issues in Scoring Health Proposals

Behavioral responses are important. When the CBO projects the impact of policy changes, it tries to predict how consumers, workers, providers, employers, state governments, and insurers would respond to those changes.[12] These behavioral responses can take many forms. For example:

• How would physicians, nurses, and other providers respond to changes in payment rates? Would they provide more or fewer services?
• How would employers respond to changes in the tax treatment of employer-provided health insurance? Would they provide more or less coverage?
• How would individuals respond to the creation of a public health insurance plan? How many individuals would sign up?
• How would consumers respond to changes in Medicare copayments, deductibles, or premiums? Would their health spending rise or fall?

There is often vigorous debate among analysts about the appropriate answers to such questions. Lacking professional consensus, the CBO often has to make difficult choices among competing, plausible views. It does this by synthesizing the various points of view--reviewing the literature, talking to experts, and, in some cases, doing internal research--and then trying to find a reasonable middle ground.[13]

Changing health habits and medical practices doesn't necessarily reduce costs. In recent years, lawmakers have often considered policies that would change health habits or the practice of medicine. These proposals come in many flavors--e.g., increased use of information technology, greater reliance on preventive care, employing disease management techniques, and comparative effectiveness--but share a common goal: improving care and reducing costs.[14]

Regrettably, there are many reasons why these policies may not reduce federal spending as much as proponents expect, if at all:

• Empirical studies often find that such changes lead to improved health, not lower costs.[15] For example, the costs of screening a large number of people for a particular ailment may offset the cost savings of identifying the ailment early in a few patients. This does not mean that screening is necessarily bad; it just means that it would not save money.
• Some health improvements might increase federal spending. For example, improving the health of people who are 55 to 64 years old might increase Medicare and Social Security spending since more of them would live long enough to become beneficiaries.
• New practices may not produce widespread benefits. Proponents often emphasize studies that show the best possible outcome, but the CBO has to consider how changes in medical practice would be implemented on a broad scale across the health care industry. A practice that improves efficiency in a tightly integrated system, for example, may be ineffective in the fragmented world of individual doctor's offices.
• Some changes in practice will not reduce spending unless they are coupled with changed incentives. Improving the quality of information that doctors have about treatments, for example, may have relatively little effect unless they have incentives (financial or otherwise) to act upon it.

Having reviewed many proposals in recent years, the CBO recently concluded:

Some of these initiatives could improve individuals' health or enhance the quality of the care that they receive, but it is not clear that they also would reduce overall health spending or federal costs. In its analysis of such initiatives, CBO considers the available studies that have been done to assess particular approaches. In many cases, those studies do not support claims of reductions in health care spending or budgetary savings.[16]

Policy impacts depend on payment rules. Another reason that changes in medical practice may not reduce federal spending as much as proponents expect is that spending is often controlled by formulas that determine provider payments and beneficiary premiums. These formulas sometimes offset or eliminate any federal savings that might result from changes in medical practices. For example:

• Some provider payment rates are set by statutory formulas. If a policy intervention reduces costs for these providers but does not reduce the payment formula, government spending will not decline (and might actually increase if the lower costs would induce providers to perform more services). To get budget savings, policymakers would also have to change formulas to lower the payment rates, which might be politically challenging.
• Spending on physician services in Medicare is controlled (at least in principle) by the sustainable growth rate mechanism. If a policy intervention would reduce physician spending in a particular year, the SGR implies that spending may be higher in later years since the SGR is targeting a cumulative spending target. To get persistent budget savings, policymakers may thus need to change the SGR.
• Beneficiary premiums in Medicare depend on the projected cost of Medicare services. If a policy intervention would reduce Medicare costs, some of the resulting cost savings would therefore go to beneficiaries as lower premiums rather than to the government as lower spending. To capture all the potential cost savings, policymakers would have to change the way that premiums are determined.

Budget scoring rules prohibit certain impacts from being scored. The CBO must follow certain rules in evaluating the budget impacts of legislative proposals. These rules were originally established in the conference report on the Balanced Budget Act of 1997. The two Budget Committees, the CBO, and the Office of Management and Budget are jointly responsible for enforcing these rules, and any change in the rules requires the agreement of all four of them.

Among other things, those scorekeeping rules prohibit the CBO from giving credit for certain types of budgetary savings. If proposed legislation would give an agency more money to combat wasteful spending, for example, the CBO would take into account the cost of the additional spending but would not take into account any of the potential spending reductions. The CBO would include any savings that would come from new authorities or powers, but it could not include any potential savings that would arise merely by spending more money. Similarly, if new discretionary spending (e.g., on preventive care) might lead to reductions in mandatory spending (e.g., in Medicaid), the CBO cannot give the appropriations bill any scoring credit for such potential reductions.

These rules can obviously create situations in which cost-saving policy changes do not get scoring credit for potential savings. However, Congress established these rules for a larger purpose: to "avoid situations where hoped-for, but quite uncertain, savings are used to offset near-term, certain spending increases or revenue decreases."[17]

Interpreting Budget Scores

Commentators often summarize CBO cost estimates as a single number: the net budget impact over the 10-year "budget window."[18] That shorthand is understandable because the 10-year costs often play an important role in the congressional budget process and are typically the focus of cost estimates.

But a single figure frequently conceals more than it reveals. To truly understand the CBO's budget scores, it is essential to look behind the headline figures to the details: the scores for individual programs and the scores for individual years.

Gross vs. net over the budget window. One way to unpack CBO cost estimates is to distinguish between provisions that worsen the budget situation (i.e., increase spending or decrease revenues) and those that improve the budget situation. The headline cost estimate reports the net of these impacts, but the ultimate policy implications usually depend on the underlying gross impacts.

The CBO's analysis of the initial Kennedy bill, as noted, found a net budget cost of $1 trillion over 10 years, but it actually showed that the policy would spend about $1.3 trillion to expand coverage and would raise about $257 billion in higher tax revenues as an offset. In that case, the gross figures-- which show the impact on spending and on revenues separately--provide a much richer description of the policy change than is provided by the net figure alone.

Gross vs. net for specific provisions. A similar type of unpacking can be useful for individual provisions whose impacts may differ across years. The recent economic stimulus bill, for example, included provisions that were intended to encourage physicians to adopt health information technology (IT). The headline cost estimate for this provision was a net budget cost of $21 billion over 10 years. As a result, the media often reported that the Administration was spending $21 billion to encourage health IT.

In reality, however, the relevant provision would actually spend at least $36 billion through 2015 in promoting health IT, although that cost was expected to be offset in later years from savings generated by penalties on non-adopters as well as expected cost savings.[19]

Limitations of 10-year budget windows. Estimating the cost of a proposed program over 10 years, as the CBO typically does under its rules, provides a cost projection for those years, but not for any subsequent years in the life of the program. Moreover, a 10-year total does not of itself indicate the pattern of costs during those years.

Thus, users of cost estimates should examine the time pattern of potential impacts within the budget window and, if possible, beyond.

• Some budget impacts may not show up inside the window.One reason is that policy changes can have impacts that stretch over multiple decades. In those cases, 10-year budget windows provide an incomplete perspective on their true budget impacts. That problem is particularly severe if the budget costs and benefits of policies appear at different times.

1. Some programs may not get credit for benefits that occur outside the budget window. For example, policies that would reduce smoking or obesity in the next few years may have benefits that do not show up for decades.
2. Some programs may not be charged for costs that occur outside the budget window. For example, the second version of the Kennedy health bill includes a provision that creates a new form of long-term care insurance. The premiums for the insurance increase faster than the payments in initial years, so the program scores as a budget saver over the 10-year window. In later years, however, the provision may persistently increase the deficit.[20]

• Ten-year costs depend on when the program starts and how fast it increases. There is a natural tendency to hear an aggregate 10-year cost estimate and divide that amount by 10 to get a sense of how much the annual budget impact will be. That approach works only if the program comes into existence fully fledged in the first year and if cost impacts do not change too much in subsequent years.

In practice, however, legislation often allows several years to set up a program, and the program is phased in only after that period. With health care, costs can increase very rapidly from year to year, so the average cost for the first 10 years may be much lower than later annual costs.

The Medicare Modernization Act of 2003, which created the Medicare Part D prescription drug benefit, is an excellent example of this. Chart 1 illustrates the year-by-year estimates that the CBO prepared for the conference report on the bill.

Total spending on the new drug benefit was estimated by CBO at about $400 billion over the 10-year budget window. However, this did not mean that the program would cost anything like $40 billion per year into the future. The program would not even be fully phased in until the fourth year of the forecast period, and costs were projected to keep rising rapidly in subsequent years.

As a result, the cost was actually projected at $73 billion in 2013, the final year of the 10-year budget window reported by the CBO, after which it would keep increasing, bringing costs in the second decade alone to well above $1 trillion.

Even fully offsetting spending can worsen the budget situation. Policymakers have pledged (much to their credit) to pay for any new health reform enacted this year. As the foregoing discussion indicates, however, it is not always straightforward to determine what exactly that means.

Under the usual conventions of budget discourse, "paying for health care reform" can mean merely that the 10-year budget score must net to zero: Any spending increases on expanded insurance coverage must be offset by spending reductions or revenue increases elsewhere.

That requirement is certainly necessary for budget neutrality, but it is not sufficient. Policymakers and commentators must also take account of the trajectory of spending.

Peter Orszag, director of the Office of Management and Budget, recently recognized this issue and suggested that an additional requirement should be that the net budget impact be zero in the final year of the budget window.[21] That is certainly a step in the right direction, since it makes it less likely that a program could be experiencing an accelerating deficit at the end of the budget window. But it is not sufficient, by itself, to ensure that new health spending is really paid for over the long run.

The major challenge with new health spending is that it is likely to grow as fast as health spending overall. History suggests that health spending grows faster than most other components of the budget, making it very difficult to identify offsets that will keep up with it over time. The obvious exception, of course, is offsets that are also driven by health spending: e.g., reductions in specific kinds of health spending or reductions in the value of the tax exclusion for employer-sponsored health insurance.

Yet, even if policymakers do identify spending reductions and revenue increases that would offset the long-run costs of health reform, there is still one more challenge: Using those offsets now may make the nation's long-run budget situation more difficult. The reason is that any offsets used to pay for health reform cannot be used in the future to address the long-run budget imbalance in our existing fiscal situation. As a result, health reform that is fully paid for today may still make it more difficult to get the budget under control in the future.

Supplementary Information

Both the Congressional Budget Office and its congressional clients are well aware of the challenges that may arise in understanding budget estimates. As a result, the CBO often provides additional information about the basis for its estimates and about impacts that fall outside the conventional scores. This usually happens in response to questions from lawmakers and their staffs. Some recent examples include:

• Estimates of policy impacts that are relevant to the budget score. In its estimates of recent health care legislation, for example, the CBO reported not only budget impacts, but also estimates of how insurance coverage would change.
• More detail about the factors underlying an estimate. When the CBO released its estimate of the first version of the Kennedy health bill, for example, it received many questions about the changes in employer coverage. CBO Director Elmendorf responded by providing additional information on his blog.[22]
• Impacts beyond the budget window. When the CBO examined immigration reform proposals in 2006, for example, the 10-year score did not reflect the long-run budget impacts of an increased immigrant population. In response to requests, the CBO prepared additional estimates of the budget impacts assuming that the effects of the proposals were fully phased in.[23]
• Interest payments. As usual, the CBO's cost estimate for the recent economic stimulus bill did not include any effects on interest payments. In response to a request, however, the CBO estimated how much interest payments would increase over the budget window.[24]
• Macroeconomic impacts. As usual, the CBO's cost estimate for the recent stimulus bill did not account for potential impacts on the size of the economy. In response to a request, however, the CBO estimated how large such "dynamic" effects might be.[25]

Conclusion

CBO cost estimates will play a central role in the ongoing debate about health policy. This is appropriate, given the potential cost of some proposals and the overwhelming budget challenges that we face.

It is important to remember, however, that the cost estimates cannot speak for themselves. There is a natural tendency to focus solely on the bottom line--estimated costs added up over the next 10 years--but by itself, a cumulative 10-year cost says little about the merits of particular health policies. One can easily imagine good policies and bad policies that have the same 10-year cost.

Nor do cumulative cost figures say much about the long-run budget impacts of particular policies. Those depend on the long-run trajectory of costs and potential offsets, not just on the impacts in the first 10 years.

To get a handle on those critical issues, policymakers, analysts, and journalists need to dig into the cost estimates, examine their details, understand the limitations of the congressional scoring process, and make good use of all the additional information that the CBO can and does provide.

Donald B. Marron  served as deputy director of the Congressional Budget Office from October 2005 to August 2007, including more than a year as acting director, and later served as a member of the Council of Economic Advisers. He now consults on economic, financial, and budget matters and writes the blog dmarron.com.