CRS Reports on Baseline Budget Projections Issues

[1] Between January 2001 and 2002, the Congressional Budget Office (CBO) reduced its 10-year baseline surplus projection from $5.6 trillion to $1.6 trillion for three reasons. First, the recession both reduced revenues and raised outlays automatically. This is the largest factor reducing the surplus in the first couple of years of the projection, but a smaller factor in later years. Second, projections were affected by policy changes made in 2001, the largest of which was the Economic Growth and Tax Relief Reconciliation Act (P.L. 107-16), and the second largest of which was increased outlays in response to September 11. The budgetary impact of P.L. 107-16 increases significantly over the 10-year window while the overall increase in outlays remains nearly constant. Finally, CBO made technical revisions to its projections (e.g., to account for the smaller than predicted growth in tax revenues in 2001) that reduced the surplus.

[2] Although larger than usual by historical standards, this dramatic revision in surplus projections should not come as a surprise. Baselines are meant to project the future budgetary path of current policy; they are not meant to serve as a "best guess" of future budgetary outcomes, Without a baseline projection, policymakers would be in the dark when planning the budget. Nevertheless, an overriding focus on the baseline projection will lead to conclusions that can be radically misleading. This is true for three reasons.

[3] First, baseline projections are only as accurate as the assumptions underlying them. Critics have argued that several of the underlying assumptions or rules followed in making the budget baselines are not as realistic as they could be. Applying alternative assumptions to the baseline could significantly reduce the projected size of the surplus. The baseline treatments of discretionary spending, which can change unpredictably from year to year, and expiring tax provisions are two assumptions that have been criticized.

[4] Second, budget baseline estimates and projections are highly sensitive to relatively small changes in the underlying assumptions and economic factors. These changes can have substantial effects on the surplus projections, and the effect on the projection compounds when extrapolated into the future. In particular, our understanding of the economy remains limited and economic forecasts remain subject to extremely large margins of error, even over short time periods. Thinking of the baseline projection as a certain outcome distorts the policymaking process.

[5] Third, baseline projections are limited to a 10-year period. While one would expect 10 years to be a more than adequate time horizon to assess the course of future policy, the U.S. faces a unique situation beyond that horizon: the retirement of the baby boomers. Under current policy, their retirement is likely to place an unsustainable strain on government finances. In a narrow sense, the baseline should not reflect the problem, for it is not supposed to advocate policy changes. Nevertheless, critics argue that to the extent that the baseline frames the budget debate, a baseline that makes unsustainable policy appear sustainable is misleading. This report will be updated as events warrant.

Contents

 What Baselines Can Do

Figure 1: Historical Discretionary Spending

List of Tables

Table 1: Probability Distribution of Baseline Surplus/Deficit(-) Projections Made in January 2002

Table 2: Changes to the Baseline Projections From 2001 to 2002

Table 3: Estimated Probability of Different Events Based on Information Available February 2001

Table 4: Baseline Surplus Under Alternative Assumptions

[8] The proper way to use a baseline is as a rule-of-thumb estimate for the budgetary ramifications of current policy. This offers the policymaker a means to measure the relative effects of proposed legislation in the context of the overall budget. Current policy is very narrowly defined in these projections. It does not include proposals made in adopted budget resolutions, bills passed by only one chamber,³ or even bills passed by both chambers but not yet signed into law.

[10] First, baseline projections are only as accurate as the assumptions underlying them. Critics have argued that several of the underlying assumptions or rules followed by CBO and OMB in making the budget baselines are not as realistic as they could be. Applying alternative assumptions to the baseline could significantly change the projected size of the surplus. As discussed more fully below, the baseline treatment of discretionary spending, which can radically change from year to year, and expiring tax provisions are two assumptions that have been criticized.

[11] Second, budget estimates and projections are highly sensitive to relatively small changes in the underlying assumptions and economic factors. These changes can have substantial effects on the surplus or deficit projection, and the effect on the projection compounds when extrapolated into the future. In particular, our understanding of the economy remains limited and economic forecasts remain subject to extremely large margins of error, even over short time periods. Thinking of the baseline projection as a certain outcome can distort the policymaking process.

[12] Third, baseline projections are limited to current-year expenditures (for 10 years). While one would expect 10 years to be a more than adequate time horizon to assess the course of future policy, the U.S. faces a unique situation beyond that horizon: the retirement of the baby boomers. Under current policy, their retirement would lead to a large expansion in funds dedicated to Social Security and Medicare that is likely to place an unsustainable strain on government finances. Since their retirement will mostly occur outside the 10-year window, the baseline does not reflect this problem. In a narrow sense, it should not reflect the problem, for the baseline is not supposed to advocate policy changes. Nevertheless, to the extent that the baseline frames the budget debate, critics argue that a baseline that makes unsustainable policy appear sustainable is misleading.

All three of these issues are discussed in detail below.

Discretionary Spending

[13] Discretionary spending presents a special problem to budget estimators. Accounting for about one-third of total outlays, it includes most spending in policy areas such as the military, transportation, education, and the environment. Unlike entitlements, there are few legal determinants of its levels; instead it is determined annually at the "discretion" of legislators. Since almost all discretionary funding comes through annual appropriations, Congress has significant control over the amounts involved. This means that there is no obvious growth rate of discretionary spending to use in budget forecasts. Arguably, the most useful rate of baseline discretionary growth for policymakers is whatever rate is most realistic.

[14] Although the Budget Enforcement Act (BEA) as amended requires that OMB and CBO assume discretionary spending will stay constant in inflation-adjusted terms in their respective baselines, such an adjustment is not the only reasonable one. For example, assuming discretionary spending grew at an average historical rate, remained constant on a per-person inflation-adjusted basis, or remained constant as a percentage of GDP would each produce different budget results for total discretionary spending, total outlays, and the surplus or deficit than does the inflation-adjustment requirement. A smaller rate of increase would slow overall outlay growth, increasing the size of future surplus projections. A higher rate of increase would speed total outlay growth, reducing future surplus projections.

[15] In their budget reports this year, CBO and OMB, as required, assumed that overall discretionary spending will stay constant in real, or inflation adjusted, terms. This has two implications. First, although discretionary spending is assumed to keep up with inflation, there is no adjustment for expected population growth. Under the baseline, therefore, future discretionary spending can buy the same amount of roads or military equipment or government services, but there will be fewer of them per person.

[16] Second, since the economy, as measured by gross domestic product (GDP), is assumed to grow in real terms over the next 10 years, but real discretionary spending is assumed to remain constant, discretionary spending would fall as a percentage of GDP. This implies that as society becomes wealthier, it will not want to spend any of its additional wealth on government-provided discretionary goods and services. Although there are undoubtedly some government- provided goods and services on which people may not wish to spend their additional wealth, it is not obvious why this would be true of total discretionary spending, as implied by the baseline. Over 10 years, maintaining discretionary spending at a constant percentage of GDP is estimated by CBO to increase cumulative outlays by $1,207 billion over the baseline levels. Assuming higher rates of discretionary spending growth is not inconsistent with the baseline concept of projecting current policy.

[17] Does recent history suggest what growth rate is most realistic for discretionary spending? (Of course, some of the historical change in discretionary spending is due to policy changes, which would not be reflected in a baseline.) As seen in Figure 1, total discretionary spending has, in fact, fallen slowly as a percentage of GDP for decades. Much of the decline in the 1990s came from a decline in military spending. Non-defense discretionary spending rose slightly as a percentage of GDP in the 1970s before falling slowly in the 1980s and 1990s.⁴ In 2000 and 2001, real discretionary spending grew at the rate of economic growth.

[18] While discretionary spending fell as a percentage of GDP, it grew more quickly than the rate of inflation in all but one year between FY1975 and FY1989. In the 1990s, it fell in real terms from FY1990-1995, but then grew more quickly than the rate of inflation from FY1996-2000. Removing the effect of the fall in defense spending, non-defense discretionary spending grew more quickly than inflation throughout the 1990s, with the exception of 1996. The baseline assumption that discretionary spending will stay constant in inflation-adjusted terms through FY2012 would drop total discretionary spending to its lowest share of GDP ever since discretionary spending totals were first recorded in the early 1960s. (Whether this is desirable is a separate issue not addressed in this report.)

[19] Some government spending and tax programs, especially tax credits, have expiration dates. Both CBO and OMB are required to assume in their baseline estimates that all tax measures (unless earmarked to a trust fund) and small spending programs will expire as scheduled, since that represents current policy according to the law. The baseline reflects this assumption by increasing revenues when the measure is due to expire, or reducing outlays when the program is due to expire. But most of these expiring provisions have proven very durable and are routinely extended. Some examples of expiring tax provisions include credits or deductions for clean-fuel vehicles, qualified zone academy bonds, welfare-to-work, medical savings accounts, research and experimentation, and empowerment zones.

[20] The routine renewal of minor tax provisions has made the proper baseline treatment tricky for years. But baseline treatment of expiring provisions became particularly difficult this year because of the decision to make the major tax cut of 2001, the Economic Growth and Tax Relief Reconciliation Act (EGTRRA, P.L. 107-16) expire in 2010 due to congressional budget rules. Clearly, expiration of the tax cut was not a goal of its proponents, and there have already been several proposals to make it permanent, including H.R.586, which was passed by the House on April 18. CBO estimates that if all expiring tax provisions were renewed, revenues would be reduced by $734.7 billion over 10 years. Of this, $374.4 billion is attributable to the expiration of EGTRRA.

Accuracy of Forecasts

[21] The baseline is inherently uncertain. Indeed, it should be viewed as the midpoint on a continuum of highly uncertain outcomes. Table 1 demonstrates the wide range of possible budgetary outcomes. It shows that -- without any policy changes -- there is a 25% probability of a budget deficit exceeding $73 billion in 2002, but there is also 25% probability of a budget surplus exceeding $32 billion. As one extrapolates further into the future, uncertainty increases. That is because CBO has less information on which to base predictions and earlier mistakes are compounded into larger mistakes. By 2007, the last year for which CBO constructs probability distributions, there is a 25% probability of a budget deficit exceeding $135 billion and a 25% probability of a budget surplus exceeding $467 billion.⁵ With margins of error that large, baseline projections, particularly in out years, cannot be counted on for accuracy; and like other economic projections reaching that far out, are not, by themselves, a meaningful basis for policy decisions.

[22] The probabilities are based on historical errors in CBO's projections and the uncertainty stems from two sources: errors in CBO's economic forecasts, and "technical" errors, which refer to all changes in budgetary outcomes that cannot be attributed to policy or economic changes. The economic errors imply no shortcoming on the part of CBO; the accuracy of their forecasts has typically been comparable to private sector forecasters. Historically, economic forecasts have been particularly inaccurate at spotting turning points in the business cycle, and the recent recession has been no exception. In January 2001, CBO, the Administration, the Federal Reserve, and virtually all major private forecasts predicted growth between 2.0-3.1% for the year. In reality, the economy grew at 1.0%. If anything, uncertainty concerning the economy has grown in recent years due to the sharp and unexpected acceleration in productivity growth and fall in unemployment in the late 1990s. Looking to the future, forecasters are highly uncertain how much of these changes should be considered permanent.⁶This is troublesome because even small changes in the underlying economic forecast lead to significant changes in baseline estimates.

[24] What does this uncertainty imply for policymaking? It implies that half of the time policymakers will find themselves with more money than anticipated, half the time they will run out of money. On average, CBO's surplus prediction for the coming budget year was too large or too small by 1.1% of GDP. Setting goals such as balancing the budget or saving the Social Security surplus according to the baseline does not ensure they will occur; in fact, the probability that they will occur is only 50%. To achieve, say, a 75% probability of balancing the budget in 2003, would require a planned surplus of roughly $88 billion dollars, based on the current baseline forecast.

Troubles on the Horizon -- Social Security, Medicare, and Medicaid

[25] Unlike corporations, the government considers only current-year liabilities in its budget for two reasons. First, it is typically assumed that only the current budget surplus or deficit affects aggregate demand in the current economy. Second, unlike corporations, the government has the power to alter its revenue (i.e., taxes) or spending levels as necessary to meet almost any future funding need. Critics have argued that since the baseline projects the budget for only the next ten years, it is unable to guide policymakers on longer term policy issues under the current accounting system. Nonetheless, long-range projections indicate that the government faces very large liabilities in its Social Security, Medicaid (which devotes many of its resources to the retired population), and Medicare programs beginning in the second decade of this century under current policy. The retirement of the baby boom generation will put enormous pressure on government finances. The benefits of these "pay as you go" programs are funded by current workers. Over the next 30 years, the ratio of workers per beneficiary is expected to fall from 3.4 to about 2.

[26] In 2001, total spending on Social Security, Medicaid, and Medicare equaled 7.8% of GDP. By 2030, spending is projected to exceed 14% of GDP, yet revenue from the payroll tax is projected to stay relatively constant. The difference between spending on these programs and the government's tax revenue is the "unfunded liability" that can only be financed through higher taxes, lower benefits than promised under current law, or the issuance of debt. The liabilities cannot be financed solely through debt issuance, however, because the size and persistence of the shortfall would quickly lead to unsustainably large budget deficits.

[27] The future problem could be mitigated by "financing" the liabilities today through a higher rate of government saving (i.e., a larger budget surplus). The surplus could be used to retire the national debt, which frees up private saving for greater private investment, or through some method of government investment in the private sector. Government accumulation of private assets could occur either through the funding of individual accounts or through direct government purchase. Economic theory sees little macroeconomic difference between debt retirement, government accumulation of assets, or the accumulation of assets in private accounts; the key is that the government uses the surplus to increase national saving. While increasing the national saving rate does not directly reduce these liabilities, by spurring greater capital formation it increases the future size of the economy.⁷ A larger future economy would ease the relative burden of paying for these future obligations. From a budgetary perspective, the problems would be eased as well. If the government used surpluses to purchase private assets today (centrally or through individual accounts), those assets could later be sold to help pay benefits. Similarly, if the government used surpluses to pay down the national debt, future interest payments on the debt would be lower and the savings could be used to help pay benefits.

[28] Critics can reasonably argue that although these problems are important, the baseline is not the proper forum for raising them. A short-term tool cannot accommodate the evaluation of long-term problems, they argue; longer term forecasts already in existence are a more appropriate forum. With margins of error so large even five years ahead, it would arguably be impractical to give, say, 75 year projections the institutional role in the budget debate that the baseline currently occupies. And incorporating long-term liabilities into the baseline could be seen as favoring this issue over other policy issues that people may believe to be equally or more important. Some argue that including future obligations would alter the baseline's role from one of neutrality to advocacy, undermining its reputation for impartiality. Nevertheless, it can be argued that to the extent the baseline frames the budget debate and makes unsustainable policy appear sustainable, it is also in this case misleading.

[30] Those observers who incorrectly use baseline projections as hard predictors of the future express surprise that a sudden reversal of fortune could occur. However, the tax act and changes in policy in response to September 11 (which include both defense spending and some of the increase in other discretionary spending) are two examples of policy changes that automatically reduce the baseline. Two of the shortcomings of an overreliance on the baseline described above can also be seen in the move from budget surplus to budget deficit in 2002. First, independent of policy changes, discretionary spending again grew faster than the baseline projected using the inflation adjustment.

[31] Second, the large change in the baseline for economic and technical reasons is also less surprising if one is familiar with the historical inaccuracy of forecasts. In the short term, the deterioration in the baseline forecast since last year attributable to economic factors is due to the recession which CBO, like other forecasters, failed to foresee. But in the longer term, economic factors reduce the surplus as a result of the 0.2 percentage point reduction in CBO's projected long-term economic growth rate. The CBO baseline now projects that less capital will be accumulated in the future and total factor productivity will grow more slowly than they had projected last year.

[33] The probability distributions CBO estimated for its baseline projections one year ago were the best information policymakers had at the time to judge the affordability of last year's policy changes. Table 3 offers some illustrative examples. Comparing those projections to the new projections made in January 2002, the changes made in 2002 due to economic and technical factors were larger than usual: CBO projected last year that there was only a 5% probability that the surplus projection would fall as much as CBO is now projecting (after adjusting for policy changes). This does not imply that CBO is getting worse at forecasting; as discussed above, recessions are the most difficult economic event to forecast, and only three of CBO's previous 19 forecasts occurred during a recession. Since CBO believed that there was only a 5% chance of the surplus declining so precipitously this year, there was little reason for policymakers to think that the legislative changes made in 2001 would cause a deficit in 2002.

[35] The choice of a balanced budget as a policy benchmark is arbitrary; policymakers may have had other goals for the level of surplus to maintain. For instance, maintaining on-budget balance (i.e., "saving" the Social Security surplus) had widespread bipartisan support in 2001 and was adopted in the "lockbox" amendment to P.L. 106-554, which became law on December 21, 2000 (SAMT.3690 to H.R.4577). One can use the same method to determine the probability that the legislative changes made in 2001 would have led to an on- budget deficit. Based on the 2001 probability distribution, the tax cut alone had a 20% probability of causing an on-budget deficit in 2002 and a 35% probability in 2006. All legislative changes made in 2001 had a 30% probability of causing an on-budget deficit in 2002 and a 45% probability in 2006.

[37] This report has highlighted two assumptions that could be altered: the growth rate of discretionary spending and the treatment of expiring tax provisions. Table 4 constructs an alternative baseline under the assumption that discretionary spending stays constant as a percentage of GDP and all expiring tax provisions are renewed. It also adjusts the surplus for the fact that these changes would lead to a larger national debt and higher interest payments on that debt. The table is meant to serve as a technical illustration rather than a recommended alternative. It indicates, as CBO notes, that the baseline is an inevitably arbitrary yardstick.

[39] Another conceptual mistake sometimes made in reference to the baseline is the assumption that the baseline projection will occur with certainty. The baseline is the midpoint in an array of possible outcomes. Because our understanding of the economy in general -- and causes of the business cycle in particular -- is limited, there is a high degree of uncertainty surrounding these estimates, even over short time intervals. Over longer time periods, uncertainty grows, arguably counseling against policy changes whose budgetary effect grows over time. Table 1 illustrates the probability distribution of baseline predictions. Whether these estimates are disconcerting or sanguine depends on the goals of policymakers.

[40] One rule of thumb for budgeting calls for a balanced budget. Since budget deficits decrease national saving and budget surpluses raise national saving, this view is justified on the grounds that a balanced budget would keep government influence on the national saving rate to a minimum. There is a 50% chance that this goal will not be met under current policy in 2003 and a 35% chance it will not be met from 2005-2007 using the official baseline. Using the alternative baseline constructed in Table 4, the probability of persistent budget deficits becomes much higher.

[41] Another rule of thumb calls for a balanced structural budget, which would result in deficits in recessions and offsetting surpluses in expansions. This view is justified on the grounds that in many economic models budget deficits can boost aggregate demand to mitigate a recession. Under this view, the high probability of budget deficit in 2002 is less troubling, although deficits still remain a distinct possibility after the recession has ended.

[42] An alternative rule of thumb calls for a policy of persistent budget surpluses. This view is rationalized on the grounds that, by raising the national saving rate and reducing the net liabilities of the government, surpluses would place the nation in a better position to grapple with the problems related to the future retirement of the "baby boom" generation. There is a slightly higher than 50% probability that large budget surpluses will reappear under current policy using the official baseline. Under the alternative baseline constructed in Table 4, however, there is a low probability that large budget surpluses will reappear.

²CBO, The Budget and Economic Outlook, Jan. 2001, p. 7. Instructions for creating the baseline estimates are contained in the Budget Enforcement Act (BEA) as amended.

³Accounting for these bills can result in significant revenue loss. For example, the Securing America's Future Energy Act of 2001 (H.R.4) as passed by the House would result in a 10-year revenue loss of $34 billion.

⁴This could be interpreted as evidence that, overall, people do desire less non-defense discretionary spending as their income rises.

⁵For more information, see CRS Report RL30854, Uncertainty in Budget Projections, by Philip Winters.

⁶For more information, see CRS Report RS20608, The U.S. Long-Term Growth Rate: Has It Increased?, by Craig Elwell; and CRS Report RL30771, Productivity Growth in the Current Economic Expansion, by Brian Cashell.

⁷This is the economic rationale behind proposals to "save" the Social Security surplus. Increasing the holdings of U.S. Treasuries in the Social Security trust fund does not give the government any real financial assets because the government is lending to and repaying itself. Rather, many economists believe that if the Social Security surpluses are used to retire debt, the shrinking federal debt should increase the future size of the economy through increased private investment. The problem is that even the additional economic growth implied by the increase in national saving spurred by an increased trust fund does not appear to be enough to fund the current level of various program benefits for the large number of baby-boom retirees that become eligible in the first three decades of this century.

Similar arguments can be made for reserving the Medicare and federal employee retirement trust fund surpluses for debt retirement because these trust funds are also meant to fund future liabilities. The Medicare trust fund, combining Part A and Part B, totals $405 billion over the 10-year forecast. The civilian and military retirement funds total $476 billion.

⁸Jerry Tempalski, "The Impact of the 2001 Tax Bill on the Individual AMT," working paper presented at National Tax Association symposium, October 2001.