Taxes and the Budget: How accurate are short-run and long-run budget scenarios?
Short-run budget projections tend to be highly inaccurate, because of unforeseen changes in economic activity, changes in technical assumptions, and changes in economic and other policies. Typically, the first two types of changes alter projections much more than do changes in policies. Surprisingly, longer-run projections may be more accurate than shorter-run projections.
- Over the period 1983-2005, the average absolute error in the five-year revenue projection of the Congressional Budget Office (CBO) caused by changes in the economic and technical assumptions was 1.6 percent of GDP, which would be $219 billion at the 2007 level of GDP. Between the earliest official projection of the 2007 budget balance (in the ten-year projection of 1997) and the actual outcome, the projected balance varied over a range of $844 billion because of changes in economic and technical assumptions.
- If the CBO projection is too pessimistic in one period, it is extremely probable that it will be too pessimistic in the next period as well. Similarly, too much optimism in one period is likely to be repeated in the next. Thus one observes long streaks of overly pessimistic or overly optimistic projections.
- Especially large errors tend to occur after the fifth year of CBO’s ten-year projection period, and so one might question the prominence that CBO publications give to the second half of their projection periods. However, these projections serve an important role in estimating the longer-run effects of changes in tax and spending policies. Errors in the underlying economic and technical assumptions shift projections of aggregate revenue and outlays but usually have little effect on estimates of the effects of policy changes. Even so, it might be better to confine those longer-run assumptions to appendices in CBO reports, to avoid giving the appearance of false precision.
- If short-run projections have such huge inaccuracies, can very long budget projections have any credibility? Ironically, such projections may actually be more accurate. One reason is that policy assumptions (as opposed to economic and technical assumptions) made for the long run may be more realistic - for a while at least - than those made for the shorter run. When making its short-run projections, CBO is required to assume the continuation of current law, even if doing so implies unlikely outcomes, such as sharply rising tax burdens or declining nondefense spending relative to GDP. CBO does not face the same constraint when making its long-run projections, but instead may assume that tax burdens will remain roughly constant in the long run and that the ratio of spending to GDP, outside of Social Security, Medicare, and Medicaid, will change little. Such assumptions are much more consistent with past policies than is the assumption of unchanging law.
- However, the whole point of long-run projections is to see whether policies are sustainable over time, and recent analyses indicate that they are not--certainly not much beyond 2030, if even that long. The kind of errors that make short-term projections inaccurate have little effect on the main purpose of long-term analysis.
- Shorter-run revenue projections (after adjusting for policy changes) have incurred large errors over recent periods, because tax revenue has grown faster than GDP over some extended periods and more slowly over others. The effects of such cycles tend to even out over the very long run.
- The main reason for making very long run budget projections is to show how rapidly spending on Social Security, Medicare, and Medicaid will grow if these programs are not reformed. Long-run projections of Social Security outlays relative to GDP tend to be extremely accurate compared with revenue projections. One reason is that Social Security outlays depend heavily on demographic trends, which tend to be easier to project over several decades than economic trends, which are more important in determining revenue. Also, the indexing of initial Social Security benefits to wages and the important role of wages in GDP mean that variations in Social Security outlays and variations in GDP tend to move together. Therefore the long-run projection of the ratio of Social Security outlays to GDP can be quite accurate even when the forecast of GDP goes astray. Unfortunately, it is much more difficult to forecast Medicare and Medicaid outlays, and if long-run budget projections turn out to be misleading, it is very likely because large errors were made in forecasting health expenditure.
- The fiscal disaster portrayed by long-run budget projections is so overwhelming that it is likely to occur eventually in the absence of significant program reforms, even if today’s projections turn out to be far too pessimistic. And those projections could easily turn out to be not too pessimistic at all, but rather too optimistic.