A Short Guide to Dynamic Scoring
In recent years, official scorekeepers and academic researchers have devoted increased attention to the macroeconomic effects of tax cuts. The Treasury also conducted a “dynamic analysis” of the President’s tax cuts that was included in this year’s Mid-Session Review of the budget as well as in a separate Treasury report.
The results of much of this work indicate that tax cuts can have positive or negative effects on the economy, with the “sign” of the effects depending on a number of variables, the most important of which is whether and how the tax cuts are paid for.
The Congressional Budget Office, the Joint Committee on Taxation (JCT), and academic researchers have all have found that tax cuts that are not accompanied by offsetting revenue increases or spending reductions — and are financed by borrowing instead — can harm the economy over the long term. The research, including the Administration’s own analysis, also indicates that even if tax cuts are paid for, the economic benefits generally are relatively modest, with any increased revenues that result from stronger economic growth offsetting only a small fraction of what conventional cost estimates indicate the tax cuts will cost.
This short guide to dynamic scoring makes six points:
JCT and Treasury tax estimates already include “dynamic” scoring components. The official cost estimates that the Treasury and Congress’ Joint Committee on Taxation produce for tax-cut legislation are “dynamic” scores in that they incorporate changes in behavior that are expected to occur as a result of the proposed tax cuts. Such behavioral reactions include, for instance, changes expected in the timing of the capital gains realizations and changes in the form of compensation (e.g., from benefits to wages or vice versa). At the same time, the official cost estimates do not assume that major macroeconomic conditions like GDP and employment will be altered by tax changes.
Tax cuts do not pay for themselves. Economists of all stripes have consistently found that tax cuts do not generate enough growth to fully pay for themselves. In fact, cost estimates that incorporate macroeconomic feedback from tax cuts are reasonably close to conventional cost estimates that ignore such feedback. For example, two recent dynamic-scoring estimates by CBO showed that the President’s proposal to make the 2001 and 2003 tax cuts permanent, together with the President’s other tax and spending proposals, would cost anywhere from 3.9 percent less to 0.4 percent more than the conventional cost estimates. This represents only a small margin of difference from the conventional cost estimates. The Administration’s own estimates published in the Mid-Session Review indicate that, even with favorable assumptions, dynamic feedback would pay for less than 10 percent of the cost of making the tax cuts permanent.
The long-run macroeconomic effects of tax cuts can be either positive or negative, depending on how and when they are paid for. The most critical assumption in an economic model is how and when tax cuts are paid for. Generally research has found that tax cuts that are not accompanied by other tax increases or spending cuts — such as the tax cuts that have been enacted in recent years — will increase the deficit, reduce national savings, and ultimately reduce economic growth. In contrast, tax cuts that are paid for contemporaneously can contribute to economic growth. In neither case are the economic effects very large.
The particular models and assumptions used by the Joint Committee on Taxation and the Congressional Budget Office generally find that income tax cuts are more beneficial — or less harmful — in the short run than the long run. The particular models and assumptions generally used by JCT and CBO often show modestly positive effects of income tax cuts in a five- or ten-year period, due in part to initial stimulus effects, but then show these gains being dissipated or even reversed in the long run, as the adverse effects of deficit-financing the tax cuts in the years following their enactment gradually take hold.
You cannot use actual revenue levels from particular years to estimate macroeconomic feedback effects. Some have argued that positive revenue surprises in one or two years are evidence of major dynamic scoring effects. This claim ignores the fact that federal revenues are highly volatile and can swing from year to year for a variety of reasons unrelated to tax cuts. Over the past 25 years, actual revenues in a given year have averaged $150 billion (in today’s terms) above or below the revenue levels that CBO predicted a year in advance.
In contrast, even an optimistic reading of the dynamic effects of the 2003 dividend and capital gains tax cuts would lead to an estimate of less than $5 billion in added revenue, relative to the conventional estimates of the costs of these tax cuts. It thus makes little sense to use data from one or two years to claim extraordinary dynamic scoring effects for tax cuts.
Moreover, in the 1990s, tax revenues exceeded expectations in the years that followed large tax increases. And revenues fell well below expectations in the years of this decade that followed large tax cuts. Differences between the revenue forecasts that were made and the subsequent, actual revenue levels could as easily be used to construct a case that tax cuts have dynamic effects that harm the economy and slow revenue growth as to construct the rosy case that tax-cut proponents make.
Distributional analysis of tax cuts should incorporate the same financing assumptions as dynamic analysis of tax cuts’ costs. All tax cuts must be paid for eventually, since it is not possible to explode the deficit and debt without limit. Macroeconomic models make a variety of assumptions about how tax cuts will be paid for. These same assumptions should be used in presenting distributional analysis of tax cuts.
For example, some analyses that show positive economic effects from the recent tax cuts use the assumption (without highlighting it) that the tax cuts will be fully paid for with across-the-board reductions in benefits and transfer payments, including Social Security and Medicare. Whatever assumptions are made in assessing economic impacts also should be used in assessing the distributional impact that tax cuts would have. Some tax-cut proponent use models that assume the tax cuts are paid for when claiming positive dynamic effects, but then use distributional estimates that are based on the assumption that the tax cuts are not paid for and are essentially a “free lunch.”