CRS RELEASES REPORT ON 'DYNAMIC' REVENUE ESTIMATING.

                     DYNAMIC REVENUE ESTIMATING

The question of incorporating behavioral feedback effects in revenue estimates has attracted increasing attention, in part because budget rules place restrictions on provisions that lose revenue and increase the importance of the revenue estimate in evaluating tax revisions.

The term "dynamic" is often used as the opposite of "static", with the latter presumably meaning that estimates include no behavioral changes. This terminology is confusing because current revenue estimates are not static. The issue is not whether the estimating methods should adopt techniques that account for economic effects, but rather what effects should be considered.

These effects can be classified into microeconomic changes that allow for the reallocation of investment and consumption and changes in the form of income received, and macroeconomic changes that alter the aggregate level of output. Microeconomic changes are incorporated currently, although these estimates are often hampered by the lack of consistent and reliable empirical measures of behavioral responses. Macroeconomic effects are not included.

Macroeconomic effects can be classified as cyclical (effects deriving from short-run stimulation of demand) and permanent. There are strong reasons, on conceptual grounds, for not incorporating cyclical effects. While the state of the economy might be considered in setting deficit policy, including these effects will produce inaccurate overall deficit effects if confined to revenue estimates; they would also apply to spending cuts. Also, the sum of feedback effects for individual proposals will be inaccurate, as it will differ from the feedback effects for the sum of individual proposals. (Since these effects depend on how much unemployment there is in the economy, each successive program that increases the deficit has a smaller effect on economic activity.) Notably, these effects are transitory and could be misleading if permanent tax policy's actions were being evaluated.

While there is a sound conceptual argument that permanent supply responses should be taken into account, there are some important practical considerations. Economic theory suggests the direction of these fundamental supply responses is uncertain because of offsetting income and substitution effects (i.e. an increase in the wage rate can decrease labor supply and an increase in the rate of return can decrease savings). Empirical evidence is mixed although it mostly suggests small responses that can be either positive or negative. For most types of tax revisions it is unlikely that incorporating these effects would alter revenue estimates very much, especially in the short run.

If the objective is to obtain more accurate estimates of the effects of policies on the budget, it is not clear that incorporating permanent responses would achieve that goal, given the uncertainty about the empirical estimates and the fact that the range of estimates generally spans zero. An argument can also be made that a full accounting would require the inclusion of feedback effects from expenditure and regulatory policies as well.

Note

This analysis was originally requested by the Senate Budget Committee Republican staff and is made available for general Congressional use with permission of the requester.

                              CONTENTS

In recent years the question of incorporating behavioral or economic feedback effects in revenue estimates has attracted increasing attention. This argument is often couched in terms of using "dynamic" revenue estimates.

Budget rules, designed to deal with large budget deficits, essentially require that revenue-losing legislation be offset with spending cuts or other revenue increases. The size of the revenue losses associated with tax changes has therefore assumed much greater importance in determining whether a proposal might be adopted.

The term "dynamic" is often used as the opposite of "static," with the latter presumably meaning that estimates are made on the assumption of fixed amounts of taxable income or other tax base. Dynamic estimates are also sometimes referred to as estimates that allow for "feedback" effects. Current revenue estimates are not static, however, because they take into account a variety of behavioral responses. Thus, assessing dynamic revenue estimating is concerned not with whether estimates account for behavioral changes in response to tax changes (they do), but rather with what types of dynamic responses are, and should be, taken into account.

The next section of this paper classifies dynamic effects into the fundamental categories of "Microeconomic" and "Macroeconomic" with the latter, in turn, divided into cyclical (short-run, demand- side or "Keynesian") multiplier effects and permanent supply responses. The third section discusses how current revenue estimates are now adjusted with respect to microeconomic effects and some of the challenges this type of dynamic revenue estimating presents. It also discusses practices among other countries and among the States. The fourth section explains why it is conceptually inappropriate to incorporate cyclical effects into revenue estimates, while it is conceptually appropriate to incorporate permanent supply responses. The fifth section discusses some of the practical issues associated with incorporating these latter effects into estimates. The final section sums up the findings.

CLASSIFICATION OF "DYNAMIC" EFFECTS

Dynamic revenue estimates take into account any changes in the tax base as a result of tax revisions. We can classify these changes into two broad types: microeconomic effects and macroeconomic effects.

Microeconomic effects are those behavioral responses that alter the allocation of consumption or investment, or that affect specific markets, or that change the forms of realization of income in a way that alters taxable income.

Macroeconomic effects are those that alter overall levels of output, employment, and prices. These effects can, in turn, be divided into cyclical (demand-side multiplier) effects and behavioral responses that increase or decrease the productive resources available to produce output (permanent effects). These latter effects occur in a fully employed economy and would include changes in the proportion of national income that is saved or invested or changes in the supply of labor.

MICROECONOMIC EFFECTS

While there is no precise line that distinguishes a micro from a macro effect, in this context, microeconomic changes can affect the allocation of consumption or investment, and the form of receipt of income, but would not affect overall employment, output, interest rates, or prices. These microeconomic changes can be quite significant in influencing revenue effects, however. Another characterization of this category of dynamic effects is that they are normally presumed to be limited enough that revenue estimates would not require further modeling of the interaction of these behavioral changes with the rest of the economy. These effects are also sometimes referred to as partial equilibrium effects, indicating that they do not involve an overall interaction with the economy. ¹

MACROECONOMIC EFFECTS

Macroeconomic effects refer to those dynamic effects that alter the aggregates in the economy: levels of output, employment, interest rates, investment, and prices. There are two basic types of macroeconomic effects, which are quite different in nature. One type is the short-run effect that derives from the stimulus to or contraction of aggregate demand in an economy with underemployed resources. We refer to this type of macroeconomic effect as a cyclical effect. The second effect is the change in employment and other variables that might occur as taxpayers respond to income and price effects of various tax changes. These effects are permanent (as long as the tax change remains in law) and largely involve potential changes in the supply of labor, the level of savings, and the level of investment. They might also be referred to as supply responses.

Cyclical Effects

Cyclical effects result from fiscal stimulus or contraction of underemployed economies. Cyclical effects are associated with any provision that affects aggregate demand, and thus apply to both tax changes and spending changes. In their simplest form, they occur simply because the injection of spending into the economy (via direct government purchases, transfers, or tax cuts) increases aggregate demand, and causes employment of unemployed resources. Investment spending by businesses can also be affected through cash flow effects, and also through reductions in the cost of capital. Any initial increases in spending are then multiplied through successive rounds of spending.

These demand induced effects are the primary reasons that tax cuts (and spending changes) increase output in standard short-run macroeconomic forecasting models (such as DRI/McGraw-Hill or Wharton Economic Forecasting Associates).

These demand-driven models have, on occasion, assumed some importance in the debate over tax provisions. For example, claims were made that the incremental investment credit proposed by the Clinton Administration in 1993 would pay for itself through the economic stimulus it provided to the economy. ² Similarly, arguments were made that the proposed energy taxes would contract the economy and cause the tax increase to yield less revenue than originally projected.

The crucial characteristics that distinguish this type of effect from other macroeconomic behavioral effects are (1) a cyclical effect is, by its nature, a transitory effect that fades eventually, sometimes rather quickly, and (2) the real effect varies depending on the initial level of unemployment and the nature of accompanying monetary and fiscal policy. ³ Note also that there is considerable variation among different models as to the effects of tax, and other changes, on economic output, and considerable debate within the economics community as to the causes of (and corrective actions appropriate to) business cycles and to the validity of the estimates projected by current models. This variation is especially the case when the translation of a fiscal policy change into spending is not straightforward (as in the case of investment subsidies).

A later section of this paper assesses the merits of trying to incorporate these effects into revenue and spending estimates.

Permanent Effects

Another type of response that could affect the macroeconomic aggregates is the permanent response of taxpayers to the price and income effects generated by tax policies. These effects may alter individuals' decisions to participate in the labor market or decisions about hours of work. They may alter individual savings, thereby altering investment and the capital stock. They may also attract or discourage net capital inflows or outflows from abroad into the United States.

Such changes in employment, savings, investment, and output would, if they occurred, set in motion changes that ultimately affect wage rates and interest rates, which in turn feed back to employment and savings decisions, and then in turn to wages rates and interest rates. This process continues until the economy reaches a new equilibrium, and for this reason it requires a "general equilibrium" model to account for all of these effects. ⁴

Supply effects could also occur on the spending side, where public investments in physical (and human) capital might be expected to yield future increases in economic output.

CURRENT DYNAMIC EFFECTS IN REVENUE ESTIMATES

Although current revenue estimates are sometimes identified as "static," this characterization is not accurate. Both the Treasury and the Joint Tax Committee estimates of revenue include a variety of behavioral effects of a microeconomic nature. They do not include macroeconomic effects. The following sections discuss the types of behavioral effects included in these revenue estimates. It also includes a brief discussion of the practices of other countries, and of the States.

PRACTICES OF THE JOINT TAX COMMITTEE AND TREASURY

The practices of the Joint Committee on Taxation (JCT) with respect to revenue estimates are outlined in their 1992 Joint Committee Print entitled Discussion of Revenue Estimation Methodology and Process; these methods are similar to those used by the Treasury's Office of Tax Analysis. ⁵

In many ways, the current revenue estimating system is extremely sophisticated. Many types of estimates (e.g. rate changes, changes in exemptions) are determined by simulation on a large scale micro-data model which involves a sample of the Nation's individual tax returns. Each taxpayer's change in tax liability is calculated given the change in tax rules, and the total added up (and adjusted to national totals). Such models allow interaction with other elements of the tax return. For example, a tax rate cut reduces the value of deductions and may shift taxpayers into the alternative minimum tax.

There are a number of other formal models used (corporate, partnership, estate and gift) and other estimates that must use data outside of the tax return (e.g. a change in depreciation rules requires data on investment spending; a change in excise taxes requires data on the consumption of the taxed commodity).

The Joint Tax Committee uses baselines provided by the Congressional Budget Office (e.g. income, corporate profits, and some specialized series), while the Treasury uses the Administration forecast.

These estimates incorporate microeconomic dynamic effects of many types. The behavioral responses are based on a variety of different data sources and in some cases little data, or conflicting data, are available to provide guidance.

The following are some illustrations of controversies surrounding microeconomic dynamic effects incorporated into revenue estimates, which illustrate the limitations that inevitably apply to any dynamic analysis.

One of the most prominent of these controversies over revenue estimates is that of capital gains realizations. In 1990, President Bush proposed a 30 percent capital gains tax exclusion. There was an extensive debate, in part because the JCT estimated a revenue LOSS of approximately $12 billion over five years, while the Treasury Department estimated a GAIN of $12 billion. These net figures looked very different, but concealed large and similar behavioral responses (changes in realizations as a result of the tax). The Joint Tax Committee found that a behavioral response would have offset 78 percent of the original revenue cost over the first five years, while the Treasury found that a response would have offset 106 percent. Compared to no response at all, these numbers represent similar magnitudes of dynamic feedback. ⁶

At the time the estimates were made, there were a number of empirical estimates in the economics literature that varied widely; there were competing fundamental methodologies for deriving the estimates; and there was considerable criticism of the validity of the models used for these estimates. The estimates were dynamic in both cases and the issue was not whether or not to include the responses, but rather the magnitude of the response.

Another illustration of a behavioral response of the microeconomic type can be found in the case of excise taxes. Two adjustments are normally made for excise taxes. First, demand is reduced due to the price increase, which depends on a measure of price elasticity presumably derived from the data or the economics literature. Secondly, taxable incomes are reduced by the amount of the excise tax, providing for an offsetting reduction in income taxes.

It is important, as in the case of the capital gains realization response, to distinguish between the concept of a dynamic response and the magnitude of such a response. For example, some arguments for using dynamic revenue estimates have referred to the case of the luxury boat tax imposed in 1991, where sales declined. (The luxury boat tax was imposed along with a number of other luxury taxes in 1990 and was an small share of the expected yield of the taxes which were, themselves, a small source of revenue).

Even setting aside the small size of this tax, it is important to recognize that the revenue estimators did not have any reason to predict a large response to the tax before the fact, and assessments after the fact suggested that the decline in that market was likely due to reasons other than the tax change. (One of the reasons for taking this view is that sales had already begun to decline, including a substantial drop in 1990, before the tax was imposed.) That decline may have reflected the recession or the cyclical nature of the industry, but there would have been nothing in the data at that time (or in economic theory) that would have pointed in the direction of a large decline, and indeed nothing in the data since to clearly point to a large tax induced effect. ⁷

These examples illustrate an important point: even with the current microeconomic feedback effects, revenue estimates of certain types are subject to considerable uncertainty that is inevitable given the limitations of empirical economic research. This degree of uncertainty varies from one revenue estimate to another. Even though there is already general agreement on the need to incorporate microeconomic behavioral effects, uncertainties -- and disagreements -- are likely to persist because empirical evidence is not always available, reliable, and consistent.

PRACTICES IN OTHER COUNTRIES

Information on the relevant practices of other countries on an up-to-date basis is not consistently published, but a 1988 survey by the OECD of 18 countries provided some insight into the degree of dynamic estimating used in other countries. ⁸

At that time, 14 countries, including the United States, used a formal micro-data model. Four countries at that time (Japan, Ireland, Portugal, and Turkey) did not use a formal model to make revenue estimates.

This survey identified two kinds of behavioral effects: the "tax minimization" process, and the response of taxpayers to changing relative prices. Tax minimization is the process of recomputing taxpayers' taxes, when options are available to them (for example, whether or not to itemize deductions), to give them the minimum tax bill. In most cases, not even the tax minimization process was incorporated into the models.

In general, this study states that other countries' revenue estimating models are "typically of a 'static' nature in the sense that they do not take account of behavioral reactions to changes in the tax systems and they do not encompass the 'multiplier' effects of changes in taxes (called feedback and spillover effects . . .)" ⁹

The study singles out the United States as the country that occasionally incorporates responses to relative prices (examples given were changes in charitable donations in response to tax rate changes and substitution of mortgage debt for non-mortgage debt when consumer interest deductions were disallowed in 1986). In our taxonomy, these changes would be characterized as microeconomic responses.

This study refers to the model used by the U.S. Treasury but notes that a similar version of the model is used by the Joint Committee on Taxation of the U.S. Congress.

The basic impression one would gather from this study is that the U.S. was recently farther along than other countries in dynamic revenue estimating; indeed, dynamic estimating was scarcely practiced in other countries by 1988.

PRACTICES IN THE STATES

The California legislature recently passed a law that required all revenue changes of a certain magnitude to consider economic dynamic feedback effects. It is not yet clear how estimators in California will comply with this mandate or exactly what meaning the mandate has.

Two individuals at California's Franchise Tax Board prepared a report on the subject of feedback effects prior to the enactment of this law. ¹⁰ They did an extensive telephone survey of the States that revealed that only one State, Massachusetts, regularly uses a dynamic model to estimate revenues. This model has been in operation only a short time.

The authors suggested that only a half-a-dozen States reported any attempt to estimate the effects of tax proposals, although they noted that there have been some attempts at feedback analysis in special cases. The California Franchise Tax Board's estimates do occasionally include certain behavioral effects in much the same manner as the JCT or Treasury.

In a study reporting effects in the Massachusetts model, the feedback effects offset about 7 percent of revenues for income taxes, about 5 percent for sales taxes, and about 30 percent for corporate taxes.

It is important to note that macroeconomic behavioral effects are actually potentially more important for the State than for the United States because labor and capital can more easily move across State borders than across national borders. In this sense, the supplies of capital and labor are potentially much more elastic to any given State than to the U.S. as a whole. ¹¹ Even so, there is considerable skepticism about the importance of tax policies in influencing the location of workers and businesses. ¹² This kind of analysis can present some disturbing problems if carried out by many States. If the supply of labor and capital in the U.S. is relatively fixed, then any gain in taxable income and sales in one State is largely a loss to another. If all States engaged in this type of feedback estimation for tax cuts, they would understate their revenue losses because each State would not take account of the effects of all other States' tax policies on themselves.

The authors outline a series of practical problems in developing a dynamic model including limits on data, on empirical estimates, and disputes about theoretical effects. They discuss the Massachusetts model in detail, but ultimately conclude that "there is no model currently available or that is likely to be available in the near future, to reliably estimate the feedback effects of tax law changes in California." ¹³

INCORPORATING MACROECONOMIC EFFECTS: CONCEPTUAL ISSUES

What sort of macroeconomic effects should, in theory, be incorporated into revenue estimates? To answer this question, one must begin with the basic purposes of a revenue estimate for an individual tax proposal. There are typically two objectives: a revenue estimate is combined with other spending and tax proposals to determine the overall fiscal position and the revenue estimate provides information on resource allocation.

ASSESSING THE FISCAL POSITION

Estimates of revenue gains or losses must be aggregated with all other estimates (for both tax and spending programs) to determine the overall effect of this package of decisions on aggregate receipts and expenditures, and, in particular, the budget deficit or surplus. The budget deficit or surplus defines, in part, our aggregate fiscal policy both for purposes of short run economic effects (stabilization policy) and long run fiscal stance. Currently, there are a number of rules and regulations that govern how this measure is to be set, and there is a budget resolution that determines the overall level of spending, taxes, and deficit that is allowed.

It is important to stress that this budget deficit or surplus is not a complete accounting of the nation's fiscal position for a variety of reasons. ¹⁴ We are concerned about the level of government involvement as well as the netting out of receipts or expenditures. Moreover, many laws that are considered, both tax and spending, carry commitments and consequences that extend far beyond the current budget period and could affect both the short run and the long run fiscal stance. ¹⁵ For example, many tax proposals, either by accident or design, may cause much different longer-run effects on revenues than they do in the budget period. ¹⁶

Nevertheless, this measure of the budget deficit or surplus is the initial step in gathering the data to estimate the fiscal impact on the short-term course of the economy, and to determine how the outstanding debt will change. It is the policy target that is set with the budget resolution. With this measure in hand, or with the various parts of the measure, general macroeconomic analysis can then be applied to project the expected effects on output, employment, prices, and interest rates arising from demand side multipliers, as well as the final (post-feedback) real deficit that will be added to the national debt. ¹⁷ Indeed, an Administration, when presenting its budget, may base its budget on this post-feedback economic forecast.

It should be clear from this analysis, therefore, that while microeconomic behavioral adjustments are appropriate to make in computing individual revenue estimates, it is not appropriate to make adjustments for demand-side multipliers. If revenue estimates are assigned feedback effects that reduce the cost of tax cuts, offsetting tax increases or spending reductions must also be reduced to reflect these feedback effects. Otherwise, there would be an incorrect measure of the post-feedback deficit. To the extent that a given tax reduction is to be offset with a spending cut, this exercise would not change the offset requirements under current budget rules such as pay-as-you-go. ¹⁸

While one might imagine incorporating such a feedback effect for each individual item (spending and revenue), the resulting post- feedback deficit would be incorrect. The sum of the effects (post- feedback) of two separate provisions is not the same as the effects of summing them and then calculating the effects after feedback, because the magnitude of the multipliers depends on how close the economy is to full employment. Consider two tax reductions of equal size, where either, by themselves, would lead to full employment. The true feedback effect for the provisions combined would be only half the size of the summed feedback effects for the provisions estimated independently.

For these reasons, it is not conceptually appropriate to allow dynamic adjustments for cyclical effects. Rather, the existence of these multipliers (along with positions on the long run fiscal position) should be taken into account in setting the initial budget policy in the budget resolution.

What about the permanent macroeconomic supply responses? Here, there is a conceptual case for including behavioral effects, assuming the macroeconomic model used does not independently do so. ¹⁹ If a tax provision is expected to alter the labor supply or the savings rate, an argument can be made for adjusting the revenue estimate to reflect that effect. Even in this case, however, it is better to estimate the effects of a full package of all legislation rather than providing for each independent estimate. This issue, however, becomes more an issue of application rather than concept, and is discussed in the following section.

ALLOCATIONAL EFFECTS

A second reason for providing revenue estimates is that costing out either a spending or, in certain cases, a tax proposal tells us what magnitude of resources are to be devoted to the objective. Such a costing measure cannot itself determine the desirability of a given proposal; rather it is one piece of information about the proposal. There are really no hard and fast rules about how such a cost should be measured because that need would be governed by the analytical framework.

For example, in the case of an excise tax, we may find a static estimate more informative than one with microeconomic effects incorporated, or we may wish to know both. For example, if we were to increase an excise tax, it is possible that demand could contract enough that no revenue would be raised. If we wished to increase the tax to discourage a certain activity, this no revenue effect would not be very informative, and we might wish to know the tax burden before behavioral response is incorporated. Nor would we necessarily view a lack of revenue an indication of undesirability; rather, it might signal a successful policy. If our tax is for the purpose of raising revenue, however, we would certainly wish to know that no net revenues would be gained from this proposal, but might also wish to know the static effect in order to understand the implicit economic burden placed on taxpayers.

In general, for these types of allocational issues, there is no reason to incorporate cyclical effects, since our concern is the resources devoted to this purpose, relative to other alternatives. That is, we want to compare one use of resources relative to another while holding fiscal policy constant. Another reason for not incorporating cyclical effects is that, in most cases, these provisions are likely to be permanent, while cyclical effects are transitory.

In evaluating proposals, we may also be interested in permanent macroeconomic effects, depending on the type of proposal under consideration (i.e., for a broad provision aimed at reducing taxes on capital, the savings effects may be central to the analysis, while for a narrowly targeted provision these effects may be less important).

INCORPORATING MACROECONOMIC EFFECTS: PRACTICAL ISSUES

As with microeconomic effects, the consequences of incorporating macroeconomic effects deriving from supply responses would depend on the magnitude of response. There are three major categories of response: increases in labor supply, increases in domestic savings, and increases in investment due to net increases in capital flows from abroad. We discuss the first two categories in concert as they need to be considered jointly in a modeling exercise. Also discussed is another potential source of growth, greater productivity from a given set of inputs through increased efficiency and innovation.

LABOR AND SAVINGS RESPONSES

Theoretical Considerations

Increased output can arise from an increase in the inputs into the productive process, i.e., labor and capital, the latter occurring gradually through an increase in the savings rate.

Although it may seem initially counter-intuitive, an increase in the wage rate (which would result from a tax decrease) could potentially either increase or decrease the labor supply. Similarly, an increase in the rate of return on savings could either increase or decrease savings. Empirical evidence does not clearly resolve these issues.

Labor Supply

Economists have long recognized that the response of work effort to a change in after-tax wage rates can be either positive or negative, reflecting the opposing forces of "income" and "substitution" effects. When the wage rate falls, leisure becomes less costly in terms of forgone wages, and the individual might wish to increase leisure by reducing the amount of time worked. This is the "substitution" effect, and it derives from the willingness to substitute leisure for consumption (the latter achieved by earning money). At the same time, the lower income makes individuals consume less of everything -- both consumption and leisure. Since the amount of leisure rises through one effect and falls with another, the direction of the effect is uncertain. Or, put another way, an individual with less after-tax income might try to earn more to make up for some of the loss and an individual with more income might find that he can work less and still satisfy his tastes for both consumption and leisure.

For considering tax policy changes, it is important to separate income and substitution effects, because, in a graduated rate system, tax changes do not reduce average income in the same proportion as marginal income (income from working an additional increment). For example, a proportional decrease in income tax rates will have slightly smaller percentage effects on average income than on marginal income because the average tax rate falls below the marginal tax rate. Adding a top rate, as was done in 1993, would have a much more pronounced substitution effects relative to an income effect. An increase in the personal exemption or a dependent credit, however, will have no marginal effects for most individuals, but will increase income, thereby producing only income effects that tend to reduce labor supply.

If an overall negligible effect on labor supply from a wage change is due to offsetting small income and substitution effects, then the behavioral effects will be small for any tax change. If an overall negligible effect on labor supply is the result of offsetting a large income effect with a large substitution effect, tax policy changes could still have a large effect on labor supply even if wage changes would not do so and those effects will depend strongly on the type of tax change envisioned.

Savings

As in the case of labor supply, the response of the savings rate to a change in the rate of return (which would result from a tax decrease) is theoretically uncertain. When the rate of return rises, a substitution effect might cause an individual to prefer more consumption in the future (because the price of future consumption has fallen in terms of foregone present consumption) and increase savings. At the same time, there is an income effect -- the higher rate of return can allow savings to be smaller and still increase consumption in the future (and in the present as well). (For example, if an individual were saving a certain amount for retirement, he could obtain that objective with a smaller amount of savings when the rate of return goes up).

It is also important to recognize that the process of altering the capital stock through a change in the savings rate is a very slow process that takes many years. A savings response and its translation into taxable income is a much less likely way than a labor supply response of recouping revenue losses in the short run.

Empirical Evidence

Overall, the statistical evidence suggests that the labor supply response is probably small and perhaps even negative. Most studies suggest that this is the result of offsetting small income and substitution effects (although this issue is by no means resolved). Moreover, the labor supply response is more likely to be small in the short run, relevant to the budget horizon, because adjustment processes take time. The evidence is discussed in more detail in Appendix A.

Note also that, even if we were to know the size of these income and substitution effects, to properly model them would require a separate income and substitution effect for each taxpayer in the sample, a calculation of the effects on average and marginal incomes or wages for each type of tax change, and a merger of the micro- simulation model with a computable general equilibrium model to account for feedback effects. Such a model does not presently exist and would probably be costly and difficult to develop, especially to model effects over several years. In the meantime, any current adjustments in the next few years would have to be done on a case by case basis, and would absorb a lot of resources. The incorporation of these effects would be particularly difficult given the time pressures under which revenue estimators often work and the volume of work. For example, the Joint Tax Committee has indicated that it prepared 2,380 revenue estimates in 1993.

The empirical literature on savings is much more limited than in the case of labor supply and in many ways the empirical research faces greater challenges in the case of the savings response. This literature is also discussed in Appendix A. It is more certain, however, that these effects will be very small in the short run, because of the adjustment process (as discussed below).

On the whole, the evidence on labor supply and savings does not appear to suggest a strong response to tax changes or even a necessarily positive one. Thus, one could argue that the current approach of leaving income fixed is, in fact, consistent with the evidence.

How Feedbacks Work

The calculation of feedback effects seems complicated, and indeed to obtain precision in an economy with many products, many individuals with different incomes, and a complicated tax code (even if all the behavioral parameters were known) would be an extremely complicated task.

One can obtain the general magnitude of effect, however, from a "back-of-the-envelope" calculation assuming the composite elasticities. The following illustrates some calculations of these types, first for a tax that applies solely to labor income, then for a tax on capital income. Both short-run and long-run effects are considered.

Labor Tax

In illustrating the effects of a tax on labor, first consider the expected magnitude of the labor supply response. (As shown in the appendix, the response to a labor tax is not affected by the savings response in either the short or long run, although the response to a capital tax is affected by the labor supply response). Ballard, et al., in constructing their large scale general equilibrium model, used a weighted labor supply elasticity of 0.15. ²⁰ That is, a one percent increase in the wage would increase the supply of labor by 0.15 percent. We will use that measure for illustration, although it is worth noting that more recent evidence suggests a lower, and possibly negative, elasticity.

A PARTIAL EQUILIBRIUM MEASURE. Consider a small change in an existing tax rate T, which we denote as dT. Our initial revenue loss is WLdT, where W is the wage rate and L is the labor supply. To calculate the feedback effect, first consider what would be called a "partial equilibrium" calculation, where we calculate the effect on labor supply, while holding other parts of the economy constant. The offsetting feedback effect is TWdL, that is, the tax rate times the wage rate, times the change in labor supply. That labor supply change is, in turn, determined by the percentage change in the after tax wage, which (to account for existing taxes) is W(1-T), for a percentage of dT/(1-T). If the elasticity, which we denote as Es, is the percentage change in labor supply divided by a percentage change in wage, then we can define the change in labor supply:

dL = EsL dT/(1-T)

and the revenue feedback:

Revenue Feedback = EsTWLdT/(1-T)

and finally the revenue feedback as a fraction of the original tax reduction (dividing by WLdT):

Revenue Feedback Fraction = EsT/(1-T).

As an illustration, consider the case where the tax rate is 20 percent and the elasticity is 0.15. In that case the revenue feedback fraction is 0.15 times 0.2 divided by 0.8, or 0.038. This amount is 3.8 percent of the initial cost. If our elasticity estimates are correct, this finding indicates that for every $100 of static revenue loss, there is an actual loss to the Treasury of $96.20 after feedback.

We can also see the effect of sensitivity analysis. If the value of the elasticity is zero, the feedback effect will be zero. If the elasticity is twice as large, the revenue feedback will be 7.5 percent. If the elasticity is negative, the effect will be to make the static revenue loss 3.8 percent LARGER (that is, for every $100 of static loss, there is a loss of $103.80 after feedback). None of these assumptions are outside the range of empirical estimation.

Note also that the value will also depend on the type and size of tax change and the initial tax level. The feedback revenue as a share of static cost, positive or negative, will be larger in magnitude if the initial size of the tax is larger. Note also that this formula only holds for small changes. For large changes, the effects will be smaller in magnitude because the response is likely to be smaller proportionally and because the induced change will be taxed at a smaller rate. (Recall also that even if we actually knew the elasticity appropriate for a proportional change, it would not be appropriate for many of the actual tax changes which would likely not have proportional income and substitution effects; that is, the elasticity would have to be varied for each type of change).

Things are actually slightly more complicated than this simple calculation, because such a calculation needs to be general equilibrium, that is, to allow wage rates, rates of return, and, in the long run, the capital supply, to change.

A SHORT-RUN GENERAL EQUILIBRIUM MEASURE. Consider first the short run, where the capital stock can be considered to be effectively fixed. An expansion in the labor supply would drive down the wage rate, and offset the effect of the increased labor supply on wage income in the economy and, of course, feed back into the labor supply, which would in turn affect the wage, and so forth. It would also affect the rate of return to capital. To calculate these effects we would also need another important behavioral measurement, the factor substitution elasticity. This measure indicates ability to substitute labor and capital in production (and is defined as the percentage change in the ratio of capital to labor divided by the percentage change in the ratio of the prices of capital and labor). We also must be concerned about the presence of taxes on capital income, since capital income will also change. ²¹

Appendix B presents a simple model for calculating these effects, but the basic effect of this general equilibrium calculation is to make the feedback effect smaller, depending, of course, on the value of the factor substitution elasticity. This effect is smaller because increased (or decreased) labor supply lowers (or raises) the real wage and chokes off some of the potential increase (or decrease) in labor supply. This ability to substitute labor and capital is thought to be relatively small in the short run when production technology cannot be altered, perhaps in the neighborhood of 0.2 (in absolute value). In that case the feedback percentage in our initial example is reduced slightly, to 3.2 percent assuming that taxes on capital are imposed at the same rate as those on labor. How much the effect is reduced depends on the relative size of the elasticities and whether the labor supply elasticity is positive or negative. A small factor substitution elasticity will constrain a large labor supply elasticity from producing a large feedback effect, but will have a much more limited effect on a small labor supply elasticity.

A LONG-RUN GENERAL EQUILIBRIUM MEASURE. In the long run, the size of the capital stock can also change as the income in the economy grows. Note that it will take, however, many years before this long run effect of a labor supply response is approximated. Because the additional income leads to additional savings, the effect of a positive labor supply response is larger. Or if the labor supply response to a tax cut is negative, the ultimate negative effect (which increases the static revenue cost) is multiplied as well. In fact, in the case of a labor tax, in the long run, the capital stock increases or decreases proportionally with the labor supply (relative to what it would otherwise have been) and the wage and rates of return are not altered. The calculation of feedback effects in our basic example is to increase the magnitude of the revenue feedback percentage by about a third -- in our example from 3.2 percent to 5 percent -- because capital income is about a third of labor income.

These calculations suggest that any feedback effects, assuming a relatively small labor supply elasticity, will be quite small even in the long run.

CAPITAL TAX

One computes a feedback for the capital tax in a fashion similar to the labor response, except that the time frame becomes much more important and the feedback effects are much smaller in the short run. If a tax changes results in an increase in savings, that savings is not taxed; rather the rate of return on the increased savings is taxed. This amount can be a very small number compared to the rate of return on the much larger existing capital stock which forms the existing tax base.

In the short run (say the first year), any savings elasticity should be multiplied by the ratio of the savings rate to the capital stock to generate the capital stock elasticity needed to calculate the feedback effect. That ratio is normally the steady state growth rate of the economy.

Ballard, et al., in constructing their large scale general equilibrium model used a savings elasticity of 0.4. ²² This estimate was reported in a study by Michael Boskin some years ago and would probably be viewed at the higher end of the currently available time series estimates. ²³

A SHORT-RUN MEASURE. In the short run, the effects would be extremely small. The partial equilibrium measure would be:

Revenue Feedback Fraction = gErTr/1(1-Tr).

where g is the growth rate of the economy, Er is the savings elasticity, and Tr is the tax rate on capital. Setting the growth rate to 0.03, the elasticity to 0.4, and the tax rate to 0.2 produces a revenue feedback fraction of 0.003, which is 0.3 percent of the initial cost. That is, the revenue cost with feedback would still be 99.7 percent of the static revenue cost of a capital tax cut. (A general equilibrium calculation would be of similar magnitude but slightly smaller).

This revenue feedback percentage would grow rapidly, slightly more than doubling in the first year, and more than tripling in the third, and so forth. Nevertheless, because it is so small to begin with it would take a number of years before it began to approach the short run labor supply response.

If the savings elasticity is zero, there will be no effect, and this measure is within the range of empirically estimated elasticities.

LONG-RUN EFFECTS. In the long run, after many years of capital accumulation, the feedback effects would be much larger, because the capital stock would expand in line with the elasticities. (We are talking in this case of many years, perhaps 30 to 50, before we reach a close approximation of the long run.) ²⁴ The appendix provides a formula for calculating this long run effect in a simple one-good model. For the savings elasticity of 0.4 and assuming a labor supply response of 0.15 and a long run factor substitution elasticity of 1.0, the feedback effect is calculated at 11 percent.

These calculations suggest that even in long run the feedback effects will be far from making up for the revenue loss and that a deficit financed tax cut for capital would likely cause a contraction of the capital stock.

It is also important to note that one cannot simply take the long-run effects and use them as a measure of permanent effects in assessing a capital tax cut. For example, if only 89 percent of the revenue loss from a capital income tax cut were offset in the long run by other budgetary changes, there would still be deficits created during the many years it takes to approximate the long run equilibrium and these deficits would in turn reduce the capital stock and otherwise reduce taxes. To determine what policy would be required to maintain revenue neutrality, the entire path of the adjustment would have to be traced.

A second caveat that should be kept in mind is that these calculations are only for small tax changes. In the extreme, if we cut all taxes to zero, there would be no feedback effects on revenues because there would be no tax to recover revenues. More generally, a large change would cause a smaller feedback effect because the tax rate that applied to the induced effects would be smaller.

It is important to note that this analysis does not take into account a number of other issues. For example, it begins with a static revenue cost that is relatively consistent over time compared to capital income. Proposals greatly differ in the timing of their revenue costs, and it is possible to design capital income cuts that have no revenue cost in the short run and even gain revenue, even though they lose revenue in the long run. This timing pattern will not change any of the fundamentals, in present value terms, but will alter the time path. For example, if the revenue loss is pushed to the future, feedback effects in the long run will be smaller relative to the revenue cost at that time.

Finally, economic theory indicates that the effect on savings depends on the type of capital income tax cut. Some types of tax cuts provide smaller additions to the existing capital stock, relative to revenue losses, than do others and that could affect the feedback measures because it would affect the savings response and the initial magnitude of revenue cost. ²⁵

HOW MODELS CAN PRODUCE LARGE EFFECTS. The question might arise as to how some models can produce large offsetting revenue effects. Sufficiently high elasticities can produce such a result. In fact, it is easy to show, using the formulas in the appendix, how an effect could be derived. The revenue feedback fraction in the long run for a capital tax, assuming wage income is taxed at the same rate as capital income is:

-Er (Es+S)/[(Er,+S)(1-a)] times T/(1-T)

where S is the factor substitution elasticity and a is the capital income share. In words, the savings elasticity, times the sum of the labor supply elasticity and the factor substitution elasticity, is divided by the sum of the savings elasticity and the factor substitution elasticity, and divided again by the labor share of income. The result would be multiplied by T/(1-T). Using the same parameters as before (T = 0.2 and setting a = to 0.25), this formula yields a value of 0.11 (11 percent with the parameter values used above).

Suppose, however, the savings elasticity were infinitely large, which occurs in models that use what appears to be an innocuous assumption of a fixed after-tax rate of return. In that case the revenue feedback fraction would be:

-(ES+S)/(1-a) times T/(1-T)

Some simple calculations show that you would derive a 100 percent feedback effect if the sum of the labor supply elasticity and the factor substitution were 3. If the initial tax rate were set higher, say at 0.4, then the sum of the factor substitution elasticity and the labor supply elasticity would only have to be only around 1.12 to obtain a result that feedback effects will make up revenue.

But assuming an infinite savings elasticity (or a fixed after tax rate of return) is not reasonable. It implies that a small tax cut can produce an enormous increase in the capital stock, an outcome that is not consistent with the evidence we currently have. ²⁶ This assumption is not an innocuous one, but rather is one that drives the overall effect.

INTERNATIONAL CAPITAL FLOWS

The responsiveness to taxes of the supply of capital from abroad is probably greater than the domestic savings elasticity, and might thus appear to represent an additional source of behavioral response, perhaps an important one. Nevertheless, there are some reasons to doubt the importance of international capital flows in causing significant feedbacks, especially in the short run.

Initially, it is important to note that only some types of tax changes are likely to influence international capital flows -- those that are associated with investment. Many Federal taxes that apply only to domestic and not to foreign savers (e.g. taxes on capital gains, interest, and dividends) would not affect the location of investment. Indeed, a savings increase (if any) from a tax cut for U.S. savers may partly be exported.

In addition, the evidence on the degree of mobility of capital across international borders is not clear, and at least some studies have found little evidence of significant capital mobility. ²⁷ (Although changes in capital flows into the U.S. in the 1980s may seem significant relative to previous years, they are of modest size compared to the capital stock overall or even overall savings). Mobility of capital may be constrained because investors do not see investments in different countries as perfect substitutes. In addition, since the investment flows, as well as reflows of return, require changes in the trade flows, imperfect substitutability of imports for domestically produced goods can limit the ability of capital to flow across international boundaries.

The nature of the adjustment path is not clear either, although the effects of capital flows would be constrained in the short run because existing capital is invested in physical form (i.e. buildings and equipment) which cannot be relocated.

Finally, the revenue feedback to the Treasury of any injection of foreign capital in the U.S. is somewhat limited because debt- financed capital is subject to a negative tax rate. Investment subsidies, such as accelerated depreciation, are likely to attract both debt and equity capital, and debt capital may be more mobile. Increasing taxable income that is subject to a subsidy, rather than a tax, can magnify any revenue loss. This effect is mitigated by the likelihood that some of this imported capital will enhance labor income and the fact that returns on debt-financed capital tend to be lower than the marginal product of capital, due to risk premiums.

A proposal that might be more likely to attract capital income that would enhance Treasury revenues is a reduction in the corporate tax.

In general, it is also important to note that any tax induced effect on international capital flows will also be limited by the importance of net capital inflows as a part of the United States capital stock; currently that capital stock inflow is only a small fraction of the U.S. total.

Although this source of feedback should be explored more fully, it seems unlikely that it would be very important, particularly in the short run budget horizon.

EFFICIENCY AND INNOVATION

A final source of behavioral response is the possibility that tax changes might cause given inputs of labor and capital can be used more efficiently to enhance productivity. There are two aspects of this potential effect.

First, if a tax change causes resources to be allocated more efficiently, there will be an efficiency gain. This efficiency gain is, however, unlikely to be very important, in part because virtually all such efficiency gains tend to be very small relative to output. In addition, many of these effects do not show up in increased output, but rather in a change in the mix of output that satisfies tastes. In general, this source of budgetary effects can probably be dismissed as unimportant for virtually all changes. In addition, some changes can actually magnify distortions, or have an unclear directional effect.

Another source of gain is an increase in technology or innovation. Clearly technical change is an important source of economic growth, and arguments are sometimes advanced that certain types of tax changes (e.g. capital gains) would encourage innovative activity.

The difficulty with this argument is that there is little evidence in the data to suggest a link between tax policy and the rate of innovation, or even between capital accumulation and the rate of innovation. For many years, this "residual" source of growth appeared to be constant despite significant changes in a variety of policies. A slowdown in productivity growth that began in the seventies has not been easily explained, but seems unrelated to tax policies or to capital accumulation rates.

In the case of capital gains taxes, there are additional reasons to question whether there is a link between this tax policy and technical innovation. First, most of the formal venture capital equity financing sources are supplied by sources not subject to the individual capital gains tax (such as tax exempt pension funds). ²⁸ Secondly, one type of capital gains relief often proposed is indexing the basis for inflation. This provision is unlikely to have a significant effect on successful venture capital operations because when risk-taking succeeds and the gain is very large, the basis, which would be adjusted for inflation, is very low relative to asset values.

SUMMARY

This assessment of the practical considerations in incorporating feedback effects suggests that there are significant problems. The most serious of these problems is probably the uncertainty in the empirical evidence on the magnitude of behavioral response. In the case of the fundamental labor supply and savings responses, this uncertainty extends not only to the magnitude of effects, but also the direction of effects. Indeed, the range of empirical estimates for these effects typically spans a zero response. This uncertainty is magnified because the size and direction of effect is likely to vary for specific types of tax changes and for different individuals.

Despite these problems it appears that, in most cases, and using a reasonable set of elasticities from the literature, such effects usually would not significantly change the magnitude of a revenue estimate in either direction. This effect is especially true for changes in the capital income tax in the short run that generally encompasses the budget horizon. That is, one could argue that the current practice of incorporating no permanent macroeconomic behavioral response is, in fact, reasonably consistent with the body of empirical evidence.

POLICY IMPLICATIONS

Currently, microeconomic behavioral effects are incorporated into some revenue estimates, on a case by case basis, while macroeconomic aggregates are held constant. The fundamental policy question, on which this paper seeks to shed some light, is whether there should be an expansion of the types of behavioral feedbacks taken into account in the budget process. In this regard, it is appropriate to begin by the clear directions pointed to in this study, and then to turn to the gray areas.

This analysis suggests that cyclical effects arising from an underemployed economy should not be separately incorporated into individual revenue estimates, both because to do so would lead to inaccurate results and because to do so serves neither purpose of revenue estimating (measuring fiscal stance or considering resource allocation issues).

The analysis also suggests that there is a conceptual case for including permanent macroeconomic effects (as well as microeconomic effects already included), but raises a number of questions about the practical ability to do so, due largely to the existence of uncertainty about empirical measures needed to incorporate these effects. It is the tension between those two points that leads to the uncertainty of a conclusion.

Thinking about this issue might be most readily organized by first stating the case for making that change, and then stating the case for not doing so.

The fundamental argument for including macroeconomic effects (and perhaps expanding microeconomic effects) is that this inclusion would lead to a more accurate assessment of the budgetary consequences of proposals. There are some specific types of tax revisions where these effects could be potentially quite large. An example that some might cite is the addition of another tax rate in 1993 which, at least some economists strongly felt, would lead to a significant contraction in the static revenue gain. (This view was not universally held, however, and there was considerable latitude to include microeconomic effects, such as compliance, under current practices.) Did this failure to take into account all feedback effects cause us to fail to attain our budgetary objective?

The fundamental argument for maintaining the status quo is actually very much the same argument -- that owing the lack of information on many of the issues and to the technical demands (as well as the presence of political pressures), incorporating macroeconomic effects may lead to less accurate estimates.

To develop this argument further, one might note that conceptually correct incorporation of these effects requires a full accounting of potential feedback effects. That is, effects should be considered for both large and small proposals (if considered only for large proposals it would bias our choice in the design of proposals). ²⁹ No models currently exist to estimate these effects quickly and they would, therefore, have to be constructed for each proposal until and if such a model could be developed, a task that would absorb considerable time and resources but, in many cases, might result in very little modification of estimates.

Moreover, the review of the literature on fundamental labor and savings responses indicates some variation in these empirical findings, particularly for certain types of tax changes. There is no objective standard that can be applied to choosing an elasticity from the literature, and such a choice will necessarily remain in part a subjective one. There is, perhaps, a legitimate reason to be concerned about who will make that choice, and how the choice will be made.

In addition, one could argue that such feedback effects should also be applied to spending tax policies (i.e. public investments that increase productivity and transfer programs that alter incentives) and to regulatory policies of the government that alter output and, therefore, revenues. These policies and their feedback effects have their own associated uncertainties.

While the exploration of the economic effects of tax changes, as well as spending and regulatory policies continues to be important to public policy analysis, the net gain from incorporating these effects into the formal estimating process seems in doubt, largely because of practical inadequacies of the estimating process. Certainly, the continued research into these issues and open discussion of the effects of permanent feedback effects would help to inform the debate on the merits of various proposals.

APPENDIX A: EMPIRICAL EVIDENCE ON LABOR SUPPLY AND SAVINGS

Overall, the statistical evidence suggests that the labor supply response to a change in net wage is probably small and perhaps even negative. For males, most studies find a negative labor supply elasticity that is small. ³⁰ (These studies tend, on the whole, to focus on middle income individuals, and may not be representative of incomes closer to the upper and lower ends of the income distribution). That is, increases in the net wage result in smaller amounts of work. Most studies indicate that this outcome is the result of small income effects and small substitution effects (rather than large ones). Some newer studies that found large offsetting income and substitution effects, and that have been regarded by some economists as superior studies, have recently been criticized on technical grounds. ³¹ Studies of female response are mixed and span a wide range, from small negative responses to relatively large positive ones (i.e. a rise in net wage rates would increase work). ³² The labor supply response for married women is particularly difficult to estimate because many married women do not participate in the labor force and there is no way to observe their wage directly. Indeed, a recent study suggests that errors in statistical techniques may be responsible for the large responses found in many previous studies, and that labor supply responses for married women may be more in line in male supply responses. ³³

In addition to the evidence from studies of the response to wages, there are other types of studies that may have implications for work effort. There were some studies of lower income individuals that focused on lower income men as a result of an experiment with wage subsidies. These studies also found effects of small magnitude, although they were more likely to produce a slight positive response. ³⁴

Some studies have tried to examine how total incomes changed after specific tax reductions in 1981 and 1986, sometimes concentrating on higher income taxpayers. ³⁵ While these studies suggest increases in incomes that respond to reductions in taxes, such studies of particular events encounter the problems that they cannot easily control for other influences on reported income, cannot always control for other structural tax revisions, often do not disaggregate effects by type of income -- in order to tie the appropriate income source to the specific kind of tax change --, and cannot separate real labor supply responses from changes in the form of compensation. ³⁶

For example, one problem plaguing such studies is that there has been a trend over time of increasing inequality in incomes that economists cannot explain and that is difficult to control for. There are other problems as well. A study of the 1981 tax cut that showed a significant increase in taxable income was apparently mostly due to increased capital gains realizations, and not wage income. Studies of the 1986 tax revision may not control for all of the tax changes in that legislation that broadened the tax base.

Actually, most of these studies did not appear to show a strong increase in wage income, although it is difficult to determine the effect when income responses are not disaggregated by type of income.

In addition to evidence suggesting that the labor supply response may be small, it is likely to be even smaller in the short run. First, it is likely that there will be a time lag before individuals might arrange to change their hours of work or their participation in the labor market. Furthermore, in the short run, the capital stock is more or less fixed and businesses cannot easily adjust the numbers of workers quickly. An attempt by workers to supply more labor under these circumstances may largely result in a falling wage, which will then drive workers out of the market again. These points may not be crucial to evaluating the merits of a proposal that alters the labor supply, but will affect revenue estimates in the budget horizon.

It is, however, important to recognize that there is a considerable amount of controversy about certain aspects of these estimates, including the degree to which small labor responses reflect the trading off of small or large income and substitution effects and the validity of studies that showed a large response for secondary workers. We also have little evidence that will inform us about the behavior of high income individuals' labor supplies, if these are different from the response of lower income individuals. The debates about these issues tend to revolve around complex technical issues of economic theory and econometrics which cannot be easily explained and on which legitimate differences in view point may persist.

The empirical research is much more limited in the case of savings than in the case of labor supply and in many ways faces greater challenges. A number of studies have examined the relationship between savings rates over time and rates of return. Overall, however, the empirical research also suggests relatively small effects, of uncertain sign (i.e., some evidence suggests that the effects are negative). ³⁷ Moreover, many of these studies do not incorporate data from the eighties; if they did it is more likely that a negative response or a smaller positive response would be found.

There have also been some specialized studies on the effects of pensions and other deferred compensation plans, which are tax favored, on savings. Although some studies have claimed, for example, significant savings responses from individual retirement accounts (IRAs), these studies have come under some criticism, and other studies have found little effect. ³⁸

APPENDIX B: FEEDBACK MEASURES

[Appendix B contains economic equations and cannot be reproduced online]

² These results did not hold up with an alternative model, however. See Brian W. Cashell and Jane G. Gravelle, Investment Tax Credit: Using Macroeconomic Models to Assess Short Run Effects, Library of Congress, Congressional Research Service Report 93-16 E, January 7, 1994.

³ See Brian W. Cashell, How Big is the Fiscal Policy Multiplier? Library of Congress, Congressional Research Service Report 94-403, May 9, 1994, for a discussion of some of these issues. Note also that it is important to distinguish between nominal and real multipliers. A policy that allows the price level to increase can result in a nominal increase in output or revenues, but that increase might measure price change rather than real output.

⁴ Supply responses are generally calculated in real terms. If the level of nominal activity were held constant, however, the same real effect would occur, but in would show up not as a change in the overall dollar value of output but in a change in prices.

⁵ See U.S. Congress, Joint Committee on Taxation, Discussion of Revenue Estimation Methodology and Process. Washington, D.C.: U.S. Government Printing Office, August 13 1992; See also Howard W. Nester, A Guide to Interpreting the Dynamic Elements of Revenue Estimating. Compendium of Tax Research 1987. Office of Tax Analysis, Department of the Treasury, Washington, D.C., 1987; Emil D. Sunley and Randall D. Weiss, The Revenue Estimating Process, Tax Notes, June 10, 1991, pp. 1299-1314; and a panel discussion (Rosemary Marcuss, Dan Feenberg, Gene Steuerle, Jerry Auten, and Tom Barthold) in the 1993 Proceedings of the Eighty-Sixth Annual Conference of the National Tax Association, Columbus, Ohio: 1994, pp. 3-14.

⁶ These offsets include both realizations responses and portfolio effects; there was also an additional revenue offset because of some other features of the tax revision, primarily depreciation recapture. The estimates were also affected by different baseline assumptions, which would have caused some differences in purely static estimates. Using the JCT baseline, the static loss would have been $100 billion; using the Treasury baseline, the static loss would have been $80 billion. Considering the wide range of empirical estimates, the behavioral assumptions of the JCT and Treasury were quite similar. See Jane G. Gravelle, Can a Capital Gains Tax Cut Pay for Itself? Library of Congress, Congressional Research Service Report 90-161, March 23, 1990, for a discussion.

⁷ See Dennis Zimmerman, The Effect of the Luxury Excise Tax on the Sale of Luxury Boats, Library of Congress, Congressional Research Service Report 92-149, February 10, 1992; and General Accounting Office, Luxury Excise Tax Issues and Estimated Effects, GAO/GGD-92-9.

⁸ Organization for Economic Cooperation and Development, A Comparative Study of Personal Income Tax Models, Report by the Committee on Fiscal Affairs, 1988.

⁹ Ibid., p.13.

¹⁰ See Jon David Vasche and Hoang Nguyen, The Treatment of Feedback Effects in Revenue Impact Analysis, Tax Notes, October 31, 1994, pp. 599-618. For another discussion of State revenue estimating see Federation of Tax Administrators, Revenue Forecasting and Estimation -- How It's Done, State by State, State Tax Notes, May 3, 1993, pp. 1039-1051. For a more detailed discussion of the Massachusetts model by an individual at the firm that developed the model, see John J. Hudder, Use of Models in Tax Policy and Revenue Analysis: A Great Leap Forward, State Tax Notes May 17, 1993, pp. 1181-1185.

¹¹ The cyclical (demand side multiplier) effects are probably not very important, however, because of leakages through trade that quickly depress multipliers and because of the tendency of States to strive for balanced budgets for a variety of reasons.

¹² Vasche and Nguyen, The Treatment of Feedback Effects in Revenue Impact Analysis.

¹³ Ibid., p. 615.

¹⁴ Our overall fiscal policy stance may be more appropriately measured by a standard employment ("full employment") deficit or surplus, which measures what the shortfall or surplus would be when the economy is at "full employment." A fully employed economy, in this context, would still have some unemployment due to frictional and structural unemployment. Hence, the term "standard employment deficit" is currently used.

¹⁵ See Laurence J. Kotlikoff, Generational Accounting, New York: The Free Press, 1992, for a view that a much more complex and different set of accounts is needed to assess the nation's fiscal position.

¹⁶ This issue is discussed in Jane G. Gravelle, Estimating Long-Run Revenue Effects of Tax Law Changes, Eastern Economic Journal, Vol. 19, no. 4, Fall 1993, pp. 481-494. See also Emil M. Sunley and Randall D. Weiss, The Revenue Estimating Process, Tax Notes, June 10, 1991, pp. 1299-1314 and Alan J. Auerbach, Public Finance in Theory and Practice, National Tax Journal, vol. XLVI, No. 4, December 1993, pp. 519-526.

¹⁷ This discussion is not meant to imply that the macroeconomic analysis is simply a matter of plugging in a deficit or surplus number. Some types of tax changes (e.g. investment subsidies) require more sophisticated and complex treatment to determine how the tax reduction or increase is translated into spending, and the degree of such an effect is subject to considerable dispute.

¹⁸ Actually, an argument might be made that spending reductions have slightly larger multipliers than tax cuts because they may be more likely to be spent.

¹⁹ While the main thrust of a macroeconomic model is to account for demand-side multipliers, some models may also incorporate small supply effects. In that case, one would not wish to use dynamic estimates that are already adjusted for these effects to make projections.

²⁰ Ballard, et al, A General Equilibrium Model for Tax Policy Analysis, 1985.

²¹ As shown below, the effect on the capital stock in the short run is negligible and simply complicates the exercise, so it is not included.

²² Ballard, et al., A General Equilibrium Model for Tax Policy Analysis, 1985.

²³ Michael Boskin, Taxation, Savings, and the Rate of Interest, Journal of Political Economy, vol. 86, January 1978, pp. s3-s27. For other sources of times series estimates see the survey in Jane Gravelle, The Economic Effects of Taxing Capital Income, Cambridge: MIT Press, 1994, pp. 25-28.

²⁴ One would need to develop a model that traces the time path of feedback effects to consider effects at any particular time.

²⁵ One common division is a "consumption type" tax change relative to a "wage type" tax change. See Jane G. Gravelle, The Economic Effects of Taxing Capital Income, pp. 114-118 for a discussion of this issue.

²⁶ As an illustration, a recent brief analysis by the National Center for Policy Analysis (Brief 137, October 26, 1994) indicated that the proposal to cut capital gains taxes by 50 percent and to index capital gains would increase the capital stock by $2.2 trillion by the year 2000. This study did not provide any details of how the calculation was made, but the following shows how a set of assumptions could produce such a result.

The brief indicated that the tax cut would reduce the cost of capital by 5 percent. As a rule of thumb, the capital stock is about 3.5 times GDP, for a stock of around $20 trillion. This increase, therefore represents an increase of around 10 percent. Even in the long run steady state calculations, the formula for the percentage increase in the capital stock is only Er(aEs+S)/[(Er+S)(1-a). If the 5 percent change in price is correct, then the value of this amount must be close to 2. That outcome becomes possible if E was set at infinity, which would set the formula to (aEs+S)/(1-a). That result could be obtained with elasticities of slightly over 1 for the remaining two elasticities. If we used our elasticities, which are themselves at the higher end of the spectrum, to obtain a long run result, the effects would be only 20 percent as large; thus the increase is five times as large as these empirical estimates might suggest would be the case in the LONG RUN -- a point that might not be approximated for decades. In the short run, using the partial equilibrium short-run formula, we would expect savings to go up no more than 2 percent (0.4 times .05) which would, applied to savings of approximately $330 billion, yield an initial increase of less than $7 billion. Over five years, even allowing for growth, the results would be less than 2 percent of the $2.2 trillion. Thus both timing and elasticities appear to play a role in these differences (although if the elasticity is infinite, the timing could presumably be instantaneous). Note that to obtain a $2.2 trillion increase over five years would require at least doubling current private savings.

²⁷ See Jane G. Gravelle, The Economic Effects of Taxing Capital Income, p. 232 for a brief summary of the literature on international capital flows.

²⁸ See James M. Poterba, Venture Capital and Capital Gains Taxation, Tax Policy and the Economy 3, National Bureau of Economic Research, ed. Lawrence H. Summers, Cambridge: MIT Press, 1989.

²⁹ This and other points are made by Gene Steuerle, How Estimates Can Bias the Legislative Process, Tax Notes, December 5, 1994.

³⁰ See Charles L. Ballard, et al., A General Equilibrium Model for Tax Policy Evaluation. National Bureau of Economic Research. Chicago: University of Chicago Press, 1985, pp. 136-137; John Pencavel, Labor Supply of Men: a Survey, in Handbook of Labor Economics, edited by Orley Ashenfelter and Richard Layard, New York: North-Holland 1986; Jerry A. Hausman, Taxes and Labor Supply, in Handbook of Public Economics, edited by Alan J. Auerbach and Martin Feldstein, New York: North Holland 1985.

³¹ See Thomas MaCurdy, David Green and Harry Pearsch, Assessing Empirical Approaches for Analyzing Taxes and Labor Supply, Journal of Human Resources, vol. 25, Summer 1990, pp. 425-90; Robert K. Triest, The Effect of Income Taxation on Labor Supply in the United States, Journal of Human Resources, v. 25, Summer 1990, pp. 495-516.

³² See Charles L. Ballard, et al., A General Equilibrium Model for Tax Policy Evaluation and Mark R. Killingsworth, James J. Heckman, Female Labor Supply: A Survey, in Handbook of Labor Economics, edited by Orley Ashenfelter and Richard Layard, New York: North-Holland 1986.

³³ Thomas A. Mroz: The Sensitivity of an Empirical Model of Married Women's Hours of Work to Economic and Statistical Assumptions, Econometrica, Vol. 55, No.,4, June 1987, pp. 765-799.

³⁴ John Pencavel, Labor Supply of Men: a Survey, in Handbook of Labor Economics.

³⁵ Lawrence Lindsey, Individual Taxpayer's Responses to Tax Cuts: 1982-1984, Journal of Public Economics, 1987, pp. 173-206; Martin Feldstein, The Effect of Marginal Tax Rates on Taxable Income: A Panel Study of the 1986 Tax Reform Act, National Bureau of Economic Research Working Paper 4496, October 1993; Gerald Auten and Robert Carroll, Behavior of the Affluent and the 1986 Tax Reform Act: The Role of Demand Side Characteristics, Presented at the National Tax Association Meetings, November 1994. There have also been related studies that examined the circumstances of very high income individuals. See Daniel R. Feenberg and James M. Poterba, Income Inequality and the Incomes of Very High Income Taxpayers: Evidence from Tax Returns, National Bureau of Economic Research, Working Paper 4229, December 1992 and Joel B. Slemrod, On the High-Income Laffer Curve, in Tax Progressivity and Income Inequality, edited by Joel B. Slemrod, Cambridge: Cambridge University Press, 1994.

³⁶ See Jane G. Gravelle, Behavioral Responses to Proposed High Income Tax Rate Increases: An Evaluation of the Feldstein-Feenberg Study, Congressional Research Service 93-434, April 20, 1993, for a discussion of some of the problems with drawing inferences from these studies, with particular reference to evidence from the 1981 tax cut.

³⁷ For a summary of the empirical research see Jane G. Gravelle, The Economic Effects of Taxing Capital Income, pp. 25-28.

³⁸ For a survey see Jane G. Gravelle, The Economic Effects of Taxing Capital Income, Cambridge: MIT Press, 1994, pp. 190-198.