The President's fiscal year (FY) 2007 budget sub­mission to Congress includes a number of important initiatives. Among them is a plan to create a Dynamic Analysis Division within the Office of Tax Analysis (OTA) in the U.S. Department of the Treasury.

Dynamic analysis gauges the impact on federal tax revenues of the changes in output and incomes induced by changes in tax policy. Proponents of sup­ply-side tax cuts and fundamental tax reform pro­vided much of the original push for dynamic analysis. They did so because conventional revenue estimates may take into account the microeconomic behavioral effects of a tax policy change, [1] but they exclude the effects on federal tax receipts of changes in macroeco­nomic factors like labor force participation, invest­ment, and capital accumulation. Dynamic analysis makes use of advances in computing technology and economic modeling to generate dynamic revenue estimates that include these macroeconomic factors.

The tax agenda has now shifted. In the short run, the focus will likely be on which of the expiring pro­visions of the 2001 and 2003 tax acts, if any, to extend. In the longer run, devising a tax policy that generates additional revenues without slowing eco­nomic growth will likely be an important part of the policy debate.

Dynamic analysis has a key role to play in this debate. Providing one dynamic "score"--or one dynamic estimate of the revenue effects of a tax policy change--may be difficult to envision. The choice of macroeconomic model along with the assumptions made about monetary policy in the short run and fiscal policy in the long run can influence dynamic estimates, sometimes even changing the sign of estimated gross domestic product (GDP) and reve­nue effects. However, dynamic estimates can still enhance the budgeting process by helping to iden­tify and rank those provisions of a tax proposal that are most likely to achieve policymakers' stated objectives.

Recent Efforts

Both the Joint Committee on Taxation (JCT) and the Congressional Budget Office (CBO) have been working to integrate dynamic estimates into the budgeting process. The JCT's efforts began in ear­nest with a 1997 symposium on "Modeling the Macroeconomic Consequences of Tax Policy."[2] In March 2003, the CBO published its first dynamic analysis of the President's budget.[3] Both the JCT[4] and the CBO[5] have since released papers exploring the application of macroeconomic models to the analysis of the economic and budget effects of tax policy changes.

The Treasury Department has recently begun to produce its own dynamic estimates. The President's Advisory Panel on Federal Tax Reform used dynamic analysis from the Treasury Department to help to evaluate different reform options.[6] The FY 2007 Mid-Session Review included for the first time dynamic estimates of the economic effects of one of the President's tax proposals--permanently extending some provisions of the 2001 Economic Growth and Tax Relief Reconciliation Act (EGTRRA) and the 2003 Jobs and Growth Tax Relief Reconciliation Act (JGTRRA) that are set to expire in 2010.[7]

A Dynamic Analysis Division within the OTA would help to institutionalize this work at the Trea­sury Department. The U.S. House of Representa­tives has already approved roughly $520,000 in funding for a nascent Dynamic Analysis Division. However, the funding is still pending in the U.S. Senate and could become the first casualty of the Democratic takeover of Congress in January.

This would be regrettable. A Dynamic Analysis Division at the Treasury Department could do much to inform the budget debate. One way to see this is to consider what the next steps for dynamic analysis at the department could be, assuming the Senate approves funding before the 109th Con­gress adjourns.

The Next Steps at the Treasury Department

The answers to four questions are important when considering the next steps that the Treasury Department should take in developing its dynamic analysis capabilities.

• Which macroeconomic models should the OTA use for its dynamic analysis?
• How can the OTA's dynamic estimates account for the key macroeconomic effects of a tax policy change?
• How should the OTA estimate revenue feed­backs from a tax policy change?
• What kinds of dynamic estimates should the OTA provide?

Which macroeconomic models should the OTA use for its dynamic analysis? The choice of which macroeconomic model to use depends on the goal of the proposed change in tax policy and the eco­nomic environment in which it is being considered.

Large-scale macroeconometric models are a good choice when a tax policy change is intended to stim­ulate aggregate demand in the short run. Such mod­els impose the long-run structure of a neoclassical growth model, but their baselines include unem­ployment and short-run gaps between actual and estimated potential GDP. Thus, they can be used to analyze how changes in monetary and fiscal policy affect employment, personal and corporate incomes, personal consumption and saving, residential and business fixed investment, and other key macroeco­nomic variables in the short run.

General equilibrium (GE) models are a better choice when the long-run effects of a tax policy change on labor hours, capital, and GDP are the primary focus of analysis. GE models are full-employment models in which prices are assumed to adjust instantaneously to equate supply and demand. GE models used for fiscal policy analysis typically impose the government's intertemporal budget constraint. Thus, the government cannot indefinitely finance higher spending or lower taxes with deficits. Rather, in the long run, it must cut spending or raise taxes to limit the growth rate of debt to the growth rate of GDP.

The effects of a tax policy change on labor supply and investment can vary dramatically depending on whether the government finances higher spend­ing or lower taxes today with lower spending or higher taxes tomorrow. This is because households and firms are assumed to know the future course of fiscal policy with perfect foresight and to make decisions about how much to work, save, and invest accordingly.

Large-scale macroeconometric models can also be used to analyze the effects of tax policy changes intended to boost the economy's stock of labor and capital in the long run. However, doing so often requires making assumptions about the response of hours worked and investment to changes in mar­ginal tax rates on labor and capital income. Such assumptions are sometimes necessary because macroeconometric models focus on the short-run effects of tax policy changes on disposable income and consumption. Households and firms are not assumed to know the future course of fiscal policy and thus do not adjust work, saving, and invest­ment in response to the long-run effects of changes in marginal tax rates on the returns to capital and labor and the cost of capital.

That said, analyzing marginal rate cuts in a mac­roeconometric model has the advantage of allowing one to compare the macroeconomic effects to the CBO's baseline economic and budgetary projections over the 10-year budget period. GE models do not afford the same opportunity because they are typi­cally calibrated to baselines that are consistent with full employment and a sustainable fiscal policy.

The CBO and the JCT use several macroeconomic models when producing dynamic estimates. In its most recent dynamic analysis of the President's bud­get, the CBO used two GE models, including an overlapping generations (OLG) model, and two macroeconometric models, including the Macroeco­nomic Advisers' (MA) model.[8] The JCT has a set of models with similar capabilities available for dynamic analysis, including the Tax Policy Advisers (TPA) OLG life-cycle model and a macroeconomic growth (MEG) model. The MEG model, like the MA model, can be used to evaluate changes in tax policy meant to boost aggregate demand in the short run.

A new Dynamic Analysis Division should simi­larly maintain a mix of models that will give it the flexibility to do dynamic analyses of tax policy changes intended to boost aggregate demand in the short run and the economy's stock of labor and cap­ital in the long run.[9]

How can the OTA's dynamic estimates account for the key macroeconomic effects of a tax policy change? Accounting for the key macroeconomic effects of a tax policy change presents a challenge. The extent to which a set of dynamic estimates accomplishes this depends in large part on the degree of disaggregation in the macroeconomic model being used.

The Internal Revenue Code is extremely complex, with a progressive rate structure, different tax rates for capital gains realizations and dividend income, an alternative minimum tax, and myriad large and small tax credits and deductions. Most proposed changes in U.S. tax law do not affect all taxpayers and sources of income uniformly. Rather, they are often tailored to benefit specific subsets of taxpayers or sectors. Thus, a proposal can apply different tax rate changes to different sources of income. It can also create tax incentives to reallocate income and resources from one sector or use to another.

Microsimulation models of the federal individual and corporate income tax capture the complexity of the U.S. tax code. They are based on large sam­ples of tax returns that are weighted to match the taxpayer population. As a result, they can generate finely disaggregated measures of the impact of a tax policy change on federal income tax revenues and average effective marginal income tax rates. Micro­simulation models of the federal individual income tax can also be used to estimate changes in labor supply by type of worker and income size. Outputs from microsimulation models are input into mac­roeconomic models when simulating the macro­economic and dynamic revenue effects of a tax policy change.

Macroeconomic models capture little of the complexity of the U.S. tax code. They tend to be highly stylized, often aggregating different sources of income and tax rates into a handful of economic and tax sectors. As a result, they implicitly assume that most taxpayers and sources of income are treated as equivalent in the U.S. tax code.

Taking such an approach to dynamic analysis is problematic. Highly aggregated macroeconomic models cannot explicitly account for the macro­economic effects of applying different tax rates and tax credits to different sources of income and income classes. They also cannot explicitly account for the macroeconomic effects of changes in the tax code, which create incentives to reallo­cate income and resources to tax-favored sectors of the economy. Accounting for these macroeco­nomic effects would require a macroeconomic model that is sufficiently disaggregated to take as inputs more disparate microsimulation-model estimates of the revenue, marginal rate, and labor supply effects of tax policy changes.

The JCT disaggregates income--and thus tax rates and revenues--in its macroeconomic models. For example, the MEG model includes tax rates for wages and salaries, interest income, rental income, dividend income, capital gains, proprietors' income, other individual income, and corporate income.[10] In contrast, in its GE models, the CBO does much less to disaggregate changes in tax rates. In general, tax policy changes are summarized using average effective marginal tax rates on labor and capital income. The JCT has found that its dynamic estimates of the economic and budget effects of tax policy changes are sensitive to the degree of disaggregation included in the macroeco­nomic model.

In its Mid-Session Review dynamic analysis of the President's tax proposals, the OTA takes an approach similar to that of the JCT. The OTA inputs into the TPA model estimated changes in marginal tax rates on capital gains, dividend income, wage and salary income, interest income, and business income. A new Dynamic Analysis Division should continue to disaggregate tax rates and sources of income in its macroeconomic models.

How should the OTA estimate revenue feed­backs from a tax policy change? Revenue feed­backs are the additional tax revenues that will be collected relative to a baseline forecast. They are the result of the macroeconomic effects of a tax policy change on federal receipts. Policy analysts can use such information to gauge the extent to which a tax policy change is likely to improve the government's finances by generating higher levels of economic activity, higher incomes, and additional revenues.

Disaggregation also has a role to play when esti­mating revenue feedbacks. Revenue feedbacks can be calculated as simply the difference between a dynamic revenue estimate from a macroeconomic model and a conventional revenue estimate from a microsimulation model. Alternatively, they can be calculated by simulating the macroeconomic effects of a change in tax policy and then updating the microsimulation model to reflect changes in the forecast levels of incomes and prices.

Calculating revenue estimates using an updated microsimulation model is potentially important because the microsimulation model includes a high degree of disaggregation. Thus, the microsimula­tion model estimates the change in income tax pay­able on a tax return basis. In contrast, a macro­economic model can at best estimate a change in aggregate tax revenues by income source.

In The Heritage Foundation's Center for Data Analysis, we estimate revenue feedbacks from major changes in the federal individual income tax using a microsimulation model.[11] Specifically, we iterate between a microsimulation model of the federal individual income tax and a macroeco­nomic model of the U.S. economy, in our case the Global Insight (GI) model. We update individual and business incomes in the microsimulation model using outputs from the GI model. We then use revenue and marginal rates estimates from the updated microsimulation model to adjust forecasts from the macroeconomic model so that they better reflect the effects of the tax proposal. We continue in this way until differences between the estimated changes in federal individual income tax revenues in the microsimulation model and federal personal income tax revenues in the GI model are minimal or can be accounted for by definitional and other differences in the federal income tax bases used.

Estimates of revenue feedbacks can be sensitive to how they are calculated. An analysis of permanently extending some of the 2001 and 2003 tax acts' expiring provisions put revenue feedbacks at $296.3 billion over 10 years, based on a comparison of the dynamic revenue estimate from the macroeconomic model and the conventional revenue estimate from the microsimulation model.[12] A comparison of the dynamic revenue estimate obtained by iterating between the macroeconomic and microsimulation models and the conventional revenue estimate put revenue feedbacks at $295.5 billion--a difference of less than $1 billion over 10 years.

This difference was more substantial in a dynamic analysis of a flat-tax plan that broadened the personal and corporate income tax bases.[13] A comparison of the dynamic revenue estimate from the macro­economic model and the conventional revenue esti­mate put revenue feedbacks at $968.2 billion over 10 years. A comparison of the dynamic revenue estimate obtained by iterating between the macroeconomic and microsimulation models and the conventional revenue estimate put revenue feedbacks at $768.1 bil­lion. Thus, accounting for disaggregation by taking estimates of the change in federal individual income taxes from the microsimulation model reduced revenue feedbacks by about $200 billion.[14]

What kinds of dynamic estimates should the OTA provide? Tax proposals can include individ­ual provisions that affect the economy in different ways. Dynamic analysis should help policymakers identify the macroeconomic and dynamic revenue effects of a proposal's key provisions.

Such an approach to dynamic analysis is not the norm. The JCT has typically analyzed proposed changes in current tax law only in aggregate. The CBO lumps all proposed changes in taxes and spending into a single package in its annual dynamic analysis of the President's budget.

However, for the FY 2007 Mid-Session Review, the OTA decomposed the President's proposal to extend permanently some of EGTRRA's and JGTRRA's expiring provisions into three separate components. The OTA first did a dynamic analysis of only a per­manent extension of JGTRRA's preferential rates on capital gains and dividend income. It then layered onto that a dynamic analysis of an extension of EGTRRA's lower marginal rates on ordinary income. Finally, the OTA performed a dynamic analysis of a package that included not just EGTRRA's and JGTRRA's lower rates on capital gains, dividend income, and ordinary income, but also EGTRRA's provisions primarily affecting after-tax income.

Treasury could build on this approach. A new Dynamic Analysis Division could evaluate and rank individual provisions of proposed changes in cur­rent tax law by their effects on key economic indi­cators like GDP, labor, capital, investment, and consumption. The choice of economic indicator could vary with the objective of either the individ­ual provision or the proposal in aggregate--e.g., a short-run aggregate-demand stimulus or a long-run increase in labor and capital. The ranking of the provision could vary with assumptions made about monetary and fiscal policies.

A new Dynamic Analysis Division could also rank individual provisions on the basis of their rev­enue effects. The OTA's dynamic analysis for the FY 2007 Mid-Session Review focused only on the mac­roeconomic effects of the President's tax proposals. However, the CBO and the JCT already regularly translate the macroeconomic effects of tax propos­als into revenue feedbacks. Treasury could use esti­mates of economic outcomes from its macroeconomic models to produce similar mea­sures of the dynamic budget effects of key provi­sions of proposed changes in tax policy.

Conclusion

A Dynamic Analysis Division in the OTA could build on and expand work already underway at the CBO and the JCT. It could enhance both the budget process and our understanding of the interactions of taxes and the macroeconomy by:

• Maintaining the mix of macroeconomic models needed to analyze a variety of proposed changes in tax policy (e.g., tax policy changes meant to stimulate consumption or investment in the short run and tax policy changes meant to increase the supply of labor or capital in the long run);
• Exploring the extent to which disaggregation influences estimates of the macroeconomic and dynamic revenue effects of a tax policy change; and
• Reporting the macroeconomic and dynamic revenue effects of the important provisions of proposed changes in tax policy.

However, the Treasury Department is unlikely to undertake this work without congressional approval of funding for a Dynamic Analysis Divi­sion. With not many days remaining before this Congress adjourns, time is running out.

Tracy L. Foertsch, Ph.D., is a Senior Policy Ana­lyst in the Center for Data Analysis at The Heritage Foundation.