EVENTS

Date:	Wednesday, November 5, 2003
Time:	9:00 AM -- 1:30 PM
Location:	Wohlstetter Conference Center, Twelfth Floor, AEI 1150 Seventeenth Street, N.W., Washington, D.C. 20036

November 2003
Static versus Dynamic Scoring

A source of much spirited debate, budget "scoring" has been used to estimate changes in government revenues and outlays if federal tax and spending programs are altered. Historically, the U.S. federal government has used a static-scoring methodology that assumes that all policy changes would not affect the size of the economy, regardless of how much businesses invest and hire or how much households save and work. In recent months, initial attempts by the Congressional Budget Office and the Joint Committee on Taxation raise the possibility that "dynamic scoring"--which takes into account, where appropriate, all behavioral reactions to changes in tax policy--might be more fully adopted. Is dynamic scoring a step in the right direction? And, if so, how it should be carried out? At a November 5 AEI event, budget and tax policy experts explained dynamic scoring, the pros and cons of the debate, and how dynamic scoring might be used in the future.

Doug Holtz-Eakin
Congressional Budget Office

The CBO, in what was somewhat erroneously seen as an unprecedented method of analyzing the president's fiscal policy proposal, compared the existing January baseline projection for the economy with one generated using the president's proposed budget.  Holtz-Eakin considers this dynamic scoring method to be promising, given the information that it can, theoretically, provide to policymakers. He emphasizes, however, the logistical and implicational limitations that will require our attention as we implement this new method of analysis.  In addition to the difficult analytical challenge of quantifying a broad range of tax and outlay proposals, the approach depends on the standardization of procedural controls in order to guarantee consistent results.  Additionally, the CBO is faced with the challenge of presenting its findings to Congress in a timely enough fashion that the findings remain applicable and sufficiently transparent so as to facilitate digestion by a congressional audience with limited knowledge of economic theory. Aside from these specific issues, Holtz-Eakin warns economists not to limit their time to the static versus dynamic debate, but to instead consider a broader range of empirical data. 

Dynamic Analysis at CBO

Robert Dennis
Congressional Budget Office

The fixed GDP assumption of conventional scoring methods is a cause for concern among those responsible for providing policymakers with complete and accurate scoring estimates. Scoring estimates made by the Joint Committee on Taxation (JTC) and subsequently passed to the CBO exclude factors likely to generate changes in GDP such as changes in labor supply and capital stock.   In addressing changes in GDP resulting from the 2004 tax proposal, the CBO calculated short-run Keynesian effects using macroeconomic advisor (MA) and global insight (GI) models, and effects on general equilibrium with Ramsey and Stochastic OLG.  Each model ultimately projected a resulting increase in GDP.   The cumulative impact on the budget surplus, according to these models, varies from the conventional estimate by a margin of 13-22 percent during the first five years. All but one of the macroeconomic feedbacks resulting from this exercise are smaller than the interest cost, also not included in the conventional scoring estimate.  Therefore, taking into account the macroeconomic feedback and the additional debt service, the conventional estimate does in fact come close to the original conventional proposal. 

Ben Page
Congressional Budget Office

After scoring the president's 2004 budget proposal, the CBO is hesitant to predict a net budgetary movement.  The acceleration and extension of the 2001 tax cuts will likely facilitate a positive budgetary effect.  The end of double taxation will also have some growth effects, including increased activity in the corporate sector and a resulting increase in efficiency gains.  Additionally, lower marginal tax rates on corporate income should result in increased saving tendencies.  On the other hand, increased transfers will likely bring a rise in disposable income levels and consumption, and with them a drop in work effort and investment.  The likelihood of coinciding decreases in investment and increases in labor supply makes any prediction of a net effect difficult to predict.  The issue of whether financing will be done through decreased government consumption or increased tax rates could very well mean the difference.   The results of scoring analysis in this case provide a wider range of estimates than found in the previous experiment with the tax proposal largely due to the difference in size and the disparity of the elements involved.

William Beach
Heritage Foundation

It is important that economists understand the micro foundations of macroeconomics.  The CBO should use micro-simulation techniques to gain a better appreciation of how the macro-effects of one period become micro-effects in the next.  These linkages between micro and macro, says Beach, should begin with an increased emphasis on tax analysis.  The CBO should begin to give greater consideration to the merging of treasury data and public use files with current population surveys.  Demographic trends present another challenge.  This country is quite different demographically than it was thirty years ago, and it is in the interest of the CBO to take this new reality seriously.  Lastly, the CBO should consider the value of building intervivo and intergenerational data set requests. 

Current Issues and Future Developments

Alan Auerbach
University of California, Berkeley

It is vital that we are diligent when distinguishing scoring from forecasting. Scoring is a direct input of the legislative process and is much more challenging, both because it must be done for every proposal and because there must be a singular numerical result.  We must also carefully distinguish between the three approaches to scoring.  Static-scoring methods necessarily exclude all forms of behavioral response.  Micro responses are defined by a fixed baseline, that is, their assumption of fixed macro-level variables.  Macro responses or dynamic-scoring methods allow for movements of this baseline. 

We know that policies have macro-effects and that we can, theoretically, predict these effects. The strongest and most obvious case for macro-level response is this greater breadth of information they should theoretically provide.  Moreover, micro responses may bias policymakers against tax cuts.  Macro responses are, however, subject to judgment and assumptions, and they require an impractical integration of scoring and baseline estimation.  Models currently used to estimate the baseline are simply not capable of providing enough detail.

Rudolph Penner
Urban Institute

In order for dynamic scoring to be practically applicable, economists must be able to generate a single figure for the purpose of gauging congressional success at reaching some quantitative target.  We are far from reaching this goal and, furthermore, attempting to do so will require a tremendous amount of intellectual effort.  Before economists devote their time to such an undertaking, they should decide if it is worth such an effort.  They should be aware that Congress pays little attention to analytical results and has little patience for changes in the baseline.  It is especially vital that we put this issue in perspective.  Economic forecasts have become far less accurate than they were a decade ago-to the point of crisis.  It is more important that we focus our energy on the problems associated with our current models than devote valuable intellectual resources to what is, in reality, but a marginal improvement over current methods.

Luncheon Address

Martin Feldstein
Harvard University

The recent advancements in behavioral simulation analysis toward a much more ambitious goal of dynamic scoring, and the implementation by the CBO and JTC of new research models to address this challenge, seem promising.  It is, however, more important that we focus our attention on the practical improvement in tax simulation analysis.  The first step that must be taken to improve the revenue estimates associated with a proposed tax change is to drop the convention that the revenue effect is conditional on an unchanged GDP.  The key parameter for such a calculation is in fact the elasticity of taxable income, and more research needs to be done in this area.   When it is possible, the analysis should include changes in economic well-being, or changes in the dead weight loss that will result in tax rates or tax rules.  The short-run demand effect should be calculated but reported separately so that the offsetting changes of the Federal Reserve be taken into account.  Finally, the infinite horizon general equilibrium calculation should be kept on the back burner as a long-term research project rather than a basis for current scoring. 

AEI intern Brian Long prepared this summary.