Choosing A Budgetary Rule


The role of numbers in policy analysis is to reduce the disputeabout whether to approve a specific policy change. That role willbe served only if the process for estimating the numbers is notitself subject to dispute.

The central point of my brief remarks is that any change in theprocess for estimating the revenue effects of tax changes should bebroadly understood and approved--preferably by the key members ofboth parties in each house and by the administration. Such a changeshould be considered the equivalent of a change in the bylaws of aclub or a change in the scoring rules in an athletic league. Suchchanges should be approved only by the support of most of theaffected groups, not by only those who expect to benefit most fromthe change in the short term.

Some other comparisons may help illustrate the issues bearing onthe choice between static and dynamic revenue estimates: Staticestimating is an application of arithmetic. Many politicians arenot very good at arithmetic, but it does not evoke much partisandispute. Dynamic estimating is based on some model of economicbehavior, a model that reflects some theory of how people behaveand estimates of how they respond to specific types of changes inthe conditions they face. Some of the characteristic differencesbetween parties involve differences on just these issues.

In that case, static estimating is somewhat like democracy--itmay be the best deal we can make with our neighbors.

But we should try to convince our neighbors if there is reasonwe can do better. And dynamic estimates can be much more accuratethan static estimates. In general, people will do more of someactivity if the after-tax returns are increased and less of thisactivity if after-tax returns are reduced, and that is the basisfor the higher potential accuracy of the dynamic estimates. Wewould probably make better tax policy decisions even on the basisof crude dynamic assumptions--for example, that tax increasesincrease revenue and that tax reductions reduce revenue by onlyhalf that estimated by static models.

But we should be able to make even more accurate estimates.There are still some differences in the estimates of the magnitudeand timing of the responses to tax changes, but some of thesedifferences can be resolved by focusing on the same scope ofresponses. For prime age males, for example, the response of hoursworked to a change in after-tax wages appears to be close to zero;the effect of tax rates on taxable earnings, however, is higher,reflecting the response of taxable earnings to tax-induced effectson occupation, location, and tax avoidance. Similarly, the responseof the savings rate to the after-tax interest rate appears to beclose to zero; the effect of tax rates on taxable interestpayments, however, is higher, reflecting the tax-induced effect onthe type of investment. The full behavioral response to change intaxes is often substantially higher than the first stage response,especially in the long run.

May I suggest, however, that the revenue estimators stop shortof including the potential demand-side effects of tax changes.First, there continues to be a major disagreement amongmacroeconomists as to whether tax changes have any significanteffect on aggregate demand. (On that issue, my position is thatmost changes in fiscal policy have no significant effect onaggregate demand, but I acknowledge that many of my professionalcolleagues believe otherwise.) And second, any demand-side effectscan be offset by changes in monetary policy. For these reasons, Isuggest, estimates of the dynamic effects of tax changes on taxrevenues should be based on supply-side models, not on the olderform of Keynesian macromodels.

The next steps toward making sense of this issue, I suggest, arethe following:

First, put to rest the wholly false, albeit common, charge thatthe unexpected increases in the federal deficit in the early 1980swere due to misleading dynamic supply-side revenue forecasts. Infact, all of the budget forecasts by both the administration andCongress were based on static revenue estimates; moreover, the OMBand CBO budget forecasts in 1981 were remarkably similar. Thefederal deficits of the early 1980s proved to be substantiallyhigher than expected for several reasons--the unusually deeprecession of 1981-1982, a faster-than-expected decline ininflation, and a failure to maintain spending restraint beyond thefirst Reagan budget. All of the budget forecasts during this periodsubstantially underestimated the deficit, but not because they werebased on supply-side models.

Second, those who favor higher taxes should acknowledge thatincreases in the top marginal income tax rates generate littleincreased revenue; a given increase in tax rates at this level is alarger proportionate reduction in the after-tax rate, and highincome taxpayers have more opportunities for legal tax avoidance.Similarly, those who favor lower taxes should acknowledge that sometax cuts reduce revenues by more than the state estimates. The $500tax credit for children proposed in the House Republican Contract,for example, would generate larger dynamic revenue losses to theextent that it increases birth rates or reduces the participationof women in the paid labor force. These examples illustrate thatdynamic revenue forecasts do not necessarily favor the preferredpolicies of either party.

Third, the Joint Committee on Taxation should open up itsestimating methods and invite peer review. May I suggest that youstart this process by asking the respected National Bureau ofEconomic Research to sponsor studies and a conference on the JCTmethodology and on the most important next steps to improve therevenue forecasts. Leading public finance economists should beasked to comment on the JCT methodology and report to Congress,maybe at hearings before this committee, on their evaluations andrecommendations

And finally, pending completion of this review, no change in theJCT methodology is appropriate. A substantial consensus amongleading public finance economists, I suggest, is probably necessaryto broaden the support for proposed changes to this methodologyacross parties in Congress and with the administration. And, as Iintroduced my testimony, more accurate revenue forecasts from thebest possible dynamic model would help resolve differences on taxpolicy only if the methodology by which the forecasts are generatedis endorsed by most of the major participants in the policydebate.

Thank you.