Mr. Chairman and members of the subcommittee, it is a great privilege to have the opportunity to appear before you today. I am an economist who works at the Washington-based think tank, the American Enterprise Institute. I have spent a good deal of my research time since I completed my dissertation studying the effects of taxation on the economy. I come to you today to provide thoughts on the key considerations associated with accounting for all of the dynamic effects when scoring spending and tax proposals.

Background When the Joint Committee on Taxation (JCT) and the Congressional Budget Office (CBO) provide estimates to Congress of the revenue impact of a tax package, behavioral effects are only partially accounted for. Policy changes are not scored as having an impact on the total level of aggregate activity, a key cornerstone of the budget projection. Policy changes are scored, however, as having an effect on the composition of that activity. For example, if Congress were to consider a bill that provided a tax credit for a particular type of equipment, then the JCT might assume that firms would employ more of that type of equipment and less of a type that does not qualify when calculating the cost of the proposal. Total investment spending in the economy, however, would be left unchanged by the policy.

For the majority of proposals, such a procedure is quite sound. Most new policies are small enough that they would not plausibly have a large impact on the economy as a whole. However, for some policies, this procedure clearly provides an inaccurate picture. The recent debate over the stimulus package provides an interesting case in point. The measures adopted were, in part, designed to help the economy recover from recession. The cost of the policies, however, was conditioned on the assumption that there would be no effect on the economy. If such an assumption were reasonable, then the stimulus package would be a bad idea. When designing policy, policymakers must keep a careful eye on their cost. Presumably, the stimulus package was the size that it was because of the fear that the budgetary implications of larger measures might be negative. If a more realistic scoring approach had been adopted, the stimulus bill might well have been larger.

Opponents of dynamic scoring most often argue that there is too much uncertainty concerning the effects of economic policies for one to expect revenue estimates to be reliable enough to make there use advisable. They sometimes also argue that political pressure might be used to influence the scorers. Others note, however, that this aversion to seeking the truth is accompanied by a cost. Static scoring methods may bias policymakers away from measures that reduce taxes, by making the revenue loss associated with reductions appear too high.[1] Because of this, an increasing amount of attention has been paid to the question of dynamic scoring, and a significant amount of progress has been made by those investigating these issues.

The Uses of Scoring Scoring of a proposal has two objectives. The first is to provide policymakers with a perspective on the likely impact of any proposal. The second is to provide policymakers with hard budget numbers that can be used when constructing prudent rules to constrain irresponsible spending or excessive tax reductions. As you know, rules that effectively require special overriding actions have often constrained Congress’s ability to adopt policies that have significant negative effects on the budget balance.

It is worth mentioning that these two objectives are often in conflict. There is a small body of evidence, for example, that positive surprises to government revenue may lead to higher government spending.[2] If Congress were to rely upon a dynamic score for a tax bill, and that score allowed for GDP and therefore tax revenue to be higher, then one might predict that government spending in the current year would be less constrained by a dynamic score than it would be by a static score.

Another conflict between the two objectives strikes at the core of the responsibility of this committee. In order to think rationally about the likely impact of a tax policy, one would like to be presented with a broad range of estimates, each accompanied by a careful explanation of the sources of disagreement between it and the other estimates. One would then apply one’s own judgment when deciding the proper course of action, perhaps after consultation with a disinterested professional expert (from the CBO perhaps). Such a procedure is commonly relied upon by the Federal Reserve when evaluating the impact of both monetary and fiscal policy. Professional staffers provide Board members with careful and neutral analysis, often even presenting them with more than one estimate. The members ultimately decide for themselves how to vote. This is worth repeating. The Fed’s models are subject to the same uncertainties as the CBO’s, but they are constantly used to influence policy. Why are the Fed’s procedures so reasonable and those used to evaluate tax policy so unreasonable? Most likely because the Fed is more insulated from political pressures, and these make the issue much more complicated.

In the political process, the opposing sides may decide to agree to the use of a specific number for the purposes of debate. Often, the competition for the title of "best estimate" is extremely tight, and the choice of a single number by the professional advisor is an unpleasant task. Again, any accurate statement about the likely impact of major policy changes will provide a diversity of opinion. If we are going to adopt budget rules that rely on one number, which should we choose? There are significant costs and benefits associated with any number-picking strategy. In particular, the choice of best strategy for the purposes of constructing a budget rule appears to have a strong impact on the perceptions of policymakers concerning the likely impact of the policy. Opponents of President Bush’s tax proposal last year, for example, often spoke as if the static score of that bill were an unambiguous fact established by the JCT. That is, the choice of a specific number for revenue estimating purposes necessarily imbues that number with too much credibility.

One additional point is worth making. Supporters of tax reforms have often been the strongest advocates of dynamic scoring, but one should note that the issue of dynamic scoring is not necessarily limited to tax reduction scenarios. The economic literature implies that higher government spending can increase short-run economic growth, while providing a long-run drag on the economy. If one has a short enough time horizon, it is easy to envision scenarios where the dynamic positive feedback from higher spending would be scored to be quite significant. Again, this suggests that there is a conflict between the two objectives. An accurate picture of the effect of spending policies would likely relax constraints on government spending that are associated with revenue estimates. One could even imagine short-run spending binges occurring because of dynamic scoring, whereby higher government spending increases estimated GDP and revenue, thereby leading to a further increase in government spending.

Effects of a Flawed System It is easy to see, given these conflicting forces, how we could arrive at a place where we use a flawed system, even before consideration of the role of uncertainty. The estimates are used for several purposes that are often in conflict. But the flawed system has real consequences, and it must be improved.

Some observers will certainly argue that static scoring leads to a world with too few tax reductions. Others will argue that static scoring leads to a world with too little government spending. If the negative long run growth effects of government spending were accounted for, it might even be argued that static scoring leads to too much government spending. All of these arguments, however, miss the most important distortion caused by our current system. Because economic analysis is not used to demonstrate the benefits of tax (and perhaps spending) proposals, there is virtually no force present disciplining policy makers to adopt economically sound proposals. We see the unfortunate results of this quite often.

Economists are, I believe, unanimous in the view that a well-designed tax system will have as broad a base and as low a marginal rate as possible, given a set of revenue and social welfare objectives. They believe this because such a system has important positive economic effects. A tax reform---like the 1986 Tax Reform Act---that moves us toward the economic ideal will have positive long-run growth effects. Alternatively, a proposal that narrows the tax base and raises marginal tax rates---something accomplished by the many tax credit programs---might well have negative dynamic effects. If decision-makers relied upon accurate scores of the two types of proposals, then it would be much harder than it is today to make the wrong choice, and a prudent tax policy would have a much higher chance of gaining bipartisan support.

The Role of Uncertainty There is a great deal of uncertainty among economists concerning the likely impact of any specific tax proposal on the economy. Consider, for example, the 1997 JCT Tax Symposium, where many of the economics profession’s most distinguished modelers calculated the economic effects of a switch to a consumption tax. Estimates of the impact of such a change on real GDP in 2010 ranged from a low of 1 percent higher GDP to a high of 16.9 percent higher GDP. The mean estimate of the impact of such a change was 5 percent, and the mean excluding the highest estimate was 2.1 percent. Obviously, the work of these scholars defines a fairly wide range of possibilities. Some argue that uncertainty concerning these estimates is too large for them to be useful. However, if Congress were to consider the adoption of a consumption tax, the current system would require the policy to be scored using an estimate (zero) that is outside of the range of estimates of our best models, effectively substituting an answer we are confident is wrong for our best guess.

When might such caution be sensible? Economists who have studied the impact of uncertainty on optimal decision making have found that it is also important to track the effect that errors might have in each direction. If an error in one direction can lead to an extreme negative consequence, for example, then it will be optimal to be very cautious and err in the other direction. Such effects are largest in economic models that do not allow agents to change their behavior over time. If policy decisions today were irreversible, then it might be optimal for us to rely upon extremely conservative revenue projections when setting future spending, especially if it is believed that negative consequences result from high deficits. As it is, however, policy changes every year, and a misstep today can easily be reversed in the future. In such a circumstance, Congress should optimally consider policies that maximize our expected welfare, and not be as excessively risk averse as it is under the current system. This reasoning also suggests that attempts to commit future Congresses to specific policy paths fundamentally alter the problem, and create a world where it is more likely to be optimal to be extremely risk averse and rely on static scoring.

Recommendations These considerations suggest a number of positive steps. My first recommendation is that Congress take a cue from the Federal Reserve and rely more heavily on its professional staff. When a literature provides differing opinions as to the efficacy of a certain policy, there is no substitute for a disinterested professional observer who can serve as a referee. The CBO already serves this function, updating its forecast, for example, after the President’s tax proposal became law last year, and providing a dynamic score of its effects. Congress could immediately begin a process whereby dynamic scores of new proposals are requested in a timely enough fashion that they could have an impact on the political debate. While the CBO is certainly not perfect, the able men and women of the agency would certainly respond to criticisms of their approaches over time to the extent that the criticisms contained academic merit. Any move in this direction, by the way, should include a request that the CBO’s methods be more transparent than they currently are.

Congress might alternatively consider setting up an independent body for fiscal policy evaluation, modeled after the Federal Reserve’s staff. Such a measure may significantly reduce the chance that political influence could have an impact on the analysis of the economic staff, and might also restrain the tendency for the economic analysis to be tied to unrealistic projections of future policies, as is now sometimes the case.

Congress should also recognize that revenue estimates currently serve two purposes and that such double duty is not necessary or advisable. The optimal procedure for information revelation may be quite different from the optimal procedure for establishing budget rules. Absent budget rules, however, the imprecise scoring mechanism may have more influence than it should. One could think of any number of reasonable rules, for example, that would constrain the growth of government spending without relying explicitly in real time on revenue forecasts. If, for example, spending growth targets were set on an ex ante basis, then spending would be far less likely to respond positively to a positive revenues. When setting these limits, this committee would have to debate the optimal level of government spending, and adjust estimates of this level over time in response to new circumstances. For example, a reconsideration of the spending caps might be mandatory if a deficit larger than some agreed upon size emerged. Such careful monitoring creates the conditions wherein reliance upon dynamic scoring is quite feasible, and would likely be an important part of any optimal budget system.

[1] From this perspective, the partial dynamic scoring methods used may be more biased than a strict static score. For example, an Investment Tax Credit for a type of equipment would have a higher cost after the Joint Tax Committee accounted for substitution into that type of equipment than would be implied by a static score.

[2] Von Furstenberg, Green, and Jeong (Review of Economics and Statistics, 1986) use U.S. federal budget data from 1954-82 to explore the relationship of causality between tax revenues and expenditures. They find that spending does not respond to changes in taxes but that higher spending leads to higher taxes in the future. Anderson, Wallace, and Warner (Southern Economic Journal, 1986) use U.S. federal budget data from 1946-83, and also conclude that spending causes taxes. In contrast, Manage and Marlow (Southern Economic Journal, 1986) use U.S. data from 1929-82 and find that the evidence supports the taxes lead to spending hypothesis. Ram (Southern Economic Journal, 1988) uses both annual data from 1929-83 and quarterly data from 1947-83, and concludes that causality runs from revenue to expenditure. Calomiris and Hassett (National Tax Journal, 2002) found that revisions to CBO budget forecasts had a significant effect on subsequent spending decisions.