Scoring and Dynamic Scoring
July 29, 2015 11:21 AM
Article first published online: July 22, 2015
Click here to view article
Budget “scores” are estimates of the change in the federal unified budget that would result from the passage of specific statutory language. Under current practice, the budgetary effects of all proposals are measured relative to a single, fixed baseline outlook for the budget, which is, in turn, built upon a projection for the United States economy. A key feature of scoring is that in evaluating legislation, the aggregate amount of economic activity—total production and income—is assumed to be unchanged from its baseline values. That is, the proposed legislation is assumed to have no effect on the macro economy and hence there is no accounting for potential feedback from changes in the macro economy to the budget.
It is this feature that has led some observers to refer to current scoring procedures as “static.” Unfortunately, this label has caused certain critics to mistakenly conclude that current procedures do not recognize any of the incentive effects of legislation; i.e., that firms, workers, investors, and households continue their economic lives as if nothing had changed. Nothing could be further from the truth.
For example, during my tenure at the Congressional Budget Office (CBO) the CBO scored the impact of the Medicare Modernization Act (MMA). To do so, the staff necessarily had to incorporate the decision of firms to offer insurance contracts for the cost of outpatient pharmaceuticals and bid for customers, the willingness of seniors to purchase such insurance, changes in the amount of drugs prescribed and purchased, take-up of low-income subsidies, and myriad other decisions by households, firms, and governments. However, in keeping with current practice, the overall level of gross domestic product and national income was assumed to be unchanged.
Dynamic scoring expands the range of economic impacts to include the pace of economic growth—that is, it would involve explicitly estimating the change in the aggregate level of economic output and income, and incorporating estimates of any second-round effects of these changes on budget aggregates. This has some desirable features. In estimating the impact of the legislation, analysts would (a) consider the direct impacts on program costs and tax receipts; (b) evaluate the effects on incentives to work, save, invest, and generally conduct economic affairs; (c) estimate the resulting change in the overall level of economic activity; (d) compute the impact of this higher or lower level of economic activity on program costs and tax receipts; and (e) calculate the net impact of the legislation on the unified budget. The key difference is step (d), which is in turn built upon (c).
A virtue of dynamic scoring is that it extends analysis of budget policy to include economic policy dimensions. Specifically, dynamic scoring requires that analysts incorporate into their evaluation of legislation all of the economic feedback at the individual, household, firm, and national level. For this reason, it has the potential to distinguish between those policies that are equal in their budget cost, but very different in their overall economic incentives. Indeed, one of the most attractive aspects of dynamic scoring is its promise of allowing policymakers to distinguish between economically efficient policies that promote growth, and those that work to reduce the living standards of future generations.
THE USE OF DYNAMIC SCORING
Should it be used all the time? No. If you look at the legislation that the CBO and the Joint Committee on Taxation (JCT) scores, they include renaming post offices, swapping one parcel of land for another, energy-efficiency standards for light bulbs, and a gazillion other things that have no chance of affecting aggregate growth. Dynamic scoring should be reserved for large proposals that are clearly going to have significant impact.
The House of Representatives’ current rule for dynamic scoring—which was embraced in the Senate when it passed the Fiscal 2016 budget resolution—follows this dictum, requiring a dynamic score only for “major” legislation, which is defined to be a budget impact of 0.25 percent of GDP (currently just under $45 billion) in any fiscal year. It builds on the statutory requirement that the JCT and CBO provide cost estimates of all tax and mandatory spending bills (the CBO does not formally score appropriations bills).
Dynamic scoring is neither new nor voodoo. Every budget submitted by the Administration was scored dynamically by the Office of Management and Budget (OMB)—the budget estimates assume that the president’s proposals are implemented and produce the desired effect on macroeconomic growth. President Obama is a dynamic scorer. President Bush was a dynamic scorer. And so were his predecessors.
Dynamic scoring is not new to Capitol Hill. The CBO has accounted for macroeconomic effects in its analyses of the President’s budget since 2003, permitting an apples-to-apples comparison to the Administration’s budget. The CBO and JCT also took into account macroeconomic effects when scoring Senate immigration bills in 2006 and 2013. The JCT has had a macroeconomic modeling staff for over 15 years and most recently published dynamic estimates of the tax reform draft released last February by former Ways and Means Committee Chairman David Camp.
Similarly, the CBO included in its recent analysis of the impact of increasing the minimum wage the macroeconomic effects of a change in aggregate demand due to the shift of income from higher-income individuals to lower-income people. It notes “there is a two-thirds chance that the effect of the $9.00 option would be in the range between a very slight increase in the number of jobs and a loss of 200,000 jobs. If employment increased under either option, in CBO’s judgment, it would probably be because increased demand for goods and services (resulting from the shift of income from higher-income to lower-income people) had boosted economic activity and generated more jobs than were lost as a direct result of the increase in the cost of hiring low-wage workers.”
Alternatively, note that raising the minimum wage to $10.10 option reduces the income of higher-income individuals by $17 billion. In the absence of dynamic effects, that means that something less (because of distortion costs) than $17 billion would be transferred to lower-income individuals. Instead, the CBO shows an increase of $19 billion, presumably the impact of dynamic scoring.
The bottom line is that dynamic scoring has already been used and had a significant impact on policy analyses.
Dynamic scoring is also used around the globe. As noted by Greg Ip “The Dutch equivalent to the CBO, the CPB Netherlands Bureau for Economic Policy Analysis, evaluated the effect on growth and revenue of a 2001 tax reform that lowered the top income tax rate. Britain’s counterpart, the Office for Budget Responsibility, has incorporated the effects of a lower corporate income tax and higher value-added tax in its forecasts. When the IMF advises recipients of its loans to revamp taxes, social programs and regulations, it calculates the effect of such reforms on the country’s economy and debt burden.”
The reality is that dynamic scoring is a well-accepted, common practice among budget professionals.
DYNAMIC SCORING IS NOT A FORECAST
It is equally important to recognize what it is not. It is important to emphasize that this is scoring and not forecasting. In the Congressional scoring process, all proposals are compared to the same baseline, which is fixed in March of each year. No forecaster would ever prepare her November forecast jumping off from the state of the economy and policy months earlier.
This is precisely what is done in the budget process because the key to scoring is that you do it exactly the same way for all proposals, so that you get them ranked correctly. Consider the following analogy. In football, a team gets six points for a touchdown, three points for a field goal, two points for running or passing the extra point and one point for kicking the extra point. Why? I have no idea. But because the scores are computed in the same way, they can be compared across teams in a game, across games in the league, and over time. Notice that if the scoring gods revealed to us that we had been inflating touchdowns, which were really only worth five points, the system would still work.
The same is true for dynamic scoring. There would be lots of difficult decisions to be made on exactly how to do it. And, inevitably, forecasting errors would lead some to turn out to be higher or lower than the actual impact of the law. But a disciplined approach would get the average about right and, more importantly, provide a consistent ranking of alternatives.
The final aspect of scoring is that it is a judgment exercise. There does not exist a formal model for every proposal Congress presents to the JCT and CBO—our elected representatives are a clever bunch. So the image of scoring, including dynamic scoring, as simply turning the crank on a model and having it spit out a number is just misplaced. Instead, the professionals at the CBO and JCT take into consideration a wide variety of evidence, including formal models where possible and appropriate, and draw a judgment about the likely budgetary impact.
UNCERTAINTY AND DYNAMIC SCORING
Opponents of dynamic scoring often argue that it opens a Pandora’s box of guesswork and uncertainty because there is no consensus model of the macroeconomy. The reality is quite different. To begin, there is nothing inherently more uncertain or speculative compared to traditional scoring. Indeed, uncertainty is part and parcel of all scoring.
During my tenure at the CBO, the office scored the Terrorism Risk Insurance Act. There is no consensus model that guides one to evaluate the budgetary consequences of a federal financial backstop for the consequences of a future terrorist attack at an unknown time (next year, next decade, ...), of an unknown nature (chemical, nuclear, biological…) and in an unknown location (Chicago, Nashville, …). Using a variety of approaches and considered judgment, the CBO produced a score.
Similarly, the CBO was asked to evaluate the budget costs of providing an additional $100,000 death benefit for those killed in combat, prior to the invasion of Iraq. Needless to say, the scoring was fraught with uncertainty. Notably, the CBO had to score the new Medicare Part D drug bill when no such product (insurance against the financial costs of outpatient prescription drugs) existed in nature.
It is a reality of the process that the more innovative the proposal, the less likely that there is a broad and deep research literature on which to base the scoring. As a result, uncertainty and difficult analyses are par for the course in the scoring business and a move to dynamic scoring would not change things significantly.
One thing that would happen, however, is a beneficial research dynamic (pun intended). The need to generate dynamic scores and the natural debate over the actual scores will prompt the research community to analyze the underlying policies more carefully. In the process, the research literature will expand precisely in those areas where it would be most useful to the CBO and JCT.
Skeptics of dynamic scoring often assert that its use will dramatically swing the projected costs of tax or mandatory spending proposals and, thus, undermine budgetary integrity. This concern seems out of proportion to the reality. As noted above, relatively few proposed laws would merit a dynamic score. And—sadly, in my opinion—Congress does not reflexively migrate toward disciplined, pro-growth policies. A reasonable expectation for a pro-growth dynamic score is that the growth rate of the economy could be 0.1 to 0.3 percentage points higher. It will matter, but will not relieve Congress of the need to make tough budgetary choices.
SOME MECHANICS OF DYNAMIC SCORING
Settling on the final values of a dynamic score will ultimately be an exercise in judgment. However, that judgment should be informed by the results of formal models of the macroeconomic impacts. This raises some important, practical considerations regarding dynamic scoring.
The scale of the analysis involved in preparing baseline budget projections points to the first problem with wholesale adoption of dynamic scoring: time. It is inevitable that statutory language continues to evolve throughout the legislative process: committee deliberation and reporting, floor amendments and votes, and conference committee negotiations. Often there is a need for very quick and timely scoring information. The scale of a dynamic scoring effort may be in conflict with this need. To alleviate this constraint, the relevant committees will have to stay in constant communication with the JCT and CBO in order to ensure that they have done the preparation necessary to produce a dynamic score.
Adopting a Single Approach for Estimates
A practical difficulty with dynamic scoring has been the absence of a single, consensus approach to the estimates. The attraction of dynamic scoring is its ability to reveal the impact of legislation on economic growth. However, this impact depends crucially on the overall foresightedness of U.S. households and firms. To take an extreme case, imagine legislation that cuts all marginal tax rates by five percentage points, with the cut to take effect five years from now, but sunset (that is, expire) ten years in the future. If people are extremely myopic, this policy has no impact on incentives to work, save, or invest and there is no dynamic feedback. If they are moderately forward-looking, they may anticipate lower taxes and respond to these incentives. If they are even more forward-looking, they will recognize both the tax reduction and the subsequent rise. As a result, they will work especially hard during the intervening years—yielding a larger increase in output, income, and taxes—with a sharper decline when taxes rise again.
One approach to this problem, exemplified by the CBO’s macroeconomic analysis of the president’s budget proposals, is to provide a variety of estimates, each corresponding to a different degree of foresight. However, a formal scoring process requires a single set of estimates. Thus, at the outset of its work it is necessary that a process be settled upon to consolidate the insights from several approaches into a single reported score. This should be relatively straightforward; the JCT and CBO face this as part of traditional scoring exercises.
Balancing the Budget
The example sketched above highlights another issue in the conduct of dynamic scoring: the need for an “offsetting policy.” Over the long-term, if individuals have foresight, then government debt (relative to the economy) must stabilize. Legislative proposals that upset this requirement by increasing spending or reducing taxes (at least relative to their impact on economic growth) will produce debt that will grow explosively. Similarly, spending cuts or tax increases (relative to their impact on the economy) will cause debt to spiral down. Since the government can neither borrow nor save unboundedly large amounts, it is necessary to put a stop to either spiral by introducing an offsetting budget policy at some point in the future.
The choice of offsetting policy—spending increases or decreases and the pace at which they take place, tax reductions or increases and their timing, or some combination of these—will have differential effects on the behavior of individuals and firms and influence the score. Since a primary objective of scoring is to treat all legislative proposals equally, it will be necessary to pick a single type of offsetting policy, a set time in the future to implement the offset, and use it for all proposals.
An equally important—but often overlooked—aspect of this problem is getting the debt stabilized to begin the analysis. Some approaches to dynamic scoring, particularly forward-looking growth approaches, simply will not work (i.e., the computer algorithms will not function) when the government budget is on an explosive debt trajectory. (Sadly, the federal budget is on such a trajectory.) Thus, even to begin the work of analyzing tax reform it would be necessary to assume an answer to the basic task of stabilizing the debt.
Supply-side versus Demand-side Dynamics
Another challenge in implementing dynamic scoring is the degree to which the score reflects only supply-side growth, or also includes demand-side cyclical influences. Broadly speaking, economies grow in one of two ways. Supply-side growth occurs when there is an increase in the capacity to produce goods and services though the addition of greater labor supply (labor force participation, hours worked, higher effort per hour, greater skills per worker, better efficiency in the use of labor effort and skills, and so forth), greater physical capital (more or better equipment, software, buildings, and so forth) and improved technical prowess (new technologies or superior organization and management). These responses are at the heart of pro-growth tax policies.
Demand-side growth (or contraction) reflects business cycle fluctuations and the extent to which existing labor supply, capital, and technical prowess are utilized. These are also key factors of economic growth.
As noted above, there will be a need to settle on a single dynamic score. In light of the need for growth of both types to be incorporated into the analysis, it will require adding business-cycle considerations to growth-style modeling approaches. The ultimate size, direction, and character of demand-side effects of fiscal policy changes depend as well upon the assumed path of monetary policy. In a manner similar to offsetting budget policies, it would be necessary to make assumptions regarding the response of monetary policy to the legislative changes.
Advocates and detractors of dynamic scoring need to stop hyperventilating. Dynamic scoring is a sensible thing to do in situations where the policy changes are large enough for the Congress to care about the growth implications—tax reform, Social Security reform, immigration reform, etc.—and it can be done in a systematic and professional fashion. The adoption of dynamic scoring presents challenges, but those challenges are surmountable, and its incorporation in the budget process will present policymakers with better, albeit imperfect, information.
DOUGLAS HOLTZ-EAKIN is President of the American Action Forum, 1747 Pennsylvania Avenue NW, 5th Floor, Washington, DC 20006 (e-mail: firstname.lastname@example.org)
Congressional Budget Office (CBO). (2006). Cost Estimate for S. 2611, Comprehensive Immigration Reform Act of 2006, May 16.
Congressional Budget Office. (2013). The Economic Impact of S. 744, the Border Security, Economic Opportunity, and Immigration Modernization Act, June 18.
Congressional Budget Office. (2014). The Effects of a Minimum-Wage Increase on Employment and Family Income, February 18.
Ip, G. (2015). Dynamics Scoring: A Potential Super Model. The Wall Street Journal, March 25.
Joint Committee on Taxation (JCT). (2014). Macroeconomic Analysis of the Tax Reform Act of 2014, February 26.
See CBO (2006) and CBO (2013).
See Joint Committee on Taxation (2014).
See CBO (2014), Appendix A.
See Ip (2015). Available at http://www.wsj.com/articles/dynamic-scoring-a-potential-super-model-1427307531.