Statement of
Douglas Holtz-Eakin
Director

The Tax Code's Impact on
the Reliability of Revenue Projections

before the
Committee on the Budget
U.S. House of Representatives

July 22, 2004

Mr. Chairman, Congressman Spratt, and Members of the Committee, I am grateful for the opportunity to appear before you this morning to discuss comprehensive tax reform. I want to begin by emphasizing that my remarks focus on the federal budget, with emphasis on the problems that the current tax system poses and that the reform of the system would pose for projecting revenues. These remarks do not extend to the merits or drawbacks of specific proposals.

From an economic perspective, the central role of the government is to provide services to the public that only it can provide (or provide adequately). Therefore, the threshold budgetary decisions faced by policymakers are about which services to provide and how much to spend on them. Those expenditures must be financed, and the challenge for tax policy is to provide adequate financing in a fair and efficient fashion. At present, the bulk of general tax revenues to the federal government derives from individual income taxes and corporate income taxes. In recent years, the combined contribution from those income taxes has been about 50 percent to 60 percent of federal receipts, or about 8.5 percent to 12.5 percent of gross domestic product. Payroll taxes constitute another roughly one-third of receipts. But under current law, those taxes are linked to outlays for Social Security and Medicare. For that reason, I will ignore payroll taxes, although one could envision a comprehensive tax reform that encompassed them. The remaining sources of receipts are less than a tenth of federal revenues. In light of the dominance of income taxes as a share of federal receipts, discussions of tax reform are largely about reforming income taxes.

Its name notwithstanding, the underlying base of the U.S. income tax system departs significantly from definitions of income. Indeed, a chief difference among many reform proposals is the question of the appropriate tax base. There are two useful benchmarks. One is a tax on comprehensive income (often referred to as a Haig-Simons income tax). Comprehensive income includes all labor compensation earned during the year (regardless of whether it is actually paid or deferred) and all capital income (again, regardless of whether it is realized or not). Comprehensive income measures the additions to an individual's potential to purchase consumption items. Actual purchases may fall short of that amount if he or she instead chooses to save part of the income, or purchases may exceed that amount if he or she spends out of past or expected future income.

The other benchmark is a consumption tax, or consumed-income tax. It differs from a comprehensive income tax by excluding from the tax base that portion of income that is saved and by subjecting to taxation an individual's purchases. Because comprehensive income and consumed income differ only as a result of individuals' saving and investment decisions, it is possible to implement a consumed-income tax by two means: by excluding saving from taxation or by excluding the income generated from saving and investment. Both variants have been proposed.

The current U.S. income tax system is a hybrid of those two benchmarks, albeit an incomplete one that excludes from taxation some items that would be taxed in either system. The current system relies principally on an income tax but embodies some elements of a consumption tax by excluding income saved for retirement and other uses. Some of the problems that arise with the tax system--both in terms of its effects on the economy and the difficulties it presents for budget projections--stem from that mixing of characteristics of income and consumption taxes. Many more difficulties derive from the ways in which the tax system departs from both concepts. For example, employer-provided health insurance, which is currently not taxed, would be taxed under either a pure consumption tax or a pure income tax, as would many of the expenditures that taxpayers are allowed to deduct in arriving at taxable income. Consequently, although some of the issues I discuss here might be addressed by the choice of income versus consumption as the tax base, many will arise in either case, because they are either unavoidable or the consequence of various policies that are independent of the choice of a base.

I will focus on four issues. The first three are general attributes of the current income tax system that impede reliable revenue projections: sunsets and expiring provisions, the volatility of income sources, and differential taxation. The fourth is transition effects--a consideration that must be taken into account in projections when any major change to the system occurs. Clearly, the most important consideration in evaluating the merits of a tax reform is its effects on the economy, and not the repercussions for federal revenue projections. However, knowing the effects on projections illuminates the real costs to the economy because the two types of effects often stem from the same phenomena.
 

Sunsets and Expiring Provisions

Under current law, the tax code contains a number of new provisions that take effect in phases and a number of expiring provisions. A few of the latter--notably the provision allowing partial expensing of equipment--were enacted to be temporary with a specific economic purpose in mind. The countercyclical effect of investment incentives such as partial expensing was enhanced by its temporary nature. Indeed, its future use as a tool to manage aggregate demand in the economy could be compromised by now extending it. But most of the expiring provisions were enacted as a step toward making them permanent or extending them indefinitely.

There are essentially two groups of expiring provisions. First, there are the roughly 30 so-called extenders: special provisions of the code that have been extended a few years at a time and are regularly expected to be continued (see Table 1). Most, if extended, would diminish revenues--altogether, by about $175 billion over the Congressional Budget Office's (CBO's) 10-year projection period. Among the most significant of them is the credit for research and experimentation, which has been extended nine times since its enactment in 1981.

Second, there are the many provisions in the Economic Growth and Tax Relief Reconciliation Act of 2001 (EGTRRA) and the Jobs and Growth Tax Relief Reconciliation Act of 2003 (JGTRRA). They include all of EGTRRA's income and estate tax provisions, which are set to expire at the end of 2010, and JGTRRA's reduced dividend and capital gains tax rates, at the end of 2008. But they also include provisions such as the child tax credit, the expanded standard deduction, the 10 percent and the 15 percent brackets, and the increased exemption under the alternative minimum tax, which are scheduled to be reduced in 2005 (see Table 2). While people hold diverse views on the future of those provisions, it is reasonable to expect that at least some will be extended.

Those scheduled changes in the tax code create a multitude of challenges for projecting receipts. In projecting baseline receipts, CBO is required to assume current law, a stricture that might normally coincide with "current policy." However, at present, that requirement dictates that all of the legislated provisions are assumed to be phased in and to expire as scheduled in law, even though those required assumptions may differ greatly from expectations of current policy. Thus, even when CBO's baseline projections are accurate, the presence of those expirations reduces transparency about just what to expect concerning the future course of the government's fiscal condition.

An even more problematic feature of the gap between current law and expectations of policy is that budget estimates are based on the former and economic activity is influenced by the latter. In projecting revenues, one strategy for dealing with the problem is to assume that the public believes that all changes to the tax code will occur as scheduled. That approach has the virtue of being consistent with the baseline assumption of current law. Unfortunately, it has the drawback of divorcing the underlying assumptions about the economy from reality.

Assuming that the public expects the tax code to change as scheduled also complicates revenue projections. Expected changes in the tax code induce significant shifts in taxable economic activity from year to year. Some of that activity represents changes in behavior that lead to real changes in income, prices, interest rates, and other aspects of economic performance. Some is just reporting or accounting--shifting bonuses into the next or previous year, for example. Those shifts--real or reporting--lie at the heart of the desirability of tax reform; the fact that households and firms alter their behavior in response to taxes demonstrates the importance of tax policy. But such shifts in measured activity resulting from the scheduled changes in the tax code make projections more sensitive to policy than they would otherwise be--and do so unnecessarily if, indeed, the code ultimately does not actually change.

Another way to deal with the problem of scheduled changes in the tax code in making projections is to assume that firms and individuals do not believe that tax provisions will change as scheduled but have their own view of how policy will evolve. That approach may make the underlying economic projections more realistic, but it renders them inconsistent with the rule that budget projections reflect a baseline under current law.

And here again the existence of scheduled changes in the tax code complicates projections. The approach makes it necessary to discern what the public expects and to factor in the effects of the public's uncertainty. EGTRRA and JGTRRA contain many provisions that interact in a complicated manner, and predicting which provisions the public expects to lapse and which it expects to be made permanent is a complicated matter of judgment. Regardless of which approach is taken, baseline revenue projections are made less reliable by the existence of expirations that few people expect to occur as written in current law.

Those difficulties for revenue projections are reflections of the costs to the economy introduced by expiring provisions. When households and firms shift the timing of their economic activity for tax purposes instead of market fundamentals, the efficiency of economic performance is impaired. Advancing or delaying the recognition of income and expenses uses real resources. And although it may be worth it to a taxpayer to incur those costs in order to save on taxes, the resources used are lost to society with no offsetting gain. Similarly, the uncertainty caused by expiring provisions for revenue projections mirrors the uncertainty that they introduce into all economic affairs. That uncertainty imposes real costs in the economy as well, as resources are used to avoid it and as taxpayers' choices are influenced by it.

And those costs from expiring provisions are not limited to explicit sunsets. Even without provisions that are scheduled to be phased in or to expire, frequent changes in the tax code inflict analogous costs. If EGTRRA were not scheduled to expire or if it had not been enacted, similar uncertainty about taxes could exist. The public has incentives to evaluate policymakers' propensity to make changes to the tax code and to adjust its economic behavior accordingly. And the perception itself, even if nothing is actually changed, is enough to impose costs on the economy. Those costs can be reduced only by allowing the system some repose.

The propensity to change the tax system is reinforced in part by the many exceptions and preferences already built into it. Its mixed attributes of income tax and consumption tax bases and the many exceptions and preferences that cause the system to deviate from both concepts mean that there is no bright line dividing new provisions that would be consistent with the tax system and those at odds with its underpinnings. One special interest claim for preferential treatment may seem as legitimate as any other, making changing the tax code easier and leading to further instability in it.
 

The Volatility of Tax Bases

Any comprehensive tax system will include some tax bases that are inherently more volatile than others. The current income tax embodies relatively stable sources of income such as regular wages and salaries, which yield relatively stable receipts. However, other sources of income swing more widely, which imparts volatility to receipts and makes projecting revenues more difficult. The current tax base, for example, includes capital gains, which are extremely volatile and unpredictable. Even though capital gains are taxed at substantially lower rates than most other income is, they nonetheless produce large swings in revenues. For example, as a consequence of a major adjustment in the stock market, receipts from gains realizations fell from $100 billion in fiscal year 2001 to $57 billion in fiscal 2002. Because stock prices are impossible to forecast, the decline in gains was not foreseeable. And even after the stock market's behavior became evident, the impending change in receipts was highly uncertain because stock prices do not translate directly into gains realizations.

Capital gains are taxed under the current tax system because they are a form of income. However, under a comprehensive income tax, gains would be taxed as they accrued instead of upon realization. That change would not alter the fundamental difficulty of projecting fluctuations in the stock market, but it would eliminate the necessity of predicting taxpayers' decisions to realize capital gains. In contrast, neither variant of a consumed-income tax would tax gains; therefore, both presumably would avoid that source of volatility and unpredictability. For much the same reason, both variants of a consumed-income tax would tend to avoid the volatility that characterizes profits in general.

However, volatility is not necessarily more characteristic of one approach to tax reform than another. Instead, the details of implementation matter. While consumption is less volatile than income, most proposed tax systems retain some volatile sources in their tax bases. Those that exempt saving (sales and value-added-style taxes) include durable consumer goods in their base, for example, which imparts volatility. Proposals that exclude income from capital still have volatility from bonuses and stock options, which are one of the sources of volatility under the income tax. Moreover, under some consumption taxes, business income is taxed net of full expensing for capital outlays. Because aggregate investment can be more volatile than business profits, the business income tax base under such a tax can be at least as volatile as it is under an income tax.

In addition, volatility within a year may also be a consideration. Currently, most tax liabilities are paid as they accrue. Because wages are such a large part of income, withholding serves to match payments closely with the activity that generates them. Other forms of income should result in payments of estimated taxes through the year, but those liabilities can be harder to determine.

Consequently, final settlements of liability are necessary at the time returns are filed. The current tax system results in "April surprises" that would not arise with some other tax systems, such as a wage tax or sales tax. Some consumption taxes, such as those for which the tax is collected at the point of sale, would reduce those surprises. Other consumption taxes, such as those that permit a deduction from income of contributions to qualified savings accounts, might be more difficult to pay as liabilities accrued, leading to potentially larger April surprises.
 

Differential Taxation

The existing tax code taxes different kinds of income at different effective tax rates (including, in many cases, a rate of zero). For example, in the aggregate, the effective tax rate on wage income (from the individual income tax) is about 20 percent. It would be lower if calculated on compensation, which includes untaxed fringe benefits. Yet capital gains income is taxed at about 15 percent. Corporate income tax rates also vary, with the return on investment in equipment now taxed at about half the effective rate of that on structures. The combined corporate and individual income taxes result in tax rates of about 25 percent to 30 percent on corporate income (which may be subject to both taxes), less than 20 percent on noncorporate business income (which is subject only to the individual income tax), and zero for the implicit income generated by owner-occupied housing (which is subject to neither).⁽¹⁾

As a consequence of differences in effective tax rates, even an accurate forecast of overall income can yield an inaccurate projection of receipts if it is not allocated correctly among various activities or individuals. The variation occurs across two dimensions. First, there are different tax rates depending on the different kinds of income and consumption. Compensation in the form of wages and salaries is subject to taxation, while fringe benefits, which serve the same purpose of compensating workers for their labor, are not taxed. Profits accruing to C corporations--but not those of proprietorships, partnerships, or S corporations--are subject to the corporate income tax. Capital gains are subject to lower rates and sometimes no taxation at all. Similarly, taxpayers' patterns of expenditures generate different effective tax rates. For instance, expenses for mortgage interest, charitable contributions, or state and local taxes reduce federal tax receipts. As a result, projections of receipts depend on projections of the components of income and expenditures, each introducing yet another potential source of error in receipts projections.

The corresponding economic effect of that more complicated and less accurate budget projecting is the economic inefficiency associated with those differential tax rates. Firms and workers have incentives to tilt compensation toward untaxed fringe benefits. Capital flows toward lower-taxed sectors of the economy, forgoing higher gains to society that would accrue if invested elsewhere. Such incentives distort economic activity and lower overall efficiency. In addition, the preferences require higher rates to be applied to other forms of income, exacerbating the inevitable disincentives that income taxes cause for work effort. The effect is real and potentially substantial; incomes are lower than they would be without the distortions caused by the differential tax rates.

The second dimension is who earns the income. Some owners of capital--notably tax-exempt organizations--pay no taxes, so, overall, income from interest and dividends yields much lower revenues than profits and wages do. In addition, the progressive structure of the tax rates means that the effective tax rate depends on whether the income accrues to someone of higher or lower income. Consequently, to project receipts, it is necessary to estimate how much income accrues to tax-exempt entities and to forecast the income distribution among taxpayers.

Again, the budgetary challenge reflects an underlying challenge posed by the tax code. In addition to minimizing interference with economic decisions, exhibiting fairness is another metric by which a tax code is measured--often manifesting itself in the desire to tax higher levels of income at higher rates. However, as the desired progressivity of the tax code increases, a premium is placed on accurate projections of the distribution of income. For example, the share of taxable income taxed in the highest marginal rate bracket (39.6 percent) under the personal income tax rose from 12 percent in 1994 to 25 percent in 2000; so taxes were almost $60 billion higher in 2000 than they would have been had the share taxed in each bracket remained constant (not including the effects of capital gains). While some forecasts may have anticipated an increasing concentration of incomes between those years, few would have anticipated such a large change.
 

Transition Effects

The previous three considerations point to how the current tax regime makes projecting receipts more difficult. My fourth point builds on the fact that all projection methods rely on history to a certain extent. And the current system has some track record. In contrast, any major reform involves major uncertainties about how the new tax system will function and how much it will yield. As a result, following the introduction of any significantly reformed system, revenue projections would probably become less reliable in the near term.

Again, the problems introduced for projections are reflections of factors at work in the real economy. The desirability of moving to a new tax code cannot be determined just by comparing what the world looks like now with what it would be like under a different code. The desirability of the reform also depends on the process of getting there. And the costs of the change could be significant. There are costs to reallocating resources to their best employment. In addition, some tax reform proposals, because of the effects they would have on prices, would create challenges for macroeconomic stabilization policy. Moreover, transition costs may generate perceptions of unfairness that lead reform proposals to incorporate relief to those most affected by the transition. But because such transition relief may mitigate the efficiency gains from reform, any tax reform proposal must be evaluated in light of the possible problems created by the transition from the old tax regime.
 

Conclusion

Budget planning is aided by accurate projections of receipts, but the current U.S. tax code has several features that make accurate projections difficult. Of course, ultimately the issue of tax reform is not about what the tax system or tax reform does to analysts' ability to project revenues. It is about what taxes do to the economy. Nevertheless, some of the current challenges in making projections reflect the underlying costs imposed by the tax code.

Table 1.
Tax Provisions That Expire Before 2014 (Other than EGTRRA and JGTRRA), Except Income Tax Provisions First Enacted Since 2001
(Billions of dollars)
Tax Provision		Expiration Date		Total, 2005-2014
Archer Medical Savings Accounts		12/31/03		*
Brownfields Remediation		12/31/03		-2.1
Corporate Contributions of Computers to Schools		12/31/03		-1.7
Credit for Electric Vehicles		12/31/03		*
Credit for Electricity Production from Renewable Sources		12/31/03		-8.7
Deductions for Clean-Fuel Vehicles and Refueling Property		12/31/03		-2.8
Net Income Limitation for Marginal Oil and Gas Wells		12/31/03		-0.5
Qualified Zone Academy Bonds		12/31/03		-0.6
Reduction in Policyholder Dividends for Insurance Companies		12/31/03		-0.4
Rum Excise Tax Revenue to Puerto Rico and the Virgin Islands		12/31/03		-0.8
Tax Incentives for Investment in the District of Columbia		12/31/03		-1.5
Treatment of Personal Credits Under AMT		12/31/03		-42.1
Welfare-to-Work Tax Credit		12/31/03		-0.7
Work Opportunity Tax Credit		12/31/03		-3.7
Credit for Research and Experimentation		6/30/04		-58.2
Abandoned Mine Reclamation Fees		9/30/04		2.5
Depreciation for Business Property on Indian Reservations		12/31/04		-3.6
Indian Employment Tax Credit		12/31/04		-0.8
IRS User Fees		12/31/04		0.4
Authority for Undercover IRS Operations		12/31/05		**
Puerto Rico Business Credits		12/31/05		-15.2
Transfer of Excess Assets in Defined-Benefit Plans		12/31/05		0.3
Andean Trade Preference Initiative		12/31/06		-0.3
Depreciation for Clean-Fuel Automobiles		12/31/06		-0.0
Subpart F for Active Financing Income		12/31/06		-28.9
Generalized System of Preferences		12/31/06		-5.3
Alcohol Fuels Income Credit		12/31/07		*
FUTA Surtax of 0.2 Percentage Points		12/31/07		9.9
New Markets Tax Credit		12/31/07		-4.6
Empowerment and Renewal Zones		12/31/09		-8.1
	Total			-177.5
Sources: Joint Committee on Taxation; Congressional Budget Office.
Notes: * = between -$50 million and zero; ** = between zero and $50 million; AMT = alternative minimum tax;IRS = Internal Revenue Service;FUTA = Federal Unemployment Tax Act.
These estimates assume that the expiring provisions are extended immediately rather than when they are about to expire and that provisions that have already expired are reinstated immediately and retroactively. The provisions are assumed to be extended at the rates or levels existing at the time of expiration. They do not include expiring excise taxes dedicated to trust funds, which are assumed to be extended permanently in CBO's baseline under budget rules. The estimates do not include debt-service costs; they do include effects on outlays for refundable tax credits.