Abstract
The federal tax code provides about $500 billion each year in incentives intended to encourage socially-valued activities, including homeownership, charitable contributions, health insurance, and education. The vast majority of these incentives operate through deductions or other approaches that link the size of the tax break to a household’s marginal tax bracket, which means that higher-income taxpayers receive larger incentives than lower-income taxpayers. Such an approach is often appropriate for provisions, such as deductions for business expenses, designed to measure income or ability to pay. But such an approach for incentives intended to promote socially-valued activities excludes more than a third of America, and misses an important opportunity to increase efficiency and economic growth.
We propose a dramatic change in how the government provides these tax incentives, which could be done on a revenue-neutral basis. Under our proposal, the default for all tax incentives intended to promote socially-beneficial behavior would be a uniform refundable tax credit, which would be available to qualifying households even if they owe no income tax. These tax credits would provide a much more even and widespread motivation for socially-valued behavior than the current set of tax incentives, and could help smooth out fluctuations in household income and macroeconomic demand, all of which would improve economic efficiency.
REFUNDABLE TAX CREDITS VERSUS OTHER FORMS OF TAX INCENTIVES
Policymakers have created tax incentives for homeownership, retirement saving, education, and medical expenses. Other tax incentives seek to promote work, charitable giving, and investment in life insurance, annuities, and state and local bonds. Together, these tax incentives reduce federal tax revenues by about $500 billion a year, or roughly 4 percent of Gross Domestic Product. Structuring these tax incentives most efficiently is therefore an immensely important policy matter.
Approximately $420 billion of these existing tax incentives operate through deductions, exemptions, or exclusions. Such tax incentives tie the size of the tax break to an individual’s marginal tax bracket: A deduction of $1, for example, is worth 35 cents to someone in the 35 percent marginal bracket but only 15 cents to someone in the 15 percent marginal bracket. Such incentives thus provide relatively weak incentives to those in low tax rate brackets. Furthermore, these types of tax incentives fail to reach the increasingly significant share of low and moderate-income individuals and families who do not have any federal income tax liability to offset in any given year. More than 35 percent of households during any given year have no income tax liability; these households are home to almost half of all American children.
Refundable tax credits represent a different approach. Since they are a credit, rather than a deduction or exclusion, they do not depend on a household’s marginal tax bracket. A tax credit of $1, for example, reduces taxes by $1 and thus is worth the same to households in the 35 percent bracket or the 15 percent bracket. And since they are refundable, they provide benefits to all tax filers, regardless of whether they owe income taxes on net.
THE GROWTH OF REFUNDABLE TAX CREDITS
Currently the tax code contains three main refundable tax credits: the Earned Income Tax Credit, the Child Tax Credit, and a small health insurance credit. The Earned Income Tax Credit (EITC) is the largest anti-poverty program for the nonelderly in the country. In inflation-adjusted terms, the budgetary cost of the EITC has risen by a factor of nine since it was enacted in 1975, and it tripled between 1990 and 2000 alone. More recently, the partially refundable Child Tax Credit (CTC) and the fully refundable health insurance credit were enacted, and the refundability of the CTC was expanded and accelerated.
Several factors likely contributed to the dramatic growth of refundable credits over the past three decades (see Figure 1). For example, policymakers have increasingly relied on the tax code rather than direct government expenditures to subsidize households and influence their behavior as a result of perceived or real incentives within the tax legislative process, a development that has supported the rise of refundable credits.
THE CASE FOR UNIFORM REFUNDABLE CREDITS
Despite the growth in refundable credits, most tax incentives intended to promote socially-beneficial behavior take the form of deductions or other approaches linked to marginal tax rates. Yet if policymakers want to broadly promote socially-valued behavior through the tax code, refundable credits are generally necessary. As illustrated in Figure 2, in any given year more than one-third of households do not have any federal income tax liability. About a quarter of tax units file a tax return but have no income tax liability, and another 13 percent do not file. Moreover, almost half of all children, and 80 percent of children in single parent households, live in tax units with no income tax liability in any given year.
As a result, if policymakers want to create incentives through the individual income tax for all or most tax units to engage in certain behavior every year, such as saving or obtaining education for themselves or their children, refundability should not only be considered an acceptable instrument of tax policy – it is imperative.
Furthermore, unless there is evidence that certain households are more responsive to the incentive than others or generate larger social benefits from engaging in the activity, tax incentives are most efficient if they provide the same incentive to all households – and that can only be accomplished in a straightforward manner through a uniform (and refundable) credit.
The reason that a uniform incentive is the most efficient approach in the absence of evidence regarding differences in responsiveness or social benefits is that a small number of large mistakes in under- or over-subsidizing an activity are more costly in efficiency terms than a large number of small mistakes. For example, imagine that certain behavior, perhaps charitable contributions, on average generates 5 cents of social benefits per dollar contributed per year and policymakers have determined to subsidize contributions by, on average, 5 cents per dollar. Imagine further that there is a 50 percent chance that a dollar of contributions by a high-income household generates 10 cents of social benefits, while a dollar of contributions by a low-income household generates none, and a 50 percent chance that this pattern is reversed. A uniform subsidy of 5 cents would leave 5 cents of lost social benefits in both cases. Meanwhile, a subsidy of 10 cents given to one group would result in 10 cents of lost social benefits in one case and none in the other. The uniform subsidy is more efficient – it technically minimizes the “expected deadweight loss” – because a small number of big errors (one case of 10 cents) is more costly than a large number of small errors (two cases of 5 cents).
We acknowledge that many behavioral tax incentives may be bad policy regardless of whether they take the form of uniform refundable credits, perhaps because the behavior in question does not actually generate social benefits or because such social benefits are best addressed through direct government provision of the good or regulation. Even taking these limitations into account, however, assuming the continued existence of a tax incentive, our default structure is generally preferable because it minimizes the expected social losses from the tax incentive, regardless of whether the behavior actually is socially beneficial.
We also acknowledge that tax incentives should not provide the same incentive to all households in all circumstances. If there is evidence that the associated social benefits vary systematically by income class, or that different income groups exhibit different levels of responsiveness to the subsidy, the tax incentive should not be identical for all households. Indeed these differences between various income groups surely exist in reality. But when, as is frequently the case, the evidence on these issues is non-existent or inconclusive, the most efficient form for a tax incentive is a uniform refundable credit. The burden of proof should therefore be on those who prefer some other form of tax incentive to demonstrate that such deviations from a uniform refundable credit are justified by empirical evidence.
Thus, if policymakers wish to use the tax system to create incentives for certain socially-valued behavior, it makes no sense to exclude more than a third of American individuals and families from their reach, or to provide smaller benefits to some households than others, absent evidence that those Americans would be relatively unresponsive or that their behavior generates fewer societal benefits. Moreover, even when there is empirical evidence suggesting that the optimal tax incentive should not be the same for all households, the most efficient incentive is almost certainly still some type of refundable credit. It is extremely unlikely that there is a sharp break in social benefits or responsiveness to an incentive exactly at the point of no income tax liability, and these types of discontinuities are inherent in the application of all other basic forms of tax incentives.
The potential benefits of refundable credits are magnified further by a second feature: Their ability to help smooth household income. That is, during hard years, transforming existing tax incentives into uniform refundable credits would boost after-tax income, and thus help to cushion the blow of a drop in earnings, unemployment, or other hardships. Such income smoothing is desirable for several reasons. It can reduce the costs associated with economic instability and offset failures in insurance markets. It also allows families to plan their expenditures more confidently and avoids the additional costs (such as moving costs and credit card debt) of financing constant changes in household living standards. Income smoothing is particularly beneficial for lower-income households because they generally don’t have easy access to credit to make it through tough times, because they tend to have more volatile incomes than other families in general, and because income shocks can result in declines in their economic circumstances that persist over a long periods of time and are passed on to their children.
The final element of the case for uniform refundable credits is their ability to smooth the macroeconomy. Like household income smoothing, macroeconomic smoothing can enhance economic efficiency. In particular, macroeconomic demand fluctuations make it difficult for companies to optimize their investment and production functions, resulting in adjustment costs. These difficulties can inhibit domestic and foreign investment, which is correlated with economic growth. As a result, there is broad consensus in support of taxing and spending policies that are automatically countercyclical. Uniform refundable credits can help stabilize macroeconomic demand fluctuations by raising cash payments to families during recessionary periods, which then helps to boost spending – precisely the desired response during such periods.
A SPECIFIC EXAMPLE: RETIREMENT CONTRIBUTIONS TO 401(K)S AND IRAS
William Gale of Brookings, Jonathan Gruber of MIT, and Peter Orszag of Brookings have recently proposed a specific example that is similar in spirit to our broader policy suggestion. They note that current incentives for contributions to 401(k) plans and IRAs deliver their largest immediate benefits to higher-income individuals in the highest tax brackets.
Gale, Gruber, and Orszag would replace the existing tax deductions for contributions to retirement saving accounts with a 30 percent government matching contribution. Unlike the current system, workers’ contributions to employer-based 401(k) accounts would no longer be excluded from income subject to taxation, and contributions to IRAs would no longer be tax deductible. Furthermore, any employer contributions to a 401(k) plan would be treated as taxable income to the employee (just as current wages are). However, all qualified employer and employee contributions would be eligible for the 30 percent government matching contribution regardless of the employee’s income. This proposal would be roughly revenue neutral for the federal government, according to estimates from the Tax Policy Center microsimulation model.
This proposal provides a specific example of how a tax deduction or exclusion could be transformed, on a revenue neutral basis, into a uniform refundable credit. Our analysis generally supports this transformation as the default structure for retirement savings incentives, assuming such incentives are intended to promote the social benefits generated by retirement savings and that no other aspects of the tax code are intended to play that role.
OPPOSITION TO REFUNDABLE CREDITS
Opponents of refundable credits typically raise four main concerns. First, some question the extent to which government should engage in redistribution between different income groups. Second, some argue that the tax system should be used only to raise revenue, not to provide subsidies. Third, some believe that all Americans should pay at least some tax, even if just one dollar, as a duty of citizenship and so that they feel some stake in governmental decisions. Finally, some argue that refundable credits would increase administrative and compliance costs on net and are particularly subject to fraud and abuse.
Concerns about the extent of governmental redistribution, however, do not justify rejecting refundable credits that are enacted to enhance economic efficiency by subsidizing socially-beneficial behavior. And concerns about delivering subsidies through the tax system instead of the transfer system are generally objections to tax incentives overall, not to structuring tax incentives as refundable credits specifically.
The third objection – that all Americans should pay some tax – ignores the fact that most households claiming refundable credits pay a variety of federal, state and local taxes other than income taxes. Moreover, if one is interested strictly in federal income taxes, it seems likely that most refundable credit beneficiaries pay a positive amount of federal income tax over time as a result of the income variations that people tend to experience over their lives. Indeed, a simplified model of 2003 federal income tax law using data from the Panel Survey of Income Dynamics suggests that about three-quarters of tax units who are eligible for the refundable element of the EITC or CTC at some point during a 20-year period would nevertheless have positive net federal income tax liability over that period if historic earnings patterns are any guide. Thus, even if one accepts the principle that paying some income tax is necessary for feeling a stake in government decisions (which we do not), this principle would not necessarily preclude refundable credits once income tax liabilities are examined over longer time periods.
The final objection to refundable credits is that they could increase fraud and related compliance problems. Yet there is no reason in theory, and no empirical evidence in practice, why there should be a “cliff effect” in fraud precisely at the point of positive income tax liability. If anything, fraud may be easier to hide when it comes in the form of a deduction or exclusion, which reduces taxable income, as opposed to a refundable credit. Instead, reducing fraud and related compliance problems for all tax incentives, including refundable credits, requires structuring the incentives simply, relying on third-party reporting, and investing in enforcement staffing.
We recognize that increasing the prevalence of refundable credits may create incentives for tax units who are currently non-filers to begin filing, thereby increasing administrative costs for the government and compliance costs for these households. These costs are real and should be taken into account. Nevertheless, they should not be overstated. Currently only about 13 percent of tax units are non-filers. As a result, non-filers represent a relatively small share of the households who stand to gain from structuring tax incentives as uniform refundable credits. Moreover, all tax incentives are elective and, even for non-filers, the administrative and compliance costs associated with claiming them are likely to be swamped in many instances by the dollar value of the credit.
CONCLUSION
Uniform refundable tax credits are generally the most efficient structure for a tax incentive to encourage desired behavior when, as frequently occurs, evidence of how the desired behavior and its associated social benefits vary across the income distribution is unavailable or inconclusive. Indeed refundable tax credits are generally the only way to ensure a tax incentive reaches the roughly two-fifths of tax units with no positive income tax liability in a given year. These efficiency benefits are magnified by the ability of refundable credits to help smooth income at a household level and by their ability, to a greater or less extent, to bolster the role of the tax system as an automatic stabilizer of macroeconomic demand. The United States spends almost 4 percent of GDP each year subsidizing socially-valued activities through the tax code. Our proposal would dramatically improve the effectiveness and fairness of this substantial investment.
