The superiority of an ideal consumption tax over an ideal income tax.

Publication: Stanford Law Review

Publication Date: 01-MAR-06

Bookmark this article Print this article Link to this article Email this article Digg It! Add to del.icio.us RSS

COPYRIGHT 2006 Stanford Law School

INTRODUCTION



I. THE CORE ARGUMENT A. Basic Definitions and Relationships Between the Bases B. Arguments for an Income Tax C. The AS 1976 Argument--Efficiency D. The AS 1976 Argument--Redistribution II. RISK AND PROFITS III. WEALTH WITHOUT LABOR INCOME IV. SAVINGS HETEROGENEITY A. Rational Savings Decisions B. Savings Myopia and Similar Problems V. DOES SAVINGS BRING VALUE BEYOND FUTURE CONSUMPTION? VI. QUALIFICATIONS CONCLUSION

INTRODUCTION

Perhaps the single most important tax policy decision is the choice between an income tax and a consumption tax. The topic has been discussed and argued over since at least the time of Hobbes and Mill, without apparent resolution. (1) Consumption and income taxes both represent substantial sources of revenue in all modern economies.

This Article considers the choice between an income tax and a consumption tax, focusing on an argument first made by Anthony Atkinson and Joseph Stiglitz in 1976 ("AS 1976"). AS 1976 shows (under the assumptions of the model) that taxes should be imposed on all commodities at the same rate-that is, taxes should be neutral. For reasons illustrated below, this conclusion implies that a consumption tax is superior to an income tax. AS 1976 has recently attracted substantial attention in the economics literature but, perhaps because the arguments are technical, has yet to receive any attention in the legal literature. (2) Our primary task here is to explain the intuition behind AS 1976 and explore its implications for the income tax versus consumption tax debate. In addition, we examine what happens when some of the strict assumptions of AS 1976 are relaxed and, in doing so, revisit a number of arguments that have been made in favor of income taxes. We conclude that, based on current understanding, ideal consumption taxes are superior to ideal income taxes.

We will generally compare only the ideal forms of income and consumption taxation. The actual choice of a tax system has to be based on how the system would be implemented, focusing on administrative and compliance costs. Neither an income tax nor a consumption tax would likely be implemented in its pure form, and differences in administrative and compliance costs might be dispositive in the choice between the two. Nevertheless, it is worth examining the ideal forms for two reasons. First, determining which ideal form is most desirable helps us design actual systems and helps us understand the flaws of actual systems--ideals matter in tax reform.

Second, the case for the income tax is likely to be strongest if the comparison is made between ideal forms. This is true because the income taxes we have had for almost a century are much worse than the ideal income tax, and they contain structural features that make reform difficult. For example, an ideal income tax would tax the change in the value of investments each year. Under existing law, the change in investment value is taxed only if it is "realized" in the form of a sale or exchange. The so-called realization requirement is responsible for much of the current complexity and distortion. Elimination of that requirement, however, raises difficult liquidity and valuation issues and, in part for those reasons, has never been seriously considered. An ideal income tax would also measure gain and loss on an inflation-adjusted basis. Inflation adjustments, while possible, would be difficult and also have never been seriously considered. A consumption tax raises neither of these difficulties, and most scholars believe that a consumption tax is easier to administer, and can be administered in purer form, than an income tax. By comparing ideal systems and ignoring administration costs, we are deliberately making the best possible case for the income tax. If a consumption tax is superior to an income tax even ignoring the major implementation problems of an income tax, it follows that a consumption tax will be even more desirable once those problems are taken into account.

Part I of this Article presents the core argument, focusing on the simplest case, in which investments produce only risk-free, time-value returns, and individuals vary by their ability. Income taxes tax the risk-free return while consumption taxes do not. In this simple world, the AS 1976 arguments show that a consumption tax can be structured to be a Pareto improvement over an income tax. Importantly, this argument addresses both efficiency and redistributive concerns. Everyone is equally well off or better off under a properly designed consumption tax. It is either more efficient, more redistributive, or both.

The AS 1976 model, like all models, contains assumptions and simplifications. To understand the practical impact of the AS 1976 arguments, we need to understand the realism of the assumptions and the effect of relaxing them. The remaining Parts of the Article consider these issues. We consider the four most prominent issues and show that the conclusions from the simplified world of only risk-flee investments carry through, almost in their entirety, to more realistic cases.

Part II considers the taxation of risky returns and economic profits. Extension of the basic case to risky returns and profits is straightforward. There is a long line of literature showing that ideal, flat-rate income and consumption taxes treat risky returns and economic profits the same way, leaving the riskfree rate of return as the only difference, as discussed in Part I. Part II very briefly explains this literature and then discusses whether imposing graduated rates on capital income changes the conclusions.

Part III considers how labor income and wealth are related and the extent to which the possibility of wealth without labor income affects the arguments. One might think, for example, that because they tax capital income, income taxes are better at capturing the benefits of inheritances. Part III shows that if correctly implemented, a consumption tax can tax such wealth; therefore, such wealth should not affect the choice between the two tax bases.

Part IV considers the difference between spenders and savers and whether savers are better off in a manner that would support an income tax. The basic argument given in Part I assumes that within an earnings or ability class, individuals make similar savings decisions. In the real world, there may be significant heterogeneity in savings, and this heterogeneity has been thought by some to support an income tax. Part IV argues that it does not.

Part V examines the argument that savings brings prestige, power, and security and that the benefit of savings is more than future consumption. This extra benefit of savings is thought by some to support an income tax. Part V shows that this is not the case. Consumption taxes properly tax the benefits from savings.

The AS 1976 model, like all models, is subject to a number of qualifications and extensions. The economics literature examining and extending AS 1976 is large and complex. Our goal here is to explore the core arguments that arise from the literature and their practical implications. Newer models show that a complete, optimal tax analysis could produce exotic taxes that look like neither a pure consumption tax nor a pure income tax. These models may also help explain deviations from pure income and consumption taxes (such as deductions granted to particular types of individuals or activities) that might otherwise seem troubling. In Part VI, we will briefly discuss the possibility that newer models might show that a tax on savings is desirable.

Before we begin the analysis, we should clarify our terminology and the origins of the ideas explored here. Throughout the Article, we will refer to the argument as originating with AS 1976 because that paper was the first in a line of papers on the topic. AS 1976 and many later papers in the economics literature analyzed the problem of taxation by assuming that there was a perfectly designed and implemented labor income or consumption tax in place and asked whether any small perturbations from such a tax were desirable. (3) An alternative method of analyzing the problem was first developed by Hylland and Zeckhauser and substantially strengthened and extended by Kaplow. (4) This method uses a "replicating tax" argument. It starts with a nonneutral commodity tax and shows how to construct a Pareto superior neutral tax. This latter method of analyzing the problem has two key strengths. First, it extends the result to cases in which a labor income or commodity tax is not optimal, which is extremely important for applying the argument to the real world. Second, the analysis is more direct and intuitive. We follow the Hylland/Zeckhauser and Kaplow method of analysis here; to avoid constant parsing of which paper in the economics literature developed which idea, however, we simply refer to the entire literature as AS 1976.

I. THE CORE ARGUMENT

A. Basic Definitions and Relationships Between the Bases

We begin with the simplest case. We assume in this Part that investments produce only the risk-free, time-value return and that individuals vary by their ability to earn. All of the AS 1976 intuitions can be illustrated in this simple case. We relax these strict assumptions in later Parts.

As is well known, the difference between an income tax and a consumption tax is the taxation of the return to savings or capital income. In a consumption tax, the risk-free return to investing is exempt, while in an income tax, the return is taxed.

A consumption tax, as a matter of legal implementation, is imposed on consumption and not on labor, but it is economically equivalent to a tax on labor earnings. The reason is that on a going-forward basis, there are two sources of consumption: earnings from labor (wages) and earnings from capital. If, under a consumption tax, capital income is not taxed, all that is left to tax is wages. (5)

Another way to see that a consumption tax is a tax on labor earnings is to imagine a consumption tax imposed when consumption goods are purchased, as is the case in a retail sales tax. The tax on purchases will reduce the value of a dollar earned in exactly the same way that a direct tax on earnings would. For example, suppose all commodities face a 30% tax when purchased. (6) If an individual has $100 of labor earnings, he can consume only $70 of goods. The benefit of working hard enough to earn $100 has been reduced by 30%. We could equivalently have taxed the $100 when earned, leaving the individual with $70 to spend as he pleases.

Note that a tax on consumption purchases does not burden capital income. Suppose, for example, the individual, subject to the retail sales tax, waits until next year to consume, investing his $100 in the market at a 10% rate of return. He will have $110 next year and will be able to consume $77 after paying the 30% taxes on his purchases. This result is the same as if we taxed his labor income when earned at 30%, and he invested his after-tax $70 in the market at a 10% rate of return.

When we refer to an ideal, neutral, or uniform consumption tax, we mean that the consumption tax is imposed at the same rate on all consumption. Note that this includes consumption occurring in different time periods as well as different forms of consumption in the same period. For example, the 30% retail sales tax considered above imposes the same 30% rate on consumption whenever it occurs. A nonneutral or nonuniform consumption (or commodity) tax imposes different tax rates on different commodities or forms of consumption. For example, a nonneutral consumption or commodity tax might impose a 20% rate on one type of good and a 40% rate on another.

Neutral consumption taxes can be progressive. Individuals with more consumption can face higher average or marginal tax rates even while those rates are imposed uniformly on all of their purchases. The easiest way to envision this progressivity is through a wage tax with graduated rates, but there are other methods of implementing such a system. (7)

An ideal income tax, like an ideal consumption tax, will impose the same nominal rate on the entire tax base (and, like a consumption tax, can be made progressive by imposing graduated marginal rates, among other methods). Because it taxes the returns to savings, however, an income tax can be thought of as imposing a higher rate of tax on future consumption than on current consumption. To see this, recall the Haig-Simons definition of income as the sum of consumption plus change in wealth in a given time period. The first component, consumption, is just like an ideal consumption tax (uniform on all consumption) and taxes all consumption, whether present or future, at the same marginal rate. The second component, the tax on the return to savings, reduces the benefit of savings, making future consumption relatively more expensive than current consumption.

We can (and will) think of an income tax as a nonneutral consumption (or sales or commodity) tax in the sense that it imposes different rates on consumption choices in different time periods. That is, the choice between an income tax and consumption tax can be seen as part of the more general question of whether any uneven or nonneutral commodity tax is desirable.

To illustrate this numerically, consider an individual, Z, who earns $100 in period one and is considering whether to spend the sum in period one or two. Assume arbitrarily that the pretax rate of interest is 5%. Absent taxes on interest income, Z could either consume $100 of goods in period one or save the $100, earn 5%, and consume $105 of goods in period two. The $105 of goods in period two have a present value to the individual of $100. Assume now that the return to savings is taxed at a 40% rate and is thus reduced to 3%. Z now must choose between consuming $100 in period one or $103 in period two. This reduction from $105 to $103 has the same effect as a sales tax of about 2% on consumption in period two. If discount rates remain constant at 5%, the market value of available period two consumption drops to $98.10. (8)

The effective tax rate levied on future-consumed goods increases as the time of consumption grows more distant. If, in the above example, consumption is deferred for three years, the tax reduces available consumption from $116 to $109--the equivalent of a sales tax of 6.4%. After thirty years, the amount available is reduced from about $430 to $240. This is equivalent to a sales tax of approximately 80%. The choice between an income tax and a consumption tax can be restated as the question of whether such a sales tax is desirable. As such, this question is part of the general issue of whether and when nonneutral commodity taxes are desirable.

B. Arguments for an Income Tax

There is a large literature on the choice between an income tax and a consumption tax, split in its support of one or the other. (9) While there are numerous arguments on the issue, we believe that there are three reasons why many prefer an income tax to a consumption tax. The first is an efficiency argument, which concludes that the determination of whether a consumption tax is more efficient than an income tax depends on empirically unknowable or indeterminate facts and that, therefore, there should be no presumption that one is more efficient than the other. The second is that an income tax is better at redistribution. This argument states that since the efficiency effects of the choice between a consumption and an income tax are ambiguous and possibly unknowable but there are clear distributive gains under the income tax system, we should support an income tax. The third argument is that wealth is thought to bring a host of benefits, such as power, prestige, and security and an income tax is more effective than a consumption tax at taxing these benefits.

The efficiency argument, which we will call the "tradeoff theory," compares the relative distortions of an income tax and those of a consumption tax. A consumption tax does not tax the return to savings. This means that savings decisions are undistorted, and individuals choose the optimal amount to consume at each date. A consumption tax does, however, tax labor earnings, which means that decisions about how much to work are distorted. An income tax taxes the return to savings, which means that future consumption is relatively more expensive, and savings decisions will be distorted. The claimed advantage of an income tax, however, is that, by taxing the returns to savings, the tax rate on labor earnings can be lower; thus, work decisions are distorted less under an income tax than they are under a consumption tax. Whether a consumption tax or an income tax is more efficient depends on the relative elasticities of savings and work effort. As stated by one prominent economist:

The efficiency effects [of the choice between an income tax and a consumption tax] depend on assumptions about behavioral effects. If individuals are relatively unwilling to substitute consumption over time and relatively willing to substitute leisure for consumption of goods, then a significant tax on capital income would constitute part of an optimal tax system. These behavioral effects are difficult to estimate empirically. (10)

This same argument is repeated in the most recently published public finance textbook, which is intended to summarize economists' basic understanding of these issues. (11)

The second reason for supporting an income tax is distributive. Income taxes are thought to have better distributive consequences than consumption taxes. One version of this argument is that failure to tax returns to savings leaves enormous pools of wealth untaxed, creating vast inequalities in our society. Much of that wealth is created because of general societal conditions--such as property rights, an effective government, the legal system, educated workers, natural resources, and protection from invasions--which have nothing to do with the fortunate (although also skilled and hard-working) individual who earns great wealth as a result. Society has a right to distribute that wealth as it sees fit, and it is just and fair to use it to reduce inequality. (12)

The more technical version of this argument is that transferring a dollar from the wealthy to the poor increases welfare because the marginal utility of that dollar to a wealthy person is likely to be lower than it is to a poor person. (13) If utility goes up with income from capital as well as with income from labor, both should be used as a basis for redistributing. This would seem to be true--someone with a large trust fund is unlikely to value another dollar as much as someone working two jobs just to scrape by. Redistributing one dollar from the trust-fund baby to the working poor is likely to increase overall welfare. Paris Hilton very likely has a much lower marginal utility of money than someone slaving in the salt mines sixty hours a week to support his family. Redistribution from Paris Hilton to the worker makes sense.

The third, often-repeated argument for an income tax is that wealth brings benefits beyond the value of future consumption. For example, wealth is said to bring security, prestige, and power. Some have argued that only an income tax can tax this wealth and corresponding benefits and, therefore, redistribute in ways that even a highly progressive consumption tax cannot. Given the importance that these commentators put on redistribution, they conclude that an income tax is desirable.

AS 1976 shows that a properly designed consumption tax is Pareto superior to an income tax. It is either more efficient (holding distribution constant), more redistributive (holding efficiency constant), or both, which means that the efficiency and distributive arguments are incorrect. We also argue that the "wealth as more than future consumption" argument is incorrect, reserving discussion of this issue for Part V.

C. The AS 1976 Argument--Efficiency (14)

The tradeoff theory argues that an income tax might possibly be more efficient than a consumption tax because it reduces the tax on labor income while increasing the tax on capital income. Depending on the relevant elasticities, an income tax might be preferable. AS 1976 shows that the tradeoff theory is incorrect because it misses one of the effects of a tax on the return to savings. In particular, a tax on the return to savings, or any nonneutral commodity tax, has two effects. As widely noted, a tax on the return to savings...

Read the full article for free courtesy of your local library.