I. Introduction
One of the most vexed issues in the history of the federal income tax is how family or household status should affect tax liability. The relative treatment of different types of households, such as singles versus married individuals as well as those with and without children, has fluctuated significantly over time and between countries, reflecting not just varying political trends but the general lack of an underlying theoretical consensus concerning how to think about household issues.1
Classic articles from the 1970s that note the long-standing lack of political stasis and theoretical consensus include Boris I. Bittker, “Federal Income Taxation and the Family,” Stanford Law Review 27 (1975): 1389–1463; and Michael J. McIntyre and Oliver Oldman, “Taxation of the Family in a Comprehensive and Simplified Income Tax,” Harvard Law Review 90 (1977): 1573–1630. Among the very numerous more recent treatments of the subject are Louis Kaplow, “Optimal Distribution and the Family,” Scandinavian Journal of Economics 98, no. 1 (1996): 75–92; Edward J. McCaffery, Taxing Women (Chicago: University of Chicago Press, 1996); and Lawrence Zelenak, “Doing Something about Marriage Penalties: A Guide for the Perplexed,” Tax Law Review 54 (2000): 1–76.
See, e.g., Grace Ganz Blumberg, “Sexism in the Code: A Comparative Study of Working Wive and Mothers,” Buffalo Law Review 21 (1980): 49–98; Marjorie E. Kornhauser, “Love, Money, and the IRS: Family, Income-Sharing, and the Joint Income Tax Return,” Hastings Law Journal 45 (1993): 63–111; Nancy C. Staudt, “Taxing Housework,” Georgetown Law Journal 84 (1996): 1571–1647; McCaffery, Taxing Women; and Anne L. Alstott, No Exit: What Parents Owe Their Children and What Society Owes Parents (Oxford: Oxford University Press, 2004).
The discussion proceeds as follows. In Section II, I discuss the need for a normative framework and my use of utilitarianism to provide such a framework. In Section III, I discuss why households matter in setting the treatment of individuals by the fiscal system. In Section IV, I discuss the definition of a household. In Sections V and VI, I discuss the commonly cited but often irreconcilable principles of equal taxation of equal-income couples and marriage neutrality. In Section VII, I discuss the relative treatment of single individuals, one-earner couples, and two-earner couples. In Section VIII, I discuss the fiscal system's incentive effects on work decisions by secondary earners in a household. In Section IX, I discuss the distributional and incentive issues raised by the treatment of households with children. In Section X, I conclude by noting some possible practical implications of the prior analysis.
II. A Normative Framework
The main advance in the tax policy literature on which I draw in this essay is the shift to a more sophisticated normative stance. Rather than assigning canonical status to the Haig-Simons income concept, which defines income as the market value of the taxpayer's consumption and change in net worth during the tax year,3
Henry Simons, Personal Income Taxation (Chicago: University of Chicago Press, 1938), 50. The adoption of an income tax in various Western European countries in the late nineteenth century, and in the United States in 1913, prompted extensive scholarly discussion of what “income” (typically assumed to be a desirable tax base on poorly specified grounds relating to the “ability to pay” concept) really meant. Simons championed the term's generalization to cover the market value of all consumption and changes in net worth, even in the absence of an observed market transaction such as a sale of appreciated property. The Haig-Simons income ideal held a dominant position in U.S. legal tax scholarship from at least the 1950s through the 1980s, although more recently it has increasingly been challenged by support for consumption taxation.
Liam Murphy and Thomas Nagel, The Myth of Ownership: Taxes and Justice (Oxford: Oxford University Press, 2002), 6.
McIntyre and Oldman, “Taxation of the Family,” 1576.
Boris I. Bittker, “A ‘Comprehensive Tax Base’ as a Goal of Income Tax Reform,” Harvard Law Review 80 (1967): 974.
Bittker, “Federal Income Taxation and the Family.”
More recent work has rejected the traditional doctrinal approach in favor of focusing on such central issues in household taxation as the tax system's discouraging married women from working and its favoring traditional one-earner families over those in which both spouses work.8
See, e.g., McCaffery, Taxing Women. I have also examined these issues in earlier work. See Daniel N. Shaviro, Making Sense of Social Security Reform (Chicago: University of Chicago Press, 2000), 65–66 and 72–73; and Daniel N. Shaviro, Who Should Pay for Medicare? (Chicago: University of Chicago Press, 2004), 69–71, 128–30, and 134–35.
See Kaplow, “Optimal Distribution and the Family”; Kaplow, “Families, Tax Treatment of,” in Joseph Cordes et al., Encyclopedia of Taxation and Tax Policy (Washington, DC: Urban Institute Press, 1999); and Louis Kaplow, “Taxation,” in A. Mitchell Polinsky and Steven Shavell, eds., Handbook of Law and Economics (forthcoming).
The need to specify an underlying, inevitably controversial normative stance undermines the possibility of achieving broad consensus, but at least it means that one's cards are on the table. My own preference, utilitarianism, has come nowhere near universal acceptance. Nor even has welfarism, or the broader set of positions holding that only individuals' well-being matters, whether or not toted up in utilitarian fashion.10
But see Louis Kaplow and Steven Shavell, Fairness Versus Welfare (Cambridge, MA: Harvard University Press, 2002), arguing for welfarism.
The form of utilitarianism that I apply is hedonic rather than preference-based, and thus permits overriding people's observed preferences if we have sufficient grounds for believing that they are making mistakes from the standpoint of their own welfare. (Their preferences would, however, typically serve as powerful empirical evidence regarding what best serves their welfare.) In addition, I assume a utilitarian aim of maximizing total rather than average welfare. I concede that there are grave difficulties with the view that people's mental states can generally (even assuming full information) be given cardinal or even ordinal welfare rankings, reflecting the diversity of our psychological sensations and experiences.
The basic idea I use in operationalizing utilitarianism is that we should think of the fiscal system, including the income tax, in terms of an expanded notion of “social insurance.”12
See Shaviro, Making Sense of Social Security Reform, 44–58, for a broader discussion of social insurance.
On the demand side, however, the role of the law of large numbers is incidental rather than fundamental. Indeed, even offsetting a bad financial outcome is merely a means, not the end. A rational consumer buys insurance, even though (assuming an actuarially fair bet with a compensated service provider) it reduces her expected financial return, because it increases her expected utility. The idea is to direct more dollars to states of the world where the consumer expects to value them more, by agreeing to accept fewer dollars in states of the world where she expects to value them less.
Due to the declining marginal utility of wealth, this often involves compensating losses. For example, I may value getting a million dollars from the insurance company much more in the state of the world where I owe an injured pedestrian one million dollars than in that where I have avoided collisions. But some bad outcomes are not worth betting against, because they would not increase the marginal value of a dollar. Thus, for example, parents typically do not insure against the death of a child, because, terrible though it would be, it would not add to the value of getting cash. Moreover, we sometimes “bet” against good outcomes rather than bad ones, by using insurance or similar contracts to direct dollars to states of the world where we are better off but expect to derive greater utility from a marginal dollar. One example is a life annuity, which insures you against the “risk” of living longer—a good outcome, but one that means you will need more money to meet your lifetime material needs. Similarly, medical insurance, when conditioned on available treatments, offers you a bigger payoff if expensive new treatments that could help you are newly developed—again, a good outcome, but one that makes money more valuable to you.
With rational consumers and complete, well-functioning markets, private insurance would advance utilitarian aims because, by definition, every arrangement that increased expected utility would arise in response to consumer demand. Indeed, even arrangements that were Kaldor-Hicks efficient rather than Pareto-efficient—that is, arrangements that made people on average better-off but with some losers—would arise if there were a stage where people could make insurance bargains behind a veil of ignorance regarding how their odds would change once they knew more about their circumstances. Thus, suppose that transferring ten dollars from A to B would increase B's utility more than it reduces A's. While A might not like this arrangement once she knew she was A, if there were a stage where she could decide whether to mandate the transfer before knowing whether she was more likely to end up as A or as B, then it would be no different from the case where she is buying insurance for her future self. Hence the insight of economist John Harsanyi, writing before John Rawls's more famous use of the behind-the-veil rubric,13
See John Rawls, A Theory of Justice (Cambridge, MA: The Belknap Press of Harvard University Press, 1971).
John Harsanyi, “Cardinal Utility in Welfare Economics and in the Theory of Risk-Taking,” Journal of Political Economy 61 (1953): 434. To be sure, positing a veil of ignorance is an extra step relative to simply valuing everyone's welfare equally in the world we actually observe, a step requiring justification and potentially raising such questions about its “thickness” as what degree of risk aversion people should be deemed to have if they do not know who they are. From a utilitarian standpoint, however, the veil of ignorance is merely a way of expressing morally required indifference between oneself and other people. Treating oneself as equally likely to be A or B is a way of treating A's and B's welfare as equally important, rather than importing selfish bias if one knows that in fact one is A. See ibid., 453; and Shaviro, Making Sense of Social Security Reform, 52.
Where market failure leaves private demand for insurance unsatisfied, including due to the lack of a behind-the-veil stage, there is a case for government intervention if the government can better address the failure. The two classic insurance problems that might lead to private market failure, but that government intervention could try to address, are moral hazard and adverse selection. Moral hazard arises when the insured reduces her effort to avoid or mitigate the circumstances that would give rise to a claim, because of how the insurance affects her incentives. An example would be driving less carefully due to one's car insurance coverage. Adverse selection involves using superior information about one's own risk profile to buy coverage disproportionately in situations where the odds are against the insurance provider. An example would be buying generous health insurance coverage based on your knowledge, which the insurance company does not share, that you have a condition calling for costly surgery.
In general, governments may have a hard time outperforming private firms with respect to moral hazard. Greater power to compel information disclosure, thus addressing the information asymmetry about the insured's decisions that underlies the problem, is the only extra tool governments have, and this tool may not always help very much. A government can, however, powerfully address adverse selection if exit from its zone of control is costly. By taxing residents to fund benefits, it can limit the effect of people's superior information about their own circumstances on enrollment outcomes.
Government insurance provision also has a major disadvantage. The weakness of competitive constraints on government actors creates incentive problems regarding the provision of welfare-enhancing insurance. Unlike businesses in a competitive market with informed consumers, governments do not have to make enrollees better off unless the governments themselves face either competitive pressures or control by informed voters (that is, exit or voice).
The most prominent instance where the case for social insurance is widely accepted pertains to lifetime income risk, or the risk that one's lifetime income will be low rather than high for reasons outside one's control.15
As I discuss later with respect to having children, a social insurance rationale potentially applies even to risks within one's control, but this may greatly raise the stakes regarding moral hazard issues.
See Shaviro, Making Sense of Social Security Reform, 51.
Income risk is addressed by means-based tax systems such as the income tax, which cause one's tax liability to rise with some measure of one's material well-being, and by means-based transfer systems such as, in the United States, Temporary Aid to Needy Families (TANF, the current official name for welfare), food stamps, and Medicaid. These various systems could all be described as engaging in vertical redistribution, or redistribution from the presumptively better-off to the presumptively worse-off, based, under a utilitarian social welfare function, on the assumption of declining marginal utility with respect to material resources.17
A transfer system such as Medicaid further reflects a view that medical needs are distributionally important independent of income.
From a utilitarian standpoint, however, the assumption of declining marginal utility that supports vertical redistribution by no means exhausts the set of cases where a transfer of money from A to B would be expected to increase total welfare. There also are cases where we would expect B to gain more welfare than A loses from the transfer, based on information wholly apart from their relative levels of well-being.
Suppose, for example, the following: A and B have the same preferences; they are comparably threatened with imminent heart attacks, but only A's condition can be treated; this treatment is very expensive, and A cannot afford it; A would choose this treatment if given the money; and B, if given the same amount of money, would have nothing better to spend it on than eating out more and going on nice vacations during the period before the heart attack struck. On the surmise that A would benefit more from the money because her condition is treatable, we might prefer giving the money to her rather than to B, without regard to which of them is at a higher well-being level.18
More precisely, this reason for giving the money to A would apply without regard to the parties' relative well-being levels. This is not to deny that one's overall distributional decision might reflect both this reason for giving money to A and information about the parties' relative well-being levels.
Or suppose that C and D are both retired, and no longer physically capable of paid work, but that neither has entered retirement with significant savings. We might decide, in the fashion of Social Security, to give each of them a life annuity. If C had a greater life expectancy, this would amount to giving him a larger expected benefit than D. The motivation for favoring C would be that, even though living longer is presumably a good thing, it increases one's post-retirement needs.19
This example ignores the possibility of inducing people with higher life expectancies to work longer or save a larger proportion of their lifetime earnings for retirement. It also does not take account of moral hazard, which could take the form of under-saving given one's above-average life expectancy because one knows that the retirement system will make up the difference.
Where an assumption other than declining marginal utility supports the view that a dollar would do more to benefit people in circumstance 1 than in circumstance 2, the utilitarian case for redistributing wealth to the former group from the latter could be described as involving lateral redistribution, rather than vertical redistribution. We are not, to the extent of this redistribution, systematically engaged in transferring resources from the better-off to the worse-off. Rather, we are using other types of information to guide our distribution policy. We might even, on average, be favoring better-off individuals through this redistribution. For example, it is presumably better to be longer-lived and have a treatable rather than an untreatable condition, and both life expectancy and treatability might be positively correlated with an individual's earnings. While this would not have to imply reduced vertical redistribution, it would mean that some notional portion of our total redistributive effort was purely in response to what I call lateral differences, or those based on observed characteristics other than well-being level.20
Since part of my aim here is to offer an analysis that is relevant to people who have sympathy for increasing total well-being even if they are not utilitarians, it is worth noting two common criticisms of utilitarianism. The first relates to hypothetical “utility monsters who get enormously greater gains in utility from any sacrifice of others than these others lose…. [Utilitarianism] seems to require that we all be sacrificed in the monster's maw, in order to increase total utility.” Robert Nozick, Anarchy, State, and Utopia (New York: Basic Books, 1974), 41. This objection will not arise here, as I will be emphasizing differences in people's preferences and circumstances that do not raise issues of differing intensity as to their overall sets of preferences. The second common criticism pertains to expensive tastes, or the concern, first raised by Kenneth Arrow, and developed most notably by Ronald Dworkin, that it is unfair to favor individuals by reason of their having freely chosen tastes that are unusually expensive to satisfy. See Kenneth J. Arrow, “Some Ordinalist-Utilitarian Notes on Rawls's Theory of Justice,” Journal of Philosophy 70 (1973): 245; Ronald Dworkin, “What Is Equality? Part 2: Equality of Resources,” Philosophy and Public Affairs 10, no. 3 (1981): 283. Anne Alstott, writing about households with children from the standpoint of liberal egalitarianism, treats this as an objection that needs to be met on nonutilitarian grounds that relate to parents' responsibility for their choices and to effects on children's development. Alstott, No Exit, 61–66. One should keep in mind, however, that the apparent underlying concerns about the social costs of accommodating expensive tastes and encouraging people to develop such tastes are relevant under utilitarianism. Thus, while concern about rewarding expensive tastes, depending on its underlying basis, might affect a nonutilitarian's degree of acceptance of the approach taken in this essay, it does not necessarily suggest reaching radically different conclusions.
With this framework in place, I turn to a number of the main issues posed by imposing tax and transfer rules on the individuals in different kinds of households. My aim is to provide a basic of survey of how a utilitarian might think about these issues, all of which could be explored in another setting at much greater length.
III. Why Do Households Matter?
Under the U.S. federal income tax, the question of whether household status should matter, via the use of joint returns for married couples, is typically posed as one of what is the “appropriate taxable unit.”21
Kornhauser, “Love, Money, and the IRS,” 92.
Bittker, “Federal Income Taxation and the Family,” 1391.
This formulation can lead to confusion. For one thing, the “taxable unit” is only of administrative interest. As an illustration, it commonly is recognized that joint returns, where the dollar amounts in each rate bracket are double those applying to single individuals' tax returns, are arithmetically equivalent to separate returns for married couples that are prepared under the assumption that each couple splits its total income fifty-fifty. Joint returns are merely one way of bringing household information to bear on tax and transfer outcomes, which can also be done through separate returns that make use of information about other household members.23
An example of using household information without joint filing is the federal income tax treatment of minor children, who must file their own returns if they have sufficient income, but may be claimed by their parents as dependents and also may be taxed on certain unearned income at the parents' marginal rate. A proposed example of using household information along with separate filing is McCaffery's proposal that spouses file separate returns but with a “more generous rate schedule, or greater deductions, for the lesser-earning spouse.” McCaffery, Taxing Women, 277.
Two simple examples may help to dramatize the importance of household information to distribution policy. First, suppose that we offer net transfers to people with low earnings and assets, and consider only individual rather than household information in deciding who needs the transfers. If Mrs. Bill Gates did not work and did not own any of the Gates assets but lived the Gates lifestyle, would it make sense to think she needed a net transfer?
Second, suppose we observe two same-aged single individuals earning $20,000 a year, one of them with no children and the other with ten. Even if we just focused on these two individuals, without considering the needs of the ten children, would it really make sense to surmise that their material circumstances, like their earnings, are the same, and thus to figure that neither of them would value an extra dollar more than the other? If you started out with no children and suddenly found yourself with ten, doesn't it seem likely that this would increase the value that you placed on an extra dollar?
Put more generally, these examples illustrate two broad points. The first is that, in assessing an individual's well-being level and likely marginal utility of a dollar, we must evaluate the significance, not only of self-owned resources, but of those owned by other members of the individual's family or household. The second is that, in making this same assessment, we must consider not only the individual's own needs but also those of others in the same family or household. The reason we need to make these broader assessments is that resources actually flow between members of the household in ways that we would not ordinarily expect as between third parties. Thus, we may need to look at other household members' resources and wants in order to understand either the current circumstances of any household member or how those circumstances would change if an extra dollar were given to her or taken away. Moreover, since we cannot look at each household individually to determine how it actually operates, we may need to apply generalizing assumptions about various types of households' internal distribution behavior that we determine are more realistic, on average, than assuming that each household member is an island.
One further implication is that we cannot understand tax and transfer incidence, or who economically gains or loses from the government's extracting or providing a dollar, without looking inside the household. Suppose, for example, that we wanted to penalize sexist husbands while aiding their wives, or that we wanted to aid the children, as distinct from the parents, in a given set of households. Making the husband write a check while sending a check to the wife, or writing government checks that were made out to the children rather than the parents, would not necessarily accomplish the desired intrahousehold distributional aim. Maybe it would, but we would have to understand or observe the household's internal decision rules—how it allocates its various resources—in order to know.
IV. What Is a Household, and How Might It Be Officially Defined?
A. Households in concept
The reasons why households matter for tax and transfer purposes suggest how, as a basic conceptual matter (leaving aside, for now, practical implementation), they ought to be defined. Households are groups of people who, to a significant degree, pool their individually owned resources for allocation among group members based on some set of rules or norms that are not limited to respecting individual ownership.
Equal sharing of resources among household members without regard to who contributed them would be one example of a norm that a household might use, but is by no means exclusive.24
In some past literature, relevant pooling and allocation have been defined more narrowly. For example, Kornhauser, “Love, Money, and the IRS,” 97, argues that viewing households as economic units requires “assum[ing] that the family is a monolithic, homogeneous group in which all members share the same tastes and resources, including income, equally…. True pooling presumes equality, if not in contributions to the pool, then at least in free access to the pool.”
For reasons that merge human biological nature with contemporary Western cultural nurture, household affiliations between individuals in contemporary Western societies are most commonly of two types. The first involves couples, defined as pairs of individuals who decide to form lasting personal relationships.25
Given the practice of polygamy at various times and places in world history, and indeed its apparent survival in portions of Utah (see Jon Krakauer, Under the Banner of Heaven: A Story of Violent Faith [New York: Doubleday, 2003]), a more universal definition would have to permit extension of this definition beyond the case of two individuals who form a couple.
I use the word “household” rather than “family,” although neither is a perfect fit, because blood relationships do not always involve significant resource pooling and reallocations (which typically are quite limited, for example, between adult siblings), and because the question of whether a given couple should be called a “family” is a political hot-button issue for reasons that have nothing to do with how individuals share their resources. The term “household” can also be misleading, however. In common usage, it typically includes roommates who physically live together but do not significantly pool their resources, while excluding spouses who have separate primary residences but engage in resource pooling.26
See Kornhauser, “Love, Money, and the IRS,” 67.
While resource pooling and allocation are the main features of interest here, it is true that living in one residence versus two may be relevant to distributional policy. Economies of scale, which roommates may enjoy without any pooling beyond that of common living space, appliances, and various food items, are often mentioned as potentially increasing one's level of material well-being.27
See, e.g., Bittker, “Federal Income Taxation and the Family,” 1422–25.
See Kaplow, “Optimal Distribution and the Family,” 80–81.
B. Households in practice
In identifying couples, the federal income tax, Social Security, and Medicare rely purely on marriage, but welfare rules often cast their net more broadly for purposes of disqualifying potential recipients through asset or income limits and work requirements. One of the well-known demographic trends of recent decades is a decline in marriage as a predictor of couple status, both because heterosexual couples face less pressure to marry than in earlier times and because acknowledged same-sex couples are now freer to form but, in nearly all states and generally for federal legal purposes, no freer to marry. Marriage has therefore become an increasingly poor proxy for identifying couples who function as a household unit.
The case for trying to identify unmarried couples who should be viewed as members of the same household is therefore growing ever stronger. Helping to make it quite difficult, however, is the fact that, just as the federal income tax has marriage penalties in some situations and marriage bonuses in others, so the verification and enforcement problem with respect to unmarried couples could go in either direction. One might, in principle, need either to verify claims of couple status (as in the case today where immigration authorities try to root out sham marriages between Americans and foreigners), or to assert it in cases where the involved individuals were not doing so.
As an example of the different ways these issues may play out, same-sex couples typically benefit under the federal income tax from not being identified as married if both individuals are working and their earnings are comparable. Likewise, same-sex couples may benefit under the welfare laws from not being amalgamated for purposes of income and asset tests and work requirements. In contrast, if only one member of the same-sex couple works, then they are denied the marriage bonus that the federal income tax, Social Security, and Medicare extend to one-earner couples.
If we find ourselves at some point living in more enlightened times, it is possible that some of these issues will be addressed, at least for same-sex couples that at present cannot marry, and at least where claiming couple status favors the claimants. In principle, however, tax and transfer recognition of a couple should not depend upon whether it helps or hurts the claimants. While assertion of couple status by the government may seem to raise the specter of IRS bed-checks to determine whether roommates are more than just roommates, the fiscal system could piggy-back on other filings, such as filings to claim partner benefits from employers, in cases where the other benefits exceeded any tax or transfer penalties that might result from acknowledging couple status. (This would, however, affect incentives to claim and indeed to offer such benefits, making piggy-backing a trade-off rather than an unambiguously desirable approach.)
With regard to children, determining household status is more straightforward. Where there is possible uncertainty regarding a child's status as a dependent or regarding the identity of the main supporting parent, the available cues include the child's age, residence, custody, and who provides financial support. Such tests can likewise be used to determine whether other family members, such as siblings or aged parents, are dependents in one's household.29
See Internal Revenue Code section 152, defining dependents for federal income tax purposes relating to the allowance of deductions for personal exemptions, for an example of a rule addressing children's and other relatives' household status.
The general social expectation in the United States of achieving financial independence when one reaches adulthood lessens the importance of tracing household connections between adults and their surviving parents. Nevertheless, the fact that there are economic links, whether through bequests from the older generation or support from the younger, helps make intergenerational tax and transfer incidence a challenging subject. For example, one's estimate of the transfer through Medicare to older generations would be a lot lower if one thought that seniors' adult children would otherwise have borne most of the seniors' medical expenses. The links between adults and their surviving parents also create difficult issues under the Medicaid rule denying subsidized medical benefits to people with assets above a specified level, since generally only one's own assets are counted and the rule therefore encourages asset transfers to family members, such as one's children, as a means of establishing eligibility.30
Intrafamily asset transfers for this purpose became sufficiently prominent that Congress in 1996 enacted a new law creating criminal penalties in cases where the transfers were designed to circumvent the Medicaid asset limit. See 42 U.S.C. sections 1396a and 1396p; and Helen Hershkoff and Stephen Loffredo, The Rights of the Poor (Carbondale, IL: Southern Illinois University Press, 1997), 184.
V. Equal Taxation of Equal-Income Couples
Boris Bittker, the renowned Yale Law School professor and tax expert, first made the point that, with graduated marginal tax rates, one cannot have both (1) a marriage-neutral system, in which the combined tax on two individuals' income is not affected by whether they are married or not, and (2) a regime of equal taxes for equal-income married couples.31
Bittker, “Federal Income Taxation and the Family,” 1395.
(a) If we ignore marriage and have separate returns, then C pays $25,000 of tax while none of the others pays anything. Thus, we violate equal taxation of equal-income couples by failing to tax C and D the same as A and B.
(b) If we have separate returns and mandatory income-splitting, so that C and D are taxed as if they were unmarried individuals earning $50,000 each, then we have equal taxation of equal-income married couples, but we also have a marriage bonus for C and D rather than marriage neutrality. Marriage lowers their collective tax bill from $25,000 to zero.
(c) If we have joint returns, then we will have equal taxation of equal-income married couples, but we will also have marriage bonuses or penalties, depending on where we set the zero bracket relative to that for single individuals. If we set the zero bracket at $100,000 or double that for singles, the result is the same as in (b) above. If the zero bracket is any lower than $100,000, we have a marriage penalty for A and B, who now owe tax that they could avoid by being divorced, but we also have a marriage bonus for C and D so long as the zero bracket for married couples exceeds, by even a dollar, the zero bracket for singles.
Whether this inconsistency is a problem depends on what we think of Bittker's two objectives: marriage neutrality and equal taxation of equal-income married couples. Starting with the second objective,32
On the importance of this principle, compare ibid., 1438, describing proposals to abandon it as “nothing less than astonishing,” with McCaffery, Taxing Women, 25: “Taxing equal-earning couples equally is terribly unfair, because in fact it leads to massive discrimination against women.”
The objection that is devastating where it applies can be illustrated through the preceding example. If C and D with one spouse working can earn as much as A and B with both working, then it seems pretty clear that C and D are better off economically, rather than in the same position as A and B. Suppose both households have young children. D is available to offer childcare during working hours that A and B must pay a lot of money to procure. Or to put it another way, suppose D, like A and B, could earn $50,000 if she (or he, to take the statistically less common case) so chose. We would recognize that C and D were better off than A and B if D took the job and her household therefore was taking in $150,000. But C and D evidently prefer the actual state of affairs, where D does not work, so even at $100,000 they apparently consider themselves better off still.
From a social insurance standpoint, we use an income tax and means-based transfers to address income risk by redistributing from high earners to low earners. However, people who voluntarily choose lower earnings because some nonpaid use of their time has greater value to them presumably think they are better off by reason of the choice. Moreover, the mere fact that one-earner couples are voluntarily earning less need not suggest that their marginal utility for a dollar is greater. Secondary earners within couples, especially though not uniquely when they have minor children in need of extensive care, are a group in which low earnings are unusually poorly correlated with ability to earn.
Thus, the notion that equal-earning couples are relevantly alike and should be taxed the same can be ruled out immediately if we are not comparing likes in terms of their levels of labor market involvement. Even where they are alike in this sense, however, the second objection that I mentioned above must be considered. Suppose we compare E and F, spouses who each earn $100,000, to G and H, where G earns $180,000 and H, despite working full time, earns only $20,000. Keeping in mind that we are interested in evaluating individuals in light of household information, rather than households as such, do we really have the posited equals here? Suppose that both households apply an internal distributional norm under which the earner of a given dollar has extra say about its use. Then it is plausible that, as indicated by the differences in individual earnings, G is better off than E or F, while H is worse off. Unfortunately, however, we cannot use this information to transfer resources away from G and toward H, nor can we even tell which of the two households should gain or lose overall, until we know more about how the G/H household determines its internal resource allocation.33
See Kaplow, “Optimal Distribution and the Family,” 78–80, for a similar view of the significance of unequal sharing within a household.
VI. Marriage Neutrality
The other prong of the dilemma that Bittker considered central to taxation of the household was marriage neutrality, or the idea that the combined tax on two individuals' income should not be affected by whether they are married or not. Under a system that did a good job of identifying unmarried couples who pool and allocate their collective resources, this issue would be transformed into couples neutrality (or neutrality concerning whatever characteristics were now being used to assess couple status). The fact that, at present, the only neutrality practically at issue, at least in the income tax, is that concerning possession of a marriage certificate importantly affects what is at stake in the analysis.
Neutrality is an old but still potent idea in tax policy, reflecting the fact that, under specified conditions, it advances both efficiency and equity. When you are choosing between activities or assets X and Y and will internalize all of the consequences of your choice, tax and transfer neutrality as between the two choices keeps your personal incentives aligned with increasing total welfare. Likewise, under a utilitarian view of equity, if people who choose X are not relevantly different (i.e., in marginal utility of a dollar) from those who choose Y, favoring one group over the other is likely to reduce total welfare under the assumption of declining marginal utility.34
As Bittker famously pointed out, marketplace responses to a non-neutral tax rule may eliminate the inequity problem, leaving only inefficiency. See Boris I. Bittker, “Equity, Efficiency, and Income Tax Theory: Do Misallocations Drive Out Inequities?” San Diego Law Review 16 (1979): 735–48.
Neither of these assumptions is necessarily correct. Even if we disregard most of the reasons why third parties, ranging from parents to friends to romantic rivals to pure busybodies, might care about the marital or relationship status of a given couple, there is the point that children's welfare may be affected. It is commonly agreed that divorce, on average, has bad effects on children, although in abusive households separation may benefit the children if a nonabusive spouse gets custody. Decisions to marry or form a couple may also have external effects through the fiscal system—for example, by easing burdens on other taxpayers where an earning spouse supports a non-earner, or increasing burdens if a worker who would have paid net taxes decides to quit working by reason of spousal support.
The effects of marriage (or forming a couple) on marginal utility are likewise ambiguous. For example, in the case of a high-earner married to a low-earner, if resources in the household are shared more equally than the split in earnings, we may be less inclined both to tax the high-earner and to support the low-earner than we would have been had they remained separate. Moreover, as I noted above, the economies of scale that may result from cohabitation are ambiguous, as they increase our estimate both of the cohabitants' welfare (relevant under the assumption of declining marginal utility) and of their efficiency as consumers.
These ambiguities may tend to push one back toward favoring marriage (and couple) neutrality, effects on children's welfare aside, if only in the weak sense of being unsure in which direction we should lean as a general matter. However, in a system with rampant marriage penalties and bonuses, which do not balance each other out because they occur in very different settings, the neutrality rubric may be too abstract and generalized to help very much. We can better come to grips with the issues by looking at the three main groups involved: one-earner couples (who get marriage bonuses), two-earner couples (who get marriage penalties), and single individuals. The issues of main interest here, which I address in the next two sections, are (1) how these different types of households ought to fare distributionally relative to each other, and (2) the incentive effects on women's labor supply decisions of strongly penalizing the decision to be a couple's second earner.35
A further important issue is the effect of taxation on decisions to marry, involving encouragement of forming a one-earner couple and discouragement of forming a two-earner couple.
VII. Single Individuals versus One-Earner Couples versus Two-Earner Couples
The mechanism by which the income tax yields a marriage bonus for one-earner couples and a marriage penalty for couples with two relatively equal earners is the use of joint returns on which various income-limited tax benefits, such as lower rate brackets, terminate at less than double the income levels used for single individuals. Thus, in terms of my earlier example with individuals A, B, C, and D, it as if the zero bracket of $50,000 for singles were adjusted only to $75,000 for marrieds. Accordingly, A and B, who earn $50,000 each, lose $12,500 per year as a result of being married, while C and D save $12,500 due to marriage if we assume that their labor supply decisions would have been the same in any event.
Adding a complication is the possibility that D would lose welfare benefits if unemployed and not married to C. However, for middle-class individuals with decent employment prospects, this assumption would often be unrealistic, and the income tax bonus from marrying therefore would not be eliminated by looking at a broader set of fiscal rules. In Social Security and Medicare, moreover, it is unambiguous that one-earner couples get a transfer from singles and two-earner couples. Both of these systems provide spousal benefits that nonworking spouses get to claim at retirement, but neither system makes the one-earner couples pay more for the extra retirement coverage. One-earner couples therefore get a two-for-one: two sets of retirement benefits for only one set of payroll taxes. A system that was designed to avoid this redistribution could still provide retirement benefits for nonworking spouses, who have at least as good a chance as anyone else of needing the support when they get to that stage, but it would require that the benefits be paid for. For example, workers might be required to make extra payroll tax contributions when they have nonworking spouses, so that the system would be more actuarially fair between different types of households.
To be sure, actuarial fairness is not an end in itself. Indeed, social insurance is meant to be actuarially unfair, in the sense of transferring expected resources to people who are expected to have greater needs at the time when their participation begins. (Examples include the income tax and means-tested transfers, which favor people with low as compared to high earning ability.) The transfer to one-earner couples may initially look like social insurance for income risk, since part of what triggers it is the nonworker's low earnings. However, this brings up again the point that low earnings are empirically much weaker evidence of bad circumstances, as opposed to rational optimizing amid good circumstances, in the secondary earner setting as compared with other settings. Moreover, while a two-person household has greater total needs than a one-person household, a nonworking spouse may still be an economic producer for the household's benefit, performing tasks that have value and that in some cases would require paying a third party if both spouses worked.
While earnings are therefore a bad measure of relative material well-being for one-earner couples as compared with two-earner couples, one should not too swiftly assume equivalent material well-being. Potential nonworking spouses are more likely to take that path when the earnings they would forgo are low rather than high. Moreover, even where nonworking spouses had good earnings prospects at the start, once they have been on the sidelines they may find that, as a prospective matter, their economic opportunities are limited. Thus, we should not exaggerate the point that one-earner households' relative economic resources are being underestimated if we rely on a measure of earnings. The point retains significant force, however, and might support increasing these households' relative fiscal burdens even without regard to the issue of effects on secondary earners' labor supply decisions (which I discuss next).
VIII. The Secondary Earner Problem
Perhaps the most important defect in existing fiscal rules for households is how they affect secondary earners' (mainly married women's) incentives to participate in the labor market. These incentive effects and their significance have been thoroughly and ably discussed elsewhere, such as by the economist Michael Boskin in the 1980s,36
See Michael J. Boskin and Eytan Sheshanski, “Optimal Tax Treatment of the Family: Married Couples,” Journal of Public Economics 20 (1983): 281–97; and Michael J. Boskin and Douglas J. Puffert, “Social Security and the American Family,” in Lawrence H. Summers, ed., Tax Policy and the Economy, Vol. 1 (Cambridge, MA: MIT Press, 1987).
An initial point to keep in mind is that married women, who most often are their households' secondary earners in the sense of earning less and being less committed to market work, tend to be highly price-responsive in making work decisions. In other words, small changes in their net economic return from work can have large effects on what they decide to do.37
See McCaffery, Taxing Women, 179–82, and sources cited therein. This is especially true for couples with children, because nonmarket work in the home is such an important form of household economic production.
See ibid., 170–75. “Compensated elasticity” refers to that which is measured holding income constant, so as to focus on substitution effects, or responses to incentives at the margin, as opposed to people's changing preferences as income changes.
Instead, we apply much higher marginal tax rates to secondary-earner married women than to nearly anyone else. The main causes are the following:
(a) While a joint return, on its face, applies a single income tax rate schedule to all of the couple's income, this may not be how the couple looks at it if the man is certain to work while the woman faces a genuine choice. Under this circumstance, the woman may view her first dollar of earnings as facing the marginal tax rate into which the man's work was already expected to place the couple. Thus, the tax rate schedule she faces in making labor supply decisions may start at more than 30 percent, whereas for single individuals and primary earners in a couple it starts at zero and gradually proceeds through increasing rate brackets.
(b) Especially in the case where the couple has children, the woman's decision to work may result in the couple's incurring increased expenses, for items such as childcare, commuting, and work-related clothing, that generally are nondeductible for income tax purposes because they are viewed as consumption expenditures.39
Internal Revenue Code section 21 provides a fairly limited childcare credit for two-earner married couples and heads of households with children, which I ignore in the text for ease of exposition.
(c) Social Security may add to the problem, due to its 12.4 percent tax rate (counting both the employer's and the employee's nominal shares) on earnings up to an annual ceiling. For primary earners and single individuals, this tax is somewhat offset by the fact that extra Social Security earnings may increase the retirement benefit for which one would qualify under present Social Security law. In effect, then, the wage tax is somewhat offset by a wage subsidy, in the form of retirement benefits that are based at the margin on earnings, although it is unclear to what extent workers take this into account when making labor supply decisions. If one is a secondary earner, however, there is no wage subsidy (in the form of increasing benefits) counteracting the wage tax until one reaches the point where one's expected own benefits exceed one's expected spousal benefits.
The end result, very high marginal tax rates affecting decisions by people who are highly tax-responsive, is a significant efficiency problem in a straightforward labor market sense, and also raises additional issues. The problem becomes a lot worse if one believes that reducing women's workforce participation has important adverse effects on our society as a whole that lie outside each worker's individual calculus in making labor supply decisions. Reduced workforce participation might, for example, entrench women's subordination, gender stereotyping, and gender bias in labor markets while reducing the availability of part-time work (by reducing employers' capacity to benefit from restructuring their workplaces to accommodate it).40
See McCaffery, Taxing Women, 240–64; and Alstott, No Exit.
IX. Children and Distribution Policy
The topic of how the presence of children in a household should affect its taxes and transfers is, if anything, even more unsettled, albeit not as intensely controversial, as the various issues discussed above relating to adult couples. The issues are sufficiently complex to merit separate discussion, first of distributional considerations and then of possible incentive effects on having children, before evaluating the policies that we actually have.
A. The case for transfers to households with children (or more children)
Suppose initially that having children was entirely a random event, not reflecting any element of choice by those who became parents. I start with this assumption simply to isolate the static distributional effects before I turn to the incentive effects.
Under this scenario, people would likely want insurance against the risk of having children, whether they wanted the children or not, because having them would be economically costly in various respects. The case differs from that of affiliating with a nonworking adult partner in a new couple, because children cannot, for many years, be economic producers. Thus, unlike a nonworking spouse, they bring extra mouths to feed without also bringing extra pairs of hands that can be used productively, either in the labor market or in housework.
One further complication to the analysis is that children may be able to help support their parents after growing up. However, parents cannot borrow against their children's future earning capacity, except in the very limited sense that, if they are feeling very brave (and have not read King Lear), they might save less of their own resources for retirement based on this expectation. The proper way to account for children's earning capacity, therefore, is through consideration of the degree to which transfers to seniors should take account of their adult children's resources.
The case for the desirability (in marginal utility terms) of significant transfers to households with children (or with more children than other households) should therefore be clear enough if one puts oneself in the shoes of a prospective parent and asks how it would affect one's economic wants. (Considered purely at the parental level, the case arguably is one of lateral redistribution, since a voluntary consumption choice to have children, even if costly, presumably is expected to make one better off.) Yet focusing just on parents undervalues the overall case for the transfer. From behind the veil of ignorance, you also might be the child, who also would likely prefer transfers to her household. From a utilitarian standpoint, it is not “double-counting” to value the child's welfare both directly and via the parent's altruistic preferences, since one is simply, as usual, counting the distinct preferences of each individual in the society.
Why, then, don't we observe private insurance markets offering coverage that pays off if you have children? To a small degree, we actually do observe this on a group basis, because employers frequently offer employees health insurance and other benefit packages that become more valuable as family size increases. Free public schools can also be seen in this light, if we think of neighboring jurisdictions as competing for residents in classic Tiebout fashion41
See Charles M. Tiebout, “A Pure Theory of Local Expenditures,” Journal of Political Economy 64 (1956): 416.
Large-scale governments can solve this adverse selection problem by requiring all residents effectively to enroll in an implicit insurance program that uses highly child-adjusted taxes and benefits to provide the coverage. This, however, raises the question of moral hazard, which governments often cannot solve much better than private firms. Here moral hazard takes the form of having more children because of the coverage, thus increasing the payoff from households that are childless or have fewer children.
B. Possible incentive effects on having children
Suppose that we believe people would respond to the transfer policy described above by having significantly more children. While this is a necessary prerequisite to discerning an important moral hazard problem here, it is far from sufficient. The externalities of having a child are so great that it is difficult to see the parent's personal calculus of the costs and benefits as a good proxy for the social costs and benefits.
I start with two important externalities suggesting that, if anything, prospective parents' incentive to have a child is too weak, rather than too strong. Fiscally, from the standpoint of people in other households, if the child is a net taxpayer during her life, her birth is likely to be a net revenue benefit over time, despite the transfers to her household early in her life. Thus, other households may benefit financially, on balance, from offering the implicit insurance coverage. Lifetime measures typically show that nearly everyone is a net taxpayer. Concededly, this could be misleading, because it involves ignoring the value of public goods that, even if highly invariant in the cost of provision to one extra person, may need to be gradually scaled up as population size increases. However, in the United States fiscal system and that of most other affluent countries, the financing structure of retirement programs such as Social Security and Medicare adds a positive revenue externality to population size. These programs were set up to provide large transfers to the initial cohorts of participants, financed by passing the burden forward to younger generations. This structure creates a general fiscal benefit to increasing population size. The benefit is divided between the members of different age cohorts based on how taxes and benefits are adjusted.
Second, and perhaps a bit more fundamentally, consider the benefit to the child of being born. Again, from a utilitarian standpoint this benefit should count independently of the parent's altruistic or other interest in having a child. The fact that this line of argument might lead us, on total welfare-based utilitarian grounds, quite far past the idea of merely treating households with children favorably in the fiscal system, is food for thought, but not a refutation of the argument as used in this more limited way.
Admittedly, this barely starts the process of examining the welfare consequences of increasing birthrates. Important negative externalities that come to mind are congestion and the potentially dire environmental effects of population increases. There may also be important effects of various kinds on wage and price levels. It seems clear, however, that the parent's cost versus benefit calculus in deciding whether to have a child falls very far short of capturing the social metric. So neutrality as to this decision has little appeal as a normative benchmark. A more pertinent question is whether we think the overall government policy should be pro-natalist or anti-natalist. If one concludes that it should be anti-natalist, then concededly there is a trade-off posed by favoring households with children to the degree suggested by the straight distributional or social insurance analysis, and the size of the transfer should be, at the least, reduced (depending on the elasticity of having children as well as the significance of the anti-natalist considerations).
Other than under that scenario, however, the case for significant transfers to households with children (or more children) remains strong, and might even be stronger than that suggested by the distributional analysis alone, if one is pro-natalist and believes that the behavioral response would be significant. It is therefore worth asking to what degree actual United States policies conform to the suggested approach.
C. Current policy toward households with children
If someone who was familiar with American political rhetoric but not with the actual details of the United States fiscal system were asked to guess whether we have the generous policy toward households with children that I suggest above, that person would very likely guess that we do. After all, concern for children is, if anything, an even more widespread sentiment than dislike for marriage penalties. While children's issues share some of the marriage penalty issue's culture-war overtones of crusading for traditional families, their appeal is arguably broader. For example, Democrats such as Hillary Clinton are no less eager than conservative Republicans to advocate pro-child policies, although the style of the policies they propose may differ.
Despite these reasons for expecting consistent pro-child policies, actual United States fiscal policy presents a mixed picture. At the bottom of the income distribution, households with children are indeed very strongly favored relative to those without children. Examples include the earned income tax credit for low-wage workers, which becomes much more generous as the number of children in a household increases from zero to two, and various welfare benefits, such as TANF, food stamps, and Medicaid, that aid children and/or households with children. In addition, public schools offer an important child benefit—free education—that, while not expressly income-linked (and probably tending to improve in quality as neighborhood income rises), is used less by higher-income people who can more easily afford private school.42
See C. Eugene Steuerle, “Can the Progressivity of Tax Changes Be Measured in Isolation?” Tax Notes 100 (2003): 1187–88. As Steuerle notes, however, higher-income households get greater college education subsidies than lower-income households.
As income rises, however, the relative benefits for households with children tend to shrink or even disappear. In the federal income tax, for example, personal exemptions and child tax credits are allowed to taxpayers who otherwise would have positive income tax liability, but both benefits are reduced or phased out completely as adjusted gross income rises through the low six-figures range. Moreover, there is an ongoing trend toward increasing the relative tax burdens of households with children in the low six-figures range, through the rapidly rising applicability of the alternative minimum tax, which offers no adjustment for dependents.
This limitation of the policy favoring households with children to lower income levels is hard to rationalize. From a utilitarian standpoint, the distributional issues posed by poor households are clearly continuous with those posed by households above the poverty level. Take two households with the same economic resources and add children to the first but not the second, and the result will be that the first has fewer resources than the second relative to its needs. The children need not be facing the threat of inadequate food or shelter in order for this to be true. Thus, there is a strong argument for favoring households with children at middle and upper income levels to a degree that is roughly comparable to that prevailing with respect to poor and near-poor households.
X. Conclusion
Although this essay aims to be conceptual and exploratory rather than being focused on concrete policy proposals, it may be worth describing the main types of policy changes that its analysis suggests would likely be desirable. These include the following:
(1) There should be greater consistency in the rules used by different parts of the fiscal system to identify couples, since the distributional issues posed by, say, TANF and income taxation are similar. Since marriage is an increasingly poor marker of couple status, other indicators of couple status should be used as well. These could include domestic partner–type statutes provided by the states, or piggybacking on claims of couple status with respect to employee benefits. Non-couples' incentive to make false claims of couple status where it was favorable would be an admitted problem, as would the issue of the government's asserting couple status where such status was denied by the individuals involved because it was unfavorable. Despite these problems, however, it is difficult to see marriage as retaining, in the twenty-first century, its adequacy as an exclusive marker of couple status.
(2) The United States fiscal system's present discouragement of work by secondary earners in couples, along with its distributional bias toward one-earner couples, should be mitigated. One means of doing this might be to require one-earner couples who are accruing spousal benefits in Social Security and Medicare to owe a minimum contribution for the lower earner, which might be collected through the income tax return to the extent it exceeded the payroll tax actually paid for the year with respect to the lower earner. Paid-for spousal benefits might even be defined as belonging, for future benefit computation purposes, to the spouse to whom they relate in the event of divorce, possibly offering the added advantage of increasing that individual's ability to leave the relationship without as much concern about facing future destitution.
The biases within the income tax in favor of one-earner couples and against work by secondary earners could also be addressed—for example, by reducing the tax burden on the earnings of the lower-earning spouse. One way to do this would be through an income tax exclusion for the first $X of this individual's earnings, similar to that which existed in the federal income tax from 1981 through 1986. In addition or alternatively, one could increase deductions or other tax benefits (such as credits) for the work-related expenses, such as childcare, of two-earner couples.
A further, frequently discussed possibility is eliminating joint returns. As I have noted, however, the official filing unit is mainly of administrative interest, rather than being as fundamental as is sometimes thought. Making use in some fashion of household information in tax and transfer filing seems unavoidable no matter how the filing is handled. And the aim of avoiding the first-dollar taxable rate effect on secondary earners that results under present law could be accomplished, with or without joint filing, through special rate schedules such as the 1981 to 1986–style secondary earner exclusion.
(3) Child benefits within the income tax probably should be increased and should not be subject to phase-out. In addition, if the alternative minimum tax is retained, it should be amended to allow child benefits. In principle, all this should be done on a revenue-neutral basis and without reducing overall progressivity, since the idea would be to change how we redistribute as opposed to addressing the separate question of how much should be redistributed. Obviously, the present budgetary environment, along with the political difficulty of imposing losses on anyone relative to prior law, makes this difficult, but it is an aspiration to keep in mind and one that, in principle, can raise just as much revenue as the existing system at a comparable efficiency cost (or, if at a greater efficiency cost, then in reasonable exchange for improving distribution policy).
