Recently several scholars have argued that tax preferences are equivalent to welfare spending, and that our understanding of welfare states should be revised to take tax preferences into account. A tax preference is a provision in the tax code (an exemption, deduction, or credit) by which taxpayers are able to reduce their tax obligations in return for behaving in some specific way. For example, the mortgage interest tax deduction allows taxpayers to deduct the interest on their mortgage payments from their taxable income, thereby reducing the taxes they would otherwise owe as long as they are making mortgage payments.

The tradition of equating tax preferences with welfare spending is several decades old. Stanley Surrey, then assistant secretary of the Treasury, compiled a list of tax preferences for the first time in the 1960s, arguing that they should be examined on the same basis as welfare spending. His argument was that whether welfare is accomplished by first transferring revenue from society to the state and then back to society (welfare spending), or whether it is accomplished by foregoing the tax revenue in certain targeted ways (tax preferences), is equivalent in accounting terms, and should therefore be analyzed equivalently. Surrey insisted on calling tax preferences “tax expenditures,” to highlight their supposed similarity to social expenditures, and this terminology persists today. In this paper, for reasons that will become clear, I avoid the “tax expenditure” terminology and use the term “tax preferences.”

One of Surrey’s reasons for focusing on tax preferences was that they disproportionately benefited the upper tax brackets. By shining bureaucratic and political light onto them he hoped to bring the practice of tax preferences in line with public opinion regarding the operation of the tax code, and preferably to replace them with direct spending or to eliminate some of them altogether. The Treasury began to compile annual lists of tax preferences, and tax preferences became subject to congressional scrutiny. Surrey’s efforts and his list called attention to the large amounts of revenue that were lost because of tax preferences, and in the intervening decades they have become only more important.Footnote ¹ Surrey and scholars who followed in his footsteps identified a host of reasons why direct spending was preferable to tax preferences, as summarized in a recent article: “Tax expenditures increase tax complexity and compliance burdens (administrative concerns); they are more opaque and esoteric than direct expenditures and so relatively more immunized from public debate or scrutiny (visibility concerns); they are more likely than direct expenditures to have regressive first-order effects (equity concerns); their uses may evolve in unexpected ways (predictability concerns); and they may stimulate unproductive activity or reward behavior that would have occurred anyway (efficiency concerns).”Footnote ² Although these claims have been questioned by others, many tax practitioners inside and outside government continue Surrey’s call to replace tax preferences with direct spending.

In the 1990s several scholars brought this analytical tradition into policy history, with Christopher Howard and Jacob Hacker being the most well-known proponents of it. Howard’s brilliant and path breaking book, The Hidden Welfare State, carefully traces the history of four major tax preferences in the American tax code: the tax preferences for home mortgage interest and employer pensions as well as the Earned Income Tax Credit and the Targeted Jobs Tax Credit.Footnote ³ Howard shows that the politics of tax preferences are distinct from the politics of social spending in that tax preferences are introduced with less struggle than direct welfare spending, often with the support of conservative politicians, and without as much input from interest groups.Footnote ⁴ Howard explicitly anticipates the argument that Jacob Hacker would later make famous,Footnote ⁵ that tax preferences for health and pensions led to the proliferation of a “private” welfare state in the United States, which then undermined attempts to develop a public welfare state.Footnote ⁶

Howard has recently pushed this argument to its limits, arguing that “the American welfare state is actually 40 percent larger than commonly believed. Imagine a map of the United States that started on the Eastern seaboard and extended west to the Rocky Mountains, then stopped. That is what the American welfare state would look like without tax expenditures.”Footnote ⁷

The concept of tax preferences equaling welfare state spending has become extremely popular, particularly in analyses of American political development. It has been cited in hundreds of books and articles, across disciplines, and has begun to anchor a revisionist interpretation of the American welfare state. While research on the main preferences that Howard identified continues, more recent work has taken the argument into new areas of the tax code. For example, David Pozen has attempted to rethink the concept of foreign aid by analyzing “hidden” forms of foreign aid in the tax code, concluding that including tax preferences for foreign aid would raise the estimation of foreign aid that goes to developing countries from the United States by 20 percent, between $1.06 billion and $2.12 billion.Footnote ⁸ Meredith Stead examines disaster relief provided through the tax code in the aftermath of Hurricane Katrina as a form of welfare spending.Footnote ⁹ Erin L. Kelly examines the effect of tax preferences for child care in the 1981 Economic Tax Recovery Act and Kimberly J. Morgan examines the earlier history of child-care tax deductions.Footnote ¹⁰ William Novak cites tax preference scholarship as one piece of his argument that the American state is not “weak” at all.Footnote ¹¹ James A. Morone writes similarly that “behind the scenes, an active state promulgates rules, writes regulations, and offers incentives [including tax preferences] that organize and maintain the private-sector safety net.”Footnote ¹²

The concept that tax preferences are equivalent to welfare state spending has been extremely important in bringing attention to corners of the tax code that have hitherto remained obscure, and to the degree that equating tax preferences to welfare states has helped to bring scholarly attention to taxes, it has served its purpose. But the equation may now have begun to obscure as much as it reveals.

There are indeed important similarities between tax preferences and welfare spending. For example, both allow politicians to give things to constituents, and sometimes to do so at their discretion, which can be an important tool for politicians in gaining things that they want, such as campaign funding. And as Surrey argued, tax preferences and welfare state spending are both ways for the government to intervene in the market by favoring some and disfavoring others. But many other policies can have these effects; what makes tax preferences different, and what makes them equal welfare state spending is, as Surrey wrote: “A dollar is a dollar—both for the person who receives it and the government that pays it, whether the dollar comes with a tax credit label or a direct expenditure label.”Footnote ¹³ At the end of the day, whether the government has given me a dollar, or refrained from taking a dollar that I already have, I still have an extra dollar that I would not have otherwise had. Moreover, the state can first take x dollars from me in taxes, and then give me x dollars worth of health care; or the state can give me a tax credit of x dollars—that is, refrain from taxing me that amount—if I use those dollars to purchase health care. In both cases, I have lost x dollars and received x dollars of health care, and all other aspects of the situation are the same. The argument is clear, logical, and unimpeachable.

Still, there is something unsettling about the equation of tax preferences and welfare state spending, since it requires us to argue that not taking equals giving, that is, that the state’s not taking tax revenues from citizens equals giving them welfare spending. If I steal your bicycle, and you complain to the police, I cannot reply that I did not take your car, which is equivalent to giving you a car, and having taken a measly bicycle in return is small recompense. Moreover, if tax preferences are the same thing as welfare state spending, then larger tax preferences equal a larger welfare state. Following out this analogy, this means that the more tax revenue foregone—the less money actually collected—the larger the welfare state. At the extreme, this means that the largest welfare state would be the one that collects no taxes at all. Since the scholars who make the tax preferences argument have not given us any reason why the logic should not be taken to its limits, by their logic the largest welfare state would be the one that has a 100 percent tax rate and a 100 percent tax preference, and no actual welfare spending. This seems to equate a welfare state with the complete absence of a welfare state. Can this be the case? Can a state equal the absence of a state?

It seems that the answer to that question should be no. It seems to be a reductio ad absurdum that points to some flaw in the logic of the equivalence of tax preferences and welfare state spending. But things are not quite that simple, and it is worth examining this question in more detail, as doing so teaches us not only about the nature and purposes of tax preferences, but also about the nature and purposes of taxes, and of welfare states. For answering this question requires us to answer: In what ways, specifically, does a welfare state not equal the absence of a welfare state? In what ways, specifically, is not taking different from giving?

To address these questions, let us begin with the example of the stolen bicycle and the “preference” for the car. While it seems on the surface not to be logical, there is one kind of world in which the logic holds: a world in which the thief has rights to the car. Imagine that we are all members of a car cooperative, and I happen to know that you want the use of the car at some particular time. Imagine that, because I know that, I refrain from using the car even though I have the keys and could take it away before you get here. In that case, not taking the car is the equivalent of giving you something, because I have given you use of the car even though we both have rights to it. This example may seem far-fetched, but in fact it brings us to the heart of the question about tax preferences, because it asks us to consider who has rights to the tax revenue that the state has foregone in a tax preference. As Ronald Reagan pointed out, inherent in the idea of a “tax expenditure” is the belief that the government has rights to the money that has been exempted: “When you hear the phrase ‘tax expenditure’ do you think it means some govt. spending program? Wrong! That is Wash.’s term for money it lets you keep. . . . The govt. calls it govt. spending when it allows you to keep your own money.”Footnote ¹⁴ So one question in asking whether tax preferences equal welfare state spending is whether the state has a right to the tax revenue it is giving up, and a positive answer to this question is built into the equivalence argument.

Now, one clear reason to argue that the state has a right to the revenue is if the state is providing something in exchange for the revenue. This is the implicit or explicit ethical justification for taxation in the first place—that the state uses that revenue in ways that benefit the society at large, if not the specific taxpayer. Even political theories that point out the socially constructed nature of the idea of private property, such as the recent contribution by Liam Murphy and Thomas Nagel, assume thisFootnote ¹⁵; indeed, I know of no political theory that argues that taxation that benefits only state actors is justified. If this is the ultimate justification for taxation—that taxes provide benefits for society—then giving up the revenue is equivalent to spending if the state continues the provision of the services the revenue was financing. Of course, since tax revenues are usually fungible, in most cases it is not the provision of one specific service that is at issue, but the provision of everything the state does—state capacity in general—that is at issue. In the absence of the tax revenue from the preference, the state may finance this capacity in other ways, such as through other taxes or through borrowing. If we can show that state capacity continues through increased taxation of other kinds, then we do have a situation of welfare spending that fits the traditional definition of welfare spending—collection of some taxes for provision of some services—and the tax preference just shifts around the ultimate liability. If we can show that state capacity continues through borrowing, then we have a situation of welfare spending financed by future generations, who will have to repay the borrowing or live with the lower economic growth caused by not repaying it. But if it is not possible to show that state capacity continues unharmed—if the cut in taxes means a cut in spending somewhere—then we do not have a situation of social spending at all, but of the end of social spending, and of reprivatization of tasks that were previously performed by the state. In practice, such demonstrations would require tricky and perhaps impossible counterfactual exercises, but in theory the first case in which a tax preference is unequivocally the same thing as welfare spending is if the state continues whatever it was doing with the tax revenue, or preserves its capacity in general, through other means of financing.

But there are also ways in which tax preferences can equal welfare spending even if one is not committed to the perspective of the state’s right to the foregone tax revenue. This can be seen by examining what it is that welfare states do, and whether tax preferences accomplish similar functions. I noted above that the general insight driving the equivalence argument is that whether welfare is accomplished by first transferring revenue from society to the state and then back to society, or whether it is accomplished by foregoing the tax revenue in certain targeted ways, is equivalent in accounting terms. But, of course, welfare is not simply a matter of first taking revenue and then giving it back. It is a matter of taking revenue from some and giving to others, or of taking revenue and spending it in ways that could not be done by individuals, such as on public goods. Thus, tax preferences are equivalent to a welfare state only to the degree that they can approximate these basic functions of the welfare state.

There is no consensus on what the basic functions of the welfare state are, but the vast literature on the welfare state does provide some guidance on this question. Some definitions of the welfare state are less useful than others for our purposes. For example, it would not be particularly enlightening to define welfare as direct provision by the state in cash or kind, and then to conclude that tax preferences are not welfare because they are not direct provision by the state in cash or kind, because the conclusion simply follows from the definition, without providing analytical insight. By contrast, a maximal definition that defines the welfare state as any state policy that affects the welfare of citizens is excessively broad, because it would include tax preferences as part of the welfare state as simply a matter of definition.Footnote ¹⁶ To move beyond this kind of argument by definitional fiat, we need to answer more carefully the question asked above: How is a welfare state not equal to the absence of a welfare state? That is, what does the presence of a welfare state change that the counterfactual absence of the welfare state would not?

The literature on the welfare state gives two clear answers to this question. First, the welfare state redistributes resources—not necessarily from rich to poor, but from young to old, from healthy to unhealthy, and from lucky to unlucky, as Arthur Stinchcombe has put it.Footnote ¹⁷ This can have the effects of reducing poverty and inequality and of “decommodifying” labor.Footnote ¹⁸ One test for whether a tax preference equals a welfare state, then, is whether it can be redistributive. Collecting taxes and then spending them allows for redistribution of wealth. But simple tax preferences can only be given to those who are paying taxes to begin with. Foregoing tax collection in targeted ways does not apply to those who are not paying taxes, or who are paying less than the amount of taxes to be foregone. Consider, for example, two states—let’s call them Exemptia and Welfaria—that provide education through preferences and welfare spending, respectively. In Welfaria, 10 percent of GDP is collected in taxes for the purposes of financing public schools and colleges. In Exemptia, citizens are required to spend 10 percent of their income on education, but no tax is collected and no spending is made for this purpose. In Exemptia, private schools for the wealthy will proliferate, offering well-trained teachers and enriching curricula for those who can afford it, and stingy services for those for whom 10 percent of income is a paltry sum; for those who do not have any income—who live off the state, for example—it is zero, and they will not benefit from the tax preference. It seems clear that in this case welfare states and tax preferences have very different effects.

Some tax preferences, however, do redistribute, with the Earned Income Tax Credit (EITC) being the most prominent current example: the EITC goes into negative territory, that is, taxpayers receive a payment if they are eligible for a tax credit that is greater than the amount of taxes they owe.Footnote ¹⁹ Refundable tax preferences of this kind—tax preferences that disburse money to those who owe fewer taxes than the amount of the preference—are indeed the same thing as welfare spending, and the tax preferences literature has done an enormous service to social science in calling attention to them. The fact that they are accomplished through the tax code has important political consequences, but the basic logic is of welfare spending since a refundable tax preference can accomplish the redistributive goals of a welfare state. But very few tax preferences are of this kind: in the United States, only the EITC and the Child Tax Credit (CTC) are refundable or partially refundable.Footnote ²⁰

The second main thing the welfare state does is pool risks.Footnote ²¹ Even putting aside the question of redistribution, the great accomplishment of the welfare state in the twentieth century has been the ability to stabilize life chances by pooling risks on a society-wide scale. François Ewald’s magisterial history began this line of thinking about the welfare state in the social sciences,Footnote ²² and it has seen a recent revival. As Jacob Hacker notes, “Not only are the largest social programs—pensions, health insurance, unemployment compensation, survivors’ benefits—centrally about insuring against risks to income, but also many aspects of the welfare state that we do not typically think of as risk protection (such as child care and worker retraining) contain important insurance elements insofar that they cushion families against the income shock of major life events.”Footnote ²³ In Peter Baldwin’s terms, “With the development of the welfare state, society decisively improved on its ability to decide autonomously its rules of association without regard to the dictates of nature, fate or circumstance.”Footnote ²⁴

But risk pooling can only be accomplished through the collection and aggregation of revenues. Tax preferences are a re-privatization of that collective risk. Imagine that in Welfaria, 10 percent of GDP is collected in taxes for the purposes of financing health care spending. In Exemptia, tax preferences are given for spending on health care of up to 10 percent of income, but no tax is collected and no spending is made for this purpose. Citizens of Exemptia may seek out savings instruments that allow them to save their 10 percent against the time when they need them, but of course in this case those with greater income will be more able to withstand health risks than others. Citizens of Exemptia may understand the benefits of risk pooling, but if so, they will also understand that their own benefits will be highest if they can exclude from the risk pool those most at risk of needing benefits and include those who are wealthy. Consequently, instruments will proliferate that offer risk pools limited to those who are healthy and rich, and these private instruments will offer much higher benefits to their members, with the risk pools of the poor and the sick limited to sums that may not cover their health expenses at all. This example suggests that tax preferences are not equivalent to welfare state spending unless the tax preference is given in exchange for participation in a scheme of insurance that approximates a society-wide risk pool.Footnote ²⁵

Using this tripartite schema to categorize tax preferences—of rights (whether the state has rights to the revenue foregone, for example, because it is continuing to provide a service that was financed by that revenue), redistribution (whether those whose tax obligations are lower than the amount of the preference receive tax credits), and risk pooling (whether the tax preference is coupled with a requirement to behave in ways that approximate the society-wide risk-pooling features of a welfare state) —leads to a different picture of how to think of tax preferences. In this framework the Earned Income Tax Credit qualifies as welfare spending, as the EITC is refundable, and therefore can redistribute. And while tax preferences for employer-provided health care and pensions are similar to welfare spending, as they do approximate the functions of the welfare state by pooling risks, they are not equivalent to welfare spending because those risk pools remain limited to the employed and their dependents rather than extending to the whole population. The most important tax preference that would not qualify as welfare spending at all is the mortgage interest tax deduction, because it is hard to argue that the money that is lost from it is recovered in some other way—indeed, it seems to be a central source of reduced state capacity in the United States, even despite high levels of borrowing—and because it only applies to those who are paying taxes and does not require risk pooling.

So far, I have given three ways in which tax preferences can be considered to be equivalent to welfare states, and I have argued that except for these three cases, tax preferences should not be equated with welfare states. The argument so far would perhaps not disturb scholars of tax preferences very much. They would surely argue that calling attention to these distributive asymmetries is precisely the purpose of their focus on tax preferences, and they might argue that these are just definitional or analytical quibbles that do not have much to do with the empirical analysis of political phenomena.

But there are three empirical issues that are affected by these arguments. The first is how to interpret Republican and conservative support for tax preferences. Christopher Howard notes that Republican lawmakers have supported tax preferences, and that even the peak moment of Republican backlash, 1994’s Contract with America, which would have slashed all kinds of welfare, included a child tax credit. If tax preferences equal welfare spending, then “[t]his doesn’t exactly sound like the party of small government,” Howard comments; it sounds like a party that supports the welfare state.Footnote ²⁶ Similarly, several scholars have noted the role that conservative politicians and business interests played in support for the EITC.Footnote ²⁷ Howard resolves this “paradox” by arguing that Republican lawmakers end up supporting the welfare state because Republican voters do: “Republican officials have had to find a way to cut taxes and spend more on social programs” (96). But it is also possible that Republicans have embraced tax preferences not because they are trying to find secret ways to support the welfare state, but because they see tax preferences as contributing to their overarching and very public goal of cutting taxes. Tax preferences are a way to nibble away at state capacity, and particularly the redistributive and risk-pooling functions described above. Tax preferences are popular because they have given Republicans a way to be true to their principles of smaller states while delivering targeted benefits—in the form of lower taxes rather than higher spending—to particular constituents to maintain electoral popularity.

Another, more consequential empirical issue affected by this analysis is the origin of neoliberalism. The problem here is easiest to see in the work of Isaac Martin, who argues that the California property tax revolt of 1978, which has come to be seen as the symbol of American antigovernment sentiment, should actually be seen as the exact opposite, a demand for protection from the market. Martin extends the tradition of tax preference research to argue that informal tax privileges—in this case, the practice of “fractional assessment,” the nearly universal (and informal) practice of taxing homeowners on only a fraction of the market value of their property—should also be seen as part of the welfare state. The property tax revolt resulted, according to Martin, when modernization of the property tax ended the tax assessor discretion that had allowed fractional assessment. For the first time, homeowners were expected to pay taxes based on the full and fluctuating market value of their homes, and they rebelled against this expectation. Thus, if informal tax privileges such as fractional assessment are part of the welfare state, then the property tax revolt was really an attempt to defend the welfare state.Footnote ²⁸

Martin’s book is clearly the definitive history of the property tax revolt, and as such, it is the definitive history of the event that many believe kicked off American neoliberalism, the central political phenomenon of the past three decades. It is a careful analysis of a contentious episode, a brilliant and beautifully written piece of research that will be discussed for years to come. Nevertheless, the argument that the politics of tax cutting at its origins had nothing to do with a revolt against the state—that it was rather a defense of social protection—is unconvincing. It does not seem to be the case that the services the tax paid for continued despite the preference (this judgment is difficult to make in the absence of targeting, but we do know that after permanently foregoing those taxes California entered a decades-long revenue crisis); the preference was not available to those who owed fewer taxes than the amount of the preference; and it was not approximating the risk-pooling functions of the welfare state.

One might attempt to save the argument by arguing that home-ownership is the American form of the welfare state, as home-ownership serves as a source of savings that can aid citizens’ well-being in the short term, as well as collateral that can aid capital accumulation and welfare in the long term.Footnote ²⁹ Thus, one might argue that this tax preference allows those who receive it to use the money to finance their retirements and their health care. But this would exaggerate the term “welfare state” beyond utility, as it would simply dub any pattern of political economy a “welfare state.” By this logic, a government that forces citizens to save for their own retirements is a “personal savings welfare state” and a government that passes legislation requiring citizens to pay out of their own pockets for health care is providing “social protection.” In fact such legislation simply makes citizens dependent on the resources that they can raise from the market. Similarly, policies to subsidize home ownership reinforce the market processes that allow some citizens to achieve home-ownership in the first place.Footnote ³⁰ They do not allow for redistribution to those who are not paying taxes, and they do not require the collectivization of risk. Of course, individuals may voluntarily choose to spend some of their tax savings on charity or on health or pension plans—just as individuals benefiting from any tax cut may choose to do so—but we cannot call tax preferences part of the welfare state for this reason unless we are willing to call all tax cuts part of the welfare state.

As for the property tax rebels, what they were willing to tolerate was lower tax payments for themselves. This is not the same thing as tolerating the redistribution and collectivization of risk that are the primary achievements of the welfare state or of social protection as we traditionally conceive of it. To call a preference for a lower tax payment the same thing as a preference for social protection seems to lose any analytical content these terms may have—since all reductions in taxes change the distribution of who pays taxes, and since all reductions in taxes subsidize the savings of those whose taxes are reduced, by the logic of Martin’s book all reductions in taxes are social protection. It makes more sense to argue that the abolition of informal tax privileges should be seen as increasing the size of the state, because it increases the state’s capacity to redistribute and to pool risks. Protests against that abolition—“neoliberal” protests—should then be seen as protests against state capacity.

That distinction is important to maintain for analytical purposes if we want to understand the origins of neoliberalism. Martin’s argument leads to the conclusion that the tax revolt at its origins was not caused by a revolt against the state, or to the conclusion that it was a contingent phenomenon. But if we reject the equation of tax preferences and welfare states, then we go back to having to ask why Americans were revolting against the state, which leads us to discover—for example—that most other countries do not rely so heavily on property taxes to begin with, and that property taxes are more visible than the kinds of taxes that other states rely on, such as sales and payroll taxes.Footnote ³¹ As Milton Friedman noted, in the United States the property tax “is the only remaining tax for which individuals write a single large check. The collection of all other taxes has unfortunately been rendered painless.”Footnote ³² These broader questions and insights are obscured by the attempt to redefine lower taxes to equal social protection or a “larger” welfare state.

The third empirical issue that is affected by the equation of tax preferences with welfare states is the question of how to understand the American welfare state. Peter Baldwin summarizes what is emerging as the new orthodoxy on the American welfare state as “different from those of northern Europe without necessarily being less extensive” because “the percentage of GNP devoted to social policy, broadly understood, is much higher than that which is channeled only through the state. In its totality, the amount of money spent through direct and indirect social policies (statutory, local, and private) bears comparison with most European nations. American policies are, however, marshaled and employed differently. Assistance is distributed more through tax expenditures (credits, exemptions, loopholes) than elsewhere, and it is more reliant on private and voluntary measures.”Footnote ³³ Stephen Pimpare calls it “narrowness” to consider the welfare state as consisting of direct expenditures rather than tax expenditures.Footnote ³⁴ Jens Alber argues that “[o]nce the impact of taxes and publicly mandated schemes is taken into account, the United States no longer falls far behind most European countries but moves closer to the middle of the pack, becoming almost indistinguishable from such European countries as Spain, the Czech Republic, Poland, or the Netherlands and far ahead of Ireland and Slovakia.”Footnote ³⁵ And through the work of Willem Adema, the OECD has committed to producing estimates of welfare states that include tax preferences.Footnote ³⁶

The equivalence argument has had consequences, and is anchoring a reinterpretation of the American welfare state as “larger” than traditionally believed, a reinterpretation that has now gone beyond academic walls. Consider an instructive debate that erupted in the public sphere in early 2010. Economic historian Price Fishback used the tax preferences argument to argue in an unpublished but widely available paper that “Americans spend more per capita on social welfare than do any of the Nordic countries.”Footnote ³⁷ Fishback was careful with his argument, including only tax preferences for specific social purposes and explaining why these tax preferences have different effects than welfare spending. But when he also presented the argument on a blog on the New York Times website,Footnote ³⁸ it was seized by websites on the right of the political spectrum, including from researchers affiliated with the American Enterprise Institute and the Cato Institute, who were not as careful as Fishback. They argued that “when it comes to caring for the poor and least advantaged, the United States currently has little it should apologize for” and “the U.S. welfare system is very generous. The hackneyed stereotypes are wrong.”Footnote ³⁹ The tax preferences argument has come full circle politically: Surrey wanted to call attention to tax preferences because he was disturbed by the fact that they favored the wealthy. But through the rock tumbler of political rhetoric, his argument is now being deployed to argue that the United States does not need a more extensive welfare state.

Before proceeding further down this path, it may be helpful to take a step back, and remember that the logic of all these statements requires the conclusion that the largest state is the one that collects no tax revenue at all—the one with a 100 percent tax rate and a 100 percent tax preference. If tax preferences make the American welfare state “no less extensive” than its European counterparts, then even larger tax preferences would make it even more extensive, and a 100 percent tax preference—no taxes collected at all—would make it the largest welfare state of all. Such a state would move to the head of Alber’s pack, and such a state would show up as the leader in Adema’s calculations. If we do not accept the conclusion that a state that collects no taxes at all can be the largest welfare state of all, then we cannot accept any of these statements that equate tax preferences to welfare states. If we accept the definition of a welfare state as different from the absence of a welfare state because it redistributes and pools risks, then tax preferences that do not do these things do not make the welfare state larger. The American state is indeed stronger than traditionally believed,Footnote ⁴⁰ but not because of tax preferences. This is not an argument from the left or from the right, but simply a matter of analytical precision.

In moving forward with research on tax preferences, we should limit the equation of tax preferences and the welfare state to the following situations: if the tax revenue was paying for some purpose that continues even after the tax preference, or we can demonstrate that state capacity was not affected by the decline in tax revenue; if the tax preference is refundable if the amount of the preference exceeds the amount the taxpayer has paid; and if the tax preference comes with a requirement to spend the foregone revenue in a way that approximates the society-wide risk-pooling functions of the welfare state. In other situations, tax preferences should not be equated with welfare state spending. They are its exact opposite, the reprivatization of risks that had previously been met collectively, and the waning of state capacity. Tax preferences are extremely important, and we should continue to pay attention to them. But aside from the circumstances above, we should examine them as a form of tax cut, not as a form of welfare state spending.