II. The Benthamite Case for Wealth Redistribution

Jeremy Bentham famously proposed “the principle of utility,” which holds that “the greatest happiness of the greatest number is the foundation of morals and legislation.” That is, both private and government policies should be judged solely or primarily by their effects on aggregate net happiness.^{Footnote 5} Many prominent subsequent economists including John Stuart Mill, Henry Sedgwick, Alfred Marshall, A. C. Pigou, and Abba Lerner have used this foundation to make a case for wealth redistribution by the state. As Pigou colorfully put the case, “the ninth course of the plutocrat’s dinner . . . yields much less satisfaction on the whole than the milk which the cost of it might have secured for a poor man’s child.” Taking this proposition to its logical conclusion, Lerner famously provided a mathematical derivation of the proposition that, if the size of the income to be divided does not depend on how it is divided, then “the maximization of probable total satisfaction is attained by an equal distribution of income.” Lerner concluded that “if it is desired to maximize the total satisfaction in a society, the rational procedure is to divide income on an equalitarian basis.”^{Footnote 6} Lerner and others treat the Benthamite principle (when combined with what they consider reasonable auxiliary assumptions) as providing an economic efficiency rationale for government policies to redistribute wealth from rich to poor. Such treatments are in error, because the economic logic of choice is not about maximizing happiness or anything else that is interpersonally addable, and so a fortiori is not an exercise in applying Benthamite utilitarianism.

Lacking an actual method for measuring happiness,^{Footnote 7} Bentham could only assess the aggregate net happiness impact of any policy measure (including government policies to redistribute wealth from the rich to the poor) from his armchair, that is, by making assumptions and deriving their implications. Bentham made the (introspectively plausible) assumption of “diminishing marginal utility of wealth,” meaning in this context that successive equal additions to an individual’s wealth yield successively smaller additions to his or her happiness. An additional dollar of wealth adds less to an individual’s happiness if he is rich than if the same individual is poor. In the absence of measurement it seems natural also to assume, as Bentham did, that any two individuals typically experience the same happiness from any given level of wealth (they have the same “utility of wealth functions”). The combination of these two assumptions implies that a dollar given to any rich man will add less to his happiness, and thus to aggregate happiness, than the same dollar would add to the happiness of any poor man if given to him instead. It follows that to promote the greatest happiness, at least when the question is how a benevolent allocator should distribute manna from heaven, the allocator should always give manna to the poorest.

If we add as a third assumption that taking a dollar from a man with $X in wealth has the same quantitative effect on his happiness (with the opposite sign) as giving him a dollar when he has $X-1 to begin with, then it further follows that aggregate happiness would be increased by a policy of taking wealth from the rich and transferring it to the poor provided (a fourth assumption) that the transfer can be made with sufficiently small administrative costs, and provided (a fifth assumption) that such a policy does not have indirect effects (such as disincentives to wealth production) that reduce aggregate happiness more than the direct effect increases it.

Bentham himself rejected the third assumption (although many of his followers tacitly seem to accept it). He thought that everyday experience showed that a man loses much more happiness at having his money taken than he gains in happiness by receiving the gift of an equivalent sum.

Bentham also rejected the fifth assumption at some point before the last dollar of transfers sufficient to equalize wealth, on the grounds that complete leveling would severely discourage the production of wealth. If income or wealth taxes were kept to the minimum rate necessary to bring the extremely poor up to subsistence, he supposed, the direct happiness gain would exceed the direct and indirect losses. But he argued that complete leveling would so severely reduce total output as to reduce aggregate happiness. In small experimental societies where output is equally shared, he observed, too many become “idle cheats” who “cause themselves to be supported in their idleness by the dupes by whom they surround themselves; whilst the cry for equality is only a pretext to cover the robbery which idleness perpetrates upon industry,” to the detriment of harmony, productivity, and general happiness.^{Footnote 8}

Bentham’s arguments about negative incentive effects of income or wealth taxation use the economic way of thinking (see below). But his arguments about happiness do not. The modern economic theory of allocative choice is non-hedonic. To make this point is not to disparage Benthamite political philosophy any more than it is to disparage any other political philosophy. It is only to insist that economics does not derive from Benthamism, nor does Benthamism derive from economics.

III. Economics, Happiness, and “Utility”

Many social scientists trained as economists have in recent years turned their attention to happiness research.^{Footnote 9} Some have unfortunately claimed more economic content for such research than is warranted. In a nutshell, the core of neoclassical microeconomics — as understood since Lionel Robbins (1932), who built on the work of Vilfredo Pareto — is a pure logic of allocative choice.^{Footnote 10} It has no more to say about happiness in the psychological or hedonic sense (the subject matter of the recent literature) than about love or nutrition, important as those things are. Robbins correctly noted that deriving the individual’s demand curve or the market’s demand curve, and thus price theory, does not require any reference to an individual’s happiness, much less a measure of her happiness. The identification of choice-theoretic utility with happiness is therefore excluded by Occam’s Razor. (Benthamism, by contrast, needs a measure of happiness that can be summed across individuals in order to quantitatively judge a law’s impact on society’s happiness in the aggregate.)

The consumer choice theory of microeconomics is about preference, not about experienced happiness. “I prefer x to y” is not synonymous with “x makes me happier than y.” Unlike happiness, preference is not about how events make you feel ex post. Preference is instead about how you rank your perceived options ex ante. Thus, microeconomic theory does not speak in any direct way to the psychological or hedonic results of choices (although of course preferences may in many cases be based on anticipated hedonic results). For example, a person who voluntarily sacrifices leisure to earn more income reveals only that she prefers more income to that leisure, not that more income makes her feel happier.

To insist on the non-hedonic nature of microeconomics is not to disparage happiness research any more than it is to disparage any other research in psychology or sociology. It rather means being clear about the (limited and merely formal) relevance of economic theory to such research. It does mean, in the spirit of Robbins, denying the status of “based on economic reasoning” to policy prescriptions based on extra-economic postulates – in particular, the postulates of Benthamite utilitarianism. When Paul Samuelson writes about adding utilities across persons to arrive at total social utility,^{Footnote 11} he has ipso facto removed his economic-theorist hat and donned his utilitarian hat.

Robbins’s point has been widely acknowledged. David Schmidtz is correct when he observes that “the concept of aggregated interpersonal utility used by Hare and Nagel” to argue on utilitarian grounds for egalitarian redistribution “has largely disappeared from economic discourse.”^{Footnote 12}

However, although many economists nod to Robbins, some walk with Bentham and do not keep the distinction between the two approaches clear. Much confusion arises from ambiguous use of the term “utility.” Bentham introduced his term as a hedonic measure, the balance of experienced pleasure over pain sensations, chosen or unchosen. Choice-theoretic utility, by contrast, is not a measure of experienced sensations, but rather an analytical construct for characterizing an individual’s preferences in a choice situation. The earliest economists who developed the neoclassical (also known as marginalist or subjectivist) theory of consumer demand (Carl Menger, William Stanley Jevons, Leon Walras, and their early followers) unfortunately did not consistently recognize that such a personal preference-ranking indicator is distinct from a hedonic measure, and that the latter can be discarded.^{Footnote 13} Translators furthered the confusion between the two concepts in the early Anglophone literature when they identified the German-language economists’ term grenzwert (the ordinal subjective value or preference-ranking a chooser assigns to an object’s marginal use) with Bentham’s hedonic term “marginal utility”.^{Footnote 14}

There have been attempts over the years to change the economists’ language. Vilfredo Pareto, one of the first to see that consumer demand theory requires only a preference ranking and not any hedonic measure, proposed unsuccessfully that economists should speak of “ophelimity” rather than utility. John Hicks and R. G. D. Allen successfully persuaded the profession to speak of “the marginal rate of substitution” between two goods (purely a ranking indicator) rather than of a good’s “marginal utility,” a term that lends itself to interpretation as the first derivative of hedonic “total utility.” But economists continue today to call the algebraic representation of a preference ranking “the utility function.” The contrast between the two types of “utility” was decisively clarified by Lionel Robbins in 1932. In an earlier article I suggested that we distinguish between “Benthamite utility” and “choice-theoretic utility.” More simply, we can call the first concept “happiness” and the second “preference.”^{Footnote 15}

When “utility” is strictly interpreted as a preference indicator, the “utility function” of an individual is completely distinct from the happiness magnitudes required by Benthamite utilitarianism. Neither derives from the other. Robbins made the distinction in the following terms:

I elaborated on Robbins’s point as follows:

A recent example serves to illustrate the confusion possible when the term “utility” is used without a qualifier. The economist Tyler Cowen blogged about “the idea of redistributing wealth to take advantage of differences in the marginal utility of money across varying wealth classes.” Another economist, David R. Henderson, quickly objected that “you can’t measure differences in the marginal utility of money across people.” Cowen was presumably using “utility” in the sense of happiness, but he did not give an explicit indication of that, leaving Henderson to think that Cowen was making erroneous statements about “utility” in the microeconomic theory sense of preference.^{Footnote 17}

Economist Richard Layard of the LSE conflates the two concepts all by himself. He asserts that the empirical finding from happiness surveys that “among industrialized countries with incomes over $20,000 per head there is no relation between average income and average happiness” is “contrary to standard economic theory.” He represents the “standard theory” as though the preference indicator of economic theory were a happiness measure. In his (ironically titled) Lionel Robbins lectures, Layard avers that “In the standard economic model, private actions and exchanges get us to a Pareto optimum where no one could be happier without someone else being less happy” and adds that “broadly speaking, the economic model says that the higher the real wage the happier the population.” But these are misrepresentations of standard microeconomic theory.^{Footnote 18} The standard economic model does not define a Pareto Optimum by reference to happiness but rather to preferred positions — not the same thing. Standard economic theory assumes that more income is preferred to less, when it can be had without cost, but it does not say that the recipients of higher average incomes or real wages experience greater happiness.

IV. Willingness to Pay and the Kaldor-Hicks Criterion

As noted, the Pareto-Robbins insight that the logic of consumer choice theory is founded on ex ante preference rankings, and not on happiness or any interpersonally comparable thing, has been widely acknowledged. From the foundational role of preference-ranking emerges the standard concept of economic efficiency: an allocation of goods is efficient (or Pareto-efficient, or Pareto-optimal) if there is no reallocation still available (such as a mutually beneficial trade of goods between J and K) that would move at least one party to a position he prefers (and move no party to a position she prefers less) to the starting point. In other words, no potential gains from trade remain uncaptured.

Correspondingly, the Pareto criterion for rating a trade an unambiguous improvement (a gain in efficiency) is that someone prefers to make the trade (there are winners) while nobody is moved to a dis-preferred position (there are no losers by the lights of their own preferences, including third parties). The Pareto criterion does not label a move an improvement if any person for any reason (it does not inquire whether envy, spite, or malice is at work) would prefer that the move not be made (meaning, would be willing to pay even a penny to stop it). Side payments from the winners out of their gains are a hypothetical way to buy off would-be objectors so that everybody prefers the package deal, but the offer of side-payments to objectors (absent a way to identify and rule out merely opportunistic objectors) invites ever more and higher-priced objections.

To limit the number of people with “veto power” over change, a modification of the Pareto criterion is sometimes suggested: a reallocation is said to be Pareto-improving if it benefits one or more parties J and K, and no party L is harmed, where what counts as harm is more than L’s distaste. The 2015 Economics Nobel laureate Angus Deaton, for example, has suggested ruling out envy as a relevant harm:

The principle that Deaton offers here, that envy should not be counted as a harm, may be reasonable as a normative principle. But it is not the Pareto criterion, which refers only to preferences in judging whether a change is an unambiguous improvement.

Likewise, the Pareto criterion as such does not say that mutually agreeable trades are unambiguous improvements so long as they do not interfere with any third party’s property rights. Thus, Robert Nozick’s scenario in which Wilt Chamberlain grows rich from mutually beneficial payments to watch him play basketball, where the payments lead from an egalitarian starting point to inequality of income and wealth, does not illustrate a Pareto improvement if some are moved to a less preferred position (would pay to reduce the inequality).^{Footnote 20} Normative disputes about the proper set of harms or rights to recognize are extrinsic to the Pareto criterion and to economic theory.

Equally, the Pareto-efficiency criterion provides no basis for considering it an improvement to take stuff from Peter to benefit Paul if Paul prefers not to have his stuff taken.^{Footnote 21} Pareto-improvement is about capturing gains from trade that are available starting from a given set of who owns what. Tax-financed transfer programs that proceed without the consent of every taxpayer change the initial pre-trade set of who owns what.^{Footnote 22} The Pareto criterion does not enable a ranking of two initial ownership sets.

The search for something more decisive than the Pareto criterion gave rise to the Kaldor-Hicks (KH) criterion, which takes a step back toward utilitarianism by reintroducing interpersonal aggregation in terms of dollars-worth of willingness-to-pay. A reallocation of goods is warranted if the gainers would be willing to pay more dollars to bring it about than the losers would pay to stop it. Unlike the requirement for a Pareto improvement, the gainers need not actually pay the losers enough to gain their consent. Aggregate dollar benefit-cost judgments replace unanimous consent. A reallocation is a Kaldor-Hicks improvement if on net it moves resources to uses valued more highly by bidders, that is, increases aggregate wealth, even if there are losers in the process. If Uber riders and drivers together gain more than taxi monopolies lose (and they must to the extent that taxis cannot compete on price because their costs are higher), allowing Uber to erode the taxi monopoly is a gain in KH-efficiency or aggregate wealth.

When it comes to evaluating wealth redistribution, the Kaldor-Hicks criterion has nothing positive to say because redistribution does nothing to increase aggregate wealth. Re-slicing the pie does not make for a bigger pie. By contrast to the Pareto criterion, the problem is not simply that there are losers, but that the gains do not exceed the losses. Having aggregate wealth as the thing to be maximized, rather than the aggregate happiness of Benthamite utilitarianism, makes all the difference. A rich man is willing to pay a dollar to keep a dollar, exactly what a poor man is willing to pay to receive a dollar. No wealth is created by a transfer. Wealth maximization does, on the other hand, argue against a redistributive policy that dissipates any wealth (incurs administrative costs) in making transfers, or that has any disincentive effects on wealth creation. (Disincentive effects are discussed below.) If transfer costs are non-zero, or production-discouragement effects are not zero, redistributive transfers unambiguously reduce aggregate wealth and are inefficient by the Kaldor-Hicks criterion.

V. The Social Welfare Function

Another technique for achieving greater decisiveness in policy recommendations, pioneered by Abram Bergson, has been to simply ignore the Pareto-Robbins strictures and posit a “social welfare function” according to which a hypothetical social planner chooses to maximize some weighted aggregate of individual welfare scores.^{Footnote 23} Here “welfare” plays the role that happiness plays in Benthamite utilitarianism. Bergson himself characterized his approach as a formalization of what the utilitarians Marshall and Pigou had been arguing all along. The “utilitarian” aggregation scheme (all individual scores equally weighted) is only one of many possible social welfare functions.

Bergson-style analysis of social welfare uses part of the economic way of thinking, in that the social planner solves a constrained optimization problem. But it neglects the second defining characteristic of the microeconomic method, namely that economy-wide or social-level phenomena are built up from the interactions of decentralized actors. Microeconomics is characteristically about individual-agent-level optimization and market-level equilibration, not about social-level optimization. The social welfare function is supposed to represent “society’s” preference function, but where individuals have different preference rankings there is not generally a non-arbitrary way to combine them into one coherent ranking.^{Footnote 24} In practice “social welfare” is simply posited to be a particular function of assignable individual welfare scores.

The social welfare function approach has dominated the “optimal taxation” literature since a 1971 article by James Mirrlees. As Diamond and Saez summarize the current literature:

This is clearly not a Paretian but a social-planner or quasi-Benthamite use of the terms “utility,” “welfare,” and “optimal.” Like Benthamism, optimal tax analysis posits a social maximization problem in which individual persons are regarded not as agents whose agreement is needed but as vessels of assigned point-scores. Deriving policy recommendations from an arbitrary function of individual scores — whether the score is called happiness or welfare, and irrespective of how the analyst stipulates that the “social welfare function” is weighted — has nothing to do with efficiency or optimality or utility as standard price theory uses the terms. Analysts who derive specific tax or redistribution recommendations by beginning from the statement that “society values more equality of income” per se^{Footnote 26} are evading the challenge of providing a compelling basis for such a valuation, such as deriving it from individuals’ preferences as they empirically are in our society. Put another way, Diamond and Saez do not make “the case for a progressive tax” but simply take it for granted that policy-makers have accepted some such case and need advice on fine-tuning its application.

VI. What Economics Can Say: Production Disincentives from Redistribution

Jeremy Bentham, John Stuart Mill, and the later utilitarian economists Alfred Marshall and Arthur C. Pigou, all advocated redistribution, but only up to the hypothetical point where aggregate happiness maxes out. All recognized that taxing wealth to redistribute it would discourage wealth creation.^{Footnote 27} They considered the sensitivity of aggregate happiness to wealth transfers, and of production to tax rates, to be empirical questions. But all presumed that the aggregate happiness criterion calls for some redistribution. Starting from the status quo, they judged that the first dollar transferred would surely increase the happiness of the recipient poor more than it would reduce the happiness of the taxed wealthy plus any happiness losses due to reduced total wealth. With additional transfers the marginal gains would shrink (because the poor are growing wealthier) and the losses grow (because the wealthy are growing poorer, and because wealth-discouragement effects are rising with higher tax rates). Losses would overtake gains before the point of complete wealth leveling was reached.

Like Bentham, Mill noted that the complete equalization of wealth would severely reduce the incentive to produce. The results of one’s extra effort would go into the common pool and thereby (in any large society) go virtually entirely to strangers. The sharp disincentive to effort from such a policy would drastically reduce the aggregate amount of income available for distribution. Mill cautioned: “Society can subject the distribution of wealth to whatever rules it thinks best, but what practical results will flow from the operation of those rules, must be discovered, like any other physical or mental truths, by observation and reasoning.”^{Footnote 28} The key practical result to be discovered was the exact extent of redistribution at which aggregate happiness was maximized and beyond which it began to decline.

Pigou did not offer a practical estimate, but asserted that the modern consensus among professional economists was that happiness-maximizing extent of redistribution was greater than economists used to think, by which he presumably meant to suggest that the government of his day should redistribute more:

Supposing that the size of the social pie to be divided is not fixed, that wealth does not fall from the sky willy-nilly but must be produced by disagreeable human effort, neoclassical economics can and does, without requiring interpersonal comparisons or happiness measures, analyze the disincentive effects of wealth taxes or progressive income taxes on wealth creation. Mappings from wealth to happiness, as previously argued, are no part of economic reasoning.

There are two main locations of disincentive effects. The first is along the income-leisure tradeoff.^{Footnote 30} Where an individual faces a marginal choice between how much time and effort to spend generating income and how much to spend otherwise, the individual’s personal optimum occurs where the income from additional work stops being preferred to the sacrificed leisure. A tax that reduces the share of the income that an individual gets to keep from the last hour worked shifts the tradeoff by lowering the sacrificed-income cost of leisure. It creates a disincentive to productive work, an incentive to choose less work effort (assuming that for hours or effort one does not expend on earning income, one gets to enjoy 100 percent of the leisure). Individuals facing such taxes will work less and produce a smaller pie.^{Footnote 31} They would have preferred a bigger pie and less leisure, if not for the tax.

The work-disincentive effect of income taxation might be considered unimportant if in practice the effect is so weak that higher marginal income tax rates reduce national income only negligibly. How strong is the effect empirically? Different empirical studies have estimated different magnitudes by taking different sets of indirect effects into account, employing different data samples, and using different estimation techniques. In a cross-country regression study that usefully isolates the effect we seek, Reinhard B. Koester and Roger C. Kormendi find that “a 10 percentage point decrease in marginal tax rates, holding average tax rates constant [so that there is no change in disposable income and thus no income effect], would increase per capita income by . . . approximately 13 percent of the median per capita income.” When they divide the sample between low-income and high-income countries, the estimated effect is 15.2 percent of median per capita income in low-income countries and 7.4 percent in high-income countries.^{Footnote 32} Their results are not just statistically significant, but the size is economically important. The U.S. Census Bureau reports that 2015 median pretax household income in the United States was $ 56,500. A drop of 7.4 percent, due to the work-disincentive effects of higher marginal income tax rates, would be a loss of $4,181 per year per household.

The second margin where redistributive taxes reduce output is the consumption-investment tradeoff. In a multi-period world, where we consider an individual’s choice between consuming more today or instead saving and investing more for the sake of consuming (or allowing heirs to consume) at future dates, a tax on income from saving or investment creates a disincentive to invest. (Here saving means consuming less than one’s current income, which provides the resources for investment. Investment means the use of saved resources to build productive capacity.) Less saving and investment means less real capital formation, leaving workers in future periods with fewer tools or less training. Potential savers and investors who face higher such taxes will save and invest less, resulting in a smaller pie in every future period.^{Footnote 33} As David Schmidtz has noted, taxing and transferring wealth from rich to poor moves resources from those who tend to save a larger share of those resource to those who tend to save a smaller share.^{Footnote 34} It shrinks the supply of savings, reducing resources for capital formation, implying smaller pies in the future.

Eric Engen and Jonathan Skinner, surveying an array of empirical studies, provide an estimate of the effect of tax rates on the growth rate of the economy’s aggregate output, assuming a 5 percent marginal tax rate change with some effect on the average tax rate:

As they note, even such seemingly modest effects on the growth rate “can have an important long-term impact on living standards.”^{Footnote 35} A tax structure with rates 5 percentage points higher, if it slows annual growth by the midpoint estimate of 0.25 percentage points per annum, after thirty years implies a 7.78 percent lower level of annual national income (GDP). The International Monetary Fund projects that the United States’ GDP will be 19.377 trillion dollars in 2016, so that a 7.78 percent reduction would be a loss of 1.51 trillion dollars for the year (and more in subsequent years).

Bentham, Mill, Marshall, and Pigou all understood the production-discouragement effects. And accordingly they anticipated the point that Schmidtz makes against Thomas Nagel (1991), that even within the Benthamite framework, fully egalitarian redistribution does not follow from the goal of happiness maximization and the assumption of diminishing marginal happiness (DMH) from wealth once we consider a world where wealth must be produced (rather than received as manna from heaven), and where taxing the production of wealth reduces its volume.^{Footnote 36} To paraphrase Schmidtz (by speaking of “happiness” rather than “utility,” and by strengthening the claim as warranted by economic reasoning about production), in a world where resources and effort have to be expended to produce happiness-yielding goods it is not generally the case that multi-period (sustainable or lifetime) happiness is maximized by transferring command over goods to those for whom the goods yield the most happiness in the current period. Reaching the highest sustainable path of happiness over time will instead require taking into account the cumulative benefits of leaving resources in the hands of those who use them most productively.

Note that talking about happiness as a function of wealth or income seems less appropriate, in a world of saving and production, than talking about happiness as a function of consumption. Not all wealth or income is consumed. If saved income does not yield happiness until the period when it is dis-saved and consumed, differences in consumption are a better proxy than differences in incomes for the degree of inequality in happiness. In practice, differences in consumption spending are considerably smaller than the differences in disposable income across income quintiles.^{Footnote 37} This should not be surprising, because it follows immediately from higher-income individuals saving considerably more. If we consider differences in what Adam Smith called the “necessaries of life,” like nutritional intake, the differences are even smaller. Smith noted in The Theory of Moral Sentiments ^{Footnote 38} that, when it comes to foodstuffs, the rich “consume little more than the poor,” although “they select from the heap what is most precious and agreeable.”

VII. The Public-Goods Argument for Wealth Redistribution

Formalized by Paul Samuelson,^{Footnote 39} the theory of public goods offers a preference-based or Pareto-efficiency case for collective action. A “public good” is one whose potential users are in the aggregate willing to pay enough to cover its production costs, yet its production cannot be financed on the usual user-pays basis because (with current technology) non-payers cannot be excluded from enjoying it, and therefore not paying is a dominant strategy (the “free-rider problem” prevails). In such a case, so the argument goes, there are amounts we would each contractually agree to pay, rather than go without the good, such that the pooled payments would be great enough to cover production costs. No individual would pay more than he or she is willing to pay. The well-calibrated collection of funds to pay for a public good does not levy an involuntary tax on the payer, but merely enforces a (hypothetical) voluntary contract.^{Footnote 40} The provision of public goods in this way is therefore a Pareto improvement. The failure to provide genuine public goods is a failure to capture potential gains from trade.

The proposition that egalitarian redistribution is a public good uses public goods theory to make an efficiency case for egalitarian redistribution by the state.^{Footnote 41} Suppose that some individuals are willing to pay for a reduction in observed inequality, or alternatively a minimum living standard for the poor. They may value the transfer for its own sake, or for the sake of an expected consequence like social stability. Each such individual prefers contributing some amount to help fund a large collective transfer program, over buying other goods available at an equal price to him, provided that enough others also contribute to achieve the desired large impact. If we generally view public goods provision by the state as a collective means of satisfying individual preferences in cases where the free-rider problem blocks user-pays provision, and suppose that payments can be calibrated not to exceed anyone’s willingness to pay, then we arrive at a preference-based Pareto-efficiency case for an egalitarian redistribution policy. Unlike the Benthamite case, it does not require interpersonal aggregation of happiness. The assumptions establish that there is some redistributive policy that can satisfy the unanimity test for a Pareto improvement. Harold M. Hochman and James D. Rodgers sketched the case as follows:

Lester Thurow similarly argued:

Citing Samuelson’s definition of a public good, Thurow argued that individuals cannot effect a desired change in the distribution of income noncollectively because the benefit of observing a more equal distribution is non-excludable, and so the free-rider problem prevails:

Here an individual’s “optimal share” is the amount that captures all available net benefits from transfers subject to the constraint that he not pay more than he is willing.

The argument just sketched is purely an exercise in armchair speculation: if we imagine that assumptions h ₁ through h _m hold, then it follows that everybody would agree to redistribution program Y. To insure that any real-world tax-financed redistribution program actually brings about a Pareto-improvement, policymakers would have to go beyond supposing that taxpayers are each willing to pay enough for a specified transfer program such that aggregate payments would cover the program’s cost, to establishing that each taxpayer really does agree to pay his assigned share (conditional on the program failing to go forward without it). Establishing actual willingness to pay, Thurow acknowledged at the end of his article, is blocked by the free-rider or demand-revelation problem: “It is possible to imagine attempts to measure individual preferences concerning the distribution of income, but these would run into the familiar revealed preference problems common to all public goods.”^{Footnote 44}

The Hochman-Rodgers and Thurow discussions of redistribution illustrate the indefinite extendibility of the public goods argument. It can be used to recommend any change from the laissez-faire outcome that an analyst supposes will provide benefits for which individuals are willing to pay but are currently blocked by the free-rider problem. The benefits need not be tangible. The individual’s assumed benefit from a change in the income distribution comes just from knowing something, and need not be manifested in any cost-bearing behavior. The lack of any evidence that individuals actually are willing to pay for the change is attributed to the free-rider problem.

So how can we ever know that individuals actually are willing to pay enough to cover the cost of a proposed public good like income redistribution? In the nature of the case, we never can. As Samuelson immediately recognized, when the payment organizer asks how highly one values the good, “it is in the selfish interest of each person to give false signals, to pretend to have less interest in a given collective consumption activity than he really has.”^{Footnote 45} When an individual claims low willingness to pay, however, he may well be telling the truth. Public-provision advocates may claim greater aggregate private willingness to pay than actually exists. When we see a good not being produced in the absence of collective action (and we always see more of a good not being produced), it needn’t be that free-rider problems are blocking what all would agree is worth doing. It could be that the good in question isn’t really worth producing in the eyes of all those who would bear the cost. We don’t know actual magnitude of willingness to pay, and by the inbuilt logic of the assumed free-rider problem, exclusion of non-payers being infeasible, we have no way of finding out.

The claim that “redistribution of income is a public good” is thus non-falsifiable. It cannot be tested without a demand revelation mechanism.^{Footnote 46} Without some method for testing suppositions about other people’s willingness to pay for income redistribution, claims to the effect that “redistribution is an underprovided public good” are without any evidentiary foundation. Equally unfounded are claims that a transfer program achieves a Pareto improvement, because the Pareto criterion requires that the payments taken from any individual are no greater than he would have agreed to make.

VIII. Buchanan on Distributive Justice

James M. Buchanan has suggested that a representative person behind the veil of ignorance would rationally choose constitutional rules empowering government to provide certain redistributive measures, namely inheritance taxation and public funding of basic education. These measures do not aim to reduce inequality in the wealth or income that members of a market society end up with at retirement, but rather to reduce inequality in the wealth and human capital with which (through no merit or fault of their own) they enter the market society at school-leaving age. Great inequality in starting points, he proposes, violates “our ordinary sense of ‘fairness’.” Here Buchanan clearly argues as a contractarian political theorist, not as a neoclassical economist. He bases the constitutional deliberator’s supposed preference on supposedly shared moral intuitions about fairness, rather than on benefits associated with capturing available gains from trade. The economic analysis of gains from trade takes an initial (pre-trade) assignment of property holdings as given. Buchanan is proposing to reallocate holdings prior to an individual’s starting to trade, which can only be for reasons other than capturing available gains from trade under the given assignment.^{Footnote 47}

There are other possible ways to motivate Buchanan’s constitutional deliberator. One is an appeal to strong risk aversion. Buchanan’s redistribution policies can then be viewed as insurance against an unfavorable draw in the starting-point lottery, the cost of which (lower mean wealth due to administrative costs and output-discouraging effects of redistribution) the deliberator would agree to bear ex ante for the sake of avoiding the risk of starting far below the mean.^{Footnote 48}

In either framing, Buchanan’s justification for redistribution has one feature in common with a public goods argument. Claims about what “we” would agree to in a hypothetical constitutional deliberation behind the veil, like claims about individuals’ willingness to pay, for which there is no demand-revelation mechanism, are unfalsifiable. I can introspect about my own actual preferences, but I cannot establish introspectively that my preferences are those of a representative individual behind the veil.

Other preferences that might with equal plausibility be attributed to the constitutional deliberator do not imply choice of inheritance taxes or public funding of basic education. For example, the representative deliberator might be concerned only with complete legal equality of opportunity and not with material starting points. She might regard even constitutionally authorized confiscation of inheritances above some size as a policy that violates the legal principle of nondiscrimination and gives government too much power.^{Footnote 49}

Or a reasonable deliberator might prefer government only to insure a minimum starting point to all citizens, and not to deny large inheritances to the offspring of the rich in pursuit of reduced inequality in starting points for its own sake. This is particularly likely if she recognizes that high rates of inheritance taxation capture little revenue to be redistributed to the disadvantaged, just like high rates of income taxation as discussed at the outset of this paper, but shrink the pie, because they divert potentially productive labor into the creation of legal vehicles for inheritance-tax avoidance. She might then regard the cost of pursuing equalization of starting points through inheritance taxes to exceed the benefit, and prefer to fund social insurance another way.

IX. Conclusion

To summarize: The economic way of thinking, contrary to what some economists have claimed or suggested, does not provide a compelling efficiency rationale for redistribution of income or wealth. The concept of diminishing marginal happiness from income or wealth, which underlies the utilitarian case for redistribution, plays no part in economic theory. The economic logic of choice rests instead on preferences, which are ex-ante and non-hedonic. The concept of a “social welfare function,” still popular in optimal tax theory, is likewise extrinsic to the logic of choice by individual agents.

Economic logic is used by those who propose that redistribution is a “public good” in the sense of economic theory, but their proposition that consumers have sufficient (although unrevealed) willingness to pay cannot, in the nature of the supposed case, be falsified or supported with evidence. James Buchanan offered a related case for taxing inheritances based on suppositions about our common preferences that are equally unfalsifiable, and no more plausible than alternative suppositions about our preferences.

The economic way of thinking does speak clearly, however, about the negative consequences of egalitarian income taxes on the generation of income, and the negative consequences of wealth taxes on the production of wealth through effort and saving. The size of income and wealth losses is an empirical question. The empirical evidence suggests large losses for large redistributive programs. That a price in reduced overall prosperity must be paid for egalitarian policies is not a decisive objection for those who place high value on egalitarian efforts, of course. But the price must be understood in order for citizens or policymakers to make an informed decision.