This is a massive book of great technical and philosophical sophistication. Its central goal is to develop and defend rigorously the view that a continuous prioritarian social welfare function provides the best framework for evaluating social policies. What Adler means by a social welfare function is a ranking of policies that is person-centred and assesses policies in terms of their interpersonally comparable welfare consequences. Reliance on a social welfare function is thus a generalization of utilitarianism. What distinguishes Adler's view from utilitarianism is that it places an increasing weight on the well-being of those who are increasingly worse off.

Adler requires that this prioritarian social welfare function be capable of evaluating partially specified alternatives rather than cognitively intractable complete states of the world, and it must be able to cope with incompleteness and imprecision in individual appraisals. But that's only a portion of what he attempts. In addition, Adler defends his own account of well-being and interpersonal comparisons and of the meaning and epistemology of moral claims. He argues for a specific continuous form of prioritarianism that draws on Atkinson's inequality metric, for methods for estimating welfare, and for much more. He intends his account to be modular so that those who, for example, reject his account of well-being (as I do) can plug their own account into his framework. Even though its framework is consequentialist and welfarist, Well-Being and Fair Distribution defends a vision of ethics in which fairness is fundamental.

Well-Being and Fair Distribution is a remarkable achievement. It is several books in one, drawing on many different literatures in philosophy, economics, decision theory and law. The vast knowledge that it demonstrates and the care, persistence and precision of its argumentation are hard to convey. I know of no work in its field that compares with it. Yet, I shall argue that Well-Being and Fair Distribution fails to provide a plausible framework for policy evaluation and, in that failure, demonstrates the inadequacy of its starting point. By dint of Adler's exceptional thoroughness and rigour in developing an account of policy evaluation in terms of a social welfare function that is person-centred, consequentialist, welfarist and prioritarian, Well-Being and Fair Distribution shows that no such account can be defended. A less brilliant and comprehensive treatment would leave one wondering if the project might be made to work by a smarter and more diligent author. Adler's remarkable efforts leave those who wish to rely on social welfare functions with no such hopes. Despite the genius with which the approach is developed, the approach is, I believe, hopeless; and the very genius with which it is developed establishes this conclusion.

Before laying out the considerations that support this conclusion, here is a sketch of the book's contents: Chapter One presents the general view of morality as person-centred and offers some considerations in favour of consequentialism and welfarism, but for the most part the book is an attempt to determine the most defensible version of a welfarist consequentialism rather than an argument for the general approach. Chapter 2 defines a social welfare function and makes clear that, unlike a good deal of the literature in economics, Adler will take social welfare functions to make use of information concerning interpersonal comparisons of well-being. Chapter 2 also criticizes other methods of policy evaluation, both comprehensive methods such as cost-benefit analysis, and more narrowly focused methods such as poverty metrics, tax-incidence analysis, and the use of cost-effectiveness analysis in the evaluation of health policies. At the same time that he criticizes these other approaches, Adler draws from them subsidiary principles that he believes that any satisfactory method of policy analysis should satisfy. For example, he maintains that every admissible method of policy appraisal must satisfy the Pareto principles (generalized so as to avoid relying specifically on preferences), and he believes that all fair methods of policy appraisal must conform to the Pigou-Dalton principle, which endorses transfers from those who are well-off to those who are badly off.

Chapter 3 presents Adler's view of well-being, which maintains roughly that a ‘life-history’ (x; i) – that is, being individual i in outcome x – has greater well-being than a life-history (y; j) (whether or not j = i) if and only if all members of the society would prefer (x; i) to (y; j) if their extended preferences were fully informed, fully rational and governed by a concern for the well-being of i and j. Put this way, Adler's account sounds circular, and I believe that ultimately it is. But Adler devotes a good deal of fancy footwork that I cannot discuss here into adapting Harsanyi's notion of extended sympathy (Reference Harsanyi1977, pp. 53–4) to his purposes. Notice that on Adler's view I am better off in outcome x than in y only if there is consensus among extended idealized preferences. Given the depth of evaluative disagreements in a pluralistic society such as the United States, I am inclined to think that well-being rankings will consequently be massively incomplete. It also seems odd to maintain that whether I am better off in outcome x than in y depends on who my compatriots are. Adler's social welfare function requires a ratio, rather than an interval scale to measure welfare, and so he needs to defend a non-arbitrary zero-utility state (which, he maintains, is non-existence) (219f).

Chapter 4 is concerned with estimating idealized extended preferences and argues that they can be inferred from ordinary (idealized) preferences and other data under certain conditions. Chapter 5 argues that if one has decided to appraise policies by means of a social welfare function, then it should be a continuous prioritarian one. Adler presents a number of conditions that a morally plausible social welfare function ought to satisfy and argues that a social welfare function of the form u(x) = (1 − γ)⁻¹ ∑_{i = 1}^Nu_i (x)^{(1 − γ)} best satisfies them. Chapter 6 argues that evaluation should depend on lifetime well-being (as opposed, for example, to well-being during some time interval or life-stage), and Chapter 7 tackles some of the problems due to uncertainty, incompleteness, and imprecision. Chapter 7 also responds to an important objection that I will discuss below. Chapter 8 concludes with brief discussions of application issues and of remaining problems particularly concerning variable populations and responsibility.

The rest of this review essay will focus on Adler's account of fairness, which he takes to be central to his project. Indeed he suggests that concerns about fairness completely exhaust morality! Although one classic criticism of utilitarianism is that it fails to take fairness into account, Adler argues that there is no incompatibility between fairness and his person-centred consequentialist and welfarist approach. Consequentialism, in his view, follows from basic premises concerning rationality and so must be compatible with a concern for fairness. Welfare, or, more precisely, opportunity for welfare is, he maintains, the appropriate ‘currency’ in which considerations about fairness can be cashed out (318–319). Opportunity for welfare does not coincide with welfare when there are differences in individual responsibility, and so taking fairness as a matter of the distribution of welfare is only an approximation (319–320). In Adler's view, our concerns about fairness and the separateness of persons can be captured fully by a continuous prioritarian social welfare function – at least when it is supplemented by some treatment of the role of responsibility, which he leaves as an issue for detailed treatment elsewhere. The claim that concerns about fairness exhaust morality is startling, and it leads Adler to defend counterintuitive implications, such as the claim that the duty to help those who are in urgent temporary need, perhaps because of some accident, is not a part of morality (472–475).

The claim that a social welfare function can capture the requirements of fairness and properly acknowledge the separateness of persons is vulnerable to many of the same objections that have been made against utilitarianism. Consider two societies or two proposals for organizing the basic structure of a single society in both of which there is the same amount and distribution of well-being. Every social welfare function, whether prioritarian or not, would have to rank these two proposals or these two societies as equally good. But one might have equal opportunities, while the other does not. Fundamental rights and liberties might differ. One might recognize traditional castes or even slavery, when the other does not. There is thus, I submit, good reason to doubt that one can formulate an acceptable account of fairness in terms of a social welfare function. The only response open to Adler, like the response of utilitarians to similar criticisms, is to deny that people in societies with highly unequal opportunities, rights, status, and so forth could be as well-off in terms of well-being as people in societies with less unequal rights, status and opportunities. It would be convenient for a consequentialist if this claim were true, but it is implausible that what makes inequalities in rights and liberties and large inequalities in opportunities and status unfair is that they lower the weighted sum of well-being. Examples such as this one suggest that fairness does not depend only on well-being and its distribution. It also depends on liberty, rights, respect and opportunity – where the measure of these is not the expected or realized value of the welfare that results.

Nearly a half-century ago, Peter Diamond published a simple critique of purely consequential evaluation of social states (Diamond Reference Diamond1967). Diamond asks us to compare two policies governing a society of two individuals. One assigns greater benefits to the first individual. The other holds a lottery in which both individuals have an equal chance of receiving these greater benefits. If the appraisal of the policy depends exclusively on the distribution of welfare within the outcome, then regardless of the character of the social welfare function – whether it is prioritarian, utilitarianism, sufficientarian, or looney – the policies must be equally good or bad, because the quantity and distribution of welfare are the same. But the policy that gives both individuals an equal chance of winding up with the greater benefit seems to Diamond to be the fairer policy. Diamond writes, ‘It seems reasonable for the individual to be concerned solely with final states, while society is also interested in the process of choice’ (1967, p. 766).

Adler addresses Diamond's example and ultimately denies that the policy that gives both individuals an equal chance at the benefits is fairer. He recognizes the intuitive pull of the case, but he insists that our approach to policy evaluation should be informed by a systematic appraisal of well-developed alternatives. His defence against Diamond's critique accordingly consists in two parts. First, he demonstrates with great lucidity and precision the considerable virtues of his approach and its incompatibility with the assessment of policies in terms of expectations rather than realizations of profiles of individual well-being. Second, he shows that ranking policies by expectations of benefit violates virtually every plausible moral constraint one would want to place upon the assessment of outcomes. For example, suppose that policy I for a two-person society has two equally likely outcomes: 50 units of welfare to both individuals A and B or a welfare distribution of 50 to A and 55 to B. Policy II also has two equally probable results: 80 for A and 20 for B or 20 for A and 80 B. Given a sufficiently prioritarian social welfare function, each of the outcomes of Policy I is superior to each of the outcomes of Policy II. But if one instead relies on a prioritarian assessment of the expectations of the two outcomes, then Policy II may be favoured because its expectations are equal. Adler argues persuasively that no attempt to replace a consequentialist social welfare function with one that appraises policies by the distribution of the expectations they offer will be satisfactory.

If one insists that policies must be evaluated by some sort of welfarist and consequentialist approach, then I think that Adler wins the argument. But, I do not think that one can deny in Diamond's case that equal chances are fairer than unequal chances. The right conclusion to draw is that one should abandon the hope of appraising social outcomes, policies, and institutions in terms of a consequentialist and welfarist social welfare function. Such a function may be a useful input into policy evaluation, but it does not capture concerns about fairness.

Before drawing this conclusion, let us examine in more detail Adler's account of fairness and of what it means for a theory of justice to acknowledge and respect the separateness of persons. In A Theory of Justice Rawls famously criticized utilitarianism for failing to respect the separateness of persons. Just as individuals should be willing to sacrifice benefits at some times in order to enjoy greater rewards at other times, so utilitarianism allegedly takes the greater welfare enjoyed by some people to compensate for the lesser welfare achieved by others. But, Rawls objects, benefits and costs in different lives are not fungible. Individuals have their own separate lives to live, and no account of justice is adequate that regards the suffering of some as redeemed by the happiness of others (Reference Rawls1971, pp. 29–30).

How exactly does utilitarianism fail to respect the separateness of persons? After all, as Adler notes (322), it apparently shows equal respect for their interests (at least insofar as their interests are represented by their well-being). Indeed, in Chapter 5 of Utilitarianism, Mill makes an argument that utilitarianism encompasses fairness that is much like Adler's claim on behalf of his prioritarian social welfare function.

What could be more respectful than adhering to ‘the highest abstract standard of social and distributive justice” that counts “everybody . . . for one, nobody for more than one’? But weighing each person's well-being equally permits the utilitarian to take one person's pluses as compensating for another person's minuses, and fairness collapses into welfare maximization.

In Rawls’ view, taking the separateness of persons seriously means instead that a theory of justice refuses to sacrifice the interests of some individuals in order to benefit others. Justice requires that the basic institutions of society promote the interests of all. Hence inequalities are permissible only insofar as they promote the interests of the least well-off representative individuals. Rawls’ veil of ignorance may be a good test of fairness and of respect for the separateness of people, but what defines social arrangements that respect the separateness of individuals are that they promote everyone's interests and do not sacrifice anyone's interests in order to benefit others or to achieve some aggregate result.

Adler agrees with Rawls that fairness requires respecting the separateness of persons, but he does not address the above interpretation of what Rawls means, and he denies that the way to respect the separateness of persons is to defend principles that everyone would agree to from the impartial perspective defined by the veil of ignorance. Choosing behind a veil of ignorance leads, Adler maintains, to utilitarianism, rather than to a fair outcome (322–326). According to Adler, the right way to understand fairness and the separateness of persons – to take ‘to heart the value of every person's life and welfare’ (330 quoted from Nagel Reference Nagel1995, p. 116) – is to recognize that each individual has a separate claim across different outcomes on behalf of those outcomes in which he or she is better off. ‘[A]n individual has a claim in favor of the outcome where she is better off. Outcome x is fairer than outcome y if, on balance the claims in favor of x are stronger than the claims in favor of y’ (313; see also 331).

Understanding fairness and the separateness of persons this way could lead one back to utilitarianism. As Adler notes (340–341), if the strength of the claim that an individual has in favour of outcomes in which he or she is better off is proportional to how much better off he or she would be, then the balance of claims in favour of x will be at least as strong as the balance of claims in favour of any other outcome if and only if the total welfare in x is at least as large as the total welfare in any other outcome.

Adler rejects utilitarianism. He maintains instead that fairness supports a continuous prioritarian social welfare function in which the strength of an individual's claims in favour of outcome x over y depends on how much better off he or she would be in outcome x weighted by a factor that is larger the worse off the individual is. The goal is still to maximize some function of the profile of individual well-being, but rather than merely adding up individual welfare, this function weights the welfare of those who are worse off more heavily that the welfare of those who are better off.

I do not understand why Adler believes that assigning a greater weight to the interests of those who are worse off respects the separateness of persons. It is still the case that the suffering of some can be justified by benefits to others. Because of the prioritarian weighting, it will take more benefits to the better off to justify imposing suffering on the worse off, but it is hard to see how that changes anything fundamentally – especially since Adler does not specify how strong the prioritarian weights should be. Indeed Adler explicitly defends what he calls, ‘numbers win’ (358) which says, roughly, that small benefits to enough well-off individuals can compensate for harms to very badly off individuals.

Rawls can say to those who are worst off, ‘There are no feasible social relations in which the worst off representative individual is better off than you are. Your interests have not been sacrificed in order to benefit others’. Adler cannot say this. All he can say to the worst off is, ‘Because you are badly off, we placed a greater weight on your interests, so that it took more than an equivalent amount of benefit to the better off to justify imposing losses on you. But what justifies your being badly off are, ultimately, the benefits to others.’

Prioritarianism is currently a popular position in social and political philosophy – undeservedly so, in my view. The argument on its behalf is that it explains intuitions, such as the view (which Adler and I and many others share) that it is better to provide benefits to homeless children than to the pampered children of millionaires. Prioritarianism explains this intuition. But so does egalitarianism. So does utilitarianism. So does a simple concern to alleviate suffering. Since there are competing explanations of such intuitions, there is not much of an argument here in defence of prioritarianism. Consider a case where there is no suffering and where it is questionable whether one increases the sum of individual welfare by favouring those who are less well off. For example, if one assumes that Donald Trump is worse off than Mitt Romney (which seems plausible to me), then the prioritarian should place more weight on improving Trump's well-being than Romney's. How plausible is this implication? Why should an impartial third party care more about improving the welfare of one rather than the other? Furthermore, there is absolutely nothing within prioritarianism to guide one in deciding how heavily to weight the interests of those who are not well off.

I do not mean to single Adler out for criticism in this regard. My complaint is a general one, and, though ultimately unsuccessful, Adler's attempt to link prioritarianism to an interpretation of what it is to respect the separateness of persons and his critique of egalitarianism (334–336) show an awareness on his part of the need to defend prioritarianism. Though Adler does a brilliant job of arguing that his continuous prioritarian social welfare function is the best formulation of a consequentialist and welfarist approach to the evaluation of social policies, in so doing he reveals that the approach cannot encompass concerns about fairness.

Even with all the bells and whistles, ribbons and bows, that Adler attaches to the utilitarian aspiration of assessing policies by adding up their welfare consequences, such an approach to policy evaluation is largely orthogonal to an assessment of policies in terms of fairness. Only by chance will the two approaches coincide. Neither fairness nor maximization of prioritarian weighted welfare by itself is, in my view, an adequate basis for policy assessment. We should care about well-being and its distribution, but policy appraisal must now stop there.