I.

“Justice,” wrote John Rawls, at the beginning of his magisterial work on the subject, “is the first virtue of social institutions, as truth is of systems of thought.”

The comparison is between scientific theory and social structure. Scientists often describe their theories as beautiful, exciting, or powerful, meaning (among other things) that their simple, elegant formulae generate startling and wide-ranging results. But whatever its beauty, power, and excitement, the scientific community must in the end reject a theory if it makes predictions that fail to materialize. We may resort to all sorts of maneuvers to avoid this result; we may question the reliability of our instruments or the results achieved by rival laboratories. Ultimately, however, truth is the tribunal to which scientific theory must answer. And analogously, Rawls is saying, justice is the tribunal at which laws and social institutions are evaluated. A society may be prosperous, stable, cultured, and powerful, but if it is unjust, these achievements are not enough to save it from a crushing moral indictment. We may want our just society to be powerful and prosperous too. But we should not want power and prosperity at the expense of justice.

In this paper I would like to consider this thesis of the primacy that justice is supposed to have over other bases on which a society might be evaluated. I take it as a thesis about the concept of justice, not just John Rawls's particular conception. In A Theory of Justice, Rawls defends the priority of his two principles over the principle of aggregate utility. He argues that parties in what he calls “the original position” would give the two principles this priority [TJ 130–160, 266] and he uses the same method to establish priorities among the two principles themselves (the priority of the principle of equal liberty over the principle regulating social and economic inequalities) and between various elements of each of the two principles (e.g., the priority of fair opportunity over the difference principle) [TJ 214–220, 266–277]. These are not topics I shall study in this paper. I am not interested in the priorities established within Rawls's conception of justice. I am interested in primacy as a thesis that is supposed to be true of any plausible conception; indeed, I want to use the primacy thesis to bring the concept of justice into clearer focus for us.

Since 1971, almost all the discussion of justice in political philosophy has been about the merits of various rival conceptions—Rawls's in particular, but also Nozick's, Walzer's, Ackerman's, Dworkin's, and so on.²

Very little has been said about the concept of justice itself,³ though Rawls was careful at the beginning of his book to establish a distinction between discussion at the level of concept of justice and discussion at the level of conceptions of that concept [TJ 5]. Of course the logic of the concept/conception distinction is such that a particular conception purports to tell us all we need to know about the concept; it purports to tell us the truth and the whole truth about justice; the relation between concept and conception is not like a whole/part relation, which would leave some important truths about justice unsaid once the favored conception had been specified.⁴ Even so, there ought to be a perspective from which we can say something about the agenda that the various rival conceptions share in common—the agenda that distinguishes the enterprise of producing a conception of justice from the enterprise of producing a conception of something else. Ronald Dworkin once observed that “it is difficult to find a statement of the concept [of justice] at once sufficiently abstract to be uncontroversial among us and sufficiently concrete to be useful.”⁵ Argumentation about justice seems to get going too quickly to leave the sort of breathing space that we need in order to say something about the concept of justice as opposed to rival conceptions of it. If we were to say, for example, that justice addresses issues such as moral desert, we already seem to be taking sides with a particular conception of justice or a particular cluster of conceptions; similarly, if we say justice is about needs or equality, we seem to be taking sides in another direction. I do not mean that these are implausible characterizations; but it seems that we have zipped over very quickly into the elaboration of a theory of justice rather than providing a more neutral statement of what it is that a theory of justice is supposed to be a theory of or about. Somehow we have to slow this process down.

H.L.A. Hart has suggested that maybe the concept of justice can be found in the empty formalism of the principle “Treat like cases alike,” and that it is the task of particular conceptions to fill out the content of relevant likeness and appropriate treatment.⁶

That is all right as far as it goes but it is certainly not sufficient, for there may be ways of filling out the principle “Treat like cases alike” that have little to do with justice.⁷ John Rawls says near the beginning of A Theory of Justice that he proposes to follow Hart's suggestion; but the more helpful part of Rawls's discussion in that passage is his suggestion that rival principles of justice address the assignment of basic rights and duties and “the proper distribution of the benefits and burdens of social cooperation” [TJ 5]. My aim in this paper is to elaborate that formulation. After some preliminaries, I shall argue in Sections IV through VI that principles of justice have a specific role to play in social theory, which has to do with the distribution of individualized benefits and burdens. The primacy thesis will be defended as a matter of the primacy of distributive over aggregative measures. I shall then use that account in Section VII to characterize the priority that justice has over prosperity, liberty, and Hobbesian values such as security and to contrast this with a different sort of priority that justice has over collective goods such as the preservation of culture. Finally, in Section VIII I shall relate what I have said about the importance of distributive assessments to some familiar critiques of the Law and Economics movement.

II.

In Sections II and III I consider two clusters of preliminary points that, I hope, will clarify what is being claimed and what the range of the primacy thesis is supposed to be.

The first cluster concerns institutions. Justice is being called the first virtue of social institutions, not necessarily the queen of the virtues as such. I know that Plato and Aristotle regarded justice as a virtue of individuals as well, and Plato in particular made much of the analogies between the two levels—social justice and individual justice.⁸

But I shall not make any argument at all at the individual level.

A second point is that Rawls's claim that justice is the first virtue of institutions evidently presupposes that there are institutions—that is, that they already have the perfection of existing. Later I shall say something about the Hobbesian claim that elementary peace and stability are prime virtues at the social level.⁹

There are several responses to this point, but one of them is that there may be no disagreement here because any claim about priorities among the virtues of institutions already presupposes that we have the modicum of peace and stability that allows social institutions to operate in any way at all.¹⁰

A third point has to do with the way that institutions form a social system. The claim that justice is the first virtue of social institutions might be read as a generalization that, for every single social institution, justice is the first virtue of that institution. This would be a very ambitious claim. More modestly, it might be read as the claim that justice is the first virtue of the array of institutions that together constitute the basic structure of society. Rawls's own holism in this regard—his emphasis throughout A Theory of Justice on basic structure as the subject of justice [TJ 6–10, 147–148]—suggests the second reading. I shall follow him in that; any claim about individual institutions will be treated as derivative from the claim about the basic structure and as most plausible in case of institutions (like the legal system) that are key to the basic structure.¹¹

This brings us to a final point about institutions. Rawls's suggestion is that justice is the first virtue of social institutions in exactly the way that truth is the first virtue of scientific theories. One difficulty with this analogy is that truth and justice stand in slightly different relations to theory and institutions, respectively. A scientist is aware from start to finish that the criterion of truth is the raison d'être of his enterprise; the whole point of producing a theory is to secure insight into the workings of the world. Truth stands in an internal relation to scientific theorizing; it is bound up with the logic of the practice.¹²

Not so with social institutions. There we often have to bring justice in as some sort of external standard. If we want to use justice as a criterion, we have to bring it in from the outside; often we have to drag the institution kicking and screaming to the tribunal of justice.

However, this disanalogy may not apply to all social institutions. Consider the institutions of the legal system. It is not implausible to say that the law has an internal relation to justice comparable to the relation between science and truth. Think how often the word “justice” occurs in the self-descriptions of the legal system. There is the Department of Justice, and many countries have a Minister of Justice; the men and women on the U.S. Supreme Court are called “Justices.” When we refer to the prosecution and punishment of crime, we talk about “the criminal justice system.” In London, the Royal Courts of Justice can be found in the Strand; in Washington, the building that houses the U.S. Supreme Court has the slogan “Equal Justice for All” inscribed on its pediment. Law does not present itself to us as an arbitrary array of decrees and regulations any more than science presents itself as a set of idle assertions. A legal system presents itself as having a commitment to justice from start to finish. I do not mean by this that every legal system is just, any more than Rawls thinks every scientific theory is true. What I mean is that you cannot begin to understand the legal system without recognizing that it presents itself as having this aspiration, any more than you can understand what scientists are doing apart from their professed aim of verisimilitude.¹³

I shall return to this in my remarks about the economic analysis of law at the end of the paper.¹⁴

III.

The second cluster of preliminary points concerns the nature of the primacy that is being claimed. What is it for something to be the first virtue of social institutions, to be at the top of this list?

A natural interpretation would read primacy in terms of the weight of reasons. We imagine reasons ranked in an order of importance, with the more important ones outweighing the less important ones (which means, roughly, that they matter more and that if we have to choose, we would choose to follow the one reason rather than the other). We use metaphors of “weight” to express this.¹⁵

We have reasons for action in respect of justice—reasons to make our own society more just, reasons to react in certain ways to the injustice of other societies—and reasons for action also in respect of other institutional virtues. From time to time these reasons may conflict—that is, they may argue for incompatible courses of action. In making these hard choices, we should give priority to the reasons of greater weight—though of course the metaphor of weight also suggests that a larger number of reasons relating to a “lighter” virtue may outweigh a smaller number of reasons relating to a “heavier” virtue. That justice-reasons are more weighty is quite compatible with large gains in security or prosperity outweighing smaller gains in respect of justice.

I think the primacy claim should be read as stronger than this. In a Rawlsian context, something like lexical priority suggests itself [TJ 37–38], but we must not forget how difficult it is to justify a lexical ordering. As Rawls says, the concept of a lexical ordering “seems to offend our sense of moderation and good judgment” particularly because it implies that when the reasons conflict or draw on the same resources, nothing should be done for the sake of the subordinated virtue as long as something remains to be done in respect of the virtue ranked above it [TJ 38]. One possibility is that addressing the higher-ranked virtue is morally inescapable once any attempt is made to deal with the virtue ranked lower; in any attempt to act on reasons relating to security or prosperity, for example, we are driven to consider reasons relating to justice. That is how I shall defend the primacy thesis for most cases. These are cases where the competing virtue refers to individualized values, and the role of reasons relating to justice is to rivet our attention on questions about the proper distribution of those values. But, as we shall see, there are some cases for which this interpretation does not work—namely, those where justice competes with genuinely collective goods. For those cases we will fall back on a more direct priority claim: Reasons relating to justice are simply and in all circumstances more important than reasons relating to collective goods such as the preservation of culture or a language. Certainly this claim is controversial, and I hope the value of my account will be to pinpoint where the controversy lies.¹⁶

I want to rule out the possibility that the primacy of justice is true by definition. Someone may say that to look for justice-reasons in regard to institutions is precisely to search for the most important reasons in regard to institutions (whatever they are). This is implausible. Suppose someone thinks that the most important dimension on which one could assess American institutions is the contribution they make to the worldwide projection of military might. Should we really infer from this that such a person must believe that might is the basis of justice? That justice is a serious matter, not a trivial one, is an important truth and it has an analytic aspect which I will explore later.¹⁷

Mainly, though, justice-reasons are important because of the sort of reasons they are, not because of a definitional connection between “justice” and “important.”

A slightly different version of the analytic suggestion is put forward by Tom Campbell, who says that “the idea that justice must be overriding attaches to the rather vague use of ‘justice,’ in which it is equivalent to rightness, rather than to its narrower and specific meaning.”¹⁸

Since Aristotle we have known that “justice” has a general sense in which it takes up almost all virtues that affect how one deals with other people as opposed to the narrower range of issues raised under the specific headings of “distributive justice,” “commutative justice,” and so on.¹⁹ If the primacy thesis is true only of justice in the broader sense, then it amounts to little more than normativity or at most to the priority of the right over the good. But I will try to show that the primacy thesis is true of the narrower sense of justice. That narrower sense includes retributive justice, distributive justice, commutative justice, and so on but it is not equivalent to the whole truth about right.

IV.

After these preliminaries, let me turn to the main argument. What sorts of reasons are justice-reasons? Let us call someone who is interested in justice “a justicier.” What sorts of questions does a justicier ask? What sort of information about social arrangements does a justicier want to obtain? Can we answer these questions at a general level without begging the question in favor of a particular conception?

Let me start with an analogy. Sometimes one hears stories like the following on the television news:

Television presenters love big numbers. But the story as it stands is uninformative, especially in the United States, where it is common for prison sentences to exceed the expected lifetime of the person sentenced. Five gangsters were sentenced to two hundred years. Does this mean each of them was sentenced to forty years? Or does it mean that four of them got five years each and the other was sentenced to 180 years? Does the story even imply that all five went to prison? It is a measure of its uninformativeness that we have to rely on the pragmatics rather than the strict meaning of the statement to infer that each of the gangsters actually got some jail time—that is, that it was not a case of one man getting two hundred years and the other four going free.

I find stories of this sort exasperating. One can imagine a person for whom the aggregate figure—the two hundred years, however distributed—is the important information: the bureaucrat who has to plan prison accommodation, for example, and therefore has to think in terms of aggregate man/cell hours. But for most of us—including, one hopes, the judge and certainly the defendants—the important question is: Who (in particular) got what sentence (in particular) and why? Prison sentences are served by individuals, and one wants to know what the individuals got. There is a big difference between a five-year sentence, a forty-year sentence, and a sentence measured in three figures. Did anyone get a sentence measured in three figures—that is, a sentence of a hundred years or more? The story does not say. A five-year sentence affords a convict some measure of hope that he can pick up his life again afterwards. Were any of the gangsters given that sort of hope? The story does not say. Suppose the gangsters killed someone, and we believe (on retributive grounds) that no murder can be expiated by anything less than a sentence of three score years and ten. Has this principle been satisfied in the outcome reported? The story does not say. Certainly, in some sense, more than seventy years of penal servitude have been handed out. But the retributive principle we are imagining is not concerned with what is clocked up on the sentencing aggregate. It would not be satisfied by sentencing 140 murderers to six months each.

A report that five men got a total of two hundred years rules some things out. No one was sent down for a thousand years. And it is not the case that they all got off on probation. If one of them got a light sentence, then someone must have got quite a heavy sentence. A logician could set out the possibilities in the form of a massive disjunction: Either one person got two hundred years, and the other four got nothing, or, one person got 199 years, and the other four got three months each, or…. This would be tedious beyond belief; but the tedium is a measure of how little determinate information the television story gives us.

Imagine a citizen concerned about a rising wave of crime who complains that the sentences that were handed down were not long enough. If a spokesman for the Justice Department responded by pointing proudly to the enormous total number of years in prison imposed in recent months, the citizen would be unimpressed. “That's not the point,” he would say, “It is not the magnitude of the aggregate total of sentences. The important thing is that individual criminals get long sentences (for serious crimes).” He would say this because he would know that a high aggregate total is perfectly consistent with low individual sentences (given a high arrest rate). He might express his position dramatically by talking about the primacy of the individual sentence, insisting that this dimension of assessment of the criminal justice system—Who in particular gets what in particular, and why?—has priority over any aggregative basis of assessment.

Well, it is in that spirit that I want to argue that justice has priority over some other social virtues such as prosperity, for example. Consider the way we present information about social prosperity. When we compare societies in regard to prosperity (or the same society over time), we consider things such as gross domestic product (GDP) or economic growth. When we say that the GDP of the United States was so many trillion dollars in the year 2000 or that it grew by so many dollars from 1999, we give information in a form remarkably similar to the form used by our television presenter telling us about prison sentences. It makes no difference when the information is expressed as an average in the sense of an arithmetical mean (per capita GDP). If our crime reporter had told us that five crooks were sentenced to an average of forty years each, his report would be no more informative. The average form is spuriously individualistic but it does not really convey individualized information. The same defect is true of both the crime report and GDP measures: Information that is primarily information about individuals is presented in a quite unilluminating form. This is a defect, in the one case because individuals serve prison sentences one by one, and in the other case because individuals or, at most, families enjoy prosperity, not whole populations considered as abstract entities. In the case of the gangsters, we want to know who got what as individuals; and that too, I believe, is what we should want to know in the case of national prosperity. With the five gangsters, we may even want to associate particular outcomes with proper names; we may want to know what sentence Mack the Knife got and how long Eddie Razors is going to spend in prison. This is not normally the form in which we call for individualized prosperity information in the case of the quarter-billion people living in the United States. We do not want to be deluged with a breakdown of social wealth that begins with Alexander Aab's prosperity and ends with Irving Zywotow's.²⁰

But ultimately that is what it comes down to: Social justice is about how Alexander Aab is doing and how Irving Zywotow is doing, and everyone in between. It is not about averages or totals.

My thesis, then, is that an interest in justice—defined generally—is an interest in distributive information across individuals. Justiciers look for information about distributions; these are the states of affairs they want to interrogate.²¹

And—for reasons I shall try to explain—the justicier thinks this interest has priority over any other form of curiosity one might have. No doubt there are many ways of compressing and summarizing the information that a justicier is interested in, short of writing out the phone book. We can give percentile analyses, saying how the top 1 percent are doing, and the bottom 10 percent, and so on. Or we may give Gini coefficients or measures of social mobility.²² How fine-grained we want this information to be is of course relative to the practicalities of our justice inquiry. If our practical efforts are confined to the broad shape of society's basic structure, then we are unlikely to want much more information than how representative members of various social groups are doing; but if it is our aim to administer a just welfare system, for example, we may want to ensure that the system is made responsive to microdifferences in individual circumstances, so that unusual individual cases will not fall between the cracks of a coarse-grained analysis. This helps explain the focus of Rawls's book. Though Rawls emphasized the importance of taking individual persons and the distinction between persons seriously [TJ 24], he oriented his principle of economic justice to “representative individuals” and in particular to the “worst-off group” [TJ 81ff., my emphasis]. This has been criticized in the literature as an inconsistency.²³ But I think it is an artifact of the practical implications Rawls envisages for his theory. “The primary subject of justice,” he emphasizes, “is the basic structure of society.”

Given this practical orientation, the Rawlsian theory does not need fine-grained information (though it would need such information if it could be shown empirically that life chances varied radically on the basis of a much greater diversity of individual starting points). Still, Rawls's theory does need distributive information; an inquiry that failed to differentiate between any of the places (or kinds of places) that individuals began from or between any of the places (or kinds of places) that they could reach from these different starting points would be failing in its task of giving scrutiny to the issues that matter—namely, how things go for individual men and women.

V.

So far I have characterized an interest in justice as an interest in distributive information: Who gets what.²⁴

In Section VI, I will explain why this interest has the importance accorded to it by the primacy thesis. But first I want to clear up some possible misunderstandings about distribution.

The term “distributive” can be misleading. In the criminal justice system, prison sentences are handed out to individuals by agents of the state. And it is tempting to assume that distributive justice works in something like the same way except that it is concerned with the handing out of benefits rather than burdens. Critics of talk of distributive justice have seized upon this “handout” element as a reason for avoiding such talk. They think that talk of distributive justice leads people to presuppose a distributor—such as the state—with all of society's largesse at its disposal and to presuppose that the question is who is to get what of the goods and resources in the state's gift.²⁵

They say that may be appropriate when Athens is handing out booty captured by its navy. But it is not an appropriate model for thinking about a market society, where goods are not in the hands of the state in the first instance and where the allocation of goods is a result of millions of individuals' decisions in the marketplace.

All this can be conceded (or not). These are not issues I mean to raise by insisting that the primacy of justice amounts in part to the primacy of distributive over aggregative bases of evaluation. By distributive information, I mean only to refer to information about the way outcomes are in fact distributed across individuals, whether any agency did the distributing deliberately or not.²⁶

My characterization is supposed to apply as much to the interest of commutative justice in the way gains from trade are shared among individual bargainers as to distributive justice (in the very narrow sense), retributive justice, and corrective justice, where there is an obvious interest in who ends up with what. Even if we grant that a lot of distribution is done “by the market,” we might still be interested in how individualized outcomes play out—that is, we might be interested in things such as rises or declines in equality and social mobility.

F.A. Hayek asserted that if a distribution is not done deliberately, it cannot possibly be a subject for evaluation in terms of justice.²⁷

This is a mistake. Hayek is perhaps right to insist that it may be nobody's fault that A happens to prosper in the marketplace while B descends into poverty, but still we might want this information in order to figure out what to do about it.²⁸ Moreover, though a market economy does not produce its outcomes intentionally, we can still make predictions about what range of social outcomes a given set of market institutions is likely to produce (from a given set of antecedent holdings). For example, those who faced the task of reconstructing economic institutions in Russia after 1989 had to ask themselves what kinds of market institutions they would try to foster. How would the market in securities be regulated? Would bank deposits be insured by the state? What consumer protection would there be? Would there be minimum wage institutions, employment protection, unionization, and collective bargaining? To address these questions sensibly, the Russian people needed some idea of how individual men and women were likely to fare under the various possible institutional regimes. How badly off would the worst-off be in an entirely unregulated system? It is obvious that structural changes ought to be assessed in these terms. Such assessment is utterly independent of any assumption about the intentionality of such effects; the most it assumes is that social institutions and the effects of their operations are under our control, at least in the sense that we can modify them to avoid ranges of outcomes deemed undesirable (on justice-related or other grounds).

I am not saying that an interest in justice is necessarily a commitment to equality. Justice is interested in equalities and inequalities (and in other distributive features such as mobility). However, our interest in these distributive features of social outcomes need not be an interest in what Robert Nozick calls “current time-slice” theories of justice.²⁹

Nozick has in mind theories of justice that assess what each person has ended up with at some point in time, considered as a distributive matrix of numbers without reference to the way these outcomes were reached. Suppose we find out at the end of this tax year that some individuals have bank balances of millions of dollars while others have bank balances of almost nothing. Is that discrepancy alone enough to condemn the overall state of affairs as unjust? Not according to most people, says Nozick:

The point can be conceded. At this stage we are just asking what sort of information the justicier is interested in. If Nozick is right, the justicier should be looking not only at the individual outcomes but also at their history. If the current time-slice theorists (whom Nozick criticizes) are correct, the justicier should look only at the array of individual outcomes. My point is that either way, the individual outcomes are the primary focus. We have to have this distributive information before we can assess it on any basis—whether the basis we want to assess it on is Nozickian historical entitlement or patterned theories of desert or simply the shape of the distributional matrices themselves.

VI.

Why are we so interested in distributive information? Why is it important to go beyond the aggregate information that economists offer us and move down to the level of individual outcomes?

The answer may strike some as disappointingly simple. This emphasis on distributive information is a reflection of our commitment to individualism. By “individualism,” I do not mean the methodological view that social science should regard individuals as the ultimate constituents of social reality; nor do I mean the political view that society will work best if individuals are in some sense left to their own devices. I mean a more abstract ethical view about the nature of value.³¹

Ultimately what matters most is how individuals are doing—that is, whether individual men, women, and children survive and prosper, and how things go for them as far as fulfillment and suffering are concerned. Pursued fanatically, individualism may amount to the view that nothing matters except how individuals are doing. (Joseph Raz defines moral individualism as “the doctrine that only states of individual human beings, or aspects of their lives, can be intrinsically good or valuable,” and he attacks it by adducing counterexamples.)³² But the primacy of distributive measures need not depend on anything that extreme. Many believe that groups, cultures, ways of life, even languages have an importance that goes beyond the effect they have on the well-being of individuals. This does not have to be denied in order to maintain the primacy thesis. All that the primacy thesis supposes is that how individuals are doing matters more than how collectives are doing, for all that states of the collective may also have some value in themselves.

I hinted earlier that “primacy” might be interpreted in slightly different ways, depending on whether the subordinated virtue was an aggregate virtue, like prosperity, or a genuinely collective good, like culture.³³

Let me now explain this. I want to distinguish two patterns of analysis that may be deployed in defense of the primacy claim.

The first pattern of analysis concerns the relation between justice and aggregate goods like prosperity or wealth maximization. Our best understanding of these aggregate goods is not that they are non- or anti-individualistic but that they present information about how individuals are doing in a form that is extraordinarily unhelpful (to an individualist). Consider the case of classical utilitarianism, for example. Despite the fulminations of Rawls and others—“Utilitarianism does not take seriously the distinction between persons” [TJ 24]—there is actually nothing anti-individualistic in the information presented in the utilitarian calculus. The utilitarian calculus sums individual pleasure and pain (or individual happiness or unhappiness, or the satisfaction and frustration of individual preferences), and nothing else; it does not involve attribution of the sum of pains and pleasures to any superindividual entity. What is wrong with the utilitarian calculus is that it presents this information in a way that is oddly indifferent to its real significance.³⁴

I said earlier, when we discussed the sentencing example, that it is very strange to present prison sentences for many individuals as sum totals when we know that the main point of measuring prison sentences is to see how much time people will have to serve one by one. Similarly there is something odd about saying that a social change will cause three million negative hedons of suffering across a whole population, when the whole point about suffering is that it is experienced by individual persons one by one, and what matters in the case of each person is how intense and prolonged his or her suffering is. In these cases, the primacy thesis is best understood as a sort of insistence that aggregative assessments of social structures are unsatisfactory unless they are accompanied by a determination to break down the information that has been presented in an aggregate form in a way that allows us to assess how individuals are doing. The primacy of justice is hence a way of drawing attention to the mistake someone would be making if he were to refuse to do this; that is, if he thought that the aggregate information as such was more important than any distributive breakdown. The critique implied in this version of the primacy thesis is partly internal. We criticize him not because he collected the wrong information or information with a lower priority but because of the way he presents the individualized information he has collected. (But our criticism has an edge because we suspect there is something about the distributive presentation of this information that he wants to hide. We say this because we cannot think of any other reason for presenting individualized information in an aggregate form.)

This first pattern of analysis can be used for a number of putative rivals to justice. If someone says that the prime virtue of any social structure is that it secures Hobbesian values of security and survival for its citizens, we want to ask who exactly benefits from these goods and why. As H.L.A. Hart observed at the end of his discussion of “the minimum content of natural law,” social institutions such as a legal system are capable of conferring these elementary protections quite unevenly among different classes of people.³⁵

So to the extent that we regard a society's ability to provide security as important, we should already be committed to scrutinizing this provision from a distributive perspective. As in the case of prosperity, it is barely intelligible from the moral point of view that we should not want to do this (although in a nonmoral perspective we know that aggregate assessments are used to distract critical attention from the fact that some are receiving this benefit and others are not). Maybe there is some elementary level of security which is necessarily secured as a public good and therefore does not give rise to questions of justice but may compete with the securing of justice in respect of other goods. But any priority accorded to the provision of this level of security is better understood—as arguably Thomas Hobbes understood it³⁶—as a condition for the existence of social institutions, something which, as we saw earlier is presupposed by the primacy thesis.³⁷

Something similar can be said about the suggestion that justice competes with liberty. Although we talk loosely of “a free society,” we know that freedom too needs to be disaggregated; different kinds of freedom can be evenly or unevenly distributed across a population. Since freedom (at least in its negative sense)³⁸

is an individual value, it would be preposterous to say that we were interested only in the total amount of freedom in society (whatever that means) and not in how much freedom each man or woman had and why.

This, then, is the first pattern of analysis: Given any suggestion that the provision of some good is a prime virtue of social institutions, we insist on asking hard questions—justice-questions—about the distribution of that good to individual persons.³⁹

Since these questions are morally inescapable from an individualist perspective, the virtue of providing such a good cannot compete with justice for priority. No doubt the argument becomes harder when justice in the provision of some one good (say, freedom) competes with justice in the provision of goods generally. But here it is worth noting that an adequate theory of justice will specify the whole list of important goods that matter from this perspective and establish their priority, one over another, as a matter of justice. Rawls's theory does this, and without wanting to get into the details of his particular account (for example, his account of the priority of equal liberty [TJ 214–220, 474–480]), we can say that this will be a necessary part of the agenda for any theory of justice that concerns itself holistically with the basic structure of society.⁴⁰

Our first pattern of analysis works for virtues that have to do with the provision of goods whose value is already understood individualistically. A different pattern of analysis is required for competition between justice and institutional virtues that concern genuinely collective goods. Here the argument is more direct and controversial: We say that justice has primacy because how individuals are doing simply matters more than how things go with collective entities.⁴¹

Some will deny the primacy thesis in this context and say either that some collective goods really are more important or that the relative importance of various goods is a matter of fine balance rather than the blunderbuss set of priorities that the primacy thesis assumes.⁴² We can do a certain amount to make this denial look less plausible. We can emphasize that the most persuasive accounts of collective goods understand their value ultimately in individualistic terms even if they do not see it as instrumental value.⁴³ We may even want to argue that it is part of the logic of many collective goods that they be understood in this way. Cultures and languages exist mainly for people, after all, rather than the other way round; or to the extent that they have value that goes beyond this, that value does not seem particularly prominent compared to their value to individuals. To the extent that cultures and languages exist for people, we can revert to the first pattern of analysis: It does seem important to interrogate whether all the members of a given society benefit from its promotion of a given culture or language, or only some; and if it is only some, then we will want to embark on something like a justice assessment of who benefits, who does not, and why.⁴⁴ But there may be limits on how far we can press this line. With some collective goods, it may not be possible to separate the benefit that accrues from them to distinct individuals.⁴⁵ In that case, it is not clear whether we should say that there is no issue of justice in relation to these goods (so the primacy thesis does not arise) or whether we should say that this shows that goods of this kind can be genuine competitors to justice as I have characterize it.

VII.

I have said that a justicier is interested in how individuals fare within the framework or under the impact of a system of social institutions and that his interest in this reflects—obviously enough—an underlying individualism about value. But this may not be quite enough to characterize justice as a virtue of social institutions. A justicier does not simply stare at an array of individual outcomes; she assesses them according to a principle. Now, because of our focus on the concept of justice, we do not want to beg any questions about what the contents of this principle might be—desert, merit, need, equality, or whatever. But we might say that it has to be a principle of a certain kind—a principle that assesses individual outcomes in relation to characteristics of the individuals concerned rather than in relation to criteria that have nothing to do with the individual persons as such. In other words, I am suggesting that justice is individualistic in a double sense; justice is assessment of individual outcomes by individualized criteria. We assess how A is faring in the distribution of some good by reference to the intrinsic importance of certain facts about A, such as A's needs or deserts or merits or A's basic moral standing as a person.

This point can be explained in part by contrasting it with approaches to individual outcomes that do not use individualized criteria in this way. In Mimi Leder's movie Deep Impact,⁴⁶

when a comet threatens the earth with a winter of mass extinction, the American government decides to shelter enough individuals in the limestone caves of Missouri to preserve the American way of life. Now, it is certainly important to determine who in particular would be sheltered in this way; it is important for all concerned that the government get this distribution right. But the distribution is not done by reference to the intrinsic importance of individual criteria; if individual criteria are involved at all, it is only as a means to the promotion of a collective good—the preservation of a certain way of life.

Or consider the utilitarian principle for determining individual punishments: The utilitarian assesses punishments by considering how they promote social welfare—by general deterrence, for example—so we assess A's sentence of ten years by considering whether society would have been better off for the imposition of, say, nine years or eleven years or no years at all. The trouble with this mode of assessment is that although it assesses an individual outcome, it does not conduct that assessment by reference to anything specifically about A except in the trivial sense that it is a fact about A that society may be better off for imposing this punishment upon him.⁴⁷

I suspect that what is going on here is that in the case of a genuine justice assessment, we do not have merely a concern for individual outcomes or an interest in how individuals are doing; we focus that interest in a way that also purports to respect the persons we are dealing with by assessing how they are doing in relation to features of their lives, actions, and moral personality that are or ought to be important to them. In a large social setting involving hundreds or millions of people, it is both important and difficult for decision-makers to be responsive to individuals in this way. It is easy to lose sight of the peculiar features of particular persons' situations—their actions, interactions, characteristics, and circumstances, features that distinguish them from other persons—when one sees them over and over again. The primacy of justice is established on the basis that some of these individual differences and some of these individual similarities are morally very important for the proper evaluation of social arrangements. The thesis reminds us of this, which is something that dealing with people en masse might otherwise incline us to forget. Now, of these important differences and similarities between individuals, some will seem like differences and similarities that we may pay attention to and respond to in our social decision-making; and others will seem like similarities and differences that we must pay attention to and respond to in our social decision-making. I mean “must” in the sense that it is a nonoptional, perhaps mandatory requirement of respect for persons. Justice, I think, is focused on the latter, which may be the grain of truth in the analytic version of the primacy thesis that we considered in Section III.⁴⁸

To tighten our characterization further, we may want to add reference to a crucial point we have neglected till now, which has to do with comparative or adjudicative character of reasons related to justice. A reason of justice for allocating X to A in particular (or for approving of a social system in which A ends up with X) is characteristically a reason focused not just on A's situation, but on A's situation as compared with B's and C's and D's.⁴⁹

In explaining why A in particular should get X, a justice-reason has to be able to take seriously and rebut the claims of other individuals with an interest in X. In previous sections I emphasized that justice is concerned with who gets what and why. But when there is competition for a scarce good, the inquiry into who and why involves allocative choices among several individuals; and the reasons have to be comparative in regard to the rival and competitive claims of different individuals. Justice-reasons for giving scarce good X (or a certain share of X) to A must be reasons that say something to B's, C's, and D's interests as well. We may express this by saying they must be adjudicatory—that is, they must explain why A's interest is the one that prevails in this particular situation.⁵⁰

This yields a sort of local primacy claim as far as individual attitudes towards justice are concerned. There are, let us imagine, good reasons for giving X to A and good reasons for giving X to B. To notice these reasons is to begin our justice-reasoning, but we have not finished it until we have figured out which of them is to prevail. Once that has been figured out, then the conclusion that is the upshot of that reasoning—say, that justice requires X to be allocated to A—obviously prevails over any considerations that grounded the contrary suggestion. Justice consists in settling this matter between the two or more individuals involved; and if it does settle the matter from a moral point of view, then a fortiori that settlement has primacy over the grounds of any of the competing claims. We sometimes ask: Which should take priority, justice or self-interest? But this is a silly question (at least for anyone who takes justice seriously enough to raise it). When the interests of two or more persons are opposed in some situation, it is the function of justice-reasoning to adjudicate the matter. The parties can hardly accept this understanding of justice but still regard it as an open question whether the basis of one of their claims among which justice has to make its adjudication is to prevail over the adjudication! But that is what regarding the claims of self-interest versus the claims of justice as an open question would entail. Of course, one can challenge any particular determination (over competing interests) made in the name of justice. But then one would be deploying an alternative conclusion of justice; one would not challenge it in the name of one of the rival claims.

VIII.

I have argued that justice is a virtue that pays attention to the reasons that justify the assignment of benefits and burdens to one person rather than another, and that justice is therefore in the business of adjudicating the rival claims that individuals may present. I also noted near the beginning of the paper that the claim that justice is the first virtue of social institutions is perhaps most plausible in the case of the legal system; not only is its operation key to what Rawls would call the basic structure of society, but its own self-representation is insistently oriented to justice.⁵¹

I would like to end discussion by considering the view taken on these matters in the economic analysis of law (EAL),⁵² for as things stand, EAL represents the most powerful tool used in the legal academy to deflect interest from distributive issues.⁵³

EAL is organized around the principle that disputed rights and resources should be allocated to those who value them most. This is what voluntary market transactions do, and the EAL is committed to the idea that law should facilitate and in some cases mimic market outcomes. When I buy your peach for a dollar, I lose a dollar and you lose a peach; but I gain something I value more highly than my dollar and you gain something you value more highly than the peach. Hence the transaction increases the amount of human value accruing from the possession and use of peaches and dollars on both sides. It maximizes wealth; and wealth-maximization is supposed to be an overriding goal.⁵⁴

Consider now a transaction that is slightly more complicated than the exchange of a peach for a dollar. Imagine that your peach trees are dying because the water you need to irrigate them is being used instead by my factory. And suppose that with this water my factory produces more for me than your orchard could possibly produce for you; indeed, it produces so much more for me that I could pay you all you might earn from an irrigated orchard and still have plenty left over for myself. Obviously, then, the amount of human value accruing from the use of this water is greater if it is used in my factory than in your orchard. And that would be the market outcome, at least in an ideal world; if I had a legal right to the water, you would not be able to offer me enough to divert it to your orchard; and if you had the legal right to the water, I would be able to offer enough to buy that right from you and use it in my factory. Value would be maximized in either case. The only additional “real-world” question is: How costly will the process of bargaining be and how much of the net gain will be eaten up in lawyers' fees, negotiating time, and the costs of drawing up the appropriate conveyances? According to the EAL, the law should be such that these processes are as near costless as possible. In the example we are considering, in which the facts about relative profitability are known, an initial legal assignment of the water right to the factory owner will minimize the time and trouble of transacting. So if a court ever faces the question about where the water right should be assigned, this should be the basis on which that question is answered.

Now, patently, this question—What can the law do to facilitate market outcomes?—does not address issues of distribution. And one would think those issues are important in our example. After all, it makes a huge difference to the individual wealth of the respective parties (under the efficient outcome) where the water right is initially located. If the water right is initially located with the orchardist, then even though the factory owner eventually gets the water, the orchardist will end up with at least as much money as he would have received from his irrigated crop; but if (as the model suggests) the court assigns the water right initially to the factory owner, then the orchardist will end up with nothing, or nothing but the value of parched land and dead trees. True, the factory owner will have enough from his profits to compensate the orchardist. But it is no part of the wealth-maximization model that this compensation should actually be paid. Any insistence that it should be paid is at best a distraction, according to EAL, and at worst, a recipe for multiplying transaction costs.⁵⁵

This carelessness about distributive outcomes has two sources: (1) the so-called Coase theorem, foundational to EAL, which holds that the initial assignment of entitlements is irrelevant to the maximization of wealth; and (2) the assimilation of what is sometimes called a Kaldor-Hicks improvement to a Pareto improvement. Both points require criticism.

(1) In a classic article, Ronald Coase showed that under ideal assumptions the initial distribution of rights makes no difference to the pursuit of efficiency.⁵⁶

And that is what we see in our example; the water rights always end up with the factory owner (provided transaction costs are low and the parties are economically rational). It would seem to follow that law in a nonideal world should not be too preoccupied with how rights fall out between the parties. Coaseans acknowledge that the way rights are distributed at the beginning of the efficiency-seeking process will certainly make a difference to the distribution of wealth at the end. But they seem to think that a preoccupation with the distributive features of the outcome is tainted by the same irrelevance that attaches to a preoccupation with the initial distribution of entitlements. Of course one would affect the other. But what matters is efficiency, they say, and in an ideal world that is not affected by the distributive question at either end of the process.

The fallacy here can be stated very simply. From the fact that the initial distribution of rights does not matter to the pursuit of efficiency, it cannot be inferred that the initial distribution of rights does not matter. And it cannot be inferred that efficiency ought to be pursued without regard to the distribution of rights. The initial allocation of rights may matter for reasons that have nothing to do with the pursuit of efficiency but reasons that nevertheless ought to have priority over aggregate measures of efficiency in the way that this paper has tried to explain.

An efficiency analysis purports to present information about how individuals are doing; it is not an inherently collectivist idea. But if it sidelines questions of distribution, it acts as if information about how individuals are doing need not be broken down into information about how this individual is doing, and how that individual is doing, and so on. Or it acts as though this disaggregated information were not worth assessing. It is, I submit, difficult to understand the moral attitude towards individuals that is supposed to license such indifference.⁵⁷

(2) The wealth-maximization criterion deployed in EAL is not presented directly as an application of the traditional utilitarian calculus. We do not simply add up the utilities on both sides and say that the assignment of the water rights to the factory is justified because it promotes total (or average) utility. Economists, like almost everyone else, say they are skeptical about such calculations. But—also like almost everyone else—they think that this skepticism should not obstruct the pursuit of Pareto improvements—that is, changes in which at least one person gains and nobody loses. Market exchanges are paradigms of Pareto improvement, and as we saw, the EAL tries as far as possible to present the water-rights transaction in that light. Could there be objections on grounds of justice to a Pareto improvement? It is not inconceivable; some egalitarians have thought that equality should be protected even against inequalities that leave everyone better off.⁵⁸

But it would be very hard to defend that position and accord it the primacy associated with justice on the sort of grounds on which I have defended the primacy thesis. If one thinks that the key to this whole matter is how individuals are faring, it is hard to oppose a change that involves nothing but individual-level improvements (unless, of course, one thinks that the outcomes comprised in the Pareto improvement are somehow misleading as to how individuals are really doing). So let us stipulate that Pareto improvements do not conflict with any plausible principle of justice and that therefore no issue of primacy arises.

Now, clearly the legal change that assigns disputed water rights to the factory owner in our example is not a Pareto improvement; although this change benefits the factory owner, the orchardist ends up with less than he otherwise would. But since the factory owner benefits more than the orchardist loses (measured by the fact that his gain would be more than enough to persuade the orchardist to give up the rights if they were assigned to him), the EAL insists that it is like a Pareto improvement, inasmuch as the increase in overall wealth is no less than what a Pareto improvement would allow. In effect this is the famous Kaldor-Hicks criterion of social improvement; a change which improves the situation of some by worsening the situation of others is permitted under this criterion if the winners win more than enough to compensate the losers.⁵⁹

But though a Kaldor-Hicks improvement is like a Pareto improvement in certain respects (e.g., in the amount of wealth gain required), it is not relevantly like a Pareto improvement in the respects that would encourage a justicier to support Pareto improvements. True, we ensure that the winner gains enough to compensate the loser, so that no-one need be worse off. But this assessment is not associated in any way with a practical determination to assign the loser any portion of what the winner gains. We can see the justice-based objection to this both from a point of view that considers the preferences of the loser—the orchardist, in our example—and from the point of view that considers the lack of consent in the transaction. The preferences of the orchardist are certainly taken into account in considering whether the water-rights transaction is an improvement; it must be the case that the orchardist would prefer a portion of the factory owner's gain to what he could secure as profits from running his orchard. But justice is not interested in facts about preferences per se. Justice is interested in facts about the satisfaction of preferences, and there is no proposal actually to satisfy the orchardist's preference for a portion of the factory owner's profits. Similarly, the legal imposition of a Kaldor-Hicks improvement as between the orchardist and the factory owner has none of the consensual respectability of a Pareto improvement (e.g., as in the case of the person with the peach and the person with the dollar). When we impose a Kaldor-Hicks improvement, we are not in any way honoring the voluntary consent of the losing party. Since the compensation that the factory owner could pay is not going to be paid, the orchardist is unlikely to consent to anything except keeping the water right. That the orchardist would consent (under condition C) adds nothing to the respectability of an actual transaction that involves no actual consent (because condition C is not satisfied). It is true that in some cases, we legitimate action on the basis of hypothetical consent; but not when the difference between the hypothetical and the real world is the presence of one of the conditions on which the hypothetical consent is predicated.

We see, then, that the Kaldor-Hicks idea partakes of all the objections made earlier about the indifference of aggregate measures to the actualities of distribution and that it evades none of them by addressing the distributive issue hypothetically. To oppose a Kaldor-Hicks imposition in the name of justice is therefore not to oppose justice to human welfare, nor is it to oppose justice to people's revealed preferences, nor is it to oppose justice to a scheme that makes everyone better off. It is rather to insist on the importance of respecting actual individuals with their actual preferences in the actual world, and it is to oppose the imposition on individuals of actual losses for which nothing but hypothetical compensation is envisaged.⁶⁰

Law-and-economics scholars seek various ways of bringing these modalities back into alignment. For example, it has sometimes been suggested that the effects of a wealth-maximization analysis are random, and that this should count in favor of EAL from a distributive point of view.⁶¹

The orchardist may lose in the water-rights case I postulated, but the wealth-maximization approach may favor him to the detriment of the factory owner in the next case that comes along. What he loses on the swings he may gain on the roundabout. But, first, it is a purely contingent matter whether things do in fact “even out” in this way. Second, the account assumes that we know (without asking) what form of “evening out” is appropriate—and that we can just say this in cavalier fashion without embarking on a discussion of justice. Third, the account neglects the possibility that some individual losses suffered in particular cases may be so ruinous as to be incapable of random redemption in this manner.⁶² And fourth, the approach neglects the point that justice is not just person-specific but situation-specific. From the point of view of justice, it matters not merely who gets what but also who gets what when and for what reason.⁶³ The “when” is not just a matter of the old proverb “Justice delayed is justice denied”—though that is true, too, and may be the motto of those losers who are waiting for the Posnerian odds to even out in their favor. It is also a matter of the way in which a justicier is supposed to pay attention to the reasons that pertain to individual persons in the particular situation that confronts them. A cavalier promise that “Well, hopefully things will even out in the long run” is insultingly neglectful of the occurrent importance of the reasons that pertain to distributions from a justice point of view.

Also, it is sometimes said that the virtue of an emphasis on wealth maximization is that it increases the sum of value in society available for distribution, and that surely increasing the size of the pie ought to have priority over sharing it out.⁶⁴

Now if there were a clear determination to distribute the maximized wealth in a way that was sensitive to considerations of justice, then there would be some force to this point. But as long as there is no such determination, the point is uninteresting. Indeed, it is as uninteresting as the claim that Al Capone should be entitled to his ill-gotten gains on the ground that he is thereby in a position to be a more generous Robin Hood, should the distributive mood ever come upon him. As long as no particular scheme of distribution is envisaged by those who talk about wealth maximization, the claim that we should give priority to the maximization of wealth over the proper distribution of such wealth as already exists in society is preposterous.

It is sometimes said that EAL applies mainly to decisions by courts and that EAL makes no judgment about the appropriate basis of decisions by legislatures, which, it is said, are the appropriate forums for considering redistribution.⁶⁵

That is right; legislatures are the appropriate forums for considering redistribution relative to the set of entitlements already officially acknowledged. But it does not follow that courts should be insensitive to the distributive—not redistributive but distributive—dimensions of the issues that come before them. Proponents of EAL maintain that wealth-maximization calculations provide the best model we have of what ought to go on in courtrooms. Judges, they say, ought to approach the rival claims of litigating parties as though nothing mattered more than the maximization of wealth; far from paying any attention to inherent merits of the distributive issues that the parties raise, they should finesse those issues as far as possible by using devices such as the Kaldor-Hicks criterion. Courts should be commissions for efficiency, letting the disputed costs and benefits fall where they may between plaintiffs and defendants.

I said at the beginning of this paper that the legal system is perhaps the one set of social institutions we have that maintains an orientation to justice analogous to Rawls's image of the orientation of science to truth. I want to reiterate that now in a way that highlights the objectionable nature of the EAL approach. Recent jurisprudence has emphasized that parties come to court not as lobbyists with various bright ideas for legal reform but—at least in their own eyes—as right-bearers: Each party thinks he is entitled to the outcome for which he is rooting. Ronald Dworkin has made much of this in his legal theory.⁶⁶

Subsequent discussion has oriented the point mainly to a slightly different issue, namely, what Dworkin says about right answers: Each party thinks the law already justifies the outcome he seeks, and Dworkin believes jurisprudence should attempt to make sense of that thought. But the original Dworkin point may also be oriented toward the distributive issue; plaintiffs and defendants approach litigation in the spirit that nothing matters more than the distributive question of who in particular ends up with what.⁶⁷ I do not just mean they are greedy and self-interested. For both of them, their position is a matter of what is legally just; plaintiff insists that he in particular is legally entitled to some benefit or resource or compensation, and defendant insists that he is not. To say that courts should try to finesse the distributive issue is to ignore the fact that which way a certain benefit goes on the distributional matrix is almost invariably what the lawsuit is exactly about. And this—I want to say—is exactly what we should conclude from the inherent orientation of our legal system to justice, if the account I have given of the nature and primacy of justice is correct.

This point about courts is sometimes connected with the observation, also made by proponents of EAL, that distributive questions are very difficult and controversial, and that as the calculus of wealth maximization is much simpler, we should indefinitely postpone any consideration of the former.⁶⁸

The premise is correct; issues of justice are often difficult. But this does not mean that they are not urgent or should not have priority. Our moral theory of the priorities that exist among the social virtues does not promise that the easy ones or the uncontroversial ones will come first; it does not promise intellectual efficiency in that rather infantile sense. The primacy thesis has to do with the character of the failings and shortcomings that a society may have if it neglects justice. The point here is exactly analogous to one made earlier about the television report telling us that five gangsters got two hundred years. It may—for all I know—be easier just to assign two hundred years to the five of them, and leave it for someone else to sort out who should get what sentence exactly within those parameters. After all, figuring out who exactly deserves what exactly may be difficult and controversial. Still, it is at the level of the individual allocation of sentences that the most terrible injustices occur, and it is wrongs of that sort that we believe must be avoided at all costs. Claims about wealth maximization offer to make the same mistake about distributive justice—as though the overall total mattered more than the justice of the prosperity or ruin of particular individuals. There is nothing wrong with someone playing with their efficiency calculations. But teaching our students that they are all-important because they are easy—and that this is why justice should be neglected—is quite another matter.